Semi-Rational Design of (N)-Methanocarba Nucleosides as Dual Acting A₁ and A₃Adenosine Receptor Agonists: Novel Prototypes for Cardioprotection

Abstract

Ring-constrained adenosine analogues have been designed to act as dualagonists at tissue-protective A₁ and A₃ adenosine receptors (ARs). 9-Ribosides transformed into the ring-constrained (N)-methanocarba-2-chloro-5′-uronamides consistently lost affinity at A₁/A_2AARs and gained at A₃AR. Among 9-riboside derivatives, only N⁶-cyclopentyl and 7-norbornyl moieties were extrapolated for mixed A₁/A₃ selectivity and rat/human A₃AR equipotency. Consequently, 2 was balanced in affinity and potency at A₁/A₃ARs as envisioned and dramatically protected in an intact heart model of global ischemia and reperfusion.

There are four subtypes of adenosine receptors (ARs): A₁, A_2A, A_2B, and A₃, and their selective agonists are under development as therapeutic agents.^1–3 Activation of one or more of the ARs and receptor overexpression have been shown to have a cytoprotective role in ischemic models.^3–8 Specifically, activation of either A₁ or A₃ARs in cardiac myocytes in several species has been shown to mimic the cardioprotective effect of ischemic preconditioning.^9–11 The co-activation of A₁ or A₃ARs in the cardiac myocyte has been shown to be protective to a greater degree than activation of either subtype alone. In vivo experiments have also demonstrated the cardioprotective effects of A₁ and A₃ARs in certain species.¹² Activation of A₁ and A₃ARs leads to activation of PLC and PLD (phospholipase C and D), respectively, in cultured cardiac myocytes.⁹ PLC and PLD converge on activation of protein kinase C (PKC), which mediates cardioprotection.^9,13 In the brain, activation of A₁ or chronic activation of A₃ARs has been shown to protect neurons against ischemia in a variety of models.^14,15 In a model of global ischemia in gerbils, the A₁ agonists CPA (N⁶-cyclopentyladenosine) and ADAC (N⁶-[4-[[[4-[[[(2-aminoethyl)amino]carbonyl]methyl]anilin]carbonyl]-methyl]phenyl]-adenosine) and the chronically administered A₃ agonist IB-MECA (1-[6-[[(3- iodophenyl)methyl]amino]-9H-purin-9-yl]-1-deoxy-N-methyl-β-D-ribofuran uronamide) were cytoprotective at very low doses.

The concurrent activation of A₁ and A₃ARs has been carried out either by co- administering selective agonists for each subtype, or using novel conjugates of functionalized congeners of A₁ and A₃ agonists.¹⁶ Some agonists of balanced potency have been reported,^17–19 however they are often partial agonists and of selectivities limited to a particular species. The careful design and covalent joining of these functionalized congeners provide binary conjugates that are balanced in their ability to activate A₁ and A₃ARs. However, due to the high molecular weights (in excess of 1000) and the presence of multiple hydrogen bond donors, these molecules do not satisfy the criteria proposed by Lipinski for prediction of oral bioactivity²⁰ and are of limited application in vivo.

We have taken a new approach, i.e. based on differential effects on ARs of the ring-constrained (N)-methanocarba ring system,²¹ to design dual acting A₁/A₃AR agonists of relatively low molecular weight. The new agonists (1 – 3, Scheme 1) were synthesized subsequent to careful SAR analysis of a large number of adenosine derivatives at ARs with respect to both binding affinity and relative efficacy.^22-24 We have incorporated specific molecular features that provide a balance in potency at A₁ and A₃ARs and that also maintain full efficacy at the A₃AR.^21,22 In the present study, several derivatives have been shown in binding and functional assays to be dual acting at the two receptor subtypes. Furthermore, a potent anti–ischemic cardioprotective effect in an intact mouse heart model of global ischemia and reperfusion injury was demonstrated.^6,13 This mammalian heart model has been shown to express both AR subtypes.

Scheme 1.

Reagents: a) RNH₂, MeOH, triethylamine; b) CH₃NH₂, MeOH; c) TFA, H₂O, MeOH, 70°C; d) 10% Pd/C/ H₂, MeOH.

The selection of N⁶-, C2, and 5′-uronamide substituents in the target compounds has been based largely on our recent findings relating to AR affinity, selectivity, and relative efficacy for specific adenine-9-riboside derivatives as well as a series of adenine- (N)-methanocarba-5′-alkyl uronamide derivatives.^21,22 At the A₃AR, in contrast to the A₁AR, relative efficacy is easily diminished by substitution of the N⁶ and C2 positions while preserving affinity. This reduction in efficacy is readily overcome by a flexible 5′-methyluronamide moiety.²² The desired analogues would be balanced in high affinity at both human and rat A₁ and A₃ARs and would display full agonism. According to published SAR findings (correlated in Figure 1) the substitution of adenine-9-ribosides to obtain the corresponding 2-chloro-(N)-methanocarba-5′-alkyl uronamides produced consistent effects on AR affinity. Upon undergoing these modifications for various hydrophobic N⁶-substituents there was roughly one order of magnitude loss of binding affinity at the A₁AR and slightly less loss at the A_2AAR. The effect of this transformation on binding affinity at the A₃AR resulted in either equal or greater affinity (up to 14-fold), due to the conformational preference of the A₃AR binding site.²¹ Thus, we examined a large published series of seventy-four adenosine derivatives,^23,24 mono-substituted at N⁶, for the ideal candidates predicted to have balanced binding affinity when adapted to the 2-Cl-(N)-methanocarba series. The following criteria were sought: 1) equipotency at rat and human A₃ARs; 2) roughly two orders of magnitude greater affinity at A₁ARs in comparison to A₃ARs; and 3) selectivity for A₁ and A₃ARs in comparison to both A_2A and A_2BARs. Few N⁶-substituents satisfied all criteria; for example, although the affinities at the rat and human A₁ARs were generally similar,²⁴ at the A₃ARs the species difference was as high as 1100-fold.²³ The most likely candidates identified were N⁶- cycloalkyl groups of the A₁-selective agonists 4 – 6.

Figure 1.

Correlation of K_i values at human A₁ (■), A_2A (◆), and A₃ (▲) receptors of adenosine derivatives in two structural series were compared. The two series compared are: mono-substituted adenosine (9-riboside) derivatives, and tri-subsituted 2-chloro-(N)-methanocarba derivatives. In each case, pairs of compounds in which both members have the same N⁶-sustitution (as indicated) are correlated. The five (N)-methanocarba analogues depicted in this graph contained N⁶-benzyl and phenylethyl-type groups, however the general effects on affinity at each of the three adenosine receptor subtypes were generalized to design new N⁶-cycloalkyl analogues having desired pharmacological properties. For the N⁶ substitutions shown, there was consistent loss of affinity at A₁ and A_2AARs and the effect at the A₃AR ranged from no change to a 14-fold gain of affinity. The (N)-methanocarba compounds shown here were all selective for the A₃ receptor, while the desired property in the new analogues was balanced affinity at A₁ and A₃ receptors.

The synthetic route to three adenosine agonists 1 – 3 that were designed for highaffinity at the A₁ and A₃ adenosine receptors and low affinity at the A₂ receptors is shown in Scheme 1. The synthesis of the target cyclopentyl derivative 2 and the 7-norbornyl derivative 3, was according to the general route presented for 5′-uronamido-(N)-methanocarba derivatives.²¹ The synthetic approach of Joshi et al.²⁵ has been followed to incorporate the 5′-uronamido-(N)-methanocarba ring system. The requisite 7-norbornylamine was prepared in three steps from 7-norbornyl bromide using a Curtius rearrangement. Accordingly, the 2,6-dichloro 5′-ester 7²¹ was treated first with a cycloalkylamine, which substituted selectively at the 6-position. Subsequent treatment with a large excess of methylamine converted the ester group to the corresponding amide. The final step was acidic deprotection of the isopropylidene protecting group at the 2′,3′-hydroxyl groups. Both 2 and 3 contain the 2-chloro and 5′-uronamido-(N)-methanocarba substituents. The 2-Cl group in 2 was hydrogenolyzed to give 1 in good yield.

Binding and functional assays were carried out at human A₁, A_2A and A₃ARsexpressed in CHO (Chinese hamster ovary) cells. The results confirmed that 1 and 2were highly selective for A₁ and A₃ARs, and that the affinities were nearly balanced. However, 3 was considerably more potent in binding to the A₃ than to the A₁AR. For all three derivatives, the K_i values at the human A_2AAR were at least several hundredfold greater than at the A₁ or the A₃AR. Compound 2 was equipotent in binding to human and rat A₁ARs with K_i values of 18.3 and 17.4 nM, respectively. Also, the affinity of compound 2 was similar at human and rat A₃ARs with K_i values of 3.7 and 5.8 nM, respectively.

In functional assays consisting of measuring inhibition of forskolin-stimulatedproduction of 3′,5′-cyclic-adenosine monophosphate (cAMP) in intact transfected CHO cells, single concentration determinations (Table 1) indicated that full A₃AR agonism was maintained in compounds 1 – 3. Concentration response curves indicated that compound 2 was a dual acting full agonist with nearly equivalent functional potencies at human A₁ (EC₅₀ = 8.2 nM) and A₃ (EC₅₀ = 2.8 nM) ARs.

Table 1.

Potency of adenosine derivatives at human A₁, A_2A, and A₃ARs and the rat A₃AR _a and maximal agonist effects at human A₃ARs expressed in CHO cells.^a

No.	N6-R′	C2-R	Ki (hA1AR) nMa	Ki (hA2AAR) nMa	Ki (hA3AR) nMa	% Activation (hA3AR)b at 10 μM	Ki (rA3AR)
1c	CP	H	34.1 ± 6.1	6420 ± 630	13.1 ± 5.1	93 ± 7	10.2 ± 2.1
2c	CP	Cl	18.3 ± 6.3g	3250 ± 300	3.7 ± 0.9	101 ± 10	5.8 ± 1.6
3c	NB	Cl	190 ± 37	>10,000	14.6 ± 3.2	92 ± 6	9.6 ± 2.7
4d	CP	H	0.45±0.04b	462	240±36	72±12	97±4
5d	CP	Cl	0.83e	2270e	38 ± 6	0	237±71f
6d	NB	H	0.48±0.01	>10,000	229±76	112±25	103±1

^aAll AR experiments were performed using adherent CHO cells stably transfected with cDNA encoding the human or rat ARs. Percent activation of the human A₃AR was determined at 10 μM. Binding at human A₁ and A_2AARs in this study was carried out as described in Methods using [³H]R-PIA or [³H]CGS 21680 as a radioligand. Values from the present study are expressed as mean ± s.e.m., n = 3–5.

^bPercent activity at 10 μM, relative to 10 μM Cl-IB-MECA (A₃).

^c1, MRS3706; 2, MRS3630; 3, MRS3638.

^dData from Gao et al., unless noted.²²,²³

^eData from Klotz et al.²⁸

^fData from van Galen et al.²⁹

^gK_i at rat A₁ AR is 17.4 ± 2.7 nM.

ND not determined. CP, cyclopentyl; NB, 7-norbornyl.

Compounds 1 – 3 were assayed for activation of the human A_2BAR stably expressed in CHO cells.¹⁸ Each adenosine derivative was tested at a concentration of 10 μM. As for the ribosides 4 – 6, the EC₅₀ values at the human A_2BAR of the (N)- methanocarba-5′-uronamide N⁶-substituted nucleosides 1 – 3 were all >10 μM. Compound 2 also showed negligible effect in stimulation of adenylate cyclase at the murine A_2A or A_2BARs endogenously expressed in PC12 (rat) and NIH/3T3 cells (mouse), respectively.

Since 2 was the most potent and still nearly matched in binding affinity and in function at the two AR subtypes known to be cardioprotective, this compound was chosen for further pharmacological studies in an intact mouse heart model of ischemia and reperfusion.^6,13 In this model, compound 2 at 30 nM exerted a potent anti-ischemic cardioprotective effect (Table 2). The mixed agonist was perfused until the induction of ischemia. The recovery of left ventricular developed pressure (LVPD), +dP/dt, -dP/dt and heart rate (HR) all improved significantly following treatment with the mixed agonist 2. The infarct size determined using computer morphometry²⁶ after staining with triphenyltetrazolium chloride (TTC) was significantly reduced in the group treated with 2 (Figure 2). The percent necrosis in the group treated with 2 was 15 ± 7% compared to 23 ± 8% in the vehicle-treated controls, n = 6. In the same model, the classical A₁AR agonist 5 at a higher concentration (100 nM) could also reduce myocardial infarct size. The percent necrosis following infusion of 5 was 15 ± 10%, n=15.

Table 2.

Recovery of left ventricular function in a mouse heart model of ischemia/reperfusion.^a

Parameter	Vehicleb	Compound 5c	Significance
LVDPd	8.5 ± 5.3	26.0 ± 4.8	t=7.1, P < 0.0001
+dP/dtd	6.3 ± 3.9	21.1 ± 4.9	t=7.4, P < 0.0001
-dP/dtd	8.2 ± 3.7	24.6 ± 7	t=7.22, P < 0.0001
HRd	37.1 ± 32.6	93.8 ± 19.3	t=4.1, P < 0.0005
% necrosis	23.4 ± 7.8	15.0 ± 6.9	t=2.32, Pe=0.029

^aValues were obtained after 35 global ischemia followed by reperfusion.

^bDMSO, n = 16.

^cA concentration of 30 nM 2 (initially dissolved in DMSO) was used, n = 6. During buffer perfusion of heart via the side port, the spontaneous heart rate dropped by 24 ± 8.2% (SEM) due to a decrease in the perfusion pressure. Perfusion of buffer containing 30 nM 2 was associated with a larger decrease in the heart rate of 57 ± 6.7%, likely because of a negative inotropic effect of 2.

^d% of baseline prior to ischemia/reperfusion.

^eTwo-tailed.

Figure 2.

Anti-ischemic infarct-reducing effects of adenosine receptor agonists. Murine hearts were excised and subjected to normothermic global ischemia and reperfusion with or without (A) adenosine receptor agonists as described in Methods. The adenosine receptor agonists were: (B) A₃ agonist Cl-IB-MECA, 30 nM; (C) A₁ agonist 5, 100 nM; (D) Mixed A₁/A₃ agonist 2, 30 nM. Agonists were infused for five min till the induction of ischemia. The heart was stained with TTC after 35 min of ischemia and 120 min of reperfusion and the infarcted areas were visualized as TTC-negative (pale, white). The infarct was quantified by morphometry and normalized to the whole heart as % necrosis. Data were representative of 6 (adenosine receptor agonist-treated) and 16 (DMSO/vehicle-treated) mice. A vehicle control not subjected to ischemia showed no pale or TTC-negative area.

Thus, we have designed novel cardioprotective agents based on mechanistic andstructural considerations. The adenosine N⁶-substituents, cyclopentyl and 7-norbornyl,were selected based on predictions made from the binding affinities of the corresponding adenosine derivatives,²² and from the consistent effects on AR affinity of replacing the 9-riboside moiety with a 5′-uronamido-(N)-methanocarba-pseudoribose moiety in combination with the 2-Cl substituent. The sum of these effects on affinity at each of the three AR subtypes was generalized to design new N⁶-cycloalkyl analogues having desired pharmacological properties. The results of Tchilibon et al.,²¹ for substituted N⁶-benzyl and N⁶-phenylethyl derivatives suggested that in each case, in comparison to the corresponding adenine-9-riboside, the affinity at the human A₁AR decreased by at least one order of magnitude while the affinity at the human A₃AR tended to increase by typically one order of magnitude. In the case of N⁶-cyclopentyl-and N⁶-(7-norbornyl)-adenine 9-ribosides, 4 and 6, respectively,²² the affinity of each was similar at rat and human A₃ARs, but the effects of such N⁶-cycloakyl substitution had not yet been probed in the 5′-uronamido-(N)-methanocarba series. The affinity of both 9-ribosides at the human A_2A and A_2BARs was weak, thus the corresponding 5′-uronamido-(N)-methanocarba derivatives were expected to be highly selective for A₁ and A₃ARs in comparison to A_2A and A_2BARs. We have confirmed the anticipated selectivity for the N⁶-cyclopentyl derivatives 1 and 2. Also, based on the 9-ribosides a large species difference at the A₃AR common among N⁶-substituted adenosine derivatives²⁴ was predicted to be absent in the new analogues. This prediction was confirmed in binding assays of all three newly-synthesized derivatives.

We have examined the mixed A₁/A₃ agonist 2 in an intact mouse heart model of ischemia and reperfusion injury,^6,13 in which either an A₁- or A₃-selective agonist acts as a potent cardioprotective agent. The initial findings validate the model for studying AR-dependent protection and illustrate the highly cardioprotective effect of 2. The role of cardiac A₃ARs is complex, with protective effects demonstrated in models of preconditioning, delayed cardioprotection,¹³ and ischemia-reperfusion.^6,8 The activation of the A₃AR in the rat coronary circulation has been proposed to mediate vasodilation.²⁷ Also, potential side effects of adenosine agonists, such as hypotension and sedation, must be considered.¹ Therefore, additional pharmacological examination of 2 and similar mixed agonists will be needed.

In conclusion, we have used a semi-rational approach based on SAR analysis to focus on a small number (3) of candidate structures predicted to display the desired pharmacodynamic properties. Thus, a series of (N)-methanocarba nucleosides previously characterized as selective A₃AR agonists has now been adapted to mixed AR selectivity desired for cytoprotection in a variety of tissue systems. Indeed, two of the three compounds synthesized reached this goal as judged by functional analyses and/or in vitro binding. Further chemical optimization of the structure to enhance the affinity and potency of mixed A₁/A₃ agonists is now feasible. One of the candidates was already shown to be highly cardioprotective in the mouse. These compounds may serve as prototypical examples for more detailed pharmacological studies leading to the development of novel dual acting cardioprotective AR agonists.

Supplementary Material

1si20051101_12. Supporting Information Available.

Experimental details for the synthesis and biological evaluation of compounds. This material is available free of charge via the Internet at http://pubs.acs.org.