Using large unilamellar vesicles containing 5 mol% PtdIns(3,4)P₂, we found that simultaneous addition of IP₆ was able to prevent the binding of C-PH to PtdIns(3,4)P₂. As shown in Figure 1, the binding of C-PH to PtdIns(3,4)P₂ was markedly inhibited by addition of 10 μM IP₆ and completely abolished by 100 μM IP₆. Only trace non-specific binding of C-PH to phosphatidylcholine was observed (Figure 1, lane 2). Using a range of IP₆ concentrations, the apparent IC₅₀ of the IP₆ used was determined to be 7.5 μM.

To investigate further the nature of the myo-inositol polyphosphate interaction with C-PH, crystals were prepared using recombinant C-PH and IP₆, which permitted the determination of a high-resolution crystal structure. Crystals of the complex were grown by the hanging drop vapor diffusion method after incubation of C-PH with 1 mM IP₆ (see materials and methods). The resulting crystals were of exceptional quality and diffracted to 1.35 Å. Data collection and model refinement statistics are presented in Table 1. The structure was solved by molecular replacement using a search model based on the previously determined C-PH crystal structure solved in the absence of ligand (PDB code: 1ZM0). The final model (Figure 2) was well ordered with the exception of an approximately eight-residue loop region between the fifth and sixth beta strands, corresponding to residues 303–310. Due to the complete disorder of these residues they could not be built into the final model. The overall structure of C-PH containing bound ligand maintained the folding pattern that is characteristic of the PH domain family. Thus, the inositol phosphate ligand was found bound in a loop region of C-PH formed between the first and second beta strands which has been well-characterized in a number of PH domains as the phosphoinositide binding region. Examination of the electron density in this region revealed that the bound D-myo-inositol polyphosphate contained only 5 phosphate groups rather than the 6 found in IP₆. Model building of the different possible D-myo-inositol pentakisphosphates into the electron density indicated the identity of the bound ligand to be Ins(1,2,3,5,6)P₅, as no other isomer fits the electron density. The identity of the bound ligand was determined with a high degree of confidence, based on the high-resolution data (1.35 Å), the well-ordered electron density in the binding loop and the fact that only the phosphate at the 2 position on the inositol ring adopts an axial conformation. The presence of Ins(1,2,3,5,6)P₅ was unexpected as a commercial source of IP₆ was used in the crystallization experiments. However, if the IP₆ sample contained IP₆ hydrolysis products this could account for the presence of Ins(1,2,3,5,6)P₅ in the final C-PH structure. This unexpected finding suggests that C-PH has a much higher affinity for Ins(1,2,3,5,6)P₅ than for IP₆. In support of this view, a separate study 21 showed that the IC₅₀ for the inhibition of binding of PtdIns(3,4)P₂ to C-PH by Ins(1,2,3,5,6)P₅ was 2.2 μM. HPLC analysis of the IP₆ sample used in these studies using a reverse-phase C₁₈ column (Vydac 218TP54 RP column) revealed the IP₆ sample was in fact not pure. Conditions used to generate IP₅-C-PH co-crystals were mild and therefore not likely to have resulted in conversion of IP₆ to Ins(1,2,3,5,6)P₅. Regardless of the source, Ins(1,2,3,5,6)P₅ was found tightly bound within the phosphoinositide binding pocket of C-PH and provided important information on the binding of D-myo-inositol polyphosphates to PH domains.

The final model of IP₅-bound C-PH was determined at exceptionally high resolution and consequently the interactions observed between ligand and protein (Figure 3) can be reported with a high degree of certainty. Table 2 lists the specific interactions and distances observed within the crystal structure between Ins(1,2,3,5,6)P₅ and C-PH. Ins(1,2,3,5,6)P₅ was bound to C-PH in the β1–β2 loop region (residues 253–264 and 277), making numerous interactions through all but one of its phosphate groups (Figure 3). The only phosphate group that does not interact with any residues of C-PH is that in the 1-position of myo-inositol. Although exposed to solvent, this phosphate remained ordered to the point of generating clear electron density. All amino acids that form stabilizing interactions with Ins(1,2,3,5,6)P₅ are located in the β1–β2 loop region with the exception of Y277 which contributes a hydrogen bond and is situated on the β3 strand of C-PH. The 2-phosphate is the only phosphate that adopts an axial position. This configuration is strongly stabilized through interactions with the side chains of R264 and K253. Unlike the 2-phosphate, the 3-phosphate makes only a single interaction, in this case with Y277. As already mentioned, the 4-phosphate is not present in the structure. However the remaining 4-OH group, which represents the location where the additional phosphate group of IP₆ would be, does make an interaction with the side chain of H256 at a distance of 3.5 Å. All attempts to model in a phosphate group at this position resulted in steric clashes with H256 and the main chain of residues R257 or G255, depending on the position of the phosphate chosen. In addition, the electron density for H256 is well-ordered indicating that this side chain is not mobile and could not adopt a conformation that would accommodate an additional phosphate group at the 4-position. The 5-phosphate of Ins(1,2,3,5,6)P₅ interacts with main chain atoms of H256, R257 and N260 (water-mediated), as well as the side chain of R257. The final position on the inositol ring is occupied by the 6-phosphate and interacts with the side chain of R258 in addition to the main chain of N260 through a water molecule. C-PH residues shown here to interact with Ins(1,2,3,5,6)P₅ are also known to be involved in binding PtdIns(3,4)P₂ 18. This demonstrates that Ins(1,2,3,5,6)P₅ competes directly with PtdIns(3,4)P₂ for binding to C-PH. While it is not known exactly how PtdIns(3,4)P₂ interacts with residues in the binding loop, it seems reasonable that two of the phosphates from Ins(1,2,3,5,6)P₅ will directly compete with the phosphates from PtdIns(3,4)P₂ while the remaining phosphates will provide additional stabilizing interactions.

Although the structure presented here represents the first high resolution structure of Ins(1,2,3,5,6)P₅ bound to a PH domain, the structure of another PH domain complexed with Ins(1,3,4,5,6)P₅ has been reported 22. The PH domain of Grp1 binds 3-phosphoinositides allowing the protein to participate in regulating the actin cytoskeleton in response to phosphoinositide-3-kinase signaling pathways. As such, the structure of the PH domain of Grp1 complexed with Ins(1,3,4,5,6)P₅ represents another instance in which specific interactions of an inositol phosphate and the phosphoinositide binding cleft of a PH domain might be expected to regulate the association of the PH domain with a specific phosphoinositide. Comparison of the Grp1 PH domain complexed with Ins(1,3,4,5,6)P₅ (PDB code: 1FHW) with that of C-PH bound to Ins(1,2,3,5,6)P₅, revealed some intriguing similarities and differences (Figure 4). Although both D-myo-inositol pentakisphosphates are bound in the β1–β2 loop region, they adopt slightly different orientations with respect to their inositol rings (Figure 4). Remarkably, despite this difference in the overall orientation of the inositol ring, four of the five phosphate groups from each ligand occupy overlapping positions in the β1–β2 binding cleft. The 2-phosphate of Ins(1,2,3,5,6)P₅ occupies the same position as the 3-phosphate from Ins(1,3,4,5,6)P₅. In both structures, the phosphate group at this position interacts with a highly conserved arginine residue (R264 in C-PH and R284 in Grp1). The arginine at this position is conserved amongst a number of PH domains including that of Bruton's tyrosine kinase (Btk), in which its mutation to either histidine or proline causes the hereditary immune disease X-linked agammaglobulinaemia 23. In the crystal structure of the Btk PH domain bound to Ins(1,3,4,5)P₄, it was shown that the conserved arginine (R28) interacts with the 3-phosphate, suggesting that mutation of this residue decreases the affinity of the PH domain for its natural ligand, PtdIns(3,4,5)P₃ 24.

The second overlapping phosphate position is occupied by the 6 and 1-phosphates of Ins(1,2,3,5,6)P₅ and Ins(1,3,4,5,6)P₅, respectively. In the C-PH structure, the 6-phosphate interacts with R258, a residue that is not conserved in the Grp1 PH domain. The corresponding 1-phosphate of Ins(1,3,4,5,6)P₅ interacts with K282 in the Grp1 PH domain which is equivalent to K262 in C-PH, although this residue does not interact with the 6-phosphate of Ins(1,2,3,5,6)P₅ due to the location of its side chain. The 5-phosphate of Ins(1,2,3,5,6)P₅ and the 6-phosphate of Ins(1,3,4,5,6)P₅ occupy the third overlapping position within the phosphoinositide-binding pocket. The 5-phosphate is in position to interact with the main chains of H256, R257 and N260 as well as the side chain of R257 in C-PH. Although this arginine is conserved in the Grp1 PH domain (R277), it does not interact with the 6-phosphate, in fact the 6-phosphate makes less optimal interactions with main chain residues compared to those observed in the C-PH structure. In the fourth and final overlapping position, the 3-phosphate of Ins(1,2,3,5,6)P₅ and the 4-phosphate of Ins(1,3,4,5,6)P₅ both interact with a conserved tyrosine at position 277 in C-PH and 295 in the Grp1 PH domain. In addition to this, the 4-phosphate makes further stabilizing contacts with the side chains of H355 and K273, interactions that are not observed in the C-PH/Ins(1,2,3,5,6)P₅ structure.

From this structural comparison it can be seen that in both cases the D-myo-inositol pentakisphosphate ligands interact with conserved side chains that would be involved in binding phosphoinositides. This suggests that the D-myo-inositol pentakisphosphates would directly compete with phosphoinositides for binding to the PH domains. In addition to this, it was found that four out of the five phosphates occupy overlapping positions within the binding cleft despite the differences in orientation and even stereochemistry. This overlap in phosphate positions is not simply coincidence, since a structural alignment of various PH domains solved in complex with either Ins(1,3,4,5)P₄ or the two D-myo-inositol pentakisphosphates used in the comparison above reveals that the overlapping positions are conserved in these structures as well (Figure 5). The mode of interaction observed in our structure is therefore consistent with the possibility that myo-inositol polyphosphates could act to regulate PH domain-phosphoinositide associations by directly competing with phosphoinositides for binding to PH domains.

The physiological significance of the interaction reported here with respect to platelet signaling pathways is currently unknown and calls for further careful examination. With the exception of IP₃, inositol phosphates have only been thought to function with minor roles as second messengers 25. However, as more information becomes available it appears as though these molecules have a much broader role in signaling then was originally appreciated 2627. The most well-characterized inositol phosphate second messenger is IP₃, produced by phospholipase C-mediated hydrolysis of PtdIns(4,5)P₂. In addition to IP₃, it has also been proposed that Ins(1,3,4,5)P₄ plays a role in the regulation of cellular Ca⁺² fluxes 27. The IP₅ family of D-myo-inositol pentakisphosphates represents one of the most abundant forms of inositol phosphates present in mammalian cells. Although six possible isomers exist, Ins(1,3,4,5,6)P₅ represents the predominant form observed in mammalian cells, including platelets 27282930. Other IP₅ isomers including Ins(1,2,4,5,6)P₅ and Ins(2,3,4,5,6)P₅ have been detected in a number of cell lines 2931, demonstrating that these molecules are indeed present in mammalian cells and as such could play roles in regulating biological functions. In agreement with this, several recent studies have provided evidence suggesting specific roles for IP₅ isomers ranging from the regulation of chromatin remodeling 3233 to Salmonella pathogenesis 3435. In further support of this idea, Ins(1,3,4,5,6)P₅ was recently shown to be capable of promoting apoptosis and to possess antiangiogenic and antitumor effects as a result of its ability to inhibit the phosphoinositide 3-kinase/Akt signaling pathway 3637. This inhibition was shown to be a consequence of Ins(1,3,4,5,6)P₅ competing with PtdIns(3,4,5)P₃ for binding to the protein kinase B (Akt) PH domain, ultimately preventing the phosphorylation and activation of Akt. Based on these intriguing results the authors have proposed that Ins(1,3,4,5,6)P₅ and potentially other inositol phosphates can be used as experimental or possibly therapeutic tools that interfere with the binding of PH domains involved in signaling pathways. Inositol pyrophosphates, specifically IP₇ and IP₈, have also been shown to act as signaling molecules that regulate PH domain interactions with phosphoinositides 38. In this study IP₇ and IP₈ were shown to compete with PtdIns(3,4,5)P₃ for binding to several PH domain-containing proteins in Dictyostelium, resulting in an effect on chemotaxis. As the body of evidence continues to grow, it appears that PH domains will emerge as versatile domains capable of mediating interactions with a range of different ligands.

It therefore seems likely that as more studies are aimed at examining the roles of IP₅ isomers, it will become apparent that this family of inositol phosphates is involved in signaling pathways that affect a range of physiological processes. The identification of PH domains capable of interacting specifically with the various IP₅ isomers will aid in characterizing the roles of the latter. However, we are not aware of any report of the presence of Ins(1,2,3,5,6)P₅ in platelets or other cells at the present time. It will therefore be important to investigate the effects of other IP₅ isomers on the binding of pleckstrin C-PH domain to PtdIns(3,4)P₂. It has not escaped our attention that in addition to acting as antagonists of pleckstrin binding by specific phosphoinositides, myo-inositol polyphosphates with high affinities for C-PH could act as regulators of the interaction of pleckstrin with other proteins.

The binding assays were performed as described in 19. Briefly, 0.3 μM recombinant C-PH was incubated with large unilamellar vesicles (LUVs) consisting of phosphatidylcholine supplemented with 5 mol% PtdIns(3,4)P₂. Simultaneous additions of IP₆ (Aldrich) (0–100 μM) were used to inhibit the binding of C-PH to the PtdIns(3,4)P₂. LUVs with bound protein were isolated by ultracentrifugation, analyzed by SDS-PAGE using 10% acrylamide and immunoblotted with antibody to the C-terminal 13 residues of pleckstrin. A logit plot was used to calculate the apparent IC₅₀ of IP₆ (y = 100[1 + (x/IC₅₀)^s]^-1, where y is the % of added C-PH bound at the inhibitor concentration x and s is a slope factor).

C-PH (pleckstrin amino acids 240–350) used for LUV binding studies was expressed and purified as described previously 1920. For crystallography, C- and N-terminal residues previously found to lack structure 20 were omitted. Thus, the carboxy terminal PH domain (amino acids 243–347) of human pleckstrin was expressed and purified as a hexahistidine fusion protein in E. coli BL21(DE3) cells using the pDEST17 expression vector (Invitrogen). Cells were grown in standard LB medium supplemented with 10 mg ml^-1 ampicillin at 310 K with shaking (225 rev min^-1) until the light absorbance at 600 nm reached 0.5. Protein expression was then induced using 0.1 mM IPTG and the incubation temperature was lowered to 300 K. After a 5-hour induction period cells were harvested by centrifugation at 4 000 rpm for 10 minutes at a temperature of 277 K. Each 1L cell pellet was then resuspended in 8 ml of 1× phosphate buffered saline (PBS) and centrifuged at 3 500 rpm for 10 minutes at 277 K. The resulting cell pellets were then frozen in liquid nitrogen and stored at 193 K. Prior to cell lyses using high pressure, cell pellets (2 L) were resuspended to a final volume of 35 ml using nickel A buffer (20 mM Hepes pH 7.5, 0.5 M NaCl, 0.06% LDAO, 2 mM β-mercaptoethanol and 10 mM imidazole). Following cell lyses samples were centrifuged at 20 000 rpm for 40 minutes at 277 K and the resulting supernatant was applied to a HiTrap Nickel affinity column (Amersham Biosciences). The C-PH was eluted from the column with 210 mM imidazole following sequential washes with 15 and 30 mM imidazole which was accomplished by mixing appropriate volumes of nickel A buffer and nickel B buffer (20 mM Hepes pH 7.5, 0.5 M NaCl, 0.06% LDAO, 2 mM β-mercaptoethanol and 300 mM imidazole). The protein sample was then diluted in S-A buffer (10 mM Hepes pH 7.5, 1 mM dithiothreitol and 1 mM EDTA) to lower the salt concentration to approximately 75 mM. This sample was then applied to a HiTrap SP Sepharose HP ion exchange column (Amersham Biosciences) to further purify the protein from any remaining contaminants. The C-PH was subsequently eluted using a gradient of increasing salt concentration generated by the application of S-B buffer (10 mM Hepes pH 7.5, 1 mM dithiothreitol, 1 mM EDTA and 0.5 M NaCl). The hexahistidine tag was removed from the C-PH by cleavage with TEV protease. This results in four residues (Gly, Ser, Phe and Thr) being left N-terminal to the first residue in C-PH. Of these four amino acids only the phenylalanine and threonine residues could be reliably built into the available electron density. In order to purify C-PH from the cleaved hexahistidine tag and the TEV protease the digested sample was re-applied to the HiTrap SP Sepharose column and purified as described above. The resulting C-PH sample was buffer exchanged into the final crystallization buffer (5 mM pottasium glutamate, 120 mM sodium glutamate, 20 mM Hepes pH 7.5, 2.5 mM EDTA, 2.5 mM EGTA and 3.15 mM MgCl₂) using a HiPrep 26/10 Desalting column (Amersham Biosciences). The resulting protein sample was concentrated to 4.0 mg ml^-1 as determined by the Bradford assay, using a centrifugal filter. SDS-PAGE analysis indicates that C-PH prepared by this method is greater then 95% pure.

C-PH (2.5 mg ml^-1) was crystallized in the presence of 1 mM IP₆ (Aldrich) using the hanging drop vapour diffusion method under the following conditions. A 2 μl drop containing 1 μl of C-PH (2.5 mg ml^-1) and 1 mM IP₆ in 5 mM potassium glutamate, 120 mM sodium glutamate, 20 mM Hepes pH 7.5, 2.5 mM EDTA, 2.5 mM EGTA and 3.15 mM MgCl₂ and 1 μl of 0.1 M Bis-Tris pH 6.5 and 28% polyethylene glycol 2 000 monomethyl ether was suspended over 500 μl of 0.75 M ammonium sulfate and incubated at 293 K. Rectangular rod-shaped crystals grown in these conditions were harvested and crushed to generate micronuclei that were subsequently used to streak seed into the same crystallization condition as well as slight variations of this condition. This resulted in large rectangular rod-shaped crystals with dimensions of approximately 400 × 100 × 100 μm. The crystal used to collect data was grown in 0.1 M Bis-Tris pH 6.5 and 28% polyethylene glycol 2 000 monomethyl ether. Prior to cryocooling in a nitrogen cold stream, the crystal was soaked for approximately 10 seconds in a cryoprotectant containing 0.1 M Bis-Tris pH 6.5, 28% polyethylene glycol 2 000 monomethyl ether and 17% glycerol. Both high (1.35 Å) and low (1.5 Å) resolution data sets were collected independently for this crystal and then merged to yield a single data set. The data was collected at a wavelength of 1.0 Å at beamline X25 of the Brookhaven National Laboratory using a ADSC Q315 CCD x-ray detector and processed using the HKL2000 program suite 39.

The crystal structure of the C-PH in complex with Ins(1,2,3,5,6)P₅ was solved by molecular replacement using the program MOLREP 40. The search model used in molecular replacement was the crystal structure of the C-PH (PDB code 1ZM0), which was solved in the absence of any ligand. Iterative cycles of model building and refinement were carried out using the programs WinCoot 41 and Refmac5 4243 respectively. All figures describing protein structures presented were generated using the molecular graphics program PyMol 44.