Structural Determinants of Adenophostin A Activity at Inositol Trisphosphate Receptors

Experimental Procedures

Materials.

InsP₃ (1) was from American Radiolabeled Chemicals Inc. (St. Louis, MO). Thapsigargin was from Alamone Laboratories (Jerusalem, Israel), and ionomycin was from Calbiochem (Nottingham, UK). The analogs of adenophostin A were synthesized as follows: adenophostin A (2) (Marwood et al., 2000a,c); purinophostin (3) and imidophostin (5) (Marwood et al., 2000b); ribophostin (6) (Jenkins et al., 1997); Glc(2′,3,4)P₃ (8) (Jenkins and Potter, 1996); 27-29 (Marchant et al., 1997a); furanophostin (7) (Marwood et al., 1999); 17,18, and 26 (Shuto et al., 1998); 9 and10 (Rosenberg et al., 2000); acyclophostin (19) and 25 (Van Straten et al., 1997);20-23 (Rosenberg et al., 2001);xylo-adenophostin (11) andmanno-adenophostin (12) (Marwood et al., 2000c); uridophostin (13); and 4, 14, and30 (Marwood et al., 2000d). The syntheses of 15,16, and furanophostin 3,4-bisphosphorothioate (24, furanophostin-PS₂) will be reported elsewhere. The structures of the analogs used, together with their abbreviations, are shown in Fig. 2.

⁴⁵Ca²⁺ Release from the Intracellular Stores of Permeabilized Cells.

Hepatocytes were isolated from the livers of male Wistar rats (200–300 g) as described previously (Marchant et al., 1997a) and permeabilized in Ca²⁺-free cytosol-like medium (140 mM KCl, 20 mM NaCl, 2 mM MgCl₂, 1 mM EGTA, 20 mM PIPES, pH 7.0) by incubation with saponin (10 μg/ml) at 37°C for 8 min. The cells were resuspended (2.2 × 10⁶cells/ml) in cytosol-like medium supplemented with 300 μM CaCl₂ (free [Ca²⁺] = 200 nM), ATP (1.5 mM), creatine phosphate (5 mM), creatine phosphokinase (1 unit/ml), the mitochondrial inhibitorp-trifluoromethoxyphenylhydrazone (10 μM), and⁴⁵CaCl₂ (5 μCi/ml). After 5 min at 37°C, during which the intracellular stores loaded to steady state with ⁴⁵Ca²⁺, InsP₃, adenophostin, A or an analog were added, together with thapisgargin (1 μM) to inhibit further Ca²⁺ uptake. After a further 1 min, the⁴⁵Ca²⁺ contents of the stores were determined by filtration through Whatman GF/C filters followed by washing with ice-cold sucrose (310 mM) and sodium citrate (1 mM). The actively accumulated⁴⁵Ca²⁺ content of the stores was defined as that which could be released by ionomycin (10 μM).

[³H]InsP₃ Binding to Hepatic Membranes.

Membranes were prepared from perfused rat livers using Percoll-gradient centrifugation as described previously (Beecroft et al., 1999). Briefly, after perfusion in situ with cold buffered saline, the liver was removed, homogenized with a Dounce homogenizer in cold buffered sucrose (25 ml; 250 mM sucrose, 5 mM Hepes, 1 mM EGTA, pH 7.4), filtered through gauze, and then centrifuged (25,000g, 10 min). The pellet was resuspended in buffered sucrose (48 ml) containing Percoll (11.8% final v/v; Amersham Pharmacia Biotech, Uppsala, Sweden), recentrifuged (35,000g, 30 min), and the membranes harvested as a fluffy band just beneath the fatty layer at the top of each tube. The membranes in hypo-osmotic medium (1 mM EGTA, 5 mM HEPES, pH 7.4, 2°C) were centrifuged (48,000g, 10 min) and the pellet resuspended in binding medium (BM: 50 mM Tris, 1 mM EDTA, pH 8.3, 2°C) at ∼20 mg of protein per millilter, before freezing in liquid nitrogen and storage at −80°C.

For measurements of [³H]InsP₃ binding, liver membranes (0.1 mg of protein in 500 μl of BM) were added to [³H]InsP₃ (final concentration 1.3–1.8 nM, specific activity 35–60 Ci/mmol) and the appropriate concentration of competing ligand. After 5 min at 2°C, during which equilibrium was attained (Beecroft et al., 1999), bound and free ligand were separated by centrifugation (20,000g, 5 min, 4°C). Total binding was typically 3000 dpm/tube, of which ∼30% was specific.

Analysis.

Equilibrium-competition binding curves were fitted to logistic equations using nonlinear curve-fitting routines (Kaleidagraph; Synergy Software, Reading, PA):where B is the amount of [³H]InsP₃ bound in the presence of a concentration of the competing ligand, [L]; T and N are the total and nonspecific [³H]InsP₃ binding, respectively; n _H is the Hill coefficient; and IC₅₀ is the concentration of competing ligand causing 50% displacement of specific [³H]InsP₃ binding. TheK _d value for each ligand was then calculated from the IC₅₀ value. A similar equation was used to analyze concentration-effect relationships for agonist-evoked Ca²⁺ mobilization from which the maximal effect, EC₅₀, andn _H of each ligand were determined.

The potencies of analogs are expressed relative to the potency of InsP₃ determined in exactly parallel experiments (see below). The variance of the potency ratio (with meanK _d values of a and b for the two agonists) was calculated from the variances (var) of each using the following equation (Colquhoun, 1971):All results are shown as means ± S.E.M.

Results and Discussion

Adenophostin A Is a Potent Agonist of Hepatic InsP₃Receptors.

The results shown in Fig.3 confirm previous work (Marchant et al., 1997a) by demonstrating that adenophostin A is a full agonist of hepatic InsP₃ receptors with about 10-fold greater affinity than InsP₃. In assays of Ca²⁺ mobilization, adenophostin A was 9.9 ± 1.6-fold more potent than InsP₃, and in radioligand binding assays it had 6.4 ± 1.1-fold greater affinity (Table 1). Similar results have been obtained with other tissues (see above). Whereas our previous work (Marchant et al., 1997a) used adenophostin A purified fromP. brevicompactum (Takahashi et al., 1994a), the present results were obtained with synthetic adenophostin A (Marwood et al., 2000a). However, in parallel comparisons synthetic (EC₅₀ = 9.6 ± 1.0 nM,n _H = 2.46 ± 0.19, n = 3) and natural (EC₅₀ = 9.2 ± 1.7 nM,n _H = 2.99 ± 0.32, n = 3) adenophostin A had indistinguishable effects on Ca²⁺ release from permeabilized hepatocytes (Marwood et al., 2000a). The similarity is important in providing the justification for subsequent analyses of the effects of synthetic analogs of adenophostin A.

All the adenophostin A analogs used in the present study (Fig. 2) that were sufficiently potent to evoke a maximal response at concentrations ≤10 μM released the same fraction of the intracellular stores as a maximal concentration of InsP₃ (∼50%). Furthermore, when a maximally effective concentration of any one of the analogs was combined with 10 μM InsP₃, the response was no greater than that evoked by the InsP₃ alone (data not shown). Although the fraction of the stores released by a maximally effective concentration of InsP₃ varied somewhat between experiments (29–55%), when averaged across all the experiments reported here, 50 ± 4% of the actively accumulated⁴⁵Ca²⁺ was released by 10 μM InsP₃ (Table 1). An indistinguishable response was evoked by a maximal concentration (1 μM) of adenophostin A (Table 1). The responses to InsP₃, adenophostin A, and each of the active analogs were positively cooperative (Hill coefficients, n _H > 1) (Table2). None of the inactive (18, 23) or very weakly active (10,22, 25, 26) analogs behaved as competitive antagonists: the response to a submaximal concentration of InsP₃ (200 nM) was undiminished in the presence of 10 μM of any of these analogs (data not shown).

The results shown in Table 2 summarize experiments performed over a lengthy period during which there were modest variations in the sensitivity of the permeabilized hepatocytes to InsP₃ (the EC₅₀ value varied between 90 ± 7 and 201 ± 20 nM). Similar variations were observed in the binding experiments (theK _d value for InsP₃varied between 2.72 ± 0.88 and 9.65 ± 0.89 nM). To be confident that even the small (although statistically significant) differences in potency (or affinity) observed between some of the analogs were meaningful, we have compared the potency of each analog with the potency of InsP₃ measured in exactly parallel experiments. All figures (Figs. 4-8) show results analyzed in this way (i.e., with potencies expressed relative to InsP₃), whereas Table 2 shows the actual EC₅₀ values determined for each analog.

Role of the Adenosine Moiety in High-Affinity Binding of Adenophostin A.

Adenophostin A (2, Fig. 1) can be viewed as a phosphorylated glucose glycosidically linked at its 1"-position to an adenosine phosphorylated at the 2′-position. In the first series of experiments, we examined the functional consequences of systematically trimming the adenosine (Fig.4). The results show that as elements of the adenine (2-6, 17) and then of the ribose (7-10) ring were successively removed, there was a progressive decrease in the potency of the analog to evoke Ca²⁺ mobilization. Analogs (ribophostin,6; furanophostin, 7; and 17) without an adenine moiety were only slightly less potent than InsP₃, but after complete removal of the adenine and opening of the five-membered ring [Glc(2′,3,4)P₃, 8] the potency fell to 12 ± 1-fold less than that of InsP₃. Shortening of the flexible O-C-C chain in 8 by two atoms to give the α-C-glycoside 9 caused a further slight reduction in potency. Interestingly, the isomeric β-C-glycoside 10 was much weaker, despite its apparent resemblance to InsP₃. Also, in relation to 6 (ribophostin), some of us have recently reported that elaboration of the methyl group in 6 to propyl or phenylpropyl groups did not appreciably affect potency (De Kort et al., 2000).

The importance of the 2′-phosphate of adenophostin A is apparent from the massive difference in potency between compounds 17 and18. Both ligands lack the adenine of adenophostin A, but17 had only 5.46 ± 0.88-fold lesser potency than InsP₃ (Fig. 4), whereas 18, which lacks the 2′-phosphate, was inactive (Fig.5C). These results are consistent with those from radioligand binding analyses of cerebellar membranes where18 bound with ∼300-fold lesser affinity than 17(Shuto et al., 1998), and removal of the 2′-phosphate from adenophostin A reduced affinity by about 100-fold (Takahashi et al., 1994b). There is a similar 100- to 300-fold difference in the affinities of Ins(4,5)P ₂ and Ins(1,4,5)P₃ for InsP₃receptors (Nerou et al., 2001), in keeping with the idea that the 2′-phosphate of adenophostin A mimics the 1-phosphate of Ins(1,4,5)P₃ (Fig. 1). The conclusion gains further support from the 159 ± 19-fold lesser potency of25 relative to acyclophostin (19) (Fig. 5B); the major difference between them being the loss of the 2′-phosphate from25 (Table 2).

Effects of Modifying the Nucleoside and Its Links with the Glucose Ring.

The effects of modifying the base (adenine) of adenophostin A on biological activity are shown in Fig. 5A. Removal of the amino group from the adenine (to give purinophostin, 3) had a minimal effect. Replacement of the pyrimidine ring of the adenine by a benzene ring (4) reduced potency by 4.0 ± 1.0-fold, and its complete removal reduced the potency of the analog (imidophostin, 5) to almost that of InsP₃ (Fig. 5A). Even replacement of the purine moiety (adenine) of adenophostin A with a much smaller pyrimidine (uracil, to give uridophostin, 13) caused the potency to decrease by only 2.31 ± 0.40-fold. Finally, replacement of the entire adenine moiety of adenophostin A with unrelated single (14) or double (15, 16) aromatic rings with a β-C-ribosyl glycosidic linkage produced ligands with activities comparable with that of 4 and significantly greater than that of InsP₃ (Fig.5A). As expected, changing the stereochemistry of theC-glycosidic linkage of 14 from β to α (30) massively decreased potency (by 29 ± 11-fold relative to 14, and to 11.8 ± 2.8-fold less than that of InsP₃). Hence, although large aromatic groups attached to the 1′-position in a β-orientation improve potency (14 is 10.2 ± 1.9-fold more potent than17), they have the opposite effect when attached in the α-orientation (30 is 2.8 ± 0.7-fold lesspotent than 17). The effect may be caused by steric hindrance around the important 2"-OH and/or the 2′-phosphate group, which may occur with the α- but not the β-oriented rings.

These results indicate that although an adenine attached to the 1′-position of the ribose ring through a β-glycosidic linkage gives the most potent agonist (adenophostin A), even radical changes to the structure of the attached group are tolerated. Each of the analogs with a β-glycosidically linked aromatic system (3-5and 13-16) is significantly more potent than17 in which there is no substitution at the 1′-position.

In the adenophostins, the 2′-phosphate group is separated from the α-glucopyranosyl ring by a three-atom -O-C-C- chain. It seems that some conformational restraint of this chain by incorporation of its two C atoms into a five-membered ring of appropriate stereochemistry (e.g., ribose) is necessary for optimal activity. However, removal of part of the ribose ring to give an apparently flexible structure (acyclophostin, 19) produces an analog with significant activity, providing the adenine is retained. Acyclophostin is unusual in that its efficacy is affected by pH (Beecroft et al., 1999), but under the conditions used to compare analogs in this study, it was only 1.37 ± 0.13-fold less potent than InsP₃. Acyclophostin (19) certainly has much greater potency than8 (which lacks the adenine moiety of 19) or25, (which lacks the 2′-phosphate) (Fig. 5B). These results suggest that even when the critical 2′-phosphate is positioned on a flexible link between the adenine and glucose ring, it can still contribute significantly to binding affinity.

Other ring structures can effectively replace the ribose of adenophostin A to give analogs (27-29) (Fig. 2) more potent than 18. However, even the most active of these analogs (28), which is more potent than 8, is still 6.0 ± 0.3-fold less potent than ribophostin (6), where the 2′-phosphate is attached to a ribose ring (Fig. 5C).

It is interesting that analog 26, which is a regioisomer of the relatively potent 17, shows very low activity. The likely reason is that the stereochemical differences between26 and 17 at the positions corresponding to C-2′ and -3′ of adenophostin A result in a different three-dimensional positioning of the 2′-phosphate group at the receptor binding site. Steric hindrance from the differently orientated hydroxymethyl group and five-membered ring may also be involved. Similar factors may account for the low activity of 27, the fructose component of which retains the required stereochemistry, but has a different type of glycosidic linkage to glucose and an extra hydroxymethyl group.

We conclude that of the analogs examined so far, a ribose ring (or close relative) between the pyranosyl and adenine moieties positions the 2′-phosphate most effectively for optimal binding.

Modifications to the Glucose Ring of Adenophostin A.

Whereas InsP₃ is based on a myo-inositol ring structure, the analogous 4,5-bisphosphate and 6-hydroxyl groups in adenophostin A are attached to a glucose ring. The functional consequences of modifying this sugar are shown in Fig.6A.

Xylo-adenophostin (11), in which the glucose of adenophostin A has been replaced by xylose, was only 1.9 ± 0.6-fold less potent than adenophostin A. The structures of the two analogs differ only in that adenophostin (2) has a CH₂OH attached to the 5"-position, which presumably mimics the 3-hydroxyl of InsP₃ (Fig.1). The small difference in potency between 2 and11 is therefore consistent with the relatively minor contribution of the 3-hydroxyl of InsP₃ to binding (Kozikowski et al., 1993; Nerou et al., 2001).

Manno-adenophostin (12) in which mannose replaces the glucose of adenophostin A, is substantially less potent (12.2 ± 2.3-fold) than the parent compound, although still only 1.51 ± 0.15-fold less potent than InsP₃ (Fig. 6A). The potency of 12 is surprising because although it differs from adenophostin only in the orientation of the hydroxyl equivalent to the 6-hydroxyl of InsP₃ (Fig. 1), inversion or removal of the 6-hydroxyl in inositol phosphates invariably reduces potency by at least 100-fold (Kozikowski et al., 1993; Wilcox et al., 1994; Nerou et al., 2001). There are no other known high-affinity agonists of InsP₃ receptors lacking a hydroxyl equivalent to the equatorial 6-hydroxyl of InsP₃. It has been suggested that the 6-OH group of InsP₃ may donate an H-bond to the receptor (Kozikowski et al., 1993), an interaction that is presumably disrupted when this group is axial. However, an additional role for the equatorial 6-OH has been proposed, in which it may influence the behavior of the crucial bisphosphate by mediating its interaction with the 1-phosphate (Felemez et al., 2000). Changing the orientation of the 6-OH to axial (InsP₃ toepi-InsP₃) alters the communication between the 4,5-bisphosphate and 1-phosphate so that they behave effectively as if isolated from one another. This “inframolecular” effect may partly underlie the greatly decreased activity ofepi-InsP₃ (Felemez et al., 2000). In adenophostin A, the analogous 3",4"-bisphosphate and 2′-phosphate group are located on separate rings. The effect of modifying the mediating hydroxyl group (2"-OH) may therefore be different from that in InsP₃.

Stereochemistry of the O-Glycosidic Linkage.

Because xylo-adenophostin (11) is almost as potent as adenophostin A (2), and xylose-based analogs are both simpler and require fewer chemical steps for their synthesis than glucose-based analogs, we chose to explore the functional consequences of modifying the stereochemistry of the O-glycosidic linkage between the pyranosyl and furanosyl moieties using xylose-based analogs (Fig. 6B).

As shown above, 11 is only 1.9 ± 0.6-fold less potent than 2. Just as furanophostin (7), which lacks the adenine and 4′-CH₂OH of adenophostin A, is 22 ± 5-fold less potent than its parent (2), so20, the xylose-based analog of furanophostin, is 17 ± 3-fold less potent than its parent (11). The similarities establish that modifications to the adenosine moiety have similar consequences for glucose- and xylose-based analogs. These results provide the justification for using 20 to explore the functional consequences of changing the stereochemistry of theO-glycosidic linkage between the furanosyl and pyranosyl rings (Fig. 6B).

Changing the stereochemistry of the O-glycosidic linkage to xylose from α (20) to β (23) massively decreased potency, so that 23 was effectively inactive. It is likely that in low-energy conformations of 23, the three-dimensional location of the third phosphate group is very different from that in 20, and that it is therefore unable to mimic the 1-phosphate of InsP₃ or the 2′-phosphate of adenophostin A. Analog 22 is stereochemically similar to the glucose-based 26, and they show similar low potency. Unlike 26, however, 22has no hydroxymethyl groups and so it is likely that its weak activity (and by implication that of 26) is again simply attributable to incorrect spatial positioning of its nonvicinal phosphate group. It is interesting that analog 21, which combines the stereochemical changes to the tetrahydrofuranyl ring of 22with a change from an α- to a β-O-glycosidic linkage shows significant activity (19 ± 2-fold weaker than InsP₃). Molecular models suggest that the combined effect of the stereochemical changes in 21 is to return the nonvicinal phosphate group to a spatial orientation more closely approaching that in 20 than in 22,23, or 26 (Rosenberg et al., 2001).

Phosphorothioate Modifications to the Phosphate Groups.

Substitution of phosphates by phosphorothioate groups in certain inositol phosphates can give partial agonists (Potter and Lampe, 1995;Murphy et al., 2000), and might thereby provide a step toward the development of antagonists. However, the strategy has so far been limited by the relatively low affinity of these phosphorothioate analogs. It seems that partial agonists of this type can be generated by combining a slight structural perturbation of the InsP₃ molecule (particularly at position 3) with phosphorothioate substitution (Potter and Lampe, 1995). Replacement of the 4- and 5-phosphate groups of 3-deoxy-3-fluoro-Ins(1,4,5)P₃ with phosphorothioates, for example, produced a partial agonist with an affinity only 10-fold less than InsP₃ for type 1 InsP₃ receptors (Wilcox et al., 1997). Thus, it seems that higher affinity partial agonists may be created by limiting phosphorothioate substitution to the bisphosphate, leaving the 1-phosphate group, with its ability to enhance affinity, unchanged.

We reasoned that adenophostin analogs such as 7(furanophostin) might be promising candidates for this approach in that they are related to position 3-modified InsP₃analogs (hydroxymethyl of glucose replacing 3-hydroxyl of InsP₃), and retain InsP₃-like affinity for the receptor. Furthermore, their structure (nonvicinal phosphate located on a separate ring) has advantages in terms of synthetic strategy for creating novel ligands with the 2′-phosphate and 3",4"-bisphosphorothioate pattern.

In 24, the two phosphate groups on the glucose ring of furanophostin (7) have been replaced by phosphorothioates to give the first phosphorothioate-containing adenophostin analog, furanophostin-PS₂ (Fig. 2). The result is a substantial decrease in potency (5.1 ± 1.3-fold relative to7) (Table 2). The decrease is consistent with similar decreases in potency after phosphorothioate substitution of InsP₃: inositol 1,4,5-trisphosphorothioate was 2.8-fold less potent than InsP₃ (Taylor et al., 1989). However, maximal concentrations of InsP₃and furanophostin-PS₂ caused similar amounts of Ca²⁺ to be released from the intracellular stores (41 ± 2%, n = 3 for each). Furthermore, in both functional and binding assays furanophostin-PS₂was about 12-fold less potent than InsP₃ (Tables2 and 3). Although rapid measurements of rates of Ca²⁺ release would be required to unequivocally establish the efficacy of furanophostin-PS₂ (Marchant et al., 1997b), our results are certainly consistent with the conclusion that furanophostin-PS₂ is a full agonist of hepatic InsP₃ receptors. A possible explanation for the unexpectedly high efficacy of furanophostin-PS₂may be that its 5"-hydroxymethyl group rather closely mimics the 3-hydroxyl of InsP₃ and therefore fails to provide sufficient structural perturbation around the pseudo-3 position to significantly reduce the efficacy of furanophostin-PS₂. Future attempts to develop high-affinity ligands with low intrinsic activity may therefore require more substantial modifications to the 5"-position of adenophostin analogs.

Binding of Adenophostin A Analogs to Hepatic InsP₃Receptors.

Because InsP₃-binding sites are present at low density in hepatocytes and the only available radioligand ([³H]InsP₃) has relatively low specific activity, radioligand binding analyses are more demanding and costly than functional assays; they must also be performed in a medium (BM) with high pH and low ionic strength. For these reasons, our radioligand binding analyses of adenophostin A analogs have focused on only the key ligands.

The results are summarized in Table 3. In Fig.7 the affinity of each analog is expressed relative to the affinity of InsP₃measured in parallel experiments (under Experimental Procedures); the analogous comparisons from functional assays are shown alongside. Other than acyclophostin (19), which we previously suggested was a partial agonist at pH 8.3 (the pH used for binding assays) but a full agonist at pH7 (Beecroft et al., 1999), there were no substantial disparities between the binding and functional assays. The rank orders of the analogs shown, which included modifications to each of the major elements of adenophostin A, were similar in binding and functional assays (Fig. 7). The results are therefore consistent with each of the analogs (1-3, 5, 6, 12,13, and 24) being a full agonist, differing from adenophostin A only in its affinity for hepatic InsP₃ receptors.

Conclusions

Adenophostin A is the most potent known agonist of InsP₃ receptors. None of the modifications we have made to the structure of adenophostin A has succeeded in either increasing its affinity for InsP₃ receptors or in changing its efficacy. We have, however, designed new ligands based upon adenophostin A with potencies higher than that of InsP₃. Using both these ligands and simplified analogs of adenophostin A, we have established which structural elements of adenophostin A determine its high affinity for InsP₃ receptors (Fig.8) in hepatocytes, which express predominantly type 2 InsP₃ receptors (Taylor et al., 1999).

The glucose 3",4"-bisphosphate and adjacent 2"-hydroxyl of adenophostin A are thought to mimic the critical 4,5-bisphosphate and 6-hydroxyl of InsP₃ (Fig. 1) (Takahashi et al., 1994a). The 2′-phosphate of adenophostin A is thought to mimic the 1-phosphate of InsP₃ and for both ligands removal of this phosphate causes the affinity to decrease by at least 100-fold (Fig. 4) (Shuto et al., 1998). The 3-hydroxyl of InsP₃contributes little to binding (Kozikowski et al., 1993; Nerou et al., 2001), and likewise removal of the analogous hydroxyl from adenophostin A (to give xylo-adenophostin, 11) has minimal effect. These results suggest that structure-activity relationships derived from inositol phosphates can be used to predict the activity of adenophostin analogs, at least for some substituents of the pyranosyl ring. However, the unexpectedly high potency ofmanno-adenophostin (12) suggests that the roles of the 2′-hydroxyl of adenophostin A and the 6-hydroxyl of InsP₃ are not exactly analogous: this hydroxyl group is clearly more important for InsP₃ than adenophostin A binding.

We have explored in detail the contribution of the adenine group to the high-affinity binding of adenophostin A to InsP₃receptors. Progressive removal of elements of the adenine cause progressive decreases in binding affinity (Fig. 4), but the adenine can be replaced by completely unrelated aromatic ring systems (13, 14, and 16) with only modest decreases in potency (Fig. 5A). The stereochemistry of the link between the furanosyl ring and the aromatic rings is, however, crucial: β-1′-aromatic substituent (as occurs in adenophostin A) enhances affinity, whereas a similar group in the α-orientation has the opposite effect (Fig. 5A). A phosphate group at the position equivalent to the 1-position of InsP₃ (2′ in adenophostin A) is essential (see above) and optimally active when attached to a furanosyl ring (Fig. 5). The stereochemistry of theO-glycosidic link between the pyranosyl and furanosyl rings is crucial, with an α-glycosidic linkage (as occurs in adenophostin A) optimally positioning the pseudo 1-phosphate.

As the most potent stable agonist of InsP₃receptors, adenophostin A is already widely used to examine the mechanisms underlying intracellular Ca²⁺regulation. Our analyses of the key determinants of the high-affinity interaction between adenophostin A and InsP₃receptors should facilitate development of further related analogs with properties tailored to meet specific biological requirements.