Hit-to-Lead Optimization of Heterocyclic Carbonyloxycarboximidamides as Selective Antagonists at Human Adenosine A3 Receptor

Abstract

Antagonism of the human adenosine A₃ receptor (hA₃R) has potential therapeutic application. Alchemical relative binding free energy calculations of K18 and K32 suggested that the combination of a 3-(2,6-dichlorophenyl)-isoxazolyl group with 2-pyridinyl at the ends of a carbonyloxycarboximidamide group should improve hA₃R affinity. Of the 25 new analogues synthesized, 37 and 74 showed improved hA₃R affinity compared to K18 (and K32). This was further improved through the addition of a bromine group to the 2-pyridinyl at the 5-position, generating compound 39. Alchemical relative binding free energy calculations, mutagenesis studies and MD simulations supported the compounds’ binding pattern while suggesting that the bromine of 39 inserts deep into the hA₃R orthosteric pocket, so highlighting the importance of rigidification of the carbonyloxycarboximidamide moiety. MD simulations highlighted the importance of rigidification of the carbonyloxycarboximidamide, while suggesting that the bromine of 39 inserts deep into the hA₃R orthosteric pocket, which was supported through mutagenesis studies 39 also selectively antagonized endogenously expressed hA₃R in nonsmall cell lung carcinoma cells, while pharmacokinetic studies indicated low toxicity enabling in vivo evaluation. We therefore suggest that 39 has potential for further development as a high-affinity hA₃R antagonist.

Introduction

Adenosine, a naturally occurring purine nucleoside, is an endogenous agonist of adenosine receptors (ARs).¹ ARs are G protein-coupled receptors (GPCRs) comprising four subtypes: A₁, A_2A, A_2B, and A₃. While the A_2AR and A_2BR subtypes activate G_s to stimulate adenylyl cyclase, increasing 3′,5′-cyclic adenosine monophosphate (cAMP) levels, A₁R and A₃R conversely couple to G_i/o subunits inhibiting adenylyl cyclase. Beyond inhibiting cAMP accumulation, A₃R has been suggested to modulate mitogen-activated protein kinase (MAPK) activity which may explain the role of this receptor on cell proliferation and differentiation^2,3 and in tumor development and progression. Evidence suggests that human A₃R (hA₃R) antagonists might become new therapeutic tools for the treatment of both chronic renal disease⁴ and acute renal ischemia and reperfusion injury.⁴ Furthermore, hA₃R antagonists have demonstrated efficacy in eye pathologies.² Indeed, it has been reported that the potent A₃R antagonist MRS1220 (N-[9-chloro-2-(2-furanyl)-1,2,4-triazolo[1,5-c]quinazolin-5-yl]benzeneacetamide) prevented oligodendrocyte damage and myelin loss triggered by ischemia or by activation of A₃R in the rat optic nerve.⁵ Hence, blockage of hA₃R has proven to be useful for the treatment of diverse diseases; however, its role is still to be elucidated in other pathophysiological conditions, such as inflammation, cancer, or pain.² The identification of new potent and selective ligands which can clarify the therapeutic potential arising from blocking or stimulating the hA₃R remains an attractive objective.^2,6

Experimental structures have been resolved for all the four subtypes of hARs that have become established drug targets, although hA₃R only has an active structure reported.⁷ Such experimental structures can help to understand the binding interactions of ARs with ligands and provide templates for structure-based drug design (SBDD) as others⁸⁻¹⁴ and we¹⁵ have shown. Free energy perturbation coupled with the molecular dynamics simulation (FEP/MD) method^16,17 has been applied for lead fragment optimization against A_2AR^18,19 or for structure–activity relationship (SAR) interpretation, e.g., of 3-deazaadenosine agonists²⁰ and thiazolo[5,4-d]pyrimidines antagonists²¹ against A_2AR, or 4-substituted-1,4-dihydrobenzo[4,5]imidazo[1,2-a]pyrimidine-3-carboxylate antagonist against A_2BR.²² Furthermore, we have performed the equivalent with FEP/MD thermodynamic integration coupled with the molecular dynamics simulation method (TI/MD)²³ to describe accurately the structure–affinity relationships of antagonist of human A₁R (hA₁R).

In a previous work, from virtual screening (VS) of an ∼18,000 compound library, we identified hits of different structures as antagonists of ARs with a new structure having low micromolar affinities using radiolabeled assays.¹⁵ Of particular interest for further development were the hits K5, K9, K10, K11, K15, K17, K18, and K32 (Scheme 1) that are heterocyclic carbonyloxycarboximidamide derivatives.^15,24,25 These hit compounds (which we purchased from commercial libraries but are synthetically feasible) showed selective low micromolar affinity against hA₃R using radiolabeled assays, and we showed that the affinities were, in most cases, consistent with antagonistic receptor activities determined using inhibition of cAMP accumulation.²⁵

Scheme 1. Chemical Structures, Dissociation Constants with Radio-Labeled Assays (K_i in μΜ), and Antagonistic Potencies (pA₂) of K18/K32 Analogues Reported in Ref (25); n.a., Not Active.

Selecting ligands based on their affinity, an equilibrium parameter, does not necessarily predict in vitro activity, and a ligand’s kinetic properties may provide a better indication of how a ligand will perform in vivo. Kinetic profiling in the drug discovery process allows the resolution of ligand–receptor interactions into both molecular recognition (dependent on association rate constant K_on) and complex stability (dependent on ligand’s dissociation rate constant K_off). Significantly, this enables estimates of the residence time (RT = 1/K_off) of that ligand upon its target.²⁶ By testing the compounds shown in Scheme 1, we observed^15,24,25 that adding chlorine atoms in the phenyl ring of compound K5 increased affinity and antagonistic potency in K17 and K18. An additional interesting finding was that replacement of the five-membered thiazole ring of compound K17 with the six-membered pyridine ring maintained the binding affinity in K10, K11, and K32, as well.

Here, we presented a hit-to-lead study through SBDD using TI/MD^27,28 for the calculation of relative binding free energies and a previous SAR study.^15,24,25 The accuracy of perturbative binding free energy methods for ligands AR or other class A GPCR systems was previously shown using the FEP/MD^{20,22,29−31} as well as by TI/MD calculations on complexes of A₁R.³² Both of the FEP/MD³³ and TI/MD^27,28 methods can provide accurate results for relative binding free energies with a method error of 1 kcal mol^–1.²⁸

The SBDD and synthesis led to 25 new compounds (37–61) which we tested for their affinity at hA₃R using nanoluciferease-based bioluminescence resonance energy transfer (NanoBRET) binding assay. Among these, seven compounds 37, 39, 40, 47, 48, 59, and 60 displayed similar or significantly higher affinity than K18. Their binding kinetic parameters including K_on, K_off, and RT were determined, as well as the hAR subtype selectivity. Compounds 37 and 39 showed ∼10-fold increased potency against hA₃R and ∼20-fold higher RT compared to K18 or K32. Based on 37 and 39, we explored the tolerance of the chlorine atoms in the 2,6-dichlorophenyl group by synthesizing and testing 4 more new analogues (74–77) with bromine or methyl groups instead and found that 39 still showed the highest affinity and selectivity toward hA₃R. We then applied the TI/MD method²⁷ to confirm quantitatively with an accurate method the observed structure–affinity relationships. We performed 500 ns MD simulations of the representative compounds 37–39, 56–57, and 60 to describe the interactions with residues in the orthosteric binding pocket and conducted an exhaustive analysis with in vitro mutagenesis experimentation for our lead compound 39 in comparison with 37 that also informs for SARs^26,41 Further, since hA₃R has often been referenced as a promising target for treating cancer,³⁴ we explored the ability of 37 and 39 to selectively antagonize hA₃R in a nonsmall cell lung cancer cell line that endogenously expresses all 4 AR subtypes.³⁵ The ability of 39 to inhibit cancer cell proliferation led us to perform a preliminary pharmacokinetic study, which displayed good lipophilicity and permeability across intestinal cells but a relatively low aqueous solubility and metabolic stability. We therefore present the high-affinity hA₃R antagonist 39 as a new lead compound for future development.

Results

Structure-Based Drug Design

Previously using mutagenesis experiments, and MD simulations using the amber14sb force field (ff14sb)³⁶ or OPLS2005³⁷ with Poisson–Boltzmann or generalized Born and surface area continuum solvation (MM/PBSA³⁸ MM/GBSA calculations³⁸) binding free energy calculations,^24,25 we have suggested a preferred binding pose for K18 or K17 bearing the 1,3-thiazolyl and 2,6-dichlorophenyl or 2-chlorophenyl groups, respectively, and K32 (compound 42 in this study) having the 2-pyridinyl and 2-chlorophenyl groups at the two ends of the carbonyloxycarboximidamide linker. Τhe binding pose (after 500 ns MD simulations with ff19sb,³⁹) for compound K18 is shown in Figure 1A.

Figure 1.

500 ns MD simulations for the complex of compound K18 with the wild-type (WT) hA₃R using the amber ff19sb.³⁹ (A) Representative frame of K18 inside the orthosteric binding area; (B) receptor–ligand interaction frequency histograms; bars are plotted only for residues with interaction frequencies ≥0.2. Color figure in frames or bar plots: ligand is shown with pink sticks and ligand’s starting position with a pink wire, receptor is shown with a white cartoon and sticks, hydrogen bonding interactions are shown with yellow dashes or bars, π–π interactions are shown with green dashes or bars, hydrophobic interactions are shown with gray bars, and water bridges are shown with blue bars. (C) Root-mean-square deviation (rmsd) plots of Ca carbons of the protein (gray line) and of heavy atoms of the ligand (magenta line). For MD simulations, we used a revised model of the inactive form of hA₃R we have recently published,⁴⁰ generated using the multistate Alphafold 2(AF2) method^41,42 of hA₃R generated from GPCRdb web-tool;⁴³ the complexes of the starting structure (docking pose) and final snapshot from the MD simulations are available as pdb files (see https://github.com/annachor/inactive_A3R_AF2-carbonyloxycarboximidamides_MDs).

In this binding pose, the dichlorophenyl group in K18 is oriented toward transmembrane (TM) 5 and TM6, instead of TM1 and TM2, thus interacting with V169^5.30 and I249^6.54. This preference was increased with the number of chlorine atoms (as is reflected by the binding affinity constants of the compounds K18, K17, and K5(15,24,25) (Scheme 1). Compared to 1,3-thiazolyl in K17, the more basic 2-pyridinyl group⁴⁴ in K32 could form a stronger hydrogen bonding interaction with N250^6.55 leading to a ∼2-fold higher affinity of K32.^15,24,25 Replacement of isoxazole in K5 with a phenyl group in K39 reduced the binding affinity and replacement of thiazolyl or pyridinyl in K17 or K10, K11, and K32 by phenyl in K15 also reduce their binding affinity.^24,25

Based on these observations, we assumed that in compound 37 (Scheme 2), the combination of the 2,6-dichlorophenyl and the 2-pyridinyl groups at the ends of the carbonyloxycarboximidamide linker would enhance affinity. To quantify these predictions using SBDD, we performed TI/MD simulations with ff14sb³⁹ in heterocyclic carbonyloxycarboximidamide–hA₃R complexes embedded in 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC) bilayers. We used for the MD simulations a revised model of the inactive hA₃R generated based on the multistate AF2 method^41,42 that we recently published.⁴⁰ Interestingly, in the experimental structure of the active hA₃R (PDB IDs 8X16, 8X17(7)), the orientation of the M172^5.33, R173,^5.34 and M174^5.34 motifs, which we showed in ref (40) as important for the residence time of antagonists, matches the conformation we adopted in our revised⁴⁰ multi-AF2-based model for the inactive hA₃R (Figure S1), which we used for all the MD simulations in the present work. We performed the calculations using a 1-step protocol which changes the partial charges and the van der Waals interactions in a single simulation by activating both Lennard-Jones and Coulomb softcore potentials simultaneously.

Scheme 2. Chemical Structure of the 19 Heterocyclic Carbonyloxycarboximidamides 37–55 and of the 6 Cyclic Derivatives 56–61 Designed and Synthesized in the Hit-to-Lead Optimization as hA₃R Antagonists Based on the Binding Profile of K18/K32^15,24,25.

Indeed, our TI/MD calculations suggested that the alchemical transformation K18 → 37, where 1,3-thiazolyl is changed to 2-pyridinyl, was favored by a relative binding free energy (ΔΔG_b,TI/MD) equal to −0.77 ± 0.08 kcal mol^–1. Also, the alchemical transformation 44 → 37 that changes the 2-chlorophenyl to 2,6-dichlorophenyl was favored by ΔΔG_b,TI/MD = −1.83 ± 0.06 kcal mol^–1. A striking observation from our previous experiments²⁵ demonstrated that mutation of L90^3.32 in the low region of the receptor area, which makes a direct interaction with K18, and the remote L264^7.35 in the middle/upper region (both to alanine) increased the antagonistic affinity of K18. This suggests an available empty space in these regions of the orthosteric area that can be filled with a sizable hydrophobic group for increasing the ligand’s affinity. Based on the commercially available synthetic fragments, we tested the addition of a bromine atom at the 5-position of the 2-pyridinyl group in 37 so that it could fit into the bottom area of the receptor. Thus, using TI/MD calculations, we observed that the alchemical transformation 37 → 39 where a bromine group was added was favored by ΔΔG_b,TI/MD = −0.67 ± 0.04 kcal mol^–1. However, the addition of two chlorine atoms at 3,5-positions of the 2-pyridinyl group in 37 was not favored since for the alchemical transformation 37 → 38, we calculated ΔΔG_b,TI/MD = 0.58 ± 0.04 kcal mol^–1.

We performed additional TI/MD calculations to determine whether replacement of the 2-pyridinyl group in compound 37 by other groups could further improve affinity. For this purpose, we decided to introduce a methyl group at position 6- of the pyridine ring of 37 to obtain 40, or transfer its nitrogen atom to obtain the 3-pyridinyl analogues 46–48. In addition, we replaced the pyridine ring with the pyrimidin-2-yl moiety to obtain derivative 45, with the 3-tolyl group to obtain compound 49 or with a bulkier, nitrogen-containing, bicyclic system such as the indole or the purine ring, leading to compounds 52 and 53, respectively. Finally, we introduced a methylene linker between the carbonyloxycarboximidamide moiety and the pyridine ring to obtain analogues 54 and 55. In order to establish our previous observations,^15,24,25 with compounds K5, K17, and K18, which suggested that by increasing the number of chlorine atoms on the phenyl ring of the isoxazole group the binding activity was increased, we decided to study selected analogues bearing one chlorine atom or unsubstituted on this phenyl ring (compounds 41–44 and 50–51). Thus, we performed TI/MD binding free energy calculations for most of the above-mentioned alchemical transformations (Table S1). Even when the alchemical calculations did not suggest improved binding to hA₃R, we still felt that their synthesis was required since it would provide further experimental validation (using in vitro pharmacological techniques) of our model. We therefore synthesized compounds 37–55 (Scheme 2).

Furthermore, in compounds 56–61 (Scheme 2), we sought to investigate the effect on the antagonistic activity caused by the incorporation of the carbonyloxycarboximidamide pharmacophore between the 2,6-dichlorophenyl isoxazole and the aromatic nitrogen heterocyclic ring, into a rigid 1,2,4-oxadiazole ring. This transformation might be achieved through ring closure of the corresponding carbonyloxycarboximidamide analogues. The 500 ns MD simulations with ff19sb³⁹ suggested that compound 56, the cyclic analogue of compound 37, is stabilized inside the orthosteric binding area through hydrogen bonding interactions between the amide side chain of N250^6.55 and the pyridinyl and 4-oxadiazolyl nitrogen atoms.

Chemical Synthesis

For the synthesis of the target derivatives, a number of commercially available aryl or aralkyl carbonitriles were used, namely, pyridine-2-carbonitrile (1), 3,5-dichloropyridine-2-carbonitrile (2), 5-bromopyridine-2-carbonitrile (3), 6-methylpyridine-2-carbonitrile (4), pyrimidine-2-carbonitrile (5), 2-chloropyridine-3-carbonitrile (6), 6-methylpyridine-3-carbonitrile (7), pyridine-3-carbonitrile (8), 3-methylbenzolocarbonitrile (9), 2-(pyridin-2-yl)acetonitrile (10), 2-(pyridin-3-yl)acetonitrile (11), as well as 1-methyl-1H-indole-5-carbonitrile (13) that was prepared according to a published procedure upon methylation of 1H-indole-5-carbonitrile (12)⁴⁵ and 9-methyl-9H-purine-6-carbonitrile (17), which was prepared from 6-chloropurine (14) by means of tris(dibenzylideneacetone)dipalladium(0) [Pd₂(dba)₃)] and 1,1′-ferrocenediyl-bis(diphenylphosphine) [dppf] in dimethylacetamide (DMA) as depicted in Scheme 3.

Scheme 3. Preparation of Carbonitrile 17.

Reagents and conditions: (a) (i) 1.4 equiv of NaH, DMF dry, 0 °C, 1 h, (ii) 1.5 equiv of iodomethane, room temperature (rt), 20 h, 58% for 15 and (b) 0.11 equiv of Zn, 0.6 equiv of Zn(CN)₂, 0.02 equiv of Pd₂(dba)₃, 0.04 equiv of dppf, DMA, reflux, 3 h, 79%.

All the above-mentioned aryl or aralkyl carbonitriles were treated with hydroxylamine hydrochloride in the presence of a base or aqueous hydroxylamine and were converted to the aryl amidoximes 18–30 (Scheme 4Table S2). The amidoximes reacted with the acyl chlorides 34–36, easily prepared from the commercial carboxylic acids 31–33, respectively, to result in the corresponding target derivatives 37–55.

Scheme 4. Preparation of Carbonyloxycarboximidamides 37–55.

Reagents and conditions: (a) 2.9 equiv of HONH₂ (50 wt % solution in water), EtOH, reflux, 2 h or 1.5 equiv of HONH₂·HCl, 1.5 equiv of NaHCO₃, EtOH, rt 90 min and then reflux 2 h, 67–96%; (b) SOCl₂, reflux, 3 h; and (c) 1 equiv of amidoxime 18–30, 1.1 equiv of Et₃N, THF dry, rt, 2–16 h, 70–90% (for 2 steps).

Finally, the target compounds for which the acyloxy imidamide moiety has been incorporated into an 1,2,4-oxadiazole ring (56, 57, 58, 59, 60, and 61) were prepared upon treatment of the derivatives 37, 39, 45, 48, 47, and 55, respectively, with potassium hydroxide in anhydrous dimethyl sulfoxide (DMSO), through an intramolecular ring closure reaction, followed by dehydration of the resulting intermediate (Scheme 5).

Scheme 5. Preparation of 1,2,4-Oxadiazole Ring Closed Analogues 56–61.

Reagents and conditions: (a) 1 equiv of KOH, DMSO dry, rt, 30–45 min, 81–93%.

In Vitro Pharmacological Characterization

Quantifying the Binding Affinity and Kinetics of Potential Antagonists at hA₃R

We have previously reported the characterization of hit compound K18, a specific hA₃R (K_i < 1 μM) competitive antagonist with a new scaffold as hA₃R ligand.^15,25 Based on this scaffold, 25 analogues were synthesized to further investigate the SAR, aiming to develop A₃R antagonists with high affinity as well as high specificity. The compounds were grouped A–D based on their backbone (Tables 1, S3).

Table 1. Chemical Structure and Binding Affinity (pK_i^a) of the Seven Novel Heterocyclic Compounds, Carbonyloxycarboximidamides (Group A) or 1,2,4-Oxadiazole Derivatives (Group D), which Displayed Equal or Increased Affinity to Their Precursor K18.

^aAll the equilibrium binding affinities (pK_i) were determined with NanoBRET ligand binding assay and represented as mean ± standard error of the mean (SEM) of n independent repeats with experiment conducted in duplicates. Data of K18 was taken from ref (25). One-way ANOVA with Dunnett’s post-test was used to determine the statistical significance (*p < 0.05) compared to the pK_i of K18.

To assess the binding affinities of these compounds at hA₃R, NanoBRET-based ligand binding assays were used in a human embryonic kidney 293T (HEK293T) cell line stably expressed nanoluciferase (Nluc)-tagged hA₃R. The saturation binding affinity (pK_i) was determined by using nonfluorescent compounds to displace 5 nM fluorescent A₃R antagonist CA200645 at Nluc-hA₃R. According to the pK_i values determined (Figure 2 and Tables 1 and S3), there were two out of the 25 compounds that displayed significantly higher affinity, (37, pK_i = 7.33 ± 0.16, p = 0.0071; and 39, pK_i = 7.92 ± 0.06, p < 0.0001) to hA₃R compared to the reference compound K18 (pK_i = 6.92 ± 0.10, K_i ∼ 120 nM²⁵), while five others showed near-equal affinity, 40 (pK_i = 6.79 ± 0.11), 47 (pK_i = 7.19 ± 0.05), 48 (pK_i = 6.99 ± 0.04), 59 (pK_i = 7.11 ± 0.05), and 60 (pK_i = 7.23 ± 0.06) to K18. Among these seven compounds, 39 showed the highest affinity (K_i = 12.0 nM), which is about 10-fold higher than compound K18, followed by compound 37 (K_i = 46.8 nM), and the oxadiazole derivatives 60 (K_i = 58.9 nM) and 59 (K_i = 91.2 nM). Since we were aiming to develop hA₃R-selective antagonists with equivalent or higher affinity to K18, only these seven compounds were selected for further investigation.

Figure 2.

Binding affinity (pK_i) of 25 heterocyclic carbonyloxycarboximidamide analogues or derivatives at hA₃R determined in NanoBRET binding assay. 5 nM CA200645 was added to HEK293 cells stably expressing Nluc-hA₃R, interacting with Nluc and produce BRET signal. The BRET ratio values were baseline-corrected with the response induced by high-concentration (1 μM) A₃R antagonist MRS1220. Each data point represents the mean ± SEM of at least three experiments performed in duplicates. The pK_i values determined were compared with the pK_i of K18 previously determined in ref (25). One-way ANOVA with Dunnett’s post-test was used to determine the statistical significance (*p < 0.05) compared to the pK_i of K18 with black * indicating the affinity significantly higher and the gray * indicating the significantly lower one.

As we²⁵ and others⁴⁶ previously showed, NanoBRET ligand binding assay can also be used to determine the parameters of the ligand binding kinetics. We have previously determined the binding kinetic parameters of fluorescent ligand CA200645 with K_on(k₁) = 3.25 ± 0.03 × 10⁶ M^–1 min^–1 and K_off(k₂) = 0.019 ± 0.003 min^–1.²⁵ Using the “kinetics of competitive binding” model (built in Graphpad prism 9.3.1), we determined the K_on(k₃) and K_off(k₄) of the new seven compounds (Table 2). From these parameters, the RT was determined as 1/K_off (Table 2). As shown in Table 2, among these seven compounds, 39 showed largest K_on (5.95 ± 0.42 M^–1 min^–1) and the smallest K_off (0.046 ± 0.002 min^–1) and therefore longest residence time (22.1 ± 1.0 min), which contributed to its high binding affinities.

Table 2. Kinetic Parameters of Binding for the Seven High-Affinity Novel K18 Derivatives at hA₃R^a.

compound	Kon(k3) (x106)/M–1 min–1	Koff(k4)/min–1	RT/min
37	1.60 ± 0.34	0.075 ± 0.002	13.5 ± 0.4
39	5.95 ± 0.42	0.046 ± 0.002	22.1 ± 1.0
40	0.23 ± 0.05	0.057 ± 0.001	17.5 ± 0.2
47	1.49 ± 0.44	0.049 ± 0.004	21.3 ± 2.2
48	1.01 ± 0.20	0.054 ± 0.003	14.7 ± 3.4
59	0.65 ± 0.07	0.050 ± 0.004	20.5 ± 1.9
60	0.35 ± 0.06	0.102 ± 0.015	10.7 ± 2.0
MRS1220	325 ± 2.8b	0.025 ± 0.005b	40.32b

^aK_on(k₃) and K_off(k₄) for each compound determined using NanoBRET binding assay at Nluc-hA₃R and fitted with the “kinetics of competitive binding model”. RT as determined by 1/K_off.

^bIndicated values from ref (25).

Seven High-Affinity Antagonists Displayed High Selectivity for hA₃R over the Other AR Subtypes

To assess the competitive antagonistic action of the seven most potent compounds 37, 39, 40, 47, 48, 59, and 60 at all human AR subtypes, cAMP accumulation assays were performed in CHO-K1 cells stably expressing the individual receptors, hA₁R, hA_2AR, hA_2BR, or hA₃R. Increasing concentrations of the nonselective AR agonist 5′-N-ethylcarboxamidoadenosine (NECA) and 10 μM antagonist or DMSO control were coincubated for 30 min. Also, for the G_i/o-coupled hA₁R and hA₃R, 1 μM forskolin (a plant toxin that activates adenylyl cyclase independent of the G protein) was added to stimulate cAMP production. For the G_s-coupled hA_2AR and hA_2BR, forskolin was not added since the receptors are able to stimulate cAMP accumulation alone. As shown in Figure 3 and Table S4, among these 7 compounds, 37, 39, 40, 47, and 48 selectively antagonize hA₃R but not the other AR subtypes. Apart from strong potency at hA₃R, 60 showed weak antagonism at hA_2AR and hA_2BR, while 59 had weak antagonistic effects at hA₁R, hA_2AR, and hA_2BR.

Figure 3.

Characterization of the seven selected compounds at all human AR subtypes in cAMP accumulation assay. CHO-K1 cells stably expressing individual AR subtypes were treated with different concentrations of NECA or vehicle (V) and 1 μM forskolin in the case of G_i/o-coupled hA₁R and hA₃R or DMSO control in the case of G_s-coupled hA_2AR and hA_2BR, as well as 10 μM test compound (red) or DMSO control (blue) for 30 min. In hA_2AR and hA_2BR, cAMP response was normalized against the response induced by 100 μM forskolin, whereas in hA₁R and hA₃R, responses were represented as the percentage of the inhibition of cAMP response generated by 100 μM forskolin. Vertical arrow and horizontal arrow denote the significance of the change in efficacy and potency, respectively. One-way ANOVA was performed to compare the changes between DMSO only and the presence of tested compound (*p < 0.05). All values are represented as mean ± standard error of the mean (SEM), obtained in n = 3 independent experimental repeats, conducted in duplicates.

For the most potent compound 39, different concentrations of the antagonist were used at hA₃R to perform a full Schild regression analysis (Figure S2). From this analysis, the resulted estimation of antagonist affinity (pA₂) was 7.95 ± 0.15, and the Schild slope was found to be close to unity, indicating that 39 acts as a competitive antagonist at hA₃R.

Further Exploration of the Role of the 2,6-Dichlorophenyl Group in the Binding Affinity toward A₃R

All seven high-affinity antagonists shared a common structure of the 2,6-dichlorophenyl group at position 3 of the isoxazole ring, and the decrease of the number of chloro-substituents leads to reduction in binding affinity (compounds of groups B and C, Table S3). Therefore, we further explored the importance of these two chloro-groups by synthesizing the structural analogues of our most active lead compounds 37 and 39, where the two chloro-groups were replaced by either bromo- or methyl-groups, leading to the corresponding derivatives 74–77 (Scheme 6). The results from TI/MD calculations for the alchemical transformations 37 → 74 or 37 → 76 that change 2,6-dichlorophenyl to 2,6-dibromorophenyl or 2,6-dimethylphenyl were ΔΔG_b,TI/MD = −0.53 ± 0.04 or −0.64 ± 0.05 kcal mol^–1, respectively. The corresponding alchemical transformations for 39, 39 → 75, or 39 → 77 were ΔΔG_b,TI/MD = −0.45 ± 0.04 or −0.46 ± 0.04 kcal mol^–1, respectively. The calculations suggested a possible small improvement in binding affinity, although the error of the method is 1 kcal mol^–1 which corresponds to an ∼5-fold difference in binding affinity.

Scheme 6. Preparation of 2,6-Dibromo and 2,6-Dimethylphenylisoxazole Analogues 74–77.

Reagents and conditions: (a) 1.15 equiv of NH₂OH·HCl, 1.15 equiv of NaOH, EtOH, reflux, 2 h, 90–95%; (b) 1 equiv of NCS, DMF dry, rt, 2 h, 61–84%; (c) 1 equiv of methyl acetoacetate, 1 equiv of MeONa, MeOH, rt, 16 h, 70–72%; (d) 1.2 equiv of NaOH, MeOH·H₂O, 65 °C, 3 h, 77–83%; (e) SOCl₂, reflux, 3 h; and (f) 1 equiv of amidoxime 18 (for 74 and 76) or 1 equiv of amidoxime 20 (for 75 and 77), 1.1 equiv of Et₃N, THF dry, rt, 2 h, 90–97% (for 2 steps).

Nevertheless, we synthesized compounds 74–77 (Scheme 6) and then characterized these 4 new compounds using NanoBRET ligand binding assay at hA₃R to determine both their affinities and kinetic parameters (Tables 3 and 4). The synthesis of the novel derivatives 74–77 was performed starting from commercially available 2,6-disubstituted benzaldehydes 62 and 63 that were converted to the isoxazole methyl esters 68 and 69, respectively, upon oxime formation, chlorination with N-chlorosuccinimide (NCS), and subsequent ring closure of intermediates 66 and 67 with methyl acetoacetate.⁴⁷ The methyl esters 68 and 69 were hydrolyzed under basic conditions to afford the carboxylic acids 70 and 71 that were converted to the corresponding acyl chlorides 72 and 73. Finally, the latter were coupled with amidoximes 18 and 20 to afford the target derivatives 74–77 (Scheme 6). The substitution to bromine from chlorine significantly increased the affinity of 37 but reduced the affinity of 39, whereas the substitution to the methyl group reduced the affinities of both compounds significantly. However, these compounds still displayed equal or higher affinity when compared with K18. Also, we employed cAMP accumulation assay to study their subtype selectivity. These four compounds maintain the hA₃R specificity observed for their precursors 37 and 39 (Table S5 and Figure S3), all failing to antagonize hA₁R, hA_2AR, and hA_2BR. Finally, interspecies differences of adenosine A₃R are higher than the other ARs, and this results in the difficulties in developing A₃R antagonists which have cross-species activity.⁴⁸ For completeness, we examined the affinity of the 11 compounds 37, 39, 40, 47, 48, 59, 60, 74, 75, 76, and 77 at rat A₃R (rA₃R) using the NanoBRET ligand binding assays using Nluc-rA₃R. None of the compounds show affinity K_i > 1 μM at Nluc-rA₃R, suggesting that all are species selective (Figure S4).

Table 3. Chemical Structure and Binding Affinity of the Four Analogues Based on 37 and 39.

All the equilibrium binding affinities (pK_i) were determined using NanoBRET binding assay and represented as mean ± SEM of n independent repeats with experiment conducted in duplicates. Data of K18 was taken from ref (25). One-way ANOVA with Dunnett’s post-test was used to determine the statistical significance (*p < 0.05) compared to the pK_i of K18 (p1) and 37^# or 39^∧ (p2) as appropriate.

Table 4. Kinetics of Binding at Human A₃R for Four High-Affinity Compounds 74–77^a.

compound	Kon(k3) (x106)/M–1 min–1	Koff (k4)/min–1	RT/min
74	3.83 ± 0.54	0.092 ± 0.006	11.0 ± 0.7
75	2.41 ± 0.26	0.057 ± 0.006	18.4 ± 2.1
76	1.12 ± 0.23	0.081 ± 0.007	12.6 ± 1.1
77	1.51 ± 0.28	0.066 ± 0.002	15.1 ± 0.4

^aK_on(k₃) and K_off(k₄) for each compound determined using NanoBRET binding assay at Nluc-hA₃R and fitted with the “kinetics of competitive binding model”. RT as determined by 1/K_off.

As summarized in Table 5, the affinities of these 11 compounds determined from (a) the equilibrium binding affinity measured in NanoBRET assay (pK_i), (b) the kinetic dissociation constant (pK_d), and (c) the functional dissociation constant (pK_B/pA₂) using either the dose ratio eq 1 or for 39 the Schild regression analysis showed excellent agreement, confirming that these 11 compounds are high-affinity hA₃R-selective antagonist candidates, especially the lead compounds 39.

Table 5. Binding Affinities of 11 Compounds Determined from Different Assays at hA₃R Showed Good Agreement to Each Other.

compound	pKi (equilibrium)a	pKd (kinetics)b	pKB (cAMP)c
37	7.33 ± 0.06	7.30 ± 0.10	7.45 ± 0.19
39	7.92 ± 0.06	8.11 ± 0.04	#7.95 ± 0.15
40	6.79 ± 0.07	6.57 ± 0.09	7.05 ± 0.13
47	7.14 ± 0.04	7.43 ± 0.11	6.99 ± 0.18
48	7.01 ± 0.05	7.25 ± 0.08	7.09 ± 0.22
59	7.04 ± 0.08	7.11 ± 0.09	6.95 ± 0.13
60	7.23 ± 0.06	6.54 ± 0.09	6.89 ± 0.07
74	7.53 ± 0.06	7.62 ± 0.02	7.32 ± 0.05
75	7.56 ± 0.04	7.60 ± 0.12	7.39 ± 0.09
76	7.36 ± 0.07	7.34 ± 0.07	7.30 ± 0.14
77	6.90 ± 0.03	7.11 ± 0.10	6.55 ± 0.21

^aSaturation binding affinity (pK_i) determined through NanoBRET competition binding assay using the Cheng-Prusoff equation. Data taken from Tables 1 and 3.

^bKinetic dissociation constant (pK_d) for each compound determined from K_off/K_on in Tables 2 and 4.

^cDissociation constant (pK_B) as determined through cAMP response using dose ratio equation for each compound or ^# Schild analysis for 39.

Binding Profile Investigation

MD Simulations

To investigate the binding profile of the best antagonists shown in Scheme 2 at hA₃R, we performed MD simulations with representative compounds, i.e., 37, 38, 39, and 56 in complex with hA₃R embedded in POPC bilayers. We used a multistate AF2 method^41,42 of hA₃R generated with the GPCRdb⁴³ web-tool. We optimized this model of hA₃R in ref (49) to achieve best agreement with thermodynamic and kinetic data. In the starting docking pose of these compounds, the dichlorophenyl group was oriented toward TM5 and TM6 and is the same as we previously observed for K18 and K32 (compound 44 in this study).^15,24,25

Using 500 ns MD simulations of the antagonists 37 and 39 in complex with hA₃R, we were able to reveal the important residues in the orthosteric binding area essential for 37 and 39 binding (Figure 4). We observed that with compound 37, its dichlorophenyl group orientates toward TM5 and TM6 to form dispersion interactions interacting with V169^5.30, M172^5.33, and I249^6.54, F182^5.43, and I253^6.58. Moreover, the isoxazole forms aromatic π–π stacking interaction with the phenyl group of F168^5.29 (Figure 4A,B). The amide side chain of N250^6.55 is suggested to form a bidentate hydrogen bond with the carboximidamide amino group, while the 2-pyridinyl nitrogen can form water-bridged interactions with N250^6.55. Nitrogen and oxygen atoms of isoxazole can form hydrogen bonds with the NH groups of F168^5.29 or V169^5.30, and the carboximidamide carbonyl group is suggested to form water-bridged interactions with H272^7.43. There are also other water-bridged interactions between the isoxazol-4-carbonyloxycarboximidamide and example V72^2.64, F168^5.29, M172^5.33, L246^6.51, and I268^7.39. The 2-pyridinyl group forms dispersion interactions with L90^3.32, L91^3.33, T94^3.36, M177^5.38, F182^5.43, W243^6.48, L246^6.51, and I253^6.58. The methyl group in the isoxazole ring can interact with V72^2.64, Y15^1.35, A69^2.61, V72^2.64, L264^7.35, I268^7.39, and H272^7.43 through dispersion interactions. By modifying 37 into 39 through a 5-bromo substitution to the 2-pyridinyl group, we observed that 39 is tilted from TM7 and ΤΜ4 to TM5 showing additional dispersion interactions with M177^5.38, V178^5.39 and loosing interactions with F168^4.52, L264,^7.35 and H272^7.43 (Figure 4D,E). Significantly, compared to 37, 39 moves deeper into the bottom of the binding area forming hydrogen bonding interactions with T87^3.29. In 38, the 3,5-dichloro substitution in the 2-pyridinyl group diminishes affinity by ∼316-fold, and the 500 ns MD simulations suggested that this ligand escaped from the binding area (Figure S5A–C). We further expanded the MD simulation analysis in the oxadiazole series 56–61 by studying compounds 56, 57, and 60 (see Discussion in the Supporting Information and Figure S6).

Figure 4.

500 ns MD simulations for the complex of compounds of 37 and 39 with the WT hA₃R using the amber ff19sb.³⁹ (Α,D) Representative frame of the ligand inside the orthosteric binding area. (B,E) Receptor–ligand interaction frequency histograms; bars are plotted only for residues with interaction frequencies ≥0.2. Color figure in frames or bar plots: ligand is shown with pink sticks and ligand’s starting position with a pink wire, receptor is shown with a white cartoon and sticks, hydrogen bonding interactions are shown with yellow dashes or bars, π–π interactions are shown with green dashes or bars; hydrophobic interactions are shown with gray bars; and water bridges are shown with blue bars. (C,F) rmsd plots of Ca carbons of the protein (gray line) and of heavy atoms of the ligand (magenta line). For MD simulations, we used a revised model of the inactive form of hA₃R we have recently published,⁴⁰ generated using the multistate AF2 method^41,42 of hA₃R generated from GPCRdb web-tool;⁴³ the complexes of the starting structure (docking pose) and final snapshot from the MD simulations are available as pdb files (see the Ancillary Information).

TI/MD Calculations

MD simulations can describe qualitatively SARs based on the inspection of MD simulation trajectories and protein–ligand interaction frequency plots. While, the TI/MD simulations can accurately calculate the changes in binding affinity between different substituents. The set of the studied compounds K18, 37–42, 44–46, 48, 52, 54, 55, and 74–77 display 3 orders of magnitude differences between their affinity range. The results from the TI/MD alchemical calculations for 23 pairs of ligands in complex with hA₃R using our optimized model of hA₃R are shown in Table S1. As is shown in Figure 5, the correlation coefficient between the TI/MD calculated relative binding free energies and experimental values was r = 0.68 (p = 0.026) with mean unsigned error (MUE) = 0.89 kcal mol^–1 (Table S1).

Figure 5.

Computed ΔΔG_b,_TI/MD values plotted against ΔΔ_Gb,exp values estimated by the experimental binding affinities pKi (Table S1) for hA3R using NanoBRET binding assay; r: correlation coefficient, s: slope. For TI/MD simulations, we used a revised model we recently published⁴⁰ of the inactive form of hA₃R generated based on the multistate AF2 method^41,42 of hA₃R.

Mutagenesis Studies of Compound 39 at hA₃R in Comparison with 37

Based on the interaction frequency of the amino acid residues in the orthosteric binding area in contact with antagonist, (Figure 4B,D) predicted from the 500 ns MD simulations, we next employed mutagenesis (alanine substitution except where alanine was present, then glycine was used) with NanoBRET binding assay to experimentally investigate residues that were suggested to be important for the binding of 37 and 39. First, since mutation of residues in GPCRs can have a detrimental effect on receptor trafficking, we initially determined the cell–surface expression of WT and mutant Nluc-hA₃R using fluorescence-activated cell sorting (FACS) in flow cytometry^50,51 and presented as % WT (Table S6). 4 out of 24 of the hA₃R mutants showed significantly reduced cell surface expression (compared with the WT). Conversely, the hA₃R mutants V72^2.64A, F168^5.29A, and F182^5.43A displayed increased cell surface expression. Despite these changes in the cell surface expression levels, all the mutants expressed sufficiently to enable NanoBRET ligand binding experiments to be formed. The equilibrium dissociation constant (K_d) of the fluorescent ligand CA200645 was determined for each mutant and the WT hA₃R—all expressed individually in HEK293T cells (Table S6). While the affinity of CA200645 remained unchanged for most of the hA₃R mutants, Y15^1.35A, L91^3.33A, M172^5.33A, M177^5.38A, L246^6.51A, and I268^7.39A did display significantly lower affinity for CA200645 with the worst L246^6.51A (K_i ∼ 258 nM) being about 10-fold lower when compared with the WT (∼24 nM). For hA₃R, F168^5.29A, N250^6.55A, and H272^7.43A, CA200645 was unable to bind, so these three could not be further investigated in competition binding assays. Importantly, the reductions in the binding affinity of CA200645 did not correlate with the changes in the cell surface expression, indicating the sufficient expression of all mutants.

In the NanoBRET competition binding assays, 37, 39, and agonist NECA were tested at WT and the 21 mutant hA₃R, and the mutational effect was represented as the change in affinity compared with the WT (ΔpK_i) (Table S6 and Figure 6). There were 8 mutants (Y15^1.35A, L91^3.33A, M172^5.33A, M177^5.38A, V178^5.39A, F182^5.43A, L246^6.51A, and I268^7.39A) which showed significantly reduced affinity for both 37 and 39, while V65^2.57A and L90^3.32A displayed increases in affinity for both antagonists. This indicated the importance of these residues in composing the orthosteric binding pocket of 37 and 39. Noteworthy, while the effects of mutating the residues Y15^1.35, L91^3.33, M172^5.33, M177^5.38, L246^6.51, and I268^7.39 to alanine was the same between 37, 39, and CA200645, the residues V178^5.39 and F182^5.43 appear to be unique for 37 and 39 forming dispersion interactions with the 2-pyridinyl group of these antagonists.

Figure 6.

Changes in the binding affinities for compounds 37 and 39 and NECA measured using NanoBRET binding against WT and mutants hA₃R. The binding affinity of 37 and 39 and NECA at hA₃R WT and mutants were determined using NanoBRET binding assay performed in HEK293T transiently transfected with each construct. The change in affinity (ΔpK_i) is calculated as the difference of pK_i between the mutant and WT. Data is represented as mean ± SEM of n = 3 independent repeats conducted in duplicates. Statistical significance (*p < 0.05) compared with WT was determined using one-way ANOVA with Dunnett’s post-test.

Also, 37 and 39 displayed different extents of changes in the affinities at several mutants which might explain the increased affinity of 39 when compared to 37. Compound 39 showed greater reductions in the affinity at L91^3.33A and L246^6.51A (∼97-fold and ∼115-fold, respectively) when compared with 37 (∼58-fold and ∼22-fold, respectively). This correlates well with the simulation results (the additional Br going deeper into the binding pocket composite of L91^3.33A and L246^6.51A and therefore increasing the affinity). Additionally, W243^6.48A showed significant reduction (p = 0.0006) in the affinity of 39 but not of 37 (p = 0.2192) when compared with the WT, indicating that there is a stronger π–π interaction of 39 with the extra bromine. On the other hand, I253^6.58A at the top of the binding pocket had significant reduced affinity for 37 (p = 0.0030) but not for 39 (p = 0.7289), suggesting that 37 interacts more with the top of the binding pocket than 39.

For the completeness of the study, we have also assessed the binding affinity of agonist NECA at all the hA₃R mutants and compared with WT. Among the 21 mutants tested, there were 12 mutants previously studied with NECA using cAMP accumulation assay.⁵² The mutational effects of these mutants in the potency of cAMP responses induced by NECA showed a good agreement with their effects at the change of binding affinity of NECA in this study. For example, T94^3.36A, L246^6.51A, I268^7.39A, and M177^5.38A showed significant reduction in the binding affinity of NECA, while M177^5.38A also showed reduction in the cAMP response potency, and the other three mutants showed no cAMP response. Also, the agonist NECA and the antagonists 37 and 39 showed differences in the change of affinity at several mutants. 37 and 39 showed increased affinity at both V65^2.57A and L90^3.32A while NECA displayed no changes. Also, NECA had higher affinity at I249^6.54A and lower affinity at V72^2.64A and T94^3.36A, while 37 and 39 had no change at these mutants. The agreement between two studies with different types of pharmacological assays and the ability to reveal the differences between the binding pockets of agonist and antagonisshowed the robustness of the mutagenesis study.

Disease Model Validation and In Vitro Pharmacokinetic Profiling

Validation of Compound Selectivity at Endogenously Expressed Receptors in a Disease Model of Lung Cancer

We next sought to confirm that the lead compounds, 37 and 39, were able to selectively antagonize hA₃R when endogenously expressed. We utilized LK-2 and NCI-H1792 cells, which are both nonsmall cell lung carcinoma cells. LK-2 cells display low hA₁R and hA_2BR expression, whereas NCI-H1792 cells express all the 4 adenosine receptor subtypes.³⁵ Neither N⁶-cyclopentyl-adenosine (CPA; A₁R selective agonist) nor 1-deoxy-1-[6-[[(3-iodophenyl)methyl]amino]-9H-purin-9-yl]-N-methyl-β-D-ribofuranuronamide (IB-MECA, A₃R selective agonist) was able to inhibit forskolin-mediated cAMP accumulation in LK-2 cells, consistent with their low adenosine receptor expression (Figure S7). CPA inhibited forskolin-mediated cAMP production in NCI-H1792 cells (pEC₅₀ 6.51 ± 0.63), but neither 37 nor 39 was able to antagonize the response (pEC₅₀ values of 6.83 ± 0.52 and 7.13 ± 0.60, respectively) (Figure 7 and Table S7). The response to IB-MECA was more potent (pEC₅₀ of 8.50 ± 0.32), with both 37 and 39 significantly reducing the potency of the response (6.69 ± 0.57 and 6.24 ± 0.42, respectively). Calculated pK_B values were similar to those measured in the heterologous expression system (7.17 ± 0.13 and 7.46 ± 0.20). This confirmed the ability of both compounds to selectively antagonize hA₃R when endogenously expressed alongside other adenosine receptors.

Figure 7.

Selective inhibition of hA₃R in nonsmall cell lung carcinoma cells inhibits proliferation. (A) Inhibition of forskolin-mediated cAMP accumulation in NCI-H1792 cells in response to CPA or IB-MECA, costimulated with DMSO (blue circles) or 10 μM 37 or 39 (red squares). (B) pEC₅₀ and E_max values for the inhibition of LK-2 and NCI-H1792 cell proliferation for 37 and 39. Statistical significance determined using an unpaired Student’s t-test.

hA₃R has often been referenced as a promising target for treating cancer.³⁴ We therefore aimed to see the effect of selectivity antagonizing the receptor on cell proliferation. Compound 37 significantly impaired proliferation, displaying toxicity even in the absence of A₃R (Figure S7 and Table S8). While 39 still reduced proliferation in LK-2 cells, showing some nonspecific toxicity, the effect was increased in NCI-H1792 cells, with an increase in potency and an increase in the percentage of cell death (Figure 7).

Pharmacokinetic Assessment of Lead Compound 39

The highest affinity compound 39 was evaluated in the in vitro pharmacokinetic study including solubility, absorption, and metabolism (Table 6). For the solution properties, 39 showed low aqueous solubility with a mean solubility of 0.10 μM in phosphate buffer solution (PBS), 0.16 μM in simulated gastric fluid, and 43.13 μM in simulated intestinal fluid. It displayed a comparable partition coefficient (Log D = 2.82) with the reference compounds haloperidol (Log D = 2.49) and phenytoin (Log D = 2.28), showing adequate lipophilicity. Compound 39 also showed high protein binding in human plasma, with 99% of the proteins bound, which is similar to that of the reference compounds sertraline (99%) and warfarin (95%).

Table 6. In Vitro Pharmacokinetic Profile of Compound 39 and Reference Compounds.

solution properties
aqueous solubilitya
		39	diethyl stilbestrol	disulfiram
	In PBS, pH 7.4 (μM)	0.10	5.22	45.64
	in simulated gastric fluid (μM)	0.16	N.D.	N.D.
	in simulated intestinal fluid (μM)	43.13	N.D.	N.D.
partition coefficientb (n-octanol/PBS, pH 7.4)
		39	haloperidol	phenytoin
	Log D	2.82	2.49	2.28
protein bindingc
		39	sertraline	warfarin
	% protein bound	99	99	95
	% recovery	105	94	112
In Vitro Absorption
permeability in Caco-2 celld
		39	propranolol	labetalol
	PappA-B (10–6 cm/s)	7.2	32.2	8.3
	PappB-A (10–6 cm/s)	0.6	32.6	38.2
	% recovery (A-B)	26	74	99
	% recovery (B-A)	19	96	103
	uptake ratio	12	0.99	0.22
In Vitro Metabolism
intrinsic clearance (human liver microsomes)e
		39	terfenadine	verapamil
	t1/2 (min)	9.5	16.9	26.6
	CLint (μL/min/mg of microsomes)	729	410.1	261.1

^aAqueous solubility (μM) in PBS at pH 7.4/simulated gastric fluid/simulated intestinal fluid determined with high-performance liquid chromatography–ultraviolet spectroscopy.

^bThe partition efficient of the compound between n-octanol and PBS at pH7.4, measuring the lipophilicity of the tested compound. Log D was calculated as Log10(the amount of compound in n-octanol/the amount of compound in PBS).

^cMeasure of percentage of protein binding and percentage of compound recovery during the assay determined in equilibrium dialysis using human plasma.

^dThe permeability of compounds assessed in bidirectional Caco-2 cell permeability assay with pH = 6.5 for donor chamber and pH = 7.4 for receiver chamber. The extent of permeability is measured as apparent permeability coefficient (P_app) from apical (A) to basolateral (B) or in reverse direction. The percentage recovery of the compound is calculated as the total amount of compound in the donor and the receiver at the end of experiment/the amount of initial compound present. The uptake ratio of the compound is calculated as P_appA-B/P_appB-A.

^eThe metabolic stability of compounds was determined in 0.1 mg/mL human liver microsomes, measured as the half-life (t_1/2) and apparent intrinsic clearance (CL_int).

The in vitro absorption of 39 was assessed using the bidirectional permeability assay (P_app) in Caco-2 cells. Compound 39 had a P_app of 7.2 × 10^–6 cm/s from the apical side (A) to the basolateral side (B) and a P_app of 0.6 × 10^–6 cm/s from B to A, resulted in an uptake ratio (P_appA-B/P_appB-A) of 12, suggesting no drug efflux across the membrane occurred. However, the percent recovery was lower (26% A to B and 19% B to A) compared to the reference compound in both directions (74% A-B and 96% B-A for propranolol), indicating potential problems like poor solubility or nonspecific binding during the assay.

The in vitro metabolism of 39 was assessed as its intrinsic clearance in human liver microsomes. Compound 39 had a shorter half-life (9.5 min) than the precursor compound K18 (24 min²⁵) as well as the reference compounds (terfenadine, 16.9 min and verapamil 26.6 min). The resulted intrinsic clearance (CL_int) was 729 μL/min/mg of microsome, which was higher than K18 (287.2 μL/min/mg of microsome²⁵) terfenadine (410.1 μL/min/mg of microsome) and verapamil (261.1 μL/min/mg of microsome). These indicated that 39 has a relatively rapid metabolism in hepatic microsomes and therefore will cause less accumulation and potential toxicity.

Discussion

Here, we have reported a hit-to-lead study by computational design, synthesis, and screening using NanoBRET binding assay of 25 novel derivatives targeting selectively hA₃R, as analogues or derivatives of the selective A₃R carbonyloxycarboximidamide heterocyclic hits K17, K18,¹⁵ or K32(24,25) from a previous structure-based VS.^15,24,25 In these previously reported hits, the carbonyloxycarboximidamide moiety was connected with a 3-(2-chlorophenyl)-isoxazolyl, 3-(2,6-dichlorophenyl)-isoxazolyl or 3-(2-chlorophenyl)-isoxazolyl group, respectively, and the carboximidamide carbon was connected with the 1,3-thiazolyl or 2-pyridinyl group.^15,24,25 Their NanoBRET-based determined dissociation constants for K17, K32, and K18 were K_i = 600, 250, and 120 nM, respectively, showing the importance of attaching two ortho-chlorine substituents in the phenyl group attached to isoxazole ring and that as regards affinity and antagonistic potency 2-pyridinyl (K32) > 4-(1,3-thiazolyl) (K17) > 3-pyridinyl (K10) > 4-pyridinyl (K11).^24,25 However, the residence time of these hits was low and unmeasurable (<1 min).²⁵

We used, for the SBDD of the novel carbonyloxycarboximidamide derivatives, the same binding pose of K18 (or K32) in the orthosteric binding area of hA₃R. This binding pose was previously suggested¹⁶ after refinement of the docking pose of K18 from VS¹⁵ using 100 ns MD simulations with the OPLS2005³⁷ force field. The binding pose in K18(14) was further confirmed with a combination of 500 ns MD simulations using the ff14sb and alanine mutagenesis supported by affinity experiments.^24,25 In K18 (or K32), the phenyl group that is attached at the 3-position of the isoxazole ring is oriented to the upper region of the receptor, while the 1,3-thiazolyl or 2-pyridinyl group was oriented deep in the receptor. There is an increasing propensity of the phenyl group to turn toward TM5 and TM6 by increasing the chlorine substituents attached to the available ortho positions of the phenyl group junction. Based on these observations, we used alchemical binding free energy calculations applied with the TI/MD and ff19sb simulations in complexes of ligands with a revised model for hA₃R we recently published⁴⁰ generated from the multi-AF2-based model for inactive hA₃R (Table S1) to design and synthesize more potent compounds than K18 (or K32). At the time we submitted this paper, the structure of the active state of hA₃R in complex with agonists and Gi heterotrimer was reported.⁷ Interestingly, in the experimental structure, the orientation of R173^5.34 matches the conformation we used in our revised multi-AF2-based model for the inactive hA₃R that we use for all the MD simulations performed here.⁴⁰ The purchase of compounds like K18 (or K32) for testing^15,24,25 from companies was costly, and the synthesis of these series with sufficient diversity in structure for exploring SARs was accomplished simply and in good yields using commercially available carbonitriles (Schemes 3 and 4). The heterocyclic carbonyloxycarboximidamides 37–55 and 74–77, and their structurally related 1,2,4-oxadiazole derivatives 56–61, provide novel and highly selective hA₃R antagonists, which are synthetically feasible and easily accessed drug molecules amenable for further development.

Among the 29 synthesized compounds (belonging in groups A–F; Tables 1, 3, and S3), there are three groups of compounds (A–C; Tables 1 and S3) which only differ by the number of chloro-substituents in the phenyl ring of 3-phenyl-isoxazole, (a) 37, 42, and 44; (b) 40, 41, and 43; and (c) 49, 51, and 50. By comparing their binding affinities at hA₃R, we showed that the affinities of antagonists increased as the number of chloro-substituents increased, which is in line with our findings.^15,24,25 In agreement with our previous studies,²⁵ the addition of chlorine atoms at the ortho positions of the phenyl group in the 3-phenyl-isoxazole moiety enhanced affinity since the dichlorophenyl group enhances the hydrophobic interactions toward TM5 and TM6 with residues V169^5.30 and I249^6.54. This was shown in the relative binding free energy values which are ΔΔG_b,exp = 1.83 ± 0.06 kcal mol^–1 and ΔΔG_b,TI/MD = 0.64 ± 0.04 kcal mol^–1 with a deviation of 1.19 kcal mol^–1 for 37 → 44, ΔΔG_b,exp = 0.75 ± 0.05 kcal mol^–1 and ΔΔG_b,TI/MD = 0.98 ± 0.04 kcal mol^–1 with a deviation of 0.23 kcal mol^–1 for 42 → 44, ΔΔG_b,exp = −1.08 ± 0.04 kcal mol^–1 and ΔΔG_b,TI/MD = −1.33 ± 0.03 kcal mol^–1 with a deviation of 0.25 kcal mol^–1 for 42 → 37, and ΔΔG_b,exp = −1.01 ± 0.06 kcal mol^–1 and ΔΔG_b,TI/MD = −0.47 ± 0.05 kcal mol^–1 with a deviation of 0.54 kcal mol^–1 for 41 → 40.

Replacement of 2,6-dichlorophenyl group attached to the isoxazole ring with 2,6-dibromophenyl or 2,6-dimethylphenyl in 37 or 39, respectively, resulted in compounds 74, 76 or 75, 77, respectively (groups E and F; Table 3). Compared to 37, the binding affinity of 74 was improved by 1.7-fold; conversely, it was reduced by 2.7-fold for 76. Interestingly, the modifications on both 75 and 77 resulted in reduced affinities at hA₃R compared to 39. This was shown in the relative binding free energy values which are for 37 → 74 ΔΔG_b,exp = −0.14 ± 0.05 kcal mol^–1 and ΔΔG_b,TI/MD = _– 0.53 ± 0.04 kcal mol^–1 with a deviation of 0.39 kcal mol^–1, for 37 → 76 ΔΔG_b,exp = 0.79 ± 0.05 kcal mol^–1 and ΔΔG_b,TI/MD = _– 0.64 ± 0.05 kcal mol^–1 with a deviation of 1.43 kcal mol^–1; for 39 → 75, we calculated ΔΔG_b,exp = 0.57 ± 0.07 kcal mol^–1 and ΔΔG_b,TI/MD = −0.45 ± 0.04 kcal mol^–1 with a deviation of 1.02 kcal mol^–1 and for 39 → 77 and ΔΔG_b,exp = 0.81 ± 0.08 kcal mol^–1 and ΔΔG_b,TI/MD = −0.46 ± 0.05 with a deviation of 1.27 kcal mol^–1.

Within group A, all compounds having a dichlorophenyl group attached at the 3-isoxazole position only differ by the aryl group attached to the carbonyloxycarboximidamide moiety, when compared to K18. This aryl group showed crucial effects on the binding affinity. Replacement of 2-pyridinyl group in 37 (Ki = 34.7 nM) with the less basic by ∼10³-fold 1,3-thiazolyl group in K18 (Ki ∼ 120 nM) or the less basic by ∼10⁴-fold 2-pyrimidinyl group in 45 (K_i ∼ 537 nM) led to a reduced affinity by 3.5-fold or ∼15.5-fold due to the stronger hydrogen bond that can be formed between the N250^6.55 amido side chain and 2-pyridinyl group.⁴⁴ The corresponding relative binding free energy values for K18 → 37 are ΔΔG_b,exp = −0.77 ± 0.08 kcal mol^–1 and ΔΔG_b,TI/MD = −1.51 ± 0.08 kcal mol^–1 with deviation |ΔΔG_b,TI/MD - ΔΔG_b,exp| = 0.74 kcal mol^–1 and for 37 → 45 are ΔΔG_b,exp = 1.69 ± 0.05 kcal mol^–1 and ΔΔG_b,TI/MD = 0.50 ± 0.04 kcal mol^–1 with a deviation of 1.19 kcal mol^–1.

It is worth noting that the five compounds with high affinities in this group all have a nitrogen atom at a similar position as the nitrogen in the thiazole ring of K18, see compound 37, which is in line with our findings.^15,24,25 Changes in this nitrogen position reduce the ability of aryl nitrogen to form a hydrogen bond with N250^6.55 amido side chain. Thus, compared to the 2-pyridinyl analogue 37 (Ki = 34.7 nM), the 3-pyridinyl analogue 48 (K_i ∼ 100 nM) had ∼3-fold lower affinity similar to previous observations for the monochloro derivatives K32 and K10 (Scheme 1). The corresponding relative binding free energy values for 37 → 48 were ΔΔG_b,exp = 0.67 ± 0.05 kcal mol^–1 and ΔΔG_b,TI/MD = 2.14 ± 0.08 kcal mol^–1 with a deviation 1.47 kcal mol^–1. Other changes that reduce the ability of pyridinyl nitrogen to form a hydrogen bond with N250^6.55 amido side chain are discussed below. Replacement of 2-pyridinyl in 37 with 2-pyridinylmethylene in 54 (K_i ∼ 316 nM) drives nitrogen away from N250^6.55 and causes reduction in binding affinity, correspondingly, by ∼9-fold. The corresponding relative binding free energy values for 37 → 54 were ΔΔG_b,exp = 1.36 ± 0.04 kcal mol^–1 and ΔΔG_b,TI/MD = 2.55 ± 0.06 kcal mol^–1 with a deviation of 1.19 kcal mol^–1. For the same reason is observed reduction in binding affinity by ∼8.7-fold when the 3-pyridinyl group in 48 (K_i ∼ 102 nM) is changed to the 3-pyridinyl methylene group in 55 (K_i∼ 891 nM). Compared to 48 (K_i ∼ 102 nM) bearing the 3-pyridinyl group, compound 46 (K_i ∼ 2089 nM) with the 2-chloro-3-pyridinyl group has 20.5-fold lower binding affinity since the 2-chloro-substituent hampers the hydrogen bonding interaction of 3-pyridinyl nitrogen. This was shown in alchemical perturbation 48 → 46 ΔΔG_b,exp = −1.33 ± 0.05 kcal mol^–1 and ΔΔG_b,TI/MD = −2.64 ± 0.11 kcal mol^–1. The 3-tolyl group in 48 (Ki ∼ 209 nM) does not form a hydrogen bond, and compared to 3-methyl-2-pyridinyl in 40 (Ki ∼ 100 nM), its binding affinity was reduced by 2.1-fold; the corresponding alchemical transformations 55 → 48 showed ΔΔG_b,exp = −1.33 ± 0.05 kcal mol^–1 and ΔΔG_b,TI/MD = −2.64 ± 0.11 kcal mol^–1 with a deviation of 1.31 kcal mol^–1 and 40 → 49 showed ΔΔG_b,exp = 0.30 ± 0.08 kcal mol^–1 and ΔΔG_b,TI/MD = 1.60 ± 0.09 kcal mol^–1 with a deviation of 1.30 kcal mol^–1. Interestingly, while the addition of methyl group at α-position to nitrogen increases pyridine basicity by 5.4-fold,⁵³ the 6-methyl-2-pyridinyl derivative 40 (K_i∼ 128 nM) has a 3.7-fold lower affinity compared to 37. Therefore, we assumed that the methyl group points toward an area of hA₃R where the water density is increased. Previously, we suggested⁶⁰ that such an area may reside between a polar substituent of the ligand and TM2 and TM3. The corresponding relative binding free energy values for 37 → 40 were ΔΔG_b,exp = 0.81 ± 0.06 kcal mol^–1 and ΔΔG_b,TI/MD = 0.57 ± 0.04 kcal mol^–1, with a deviation of 0.24 kcal mol^–1.

Compound 39 (K_i ∼ 11.7 nM) displayed about a 10- or 3-fold increase in affinity compared to K18 (K_i ∼ 120 nM) or 37 (K_i = 34.7 nM), respectively, whereas 38 (K_i ∼ 1072 nM) was significantly (p < 0.0001) lower in affinity. Among these high-affinity candidates, K18, K38, and K39, compound 39 (the most potent) uniquely has a bromine group in the 2-pyridinyl group at position 5. This brings a 3.9-fold increase in affinity when compared with 37 which only has the pyridinyl group instead since the bromine group at 5-position of the 2-pyridinyl group forms dispersion interactions with W243^6.48 and F182^5.43. For the alchemical transformation 37 → 39 in which a bromine group is added at 5-position of the 2-pyridinyl group in 37, ΔΔG_b,exp = −1.33 ± 0.05 kcal mol^–1 and ΔΔG_b,TI/MD = −0.67 ± 0.04 kcal mol^–1 with a deviation of 0.34 kcal mol^–1. Compared to 37, the 3,5-dichloro-2-pyridinyl group in 38 (K_i ∼ 1072 nM) reduces the affinity by ∼31-fold since the 3-chlorine atom is not favored lying in an area where the water density is increased. Indeed, for alchemical perturbations, 37 → 39 ΔΔG_b,exp = −0.67 ± 0.07 kcal mol^–1, ΔΔG_b,TI/MD = −1.01 ± 0.04 kcal mol^–1, and 37 → 38 ΔΔG_b,exp = 2.11 ± 0.04 kcal mol^–1, ΔΔG_b,TI/MD = 0.58 ± 0.04 kcal mol^–1. Additionally, both 37 and 39 had increased RT (14 and 22 min, respectively) inside the orthosteric binding pocket when compared to K18 or K32 (<1 min).

Moreover, a rigidification of the carbonyloxycarboximidamide moiety led to the oxadiazole derivatives with 59 and 60 having slightly higher affinities compared to their precursors 48 and 47. The 500 ns MD simulations showed that rigidification of the carbonyloxycarboximidamide moiety to oxadiazole derivatives 56–61 (class D, Tables 1 and S3) causes reorientation of the dichlorophenyl group through rotation around the oxazole-oxadiazole C–C bond as we showed for compounds 56, 57, and 60 (see Figure S6). Thus, the dichlorophenyl group faces TM2 in compounds 56 and 57 and TM7 in compound 60. Between compounds 56 and 61 (class D, Tables 1 and S3), the 3-pyridinyl derivative 59 is a stronger binder by 2-fold compared to 2-pyridinyl derivative 56, while for the acyloxyimidamide 2-pyridinyl derivative, 37 has 2-fold higher affinity compared to 48. Compared to 59 (K_i = 91.2 nM and RT = 20.5 min), the methyl substituent at 4-position of the 3-pyridinyl group increased the affinity in 60 (K_i = 58.9 nM and RT = 10.7 min) since is oriented toward M243^6.48 favoring hydrophobic interactions (see Figure S6,G–I), providing an additional lead for further improvement; the TI/MD results for this series are discussed in the Supporting Information. However, in functional cAMP experiments, we observed that they lost their hAR subtype selectivity. This contrasted with compounds 37, 39, 40, 47, 48, and 74–77 which were shown to display high hA₃R selectivity. Moreover, none of the compounds tested showed high (K_i > 1 μM) affinity toward rA₃R.

We previously determined the effects of receptor mutation on antagonist potency (pA₂) for A₃R L90A^3.32, V169^5.30A/E, M177^5.40 A, I249^6.54A, and L264^7.34A by functional assays.²⁵ We reported^24,25 that M177^5.38A caused the most significant reduction on the K18 antagonist effect. Interestingly, we found that L90A^3.32 in the low region and L264A^7.34 in the middle/upper region increased K18 potency, while I249^6.54A had little effect (compared to WT hA₃R).²⁵ We suggested that V169^5.30 was not a selectivity filter for hA₃R agonists or antagonists. Here, we also showed that^32,44,45 V169^5.30A did not cause a significant change in affinity for both 37 and 39 (Table S6). Moreover, we did observe significant reductions in affinity for both 37 and 39 with 8 of our mutants hA₃R (Y15^1.35A, L91^3.33A, M172^5.33A, M177^5.38A, V178^5.39A, F182^5.43A, L246^6.51A, and I268^7.39A), while 2 other mutations (V65^2.57A and L90^3.32A) produced an increase in affinity for both antagonists 37 and 39, suggesting that these 10 residues comprise the orthosteric binding pocket (Table S6).

However, there are also differences in binding profiles between compounds 37 and 39. Indeed, 39 showed a greater reduction in the affinity at L91^3.33A (∼97-fold) and L246^6.51A (∼115-fold) when compared with 37 (∼58-fold and ∼22-fold, respectively). The effects of the Y15^1.35A, Y265^7.36, or W243^6.48A mutations were also more significant for 39 than for 37, which showed excellent agreement with our simulation results. It appears that the sizable hydrophobic bromine can orientate in the orthosteric binding area through the increase in hydrophobic interaction with W243^6.48 and L91^3.33. This was also validated through analysis of I253^6.58A which is located on top of the binding pocket. The binding affinity of 37 showed a greater reduction in affinity compared to 39, suggesting that 37 occupies a position near the top of the orthosteric pocket.

As is shown in Figure 5 using an AF2-generated model that we recently published,⁴⁰ the correlation between the 23 TI/MD calculated relative binding free energies and experimental values measured with the NanoBRET binding assay was very good with r = 0.68 (p = 0.026) and MUE = 0.89 kcal mol^–1 (Table S1). This shows that we can use this procedure for further optimization of this series of compounds which is ongoing research.

Finally, for future optimization and in vivo studies, lead compound 39 was assessed for its affinity for endogenously expressed receptors and in vitro pharmacokinetic properties. Compound 39 selectively antagonized IB-MECA, but not CPA, when looking at the inhibition of cAMP production, demonstrating its hA₃R selectivity in cells expressing both inhibitory hARs. Furthermore, despite cell toxicity in both cell lines, 39 exerted a greater potency and efficacy for the inhibition of cellular proliferation in lung cancer cells expressing the A₃R, reinforcing the receptor as a potential target for the treatment of cancer. When looking at the pharmacokinetic suitability of 39, it showed 400-fold higher solubility in simulated intestinal fluid than in PBS or simulated gastric fluid. The partition coefficient and percentage of protein binding and permeability in Caco-2 cells were all comparable to reference compounds haloperidol, sertraline, or labetalol. In terms of in vitro metabolism, compound 39 showed a much larger intrinsic clearance in liver microsomes which resulted in a short half-life (∼9.5 min). Future optimization based on this compound should focus on improving its aqueous solubility and metabolic stability to increase the in vivo efficacy.

Experimental Section

Biological Methods

Compounds

NECA, CPA, IB-MECA, and MRS1220 were purchased from Tocris Bioscience (Wiltshire, UK). Rolipram was obtained from Cayman Chemicals (Michigan, USA). All the ligands above-mentioned were dissolved in DMSO as 10 mM stock and stored at −20 °C until use. Forskolin was purchased from Tocris Bioscience (Wiltshire, UK), made up as 10 mM stock in DMSO, and stored at room temperature. CA200645 and CA200623 were purchased from HelloBio (Bristol, UK), dissolved in DMSO as 100 μM stock, and stored at −20 °C. The purity of the tested compounds was >95% (see Chemistry methods, General Information).

Constructs

The FLAG tag (DYKDDDDK) and Nluc-hA₃R (gifted by Stephen Briddon) were cloned into vector pcDNA3.1(−). Site-directed mutagenesis was performed to make A₃R mutants using the QuikChange Lightning kit (Agilent Technologies, US) according to the manufacturer’s protocol. All the construct sequences were confirmed by the DNA sequencing performed by the DNA sequencing facility at the Department of Biochemistry, University of Cambridge (Cambridge, UK).

Cell Culture and Transfection

The source of cells was as described previously.^25,35 These cells were maintained in Dulbecco’s modified Eagle’s medium (DMEM)/Hams F-12 nutrient mix (F12) GlutaMAX media (ThermoFisher, UK), supplemented with 10% heat-inactivated fetal bovine serum (FBS) (Sigma-Aldrich, Poole, Dorset, UK) and 1% antibiotic-antimycotic (AA) (Sigma, UK). In the cAMP accumulation assay, the Chinese hamster ovary (CHO) K1 cell line was used because of its lack of endogenous ARs expression.⁶¹ Stable cell lines of CHO-K1-A₁R, CHO-K1-A_2AR, CHO-K1-A_2BR, and CHO-K1-A₃R were cultured with the Hams F-12 nutrient mix (ThermoFisher, UK) supplemented with 10% FBS and 1% AA. 600 μg/mL Geneticin (ThermoFisher, UK) was added to the stable HEK293T and CHO-K1 cell lines for selection, and the medium was changed every 2 days. In the mutagenesis study, HEK293T cells were plated to a 24-well plate and grown overnight. The seeded cells were then transfected with 250 ng of FLAG-Nluc-A₃R WT or mutant receptor using polyethylenimine 25 kDa (PEI, Polysciences Inc., Germany) in 6:1 ratio, diluted in 150 mM NaCl. LK-2 and NCI-H1792 cells were cultured in RPMI 1640 medium supplemented with 10% FBS and 1% AA. All cells were maintained at 37 °C with 5% CO₂ in a humidified atmosphere.

NanoBRET Binding Assay

NanoBRET ligand competition binding assays were performed to identify the saturation binding affinity (pK_i) and the binding kinetic parameters of the potential antagonists. HEK293 cells stably expressing Nluc-hA₃R/rA₃R or HEK293T cells transfected with FLAG-Nluc-A₃R WT or mutants for 24 h were seeded onto white 96-well plates (Greiner, UK) at a density of 10,000 cells/well and grown overnight. On the assay day, the medium was discarded and replaced with 100 μL of phosphate-buffered saline (PBS) buffer (ThermoFisher, UK) containing 0.1% bovine serum albumin (BSA, Sigma-Aldrich, Poole, Dorset, UK) and 0.1 μM Nano-Glo Luciferase substrate (Promega, UK). After 5 min of incubation in the dark, tested compounds in different concentrations together with 5 nM CA200645 for Nluc-hA₃R or 100 nM CA200623 for Nluc-rA₃R or CA200645 at its Ki concentration for individual FLAG-Nluc-hA₃R mutants were added. The BRET signal was measured with Mithras LB940, recording the light emission at 460 nM (Nluc) and 610 nM (fluorescent ligands) for 30 min. The raw BRET ratio was calculated by dividing the emission at 610 nM with the 450 nM emission. Nonspecific binding was determined with the addition of high concentration of unlabeled antagonist, 10 nM MRS1220, and the corresponding BRET ratio was used for baseline correction. The baseline-corrected BRET ratio at 10 min poststimulation was used for calculating the affinity constant.

cAMP Competition Assay

In the cAMP inhibition (A₁R or A₃R) or accumulation (A_2AR and A_2BR) experiments, CHO-K1 cells stably expressing the corresponding receptor were performed as described previously.²⁵ Briefly, cells expressing the receptor of choice were harvested and resuspended in stimulation buffer (PBS containing 0.1% BSA and 25 μM rolipram) and seeded at a density of 2000 cells/well in the white 384-well Optiplate. Potential antagonists or DMSO control were added with different concentrations of AR agonist NECA. In the experiment with A₁R/A₃R, cells were costimulated with 1 μM forskolin (an adenylyl cyclase activator) to detect the inhibition of cAMP response. LK-2 and NCI-H1792 cells were stimulated as above, but using 1000 cells/well, and 0.1 μΜ forskolin, with different concentrations of CPA or IB-MECA. After 30 min of stimulation, the cAMP levels were determined with a LANCE cAMP kit (PerkinElmers, MA, US).

Flow Cytometry

To assess the cell surface expression level of FLAG-Nluc-A₃R WT and mutants, HEK293T cells transiently transfected with FLAG-Nluc-A₃R WT or mutants were analyzed with FACS in flow cytometry. After 48 h of transfection, 300,000 cells were harvested from each sample and washed twice with FACS buffer (PBS containing 1% BSA and 0.03% sodium azide). The cells were then incubated with 50 μL of FACS buffer containing 1:120 phycoerythrin (PE) anti-DYKDDDDK(FLAG) tag antibody (BioLegend, San Diego, US) in the dark for 1 h. After incubation, the cells were washed twice again and resuspended in 50 μL of FACS buffer. Analysis was performed using a BD AccuriTM C6 flow cytometer with an excitation at 488 nm and an emission wavelength at 585 nm. The resulting median intensity of cells was normalized against cell transfection with pcDNA3.1(−) as 0% and FLAG-Nluc-hA₃R WT as 100%.

Proliferation Assay

Proliferation assays were performed using the cell counting kit-8 (CCK-8) as described previously.³⁵ Briefly, LK-2 or NCI-H1792 cells were plated at 2500 cells/well of a clear 96-well plates in complete RPMI media and cultured for 24 h. The cells were then treated with compounds over a concentration range for 72 h. CCK-8 reagent was added, and after 3 h, plates were read using a Mithras LB 940 multimode microplate reader.

In Vitro Pharmacokinetic Assessment

The pharmacokinetic assessment of 39 was outsourced and performed by Eurofins Panlabs (MO, USA). The assessments included the determination of aqueous solubility, protein binding ability, partition coefficient, permeability across Caco-2 cells, and the metabolic stability in human liver microsomes. The aqueous solubility was determined by comparing the peak area of standard (200 μM calibration standard dissolved in a solvent made up of 60% methanol and 40% water) with the peak area of the corresponding peak in an individual buffer (PBS at pH = 7.4, simulated gastric buffer, and simulated intestinal fluid) as the method shown in ref (54). A chromatogram of 200 μM test compound along with a UV/vis spectrum with labeled absorbance maxima was generated. For protein binding assays, equilibrium dialysis was performed according to ref (55). The partition coefficient was determined from the amount of compound in the organic phase and in the aqueous phase. The amount of compound in buffer was determined as the combined, volume-corrected, and weighted areas of the corresponding peaks in the aqueous phases of three organic-aqueous samples of different compositions following ref (55). In the Caco-2 bidirectional permeability assay, the compounds were tested according to ref (56) in the human colon carcinoma cell line Caco-2. For the intrinsic clearance, the remaining tested compounds in the human liver microsomes (0.1 mg/mL) were quantified using HPLC-MS after 0, 15, 30, 45, and 60 min based in ref (57).

Data Analysis

All the assay data was analyzed with Prism 9.3.1 (Graphpad, San Diego, CA). The affinity (pK_i) of the potential antagonists in NanoBRET ligand binding assay was determined by fitting the baseline-corrected BRET ratio response curve with the “one-site K_i model” based on the Cheng and Prusoff equation⁵⁸ with both the concentration (HotNM) and K_d (HotKDNM) values of the “Hot ligand” CA200645 set to 5 nM for hA₃R and HotNM/HotKDNM as 70 nM/100 nM or 100 nM/300 nM for CA200623 at hA₃R or rA₃R. For the A₃R mutants, radioligandNM/HotKDNM was input as the corresponding K_d values shown in Table 4. Receptor binding kinetics was determined as described previously^25,32 using the Motulsky and Mahan method⁵⁹ (built into Prism 9.3.1) to determine the test compound association rate constant and dissociation rate constant using rate constants previously described.²⁵ In the cAMP competition experiments, the responses were normalized to the response of 100 μM forskolin as maximum and fitted with a three-parameter logistics equation. pK_B values for the potential antagonists were determined based on eq 1

1

where A′ and A are the EC₅₀ of the response induced by NECA with the presence of the antagonist or DMSO control and K_B = the affinity of the antagonist used.⁵⁸

All the statistical significance (*p < 0.05) was calculated by nonparametric Kruskal–Wallis test or one-way ANOVA with a Dunnett’s multiple comparison test and determined as described in ref (60).

The data analysis of in vitro pharmacokinetic assessment was performed by Eurofins Panlabs (MO, USA). In the protein binding assay, % protein binding and % recovery were determined using eqs 2 and 3

2

3

where area = peak area of the analyte in the protein matrix(p), buffer(b), and control sample(c).

The apparent permeability coefficient (P_app) and % recovery of the test compounds were calculated using eqs 4 and (5)

4

5

where V_R/D is the volume of the receiver/donor chamber. C_R/D,end is the concentration of the test compound in the receiver/donor chamber at the end time point, Δt is the incubation time, and A is the surface area of the cell monolayer. C_R/D,mid is the calculated midpoint concentration of the test compound in the receiver/donor side. C_D0 is the concentration of the test compound in the donor sample at time zero.

For the intrinsic clearance, the half-life (t_1/2) was determined from the slope of the initial linear range of the logarithmic curve of compound remaining (%) against time, assuming the first-order kinetics. Also, the apparent intrinsic clearance (CLint) was calculated using eq 6

6

Computational Medicinal Chemistry

Preparation of Model of the Unresolved Inactive hA₃R

Residues are described by their amino acid identity (single letter code) and position (amino acid number) within the specific GPCR with the Ballesteros and Weinstein numbering,⁶¹ a scheme for class A GPCRs, whereby X.50 represents the defined centrally conserved residue on helix X, in superscript. Αll His residues were protonated on the Nε.⁶²

We used for the inactive hA₃R a ML-based model derived from GPCRdb web-tool⁴³ that contains predictions for GPCRs in active and inactive forms via the advanced multistate AF2 method^41,42 of hA₃R. We revised this model of the inactive state of hA₃R by changing the orientation of R173^5.34, M172,^5.33 and M174^5.35 as we previously described.⁴⁰

We superimposed the experimental crystal structure ZM241385—A_2AR complex (PDB ID 3EML)⁶³ to our revised⁴⁰ ΑF2 model of WT hA₃R model N(1.32)–H(7.75). (Residue numbers in parentheses refer to the Ballesteros–Weinstein numbering.⁶⁴) Then, the A_2AR protein (PDB ID 3EML)⁶³ and the crystal waters were removed resulting in the AF2 model of WT hA₃R in complex with ZM241385 which was used as a template for the docking calculations. The model of hA₃R in complex with ZM241385 was optimized using the Protein Preparation Wizard in Schrödinger suite 2021 (Protein Preparation Wizard; Epik, Schrödinger, LLC, New York, NY, 2021)⁶⁵ as we previously described.³²

Molecular Docking Calculations

Ligands 37-61 and 74-77 preparation was achieved as described in ref (32). Molecular docking calculations were performed using the induced-fit docking protocol of Schrödinger suite 2021 (Induced-fit Docking, Schrödinger, LLC, New York, NY, 2021) in a standard protocol (standard precision) which allows flexibility of both the ligand and the entire binding site. The AF2 model of the WT hA₃R model in complex with ZM241385 was used as a template structure. Thus, the grid boxes for the binding site were built considering the coordinates of ZM241385. Docking of compounds 37-61 and 74-77 was performed using a softened potential in which the van der Waals scaling factor was set at 0.5 for both receptor and ligand. The Prime refinement step was set on side chain prediction of amino acid residues within 5 Å of the ligand. Subsequently, a minimization of the same set of residues and the ligand for each protein/ligand complex pose was performed. After this stage, any receptor structure in each pose reflects an induced fit to the ligand structure and conformation. For each ligand docked, a maximum of 20 poses was retained. The binding conformations of compounds 37-61 and 74-77 were analyzed, and the top-scoring docking poses were used for the MD simulations to investigate the binding profile of the tested compounds to the inactive hA₃R.

MD Simulations

Each complex of ligands K18, 37–39, 56, 57, and 60 with hA₃R from docking calculations was inserted in a pre-equilibrated hydrated POPC membrane bilayer according to the Orientation of Proteins in Membranes (OPM) database.⁶⁶ The orthorhombic periodic box was set 12 Å away from the protein, and the 10 × 10 × 18 Å box consisted ca. 130 lipids and 13,000 TIP3P water molecules,⁶⁷ using the System Builder utility of Desmond v4.9 (Schrödinger Release 2021-1: Desmond Molecular Dynamics System, D. E. Shaw Research, New York, NY, 2021. Maestro-Desmond Interoperability Tools, Schrödinger, New York, NY, 2021). Sodium and chloride ions were added randomly in the water phase to neutralize the systems and reach the experimental salt concentration of 0.150 M NaCl. The total number of atoms of the complex was approximately 75,000, and the simulation box dimensions was 71 × 81 × 105 ų. We used the LEaP, the main program for preparing simulations in Amber Software, antechamber, and Parmchk2 of AmberTools22 to assign the amberff19sb force field parameters³⁹ for the calculation of the protein and intermolecular interactions, lipids21 parameters for lipids,⁶⁸ the Generalized Amber Force Field parameters for the ligands,⁶⁹ and the TIP3P model for waters and ions.⁷⁰⁻⁷² Partial charges for ligands were obtained using RESP⁷³ fitting of the electrostatic potentials calculated with Gaussian03⁷⁴ at the Hartree–Fock/6-31G*⁷⁵ level of theory and the antechamber of AmberTools22.⁷⁶

MD simulation protocol starts with energy minimization by applying 2500 steps of steepest descent to remove bad contacts and 7500 steps of conjugated gradient minimization in the presence of a harmonic restraint with a force constant of 5 kcal mol^–1 Å^–2 on all atoms of protein and ligand and a nonbonded cutoff of 12.0 Å. The next stage in the MD simulation protocol is to allow the system to heat up from 0 to 310 K using the Langevin thermostat (dynamics)⁷⁷ for temperature control, as implemented in the Amber22 program,⁷⁸ employing a Langevin collision frequency of 2.0 ps. Heating was accomplished in two consecutive steps in the presence of a harmonic restraint with a force constant of 10 kcal mol^–1 Å^–2 on all membrane, protein, and ligand atoms. In the first step, systems were heated to 100 K in a NVT of 50 ps length. In the second step, the temperature was raised to 310 K in a NPTγ (with γ = 10 dyn cm^–1) simulation of 500 ps length. The Berendsen barostat⁷⁹ was used to adjust the density over the 500 ps simulation at constant pressure (NPTγ) (with γ = 10 dyn cm^–1), with a target pressure of 1 bar and a 2 ps pressure relaxation time. Subsequently, the systems were equilibrated without restraints in a NPTγ simulation of 1 ns length with T = 310 K and γ = 10 dyn cm^–1. In the NPTγ simulations, semiisotropic pressure scaling to p = 1 bar was applied using a pressure relaxation time of 1.0 ps. The temperature of 310 K was used in MD simulations to ensure that the membrane state is above the main phase transition temperature of 298 K for POPC bilayers.⁸⁰

Bonds involving hydrogen atoms were constrained by the SHAKE algorithm,⁸¹ and a time step of 2 fs was used for the integration of the equations of motion. Long-range electrostatics were calculated using the Particle mesh Ewald summation,⁷⁹ with a 1 Å grid, and short-range nonbonding interactions were truncated at 12 Å with a continuum model long-range correction applied for energy and pressure.

The equilibration phase was followed by production MD simulation for 500 ns for 7 representative ligands (K18, 37–39, 56, 57, and 60) using the same protocol as in the final equilibration step. Snapshots were recorded every 100 ps during the production phase. Within the 500 ns MD simulation time, the total energy and rmsd of the protein backbone Cα atoms reached a plateau, and the systems were considered equilibrated and suitable for statistical analysis (see Figure 1, 4, S5, and S6).

Two MD simulation repeats were performed for each complex using the same starting structure and applying randomized velocities. The visualization of the MD simulation trajectories was performed using VMD.⁸² We used ptraj and cpptraj⁸³ of AmberTools22,⁷⁶ MDAnalysis,^84,85 and Matplotlib programs⁸⁶ and ProLIF,⁸⁷ NumPy⁸⁸ libraries to perform analysis of the MD simulations trajectories.

Particle mesh Ewald molecular dynamics (PMEMD) is the primary engine for running MD simulations with AMBER22 software.⁷⁶ All the MD simulations with AMBER22 software⁷⁶ were run on GTX 3070i GPUs in lab workstations. The pmemd.CUDA executable provides the ability to use NVIDIA GPUs to run the MD simulations.

Alchemical TI/MD Binding Free Energies Calculated with the MBAR Method

For the TI/MD calculations, the relaxed complex of compound 37 at hA₃R from the 500 ns MD simulations in a POPC lipid bilayer with the ff19sb³⁹ was used as the reference structure for the calculations. Thus, binding poses of ligands aligned with 37-inactive hA₃R complexes were used as starting structures for the alchemical calculations with our optimized⁴⁰ multistate AF2 method^41,42 of hA₃R generated from GPCRdb⁴³ web-tool. These alchemical perturbations are described in Table S1. We found that the final snapshots from the 2 ns TI/MD simulations matched the final snapshots of relevant ligands from the 500 ns MD simulations (compounds K18, 37–39, 56, 57, and 60); the ligand did not change its binding pose during the much longer MD simulations. Thus, the TI/MD binding free energy simulations calculated the binding free energy change of the binding poses between the examined two ligands without any of the ligands changing its conformation inside the receptor. TI/MD calculations were also performed for the ligands in solution.

The calculation of the relative binding free energies ΔΔA_b,0→1 or ΔΔA_b,0,1 for two ligands 0 and 1 bound to A₃R (for the 23 pairs of ligands shown in Table S1) can be performed using the MBAR method⁸⁹ and applying a thermodynamic cycle,⁹⁰⁻⁹² i.e., using the ΔA values obtained for the alchemical transformations of the ligands in the bound (b) and the solvent (s; water) state ΔA_b,0,1 and ΔA_s,0,1(s), respectively, according to eq 7

7

The TI estimator computes the free energy change of the transformation 0 → 1 ΔA_0→1 or ΔA_0,1 by integrating the Boltzmann averaged dU(λ)/dλ as is described in eqs 8 and 9

8

9

MBAR⁸⁹ calculates the free energy difference between neighboring intermediate states ΔA_λ→λ+1 using eq 10

10

where w is a function of Α(λ) and Α(λ+1). The equation is solved iteratively to give the free energy change of neighboring states ΔA_λ→λ+1 which via combination yield the overall free energy change. The MBAR method has been shown to minimize the variance in the calculated free energies, by making more efficient use of the simulation data.^89,93−95 Details of the TI/MD theory¹⁷ and the TI/MD protocol³² have been described.

Experimental relative binding free energies were estimated using the experimental binding affinities pK_d in Table S1 according to eq 11

11

Chemistry

General Information

Melting points were determined on a Büchi apparatus and are uncorrected. ¹H NMR and ¹³C NMR spectra were recorded on a Bruker AVANCE III 600 or a Bruker AVANCE DRX 400 instrument in deuterated solvents and were referenced to TMS (δ scale) (Figure S8). Mass spectra were recorded with a LTQ Orbitrap Discovery instrument, possessing an Ionmax ionization source. Flash chromatography was performed on Merck silica gel 60 (0.040–0.063 mm). Analytical thin-layer chromatography (TLC) was carried out on precoated (0.25 mm) Merck silica gel F-254 plates. The purity of the target derivatives (>95%) was determined on a Thermo Finnigan HPLC System (P4000 Pump, AS3000 Autosampler, UV Spectra System UV6000LP detector, Chromquest 4.1 Software); Phenomenex HYPERSIL C18-BDS (250 mm, 4.0 mm, 5 μm); mobile phase: Method A: 0.2% formic acid in water/acetonitrile; flow rate 0.8 mL/min or Method B: 1% formic acid in water/acetonitrile/methanol (9:1); flow rate 1 mL/min; column temperature 25 °C; injection volume 5 μL (Table S9 and Figure S9).

6-Chloro-9-methyl-9H-purine (15) and 6-Chloro-7-methyl-7H-purine (16)

Sodium hydride (60% dispersion in mineral oil, 1.1 g, 27.5 mmol) was added in two portions into a suspension of 6-chloropurine (14, 3 g, 19.39 mmol) in anhydrous N,N-dimethylformamide (40 mL) at 0 °C, and this mixture was stirred at r.t. under argon for 1 h. Then, the reaction was cooled at 0 °C, iodomethane (1.8 mL, 28.86 mmol) was added, and this mixture was stirred at r.t. for 20 h. Upon completion of the reaction, the mixture was diluted with water and extracted with dichloromethane (4 × 250 mL), and the combined organic layers were extracted with brine (2 × 400 mL), dried over sodium sulfate, and evaporated. The crude mixture was purified by column chromatography using a mixture of cyclohexane/ethyl acetate as the eluent (from 40/60 up to 10/90, v/v) to provide the pure isomers 15 and 16.

Data for 15: Yield 58%. White solid, mp 156–157 °C (CH₂Cl₂/n-pentane) (reported⁹⁶ 142–143 °C). ¹H NMR (600 MHz, CDCl₃): δ 8.68 (s, 1H), 8.07 (s, 1H), 3.89 (s, 3H). ¹³C NMR (151 MHz, CDCl₃): δ 152.3, 152.1, 151.0, 145.8, 131.6, 30.4.

Data for 16: Yield 34%. White solid, mp 209 °C (CH₂Cl₂/n-pentane) (reported⁹⁶ 180–182 °C). ¹H NMR (600 MHz, CDCl₃): δ 8.88 (s, 1H), 8.17 (s, 1H), 4.17 (s, 3H). ¹³C NMR (151 MHz, CDCl₃): δ 162.1, 152.7, 149.6, 143.6, 123.2, 34.5.

9-Methyl-9H-purine-6-carbonitrile (17)

Zinc cyanide (83 mg, 0.71 mmol), zinc dust (8.6 mg, 0.13 mmol), tris(dibenzylideneacetone)dipalladium(0) (Pd₂(dba)₃, 22 mg, 0.024 mmol), and 1,1′-ferrocenediyl-bis(diphenylphosphine) (dppf, 26 mg, 0.049 mmol) were added into a solution of the chloroderivative 15 (200 mg, 1.19 mmol) in anhydrous N,N-dimethylacetamide (1.5 mL), and this mixture was refluxed, under argon, for 3 h. Upon completion of the reaction, the mixture was filtered through a Celite pad and washed with dichloromethane. The filtrate was diluted with water and extracted with dichloromethane (4 × 20 mL), and the combined organic layers were extracted with brine (2 × 50 mL), dried over sodium sulfate, and evaporated. The crude product was purified by column chromatography using a mixture of cyclohexane/ethyl acetate as the eluent (from 70/30 up to 20/80, v/v) to provide the pure nitrile 17 (150 mg, yield 79%) as a brown solid. mp 171–172 °C (CH₂Cl₂/n-pentane), reported⁹⁷ 153–154 °C. ¹H NMR (600 MHz, CDCl₃): δ 9.06 (s, 1H), 8.30 (s, 1H), 3.99 (s, 3H). ¹³C NMR (151 MHz, CDCl₃): δ 153.5, 152.9, 149.0, 135.1, 131.0, 113.8, 30.6.

Method A

An aqueous solution of hydroxylamine (50 wt %, 0.75 mL, 11.4 mmol) was added into a solution of the corresponding arylnitrile (4 mmol) in ethanol (20 mL), and this mixture was refluxed for 2 h. Upon completion of the reaction, the solvent was evaporated at half volume, and the solid was filtered under vacuum, washed with a small amount of ethanol, and air-dried. The solid product was collected and recrystallized to provide the pure amidoxime.

Method B

Hydroxylamine hydrochloride (417 mg, 6 mmol) and sodium bicarbonate (504 mg, 6 mmol) were added into a solution of the corresponding arylnitrile (4 mmol) in ethanol (20 mL), and this mixture was stirred at room temperature for 90 min and then refluxed for 2 h. Upon completion of the reaction, the solvent was evaporated, water was added into the flask, and the solid was filtered under vacuum, washed with water, and air-dried. The solid product was collected and recrystallized to provide the pure amidoxime.

(Z)-N′-Hydroxypicolinimidamide (18)

This compound was prepared according to method A, starting from pyridine-2-carbonitrile (1). Yield 82%. White solid, mp 128 °C (EtOH) (reported⁹⁸ 119–120 °C). ¹H NMR (600 MHz, DMSO-d₆): δ 9.88 (br s, 1H, D₂O exch.), 8.56 (d, J = 4.6 Hz, 1H), 7.86 (d, J = 8.0 Hz, 1H), 7.82–7.78 (m, 1H), 7.39 (dd, J = 6.5, 5.4 Hz, 1H), 5.82 (br s, 2H, D₂O exch.). ¹³C NMR (151 MHz, DMSO-d₆): δ 150.0, 149.4, 148.2, 136.5, 124.0, 119.3.

(Z)-3,5-Dichloro-N′-hydroxypicolinimidamide (19)

This compound was prepared according to method B, starting from 3,5-dichloropyridine-2-carbonitrile (2). Yield 88%. White solid, mp 151–152 °C (EtOH). ¹H NMR (400 MHz, DMSO-d₆): δ 9.90 (br s, 1H, D₂O exch.), 8.64 (d, J = 1.9 Hz, 1H), 8.28 (d, J = 1.9 Hz, 1H), 5.87 (br s, 2H, D₂O exch.). ¹³C NMR (151 MHz, DMSO-d₆): δ 148.9, 147.5, 145.7, 137.8, 130.8, 130.0.

(Z)-5-Bromo-N′-hydroxypicolinimidamide (20)

This compound was prepared according to method A, starting from 5-bromopyridine-2-carbonitrile (3). Yield 76%. White solid, mp 181–182 °C (EtOH) (reported⁹⁹ 162–164 °C). ¹H NMR (600 MHz, DMSO-d₆): δ 10.02 (br s, 1H, D₂O exch.), 8.68 (dd, J = 2.4, 0.7 Hz, 1H), 8.04 (dd, J = 8.6, 2.4 Hz, 1H), 7.81 (dd, J = 8.6, 0.6 Hz, 1H), 5.83 (br s, 2H, D₂O exch.). ¹³C NMR (151 MHz, DMSO-d₆): δ 148.89, 148.87, 139.2, 121.1, 120.3.

(Z)-N′-Hydroxy-6-methylpicolinimidamide (21)

This compound was prepared according to method A, starting from 6-methylpyridine-2-carbonitrile (4). Yield 80%. White solid, mp 152–153 °C (EtOH) (reported¹⁰⁰ 179–180 °C). ¹H NMR (400 MHz, DMSO-d₆): δ 9.84 (br s, 1H, D₂O exch.), 7.69–7.65 (m, 2H), 7.25 (dd, J = 7.0, 1.2 Hz, 1H), 5.78 (br s, 2H, D₂O exch.), 2.49 (s, 3H, overlapping with DMSO-d₆). ¹³C NMR (151 MHz, DMSO-d₆): δ 156.6, 149.5, 149.2, 136.8, 123.2, 116.3, 23.9.

(Z)-N′-Hydroxypyrimidine-2-carboximidamide (22)

This compound was prepared according to method A, starting from pyrimidine-2-carbonitrile (5). Yield 93%. White solid, mp 228–229 °C (EtOH) (reported¹⁰¹ 211–212 °C). ¹H NMR (600 MHz, DMSO-d₆): δ 10.15 (br s, 1H, D₂O exch.), 8.83 (d, J = 4.8 Hz, 2H), 7.50 (t, J = 4.8 Hz, 1H), 5.81 (br s, 2H, D₂O exch.). ¹³C NMR (151 MHz, DMSO-d₆): δ 158.3, 157.2, 149.0, 121.1.

(Z)-2-Chloro-N′-hydroxynicotinimidamide (23)

This compound was prepared according to method A, starting from 2-chloropyridine-3-carbonitrile (6). Yield 79%. White solid, mp 165–166 °C (EtOH) (reported¹⁰¹ 139–140 °C). ¹H NMR (600 MHz, DMSO-d₆): δ 9.59 (br s, 1H, D₂O exch.), 8.43 (dd, J = 4.8, 1.9 Hz, 1H), 7.84 (dd, J = 7.5, 1.9 Hz, 1H), 7.46–7.43 (m, 1H), 5.93 (br s, 2H, D₂O exch.). ¹³C NMR (151 MHz, DMSO-d₆): δ 149.7, 149.5, 149.0, 140.2, 130.3, 122.8.

(Z)-N′-Hydroxy-6-methylnicotinimidamide (24)

This compound was prepared according to method A, starting from 6-methylpyridine-3-carbonitrile (7). Yield 86%. White solid, mp 184–185 °C (EtOH) (reported¹⁰⁰ 172–173 °C). ¹H NMR (600 MHz, DMSO-d₆): δ 9.71 (br s, 1H, D₂O exch.), 8.72 (d, J = 2.1 Hz, 1H), 7.89 (dd, J = 8.1, 2.3 Hz, 1H), 7.25 (d, J = 8.1 Hz, 1H), 5.88 (br s, 2H, D₂O exch.), 2.47 (s, 3H). ¹³C NMR (151 MHz, DMSO-d₆): δ 158.3, 149.0, 145.9, 133.1, 126.2, 122.4, 23.8.

(Z)-N′-Hydroxynicotinimidamide (25)

This compound was prepared according to method B, starting from pyridine-3-carbonitrile (8). Yield 85%. White solid, mp 138 °C (MeOH/n-pentane) (reported¹⁰² 134–136 °C). ¹H NMR (400 MHz, DMSO-d₆): δ 9.84 (br s, 1H, D₂O exch.), 8.86 (dd, J = 2.2, 0.7 Hz, 1H), 8.56 (dd, J = 4.8, 1.6 Hz, 1H), 8.04–7.98 (m, 1H), 7.42–7.38 (m, 1H), 5.97 (br s, 2H, D₂O exch.). ¹³C NMR (100 MHz, DMSO-d₆): δ 149.8, 149.0, 146.6, 132.9, 129.1, 123.3.

(Z)-N′-Hydroxy-3-methylbenzimidamide (26)

This compound was prepared according to method A, starting from 3-methylbenzolocarbonitrile (9). Upon completion of the reaction, the solvent was evaporated, and the oily residue was purified by column chromatography, using a mixture of dichloromethane/methanol (from 95/5 up to 90/10, v/v) as the eluent, to provide the pure amidoxime 26 as a pale yellow solid. Yield 96%. mp 94–95 °C (EtOAc/n-pentane) (reported¹⁰³ 90–91 °C). ¹H NMR (400 MHz, DMSO-d₆): δ 9.56 (br s, 1H, D₂O exch.), 7.49 (s, 1H), 7.45 (d, J = 7.7 Hz, 1H), 7.25 (t, J = 7.6 Hz, 1H), 7.18 (d, J = 7.5 Hz, 1H), 5.74 (br s, 2H, D₂O exch.), 2.31 (s, 3H). ¹³C NMR (100 MHz, DMSO-d₆): δ 151.0, 137.2, 133.3, 129.5, 128.0, 126.0, 122.6, 21.1.

(Z)-N′-Hydroxy-2-(pyridin-2-yl)acetimidamide (27)

This compound was prepared according to method B, starting from 2-(pyridin-2-yl)acetonitrile (10). Yield 71%. Beige solid, mp 114–115 °C (EtOH) (reported¹⁰⁴ 112–114 °C). ¹H NMR (600 MHz, DMSO-d₆): δ 8.92 (br s, 1H, D₂O exch.), 8.47 (d, J = 4.8 Hz, 1H), 7.71 (td, J = 7.7, 1.8 Hz, 1H), 7.33 (d, J = 7.8 Hz, 1H), 7.23 (dd, J = 6.8, 5.1 Hz, 1H), 5.40 (br s, 2H, D₂O exch.), 3.44 (s, 2H). ¹³C NMR (151 MHz, DMSO-d₆): δ 157.6, 150.8, 148.7, 136.4, 122.9, 121.6, 39.4 (overlapping with DMSO-d₆).

(Z)-N′-Hydroxy-2-(pyridin-3-yl)acetimidamide (28)

This compound was prepared according to method A, starting from 2-(pyridin-3-yl)acetonitrile (11). Yield 78%. Pale yellow solid, mp 173 °C (EtOH) (reported¹⁰⁵ 153 °C). ¹H NMR (400 MHz, DMSO-d₆): δ 8.97 (br s, 1H, D₂O exch.), 8.47 (d, J = 1.8 Hz, 1H), 8.41 (dd, J = 4.8, 1.5 Hz, 1H), 7.69–7.64 (m, 1H), 7.31 (dd, J = 7.7, 4.9 Hz, 1H), 5.50 (br s, 2H, D₂O exch.), 3.29 (s, 2H). ¹³C NMR (151 MHz, DMSO-d₆): δ 151.5, 149.8, 147.5, 136.1, 133.5, 123.2, 34.3.

(Z)-N′-Hydroxy-1-methyl-1H-indole-5-carboximidamide (29)

This compound was prepared according to method A, starting from 1-methyl-1H-indole-5-carbonitrile (13). Upon completion of the reaction, the solvent was evaporated, and the residue was purified by column chromatography, using a mixture of dichloromethane/methanol (from 95/5 up to 88/12, v/v) as the eluent, to provide the pure amidoxime 29 as a white solid. Yield 72%. mp 175–176 °C (EtOH) (reported¹⁰⁶ 169–171 °C). ¹H NMR (600 MHz, DMSO-d₆): δ 9.36 (br s, 1H, D₂O exch.), 7.86 (s, 1H), 7.52 (dd, J = 8.6, 1.2 Hz, 1H), 7.39 (d, J = 8.6 Hz, 1H), 7.32 (d, J = 2.9 Hz, 1H), 6.44 (d, J = 2.6 Hz, 1H), 5.68 (br s, 2H, D₂O exch.), 3.78 (s, 3H). ¹³C NMR (151 MHz, DMSO-d₆): δ 152.0, 136.7, 130.2, 127.5, 124.4, 119.1, 117.6, 109.2, 100.9, 32.5.

(Z)-N′-Hydroxy-9-methyl-9H-purine-6-carboximidamide (30)

This compound was prepared according to method B, starting from 9-methyl-9H-purine-6-carbonitrile (17). Yield 67%. Beige solid, mp 245 °C (CH₂Cl₂/n-pentane). ¹H NMR (400 MHz, DMSO-d₆): δ 10.53 (br s, 1H, D₂O exch.), 8.92 (s, 1H), 8.56 (s, 1H), 6.10 (br s, s2H, D₂O exch.), 3.85 (s, 3H). ¹³C NMR (151 MHz, DMSO-d₆): δ 152.6, 151.0, 148.7, 147.5, 147.2, 129.6, 29.5.

General Procedure for the Preparation of the Acyl Chlorides 34–36

A solution of the corresponding carboxylic acid 31–33 (1 mmol) in thionyl chloride (3 mL) was refluxed for 3 h under argon. Then, the solvent was evaporated under reduced pressure, and the resulting acyl chlorides 34–36 were used immediately to the next step, with no further purification.

General Procedure for the Preparation of the Target Derivatives 37–55

The corresponding amidoxime 18–30 (1 mmol) and triethylamine (0.15 mL, 1.1 mmol) were added into a solution of the acyl chloride 34, 35, or 36 (1 mmol) in anhydrous tetrahydrofuran (5 mL), under argon, and this reaction mixture was stirred at room temperature for 2–16 h. Upon completion of the reaction, the mixture was diluted with ethyl acetate (40 mL) and extracted with water (40 mL). The aqueous layer was extracted two more times with ethyl acetate (40 mL). The combined organic layers were washed with brine (100 mL), dried over sodium sulfate, and concentrated under reduced pressure. The resulting crude products were recrystallized to provide the pure target derivatives 37–55.

(Z)-N′-((3-(2,6-Dichlorophenyl)-5-methylisoxazole-4-carbonyl)oxy)picolinimidamide (37)

This compound was prepared according to the general procedure described above, upon reaction of the acyl chloride 34 with amidoxime 18. Reaction time: 4 h. The crude product was recrystallized from ethyl acetate, to provide the pure derivative 37 in 76% yield, as a beige solid. mp 184–185 °C (EtOAc). ¹H NMR (400 MHz, DMSO-d₆): δ 8.65 (d, J = 4.8 Hz, 1H), 7.95–7.86 (m, 2H), 7.70–7.65 (m, 2H), 7.63–7.52 (m, 2H), 6.99 (br s, 1H, D₂O exch.), 5.95 (br s, 1H, D₂O exch.), 2.88 (s, 3H). ¹³C NMR (100 MHz, DMSO-d₆): δ 176.1, 158.3, 158.1, 154.9, 148.9, 147.7, 137.4, 134.3, 132.7, 128.5, 127.3, 126.0, 121.1, 108.3, 13.6. HRMS (ESI) m/z: calcd for C₁₇H₁₃Cl₂N₄O₃ [Μ + Η]⁺, 391.0360; found, 391.0354. HPLC analysis (Method A): t_R = 11.35 min, purity 99.75%.

(Z)-3,5-Dichloro-N′-((3-(2,6-dichlorophenyl)-5-methylisoxazole-4-carbonyl)oxy)picolinimidamide (38)

This compound was prepared according to the general procedure described above, upon reaction of the acyl chloride 34 with amidoxime 19. Reaction time: 4 h. The crude product was recrystallized from dichloromethane/diethyl ether, to provide the pure derivative 38 in 87% yield, as a pale yellow solid. mp 190 °C (CH₂Cl₂/Et₂O). ¹H NMR (600 MHz, DMSO-d₆): δ 8.68 (d, J = 1.9 Hz, 1H), 8.38 (d, J = 1.9 Hz, 1H), 7.69–7.65 (m, 2H), 7.62–7.58 (m, 1H), 6.57 (br s, 2H, D₂O exch.), 2.88 (s, 3H). ¹³C NMR (151 MHz, DMSO-d₆): δ 175.8, 158.3, 157.7, 154.7, 146.5, 146.2, 137.6, 134.3, 132.44, 132.38, 130.8, 128.3, 127.3, 108.2, 13.5. HRMS (ESI) m/z: calcd for C₁₇H₁₁Cl₄N₄O₃ [Μ + Η]⁺, 458.9580; found, 458.9578. HPLC analysis (Method A): t_R = 11.52 min, purity 99.00%.

(Z)-5-Bromo-N′-((3-(2,6-dichlorophenyl)-5-methylisoxazole-4-carbonyl)oxy)picolinimidamide (39)

This compound was prepared according to the general procedure described above, upon reaction of the acyl chloride 34 with amidoxime 20. Reaction time: 2 h. The crude product was recrystallized from ethyl acetate, to provide the pure derivative 39 in 82% yield, as a white solid. mp 249 °C (EtOAc). ¹H NMR (600 MHz, DMSO-d₆): δ 8.79 (d, J = 2.2 Hz, 1H), 8.17 (dd, J = 8.5, 2.3 Hz, 1H), 7.83 (d, J = 8.5 Hz, 1H), 7.69–7.65 (m, 2H), 7.61–7.58 (m, 1H), 7.05 (br s, 1H, D₂O exch.), 6.02 (br s, 1H, D₂O exch.), 2.88 (s, 3H). ¹³C NMR (151 MHz, DMSO-d₆): δ 176.1, 158.3, 157.9, 154.4, 149.6, 146.6, 140.1, 134.3, 132.7, 128.4, 127.3, 122.7, 122.5, 108.2, 13.6. HRMS (ESI) m/z: calcd for C₁₇H₁₂BrCl₂N₄O₃ [Μ + Η]⁺, 468.9465; found, 468.9465. HPLC analysis (Method A): t_R = 11.87 min, purity 97.21%.

(Z)-N′-((3-(2,6-Dichlorophenyl)-5-methylisoxazole-4-carbonyl)oxy)-6-methylpicolinimidamide (40)

This compound was prepared according to the general procedure described above, upon reaction of the acyl chloride 34 with amidoxime 21. Reaction time: 4 h. The crude product was recrystallized from ethyl acetate/n-pentane, to provide the pure derivative 40 in 74% yield, as a white solid. mp 167–168 °C (EtOAc/n-pentane). ¹H NMR (600 MHz, DMSO-d₆): δ 7.78 (t, J = 7.7 Hz, 1H), 7.69–7.66 (m, 3H), 7.63–7.59 (m, 1H), 7.40 (d, J = 7.6 Hz, 1H), 6.76 (br s, 1H, D₂O exch.), 5.80 (br s, 1H, D₂O exch.), 2.88 (s, 3H), 2.53 (s, 3H). ¹³C NMR (151 MHz, DMSO-d₆): δ 176.1, 158.1, 158.0, 157.5, 154.8, 146.8, 137.5, 134.2, 132.6, 128.4, 127.3, 125.3, 118.0, 108.3, 23.9, 13.5. HRMS (ESI) m/z: calcd for C₁₈H₁₅Cl₂N₄O₃ [Μ + Η]⁺, 405.0516; found, 405.0507. HPLC analysis (Method A): t_R = 11.69 min, purity 99.49%.

(Z)-N′-((3-(2-Chlorophenyl)-5-methylisoxazole-4-carbonyl)oxy)-6-methylpicolinimidamide (41)

This compound was prepared according to the general procedure described above, upon reaction of the acyl chloride 35 with amidoxime 21. Reaction time: 4 h. The crude product was recrystallized from dichloromethane/diethyl ether, to provide the pure derivative 41 in 70% yield, as a white solid. mp 141–142 °C (CH₂Cl₂/Et₂O). ¹H NMR (600 MHz, DMSO-d₆): δ 7.77 (t, J = 7.7 Hz, 1H), 7.68 (d, J = 7.8 Hz, 1H), 7.65–7.61 (m, 1H), 7.60–7.56 (m, 2H), 7.51 (t, J = 7.1 Hz, 1H), 7.38 (d, J = 7.6 Hz, 1H), 6.70 (br s, 1H, D₂O exch.), 5.43 (br s, 1H, D₂O exch.), 2.81 (s, 3H), 2.51 (s, 3H, overlapping with DMSO-d₆). ¹³C NMR (151 MHz, DMSO-d₆): δ 175.7, 160.1, 158.5, 157.6, 154.7, 146.8, 137.6, 132.7, 131.9, 131.3, 129.6, 128.2, 127.5, 125.3, 118.1, 108.7, 23.9, 13.2. HRMS (ESI) m/z: calcd for C₁₈H₁₆ClN₄O₃ [Μ + Η]⁺, 371.0906; found, 371.0905. HPLC analysis (Method A): t_R = 11.45 min, purity 98.53%.

(Z)-N′-((3-(2-Chlorophenyl)-5-methylisoxazole-4-carbonyl)oxy)picolinimidamide (42)

This compound was prepared according to the general procedure described above, upon reaction of the acyl chloride 35 with amidoxime 18. Reaction time: 8 h. The crude product was recrystallized from ethyl acetate/n-pentane, to provide the pure derivative 42 in 80% yield, as a beige solid. mp 168 °C (EtOAc/n-pentane). ¹H NMR (600 MHz, DMSO-d₆): δ 8.63 (d, J = 4.8 Hz, 1H), 7.92–7.88 (m, 2H), 7.63 (d, J = 7.9 Hz, 1H), 7.60–7.56 (m, 2H), 7.55–7.53 (m, 1H), 7.52–7.49 (m, 1H), 2.82 (s, 3H). ¹³C NMR (151 MHz, DMSO-d₆): δ 175.5, 160.0, 158.3, 154.6, 148.7, 147.6, 137.3, 132.6, 131.7, 131.2, 129.5, 128.1, 127.3, 125.8, 121.0, 108.6, 13.1. HRMS (ESI) m/z: calcd for C₁₇H₁₄ClN₄O₃ [Μ + Η]⁺, 357.0749; found, 357.0743. HPLC analysis (Method A): t_R = 11.03 min, purity 99.35%.

(Z)-6-Methyl-N′-((5-methyl-3-phenylisoxazole-4-carbonyl)oxy)picolinimidamide (43)

This compound was prepared according to the general procedure described above, upon reaction of the acyl chloride 36 with amidoxime 21. Reaction time: 3 h. The crude product was recrystallized from ethyl acetate/n-pentane, to provide the pure derivative 43 in 73% yield, as a beige solid. mp 148–149 °C (EtOAc/n-pentane). ¹H NMR (600 MHz, DMSO-d₆): δ 7.79 (t, J = 7.7 Hz, 1H), 7.72 (d, J = 7.8 Hz, 1H), 7.65 (d, J = 6.4 Hz, 2H), 7.55–7.49 (m, 3H), 7.40 (d, J = 7.6 Hz, 1H), 6.67 (br s, 1H, D₂O exch.), 6.18 (br s, 1H, D₂O exch.), 2.78 (s, 3H), 2.53 (s, 3H). ¹³C NMR (151 MHz, DMSO-d₆): δ 175.6, 161.8, 159.1, 157.4, 155.0, 146.9, 137.5, 130.0, 128.9, 128.4, 128.3, 125.2, 118.1, 107.7, 23.9, 13.3. HRMS (ESI) m/z: calcd for C₁₈H₁₇N₄O₃ [Μ + Η]⁺, 337.1296; found, 337.1294. HPLC analysis (Method A): t_R = 11.30 min, purity 99.15%.

(Z)-N′-((5-methyl-3-phenylisoxazole-4-carbonyl)oxy)picolinimidamide (44)

This compound was prepared according to the general procedure described above, upon reaction of the acyl chloride 36 with amidoxime 18. Reaction time: 4 h. The crude product was recrystallized from ethyl acetate, to provide the pure derivative 44 in 75% yield, as a beige solid. mp 169 °C (EtOAc). ¹H NMR (400 MHz, DMSO-d₆): δ 8.65 (dt, J = 4.8, 1.3 Hz, 1H), 7.96–7.88 (m, 2H), 7.69–7.64 (m, 2H), 7.57–7.48 (m, 4H), 6.84 (br s, 1H, D₂O exch.), 6.30 (br s, 1H, D₂O exch.), 2.78 (s, 3H). ¹³C NMR (151 MHz, DMSO-d₆): δ 175.5, 161.8, 159.0, 155.0, 148.7, 147.8, 137.3, 129.9, 128.9, 128.33, 128.26, 125.8, 121.0, 107.7, 13.3. HRMS (ESI) m/z: calcd for C₁₇H₁₅N₄O₃ [Μ + Η]⁺, 323.1139; found, 323.1156. HPLC analysis (Method A): t_R = 10.81 min, purity 99.71%.

(Z)-N′-((3-(2,6-Dichlorophenyl)-5-methylisoxazole-4-carbonyl)oxy)pyrimidine-2-carboximidamide (45)

This compound was prepared according to the general procedure described above, upon reaction of the acyl chloride 34 with amidoxime 22. Reaction time: 2 h. The crude product was recrystallized from dichloromethane/n-pentane, to provide the pure derivative 45 in 73% yield, as a white solid. mp 219–220 °C (CH₂Cl₂/n-pentane). ¹H NMR (400 MHz, DMSO-d₆): δ 8.92 (d, J = 4.9 Hz, 2H), 7.70–7.64 (m, 3H), 7.63–7.58 (m, 1H), 2.88 (s, 3H). ¹³C NMR (100 MHz, DMSO-d₆): δ 176.2, 158.1, 157.9, 157.7, 157.1, 154.5, 134.3, 132.7, 128.4, 127.3, 122.7, 108.3, 13.5. HRMS (ESI) m/z: calcd for C₁₆H₁₂Cl₂N₅O₃ [Μ + Η]⁺, 392.0312; found, 392.0304. HPLC analysis (Method A): t_R = 10.58 min, purity 99.76%.

(Z)-2-Chloro-N′-((3-(2,6-dichlorophenyl)-5-methylisoxazole-4-carbonyl)oxy)nicotinimidamide (46)

This compound was prepared according to the general procedure described above, upon reaction of the acyl chloride 34 with amidoxime 23. Reaction time: 2 h. The crude product was recrystallized from dichloromethane/n-pentane, to provide the pure derivative 46 in 80% yield, as a beige solid. mp 191–192 °C (CH₂Cl₂/n-pentane). ¹H NMR (400 MHz, DMSO-d₆): δ 8.52 (dd, J = 4.8, 1.9 Hz, 1H), 7.93 (dd, J = 7.5, 1.9 Hz, 1H), 7.70–7.65 (m, 2H), 7.62–7.57 (m, 1H), 7.50 (dd, J = 7.6, 4.8 Hz, 1H), 6.71 (br s, 2H, D₂O exch.), 2.88 (s, 3H). ¹³C NMR (151 MHz, DMSO-d₆): δ 175.7, 158.4, 157.9, 155.7, 150.9, 148.7, 140.3, 134.3, 132.4, 128.3, 127.9, 127.3, 122.9, 108.3, 13.6. HRMS (ESI) m/z: calcd for C₁₇H₁₂Cl₃N₄O₃ [Μ + Η]⁺, 424.9970; found, 424.9969. HPLC analysis (Method A): t_R = 10.80 min, purity 99.76%.

(Z)-N′-((3-(2,6-Dichlorophenyl)-5-methylisoxazole-4-carbonyl)oxy)-6-methylnicotinimidamide (47)

This compound was prepared according to the general procedure described above, upon reaction of the acyl chloride 34 with amidoxime 24. Reaction time: 8 h. The crude product was recrystallized from ethyl acetate, to provide the pure derivative 47 in 77% yield, as a pale yellow solid. mp 195–196 °C (EtOAc). ¹H NMR (400 MHz, DMSO-d₆): δ 8.71 (d, J = 1.9 Hz, 1H), 7.92 (dd, J = 8.1, 2.4 Hz, 1H), 7.70–7.66 (m, 2H), 7.63–7.58 (m, 1H), 7.33 (d, J = 8.1 Hz, 1H), 6.51 (br s, 2H, D₂O exch.), 2.87 (s, 3H), 2.51 (s, 3H, overlapping with DMSO-d₆). ¹³C NMR (100 MHz, DMSO-d₆): δ 176.0, 160.5, 158.3, 158.1, 155.7, 147.0, 134.7, 134.3, 132.6, 128.4, 127.4, 124.2, 122.8, 108.3, 24.0, 13.6. HRMS (ESI) m/z: calcd for C₁₈H₁₅Cl₂N₄O₃ [Μ + Η]⁺, 405.0516; found 405.0516. HPLC analysis (Method B): t_R = 8.64 min, purity 99.29%.

(Z)-N′-((3-(2,6-Dichlorophenyl)-5-methylisoxazole-4-carbonyl)oxy)nicotinimidamide (48)

This compound was prepared according to the general procedure described above, upon reaction of the acyl chloride 34 with amidoxime 25. Reaction time: 4 h. The crude product was recrystallized from dichloromethane/n-pentane, to provide the pure derivative 48 in 83% yield, as a beige solid. mp 183 °C (CH₂Cl₂/n-pentane). ¹H NMR (400 MHz, DMSO-d₆): δ 8.84 (dd, J = 2.2, 0.7 Hz, 1H), 8.69 (dd, J = 4.8, 1.6 Hz, 1H), 8.06–8.02 (m, 1H), 7.70–7.66 (m, 2H), 7.63–7.58 (m, 1H), 7.51–7.46 (m, 1H), 6.61 (br s, 2H, D₂O exch.), 2.88 (s, 3H). ¹³C NMR (151 MHz, DMSO-d₆): δ 176.0, 158.3, 158.1, 155.7, 151.6, 147.6, 134.6, 134.3, 132.6, 128.4, 127.4, 127.0, 123.5, 108.3, 13.6. HRMS (ESI) m/z: calcd for C₁₇H₁₃Cl₂N₄O₃ [Μ + Η]⁺, 391.0360; found, 391.0359. HPLC analysis (Method A): t_R = 10.85 min, purity 99.70%.

(Z)-N′-((3-(2,6-Dichlorophenyl)-5-methylisoxazole-4-carbonyl)oxy)-3-methylbenzimidamide (49)

This compound was prepared according to the general procedure described above, upon reaction of the acyl chloride 34 with amidoxime 26. Reaction time: 2 h. The crude product was recrystallized from ethyl acetate, to provide the pure derivative 49 in 90% yield, as a white solid. mp 197–198 °C (EtOAc). ¹H NMR (600 MHz, DMSO-d₆): δ 7.69–7.66 (m, 2H), 7.62–7.59 (m, 1H), 7.49 (s, 1H), 7.46–7.44 (m, 1H), 7.34–7.30 (m, 2H), 6.23 (br s, 2H, D₂O exch.), 2.87 (s, 3H), 2.33 (s, 3H). ¹³C NMR (151 MHz, DMSO-d₆): δ 175.9, 158.1, 157.3, 137.7, 134.3, 132.6, 131.3, 130.9, 128.4, 128.3, 127.5, 127.3, 123.9, 108.4, 20.9, 13.5. HRMS (ESI) m/z: calcd for C₁₉H₁₆Cl₂N₃O₃ [Μ + Η]⁺, 404.0564; found, 404.0558. HPLC analysis (Method A): t_R = 11.60 min, purity 98.45%.

(Z)-3-Methyl-N′-((5-methyl-3-phenylisoxazole-4-carbonyl)oxy)benzimidamide (50)

This compound was prepared according to the general procedure described above, upon reaction of the acyl chloride 36 with amidoxime 26. Reaction time: 2 h. The crude product was recrystallized from ethyl acetate, to provide the pure derivative 50 in 81% yield, as a white solid. mp 163–164 °C (EtOAc). ¹H NMR (600 MHz, DMSO-d₆): δ 7.67–7.63 (m, 2H), 7.56–7.50 (m, 4H), 7.47 (d, J = 6.4 Hz, 1H), 7.34–7.30 (m, 2H), 6.32 (br s, 2H, D₂O exch.), 2.76 (s, 3H), 2.34 (s, 3H). ¹³C NMR (151 MHz, DMSO-d₆): δ 175.5, 161.8, 159.2, 157.5, 137.6, 131.2, 131.1, 129.9, 128.9, 128.42, 128.38, 128.26, 127.3, 123.9, 107.9, 20.9, 13.3. HRMS (ESI) m/z: calcd for C₁₉H₁₈N₃O₃ [Μ + Η]⁺, 336.1343; found, 336.1347. HPLC analysis (Method A): t_R = 11.21 min, purity 98.94%.

(Z)-N′-((3-(2-Chlorophenyl)-5-methylisoxazole-4-carbonyl)oxy)-3-methylbenzimidamide (51)

This compound was prepared according to the general procedure described above, upon reaction of the acyl chloride 35 with amidoxime 26. Reaction time: 3 h. The crude product was recrystallized from ethyl acetate, to provide the pure derivative 51 in 86% yield, as a white solid. mp 176–177 °C (EtOAc). ¹H NMR (400 MHz, DMSO-d₆): δ 7.67–7.63 (m, 1H), 7.61–7.56 (m, 2H), 7.54–7.50 (m, 1H), 7.48 (s, 1H), 7.46–7.42 (m, 1H), 7.34–7.28 (m, 2H), 6.05 (br s, 2H, D₂O exch.), 2.81 (s, 3H), 2.32 (s, 3H). ¹³C NMR (100 MHz, DMSO-d₆): δ 175.5, 160.0, 158.5, 157.2, 137.7, 132.7, 131.7, 131.30, 131.25, 130.9, 129.5, 128.29, 128.26, 127.34, 127.26, 123.9, 108.8, 20.9, 13.2. HRMS (ESI) m/z: calcd for C₁₉H₁₇ClN₃O₃ [Μ + Η]⁺, 370.0953; found, 370.0957. HPLC analysis (Method A): t_R = 11.35 min, purity 99.82%.

(Z)-N′-((3-(2,6-Dichlorophenyl)-5-methylisoxazole-4-carbonyl)oxy)-1-methyl-1H-indole-5-carboximidamide (52)

This compound was prepared according to the general procedure described above, upon reaction of the acyl chloride 34 with amidoxime 29. Reaction time: 16 h. The crude product was recrystallized from dichloromethane/diethyl ether, to provide the pure derivative 52 in 82% yield, as a white solid. mp 188 °C (CH₂Cl₂/Et₂O). ¹H NMR (600 MHz, DMSO-d₆): δ 7.91 (s, 1H), 7.70–7.66 (m, 2H), 7.63–7.60 (m, 1H), 7.49–7.45 (m, 2H), 7.38 (d, J = 3.0 Hz, 1H), 6.49 (d, J = 3.0 Hz, 1H), 6.13 (br s, 2H, D₂O exch.), 3.80 (s, 3H), 2.88 (s, 3H). ¹³C NMR (100 MHz, DMSO-d₆): δ 176.0, 158.29, 158.28, 158.15, 137.5, 134.3, 132.6, 130.9, 128.5, 127.6, 127.5, 121.5, 119.8, 119.4, 109.6, 108.6, 101.2, 32.6, 13.5. HRMS (ESI) m/z: calcd for C₂₁H₁₇Cl₂N₄O₃ [Μ + Η]⁺, 443.0673; found, 443.0666. HPLC analysis (Method A): t_R = 11.50 min, purity 97.63%.

(Z)-N′-((3-(2,6-Dichlorophenyl)-5-methylisoxazole-4-carbonyl)oxy)-9-methyl-9H-purine-6-carboximidamide (53)

This compound was prepared according to the general procedure described above, upon reaction of the acyl chloride 34 with amidoxime 30. Reaction time: 16 h. The crude product was recrystallized from dichloromethane/diethyl ether, to provide the pure derivative 53 in 74% yield, as a white solid. mp 222–223 °C (CH₂Cl₂/Et₂O). ¹H NMR (600 MHz, DMSO-d₆): δ 9.01 (s, 1H), 8.65 (s, 1H), 7.69–7.67 (m, 2H), 7.63–7.59 (m, 1H), 6.64 (br s, 2H, D₂O exch.), 3.87 (s, 3H), 2.91 (s, 3H). ¹³C NMR (151 MHz, DMSO-d₆): δ 176.0, 158.3, 157.7, 154.2, 153.1, 151.1, 148.6, 145.4, 134.3, 132.5, 130.6, 128.4, 127.3, 108.2, 29.7, 13.5. HRMS (ESI) m/z: calcd for C₁₈H₁₄Cl₂N₇O₃ [Μ + Η]⁺, 446.0530; found, 446.0539. HPLC analysis (Method A): t_R = 10.58 min, purity 98.43%.

(Z)-N′-((3-(2,6-Dichlorophenyl)-5-methylisoxazole-4-carbonyl)oxy)-2-(pyridin-2-yl)acetimidamide (54)

This compound was prepared according to the general procedure described above, upon reaction of the acyl chloride 34 with amidoxime 27. Reaction time: 2 h. The crude product was recrystallized from ethyl acetate, to provide the pure derivative 54 in 76% yield, as a beige solid. mp 156–157 °C (EtOAc). ¹H NMR (400 MHz, DMSO-d₆): δ 8.49 (d, J = 4.1 Hz, 1H), 7.74 (td, J = 7.7, 1.8 Hz, 1H), 7.68–7.62 (m, 2H), 7.61–7.56 (m, 1H), 7.35 (d, J = 7.8 Hz, 1H), 7.26 (dd, J = 7.0, 5.3 Hz, 1H), 6.05 (br s, 2H, D₂O exch.), 3.54 (s, 2H), 2.81 (s, 3H). ¹³C NMR (151 MHz, DMSO-d₆): δ 175.6, 158.1, 158.0, 157.5, 156.0, 148.9, 136.7, 134.2, 132.4, 128.3, 127.4, 123.0, 122.0, 108.4, 38.9 (overlapping with DMSO-d₆), 13.4. HRMS (ESI) m/z: calcd for C₁₈H₁₄Cl₂N₄NaO₃ [Μ + Na]⁺, 427.0336; found, 427.0331. HPLC analysis (Method B): t_R = 8.46 min, purity 98.10%.

(Z)-N′-((3-(2,6-Dichlorophenyl)-5-methylisoxazole-4-carbonyl)oxy)-2-(pyridin-3-yl)acetimidamide (55)

This compound was prepared according to the general procedure described above, upon reaction of the acyl chloride 34 with amidoxime 28. Reaction time: 16 h. The crude product was recrystallized from ethyl acetate, to provide the pure derivative 55 in 71% yield, as a pale yellow solid. mp 175–176 °C (EtOAc). ¹H NMR (400 MHz, DMSO-d₆): δ 8.50 (s, 1H), 8.44 (d, J = 3.9 Hz, 1H), 7.72–7.62 (m, 3H), 7.60–7.55 (m, 1H), 7.34 (dd, J = 7.6, 4.9 Hz, 1H), 3.40 (s, 2H), 2.80 (s, 3H). ¹³C NMR (100 MHz, DMSO-d₆): δ 175.7, 158.3, 158.2, 158.1, 149.8, 147.9, 136.1, 134.3, 132.5, 132.1, 128.4, 127.4, 123.5, 108.4, 33.7, 13.5. HRMS (ESI) m/z: calcd for C₁₈H₁₅Cl₂N₄O₃ [Μ + Η]⁺, 405.0516; found, 405.0505. HPLC analysis (Method B): t_R = 8.17 min, purity 97.87%.

General Procedure for the Preparation of the Target Derivatives 56–61

Potassium hydroxide (14 mg, 0.25 mmol) was added into a solution of the corresponding derivative 37, 39, 45, 47, 48, or 55 (0.25 mmol) in anhydrous DMSO (0.4 mL), and each reaction mixture was stirred at room temperature for 30–45 min. Upon completion of the reaction, the mixture was diluted with water (30 mL) and extracted with dichloromethane (3 × 30 mL). The combined organic layers were washed with brine (2 × 20 mL), dried over sodium sulfate, and evaporated. The crude products were purified with column chromatography to provide the pure target derivatives 56–61.

5-(3-(2,6-Dichlorophenyl)-5-methylisoxazol-4-yl)-3-(pyridin-2-yl)-1,2,4-oxadiazole (56)

This compound was prepared according to the general procedure described above, starting from derivative 37. Reaction time: 30 min. Purification was carried out by using a mixture of cyclohexane/ethyl acetate as the eluent (from 90/10 up to 70/30, v/v) to provide the pure compound 56 in 91% yield, as a white solid. mp 189–190 °C (EtOAc/Et₂O). ¹H NMR (400 MHz, CDCl₃): δ 8.79 (d, J = 4.2 Hz, 1H), 7.96 (d, J = 7.9 Hz, 1H), 7.83 (t, J = 7.7 Hz, 1H), 7.49–7.40 (m, 4H), 3.03 (s, 3H). ¹³C NMR (151 MHz, CDCl₃): δ 174.5, 169.6, 168.1, 157.8, 150.4, 146.1, 137.7, 136.02, 131.95, 128.3, 127.2, 126.0, 123.7, 104.4, 13.8. HRMS (ESI) m/z: calcd for C₁₇H₁₁Cl₂N₄O₂ [Μ + Η]⁺, 373.0254; found, 373.0253. HPLC analysis (Method A): t_R = 11.74 min, purity 99.59%.

3-(5-Bromopyridin-2-yl)-5-(3-(2,6-dichlorophenyl)-5-methylisoxazol-4-yl)-1,2,4-oxadiazole (57)

This compound was prepared according to the general procedure described above, starting from derivative 39. Reaction time: 45 min. Purification was carried out by using a mixture of cyclohexane/ethyl acetate as the eluent (from 90/10 up to 80/20, v/v) to provide the pure compound 57 in 81% yield, as a white solid. mp 209–210 °C (EtOAc/Et₂O). ¹H NMR (600 MHz, CDCl₃): δ 8.82 (d, J = 2.0 Hz, 1H), 7.96 (dd, J = 8.3, 2.3 Hz, 1H), 7.83 (d, J = 8.3 Hz, 1H), 7.49–7.39 (m, 3H), 3.02 (s, 3H). ¹³C NMR (151 MHz, CDCl₃): δ 174.4, 169.6, 167.7, 157.7, 151.8, 144.7, 139.9, 135.9, 131.9, 128.3, 127.0, 124.5, 123.3, 104.2, 13.7. HRMS (ESI) m/z: calcd for C₁₇H₁₀BrCl₂N₄O₂ [Μ + Η]⁺, 450.9359; found, 450.9369. HPLC analysis (Method A): t_R = 12.42 min, purity 99.91%.

5-(3-(2,6-Dichlorophenyl)-5-methylisoxazol-4-yl)-3-(pyrimidin-2-yl)-1,2,4-oxadiazole (58)

This compound was prepared according to the general procedure described above, starting from derivative 45. Reaction time: 30 min. Purification was carried out by using a mixture of cyclohexane/ethyl acetate as the eluent (from 90/10 up to 60/40, v/v) to provide the pure compound 58 in 89% yield as a pale yellow solid. mp 213 °C (CH₂Cl₂/n-pentane). ¹H NMR (400 MHz, CDCl₃): δ 8.90 (d, J = 4.9 Hz, 2H), 7.44–7.34 (m, 4H), 3.00 (s, 3H). ¹³C NMR (100 MHz, CDCl₃): δ 174.5, 170.2, 167.9, 158.1, 157.5, 156.1, 135.7, 132.0, 128.3, 126.7, 122.3, 104.2, 13.7. HRMS (ESI) m/z: calcd for C₁₆H₁₀Cl₂N₅O₂ [Μ + Η]⁺, 374.0207; found, 374.0210. HPLC analysis (Method A): t_R = 11.05 min, purity 99.42%.

5-(3-(2,6-Dichlorophenyl)-5-methylisoxazol-4-yl)-3-(pyridin-3-yl)-1,2,4-oxadiazole (59)

This compound was prepared according to the general procedure described above, starting from derivative 48. Reaction time: 30 min. Purification was carried out by using a mixture of cyclohexane/ethyl acetate as the eluent (from 90/10 up to 70/30, v/v) to provide the pure compound 59 in 93% yield as a white solid. mp 183–184 °C (CH₂Cl₂/n-pentane). ¹H NMR (400 MHz, CDCl₃): δ 9.19 (br s, 1H), 8.73 (br s, 1H), 8.27 (dt, J = 8.0, 1.8 Hz, 1H), 7.50–7.38 (m, 4H), 3.01 (s, 3H). ¹³C NMR (100 MHz, CDCl₃): δ 174.4, 169.2, 166.5, 157.8, 152.1, 148.7, 135.9, 135.1, 132.0, 128.3, 127.1, 124.0, 123.1, 104.2, 13.7. HRMS (ESI) m/z: calcd for C₁₇H₁₁Cl₂N₄O₂ [Μ + Η]⁺, 373.0254; found, 373.0255. HPLC analysis (Method A): t_R = 12.31 min, purity 99.54%.

5-(3-(2,6-Dichlorophenyl)-5-methylisoxazol-4-yl)-3-(6-methylpyridin-3-yl)-1,2,4-oxadiazole (60)

This compound was prepared according to the general procedure described above, starting from derivative 47. Reaction time: 30 min. Purification was carried out by using a mixture of cyclohexane/ethyl acetate as the eluent (from 90/10 up to 80/20, v/v) to provide the pure compound 60 in 92% yield as a white solid. mp 185–186 °C (CH₂Cl₂/n-pentane). ¹H NMR (400 MHz, CDCl₃): δ 9.04 (d, J = 1.7 Hz, 1H), 8.15 (dd, J = 8.1, 2.1 Hz, 1H), 7.49–7.40 (m, 3H), 7.26 (d, J = 8.0 Hz, 1H, overlapping with CDCl₃), 3.01 (s, 3H), 2.63 (s, 3H). ¹³C NMR (151 MHz, CDCl₃): δ 174.3, 169.1, 166.6, 161.6, 157.8, 148.0, 136.0, 135.4, 132.0, 128.3, 127.2, 123.7, 120.4, 104.3, 24.6, 13.7. HRMS (ESI) m/z: calcd for C₁₈H₁₃Cl₂N₄O₂ [Μ + Η]⁺, 387.0411; found, 387.0410. HPLC analysis (Method A): t_R = 14.32 min, purity 99.73%.

5-(3-(2,6-Dichlorophenyl)-5-methylisoxazol-4-yl)-3-(pyridin-3-ylmethyl)-1,2,4-oxadiazole (61)

This compound was prepared according to the general procedure described above, starting from derivative 55. Reaction time: 45 min. Purification was carried out by using a mixture of cyclohexane/ethyl acetate as the eluent (from 90/10 up to 65/35, v/v) to provide the pure compound 61 in 84% yield as a pale yellow solid. mp 109–110 °C (CH₂Cl₂/n-pentane). ¹H NMR (600 MHz, CDCl₃): δ 8.53 (s, 1H), 8.51 (d, J = 4.1 Hz, 1H), 7.59 (d, J = 7.8 Hz, 1H), 7.44–7.36 (m, 3H), 7.23 (dd, J = 7.8, 4.9 Hz, 1H), 4.02 (s, 2H), 2.89 (s, 3H). ¹³C NMR (151 MHz, CDCl₃): δ 174.1, 169.0, 168.8, 157.7, 150.0, 148.4, 137.1, 135.9, 131.9, 131.3, 128.2, 127.1, 123.7, 104.3, 29.7, 13.5. HRMS (ESI) m/z: calcd for C₁₈H₁₃Cl₂N₄O₂ [Μ + Η]⁺, 387.0411; found, 387.0411. HPLC analysis (Method A): t_R = 17.34 min, purity 99.47%.

General Procedure for the Preparation of the Carboxylic Acids 70 and 71

Sodium hydroxide (160 mg, 4 mmol) was dissolved in a mixture of methanol (13 mL) and water (13 mL), and then the corresponding methyl ester 68 or 69 (3.4 mmol) was added. This reaction mixture was heated under stirring at 65 °C for 3 h. Upon completion of the reaction, the mixture was poured into cold water (50 mL) and acidified with 1 N HCl until pH 3. The solid was filtered under vacuum and air-dried. The crude product was then recrystallized from methanol to provide the pure derivative 70 or 71.

3-(2,6-Dibromophenyl)-5-methylisoxazole-4-carboxylic Acid (70)

This compound was prepared according to the general procedure described above, starting from derivative 68, in 83% yield, as a white solid. mp 233–235 °C (MeOH). ¹H NMR (600 MHz, DMSO-d₆): δ 13.06 (br s, 1H, D₂O exch.), 7.78 (d, J = 8.1 Hz, 2H), 7.37 (t, J = 8.1 Hz, 1H), 2.75 (s, 3H). ¹³C NMR (151 MHz, DMSO-d₆): δ 175.5, 161.8, 161.6, 132.6, 131.6, 131.3, 123.8, 109.4, 12.9. HRMS (ESI) m/z: calcd for C₁₁H₈Br₂NO₃ [Μ + Η]⁺, 361.8845; found, 361.8892.

3-(2,6-Dimethylphenyl)-5-methylisoxazole-4-carboxylic Acid (71)

This compound was prepared according to the general procedure described above, starting from derivative 69, in 77% yield, as a white solid. mp 177–178 °C (MeOH). ¹H NMR (600 MHz, DMSO-d₆): δ 12.81 (br s, 1H, D₂O exch.), 7.24 (t, J = 7.5 Hz, 1H), 7.11 (d, J = 7.5 Hz, 2H), 2.73 (s, 3H), 2.01 (s, 6H). ¹³C NMR (151 MHz, DMSO-d₆): δ 175.6, 162.4, 161.3, 136.5, 128.73, 128.66, 127.0, 109.3, 19.6, 13.1. HRMS (ESI) m/z: calcd for C₁₃H₁₄NO₃ [Μ + Η]⁺, 232.0969; found, 232.0999.

3-(2,6-Dibromophenyl)-5-methylisoxazole-4-carbonyl Chloride (72) and 3-(2,6-Dimethylphenyl)-5-methylisoxazole-4-carbonyl Chloride (73)

These compounds were synthesized according to the general method described for the preparation of derivatives 34–36 and were used immediately to the next step, with no further purification.

(Z)-N′-((3-(2,6-Dibromophenyl)-5-methylisoxazole-4-carbonyl)oxy)picolinimidamide (74)

This compound was prepared according to the general procedure described for the synthesis of the target derivatives 37–55, upon reaction of the acyl chloride 72 with amidoxime 18. Reaction time: 2 h. The crude product was recrystallized from ethyl acetate, to provide the pure derivative 74 in 97% yield, as a white solid. mp 198–199 °C (EtOAc). ¹H NMR (400 MHz, DMSO-d₆): δ 8.67–8.63 (m, 1H), 7.95–7.88 (m, 2H), 7.86 (d, J = 8.1 Hz, 2H), 7.58–7.53 (m, 1H), 7.44 (t, J = 8.1 Hz, 1H), 6.98 (br s, 1H, D₂O exch.), 5.67 (br s, 1H, D₂O exch.), 2.89 (s, 3H). ¹³C NMR (100 MHz, DMSO-d₆): δ 176.1, 161.2, 157.9, 154.7, 148.8, 147.6, 137.4, 133.2, 132.0, 131.0, 126.0, 123.7, 121.1, 107.9, 13.5. HRMS (ESI) m/z: calcd for C₁₇H₁₃Br₂N₄O₃ [Μ + Η]⁺, 480.9329; found, 480.9389. HPLC analysis (Method B): t_R = 9.14 min, purity 95.34%.

(Z)-5-Bromo-N′-((3-(2,6-dibromophenyl)-5-methylisoxazole-4-carbonyl)oxy)picolinimidamide (75)

This compound was prepared according to the general procedure described for the synthesis of the target derivatives 37–55, upon reaction of the acyl chloride 72 with amidoxime 20. Reaction time: 2 h. The crude product was recrystallized from ethyl acetate, to provide the pure derivative 75 in 91% yield, as a white solid. mp 229–230 °C (EtOAc). ¹H NMR (400 MHz, DMSO-d₆): δ 8.78 (d, J = 1.9 Hz, 1H), 8.17 (dd, J = 8.5 Hz, 2.3 Hz, 1H), 7.88–7.82 (m, 3H), 7.43 (t, J = 8.1 Hz, 1H), 6.91 (br s, 1H, D₂O exch.), 5.71 (br s, 1H, D₂O exch.), 2.88 (s, 3H). ¹³C NMR (100 MHz, DMSO-d₆): δ 176.1, 161.2, 157.8, 154.1, 149.6, 146.5, 140.1, 133.2, 132.0, 130.9, 123.7, 122.7, 122.5, 107.8, 13.5. HRMS (ESI) m/z: calcd for C₁₇H₁₂Br₃N₄O₃ [Μ + Η]⁺, 558.8434; found, 558.8474. HPLC analysis (Method B): t_R = 9.87 min, purity 97.35%.

(Z)-N′-((3-(2,6-Dimethylphenyl)-5-methylisoxazole-4-carbonyl)oxy)picolinimidamide (76)

This compound was prepared according to the general procedure described for the synthesis of the target derivatives 37–55, upon reaction of the acyl chloride 73 with amidoxime 18. Reaction time: 2 h. The crude product was recrystallized from ethyl acetate/n-pentane, to provide the pure derivative 76 in 90% yield, as a pale gray solid. mp 159–161 °C (EtOAc/n-pentane). ¹H NMR (400 MHz, DMSO-d₆): δ 8.65–8.59 (m, 1H), 7.94–7.83 (m, 2H), 7.58–7.49 (m, 1H), 7.33 (t, J = 7.8 Hz, 1H), 7.22 (t, J = 7.8 Hz, 2H), 6.76 (br s, 1H, D₂O exch.), 4.82 (br s, 1H, D₂O exch.), 2.84 (s, 3H) 2.08 (s, 6H). ¹³C NMR (100 MHz, DMSO-d₆): δ 176.7, 160.5, 158.4, 154.3, 148.8, 147.6, 137.4, 136.6, 129.5, 128.6, 127.5, 125.9, 121.0, 107.9, 19.6, 13.4. HRMS (ESI) m/z: calcd for C₁₉H₁₉N₄O₃ [Μ + Η]⁺, 351.1452; found, 351.1495. HPLC analysis (Method B): t_R = 9.14 min, purity 97.20%.

(Z)-5-Bromo-N′-((3-(2,6-dimethylphenyl)-5-methylisoxazole-4-carbonyl)oxy)picolinimidamide (77)

This compound was prepared according to the general procedure described for the synthesis of the target derivatives 37–55, upon reaction of the acyl chloride 73 with amidoxime 20. Reaction time: 2 h. The crude product was recrystallized from ethyl acetate, to provide the pure derivative 77 in 93% yield, as a white solid. mp 234–235 °C (EtOAc). ¹H NMR (400 MHz, DMSO-d₆): δ 8.75 (d, J = 1.9 Hz, 1H), 8.15 (dd, J = 8.5 Hz, 2.2 Hz, 1H), 7.80 (d, J = 8.5 Hz, 1H), 7.32 (t, J = 7.6 Hz, 1H), 7.20 (d, J = 7.6 Hz, 2H), 6.81 (br s, 1H, D₂O exch.), 4.83 (br s, 1H, D₂O exch.), 2.83 (s, 3H), 2.06 (s, 6H). ¹³C NMR (100 MHz, DMSO-d₆): δ 176.8, 160.5, 158.3, 153.8, 149.6, 146.6, 140.1, 136.6, 129.6, 128.6, 127.5, 122.6, 122.4, 107.8, 19.6, 13.4. HRMS (ESI) m/z: calcd for C₁₉H₁₈BrN₄O₃ [Μ + Η]⁺, 429.0557; found, 429.0606. HPLC analysis (Method B): t_R = 9.95 min, purity 97.03%.

Glossary

Abbreviations

ΔΔG_b,TI/MD

relative binding free energy

A_2AR

adenosine A_2A receptor

A_2BR

adenosine A_2B receptor

AF2

Alphafold 2

ARs

adenosine receptors

cAMP

3′,5′-cyclic adenosine monophosphate

CHO-K1

Chinese hamster ovary K1

CCK-8

cell counting kit-8

CL_int

intrinsic clearance

CPA

N⁶-cyclopentyl-adenosine

dppf

1,1′-ferrocenediyl-bis(diphenylphosphine)

DMA

dimethylacetamide

DMSO

dimethyl sulfoxide

FACS

fluorescence-activated cell sorting

FEP/MD

free energy perturbation coupled with molecular dynamics simulations

Ff14sb

amber14sb force field

Ff19sb

amber19sb force field

GPCRs

G protein-coupled receptors

hA₁R

human adenosine A₁ receptor

hA₃R

human adenosine A₃ receptor

HEK293T

human embryonic kidney 293T

IB-MECA

1-deoxy-1-[6-[[(3-iodophenyl)methyl]amino]-9H-purin-9-yl]-N-methyl-β-d-ribofuranuronamide

MAPK

mitogen-activated protein kinase

MD

molecular dynamics

MM/PBSA

Poisson–Boltzmann surface area continuum solvation

MM/GBSA

generalized Born surface area continuum solvation

MRS1220

(N-[9-chloro-2-(2-furanyl)-1,2,4-triazolo[1,5-c]quinazolin-5-yl]benzeneacetamide)

PMEMD

particle mesh Ewald molecular dynamics

P_app

permeability assay

NECA

5′-N-ethylcarboxamidoadenosine

NanoBRET

Nanoluciferease-based bioluminescence resonance energy transfer

Nluc

Nanoluciferease

RT

residence time

rmsd

root-mean-square deviation

SAR

structure–activity relationships

SBDD

structure-based drug design

SEM

standard error of the mean

TI/MD

thermodynamic integration coupled with molecular dynamics simulations method

TM

transmembrane

VS

virtual screening

WT

wild type

Data Availability Statement

The following link is provided to access the starting structures (docking poses) and output frames of the complexes between our revised model of the inactive hA₃R generated using the multistate AF2 method and ligands K18, 37–39, 56, 57, and 60 from the MD simulations: https://github.com/annachor/inactive_A3R_AF2-carbonyloxycarboximidamides_MDs.

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acs.jmedchem.4c01092.

• Molecular formula strings (CSV)

• Information on the pharmacological, computational, and chemical characterization of the tested compounds (PDF)

Author Contributions

^§ X.H., A.C., and P.K. contributed equally to this work. A.K., N.L., and G.L. contributed equally to this work. The manuscript was written through contributions of all authors. All authors have given approval to the final version of the manuscript.

The authors declare no competing financial interest.