Amino-3,5-Dicyanopyridines Targeting the Adenosine Receptors. Ranging from Pan Ligands to Combined A1/A2B Partial Agonists

The amino-3,5-dicyanopyridine derivatives belong to an intriguing series of adenosine receptor (AR) ligands that has been developed by both academic researchers and industry. Indeed, the studies carried out to date underline the versatility of the dicyanopyridine scaffold to obtain AR ligands with not only a wide range of affinities but also with diverse degrees of efficacies at the different ARs. These observations prompted us to investigate on the structure–activity relationships (SARs) of this series leading to important previously reported results. The present SAR study has helped to confirm the 1H-imidazol-2-yl group at R2 position as an important feature for producing potent AR agonists. Moreover, the nature of the R1 substituent highly affects not only affinity/activity at the hA1 and hA2B ARs but also selectivity versus the other subtypes. Potent hA1 and hA2B AR ligands were developed, and among them, the 2-amino-6-[(1H-imidazol-2-ylmethyl)sulfanyl]-4-[4-(prop-2-en-1-yloxy)phenyl]pyridine-3,5-dicarbonitrile (3) is active in the low nanomolar range at these subtypes and shows a good trend of selectivity versus both the hA2A and hA3 ARs. This combined hA1/hA2B partial agonist activity leads to a synergistic effect on glucose homeostasis and could potentially be beneficial in treating diabetes and related complications.


Introduction
Adenosine is a ubiquitous autacoid that has been reported to play important roles in various tissues by acting through four different adenosine receptors (ARs) classified as A 1 , A 2A , A 2B , and A 3 receptors [1]. These ARs belong to the G protein-coupled receptors superfamily, and each of them has different functions although with some overlap. Depending on their coupling to G-protein, they can either increase or decrease intracellular cAMP levels by affecting adenylate cyclase (AC) activity. However, other effector systems have been proposed [2]. Adenosine has the highest affinity for the A 1 and A 2A ARs, the lowest for A 2B AR, and intermediate affinity for the A 3 subtype. Selective agonists and antagonists have been developed for each of the ARs [3,4], and their discovery opened up new avenues for potential drugs treatment of a variety of conditions such as neurodegenerative disorders, On this basis, the 1H-imidazol-2-yl group was used to design other amino-3,5-dicyanopyridine derivatives in order to deeply investigate the influence of the R 1 substituent at the 4-position on AR affinity and selectivity. Moreover, other modifications were performed at R 2 to enrich what is known about how the nature of the substituent at this position modulates AR affinity and selectivity.
To replace the thiomethyl linker at the 6-position with an aminomethyl bridge, compounds 14-16 were synthesized (Scheme 2) starting from the 6-chloro derivative 42, obtained with a reported procedure [27], and the suitable commercially available amine. The reaction was carried out in the presence of triethylamine and in a refluxing mixture of THF/EtOH. The same procedure was applied for the synthesis of the 6-(1H-imidazol-1-yl)-derivative 17, which was obtained in high yield. From a structure-activity point of view, the new findings confirm the 1H-imidazol-2-yl group in the R 2 pendant as an important feature for producing potent AR agonists. Certain 2-amino-4-aryl-6-(1H-imidazol-2-yl-methylsulfanyl)-pyridine-3,5-dicarbonitrile compounds, belonging to the LUF series (Figure 1), were already reported to display nanomolar affinity for all the ARs, including the A 2B subtype for which they showed, in general, considerable affinity and efficacy [14]. The growing interest in this class is also dictated by the fact that non-nucleoside AR agonists seem to be more versatile for pharmacological studies, showing fewer species differences than the adenosine-like ones [4]. Moreover, the studies carried out to date are sufficient to underline the versatility of the amino-3,5-dicyanopyridine scaffold for producing AR ligands with not only a wide range of affinities but, interestingly, with different degrees of efficacy at the different ARs [11,12,[14][15][16][17][18][19].
On this basis, the 1H-imidazol-2-yl group was used to design other amino-3,5-dicyanopyridine derivatives in order to deeply investigate the influence of the R 1 substituent at the 4-position on AR affinity and selectivity. Moreover, other modifications were performed at R 2 to enrich what is known about how the nature of the substituent at this position modulates AR affinity and selectivity.

Pharmacological Assays
The amino-3,5-dicyanopyridines 1-20 were tested for their affinity at hA1, hA2A, and hA3 ARs, stably transfected in Chinese Hamster Ovary (CHO) cells, and were also studied as hA2B agonists by evaluating their stimulatory effect on cAMP production in CHO cells, stably expressing the hA2B AR. Compounds 3, 8, and 11, the most interesting in terms of combined affinity/activity at hA1 and hA2B ARs, were evaluated for their A1 pharmacological profile in the cAMP assay, where each compound was tested to assess its capability to modulate Forskolin-stimulated cAMP levels in the absence or presence of the A1AR agonist 2-chloro-N 6 -cyclopentyladenosine (CCPA) (1 nM). All pharmacological data are reported in Tables 1-2.

Molecular Docking Studies
To simulate the binding mode of the herein reported dicyanopyridines at hA1 and hA2B ARs, molecular docking studies were performed on the cryo-EM structure of the agonist-bound hA1AR (pdb code: 6D9H; 3.6-Å resolution [28]) and on a homology model of the hA2BAR developed by using the crystal structure of agonist-bound hA2AAR as a template (pdb code: 2YDO; 3.0-Å resolution [29]). The obtained hA2BAR homology model was checked using the Protein Geometry Monitor application within MOE (Molecular Operating Environment 2019.0101) [30] by inspecting the structural quality of the protein model (backbone bond lengths, angles, and dihedrals, Ramachandran φ-ψ dihedral plots, and quality of side chain rotamer and non-bonded contact). The two AR structures were then   [24] by exploiting the good property of the 6-phenylthio function as the leaving group in nucleophilic substitution [15]. The reaction was performed at 100 °C in DMF using an excess of 2-aminoethanol as nucleophile. Scheme 3. Reagents and conditions. (a) 2-aminoethanol, DMF, 100 °C.

Pharmacological Assays
The amino-3,5-dicyanopyridines 1-20 were tested for their affinity at hA1, hA2A, and hA3 ARs, stably transfected in Chinese Hamster Ovary (CHO) cells, and were also studied as hA2B agonists by evaluating their stimulatory effect on cAMP production in CHO cells, stably expressing the hA2B AR. Compounds 3, 8, and 11, the most interesting in terms of combined affinity/activity at hA1 and hA2B ARs, were evaluated for their A1 pharmacological profile in the cAMP assay, where each compound was tested to assess its capability to modulate Forskolin-stimulated cAMP levels in the absence or presence of the A1AR agonist 2-chloro-N 6 -cyclopentyladenosine (CCPA) (1 nM). All pharmacological data are reported in Tables 1-2.

Molecular Docking Studies
To simulate the binding mode of the herein reported dicyanopyridines at hA1 and hA2B ARs, molecular docking studies were performed on the cryo-EM structure of the agonist-bound hA1AR (pdb code: 6D9H; 3.6-Å resolution [28]) and on a homology model of the hA2BAR developed by using the crystal structure of agonist-bound hA2AAR as a template (pdb code: 2YDO; 3.0-Å resolution [29]). The obtained hA2BAR homology model was checked using the Protein Geometry Monitor application within MOE (Molecular Operating Environment 2019.0101) [30] by inspecting the structural quality of the protein model (backbone bond lengths, angles, and dihedrals, Ramachandran φ-ψ dihedral plots, and quality of side chain rotamer and non-bonded contact). The two AR structures were then To replace the thiomethyl linker at the 6-position with an aminomethyl bridge, compounds 14-16 were synthesized (Scheme 2) starting from the 6-chloro derivative 42, obtained with a reported procedure [27], and the suitable commercially available amine. The reaction was carried out in the presence of triethylamine and in a refluxing mixture of THF/EtOH. The same procedure was applied for the synthesis of the 6-(1H-imidazol-1-yl)-derivative 17, which was obtained in high yield. Elimination of the 6-substituent was realized in compound 18 which was synthesized by catalytic reduction (10% Pd/C) of 42 with hydrogen at 30 Psi in absolute EtOH.
Finally, the 6-[(2-hydroxyethyl)amino]-derivative 19 was synthesized (Scheme 3) starting from the intermediate 27 [24] by exploiting the good property of the 6-phenylthio function as the leaving group in nucleophilic substitution [15]. The reaction was performed at 100 • C in DMF using an excess of 2-aminoethanol as nucleophile.

Pharmacological Assays
The amino-3,5-dicyanopyridines 1-20 were tested for their affinity at hA 1 , hA 2A , and hA 3 ARs, stably transfected in Chinese Hamster Ovary (CHO) cells, and were also studied as hA 2B agonists by evaluating their stimulatory effect on cAMP production in CHO cells, stably expressing the hA 2B AR. Compounds 3, 8, and 11, the most interesting in terms of combined affinity/activity at hA 1 and hA 2B ARs, were evaluated for their A 1 pharmacological profile in the cAMP assay, where each compound was tested to assess its capability to modulate Forskolin-stimulated cAMP levels in the absence or presence of the A 1 AR agonist 2-chloro-N 6 -cyclopentyladenosine (CCPA) (1 nM). All pharmacological data are reported in Tables 1 and 2. performed using the MOE docking tool (induced fit docking and optimization protocol) and Gold software [31]. For each compound, the top-score docking pose at each AR model, according to at least two out of three scoring functions, was selected for final ligand-target interaction analysis.
Docking conformations at hA1 and hA2A ARs of selected dicyanopyridine compounds are reported in Figure 2 and Figure 3, respectively (see the Discussion Section). used as target for the docking analysis of the synthesized derivatives. The docking studies were performed using the MOE docking tool (induced fit docking and optimization protocol) and Gold software [31]. For each compound, the top-score docking pose at each AR model, according to at least two out of three scoring functions, was selected for final ligand-target interaction analysis. Docking conformations at hA1 and hA2A ARs of selected dicyanopyridine compounds are reported in Figure 2 and Figure 3, respectively (see the Discussion Section). used as target for the docking analysis of the synthesized derivatives. The docking studies were performed using the MOE docking tool (induced fit docking and optimization protocol) and Gold software [31]. For each compound, the top-score docking pose at each AR model, according to at least two out of three scoring functions, was selected for final ligand-target interaction analysis. Docking conformations at hA1 and hA2A ARs of selected dicyanopyridine compounds are reported in Figure 2 and Figure 3, respectively (see the Discussion Section). used as target for the docking analysis of the synthesized derivatives. The docking studies were performed using the MOE docking tool (induced fit docking and optimization protocol) and Gold software [31]. For each compound, the top-score docking pose at each AR model, according to at least two out of three scoring functions, was selected for final ligand-target interaction analysis. Docking conformations at hA1 and hA2A ARs of selected dicyanopyridine compounds are reported in Figure 2 and Figure 3, respectively (see the Discussion Section). used as target for the docking analysis of the synthesized derivatives. The docking studies were performed using the MOE docking tool (induced fit docking and optimization protocol) and Gold software [31]. For each compound, the top-score docking pose at each AR model, according to at least two out of three scoring functions, was selected for final ligand-target interaction analysis. Docking conformations at hA1 and hA2A ARs of selected dicyanopyridine compounds are reported in Figure 2 and Figure 3, respectively (see the Discussion Section). used as target for the docking analysis of the synthesized derivatives. The docking studies were performed using the MOE docking tool (induced fit docking and optimization protocol) and Gold software [31]. For each compound, the top-score docking pose at each AR model, according to at least two out of three scoring functions, was selected for final ligand-target interaction analysis. Docking conformations at hA1 and hA2A ARs of selected dicyanopyridine compounds are reported in Figure 2 and Figure 3, respectively (see the Discussion Section). used as target for the docking analysis of the synthesized derivatives. The docking studies were performed using the MOE docking tool (induced fit docking and optimization protocol) and Gold software [31]. For each compound, the top-score docking pose at each AR model, according to at least two out of three scoring functions, was selected for final ligand-target interaction analysis. Docking conformations at hA1 and hA2A ARs of selected dicyanopyridine compounds are reported in Figure 2 and Figure 3, respectively (see the Discussion Section). used as target for the docking analysis of the synthesized derivatives. The docking studies were performed using the MOE docking tool (induced fit docking and optimization protocol) and Gold software [31]. For each compound, the top-score docking pose at each AR model, according to at least two out of three scoring functions, was selected for final ligand-target interaction analysis. Docking conformations at hA1 and hA2A ARs of selected dicyanopyridine compounds are reported in Figure 2 and Figure 3, respectively (see the Discussion Section). used as target for the docking analysis of the synthesized derivatives. The docking studies were performed using the MOE docking tool (induced fit docking and optimization protocol) and Gold software [31]. For each compound, the top-score docking pose at each AR model, according to at least two out of three scoring functions, was selected for final ligand-target interaction analysis. Docking conformations at hA1 and hA2A ARs of selected dicyanopyridine compounds are reported in Figure 2 and Figure 3, respectively (see the Discussion Section).

g
Pharmaceuticals 2019, 12, x FOR PEER REVIEW 6 of 22 used as target for the docking analysis of the synthesized derivatives. The docking studies were performed using the MOE docking tool (induced fit docking and optimization protocol) and Gold software [31]. For each compound, the top-score docking pose at each AR model, according to at least two out of three scoring functions, was selected for final ligand-target interaction analysis. Docking conformations at hA1 and hA2A ARs of selected dicyanopyridine compounds are reported in Figure 2 and Figure 3, respectively (see the Discussion Section).
Pharmaceuticals 2019, 12, x FOR PEER REVIEW 6 of 22 used as target for the docking analysis of the synthesized derivatives. The docking studies were performed using the MOE docking tool (induced fit docking and optimization protocol) and Gold software [31]. For each compound, the top-score docking pose at each AR model, according to at least two out of three scoring functions, was selected for final ligand-target interaction analysis. Docking conformations at hA1 and hA2A ARs of selected dicyanopyridine compounds are reported in Figure 2 and Figure 3, respectively (see the Discussion Section).
Pharmaceuticals 2019, 12, x FOR PEER REVIEW 6 of 22 used as target for the docking analysis of the synthesized derivatives. The docking studies were performed using the MOE docking tool (induced fit docking and optimization protocol) and Gold software [31]. For each compound, the top-score docking pose at each AR model, according to at least two out of three scoring functions, was selected for final ligand-target interaction analysis. Docking conformations at hA1 and hA2A ARs of selected dicyanopyridine compounds are reported in Figure 2 and Figure 3, respectively (see the Discussion Section).
Pharmaceuticals 2019, 12, x FOR PEER REVIEW 6 of 22 used as target for the docking analysis of the synthesized derivatives. The docking studies were performed using the MOE docking tool (induced fit docking and optimization protocol) and Gold software [31]. For each compound, the top-score docking pose at each AR model, according to at least two out of three scoring functions, was selected for final ligand-target interaction analysis. Docking conformations at hA1 and hA2A ARs of selected dicyanopyridine compounds are reported in Figure 2 and Figure 3, respectively (see the Discussion Section).
Pharmaceuticals 2019, 12, x FOR PEER REVIEW 6 of 22 used as target for the docking analysis of the synthesized derivatives. The docking studies were performed using the MOE docking tool (induced fit docking and optimization protocol) and Gold software [31]. For each compound, the top-score docking pose at each AR model, according to at least two out of three scoring functions, was selected for final ligand-target interaction analysis. Docking conformations at hA1 and hA2A ARs of selected dicyanopyridine compounds are reported in Figure 2 and Figure 3, respectively (see the Discussion Section).
Pharmaceuticals 2019, 12, x FOR PEER REVIEW 6 of 22 used as target for the docking analysis of the synthesized derivatives. The docking studies were performed using the MOE docking tool (induced fit docking and optimization protocol) and Gold software [31]. For each compound, the top-score docking pose at each AR model, according to at least two out of three scoring functions, was selected for final ligand-target interaction analysis. Docking conformations at hA1 and hA2A ARs of selected dicyanopyridine compounds are reported in Figure 2 and Figure 3, respectively (see the Discussion Section).  used as target for the docking analysis of the synthesized derivatives. The docking studies were performed using the MOE docking tool (induced fit docking and optimization protocol) and Gold software [31]. For each compound, the top-score docking pose at each AR model, according to at least two out of three scoring functions, was selected for final ligand-target interaction analysis. Docking conformations at hA1 and hA2A ARs of selected dicyanopyridine compounds are reported in Figure 2 and Figure 3, respectively (see the Discussion Section).
Pharmaceuticals 2019, 12, x FOR PEER REVIEW 6 of 22 used as target for the docking analysis of the synthesized derivatives. The docking studies were performed using the MOE docking tool (induced fit docking and optimization protocol) and Gold software [31]. For each compound, the top-score docking pose at each AR model, according to at least two out of three scoring functions, was selected for final ligand-target interaction analysis. Docking conformations at hA1 and hA2A ARs of selected dicyanopyridine compounds are reported in Figure 2 and Figure 3, respectively (see the Discussion Section).  used as target for the docking analysis of the synthesized derivatives. The docking studies were performed using the MOE docking tool (induced fit docking and optimization protocol) and Gold software [31]. For each compound, the top-score docking pose at each AR model, according to at least two out of three scoring functions, was selected for final ligand-target interaction analysis. Docking conformations at hA1 and hA2A ARs of selected dicyanopyridine compounds are reported in Figure 2 and Figure 3, respectively (see the Discussion Section).
Pharmaceuticals 2019, 12, x FOR PEER REVIEW 6 of 22 used as target for the docking analysis of the synthesized derivatives. The docking studies were performed using the MOE docking tool (induced fit docking and optimization protocol) and Gold software [31]. For each compound, the top-score docking pose at each AR model, according to at least two out of three scoring functions, was selected for final ligand-target interaction analysis. Docking conformations at hA1 and hA2A ARs of selected dicyanopyridine compounds are reported in Figure 2 and Figure 3, respectively (see the Discussion Section).  used as target for the docking analysis of the synthesized derivatives. The docking studies were performed using the MOE docking tool (induced fit docking and optimization protocol) and Gold software [31]. For each compound, the top-score docking pose at each AR model, according to at least two out of three scoring functions, was selected for final ligand-target interaction analysis. Docking conformations at hA1 and hA2A ARs of selected dicyanopyridine compounds are reported in Figure 2 and Figure 3, respectively (see the Discussion Section).  used as target for the docking analysis of the synthesized derivatives. The docking studies were performed using the MOE docking tool (induced fit docking and optimization protocol) and Gold software [31]. For each compound, the top-score docking pose at each AR model, according to at least two out of three scoring functions, was selected for final ligand-target interaction analysis. Docking conformations at hA1 and hA2A ARs of selected dicyanopyridine compounds are reported in Figure 2 and Figure 3, respectively (see the Discussion Section). used as target for the docking analysis of the synthesized derivatives. The docking studies were performed using the MOE docking tool (induced fit docking and optimization protocol) and Gold software [31]. For each compound, the top-score docking pose at each AR model, according to at least two out of three scoring functions, was selected for final ligand-target interaction analysis. Docking conformations at hA1 and hA2A ARs of selected dicyanopyridine compounds are reported in Figure 2 and Figure 3, respectively (see the Discussion Section). used as target for the docking analysis of the synthesized derivatives. The docking studies were performed using the MOE docking tool (induced fit docking and optimization protocol) and Gold software [31]. For each compound, the top-score docking pose at each AR model, according to at least two out of three scoring functions, was selected for final ligand-target interaction analysis. Docking conformations at hA1 and hA2A ARs of selected dicyanopyridine compounds are reported in Figure 2 and Figure 3, respectively (see the Discussion Section).

Structure-Activity Relationships
The biological and pharmacological data for the newly synthesized amino-3,5-dicyanopyridine derivatives 1-20 are reported in Tables 1-2 together with those of the reference compounds LUF5833

Structure-Activity Relationships
The biological and pharmacological data for the newly synthesized amino-3,5-dicyanopyridine derivatives 1-20 are reported in Tables 1-2

Structure-Activity Relationships
The biological and pharmacological data for the newly synthesized amino-3,5-dicyanopyridine derivatives 1-20 are reported in Tables 1-2 together with those of the reference compounds LUF5833

Structure-Activity Relationships
The biological and pharmacological data for the newly synthesized amino-3,5-dicyanopyridine derivatives 1-20 are reported in Tables 1-2

Structure-Activity Relationships
The biological and pharmacological data for the newly synthesized amino-3,5-dicyanopyridine derivatives 1-20 are reported in Tables 1-2 together with those of the reference compounds LUF5833  [12]. h Reference [14].

Structure-Activity Relationships
The biological and pharmacological data for the newly synthesized amino-3,5-dicyanopyridine derivatives 1-20 are reported in Tables 1-2 together with those of the reference compounds LUF5833  [12]. h Reference [14]. showed the higher capability to increase cAMP production. c Potency of the novel compounds to increase Forskolin-stimulated cAMP levels.

Structure-Activity Relationships
The biological and pharmacological data for the newly synthesized amino-3,5-dicyanopyridine derivatives 1-20 are reported in Tables 1-2 [12]. h Reference [14]. showed the higher capability to increase cAMP production. c Potency of the novel compounds to increase Forskolin-stimulated cAMP levels.

Structure-Activity Relationships
The biological and pharmacological data for the newly synthesized amino-3,5-dicyanopyridine derivatives 1-20 are reported in Tables 1-2 together with those of the reference compounds LUF5833  [12]. h Reference [14]. showed the higher capability to increase cAMP production. c Potency of the novel compounds to increase Forskolin-stimulated cAMP levels.

Structure-Activity Relationships
The biological and pharmacological data for the newly synthesized amino-3,5-dicyanopyridine derivatives 1-20 are reported in Tables 1-2   1.52 ± 0.12 a Potency values (EC 50 ) are expressed as means ± SEM of four independent cAMP experiments, each performed in triplicate. The compounds have been tested on Forskolin at 5 µM concentration. b Efficacy of the novel compounds were normalized by using the efficacy value of the reference compound DPCPX (set at 100%) that showed the higher capability to increase cAMP production. c Potency of the novel compounds to increase Forskolin-stimulated cAMP levels.

Molecular Docking Studies
To simulate the binding mode of the herein reported dicyanopyridines at hA 1 and hA 2B ARs, molecular docking studies were performed on the cryo-EM structure of the agonist-bound hA 1 AR (pdb code: 6D9H; 3.6-Å resolution [28]) and on a homology model of the hA 2B AR developed by using the crystal structure of agonist-bound hA 2A AR as a template (pdb code: 2YDO; 3.0-Å resolution [29]). The obtained hA 2B AR homology model was checked using the Protein Geometry Monitor application within MOE (Molecular Operating Environment 2019.0101) [30] by inspecting the structural quality of the protein model (backbone bond lengths, angles, and dihedrals, Ramachandran ϕ-ψ dihedral plots, and quality of side chain rotamer and non-bonded contact). The two AR structures were then used as target for the docking analysis of the synthesized derivatives. The docking studies were performed using the MOE docking tool (induced fit docking and optimization protocol) and Gold software [31]. For each compound, the top-score docking pose at each AR model, according to at least two out of three scoring functions, was selected for final ligand-target interaction analysis.  and His251 6.52 . This double polar interaction with the receptor is also given by the 5′-amide groups of the hA2AAR agonists NECA [29], UK-432097 (6-[2,2-di(phenyl)ethylamino]-9-[(2R,3R,4S,5S)-5-(ethylcarbamoyl)-3,4-dihydroxyoxolan-2-yl]-N-[2-[(1-pyridin-2-ylpiperidin-4yl)carbamoylamino]ethyl]purine-2-carboxamide) [37] and CGS-21680 (4-[2-[[6-amino-9-(N-ethyl-β-D-ribofuranuronamidosyl)-9H-purin-2-yl]amino]ethyl]benzenepropanoic acid) [38], as observed from X-ray structures of the hA2AAR in complex with these two compounds. The insertion of an amide group at the 5′-position of nucleosides is a typical modification of the Ado scaffold to improve its affinity and potency at the hA2A and hA2B ARs. This observation helps to explain the high activity of compound 6 at the hA2BAR.
The potential use of derivatives belonging to this series as therapeutic agents depends not only on their pharmacodynamics but also pharmacokinetics, the latter including Adsorption, Distribution, Metabolism, Excretion, and Toxicity (ADME/T). A preliminary study of the theoretical ADMET profiles of compound 3 was performed using the methodology developed by pKCSM, a novel approach to the prediction of pharmacokinetic and toxicity properties [39]. A complete summary of all the parameters evaluated for the selected compound is reported in the Supplementary Materials. It is worth noting that no violation of the Lipinski's rule of five (RO5) (molecular weight (MW) < 500; log P < 5; number of H.-bond donors (HBD) < 5, and acceptors (HBA) < 10) was evidenced, making derivative 3 a promising drug-like candidate [40]. Nevertheless, while some bioavailability and toxicity parameters fall in the safety range, others indicate some points of concerns for compound 3. In interpreting the contrasting results, it is important to consider that in silico evaluation of the ADMET parameters for a group of approved antidiabetic drugs [41] evidenced not only violation of some of the RO5 (25%) but also the toxicity parameters as the major failures (2-33%).

Structure-Activity Relationships
The biological and pharmacological data for the newly synthesized amino-3,5-dicyanopyridine derivatives 1-20 are reported in Tables 1 and 2 together with those of the reference compounds LUF5833 (2-amino-6-[(1H-imidazol-2-ylmethyl)sulfanyl]-4-phenylpyridine-3,5-dicarbonitrile) [14] and P453 [12]. All the derivatives that interact with the hA 2B AR (2, 3, 5-8, 10-12) display EC 50 values from 12 to 260 nM behaving as partial agonists, with the only exception being compound 6, previously reported as trifluoroacetate salt [32], which shows a full agonist profile. Some compounds (17)(18)(19)(20), selected among those that had no activity at the hA 2B AR were also evaluated to investigate their antagonistic effect on cAMP production stimulated by 5 -(N-ethylcarboxamido)adenosine (NECA) 100 nM in hA 2B CHO cells. These functional studies reveal their inability to inhibit NECA-stimulated cAMP levels (I% ≤ 3) (data not shown). Most of the compounds have generally modest to null hA 2A and hA 3 AR affinity with the only exception of the two pairs of compounds 6 and 10 and 5 and 6 that bind, respectively, the hA 2A and hA 3 ARs with nanomolar affinity. Best results have been obtained in the hA 1 AR binding experiments. In particular, all the 6-(imidazole-2-ylmethyl)sulfanyl derivatives reported herein (1-13), with the only exception being compounds 1, 2, 9, and 12, have from high to good hA 1 AR affinity. In particular, some compounds showed K i values in the low nanomolar range and in general below 10 nM. Functional studies at the hA 1 subtype (data are reported in Table 2) indicated that the tested compounds 3, 8, and 11 behave as partial agonists and that their relative potencies correlate well with the K i values obtained in the hA 1 binding experiments. All the newly synthesized compounds showed different and not easily predictable behavior depending on both R 1 and R 2 substituents. However, the data shown in Table 1 indicate that the binding at the hA 2B AR subtype depends strictly on the substitution at the R 1 position [12].
Modifications of the para-substituent at the R 1 pendant of the lead compound P453 were performed to evaluate the influence on hA 2B AR activity and selectivity. 6-(Imidazole-2-ylmethyl)sulfanyl derivatives 1-5, bearing different cycloalkyl-or cycloalkenyl-methyloxy substituents at R 1 position, were synthesized. Only compounds 3 and 5 maintain the hA 2B AR activity in the nanomolar range, while derivative 2 and derivatives 1 and 4 are, respectively, scarcely active or inactive, probably due to the different steric hindrance of the para-substituent with respect to P453. In particular, derivative 5 is endowed with hA 2B AR activity comparable to that of the lead P453, despite a low or null selectivity versus the other AR subtypes. Introduction of a para-acetamido substituent on the 4-phenyl moiety led to compound 6, which is the only amino-3,5-dicyanopyridine in the whole series having a full agonist profile at the hA 2B AR. The para-acetamido substituent, in addition to possessing an oxygen atom as H-bond acceptor, also contains the NH donor group that could have some influence in determining the pharmacological profile of this derivative as observed for the dicyanopyridine of the LUF series ( Figure 1) [33,34].
Activity at the hA 2B AR is also affected by the electronic effects of the para-substituent on the 4-phenyl ring. In fact, the corresponding EC 50 values fall in the low nanomolar range for compounds 8 (para-fluoro) and 11 (para-thiomethyl) and move to higher values by passing to derivatives 10 (para-methyl) and 7 (para-chloro), up to the total inactivity of 9, bearing a para-trifuoromethyl group. Homologation of the 4-phenyl moiety of LUF5833 led to the 4-benzyl-susbstituted derivative 12, which maintains modest hA 2B activity. In contrast, its replacement with a methyl group was highly detrimental leading to compound 13, which is completely avoid of any activity at the hA 2B AR subtype. In this first set of compounds (1-13), bearing a 6-(imidazole-2-ylmethyl)sulfanyl moiety on the dicyanopyridine scaffold, the hA 1 AR affinity depends on the para-phenyl substitution as observed for hA 2B receptor activity trend. In fact, many compounds show mixed hA 1 /hA 2B combined affinity/activity. Looking at Table 1, we can observe different activity profiles-for example, compound 6, whose biological data fall in the same range of values (EC 50 at hA 2B AR and K i at the other subtypes) can be considered in effect as a pan ligand. Compound 3 is interesting in terms of its combined hA 1 /hA 2B AR partial agonist activity, exhibiting a good trend of selectivity versus both the hA 2A and hA 3 subtypes. The data of derivatives 8 and 11, which lose a little in selectivity but behave as hA 1 /hA 2B partial agonists, are also noteworthy. Moreover, compound 8, binding the hA 1 AR in the sub-nanomolar range, is the most potent hA 1 AR ligand among the herein reported dicyanopyridine series. When the 6-sulfanyl linker of LUF5833 was replaced by a 6-amino bridge, compound 14 was obtained. A dramatic decrease of hA 2B AR activity was observed with respect to the LUF derivative but the effect was detrimental also for the other ARs. Replacement of the imidazolyl moiety at R 2 position with different substituents in terms of both steric hindrance and capability to engage hydrogen-bonding led to derivatives 15-20, which show null affinity/activity for each of the AR subtypes. The only exception are compounds 15 and 20 endowed with hA 1 AR binding affinity in the micromolar range. Compounds 17-20, besides being inactive as hA 2B agonists, do not even work as antagonists.

Molecular Docking Investigation and In Silico ADMET Prediction.
The simulated binding mode at both the hA 1 and hA 2B AR structures, generally associated to the best score, presents the herein reported compounds oriented similarly to analogous derivatives previously reported as hAR ligands (Figure 2A) [12,35,36]. In detail, the pyridine scaffold is located in the AR cavity in correspondence to the purine moiety of Ado observed at the hA 1 AR cryo-EM structure and at the hA 2A AR template, and interacts with the side chains of hA 1 AR residues Phe171 extracellular loop (EL) 2 and Ile274 7.39 (Phe173 and Ile276 7.39 in hA 2B AR). For clarity, in this section, derivative 8 has been considered the reference compound to define the position of the substituents on the pyridine nucleus. Thus, starting from the N1 position, the amino and the sulfanyl functions occupy positions 2 and 6, respectively. The 3-cyano and 2-amino groups make H-bonds with the amide function of Asn254 6.55 . An additional H-bond is observed between the 2-amino group and the hA 1 AR residue Glu172 (EL2, Glu174 in hA 2B AR), while the 5-cyano group points toward transmembrane (TM) 2 (i.e., Ala66 2.61 in hA 1 AR, Ala64 2.61 in hA 2B AR) and TM7 (i.e., His278 7.43 in hA 1 AR, His280 7.43 in hA 2B AR) residues. The 2-sulfanyl substituent is located at the entrance of the binding cavity, providing interaction with residues of TM1, TM2, and TM7. Binding data show that in general an aryl moiety at R 1 is the type of substituent providing highest affinity at ARs. Its replacement with a smaller methyl group (compound 13) or a larger (and differently oriented) benzyl substituent (12) leads to a significant decrease (if not loss) of affinity. In this sense, the 4-aryl moiety well fits the depth of the binding cavity at both receptors, by occupying the available space in a more complete and efficient way with respect to the methyl (13) or benzyl group (12) (Figure 2B). Looking at the para-position of the 4-aryl moiety, biological data show that the size of the substituent at this level is critical for compound affinity, with the highest obtained with a small group. At the hA 1 AR receptor, the space available around the para-position is limited, allowing the insertion of small residues and not well tolerating ramified substituents. Accordingly, compounds bearing a bulky, non-linear para-substituent (compounds 4 and 6) are endowed with lower hA 1 AR affinity with respect to other compounds of the series active in the low nanomolar range.
At the hA 2B AR, the impact of the size of the para-substituent on compound affinity appears analogous to that observed at the hA 1 AR, with the exception of compound 6 bearing a para-acetamido group. This compound is among the most potent derivatives of the series at the hA 2B AR. Figure 3 shows the binding mode of compound 6 at the hA 2B AR, with indication of the key residues involved in ligand interaction. The para-acetamido function engages a double H-bond interaction with Thr89 3.36 and His251 6.52 . This double polar interaction with the receptor is also given by the 5 -amide groups of the hA 2A AR agonists NECA [29], UK-432097  [38], as observed from X-ray structures of the hA 2A AR in complex with these two compounds.
The insertion of an amide group at the 5 -position of nucleosides is a typical modification of the Ado scaffold to improve its affinity and potency at the hA 2A and hA 2B ARs. This observation helps to explain the high activity of compound 6 at the hA 2B AR.
The potential use of derivatives belonging to this series as therapeutic agents depends not only on their pharmacodynamics but also pharmacokinetics, the latter including Adsorption, Distribution, Metabolism, Excretion, and Toxicity (ADME/T). A preliminary study of the theoretical ADMET profiles of compound 3 was performed using the methodology developed by pKCSM, a novel approach to the prediction of pharmacokinetic and toxicity properties [39]. A complete summary of all the parameters evaluated for the selected compound is reported in the Supplementary Materials. It is worth noting that no violation of the Lipinski's rule of five (RO5) (molecular weight (MW) < 500; log P < 5; number of H.-bond donors (HBD) < 5, and acceptors (HBA) < 10) was evidenced, making derivative 3 a promising drug-like candidate [40]. Nevertheless, while some bioavailability and toxicity parameters fall in the safety range, others indicate some points of concerns for compound 3. In interpreting the contrasting results, it is important to consider that in silico evaluation of the ADMET parameters for a group of approved antidiabetic drugs [41] evidenced not only violation of some of the RO5 (25%) but also the toxicity parameters as the major failures (2-33%).

Conclusions
This study, which has led to some amino-3,5-dicyanopyridines differently substituted at R 1 and R 2 positions, provides greater insight into the SARs of this series as AR ligands. In particular, it has emerged that the affinity/activity at the hA 1 and hA 2B ARs depends strictly on the R 1 substitution which also affects the selectivity versus the other subtypes. Moreover, the 1H-imidazol-2-yl group at R 2 position was confirmed to be an essential feature for potent AR agonists. In fact, looking at the biological and pharmacological data of this series, it is usual to find a combined AR affinity/activity, even if most of the compounds have generally modest to null hA 2A and hA 3 AR affinity. Many of the dicyanopyridines herein reported interact with the hA 2B subtype with IC 50 values from 12 to 260 nM behaving as partial agonists. The best results were obtained in the hA 1 AR binding assays where most of the compounds show from high to good affinity at this receptor, the 2-amino-4-(4-fluorophenyl)-6-[(1H-imidazol-2-ylmethyl)sulfanyl]pyridine-3,5-dicarbonitrile (8) emerging as the most active compound with a partial agonist profile. The compounds showing a combined hA 1 /hA 2B AR partial agonist activity, as derivatives 3, 8, and 11, deserve the greatest attention.
In particular, the 2-amino-6-[(1H-imidazol-2-ylmethyl)sulfanyl]-4-[4-(prop-2-en-1-yloxy)phenyl]pyridine-3,5-dicarbonitrile (3) is active in the low nanomolar range at both these subtypes and shows a good trend of selectivity versus both the hA 2A and hA 3 ARs. This combined hA 1/ hA 2B partial agonist activity, leading to a synergistic effect on glucose homeostasis, could be of interest in the development of potentially useful agents for treating diabetes and its complications. The in silico calculated ADMET profiles indicated that some bioavailability and toxicity parameters fall in the safety range, while others suggest some points of concerns for compound 3. Thus, further studies will be devoted in the future to thoroughly investigate on the toxicity of compound 3. We are however confident in considering 3 as a promising lead for the development of new antidiabetics with improved pharmacokinetics and safer therapeutic potential.

General Methods
Analytical silica gel plates (Merck F254, Sigma-Aldrich, Milan, Italy), preparative silica gel plates (Merck F254, 2 mm), and silica gel 60 (Merck, 70-230 mesh) were used for analytical and preparative TLC, and for column chromatography, respectively. All melting points were determined on a Gallenkamp (U.K) melting point apparatus and are uncorrected. Elemental analyses were performed with a Flash E1112 Thermo Finnigan elemental analyzer for C, H, abd N (Thermo Fisher Scientific, Milan, Italy), and the results were within ±0.4% of the theoretical values. All final compounds revealed purity not less than 95%. The IR spectra were recorded with a Perkin-Elmer Spectrum RX I spectrometer (Perkin-Elmer, Milan, Italy) in Nujol mulls and are expressed in cm -1 . NMR spectra were recorded on a Bruker Avance 400 spectrometer (Bruker, Milan, Italy) (400 MHz for 1 H NMR and 100 MHz for 13 C NMR). The chemical shifts are reported in δ (ppm) and are relative to the central peak of the residual non-deuterated solvent, which was DMSO-d 6 . The following abbreviations are used: s = singlet, d = doublet, t = triplet, q = quartet, m = multiplet, br = broad, Ar = aromatic protons. Compounds 28-32 were synthesized as reported in reference [12]. When available, melting point and/or 1 H NMR values were in accordance to literature data.

Data Analysis
The protein concentration was determined according to a Bio-Rad method with bovine albumin as a standard reference. Inhibitory binding constant (Ki) values were calculated from those of IC 50 according to Cheng & Prusoff equation Ki = IC 50 /(1+[C*]/K D *), where [C*] is the concentration of the radioligand and K D * its dissociation constant [44]. K i and IC 50 values were calculated by non-linear regression analysis using the equation for a sigmoid concentration-response curve (Graph-PAD Prism, San Diego, CA, USA).

Molecular Modelling
Homology modelling and energy minimization studies were carried out using MOE (Molecular Operating Environment, C.C.G., Inc., 1255 University St., Suite 1600, Montreal, Quebec, Canada, H3B 3X3, version 2019.0101) suite [30]. A homology model of the hA 2B AR was built using the X-ray structure of the hA 2A AR in complex with Ado as template (pdb code: 2YDO; 3.0-Å resolution [29]). A multiple alignment of the hAR primary sequences was built within MOE as preliminary step. The boundaries identified from the used X-ray crystal structure of hA 2A AR were then applied for the corresponding sequences of the TM helices of the hA 2B AR. The missing loop domains were built by the loop search method implemented in MOE. Once the heavy atoms were modelled, all hydrogen atoms were added, and the protein coordinates were then minimized with MOE using the AMBER10 force field until the Root Mean Square (RMS) gradient of the potential energy was less than 0.05 kJ mol −1 Å −1 . Reliability and quality of the model were checked using the Protein Geometry Monitor application within MOE, which provides a variety of stereochemical measurements for inspection of the structural quality in a given protein, like backbone bond lengths, angles and dihedrals, Ramachandran ϕ-ψ dihedral plots, and quality of side chain rotamer and non-bonded contact. The cryo-EM structure of the agonist-bound hA 1 AR (pdb code: 6D9H; 3.6-Å resolution [28]) was added of hydrogen atoms and energetically minimized following the same protocol above described; the coordinates of the heavy atoms were kept fixed in this task.
All compound structures were docked into the binding site of the hA 2B and hA 1 ARs structures using as docking tools the Induced Fit docking protocol of MOE and the genetic algorithm docking tool of CCDC Gold [31]. The Induced Fit docking protocol of MOE is divided into the following several stages:

•
Conformational Analysis of ligands. The algorithm generated conformations from a single 3D conformation by conducting a systematic search. In this way, all combinations of angles were created for each ligand. • Placement. A collection of poses was generated from the pool of ligand conformations using Alpha Triangle placement method. Poses were generated by superposition of ligand atom triplets and triplet points in the receptor binding site. The receptor site points are alpha sphere centres which represent locations of tight packing. At each iteration, a random conformation was selected, a random triplet of ligand atoms and a random triplet of alpha sphere centres were used to determine the pose. • Scoring. Poses generated by the placement methodology were scored using the Alpha HB scoring function, which combines a term measuring the geometric fit of the ligand to the binding site and a term measuring hydrogen bonding effects. • Induced Fit. The generated docking conformations were subjected to energy minimization within the binding site and the protein sidechains are included in the refinement stage. In detail, the protein backbone is set as rigid while the side chains are not set to "free to move" but are set to "tethered", where an atom tether is a distance restraint that restrains the distance not between two atoms but between an atom and a fixed point in space.

•
Rescoring. Complexes generated by the Induced Fit methodology stage were scored using the Alpha HB scoring function. Gold tool was used with default efficiency settings through MOE interface, by selecting GoldScore as scoring function [31].

Conflicts of Interest:
The authors declare no conflict of interest.