Synthesis of phosphonoacetate analogues of the second messenger adenosine 5′-diphosphate ribose (ADPR)

Adenosine 5′-diphosphate ribose (ADPR) is an intracellular signalling molecule generated from nicotinamide adenine dinucleotide (NAD+). Synthetic ADPR analogues can shed light on the mechanism of activation of ADPR targets and their downstream effects. Such chemical biology studies, however, are often challenging due to the negatively charged pyrophosphate that is also sensitive to cellular pyrophosphatases. Prior work on an initial ADPR target, the transient receptor potential cation channel TRPM2, showed complete pyrophosphate group replacement to be a step too far in maintaining biological activity. Thus, we designed ADPR analogues with just one of the negatively charged phosphate groups removed, by employing a phosphonoacetate linker. Synthesis of two novel phosphonoacetate ADPR analogues is described via tandem N,N′-dicyclohexylcarbodiimide coupling to phosphonoacetic acid. Neither analogue, however, showed significant agonist or antagonist activity towards TRPM2, underlining the importance of a complete pyrophosphate motif in activation of this particular receptor.


Introduction
The linear diphosphate adenosine 5 0 -diphosphate ribose (ADPR, Fig. 1a) is formed from nicotinamide adenine dinucleotide (NAD + ) by two routes; either hydrolysis of the nicotinamide ribose by NAD + glycohydrolases such as CD38 1 or via the sequential action of poly-ADPR polymerase (PARP) and poly-ADPR glycohydrolase (PARG). 2 PARP inhibitors are currently highly topical as drugs in oncology. 3 ADPR's cyclic incarnation as cyclic ADP-ribose (cADPR) is a second messenger in cellular calcium signalling. 4,5 cADPR is also hydrolysed to ADPR as a deactivation mechanism. ADPR targets include widespread macro domains 6 and the non-selective cation channel, transient receptor potential cation channel, subfamily M, member 2 (TRPM2). 7,8 They are associated with a diverse range of physiological processes that are still not yet fully understood. However, chemical biological studies of ADPR-mediated processes in whole cells are oen hampered by ADPR's negative charges and lack of membrane permeability. Studies of these processes oen require technically demanding and time-consuming patch clamp or microinjection experiments on single cells to deliver charged species physically through the cell membrane. Furthermore, ADPR is subject to hydrolysis of the pyrophosphate by enzymes (intracellular enzymes, ectoenzymes in the plasma membrane and enzymes present in biological uids), including hydrolases of the NUDIX family. This means that ADPR itself is unlikely to be stable for any length of time in a biological setting, complicating the direct study of its effects. Thus, synthetic ADPR analogues that are both membrane permeant and stable in the biological environment would be valuable tools to unravel the downstream mechanisms and function of this ligand. Stable ADPR analogues may also be useful when clarication of the precise second messenger that is acting under physiological conditions has proved challenging. 9,10 Small changes to the structure of ADPR have demonstrated profound effects on activity and, as a result, synthetic analogues closely based on the ADPR structure have been very informative. [11][12][13] Analogues with modications in the adenosine 11,13 and terminal ribose [14][15][16][17] motifs have been prepared via coupling of a modied 5 0 -O-phosphoryl adenosine or ribose to its 5 0 -Ophosphoryl counterpart using morpholidate, diphenylphosphonate or imidazolide methods to activate one phosphate towards nucleophilic attack and thus generate the pyrophosphate. Notwithstanding their utility for studying ADPR processes using single cell techniques such as patch clamping, further studies even with many of these designed analogues are limited by the negative charge and potential biological instability of the pyrophosphate moiety. Such pyrophosphatecontaining ADPR analogues are thus not suitable for use in whole cell studies or for in vivo applications. The cell permeable peptide tat-M2NX has been shown to inhibit TRPM2 and reduce ischemic injury in male mice, 18 suggesting that membrane permeable analogues closely related to ADPR in structure would nd similar utility.
The negative charge of ADPR may be reduced by masking the charged phosphate groups. Indeed, ADPR that was acetylated and bioreversibly masked with a lipophilic group on the adenosine phosphate has been reported. 19 However, pyrophosphates masked in this way can be more unstable towards hydrolysis by nucleophiles than their negatively charged counterparts, and this does not address the issue of degradation due to cellular pyrophosphatases. The use of phosphate bioisosteres in medicinal chemistry is an alternative method to improve cellular stability and potentially also improve membrane permeability. 20 Previous studies using pyrophosphate replacement in ADPR have provided some new insights into ADPR binding. The methylene bis-phosphonate analogue of ADPR (a-b methylene ADPR, AMPCPR, Fig. 1b) contains a non-hydrolysable P-C-P pyrophosphate substitution that is stable towards pyrophosphatases and has therefore generated a number of crystal structures of ADPR targets bound to AMPCPR. 21 Tóth et al. showed that AMPCPR still supports TRPM2 channel gating, albeit as a lower affinity partial agonist. 22 AMPCPR has been used to uncouple the TRPM2 channel gating mechanism from its enzymatic activity by virtue of its pyrophosphatase resistant methylene bis-phosphonate linker. 22,23 Two further ADPR analogues with similarly modied pyrophosphates (one methylene bisphosphate and one diuoromethylene bisphosphate, see Fig. 1e) that were also protected on some or all of the ADPR hydroxyl groups selectively inhibited TRPM2. 24 These hydrolysis resistant analogues demonstrate that modication of the pyrophosphate can impart valuable attributes to the ADPR analogue. However, such methylene bis-phosphonate analogues still carry two negative charges at physiological pH, and thus are not membrane permeable.
Our previous highly ambitious complete replacements of the pyrophosphate group in ADPR ( Fig. 1c and d), that are unlikely to be charged at physiological pH, unfortunately did not show TRPM2 agonist or antagonist activity. 11, 13 We therefore pursued a more conservative approach in this study, designing two ADPR analogues 1 and 2 with only one phosphorus centre replaced (Fig. 1f). We chose the isosteric phosphonoacetate group as a pyrophosphate mimic between the adenosine and the terminal ribose of ADPR, as this analogue would only carry a single negative charge at physiological pH. A phosphonoacetate motif should also be resistant to the intrinsic cellular pyrophosphatase activity of some ADPR targets as the P-C bond is more chemically stable than the pyrophosphate P-O bond. 25 Therefore, such analogues could potentially be used to probe ADPR-mediated activities in isolation from pyrophosphatase activity. The rst phosphonoacetate analogues of the 5-O-pyrophosphate in diphosphoinositol polyphosphates (PP-InsPs) were recognised and phosphorylated by the kinase domain of diphosphoinositol pentakisphosphate kinase 2 (PPIP5K2), showing that such substitution of a pyrophosphate did not prevent recognition of the analogue by its target enzyme and allowing downstream effects unrelated to dephosphorylation to be studied. 26 We therefore report here the synthesis of phosphonoacetate ADPR analogues 1 and 2, studies on the stability of both analogues and their initial biological evaluation at TRPM2.

Synthesis of ADPR analogues
Synthesis of adenosine-5 0 -O-(2-phosphoryl) acetate ribose (Fig. 1f, 1) started with commercially available triethylphosphonoacetate 3 (Scheme 1). Selective deprotection of the carboxylic ester in the presence of the phosphate esters 27 led to the corresponding carboxylic acid 4 (97%). Carboxylic acid 4 was esteried, by N,N 0 -dicyclohexylcarbodiimide (DCC) mediated coupling, with 2 0 ,3 0 -O-isopropylidene adenosine 11 to obtain compound 5 (82%). Diethylphosphonate ester 5 was then transsilylated by bromotrimethylsilane (TMSBr), followed by hydrolysis to obtain phosphonic acid 6 (best yield 46%). Selective removal of the phosphonate esters in the presence of the isopropylidene group was achieved by carrying out the reaction in pyridine, as the basic conditions buffer the HBr generated. Carrying out the same conversion with DCM as a solvent generated a complex mixture due to concomitant deprotection of the isopropylidene protecting group. Purication of the free phosphate 6 was carried out using normal phase ash chromatography on silica gel. Impurities were eluted rst with an isocratic solvent system of ethyl acetate, methanol and water (7 : 2 : 1 v/v/v) containing 0.1% triethylamine. The desired product 6 was then eluted by switching to a second isocratic solvent system of isopropyl alcohol, water and aqueous ammonia (7 : 2 : 1 v/v/v) containing 0.1% triethylamine. First attempts to obtain compound 7 via DCC mediated coupling to 1 0 -O-methyl-2 0 , 3 0 -O-isopropylidene-b-D-ribofuranoside 12 were unsuccessful. However, aer a short optimization, we found that at elevated temperature (50-70 C) compound 6 forms a mixture of anhydrides, including most likely adenosine phosphonoacetate anhydride 28 which only reacts with 12 at temperatures above 100 C. Thus, the higher temperature of 120 C (compared to the DCC mediated reaction of 4 and 11 that occurred at rt) plays a crucial role in the conversion of 6 to 7. Finally, with the optimized conditions, the DCC mediated coupling of 6 with 1 0 -O-methyl-2 0 , 3 0 -O-isopropylidene-b-D-ribofuranoside 12 generated the desired ADPR analogue 7 (best yield 57%). Deprotection of 7 led to target compound 1, which was obtained aer HPLC purication in 0.1 M TEAB buffer with a gradient of acetonitrile as the triethylammonium salt (36%). Until deprotection of analogue 7, all reactions are stereospe-cic. The nal ADPR analogue 1 is a mixture of a and b anomers at the terminal ribose (ratio a : b ¼ 2 : 3, as detailed in the Experimental section).

Stability of target phosphonoacetates 1 and 2
The stability of 1 and 2 in solution was monitored by analytical HPLC, comparing freshly made samples from the solid compounds 1 and 2, stored at À20 C. Both nal analogues were synthesised as the corresponding triethylammonium salts (1.0 Â Et 3 N and 1.2 Â Et 3 N equivalents respectively, calculated from 1 H-NMR spectra). While phosphonoacetate 1 is stable in this form in MilliQ water, TEAB (0.1 M) or MeCN solutions, we were surprised to observe that phosphonoacetate 2, degrades slowly in TEAB (pH 8-9, 0.1 M TEAB, ESI, Fig. S1 †). HPLC of 2 (R t ¼ 3.9 minutes) aer 16 h in 0.1 M TEAB buffer showed an additional peak at R t ¼ 2.7 minutes. This decomposition product showed a characteristic adenosine UV absorption (l max ¼ 259) and was analysed using mass spectrometry (ESI + found [M À H] À 388.0676, C 12 H 15 N 5 O 8 P requires 388.0664) suggesting loss of the terminal ribose to generate adenosine 5 0 -phosphonoacetate (ESI, Fig. S1 †). We suspect it is most likely that 2 undergoes basic hydrolysis of the terminal ribose from the phosphonoacetate ester mediated by triethylamine from the TEAB buffer.
We then measured the stability of phosphonoacetate 2 under various aqueous conditions (Fig. 2). According to these ndings, analogue 2 was stable in pure MilliQ water (pH 7) as the triethylammonium salt when stored in the fridge. Importantly, this should be sufficient for biological assays using the pure phosphonoacetate 2 when used immediately from a frozen sample. Increased pH 8-9 (in 0.1 M TEAB buffer) resulted in slow decomposition of 2 either at room temperature or when stored in the fridge. We prepared the sodium salt of analogue 2 in an attempt to overcome the possibility of triethylammonium counter ions generating a basic solution and thus causing the decomposition of 2. The triethylammonium salt of 2 was stirred with Chelex (Na + form) resin, which was removed by ltration to generate the sodium salt of 2. However, we were surprised to nd that the corresponding sodium salt of 2 was even more unstable and degraded completely either at room temperature or when stored in the fridge (Fig. 2).
It was rather unexpected that ADPR analogue 1 is stable under various aqueous conditions whereas ADPR analogue 2 is not. This led us to consider which feature of 2 might be responsible for the observed degradation. Our observations suggest that 2 is sensitive to basic hydrolysis of the phosphonoacetate ester in TEAB buffer but that, under the same conditions, the phosphonoacetate ester of 1 is stable. This suggests that the terminal ribose 5 00 -O-ester may inuence the stability in a way that is not observed with the adenosine ribose  This journal is © The Royal Society of Chemistry 2020 RSC Adv., 2020, 10, 1776-1785 | 1779 5 0 -O-ester. The triethylamine base present in TEAB buffer may catalyse hydrolysis of the ester bond between the phosphonoacetate ester and the 5 00 -O-terminal ribose. Indeed, we have previously observed that modications to the terminal ribose can generate ADPR analogues that are particularly prone to instability. 14 Analogue 1 is not susceptible to hydrolysis because the more stable phosphonate monoester is adjacent to the terminal ribose. Phosphate monoesters are stable under basic conditions due to the negative charge on the phosphonate oxygen preventing nucleophilic attack. Presumably, the 5 0 -Oester in analogue 1 is more sterically hindered, due to its proximity to the less exible adenosine, and thus is less susceptible to hydrolysis under mildly basic conditions. In future, synthesis of the attractive corresponding phosphonamide ADPR analogues in which the ester is replaced by an amide linker, may be one method to ensure stable analogues are generated, as exemplied in a recent synthesis of an inositol polyphosphate pyrophosphate analogue. 29 Biological evaluation of 1 and 2 Due to our stability studies, we were aware of the possibility of degradation of the unstable ester bond of compound 2. This would have been undesirable in our assay to determine the agonist and antagonist effects of 2. In order to prevent degradation, the compounds were reconstituted using MilliQ water (pH 7), which was shown to generate a stable solution of both compounds (Fig. 2). The assay was carried out at pH 7.2. Immediately before the assay for agonist or antagonist effects, the solution of the analogue to be tested was veried by HPLC. No degradation was detected during the entire sequence of testing, or even when the pH 7 solution of the analogue was stored at À80 C for 1 year.
ADPR analogues 1 and 2 were tested for their agonist and antagonist activity on human TRPM2 (hTRPM2) using patch clamp experiments in HEK293 clones with stable expression of hTRPM2. 13 Because the compound is infused into the cytoplasm of the cell via the patch-pipette, this method allows evaluation of the activity of the compounds independently of their membrane permeability. Following the results of our stability testing, phosphonoacetate 2 was stored as a lyophilized dry solid at À20 C and made up in fresh MilliQ water immediately before testing to ensure that decomposition did not occur (vide supra). When applied at 100 mM or 1000 mM, neither 1 nor 2 had signicant agonist effects at hTRPM2 (Fig. 3a). When applied with ADPR in the patch pipette (100 mM of ADPR with 300 mM or 900 mM 1 or 2), neither analogue showed signicant antagonist effects on the ADPR-induced TRPM2 current (Fig. 3b). If activity was observed, we would have determined the membrane permeability of 1 and 2 by applying them extracellularly to whole cells, however this was not possible.
These results highlight the sensitivity of hTRPM2 channel activation to modications of the ADPR molecular structure. It is possible that both phosphate motifs of the pyrophosphate bridge are required for binding and activation or inhibition of the channel, so that even a subtle change in this part of the molecule may lead to the complete loss of activity. The single atom substitution in the methylene bis-phosphonate analogue AMPCPR corresponded to a roughly 40 fold reduction in potency at hTRPM2 and AMPCPR is also only a partial hTRPM2 agonist. 22 Alternatively the substitution of the pyrophosphate with the phosphonoacetate bioisostere may alter the presentation of other critical binding partners in the hTRPM2 binding pocket. Our molecular modelling suggests the phosphonoacetate linker theoretically seems to be a very good spatial mimic of the pyrophosphate unit, with sufficient exibility to allow the adenosine and terminal ribose binding partners to position themselves in the same orientation as ADPR (ESI, Fig. S2 †), but this may not reect the reality once solvent effects and binding interactions are taken into account.
Cryo-EM structures of hTRPM2 have recently improved understanding of the zebrash hTRPM2 apo and ADPR-bound states 30 importantly for hTRPM2 also in the presence of bound ligands. 31,32 However, only recently has their resolution become detailed enough to place ADPR in the binding pocket and has revealed, for the rst time, ADPR binding sites in hTRPM2 that may act synergistically. 32 The requirement for an ADPR mimic to bind to both sites, which possess very distinct shapes, 30 may explain why very few synthetic ADPR analogues have activated the channel. 11,15 Despite this recent advance, the structure activity relationship for hTRPM2 is not fully understood, and ADPR analogues remain essential for ligand-based drug design.
The crystal structure of the macro domain of thermophilic protein Af1521 bound to ADPR was solved in high resolution (1.5Å, PDB 2BFQ). 6 With the exception of the adenosine ribose, each ADPR structural feature interacts closely with the macro domain structure. The adenosine 5 0 -O-phosphate has two hydrogen bonds to the backbone N-H of valine 43 and glycine 143. The ribose 5 0 -O-phosphate has hydrogen bonds to the backbone N-H of serine 141 and tyrosine 145. Furthermore, the pyrophosphate is stabilized by the helix dipole. Thus, the demonstrated highly specic binding interactions of macro domains and apparent similar selectivity of hTRPM2 binding require very specic ADPR binding partners. Our initial results with hTRPM2 suggest that phosphonoacetate analogues 1 and 2 may not be able to full all these demands, but 1 and 2 may potentially nd applicability in other systems that bind ADPR.

Conclusion
We have synthesised two phosphonoacetate analogues of ADPR, in which the pyrophosphate is substituted by a phosphonoacetate linker. These analogues retain all other ADPR features but would be predicted to carry only a single negative charge at physiological pH. Suitably protected ribose and adenosine building blocks were sequentially coupled to the phosphonoacetate core by carbodiimide-mediated coupling, followed by a global deprotection to give phosphonoacetates 1 and 2. Analogue 2, in which the phosphonoacetate ester is attached to the terminal ribose 5 00 -hydroxyl, has unexpected instability in TEAB buffer. Future efforts would seek to avoid using ester links in such analogues, but likely employ amide linkages. Neither analogue showed any agonist activity at hTRPM2, nor antagonist effects on the ADPR-induced current at TRPM2, suggesting the full pyrophosphate motif is required for activity and adding to the growing picture of a highly specic binding interaction with many requisite ADPR features.

Experimental section
General Reagents and solvents were purchased from commercial sources and used without further purication, unless described otherwise. Triethylamine was dried with potassium hydroxide, puried by distillation, and stored over potassium hydroxide. MilliQ quality water was used for all aqueous experiments and purications. All reactions were performed under an inert atmosphere of argon unless otherwise stated. For NMR experiments; 1 H, 13 C, and 31 P NMR spectra were collected using either a Varian Mercury 400 MHz or Bruker Avance III 500 MHz Spectrometer. All 1 H and 13  Diethylphosphonoacetic acid (4). Compound 4 was prepared from commercially available triethyl phosphonoacetate following the described procedure. 27 Yield 97% of compound 4.  (5). Diethylphosphonoacetic acid 4 (0.9 g, 4.59 mmol) and 2 0 ,3 0 -O-isopropylidene adenosine (1 g, 3.25 mmol) were dissolved in THF (50 mL). N,N 0 -Dicyclohexylcarbodiimide (DCC, 1 g, 4.8 mmol) was added to the stirred solution in one portion, followed by catalytic amount of DMAP (20 mg, 0.16 mmol). The resulting mixture was stirred at room temperature overnight under argon atmosphere. The precipitate (N,N 0 -dicyclohexylurea) was ltered off and the solution was evaporated. The residue was dissolved in EtOAc, cooled to 5 C, and ltered through a sintered glass lter. The solution was evaporated and the crude product was puried by Isco-Flash chromatography eluted with petroleum ether-EtOAc (6 : 4 v/v) and a gradient of MeOH (0-10%). This procedure afforded the title compound 5 as colourless glass (0.98 g, 62%). 1  14 mmol) in dry pyridine (6 mL) at 0 C. The mixture was stirred at room temperature for 5 h and then another portion of bromotrimethylsilane (0.5 mL) was added dropwise. The mixture was stirred at room temperature and monitored via phosphorus NMR, 31 P NMR (161 MHz, d 6 -DMSO) d 8.14. Aer 16 h, the solvent was evaporated and the residue was coevaporated with aqueous ammonia. The crude material was puried by silica gel chromatography. Impurities were eluted with EtOAc-MeOH-H 2 O (7 : 2 : 1 v/v/v + 0.1% Et 3 N) and the product was eluted using iPrOH-H 2 O-aq. ammonia (7 : 2 : 1 v/v/v + 0.1% Et 3 N). This procedure afforded the title compound 6 as a white foam, as the mono-triethylammonium salt (179 mg, 30%). 1   00 ,3 00 -O-isopropylidene-b-D-ribofuranoside 7 (10.7 mg, 0.0165 mmol) was dissolved in 75% aqueous CF 3 COOH (8 mL) and the mixture was stirred at room temperature for 2 h. The reaction mixture was quickly evaporated at 30 C to dryness and the residue was co-evaporated (3Â) with Milli-Q water. The crude product was dissolved in triethylammonium bicarbonate buffer (TEAB, 2 mL, 0.1 M) and puried by semi-preparative, reversephase HPLC using a gradient of 0.1 M TEAB-acetonitrile (95 : 5 / 35/65 v/v). The title compound 1 was afforded as colourless glass, as a mixture of a and b anomers in the ratio a : b ¼ 2 : 3, (3.1 mg, 36%, 1 3 (5 mL). N,N 0 -Dicyclohexylcarbodiimide (142 mg, 0.68 mmol) was added to the stirred solution in one portion followed by catalytic amount of DMAP (2 mg, 0.016 mmol). The resulting mixture was stirred at room temperature for 16 h under argon. The precipitate (N,N 0 -dicyclohexylurea) was ltered off and the solution was evaporated to dryness. The residue was dissolved in EtOAc, cooled to 5 C, and ltered through a sintered glass lter. The solution was evaporated to dryness and the crude product was puried by Isco-Flash chromatography eluting with petroleum ether-EtOAc (1 : 0 / 0 : 1 v/v) to afford the title compound 8 as a colourless glass (112 mg, 60%). 1  at room temperature. Patch pipettes were prepared from borosilicate glass tubing 1.05 Â 1.50 Â 80 mm, Science Products (Hoeim, Germany) using a Sutter P-97 horizontal puller resulting in a pipette resistance of 1.5 MU to 3.5 MU. An EPC-10 amplier was used in conjunction with PatchMaster (v2x32, HEKA Elektronik, Lambrecht/Germany) to record the data. Aer break-in series resistance was compensated to 70% and the holding potential set to À50 mV. A total of 90 voltage ramps from À85 mV to +20 mV over 140 ms were applied every 5 s to follow activation of TRPM2 activation by ADPR or ADPR analogues added to the pipette solution (in mmol L À1 : 120 KCl, 8 NaCl, 1 MgCl 2 , 10 HEPES, 10 EGTA, 5.6 CaCl 2 , pH 7.2 with KOH, 200 nmol L À1 free Ca 2+ as calculated with MaxChelator (https://somapp.ucdmc.ucdavis.edu/pharmacology/bers/ maxchelator/CaEGTA-NIST.htm)). The peak current at +15 mV during the course of the recording was taken for further quantitative analysis of TRPM2 activity. Statistics. For statistical evaluation of patch-clamp data GraphPad Prism (version 7.04, GraphPad Soware Inc.) was used. Currents from the patch-clamp recordings showed a positively skewed distribution and were normalized by log-transformation. The normalized data were tested using one-way-ANOVA with posthoc testing against the appropriate control using Bonferroni correction for multiple testing (a ¼ 0.05).