Crystallographic and kinetic analyses of human IPMK reveal disordered domains modulate ATP binding and kinase activity

Inositol polyphosphate multikinase (IPMK) is a member of the IPK-superfamily of kinases, catalyzing phosphorylation of several soluble inositols and the signaling phospholipid PI(4,5)P2 (PIP2). IPMK also has critical non-catalytic roles in p53, mTOR/Raptor, TRAF6 and AMPK signaling mediated partly by two disordered domains. Although IPMK non-catalytic functions are well established, it is less clear if the disordered domains are important for IPMK kinase activity or ATP binding. Here, kinetic and structural analyses of an engineered human IPMK lacking all disordered domains (ΔIPMK) are presented. Although the KM for PIP2 is identical between ΔIPMK and wild type, ΔIPMK has a 1.8-fold increase in kcat for PIP2, indicating the native IPMK disordered domains decrease IPMK activity in vitro. The 2.5 Å crystal structure of ΔIPMK is reported, confirming the conserved ATP-grasp fold. A comparison with other IPK-superfamily structures revealed a putative “ATP-clamp” in the disordered N-terminus, we predicted would stabilize ATP binding. Consistent with this observation, removal of the ATP clamp sequence increases the KM for ATP 4.9-fold, indicating the N-terminus enhances ATP binding to IPMK. Together, these structural and kinetic studies suggest in addition to mediating protein-protein interactions, the disordered domains of IPMK impart modulatory capacity to IPMK kinase activity through multiple kinetic mechanisms.

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 glucose 14,15 and metformin 16 . Further, IPMK binds and stabilizes the TRAF6 ubiquitin ligase signaling scaffold in macrophages, preventing ubiquitination of the TRAF6 protein to maintain signaling 17 . Thus, these important studies have established several roles IPMK plays via kinasedependent 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 date [18][19][20][21][22][23] . Although it is well established that disordered protein domains generally participate in proteinprotein interactions [24][25][26][27] , what role the IPMK disordered regions might play in modulating IPMK kinase activity has not been thoroughly explored 28 .
Here, a highly engineered human IPMK lacking its two disordered domains was designed, and designated as ∆IPMK. To determine if the absence of these disordered domains alters IPMK enzyme activity on the well-known substrate PIP 2 , detailed kinetics analyses were performed and compared to wildtype activity levels. Surprisingly, ∆IPMK has a 1.8-fold higher turnover number (k cat ) on PIP 2 than wild type IPMK. In addition, a ∆IPMK construct with an additional 33 N-terminal amino acid extension to G37 ( ext ∆IPMK) also has a 1.8-fold increase in k cat , 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 ∆IPMK maintains all the consensus features of the IPK-superfamily fold. Comparison of the ∆IPMK 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 ∆IPMK, and the K M for ATP increased 4.9-fold for ∆IPMK compared to wild type IPMK. In addition, adding back a portion of the N-terminus containing the ATP clamp partially rescued the increase in K M 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.

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 members [18][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 site 29 . Both the N-terminal and internal loop disordered domains participate in IPMK interactions with p53 12 , TRAF6 17 , AMPK 14 , and Raptor 13 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 noncatalytic functions are well established, it is unclear if they modulate IPMK kinase activity.
To examine if removal of the two IPMK disordered domains alters IPMK kinetic properties on the PIP 2 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] (Fig 1A, S1). This construct was designated ∆IPMK, and was easily expressed in bacteria and purified to homogeneity for crystallographic and enzyme kinetic analyses. Further, a ∆IPMK construct with an additional 33 N-terminal amino acid extension to G37 ( ext ∆IPMK) 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 ∆IPMK has an almost identical K M for PIP 2 as the wild type enzyme (both about ~38 μM), the k cat increases 1.8-fold for ∆IPMK to 7.4 s -1 (Table 1) When compared to the wild type IPMK k cat for PIP 2 (4.4 s -1 ), the increased k cat drives a 1.8-fold higher catalytic efficiency (k cat /K M ) of ∆IPMK over wild type IPMK (Fig 1B, Table 1). The ext ∆IPMK construct behaved almost identical to ∆IPMK with respect to the PIP 2 substrate in these assays. Thus, both ∆IPMK and ext ∆IPMK (Fig 1A) appear to be more catalytically efficient than wild type IPMK at phosphorylating PIP 2 to PIP 3 .
Crystallography shows ∆IPMK maintains the ATP-grasp fold. In parallel with the kinetic studies, crystals of the ∆IPMK protein were produced for X-ray structure determination of the enzyme. The growth of these crystals required the presence of the ATP-analog adenylylimidodiphosphate (AMP-PNP), inositol-(1,3,4,6)-phosphate (IP 4 ) substrate and 1mM MnCl 2 , 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 ∆IPMK 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-∆IPMK by molecular replacement using the yeast IPMK as the search model. Table 2 contains data collection and structure refinement statistics.
Analysis of the fold of ∆IPMK revealed the expected ATP-grasp domain structure (Fig  2A). As with all other members of this superfamily, ∆IPMK 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 ∆IPMK monomers are in each asymmetric unit of the P2 1 2 1 2 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 ∆IPMK 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).

∆IPMK core-kinase structure is similar to other IPK-superfamily members.
Comparing ∆IPMK to other IPK-superfamily members, ∆IPMK 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 Å 18 . Superposition over the Cα backbone of ∆IPMK with the human H. Sapiens IP3K gave an RMSD of 1.041 Å (Fig 3A). As expected, human IP3K diverges most significantly with ∆IPMK 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 ∆IPMK 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 1.10Å (Fig 3D). A more thorough comparison of the differences between these structures has been extensively addressed elsewhere 18 , however we note the larger IP-binding loops of IP3Ks have been proposed to prevent IP3K activity on membrane phosphoinositides 23 , and our ∆IPMK structure further supports that hypothesis.
∆IPMK has higher K M 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 (4.2Å -3.8Å) (Fig 4C), but the equivalent sequence in the ∆IPMK construct was removed and replaced with an N-terminal histidine tag (Fig 4D and S1), further suggesting an ATP-stabilizing effect of the I65-P69 sequence. This sequence is clearly not required for catalysis on the PIP 2 substrate, as the removed I65-P69 sequence in ∆IPMK 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. ∆IPMK would be predicted to have an increased K M for ATP compared to wild type, since the I65-P69 sequence was removed from ∆IPMK. Indeed, ∆IPMK has a K M for ATP of 300 μM (Fig 4E), while the ATP K M for wild type IPMK is 61 μM ( Table 1), demonstrating ∆IPMK has a 4.9-fold increased K M for ATP than wild type, despite ∆IPMK having an identical K M and 1.8-fold increased k cat for the PIP 2 kinase substrate ( Table 1). When the I65-P69 sequence was added back to ∆IPMK using the ext ∆IPMK construct (Fig 1A) the K M for ATP was rescued to 94 μM, only a 1.5-fold increase from the wild type IPMK (Table 1).
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 K M for ATP only increased about 2.3-fold for this mutant 28 , suggesting removal of the kinase-domain internal loop region alone can also change the K M for ATP. Because the change in K M for ATP in ∆IPMK (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 removed 28 , 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 (∆IPMK) results in an enzyme that is more catalytically efficient than wild type IPMK, driven by a 1.8-fold increase in k cat . Since the K M for PIP 2 is not significantly different between ∆IPMK, ext ∆IPMK, or wild type IPMK (  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 PIP 2 kinetics experiments. However, the K M for ATP is actually increased 1.5-fold for ext ∆IPMK and 4.9-fold for ∆IPMK, despite the enzyme becoming faster. Thus, the data suggest the disordered domains actually increase IPMK binding affinity for ATP because the K M for ATP decreases with increasing proportions of the disordered domains present (i.e. K M,WT < K M,ext∆ << K M, ∆ ) .
The structure of ∆IPMK 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.
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. PIP 2 ), increases ATP nucleotide binding affinity (i.e. decreases K M ), and decreases the turnover number (i.e. k cat ) (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 IPMK 21 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.
A recently published structure of apo-human IPMK had a similar but non-identical construct design as reported here 18 , and comparison of our 2.5 Å apo-∆IPMK 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 Nterminal sequence (I65-P69) was ordered directly above the ATP binding site 18 . 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 (4.2Å -3.8Å) 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 IP3K 20,23 and the amoeba IP6K 22 , 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 nucleotidebinding pocket (Fig 6E), and that removal of the clamp would increase the K M for ATP (Fig 6D). The I65-P69 ATP-clamp sequence was deleted from the ∆IPMK construct reported here, and accordingly observed a 4.9-fold increase in the K M for ATP in ∆IPMK, but only a 1.5 fold increase in ATP K M was observed when the ATP-clamp sequence was restored using the ext ∆IPMK 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 mM 30-32 , one might expect shifting the K M 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 μM 30,33 , and ATP levels can change in subcellular compartments as measured with biosensors by over 1,000fold 31,32,34 . IPMK shuttles between the nucleus and the cytoplasm 29 , but often localizes to the nucleus with a speckled, uneven distribution 3 . 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 substratebinding 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 reaction 35 . 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 diC 4 -PIP 2 ) and the apo-∆IPMK reported here has an RMSD over the Cα backbone of 0.414Å. Similarly, superposition of ∆IPMK vs. ADP/Mn 2+ -bound S. cerevisiae IPMK 21 has an RMSD of 0.995Å 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 PIP 2 not by altering PIP 2 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.

Protein Expression and Purification:
The human ∆IPMK 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 (Gly 4 -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 4L of Terrific broth, grown at 37°C to OD 600 =0.8, and induced with 1mM IPTG at 15°C overnight. Cells were resuspended in Buffer A (20 mM Tris pH 8.0, 300 mM NaCl, 20mM 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 + 200mM 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 5mM 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 + 10mM NaCl. ∆IPMK peaks were concentrated to 10mg/mL and used for both enzyme assays and crystallization experiments.

IPMK Kinase Assays:
IPMK activity on PI(4,5)P 2 (PIP 2 ) kinase assays were performed using radiolabeled nitrocellulose capture assays as previously described 8 . In all kinetics experiments, lyophilized dipalmitoyl PIP 2 under a vacuum was resuspended in water and sonicated with dioleylphosphatidylserine (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 MgCl 2 ). Cold ATP was used at 10 mM final concentration to saturate the enzyme with respect to ATP when measuring PIP 2 kinetics, and 300 μM PIP 2 was used to saturate IPMK with respect to PIP 2 when determining ATP kinetics. Cold ATP was spiked with purified 0.25 μCi/μL [ 32 P]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:
∆IPMK 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 0.1M sodium citrate pH 3.5, 0.8M (NH 4 ) 2 SO 4 in the presence of 1 mM AMPPNP, 1 mM MnCl 2 , and 0.73 mM Ins(1,3,6)P 4 . 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 others 18 . 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.

Data availability:
The datasets generated and analysed during the current study are available in the Protein Data Bank repository as 6E7F, at https://www.rcsb.org/.  Table 1, 10nM of each enzyme was used in these assays. Velocities were fit using Graphpad prism.   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 ADP 18 , highlighting Van der Waals interaction distances of I65 sidechain (grey sticks) to 5' (4.2Å) and 4' (3.8Å) carbons of the ribose moiety in ADP (sticks). D. N-terminus of ∆IPMK structure (blue) superposed on IPMK structure (magenta). Orange box shows the ordered ∆IPMK 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, ext ∆IPMK, and ∆IPMK kinetic parameters on ATP with saturating PIP 2 . Velocities were fit using Graphpad prism.  Surface and B. cartoon ribbon representations of ∆IPMK, with ADP and PIP 2 modeled into the structure. The IPMK interaction site with p53 has been mapped to IPMK exon 4 12 , 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.      Authors: Corey D. Seacrist and Raymond D. Blind Figure S1. Sequence alignment of full length human IPMK and ΔIPMK. ΔIPMK does not possess residues 1-69 and 279-373 of the full-length human IPMK sequence. 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. Table S1. Human IPMK construct expression in bacteria and crystallization. Figure S2. Protein-protein interaction sites superimpose onto IPMK ATP-and substrate-binding sites. Surface (left) and ribbon (right) representations of ΔIPMK, with ADP and PIP2 modeled into the structure. A. IPMK interaction site with AMPK has been mapped 14 to IPMK exons 4 and 6, depicted as orange. With the remainder of the IPMK kinase domain is depicted in light blue. Both ADP and an inositol phosphate kinase substrate were modeled into the structure for reference. C. IPMK interaction site with TRAF6 has been mapped 17 to exons 4 and 6, depicted as orange. Note that in both cases, binding of these proteins to IPMK would be predicted to sterically interfere with IPMK kinase activity.