|Home | About | Journals | Submit | Contact Us | Français|
Inositol 1,3,4,5,6-pentakisphosphate 2-kinase (IPK1) converts inositol 1,3,4,5,6-pentakisphosphate(IP5) to inositol hexakisphosphate (IP6). IPK1 shares structural similarity with protein kinases and is suspected to employ a similar mechanism of activation. Previous studies revealed roles for the 1- and 3-phosphates of IP5 in IPK1 activation and revealed that the N-lobe of IPK1 is unstable in the absence of inositol phosphate (IP). Here, we demonstrate the link between IPK1 substrate specificity and the stability of its N-lobe. Limited proteolysis of IPK1 revealed that N-lobe stability is dependent on the presence of the 1-phosphate of the substrate, whereas overall stability of IPK1 was increased in ternary complexes with nucleotide and IPs possessing 1- and 3-phosphates that engage the N-lobe of IPK1. Thus, the 1- and 3-phosphates possess dual roles in both IPK1 activation and IPK1 stability. To test whether kinase stability directly contributed to substrate selectivity of the kinase, we engineered IPK1 mutants with disulfide bonds that artificially stabilized the N-lobe in an IP-independent manner thereby mimicking its substrate-bound state in the absence of IP. IPK1 E82C/S142C exhibited a DTT-sensitive 5-fold increase in kcat for 3,4,5,6-inositol tetrakisphosphate (3,4,5,6-IP4) as compared with wild-type IPK1. The crystal structure of the IPK1 E82C/S142C mutant confirmed the presence of the disulfide bond and revealed a small shift in the N-lobe. Finally, we determined that IPK1 E82C/S142C is substantially more stable than wild-type IPK1 under nonreducing conditions, revealing that increased stability of IPK1 E82C/S142C correlates with changes in substrate specificity by allowing IPs lacking the stabilizing 1-phosphate to be used. Taken together, our results show that IPK1 substrate selection is linked to the ability of each potential substrate to stabilize IPK1.
Inositol 1,3,4,5,6-pentakisphosphate 2-kinase (IPK1,2 IP5 2-K) is an inositol phosphate kinase (IPK) that catalyzes the phosphorylation of inositol 1,3,4,5,6-pentakisphosphate (IP5) to inositol hexakisphosphate (IP6) (1). IP6 is involved in diverse cellular processes such as mRNA export (2), chromatin remodeling (3), apoptosis (4), and development (5). IPK1 itself has been demonstrated to affect apoptosis, and its expression is markedly elevated in renal samples from patients suffering from diabetic nephropathy, indicating that it may play functional roles in disease states (6, 7). Recent crystal structures have revealed that some IPKs possess an N- and C-lobed structure similar to that of protein kinases (PKs), in which both lobes contribute to a conserved ATP binding pocket (8). Both PKs and IPKs require mechanisms to recognize their physiological substrates from among pools of similar molecules. Although PKs recognize consensus sequences in their targets, IPKs must recognize their substrates from among more than 30 IPs based on differences in stereochemistry and numbers and patterns of phosphate groups (9, 10). This mechanism must be able to differentiate physiological substrates from non-substrates that may differ by the presence, absence, or position of a single phosphate. Structures of inositol 1,4,5-trisphosphate 3-kinase (IP3K), inositol polyphosphate multikinase (IPMK), inositol 1,3,4-trisphosphate 5/6-kinase/inositol 3,4,5,6-tetrakisphosphate 1-kinase (ITPK1), and diphosphoinositol pentakisphosphate kinase (PPIP5K) all reveal mechanisms that define substrate specificity (11–14). The first crystal structures of IPK1 revealed how IP5 is phosphorylated on the 2′-hydroxyl of the inositol ring to yield IP6, but the recognition of all five phosphate groups and the axial hydroxyl group did not offer explanations for how it excludes IPs with fewer than five phosphates from its pool of potential substrates (1, 15). For PKs, the stability of the N- and C-lobes is a critical precursor for kinase activation (16–18), and these IPK1 structures indicated that a similar mechanism might be employed for some IPKs. Comparison of IP-bound and free crystal structures of IPK1 suggested that stabilization of its N-lobe was triggered by IP binding (19); however, this mechanism of IP recognition for IPK1, termed “IP-induced stabilization,” has yet to be validated. Recently, we demonstrated that the positions of the phosphates on IPK1 substrates play different roles in directing IP binding to IPK1 and the recognition of IPs as substrates (20).
The aim of our current work is to define the contributions of IP phosphates on IPK1 stabilization and define how this relates to enzyme activation. The results presented here demonstrate that the stable conformation of IPK1 is achieved when IP binding sites in the N-lobe are occupied by the 1- and 3-phosphates of the IP and that substrate specificity of IPK1 can be manipulated through artificial stabilization of the N-lobe in an IP-independent manner.
Inositol phosphates (1,3,4,5,6-IP5; 1,3,4,5-IP4; 1,4,5,6-IP4; 1,3,4,6-IP4; 3,4,5,6-IP4) were purchased from Cayman Chemical Company (Ann Arbor, MI). Trypsin, ADP, and IP6 were purchased from Sigma-Aldrich. AMPPNP was purchased from Jena Bioscience (Jena, Germany). ATP was purchased from Fisher. All IPs were dissolved in Tris-HCl (pH 8.0) to a stock concentration of 20 mm and stored at −20 °C until use.
IPK1 was purified as described previously (19). Limited proteolysis of IPK1 was performed in 50 mm Tris-HCl (pH 8.0), 150 mm NaCl, and 2.5 mm DTT. 9 μm of purified IPK1 was incubated with 2 mm MgCl2 and 1 mm IP (IP4, IP5, or IP6) with and without nucleotides (AMPPNP, ADP) for 20 min at 4 °C. 0.08 μg of trypsin was then added to each reaction, except the undigested control. The reactions were incubated at 20 °C, and samples were taken at 2, 4, 6, and 8 h. Samples were analyzed by SDS-PAGE, stained with Coomassie Blue, and analyzed as described by Gosein et al (19).
Densitometry of IPK1 digestion fragments was performed using ImageJ (21). The Analyze Gel module was used as described in the ImageJ manual. The area was measured for the uncut control fragment where no digestion had occurred, and for the full-length fragment and fragments comprising amino acids 52–451 and 130–451 bands in digested conditions. The areas of the digested fragments were plotted as a percentage of the area of the control band.
DSF was performed on a Corbett Life Science Rotor-Gene 6000. All reactions were prepared individually in 0.2-ml PCR tubes at a final volume of 50 μl. The reactions consisted of 9 μm of purified IPK1 in 50 mm HEPES (pH 7.5), 5 mm MgCl2, 50 mm NaCl, and 2.5 mm DTT buffer, incubated with 1 mm nucleotide (ADP, AMPPNP) and/or 1 mm IP (IP4, IP5, or IP6) for 5 min on ice. SYPRO Orange (Life Technologies) was then added to 5× under dark conditions. The final dimethyl sulfoxide (DMSO) concentration was 0.1%. A temperature melt was carried out between 28 and 80 °C with 0.15 °C/s increments, and the gain was set to 2. The high resolution melt module was used with an excitation filter of 460 nm and emission filter of 510 nm. Each condition was performed in triplicate. Data were analyzed using the Rotor-Gene software. The first derivative of the raw data was analyzed for peaks, which corresponded to the melting temperature (Tm) of IPK1. DSF was performed similarly using IPK1 E82C/S142C mutant purified in the absence of reducing agents; however, the reactions consisted of IPK1 E82C/S142C incubated with nucleotide and/or IP in the absence or presence of 2.5 mm DTT prior to the DSF run.
The structure of IPK1 was processed using the Disulfide by Design software to identify pairs of residues that could be mutated to cysteines to form disulfide bonds (22). Cysteine mutants of these residues were generated by site-directed mutagenesis using the QuikChange method (Stratagene) in two subsequent steps. A pET28a vector containing wild-type Arabidopsis thaliana IPK1 and a hexahistidine tag was used a template (a generous gift from Dr. C. A. Brearley). All mutations were verified by DNA sequencing.
IPK1 kinase activity was measured using the Kinase-Glo Max luminescent kinase assay (Promega) as per the manufacturer's instructions. Kinase reactions were performed in 25-μl volumes in black 96-well plates at 25 °C and contained 50 mm HEPES (pH 7.5), 6 mm MgCl2, 50 mm NaCl, and 300 μm ATP. 0.1 μm of each IPK1 mutant, purified in the absence of reducing agents, was tested with 80 μm IP in the presence or absence of 2.5 mm DTT. 25 μl of Kinase-Glo reagent was added to stop the reaction. Luminescence was measured after 20 min on a Berthold Orion II microplate luminometer.
Initially, 80 μm IP was used, and the amount of IPK1 E82C/S142C enzyme was varied to determine conditions in which product formation was linear over 30 min. Subsequently, an array of reactions with varying concentrations of IP (20, 40, 60, 80, 100, 120, and 140 μm) stopped at various time points (2, 5, 10, 20, and 30 min) was performed in triplicate. The process was performed for both IP5 and 3,4,5,6-IP4, in the absence of reducing agent. The rate of product formation versus IP concentration was plotted and fitted to the Michaelis-Menten equation using nonlinear regression to determine Km and Vmax (GraphPad). Values were reported as mean ± S.D. The kcat values were calculated using the equation kcat = Vmax/[E], where [E] is the concentration of enzyme in μm.
IPK1 E82C/S142C was expressed and purified as described previously for wild-type IPK1; however, DTT and β-mercaptoethanol were omitted in all buffers (19). IPK1 E82C/S142C at 5 mg/ml crystallized with 5 mm ADP/MgCl2, 5 mm IP6 in 0.08 m MES (pH 6.5), 19.85% PEG 3000, 0.17 m NaCl, 2.35% benzamidine HCl at 20 °C within 6–72 h using the sitting-drop vapor diffusion method. X-ray diffraction data were collected on a Rigaku MicroMax-007 HF microfocus x-ray generator fitted with Varimax x-ray optics and a Saturn 944+ CCD detector. Crystals were cryoprotected with reservoir solution that included 10% PEG 400, and data were collected under cryogenic conditions. Diffraction data were analyzed and processed with the HKL2000 software and refined with Phenix and Coot (23–25). Molecular replacement was performed with Protein Data Bank code: 2XAM. All model images were created using PyMOL (27).
We used limited proteolysis as a probe for IPK1 stability (19). As reported previously (19), IPK1 can be cleaved by trypsin at Arg-130 when the N- and C-lobes are not stabilized, but not when the N- and C-lobes have stabilized (Fig. 1A). To identify phosphate groups of the IP substrate that promote IPK1 stabilization, we performed limited proteolysis of IPK1 in complex with IPs alone or in ternary complexes (AMPPNP+IP4, AMPPNP+IP5, or ADP+IP6). The resulting proteolytic patterns observed on SDS gels were compared (Fig. 1B). In contrast to IP alone, IPK1 in complex with both nucleotide and IP protected an additional 46-kDa fragment with Lys-52 at its N terminus, indicating that the Arg-130 site was protected under these conditions (Fig. 1B). To assess the relative stability of IPK1 in the complexes, we performed densitometry to compare the ratios of the stabilized fragments to uncut IPK1 (Fig. 1C). The amounts of full-length fragment in each of the ternary complexes were similar, indicating that digestion occurred to a similar extent in each of these complexes (Fig. 1C, green bars). There was little variability in the amount of the Arg-130-Ser-451 band between the different complexes, indicating that C-lobe stability was very similar in each of the complexes (Fig. 1C, blue bars). Finally, the amount of the Lys-52-Ser-451 band varied between each of the ternary complexes, indicating that IPK1 was differentially stabilized specifically at the N-lobe when IPK1 was bound to different IPs (Fig. 1C, red bars). Protection of the Lys-52 cleavage site is substantially decreased with IPK1 in complex with AMPPNP and 3,4,5,6-IP4 as compared with other ternary complexes, which demonstrates that the 1-phosphate of the IP is singularly important for N-lobe stabilization.
To further characterize the overall stability of IPK1, we used DSF to measure the melting point of IPK1 in the ligand-free state, bound to nucleotide (AMPPNP or ADP), bound to IP (IP4s, IP5, or IP6), or as ternary complexes bound to both nucleotide and IP (Fig. 2). In the free state, IPK1 exhibited a Tm of 35 °C. When bound to either AMPPNP or ADP, the Tm of IPK1 increased to 38 °C, indicating that nucleotide contributes to the overall stability of IPK1. When bound to IPs only, except 1,4,5,6-IP4, IPK1 exhibited Tm values of 40 °C, revealing that the stability of IPK1 is impacted more by the binding of IP than nucleotide. Finally, in the ternary complexes, IPK1 exhibited varying Tm values that were dependent on the phosphorylation pattern of the IP. When also bound to nucleotide, IPK1 displayed markedly lower Tm values in the 3,4,5,6-IP4 and 1,4,5,6-IP4 conditions as compared with the 1,3,4,6-IP4, 1,3,4,5-IP4, IP5, and IP6 conditions. Our results indicate that the 1- and 3-phosphate groups contribute more to the overall stability of IPK1 than other phosphates when IPK1 is in the nucleotide-bound state.
Observing that binding of different IPs led to differential stabilization of IPK1, we hypothesized that IPK1 stabilized in an IP-independent manner would exhibit altered specificity for IP substrates, diminishing the requirement for N-lobe-interacting 1- and 3-phosphates that stabilize IPK1. To test this role of IPK1 stabilization, we engineered and tested a series of disulfide bonds to artificially stabilize the N-lobe of IPK1. We created double cysteine mutants of IPK1 at different sites near residues 110–140, the region of the N-lobe that was previously observed to be unstable in the absence of IP (19). We tested the kinase activity of these mutants using its physiological substrate, IP5, and 3,4,5,6-IP4, a poor IPK1 substrate (20), in reducing and nonreducing conditions (Fig. 3). Wild-type IPK1 displayed very high activity for IP5 and very low activity for 3,4,5,6-IP4 and was unaffected by DTT. The IPK1 E82C/S142C mutant showed moderate activity for IP5 and greater activity for 3,4,5,6-IP4 than the wild-type enzyme under nonreducing conditions, and wild-type activity was restored in the presence of DTT (Fig. 3). This demonstrated the mutations themselves, which are remote from the active site, are not responsible for the altered activity and that the changes were due primarily to the formation of the disulfide bond.
To further examine the effect of the disulfide bond on IPK1 kinase activity, we performed a kinetic analysis of IPK1 E82C/S142C with IP5 or 3,4,5,6-IP4, under nonreducing conditions (Fig. 4). IPK1 E82C/S142C exhibited similar Km and kcat values to that of wild-type IPK1 with IP5 as a substrate, both possessing kcat values of ~44 nmol/min (Table 1). In contrast, IPK1 E82C/S142C displayed a 5-fold increase in kinase activity (kcat = 32.93 nmol/min) with 3,4,5,6-IP4 as a substrate as compared with wild-type IPK1 (kcat = 6.30 nmol/min) (Table 1). The ratio of kcat/Km for the disulfide mutant and wild-type IPK1 for each IP revealed changes in substrate selectivity due to the presence of the disulfide bond. We observed that wild-type IPK1 had a 5.3:1 preference for IP5 over 3,4,5,6-IP4 when comparing IP selectivity. In contrast, IPK1 E82C/S142C displayed a 2.5:1 preference for IP5 over 3,4,5,6-IP4, demonstrating that the disulfide bond altered substrate specificity of IPK1.
To explore structural changes associated with the introduction of the disulfide bond, we crystallized the disulfide mutant in the presence of ADP and IP6 under nonreducing conditions to obtain a structure of IPK1 E82C/S142C (Fig. 5, Table 2). Overall, the structure of IPK1 E82C/S142C was very similar to wild-type IPK1 (root mean squared deviation = 0.5 Å). The presence of the disulfide bond shifted the N-lobe in the mutant structure as compared with the structure of wild-type IPK1 (Fig. 5A). Inspection of the electron density indicated the presence of the disulfide bond between Cys-82 and Cys-142 (Fig. 5B). To confirm the presence of a disulfide bond between these residues, we generated an omit map lacking the side chains at these residues and observed a large unoccupied density in the Fo − Fc map (σ = 3.0) located between both residues, consistent with the presence of a disulfide bond (Fig. 5C). We performed the same refinement using diffraction data from a wild-type structure of IPK1 (Protein Data Bank code: 3UDZ) and did not observe unoccupied density between the two residues in the Fo − Fc difference map (σ = 3.0), confirming that the density observed is unique to the mutant structure (Fig. 5D).
To investigate the effect of the disulfide bond on the overall stability of IPK1, we employed DSF to measure the Tm of wild-type IPK1 and IPK1 E82C/S142C. We compared the Tm values of wild-type IPK1 and IPK1 E82C/S142C in the free state, bound to nucleotide (AMPPNP or ADP) or IP (3,4,5,6-IP4, IP5, or IP6), or in ternary complexes bound to both nucleotide and IP (Fig. 6). We performed this experiment under reducing conditions (with DTT) and nonreducing conditions (without DTT) to detect the effect of the disulfide bond. In the presence or absence of DTT, wild-type IPK1 exhibited similar Tm values as noted above, with the lowest Tm in the apo condition, a slight 2 °C increase in Tm when bound to nucleotide only, a moderate 5 °C increase in Tm when bound to IP only, and a large 12 °C increase in the ternary complex with IP5 and IP6 (Fig. 6, first and second sets). AMPPNP + 3,4,5,6-IP4 failed to stabilize IPK1 to the same extent as AMPPNP + IP5 and ADP + IP6 as shown by a 5 °C difference in Tm between the ternary complexes. DTT did not markedly affect any of the Tm values of wild-type IPK1 under any conditions, revealing that the overall stability of wild-type IPK1 is not dependent on disulfide bonds. This is consistent with the structure of wild-type IPK1 that does not have disulfide bonds. In contrast, IPK1 E82C/S142C exhibited an average increase in Tm of 5.6 °C under each condition as compared with wild-type IPK1 in the absence of DTT, and this increase in Tm is reversible in the presence of DTT, confirming that the disulfide bond is increasing the overall stability of IPK1 (Fig. 6, third and fourth sets). In our kinetic studies, we observed that IPK1 E82C/S142C exhibited increased selectivity for 3,4,5,6-IP4 (Table 1). Thus, the engineered disulfide bond that was stabilizing the N-lobe of IPK1 in an IP-independent manner conferred specificity to 3,4,5,6-IP4, a poor substrate that lacks the 1-phosphate that is key for N-lobe stabilization, indicating that the stability of the N-lobe is a determinant for substrate specificity of IPK1.
Here, we investigated the role of the IP phosphates in stabilizing IPK1. We determined that the 1-phosphate is important for localized stability of the N-lobe (Fig. 1), consistent with the IP-free crystal structure of IPK1 that revealed localized destabilization of the region surrounding Arg-130, which interacts with the 1-phosphate (19). We recently investigated specific roles of phosphate groups in binding and activation of IPK1, and we determined that the 5- and 6-phosphates were important for binding of the IP, whereas the 1- and 3-phosphates were important for activation of IPK1 (20). In our current study, we demonstrated that the overall stability of IPK1 is dependent on the N-lobe binding 1- and 3-phosphates (Fig. 2). Thus, the dual roles of the N-lobe binding 1- and 3-phosphates in both IPK1 stabilization and IPK1 activation indicate that the stabilization of the N-lobe is an important component of IPK1 activation. We have also demonstrated that the IP, and not the nucleotide, contributes substantially to the stabilization of the N-lobe of the kinase. This is consistent with our earlier observation that the N-lobe is destabilized in the absence of IP and the structures of Gonzales et al. (15), which demonstrated structures bound to IP, but not to nucleotide adopt the stable state (19).
Changes in protease accessibility and increased protein stability upon substrate binding suggested that activation of IPK1 may be linked to the ability of the IP substrate to stabilize IPK1, so we hypothesized that artificial stabilization of the N-lobe of IPK1 in an IP-independent manner would alter the substrate specificity of IPK1. We engineered the IPK1 E82C/S142C disulfide mutant by introducing a cysteine in the N-lobe of IPK1 at a site that is unstable in the IP-free crystal structure and a second cysteine at a site that is well ordered and apparently stable in all IPK1 crystal structures (Fig. 5) (15, 19). By covalently linking stable and unstable regions, we endeavored to stabilize some or all of the N-lobe. This increase in stability was reflected in our DSF data with IPK1 E82C/S142C, which shows increased Tm values under nonreducing conditions (Fig. 6). This mutant also displays increased selectivity for 3,4,5,6-IP4 under nonreducing conditions as compared with wild-type IPK1 (Table 1). Given that 3,4,5,6-IP4 lacks the ability to stabilize wild-type IPK1 as much as IP5 in ternary complexes (Fig. 1), our data collectively show that substrate specificity is linked to N-lobe stability.
We recently proposed a model for IP recognition wherein interactions with the IP substrate stabilize the N-lobe and C-lobe to promote kinase activation (19). In this model, the stable C-lobe recognizes the 4-, 5-, and 6-phosphates of the IP, and the 1- and 3-phosphates of the IP induce N-lobe stabilization. The key interaction between Arg-130 and the 1-phosphate of the IP stabilizes the N-lobe. Thus, our model possesses a triad of elements that are interconnected for IPK1 activity: 1) substrate specificity, 2) stability, and 3) activation. Our current study provides the necessary evidence for the first time to reconcile all three elements and validates our proposed model. We have linked substrate specificity to kinase stability by demonstrating that IPs possessing the 1-phosphate markedly stabilize the N-lobe as compared with 3,4,5,6-IP4, which lacks a 1-phosphate, whereas artificial stabilization of the N-lobe of IPK1 alters substrate specificity. We have also found that the 1- and 3-phosphates of the IP provide increased stability to IPK1, whereas the 5- and 6-phosphates do not affect the overall stability of IPK1, consistent with an unstable N-lobe and a stable C-lobe in the absence of IP. Moreover, previous kinetic studies of IPK1 link substrate specificity to kinase activation (20). We contend that IPK1 stability is linked to IPK1 activation because IPK1 stability correlates with the active state in previous crystal structures (19), IPs that display low IPK1 kinase activity (kcat) correlate with a decreased ability to stabilize IPK1 (20), and kinase stability has been validated to be a precursor for activation of PKs, for which IPK1 shares similar features (9, 15). In short, our current data provide strong evidence to support our model that IPK1 substrate specificity is coupled to IPK1 stability and subsequent activation (Fig. 7).
Conformational changes that stabilize links between protein kinase N- and C-lobes that occur during activation have been well documented (9). Stabilization of the kinase structure is promoted by the assembly of a hydrophobic spine spanning the two lobes, one of which consists of the Phe residue from the DFG motif, a hydrophobic residue from the αC helix, and two residues buried in the cores of the N- and C-lobes. This structural spine, called the R-spine, assembles only when the DFG and αC helix regulatory motifs sit in their active conformations (8). In IPK1, the αC helix sits at a 46° angle as compared with PKA that precludes the αC helix from contributing a hydrophobic residue to the R-spine (Fig. 8, A and B). Further, the IPK1 DLS motif lacks the large hydrophobic Phe residue, which likely impairs its ability to assemble a stabilizing R-spine (Fig. 8, C and D). These structural differences highlight that IPK1 activation occurs through a different mechanism than found in many PKs. In the proposed model, it is the IP substrate that acts as the bridge between the N-lobe and C-lobe to stabilize the active conformation of the kinase rather than the intrinsic R-spine structure (Fig. 7). The phosphate profile of the IP dictates the ability of the substrate to stabilize the N-lobe of IPK1 and promote kinase activation. This is consistent with previous suggestions that conformational dynamics may play a role in the catalytic cycle of IPK1 (19, 26). Thus, IPK1 retains important features of PK activation such as N- and C-lobe stabilization, yet distinguishes itself from the PKs. These distinct features could be exploited to selectively inhibit IPK1.
In this study, we demonstrate the missing link between IP substrate specificity and the N-lobe stability of IPK1. Our results conclusively support IP-induced stabilization of IPK1 as a proposed model of IP discrimination. IP substrates are very similar, and the elucidation of the unique substrate recognition mechanism of IPK1 among IPKs may provide a basis for selective pharmacological targeting of IPK1. Inhibitors of IPK1 would be useful for ascertaining the functional roles of higher IPs and validation of this IP signaling axis as a therapeutic target.
We thank Dr. Charles A. Brearley (University of East Anglia) for the gift of the AtIPK1-pET28 vector and Dr. Dan Bernard (McGill University) for access to both the Rotor-Gene 6000 used for DSF and the luminometer used in enzyme assays.
*This work was supported by a Canadian Institutes of Health (CIHR) Research Operating Grant MOP-93687 (to G. J. M.) and a CIHR Strategic Training Initiative in Chemical Biology (to V. G.).
2The abbreviations used are: