Species differences and mechanism of action of A₃ adenosine receptor allosteric modulators

Abstract

Activity of the A₃ adenosine receptor (AR) allosteric modulators LUF6000 (2-cyclohexyl-N-(3,4-dichlorophenyl)-1H-imidazo [4,5-c]quinolin-4-amine) and LUF6096 (N-{2-[(3,4-dichlorophenyl)amino]quinolin-4-yl}cyclohexanecarbox-amide) was compared at four A₃AR species homologs used in preclinical drug development. In guanosine 5′-[γ-[³⁵S]thio]triphosphate ([³⁵S]GTPγS) binding assays with cell membranes isolated from human embryonic kidney cells stably expressing recombinant A₃ARs, both modulators substantially enhanced agonist efficacy at human, dog, and rabbit A₃ARs but provided only weak activity at mouse A₃ARs. For human, dog, and rabbit, both modulators increased the maximal efficacy of the A₃AR agonist 2-chloro-N ⁶-(3-iodobenzyl)adenosine-5′-N-methylcarboxamide as well as adenosine > 2-fold, while slightly reducing potency in human and dog. Based on results from N ⁶-(4-amino-3-[¹²⁵I]iodobenzyl)adenosine-5′-N-methylcarboxamide ([¹²⁵I]I-AB-MECA) binding assays, we hypothesize that potency reduction is explained by an allosterically induced slowing in orthosteric ligand binding kinetics that reduces the rate of formation of ligand-receptor complexes. Mutation of four amino acid residues of the human A₃AR to the murine sequence identified the extracellular loop 1 (EL1) region as being important in selectively controlling the allosteric actions of LUF6096 on [¹²⁵I]I-AB-MECA binding kinetics. Homology modeling suggested interaction between species-variable EL1 and agonist-contacting EL2. These results indicate that A₃AR allostery is species-dependent and provide mechanistic insights into this therapeutically promising class of agents.

Introduction

Selective A₃ adenosine receptor (AR) agonists are currently being developed as pharmacological therapy for a number of inflammatory diseases including rheumatoid arthritis, psoriasis, and cancer [1]. However, the full therapeutic benefit of these drugs could be limited by dose-limiting side effects [2]. Clinical development of A₃AR ligands is further challenged by substantial differences in A₃AR pharmacology among species [3–6]. This complication hinders assessment of therapeutic efficacy of A₃AR ligands in non-primate models of human disease.

Allosteric modulators of G protein-coupled receptors (GPCRs) are ligands that alter the affinity or intrinsic activity of orthosteric ligands by interacting with a distinctly different binding site on the receptor [7–10]. From a therapeutic perspective, allosteric modulators have the advantage of regulating activity of an endogenous ligand in a spatially and temporally specific manner [7–10]. Allosteric ligands typically mediate their effects by inducing conformational changes in the receptor that alter orthosteric agonist binding affinity and/or intrinsic activity thereby causing positive or negative allosteric modulation. These effects can vary depending on the nature of the orthosteric ligand (“probe dependence”) and have the capability to selectively modulate specific intracellular signaling pathways (“signaling bias”).

Several classes of compounds have been reported to be allosteric modulators for the human A₃AR, including the imidazoquinolinamines and the 2,4-disubstituted quinolines [11–18]. Derivatives of these two chemical classes function as positive allosteric modulators (PAMs) that increase the intrinsic activity of orthosteric agonists, while tending to reduce potency. 2-Cyclohexyl-N-(3,4-dichlorophenyl)-1H-imidazo [4,5-c]quinolin-4-amine (LUF6000) (Fig. 1) emerged from structure-activity relationship (SAR) studies of the imidazoquinolinamine template as having a pronounced effect to increase agonist efficacy with less propensity to decrease potency (Fig. 1; [16]) and to exhibit selectivity versus the other three AR subtypes [15, 16]. Modification of LUF6000 involving scission of the imidazole ring led to the development of the 2,4-disubstituted quinoline LUF6096 (Fig. 1), which similarly retained desirable allosteric effects and A₃AR selectivity [17]. Both of these compounds have begun to be tested for therapeutic efficacy in animal models of disease [19, 20]. These compounds were developed for activity at the human A₃AR, however, and it remains to be determined whether they exhibit similar modulating activity at the A₃AR from other species.

Fig. 1.

Chemical structures of A₃AR allosteric modulator compounds used in the present investigation

The goal of this investigation was to characterize the modulatory activity of LUF6000 and LUF6096 for the A₃AR from a panel of animal species commonly used in drug development research. Functional modulation was assessed using a guanosine 5′[γ-[³⁵S]thio]triphosphate ([³⁵S]GTPγS) binding assay that detects receptor-induced G protein activation with membranes prepared from human embryonic kidney cells (HEK 293) expressing recombinant A₃ARs from the various species. Orthosteric effects were explored in radioligand binding assays with the agonist radioligand N ⁶-(4-amino-3-[¹²⁵I]iodobenzyl)adenosine-5′-N-methylcarboxamide ([¹²⁵I]I-AB-MECA). Based on the results of initial studies, we focused on examining the potential role of the first extracellular loop (EL1) of the A₃AR in mediating activity of the modulators using a mutagenesis and molecular modeling approach.

Materials and methods

Materials

Cell culture reagents were purchased from Invitrogen (Carlsbad, CA). [³⁵S]GTPγS (specific activity 1250 Ci/mmol) and carrier-free Na¹²⁵I (2200 Ci/ml) were from PerkinElmer (Boston, MA). LUF6000 and LUF6096 were synthesized as reported [16, 17]. Plasmid purification kits were purchased from Qiagen (Valencia, Ca). Adenosine deaminase (ADA) was obtained from Roche Applied Sciences (Indianapolis, IN) and glass fiber filters (GF/B and GF/C) were obtained from Whatman (Sanford, ME). All remaining drugs and reagents were purchased from Sigma-Aldrich (St. Louis, MO).

Creation of stable HEK 293 cell lines expressing A₃ARs

Full-length dog, rabbit, and mouse A₃AR cDNAs subcloned into the mammalian expression vector pCDNA3.1 were obtained, as previously described [21–23]. The human A₃AR cDNA in pCDNA3.1 was purchased from the cDNA Resource Center (www.cdna.org). Site-directed mutations (G75Q/I76V/T77K/I78M) were introduced into the human A₃AR cDNA using the QuikChange Site-Directed Mutagenesis kit (Stratagene), according to the manufacturer’s instructions. The following primers were used to create the mutations: 5′-GCC TTT GGC CAT TGT TGT CAG CCT GCA GGT CAA AAT GCA CTT CTA CAG CTG CCT TTT TAT GA-3′ and 5′-TCA TAA AAA GGC AGC TGT AGA AGT GCA TTT TGA CCT GCA GGC TGA CAA CAA TGG CCA AAG GC-3′. Following amplification in Escherichia coli, plasmids were purified using Qiagen (Valencia, CA) purification kits and sequenced through the entire coding region to establish the presence of the desired mutations and the absence of unintended mutations potentially generated during the amplification steps. The plasmids were transfected into human HEK 293 cells using Lipofectamine 2000 reagent (Invitrogen) and selected with 2 mg/ml of G418. Cell lines derived from individual clones were maintained in cell culture media (Dulbecco’s modified Eagle’s medium with 10% fetal bovine serum and antibiotics) containing 0.6 mg/ml G418. The level of expression of the receptors in each of the cell lines was equivalent based on saturation radioligand binding analyses (not shown).

Membrane preparations

Cell membranes were prepared as described previously [16, 17, 20]. Briefly, HEK 293 cells stably expressing recombinant wild-type or mutant human A₃ARs were washed in phosphate-buffered saline followed by homogenization in hypotonic lysis buffer containing Tris-HCl buffer (50 mM, pH 7.4) containing 1 mM EDTA and 5 mM MgCl₂ and then centrifuged at 27,000×g for 30 min at 4 °C. Cell pellets were washed twice in the same buffer, after which the resultant pellets were re-suspended in 50 mM Tris-HCl buffer (pH 7.4) containing 10% sucrose and stored at − 80 °C. Because the modulators failed to increase [³⁵S]GTPγS binding in preliminary studies with the mouse A₃AR using crude P1 preparations, for some studies, plasma membranes were enriched by preparing P2 pellets. Cells were homogenized in lysis buffer containing 10% sucrose and centrifuged at 500×g for 10 min to remove nuclear and other cellular debris. The pellet was re-suspended in sucrose buffer and centrifuged again at 500×g. The pooled supernatants were diluted at least threefold, pelleted, and washed twice by centrifugation at 27,000×g for 30 min in 50 mM Tris-HCl buffer (pH 7.4). The resultant pellets were re-suspended and frozen in sucrose buffer.

[³⁵S]GTPγS binding assays

[³⁵S]GTPγS binding assays were conducted as described [15, 20]. In brief, 200 μl of buffer containing 50 mM Tris-HCl (pH 7.4), 1 mM EGTA, 10 mM MgCl₂, 1 μM GDP, 1 mM dithiothreitol, 100 mM NaCl, 0.2 nM [³⁵S]GTPγS, 300 nM 8-[4-[4-(4-chlorophenyl)piperazide-1-sulfonyl)phenyl]]-1-propylxanthine (PSB-603) to block A_2BARs expressed endogenously in HEK 293 cells; 0.005% 3-[(3-cholamidopropyl) dimethylammonio]-1-propanesulfonate (CHAPS); and 0.5% bovine serum albumin was added to polypropylene incubation tubes. Adenosine deaminase (3 U/ml) was also included in all assays except those examining the effects of adenosine. Then, reactions were started by addition of the membrane suspension (5 μg protein) to the tubes and carried out in quadruplicate for 2 h at room temperature. In studies with the modulators, the membranes were pre-incubated with the compounds for 30 min prior to initiating the assay. The reactions were stopped by rapid filtration through Whatman GF/B filters presoaked in 50 mM Tris-HCl buffer (pH 7.4) containing 5 mM MgCl₂ and 0.02% CHAPS. The filters were washed three times with 3 ml of wash buffer (10 mM Tris-HCl and 1 mM MgCl₂, pH 7.4, 4 °C), and radioactivity trapped in the filters was measured by liquid scintillation counting. Non-specific binding of [³⁵S]GTPγS was measured in the presence of 10 μM unlabeled GTPγS.

Binding assays with [¹²⁵I]I-AB-MECA

Cell membranes (50 μg) were incubated in 100 μl binding buffer (50 mM Tris-HCl [pH 7.4], 1 mM EDTA, 10 mM MgCl₂, and 3 units/ml adenosine deaminase) containing ~ 0.3 nM [¹²⁵I]I-AB-MECA and indicated concentrations of the A₃AR allosteric modulator compounds. The reactions were incubated at room temperature for the times indicated after which bound and free radioligands were separated by rapid filtration through GF/C glass fiber filters. Radioactivity trapped in the filters was measured using a gamma counter. For dissociation studies, [¹²⁵I]I-AB-MECA (~ 0.3 nM) was incubated with HEK 293 cell membranes (50 μg) expressing ARs for 2 h at room temperature in binding buffer (100 μl), after which the assay was initiated by the addition of 100 μM adenosine-5′-N-ethylcarboxamide (NECA, a nonselective agonist) along with the enhancer compounds or equivalent vehicle. The time-course of dissociation of specific [¹²⁵I]I-AB-MECA binding was measured by rapid filtration at the indicated time intervals. For association/equilibration binding assays, membranes (50 μg) were incubated with [¹²⁵I]I-AB-MECA (~ 0.3 nM) and the modulator compounds for the indicated times. For all assays, non-specific binding was determined by incubation in the presence of 100 μM NECA. [¹²⁵I]I-AB-MECA was prepared by radioiodination of AB-MECA using the chloramine-T method [21, 24, 25].

Molecular modeling

A₃ARs homology models

Homology models of the human A₃AR, mouse A₃AR, and mEL1-hA₃AR were built based on the human A_2AAR crystal structure of highest available resolution (PDB ID: 4EIY; [26]) and by using the alignment and homology modeling tools implemented in the MOE suite [27]. In particular, A₃AR homology models were built using the automated Homology Modeling protocol. The AMBER99 force field was used for protein modeling and the Protonate 3D methodology was used for protonation state assignment [28]. The final models were refined through energy minimization until a root mean square (RMS) gradient of 0.1 kcal/mol/Å was achieved. The stereochemical quality of the models was checked using several tools (Ramachandran plot; backbone bond lengths, angles, and dihedral plots; clash contacts report; rotamer strain energy report) implemented in the MOE suite.

Loop refinement

For the final A₃AR models, EL1 and EL2 were subjected to refinement by means of the Prime [29] package portion of the Schrödinger Suite [30] to better explore their conformational space. The two loops were cooperatively refined in pairs using an implicit membrane whose orientation and thickness were derived from the spatial arrangement of the human A_2AAR structure stored in the Orientations of Proteins in Membranes (OPM) database [31]. For the human A₃AR model, EL1 was refined from Ser73 to His79 and EL2 from Gly148 to Cys166. Corresponding residues were refined for the other A₃AR models.

Molecular docking of Cl-IB-MECA at the A₃AR

The Cl-IB-MECA structure was built using the builder tool implemented in the MOE suite [27] and subjected to energy minimization using the MMFF94x force field, until a RMS gradient of 0.05 kcal/mol Å was achieved. A homology model of the human A₃AR was built based on an agonist-bound human A_2AAR crystal structure (PDB ID: 3QAK; [32]) following the same procedure described above. Molecular docking of the ligand at the human A₃AR model was performed by means of the Glide [33] package portion of the Schrödinger Suite [30]. The docking site was defined using key residues in the binding pocket of the human A₃AR model, namely Phe (EL2), Asn (6.55), Trp (6.48), and His (7.43), and a 20 Å × 20 Å × 20 Å box was centered on those residues. Docking of the ligand was performed in the rigid binding site using the extra precision (XP) protocol. The top scoring docking conformations were subjected to visual inspection and comparison with the nucleoside crystallographic pose at the human A_2AAR [32] to select the final binding conformation. Due to lack of knowledge of the binding pocket, we did not attempt docking analyses with the allosteric modulators at this time.

Data analysis

EC₅₀ and E _max values were calculated by fitting the data to: E = (E _max × x) / (EC₅₀ + x), in which x is the concentration of the test compound. In the kinetic binding assays, dissociation rate constants were determined by fitting the specific binding data to a two-phase exponential decay model: Y = Y ₀ × f ₁ × e ^{(−k1 × t)} + Y ₀ × f ₂ × e ^{(−k2 × t)}, where Y _o denotes specific binding at time 0, f ₁ and f ₂ denote the fraction of receptors in the two phases, k ₁ and k ₂ denote the dissociation rate constants, and t denotes the elapsed time. For association assays, association binding rates were determined by fitting the specific binding data to a one-phase exponential association model: Y = Y _max × (1 − e ^{(−k × t)}, where Y _max denotes maximal binding and k denotes the observed association rate constant. The association half-time was calculated as 0.693/k. All values were statistically compared using an unpaired Student’s t test or a one-way ANOVA followed by Student’s t test with the Bonferroni correction, as appropriate. All values are presented as the mean ± SEM. A p value < 0.05 was considered to be statistically significant.

Results

[³⁵S]GTPγS binding

In previous functional studies with the human A₃AR [15, 17], both LUF6000 and LUF6096 have been shown to increase the maximal efficacy of the A₃AR agonist 2-chloro-N ⁶-(3-iodobenzyl)-adenosine-5′-N-methylcarboxamide (Cl-IB-MECA). As shown in Fig. 2, this was confirmed in the present study using the [³⁵S]GTPγS binding assay. In the presence of 10 μM of either LUF6000 or LUF6096, the E _max of Cl-IB-MECA was increased ~ 2–3-fold, while its EC₅₀ was increased 5–6-fold. Neither of the compounds directly stimulated [³⁵S]GTPγS binding, supporting an allosteric mechanism. Thus, with the human A₃AR, both compounds function as PAMs with respect to efficacy enhancement but also slightly reduce agonist potency. Similar findings were observed in assays with the dog A₃AR (Fig. 2), whereby the compounds increased the efficacy of Cl-IB-MECA ~ 2–3-fold while decreasing its potency (5–6-fold). In assays with the rabbit A₃AR, however, the efficacy of Cl-IB-MECA was increased without a reduction in potency, and in assays with the mouse A₃AR, no enhancement was detected (Fig. 2). A nearly identical profile was observed among the species with LUF6096 (10 μM) on efficacy and potency in assays using adenosine as the orthosteric agonist (Table 1). To compare the potency of the enhancers among the different species, concentration-response curves with LUF6096 were conducted in the presence of a maximal concentration (10 μM) of Cl-IB-MECA. As shown in Fig. 3, the calculated EC₅₀ values for LUF6096 for the responding species (human, dog, rabbit) differed only within a factor of 3.

Fig. 2.

Modulatory effect of 10 μM LUF6000 or LUF6096 on Cl-IB-MECA-induced [³⁵S]GTPγS binding in assays with HEK 293 cell membranes expressing human (a), dog (b), rabbit (c), or mouse (d) A₃ARs. Incubations were started by addition of the membrane suspension (5 μg protein) to reactions containing ~ 0.1 nM [³⁵S]GTPγS and Cl-IB-MECA, as described in “Materials and methods,” and were carried out in quadruplicate for 2 h at room temperature. The membrane solutions were pre-incubated with the modulators for 30 min before beginning the assays. EC₅₀ and E _max values are reported in the tables. Data are the mean ± SEM of four to eight experiments performed in quadruplicate. Asterisk: P < 0.05 versus control by one-way ANOVA followed by post hoc analysis by a Student’s t test for unpaired data with the Bonferroni correction

Table 1.

Effect of LUF6096 on adenosine-induced [³⁵S]GTPγS binding

	EC50 (nM)	E max (%)
Human
Vehicle	104 ± 60	16 ± 2
LUF6096	405 ± 124*	25 ± 2*
Dog
Vehicle	3971 ± 197	23 ± 2
LUF6096	10,260 ± 1766*	64 ± 5*
Rabbit
Vehicle	280 ± 24	20 ± 4
LUF6096	233 ± 109	60 ± 6*
Mouse
Vehicle	3643 ± 408	21 ± 4
LUF6096	3169 ± 117	15 ± 2

EC₅₀ and E _MAX values (mean ± SEM; n = 5–8 performed in quadruplicate) calculated from concentration-response curves, as described in “Materials and methods”

*P < 0.05 vs. vehicle

Fig. 3.

Modulatory effect of increasing concentrations of LUF6096 on [³⁵S]GTPγS binding in response to a maximal concentration (10 μM) of Cl-IB-MECA in assays using HEK 293 cell membranes expressing human, canine, rabbit, or mouse A₃ARs. Assays were conducted as described in Fig. 2 legend. The membrane preparations were pre-incubated with LUF6096 for 30 min before beginning the assays. EC₅₀ values and E _max values calculated from the data for each species of A₃AR are reported in the table. Data are the mean ± SEM of four to eight experiments performed in quadruplicate

Due to lack of an effect of the modulators on mouse A₃ARs, we repeated the [³⁵S]GTPγS binding assays using enriched P2 membrane preparations (see “Materials and methods”). As shown in Fig. 4, in these experiments, we observed a small efficacy enhancing effect of both modulator compounds. In the presence of 10 μM of either LUF6000 or LUF6096, the E _max of Cl-IB-MECA was increased 20–30%, with no change in the EC₅₀. Similarly, LUF6096 produced a detectable efficacy enhancing effect (46%) in assays using adenosine as the orthosteric agonist (Fig. 4). The EC₅₀ of LUF6096 to produce efficacy enhancement, determined from concentration-response curves in the presence of 10 μM Cl-IB-MECA, was calculated to be 1.33 ± 0.83 μM (n = 5), suggesting reduced (~ 25-fold) potency at mouse A₃ARs compared to the other species. Overall, these observations indicate that the modulatory effects of LUF6000 and LUF6096 vary among the species tested with respect to efficacy enhancement, potency, and the propensity to reduce agonist potency.

Fig. 4.

Modulatory effect of 10 μM LUF6000 or LUF6096 on a Cl-IB-MECA-induced or b adenosine-induced [³⁵S]GTPγS binding in assays with enriched P2 HEK 293 cell membranes expressing mouse A₃ARs. Assays were conducted as described in Fig. 2 legend. EC₅₀ and E _max values are reported in the tables. Data are the mean ± SEM of three to seven experiments performed in quadruplicate. Asterisk: P < 0.05 versus control by one-way ANOVA followed by post hoc analysis by a Student’s t test for unpaired data with the Bonferroni correction

[¹²⁵I]I-AB-MECA binding

We compared among the species the effects of LUF6000 and LUF6096 on binding of the orthosteric agonist [¹²⁵I]I-AB-MECA. We began with a dissociation binding assay, because in previous studies with the human A₃AR both modulators produce prominent slowing of the rate of [¹²⁵I]I-AB-MECA dissociation [17]. For these studies, membranes from HEK 293 cells expressing A₃ARs from the various species were incubated with [¹²⁵I]I-AB-MECA for 2 h to reach equilibrium, after which displacement was initiated by the addition of NECA (100 μM) either alone or in combination with either modulator (10 μM). When the A₃AR is expressed in HEK 293 cells, [¹²⁵I]I-AB-MECA binds to two affinity states, which relates to G protein-coupling status [21, 34]. Thus, the dissociation of [¹²⁵I]I-AB-MECA is biphasic, whereby the slow component corresponds to the high-affinity, G protein-coupled state of the A₃AR and the fast component corresponds with binding to the low-affinity, uncoupled state of the receptor [21, 34].

The results of the [¹²⁵I]I-AB-MECA dissociation kinetic experiments are shown in Fig. 5, and the calculated dissociation rate constants along with the fraction of receptors for the slow phase (high-affinity state) are reported in Table 2. At a concentration of 10 μM, both LUF6000 and LUF6096 reduced the rate of dissociation of [¹²⁵I]I-AB-MECA from human and dog A₃ARs. The modulators did not influence the rate of dissociation of [¹²⁵I]I-AB-MECA from rabbit or mouse A₃ARs. When an effect on the dissociation rate was observed, the modulators produced a significant decrease in the slow [¹²⁵I]I-AB-MECA dissociation rate. The modulators also tended to decrease the fast [¹²⁵I]I-AB-MECA dissociation rate, although this did not reach the level of significance, likely due to difficulty in obtaining accurate measurements of this rapid phase. Thus, the allosteric effect of the modulators to alter agonist dissociation is also species-dependent.

Fig. 5.

Effect of LUF6000 or LUF6096 on dissociation of [¹²⁵I]I-AB-MECA binding to HEK 293 cell membranes expressing human (a), dog (b), rabbit (c), or mouse (d) A₃ARs. Membranes (50 μg) were incubated with [¹²⁵I]I-AB-MECA (~ 0.3 nM) for 2 h at room temperature after which dissociation was initiated by adding 100 μM NECA mixed with either vehicle or modulator compound. The fraction of specific binding at various times after the addition of NECA is shown. The data were fitted to a two-phase exponential decay model. Data are the mean ± SEM of five to eight experiments performed in triplicate. See Table 1 that reports dissociation rate constants

Table 2.

Influence of LUF6000 and LUF6096 on the rate of dissociation of [¹²⁵I]I-AB-MECA binding from A₃ARs

	k 1 (10−2 min−1)	k 2 (10−2 min−1)	Fraction of high affinity state
Human
Vehicle	1.21 ± 0.05	8.05 ± 2.22	0.56 ± 0.08
LUF6000 (10 μM)	0.49 ± 0.08*	3.57 ± 1.09 (P = 0.11)	0.56 ± 0.14
LUF6096 (10 μM)	0.75 ± 0.08*	3.68 ± 1.13 (P = 0.10)	0.65 ± 0.08
Dog
Vehicle	1.81 ± 0.22	40.1 ± 8.89	0.60 ± 0.04
LUF6000 (10 μM)	1.02 ± 0.02*	21.3 ± 4.86 (P = 0.14)	0.63 ± 0.05
LUF6096 (10 μM)	1.16 ± 0.06*	28.9 ± 4.92 (P = 0.27)	0.69 ± 0.05
Rabbit
Vehicle	0.60 ± 0.32	6.74 ± 1.50	0.74 ± 0.19
LUF6000 (10 μM)	0.62 ± 0.42	10.9 ± 2.39	0.52 ± 0.18
LUF6096 (10 μM)	0.60 ± 0.28	19.2 ± 6.93	0.82 ± 0.12
Mouse
Vehicle	1.00 ± 0.06	22.7 ± 3.04	0.69 ± 0.12
LUF6000 (10 μM)	0.95 ± 0.04	20.3 ± 9.74	0.64 ± 0.17
LUF6096 (10 μM)	0.93 ± 0.01	25.5 ± 11.0	0.62 ± 0.03

Data are dissociation rate constants (mean ± SEM; n = 3–6) calculated for the slow (K ₁) and fast (K ₂) components, as described in “Materials and methods”

*P < 0.05 vs. vehicle

We next performed a preliminary association binding assay with the human A₃AR. Membranes from HEK 293 cells were incubated with [¹²⁵I]I-AB-MECA (0.3 nM) for up to 18 h in the presence or absence of 10 μM LUF6096. As shown in Fig. 6, inclusion of LUF6096 markedly delayed the rate at which [¹²⁵I]I-AB-MECA binding reached equilibrium. In control assays, binding reached equilibrium in ~ 2 h, whereas in the presence of LUF6096, binding reached equilibrium at a time well beyond 5 h. After fitting the data to a single-site association binding model, the association half-time of [¹²⁵I]I-AB-MECA was calculated to be increased by LUF6096 more than fivefold from ~ 31 min in the control assays to ~ 185 min in the presence of LUF6096. Maximal [¹²⁵I]I-AB-MECA binding in the presence of LUF6096 was not reduced.

Fig. 6.

Effect of LUF6096 on association kinetic binding of [¹²⁵I]I-AB-MECA to HEK 293 cell membranes expressing the human A₃AR. Membranes (50 μg) were incubated with [¹²⁵I]I-AB-MECA (~ 0.3 nM) for the times indicated at room temperature with either vehicle or 10 μM LUF6096. Non-specific binding was determined in the presence of 100 μM NECA. The fraction of specific binding obtained at the 18-h time-point in the presence of vehicle is shown. The binding data were fitted to a one-phase exponential association model. Data are the mean ± SEM of four to six experiments performed in triplicate

Based on these findings, we utilized a hybrid association/equilibrium binding experiment to efficiently compare activity of the modulators on [¹²⁵I]I-AB-MECA binding among the A₃ARs from the different species. For these assays, transfected HEK 293 cells expressing A₃ARs were incubated with 0.3 nM [¹²⁵I]I-AB-MECA in the presence of increasing concentrations of either LUF6000 or LUF6096 and the extent of specific binding was compared at 3 h with binding obtained at 18 h. Assessment at the 18-h time-point will detect an effect of the modulators on [¹²⁵I]I-AB-MECA binding under equilibrium conditions, whereas assessment at the 3-h time-point will detect whether the modulators might have an impact on the rate at which [¹²⁵I]I-AB-MECA binding achieves equilibrium. As shown in Fig. 7, both of the modulators produced a concentration-dependent reduction in [¹²⁵I]I-AB-MECA binding to the A₃AR from all of the species tested at the 3-h time-point, while having no effect on binding at the 18-h time-point. The magnitude of this effect (i.e., the extent of reduction in [¹²⁵I]I-AB-MECA binding at 3 h) among the species was human = dog > rabbit = mouse. These results demonstrate a general effect of the modulators to allosterically impact [¹²⁵I]I-AB-MECA binding kinetics without affecting the extent of binding at equilibrium.

Fig. 7.

Effect of LUF6000 and LUF6096 on binding of the orthosteric radioligand [¹²⁵I]I-AB-MECA in assays with HEK 293 cell membranes expressing human, dog, rabbit, or mouse A₃ARs. Membranes (50 μg) were incubated with [¹²⁵I]I-AB-MECA (~ 0.3 nM) for either 3 h (a and c) or 18 h (b and d) at room temperature in the presence of increasing concentrations of LUF6000 (a and b) or LUF6096 (c and d). Non-specific binding was determined in the presence of 100 μM NECA. The fraction of specific binding obtained in the absence of the modulator compounds is shown. Data are the mean ± SEM of four to six experiments performed in triplicate

Participation of the first extracellular loop

The amino acid sequences of those species whereby the modulators produced prominent efficacy enhancement (human, dog, rabbit) were compared to those that were minimally responsive (rat, mouse) to potentially identify regions of the A₃AR that might participate in mediating the actions of the modulator compounds. We included the rat A₃AR as a poor responding species in this analysis, since we found that both LUF6000 and LUF6096 produced no detectable efficacy enhancing activity in [³⁵S]GTPγS binding assays (P1 membranes) with the rat A₃AR (not shown). As depicted in Fig. 8, this analysis revealed that the first extracellular loop (EL1) region, composed of five amino acids of which four were conserved in human, dog, and rabbit (75G-I-T-I-H/Q/G79), was markedly different in mouse and rat and notably contained charged residues (E or K). Based on this analysis, assays were conducted with a stably transfected HEK 293 cell line expressing a mutated form of the human A₃AR, named the mEL1-hA₃AR, in which the EL1 region was changed to the mouse sequence (76Q-V-K-M-H80). A clonal cell line was selected that expressed the mEL1-hA₃AR at levels similar to that of the wild-type human A₃AR-expressing HEK 293 cell line. As determined in saturation equilibrium radioligand binding assays, mutation of EL1 of the human A₃AR to the mouse sequence did not influence the binding affinity of [¹²⁵I]I-AB-MECA to either the high- (G protein-coupled) or low- (uncoupled) affinity states and also did not influence the fraction of receptors in the high-affinity state (Table 3).

Fig. 8.

“Snake” diagram of the human A₃AR sequence. Amino acid residues that are conserved in species found to be responsive to both modulator compounds (human, dog, rabbit) but are not conserved in weakly responding species (rat, mouse) are highlighted in red. According to this criterion for comparison, marked variability was identified in the first extracellular loop region. Detailed amino acid sequence comparison of this region is shown in the lower left-hand section of the illustration. The dendrogram in the lower right-hand corner reveals that A₃AR sequences from responsive and less responsive species cluster together. This diagram was prepared using the GPCRdb database (http://www.gpcrdb.org).

Table 3.

Saturation binding data with [¹²⁵I]I-AB-MECA for the wild-type human A₃AR and the mEL1-hA₃AR

	Wild-type	mEL1-hA3AR
B max Hi (fmol/mg)	418.1 ± 44.2	471 ± 59.6
B max Lo (fmol/mg)	447.9 ± 56.1	424 ± 57.4
K d Hi (nM)	0.3 ± 0.1	0.4 ± 0.1
K d Lo (nM)	1.7 ± 0.7	1.2 ± 0.3

Data are the ± SEM (n = 6). Parameters calculated with the mEL1-hA₃AR were not significantly different from the wild-type human A₃AR by Student’s t test

In [³⁵S]GTPγS binding assays, LUF6096 increased the efficacy of Cl-IB-MECA (10 μM) with HEK 293 cell membranes expressing the mEL1-hA₃AR to a similar extent as compared to assays with the wild-type human A₃AR (Fig. 9). However, the EC₅₀ of LUF6096 was reduced 17-fold, determined from [³⁵S]GTPγS binding assays in which concentration-response curves with LUF6096 were conducted in the presence of 10 μM Cl-IB-MECA. As shown in Fig. 9, LUF6096 did not reduce the extent of [¹²⁵I]I-AB-MECA binding to the mEL1-hA₃AR at the 18-h time-point in the hybrid association/equilibrium binding assay described earlier. At the 3-h time-point, however, LUF6096 produced a greater reduction in [¹²⁵I]I-AB-MECA binding to the mEL1-hA₃AR compared to the wild-type human A₃AR (Fig. 9). The potency (EC₅₀) of LUF6096 to reduce binding to the mEL1-hA at the 3-h time-point was comparable to the wild-type receptor (wild-type = 1.91 ± 0.35 μM; mEL1-hA₃AR = 2.54 ± 0.5 μM; Fig. 9). LUF6096 also produced a greater effect to slow [¹²⁵I]I-AB-MECA dissociation from the mEL1-hA₃AR (Fig. 9). These results indicate that changing the EL1 region of the human A₃AR to the mouse sequence results in a greater effect of LUF6096 to (i) reduce the potency of Cl-IB-MECA and (ii) slow [¹²⁵I]I-AB-MECA binding kinetics.

Fig. 9.

Modulatory effect of LUF6096 in assays with HEK 293 cell membranes expressing the mEL1-hA₃AR mutant. a Effect of LUF6096 (10 μM) on [³⁵S]GTPγS binding while stimulating with a maximal concentration of Cl-IB-MECA (10 μM). b Effect of increasing concentrations of LUF6096 on [³⁵S]GTPγS binding in response to Cl-IB-MECA (10 μM). Data in this panel are presented as a percentage of maximal [³⁵S]GTPγS binding. c Effect of increasing concentrations of LUF6096 on binding of [¹²⁵I]I-AB-MECA in the hybrid association/equilibrium binding assay involving comparison of specific binding at 3 and 18 h. d Effect of LUF6096 on [¹²⁵I]I-AB-MECA dissociation kinetics. All of the assays were conducted as described in “Materials and methods” and in previous figure legends. Data are the mean ± SEM of four to eight experiments performed in triplicate ([¹²⁵I]I-AB-MECA binding assays) or quadruplicate ([³⁵S]GTPγS binding assays). Asterisk: P < 0.05 versus the wild-type human A₃AR by an unpaired Student’s t test

Molecular modeling

In an attempt to analyze the structural details of the EL1 region, we built homology models of the human A₃AR, mouse A₃AR, and mEL1-hA₃AR and performed a conformational search of the extracellular loop areas. The models were based on the reported crystal structure of the human A_2AAR [26]. As part of the modeling, EL1 and EL2 were cooperatively refined due to their close proximity and the potential for interaction. Loop modeling is still a challenging task due to marked flexibility, making it difficult to obtain unequivocal information. Figure 10 shows the top scoring conformations obtained for the three different receptors.

Fig. 10.

Comparison of human A₃AR (magenta), mouse A₃AR (orange), and mEL1-hA₃AR (yellow) homology models, after loop refinement. Ribbon region corresponding to the mutated residues in EL1 is colored green. Docking pose of Cl-IB-MECA is shown in cyan

The following information regarding EL1 and EL2 was obtained from the analyses. The general structure of EL1 was predicted to be similar between the human and mouse receptors because of its short length (five residues) despite marked dissimilarity in the amino acid sequence in this region. In some cases, it formed a β-sheet with part of EL2, as observed in several human A_2AAR crystal structures [26, 32]. The EL2 residues predicted from the model to make contact with the EL1 sequence are different between the species: 162T-F-L-S-C-Q167 for the human A₃AR and 163S-T-L-L-C-H168 for the mouse A₃AR. Therefore, replacement of the EL1 region of the human A₃AR with the mouse sequence affects the interaction pattern between the two loops in comparison with the native receptor, which could result in modification of EL2 structure and function of the mEL1-hA₃AR. Notably, EL1 does not make direct contact with Cl-IB-MECA in the binding cavity while EL2 does, as shown from the Cl-IB-MECA docking pose reported in Fig. 10. In fact, some EL2 residues have a role in shaping the binding cavity and stabilizing the ligand. For example, a phenylalanine that is conserved among ARs (F168 at the human A₃AR) forms a π-π stacking interaction with the adenine core of the ligand in agreement with the binding mode of AR agonists in the A_2AAR crystallographic structures [32]. EL2 appears to be very flexible given the wide range of possible conformations it assumed during the search. At the human A_2AAR, EL2 is involved in three disulfide bridges that constrain its structure, while in the A₃AR, only one disulfide is present. Therefore, the conformation of EL2 of the A₃AR could be quite different from that of the A_2AAR.

Discussion

This study compared the modulatory activity of LUF6000 and LUF6096 at the A₃AR from four different mammalian species (human, dog, rabbit, and mouse) expressed in HEK 293 cells. These ligands, which are currently under investigation in preclinical animal models [19, 20], emerged from SAR studies of the imidazoquinolinamine and 2,4-disubstituted quinoline classes of allosteric modulators of the human A₃AR that enhance the intrinsic activity of orthosteric agonists [11–18]. We demonstrate that the pharmacology of the modulators is unique for each of the species investigated. Most notably, both modulators showed weak efficacy enhancing activity on mouse A₃ARs.

The observation that both LUF6000 and LUF6096 exert low efficacy-enhancing activity versus rodent A₃ARs was not unexpected. Previous studies have documented that orthosteric pharmacology of rodent A₃ARs differs substantially from other mammalian species, explained by a greater degree of sequence divergence that influences the orthosteric ligand binding pocket and receptor activation mechanisms [3, 4, 6]. The A₃AR is known to be capable of stimulating G protein-independent translocation of β-arrestin2 upon ligand binding that contributes to rapid desensitization of the receptor and may also lead to activation of additional intracellular signaling pathways [14]. Since LUF6000 has been reported to modulate β-arrestin translocation mediated by the human A₃AR [14], it would be interesting to compare activity of the modulator compounds on this signaling response in other species.

Other notable species differences include a variable effect of the modulators to decrease agonist potency in the [³⁵S]GTPγS binding assay and to influence [¹²⁵I]I-AB-MECA binding properties. Interestingly, both of the modulators decreased agonist (Cl-IB-MECA and adenosine) potency in the [³⁵S]GTPγS binding assays in the same species (human and dog) that they substantially slowed [¹²⁵I]I-AB-MECA binding kinetics. These findings raise the possibility that potency reduction by the modulators might be explained by a slowed rate of receptor occupancy. The effect of the modulators to alter [¹²⁵I]I-AB-MECA binding kinetics appears to be complex, whereby the on and off rates are influenced in a manner that results in slowed ligand-receptor complex formation but no change in binding at equilibrium. In general, the properties of the two different modulators were comparable. Both produced a similar degree of efficacy enhancement as well as effects on [¹²⁵I]I-AB-MECA binding, and both displayed a similar activity profile among the different species of receptors. While the two different modulators are structurally dissimilar, the data suggest that they both likely interact with the same allosteric site on the receptor.

The mutagenesis studies suggest that the EL1 region of the A₃AR participates in mediating the allosteric effect of LUF6096 to alter orthosteric ligand binding kinetics. It is possible that the EL1 region provides structural support that allows access of agonists to the orthosteric ligand binding site. Based on findings from the molecular modeling analyses of the extracellular loop regions, this may occur through interactions with EL2. These analyses predicted that EL2, which is much larger, can assume different conformations that are influenced by EL1, as shown by the formation of a β-sheet between the two loops in some cases. It is interesting that the EL1 region of the M₄ muscarinic receptor has been implicated to serve a similar function. Nawaratne and colleagues [35] identified that a single mutation of an isoleucine^93(2.65) of EL1 of the M₄ muscarinic receptor to a threonine residue selectively reduced the modulatory activity of the allosteric ligand LY2033298 to increase orthosteric agonist binding while leaving efficacy enhancing activity unchanged. Another mutagenesis study of the human A₃AR by our group determined that specific amino acid residues in transmembrane regions 1, (asparagine^30(1.50)), 2 (aspartic acid^58(2.50)), 3 (aspartic acid^107(3.49)), 5 (phenylalanine^182(5.43)), and 7 (asparagine^274(7.45)) abolished the effect of the imidazoquinolinamine A₃AR allosteric modulator DU124183 to decrease the rate of dissociation of [¹²⁵I]I-AB-MECA, implicating these regions as additional sites required for allosteric interactions [12]. All of these amino acid residues are conserved among the human, dog, rabbit, mouse, and rat sequences. In this earlier study, the effect of the mutations on functional enhancing activity was not addressed.

In conclusion, species variation in the actions of GPCR allosteric ligands has been underappreciated. Unlike the orthosteric binding site, allosteric domains of GPCRs are not subject to evolutionary pressures to accommodate endogenous ligands and can therefore exhibit greater variation between receptor subtypes and between species. The possibility for species variability lies not only with the ability of modulator compounds to bind to an allosteric site but also with the magnitude of allosteric interactions that influence signaling responses. This study provides evidence that allosteric pharmacology of the A₃AR displays prominent species variability. This information is important for future development of this class of compounds for therapeutic use [8, 20] and should be considered when validating A₃AR modulation by PAMs in animal models. It will be useful to focus efforts in future investigations to develop pan-species A₃AR PAMs with minimal negative allosteric properties.

Abbreviations

AR

Adenosine receptor

CHAPS

3-[(3-Cholamidopropyl)dimethylammonio]-1-propanesulfonate

Cl-IB-MECA

2-Chloro-N ⁶-(3-iodobenzyl)adenosine-5′-N-methylcarboxamide

GPCR

G protein-coupled receptor

[³⁵S]GTPγS

Guanosine 5′-[γ-[³⁵S]thio]triphosphate

HEK 293

Human embryonic kidney 293

[¹²⁵I]I-AB-MECA

N ⁶-(4-amino-3-[¹²⁵I]iodobenzyl)adenosine-5′-N-methylcarboxamide

LUF6000

2-Cyclohexyl-N-(3,4-dichlorophenyl)-1H-imidazo [4,5-c]quinolin-4-amine

LUF6096

N-{2-[(3,4-dichlorophenyl)amino]quinolin-4-yl}cyclohexanecarbox-amide

NECA

Adenosine-5′-N-ethylcarboxamide

PAM

Positive allosteric modulator

Compliance with ethical standards

Conflicts of interest

Lili Dua declares that she has no conflict of interest.

Zhan-Guo Gao declares that he has no conflict of interest.

Silvia Paoletta declares that she has no conflict of interest.

Tina C. Wan declares that she has no conflict of interest.

Elizabeth T. Gizewski declares that she has no conflict of interest.

Samantha Barbour declares that she has no conflict of interest.

Jacobus P. D. van Veldhoven declares that he has no conflict of interest.

Adriaan P. IJzerman declares that he has no conflict of interest.

Kenneth A. Jacobson declares that he has no conflict of interest.

John A. Auchampach declares that he has no conflict of interest.

Ethical approval

This article does not contain any studies with human participants or animals performed by any of the authors.