Introduction

As an essential gene in mammalian development with ubiquitous expression in all eukaryotes1, IPMK has important functions at the nexus of many important signaling, metabolic and regulatory pathways, exerting influence through both kinase-dependent and independent roles2. The kinase activity of IPMK is a well-characterized node connecting inositol polyphosphate to lipid phosphoinositide signaling, and is critical for AKT activation3, Wnt signaling4, PLC-mediated IP3/Calcium signaling1,5,6,7, nuclear receptor transcriptional activation8, mRNA export9, and many other roles2,10. However, IPMK also has enzyme-independent roles, which utilize protein-protein interactions to regulate several crucial signaling proteins. IPMK acts in an enzyme-independent manner interacting with the p53 tumor suppressor, acting as a transcriptional co-activator for p5311,12. IPMK participates in mTOR signaling by serving as a classic adaptor protein, connecting mTOR and Raptor in mammalian cells13. IPMK has been shown to mediate LKB1 regulation of AMPK by directly interacting with AMPK, which prevents AMPK phosphorylation and activation in response to several metabolic signals, including glucose14,15 and metformin16. Further, IPMK binds and stabilizes the TRAF6 ubiquitin ligase signaling scaffold in macrophages, preventing ubiquitination of the TRAF6 protein to maintain signaling17. Thus, these important studies have established several roles IPMK plays via kinase-dependent and –independent mechanisms.

IPMK non-catalytic functions are mediated in part by regions of the IPMK protein that are not resolved in all crystal structures of IPMK orthologues and homologues solved to date18,19,20,21,22,23. Although it is well established that disordered protein domains generally participate in protein-protein interactions24,25,26,27, what role the IPMK disordered regions might play in modulating IPMK ATP-binding and kinase activity has not been thoroughly explored28.

Here, a highly engineered human IPMK lacking its two disordered domains was designed, and designated as ΔIPMK. To determine if the absence of these disordered domains alters IPMK enzyme activity on the well-known substrate PIP2, detailed kinetic analyses were performed and compared to wildtype activity. Surprisingly, ΔIPMK has a 1.8-fold higher turnover number (kcat) on PIP2 than wild type IPMK. In addition, a ΔIPMK construct with an additional 33 N-terminal amino acid extension to G37 (extΔIPMK) also has a 1.8-fold increase in kcat, suggesting this increase in activity is due to replacement of the native internal loop, not the unstructured N-terminus. Further, removal of the disordered domains from IPMK does not dramatically alter IPMK structure, as a 2.5 Å crystal structure of ΔIPMK maintains all the consensus features of the IPK-superfamily fold. Comparison of the ΔIPMK structure with other superfamily structures suggested a short N-terminal sequence of IPMK that is disordered in most structures could participate in ATP-nucleotide binding. This sequence was removed from ΔIPMK, and the KM for ATP increased 4.9-fold for ΔIPMK compared to wild type IPMK. In addition, adding back a portion of the N-terminus containing the ATP clamp partially rescued the increase in KM for ATP. Taken together, these data suggest the disordered domains of IPMK play an important role in modulating the in vitro kinase activity of IPMK, through multiple kinetic mechanisms.

Results

Human IPMK lacking native internal loop is a more efficient PI3-kinase than wild type

All crystal structures of IPK-superfamily kinases have a consensus ATP-grasp fold structurally conserved between family members18,19,20,21,22,23. However, there are two intrinsically disordered domains in several members of the superfamily: an N-terminal disordered domain and an “internal loop” within the kinase-domain that occurs between S279 and Q373 in human IPMK (Fig. 1A). This kinase-domain internal loop contains a nuclear localization signal and a CK2 phosphorylation site29. Both the N-terminal and internal loop disordered domains participate in IPMK interactions with p5312, TRAF617, AMPK14, and Raptor13 in GST-pull down and/or functional assays, to various degrees and levels of specificity, and are usually required but not sufficient for full IPMK non-catalytic functions. Although the disordered domain’s non-catalytic functions are well established, it is unclear if they modulate IPMK kinase activity.

Figure 1
figure 1

Removal of IPMK disordered domains confers a 1.8-fold increase in PIP2 catalytic efficiency. (A) Primary organization comparison between human ΔIPMK (top), extΔIPMK (middle), and wild type IPMK (bottom). Gray represents disordered regions in all previous structures of IPK superfamily members, ΔIPMK does not possess residues 1–69 and 279–373 of the full-length WT human IPMK; extΔIPMK does not possess residues 1–36 and 279–373 of full length WT IPMK. Red represents catalytic regions, blue the IP-helices that bind substrate and green an artificial linker sequence added to maintain protein stability. The green artificial (Gly4-Ser)2 linker was inserted between residues 279 and 373. (B) Best fit non-linear Michaelis-Menten curves describing WT IPMK, extΔIPMK, and ΔIPMK kinetic parameters on PIP2 micelles with saturating ATP, actual values provided in Table 1, 10 nM of each enzyme was used in these assays. Velocities were fit using Graphpad prism.

Full size image

To examine if removal of the two IPMK disordered domains alters IPMK kinetic properties on the PIP2 substrate, the entire disordered N-terminus of human IPMK at D70 with a 6X Histidine tag was replaced, and the kinase-domain internal loop was replaced with a small synthetic Glycine-Serine linker [S279-(GGGGS)2-C373] (Figs 1A and S1). This construct was designated ΔIPMK, and was easily expressed in bacteria and purified to homogeneity for crystallographic and enzyme kinetic analyses. Further, a ΔIPMK construct with an additional 33 N-terminal amino acid extension to G37 (extΔIPMK) was similarly purified and examined. It is worth noting several variations of the boundaries for this precise IPMK construct yielded significantly less amounts of soluble IPMK protein when expressed in bacteria (Table S1).

These kinetic studies showed although ΔIPMK has an almost identical KM for PIP2 as the wild type enzyme (both about ~38 μM), the kcat increases 1.8-fold for ΔIPMK to 7.4 s−1 (Table 1) When compared to the wild type IPMK kcat for PIP2 (4.4 s−1), the increased kcat drives a 1.8-fold higher catalytic efficiency (kcat/KM) of ΔIPMK over wild type IPMK (Fig. 1B, Table 1). The extΔIPMK construct behaved almost identical to ΔIPMK with respect to the PIP2 substrate in these assays. Thus, both ΔIPMK and extΔIPMK (Fig. 1A) appear to be more catalytically efficient than wild type IPMK at phosphorylating PIP2 to PIP3.

Table 1 Kinetic parameters of WT IPMK, extΔIPMK, and ΔIPMK.
Full size table

Crystallography shows ΔIPMK maintains the ATP-grasp fold

In parallel with the kinetic studies, crystals of the ΔIPMK protein were produced for X-ray structure determination of the enzyme. The growth of these crystals required the presence of the ATP-analog adenylyl-imidodiphosphate (AMP-PNP), inositol-(1,3,4,6)-phosphate (IP4) substrate and 1 mM MnCl2, although these ligands were not detectable in the electron density maps. The ligands may have been lost in the 30-minute cryo-protectant soak prior to cryopreservation for X-ray diffraction analysis. Fifteen molecules of buffer components were ordered in the structure. Three of the 6 amino acids of the His-tag in ΔIPMK were weakly ordered in a position next to the β1 strand where nucleotide is predicted to be, if the nucleotide were not lost during cryopreservation. Nevertheless, we were able to solve the structure of human apo-ΔIPMK by molecular replacement using the yeast IPMK as the search model. Table 2 contains data collection and structure refinement statistics.

Table 2 Summary of crystal parameters, data collection and refinement statistics for PDB entry 6E7F.
Full size table

Analysis of the fold of ΔIPMK revealed the expected ATP-grasp domain structure (Fig. 2A). As with all other members of this superfamily, ΔIPMK has an N-terminal domain consisting of β-sheets 1–3 and α-helix 1, the IP-binding loop consisting of α-helix 2, and a C-terminal domain consisting of the remaining β-sheets 4–7 and α-helices 1, 3, 4, 5 and 6 (Fig. 2B). Two ΔIPMK monomers are in each asymmetric unit of the P21212 space group crystals, with a crystallographic dimerization interface mediated mainly through contacts in the IP-binding helix (Fig. 2C). The residues participating in this interface are I92, V96, P100 and L101 (Fig. 2D). We have not observed ΔIPMK to exist as a dimer in size exclusion chromatography, suggesting the dimer in our asymmetric unit is crystallographic. The coordinates for this structure have been deposited within the Protein Data Bank (PDB ID: 6E7F).

Figure 2
figure 2

Crystal structure of the catalytic core of Human ΔIPMK. (A) Ribbon diagrams of the overall structure of the catalytic core of human ΔIPMK and the unstructured kinase-domain internal loop shown in red, N-terminal kinase lobe shown in cyan, inositol phosphate-binding helix shown in blue. (B) Topology map of IPMK, color scheme identical as in A., generated using TopDraw. (C) Overall structure of the crystallographic IPMK dimer in the asymmetric unit, with the dimerization interface boxed, note that ΔIPMK dimerization is not detectable by size exclusion chromatography. (D) Magnification of panel C, showing hydrophobic residues that mediate the crystallographic dimer interface along the IP helix, hydrophobic residues indicated.

Full size image

ΔIPMK core-kinase structure is similar to other IPK-superfamily members

Comparing ΔIPMK to other IPK-superfamily members, ΔIPMK shares an expected high degree of structural homology to other superfamily members. The structure reported here is in close agreement with the other recently published crystal structures of human IPMK, with a root mean square deviation (RMSD) of 0.48 Å18. Superposition over the Cα backbone of ΔIPMK with the human H. Sapiens IP3K gave an RMSD of 1.041 Å (Fig. 3A). As expected, human IP3K diverges most significantly with ΔIPMK in the IP-binding loop, where IP3K has a much larger sequence that contains 2 alpha helices that are not seen in IPMK (Fig. 3A, boxed helix). Superpositions of the human ΔIPMK with the parasitic amoeba E. histolytica IP6K produced an RMSD of 1.145 Å (Fig. 3B), with plant A. thaliana IPMK produced an RMSD of 1.028 Å (Fig. 3C) and with the yeast S. cerevisiae IPMK structure yields an RMSD of 1.10 Å (Fig. 3D). A more thorough comparison of the differences between these structures has been extensively addressed elsewhere18, however we note the larger IP-binding loops of IP3Ks have been proposed to prevent IP3K activity on membrane phosphoinositides23, and our ΔIPMK structure further supports that hypothesis.

Figure 3
figure 3

The overall fold of ΔIPMK is conserved within the Inositol Phosphate Kinase Superfamily. (A) Superposition of ΔIPMK (blue) and human IP3K (dark green) bound to AMPPNP, Mn2+, and Ins(1,4,5)P3 (PDB: 1W2C) with a RMSD of 0.995 Å, determined using PyMOL. Human IP3K has a significantly larger inositol phosphate-binding region compared to ΔIPMK (green box), which has been ascribed to preventing IP3K from phosphorylating PIP2 in membranes. (B) Superposition of ΔIPMK (blue) and E. histolytica IP6K (orange) bound to ATP and Ins(1,4,5)P3 (PDB: 4O4D) with an RMSD of 1.170 Å. (C) Superposition of ΔIPMK (blue) and apo-A. thaliana IPMK (magenta) (PDB: 4FRF) with an RMSD of 1.045 Å. (D) Superposition of ΔIPMK (blue) and S. cerevisiae IPMK (red) (PDB: 2IF8) with an RMSD of 0.995 Å. All structures generated using PyMOL and any internal loop regions disordered theses structures are represented by dashed lines connecting to the next ordered amino acid.

Full size image

ΔIPMK has higher KM for ATP

In comparing our structure to other IPK-superfamily members, we noted two IPK-superfamily members contain extended N-terminal regions that are ordered in two of the crystal structures, and covers part of the nucleotide-binding pocket, as highlighted in the human IP3K structure (Fig. 4A,B). This region is disordered in all other IPK-superfamily members except in two recently reported structures of human IPMK (PDB:5W2H and 5W2I), where a homologous region of the N-terminus, residues I65-P69 in human IPMK, is also similarly ordered over the nucleotide-binding pocket (Fig. 4C,D)18. We found the N-terminal I65 residue is in reasonable Van der Waals contact distance to the 4′ and 5′ carbons of the ribose unit of ADP (4.2 Å – 3.8 Å) (Fig. 4C), but the equivalent sequence in the ΔIPMK construct was removed and replaced with an N-terminal histidine tag (Figs 4D and S1), further suggesting an ATP-stabilizing effect of the I65-P69 sequence. This sequence is clearly not required for catalysis on the PIP2 substrate, as the removed I65-P69 sequence in ΔIPMK results in a kinase with increased catalytic efficiency compared to the wild type enzyme (Fig. 1B, Table 1). Nevertheless, in our electron density maps we observed weak electron density occupying a similar space as the I65-P69 sequence (Fig. 4D). This density could only be attributed to 3 amino acids of the 6-histidine tag (Fig. 4D). The weak density occupies a comparable space to the ordered I65-P69 sequences in the two other structures mentioned above (Fig. 4A–C). Thus, we hypothesized the native N-terminal I65-P69 sequence might function to stabilize ATP binding in human IPMK. ΔIPMK would be predicted to have an increased KM for ATP compared to wild type, since the I65-P69 sequence was removed from ΔIPMK. Indeed, ΔIPMK has a KM for ATP of 300 μM (Fig. 4E), while the ATP KM for wild type IPMK is 61 μM (Table 1), demonstrating ΔIPMK has a 4.9-fold increased KM for ATP than wild type, despite ΔIPMK having an identical KM and 1.8-fold increased kcat for the PIP2 kinase substrate (Table 1). When the I65-P69 sequence was added back to ΔIPMK using the extΔIPMK construct (Fig. 1A) the KM for ATP was rescued to 94 μM, only a 1.5-fold increase from the wild type IPMK (Table 1).

Figure 4
figure 4

Deletion of IPMK disordered domains perturbs KM for ATP. (A) Ribbon diagram of IP3K-B bound to ATP with Ins(1,4,5)P3 modeled into the inositol binding site (PBD: 2AQX)36. (B) Enlarged view of the IP3K-B ATP binding site boxed in A. shows the ordered N-terminus potentially acting as an “ATP clamp” to stabilize ATP binding. (C) ATP binding site from the crystal structure of human IPMK bound to ADP18, highlighting Van der Waals interaction distances of I65 sidechain (grey sticks) to 5′ (4.2 Å) and 4′ (3.8 Å) carbons of the ribose moiety in ADP (sticks). (D) N-terminus of ΔIPMK structure (blue) superposed on IPMK structure (magenta). Orange box shows the ordered ΔIPMK His-tag residues (depicted as blue sticks) in a similar position as the I65 ATP-clamp highlighted in C. (E) Enzyme kinetic data fit to nonlinear Michaelis-Menten curves describing WT IPMK, extΔIPMK, and ΔIPMK kinetic parameters on ATP with saturating PIP2. Velocities were fit using Graphpad prism.

Full size image

Consistent with these results, a previous study deleted the kinase-domain internal loop alone from human IPMK (residues 266–371) but left all N-terminal IPMK sequences native, and showed the KM for ATP only increased about 2.3-fold for this mutant28, suggesting removal of the kinase-domain internal loop region alone can also change the KM for ATP. Because the change in KM for ATP in ΔIPMK (both disordered domains removed) is 4.9-fold higher versus only a 1.5-fold change when only the kinase-domain internal loop region is removed28, the data suggest changes to the N-terminal disordered domain immediately proceeding the kinase domain more dramatically determine ATP binding affinity, but also that there exists a synergy between both disordered domains which can alter ATP binding to IPMK. Taken together, these crystallographic observations and enzyme kinetic data suggest the native N-terminal sequences of human IPMK can form a previously uncharacterized and structurally conserved ATP clamp, stabilizing ATP binding to IPMK.

Discussion

Here, we report removal of the disordered domains of human IPMK (ΔIPMK) results in an enzyme that is more catalytically efficient than wild type IPMK, driven by a 1.8-fold increase in kcat. Since the KM for PIP2 is not significantly different between ΔIPMK, extΔIPMK, or wild type IPMK (Table 1), IPMK disordered domains are not likely involved with PIP2 substrate binding. In support of this claim, as \({K}_{{\rm{M}}}=\frac{{k}_{-1}+{k}_{cat}}{{k}_{1}}\), KM,WT = KM,extΔIPMK = KM,Δ for PIP2, and kcat,WT IPMK < kcat,extΔIPMK = kcat,Δ for PIP2, kcat does not substantially influence the IPMK KM for PIP2. Therefore, IPMK KM,PIP2 ≈ KD, PIP2, which suggests removal of the native IPMK internal loop and replacement with a glycine-serine linker does not alter IPMK’s binding affinity for PIP2. If the IPMK disordered domains did affect PIP2 binding, the PIP2 KM would be expected to change for both extΔIPMK and ΔIPMK compared to wild type IPMK. In agreement with our data, a previous study suggested completely removing the human IPMK internal loop domain (residues 266–371) does not alter substrate binding, since removal of the domain alone only nominally increased IPMK enzyme substrate KM for the Inositol (1,4,5) trisphosphate (IP3) substrate (0.35 μM for WT vs. 0.4 μM for mutant)28. Thus, the data clearly suggest neither of the two disordered domains in human IPMK affect substrate binding to the IP-helix.

As an alternative hypothesis, IPMK disordered domains could be predicted to alter ATP binding, even when excess ATP is present as is the case for PIP2 kinetics experiments. However, the KM for ATP is actually increased 1.5-fold for extΔIPMK and 4.9-fold for ΔIPMK, despite the enzyme becoming faster. Thus, the data suggest the disordered domains actually increase IPMK binding affinity for ATP because the KM for ATP decreases with increasing proportions of the disordered domains present (i.e. KM,WT < KM,extΔ ≪KM, Δ).

The structure of ΔIPMK allowed us to highlight regions of IPMK previously established to be required in GST pull-down assays for full co-precipitation with p53 (Fig. 5A,B). The p53-IPMK interaction was shown to modulate p53 activity, and p53 requires the IPMK IP-binding loop and ATP cleft to fully interact with IPMK. This model predicts interaction of p53 with IPMK would down-regulate IPMK kinase activity, consistent with p53 serving an anti-proliferative role during stress responses. Similar analyses using AMPK (Fig. S2A) and TRAF6 (Fig. S2B) were also consistent with protein-protein interactions inhibiting IPMK kinase activity, although this remains to be determined.

Figure 5
figure 5

p53 interaction site superimposes onto ΔIPMK ATP- and substrate-binding sites. (A) Surface and (B) cartoon ribbon representations of ΔIPMK, with ADP and PIP2 modeled into the structure. The IPMK interaction site with p53 has been mapped to IPMK exon 412, the residues of exon 4 are highlighted in orange, with the remainder of the IPMK kinase domain depicted in light blue. Both ADP and an inositol phosphate kinase substrate were modeled into the structure for reference, the position of the p53 interaction site predicts p53 could function to sterically interfere with IPMK kinase activity.

Full size image

Taken together, the crystallographic and kinetic data are consistent with a model of IPMK where the disordered domains allosterically modulate kinase activity, through a mechanism that does not alter kinase substrate binding affinity (i.e. PIP2), increases ATP nucleotide binding affinity (i.e. decreases KM), and decreases the turnover number (i.e. kcat) (Fig. 6A,B). This could represent biologically relevant regulation, given intrinsically disordered domains are well known to serve as protein-protein interaction scaffolding domains. Proteins that interact with IPMK disordered loops could compete for disordered domain binding, freeing IPMK from allosteric inhibition (Fig. 6C). However, it is not clear from a structural perspective how any allosteric mechanism could work. Both disordered domains are present in the yeast IPMK21 crystal structure (although no electron density is observed in these regions), and yet the fold of the yeast IPMK with disordered domains is very similar to human IPMK without the disordered domains. It remains possible that small perturbations in side chain interactions between the IPMK core kinase domain and disordered domains could regulate catalytic rate. Thus, solution-based biophysical dynamics data may be required to resolve this question, such as heteronuclear spin quantum coupling NMR studies to determine sites of contact and electron paramagnetic resonance (EPR) based double electron-electron resonance (DEER) spectroscopy measurements to determine long-range distances between residues.

Figure 6
figure 6

Model of intrinsic regulation of IPMK catalysis and ATP-binding. (A) Removal of disordered domains (blue) from ΔIPMK results in a catalytically faster enzyme. (B) When present, the disordered domains (blue) may allosterically modulate the core kinase domain of IPMK that results in a slower enzyme. (C) Proteins (green) could interact with the internal loop region of IPMK, removing the inhibition of the disordered domains, resulting in activation of IPMK catalysis. (D) Removal of disordered domains from ΔIPMK results in less stable ATP binding. (E) When present, the disordered domains stabilize ATP binding through the N-terminal regions that includes the I65-P69 ATP-clamp sequence (blue). (F) Proteins that interact with the disordered N-terminal region of IPMK (red) could regulate the structural positioning of the ATP clamp sequence, hence regulating IPMK ATP-binding in certain cellular compartments where ATP concentrations are low.

Full size image

A recently published structure of apo-human IPMK had a similar but non-identical construct design as reported here18, and comparison of our 2.5 Å apo-ΔIPMK structure to the 1.8 Å apo-HsIPMK structure (PDB: 5W2G) reveals almost identical results, with an RMSD of 0.48 Å over the 161 identical residues of the IPMK Cα backbone, despite different space groups and different numbers of molecules in the crystallographic asymmetric units. Further, the crystallization conditions reported here required acidic mother liquor (sodium citrate pH 3.5), while the conditions reported for H.s.IPMK are far milder (MES-imidazole pH 6.0 or HEPES pH 7.5). Despite these differences, the final structures derived from these crystals are very similar. Thus, the independently solved structures make a strong case the native structure of the human IPMK core kinase domain is accurately represented with these models, at least as examined by X-ray crystallography.

In two of the structures of human IPMK (PDB: 5W2H, 5W2I), a short 5-amino acid N-terminal sequence (I65-P69) was ordered directly above the ATP binding site18. These residues, particularly I65, form a “roof” atop the ATP binding pocket, with the well-resolved I65 side chain in reasonable Van der Waals distance (4.2 Å – 3.8 Å) from the nucleotide ribose sugar C4′ and C5′ positions (Fig. 4C)18. Similar sequences over the ATP-binding sites are ordered in the crystal structure of the IPK-superfamily members human IP3K20,23 and the amoeba IP6K22, but are disordered in all other structures. These crystallographic observations suggested residues I65-P69 of human IPMK might form an “ATP-clamp” helping to hold ATP in the nucleotide-binding pocket (Fig. 6E), and that removal of the clamp would increase the KM for ATP (Fig. 6D). The I65-P69 ATP-clamp sequence was deleted from the ΔIPMK construct reported here, and accordingly observed a 4.9-fold increase in the KM for ATP in ΔIPMK, but only a 1.5 fold increase in ATP KM was observed when the ATP-clamp sequence was restored using the extΔIPMK construct.

It is unknown if the ATP clamp sequence can alter IPMK activity in living cells. Since intracellular ATP levels normally range between 2–10 mM30,31,32, one might expect shifting the KM for ATP from 61 μM to 300 μM to have little if any regulatory capacity on IPMK kinase activity. However, ATP concentrations in certain subcellular compartments can exist below the level of detection for ATP biosensors with dissociation constants for ATP below 10 μM30,33, and ATP levels can change in subcellular compartments as measured with biosensors by over 1,000-fold31,32,34. IPMK shuttles between the nucleus and the cytoplasm29, but often localizes to the nucleus with a speckled, uneven distribution3. The I65-P69 sequence is not ordered in most IPK-superfamily structures, suggesting the I65-P69 sequence is dynamic. It is tempting to speculate IPMK ATP binding could be regulated by structural perturbations induced in the I65-P69 ATP clamp caused by protein-protein interactions with the N-terminal intrinsically disordered domain of IPMK (Fig. 6F). In this model, the positioning of the I65-P69 ATP-clamp over the nucleotide could be modulated by proteins interacting with the disordered N-terminus of IPMK (Fig. 6F). This model predicts proteins discovered to interact with the N-terminus of human IPMK might function to discourage ATP binding to IPMK, perhaps regulating IPMK kinase activity in certain sub-cellular compartments when ATP levels drop. This type of regulation is further supported by GST-pull down interaction data showing proteins require the substrate-binding helix and the ATP binding cleft to fully interact with IPMK (Fig. 5). Another intriguing possibility is the recent elucidation that the nucleotide GTP can be used by a certain phosphoinositide kinase previously thought to only use ATP for the phosphoryl-transfer reaction35. Should the ATP-clamp help select for GTP or ATP, any regulation of the clamp might have the possibility to act as a “nucleotide switch”, converting IPMK from an ATP-dependent kinase to a GTP-dependent kinase. It is currently unknown if IPMK can use GTP in the phosphoryl-transfer reaction.

From a structural perspective, it is important to note all the published crystal structures of IPK-superfamily members appear to be extremely similar with or without small molecule ligands bound to the kinase, e.g. superposition of the recently published IPMK (PDB:5W2I) bound to 3 ligands (ATP-analog, 2 magnesium metal ions and diC4-PIP2) and the apo-ΔIPMK reported here has an RMSD over the Cα backbone of 0.414 Å. Similarly, superposition of ΔIPMK vs. ADP/Mn2+-bound S. cerevisiae IPMK21 has an RMSD of 0.995 Å over the Cα backbone. These data suggest the IPK-superfamily of kinases do not undergo large conformational changes upon small molecule ligand binding. However, it remains possible these ligands induce conformational dynamics less well suited to be detected by X-ray crystallography and which could become apparent if solution biophysical approaches were applied.

The crystallographic and kinetic studies presented here show the disordered domains of human IPMK can inhibit IPMK in vitro kinase activity on PIP2 not by altering PIP2 binding, but by somehow decreasing the turnover number. We also show that removal of the IPMK disordered domains discourages ATP binding to IPMK in vitro. The data are consistent with a new model where the disordered domains of IPMK modulate its kinase activity, through multiple, distinct kinetic mechanisms. Crystallographic studies confirmed removing the disordered domains does not alter the overall fold of the IPMK kinase-domain, however the structural basis of how IPMK disordered domains alter the kinase activity remains to be determined.

Methods

Protein Expression and Purification

The human ΔIPMK expression vector was constructed by VectorBuilder (Santa Clara, CA). Briefly, human IPMK (NCBI nucleotide accession NM_152230.4) 70–279 and 373–416 interrupted with a (Gly4-Ser)2 linker was inserted into an IPTG inducible N-terminal His-TEV expression vector. This expression vector was transformed into chemically competent BL21 dE3 E. coli cells. An overnight Terrific Broth (Sigma) culture under ampicillin selection of transformed E. coli was inoculated into 4 L of Terrific broth, grown at 37 °C to OD600 = 0.8, and induced with 1 mM IPTG at 15 °C overnight. Cells were resuspended in Buffer A (20 mM Tris pH 8.0, 300 mM NaCl, 20 mM imidazole, 5% glycerol, 5 mM βME) with protease inhibitor tablets (Roche) and frozen at −80 °C until ready for use. Cells were lysed via sonication at 4 °C for 10 minutes and centrifuged at 16,000xg for 20 minutes at 4 °C. Supernatant was incubated with TALON® metal affinity resin (Clontech) for 1 hour, washed with 20 column volumes of Buffer A, and eluted with Buffer A + 200 mM imidazole. Eluted protein was loaded directly onto a SEPAX SRT-10 SEC-300 sizing column and desalted into Buffer B (20 mM Tris pH 8.0, 5% glycerol, and 5 mM βME). Protein fractions were then loaded onto a MonoQ 5/50 GL anion exchange column (GE Life Sciences) using an AKTA PURE (GE Life Sciences) and eluted using a gradient of Buffer B + 10 mM NaCl. ΔIPMK peaks were concentrated to 10 mg/mL and used for both enzyme assays and crystallization experiments.

IPMK Kinase Assays

IPMK activity on PI(4,5)P2 (PIP2) kinase assays were performed using radiolabeled nitrocellulose capture assays as previously described8. In all kinetics experiments, lyophilized dipalmitoyl PIP2 under a vacuum was resuspended in water and sonicated with dioleyl-phosphatidylserine (DOPS) to form micelles. The final concentration of all enzymes were 10 nM, added to reactions with in vitro kinase buffer (50 mM HEPES pH 7.5, 5 mM MgCl2). Cold ATP was used at 10 mM final concentration to saturate the enzyme with respect to ATP when measuring PIP2 kinetics, and 300 μM PIP2 was used to saturate IPMK with respect to PIP2 when determining ATP kinetics. Cold ATP was spiked with purified 0.25 μCi/μL [32P]γ-ATP (Perkin Elmer) in reactions. All reactions were initiated by addition of ATP, incubated at 37 °C for one minute, and quenched by dotting onto nitrocellulose paper. One minute was determined empirically to maximize signal/noise while remaining within the steady state of the reaction. Radioactivity incorporation was detected using a phosphorimaging screen, scanned on a Typhoon phosphoimager and quantified using ImageJ. Data were processed using a standard Michaelis-Menten curve fit using GraphPad Prism.

Crystallization

ΔIPMK crystallization conditions were screened with the Joint Center for Structural Genomics (JCSG) Core Suite using the TTP Labtech Mosquito® with 100nL drops via sitting drop vapor diffusion at a final protein concentration of 3.33 mg/mL. Crystals formed after 3 days in 0.1 M sodium citrate pH 3.5, 0.8 M (NH4)2SO4 in the presence of 1 mM AMPPNP, 1 mM MnCl2, and 0.73 mM Ins(1,3,6)P4. Crystals were cryo-protected in crystallization buffer + 25% ethylene glycol for 30 minutes, and flash frozen in liquid nitrogen for X-ray diffraction data collection.

Crystal Data Collection, Structure Determination, and Refinement

Diffraction data were collected using APS beamline 21-ID-F and processed with HKL2000. Initial phases were determined using the BALBES automated molecular replacement pipeline, which empirically determined the S. cerevisiae IPMK bound to ADP (PDB: 2IF8) as the highest confidence search model. The initial structure was rebuilt using COOT and refined with phenix.refine in the Phenix software suite. Structural model figures were generated with Pymol and UCSF Chimera. Of note, 6 Ramachandran outliers were observed. Two of these 6 were part of the artificial and very poorly ordered His-tag (Chain A: H-13, and Chain B: H-13). One of the remaining 4 outliers is a native IPMK amino acid (Chain B: S261), however S261 is the second to the last assignable residue preceding the highly disordered and unassignable Ser/Glycine linker region, itself in a region of low density. Inclusion of the artificial Ser/Gly linker in the construct was required for crystallization, as also noted by others18. The remaining 3 Ramachandran outliers are all native IPMK amino acids (Chain A: S188; Chain B: K136, C220, S261), are also most likely due to poor electron density in these three areas of the model. None of the outlier amino acids are in an obvious crystallographic interface between asymmetric units, or in any interface between Chains A and B. Structure factors and atomic coordinates were deposited in the Protein Data Bank (PDB) with accession code PDB: 6E7F.