Inositol Adenophostin: Convergent Synthesis of a Potent Agonist of d-myo-Inositol 1,4,5-Trisphosphate Receptors

d-myo-Inositol 1,4,5-trisphosphate receptors (IP3Rs) are Ca2+ channels activated by the intracellular messenger inositol 1,4,5-trisphosphate (IP3, 1). The glyconucleotide adenophostin A (AdA, 2) is a potent agonist of IP3Rs. A recent synthesis of d-chiro-inositol adenophostin (InsAdA, 5) employed suitably protected chiral building blocks and replaced the d-glucose core by d-chiro-inositol. An alternative approach to fully chiral material is now reported using intrinsic sugar chirality to avoid early isomer resolution, involving the coupling of a protected and activated racemic myo-inositol derivative to a d-ribose derivative. Diastereoisomer separation was achieved after trans-isopropylidene group removal and the absolute ribose–inositol conjugate stereochemistry assigned with reference to the earlier synthesis. Optimization of stannylene-mediated regiospecific benzylation was explored using the model 1,2-O-isopropylidene-3,6-di-O-benzyl-myo-inositol and conditions successfully transferred to one conjugate diastereoisomer with 3:1 selectivity. However, only roughly 1:1 regiospecificity was achieved on the required diastereoisomer. The conjugate regioisomers of benzyl derivatives 39 and 40 were successfully separated and 39 was transformed subsequently to InsAdA after amination, pan-phosphorylation, and deprotection. InsAdA from this synthetic route bound with greater affinity than AdA to IP3R1 and was more potent in releasing Ca2+ from intracellular stores through IP3Rs. It is the most potent full agonist of IP3R1 known and .equipotent with material from the fully chiral synthetic route.


■ INTRODUCTION
Inositol 1,4,5-trisphosphate receptors (IP 3 Rs) are intracellular Ca 2+ channels. In most animal cells, IP 3 Rs release Ca 2+ from the endoplasmic reticulum (ER) in response to the many extracellular stimuli that evoke formation of D-myo-inositol 1,4,5-trisphosphate (IP 3 , 1, Figure 1). IP 3 Rs thereby generate Ca 2+ signals that regulate diverse cellular processes. 1 IP 3 R activation is initiated by IP 3 binding to the IP 3 -binding core (IBC, residues 224−604) of each subunit of the tetrameric receptor. 2 The two domains (α and β) of the IBC form a clamshaped structure, lined by conserved residues that coordinate IP 3 (Figure 2a). 3 The 4-phosphate of IP 3 interacts primarily with IBC-β, whereas the 1-and 5-phosphates interact predominantly with residues in IBC-α. Interaction of the critical vicinal 4-and 5-phosphates 4,5 with opposing sides of the clam allows IP 3 to partially close the clam. 6−8 This initial conformational change then propagates through putative Ca 2+binding sites to the pore, where the movement of occluding hydrophobic residues within the pore opens a path for Ca 2+ to leak from the ER lumen to the cytosol. 1,8 Although an extensive understanding of the structure− activity relationships (SARs) of IP 3 analogues and other noninositol-based derivatives has been established, there is still much ongoing interest in the structure-based design of new ligands of IP 3 Rs. 9−12 Adenophostin A (AdA, 2) and its acetate analogue, adenophostin B (AdB, 3), are highly potent IP 3 R agonists isolated from Penicillium brevicompactum. 13 Both compounds bind to IP 3 Rs with about tenfold greater affinity than IP 3 , and they stimulate Ca 2+ release with about tenfold higher potency than IP 3 . 1414 These compounds and their analogues have elicited much synthetic and biological interest. 1,4,5 The vicinal 3″,4″-bisphosphate motif in AdA resembles the critical 4,5-bisphosphate moiety in IP 3 . SAR considerations show that the 3″,4″-bisphosphates in AdA ( Figure 5, red area), although not essential for Ca 2+ release, 5 are important since 3″dephospho-AdA ( Figure 3, compound 6) is almost 10 000-fold less potent than AdA, and 4″-dephospho-AdA ( Figure 3, compound 7) is inactive. 5,18 Loss of the 6-OH group from IP 3 causes a ca. 100-fold decrease in potency. 4 The 2″-hydroxyl of AdA may partially mimic the 6-OH of IP 3 since manno-AdA ( Figure 3, compound 9), in which the C2″ configuration is inverted, is 5−10-fold less potent than AdA. 10,14 The enhanced potency of AdA was proposed to be due to its adenosine moiety, causing a better positioning of the 2′-phosphate relative to the 1-phosphate of IP 3 , but this was not confirmed 17 (vide infra). The 80-fold loss of potency after removing the 2′phosphate from AdA to give 8 ( Figure 3) is less than the 200fold loss of potency associated with loss of the 1-phosphate from IP 3 . 1818 An attempt to move the 1-phosphate of IP 3 further from the ring did not enhance activity. 19 Removing the CH 2 OH group from the C5″-position of the glucose unit has a minimal effect as indicated for xylo-AdA ( Figure 3, compound 10), which is approximately twofold less potent than AdA. 20 The C5′ hydroxyl is also unimportant, as indicated by analogues with aromatic groups conjugated at the C5′-position (Figure 3, compound 11), both of which are equipotent with AdA. 21 This allowed attachment of a large fluorescent moiety, providing both a potent fluorescent analogue of AdA and a means of measuring low concentrations of IP 3 . 2222 An intact ribose ring is important for maintaining AdA in a conformation for binding to the IBC. Early efforts to simplify the structure of AdA led to glucopyranoside 2′,3,4-tri-sphosphate ( Figure 3, compound 12), 23,24 a full agonist with tenfold lesser potency than IP 3 , indicating the likely importance of the more constrained ribose moiety to keep its phosphate group in the correct position. Acyclophostin ( Figure 3, compound 13) was also designed to provide an analogue with an opened ribose ring, 25 with the adenine base attached to the anomeric position of glucose via a flexible three-carbon chain. Acyclophostin has a slightly higher affinity than IP 3 , but its Ca 2+ -mobilizing activity is pH-dependent. 25 Most recently, polyphosphorylated analogues derived from both D-and L-glucose were synthesized, some of which can be viewed as truncated analogues of AdA that refine SAR understanding. 26 Removal of the adenine moiety or an electron-rich aromatic ring from the C1′-position of AdA leads to analogues ( Figure  4, compounds 4, 14, 15) with reduced potency. 10,17 Furanophostin (15) and ribophostin (4), in which the adenine is replaced with a H or methoxyl group, respectively, have similar potency to IP 3 , revealing the minimal substitution at C1′ to achieve potent receptor activation. 27,28 Introduction of an imidazole ring at C1′ led to imidophostin (14), equipotent  showing that the N 6 -position may be removed. 29 A bulky group is also tolerated at the N 6 -position, consistent with the purine moiety being in a rather open area of the IBC. 15 Similarly, guanophostin ( Figure 4, compound 17) is also equipotent with AdA. 30 However, when the adenine is replaced with an indole, compounds ( Figure 4, compound 18) show reduced binding affinity; interestingly, however, the 4fluoroindole derivative ( Figure 4, compound 18, X = F) is IP 3 R1-selective. 31 More recently, a triazole ring replacement for adenine led to triazolophostins ( Figure 4, compound 19), potent agonists of IP 3 R. 32 Triazolophostin (19, X = H) is almost as potent as AdA in releasing Ca 2+ through IP 3 R1. While the imidazole analogue ( Figure 4, compound 14) is only slightly more potent  than IP 3 , the triazole equivalent (19, X = H) is 13 times better, suggesting subtle effects on the binding. Dimer analogues ( Figure 4, compound 20) are broadly equipotent to AdA. 33 SAR and modeling studies suggest the potential for enhanced binding of the adenine ring with Arg504 in the IBC through a cation−π interaction ( Figure 2b). 15,34 Studies with a mutated IP 3 R1 with Arg504 replaced by glutamine revealed reduced activity for both IP 3 and AdA; however, the detrimental effect is more marked for AdA than for IP 3 (353fold vs 13-fold 17 ). This concurs with the observation that Arg504 plays a far more important role in AdA binding, potentially via the cation−π interaction. 17,18 Recent analysis using single-particle cryo-EM, however, suggested a different binding mode for AdA (Figure 2c), one in which the adenine ring interacts with residues from R265 to S277 of β-TF2. 16 Further investigations are needed to validate this binding mode, however, as the ligand of the model presents incorrect configurations on both glucose and ribose rings and clashes between the adenine and residues in β-TF2 were observed (PDB 6MU1). Thus, the exact binding mode of AdA at IP 3 R is controversial, but our best working model is still that shown in Figure 2b.
Despite this extensive SAR insight, most studies have addressed iterative variations in the base, ribose, phosphate, and glucose moieties that are synthetically accessible ( Figure  5). While it has long been hypothesized that in AdA the glucopyranoside moiety directly mimics inositol, direct replacement with an inositol surrogate has remained unexplored until recently when we synthesized the novel InsAdA ( Figure 1) through conjugation of a protected and activated chiral myo-inositol derivative with a protected chiral ribose unit, followed by further elaboration, phosphorylation, and deblocking. 35 InsAdA, with a D-chiro-inositol substituting for an α-D-glucose moiety, was interestingly slightly more potent than AdA itself. Importantly, the extra 6″-axial hydroxyl group replacing the AdA pyranoside ring oxygen and the overall conserved high activity offer new potential for wider synthetic elaboration than just AdA itself. Moreover, although missing the adenine base, the corresponding D-chiro-inositol ribophostin counterpart (Figure 1, compound 4a) is the most potent small-molecule IP 3 receptor agonist without a nucleobase yet synthesized, with potency and binding affinity for IP 3 R approaching those of AdA. 36 We now report an alternative synthetic strategy employing intrinsic D-ribose chirality and involving the separation of the diastereoisomers of a suitably protected chiro-inositol-ribonucleoside conjugate derived from a racemic myo-inositol building block. Optimal conditions for regio-monobenzylation  on the inositol ring of protected conjugates and for the removal of different protecting groups were also explored. InsAdA was evaluated biologically, both for its IP 3 R1 affinity in equilibrium competition binding assays and its ability to release Ca 2+ from intracellular stores of permeabilized cells, and was directly compared with material from the earlier synthetic route.

■ RESULTS AND DISCUSSION
Chemistry. Our synthetic route for InsAdA involved coupling a suitably protected and activated racemic myoinositol derivative ( Figure 6A) to a protected D-ribose derivative ( Figure 6B) using a Williamson ether synthetic approach, followed by the introduction of a purine base ( Figure 6C) and subsequent separation of the diastereoisomers of the resulting fully protected conjugates. Configurational inversion upon coupling led to the conversion of the myoinositol motif into a chiro-inositol component. Thus, the nucleoside conjugates possess both D-and L-chiro-inositol. The assignment of the absolute configuration of diastereoisomers was achieved by comparison of NMR data for material synthesized through the route using chiral precursors. 35 Conditions for regioselective protection of specific hydroxyl groups on the inositol ring were explored using a stannylenemediated reaction. The fully protected conjugated isomers with benzyl protection at adjacent positions of the D-chiroinositol ring were separated successfully and transformed subsequently to InsAdA after amination, removal of paramethoxybenzyl (PMB) protection, pan-phosphorylation, and deprotection. Note, as earlier, 35 that no N 6 amino protection of the adenine moiety before the phosphorylation was planned for the synthetic route.
The six secondary hydroxyl groups of myo-inositol possess similar reactivity. 37 However, the reactivity of free cyclitol hydroxyls differs depending on several factors including ring conformation, hydrogen-bonding interactions with neighboring groups, and the conditions used for protection. 4,38 A commonly used strategy is acid-mediated ketalization of vicinal-diol groups. 37,39 Diketal (±)-1,2:4,5-di-O-isopropylidene-myo-inositol (Scheme 1, 21) can be synthesized from myo-inositol by a simple two-step process. 40,41 The 1,2:4,5diketal structure forces the ring into a rigid conformation, facilitating selective individual protection of the other two free hydroxyls as a 6-O-PMB derivative and 3-O-triflate for activation. This conformational rigidity stabilizes the triflate from undergoing β-elimination under alkaline conditions. 35 Less rigid combinations failed in the subsequent coupling step. 35 To tether the protected myo-inositol moiety ( Figure 6A) to the ribose moiety ( Figure 6B) in a second-order nucleophilic substitution (S N 2) fashion, a triflate leaving group needs to be introduced at the 3-position of 21. The C3−OH position of 21 was initially protected via a substantially selective tosylation using 1-(p-toluenesulfonyl)imidazole in the presence of CsF. The 3-mono-tosyl product and 3,6-di-tosyl byproduct were obtained in a respective 6:1 ratio, and the mixture subsequently subjected to para-methoxybenzylation, followed by detosylation using magnesium in methanol and dichloromethane 42 to give (±)-22. In the process, the 3,6-di-tosyl byproduct was converted back to (±)-21 that can be easily removed. All reactions could be achieved in an efficient one-

ACS Omega
http://pubs.acs.org/journal/acsodf Article pot process, and we found that combining these three steps without separating byproducts greatly improves efficiency. The reaction of (±)-22 with triflic anhydride in the presence of pyridine in DCM generated the desired activated precursor (±)-23 (Scheme 1). The synthesis of the diastereoisomeric conjugate between the protected DL-myo-inositol and D-ribose building blocks involved a secondary−secondary S N 2 ether formation between the 3-oxyanion of the ribose derivative and the 3′-O-triflate group of the myo-inositol unit. The structural feature of a chiroinositol−ribose conjugate is found in the nucleoside antibiotic molecule adenomycin (24), produced by Streptomyces griseoflavu 43,44 (Figure 7). However, in this case, the inositol is attached to the primary alcohol of the ribose. Other sugar− inositol conjugates found in prokaryotic and eukaryotic cells include small molecules such as galactinol (25), 45 mycothiol (26) 46 and more complex molecules such as glycosylphospha-tidylinositols (GPIs), 47−49 acting as anchors for cell-surface proteins. As the inositol moiety is tethered to the anomeric position of the sugar molecule, these compounds can usually be prepared via a straightforward O-glycosylation reaction onto the sugar donor. 50−52 The structural feature of an inositol− sugar conjugate with a highly functionalized sec−sec ether linkage as seen in InsAdA is, however, rarely reported in the literature and therefore a unique synthetic route was needed.
The D-ribose moiety ( Figure 6B) needs not only to be fully protected on all hydroxyl groups except the 3-OH, but also to behave like an N-glycosylation donor for the nucleoside synthesis. Numerous methodologies have been developed to prepare a nucleoside via an N-glycosylation reaction. 53 Among them, commonly used strategies include (a) using protected sugar halide as a donor to react with a metal salt of the nucleobase, 54,55 (b) using per-acylated sugars as a donor to react with the base under fusion conditions, 56 and (c) the silyl-Hilbert−Johnson method, modified by Vorbruggen, that has now become the preferred methodology. 57,58 This latter procedure involves a silylated heterocycle reacting with protected sugar acetate in the presence of a Lewis acid such as SnCl 4 or trimethylsilyl trifluoromethanesulfonate (TMSOTf). To further improve selectivity and yield, more efficient methodologies were devised. 59 Fraser-Reid developed a versatile n-pentenyl orthoester donor system, initially used in the synthesis of an oligosaccharide. 60,61 This was later adapted into a reverse synthetic strategy for preparing ribonucleosides, which allows different structural modifications on the ribose moiety before the N-glycosylation occurs in very mild conditions. 62 3′,5′-Dibenzoyl-D-ribofuranose-1′,2′-n-pentenyl orthoester (27) was prepared 62 and after removing benzoyl groups, the primary alcohol was selectively protected with a benzyl group using silver carbonate 63 in more than 90% yield (Scheme 2).
The sec−sec ether formation between 28 and racemic 23 was achieved by reacting the C3′-alkoxide of 28 with triflate (±)-23 under very mild conditions. The C3 S N 2 substitution of the myo-inositol moiety generated the desired conjugate of DL-chiro-inositol and D-ribose (29a, 29b) as a pair of diastereoisomers in a 1:1 ratio in 79% yield (Scheme 3). Initial attempts to separate the diastereoisomers before Nglycosylation were unsuccessful, as the removal of the 2,3trans-O-isopropylidene resulted in ring opening of the orthoester group (Scheme 3). The reaction of 29a and 29b with p-toluenesulfonic acid (pTSA) and ethylene glycol in DCM only generated a mixture of polar byproducts. Therefore, the required 6-chloropurine base element was introduced at this stage through an N-glycosylation reaction promoted by iodonium ion generated in situ from N-iodosuccinimide. 62,64 Reacting the silylated 6-chloropurine with 29a and 29b in the presence of N-iodosuccinimide, ytterbium triflate, and 3 Å molecular sieve in acetonitrile at room temperature produced compounds 30a and 30b in good yields (Scheme 3). We found that adding molecular sieve powder to the system significantly improved the yield. The diastereoisomers of 30a and 30b could not be separated with silica-based chromatography in varied solvent systems. However, selective removal of the 2″,3″-trans-O-isopropylidene group generated the diols 31 and 32 that could be differentiated slightly by silica-based thin-layer chromatography (TLC). The separation of 31 and 32 was difficult using chromatography on a silica column. The optimal separation was eventually achieved by a gradient solvent system of acetone and DCM on a silica column with a loading ratio of 1:133 (w/w, sample/silica). The identification of the less polar isomer 31 as possessing the D-chiro-inositol motif was performed by comparing the NMR data with those of the same compound obtained through a synthetic route with completely chiral precursors. 35 The selective benzylation of the inositol C2″−OH in the presence of C3″−OH needs to be conducted under mild conditions, as the purine base cannot tolerate alkylating reagents under basic conditions. 65 This was illustrated by our failed attempts to benzylate the hydroxyl groups directly in the presence of NaH, DMF/THF, and BnBr. With both C2″ and the C3″ hydroxyls being equatorial, they are positioned in a very similar environment in terms of stereochemistry and hydrogen-bonding capacity. 66 Several conditions for the regioselective alkylation of sugar molecules have been reported using an organic metal reagent as an activator, including an organotin, 67 nickel complex, 68 silver salt, 63 and, more recently, an iron complex. 69 Organotin compounds such as dibutyltin oxide and tributyltin oxide are still widely used for selective manipulation of carbohydrate molecules, 70,71 despite their potential toxicity. Regioselective alkylations were achieved through the preactivation of the two adjacent hydroxyl groups by forming a cyclic dioxolane-like intermediate with the organotin reagent. It was noted, however, that tin-mediated benzylation reactions on trans-diols are less selective as those on cis-diols. 72 To establish suitable conditions for the monobenzylation on the desired 31, we used the myo-inositol derivative racemic 33 as a model for optimizing the tin-mediated reaction (Scheme 4). Benzylation of 33 had been previously performed under either strongly basic conditions 73 or via tin-mediated one-pot reaction conditions. 74 We used a two-phase reaction condition, i.e., isolation of the stannylene-ketal intermediate 34 before adding BnBr, to avoid possible benzylation on the purine moiety. Initially, the reaction was performed in a mixed solvent of toluene and methanol for both phases (Table 1, entry 1). However, no product was observed when the second phase was conducted at room temperature after 3 days. Even though  Compound 35 was obtained in 20% and 27% yields; however, the other regioisomer 36 was also produced in 16−20% yield, indicating low selectivity. Increasing both dibutyltin oxide and BnBr to 2 equiv and changing the solvent for the first phase to acetonitrile improved the selectivity slightly to 2:1 ( Table 1, entry 4). The reaction was then conducted with 3 Å molecular sieves added to the system on the second phase and, with benzyl bromide increased to 3 equiv, compound 35 was obtained in 57% yield with a preference ratio of 3.5:1 (Table 1, entry 5). The benzylation conditions from Table 1, entry 5 were translated to the "undesired" diastereoisomer (32) containing the L-chiro-inositol motif as a model (Scheme 5). The monobenzylated products 37 and 38 were obtained in 58% yield. 1 H NMR spectroscopy indicated that the two regioisomers were present in a 3:1 ratio. These two compounds were inseparable by a silica-based chromatography system. The same benzylation condition was then used on the "desired" D-chiro-inositol derivative 31. However, the benzylation was unfortunately much slower and less selective. Compounds 39 and 40 were obtained in 26% and 31% yields, respectively, with 27% of the starting material 31 recovered (Scheme 6). It was hypothesized that the behavioral difference toward the benzylation process between the two diastereoisomers 31 and 32 was due to stereo effects positioning the two adjacent hydroxyl groups in 31 in a hindered location. An AMMP force field minimization and conformational search were conducted for 31 with the VEGA ZZ program (version 3.1.1). 75 With torsions set for C−O−C linker between the ribose and inositol parts and O-benzoyl group on the ribose ring, the lowest energy conformers with ribose in both 2′-Cendo and 3′-C-endo conformations indicated that the two hydroxyl groups were positioned between the inositol ring and the benzoyl ring, whereas for the "undesired" diastereoisomer 32, the two adjacent hydroxyl groups were positioned in a more exposed area.

ACS Omega
http://pubs.acs.org/journal/acsodf Article dicyanobenzoquinone (DDQ) 76 or ceric ammonium nitrate (CAN). 77 The conditions of 10% TFA in DCM have been successfully employed to remove PMB groups from sugar molecules without affecting the glycosidic linkage. 78 Because of the presence of the sensitive 2,3-O-isopropylidene group, we initially avoided acidic conditions when deprotecting the model molecule 45 74 (Scheme 7). However, the oxidative process using CAN only generated the target compound 33 in low yield, along with the fully hydrolyzed byproduct 47 ( Table  2, entry 1). The same reaction performed using a phosphate buffer system only improved the yield slightly ( Table 2, entry 2). Using DDQ instead of CAN in the buffer system gave a significantly improved yield of 55% without the detection of byproduct 47 (Table 2, entry 3). However, the messy procedure of working up the DDQ reaction prompted us to try the TFA deprotection method 78 on racemic 45. Surprisingly, the PMB ether of 45 was cleaved in as short a time as 3−4 min and gave an 80% yield of the product 33 ( Table 2, entry 4). Longer reaction times even at lower temperatures generated more byproduct 47 (Table 2, entry 5). Therefore, using the same conditions in Table 2, entry 4, compound 41 was converted to the desired precursor 42 in 63% yield. The protected phosphate groups at the C2′-position of the ribose component and the 3″,4″-positions of the inositol component could be introduced via a phosphitylation reaction, followed by oxidation of the resulting P(III) intermediate to the corresponding P(V) triphosphate. Phosphoramidites are among the most widely used phosphitylation reagents with the advantage of being highly reactive, easily accessible, and, importantly, being able to react with hydroxyl groups under very mild conditions. 79 Phosphoramidite reagents are usually activated by weak acids such as 1H-tetrazole; 80 5-(p-nitrophenyl)-1H-tetrazole; 81 4,5-dicyanoimidazol; 82 benzimidazolium triflate; 83 or imidazolium, anilinium, and pyridinium salts. 84 Some activators such as pyridinium hydrochloride with aniline 85 or imidazolium triflate 86 could selectively promote Ophosphorylation over N-phosphorylation, therefore avoiding the necessity of protecting NH 2 of nucleotide base.
Part of our strategy was not to protect the amino group of adenine. However, when 42 was reacted with 5.3 molar equiv of dibenzyl N,N-diisopropylphosphoramidite, preactivated with imidazole triflate, and the product subsequently oxidized with tert-butyl hydroperoxide, no significant selectivity between OH groups and the adenine NH 2 was observed; a mixture of tetrakisphosphate 43 and the desired trisphosphate 44 was obtained in a respective 4:3 ratio (Scheme 6). N 6phosphoramidate was distinguished easily from the other phosphotriesters by its 1 H NMR resonance at 8.67 ppm for NH and a 31 P NMR broad peak at 0.90 ppm. Nonselectivity was most likely due to an excess amount of phosphoramidite, which was used to ensure the complete hydroxyl phosphorylation.
N-Phosphoramidates can be hydrolyzed using strongly acidic, 87−90 strongly basic, or metallic reagents. 91 Hydrogenation of N-phosphate benzyl ester generated partially deprotected phosphoramidate that could be further converted to the fully deblocked free amino compound under controlled mild acidic conditions. 92 Our final deprotection step was performed using a catalytic hydrogen transfer reaction with Pd(OH) 2 as a catalyst and cyclohexene as a hydrogen source. After refluxing the mixed tris-O-phosphate-mono-N-phosphoramidate benzyl esters 43 in a mixture of methanol−water− cyclohexene over Pd(OH) 2 on charcoal for 18 h, all protecting groups were removed, leading to the final target compound 5 in its free acid form. The concomitant removal of the Nphosphate and O-isopropylidene groups, attributed to the intrinsic acidity of the free phosphate groups, was observed in our earlier work. 35 The strategy of removing four different types of protecting groups efficiently in a single step is particularly useful in solution-phase nucleotide synthesis to avoid cumbersome extra N-protection methods. Furthermore, the strategy is more suitable when the availability of only small amounts of material might preclude the easy use of phosphoramidite activators, e.g., imidazolium triflate for selective O-vs N-phosphitylation since this requires very careful titration of reagents and rigorous exclusion of water. 93 Purification using semipreparative reverse-phase high-perform-  In comparison with our earlier synthetic method using a chiral inositol building block as a precursor, 35 the alternative approach reported here achieved the fully chiral target molecule using just the intrinsic protected D-ribose unit chirality to avoid resolution at an early stage, thus saving several steps. The present synthetic route should therefore offer a somewhat more economical and practical approach than that reported earlier, 35 especially to the now anticipated further analogues of this highly potent ligand and those exploiting substitutions on the 6″-axial hydroxyl group that replaces the pyranoside ring oxygen of AdA. Equally, this broad synthetic approach may also be applicable to such analogues of D-chiroinositol ribophostin (Figure 1, compound 4a), 36 a related and simpler ligand that showed unexpectedly potent InsP 3 R activity in comparison to its disaccharide counterpart ribophostin and worthy of further development.
Biology. We assessed the ability of InsAdA to bind with IP 3 R and stimulate Ca 2+ release. In equilibrium competition binding assays using [ 3 H]−IP 3 , InsAdA bound to the fulllength IP 3 R1 of cerebellar membranes with greater affinity than IP 3 , and its affinity was slightly, but not significantly, higher than that of AdA (Figure 8a and Table 3). The shared features of InsAdA and AdA provide likely explanations for the high affinity of InsAdA. The cyclitol ring of InsAdA most likely adopts a conformation that locates the critical 3″,4″bisphosphates in the same area as that of the 3″,4″-bisphosphates of AdA, and the C2″−OH of inositol sits in the area occupied by the 2″-OH of AdA. The rigidity of the ribose unit plays a key role in keeping the geometrical relationship of the cyclitol and adenine rings, which is critical for the electron-rich adenine ring to interact with positively charged residues in the IBC and presumably Arg504. The exposed cis-diols in InsAdA are unlikely to have a substantial impact on the binding affinity; however, they may function as versatile handles to attach other groups for further SAR study and antagonist design.
In assays of Ca 2+ release from the intracellular stores of permeabilized HEK cells expressing only IP 3 R1 (HEK-IP 3 R1 cells), maximally effective concentrations of IP 3 , AdA, or InsAdA released the same fraction (∼70%) of the intracellular Ca 2+ stores. However, InsAdA was more potent than either AdA or IP 3 (Figure 8b and Table 4). The potency of InsAdA  Table  4. (c) Concentration-dependent effects of InsAdA obtained via two different synthetic routes on Ca 2+ release from intracellular stores. Results (%, means ± SEM, n = 6) show Ca 2+ release from the intracellular stores of permeabilized HEK-IP 3 R1 cells evoked by the indicated concentrations of InsAdA and "chiral route" InsAdA. There were no significant differences between the ligands in the values for Ca 2+ release (%) or pEC 50 (P < 0.05). pEC 50 = 7.65 ± 0.14 and 7.67 ± 0.14 for InsAdA and "chiral route" InsAdA, respectively. Ca 2+ release (%) = 74 ± 2 and 72 ± 3 for InsAdA and "chiral route" InsAdA, respectively. (A) Data for IP 3

ACS Omega
http://pubs.acs.org/journal/acsodf Article was the same as that determined for compounds obtained via a totally chiral synthetic route ("chiral route" InsAdA, Figure  8c). 35 IP 3 and AdA are full agonists; 94 we therefore used the ratio of the concentrations of ligand required to evoke halfmaximal Ca 2+ release (EC 50 ) and to occupy 50% of IP 3 Rs (equilibrium dissociation constant, K d ), the EC 50 /K d ratio, as an indication of agonist efficacy, which is the ability of the ligand to activate the IP 3 R once it has bound to it. The EC 50 / K d ratios were similar for all three ligands ( Table 5), suggesting that InsAdA is a full agonist of IP 3 R.
Thus, we have shown that a convergent synthetic route can achieve the target InsAdA, fully active in two biological assays, using the intrinsic chirality of a D-ribose building block and without an early resolution step. InsAdA prepared by this route is biologically equipotent with that from the synthetic route using only chiral precursors. 35 Now that the high potency of InsAdA has been reaffirmed, this ligand also provides importantly and, unlike its parent AdA, an axial cyclitol hydroxyl group for potential further synthetic elaboration. As will be noted from the position of the pyranoside oxygen of AdA in the binding model in Figure 2a, its replacement with an axial hydroxyl, 35 if mirrored by InsAdA binding, should offer through suitably substituted derivatives, the potential for direct targeting of the IBC clam cleft that may be useful in the future antagonist design.
It is naturally tempting to speculate why InsAdA is slightly more potent than AdA and how this might relate to the simple replacement of a glucose with chiro-inositol. This is clearly the first example of such an analogue, and even though this might be viewed as a relatively conservative modification, the consequences at an SAR level are likely to be multifactorial and the SAR profiles rehearsed above may not be definitive enough to rationalize its activity. This may best be tackled when further such InsAdA analogues with modifications to the chiro-inositol ring are synthesized and evaluated biologically. Activity ideally needs to be benchmarked to more members of a closely related structural series and perhaps also in concert with studies on IP 3 R-binding site mutants. For the present, however, we might reasonably assume that InsAdA could bind in a similar fashion to AdA, with its vicinal bisphosphates engaging the same receptor elements as for AdA and IP 3 and its adenine still interacting specifically with Arg504 as for AdA, and according to our working model (Figure 2b). Previously rehearsed possibilities for mimicry of the 1′-phosphate of IP 3 by the 2″-phosphate of AdA involving Arg568 14,17 of the αdomain may also be valid. Conformational flexibility, however, is likely to be impacted through the broad structural change initiated in InsAdA. We should also note that InsAdA represents two pharmacophoric components conjoined through an ether linkage and not via the disaccharide linkage of AdA. Overall conformation might be influenced by exoanomeric and other steric effects different from AdA and conformational populations presented to the IP 3 R could differ between the two ligands. For the present, it is encouraging to note that the high potency of AdA has been closely maintained in the new analogue, which importantly offers slightly more synthetic diversity for development. Thus, the potential for using InsAdA to design further ligands of interest is recognized, although the obvious difficulty and length of the synthetic procedures to access InsAdA analogues may well be a limiting factor in progressing this series.

■ CONCLUSIONS
We have developed an alternative synthetic strategy for InsAdA (5), a molecule that combines many structural features of IP 3 , the native ligand of IP 3 R, with AdA, the most potent agonist of IP 3 R. The core structural template with a D-chiro-inositol tethered to a ribonucleoside via a sec−sec ether linkage is rarely found in the literature. The chiro-inositol-ribonucleoside diastereoisomers (31, 32) from coupling of a suitably protected and activated racemic myo-inositol building block to a suitably protected ribose derivative were successfully separated after purine glycosidation and partial deprotection; the desired D-chiro-inositol-ribonucleoside conjugate (31), after configurational assignment, was subsequently converted to the target compound through monobenzylation of the vicinal inositol trans-diol, amination of chloropurine, PMB removal, pan-phosphorylation, and finally full deblocking of all protection. InsAdA was shown to be a high-affinity full agonist of IP 3 R1, more potent than AdA 2 in releasing Ca 2+ from intracellular stores and equipotent to the same material synthesized via a totally chiral route. 35 InsAdA is the most potent simple IP 3 R agonist known and provides a novel structural template for further exploitation of potential IP 3 R agonists/antagonists. In particular, future synthetic elaboration of the axial hydroxyl of InsAdA could provide high potency ligands that may directly target the IBC cleft and thus offer a novel approach to ligand design not achievable with AdA. A full understanding of the activity of InsAdA must ideally await the synthesis of further members of this new structural series but, most encouragingly, the corresponding D-chiro-inositol ribophostin analogue (Figure 1, compound 4a) has already shown unusually high potency in comparison to ribophostin, its disaccharide counterpart. 36 ■ EXPERIMENTAL SECTION Materials for Biological Analyses. HEK-293 cells in which the genes for all three IP 3 R subtypes had been disrupted using the CRISPR-Cas9 technique (HEK-3KO) 2 were from   3 Binding. Cerebellar membranes, which are enriched in IP 3 R1, were prepared from the cerebella of adult Wistar rats as previously described. 94 Equilibrium competition binding assays (4°C, 5 min) were performed in medium (500 μL) comprising 50 mM Tris, 1 mM ethylenediaminetetraacetic acid (EDTA), pH 8.3 with 3 H−IP 3 (19.3 Ci/mmol, 1.5 nM), cerebellar membranes, and competing ligands. 94 Bound and free ligands were separated by centrifugation (20 000g, 5 min, 4°C). Nonspecific binding, determined by the addition of 10 μM IP 3 or by extrapolation of competition curves to infinite IP 3 concentration, was <10% of total binding.
Cell Culture and Transfection. HEK cells were cultured in DMEM/F-12 GlutaMAX medium supplemented with 10% FBS at 37°C in 95% air and 5% CO 2 . Cells were passaged or used for experiments when they reached confluence. HEK cells expressing only IP 3 R1 (HEK-IP 3 R1 cells) were generated by transfecting HEK-3KO cells with the gene encoding rat IP 3 R1 (lacking the S1 splice site) cloned into pcDNA3.1(−)/Myc-His B plasmid using TransIT-LT1 reagent following the manufacturer's instructions. To generate stable cell lines, cells were passaged 48 h after transfection in a medium with G418 (1 mg/mL) and the selection was maintained for 2 weeks. Monoclonal cell lines were selected by plating cells (∼1 cell/ well) into 96-well plates in a medium containing G418 (1 mg/ mL). After 4 days, wells with only one cell were identified, and cells were grown to confluence. After expansion, expression of IP 3 R1 was confirmed by western blotting using an antibody specific for IP 3 R1. 94 Ca 2+ Release from Intracellular Stores. Mag-fluo 4, a low-affinity fluorescent Ca 2+ indicator, was used to monitor free [Ca 2+ ] within the ER lumen 93 of HEK-IP 3 R1 cells. Cells were loaded with an indicator by incubating cells with Magfluo 4 AM (20 μM, 60 min, 22°C) in N-(2-hydroxyethyl)piperazine-N′-ethanesulfonic acid (HEPES)-buffered saline (HBS; 135 mM NaCl, 5.9 mM KCl, 11.6 mM HEPES, 1.5 mM CaCl 2 , 11.5 mM glucose, 1.2 mM MgCl 2 , pH 7.3) as described. 95 After washing and permeabilization with saponin (10 μg/mL, 37°C, 2−3 min) in Ca 2+ -free cytosol-like medium (Ca 2+ -free CLM), cells were centrifuged (650g, 3 min) and incubated for 7 min in Ca 2+ -free CLM to ensure stores were fully depleted of Ca 2+ . Cells were further centrifuged (650g, 3 min) and resuspended in Mg 2+ -free CLM supplemented with CaCl 2 to give a final free [Ca 2+ ] of 220 nM after the addition of 1.5 mM MgATP. Ca 2+ -free CLM comprised 20 mM NaCl, 140 mM KCl, 1 mM ethylene glycol-bis(β-aminoethyl ether)-N,N,N′,N′-tetraacetic acid (EGTA), 20 mM piperazine-N,N′bis(2-ethanesulfonic acid) (PIPES), and 2 mM MgCl 2 , pH 7.0. Cells (∼3 × 10 6 /well) were attached to poly-L-lysine-coated 96-well black-walled plates (Greiner Bio-One, Stonehouse, U.K.), and fluorescence (excitation and emission at 485 and 520 nm, respectively) was recorded at intervals of 1.44 s using a FlexStation III plate reader (Molecular Devices, Sunnyvale, CA). MgATP (1.5 mM) was added to initiate Ca 2+ uptake, and when the ER had loaded to steady state (∼2.5 min), cyclopiazonic acid (CPA, 10 μM) was added to inhibit the ER Ca 2+ pump. IP 3 and related ligands were added after a further 60 s. The amount of Ca 2+ released was calculated as a percentage of the fluorescence signal from fully loaded stores (F full ) minus the signal from nonloaded stores (F full − F empty ).
Chemistry. Chemical reagents were purchased from either Sigma-Aldrich (Gillingham, U.K.) or Alfa Aesar (Heysham, U.K.). AR-grade solvents and anhydrous solvents were purchased from Sigma-Aldrich (Gillingham, U.K.) or Fisher Scientific (Loughborough, U.K.) and used as supplied unless specified individually. Superdry DCM and MeCN were prepared by distillation over CaH 2 and stored over 3 Å molecular sieves. (±)-1,2:4,5-Di-O-isopropylidene-myo-inositol (21) was synthesized from myo-inositol following a literature process. 40,41 The ribose analogue 28 was prepared using methods reported previously. 35 Analytical thin-layer chromatography (TLC) was performed using precoated plates (Merck aluminum sheets silica 60 F 254 ) and visualized under a UV lamp (254 nm) and/or by staining in ethanolic phosphomolybdic acid (PMA) or aqueous potassium permanganate (KMnO 4 ), followed by heating. Flash column chromatography was performed on a CombiFlash Rf Automated Flash Chromatography System (Teledyne Isco, Lincoln, NE) equipped with a UV detector using RediSep Rf disposable silica gel columns or high-performance GOLD silica columns. Analytical HPLC analyses were carried out on a Waters 2695 Alliance module equipped with a Waters 2996 photodiode array detector (210−350 nm).  13 C NMR, and 31 P NMR spectra were recorded on a Bruker AVIII HD 400 spectrometer at 400, 100, and 162 MHz, respectively. 1 H, 13 C, and 31 P NMR spectra of compound 5 were recorded with a Bruker Avance III HD 500 MHz Spectrometer at 500, 126, and 202 MHz, respectively. Chemical shifts are reported in parts per million (ppm) relative to the solvent residual peaks as internal standards.  (22). To a mixture of 1-(p-toluenesulfonyl)imidazole (4.16 g, 18.7 mmol) and 21 (4.8 g, 18.5 mmol) in DMF (20 mL) was added CsF (3.372 g, 22.2 mmol). The mixture was stirred at rt under argon for 18 h and evaporated in vacuo. The resulting syrup was partitioned between CH 2 Cl 2 (400 mL) and water (400 mL), and the organic layer was ACS Omega http://pubs.acs.org/journal/acsodf Article filtered through a bed of celite and evaporated to provide a white solid (6.91 g) that was identified by 1 (23). Compound 22 (4.39 g, 11.5 mmol) was dissolved in CH 2 Cl 2 (55 mL) and pyridine (5 mL) and cooled to 0°C. Trifluoromethane sulfonic anhydride (13.8 mL, 13.8 mmol, 1 M in CH 2 Cl 2 ) was added dropwise. The reaction mixture was stirred for 30 min at 0°C and then warmed up to rt for 2 h. The mixture was diluted with CH 2 Cl 2 (200 mL), washed with water (200 mL), and brine was added (20 mL). The organic layer was dried by filtration through solid NaCl and concentrated in vacuo to give compound 23 as a pale beige amorphous solid (5.72 g, 97%). 1 H and 13 C NMR data were identical to the previously reported data for the material prepared by the chiral route. 35 3  (29a, 29b). To a solution of 28 35 (740 mg, 1.79 mmol) in dry THF (3 mL) was added dry hexamethylphosphoramide (8 mL), followed by NaH (90 mg, 95%, 3.6 mmol). The mixture was stirred under argon at rt for 20 min and cooled to 0°C. (±)-1,2:4,5-Di-O-isopropylidene-6-O-p-methoxybenzyl-3-O-trifluoromethane sulfonyl-myo-inositol (±-23, 1.2 g, 2.34 mmol) was added, and the mixture was stirred at room temperature for 12 h and concentrated in vacuo. The mixture was then partitioned between ethyl ether (80 mL) and saturated ammonium chloride solution (60 mL). The organic solution was washed with brine, dried over Na 2 SO 4 , and concentrated. The crude product was purified using a Combiflash and eluted with a solvent gradient of petroleum ether to 20% EtOAC/petroleum ether to give a clear oil. (1.1 g, 79%). 1 H NMR indicated the two diastereoisomers were in a 1:1 ratio. 1 (30a, 30b). 6-Chloro-9-(trimethylsilyl)-9H-purine was prepared by refluxing 6-chloropurine (284 mg, 1.8 mmol) in hexamethyldisilazane (8 mL) for 1.5 h, followed by concentration in vacuo. The yellow residue was mixed with acetonitrile (4 mL) and added to a solution of the diastereoisomers 29a and 29b (1.0 g, 1.3 mmol) in acetonitrile (4 mL) at 0°C. The mixture was stirred over molecular sieves (3 Å, 500 mg) at 0°C for 20 min. A solution of Yb(OTf) 3 (360 mg, 0.58 mmol) and N-iodosuccinimide (450 mg, 2 mmol) in acetonitrile (10 mL) was also stirred with molecular sieves (3 Å, 500 mg) at room temperature for 20 min. The two suspensions were mixed and stirred under argon at room temperature for 24 h. DCM (50 mL) was added, and the mixture was partitioned between DCM (100 mL) and sodium bisulfate solution (4%, 60 mL). The organic solution was washed with brine, dried over Na 2 SO 4 , and concentrated. The crude product was purified with a CombiFlash and eluted with a solvent gradient of DCM to 20% EtOAC/DCM to give a white foamy product (0.78 g, 78%). 1  To a solution of the diastereoisomers 30a and 30b (1.2 g, 1.45 mmol) in anhydrous DCM (25 mL) was added ethylene glycol (144 mg, 2.3 mmol), followed by pTSA (25 mg). The mixture was stirred at room temperature for 30 min and partitioned between DCM (50 mL) and 5% NaHCO 3 (50 mL). The organic solution was washed with brine, dried over ACS Omega http://pubs.acs.org/journal/acsodf Article Na 2 SO 4 , and concentrated to give a foam (1.1 g). The crude product was divided into 300 mg portions, and each portion was subjected to repeated flash chromatography and eluted with a gradient solvent of DCM to 15% acetone/DCM. Compound 31, the less polar product on TLC (40% acetone/ DCM), was obtained as a clear oil (480 mg, 42%). The compound was identified as the D-chiro-inositol derivative (31) by comparing the NMR spectra with those of the same material prepared by the chiral route. 34 [α] D 23 = +4.0°(c = 0.5; CH 3 CN). 1 (36). To a solution of 33 96 (200 mg, 0.50 mmol) in dry CH 3 CN (25 mL) was added Bu 2 SnO (390 mg, 1.1 mmol). The mixture was refluxed in a flask equipped with a Soxhlet extractor containing molecular sieve powder (3 Å) under nitrogen. After 15 h, the mixture was cooled to room temperature and concentrated to dryness in vacuo. DMF (5 mL) was added to the residue, followed by CsF (266 mg, 1.75 mmol), TBAI (37 mg, 0.09 mmol), molecular sieves (3 Å, 120 mg), and BnBr (0.18 mL, 1.47 mmol). The mixture was stirred at room temperature under argon for 18 h, diluted with DCM (50 mL), and filtered through celite. The filtrate was washed with brine, dried over Na 2 SO 4 , and concentrated to give a clear oil. The crude product was subjected to flash chromatography and eluted with a gradient solvent of petroleum ether to 60% EtOAc/ petroleum ether. Compound 35, the less polar product on TLC (50% EtOAc/petroleum ether), was obtained as a clear oil, which turned into a waxy solid (140 mg, 57%). 1 (38). The title compounds were prepared from 32 using the same method as that for 35 and 36. Compounds 37 and 38 were obtained as clear oils (an inseparable mixture in ratio of 3:1, 58% yield). 1 H NMR signals are from the combination of the two regioisomers. 1 (40). To a solution of diol 31 (150 mg, 0.19 mmol) in anhydrous CH 3 CN (15 mL) was added Bu 2 SnO (150 mg, 0.60 mmol). The mixture was refluxed in a flask equipped with a Soxhlet extractor containing activated molecular sieve powder (3 Å) under nitrogen. After 18 h, the mixture was cooled to room temperature and concentrated to dryness in vacuo. DMF (3 mL) was added to the residue, followed by CsF (150 mg, 0.99 mmol), TBAI (20 mg, 0.05 mmol), molecular sieves (3 Å, 120 mg), and BnBr (0.1 mL, 0.82 mmol). The mixture was stirred at room temperature under argon for 16 h, diluted with DCM (50 mL), and filtered through celite. The filtrate was washed with brine, dried over Na 2 SO 4 , and concentrated to give a clear oil. The crude product was subjected to flash chromatography and eluted with a solvent gradient of petroleum ether to 60% EtOAc/ petroleum ether. Compound 39, the less polar product on TLC (50% EtOAc/petroleum ether), was obtained as a clear oil (43 mg, 26%). ACS Omega http://pubs.acs.org/journal/acsodf Article temperature. After 20 min, the solution was added to a flask containing compound 42 (24 mg, 0.037 mmol). The mixture was stirred for another 1 h under the same conditions. 31 P NMR spectroscopy indicated reaction completeness. The reaction was quenched by adding a drop of water and cooled to −78°C. tert-Butyl hydroperoxide (0.3 mL, 70% water solution) was added, and the mixture was brought to room temperature. After stirring for 30 min, the reaction was quenched by adding Na 2 SO 3 (5 mL, 10% water solution), followed by DCM (60 mL). After separation, the organic phase was washed with brine, dried over Na 2 SO 4 , and concentrated in vacuo to give a clear oil. The crude product was subjected to flash chromatography and eluted with a solvent gradient of DCM to 7% MeOH/DCM.