Mechanistic Basis of Differential Cellular Responses of Phosphatidylinositol 3,4-Bisphosphate- and Phosphatidylinositol 3,4,5-Trisphosphate-binding Pleckstrin Homology Domains*

Phosphatidylinositol 3,4-bisphosphate (PtdIns(3,4)P2) and phosphatidylinositol 3,4,5-trisphosphate (PtdIns(3,4,5)P3) are lipid second messengers that regulate various cellular processes by recruiting a wide range of downstream effector proteins to membranes. Several pleckstrin homology (PH) domains have been reported to interact with PtdIns(3,4)P2 and PtdIns(3,4,5)P3. To understand how these PH domains differentially respond to PtdIns(3,4)P2 and PtdIns(3,4,5)P3 signals, we quantitatively determined the PtdIns(3,4)P2 and PtdIns(3,4,5)P3 binding properties of several PH domains, including Akt, ARNO, Btk, DAPP1, Grp1, and C-terminal TAPP1 PH domains by surface plasmon resonance and monolayer penetration analyses. The measurements revealed that these PH domains have significant different phosphoinositide specificities and affinities. Btk-PH and TAPP1-PH showed genuine PtdIns(3,4,5)P3 and PtdIns(3,4)P2 specificities, respectively, whereas other PH domains exhibited less pronounced specificities. Also, the PH domains showed different degrees of membrane penetration, which greatly affected the kinetics of their membrane dissociation. Mutational studies showed that the presence of two proximal hydrophobic residues on the membrane-binding surface of the PH domain is important for membrane penetration and sustained membrane residence. When NIH 3T3 cells were stimulated with platelet-derived growth factor to generate PtdIns(3,4,5)P3, reversible translocation of Btk-PH, Grp1-PH, ARNO-PH, DAPP1-PH, and its L177A mutant to the plasma membrane was consistent with their in vitro membrane binding properties. Collectively, these studies provide new insight into how various PH domains would differentially respond to cellular PtdIns(3,4)P2 and PtdIns(3,4,5)P3 signals.

The PH domain is the first discovered PI binding domain (13)(14)(15), and this domain of 100 -120 amino acids is present in a wide range of signaling and membrane trafficking proteins. Although PH domains show low amino acid sequence homology, they have a conserved structural fold (see Fig. 1A) that consists of a 7-stranded ␤-sandwich (␤1 to ␤7), one end of which is capped by a C-terminal ␣-helix, and the open end is connected by three variable loops (␤1-␤2, ␤3-␤4, and ␤6 -␤7 loops). PH domains also show characteristic electrostatic polarization with a strong positive electrostatic potential surrounding the variable loops (see Fig. 1B) (32). Structural and biochemical studies have indicated that the variable loops form both a specific lipid binding pocket and a nonspecific membrane-binding surface (13)(14)(15). PH domains exhibit a wide range of PI affinities and specificities, which originate from high variability in the length and sequence of the variable loops, the ␤1-␤2 loop in particular (13,15).
In general, membrane binding of lipid binding domains is initiated by nonspecific electrostatic interactions between anionic lipids and cationic protein residues, which is followed by specific ligand binding and/or partial membrane penetration (9). It has been shown that the partial membrane penetration is important for the membrane recruitment and activation of lipid binding domains and their host proteins (9). For this reason, the depth of membrane penetration has been determined for various lipid binding domains by several biophysical techniques, including electron spin resonance (45) and x-ray reflectivity (46,47) measurements, and lipid binding domains have been classified into three groups, S-, I-, and H-types, depending on the degree of their membrane penetration (9). For FYVE (48), PX (49), and ENTH (50) domains, which belong to a membrane-penetrating H-or I-type, PI binding has been shown to trigger membrane penetration of the domains by inducing an electrostatic potential switch and/or local conformational changes. Although the depth of membrane penetration has not been reported for any PH domain, PH domains are generally thought to be S-type proteins whose membrane binding does not involve a significant degree of membrane penetration (9,10,13). This is largely because of the presence of a large number of cationic residues and the apparent lack of hydrophobic residues on their membrane-binding surfaces (32,51). However, recent studies have indicated that the PH domain of phospholipase C (PLC) ␦1 significantly penetrates the membrane (52,53).
To systematically study the mechanism by which PtdIns(3,4)P 2 and PtdIns(3,4,5)P 3 induce the membrane recruitment of their effector proteins, we quantitatively determined by surface plasmon resonance (SPR) and monolayer penetration analyses the PI affinities, specificities, and membrane binding properties of several PH domains that have been reported to bind PtdIns(3,4)P 2 and/or PtdIns(3,4,5)P 3 . The results reveal new PI specificities for these PH domains and also show that they have distinctively different membrane binding mechanisms, different degrees of membrane penetration in particular, due in part to variable distribution of hydrophobic residues in their membrane binding loops.

EXPERIMENTAL PROCEDURES
Materials-1-Palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC), 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphoethanolamine (POPE), and 1-palmitoyl-2-oleoyl-snglycero-3-phosphoserine were purchased from Avanti Polar Lipids, Inc. (Alabaster, AL). Phosphatidylinositol 3,4bisphosphate (PtdIns(3,4)P 2 ), phosphatidylinositol 4,5bisphosphate (PtdIns(4,5)P 2 ), and phosphatidylinositol 3,4,5-triphosphate (PtdIns(3,4,5)P 3 ) were purchased from Cayman. The concentrations of the phospholipids were determined by a modified Bartlett analysis. Octyl glucoside was purchased from Fisher. The Pioneer L1 sensor chip was FIGURE 1. Structures of PH domain-PI complexes. A, crystal structure of PLC␦ PH domain-Ins(1,4,5)P 3 complex is shown in ribbon diagram. Each secondary structural unit and three surface-exposed cationic loops are shown in different colors and labeled. Putative membrane-penetrating hydrophobic residues are shown in green spheres and labeled. B, electrostatic potential surface of the PLC␦ PH domain in the same orientation. Red and blue qualitatively indicate negative and positive electrostatic potentials, respectively. C, crystal structures of PH domains used in this study. These PH domains are oriented with their putative membrane-binding surface facing upward (notice that these orientations are different from that of the PLC␦-PH). Putative membrane-penetrating hydrophobic residues are shown in green spheres and labeled. For Btk-PH, Phe 44 near the membrane-binding surface is not fully exposed. Lys 18 (71), and 1H10 (Akt) (69). The structure of a 3 Gly variant of ARNO-PH is used in lieu of 2 Gly form used in this study. purchased from Biacore AB (Piscataway, NJ). Fatty acid-free bovine serum albumin was from Bayer, Inc. (Kankakee, IL).
Vector Construction and Mutagenesis-The cDNAs of murine Grp1, human Btk, murine ARNO, and human DAPP1 PH domains were subcloned into the vector pKTM, between the restriction sites BamHI and EcoRI. pKTM is a modified vector of pET-21a (ϩ), with an optional N-terminal His 6 tag and a Gly-Ser linker. DAPP1 PH domain mutants, L117A, V178A, and R184A, were generated by the overlap extension PCR and also subcloned into the pKTM vector. The cDNAs of the human Akt PH domain and the human TAPP1 C-terminal (TAPP1c) PH domain were subcloned into the pGEX-4T-1 (Novagen, Madison, WI) vector, containing an N-terminal glutathione S-transferase fusion. PH domain mutants were generated by the overlap extension PCR and subcloned into corresponding vectors. All of the above constructs were transformed into DH5␣ cells for plasmid isolation, and their DNA sequences were verified. Those constructs subcloned into the pGEX-4T-1 vector were transformed into Escherichia coli BL21 cells, whereas those subcloned into the pKTM vector were transformed into E. coli BL21(DE3) cells for protein expression. For the cellular translocation work in NIH 3T3 cells, the PH domains of Grp1, Btk, ARNO, DAPP1, and its L177A mutant, respectively, were subcloned between the HindIII and PstI sites of the mCerulean-C1 vector (54) and pEGFP-C2 vector, respectively, to yield the PH domains with cyan fluorescence protein (CFP) (54) and enhanced fluorescence green protein (EGFP), respectively, at the N terminus. The mCerulean-C1 vector was a generous gift of Dr. David Piston.
Protein Expression and Purification-For the expression of the Akt PH domain and the wild type and mutants of the TAPP1c PH domain, 1 liter of Luria broth containing 100 g/ml ampicillin was inoculated with BL21 cells harboring the PH domain construct and grown at 37°C until the absorbance at 600 nm reached 0.6. Protein expression was induced by the addition of 250 M isopropyl 1-thio-␤-D-galactopyranoside (Research Products, Mount Prospect, IL), and cells were harvested by centrifugation (5,000 ϫ g for 10 min at 4°C) after 14 h of incubation at 25°C. The resulting pellet was resuspended in 20 ml of 20 mM Tris buffer, pH 8, containing 160 mM KCl, 50 M phenylmethylsulfonyl fluoride, 2 mM dithiothreitol, and 0.1% Triton X-100. This solution was sonicated for 10 min (30 s of sonication followed by 30 s of cooling on ice) and then centrifuged for 30 min (48,000 ϫ g at 4°C). After filtering the supernatant into a 50-ml Falcon tube, 500 l of the glutathione S-transferase Tag TM resin (Novagen, Madison, WI) was added. After incubating this mixture on ice for 45 min with mild shaking at 80 rpm, it was poured onto a column pre-rinsed with 20 ml of 20 mM Tris buffer, pH 8, containing 160 mM KCl. After washing the nonspecifically bound protein with 50 ml of 20 mM Tris buffer, pH 8, containing 160 mM KCl, 1 ml of 20 mM Tris buffer, pH 8.4, containing 160 mM KCl, 25 mM CaCl 2 , and 1 unit of thrombin was added to cleave the glutathione S-transferase tag, and the column was sealed for the 12 h of incubation at 4°C. The protein was then eluted in five fractions using 500 l of 20 mM Tris buffer, pH 8, containing 160 mM KCl. All the fractions were collected together and purified by ion-exchange chromatography (see below).
For the expression of the Grp1, Btk, ARNO, and DAPP1 PH domains, 1 liter of Luria broth containing 100 g/ml ampicillin was inoculated with BL21(DE3) cells harboring each construct, and the culture was grown at 37°C until absorbance at 600 nm reached 0.5. At this time, 1 mM of isopropyl 1-thio-␤-D-galactopyranoside was added, and cells were then incubated at 25°C for 14 h. Cells were harvested for 10 min at 5000 ϫ g, and the resulting pellet was resuspended in 20 ml of 50 mM NaH 2 PO 4 , pH 8.0, containing 0.3 M NaCl, 10 mM imidazole, 50 M phenylmethylsulfonyl fluoride, 2 mM dithiothreitol, and 0.1% Triton X-100. The solution was then sonicated for 10 min (30 s sonication followed by 30 s cooling on ice). This was followed by centrifugation for 30 min (48,000 ϫ g at 4°C). The supernatant was filtered into a 50-ml tube, and 1 ml of nickel-nitrilotriacetic acid resin (Qiagen, Valencia, CA) was added. The mixture was incubated on ice with gentle stirring (80 rpm) for 1 h. After this time, the mixture was poured onto a column filled with 20 ml of 50 mM NaH 2 PO 4 , pH 8.0, containing 300 mM NaCl and 10 mM imidazole. After sequentially washing the nonspecifically bound protein with 50 ml of 50 mM NaH 2 PO 4 , pH 8.0, 300 mM NaCl containing 10, 15, 20, and 50 mM imidazole, respectively, the protein was eluted in five fractions using 1 ml of 50 mM NaH 2 PO 4 , pH 8.0, containing 300 mM NaCl and 300 mM imidazole. All the fractions were pooled and diluted with deionized water to 50 ml and purified using either a Q-Sepharose column for Akt-PH or a S-Sepharose column for all other PH domains, which was connected to the Á KTA FPLC system. The column was equilibrated with 20 mM MES buffer, pH 6.5 (Grp1-PH and ARNO-PH), 20 mM Tris buffer, pH 7.4 (Btk-PH and TAPP1c-PH), or 20 mM Tris buffer, pH 8.0 (Akt-PH), depending on the isoelectric point of the PH domain, and the elution was performed with the linear increase of KCl from 0 to 2 M in the same buffer. Protein purity was checked on an 18% polyacrylamide gel, and the protein concentration was determined by the bicinchoninic acid method (Pierce). The purified proteins were stored in 20 mM Tris-HCl buffer, pH 7.4, with 0.16 M KCl.
SPR Measurements-All SPR measurements were performed at 23°C using a lipid-coated L1 chip in the BIAcore X system as described previously (55). Briefly, after washing the sensor chip surface with the running buffer (20 mM HEPES, pH 7.4, containing 0.16 M KCl), POPC/POPE/PI (77:20:3), and POPC/ POPE (80:20), vesicles were injected at 5 l/min to the active surface and the control surface, respectively, to give the same resonance unit values. The level of lipid coating for both surfaces was kept at the minimum that is necessary for preventing the nonspecific adsorption to the sensor chips. This low surface coverage minimized the mass transport effect and kept the total protein concentration (P 0 ) above the total concentration of protein-binding sites on vesicles (M 0 ) (56). The control surface was also coated with 40 l of BSA (0.1 mg/ml in the running buffer) at a flow rate of 5 l/min before the injection of PH domains to minimize nonspecific adsorption of PH domains to the control surface. Equilibrium SPR measurements were done at the flow rate of 5 l/min to allow sufficient time for the R values of the association phase to reach near-equilibrium values (R eq ) (57). After sensorgrams were obtained for five or more different concentrations of each protein within a 10-fold range of K d , each of the sensorgrams was corrected for refractive index change by subtracting the control surface response from it. Assuming a Langmuir-type binding between the protein (P) and protein-binding sites (M) on vesicles (i.e. P ϩ M 7 PM) (56), R eq values were then plotted versus P 0 , and the K d value was determined by a nonlinear least squares analysis of the binding isotherm using an equation, R eq ϭ R max /(1 ϩ K d /P 0 ) (56). Each data set was repeated three or more times to calculate average and S.D. values. For kinetic SPR measurements, the flow rate was maintained at 15 l/min for both association and dissociation phases. Some kinetic data were analyzed with BIAevaluation 3.0 software (Biacore) to determine the rate constants of association (k a ) and dissociation (k d ), using a one-step (i.e. P ϩ M 7 PM) or two-step 1:1 (i.e. P ϩ M 7 PM 7 P*M) protein-membrane binding model.
Monolayer Measurements-The penetration of PH domains into the phospholipid monolayers of different lipid compositions was measured in terms of the change in surface pressure () using a 10-ml circular Teflon trough and a Wilhelmy plate connected to a Cahn microbalance as described previously (58).
Cellular Translocation Measurements-NIH 3T3 cells were seeded into 8 wells of a sterile Nunc Lak-TeKII TM chambered cover glass plate, which was filled with 400 l of Dulbecco's modified Eagle's medium (DMEM) and 10% (v/v) fetal bovine serum and incubated at 37°C with 5% CO 2 for 24 h. For transfection NIH 3T3 cells were incubated for 4 h with the mCerulean-C1 vector harboring each PH domain (1 g/ml) in the presence of the Lipofectamine 2000 reagent in OPTI-MEM (Invitrogen). Co-transfection of cells with CFP-Btk-PH and EGFP-DAPP1-PH (or EGFP-Btk-PH and CFP-DAPP1-PH) was performed with corresponding two plasmids in a 2:1 ratio to balance the emission intensities of two fluorophores. Cells were then incubated in DMEM with 10% fetal bovine serum overnight, washed with DMEM twice, and incubated in DMEM without serum for another 8 -10 h. For imaging, cells were washed twice with the Hanks' balanced salt solutions.
The membrane translocation of PH domains was monitored at fixed intervals (every 7 s) using a custom-built combination laser-scanning multiphoton microscope (59) after stimulation with 50 ng/ml of human platelet-derived growth factor-BB (PDGF-BB; Invitrogen) in 200 l of the same medium. For two-channel dual imaging of CFP-and EGFP-tagged proteins, fluorophores were excited at 840 nm, and 480/40 and 525/50 bandpass filters and a Q505 long pass beam splitter were used to separate the emission signals. Images were analyzed using SimFCS (a kind gift of Dr. Enrico Gratton) as described previously (60). Specifically, regions of interest in the cytosol were defined, and the average intensity in a square (1 ϫ 1 m) was obtained with respect to time. Membrane intensities were determined for each frame in individual cells by extending a line from the cytosol to the outside of the cell and reading off the intensity with distance along the line. Intensity values corresponding to the place on the line indicating the edge of the cell were averaged. Lines were drawn in at least three places in each cell, and membrane intensity was determined. These values were averaged, and the resultant cytosolic intensity values were converted to a ratio for each frame: membrane/(membrane ϩ cytosol). Each experiment was repeated at least three times on a given day and was repeated at least 2 different days with different transfected cells.
Effects of PI on Membrane Penetration of PH Domains-Our studies on FYVE (48), PX (49), and ENTH (50) domains have shown that PI binding triggers membrane penetration of the domains by inducing an electrostatic potential switch and/or local conformational changes. However, most PH domains are thought not to significantly penetrate the membrane. To date, only the PLC␦1 PH domain has been reported to have a significant membrane penetrating activity (52,53). To see if other PH domains have membrane-penetrating activity and how PIs may affect their membrane penetration, we measured the penetra-tion of the above PH domains into monolayers composed of various phospholipids. The lipid monolayer of a given surface pressure ( 0 ) was spread at a constant area, and the change in surface pressure (⌬) was monitored after the injection of proteins into the subphase. In general, ⌬ is inversely proportional to 0 of the phospholipid monolayer, and an extrapolation of ⌬ versus 0 yields c , which specifies an upper limit of 0 that a protein can penetrate (56). The surface pressure of cell membranes and large unilamellar vesicles has been estimated to be 31-35 dynes/cm (63)(64)(65). Thus, for a protein to effectively penetrate a particular cell membrane (or large vesicles), it should have the c value above this range for the monolayer whose lipid composition mimics that of the cell membrane.
When compared with DAPP1-PH and TAPP1c-PH, Akt-PH (data not shown), ARNO-PH (Fig. 3D), and Grp1-PH (Fig. 3E) showed smaller but definite increases in monolayer penetration, with c ranging from 27 to 29 dynes/cm, in the presence of PtdIns(3,4)P 2 or PtdIns(3,4,5)P 3 . Thus, membrane interactions of these domains are not likely to involve extensive membrane penetration but their modest membrane penetration activities may still play some role under certain conditions (see below). Among all PH domains tested, Btk-PH (Fig. 3F) exhibited the smallest increase in monolayer penetration in the presence of PtdIns(3,4,5)P 3 , indicating that the membrane penetration does not significantly contribute to its PtdIns(3,4,5)P 3 -dependent membrane interactions.

Effects of PI-induced Membrane Penetration on Membrane
Dissociation of PH Domains-To investigate how differential membrane penetrating activities of the PH domains may affect their membrane binding activities, we measured the kinetics of membrane association and dissociation of the PH domains by the SPR analysis. Our previous studies on various lipid binding domains and membrane-binding proteins have shown that nonspecific electrostatic interactions accelerate their membrane association, whereas short range specific interactions and membrane penetration slow the membrane dissociation (9, 55). It was therefore expected that membrane-penetrating PH domains, DAPP1-PH in particular, show slower membrane dissociation than weak penetrating PH domains, such as Btk-PH.
We injected the same concentration (final concentration of 200 nM) of PH domains to the sensor chip coated with POPC/ POPE/PtdIns(3,4,5)P 3 (77:20:3) (Fig. 4A) or POPC/POPE/ PtdIns(3,4)P 2 (77:20:3) (Fig. 4B) vesicles and monitored the kinetics of vesicle binding. Clearly, PH domains displayed diverse kinetic patterns. Rate constants (k a and k d ) could not be robustly determined and directly compared in this study because the observed kinetics for the PH domains followed complex patterns and did not uniformly conform to either a one-step (i.e. P ϩ M 3 PM) or a two-step (i.e. P ϩ M 3 PM 3 P*M) 1:1 binding model. Kinetics of membrane association for some PH domains, including TAPP1c-PH, could be fit with the one-step 1:1 binding model, but their membrane dissociation kinetics did not follow the same model. For this reason, only the qualitative comparison of kinetic patterns was made. As for the membrane association, different maximal resonance unit values caused by these PH domains reflect their different affinities for the vesicles (see Table 1) and different degrees of membrane penetration (see Fig. 3). As far as the rate of association is concerned, Btk-PH (Fig. 4A) and TAPP1c-PH (Fig. 4B) apparently were the fastest for PtdIns(3,4,5)P 3 -and PtdIns(3,4)P 2 -containing vesicles, respectively. This is consistent with highly positive electrostatic potentials on the membrane-binding surfaces of these PH domains (15).
When the kinetics of membrane dissociation was compared, noticeable different patterns were seen among PH domains. In the presence of POPC/POPE/PtdIns(3,4,5)P 3 (77:20:3) vesicles (Fig. 4A), Btk-PH dissociated faster than Grp1-PH and ARNO-PH, which is consistent with their monolayer penetration capabilities. In particular, ARNO-PH showed extremely slow dissociation after a rapid initial phase that accounted for ϳ20% of total dissociation. As a result, even after exhaustive elution (i.e. Ͼ20-min), about 75% of ARNO-PH still remained on the membrane surface (data not shown). When POPC/POPE/ PtdIns(3,4)P 2 (77:20:3) vesicles were used (Fig. 4B), the DAPP1 PH domain showed much slower dissociation than any other PH domain. This again shows that the ability of this PH domain to effectively penetrate densely packed lipid monolayers allowed this PH domain to reside on the membrane longer. Similarly to ARNO-PH, Akt-PH showed a rapid initial dissociation, followed by a much slower phase. Thus, it appears that the partial membrane penetration of PH domains plays a significant role in the kinetics of their membrane interactions, membrane dissociation in particular.
Structural Basis of Differential Membrane Penetration of PH Domains-To understand the basis of differential membrane penetration activities of the above PH domains, we examined their tertiary structures and performed mutational analysis. The reported crystal structures of PH domain-PI complexes (see Fig. 1) suggest that differential membrane penetrating abilities of the PH domains derive from the variable distribution of hydrophobic residues on their putative membrane-binding surfaces. Interestingly, DAPP1-PH has two prominently protruding hydrophobic residues, Leu 177 and Val 178 , side-by-side in the ␤1-␤2 loop, suggesting their critical involvement in membrane penetration (66). TAPP1c-PH also contains neighboring Val 204 and Met 205 in the ␤1-␤2 loop (67). Presumably, the lower membrane penetrating activity of TAPP1c-PH compared with DAPP1-PH derives from the lower hydrophobicity of Met 205 in the former than Leu 177 in the latter. It is interesting to find that PLC␦1-PH also has two neighboring hydrophobic residues, Val 58 and Met 59 , in its ␤3-␤4 loop (68). Thus, this PH domain is expected to be as effective as TAPP1c-PH in terms of membrane penetration. On the other hand, Akt-PH (69), ARNO-PH (70), and Grp1-PH (71) have one hydrophobic residue on the putative membrane-binding surfaces, which may contribute to their modest membrane penetrating activities. Also, Btk-PH (72) has no fully exposed hydrophobic residue on the membrane-binding surface (Phe 44 near the membrane-binding surface is not fully exposed), which accounts for its lower monolayer penetration.
Thus, it appears that the presence of two neighboring hydrophobic residues in a protruding loop is important for a PH domain to penetrate densely packed monolayers and bilayers.
To test the notion that surface hydrophobic residues are important for the membrane penetration and sustained membrane residence of PH domains, we mutated two hydrophobic residues (Leu 177 and Val 178 ) of DAPP1-PH to Ala and measured the effects of mutations on its membrane binding. As shown in Fig. 3B, the mutation of a single hydrophobic residue to Ala was enough to essentially abrogate the PtdIns(3,4)P 2enhanced monolayer penetration of the PH domain, i.e. both mutants penetrated the POPC/POPE/PtdIns(3,4)P 2 (77:20:3) monolayer only as effectively as the wild type DAPP1-PH penetrated the POPC/POPE (80:20) monolayer. Furthermore, the kinetic SPR analysis of these proteins (see Fig. 4C) revealed that L177A and V178A dissociated from the membrane much faster than the wild type, behaving similarly to other PH domains with weaker membrane penetrating activities. In particular, L177A showed Ͼ70% dissociation from the membrane. L177A and V178A also had significantly lower affinities for POPC/POPE/ PtdIns(3,4)P 2 (77:20:3) vesicles than wild type (see Table 1).
To see if surface hydrophobic residues play the same role for other PH domains, we mutated two hydrophobic residues (Val 204 and Met 205 ; see Fig. 1) of TAPP1c-PH and monitored the membrane dissociation kinetics of TAPP1c-PH and mutants, V204A and M205A. The wild type TAPP1c-PH showed about 30% dissociation after 5 min (see Fig. 4D) and less than 40% dissociation after exhaustive elution (data not shown). Under the same condition, however, V204A and M205A exhibited about 95 and 80% dissociation. Notice also that the membrane dissociation of M205A to V204A was much faster than that of the wild type TAPP1c-PH. We also introduced two hydrophobic residues into the ␤1-␤2 loop of Btk-PH by the K18F/K19L mutation to see if Btk-PH could be converted into one with DAPP1-PHlike properties. As shown in Fig. 4A, this mutant showed significantly slower membrane association than the wild type Btk-PH. This is presumably because of the removal of two lysines that are involved in nonspecific electrostatic interaction with the anionic membrane, which facilitates the membrane association. More importantly, the mutant exhibited extremely slow membrane dissociation; little dissociation was observed even after exhaustive elution. Collectively, these results support the notion that two proximal hydrophobic residues on the membrane-binding surface are essential for a PH domain to penetrate compactly packed lipid bilayers, including cell membranes, in a PI-dependent manner. The results also indicate that the partial membrane dissociation of PH domains is because of PI-induced membrane insertion of hydrophobic residues on their membrane-binding surfaces.
Differential Membrane Cellular Translocation of PH Domains-To see if cellular membrane targeting behaviors of PH domains are governed by their membrane binding properties, we monitored the PtdIns(3,4,5)P 3 -mediated subcellular translocation of Btk-PH, Grp1-PH, ARNO-PH, and DAPP1-PH each tagged with the modified CFP at their N termini, in NIH 3T3 cells. The cell populations expressing similar levels of PH domains were selected by visual inspection of CFP fluorescence intensity and used for translocation measurements. A minimum of quadruple measurements was performed for each protein with Ͼ5 cells monitored for each measurement. Typically, Ͼ80% of cell population showed similar behaviors with respect to membrane translocation of PH domains. Fig. 5 shows the time-lapse images of PH domains in representative cells. When expressed in NIH 3T3 cells grown in serum-supplemented media, all PH domains showed some degree of pre-localization at the plasma membrane of quiescent cells (data not shown), which was largely suppressed by serum starvation. As reported previously (73), nuclear localization was observed for all PH domains, with Grp1-PH showing the most pronounced nuclear distribution (Fig. 5A). When NIH 3T3 cells were treated with 50 ng/ml PDGF, which was reported to produce PtdIns(3,4,5)P 3 in the plasma membrane (74,75), cytosolic Btk-PH and Grp1-PH rapidly translocated to the plasma membrane (Fig. 5, A and C). In particular, Btk-PH showed the fastest translocation to the plasma membrane, completing the translocation within 2 min (Fig. 5C). In contrast, ARNO-PH exhibited much slower plasma membrane translocation. Interestingly, Btk-PH moved back to the cytoplasm rather quickly after reaching the maximal plasma membrane localization, presumably due to the removal of PtdIns(3,4,5)P 3 by cellular enzymes, such as PTEN. In comparison with Btk-PH, Grp1-PH showed substantially slower dissociation, whereas ARNO-PH, although translocating to the plasma membrane much more slowly than other PH domains, displayed little sign of membrane dissociation during 10 min. These different plasma membrane translocation behaviors of the three PH domains are consistent with their kinetic patterns of vesicle binding shown in Fig. 4A.
To further investigate the correlation between in vitro membrane binding properties of PH domains and their cellular membrane targeting, we then measured the dissociation of DAPP1-PH and its L177A mutant from the plasma membrane after PDGF stimulation (Fig. 5D). Although DAPP1-PH prefers PtdIns(3,4)P 2 to PtdIns(3,4,5)P 3 , it still has higher affinity for PtdIns(3,4,5)P 3 -containing vesicles than ARNO-PH (see Table  1). Accordingly, both DAPP1-PH and L177A translocated to the plasma membrane in response to PDGF stimulation faster than ARNO-PH, and slightly slower than Grp1-PH. Most important, DAPP1-PH stayed bound to the plasma membrane much longer than Btk-PH and Grp1-PH. It gradually dissociated from the membrane only after 8 min. In contrast, L177A started to dissociate from the plasma membrane as fast as Btk-PH even before it reached the full association.
Finally, we co-transfected the NIH 3T3 cells with EGFP-Btk-PH and CFP-DAPP1-PH (or vice versa) and simultaneously monitored their membrane translocation by two-channel imaging in response to PDGF activation to verify that these two PH domains show distinct membrane translocation patterns under the same conditions. As illustrated in Fig. 5, B and E, individual behaviors of EGFP-Btk-PH and CFP-DAPP1-PH in co-transfected cells were essentially the same as their behaviors in independently transfected cells, i.e. Btk-PH rapidly bound to and dissociated from the plasma membrane in response to PDGF activation, whereas DAPP1-PH showed much slower membrane dissociation. The same pattern was seen when NIH 3T3 cells were co-transfected with CFP-Btk-PH and EGFP-DAPP1-PH (data not shown). These co-transfection experiments thus rule out a possibility that differential responses of PH domains to PDGF activation are due to variable effects of their overexpression on PtdIns(3,4,5)P 3 signaling pathways in NIH 3T3 cells. Collectively, our results show that the kinetics of PtdIns(3,4,5)P 3 -mediated subcellular localization of the PH domains is well correlated with their relative affinity for PtdIns(3,4,5)P 3 -containing vesicles (Table  1) and, more importantly, the kinetics of binding to PtdIns(3,4,5)P 3 -containing vesicles (Fig. 4A). This semi-quantitative correlation supports the notion that differential cellular membrane targeting behaviors of PH domains are largely attributed to their divergent membrane binding properties.

DISCUSSION
In cell signaling, a single signal mediator, such as Ca 2ϩ and a lipid second messenger, can mediate divergent signaling pathways in a spatially and temporally specific manner. A majority of PIs serve as site-specific membrane signals that recruit and activate effector proteins, and generally, each PI is recognized by multiple effector molecules. This raises a question as to how a PI selectively recruits and activates a particular effector protein under a specific condition. A commonly proposed mechanism is the membrane recruitment by multiple interactions, involving a lipid and a protein adaptor or two different lipids, for example, which would confer an extra level of specificity and control (1,76). It is also possible that the compartmentalization through the formation of signaling complexes at the membrane allows only a small set of proteins to see the locally confined PI molecules. Another possibility is that PI effectors have such divergent membrane binding properties that they can differentially respond to emerging PI molecules. The present study investigates the last aspect using several PtdIns(3,4)P 2 -and PtdIns(3,4,5)P 3 -binding PH domains as models.
Because the lipid selectivity of many PH domains has been determined under different conditions, we reexamined the lipid selectivity of representative PtdIns(3,4)P 2 -and PtdIns(3,4,5)P 3binding PH domains by SPR analysis. This study redefines PI selectivity for some PH domains (ARNO, DAPP1, Grp1, and Akt), and confirms previous reports on other PH domains (Btk and TAPP1c). In general, our PI selectivity data are in good agreement with the reported structural information on PH domain-PI complexes. For example, the pronounced specificity of Btk-PH for PtdIns(3,4,5)P 3 -containing vesicles is consistent with its large binding pocket containing ligands for D3, D4, and D5 phosphoryl groups (72). Its extremely low affinity for other PIs indicates that partial occupancy of the binding pocket of Btk-PH is not conducive to productive PI binding. Also, high PtdIns(3,4)P 2 specificity of TAPP1c-PH has been attributed to the presence of an Ala residue in the ␤1-␤2 loop that sterically discriminates against the D5 phosphate group (67). Furthermore, the modest preference of DAPP1-PH and Akt-PH for PtdIns(3,4)P 2 over PtdIns(3,4,5)P 3 agrees with the finding that they do not have specific binding sites for the D5 phosphate in the pockets (66,69). Significant PtdIns(3,4,5)P 3 selectivity of Grp1-PH is also consistent with its structure showing a well defined binding site for the D5 phosphate. However, it is somewhat unexpected to find that a structurally similar ARNO-PH has modest selectivity for PtdIns(3,4)P 2 over PtdIns(3,4,5)P 3 because it also has a well defined binding site for the D5 phosphate. It should be noted that splice variants of ARNO/Grp1 that contain three Gly residues in the ␤1-␤2 loop have different lipid selectivity than the major forms that have two glycines in the same loop (38,70). Our sequencing analysis verified that ARNO-PH used herein has two glycines in the loop (data not shown) and thus our specificity data are not because of structural variation. It thus appears that binding to the D5 phosphate to ARNO-PH does not contribute much to the overall ligand binding energy. This notion is consistent with the finding that ARNO-PH has much lower affinity for PtdIns(3,4,5)P 3 -containing vesicles than Btk-PH or Grp1-PH.
Because PtdIns(3,4)P 2 and PtdIns(3,4,5)P 3 are lipid second messengers that accumulate only transiently in stimulated cells, their effector proteins must be able to rapidly and proficiently respond to these signals. This entails high affinity (i.e. low K d ) and rapid association kinetics (i.e. high k a ). Our results show that TAPP1c-PH and Btk-PH meet these requirement as specific PtdIns(3,4)P 2 and PtdIns(3,4,5)P 3 effector proteins. They have high PI specificities (i.e. Ͼ20-fold selectivity for their preferred PIs) and affinities (K d Ͻ 100 nM), and they also show rapid membrane association kinetics (Fig. 4, A and B). Although individual rate constants for PH domains could not be accurately measured in this study because of complex kinetic patterns, their k a values estimated from their association kinetic phases, assuming one-step 1:1 membrane-protein binding, are in the range of 10 5 M Ϫ1 s Ϫ1 , which compare favorably with those of other fast lipid/membrane binding domains (9). Grp1-PH also has high affinity for PtdIns(3,4,5)P 3 -containing vesicles, but it displays significant affinity for PtdIns(3,4)P 2containing vesicles, and its association rate constant appears to be smaller than that of Btk-PH (Fig. 4B).
On the basis of these membrane binding properties, one would expect that Btk-PH and TAPP1c-PH would respond to emerging cellular PtdIns(3,4,5)P 3 and PtdIns(3,4)P 2 signals, respectively, more effectively than other PH domains if lipidprotein interactions provide a primary driving force for membrane targeting of these proteins. Grp1-PH should serve as a proficient effector for PtdIns(3,4,5)P 3 , but it may also respond to the formation of PtdIns(3,4)P 2 . Likewise, the PH domains of Akt, ARNO, and DAPP1, which do not clearly distinguish between PtdIns(3,4)P 2 and PtdIns(3,4,5)P 3 , may serve as dual effectors for PtdIns(3,4)P 2 and PtdIns(3,4,5)P 3 but their cellular responses would be slower than Btk-PH and Grp1-PH for PtdIns(3,4,5)P 3 and TAPP1c-PH for PtdIns(3,4)P 2 . Our cell translocation results support this notion. When the formation of PtdIns(3,4,5)P 3 is induced by PDGF in NIH 3T3 cells, Btk-PH shows the fastest translocation to the plasma membrane, which is followed by Grp1-PH and ARNO-PH.
It should be noted that there is some noticeable discrepancy between our membrane affinity data and published data. For example, Grp1-PH (35)(36)(37)(38) and Akt-PH (40,41) have been reported to have higher affinity for PtdIns(3,4,5)P 3 than Btk-PH. However, our results show that Btk-PH has 2-and 10-fold higher affinity for PtdIns(3,4,5)P 3 -containing vesicles than Grp1-PH and Akt-PH, respectively. This type of discrepancy in membrane binding parameters, which is commonly found for lipid binding domains, arises mainly from the fact that affinity data were obtained by various methods (e.g. lipid overlay, vesicle pelleting, fluorometric, or SPR assay) with highly variable sensitivity and accuracy under different conditions (e.g. PIs in different physical states). This makes it difficult to directly compare reported membrane affinities. The semi-quantitative cor-relation between our in vitro membrane binding data and cell data not only validates our affinity data but also underscores the importance of rigorously and systematically determining membrane binding parameters of lipid binding domains under well defined conditions.
Another key finding in this study is that all PH domains but Btk-PH show PI-induced membrane penetration, although the degree of penetration varies widely among them. PH domains have been generally considered as S-type proteins that primarily interact with the membrane surface because of the presence of a cluster of basic residues on their putative membrane-binding surfaces (9). So far, membrane penetration has been reported only for the PH domain of PLC␦1 (52,53). In particular, Flesch et al. (53) reported that PLC␦1-PH had strong PI-independent monolayer penetration and partial dissociation from the vesicle-coated SPR chip. However, this study was performed with excessively high PI concentrations (i.e. 20 mol %) in monolayers and bilayers, which casts doubt on the validity and physiological relevance of their findings and interpretation. Our study employing physiologically more relevant PI concentrations (i.e. 1-5 mol %) clearly shows that PtdIns(3,4)P 2 and/or PtdIns(3,4,5)P 3 significantly enhance the monolayer penetration of PH domains. For DAPP1-PH and TAPP1c-PH with the highest monolayer penetrating activities, the monolayer penetration is greatly promoted by PtdIns(3,4)P 2 . In particular, 1-3 mol % of PtdIns(3,4)P 2 allows DAPP1-PH to effectively insert its hydrophobic residues on the ␤1-␤2 loop into densely packed monolayers and bilayers. Thus, PtdIns(3,4)P 2 would seem to trigger the penetration of DAPP1-PH (and perhaps TAPP1c-PH) into the cell membrane. On the basis of our previous studies on FYVE (48), PX (49), and ENTH (50) domains, we postulate that PtdIns(3,4)P 2 binding induces electrostatic attenuation of the membrane-binding surface of DAPP1-PH (i.e. reducing the positive electrostatic potential surrounding hydrophobic residues, thereby allowing them to penetrate the membrane without paying the significant dehydration penalty) and/or causes local conformational changes. Our results indicate that the presence of two neighboring hydrophobic residues (Leu 177 and Val 178 for DAPP1-PH) is critical for this strong membrane-penetrating activity as the substitution of either residue by a less hydrophobic one in DAPP1-PH mutants or in other PH domains greatly weakens the monolayer penetration and shortens the residence time at the lipid bilayers. Also, introduction of two neighboring hydrophobic residues into the membrane binding loop of Btk-PH converts it into a DAPP1-PH-like domain with strong membrane-penetrating activity.
It has been reported that sustained membrane residence, which is achieved through membrane insertion of hydrophobic residues, is essential for the activation and processive actions of membrane-binding proteins (9). Although we could not quantitatively fit the complex membrane binding kinetics of PH domains using a single model, monolayer and kinetic SPR data of wild type and mutant PH domains indicate that membrane penetration of PH domains, mediated by hydrophobic residues on their membrane-binding surfaces, plays a key role in regulating the dissociation of PH domains from model and cell membranes. DAPP1-PH with two membrane-penetrating hydrophobic residues shows extremely slow in vitro and cellular membrane dissociation, whereas removal of either of two hydrophobic residues dramatically accelerates the membrane dissociation. Even for those PH domains with modest monolayer penetrating activities, their membrane dissociation kinetics seem to be greatly influenced by the partial membrane penetration of a hydrophobic residue(s) on their membrane-binding surfaces. Btk-PH is unique among PtdIns(3,4,5)P 3 -binding PH domains in that it shows rapid dissociation for the plasma membrane presumably in response to the removal of PtdIns(3,4,5)P 3 . Thus, its rapid kinetics of membrane association and dissociation may enable Btk-PH to serve as a genuine sensor for cellular PtdIns(3,4,5)P 3 .
It should be noted that there is some discrepancy between our monolayer penetration data and membrane dissociation data. For instance, ARNO-PH and Grp1-PH show similar monolayer penetration behaviors, but ARNO-PH dissociates from model and cellular membranes more slowly than Grp1-PH. Although it is beyond the scope of this investigation, determination of the membrane-bound topology of different PH domains and the depth of their membrane penetration should provide further insight into the basis of different membrane dissociation behaviors of PH domains. Also, protein-protein interactions, which may play a significant role in subcellular localization and cellular actions of PH domains (13)(14)(15), can slow the membrane dissociation of proteins (62). This may at least partially account for the different cellular membrane dissociation behaviors between Grp1-PH and ARNO-PH, for example.
Cellular membrane recruitment has been studied for many PH domains with different PI specificities (13)(14)(15)). Yet it remains unclear how each PI can differentially regulates many PH domains. Due in part to a lack of kinetic and thermodynamic parameters of membrane binding determined under well defined conditions, however, cellular membrane translocation properties of PH domains with similar PI specificities (e.g. PtdIns(3,4,5)P 3 ) have not been systematically compared. This study provides such information for the PH domains that interact with two important lipid second messengers, PtdIns(3,4)P 2 and/or PtdIns(3,4,5)P 3 . A good overall correlation between the membrane binding properties of these PH domains and their cellular translocation indicates that the kinetics of PI-mediated cellular membrane recruitment of these PH domains is governed, to a large extent, by their membrane binding properties. This should also help account for different cellular functions and regulation of the proteins in which these PH domains reside. Furthermore, our results should form the foundation of systematic and quantitative assessment of different cellular membrane translocation properties of a large number of PH domains and their host proteins.