Inositol polyphosphates are ubiquitous molecular signals in metazoans, as are their pyrophosphorylated derivatives that bear a so-called ‘high-energy’ phosphoanhydride bond. A structural rationale is provided for the ability of Arabidopsis inositol tris/tetrakisphosphate kinase 1 to discriminate between symmetric and enantiomeric substrates in the production of diverse symmetric and asymmetric myo-inositol phosphate and diphospho-myo-inositol phosphate (inositol pyrophosphate) products. Simple tools are applied to chromatographic resolution and detection of known and novel diphosphoinositol phosphates without resort to radiolabeling approaches. It is shown that inositol tris/tetrakisphosphate kinase 1 and inositol pentakisphosphate 2-kinase comprise a reversible metabolic cassette converting Ins(3,4,5,6)P4 into 5-InsP7 and back in a nucleotide-dependent manner. Thus, inositol tris/tetrakisphosphate kinase 1 is a nexus of bioenergetics status and inositol polyphosphate/diphosphoinositol phosphate metabolism. As such, it commands a role in plants that evolution has assigned to a different class of enzyme in mammalian cells. The findings and the methods described will enable a full appraisal of the role of diphosphoinositol phosphates in plants and particularly the relative contribution of reversible inositol phosphate hydroxykinase and inositol phosphate phosphokinase activities to plant physiology.
Myo-inositol hexakisphosphate (InsP6) is the predominant form of phosphorus storage molecule in plants, where in storage organs and tissues it may accumulate to several percent of dry weight . The enzymology of InsP6 synthesis extends to families of enzymes collectively capable of phosphorylating all six hydroxyls of the inositol ring . These include inositol kinase  and assorted inositol phosphate (hydroxy) kinases, including inositol 1,3,4-trisphosphate 5/6-kinases, also known as inositol tris/tetrakisphosphate kinases, ITPKs [4–6]; inositol polyphosphate kinases , commonly called IPK2 after the yeast ortholog  and inositol pentakisphosphate 2-kinases , commonly called IPK1, again after the yeast ortholog .
ITPK1 controls phosphate homeostasis in Arabidopsis and Atitpk1 mutants accumulate Ins(3,4,5,6)P4 and/or its enantiomer . Ins(3,4,5,6)P4 is also the dominant InsP4 isomer in Atipk1 mutants which similarly over-accumulate phosphate . Recently, two groups described phosphoanhydride bond formation on the 5-phosphate of InsP6 (synthesis of 5-InsP7, 5-PP-InsP5) catalyzed by recombinant plant ITPK1 and ITPK2 [12,13]. Others have shown that synthesis of 1,5-InsP8 (also known as 1,5-bis-PP-InsP4) occurs by 1-phosphorylation of 5-InsP7 and is mediated by VIH1 and VIH2  (see Figure 1). Deletion of Vih1 and Vih2 recapitulates constitutive Pi starvation response [15,16]. These works confirm the identity of plant InsP8 previously described [14,17]. While it seems likely that the ITPK1 contribution to phosphate homeostasis may lie in the provision of the 5-InsP7 precursor of 1,5-InsP8, in yeast 5-InsP7 is the principal agent of activation of the SPX-dependent polyphosphate polymerase VTC .
Metabolic relations of inositol phosphates and diphosphoinositol phosphates in plants.
Nevertheless, the question of which PP-InsPs and InsPs competitively control phosphate homeostasis in plants is compounded for many biological and technical reasons: (1) Plants have multiple SPX-domain proteins  with potential discrete and/or overlapping functions in phosphate homeostasis and, generally, poorly defined specificity for binding of inositol and diphosphoinositol phosphates. (2) The relative levels of InsP7 and InsP8 species revealed by radiolabeling [10,12,17], InsP7 greater than InsP8, are reversed in recent gel electrophoresis determinations . In the latter study, InsP8 levels are substantially greater than InsP7 in phosphate replete Arabidopsis. (3) The biochemical activities of enzymes such as ITPK1 and VIH1/2 are not fully defined; for VIH1/2, full-length protein has not been studied, while for both ITPK1 and VIH1/2, the reversibility of the hydroxykinase- and phosphokinase functions of these enzymes is not described.
Here, we focus on ITPK1, disruption of which in Arabidopsis has profound effect on InsP6 and InsP7 as well as Ins(3,4,5,6)P4 [10,20]. To study ITPK1, we employ a suite of accessible methodologies that have not previously been applied to diphosphoinositol phosphates. They afford the opportunity to distinguish known and novel diphosphoinositol phosphates and to study the reversibility of enzymes that catalyze their interconversion. The pathways in plants in which ITPK1 participates and those revealed by this study are shown in Figure 1.
Materials and methods
The structures of all inositol phosphates and diphosphoinositol phosphates described in this study are numbered according to the 1d-numbering convention and are shown in Supplementary Figure S1.
InsP6 was obtained from Merck Millipore (Product No. 407125). An acid-hydrolysate of phytate (Sigma P-8810) was prepared and used as chromatographic standard according to Madsen et al. . All InsP5 isomers used as substrate for ITPK1 were obtained from Sichem as decasodium salts. Ins(1,3,4,5,6)P5 obtained from Sichem showed evidence of phosphate migration between cis-vicinal hydroxyls. Ins(1,3,4,5,6)P5 used as substrate for IPK1 was synthesized according to published procedures  and did not show evidence of phosphate migration. InsP4s and InsP3s were obtained from Cayman Chemical Company or were synthesized according to published procedures . Comparisons made between different enantiomers of enantiomeric pairs were performed with compounds obtained from one source only. Diphosphoinositol phosphates (Supplementary Figure S1): 1-InsP7, 3-InsP7, 5-InsP7 and 5PP-Ins(1,3,4,6)P4 were synthesized similarly to published procedures . 4-InsP7 and 6-InsP7 were synthesized by Dr Henning Jessen, Institute of Organic Chemistry, and the Centre for Integrative Biological Signalling Studies, University of Freiburg, Germany. Acid-catalyzed migration of phosphate on 32P-labeled InsPs was performed according to Stephens and Downes . 32P-labeled InsPs were used directly for HPLC without processing to remove nucleotides.
AtITPK1 was cloned into the pOPINF plasmid , to generate a construct for expression of protein with a 3C cleavable C terminal His tag. Primer sequences used were 5′-AAGTTCTGTTTCAGGGCCCGATGTCAGATTCAATCCAGGAAAG-3′ and 5′-ATGGTCTAGAAAGCTTTAGACATGATTCTTCTTAGTGAC-3′, where the sequence in italics is specific for recombination with the pOPINF vector. Linearized pOPINF plasmid, digested with HindIII and KpnI, was recombined with the PCR product using the In Fusion HD enzyme kit (Clontech) and subsequently transformed into Escherichia coli Stellar cells (Clontech). Colonies were confirmed by PCR amplification and transformed into E. coli Rosetta 2(DE3)pLysS (Novagen) for protein expression.
ITPK1:His in pOPINF was expressed at 0.1 mM IPTG in Rosetta2 E. coli at 18°C overnight and lysed in 25 mM HEPES pH 7.5, 350 mM NaCl, 1 mM DTT, 20 mM imidazole, 1% triton using a French Pressure Cell. Clarified lysate was loaded onto a 5 ml Ni-NTA column (Qiagen) equilibrated in 25 mM HEPES pH 7.5, 350 mM NaCl, 1 mM DTT, 20 mM imidazole and eluted in the same buffer with 20–250 mM imidazole over a 100 ml gradient. The recombinant protein was subsequently purified using a Superdex 75 10/300 column in 20 mM Tris pH7.5, 200 mM NaCl, 2 mM DTT, 10% glycerol, concentrated and stored at −80°C.
IPK1 was prepared according to Whitfield et al.  and assayed under the same conditions as ITPK1.
ITPK1 (typically 0.036 µM, with InsP4; or 33.6 µM, with InsP6) was incubated at 25°C in 10 mM HEPES pH 7.5, 1 mM MgCl2 with 1 mM ATP and 0.5 mM or 1 mM inositol phosphate substrate for periods up 2 h. Assays were typically of 50–100 µl volume.
Kinase assays run under regenerating conditions also contained 5 mM phosphocreatine, 3 U creatine kinase and 1 mM ATP under standard conditions. For ATP Km calculations ITPK1 was assayed with 0.08–10 mM ATP for 15 min, 40 min or 2 h, depending on the substrate used.
Reactions were terminated by the addition of an equal volume of 60 mM (NH4)2HPO4, pH 3.5, with or without incubation at 95°C for 3 min. Samples were clarified by centrifugation at 14 000×g for 5 min and subsequently diluted with an equal volume of deionized water. The injection volume was typically 50 µl.
For incorporation of 32P, assays run in 50 µl volume under ATP-regenerating conditions, 0.5 mM ATP, were supplemented with 10 kBq of [γ-32P] ATP 3000 Ci mmol−1 (NEG 502A, PerkinElmer). Assays were run for 2 h at 25°C and reactions stopped by freezing at –20°C. Typically, 25 µl aliquots of the reaction products were diluted to 100 µl with water and 50 µl injected.
HPLC was performed according to Whitfield et al.  using either 0.6 M methanesulfonic acid or 0.8 M HCl in buffer reservoirs. Inositol phosphates were detected as ferric complexes by UV detection at 290 nm . In some experiments, a second channel of UV data was collected at 254 nm to detect nucleotides. Inositol phosphates are also detected but with reduced sensitivity at this wavelength.
In some experiments, inositol phosphates were analyzed by strong anion exchange high-performance liquid chromatography using a 4.6 × 235 mm Partisphere (Whatman) SAX column according to Kuo et al. . Inositol phosphates were monitored by incorporation of 32P, detected by Cerenkov counting in a Radiomatic 515 series Flow Detector (Canberra Packard). The detector was set to an integration interval of 12 s, and a second channel of data was collected from a UV detector placed upstream of the radio-detector and set to 254 nm to monitor nucleotides.
For calculation of kinetic parameters, peak areas of inositol phosphates were integrated with ChromNav v.2 (Jasco). For the reproduction of HPLC profiles, data were exported from ChromNav v.1 or v.2 (Jasco) or Flo-One (Canberra Packard) software as x,y data and redrawn in GraFit v.7 .
Measurement of absorbance spectra of inositol phosphate ferric complexes
UV spectra were measured for Ins(3,4,5,6)P4 and InsP6 in 0.4 mM ferric nitrate, 0.4 M methanesulfonic acid, 0.67% (w/v) perchloric acid in quartz glass cuvettes.
Substrate docking calculations
The ITPK1 from Entamoeba histolytica  shares 55% sequence identity (60% including conservative substitutions) with the Arabidopsis enzyme when calculated over active site residues alone. For the human enzyme, this rises to 70% identity (80% with conservatively varied substitutions). For this reason, it was decided to build a molecular model for Arabidopsis ITPK1 by reference to the structure of the human enzyme. The amino acid sequence of Arabidopsis ITPK1 (Genbank, At5g16760) was accordingly submitted to Phyre2  and a structural model for the enzyme built by reference to PDB ID: 2QB5 (Crystal Structure of Human Inositol 1,3,4-Trisphosphate 5/6-Kinase (ITPK1) in Complex with ADP and Mn2+). Regularization of polypeptide geometry was carried out using Coot  where necessary. The final model contained ATP and two magnesium ions. Docking of experimentally verified inositol phosphate substrates (Supplementary Table S1) was carried out using Autodock Vina . A fully flexible model for the ligand and fixed model for the receptor were employed. From the energy-ranked docked conformations (poses) calculated for each substrate, the best ‘productive’ pose was selected on the basis of (i) lowest energy and (ii) positioning of the acceptor site for hydroxykinase (oxygen) or phosphokinase (phosphorus) activity within a maximum of 3.8 Å of the γ-phosphate phosphorus atom of ATP. The cutoff distance of 3.8 Å was chosen to reflect the fact that a rigid model for the protein was employed in the calculation. Productive poses are those that are reasoned to most likely lead to phosphorylation. Poses were rendered in PyMOL (The PyMOL Molecular Graphics System, Schrödinger, LLC.). Docking data and models are provided as Supplemental Information.
Accession Numbers Sequences corresponding to the subjects of this study can be found in the GenBank/EMBL databases under the following accession numbers: ITPK1 (At5g16760), IPK1 (At5g42810).
We employed anion exchange chromatography on an acid- and hydroxide-stable quaternary amine-functionalized latex (CarboPac PA200, Dionex) as we have done before . This chemistry allows the use of the widest range of eluents. The use of acid allows detection by post-column complexation of inositol phosphate with ferric ion (in acid conditions) and detection of the ferric complex at 290 nm . While the original literature does not explain the physicochemical basis of the absorbance, we may assume from its magnitude (we measured an extinction coefficient of ca. 3200 M−1 cm−1 for the complex with InsP6) that the absorbance at 290 nm arises from charge transfer transitions, i.e. ligand-field splitting effects of inositol phosphate disturbing symmetry of the octahedral geometry of hexa-co-ordinate hydrated ferric ion . Moreover, while this is a method of choice for analysis of inositol phosphates in animal feed digestive situations where inositol phosphates are abundant , we consider that the sensitivity of the method is not widely appreciated for other purposes.
We used both methanesulfonic acid and HCl as eluents. Separations with HCl of InsP6, 5-InsP7, 1-InsP7 and 4-InsP7 are shown (Figure 2A) and of InsP6, 3-InsP7 and 6-InsP7 (Figure 2B). Enantiomeric pairs, 1-InsP7/3-InsP7 and 4-InsP7/6-InsP7 co-eluted (cf. Figure 2A and B), distinct from the meso-compound 5-InsP7, well resolved from 5-PP-Ins(1,3,4,6)P4 which eluted earlier, shortly after InsP6 (Figure 2C). The peaks shown represent injections of ca. 1–2 nmol of compound. The 4-InsP7 and 6-InsP7 samples available to us showed the presence of some InsP6 impurities. The absorbance spectra of Ins(1,4,5,6)P4 and InsP6 complexes of ferric ion, are shown (Figure 2D). A calibration curve of detector response for InsP6 is shown for methanesulfonic acid eluent, which gives a flatter baseline, in Supplementary Figure S2. The method is sensitive enough for detection of ca. 50 pmol of InsP6 on-column in our hands, but methanesulfonic acid at this concentration did not elute diphosphoinositol phosphates. In measurements of the kinetic parameters of ITPK1 reported later in this manuscript we observed approximately equivalent peak areas for InsP6 and 5-InsP7 eluted with the stronger acid HCl. The estimations are explained at that juncture.
Separations of inositol phosphates and diphosphoinositol phosphates on CarboPac PA200 eluted with HCl.
The opportunity to resolve and measure diphosphoinositol phosphates beside ‘lower’ inositol phosphates without the use of radiolabel is an advance on separations on Partisphere SAX columns, whose silica-based chemistry is not stable under the acid conditions required of post-column complexation with ferric ion. The use of UV detection further allows sampling of data at rates (typically up to 100 Hz) offering resolution beyond fraction collection and scintillation counting, as exemplified [14,36], and beyond on-line radioactivity counting . The method is more convenient and ca. 50–100-fold more sensitive than 31P-NMR (as used to verify the InsP6 kinase activity of ITPK1, Figure 2, ). In our hands, the method is several-fold less sensitive than the metal dye detection-HPLC method of Mayr  and is much less compromised by the baseline changes that we have observed of gradients capable of resolving InsP2 through to InsP7 using that method. Moreover, the method we use requires minimal sample preparation as shown for complex animal digesta matrices  and, as we show in the following, can be combined with UV detection at wavelengths allowing simultaneous measurement of nucleotides to follow the reversibility of kinase reactions.
We used these methods to examine the substrate specificity of Arabidopsis ITPK1, which belongs to a class of enzyme that has been shown to possess 1-, 5- and 6-hydroxykinase and (InsP6) 5-phosphokinase activities. The availability of pairs of enantiomeric substrates, not available in the radiolabeled form, beside meso-compounds, allowed analysis of the enantiospecificity of the enzyme. Of particular interest to us were InsP4 substrates, since Ins(3,4,5,6)P4 is elevated in itpk1 and ipk1 mutants that share a common misregulation of phosphate homeostasis [10,11,38] attributed to deregulated diphosphoinositol phosphate synthesis [15,16]. Of the enantiomeric pairs Ins(1,2,4,6)P4/Ins(2,3,4,6)P4, Ins(1,3,4,5)P4/Ins(1,3,5,6)P4 and Ins(1,4,5,6)P4/Ins(3,4,5,6)P4 (all shown in Supplementary Figure S1), only the latter pair were substrates under ATP-regenerating conditions. The enantiomers were phosphorylated on the 3- and 1-positions, respectively (Figure 3A–H). The activity was robust for both enantiomers, but with much greater activity for Ins(3,4,5,6)P4 (Figure 3F). The meso-isomer Ins(1,3,4,6)P4, a canonical substrate of phosphoisomerase activity of EhITPK1and human ITPK1  was not a substrate for hydroxykinase or phosphoisomerase activity (Figure 3G), consistent also with the lack of Ins(1,3,4,5)P4 hydroxykinase or phosphoisomerase activity (Figure 3C).
Arabidopsis ITPK1 has robust enantiospecific hydroxykinase activity against Ins(3,4,5,6)P4.
We also tested many InsP3 substrates including the enantiomeric pair Ins(1,4,6)P3/Ins(3,4,6)P3 (the structures are shown in Supplementary Figure S1 and the HPLC profiles of the reaction products are shown in Supplementary Figure S3A–D). Consistent with the phosphorylation of the enantiomeric pair Ins(1,4,5,6)P4/Ins(3,4,5,6)P4, ITPK1 showed phosphorylation of Ins(1,4,6)P3 at the 3-position and Ins(3,4,6)P3 at the 1-position. The presence of an Ins(1,3,4,5)P4 peak for all three InsP3s tested, possibly indicates phosphoisomerase activity at the level of InsP4, under these (non-regenerating for ATP) conditions. The lack of Ins(1,4,5,6)P4/Ins(3,4,5,6)P4 product for Ins(1,4,6)P3 or Ins(3,4,6)P3 substrates (Supplementary Figure S3B,C) discounts 5-phosphorylation of these substrates, while the ratio of substrate to product peaks reveals that of the pair, Ins(3,4,6)P3 is the preferred substrate.
In light of the consistent pattern of phosphorylation of 3- and 1- hydroxyls of Ins(1,4,6)P3/Ins(1,4,5,6)P4 and Ins(3,4,6)P3/Ins(3,4,5,6)P4 pairs, respectively, we modeled the binding of the enantiomers Ins(1,4,5,6)P4 and Ins(3,4,5,6)P4 (to ITPK1 (Figure 4)). We used as reference the specificity subsite nomenclature described for the crystal structure of E. histolytica ITPK1 . On this basis our modeling suggests significant interaction of the 4-phosphate of Ins(1,4,5,6)P4 and of the 6-phosphate of the Ins(3,4,5,6)P4 in site F (Figure 4A,B). Similarly, we posit that the respective 1- and 3- phosphates of Ins(1,4,5,6)P4 and Ins(3,4,5,6)P4 make contacts in site C. Collectively, residues in these subsites are likely determinants of the reactivity of ITPK1. Indeed, K188A mutation abrogates the activity of ITPK1 against InsP6 in kcs1 vip1 ddp1 yeast, as does D288A mutation . Irrespective of the fine detail, the symmetry-generating capacity of ITPK1's hydroxykinase activity (Supplementary Table S2) suggests fundamental differences in the pose of enantiomeric substrates. We posit that the inositol ring can bind either in ‘obverse’ (axial 2-OH group facing up and pointing out of active site) as presented for the Ins(1,4,5,6)P4 (Figure 4A), or ‘reverse’ (2-OH group face down) orientation as presented for Ins(3,4,5,6)P4 (Figure 4B).
A structural model for enantiospecific kinase activity of
AtITPK1 towards InsP4 substrates.
Because Ins(3,4,5,6)P4 is a known physiological substrate of human ITPK1 , we next compared in more detail the relative activity of ITPK1 for Ins(3,4,5,6)P4 and the recently identified InsP6 substrate. At low enzyme, InsP6 was not phosphorylated while Ins(3,4,5,6)P4 was (Figure 5A,B). With 100-fold more enzyme, we confirmed the production of 5-InsP7 from InsP6 (Figure 5C), as reported [12,13]. Because a central tenet of the energy status signaling role of diphosphoinositol phosphates (5-InsP7, specifically) is the high (1–1.2 mM) Km for ATP of mammalian IP6K  and kcs1 [41,42], a more recent estimate for IP6K1 is 0.35 mM , we measured kinetic parameters for ATP with these two substrates (Figure 5D,E). Ins(3,4,5,6)P4 was, by more than two-orders of magnitude (Vmax 8640 ± 025 nmol min−1 mg−1 vs 40 ± 3 nmol min−1 mg−1), the stronger substrate. The Km ATP values 1.22 ± 0.37 mM and 0.77 ± 0.19 mM for InsP6 and Ins(3,4,5,6)P4, respectively, identify both InsP6 and Ins(3,4,5,6)P4 as metabolites that are responsive to energy status, but the much greater activity towards Ins(3,4,5,6)P4 focuses attention on this isomer and on historic detailed analyses of inositol phosphate isomerism in plants [2,36,44,45]. Significantly, Ins(3,4,5,6)P4 was identified in the duckweed Spirodela polyrhiza , where it is a substrate for phosphorylation by a cytosolic 1-hydroxykinase [2,46]. The pronounced effect of mutation of ITPK1 on phosphate homeostasis  gives physiological context to this activity.
Comparison of Ins(3,4,5,6)P4 and InsP6 as substrates of Arabidopsis ITPK1.
To estimate the sensitivity of the method for detection of diphosphoinositol phosphates, we took advantage of the characterized InsP6 kinase activity of ITPK1 to synthesize 5-InsP7 from the readily available precursor InsP6. We note that the lack of commercial availability of diphosphoinositol phosphates does not allow easy comparison of the chemical stability or purity of different diphosphoinositol phosphates between laboratories, or their batch-to-batch variation, particularly given the difficulty in separation, detection and quantification. Consequently, we set up an assay with sufficient enzyme to effect the substantive conversion of InsP6. The data (Supplementary Figure S4A–C) show that for HPLC runs from assays with constant starting InsP6, the sum of the integrated peak areas for substrate and product (InsP6 + InsP7) did not vary substantially despite greater than 50% reduction in InsP6 peak area. Thus, we conclude that the sensitivity of the detection of InsP6 and 5-InsP7 is very similar. Moreover, an example chromatogram (Supplementary Figure S4D) shows that with ease it is possible to detect 5-InsP7 at 1% conversion in an assay where approximately one-quarter of the assay products from a 50 µl assay with 1mM InsP6 were applied to the column (ca. 125 pmol). Close inspection of the baseline (see inset) makes apparent that the sensitivity of the method is rather better than this, probably approaching 20 pmol. For the purpose of illustrating the utility of our methods for in vivo measurements, Supplementary Figure S5 shows example chromatograms from the seeds of Arabidopsis wild type, ipk1-1 and mrp5-2 mutants (lines described ) and Supplementary Figure S6 shows example chromatograms of InsP6 fractions purified from maize and rice bran .
We also tested whether ITPK1 has activity against Ins(1,3,4,5,6)P5 (for structures of InsP5s see Supplementary Figure S1). The data of Figure 6 show that Ins(1,3,4,5,6)P5 (InsP5 [2-OH]), unlike its metabolic neighbors, Ins(3,4,5,6)P4 and InsP6 , is not a substrate. We point out that the particular substrate obtained from Sichem had almost equimolar InsP5 [2-OH] and InsP5 [1/3-OH] (Figure 6B). The lack of enzymatic product with InsP5 [2-OH]/[1/3-OH] is consistent with the lack of secondary products from Ins(1,4,5,6)P4 and Ins(3,4,5,6)P4] (Figure 3). We suggest that Ins(1,3,4,5,6)P5 binds in the same mode as Ins(3,4,5,6)P4, but that the axial 2-OH is inappropriately oriented for phosphorylation. Were it to adopt the pose of InsP6, we would expect to see a 5-PP-Ins(1,3,4,6)P4 product (structure shown in Supplementary Figure S1) which elutes after InsP6 (Figure 2). The lack of this activity distinguishes ITPK1 from kcs1  and IP6K, as noted .
Arabidopsis ITPK1 shows enantiospecific phosphokinase activity on InsP5.
Ins(1,3,4,5,6)P5 is, however, not the only InsP5 present in plants, multiple peaks of InsP5 have been detected by radiolabeling [2,11,14,36,44,45] and in some cases enantiomers resolved. Because Partisphere SAX columns do not adequately resolve Ins(1,2,3,4,6)P5, from Ins(2,3,4,5,6)P5 or its enantiomer Ins(1,2,4,5,6)P5 [2,36,44,45], we tested the ability of ITPK1 to phosphorylate all isomers of InsP5 on the CarboPac PA200 column.
InsP5 isomers bearing a hydroxyl on the 1-, 2-, 3-, 4- and 5-positions (see Supplementary Figure S1 for structures) were not substrates (Figure 6A–E), but Ins(1,2,3,4,5)P5 (InsP5 [6-OH]) yielded a novel product that is not InsP6 (cf. Figure 6F–G). Given the elution between InsP5s and InsP6, the most parsimonious explanation is that the product is a novel PP-InsP4. We suggest 5-PP-Ins(1,2,3,4)P4 (structure shown in Supplementary Figure S1), partly on consideration that neither of Ins(1,2,3,5,6)P5 or Ins(1,2,3,4,5)P5, nor any of the other InsP5 isomers, are phosphorylated on the vacant hydroxyl (Figure 6A–E), but also on account of the discrete retention time, distinct from 5-PP-Ins(1,3,4,6)P4 (cf. Figure 2C). We summarize the inositol phosphate hydroxykinase and inositol phosphate phosphokinase reactions catalyzed by ITPK1 in Supplementary Table S2 and present speculative binding modes for Ins(1,2,3,4,5)P5 and InsP6 to ITPK1 in Figure 7A,B.
A structural model for phosphokinase activity of ITPK1.
The two ligands adopt similar poses in their predicted low energy complexes: binding modes quite dissimilar to those seen for InsP3 or InsP4 ligands. Specifically, for InsP6, the axial 2-phosphate occupies site C and equatorial 3-phosphate site B of the canonical specificity subsite set described by Miller et al. . However, for both Ins(1,2,3,4,5)P5 and InsP6 ligands the inositol ring rotates to lie roughly perpendicular to that seen for docked InsP3 and InsP4 ligands and the 4-phosphate consequently occupies site E. This forces the 5-phosphate to occupy site A, the site of phosphotransfer. The 1- and 6-phosphates (where they exist) occupy new sites which we name D′ and F′, respectively. Subsite D′ involves residues Asp63, Asp150 and His156, whilst F′ involves His156 and Ser225. These predicted subsites presumably play a role in helping to ameliorate the problems arising from the need to accommodate the steric bulk and phosphate crowding of these substrates.
The commercial unavailability of PP-InsPs, especially PP-InsP4s for which there are 30 theoretical possibilities, five for each InsP5 ‘parent’, does not allow facile identification of the novel ITPK1 product on chromatographic grounds. The near micromole quantities required for NMR analysis preclude easy confirmation of identity.
Nevertheless, the identification of predominant Ins(3,4,5,6)P4 1-hydroxykinase activity for ITPK1 affords explanation of itpk1 phenotype (disruption of InsP6 synthesis and accumulation of Ins(3,4,5,6)P4 [10,20]). Because ITPK1 and IPK1 individually control phosphate homeostasis, individually accumulate Ins(3,4,5,6)P4 [10,11,38] and share substrates and products, we next tested whether ITPK1 and IPK1 constitute an ATP-responsive metabolic cassette linking Ins(3,4,5,6)P4 to 5-InsP7. We combined the two enzymes with nucleotide and Ins(3,4,5,6)P4. We added equimolar ITPK1 and IPK1 and either buffered ATP with an ATP-regenerating system (this maintains high ATP: ADP ratio) or excluded the ATP-regenerating system, varying ATP: ADP ratio.
In the ATP-regenerating system with 0.1 mM ATP, Ins(3,4,5,6)P4 was converted via Ins(1,3,4,5,6)P5 to InsP6 (Figure 8A) and at 1 mM ATP in the regenerating system to 5-InsP7 (Figure 8B). Without the regenerating system, but with 20 : 1 ATP: ADP ratio, Ins(3,4,5,6)P4 was robustly converted via Ins(1,3,4,5,6)P5 to InsP6 (Figure 8C), while InsP6 was not significantly phosphorylated (Figure 8D). These experiments show that combination of ITPK1 and IPK1 drives 5-InsP7 synthesis from Ins(3,4,5,6)P4 in an ATP-dependent manner.
ITPK1 and IPK1 comprise an energy-responsive metabolic cassette. Products of reaction coupled with the two enzymes were analyzed by CarboPac PA200 HPLC.
The reversibility of IPK1 [9,48], kcs1  and IP6K [40,49], led us to test the reversibility of ITPK1. With ITPK1 and IPK1 combined at a 1 : 20 ATP: ADP ratio, InsP6 was converted via Ins(1,3,4,5,6)P5 to Ins(3,4,5,6)P4 and/or Ins(1,4,5,6)P4 with concomitant generation of ATP (Figure 8E).
At the same nucleotide ratio, ITPK1 generated InsP6 alone from 5-InsP7 with the concomitant production of ATP (Figure 9A,B), but was without effect on InsP6 under the same conditions (Figure 9C,D). The sensitivity of the direction of the reaction catalyzed by ITPK1 to nucleotide ratio is in part similar to IP6K, but unlike for IP6K , we did not observe phosphotransfer from IP6 to ADP, that function is efficiently performed by IPK1 [9,48,50].
Arabidopsis ITPK1 shows symmetry conserving 5-InsP7-ADP phosphotransferase activity.
By labeling of kinase reaction products with [γ-32P] (see Supplementary Figure S1) and use of more common Partisphere SAX HPLC, we determined the enantiospecificity of ITPK-mediated dephosphorylation of Ins(1,3,4,5,6)P5. First, we synthesized Ins([32P]1,3,4,5,6)P5, and Ins(1,[32P]3,4,5,6)P5, from Ins(3,4,5,6)P4 and Ins(1,4,5,6)P4, respectively (Figure 10A,B). Again, we confirmed that Ins(3,4,5,6)P4 is the much stronger substrate. Next we incubated Ins([32P]1,3,4,5,6)P5 with ITPK1 at unlabeled nucleotide ratio (ATP:ADP, 1 : 20) favoring phosphotransfer to ADP. We did not observe the production of [32P]InsP4, despite synthesis of [32P]ATP and an increase in ATP peak area (in the UV254 channel). Significantly, the reaction did not generate orthophosphate product (Figure 10C,D). These data show that ITPK1 has a preferential reversible Ins(3,4,5,6)P4 1-kinase /Ins(1,3,4,5,6)P5 1-phosphotranferase to ADP (ATP synthase) activity, i.e. the phosphorylation of the preferred enantiomeric (asymmetric) substrate to meso- (symmetric) product is reversible in generating the same enantiomer of InsP4. Using this more common HPLC method we were also able to compare the efficiency of phosphorylation of Ins(3,4,5,6)P4, Ins(1,2,3,4,5)P5 and InsP6 substrates (Figure 10E–G). Ins(3,4,5,6)P4 was, again, much the preferred substrate, with Ins(1,2,3,4,5)P5 a slightly weaker substrate than InsP6. In Supplementary Figure S7, we show that the presumed 5-PP-Ins(1,2,3,4)P4 product of phosphorylation of Ins(1,2,3,4,5)P5 elutes between Ins(2,3,4,5,6)P5 (InsP5 [1-OH]) and InsP6 on Partisphere SAX HPLC.
Partisphere SAX HPLC analysis of phosphotransfer reactions catalyzed by ITPK1.
Despite historic description of inositol phosphates more highly charged than InsP6 in amoeboid organisms [51,52], animals [53,54] and plants [44,55,56], the identity of inositol phosphate kinases in plants capable of forming phosphoanhydride bonds has until recently proved enigmatic [12,13]. Because diphosphoinositol phosphates have garnered attention as a cellular signal of eukaryotic energy status, reviewed [42,57,58], the recent works [12,13] focus attention on 5-InsP7 as an agent of energy signaling in plants.
By measuring Km ATP for Arabidopsis ITPK1 we show that diphosphoinositol phosphate synthesis from Ins(3,4,5,6)P4 is responsive to nucleotide ratio, which taking account of AMP is encapsulated as a metabolic concept in the term energy charge . The low affinity of InsP6 kinase for ATP  is a central tenet of the signaling role of 5-InsP7. Similarly, the Km ATP of Arabidopsis ITPK1 (1.2 mM) poises the activity of this enzyme within the estimated physiological range of this nucleotide in the cytosol of plants [60–62]. A logical extension is that the physiological levels of all ITPK1 substrates and products, including Ins(3,4,5,6)P4 and Ins(1,3,4,5,6)P5, InsP6 and 5-InsP7, and for that matter Ins(1,2,3,4,5)P5 and the presumed 5PP-Ins(1,2,3,4)P4, are responsive to energy status. Here, by defining the reversibility of ITPK1 we show that the extent and direction of flux between Ins(3,4,5,6)P4 and 5-InsP7 is responsive to nucleotide ratio (Figure 1).
While 5-InsP7 is a reported regulator of phosphate homeostasis in yeast , it has been shown in metazoans, the HCT116 cell line, that both 5-InsP7 and 1,5-InsP8 are responsive to extracellular Pi, but 1,5-InsP8 more so . This is explained in part by inhibition of phosphatase activities of PPIP5K1 and PPIP5K2 by Pi, and activation, for PPIP5K2 of kinase activity.
The plant orthologs of PPIP5K have proved recalcitrant to study as full-length enzymes, but plants bearing deletion of Vih1 and Vih2 show constitutive Pi starvation responses recapitulating itpk1 phenotype [10,15]. The activities of the separated recombinant kinase and phosphatase domains of VIH1 and VIH2 [13–16] are shown (Figure 1). What is not clear is how full-length VIH is influenced by different nucleotides or nucleotide ratio in vitro or in vivo. Since recombinant full-length ScVip1 converts 5-InsP7 to InsP8 at supra-mM Mg2+-ATP and produces InsP6 at sub-mM Mg2+-ATP  we may expect VIH1/2 to do the same, but whether VIH1/2 (or ScVip) show diphosphoinositol phosphate-‘driven’ ATP synthase activity (from InsP7 or InsP8) in the manner of IP6K and Kcs1 [40,41,49] and AtITPK1 (this study) is not resolved. Clearly, the example of AtITPK1 illustrates how reversibility of the activity(s) of full-length VIH1/2 is critical to our understanding of phosphate homeostasis, since between them ITPK1 and VIH1/2 set the balance between 5-InsP7 and 1,5-InsP8 as ligands of SPX-domain-containing proteins.
Reversibility of plant inositol phosphate (hydroxyl) kinases is, however, well documented for ITPKs [6,65] and IPK1 [9,48], the latter echoing earlier work on a mung bean activity  for which physiological context in germinative ATP synthesis was predicted as early as 1963 . Consideration of these works shows how enzymes such as IPK1 with large equilibrium constants  can drive phosphoanhydride formation. Moreover, they show that phosphoanhydride formation is not a priori a facet of ‘high-energy’ status of either substrate, it can for reversible enzymes merely reflect prevailing substrate concentration.
Considering AtITPK1's contribution to InsP and PP-InsP metabolism, Saiardi and co-workers  have shown by heterologous expression of ITPK1 orthologs in yeast how inositol phosphate metabolism may be cryptic, or at least hidden to conventional radiolabeling. Without measurement of specific radioactivity of metabolite pools, labeling studies run the risk of missing these ‘cryptic’ pathways. The work of Stephens and Downes  is a notable exception that remarkably defined a pathway for phosphorylation of Ins(3,4,6)P3 via Ins(3,4,5,6)P5 to Ins(1,3,4,5,6)P5 that is paralleled in duckweed [2,36,46]. Given the identification of InsP7 and InsP8 in duckweed , it is likely that InsP8 synthesis therein is catalyzed through the contribution of orthologs of AtITPK1 and AtIPK1 whose reversible nature in 5-InsP7 synthesis/turnover is revealed here (Figure 1).
Moreover, our analysis suggests that ITPK1's substantially greater activity for Ins(3,4,5,6)P4 over InsP6, coupled through IPK1, is ‘matched’ to the different pool sizes of Ins(3,4,5,6)P4 and InsP6 . Indeed, InsP6 is typically two orders of magnitude more strongly labeled, but not in vegetative duckweed [2,36]. While we caution again of the limitations of radiolabeling without independent measurement of pool size, such a proposition explains well the phenotype of itpk1 mutants, viz. increase in Ins(3,4,5,6)P4 and decrease in 5-InsP7.
Most interestingly, perhaps, our work reveals how evolution has assigned the same function, an eminently reversible and nucleotide-sensitive InsP6/5-InsP7 phosphotransferase activity, to wholly different enzymes in plants and mammals. ITPK1, and perhaps ITPK2, assume the role of IP6K, or vice-versa, despite the presence of ITPK1 homolog in mammals. This places special emphasis on ITPK1. As we have shown, itpk1 mutants make less InsP6, InsP7 and, most likely, InsP8, and hyperaccumulate phosphate . Others [15,16] have collectively assigned special function to InsP8 as the cognate binding partner for SPX1 which interacts with the master transcriptional regulator, PHR1, of phosphate starvation responses. Consistent with studies in barley leaves , we have previously shown a shoot-specific near doubling in ATP levels on phosphate starvation. This increase, also in ATP/AMP ratio, was associated with a shoot-specific increase in InsP7. This is consistent with the nucleotide ratio-dependent control of ITPK1 that we demonstrate here. Others, however, have reported that whole seedlings starved of phosphate, and hence showing full phosphate starvation responses, increase nucleotides and ATP/ADP ratio approximately 2-fold on re-supply of phosphate .
Of course, plants are photosynthetic autotrophs, which yeast and mammals are not. Plants couple phosphate import across the chloroplast membrane to export of triose phosphate for growth, in an obligatory manner. These points caution that there are substantive differences in phosphate homeostasis between eukaryotes, extending to the engagement of multiple SPX-domain proteins in plants . Just as there are substantive differences in the fundaments of phosphoinositide signaling between plants, animals and yeast we may anticipate differences in control of phosphate homeostasis at the level of diphosphoinositol phosphate interaction with SPX proteins. The evolutionary diversification of ITPK [6,10,12,13,65] and SPX  families in plants say as much. Significantly, we have shown that ITPK2, despite its InsP6 kinase activity [12,13] does not regulate phosphate starvation responses, nor do ITPK3 and ITPK4 . If plants engage diphosphoinositol phosphates in as wide a range of physiological functions as animal cells, we can expect ITPK1 and ITPK2 to be engaged in other aspects of plant physiology, but for plants — integration with photosynthesis is likely to be critical. Here, tools that allow simultaneous measurement of subcellular nucleotide pools  could be brought to bear to elucidate the complex relationships between intracellular phosphate, nucleotides and diphosphoinositol phosphates. Nevertheless, the responsiveness of the direction of ITPK1 activity to the nucleotide ratio is likely a critical control point.
Open access for this article was enabled by the participation of University of East Anglia in an all-inclusive Read & Publish pilot with Portland Press and the Biochemical Society under a transformative agreement with JISC.
The authors declare that there are no competing interests associated with the manuscript.
H.W., G.W., C.S., A.M.R., A.M.H. and C.A.B. performed experiments. H.W., A.M.R., B.V.L.P., A.M.H. and C.A.B. designed the study. C.A.B. wrote the manuscript with contributions from all authors.
Funding supporting this study was obtained by C.A.B. and B.V.L.P. C.S. acknowledges the support of a BBSRC Norwich Research Park Doctoral Training Studentship [grant no. BB/M011216/1] with contribution from AB Vista. B.V.L.P. is a Wellcome Trust Senior Investigator [grant no.101010].
We thank Huifen Kuo and Tzyy-Jen Chiou (Agricultural Biotechnology Research Center, Academia Sinica, Taiwan) for helpful discussions of the work, and our reviewers for their comments that have improved the manuscript.
Arabidopsis thaliana inositol pentakisphosphate 2-kinase AtITPK1, Arabidopsis thaliana inositol tris/tetrakisphosphate kinase 1
Dictyostelium discoideum inositol tris/tetrakisphosphate kinase 1
ethylenediamine tetra-acetic acid
4-(2-hydroxyethyl)-1-piperazineethane sulfonic acid
high-pressure liquid chromatography
Homo sapiens inositol tris/tetrakisphosphate kinase 1
inositol hexakisphosphate kinase
inositol polyphosphate multikinase
- InsP6, Ins(1,2,3,4,5,6)P6
- 5-PP-Ins(1,3,4,6)P4 5-diphospho-myo-inositol 1,3,4,6-tetrakisphosphate;1-InsP7, 1-PP-InsP5
- 3-InsP7, 3-PP-InsP5
- 4-InsP7, 4-PP-InsP5
- 5-InsP7, 5-PP-InsP5
- 6-InsP7, 6-PP-InsP5
Oryza sativa inositol tris/tetrakisphosphate kinase 1
polymerase chain reaction
Arabidopsis Phosphate Starvation Response 1
diphosphoinositol pentakisphosphate kinase
SYG1/Pho81/XPR1 domain-containing protein 1
Arabidopsis thaliana diphosphoinositol pentakisphosphate kinase 1
Arabidopsis thaliana diphosphoinositol pentakisphosphate kinase 2