Characterization of Dual-Acting A₃ Adenosine Receptor Positive Allosteric Modulators That Preferentially Enhance Adenosine-Induced Gα_i3 and Gα_oA Isoprotein Activation

Abstract

The A₃ adenosine receptor (A₃AR) is a promising therapeutic target for inflammatory diseases, cancer, and chronic neuropathic pain, with agonists already in advanced clinical trials. Here we report an in-depth comparison of the pharmacological properties and structure–activity relationships of existing and expanded compound libraries of 2-substituted 1H-imidazo[4,5-c]quinolin-4-amine and 4-amino-substituted quinoline derivatives that function as A₃AR positive allosteric modulators (PAMs). We also show that our lead compound from each series enhances adenosine-induced A₃AR signaling preferentially toward activation of Gα_i3 and Gα_oA isoproteins, which are coexpressed with the A₃AR in immune cells and spinal cord neurons. Finally, utilizing an extracellular/intracellular chimeric A₃AR approach composed of sequences from a responding (human) and a nonresponding (mouse) species, we provide evidence in support of the idea that the imidazoquinolin-4-amine class of PAMs variably interacts dually with the orthosteric ligand binding site as well as with a separate allosteric site located within the inner/intracellular regions of the receptor. This study has advanced both structural and pharmacological understanding of these two classes of A₃AR PAMs, which includes leads for future pharmaceutical development.

Introduction

The biological effects of adenosine, a purine nucleoside produced locally during ischemia, hypoxia, and inflammation, are mediated by a family of four G protein-coupled receptors (GPCRs) designated A₁, A_2A, A_2B, and A₃.¹ The A₃ adenosine receptor (A₃AR) is a G_i protein-coupled receptor that is abundantly expressed in granulocytic cells (neutrophils, eosinophils, basophils, and mast calls) and microglia, where it mediates the effects of adenosine to control immune cell chemotaxis and cellular activation.²⁻⁸ Several A₃AR agonists have been developed that are being tested in phase 2 and 3 clinical trials for the treatment of psoriasis, rheumatoid arthritis, nonalcoholic steatohepatitis (NASH), and cancer.⁹⁻¹³ Due to combined anti-inflammatory and central nervous system effects, A₃AR agonists are also being investigated for the treatment of neuropathic pain.¹⁴⁻¹⁸

To circumvent complications associated with the use of direct GPCR agonists, we have pursued the development of positive allosteric modulators (PAMs) for the A₃AR. PAMs are ligands that increase the potency and/or efficacy of endogenous ligands by binding to a site (allosteric site) that is topographically distinct from the binding site for the endogenous agonist (orthosteric site).¹⁹⁻²¹ Mechanistically, binding of a PAM enhances potency by increasing the binding affinity of the orthosteric ligand, which is achieved by changing its association and/or dissociation binding rates. For efficacy modulation, the conformational change in the receptor produced by PAM binding increases signaling capacity, thereby enhancing coupling to downstream effectors. PAMs theoretically offer several pharmacological and therapeutic advantages over traditional orthosteric agonists including improved receptor subtype selectivity and spatiotemporal specificity.¹⁹⁻²¹

To discover novel A₃AR PAMs, we conducted a small compound library screen and identified derivatives of the 1H-imidazo[4,5-c]quinoline-4-amine template (Figure 1) that slow the rate of orthosteric agonist dissociation via an allosteric mechanism,²² which predictably would increase binding affinity and therefore functional potency. In functional assays, it was noted that some members of this class, such as DU124183 (Figure 1), also increased the signaling efficacy of orthosteric agonists.²² However, rather than producing potency enhancement as predicted, they greatly decreased agonist potency.²² Given that certain imidazoquinoline-4-amines have been described in earlier studies to function as competitive AR antagonists,²³ we speculated that these compounds have dual effects to compete for orthosteric ligand binding while acting simultaneously at a separate overlapping allosteric site. Structural modification of DU124183 and further structure-activity-relationship (SAR) exploration at the 2- and 4-amino positions of the imidazoquinolin-4-amine template, which aimed to separate orthosteric and allosteric effects, led to several novel compounds with A₃AR modulatory activity including the current lead PAM from this class, LUF6000 (Figure 1).²⁴⁻²⁶ Additional modification, whereby the imidazole ring of LUF6000 was opened and replaced with an amide, produced the 2,4-disubstituted quinoline LUF6096 (Figure 1).²⁷ Like DU124183, LUF6000 and LUF6096 increased the maximal efficacy of adenosine as well as the A₃AR agonist 2-chloro-N⁶-(3-iodobenzyl)adenosine-5′-N-methylcarboxamide (Cl-IB-MECA) on the order of 2–3-fold in functional assays, but with less propensity to decrease their potency.^24,26,27 Another key modification identified previously and maintained in the current series was the addition of a (3,4-dichlorophenyl)amino substitution that improved receptor subtype selectivity.²⁵⁻²⁷ Although LUF6000 and LUF6096 exhibit PAM activity at the human (h) A₃AR (as well as rabbit and dog), they are strikingly less potent at rodent (mouse and rat) receptors, indicating species variability with this class of modulators.²⁴

Figure 1.

Structures of previously reported A₃AR PAMs including the reference imidazoquinolin-4-amine LUF6000 (4) and the reference quinoline LUF6096 (17).

Presently, an A₃AR structure remains to be solved experimentally, which limits our ability to utilize structure-based techniques to inform drug-design decisions. The purpose of this investigation was to re-examine our current and expanded compound library in more detail, with an emphasis on allosteric ligand potency and influences on orthosteric ligand binding potency and efficacy. Direct comparisons were made between the imidazoquinolin-4-amine and 2,4-disubstituted quinoline scaffolds containing equivalent substitutions. In addition, we expanded the assessment of our two current lead compounds (LUF6000 and LUF6096) to include a more comprehensive analysis of effects on specific G protein isoform coupling and β-arrestin-2 recruitment. Lastly, we utilized human/mouse A₃AR chimeric receptors to investigate the molecular mechanism of action of our lead PAMs. These studies provide in depth characterization of two promising chemical classes of A₃AR allosteric modulators and reveal that our lead PAMs preferentially enhance adenosine-induced G-protein signaling by binding to an allosteric site within intracellular regions of the A₃AR.

Methods

Chemical Synthesis

Please refer to the synthetic methods in the Supporting Information (for 2,4-disubstituted quinoline derivatives) and in the submitted paper by Fallot et al.²⁸ (1H-imidazo[4,5-c]quinoline-4-amine derivatives).

Creation of Stable HEK293 Cell Lines Expressing Wild-Type and Human/Mouse Chimeric A₃ARs

The full-length mouse (as previously described²⁹), human (purchased from the cDNA Resource Center), and human/mouse chimeric (custom synthesized by TOP Gene Technologies, St. Laurent, Quebec) A₃AR cDNAs were subcloned into pcDNA3.1. The constructs were transfected into HEK293 cells (American Type Culture Collection, Manassas, VA) using Lipofectamine 2000 reagent (Invitrogen, Waltham, MA) and selected with 2 mg/mL of G418 in cell culture media (DMEM with 10% fetal bovine serum, 100 units/mL penicillin, and 100 μg/mL streptomycin). Cell lines derived from individual clones were maintained in cell culture media containing 0.6 mg/mL G418. The level of receptor expression in each of the cell lines was equivalent (∼1 500 fmol/mg) based on saturation radioligand binding analyses.

Membrane Preparations

Transfected HEK293 cells were washed with PBS followed by homogenization in buffer A containing 10 mM Na⁺-HEPES, 10 mM EDTA, 1 mM benzamidine (pH 7.4) and centrifugation at 27 000 × g for 30 min at 4 °C. Cell pellets were subsequently rehomogenized in HE buffer consisting of 10 mM Na⁺-HEPES, 1 mM EDTA, 1 mM benzamidine (pH 7.4) and recentrifuged. The supernatant was discarded, and cell pellets were resuspended in HE buffer containing 10% sucrose and stored at −20 °C.

[³⁵S]GTPγS Binding Assays

Cell membranes (5 μg of protein) isolated from transfected HEK293 cells were pretreated with modulators for 1 h in 100 μL of GTPγS binding buffer (50 mM Tris HCl [pH 7.4], 1 mM EGTA, 10 mM MgCl₂, 100 mM NaCl, 0.004% CHAPS, and 0.5% BSA) in a 96-well large-volume polypropylene assay plate. In all assays, ZM-241385 and PSB-603 (each at a final concentration of 300 nM) were included to block A_2BARs expressed endogenously in HEK293 cells.³⁰ Adenosine deaminase (ADA) (1 unit/mL) was also included to degrade any endogenous adenosine that might have been produced during the assay, except when adenosine was used as the orthosteric agonist. The reactions were initiated by the addition of ∼0.2 nM [³⁵S]GTPγS and agonist in 100 μL of GTPγS binding buffer and allowed to incubate for 2 h at room temperature. At the end of the 2 h incubation period, the membranes were harvested by rapid filtration through Whatman GF/B filters that had been presoaked for 2 h in GTPγS binding buffer containing 0.02% CHAPS using a 96-well cell harvester (Brandel, Gaithersburg, MD). Radioactivity trapped in the filters was measured by liquid scintillation counting. Nonspecific binding of [³⁵S]GTPγS was determined in the presence of 10 μM unlabeled GTPγS.

Binding Assays with [¹²⁵I]I-AB-MECA

Cell membranes (50 μg) isolated from transfected HEK293 cells as described earlier were incubated in 100 μL of binding buffer (50 mM Tris-HCl [pH 7.4], 10 mM MgCl₂, 1 mM EDTA, and 1 unit/mL ADA) 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 HEK293 cell membranes (50 μg) expressing ARs for 3 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 potent nonselective AR agonist) along with the enhancer compounds or equivalent vehicle. Specific [¹²⁵I]I-AB-MECA binding was measured by rapid filtration at the indicated time intervals. For equilibration binding assays, membranes (50 μg) were incubated with [¹²⁵I]I-AB-MECA (∼0.3 nM) and the modulator compounds for the indicated times before filtration. For all assays, nonspecific 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 and purified by high-performance liquid chromatography (HPLC).³¹

Bioluminescence Resonance Energy Transfer Assays

To measure A₃AR-mediated β-arrestin-2 recruitment by BRET¹, HEK293T cells were cotransfected (1:15) with two constructs (pcDNA3.1) encoding the human A₃AR with C-terminally fused Renilla Luciferase (Rluc8) and N-terminally fused Venus-tagged β-arrestin-2 in DMEM supplemented with 10% fetal bovine serum, 100 units/mL penicillin, and 100 μg/mL streptomycin using TransIT-293 transfection reagent (Mirus Bio, Madison, WI). To measure A₃AR G protein coupling via βγ subunit dissociation, as assessed by BRET² (TRUPATH open-source biosensor platform³²), HEK293T cells were cotransfected (1:1:1:1) in DMEM supplemented with 10% fetal bovine serum, 100 units/mL penicillin, and 100 μg/mL streptomycin with constructs encoding the indicated Rluc8-fused human Gα isoprotein, human G_β1, C-terminally fused GFP² fused to human G_γ2, and either the human or mouse A₃AR using TransIT-293. After 24 h, transfected cells were seeded in poly-l-lysine-coated, white-walled, clear-bottom, 96-well cell culture plates at a density of 30 000 cells per well in DMEM with 1% dialyzed fetal bovine serum, 100 units/mL penicillin, and 100 μg/mL streptomycin and 25 mM HEPES (pH 7.4). The next day, the cell culture medium was aspirated, and cells were washed once with assay buffer (Hank’s buffered salt solution with Mg²⁺, Ca²⁺, and 20 mM HEPES [pH 7.4]) and allowed to incubate for 1 h at 37 °C in 60 μL of fresh assay buffer. Cells were subsequently treated with 30 μL of drugs (3×; agonist and modulator compounds) prepared in drug buffer (Hank’s buffered salt solution with Mg²⁺, Ca²⁺, 0.3% BSA, 0.03% ascorbic acid, and 20 mM HEPES [pH 7.4]), and bioluminescence signals were assessed at the times indicated using a Mithras 940 Multimode Plate reader. For β-arrestin-2 recruitment assays, 10 μL of the Rluc8 substrate coelenterazine h (Promega, Madison, WI; 5 μM final concentration) was added per well 15 min prior to measurements to allow for cell penetration and was read for bioluminescence at 480 nm and fluorescence eYFP emission at 515 nm. BRET¹ ratios were calculated (emission at 515/emission at 480) first, and then the NET BRET¹ was determined by subtracting the BRET ratio from control wells containing cells transfected with only the donor plasmid (human A₃AR-Rluc8) from the BRET ratio of experimental wells. For BRET² G isoprotein-coupling assays, 10 μL of the RLuc8 substrate coelenterazine 400a (Nanolight Technology, Pinetop, AZ; 5 μM final concentration) was added per well 15 min before measurement of bioluminescence at 385 nM and fluorescent GFP² emission at 510 nm. NET BRET² ratios were calculated in the same manner as described earlier for the BRET¹ assays.

cAMP Production Assays

HEK293T cells were cotransfected (1:30) with pcDNA3.1 encoding either the human or mouse A₃AR and the pGloSensor-22F reporter plasmid (Promega, Madison, WI) in DMEM supplemented with 10% fetal bovine serum, 100 units/mL penicillin, and 100 μg/mL streptomycin using TransIT-293 transfection reagent (Mirus Bio, Madison, WI). After 24 h, transfected cells were seeded in poly-l-lysine-coated, white-walled, clear-bottom, 96-well cell culture plates at a density of 30 000 cells per well in DMEM with 1% dialyzed fetal bovine serum, 100 units/mL penicillin, and 100 μg/mL streptomycin and 25 mM HEPES (pH 7.4). The next day, the cell culture medium was aspirated, and cells were washed once with assay buffer (Hank’s buffered salt solution with Mg²⁺, Ca²⁺, and 20 mM HEPES [pH 7.4]) and allowed to incubate for 2 h at 37 °C in 50 μL of fresh assay buffer containing d-luciferin (4 mM). ZM-241385 and PSB-603 (300 nM each) were also included to block A_2BARs expressed endogenously in HEK293T cells. After incubation with the substrate, the plate was removed from the incubator and allowed to equilibrate at room temperature for 15 min, at which point a baseline chemiluminescence reading was acquired. Cells were subsequently treated with 5 μL of test compounds (12×; agonist and modulator compounds) prepared in assay buffer for 15 min followed by the addition of 5 μL of forskolin (12×; 1 μM final concentration) to stimulate production of cAMP. Chemiluminescence readings were obtained 30 min after the addition of forskolin.

Data Analysis

E_max and EC₅₀ values were calculated from data obtained from agonist concentration–response curves according to E = (E_max × x)/(EC₅₀ + x), in which x is the agonist concentration. All values were statistically compared using a one-way ANOVA followed by Bonferroni-adjusted t tests for post hoc comparison or a two-tailed Student’s t test, as indicated. For [¹²⁵I]I-AB-MECA dissociation binding assays, data were fit to a one-phase exponential decay model: Y = (Y₀ – NS)^(−k×t), in which Y₀ is the specific binding at time 0, k is the dissociation rate constant, and t is the elapsed time. Affinity estimation of the 2-cyclopropyl derivative for the orthosteric site at Human_Out/Mouse_In chimeras was calculated from data obtained from agonist concentration–response curves according to Y = bottom + (top – bottom)/(1 + 10^{((log(EC–x))×Hillslope)}), in which x is the agonist concentration with Hillslope and Schild slope values constrained to 1. All values are presented as the mean ± SEM. A p value <0.05 was considered statistically significant.

Results

Pharmacological Characterization of Novel A₃AR PAMs

Equilibrium binding assays using an orthosteric agonist radioligand at a nonsaturating concentration gauge the net effect of modulator compounds on orthosteric agonist binding due to combined allosterism and potential competitive antagonism, which should inform effects on agonist potency. In prior work,^22,24−27 we have not succeeded in identifying any derivatives that increase orthosteric agonist radioligand equilibrium binding (in fact, nearly all reduce it). However, these initial studies using [¹²⁵I]I-AB-MECA were performed using conventional incubation times (≤2 h), which we have since learned may not be sufficient to achieve equilibrium conditions because the modulators can slow the binding kinetics.²⁴ In preliminary studies, we retested a previously synthesized series of imidazoquinolin-4-amine derivatives^25,26 with hydrophobic mono- or bicycloalkyl substitutions at the 2- position and the 3,4-dichlorophenyl group at the 4-amino position (2–5, 11, and 12; Table 1), including our current lead compound 4 (LUF6000) containing a 2-cyclohexyl substituent, where the incubation time was extended to 18 h. Under these conditions, it was revealed that several derivatives at a concentration of 10 μM increased binding by as much as ∼30%. Because binding tended to increase as the 2-cycloalkyl substituent increased in size, we expanded this series to include ring systems ranging from 3 to 12 carbons, as well as a flexible hydrophobic heptan-4-yl derivative (1, 6–10, and 13; Table 1). To allow a direct comparison, we also expanded our existing library of quinoline derivatives (15–17, 21, and 22; Table 1)²⁷ containing a 2-(3,4-dichlorophenyl)amine substituent with matching substitutions (up to C-9) at the 4-amide position (18–20 and 23; Table 1). Within this series, compound 17 (LUF6096) is the cyclohexyl derivative analogous to reference compound 4 in the imidazoquinolin-4-amine series. Synthesis of the new quinoline derivatives is included in the Supporting Information. Detailed synthetic methods for the imidazoquinolin-4-amine derivates will be reported elsewhere.²⁸

Table 1. 2-Substituted Imidazoquinolin-4-amine and 4-Substituted Quinoline Derivatives.

A full comparison of the two series of modulator compounds on [¹²⁵I]I-AB-MECA equilibrium binding using membranes prepared from HEK293 cells expressing the human (h) A₃AR is presented in Figure 2A (values reported in Table S1). For the imidazoquinolin-4-amine series, it was remarkable that derivatives with larger 2-cycloalkyl substitutions, beginning with the 2-cycloheptyl derivative (5), began to increase [¹²⁵I]I-AB-MECA equilibrium binding. However, this effect was lost as the ring size increased beyond C-9 (7). The more flexible heptan-4-yl derivative (13) produced the greatest increase in binding (70%), and the tricyclic 2-adamantanyl derivative (12) increased binding modestly (20%). Notably, the C-3 derivative (1) reduced binding by up to 70%. In contrast, most of the quinoline derivatives reduced [¹²⁵I]I-AB-MECA equilibrium binding, but the direction of this effect upon changes of the 2-position substitution followed a similar SAR as the imidazoquinolin-4-amine series. Thus, the cyclohexyl (17) and cycloheptanyl (18) groups in the quinoline series had the least-negative values for the [¹²⁵I]I-AB-MECA equilibrium binding. The exception to this pattern was the 4-heptan-4-yl derivative in the quinoline series (23), which surprisingly produced the greatest decrease in binding. Thus, the SARs of the PAMs derived from the two chemical scaffolds are highly divergent.

Figure 2.

(A) Equilibrium binding assays with membranes prepared from HEK293 cells expressing the hA₃AR. Membranes were incubated with 0.3 nM [¹²⁵I]I-AB-MECA and 10 μM modulator compounds for 18 h. Data are presented as the percent change in specific binding compared to vehicle. (B) Dissociation binding assays with membranes prepared from HEK293 cells expressing the hA₃AR. Membranes were incubated with ∼0.3 nM [¹²⁵I]I-AB-MECA and 10 μM modulator compounds for 3 h when the assay was initiated by the addition of 100 μM NECA. After 60 min, the amount of specific binding remaining was determined. Data are presented as mean ± SEM. *p < 0.05 vs vehicle by Student’s t tests. Percent change from vehicle (A) and % [¹²⁵I]I-AB-MECA remaining after 60 min (B) values are reported in Tables S1 and S2.

To evaluate pure allosteric actions, we assessed the effects of the modulators in a [¹²⁵I]I-AB-MECA dissociation binding assay using transfected HEK293 cell membranes (Figure 2B and Table S2). Cell membranes were incubated with [¹²⁵I]I-AB-MECA for 3 h, and then binding was determined after the addition of test compounds and excess NECA (100 μM), which prevented [¹²⁵I]I-AB-MECA from rebinding after it had dissociated. In the absence of modulator (vehicle), ∼25% of the radioligand remained bound 1 h after the addition of NECA (dotted line in Figure 2B). An increase in binding by a modulator compound indicates an allosteric effect to slow [¹²⁵I]I-AB-MECA dissociation. Using this assay, it was observed that several of the derivatives from both series at a concentration of 10 μM slowed [¹²⁵I]I-AB-MECA dissociation with a SAR similar to that observed in equilibrium binding assays, whereby increasingly bulky cycloalkyl substitutions from C-4 to C-8/9 (2–7 and 15–19) produced maximal effects. The bicyclic (except for the adamantanyl-substituted quinoline derivative, 22) and especially the heptan-4-yl derivatives (13 and 23) also slowed [¹²⁵I]I-AB-MECA dissociation. Maximal slowing was produced by the 2-heptan-4-yl imidazoquinolin-4-amine modulator (13).

Several interesting observations arise when the equilibrium binding data in Figure 2A and the dissociation binding data in Figure 2B are considered together. Neither compound with the cyclopropyl substitution (1 and 14) affected the [¹²⁵I]I-AB-MECA dissociation binding rate and therefore lacked allosteric actions. We hypothesized that the reduction in equilibrium binding produced in both compound series presumably was caused by competitive antagonism rather than negative allosteric modulation. This hypothesis was later supported using chimeric receptor constructs (see below). For other compounds that slowed [¹²⁵I]I-AB-MECA dissociation without increasing equilibrium binding, which occurred with several of the quinoline derivatives (16, 19, 21, and 23), prominent competitive antagonism was also assumed. All compounds that increased equilibrium binding, including compounds 5, 6, 12, and 13 within the imidazoquinolin-4-amine series containing C-7/8, cycloalkyl, bicyclic, or the heptan-4-yl substitutions, produced prominent effects to slow [¹²⁵I]I-AB-MECA dissociation, implicating prominent allosterism with diminished competitive antagonism.

We next evaluated the two series of modulators in a [³⁵S]GTPγS binding assay using isolated HEK293 cell membranes expressing the hA₃AR, which detects receptor-induced G protein activation. For these studies, concentration–response curves with the orthosteric agonist Cl-IB-MECA were assessed in the presence of vehicle or 0.1, 1, or 10 μM modulator compound. As shown in Figures 3 (imidazoquinolin-4-amine series) and 4 (quinoline series), a clear SAR emerged with each series of compounds that correlated well with the single-point radioligand binding assays (EC₅₀ and E_max for Cl-IB-MECA are reported in Table S3). Indeed, the cyclopropyl derivatives from both chemical series (1 and 14) right-shifted the Cl-IB-MECA concentration–response curve in a concentration-dependent manner without altering the efficacy, indicative of competitive antagonism. As the ring size increased, efficacy enhancement became apparent, beginning with the C-4 derivatives (2 and 15), while potency reduction diminished, correlating with the effect of the modulators in radioligand binding assays to slow [¹²⁵I]I-AB-MECA dissociation while tending to improve [¹²⁵I]I-AB-MECA equilibrium binding. The greatest increase in efficacy enhancement within the imidazoquinolin-4-amine series occurred with the C-6 (4), C-9 (7), norbornanyl (11), and adamantanyl (12) derivatives; among these, the C-9 (7) and adamantanyl (12) derivatives tended to increase the potency of Cl-IB-MECA, whereas the C-6 (4) and norbornyl (11) derivatives tended to reduce it. While the magnitude of the efficacy-enhancing action of the heptan-4-yl derivative (13) was not as prominent, it is notable that it appeared to be the most potent among this series given that it produced a maximal degree of efficacy enhancement at a concentration of only 100 nM. In comparison, the lowest concentration to produce maximal activity for the other compounds exhibiting substantial PAM activity, including compound 4, was 1 μM. It is necessary to mention that efforts to fit the [³⁵S]GTPγS binding data to operational models^33,34 to calculate affinity constants and cooperativity factors were not successful, presumably because the compounds have additional pharmacological actions (i.e., competitive antagonism) or because of the limited amount of data that was collected. Within the quinoline series (Figure 4), our lead compound 17 containing the cyclohexyl substitution produced the greatest increase in efficacy without reducing the potency of Cl-IB-MECA. Correlating with the [¹²⁵I]I-AB-MECA equilibrium binding data, this series of derivatives in general produced a greater reduction in agonist potency.

Figure 3.

[³⁵S]GTPγS binding assays with the imidazoquinolin-4-amine series of 2-positioned derivatives and membranes prepared from HEK293 cells expressing the human A₃AR. Concentration–response curves with Cl-IB-MECA were conducted in the presence of vehicle or 0.1, 1.0, or 10 μM modulator compounds. Results were normalized to the E_max value obtained in the presence of vehicle. Data are presented as mean + SEM; n = 3. * p < 0.05 vs vehicle by one-way ANOVA followed by Bonferroni-adjusted t tests for post hoc comparison. EC₅₀ and E_max values are reported in Table S3.

Figure 4.

[³⁵S]GTPγS binding assays with the quinoline series of derivatives and membranes prepared from HEK293 cells expressing the human A₃AR. Concentration–response curves with Cl-IB-MECA were conducted in the presence of vehicle or 0.1, 1.0, or 10 μM modulator compounds. Results were normalized to the E_max value obtained in the presence of vehicle. Data are presented as mean ± SEM; n = 3. * p < 0.05 vs vehicle by one-way ANOVA followed by Bonferroni-adjusted t tests for post hoc comparison. EC₅₀ and E_max values are reported in Table S3.

Results of Agonist Activity at the mA₃AR for the Newly Synthesized Derivatives

We assessed activity of the two series of modulators in assays using HEK293 cells expressing the mouse (m) A₃AR. Both of our current leads (4 and 17) display minimal allosteric activity with rodent A₃AR’s,²⁴ and we are eager to identify pan-species derivatives that will facilitate efforts to characterize their biological activity in disease models. Unfortunately, none of the newly synthesized compounds showed improved activity. Figure 5 presents results of the 2-substituted cyclopropyl (1), cyclohexyl (4), cyclononanyl (7), and heptan-4-yl (13) derivatives from the imidazoquinolin-4-amine series that performed best in assays with the hA₃AR. In the [³⁵S]GTPγS binding assay using mA₃AR membranes (Figure 5A), 4 and the heptan-4-yl derivative (13) increased the efficacy of Cl-IB-MECA by only ∼35 and ∼60% at the highest concentration tested, respectively (10 μM; EC₅₀ and E_max values for Cl-IB-MECA are presented in Table S4). In comparison, in assays using hA₃AR membranes, compound 4 and the heptan-4-yl derivative at a concentration of 10 μM increased the efficacy of Cl-IB-MECA by ∼250% and ∼220%, respectively. In [¹²⁵I]I-AB-MECA binding assays, none of the derivatives from either series enhanced [¹²⁵I]I-AB-MECA equilibrium binding (Figure 5B; Table S4), nor did they slow the rate of [¹²⁵I]I-AB-MECA dissociation (Figure 5C; Table S4) in assays with the mA₃AR.

Figure 5.

Characterization of select imidazoquinolin-4-amine derivatives at the mouse A₃AR. (A) [³⁵S]GTPγS binding assays, (B) [¹²⁵I]I-AB-MECA dissociation binding assays, and (C) [¹²⁵I]I-AB-MECA equilibrium binding assays. Assays were conducted using HEK293 cells stably expressing the mouse A₃AR. Data are presented as mean ± SEM; n = 3. EC₅₀ and E_max values from the [³⁵S]GTPγS binding assays and values for the [¹²⁵I]I-AB-MECA binding assays are presented in Table S4.

Evidence That the Imidazoquinolin-4-amine PAMs Act at Two Distinct Sites on the hA₃AR

Taking advantage of species differences,²⁴ we generated stable HEK293 cell lines expressing human/mouse chimeric A₃ receptors in efforts to begin to localize the allosteric binding site. Two chimeric A₃ARs were prepared in which the lower transmembrane regions of the human sequence were changed to the mouse sequence (H_Out/M_In) and vice versa (M_Out/H_In). Division between the upper and lower segments of the transmembrane regions were at residue x.50 defined by the Ballesteros–Weinstein numbering scheme, where x.50 is the most-conserved residue in that helix. Figure 6 illustrates differences in amino acid sequences in the species chimeras, which were most prominent in the extracellular loops and extracellular portions of the transmembrane helices, but differences were also notable in the intracellular loop segments and helix 8. The residues previously identified by mutagenesis and modeling to coordinate the orthosteric nucleoside agonists are all contained in the upper segments.^16,17 In [³⁵S]GTPγS binding assays with the M_Out/H_In chimera, strikingly compound 4 retained the ability to increase the efficacy of Cl-IB-MECA to a similar extent as in assays with the wild-type hA₃AR, whereas efficacy enhancement was minimal in assays with the H_Out/M_In chimera, similar to that observed with the wild-type mA₃AR (Figure 7A; EC₅₀ and E_max values for Cl-IB-MECA are reported in Table S5). An identical profile was obtained in [¹²⁵I]I-AB-MECA dissociation binding assays, wherein compound 4 slowed dissociation from the M_Out/H_In chimera but not from the H_Out/M_In chimera (Figure 7B; Table S5). This finding that the allosteric activity of compound 4 is retained when intracellular portions of the A₃AR are composed of the human amino acid sequence suggests that the imidazoquinolin-4-amine allosteric binding site is distinct from the orthosteric site and that it resides within the inner cytoplasmic portions of the receptor.

Figure 6.

Snake diagram of the human A₃AR amino acid sequence. Residues highlighted in green indicate outside residues, while those in white indicate inside residues of the human/mouse chimeric receptors. Amino acid residues that differ between the human and mouse sequences are highlighted in red. This diagram was prepared using the GPCRdb database (http://www.gpcrdb.org).

Figure 7.

Characterization of human/mouse chimeric A₃ARs. (A, C) [³⁵S]GTPγS binding assays with Cl-IB-MECA (agonist) and either compound 4 (A) or compound 1 (C). (B) [¹²⁵I]I-AB-MECA dissociation binding assays in the presence of vehicle or 10 μM compound 4. (D, E) [¹²⁵I]I-AB- MECA equilibrium binding assays in the presence of increasing concentrations of compound 1 (D) or compound 4 (E). EC₅₀ and E_max values from the [³⁵S]GTPγS binding assays and values for the [¹²⁵l]l-AB-MECA dissociation binding assays are presented in Table S5.

We noticed in the chimeric receptor studies that compound 4 had less propensity to reduce the potency of Cl-IB-MECA in [³⁵S]GTPγS binding assays with the M_Out/H_In chimera (Figure 7A; Table S5). In fact, the EC₅₀ of Cl-IB-MECA in the presence of 10 μM compound 4 was significantly reduced 5-fold from 10.5 to 2.24 nM (p < 0.05), while in assays with the wild-type hA₃AR the EC₅₀ was increased 5-fold by 10 μM compound 4 (15.1 vs 72.5 nM; p < 0.05). Because the binding affinity of non-nucleoside typical non-nucleoside hA₃AR antagonist ligands in the mA₃AR is greatly reduced as compared to the hA₃AR,³⁵ we predicted that the enhanced potency of Cl-IB-MECA observed with the M_Out/H_In chimera is explained by reduced affinity of compound 4 for the orthosteric binding site composed of the mouse sequence. To further test this idea, we examined the 2-cyclopropyl imidazoquinolin-4-amine derivative (1) in [³⁵S]GTPγS binding assays with human/mouse chimeric receptors, which lacks efficacy-enhancing allosteric activity but greatly right-shifts the Cl-IB-MECA concentration–response relationship in assays with the hA₃AR in a manner suggestive of competitive antagonism. As shown in Figure 7C (EC₅₀ and E_max values for Cl-IB-MECA are reported in Table S5), compound 1 produced a similar rightward shift in assays with the H_Out/M_In chimera but had little effect in assays with the M_Out/H_In chimera, confirming competitive antagonism. Cl-IB-MECA concentration–response curves treated with vehicle, 0.1, 1.0, or 10 μM of the 2-cyclopropyl derivative (1) at the H_Out/M_In chimera were used to estimate the affinity of this derivative for the orthosteric site by Schild analysis, where a purely competitive antagonist effect was observed and determined to be ∼135 nM (pA2 = 6.87 ± 0.19). Lastly, we compared the effects of compounds 1 and 4 in [¹²⁵I]I-AB-MECA equilibrium binding assays using the human/mouse A₃AR chimeras. As shown in Figure 7D, the results support the idea that compound 1 functions as a competitive antagonist with greater potency for the hA₃AR as compared to the mA₃AR, and data in Figure 7E support a dual mechanism of action of compound 4, evidenced by a reduction in binding in assays with the H_Out/M_In chimera (binds with high affinity to the human orthosteric binding site but has low activity at the mouse allosteric site) and an increase in binding with the M_Out/H_In chimera (binds with low affinity to the mouse orthosteric binding site while fully activating the human allosteric site). Thus, the chimeric receptor studies not only aided in localizing the allosteric binding site to cytosolic regions of the receptor but also provided further evidence that this family of PAMs can also act dually as competitive antagonists for the orthosteric binding site and allowed us to derive modulator affinity estimates for the orthosteric site.

Preferential Enhancement of Adenosine-Induced hA₃AR G_α Isoprotein Coupling by Compounds 4 and 17

We investigated the activity of our two lead PAM compounds to modulate A₃AR-induced activation of specific G protein isoforms and β-arrestin-2 translocation using BRET technology to explore the possibility that they may preferentially enhance select signaling pathways. The G protein-coupling assays involved measurement of BRET² fluorescence after expressing various G_α-Rluc8 fusion proteins and the G_γ2-GFP² fusion protein in HEK293T cells (TRUPATH³²). With this assay, loss of BRET² signal, reflecting dissociation of the G_α and G_βγ subunits, indicates receptor-mediated activation of each specific G protein isoform. β-arrestin-2 recruitment was assessed by expressing both the hA₃AR-Rluc8 and β-arrestin-2-YFP fusion proteins in HEK293T cells and measuring the attainment of BRET¹ fluorescence.

We initially identified the specific G proteins that are activated by the hA₃AR in response to adenosine. As expected, adenosine induced activation of all three Gα_i isoforms (Gα_i1,2,3) with EC₅₀ values in the range of 1–10 μM (Figure 8A; Table S6). Coupling to Gα_oA and Gα_oB was also detectable, with the latter being highly variable. No coupling was observed to the Gα_Z, Gα_SS, Gα_SL, Gα_q, or Gα_12/13 isoforms. In these assays, bioluminescence emission values for the various Gα-Rluc constructs were similar (Table S7). To address the potential for probe-dependence, we further assessed G protein activation profiles of 2-chloroadenosine (2-CADO) and Cl-IB-MECA, which were similar to adenosine, except that Cl-IB-MECA was ∼35% more efficacious at activating Gα_i3 and 2-CADO was unable to activate Gα_oA up to a concentration of 100 μM (Figure 8C and Table S8). In BRET¹ assays for the assessment of β-arrestin-2 signaling, all three of the agonists also stimulated β-arrestin-2 recruitment (Figure 8B and C). However, Cl-IB-MECA and 2-CADO were, respectively, ∼40% and 20% less efficacious as compared to adenosine and thus appeared to function as partial agonists in this assay. A heat map comparing relative potencies and efficacies of the three agonists to stimulate G isoprotein activation and β-arrestin-2 recruitment is displayed in Figure 8D.

Figure 8.

Agonist-induced G isoprotein activation and β-arrestin-2 recruitment by the human A₃AR. (A) BRET² G protein profiling assays using adenosine as the orthosteric agonist. Data for each G isoprotein are presented relative to vehicle. (B) BRET¹ assays assessing β-arrestin-2 recruitment. (C) BRET² and BRET¹ assays assessing effects of adenosine, 2-CADO, and Cl-IB-MECA on activation of G isoproteins or β-arrestin-2 as indicated. Results were normalized to the E_max value obtained in the presence of adenosine. (D) Heat maps relating potency and efficacy of the agonists to activate the indicated G isoproteins and β-arrestin-2 recruitment. BRET assays were conducted with HEK293T cells transfected with the human A₃AR and constructs for BRET assessment as described in the Methods section. Colors indicate log EC₅₀ and efficacy values. Data are presented as mean ± SEM; n = 3. EC₅₀ and E_max values are presented in Table S8.

We next assessed the actions of compounds 4 and 17 on hA₃AR-induced G isoprotein activation and β-arrestin-2 recruitment in response to adenosine. HEK293T cells transfected with the hA₃AR were simultaneously stimulated with increasing concentrations of adenosine and 10 μM of either 4 or 17, and BRET measurements were assessed for up to 2 h. As shown in Figure 9 (EC₅₀ and E_max values for adenosine are reported in Table S9), both compounds enhanced the efficacy of adenosine to activate both Gα_i3 and Gα_oA by ∼35 and 60%, respectively, whereas neither compound affected adenosine-induced activation of Gα_i1 or Gα_i2, although compound 17 exhibited agonism in the Gα_i1 and Gα_i2 protein-coupling assays, an effect that was also observed in a cAMP production assay (Figure S1). An efficacy-enhancing effect of both compounds was observed as early as 15 min after the addition of adenosine in assays of Gα_oA, whereas efficacy enhancement was only observed at the 60 min time point in assays with Gα_i3. At the 60 min time point, compounds 4 and 17 were found to not only increase the efficacy of adenosine but also increase their potencies by ∼7- and 40-fold, respectively. Neither compound influenced adenosine-induced β-arrestin-2 recruitment. Thus, the modulators preferentially enhanced adenosine-induced activation of the Gα_i3 and Gα_oA isoforms. Figure 10 (Table S10) displays concentration–response curves with compounds 4 and 17 on Gα_i3 and Gα_oA isoprotein coupling, in comparison with β-arrestin-2 recruitment. Similar data for compounds 7 and 13 are reported in the Supporting Information (Figure S2).

Figure 9.

Effects of compound 4 and compound 17 on agonist-induced G isoprotein activation and β- arrestin-2 recruitment by the hA₃AR over time. BRET² assays to assess G protein coupling and BRET¹ assays to assess β-arrestin-2 recruitment were conducted with adenosine in the presence of vehicle, 10 μM compound 4, or 10 μM compound 17 as described in the Methods section. Results were normalized to the E_max value obtained in the presence of vehicle. Data represent mean + SEM; n = 3–5. EC₅₀ and E_max values are presented in Table S9.

Figure 10.

In-depth analysis of specific G isoproteins identified to be potentiated by compounds 4 and 17 at select time points. BRET² and BRET¹ assays to assess (A) Gaα_oA protein coupling and β-arrestin- 2 recruitment at 30 min and (B) Gaα_l3 protein coupling and β-arrestin-2 recruitment at 60 min were conducted with adenosine in the presence of vehicle, compound 4, or compound 17 as described in the Methods section. Results were normalized to the E_max value obtained in the presence of vehicle. Data represent mean ± SEM; n = 3. EC₅₀ and E_max values are presented in Table S9.

Discussion

Because adenosine is produced locally at sites of tissue injury, hypoxia, and inflammation,¹ an allosteric tactic is likely an optimal approach to therapeutically target each of the four adenosine receptor subtypes. In the present study, we have compared and expanded knowledge of the pharmacological properties of two previously identified classes of compounds that function as A₃AR PAMs, namely, 1H-imidazo[4,5-c]-imidazoquinolin-4-amines represented by our lead compound LUF6000 (4)²⁴⁻²⁶ and 2,4-disubstituted quinolines represented by LUF6096 (17).^24,27 Salient findings from this work are as follows. First, expanded SAR studies revealed that bulkier cycloalkyl substitutions at the 2-position of the imidazoquinolin-4-amine and the 4-amide position of the quinoline scaffold, each containing a 3,4-dichlorophenyl substituent at the 4- and 2-amino positions, respectively, produced substantial improvements in allosteric actions of the derivatives to promote agonist efficacy enhancement with lessened propensity to reduce agonist potency. Maximal responses were obtained with ring systems containing 6–9 carbons yielding EC₅₀ values estimated to be in the range of 0.1–1 μM. Second, more in-depth evaluation of our lead compounds using cell-based BRET assays with transfected HEK293T cells determined that, among the four G isoproteins coupled with the A₃AR (α_i1, α_i2, α_i3, and α_oA), both compounds 4 and 17 enhanced adenosine-induced activation of the Gα_i3 and Gα_oA isoproteins; neither compound influenced the potency or efficacy of adenosine to elicit β-arrestin-2 recruitment. Thus, we provide evidence that compounds 4 and 17 potentiate adenosine-mediated A₃AR signaling preferentially to specific Gα isoproteins. Two newly synthesized imidazoquinolin-4-amine modulators described in this investigation, compounds 7 and 13, exhibited similar pharmacological profiles. Finally, utilizing an extracellular/intracellular chimeric A₃AR approach composed of sequences from a responding (human) and a nonresponding (mouse) species, we have confirmed our prior prediction that the imidazoquinolin-4-amine class of PAMs retains variable competitive orthosteric antagonist activity, which likely explains the propensity of these compounds to reduce agonist potency. Loss of PAM activity when intracellular portions of the A₃AR are composed of the mouse sequence (nonresponding) provided evidence that the allosteric binding pocket is located distal to the orthosteric site within the inner regions of the receptor.

The involvement of specific Gα isoproteins in mediating the biological actions of the A₃AR remain unknown; thus, we can only speculate at the present time on the potential therapeutic advantages that may be provided by the PAMs described in this investigation. The Gα_i3 isoprotein is the least ubiquitously expressed of the Gα_i/o isoforms and is enriched in immune cell populations including granulocytic cells and microglia where expression of the A₃AR is also prominent.³⁶⁻³⁸ This has been elegantly demonstrated by single-cell mRNA sequencing of mouse and human tissues.^36,38 The Gα_i3 isoprotein has been implicated in the regulation of leukocyte chemotaxis,³⁹ and our prior work suggests that increased A₃AR signaling may indirectly influence oxidant production and chemotaxis through sequestration of cellular components involved in cellular activation, including rac1.^7,40 Taken together, preferential enhancement of adenosine-induced Gα_i3 activation by the A₃AR in leukocytes might be an effective and more tolerable means to treat inflammatory diseases. On the other hand, the Gα_oA isoprotein is known to be found in neurons in the brain as well as in the spinal cord where the A₃AR is also reported to be abundantly expressed.^36,38 Preclinical work has demonstrated efficacy of A₃AR agonists in a number of different models of neuropathic pain due to mechanical injury and chemotherapeutics that provide a benefit not only by reducing inflammatory responses but also by suppressing neuronal circuits in the spinal cord involved in pain sensation.^15,41−44 Thus, the current families of PAMs described herein might be highly effective therapies for the treatment of neuropathic pain due to selective enhancement of two analgesic mechanisms. It is unfortunate that we are currently limited in our ability to investigate the biological actions of the PAMs we have characterized thus far in animal models in detail because we have yet to identify any derivatives that display meaningful activity with rodent A₃ARs.²⁴ In a recent study, however, we reported that compound 17 was effective at limiting infarct size in a dog model of ischemia/reperfusion injury.⁴⁵ Because it produced a reduction in infarct size when it was given at the onset of reperfusion, we predict that it was effective in this model in part by suppressing inflammatory responses.⁴⁵

As part of the G isoprotein BRET studies, we have reported the G protein coupling profile of the A₃AR in a live cell system for the first time. The endogenous agonist adenosine elicited activation of Gα_i1, α_i2, α_i3, and α_oA. Two synthetic agonists commonly used in adenosinergic research, specifically 2-CADO and the selective A₃AR agonist Cl-IB-MECA, produced similar responses, although Cl-IB-MECA was more efficacious at activating Gα_i3 and 2-CADO was not able to activate Gα_oA. In comparison to adenosine and 2-CADO, Cl-IB-MECA was ∼50% less effective at activating β-arrestin-2 recruitment, and thus it appears to favor G protein activation of the A₃AR when compared to adenosine. This property of Cl-IB-MECA may explain its broad efficacy in preclinical models of inflammatory diseases.^{4,14,43,46,47}

We were surprised by our findings from the species chimera experiments that the imidazoquinolin-4-amine PAMs appear to compete for orthosteric ligand binding while acting separately within an inner region of the receptor to produce allosteric enhancement. This is supported by (1) our findings that PAM activity is only present when intracellular regions of the receptor comprise the human sequence and (2) findings with the H_Out/M_In chimera where the efficacy-enhancing activity of compound 4 is absent, yet competitive antagonistic activity of this compound as well as compound 1, which lacks PAM activity, remains. We speculate that the imidazoquinolin-4-amine PAMs, and presumably the quinoline PAMs, may interact with the orthosteric site, or perhaps with an extracellular vestibule described in other GPCRs that controls the entry and efflux of agonist ligands into the orthosteric site.^48,49 In an early guided molecular dynamics simulation and molecular modeling study, compound 4 was predicted to interact at both of these sites.⁵⁰ At the present time, we do not know the location of the imidazoquinolin-4-amine allosteric site, and we are pursuing additional mutagenesis studies in combination with molecular modeling and structural approaches to gain further insights. It is notable that there are a number of structure-based reports of small-molecule PAMs binding to sites located within inner regions of other GPCRs.^48,49 For example, compound 6FA, a PAM of the β₂ adrenergic receptor, binds to the inner surface of the receptor in a pocket created by intracellular loop 1 and transmembrane helices 3 and 4.⁵¹ Compound 6FA binding is thought to stabilize the loop in an α-helical formation that maintains an opening for insertion of the terminal α-helix of the G protein. An ago-PAM for the class B1 human glucagon-like peptide-1 (GLP-1) receptor, termed compound 1, binds covalently to the membrane-facing surface of TMD6, causing its outward movement along with conformation changes in the extracellular domains that facilitate insertion of the amino terminus into the GLP-1 binding site.⁵² Finally, a cryo-electromicroscopic structure of the A₁ adenosine receptor complexed with a Gα_i2 heterotrimer, adenosine, and the 2-amino-3-benzoylthiophene PAM MIPS521 revealed an extrahelical lipid-facing allosteric binding pocket involving transmembrane helices 1, 6, and 7.⁵³ MIPS521 binding is hypothesized to stabilize the adenosine–receptor–G protein complex.

In conclusion, we present expanded SAR information for two promising families of A₃AR PAMs that selectively enhance G isoprotein-dependent signaling. Mechanistic studies provide support for a dual mechanism of action and the presence of an allosteric binding site within the inner/intracellular regions of the receptor. This study has advanced both structural and pharmacological understanding of these two classes of A₃AR PAMs. Our findings will guide future computational and structure-based approaches to develop improved A₃AR PAMs that have potential as anti-inflammatory and analgesic therapeutics.

Glossary

Abbreviations

2-CADO

2-Chloroadenosine

ADA

Adenosine deaminase

AR

Adenosine receptor

BSA

Bovine serum albumin

BRET

Bioluminescence resonance energy transfer

cAMP

Cyclic adenosine monophosphate

CHAPS

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

Cl-IB-MECA

2-Chloro-N⁶-(3-iodobenzyl)adenosine-5′-N-methylcarboxamide

DMEM

Dulbecco’s modified Eagle medium

EDTA

Ethylenediaminetetraacetic acid

EGTA

Triethylene glycol diamine tetraacetic acid

GPCR

G protein-coupled receptor

[³⁵S]GTPγS

Guanosine 5′-[γ[³⁵S]thio]triphosphate

HEK293

Human embryonic kidney 293

HEPES

2-[4-(2-Hydroxyethyl)piperazin-1-yl]ethanesulfonic acid

[¹²⁵I]I-AB-MECA

N⁶-(4-amino-3-[¹²⁵I]iodobenzyl)adenosine-5′-N-methycarboxamide

NASH

Nonalcoholic steatohepatitis

NECA

Adenosine-5′-N-ethylcarboxamide

RLuc8

Renilla luciferase 8

PAM

Positive allosteric modulator

PBS

Phosphate-buffered saline

SAR

Structure–activity relationship

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsptsci.2c00076.

• Synthetic methods of quinoline derivatives; values from [¹²⁵I]I-AB-MECA and [³⁵S]GTPγS binding assays with human, mouse, and human/mouse chimeric A₃ARs; and values from BRET G isoprotein coupling, β-arrestin-2 recruitment, and cAMP accumulation assays (PDF)

The authors declare no competing financial interest.

Special Issue

Published as part of the ACS Pharmacology & Translational Science virtual special issue “GPCR Signaling”.