|Home | About | Journals | Submit | Contact Us | Français|
Inositol pyrophosphates (such as IP7 and IP8) are multifunctional signaling molecules that regulate diverse cellular activities. Inositol pyrophosphates have ‘high-energy’ phosphoanhydride bonds, so their enzymatic synthesis requires that a substantial energy barrier to the transition state be overcome. Additionally, inositol pyrophosphate kinases can show stringent ligand specificity, despite the need to accommodate the steric bulk and intense electronegativity of nature’s most concentrated three-dimensional array of phosphate groups. Here we examine how these catalytic challenges are met by describing the structure and reaction cycle of an inositol pyrophosphate kinase at the atomic level. We obtained crystal structures of the kinase domain of human PPIP5K2 complexed with nucleotide cofactors and either substrates, product or a MgF3− transition-state mimic. We describe the enzyme’s conformational dynamics, its unprecedented topological presentation of nucleotide and inositol phosphate, and the charge balance that facilitates partly associative in-line phosphoryl transfer.
The inositol phosphates initially came to prominence as second messengers following the discovery that inositol 1,4,5-trisphosphate (IP3) releases Ca2+ from the endoplasmic reticulum1. Since then, it has become widely recognized that IP3 is metabolized to form a large number of additional inositol phosphates that also function as cellular signals2. More recently, the size of this signaling family was further expanded with the discovery that some of its members contain diphosphate groups3,4. These so-called inositol pyrophosphates include diphosphoinositol penta-kisphosphate (IP7) and bisdiphosphoinositol tetrakisphosphate (IP8). There is now evidence that IP7 and IP8 regulate many cellular processes in eukaryotes, including apoptosis, vesicle trafficking, cytoskeletal dynamics, exocytosis, insulin signaling and neutrophil activation (reviewed in refs. 5–7 and also refs. 8,9).
The presence of diphosphate groups is a key, functionally important feature of the inositol pyrophosphates. It is the diphosphates that facilitate the ability of these molecules to compete with phosphatidylinositol-3,4,5-trisphosphate for binding to pleckstrin homology domains8,9. This phenomenon seems to explain how IP7 can antagonize signaling by insulin9 and also regulate neutrophil activity8. There has also been one description of a more classical IP7 ‘receptor’: the Pho80–Pho85–Pho81 cyclin-dependent kinase complex10. IP7 augments the ability of Pho81 to inhibit kinase activity, and the ligand’s diphosphate group contributes to the specificity of the interaction10. The diphosphate groups can also be used for the nonenzymatic phosphorylation of a range of different proteins11. This particular mechanism may help explain the multifunctionality of inositol pyrophosphate signaling.
The enzymatic synthesis of inositol pyrophosphates requires the formation of high-energy phosphoanhydride bonds12, even though they are in a class of molecules that is unrivaled throughout nature in its degree of phosphate congestion. Thus, the active site of an inositol pyrophosphate kinase must accommodate considerable steric bulk and intense electronegativity yet retain selective substrate specificity13,14. Moreover, this group of kinases must also overcome a substantial energy barrier to the transition state. There are two groups of enzymes that synthesize inositol pyrophosphates: inositol hexakisphosphate kinases 1, 2 and 3 (IP6K1, IP6K2 and IP6K3; ref. 15) and diphosphoinositol pentakisphosphate kinases 1 and 2 (PPIP5K1 and PPIP5K2; refs. 13,16); in yeasts, these groups of proteins are named Kcs1 (ref. 17) and Vip1 (ref. 18), respectively. These two groups of enzymes collaborate in the synthesis of IP8 through two complementary metabolic pathways. In one, IP6 is phosphorylated by IP6K, forming an isomer of IP7 with a diphosphate group at the 5 position (that is, 5-IP7)19. Next, 5-IP7 is further phosphorylated by PPIP5K to form IP8 (ref. 20). In the alternate pathway, IP6K adds a 5-phosphate to a second isomer of IP7, which is formed from IP6 by PPIP5K19. Previous analysis of inositol pyrophosphate structures19 have determined that this alternative IP7 isomer has a diphosphate at either the 1 or 3 position; the ambiguity reflects difficulties in discriminating between alternative products that are enantiomers. As it has not hitherto been demonstrated whether PPIP5Ks are 1- or 3-kinases, the exact structure of IP8 has also not previously been established.
The PPIP5K enzymes are of particular interest for exercising stimulus-dependent spatiotemporal control over inositol pyrophosphate synthesis. The activities of the N-terminal kinase domains of these enzymes (PPIP5K1KD and PPIP5K2KD) are modulated in response to certain cellular stresses7,10, whereas separate inositol-lipid binding domains (PPIP5K1PBD and PPIP5K2PBD) mediate the receptor-regulated modulation of the intracellular compartmentalization of these enzymes14.
In this study, we use a structural and biochemical approach to determine how the catalytic challenges of pyrophosphorylation are met. To this end, we enlisted human PPIP5K2 as a representative member of the PPIP5K family. We report nine crystal structures and mutagenesis data on PPIP5K2 that define the basis for substrate recognition and selectivity, establish the reaction performed by the PPIP5Ks and the identity of the product, and demonstrate the reaction coordinate for creation of a high-energy pyrophosphate bond in IP7 and IP8.
We first determined crystal structures of hPPIP5K2KD, including residues 37–366 or 41–366 (Supplementary Methods), which comprise the core catalytic domain of this particular protein; the length of the construct was varied to optimize crystallization (data not shown and ref. 16). These structures (Fig. 1a,b and Supplementary Results, Supplementary Table 1) reveal that the nucleotide nestles between two sets of antiparallel β-sheets, confirming the predictions13,18 of homology modeling servers that hPPIP5-K2KD uses an ATP-grasp fold. The remainder of the hPPIP5K2KD structure, mainly the N terminus, forms an αβα domain (Fig. 1a and Supplementary Fig. 1a). The structure of hPPIP5K1KD (84% sequence identity13) is most likely very similar to that of hPPIP5-K2KD; dissimilar residues are confined to the surface distal to the active site (Supplementary Fig. 1b).
Detailed interactions between nucleotide and protein are shown in Supplementary Figure 2a. Nucleotide binding may stabilize protein conformation, as we could not crystallize the apoenzyme. Nevertheless, hPPIP5K2KD hydrolyzed ATP at only a low rate (0.5 nmol mg protein−1 min−1) until a physiologically relevant substrate, 5-IP7, was also added; the rate of ATP hydrolysis then increased to 400 nmol mg protein−1 min−1. The Dali server21 identified hITPK1, another ATP-grasp inositol kinase (Fig. 1c,d), as the structure that is most homologous to hPPIP5K2KD, although they share only 11% sequence identity. Furthermore, β11 in the ATP-grasp region of hPPIP5K2KD (Fig. 1a) does not occur in ITPK1 (Fig. 1c,d). Other regions of hPPIP5K2KD also show considerable structural differences from ITPK1. For example, an insert unique to hPPIP5K2KD forms an additional three-stranded β sheet (β5, β6 and β7; Fig. 1). Similar insertions are also present in the sequences of hPPIP5K1 and the yeast homolog ScVip1.
The ATP-grasp superfamily is noted for its considerable functional and structural divergence at the active site22. Indeed, the composition of the active site of ITPK1 (refs. 23,24) differs considerably from that of hPPIP5K2KD (Fig. 2). For example, hPPIP5K2KD uses only arginine and lysine residues to bind substrate, with the exception of Ser326 (Figs. 2a and and3a3a and Supplementary Fig. 3). The inositol phosphate binding pocket in EhITPK1 makes more general use of smaller and nonbasic residues such as glycine and glutamine23 (Fig. 2b). The shapes of the two binding pockets are also substantially different (Fig. 2). EhITPK1 uses a relatively wide but flattened binding pocket that shows little substrate specificity23. hITPK1, which phosphorylates a smaller range of substrates24, has a binding pocket that is somewhat smaller but still loosely defined. In contrast, the highly specific hPPIP5K2KD binds its substrates in two near-parallel grooves (marked with black arrows in Fig. 2c) that form a staggered ‘H’ shape. This binding pocket makes a perfectly tailored aperture for accommodating six phosphate or pyrophosphate groups attached to an inositol ring in the chair conformation. The biological and mechanistic consequences of the catalytic specialization of hPPIP5K2KD are discussed below.
Substrate specificity for PPIP5Ks is biologically important in that it limits the variety of inositol pyrophosphate isomers that can be synthesized. As with hPPIP5K1 (refs. 13,14), we have found that the only important substrates for hPPIP5K2 are 5-IP7 and IP6 (the first-order rate constants are 346 ± 15 μg−1 min−1 and 16 ± 1 μg−1 min−1, respectively; n = 3, mean ± s.e.m.). There was negligible phosphorylation of 1-IP7 (0.09 ± 0.01), IP5 (0.028 ± 0.003) and PP-IP4 (0.75 ± 0.07) which, among the naturally occurring inositol phosphates, are the most similar to 5-IP7 and IP6. The stringent substrate specificity of hPPIP5K2 is achieved by the array of positively charged lysine and arginine residues lining the substrate binding pocket, which makes polar contacts with every phosphate or pyrophosphate group in both IP6 and 5-IP7, and by the Mg2+ ions that bridge certain phosphate groups (Figs. 2c and and3a3a and Supplementary Fig. 3). Lys214 and the magnesium atom Mg5 interact specifically with the 5-β-phosphate of 5-IP7; the absence of the 5-β-phosphate in IP6 (Fig. 3a) may contribute to hPPIP5K2’s preference for 5-IP7 as a substrate (see above). To further probe inositol phosphate recognition, we mutated basic residues lining the substrate binding pocket and assessed the effects of these mutations on catalytic activity against both IP6 and 5-IP7 (Fig. 3b). Most mutants showed sharply diminished activities (to less than 10% of wild-type enzyme activity). Less dramatic effects were observed for K53A and R273A mutations (25–50% of wild-type activity), as the original residues are more exposed to solvent. The K248A mutant showed the largest decrease in catalytic activity with either substrate (Fig. 3b). This lysine residue interacts with both the 1-phosphate of the substrates and the γ-phosphate of ATP throughout the catalytic cycle (Fig. 4; described further below).
Some inositol phosphate kinases (ITPK1 and IPMK) phosphorylate multiple positions around the inositol ring25. The structural rationalization of this apparent promiscuity23,26 is that the loosely fitting active sites of these kinases allow several alternate substrates to bind in different orientations. In contrast, our structures reveal that IP6 and 5-IP7 fit into the active site of hPPIP5K2KD in a near-superimposable orientation (Fig. 3a and Supplementary Fig. 4). Thus, for both substrates it is the 1-phosphate that is closest to the γ-phosphate of ATP (Fig. 4a,b and Supplementary Fig. 4). Moreover, when hPPIP5K2KD crystals were soaked with ATP and 5-IP7 substrate, we obtained complexes that contained ADP and product (Fig. 4c), indicating that the crystallized enzyme retained its kinase activity. A difference Fourier map identified the product as 1,5-IP8 (Fig. 4c and Supplementary Fig. 5). Thus, PPIP5Ks are 1-kinases. Hitherto, it has not been determined whether PPIP5Ks phosphorylate the 1-position, the 3-position or both19. These two alternative reactions would yield products that are enantiomers, which have previously been difficult to distinguish between.
Induced-fit motion also contributes to specificity27. A comparison of the AMP-PNP–enzyme crystal complexes with or without 5-IP7 indicates that substrate binding is accompanied by induced-fit motion in hPPIP5K2KD of the side chains of three active-site residues: Arg262, Arg281 and Lys329 (Fig. 4d, Supplementary Fig. 6 and Supplementary Movies 1 and 2). The use of these flexible side chains enables the energy cost upon ligand binding to be lower than it would be if backbone rearrangement was necessary27.
To gain direct insight into the reaction mechanism, we obtained a crystal complex that contains 5-IP7, ADP and MgF3− (Fig. 4), as the latter matches the charge and geometry of the transition state of a phosphoryl transfer reaction28. Indeed, in the active site, we detected MgF3− with near planar geometry (Fig. 4b and Supplementary Fig. 5d) that mimics a trigonal bipyramidal phosphoryl transition state28. The magnesium atom was sandwiched 1.9 Å and 2.3 Å from the donor oxygen of ADP and the acceptor oxygen of the 1-phosphate of 5-IP7, respectively (Fig. 4b). For this total coordination distance of <4.9 Å, a new P-O bond can be formed by nucleophilic attack of the acceptor oxygen before the original P-O bond with the donor oxygen is broken, suggesting an in-line associative reaction mechanism29. The three negative charges of the pentacoordinate (phosphorane) transition state seem to be balanced by electrostatic interactions with two Mg2+ ions, a water molecule, and the positively charged side chains of Lys248 and Arg213. This snapshot of charge neutralization in the transition state (Fig. 4b and Supplementary Fig. 7) helps explain how hPPIP5K2KD overcomes the energy barrier to efficient catalysis.
Catalysis can also be enhanced when the geometry of the substrate is directed toward that of the transition state by conformational dynamics30. A comparison between the substrate-bound ground and transition states of hPPIP5K2KD (Fig. 4e and Supplementary Movies 1 and 2) revealed that Lys329 not only alters its conformation during this transition but also forms different contacts and gains two additional hydrogen bonds. Arg262 gains one hydrogen bond with 5-IP7 in the transition state. The entire inositol ring moves under the concerted actions of Lys54, Arg213, Lys214, Arg262 and Lys329 (Fig. 4e, Supplementary Fig. 6 and Supplementary Movies 1 and 2). Thus, these conformational changes close the gap between substrate and nucleotide. The position of C1 advances 0.7 Å closer to the γ-phosphate of ATP, and the entire inositol ring rotates 7° as it moves (Fig. 4e). Our crystal structures also show that hPPIP5K2KD exploits a specific property of a phosphoester bond: its ability to rotate. The 1-α-phosphate group rotates 17° between ground and transition states (Fig. 4f) and then a further 19° after the phosphoanhydride bond is formed (Fig. 4f). These transitions are not available for inositol phosphate kinases that phosphorylate the hydroxyl groups of the inositol ring, which lack conformational flexibility. Furthermore, in hPPIP5K2KD the inositol ring of IP6 and 5-IP7 is presented perpendicular to the plane of the nucleotide’s β- and γ-phosphates, which avoids steric and electrostatic clashing between the nucleotide and the non-reacting oxygens on the 1-phosphate of the substrate (Fig. 5 and Supplementary Fig. 8). This topology distinguishes hPPIP5K2KD from all of the other inositol phosphate kinases that phosphorylate hydroxyl groups (Fig. 5). Our data also indicate that the newly formed 1-pyrophosphate group is under considerable steric strain (Supplementary Fig. 9), which could favor product release.
A final and unexpected outcome of our studies was the identification of a basic three-dimensional blueprint within the active site that is shared by some other phosphoryl transferases (Fig. 6 and Supplementary Fig. 10). Both Lys248 and the metal-coordinating residue Asp321 are spatially and functionally conserved in ITPK1 and in several other kinases that do not use an ATP-grasp domain: namely, IP3K31,32, AtIP5K33 and protein kinase A34.
Some of the signaling activities of inositol pyrophosphates rely on highly specific interactions with receptors10. Our stereoselective resolution of uncertainty concerning the structures of inositol pyrophosphates (Fig. 4) will now permit the chemical synthesis of physiologic isomers, the deciphering of their molecular interactions with receptors10 and the design of antagonists. The uniquely crowded three-dimensional phosphate topology of inositol pyrophosphates also endows them with high-energy properties12; exploitation of this free energy to nonenzymatically phosphorylate proteins may add to their multifunctionality6,11. Nevertheless, that free energy also imposes a particular challenge upon a kinase; it must not only overcome a substantial energy barrier to synthesize the inositol pyrophosphates but also must do so at a high rate, as evidenced by the rapid turnover of these signaling molecules in vivo4. Our study shows how the active sites of PPIP5Ks have adapted to meet these demands: there is an extensive charge balance that facilitates a partly associative in-line phosphoryl transfer mechanism, a unique topological presentation of substrate to nucleotide and the exploitation of the conformational dynamics of both enzyme and substrate. Clearly, hPPIP5K2KD has a highly specialized catalytic environment. Overall, the structural information that we provide here offers directions for studying the evolution of the active sites of diverse phosphoryl transferases and also offers templates for an underdeveloped area of research: the rational design of inhibitors that specifically target a particular inositol phosphate kinase.
The hPPIP5K2KD constructs (residues 1–366, 37–366 and 41–366 and the single-site mutants) were each subcloned into a vector containing N-terminal tandem His6 and maltose binding protein tags and a TEV protease cleavage site. ArcticExpress (DE3) competent cells (Stratagene) were transformed with the resultant plasmids. Recombinant proteins were purified using Ni-NTA agarose (Qiagen) and HiTrap Heparin HP columns (GE Healthcare), subjected to TEV cleavage, and further purified using Ni-NTA agarose and HiTrap Q Fast Flow and Superdex 200 columns (GE Healthcare). Further details are in Supplementary Methods.
hPPIP5K2KD (residues 37–366 or 41–366) was crystallized by vapor diffusion at 4 °C against a well buffer of 12% (w/v) PEG 3350, 20 mM MgCl2, 0.1 M HEPES pH 7.0, 1 mM ATP or AMP-PNP, with or without 2 mM CdCl2 (Supplementary Table 1). For ternary complexes, the crystals were soaked in 22% (w/v) PEG 3350, 10 mM MgCl2 at pH 5.2 for 0–3 d with either 10 mM IP6 (Calbiochem; >99% pure) or 2 mM 5-IP7 (ref. 35), which was >80% pure36. To generate the transition mimic, 10 mM NaF was also added to the soaking buffer. Diffraction data were collected with an in-house facility or with APS beamline 22-ID/22-BM and were processed with the HKL2000 program37. Our initial structure of hPPIP5K2 (residues 37–366) was solved and auto-built with the PHENIX program38 using single-wavelength anomalous dispersion techniques. Further manual rebuilding used COOT39, and the structure was refined using REFMAC40. Other structures were solved by rigid-body and direct Fourier synthesis or by molecular replacement approaches. Further details are in Supplementary Methods.
Catalytic activities were determined by HPLC using radiolabeled substrates13,14. Additional information is in Supplementary Methods. [3H]IP6 was purchased from PerkinElmer and repurified by HPLC14 (99% pure). The following [3H]-labeled inositol phosphates were prepared as previously described14: IP5 (>99% pure), PP-IP4 (93% pure), 1-IP7 (95% pure), 5-IP7 (97% pure).
Protein Data Bank: the previously determined crystal structures for EhITPK1, hIP3K, AtIP5K, hITPK1 and PKA are deposited under accession codes 1Z2P, 1W2C, 2XAN, 2Q7D and 1L3R, respectively. Newly acquired structures are deposited under accession codes 3T54, 3T99, 3T7A, 3T9A, 3T9B, 3T9C, 3T9D, 3T9E and 3T9F.
Data were collected at the Southeast Regional Collaborative Access Team 22-ID/22-BM beamline at the Advanced Photon Source, Argonne National Laboratory. Supporting institutions may be found at http://www.ser-cat.org/members.html. The use of the Advanced Photon Source was supported by the US Department of Energy, Office of Science, Office of Basic Energy Sciences, under contract no. W-31-109-Eng-38. This work was supported by the Intramural Research Program of the National Institutes of Health and National Institute of Environmental Health Sciences (NIEHS). We are grateful to L.C. Pedersen for advice and support. We also thank H. Ke for his assistance in writing the manuscript. Expression vectors were prepared by the NIEHS Protein Expression Core Facility.
Author contributionsH.W., T.M.T.H. and S.B.S. designed experiments and wrote the manuscript. H.W. and S.B.S. performed the experiments. J.R.F. synthesized substrate for hPPIP5K2.
Competing financial interests
The authors declare no competing financial interests.
Reprints and permissions information is available online at http://www.nature.com/reprints/index.html.