A3 Adenosine Receptor Antagonists with Nucleoside Structures and Their Anticancer Activity

The overexpression of the A3 adenosine receptor (AR) in a number of cancer cell types makes it an attractive target for tumor diagnosis and therapy. Hence, in the search for new A3AR ligands, a series of novel 2,N6-disubstituted adenosines (Ados) was synthesized and tested in radioligand binding and functional assays at ARs. Derivatives bearing a 2-phenethylamino group in the N6-position were found to exert higher A3AR affinity and selectivity than the corresponding N6-(2,2-diphenylethyl) analogues. 2-Chloro-N6-phenylethylAdo (15) was found to be a potent full A3AR agonist with a Ki of 0.024 nM and an EC50 of 14 nM, in a cAMP accumulation assay. Unlike 15, the other ligands behaved as A3AR antagonists, which concentration-dependently reduced cell growth and exerted cytostatic activity on the prostate cancer cell line PC3, showing comparable and even more pronounced effects with respect to the ones elicited by the reference full agonist Cl-IB-MECA. In particular, the N6-(2,2-diphenylethyl)-2-phenylethynylAdo (12: GI50 = 14 µM, TGI = 29 µM, and LC50 = 59 µM) showed the highest activity proving to be a potential antitumor agent. The cytostatic effect of both A3AR agonist (Cl-IB-MECA) and antagonists (12 and other newly synthesized compounds) confirm previous observations according to which, in addition to the involvement of A3ARs, other cellular mechanisms are responsible for the anticancer effects of these ligands.


Introduction
Ubiquitous nucleosides and nucleotides are involved in several biological functions and they play a crucial role in some mechanisms, such as cell growth, migration, differentiation, bacterial-induced inflammation and growth factor secretion [1][2][3][4]. The nucleoside adenosine (Ado, 1, Figure 1) performs its function by interacting with the four A 1 , A 2A , A 2B and A 3 adenosine receptors (ARs) belonging to the superfamily of G protein-coupled receptors (GPCRs). The activation of A 1 AR and A 3 AR mainly produces a decrease of the intracellular cyclic adenosine monophosphate (cAMP) concentration, while stimulation of the A 2A AR and A 2B AR leads to an opposite effect.
In physiological conditions, this peculiar nucleoside is present in almost all cells and tissues in low concentrations (nM range), while in stress conditions such as hypoxia, which characterizes tumors, its concentration increases (µM range) [2]. A 3 ARs overexpression in various types of cancer cell has been reported in numerous studies [5] and has been demonstrated in lung, breast, melanoma, prostate, pancreatic, and liver cancers, as well as malignant pleural mesothelioma (MPM), glioblastoma, and lymphoma [2,6,7]. However, the particular issue is focused on the role of the A 3 AR in regulating cell proliferation and death, as this receptor acts differently depending on the type of tissue in which it is expressed [8]. On this basis, the A 3 AR is an attractive target for cancer diagnosis or for its ability to counteract the tumor grow, although there is evidence that, in some cancer types, activation of the A 3 AR promotes cell proliferation and survival, while in others it activates cytostatic and apoptotic pathways [9,10].
This dual behavior seems to be due to the different regulation/dysregulation of the Wnt pathway induced by A 3 AR activation in different tumor types [10]. In this regard, the mechanisms that influence the cancer cells' proliferation or death could also be due to the modulation of Ado levels, which can be modified by affecting the pathways of Ado generation, degradation, and elimination [11,12]. Furthermore, A 3 AR ligands conjugated with metallacarborans, containing iron and cobalt, protect cells from cross-resistance showed by anticancer drugs such as cisplatin, carboplatin, and doxorubicin [13,14].
The known full A 3 AR agonist, 2-chloro-N 6 -(3-iodobenzyl)adenosine-5 -N-methyluronamide (Cl-IB-MECA, also called CF-102 or Namodenoson, 2, Figure 1), has been extensively studied on tumor cell proliferation and this compound has been reported to inhibit cancer cell proliferation in in vitro and in vivo tumor models [15,16]. In a randomized, placebocontrolled phase II trial for hepatocellular carcinoma and moderate hepatic dysfunction, Cl-IB-MECA did not meet its primary end-point even though it showed a favorable safety profile and preliminary efficacy [17].
It is worthwhile to note that compound 3 behaves as a full A 3 AR agonist endowed with higher A 3 AR affinity and selectivity compared to Cl-IB-MECA [18,19]. The lack of anticancer activity of compound 3 indicated that the active compounds exerted their cytotoxic activity not only through the interaction with A 3 ARs and suggested the possible involvement of other cellular mechanisms. Indeed, further experiments led to hypothesize that the anticancer activity of compound 4 was due to its ability to induce apoptosis and to raise the level of reactive oxygen species (ROS) [18].
Starting from these observations, the purpose of this work was to identify new potential anticancer agents and to verify that the antitumor activity of these molecules is not closely associated with the A 3 AR binding affinity, confirming the possible intervention of other mechanisms. Hence, taking into account the structure of the two Ado derivatives 4 and 5, which showed cytotoxic activity in the three tumor cell lines mentioned above, in the present work, the synthesis of a new series of 2,N 6 di-substituted Ado derivatives bearing the same substituents of 4 and 5 at the N 6 -position and a chlorine atom, different alkynyl chains and a phenethylthio group at the 2-position, was undertaken ( Figure 2). The new compounds were tested in radioligand binding and functional assays, to assess their affinity (A 1 , A 2A , and A 3 ARs) and potency (A 2B AR) at human ARs. Furthermore, they were tested in a PC3 prostate cell line, to establish their antiproliferative and cytotoxic activities, and in a functional cAMP assay to evaluate their ability to activate the A 3 AR. Since the lack of sugar modification can lead to A 3 AR antagonists [19], this last experiment was aimed at verifying whether these compounds are able or not to activate the A 3 AR and, therefore, if they behave like Cl-IB-MECA.

Chemistry
Here, we report the synthesis of the newly nucleosides 10-13, 15, and 16, together with that of the previously reported compounds 4 and 5, which was carried out using a divergent approach starting from commercial guanosine (Schemes 1 and 2).
Commercial guanosine was converted in the protected 2-amino-6-chloropurineriboside 6, in two steps, as already described (Scheme 1) [20]. By a modification of the Sandmeyer reaction, consisting in a diazotization with isoamyl nitrite and diiodomethane, the 2iodo derivative 7 was prepared from 6, using dry THF as a solvent, without iodine as previously reported [21]. The 6-chloro-2-iodopurineriboside 7 was then reacted with 2phenethylamine or 2,2-diphenylethylamine, at r. t., to selectively displace the 6-chlorine atom. Then, methanolic ammonia was added at r. t. for the complete removal of protecting groups at the sugar moiety. The corresponding 2-iodo-N 6 -substituted Ados 8 and 9 were obtained as white powders after chromatography, with very good yields (Scheme 1).
The reaction of 7 to give 2-phenethylamino-N 6 -phenethylAdo (5) was performed in two steps. First, 7 was reacted with phenethylamine at 120 • C in a sealed vial, using DMF and potassium carbonate as a solvent and catalyst, respectively. Since these conditions led to the substitution of both the halogens in the 2-and 6-positions and the partial deprotection of the sugar moiety, the mixture, in a second step, was treated with methanolic ammonia at r. t. to have the complete sugar deprotection and to obtain 8 (Scheme 2). Reaction of 2-iodo-N 6 -phenylethylAdo (8) with phenylethylthiol, in dry DMF and in the presence of potassium carbonate at 120 • C, in a steel vial, furnished the 2-thioderivative 13, with good yield after chromatography (Scheme 1).
The 2-chloroadenosine derivatives 15 and 16 were synthesized starting from 6, which was reacted with isoamyl nitrite and antimony chloride in DCM, at 0 • C, to get the 2,6dichloropurineriboside 14. This compound was, in turn, treated with 2-phenethylamine and 2,2-diphenylethylamine, using ethanol as solvent and potassium carbonate as catalyst, to obtain the desired nucleosides 15 and 16 (Scheme 2).

Binding Assay at A 1 , A 2A , and A 3 ARs and Functional Studies at A 2B ARs
The new compounds 10-13 and 15-16, together with the 2-iodo nucleoside intermediates 8 and 9, were tested in radioligand binding assay at human recombinant ARs, expressed in Chinese hamster ovary (CHO) cells, to evaluate their affinity for A 1 , A 2A , and A 3 AR subtypes. [ 3 H]CCPA (2-chloro-N 6 -cyclopentylAdo), [ 3 H]NECA (5 -N-ethylcarboxamidoAdo), and [ 3 H]HEMADO (2-hexynyl-N 6 -methylAdo) were used as respective radioligands [22,23]. The results are reported as Ki values in nM (± standard errors) ( Table 1). In the case of A 2B AR, the potency of selected compounds was determined through a functional GloSensor cAMP assay [22]. Since their EC 50 values resulted > 30 µM, these data are not shown in Table 1. Cl-IB-MECA and the already known A 3 AR ligands 4 and 5 are reported as reference compounds. Cl-IB-MECA is an A 3 AR full agonist endowed with high affinity (Ki = 1.4 nM) and a very good A 3 AR selectivity of 886 and 3829 fold vs. the A 1 and A 2A ARs, respectively. The newly synthesized compounds behave, in general, as A 3 AR ligands with high affinity (Ki in the nM and sub-nM range) and various degrees of selectivity. In particular, the already reported 2-phenethylamino-N 6 -phenethylAdo (5) possesses a very high A 3 AR affinity with a KiA 3 AR = 0.33 nM and a selectivity of thousands of times versus both the other AR subtypes. This compound was the A 3 AR ligand endowed with the most balanced affinity and selectivity values for the A 3 AR, better also than the reference Cl-IB-MECA.
The replacement of the 2-phenethylamino group of 5 with halogens led to an increase of affinity compared to the A 1 AR subtype, but different results were found on the other subtypes. In fact, the 2-iodo derivative 8 showed a reduced affinity for both A 2A and A 3 ARs, compared to 5. On the contrary, for the 2-chloro derivative 15 there was an increase in affinity for these subtypes. It is worthwhile to note that the 2-chloro-N 6 -phenethylAdo (15) possesses an A 3 AR Ki in the pM range and a very high selectivity vs. the A 2A subtype (27,500 fold), although it is less selective for the A 1 AR with respect to the parent compound 5 and Cl-IB-MECA.
Substitution of the phenethylamino group at the 2-position of 5 with different alkynyl chains, as in compounds 10 and 11, or with the isosteric phenethylthio group, as in compound 13, has resulted in derivatives that maintain high A 3 AR affinity, although with lower A 3 AR selectivity. The presence of a bulky 2,2-diphenylethyl group in the N 6 -position, as in compounds 9, 16, 4, and 12, caused a general decrease in A 3 AR affinity and selectivity (see 9 vs. 8, 16 vs. 15, 4 vs. 10, and 12 vs. 11). It should be noted that, also in this series of compounds, the 2-chloro derivative showed the highest A 3 AR affinity (16; KiA 3 AR = 0.13 nM).

Antiproliferative and Cytotoxic Assays
To assess the antitumor activity of all the synthesized compounds 4, 5, 8-13, 15, 16, their antiproliferative and cytotoxic effects were evaluated in the PC3 cell line by the sulforhodamine B (SRB) assay, according to the National Cancer Institute protocol [25]. The reference A 3 AR agonist Cl-IB-MECA was tested with the same protocol for comparison, in order. The compounds were used at concentrations of 1, 10, 25, 50, and 100 µM for 48 h at 37 • C. The antitumor activity was estimated by measurements of three parameters: Growth Inhibition 50 (GI 50 ), the compound concentration (µM) required to inhibit 50% net of cell growth; Total Growth Inhibition (TGI), the compound concentration (µM) required to inhibit 100% of cell growth; Lethal Concentration 50 (LC 50 ), the compound concentration (µM) required to kill 50% of the initial cell number. The results are shown in Table 2 and Figure S1, along with A 3 AR affinity.
The compounds demonstrated to concentration-dependently reduce cell growth of the PC3 prostate cancer cell line. Most of them showed a significant inhibitory effect on cell proliferation and a pronounced cytotoxic activity comparable to the one elicited by Cl-IB-MECA, after 48 h exposure. Compounds 5, 8, 13, and 15, bearing a phenethyl chain at the N 6 -position, exhibited a lower ability to inhibit cell proliferation and cell survival than Cl-IB-MECA, which showed a GI 50 = 18 µM, TGI = 44 µM, and LC 50 = 110 µM. On the contrary, compounds 10 and 11, bearing the same N 6 -phenethyl group and an alkynyl chain at the 2-position, exhibited a higher cytostatic effect than Cl-IB-MECA (10: GI 50 = 13 µM and 11: GI 50 = 2.5 µM and TGI = 19 µM). It should be noted that the N 6 -phenethyl-2-phenylethynylAdo (11) is endowed with the best cytostatic activity, among all the tested derivatives. Furthermore, two of these compounds, 13 and 15, at a concentration of 500 µM, were not able to kill all the cells, in fact their LC 50 was higher than 500 µM. It should be noted that, also in the series of the N 6 -(2,2diphenylethyl) derivatives 4, 9, 12, and 16, the compounds bearing an alkynyl chain at the 2-position showed a pronounced cytostatic activity, being more active than Cl-IB-MECA. These two compounds were found to be more potent than the reference also for their cytotoxic effects (4 and 12: LC 50 of 80 and 59 µM, respectively, compared to Cl-IB-MECA: LC 50 = 110 µM). The N 6 -(2,2-diphenylethyl)-2-phenylethynylAdo (12) resulted as the most active compound with a GI 50 = 14 µM, TGI = 29 µM, and LC 50 = 59 µM. The observation that the A 3 AR ligands 5 (Ki = 0.33 nM), 15 (Ki = 0.024 nM), and 16 (Ki = 0.13 nM), were not those with the highest cytostatic and cytotoxic effects, despite their remarkable A 3 AR affinity, seems to demonstrate that the antitumor activity of the tested 2,N 6 -disubstitued Ados is not strictly correlated to their affinity for the A 3 AR subtype.

Functional Activity at Human A 3 AR
Among the four AR subtypes, A 3 AR appears to be the most sensitive to small chemical changes in its ligands. Indeed, various Ado derivatives, which were previously claimed as full or partial A 3 AR agonists, behaved subsequently as A 3 AR antagonists [24]. In several papers we have reported that the MECA and NECA derivatives substituted in 2-position with alkynyl chains and bearing a methyl or a methoxy group at the N 6 -position behave as full A 3 AR agonists like Cl-IB-MECA [19,26]. Conversely, the corresponding Ado derivatives lose efficacy, proving to act as partial agonists or antagonists on this receptor subtype [19,26]. Since the here presented newly synthesized A 3 AR ligands possess an intact ribose moiety in their structure, a functional assay was performed to measure their ability to activate the A 3 AR subtype. Then, the intrinsic activity of the selected compounds 4, 10-12, and 15, chosen on the basis of their very high A 3 AR affinity or antitumor activity, was evaluated. In particular, the ligands were analyzed in a functional experiment to assess their ability to inhibit forskolin-stimulated cAMP production through the human A 3 AR, compared to the full agonist Cl-IBMECA [22].
Results showed that the reference compound Cl-IB-MECA and the 2-chloro-N 6 -phenethylAdo (15) were able to completely counteract the stimulation of adenylyl cyclase induced by 10 µM forskolin, so behaving as A 3 AR full agonists. On the contrary, compounds 4 and 10-12 did not affect cAMP levels induced by forskolin, when tested alone, but were able, to a different extent, to lower the effect of the full agonist, demonstrating behavior as A 3 AR antagonists. For all the compounds, the EC 50 or IC 50 values were calculated and the results are reported in Table 3 and Figure S2. The EC 50 of Cl-IB-MECA was in a low nM range, according to the literature data [27], while the full agonist 15, which showed the best A 3 AR affinity (KiA 3 AR = 0.024 nM), exhibited an EC 50 of 14 nM. Among the four antagonists, 10 (IC 50 = 31 nM) and 11 (IC 50 = 79 nM), bearing a phenethyl group at the N 6 -position, showed a greater potency than 4 (IC 50 = 380 nM) and 12 (IC 50 = 153 nM), which present a 2,2-diphenylethyl chain in the same position. This order of potency was in close agreement with the compounds' A 3 AR binding affinities. These data confirm that the efficacy of the A 3 AR ligands is closely related to the nature of the substituents present at the different positions of the nucleoside structure [19]. It is worth noting that the presence of a chlorine atom in 2-position of N 6 -phenethylAdo furnished the full A 3 AR agonist 15 like Cl-IB-MECA but, unlike the latter, 15 showed no cytotoxic activity at concentrations up to 500 µM. On the other hand, compounds 4 and 10-12, which behaved as A 3 AR antagonists, demonstrated comparable and even greater antitumor activity than Cl-IB-MECA in PC3 prostate cells, at least in the cAMP accumulation assay. These inconsistencies, due to the antitumor activity of both the full agonist Cl-IB-MECA and the A 3 AR antagonists 4 and 10-12 on the same tumor cell type, cannot be explained by the different role played by the A 3 AR in the regulation of cell proliferation and death, depending on the tissue type in which it is expressed, but they suggest the possible involvement of other cellular mechanisms, as we have previously hypothesized [18]. On the other hand, this is not the first report in which A 3 AR antagonists have been found to exert anticancer activity [28,29].

Membrane Preparation
All pharmacological methods followed the procedures as described earlier [22]. In brief, membranes for radioligand binding were prepared from CHO cells stably transfected with human adenosine receptor subtypes through two centrifugations at different speeds. The first low-speed (1000 g) centrifugation allowed the removing of cell fragments and nuclei, while the second, performed at high speed (100,000 g), allowed the precipitation of the crude membrane fractions.
The resulting membrane pellet was resuspended in the buffer used for the respective binding experiments, frozen in liquid nitrogen and stored in aliquots at −80 • C.

Binding Assay
The binding affinity of the novel compounds was evaluated using radioligand competition experiments in CHO cells stably expressing hA 1  The potency of compounds at the hA 2B receptor (expressed on CHO cells) was determined through GloSensor cAMP Assay.

Functional Agonism or Antagonism at A 2B or A 3 ARs in GloSensor cAMP Assay
CHO cells stable expressing A 2B or A 3 ARs and the plasmid encoding the biosensor were used to study functional agonism and antagonism of understudy ligands. The desired cell number was incubated in equilibration medium containing a 3% v/v GloSensor cAMP reagent stock solution, 10% FBS and 87% of CO 2 independent media. After 2 h of incubation, cells were dispensed in wells of 384 well plate and when a steady-state basal signal was obtained the agonism profile was studied, adding to wells the reference agonist NECA or compounds at different concentrations. In the case of the G i coupled receptor A 3 AR Forskolin (FSK) 10 µM was added 10 min after the agonists and various luminescence readings were performed at different incubation times.
The antagonist profile of the compounds was evaluated by assessing their ability to counteract an agonist-induced increase or decrease (A 2B and A 3 ARs, respectively) of cAMP accumulation [22]. The cells were incubated with different antagonist concentrations and then treated with a fixed dose of NECA (10 µM for A 2B AR and 1 µM for A 3 AR). FSK 10 µM was added 10 min after the agonists and various luminescence reads were performed at different incubation times.
Responses were expressed as percentage of the maximal relative luminescence units (RLU) and concentration-response curves were fitted by a nonlinear regression with the Prism program (GraphPAD Software, San Diego, CA, USA). The agonist or antagonist profile of compounds was expressed as EC 50 or IC 50 , respectively [30]. Each concentration was tested three-five times in duplicate and the values are given as the mean ± S.E.

Cell Growth Inhibition Assay
To determine the effects of compounds 4, 5, 8-13, 15, 16, and Cl-IB-MECA on PC3 cell line, an SRB assay was performed. Cells were seeded at 4 × 10 3 cells/well in a 96-well microplate. Twenty-four hours later, cultures were treated with increasing concentrations of compounds, 1, 10, 25, 50, 100 µm, for 48 h at 37 • C in a 5% CO 2 atmosphere and 95% relative humidity. At the same time (t = 0), and after drug treatments, 100 µL of 10% (w/v) TCA were added to each well, incubated for 1 h at 4 • C, washed with deionized water, and dried at room temperature. One hundred microliters of SRB solution were added to each well, incubated for 10 min at room temperature, rinsed four times with 1% (v/v) acetic acid, and allowed to dry at room temperature. Finally, 100 µL of 10 mM Tris base solution (pH 10.5) was added to each well, and the absorbance was measured at 515 nm in a microplate reader (BioTek Instruments, Winooski, VT, USA). The absorbance at t = 0 was compared with the absorbance at the end of the experiment to determine cell growth in treated cells compared with control cells. The antitumor activity was estimated by measurements of three parameters: Growth Inhibition 50 (GI 50 ), the drug concentration (µM) required to inhibit 50% net of cell growth; Total Growth Inhibition (TGI), the drug concentration (µM) required to inhibit 100% of cell growth; and Lethal Concentration 50 (LC 50 ), the drug concentration (µM) required to kill 50% of the initial cell number. LC 50 , GI 50 and TGI values are shown as mean ± standard deviation (SD) of three different experiments calculated using GraphPad Prism 5.0 (GraphPad Software, San Diego, CA, USA).

Statistical Analysis
The statistical significance was determined by Student's t-test by using, as control, the reference compound 2-Cl-IB-MECA. * p < 0.05.

Conclusions
Novel di-substituted Ado derivatives bearing a N 6 -phenethyl or a more steric hindered N 6 -(2,2-diphenylethyl) group combined with halogens, alkynes or a phenylethylthio chain in 2-position were synthesized and tested in radioligand binding assays at A 1 , A 2A , and A 3 ARs. Selected compounds were also tested in a functional assay at A 2B ARs and, in most cases, were found not able to activate the receptor at concentrations up to 30 µM. In the binding studies, all the compounds were found to be A 3 AR ligands with high affinity (Ki in the nM and sub-nM range) and different degree of selectivity. In general, compounds bearing a 2-phenethylamino group in N 6 -position resulted endowed with higher affinity and selectivity than the corresponding analogues bearing a more hindered 2,2-diphenylethyl group in the same position. Furthermore, in the PC3 prostate cancer cell line, all the compounds concentration-dependently reduced the cell growth and most of them showed a significant inhibitory effect on proliferation and a pronounced cytotoxic activity comparable to that of Cl-IB-MECA. In particular, the ligands with the best cytostatic properties were those bearing a N 6 -(2,2-diphenylethyl) group, indicating that the antitumor activity is not closely related to the affinity for the A 3 AR subtype. Finally, functional cAMP assays demonstrated that compounds endowed with the best cytotoxic activity behave as A 3 AR antagonists, confirming the hypothesis that other cellular mechanisms are involved in the anticancer properties of these A 3 AR ligands. Therefore, further experiments will be needed to explain the precise pathways responsible for their anticancer effects.