Engineering the Phosphoinositide-binding Profile of a Class I Pleckstrin Homology Domain*

Pleckstrin homology (PH) domains are protein modules that bind with varying degrees of affinity and specificity membrane phosphoinositides. Previously we have shown that although the PH domains of the Ras GTPase-activating proteins GAP1m and GAP1IP4BP are 63% identical at the amino acid level they possess distinct phosphoinositide-binding profiles. The GAP1m PH domain binds phosphatidylinositol 3,4,5-trisphosphate (PtdIns(3,4,5)P3), whereas the domain from GAP1IP4BP binds PtdIns(3,4,5)P3 and phosphatidylinositol 4,5-bisphosphate (PtdIns(4,5)P2) equally well. These phosphoinositide specificities are translated into distinct subcellular localizations. GAP1m is cytosolic and undergoes a rapid PtdIns(3,4,5)P3-dependent association with the plasma membrane following growth factor stimulation. In contrast, GAP1IP4BP is constitutively associated, in a PtdIns(4,5)P2-dependent manner, with the plasma membrane (Cozier, G. E., Lockyer, P. J., Reynolds, J. S., Kupzig, S., Bottomley, J. R., Millard, T., Banting, G., and Cullen, P. J. (2000) J. Biol. Chem. 275, 28261–28268). In the present study, we have used molecular modeling to identify residues in the GAP1IP4BP PH domain predicted to be required for high affinity binding to PtdIns(4,5)P2. This has allowed the isolation of a mutant, GAP1IP4BP-(K591T), which while retaining high affinity for PtdIns(3,4,5)P3 has a 6-fold reduction in its affinity for PtdIns(4,5)P2. Importantly, GAP1IP4BP-(K591T) is predominantly localized to the cytosol and undergoes a PtdIns(3,4,5)P3-dependent association with the plasma membrane following growth factor stimulation. We have therefore engineered the phosphoinositide-binding profile of the GAP1IP4BP PH domain, thereby emphasizing that subtle changes in PH domain structure can have a pronounced effect on phosphoinositide binding and the subcellular localization of GAP1IP4BP.

Several PH domain structures have now been solved . The core of each domain is a ␤-sandwich composed of two nearly orthogonal ␤-sheets, ␤1-␤4 and ␤5-␤7, that are connected by six loop regions. Three of these, ␤1/␤2, ␤3/␤4, and ␤6/␤7, have been termed the variable loops, as they display hypervariable sequences in PH domain alignments. These loops close off one corner of the ␤-sandwich, whereas an amphipathic carboxy-terminal ␣-helix closes off the opposite open corner. To date the structure of six PH domains have been solved in complex with the inositol phosphate head group of their cognate phosphoinositide (18,19,25,27,28,32). These studies have shown that although PH domains share a common fold, they have evolved their sequence and structure to provide a wide range of different phosphoinositide specificities and affinities.
In the present study, we have used molecular modeling to generate predicted structures for the PtdIns(4,5)P 2 and/or PtdIns(3,4,5)P 3 binding pockets within the PH domains of GAP1 m and GAP1 IP4BP (38). These models were analyzed to identify residues in the GAP1 IP4BP PH domain predicted to be required for high affinity binding to PtdIns(4,5)P 2 . Three possible residues, lysine 591, arginine 604, and lysine 616, were identified each of which was individually mutated into the corresponding residue from GAP1 m . Analysis of the phos-phoinositide-binding profile revealed that although the affinity for PtdIns(4,5)P 2 was retained for the lysine 616 mutant, an approximate 6-fold reduction in affinity was observed for the lysine 591 and arginine 604 mutants. Importantly, for these two mutants the drop in affinity for PtdIns(4,5)P 2 was sufficient to result in a loss in their constitutive association with the plasma membrane. However, these cytosolically localized mutants retained a high affinity for PtdIns(3,4,5)P 3 , which was manifested in vivo through their ability to undergo a PtdIns(3,4,5)P 3 -dependent association with the plasma membrane following growth factor stimulation. By using molecular modeling coupled with single site mutagenesis we have therefore engineered the phosphoinositide-binding profile of the GAP1 IP4BP PH domain and hence manipulated the subcellular localization and regulation of this protein.

EXPERIMENTAL PROCEDURES
Molecular Modeling of the GAP1 IP4BP and GAP1 m PH Domains-All analysis and modeling of the PH domains was performed using the program packages Quanta/CHARMm and InsightII. The alignment of PLC-␦ 1 and Btk was structure-based, with the GAP1 IP4BP and GAP1 m PH domains being aligned by sequence, initially using Omiga and then refined by hand where necessary. The modeling was performed as described previously (38), with GAP1 IP4BP and GAP1 m being modeled independently on the structure of Btk in complex with Ins(1,3,4,5)P 4 (25).
Expression and Purification of GAP1 IP4BP and GAP1 IP4BP Mutants-Both wild-type and mutagenic GST-GAP1 IP4BP fusion proteins were isolated as follows. An overnight 20-ml culture was used to inoculate each of 4 ϫ 2.5-liter conical flasks of 500 ml of LB containing ampicillin (50 g/ml). The cells were cultured at 25°C until an A 600 of 0.6. Protein expression was then induced by the addition of 0.1 mM isopropyl-1-thio-␤-D-galactopyranoside followed by an overnight incubation at 15°C prior to harvesting the bacteria by centrifugation (3000 ϫ g for 10 min at 4°C). All of the subsequent steps were carried out at 4°C. The bacterial pellet was gently resuspended in 40 ml of PBS buffer (pH 7.2) containing EDTA (1 mM), EGTA (1 mM), ␤-mercaptoethanol (1 mM), and Triton X-100 (0.1%) (v/v) and sonicated for four periods of 30 s with 30 s on ice between each sonication. Any cell debris was removed by centrifugation (36000 ϫ g for 30 min). The supernatant was removed, and 2 ml of a 1:1 suspension of glutathione-agarose beads (washed and preswollen with several volumes of PBS buffer) was added and incubated on a rotating wheel for 1 h at 4°C. The beads were pelleted by centrifugation and washed sequentially with 3 ϫ 20 ml of PBS containing 0.1% (v/v) Triton X-100. The beads were added to a column and washed with a further 3 ϫ 20 ml of PBS. The protein was cleaved from the GST tag while bound to the column using thrombin at room temperature overnight. The free protein was then washed off the column with PBS.
Phosphoinositide Binding as Determined Using a Protein-Lipid Overlay Assay-To assess the phosphoinositide binding properties of each GAP1 IP4BP mutant, a protein-lipid overlay assay was performed using the GST fusion proteins. Briefly, 1 l of lipid solution containing 1-450 pmol of phospholipids dissolved in a mixture of chloroform/methanol/ water (1:2:0.8 by volume) was spotted on to Hybond-C extra membrane and allowed to dry at room temperature. The membrane was incubated with blocking solution (50 mM Tris-HCl buffer (pH 7.5) containing 3% (w/v) bovine serum albumin, NaCl (150 mM), and Tween 20 (0.1% (v/v) final)) for 1 h at room temperature. The membrane was then incubated, with gentle rocking, overnight at 4°C with 0.5 g/ml of the relevant protein in blocking solution. The membranes were washed 4 ϫ 15 min in washing solution (50 mM Tris-HCl buffer (pH 7.5) containing NaCl (150 mM) and Tween 20 (0.1% (v/v) final)) and then incubated for 1 h with 1:1000 dilution of anti-GAP1 IP4BP monoclonal antibody. The membranes were washed as before prior to being incubated for 1 h with 1:1000 dilution of anti-mouse horseradish peroxidase conjugate. Finally, the membranes were washed as before, and the bound protein was detected by enhanced chemiluminescence.
Transient Transfection and Indirect Immunofluorescence-HeLa and PC12 cells, cultured as described, were plated on glass coverslips and transfected with vector DNA at 50 -60% confluence by lipofection using LipofectAMINE (Invitrogen) at a ratio of 0.15 g of DNA/l of cationic lipid. 24 h after transfection, HeLa cells were serum-starved for 2 h prior to being fixed and stained for GAP1 IP4BP expression as described previously (39). 24 h after transfection PC12 cells were either serumstarved for 2 h followed by incubation with 100 nM wortmannin for 30 min or incubated with 100 ng/ml EGF for 2 min. The cells were then fixed and stained for GAP1 IP4BP expression as described previously (39). Staining was visualized by a Leica TCS-NT confocal microscope equipped with a krypton/argon laser. The confocal was attached to a Leica DM RBE upright epifluorescence microscope.

Molecular
Modeling and Site-directed Mutagenesis of the GAP1 IP4BP PH Domain-Previously we have shown that, although the PH domains from GAP1 m and GAP1 IP4BP are ϳ63% identical at the amino acid level, they have distinct phosphoinositide-binding profiles (35). Whereas the GAP1 m PH domain is specific for PtdIns(3,4,5)P 3 , the PH domain from GAP1 IP4BP binds equally well to PtdIns(3,4,5)P 3 and PtdIns(4,5)P 2 (35). To locate the residues within GAP1 IP4BP that may be involved in interacting with PtdIns(4,5)P 2 the sequence of the PH domain of GAP1 IP4BP was aligned with the specific PtdIns(3,4,5)P 3binding PH domains from GAP1 m , ARNO, and Btk (Fig. 1). Using the coordinates from the crystal structure of the Btk PH domain in complex with the inositol head group of PtdIns(3,4,5)P 3 (25), molecular models were created for the predicted phosphoinositide-binding site within the GAP1 IP4BP and GAP1 m PH domains (Fig. 2). Given the importance of basic residues in the binding of phosphoinositides to PH domains (1, 2), these model structures allowed an examination of the distribution of the basic residues to be compared between the GAP1 IP4BP and GAP1 m PH domains. We specifically examined the location of basic residues that were present in the GAP1 IP4BP PH domain but were absent from the GAP1 m PH domain. This analysis revealed that within the GAP1 IP4BP PH domain six residues, lysine 591, arginine 604, lysine 616, lysine 641, lysine 666, and lysine 681, are present that are equivalent to non-basic residues in the GAP1 m PH domain (threonine 619, cysteine 632, proline 644, asparagine 669, asparagine 694, and asparagine 709, respectively) (Fig. 2). Of the basic residues present in the GAP1 IP4BP PH domain, lysine 641, lysine 666, and lysine 681 are located away from the phosphoinositidebinding site on the opposite corner of the PH domain structure and so are not in a position to directly interact with the bound phosphoinositide or are unlikely to interact with the membrane surface (Fig. 2). Only residues lysine 591, arginine 604, and lysine 616 are located in, or around, the proposed phosphoinositide-binding site in the model GAP1 IP4BP PH domain (Fig.  3). Lysine 591 is on the ␤1/␤2-loop in a basic run of residues, arginine 590, lysine 591, and arginine 592. In Btk the equivalent residues are all lysines, and the first of the three, lysine 17 (equivalent to arginine 590 in GAP1 IP4BP ), was shown in the crystal structure of Btk in complex with Ins(1,3,4,5)P 4 to directly interact with the 3-phosphate (25). Therefore, lysine 591 of GAP1 IP4BP is located in the correct area for potential interactions with the bound lipid. Arginine 604 and lysine 616 are located on the opposite side of the binding site. Lysine 616 is on the ␤3/␤4-loop, which is important in the PH domain of PLC-␦ 1 for binding PtdIns(4,5)P 2 (19). Based solely on the model structure, lysine 616 appears close enough for direct interaction with the bound lipid (Fig. 3). In contrast, arginine 604, which is located on the ␤2/␤3 loop, may not interact directly with the bound phosphoinositide as it is located too far from the binding site. However, it should not be overlooked that this residue may interact with the negatively charged plasma membrane surface.
Examining the Subcellular Localization of the GAP1 IP4BP Mutants-To observe the effect of the altered phosphoinositidebinding profile on the subcellular distribution of GAP1 IP4BP we initially transiently transfected HeLa cells with each mutant, examining the subcellular localization by indirect immunofluorescence (Fig. 5). Consistent with our previous observations (35,37), wild-type GAP1 IP4BP displayed a strong plasma membrane localization with very little cytosolic staining (Fig. 5). Similarly, GAP1 IP4BP -(K616P) was predominantly associated with the plasma membrane, again with very little detectable cytosolic staining (Fig. 5). In contrast, like the PtdIns(3,4,5)P 3specific binding proteins GAP1 m and ARNO (36,40), both GAP1 IP4BP -(K591T) and -(R604C) were predominantly localized to the cytosol, although a minor residual plasma membrane localization was also observed (Fig. 5). These data highlight a correlation between the affinity of these GAP1 IP4BP mutants for PtdIns(4,5)P 2 and their resultant ability to associate with the plasma membrane in serum-starved cells. Thus, whereas the GAP1 IP4BP -(K616P) retains its ability to bind with high affinity PtdIns(4,5)P 2 and to associate with the plasma membrane, the GAP1 IP4BP -(K591T) and -(R604C) mutants have a significantly reduced affinity for PtdIns(4,5)P 2 and are predominantly localized to the cytosol.
To examine the effect of PtdIns(3,4,5)P 3 production on the subcellular localization of the GAP1 IP4BP mutants we transiently co-transfected HeLa cells with the GAP1 IP4BP mutants and an expression construct encoding for p110 CAAX . This is a constitutively active phosphatidylinositol 3-kinase catalytic subunit that induces an elevation in plasma membrane PtdIns(3,4,5)P 3 . Under these conditions, as has been shown previously (36,40), both GAP1 m and ARNO became associated with the plasma membrane (Fig. 5). Interestingly, GAP1 IP4BP -(K591T) and -(R604C) also became associated with the plasma membrane under these conditions, an association that was dependent upon the catalytic activity of the p110 CAAX as incubation with the phosphatidylinositol 3-kinase inhibitor wortmannin (100 nM for 30 min) resulted in each mutant retaining their cytosolic localization (data not shown). These data show that the reduction in affinity for PtdIns(3,4,5)P 3 observed with these mutants, ϳ2and 6-fold for GAP1 IP4BP -(K591T) and -(R604C), respectively, does not affect their ability to detect the elevated plasma membrane PtdIns(3,4,5)P 3 observed in HeLa cells transiently transfected with p110 CAAX .
Effect of Growth Factor Stimulation on the Subcellular Localization of the GAP1 IP4BP Mutants-Using p110 CAAX causes a high unphysiological elevation in the levels of plasma membrane PtdIns(3,4,5)P 3 . To test the effect of more physiological levels of PtdIns(3,4,5)P 3 we transiently transfected the GAP1 IP4BP mutants into PC12 cells and examined their resultant subcellular localization prior to, and after, stimulation with epidermal growth factor. In wortmannin-treated, serumstarved PC12 cells GAP1 m and ARNO were localized to the cytoplasm whereas GAP1 IP4BP was constitutively associated with the plasma membrane (Fig. 6). Consistent with the data from HeLa cells (Fig. 5), whereas GAP1 IP4BP -(K616P) was associated with the plasma membrane, both GAP1 IP4BP -(K591T) and -(R604C) were predominantly localized to the cytosol although, as with the HeLa cells, some residual plasma membrane association was also observed (Fig. 6). Following a 2-min stimulation with 100 ng/ml of EGF, GAP1 IP4BP -(K591T) and -(R604C) became associated with the plasma membrane (Fig.  6). A similar EGF-induced plasma membrane association was also observed for GAP1 m and ARNO (Fig. 6) (36, 40). The EGF-induced plasma membrane association of GAP1 IP4BP -(K591T) and -(R604C) required the ability of the growth factor to activate phosphatidylinositol 3-kinase as membrane association was inhibited by pre-incubation with 100 nM wortmannin (data not shown). These data therefore demonstrate that the affinity of the GAP1 IP4BP -(K591T) and -(R604C) mutants for PtdIns(3,4,5)P 3 are sufficiently high to sense the changes in the plasma membrane level of this phosphoinositide following growth factor stimulation. Thus, the lowering of the affinity for PtdIns(4,5)P 2 observed in the GAP1 IP4BP -(K591T) and -(R604C) mutants results in the generation of GAP1 IP4BP mutants whose dynamic subcellular localization is regulated by PtdIns(3,4,5)P 3 in a similar manner to that of GAP1 m . DISCUSSION Previously we have described that the PH domains of the Ras GTPase-activating proteins GAP1 m and GAP1 IP4BP posses dis- , and GAP1 IP4BP -(K616P) (D) to bind PtdIns(4,5)P 2 and PtdIns(3,4,5)P 3 was analyzed using a protein-lipid overlay assay. Decreasing amounts of the relevant phosphoinositide were spotted on to a nitrocellulose membrane, which was then incubated with the purified proteins. The membranes were washed, and proteins bound to the membrane by virtue of their interaction with lipid were detected using specific antibodies. A representative of at least three separate experiments is shown. E and F show lipid binding curves for PtdIns(4,5)P 2 and PtdIns(3,4,5)P 3 , respectively, derived from the equation y ϭ B max ϫ X/ (K d ϩ X) where B max is the maximal binding, and K d is the concentration of ligand required for half-maximal binding. The data were obtained by measuring the total pixel intensity for each spot and are an average of three separate experiments.
tinct phosphoinositide-binding profiles (35). Whereas the GAP1 m PH domain is specific for PtdIns(3,4,5)P 3 , the PH domain from GAP1 IP4BP binds equally well to PtdIns(3,4,5)P 3 and PtdIns(4,5)P 2 (35). This difference in phosphoinositide specificity is manifested in the distinct subcellular localization of these proteins (36). GAP1 m is a cytosolic protein that undergoes a rapid plasma membrane association upon the agonist-stimulated production of PtdIns(3,4,5)P 3 (37). In contrast, GAP1 IP4BP is constitutively plasma membrane-associated as a direct result of its PH domain-binding PtdIns(4,5)P 2 (35,37). In the FIG. 5. The subcellular localization of wild-type GAP1 IP4BP , GAP1 m , ARNO, and the various site-directed mutants of GAP1 IP4BP in HeLa cells with and without co-transfection of p110 CAAX . Expression vectors encoding GFP-GAP1 IP4BP , GFP-GAP1 m , and GFP-ARNO and the various GAP1 IP4BP mutants were transiently co-transfected with or without p110 CAAX into HeLa cells. 24 h after transfection the cells were serum-starved for 2 h and fixed, and GAP1 IP4BP , GAP1 m , and ARNO were detected by confocal microscopy. The GAP1 IP4BP mutants were detected by indirect immunofluorescence using GAP1 IP4BP -specific antiserum as described under "Experimental Procedures." The white arrows highlight the residual plasma membrane association observed with the GAP1 IP4BP -(R604C) and -(K591T) mutants. It should also be noted that the GAP1 IP4BP -(R604C) had an increased presence in the nucleus of serum-starved cells. This was, however, absent from cells co-transfected with p110 CAAX . Image analysis was performed by measuring the intensities according to pixel brightness along cell transepts on the depicted cells. Similar data were observed in an additional 10 cells for each construct. current study we have used molecular modeling to identify residues within the GAP1 IP4BP PH domain that are required for high affinity binding to PtdIns(4,5)P 2 . We have highlighted three residues, lysine 591, arginine 604, and lysine 616, as potentially providing the necessary interactions for PtdIns (4,5)P 2 binding. By mutating each residues into the equivalent residue in GAP1 m we have shown that whereas the GAP1 IP4BP -(K616P) mutant binds PtdIns(4,5)P 2 and PtdIns(3,4,5)P 3 with affinities comparable with wild-type, GAP1 IP4BP -(R604C) has a 6-fold reduction in its affinity for PtdIns(4,5)P 2 and FIG. 6. The subcellular localization of wild-type GAP1 IP4BP , GAP1 m , ARNO, and the GAP1 IP4BP mutants in PC12 cells treated with either 100 nM wortmannin or 100 ng/ml EGF. Expression constructs encoding GFP-GAP1 IP4BP , -GAP1 m , and -ARNO, and the various GAP1 IP4BP mutants were transiently transfected into PC12 cells. 24 h after transfection cells were serum-starved for 2 h prior to incubation with either 100 nM wortmannin (37°C for 30 min) or 100 ng/ml EGF (37°C for 2 min). Wild-type GAP1 IP4BP , GAP1 m , and ARNO were then fixed, mounted, and detected by confocal microscopy. The various GAP1 IP4BP mutants were fixed and detected by indirect immunofluorescence as described under "Experimental Procedures." Again residual plasma membrane association was observed with the GAP1 IP4BP -(R604C) and -(K591T) mutants, and GAP1 IP4BP -(R604C) had an increased presence in the nucleus of serum-starved, wortmannin-treated cells. Image analysis was performed by measuring the intensities according to pixel brightness along cell transepts on the depicted cells. Similar data were observed in an additional 10 cells for each construct. PtdIns(3,4,5)P 3 . In contrast, GAP1 IP4BP -(K591T) has an affinity for PtdIns(4,5)P 2 that is reduced 6-fold, whereas the affinity for PtdIns(3,4,5)P 3 is reduced only 2-fold. Through the isolation of GAP1 IP4BP -(K591T) we have therefore engineered a PH domain that displays a phosphoinositide-binding profile reminiscent of that observed for the corresponding domain from GAP1 m . Such a conclusion is further supported by our examination of the subcellular localization of the GAP1 IP4BP mutants. Whereas GAP1 IP4BP -(K616P) retains the constitutive plasma membrane association observed for wild-type GAP1 IP4BP , the GAP1 IP4BP -(R604C) and -(K591T) mutants are no longer capable of associating with the plasma membrane. Rather, consistent with their reduced affinity for PtdIns(4,5)P 2 , these proteins are like GAP1 m in being located within the cytosol. Importantly, following growth factor stimulation GAP1 IP4BP -(K591T) becomes associated in a PtdIns(3,4,5)P 3 -dependent manner with the plasma membrane. These data emphasize that subtle changes in the phosphoinositide-binding profile of the GAP1 IP4BP PH domain can have a pronounced effect on the subcellular distribution and regulation of this protein.
Apart from the study described here, the dramatic effects that subtle changes can have on phosphoinositide binding is most apparent when considering the diglycine versus triglycine forms of Grp1 (39,41), a member of the cytohesin family of ADP-ribosylation factor nucleotide exchange factors. Here, the addition of a single glycine residue to the PH domain can have a significant effect on the phosphoinositide-binding specificity; the triglycine form shows much less discrimination between PtdIns(4,5)P 2 and PtdIns(3,4,5)P 3 than the diglycine form. How the extra glycine causes the increase in affinity for PtdIns(4,5)P 2 is difficult to comprehend. One possibility is that the third glycine opens up the binding pocket, allowing the PtdIns(4,5)P 2 to enter in an orientation, possibly more akin to the PtdIns(4,5)P 2 -binding site in the PH domain of PLC-␦ 1 (19), that is sterically hindered in the diglycine form (42,43). It is tempting to extrapolate this idea and speculate that although PtdIns(3,4,5)P 3 may bind to the GAP1 IP4BP PH domain in a similar orientation to that observed in Btk (25), the binding of PtdIns(4,5)P 2 may be similar to that observed in PLC-␦ 1 whereby the PtdIns(4,5)P 2 molecule is rotated 180 o through the P1/P4 axis of the inositol ring (19). This could lead to lysine 591 having a significant role in the binding of PtdIns(4,5)P 2 , but because of the inositol head group orientation may not play such a significant role in PtdIns(3,4,5)P 3 binding.
In summary therefore, through the use of molecular models we have successfully predicted residues that allow the PH domain from GAP1 IP4BP to bind PtdIns(4,5)P 2 with high affinity. We have demonstrated that a single targeted site-directed mutant, GAP1 IP4BP -(K591T), while retaining its ability to bind PtdIns(3,4,5)P 3 with high affinity, has a specific reduction in its affinity for PtdIns(4,5)P 2 . Furthermore, this subtle alteration in phosphoinositide binding is translated into a dynamic change in the regulation of the subcellular localization of the mutant protein following growth factor stimulation.