Truncated (N)-Methanocarba Nucleosides as A₁ Adenosine Receptor Agonists and Partial Agonists: Overcoming Lack of a Recognition Element

Abstract

A₁ adenosine receptor (AR) agonists are neuroprotective, cardioprotective, and anxiolytic. (N)-Methanocarba adenine nucleosides designed to bind to human A₁AR were truncated to eliminate 5′-CH₂OH. This modification previously converted A₃AR agonists into antagonists, but the comparable effect at A₁AR is unknown. In comparison to ribosides, affinity at the A₁AR was less well preserved than that at the A₃AR, although a few derivatives were moderately A₁AR selective, notably full agonist 21 (N⁶-dicyclopropylmethyl, K_i 47.9 nM). Thus, at the A₁AR, recognition elements for nucleoside binding depend more on 5′ region interactions, and in their absence, A₃AR selectivity predominates. Based on the recently reported agonist-bound AR structure, this difference between subtypes likely correlates with an essential His residue in transmembrane domain 6 of A₁ but not A₃AR. The derivatives ranged from partial to full agonists in A₁AR-mediated adenylate cyclase inhibition. Truncated derivatives have more druglike physical properties than other A₁AR agonists; this approach is appealing for preclinical development.

Adenosine modulates many physiological processes by activating one or more of four subtypes of G protein-coupled receptors (GPCRs).¹ The medicinal chemistry of adenosine receptors (ARs) is now well advanced in comparison to the cases of many other GPCRs, with the existence of numerous selective agonists and antagonists, allosteric modulators, prodrugs, radioligands for imaging, fluorescent probes, and macromolecular ligand conjugates.² A₁AR ligands have been considered clinically for a variety of conditions: agonists (diabetes, pain), partial agonists (arrhythmias), and antagonists (heart failure, renal protection).³

The adenosine structure has been extensively modified, on both the nucleobase and ribose, for pharmacological optimization to selectively activate ARs. One means of achieving AR subtype selectivity has been the replacement of ribose with a sterically constrained methanocarba ([3.1.0]-bicyclohexane) ring system. This bicyclic system adopts a North (N)-envelope conformation in MRS3558 (1) (Chart 1) to maintain a receptor-preferred conformation and enhanced affinity at the A₃AR.⁶ The (N)-methanocarba modification is also tolerated at A₁AR, but without affinity enhancement. Consequently, the cardioprotective MRS3630 (2) containing the A₁AR-favoring N⁶-cyclopentyl substituent is a mixed A₁/A₃AR agonist.⁷

Chart 1. Structures of Representative Ribosides and Ring-Constrained Methanocarba Nucleoside Derivatives That Have Been Characterized as Agonist, Partial Agonists, and Antagonists at the A₃AR (K_i, nM, in Binding to the hA₃AR in Italics)⁶⁻¹⁰.

Another useful modification is truncation of the ribose at the 4′ position, that is, removing the 5′-CH₂OH moiety, while retaining all other features of the ribose-like moiety and its stereochemistry. Thus, 4′-thionucleoside antagonist LJ-1251 (3) and 4′-oxo antagonist 4 preserve affinity and selectivity at the A₃AR, while removing the ability to induce the required conformational change for receptor activation.^8,9 Truncation of (N)-methanocarba nucleosides originally was reported to convert A₃AR agonists into selective antagonists in a guanine nucleotide binding assay.¹⁰ Subsequently, partial agonism in a functional assay of adenosine 3′,5′-cyclic phosphate (cyclic AMP) at the G_i-coupled A₃AR was shown for MRS5127 (5) and congeners.¹¹

In contrast to the reduced A₃AR efficacy of 5′-truncated nucleosides, at the G_s-coupled A_2AAR, full agonism is retained, as shown recently for the 4′-thio series.¹² The effects of truncation on A₁AR efficacy are unknown. Our major objective was to probe the effects of truncated (N)-methanocarba nucleosides at the human (h) A₁AR, both pharmacologically and with insight into the structural basis for receptor recognition. Therefore, we have incorporated N⁶ substituents that are expected to promote A₁AR affinity. For example, nucleoside 6, previously characterized at the A₃AR,¹² contains N⁶-cyclopentyl, which generally produces A₁AR selectivity in the riboside series. The G_i-coupled A₁AR is more homologous in primary sequence and effector coupling to the G_i-coupled A₃AR than to the G_s-coupled A_2AAR. If it more closely resembles A₃AR in ligand binding and activation mechanism, the truncated analogues such as 6 will be A₁AR antagonists or partial agonists. However, if its activation more closely resembles the A_2AAR, then these truncated derivatives will be full A₁AR agonists. With the recent structural elucidation of an A_2AAR active state,¹⁵ it is feasible to relate these findings to specific binding site interactions.

Results

With the objective of increasing A₁AR affinity, we explored N⁶ substitution of 4′-truncated (N)-methanocarba nucleoside derivatives, which were shown previously to be selective A₃AR antagonists/partial agonists.¹² Therefore, the series included N⁶-cycloalkyl (8–12) and N⁶-bicycloalkyl (13,14), and N⁶-acyclic alkyl and N⁶-cyclopropylalkyl (15–21) substitutions associated previously with A₁AR selectivity of ribosides (Table 1).¹³ Finally, certain substituted N⁶-benzyladenosine derivatives, such as 2-fluorobenzyl, were reported to have enhanced affinity at the A₁AR.²³ Therefore, a variety of fluorinated and nonfluorinated N⁶-benzyl derivatives (22–28) were prepared.

Table 1. Potency of a Series of Truncated (N)-Methanocarba Adenosine Derivatives at Three Subtypes of hARs and Relative Efficacy at hA₁AR.

^aUsing CHO or HEK293 (A_2A only) cells stably expressing a hAR (Supporting Information); affinity was expressed as K_i value (n = 3–5) or percent inhibition of radioligand binding at 10 μM.

^bValues from refs (6, 7, and 12). 5 and 6 were prepared previously.¹²

^c20 is a diastereomeric mixture.

^dMaximal efficacy (at 10 μM) in an A₁AR functional assay, determined by inhibition of forskolin-stimulated cyclic AMP production in AR-transfected CHO cells, expressed as percent inhibition (mean ± standard error, n = 3–5) in comparison to effect (100%) of full agonist CPA 30 at 10 μM. The value for NECA was 100 ± 15.

The synthetic route to the truncated derivatives involves nucleophilic displacement by the appropriate amine of a 6-chloroadenine group in a 2′,3′-isopropylidene-protected precursor 29 (Scheme S1 of the Supporting Information). All of the analogues contain a 2-chloro substitution of the adenine ring, which has been shown to increase affinity at either or both A₁AR and A₃AR. 2-Chloro substitution of the A₁AR agonist N⁶-cyclopentyladenosine (CPA, 30) also was shown to reduce agonist efficacy at the A₃ but not A₁AR.¹⁴

Several previously reported 2-chloro (N)-methanocarba derivatives (1, 2, 5, and 6) were used for comparison in the biological assays (Table 1). Binding assays at three hAR subtypes were carried out using standard radioligands and membrane preparations from Chinese hamster ovary (CHO) cells (A₁ and A₃) or human embryonic kidney (HEK) 293 cells (A_2A) stably expressing a hAR subtype (Table 1).¹² Since activity within the class of (N)-methanocarba nucleosides was previously noted to be very weak or absent at the hA_2BAR,¹⁶ we did not include this receptor in the screening protocol.

Generally, the 5′-truncated (N)-methanocarba-adenosine derivatives, in comparison to the corresponding 9-ribosides, maintained affinity at the A₃AR more effectively than at the A₁AR. Nevertheless, affinities of <100 nM at the A₁AR were achieved for certain N⁶-alkyl and cycloalkyl members of this series. Among the most potently binding 2-chloro (N)-methanocarba analogues (K_i, in nM) at the hA₁AR (with K_i of the corresponding N⁶ derivatized 9-ribosides at rat A₁AR in parentheses¹³) are the following: cyclobutyl 9, 51.6 (0.7); isopropyl 17, 72.2 (1.9); cyclopropylmethyl 19, 68.4 (0.8); and di(cyclopropyl)methyl 21, 47.9 (0.8). The A₁AR affinity of the N⁶-ethyl analogue 15 was more substantially reduced, by 141-fold, with K_i values of 930 and 6.6 nM for the truncated and 9-riboside analogues, respectively. Other N⁶ analogues that were more significantly reduced in their hA₁AR affinity in comparison to the corresponding riboside at rat A₁AR (K_i, nM) are as follows: cyclohexyl 10, 131-fold (0.9); endo-norbornyl 13, 113-fold (0.34); exo-norbornyl 14, 231-fold (0.7); and 2-fluorobenzyl 22, 800-fold (6). Because of this reduced affinity, the selectivity for the A₁AR was diminished overall. Only a few derivatives tended toward A₁AR selectivity in comparison to the A₃AR: di(cyclopropyl)-methyl 21, 10-fold; endo-norbornyl 13, 4-fold; and large cycloalkyl derivatives 10–12, 2–3-fold. The degree of selectivity vs A_2AAR was higher: 21, 74-fold; 13, 30-fold. Many of the other derivatives were equipotent in binding to A₁ and A₃ARs, and the substituted N⁶-benzyladenosine derivatives (22–28) were generally selective for the A₃AR, similar to previous observations with truncated N⁶-benzyl derivatives.^10,12

Functional data determined at a single concentration (10 μM) in an assay of adenylate cyclase (A₁AR-induced inhibition of cyclic AMP) are reported in Table 1. The potent and selective agonist CPA was used as the standard full agonist, and the nonselective AR agonist 5′-N-ethylcarboxamidoadenosine (NECA) was also a full agonist in this assay. Most of the analogues were partial agonists of the A₁AR. However, concentration–response curve for dicyclopropylmethyl analogue 21 in A₁AR-mediated inhibition of cyclic AMP indicated full agonism compared to NECA (Figure S1 of the Supporting Information). The EC₅₀ values for 21 and NECA were 40.7 ± 19.7 and 10.2 ± 3.3 nM, respectively, in close agreement with the A₁AR binding affinities.

A correlation plot of the hA₃AR affinity of truncated 2-chloro-(N)-methanocarba analogues in the present study (x-axis) vs the hA₃AR affinity of 2-unsubstituted adenine-7-riboside analogues (y-axis)¹³demonstrated a parallel in these parameters (Figure 1B). The affinity at the A₃AR subtype was relatively well maintained (correlation coefficient of 0.55). The enhanced A₃AR affinity of a 5′-truncated ribonucleoside derivative was first noted for the 2-chloro-N⁶-(3-iodobenzyl) derivative¹⁶ and has been validated consistently in SAR studies that also showed lowered efficacy in this series.⁸⁻¹⁰ However, a similar plot of A₁AR affinity comparing ribonucleosides and truncated ring-constrained nucleosides illustrated the trend of consistently lower affinity with truncation (Figure 1A), but without correlation of K_i values (correlation coefficient of 0.33). Therefore, in comparison to the case of ribosides, the affinity of the truncated ring-constrained analogues at the A₁AR was less well preserved than that at the A₃AR.

Figure 1.

(A) Correlation of the affinity of truncated 2-chloro-(N)-methanocarba analogues (x-axis, at the hA₁AR) and 2-unsubstituted adenine-7-riboside analogues (y-axis, at the rat A₁AR).¹³ (B) Correlation of the hA₃AR affinity of truncated 2-chloro-(N)-methanocarba analogues (x-axis) and 2-unsubstituted adenine-7-riboside analogues (y-axis).¹³

One derivative that tended toward A₁AR-selectivity, the full agonist 21, contained a N⁶-dicyclopropylmethyl group. Curiously, truncated analogues of N⁶-cyclopentyl and N⁶-benzyl derivatives were notably reduced in A₁AR affinity, even in the case of a 2-fluoro analogue 22, a modification previously found to enhance A₁AR selectivity in the riboside series.¹³ Compound 22 was 115-fold selective for the A₃AR in comparison to the A₁AR. A 2-fluoroethyl ether 23 was a potent and selective ligand at the A₃AR (K_i 10.3 nM), which might be sufficient affinity for use in radiofluorination for positron emission tomography.

There is a general phenomenon of reduced AR efficacy by removing H-bonding groups from the ribose moiety, especially at the A₃AR but also at the A₁AR.^5,16 A precedent for reduced A₁AR efficacy in nucleoside derivatives is the series of xanthine 7-ribosides. The 5′-uronamide analogue 5′-N-methyl 1,3-dibutylxanthine 7-β-d-ribofuronamide (DBXRM) was a full agonist at both A₁ and A₃ARs. However, removal of the 3′-hydroxyl group produced antagonism at the rat adipocytes A₁AR and partial agonism at the rat A₃AR. In examining the question of whether 5′-truncation in the present series of rigid (N)-methanocarba analogues provides agonists or antagonists at the A₁AR, we found that the answer lies in the middle; that is, the agonist efficacy is partially maintained for many of the derivatives, as shown in Table 1, but the pattern is not uniform.

To examine docking of truncated (N)-methanocarba nucleosides, homology models of the hA₁AR and hA₃AR were produced from the agonist-bound A_2AAR X-ray structure (3QAK).¹⁵ It was noted that key interactions of the nucleoside ligand with His251 in transmembrane domain (TM) 6, conserved between A₁AR and A_2AAR, and with Thr91 in TM3 are necessarily missing in the truncated derivatives. In the A_2AAR structure, this essential His residue, which is absent in the A₃AR, and Thr91 both H-bonded to a 5′-CO-NH-alkyl moiety of a cocrystallized NECA-like agonist.¹⁵ We predict that the difference in this key His residue is related to the dramatically altered binding of the truncated ring-constrained nucleosides at A₁ and A₃ARs. Thus, in general, A₃AR recognition would depend more on interactions of groups other than the nucleoside 5′ substituent; A₁AR recognition depends more on 5′ region interactions, and in their absence, A₃AR selectivity predominates. This is consistent with the truncated derivatives maintaining A₃AR, but not A₁AR, affinity.

Furthermore, the hydrophilic region of the receptor associated with ribose binding is critical for activation. This step likely involves essential residues of TM3, TM6, and TM7 throughout the AR family, as in the A_2AAR.¹⁵ Multiple H-bonding groups in this region promote agonism at the A₃AR,¹⁴ such as the interactions of the 5′-CH₂OH of the ribose moiety or the corresponding 5′-CO-NH-alkyl in NECA-like analogues. Thus, loss of the 5′ substituent is expected to reduce AR efficacy, which it clearly does at the A₃AR. However, the effect at the A₁AR is highly variable, with relative maximal efficacies of the truncated (N)-methanocarba analogues ranging from low (20–30% in compounds 14, 16, and 18) to high (80–100% in compounds 8, 19, and 21). The structural basis for the full agonism of 21, in contrast to closely related compounds such as the N⁶-(3-pentyl) derivative 18, which has greatly reduced efficacy at this subtype, must arise from conformational effects of the N⁶ group.

The docking poses of nonselective agonist NECA, full agonist 21, and low efficacy partial agonist 16 in the putative binding site of the A₁AR were compared (Figure 2 and Figure S2 of the Supporting Information). The H-bond interactions between the 5′-CO-NH-ethyl group of NECA and the two residues, His251(6.52) and Thr91(3.36), could lock the ribose moiety in an active conformation, with the 2′,3′-hydroxyl groups of the ribose ring correctly directed toward Thr277(7.42) and His278(7.43) in order to pull TM7 toward TM3 to efficiently activate the receptor.¹⁵ In the absence of an interaction with His251(6.52) and/or Thr91(3.36) due to the lack of the 5′ substituent, the orientation of the rigid methanocarba moiety could be less effective in forming the H-bond interactions with Thr277(7.42) and His278(7.43) needed to attract TM7. In the case of A₃AR, the activation process could be slightly different from that for the A₁AR, due to some differences in the key residues of the binding pocket. Position 3.32 in A₃AR consists of a nonconserved bulky and hydrophobic leucine residue, while a smaller valine is present in the other AR subtypes. Residue 3.32 was close to the ribose ring of the docked NECA in both A₁AR and A₃AR binding sites. The longer side chain of Leu90(3.32) in A₃AR, in comparison to Val87(3.32) of A₁AR, could contribute to a stronger hydrophobic interaction with the methanocarba moiety of the 5′-truncated agonists. At the same time, the different nature of the substituents at the N⁶ position of the adenine ring could influence the selectivity and the efficacy of agonists at both the A₁AR and the A₃AR. The N⁶ groups in the docking poses of the studied agonists lay in a region of the pocket formed by residues in the upper part of TM6 and TM7. The A₁AR and A₃AR differ in the nature of the residues lining this pocket. The variation of the affinity and potency of this series of 5′ truncated methanocarba analogues at the A₁AR indicates that the N⁶ substituent can greatly affect these parameters, in some cases compensating for the lack of H-bonding interactions in the ribose 4′ region. From the docking pose of full agonist 21 in the A₁AR model (Figure 2B), the favorable interactions of the N⁶-dicyclopropylmethyl substituent and the residues in the upper part of TM6 and TM7 (such as Thr257, Leu253, and Thr270) could maintain the adenine and methanocarba moieties in an efficacious active conformation with strong H-bond and hydrophobic interactions with residues in TM6, TM3, TM7, and EL2. A smaller and more flexible N⁶ group, such as the 3-fluoropropyl substituent of 16, in addition to the lack of interactions with His251(6.52) and Thr91(3.36), could negatively affect the ability to fully activate the receptor through conformational affects originating at the upper pocket of TM6 and TM7 (Figure S3 of the Supporting Information).

Figure 2.

Docking poses of NECA (A) and truncated nucleoside 21 (B) in the binding site of the hA₁AR homology model based on the X-ray structure of an agonist-bound hA_2AAR, indicating the main H-bond interactions.

The physicochemical properties of nucleosides that act as AR agonists often lead to limited in vivo bioavailability. The C-log p of compound 21 is 1.41, with the optimal for small molecular pharmaceutical substances being typically 2–3.¹⁷ The comparable parameter for the related A₁AR-selective riboside and prototypical agonist CPA is 0.14, which is less desirable. Also, the polar surface area (PSA) values for 21 and CPA are calculated to be 92.8 and 122 Ų, respectively. Most druglike small molecules have a PSA smaller than 120 Ų. Compound 21 has fewer hydroxyl groups than CPA, which would favor bioavailability. The molecular weight of 21 of 376 is comfortably within the preferred range. Therefore, by several criteria, the full agonist of the A₁AR 21 is more druglike than CPA.

The bioavailability of peripherally administered 21 in the brain, i.e. whether its altered physicochemical properties may facilitate its passage across the blood brain barrier, is undetermined. Many of the efforts to develop A₁ agonists have attempted to limit central nervous system (CNS) penetration to avoid central mediated side effects, but other envisioned applications of A₁ agonists depend on brain entry. In previous studies of the activity of A₁AR agonists in the CNS, only a small fraction of a peripherally administered agent crossed the blood brain barrier.⁴ However, a similar attempt to alter the biodistribution by removing the 2′-hydroxyl group of CPA did not enhance brain uptake.⁴

In conclusion, truncated (N)-methanocarba adenine nucleosides display highly variable degrees of binding affinity and activation at the hA₁AR. Based on the recently reported agonist-bound AR X-ray structure, this difference between subtypes likely correlates with an essential His residue in TM6 of A₁ but not A₃AR. By overcoming the lack of an important recognition element for receptor binding, i.e., the 5′ substituent, a N⁶-dicyclopropylmethyl derivative 21 was empirically identified as a moderately A₁AR selective, full agonist. This is counterintuitive given that many 5′-modified analogues are partial agonists at the A₁AR.¹⁸ A₁AR agonists hold interest therapeutically for their cardio- and neuroprotective, antiarrhythmic, antiseizure, antilipolytic, antiglaucoma, and anxiolytic actions. It is conceivable that the expanded range of physical properties in the present series of truncated derivatives would offer pharmacokinetic advantages. Therefore, this approach is appealing for preclinical development. This hypothesis will have to be evaluated in the future in vivo studies.

Glossary

Abbreviations

AR

adenosine receptor

cyclic AMP

adenosine 3′,5′-cyclic phosphate

CPA

N⁶-cyclopentyladenosine

CHO

Chinese hamster ovary

GPCR

G protein-coupled receptor

HEK

human embryonic kidney

MRS3558

(1′S,2′R,3′S,4′R,5′S)-4′-{2-chloro-6-[(3-chlorophenylmethyl)amino]purin-9-yl}-1-(methylaminocarbonyl)bicyclo[3.1.0]hexane-2,3-diol

MRS3630

(1′S,2′R,3′S,4′R,5′S)-4-(2-chloro-6-(cyclopentylamino)-9H-purin-9-yl)-2,3-dihydroxy-N-methylbicyclo[3.1.0]hexane-1-carboxamide

MRS5127

(1′S,2′R,3′S,4′R,5′S)-4′-[2-chloro-6-(3-iodobenzylamino)-purine]-2′,3′-O-dihydroxybicyclo[3.1.0]hexane

NECA

5′-N-ethylcarboxamidoadenosine

TM

transmembrane domain

Supporting Information Available

Synthetic procedures for compounds 7–28, their characterization and bioassays, and modeling procedures and results. This material is available free of charge via the Internet at http://pubs.acs.org.

The X-ray structure of NECA bound to a thermostabilized A_2AAR was recently solved (RCSB ID: 2YDV; Lebon et al. 2011, Nature, doi:10.1038/nature10136) and is very similar to our docked pose of NECA.

Funding Statement

National Institutes of Health, United States