A synthetic diphosphoinositol phosphate analogue of inositol trisphosphate

We describe the synthesis and biological evaluation of 1-PP-Ins(4,5)P2, the first diphosphate-containing analogue of the intracellular signalling molecule Ins(1,4,5)P3.


Introduction
The myo-inositol phosphates (InsPs) are a family of intracellular signalling molecules containing monophosphate (P) and diphosphate (PP) groups arranged around the hexahydroxycyclohexane ring of myo-inositol (Ins). 1 InsPs regulate many cellular processes, the best known being the release of Ca 2+ from intracellular stores by D-myo-inositol 1,4,5-trisphosphate (InsP 3 ), which binds to receptors on the endoplasmic reticulum. 2 InsP 3 is converted via a series of enzymatic phosphorylations 3 into InsP 6 ( Fig. 1), which can then be further phosphorylated to give highly charged PP-InsPs containing diphosphate (pyrophosphate) groups. 4,5 InsP 3 receptors (IP 3 Rs) are tetrameric intracellular Ca 2+ channels, expressed in most animal cells. 2 When InsP 3 binds to the N-terminal InsP 3 -binding core (IBC) of all four IP 3 R subunits, 6 conformational changes propagate to the central pore. The pore then opens, allowing Ca 2+ to flow into the cytosol, where it regulates many intracellular processes. The vicinal 4,5-bisphosphate structure of InsP 3 is crucial (if not absolutely essential 7 ) for activating IP 3 Rs because it cross-links the two domains of the clam-like IBC, pulling them together and initiating the conformational changes. The 1-phosphate group has a less direct, but enhancing, effect on activity. 8 Although PP-InsP signalling is thought to be more evolutionarily ancient than InsP 3 -mediated mobilisation of Ca 2+ , 9 much less is known about the functions and protein targets of PP-InsPs. Nevertheless, evidence is accumulating that PP-InsPs play important roles at the interface of cell signalling and metabolism in the regulation of bioenergetic and phosphate homeostasis. 4,5 Possible receptors for PP-InsPs include the PH (pleckstrin homology) domains 10,11 and SPX (SYG1/ Pho81/XPR1) domains 12,13 of proteins. PP-InsPs may also exert some of their effects by direct non-enzymatic diphosphorylation of target proteins. 14 Phosphorylating a phosphate monoester in an InsP n to give a PP-InsP n−1 not only increases the overall negative charge of the molecule, but also changes its shape, solvation and metal complexation properties. Unsurprisingly, therefore, a diphosphate group may alter ligand affinity for protein binding sites. 4 For example, some PH domains that bind InsP 6 bind 5-InsP 7 with higher affinity, 10,11 while 1-InsP 7 and InsP 8 are weaker. 11 In contrast, both 1-InsP 7 and 5-InsP 7 stimulate synthesis of inorganic polyphosphate (polyP) by the vacuolar transporter chaperone (VTC) by binding to its SPX domain, while InsP 6 is inactive and InsP 8 is 20-fold more potent. 13 PP-InsPs can be dephosphorylated back to InsPs by diphosphoinositol polyphosphate phosphohydrolases (DIPPs, Fig. 1), which specifically hydrolyse the diphosphate group, leaving a phosphate monoester and liberating inorganic phosphate. 3,15 Given that introducing a diphosphate into an InsP may modify its interaction with proteins, we were interested in the possible effects of converting one of the phosphate groups in InsP 3 into a diphosphate. The 1-phosphate group of InsP 3 has been a popular target for synthetic elaboration of InsP 3 since early structure-activity studies showed that it is much more tolerant of modification than the 4-or 5-phosphate groups. 8 Interest in the role of the 1-phosphate group was further stimulated by the discovery in 1993 of the adenophostins, fungal metabolites that are highly potent InsP 3 receptor ligands. 16 The adenophostins contain a glucopyranoside 3,4-bisphosphate structure that mimics the myo-inositol 4,5-bisphosphate of InsP 3 but intriguingly, their third phosphate group is located on a separate (ribofuranoside) ring, suggesting that repositioning this phosphate group may enhance affinity. 17 The X-ray structure 18 of the IBC of type 1 InsP 3 receptor bound to InsP 3 confirmed the area of the binding pocket around the 1-phosphate of bound InsP 3 to be relatively open. Our molecular docking experiments using this structure suggested that a 1-diphosphate should bind well to this region. We therefore set out to synthesise the 1-diphosphate an-alogue of InsP 3 , i.e. 1D-diphospho-myo-inositol 4,5bisphosphate [1-PP-InsĲ4,5)P 2 (1), Fig. 1] and examine its interaction with InsP 3 receptors.
We were also interested to examine the interaction of 1-PP-InsĲ4,5)P 2 with DIPPs. Although DIPPs can hydrolyse the PP groups of highly phosphorylated PP-InsPs (Fig. 1), inorganic polyphosphate, 5-phosphoribosyl 1-pyrophosphate and nucleotide dimers, 3,15 their catalytic mechanisms are poorly understood. 1-PP-InsĲ4,5)P 2 contains the target 1-PP structure found in the known DIPP substrate 1-PP-InsP 5 , but presented in the context of a molecule with only two phosphate monoester groups. There are no reports in the literature on whether "lower" PP-InsPs such as 1-PP-InsĲ4,5)P 2 could be recognised by the active sites of DIPPs.

Chemistry
The synthesis of 1-PP-InsĲ4,5)P 2 (1) begins from the known alcohol 2 (ref. 19 and 20) (Scheme 1). To construct the diphosphate unit at O-1, we employed a modification of a recently described strategy, 21,22 in which a temporarily protected phosphate group is introduced and then selectively deprotected to reveal a phosphate monoester. This phosphate is then phosphitylated to give a mixed PĲIII)-PĲV) anhydride, which is oxidised to a partially protected pyrophosphate unit. Removal of all protecting groups by catalytic hydrogenolysis then yields the target PP-InsP. We reasoned that it might be possible to employ methylsulfonylethyl (MSE) 23,24 as a temporary phosphate protecting group in this sequence. The MSE group can be removed by β-elimination, similar to the betterknown β-cyanoethyl (β-CE) 22,25 group. However, the MSE group is unaffected by catalytic hydrogenation, affording greater synthetic versatility, and the required phosphitylating reagent, phosphoramidite 5, is a stable crystalline solid.
Thus, the 1-OH group in 2 was reacted with 5 in the presence of 5-phenyl-1H-tetrazole to give an intermediate MSEprotected phosphite triester. Oxidation using mCPBA then gave 3, containing the MSE-protected phosphate triester at O-1. The diphosphate unit at O-1 was then constructed using a sequence of transformations carried out as described previously, 21,22,25 with slight modifications. The progress of each step was carefully monitored by 31 P NMR spectroscopy (see Experimental section and ESI †). The protected diphosphate 4 was found to be rather unstable and was immediately deprotected by catalytic hydrogenolysis at atmospheric pressure. A final purification step by gradient elution anion exchange chromatography on Q-Sepharose Fast Flow resin gave 1-PP-InsĲ4,5)P 2 (1) as the triethylammonium salt, which was accurately quantified by total phosphate assay.
Interactions of 1-PP-InsĲ4,5)P 2 with type 1 InsP 3 receptors Both InsP 3 and 1-PP-InsĲ4,5)P 2 (1) stimulated a concentrationdependent release of Ca 2+ from the intracellular stores of permeabilised DT40 cells expressing type 1 InsP 3 receptors ( Fig. 2A). The maximal Ca 2+ release evoked by each ligand and the half-maximally effective concentration (EC 50 ) were similar for 1 and InsP 3 ( Fig. 2A). Membranes from Sf9 cells expressing rat type 1 InsP 3 receptors were used for equilibrium competition binding studies with 3 H-InsP 3 , because these membranes express full-length type 1 InsP 3 receptors at ∼20-fold higher levels than cerebellar membranes, the richest endogenous source. The experiments were carried out in cytosol-like medium (CLM, pH 7.3) containing 1.5 mM Mg-ATP to match the conditions used for Ca 2+ -release assays.
In agreement with the Ca 2+ -release assays, 1-PP-InsĲ4,5)P 2 (1) bound with the same affinity as InsP 3 to InsP 3 receptors (Fig. 2B). Thus, the two compounds were essentially indistinguishable in both functional and binding assays ( Table 1). Rapid chemical hydrolysis of 1 could in principle explain the similar behaviour of InsP 3 and 1, but we saw no evidence that 1 is unstable. The 31 P NMR spectrum of 1 in D 2 O (see ESI †) was unchanged after the sample solution had been kept for several days at room temperature, followed by one year at 4°C.
Molecular docking experiments (see Experimental section and ESI † for details) using the X-ray crystal structure of the IBC of type 1 InsP 3 receptor 18 suggested that the diphosphate group in 1 should be well-tolerated by the InsP 3 -binding pocket and may be capable of forming additional hydrogen bonds with residues in the binding site (Fig. 3). Nevertheless, it is well known that attempts to optimise drug candidates by adding polar groups may fail because the expected enthalpic gains from new polar interactions are opposed by ligand desolvation penalties and unfavourable entropic effects, resulting in no gain in binding affinity. 26 Such compensatory effects may underlie the similar affinities of 1 and InsP 3 for type 1 InsP 3 receptors.
With Mg 2+ present in the buffer, 1-PP-InsP 5 and 5-PP-InsP 5 were rapidly metabolised by all four DIPPs (Fig. 4A). The rate of hydrolysis of 1-PP-InsP 5 was significantly higher than that for 5-PP-InsP 5 in each case. This finding is in agreement with a previous study. 15 As expected, the PCP analogues were not metabolised, confirming that DIPPs can hydrolyse only the diphosphate unit and not the phosphate monoesters. Ap 3 A and Ap 5 A were unaffected by all four enzymes in Mg 2+ -containing buffer, an observation that had been reported for NUDT10 and NUDT11, but not for NUDT3 and NUDT4. 29 Perhaps surprisingly, 1-PP-InsĲ4,5)P 2 (1) was also not metabolised under these conditions. The presence of a divalent cation is required for the activity of NUDT10 and NUDT11 and also for NUDT3. 3 When Mg 2+ in the buffer was replaced by Mn 2+ , 1 was now hydrolysed by the DIPPs, while 1-PP-InsP 5 and 5-PP-InsP 5 resisted hydrolysis. In addition, Ap 5 A now also behaved as a substrate for all four DIPPs (Fig. 4B). In the absence of enzyme none of the compounds, including 1, showed any sign of hydrolysis during the time course of the experiment in the pres-ence of either Mg 2+ or Mn 2+ -containing buffers. This further supports our conclusion above that 1 was not hydrolysed to InsP 3 during the InsP 3 receptor assays.
Next, we used differential scanning fluorimetry (DSF) to measure the ability of the compounds to stabilise NUDT3 (DIPP1). While the effects of Ap 3 A and Ap 5 A were not significantly different from control (Fig. 5A), 1-PP-InsĲ4,5)P 2 (1) raised the melting temperature (T m ) of NUDT3 by approx. 5°C at a concentration of 0.1 mM. As expected, the more highly phosphorylated 1-PP-InsP 5 had much stronger effects, resulting in a T m -shift of 20-25°C. Similar DSF experiments were then carried out for NUDT4, NUDT10 and NUDT11. Ap 3 A did not stabilise any of the DIPPs, which supports our results for the activity assay. The results are summarised in Fig. 5B.
We obtained further DSF data over a range of ligand concentrations for 1-PP-InsP 5 and 1-PP-InsĲ4,5)P 2 (1), constructing dose-response curves for the two compounds (Fig. 6). It is interesting to note that the effect of 1 on NUDT10 was significantly lower compared to the other DIPPs and especially compared to NUDT11 (Fig. 6B). NUDT10 and NUDT11 have identical protein sequences apart from residue 89, which is either proline (NUDT10) or arginine (NUDT11).
Noting the strong stabilisation of all the proteins by the PCP analogues, we obtained further DSF data over a range of ligand concentrations for 1-PCP-InsP 5 and 5-PCP-InsP 5 (ESI † Fig. S4 and S5) and calculated K D values from these curves (ESI † Tables S1 and S2). We found that, in some cases, the PCP analogues had binding affinities comparable to those of their natural PP-containing ligands.

Conclusions
Replacing a phosphate group in an inositol phosphate ligand with a diphosphate (PP) group can modify the interaction of the ligand with target proteins. [10][11][12][13] Structure-activity studies have previously shown that the 1-phosphate group of InsP 3 is amenable to synthetic modification, and molecular docking experiments suggested that a 1-diphosphate group should be well-tolerated by the binding site of the InsP 3 receptor. We therefore synthesised 1-PP-InsĲ4,5)P 2 (1), the first PPcontaining analogue of InsP 3 . Using assays of Ca 2+ -release through type 1 InsP 3 receptors, we found that 1 was equipotent to InsP 3 and in binding assays its affinity was indistinguishable from that of InsP 3 . Thus, the 1-diphosphate modification of InsP 3 does not affect its affinity for or activity at type 1 InsP 3 receptors. Nevertheless, 1 is the first Ca 2+ -releasing PP-InsP and also the most potent P-1 modified ligand of InsP 3 receptors yet identified. ‡ The novel diphosphate compound 1 was not metabolised by DIPPs in the presence of Mg 2+ -containing buffer, while the naturally-occurring InsP 7 isomers, 5-PP-InsP 5 and 1-PP-InsP 5  (1) in the IBC produced by molecular docking (see Experimental section and ESI † for details). For clarity, water molecules are not shown. ‡ A synthetic InsP 3 derivative featuring 4-carboxy-malachite green conjugated to the 1-phosphate group was reported to have ∼170-fold higher affinity than InsP 3 for an N-terminal fragment of type 1 InsP 3 receptors. 43 In our hands, this compound was ∼5-fold less potent than InsP 3 at each InsP 3 receptor subtype and had an affinity ∼7-fold less than InsP 3 for type 1 InsP 3 receptors. 44 were rapidly hydrolysed. Conversely, in the presence of Mn 2+ , 1 was hydrolysed while the two InsP 7 isomers were unaffected. Synthetic PCP-containing analogues of the InsP 7 s were not hydrolysed under any conditions examined, but when evaluated for their ability to stabilise DIPP proteins using differential scanning fluorimetry (DSF), they gave temperature shifts comparable to their natural PP-containing equivalents.   This strongly suggests that 1-PCP-InsP 5 and 5-PCP-InsP 5 could be promising ligands for co-crystallisation studies with DIPPs.
Could 1-PP-InsĲ4,5)P 2 be an endogenous molecule? The mammalian enzymes known to synthesise PP-InsPs are 5-diphosphoinositol pentakisphosphate kinases (PPIP5Ks) and inositol hexakisphosphate kinases (IP6Ks). Inositol phosphate multikinase (IPMK) has also been reported to synthesise PP-InsP 4 from InsP 5 in vitro, 30 but the products of InsP 3 phosphorylation by IPMK are InsĲ1,3,4,5)P 4 and/or InsĲ1,4,5,6)P 4 . 31 Phosphorylation of lower InsPs by PPIP5Ks seems unlikely, considering the constraints of the catalytic site 32 and the recently discovered capture site; 22 even InsĲ1,3,4,5,6)P 5 is not phosphorylated. 32 Recombinant Kcs1p, a yeast homologue of IP6K1, was reported to phosphorylate InsP 3 slowly, although the identities of the products could not be determined. 33 Later work confirmed that InsP 3 was phosphorylated by Kcs1 and the product was identified as InsĲ1,3,4,5)P 4 (i.e. in this case, Kcs1 functioned as a 3-kinase). 34 More recently, a study found that EhIP6KA, an IP6K homologue from Entamoeba histolytica, was capable of slowly phosphorylating InsP 3 , although the products were identified as InsĲ1,4,5,6)P 4 and InsĲ1,2,4,5)P 4 . 35 On this basis, naturally occurring 1-PP-InsĲ4,5)P 2 seems unlikely. However, in both studies where the identities of the enzyme products were assigned, 34,35 resistance to hydrolysis by DIPP1 was used to exclude the possibility that the products contained diphosphate groups. The present work shows that this criterion may not always be valid; in our hands, 1-PP-InsĲ4,5)P 2 was not metabolised in the presence of Mg 2+ by any of the DIPPs, yet it does contain a diphosphate group.
Notwithstanding the evidence for PP-InsPs playing physiological roles, 4,5 the present work indicates that a physiological function for 1-PP-InsĲ4,5)P 2 , at least in relation to the regulation of InsP 3 receptor-mediated Ca 2+ release, may be unlikely. Converting the 1-phosphate of InsP 3 into a diphosphate neither attenuates nor enhances the ability of the ligand to activate InsP 3 R. As the first example of a diphosphate analogue of a second messenger, however, the results add a new component to structure-activity relationships. Cocrystallisation studies with DIPPs using some of the nonhydrolysable substrate analogues discussed here are currently in progress.
Compound 4 (37 mg, 28 μmol) was dissolved in methanol (4 mL) and deionised water (1 mL). Powdered NaHCO 3 (14 mg, 168 μmol) was added followed by PdĲOH) 2 /C (30 mg). The suspension was stirred vigorously under H 2 (balloon) for 24 h, after which time more water (4 mL) was added. A fresh balloon of H 2 was attached and stirring was continued for a further 72 h. The catalyst was then removed by filtration through a PTFE filter, giving a colourless solution, which was concentrated under reduced pressure to give a solid white residue. Analysis of this residue by 31 P and 1 H NMR in D 2 O showed that deprotection was complete. The residue was purified by anion-exchange chromatography on Q-Sepharose Fast Flow resin, eluting with a gradient of 0 to 1.5 M triethylammonium bicarbonate (TEAB). The target compound 1 eluted at 70 to 77% 1.5 M TEAB. Fractions containing the target were identified using the Briggs phosphate assay, combined and evaporated under reduced pressure. De-ionised water was repeatedly added and evaporated until the triethylammonium salt of 1 remained as a colourless glassy solid (14 mg, 16 μmol, 57%). This material was accurately quantified using total phosphate assay 37 before biological evaluation. For 31 P and 1 H NMR analysis of 1, a small amount of EDTA (sodium salt, approx. 0.1 mg) was added to a sample of 1 (2.0 μmole in 0.4 mL D 2 O) to give sharper signals. This NMR sample containing EDTA was kept as the solution in D 2 O for >1 year at 4°C with no sign of deterioration. 1  Molecular docking of 1-PP-InsĲ4,5)P 2 (1) into type 1 InsP 3 receptor. Molecular docking experiments were carried out using the X-ray crystal structure of the N-terminal IBC of type 1 InsP 3 receptor in complex with InsĲ1,4,5)P 3 (1N4K). 18 Docking methods were optimised by docking flexible models of InsĲ1,4,5)P 3 into the 1N4K structure using GOLD 38 (version 5.6, CCDC). In the most successful protocol, the binding site was defined as a sphere of 6 Å radius centred on the centroid of bound InsĲ1,4,5)P 3 and two water molecules (waters 1139 and 1198) were included in the docking protocol. These water molecules were toggled on and off and allowed to spin in the docking runs. 39 The ligand was docked 100 times using the GoldScore scoring function, and genetic algorithm settings for very flexible ligands were used. This method accurately reproduced the observed pose of bound InsĲ1,4,5)P 3 in 1N4K; the ten highest scoring poses all closely resembled the conformation of bound InsĲ1,4,5)P 3 (mean RMSD 0.58 Å). When 1-PP-InsĲ4,5)P 2 (1) was docked using the same protocol, the highest-scoring poses were very similar to the bound conformation of InsĲ1,4,5)P 3 but often showed additional interactions of the 1-beta-phosphate group with residues in the binding site. More details are given in the ESI. † Assays of InsP 3 receptor activity. Ca 2+ release from the intracellular stores of permeabilised DT40 cells expressing rat type 1 InsP 3 receptors was measured in cytosol-like medium (CLM) using a low-affinity fluorescent Ca 2+ indicator trapped within the endoplasmic reticulum as previously reported. 40 Equilibrium competition binding of [ 3 H]-InsP 3 (1.5 nM, 19.3 Ci mmol −1 ) to membranes prepared from insect Sf9 cells expressing rat type 1 InsP 3 receptors was determined in CLM at 4°C. Bound and free ligand were separated by centrifugation and non-specific binding was determined by addition of 10 μM InsP 3 .
DIPP purification. cDNAs for all DIPPs were kind gifts from the Structural Genomics Consortium, Stockholm. cDNAs were modified as necessary in order to represent the full-length constructs, cloned into pET28a (+) and expressed as N-terminally His-tagged proteins. All proteins were expressed in BL21 (DE3) T1R pRARE2 at 18°C overnight and purified by the Protein Science Facility (PSF) at the Karolinska Institute, Stockholm. Briefly, the proteins were first purified over a HisTrap HP column (GE Healthcare), followed by thrombin cleavage of the N-terminal His-tag. After removal of the His-tag through a second run over a HisTrap HP column, the proteins were further purified by gel filtration using a HiLoad 16/60 Superdex 75 column (GE Healthcare).
Differential scanning fluorimetry (DSF). DSF 42 was performed with 5 μM purified protein in 25 mM HEPES pH 7.5, 100 mM NaCl, 0.5 mM TCEP and 5× Sypro Orange added per well of a 96-well PCR plate. Substrates and substrate analogues were dissolved in water and diluted 1 : 10 in the assay mixture. The heat denaturation curves with a temperature increase of 1°C min −1 from 25°C to 95°C were recorded on a CFX96 real-time PCR machine (Bio-Rad) by measuring the fluorescence of Sypro Orange with excitation and emission wavelengths of 470 and 570 nm, respectively. The Boltzmann equation was used to analyse the denaturation curves in GraphPad Prism. The determined melting temperature (T m ) is the inflection point of the sigmoidal denaturation curve.

Conflicts of interest
There are no conflicts to declare.