Synthesis and evaluation of adenosine derivatives as A₁, A_2A, A_2B and A₃ adenosine receptor ligands containing boron clusters as phenyl isosteres and selective A₃ agonists

Abstract

A series of adenosine and 2′-deoxyadenosine pairs modified with a 1,12-dicarba-closo-dodecaborane cluster or alternatively with a phenyl group at the same position was synthesized, and their affinity was determined at A₁, A_2A, A_2B and A₃ adenosine receptors (ARs). While AR affinity differences were noted, a general tendency to preferentially bind A₃ AR over other ARs was observed for most tested ligands. In particular, 5′-ethylcarbamoyl-N⁶-(3-phenylpropyl)adenosine (18), N⁶-(3-phenylpropyl)-2-chloroadenosine (24) and N⁶-(3-phenylpropyl)adenosine (40) showed nanomolar A₃ affinity (K_i 4.5, 6.4 and 7.5 nM, respectively). Among the boron cluster-containing compounds, the highest A₃ affinity (K_i 206 nM) was for adenosine derivative 41 modified at C2. In the matched molecular pairs, analogs bearing boron clusters were found to show lower binding affinity for adenosine receptors than the corresponding phenyl analogs. Nevertheless, interestingly, several boron cluster modified adenosine ligands showed significantly higher A₃ receptor selectivity than the corresponding phenyl analogs: 7 vs. 8, 15 vs. 16, 17 vs. 18.

1. Introduction

Adenosine is a naturally occurring nucleoside that is ubiquitously present throughout the body as a metabolic intermediate. In addition to its metabolic role within cells as a building block of RNA and many small biomolecules, such as the chemical energy source ATP, adenosine is released into the extracellular space. Extracellular adenosine, through interaction with the extracellular domain of G-protein-coupled receptors (A₁ AR, A_2A AR, A_2B AR and A₃ AR) exerts its regulatory role in the nervous, circulatory, endocrine, and immunological systems [1]. Adenosine levels are markedly increased in response to hypoxia, inflammation, or ischemia, and serve as a protectant under these conditions [2]. The prospect of harnessing these effects specifically for therapeutic purposes is attractive [3].

The availability of selective AR (adenosine receptor) ligands facilitates research on therapeutic applications of strategies modulating AR activity and, in some cases, has already provided clinical candidates [4]. Currently, the A_2A AR agonist regadenoson is being evaluated as an agent that reduces COVID-19-induced lung injury through inhibition of hyperinflammation (ClinicalTrials.gov Identifier: NCT04606069).

In the search for potent and selective AR ligands, a plethora of adenosine and nonnucleoside ligands have been synthesized and tested [1,3–6]. Although selective AR modulators have promise for numerous therapeutic applications in practice, this goal, as in the case of many other targets, is not easy to achieve. The difficulties in the use of AR ligands in clinical practice result mainly from: i) low tissue specificity of the adenosine receptors in the body, and ii) cross-activation of nontarget AR receptors by highly potent selective ligands. One of the possible approaches to solve this problem is to develop ligands that do not cause 100% activation of the receptors but activate only a smaller percentage. In this way, not all adenosine receptors are targeted, but only those essential for achieving desirable clinical outcomes.

The development of nucleoside-boron cluster conjugates, including adenosine derivatives, is a new avenue of research in this field that is attracting considerable attention. Medicinal chemists are increasingly utilizing boron clusters (polyhedral boron cages) as a new generation of 3-dimensional (3D), abiotic privileged scaffolds, modifiers and pharmacophores in diverse bioactive molecular designs [7–10]. As we show in this article, the use of boron clusters as adenosine-modifying units may be one of the possible solutions enabling the construction of more selective AR ligands with sufficient potency at the same time.

The properties of boron clusters that are useful in drug design are listed below. (1) These clusters can form unique noncovalent interactions such as dihydrogen bonds due to the hydride character of H atoms, σ-hole bonding or ionic interactions; as a result, the types of interactions of boron cluster-containing compounds with biological targets may differ from purely organic molecules. (2) Boron clusters exhibit a spherical or ellipsoidal geometry and rigid 3D arrangement, which offer versatile platforms for 3D molecular construction. (3) The lipophilicity, amphiphilicity or hydrophilicity depend on the type of boron cluster used and allow the tuning of pharmacokinetics and bioavailability. (4) Boron clusters exhibit chemical stability and simultaneous susceptibility to functionalization, as well as (5) bioorthogonality, stability in biological environments and a decreased susceptibility to metabolism [7,9,11]. Due to the above reasons, the interactions of boron cluster drugs with their protein targets may be slightly different from the lock and key system that has evolved for purely organic ligands. Therefore, the use of boron cluster modification can be expected to allow us to challenge known drug targets in a new way and test new targets that have not been accessible thus far.

All ARs belong to the same class of purinergic G protein-coupled receptors; however, although they build on the same molecular blueprint, they differ in structural details and function. Due to the unique 3D structure of the boron cluster and the hydric character of hydrogen atoms of B–H bonds allowing for the formation of unconventional hydrogen bonds, the boron cluster modifier seems to allow for better sensing of these subtle differences between A₃ and A₁, A_2A, A_2B ARs than in the case of derivatives bearing purely organic groups.

We previously reported the chemical synthesis of a range of nucleoside-boron cluster conjugates, including those formed from adenosine, and evaluated their activity as blood platelet aggregation inhibitors [12], reactive oxygen species inhibitors [13], antivirals [14,15] and antitumor agents [16]. Some of these compounds are potential ligands for purinergic receptors and the observed biological activities might be attributed to the modulation of AR activity [13].

Here, a series of 30 adenosine derivatives that we designed modified with either inorganic, boron clusters (carborane groups) or organic phenyl groups for comparison were synthesized, and the effects of organic and inorganic modifications on the affinity of the ligands for the A₁, A_2A, A_2B and A₃ ARs were evaluated and compared. While no chemical modification can guarantee the transformation of any compound into drug candidates, we believe that our investigation of the chemistry and biology of innovative adenosine derivatives containing boron clusters expands the range of AR ligand types available for testing. Moreover, the effect of modification with a boron cluster on the increased selectivity of adenosine ligands towards A₃ AR demonstrated herein is a finding of practical importance.

2. Results

2.1. The synthesis of boron cluster donors

The synthesis of 1-(3-azidopropyl)-1,12-dicarba-closo-dodecaborane (2) [13] was performed using a method analogous to the method described previously for 1,7-dicarba-closo-dodecaborane isomer [17] by treating bromide 1 [18] with sodium azide in a dimethylformamide (DMF) solution to provide compound 2 with a 93% yield (Scheme 1). Alternatively, a tosyl derivative instead of bromide can be used analogously, as described for 1-(3-azidopropyl)-1,2-dicarba-closo-dodecaborane synthesis [19]. The 1-(3-aminopropyl)-1,12-dicarba-closo-dodecaborane (4) [13] was obtained from azide 2 via catalytic hydrogenation according to the procedure published for isomeric 1,2- or 1,7-dicarba-closo-dodecaborane [17,19].

Scheme 1.

Methods of amine 4 synthesis: i. NaN₃, DMF, rt; ii. QHSO₄ in 2 M NaOH, HN(Boc)₂ in CH₂Cl₂, reflux; iii. H₂/10% Pd/C, CH₃OH, rt; iv. HCl in CH₃OH, rt.

Another method used to synthesize amine 4 is based on the conversion of bromide 1 to di-Boc-protected amine 3, as originally described for the synthesis of an analogous derivative of the 1,2-dicarba-closo-dodecaborane isomer. Amine 4, in the form of a hydrochloride, was obtained by removing the Boc groups with methanolic hydrogen chloride (Scheme 1) [20,21]. Next, 1-(4-aminobutyl)-1,12-dicarba-closo-dodecaborane (6), an aminoalkyl derivative of 1,12-dicarba-closo-dodecaborane with an extended butyl linker, was synthesized as described in a previous study [22] using a method based on phthalimide chemistry [19,22].

2.2. Synthesis of adenosine derivatives bearing boron clusters

Previously described compounds (34 [23], 37 [24], 38 [25], 39 [12], 40 [26], 41 [13], 42 [27], 43 [12], 47 [16] and 49 [16]), new adenosine derivatives containing 1,12-dicarba-closo-dodecaborane clusters (45 and 51) or phenyl groups (44, 46, 48, 50 and 52, Table 1) are described elsewhere [28], and a series of adenosine derivative pairs containing inorganic boron cluster modifications or alternative organic phenyl modifications (7, 8, 15–18, 23, 24, 27, 28 and 33, 35, 36) described here were synthesized (Table 1 and Schemes 2–5).

Table 1.

Binding affinity of compounds 7, 8, 15–18, 23, 24, 27, 28, 33–52, and unmodified adenosine and 2’-deoxyadenosine at the human A₁, A_2A, A_2B and A₃ adenosine receptors.

Compound/reference	Structure	Ki [nM] or percent of inhibition [10 μM]
		A1 AR	A2A AR	A2B AR	A3 AR
7		29 ± 1 %a	2 ± 1 %a	1 ± 1 %a	444a
8 26		864a	11 ± 2 %a	19 ± 3%a	32.9a
15		40 ± 2 %a	9 ± 1 %a	7 ± 2 %a	437a
16		803a	7100a	3330a	9.2a
17		18 ± 3 %a	1 ± 1 %a	4 ± 4 %a	34 ± 4 %a
18		49 ± 2 %a	42 ± 1 %a	1380a	4.5a
23		38 ± 2 %a	1 ± 1 %a	5940a,b	2200a
24		3340a	27 ± 5 %a	24 ± 1 %a	6.44a
27		25 ± 1 %a	1 ± 1 %a	4670a	10 ± 3 %a
28		13,300a,b	14 ± 2 %a	1 ± 1 %a	8770a,b
33		29 ± 7 %	17.1 ± 8 %	nd	48 ± 12 %
34 24		11 ± 3 %c	3 ± 2 %c	n.d.	856 ± 291c
35		16 ± 7 %	18 ± 4 %	n.d.	9 ± 13 %
36		11 ± 11 %	15 ± 10 %	n.d.	8 ± 13 %
37 25		6 ± 10 %	7 ± 2 %	n.d.	19 ± 14 %
38 26		4 ± 10 %	3 ± 0 %	n.d.	29 ± 12 %
39 14		32 ± 1 %	0 ± 1 %	n.d.	0 ± 12 %
40 27		78.8 ± 34.3	2050 ± 40	n.d.	54.6 ± 5.8
41 14		2870 ± 1690	3070 ± 530	n.d.	206 ± 31
42 28		646 ± 4d	1040 ± 200d	n.d.	42 ± 2d
43 13		6 ± 6 %	10 ± 5 %	n.d.	1 ± 8 %
44 29		40 ± 9 %	1990 ± 180	n.d.	879 ± 151
45 29		20 ± 6 %	7080 ± 240	n.d.	2780 ± 800
46 29		1940 ± 1340	22 ± 9 %	n.d.	8610 ± 1390
47 17		15 ± 1 %	17 ± 2.7 %	n.d.	4390 ± 2020
48 29		27.6 ± 7.2 %	15 ± 22 %	n.d.	1230 ± 220
49 17		27 ± 9 %	12 ± 14 %	n.d.	25 ± 8 %
50 29		15 ± 20 %	−5 ± 6 %	n.d.	54 ± 6 %
51 29		16 ± 4 %	7 ± 2 %	n.d.	34 ± 7 %
52 29		28 ± 28 %	20 ± 6 %	n.d.	44 ± 9 %
Ade		13 ± 1 %	3 ± 1 %	4 ± 2 %	10 ± 1 %
dAde		5 ± 1 %	2 ± 1 %	12 ± 3%	5 ± 1 %

^aThe radioligand replacement assay of compounds 7, 8, 15–18, 23, 24, 27 and 28 was performed under a contractual service agreement with Plataforma de Screening de Farmacos (USEF), 15782 Santiago de Compostela, Spain; Binding affinity assays of compounds 33–52 were performed in authors’ laboratory. Compounds 37–40 were double tested, for USEF data please see Supplementary Information, Table S2.

^bCompound 28 for A₁ AR and A₃ AR and 23 for A_2B AR does not completely displace the radioligand. K_i value extrapolated by the analysis software might be not accurately estimated. n.d. = not determined.

^cCompound 34: A₁ K_i > 100 μM (human), A_2A K_i > 100 μM (human), A_2B K_i > 100 μM (human), A₃ K_i = 790 nM (human).[23]

^dCompound 42: A₁ K_i = 701 ± 13 nM (rat), A_2A K_i = 109 ± 28 nM (rat); ref. 6: A₁ K_i = 391 nM (human), A_2A K_i = 363 nM (human), A_2B K_i > 100 μM (human), A₃ K_i = 16 nM (human).[27]

Scheme 2.

Synthesis of adenosine derivatives with boron cluster 7 and phenyl 8 modifications, with an extended linker between the modification and exo-amino group of adenosine: i. Et₃N in n-BuOH was used as the solvent.

Scheme 5.

Synthesis of adenosine derivatives with a boron cluster (33, 35) or a phenyl group (34, 36) attached to the carbon in the eight position of adenine nucleobase: i. Et₃N/CuI/Pd(PPh₃)₂Cl₂ (29), ii. H₂/10% Pd/C, CH₃OH.

Based on the results of an in silico virtual screen (data not shown), adenosine derivatives with an extended butyl linker between the boron cluster or phenyl modification and the exocyclic amino group of adenosine were synthesized. Adenosine was modified at position 6 via the substitution of the chlorine in commercially available 6-chloroadenosine (5) with a 1-(4-aminobutyl)-1,12-dicarba-closo-dodecaborane (6), which was used as the boron cluster donor, or alternatively, with its fully organic counterpart, 4-phenylbutylamine (9).

The substitution of chlorine at position 6 of compound 5 with a boron cluster containing amine 6 was performed using a method developed in a previous study for the synthesis of a homolog of compound 7 bearing a propyl linker [13]. N⁶-(4-Phenylbutyl) adenosine (8) was synthesized using a published procedure (Scheme 3) [26,29]. 5′-N-Ethylcarboxamidoadenosine (NECA) derivatives (Scheme 3) modified at position 6 with a boron cluster or phenyl group were synthesized by substituting chlorine at position 6 using a method analogous to the procedure described for compounds 7 and 8 above. The key intermediate 5′-ethylamino-5′-oxo-6-chloroadenosine (13) was obtained using published procedures with some modifications.

Scheme 3.

Synthesis of adenosine derivatives with boron cluster (15, 17) or phenyl (16, 17) modification, with a propyl (17, 18) or butyl (15, 16) linker between the modification and exo-amino group of adenosine, and with 5′-ethylamino group: i. Et₃N in n-BuOH; ii. 2 M HCl in H₂O/1,4-dioxane (1:1); iii. DMP, PTSA in DMF; iv. TEMPO, BAIB in CH₃CN/H₂O; v. SOCl₂ in DMF/CHCl₃, vi. EtNH₂/THF.

First, 6-chloroadenosine (5) was treated with 2,2-dimethoxypropane in the presence of catalytic amounts of p-toluenesulfonic acid [30] to produce 2′,3′-O,O-isopropylidene-6-chloroadenosine (10) [31]. The remaining 5′-hydroxyl functional group of compound 10 was oxidized upon treatment with TEMPO/BAIB (2,2,6,6-tetramethyl-1-piperidinyloxyl/bisacetoxyiodobenzene) providing 2′,3′-O,O-isopropylidene-6-chloro-5′-oxoadenosine (11) [32]. The 4′-carboxylic group in compound 11 was amidated by first transforming it into C4′-acid chloride and then performing in situ ethylamine-mediated aminolysis to produce compound 12 [33]. The final removal of the isopropylidene protection in compound 12 was achieved under acidic conditions using 2 M hydrochloric acid, providing 5′-ethylamino-5′-oxo-6-chloroadenosine adenosine (13) [33]. The introduction of various amino substituents into the purine 6-position was achieved by treating compound 13 with boron cluster-containing amines 4 or 6 or alternatively with their fully organic counterparts, amines 9 or 14, yielding modified NECA derivatives 15–18.

The synthesis of 2-chloroadenosine or 2-chloro-2′-deoxyadenosine modified at the 6-position with a 1-(3-aminopropyl)-1,12-dicarba-closo-dodecaborane (4) or fully organic counterpart, 3-phenylpropylamine (14) (Scheme 4), was performed using the methods described above for compounds 7, 8 and 15–18. The much higher reactivity of the chlorine substituent at position 6 compared to position 2 was exploited [34,35]. Sonogashira-type cross coupling [36] between 8-bromo-2′-deoxyadenosine (30) and 2-ethynyl-1,12-dicarba-closo-dodecaborane (31) [37], a boron cluster donor, or phenylacetylene (32), its organic counterpart containing a triple bond, in the presence of bis(triphenylphosphine) palladium(II) dichloride (Pd(PPh₃)₂Cl₂; 29) was used to tether the modification to nucleoside acceptor 30 (Scheme 5).

Scheme 4.

Synthesis of 2-chloro-adenosine and 2-chloro-2′-deoxyadenosine derivatives with boron cluster (23, 27) or phenyl (24, 28) modification: i. DMF/Et₃N; ii. NH₃ in CH₃OH.

Although almost all adenosine derivatives described in the present study are modified at different positions of the adenine nucleobase, compounds 37 and 38 containing boron cluster modification or alternative organic phenyl modifications at the 2′-position of the sugar residue were also obtained (Scheme S1).

Compound 37 bearing a boron cluster was synthesized via the formation of a formacetal linkage using a previously described method [24]. Compound 38 containing a phenyl group at the 2’ position was obtained analogously using 3-phenyl-1-propanol instead of 1-(3-hydroxypropyl)-1,12-dicarba-closo-dodecaborane (Scheme S1 showing the synthesis of compounds 37 and 38 is available in Supplementary Information) [24,25,38].

2.3. Affinity and selectivity for adenosine receptors

The affinity and selectivity of the compounds were assayed by the radioligand binding method with the use of transfected cell lines overexpressing only one AR subtype. The radioligand replacement assay based on the competitive binding of the tested compound and a ligand with known affinity to the receptor was performed under a contractual service agreement with Plataforma de Screening de Farmacos (USEF), 15782 Santiago de Compostela, Spain (compounds 7, 8, 15–18, 23, 24, 27, and 28) or in the NIH laboratory (compounds 33–52) (Table 1). Compounds 37–40 were double tested; for USEF data, please see Supplementary Information, Table S2. Adenosine and 2′-deoxyadenosine were tested under the same conditions for comparison.

For the assessment of compounds 7, 8, 15–18, 23, 24, 27 and 28 as radioligands for A₁ and A_2B receptors, 1,3-[³H]-dipropyl-8-cyclopentylxanthine ([³H]-DPCPX), a selective antagonist of A₁, was administered at different concentrations, 1 nM and 25 nM. An adenine isostere and a high-affinity selective antagonist, 2,8-substituted [1,2,4]triazolo[1,5-a] [1,3,5]triazine ([³H]-ZM241385), were selected for the A_2A receptor. For the A₃ receptor, 1-[³H]-(6-amino-9H-purin-9-yl)-1-deoxy-N-ethyl-β-d-ribofuronamide ([³H]-NECA) was used as a radioligand in the replacement assay. Nonspecific binding was determined in the presence of 6-N-R-phenylisopropyladenosine (R-PIA) for A₁ and A₃, and NECA for A_2A and A_2B receptors. CGS15943, ZM241385, MRS1220 and XAC were used as the standard controls for the A₁, A_2A, A_2B and A₃ receptors, respectively. CHO-A₁, HeLa-A_2A, HEK-A_2B and HeLa-A₃ cells expressing human A₁, A_2A, A_2B or A₃ receptors were used.

Radioligand binding assays for compounds 33–52 were performed using HEK293 cells expressing human A₁, A_2A, or A₃ receptors and standard radioligands, as previously described [39]. Unlike the binding performed by USEF, which utilized antagonist radioligands for A₁ AR and A_2A AR, these determinations at human A₁, A_2A and A₃ ARs used agonist radioligands (concentration, K_D value; nM): [³H]-N⁶-R-phenylisopropyladenosine (1.0, 1.5), [³H]-2-[p-(2-carboxyethyl)phenyl-ethylamino]-5′-N-ethylcarboxamidoadenosine (10, 16.2), [¹²⁵I]-N⁶-(4-amino-3-iodobenzyl)adenosine-5′-N-methyluronamide (0.2, 1.22), respectively. Nonspecific binding was determined using adenosine 5′-N-ethyluronamide (10 μM) [39].

2.4. Assessment of the functions of compounds 7, 8, 15, 16, 23 and 24 as ligands for the human A₃ receptor

Among compounds that interact with the A₃ receptor, a set of three pairs of boron cluster- or phenyl-modified adenosine derivatives with the highest affinity towards the A₃ AR in their categories (“boronic” compounds vs. “phenylic” compounds) was selected for the functional study. The set of boronic compounds included compounds 7 (A₃ K_i = 444 nM), 15 (A₃ K_i = 437 nM) and 23 (A₃ K_i = 2200 nM), and the phenylic compounds included compounds 8 (A₃ K_i = 33 nM), 16 (A₃ K_i = 9.2 nM) and 24 (A₃ K_i = 6.4 nM). The interaction of agonists with the A₃ receptor induces cAMP accumulation in the CHO-A₃ cell line. The cAMP level was determined in CHO-A₃ cells incubated with increasing concentrations of selected compounds to describe the A₃ AR agonistic profiles of the compounds. Subsequently, the ability of the compounds to inhibit NECA-induced cAMP accumulation was evaluated to determine their antagonistic features. The values of the equilibrium dissociation constant (K_B) for antagonists and agonist potency (50% maximal response, EC₅₀) and maximal agonist effects (E_max) were determined. Based on the results of the functional study, all tested compounds exclusively exhibited A₃ receptor agonist characteristics, and none inhibited the function of the A₃ receptor (Table 2). Phenyl derivatives were the most potent agonists of the A₃ receptor compared to boronic agonists. Among the boronic compounds, compound 7 was distinguished as the most effective activator of A₃ receptors. The assessments of the affinity and biological activity of the phenylic compounds produced consistent results. Compound 24 displayed the highest A₃ receptor affinity (K_i = 6.4 nM) and the highest potency in inducing cAMP production in CHO-A₃ cells (EC₅₀ = 10.4 nM), whereas the lowest affinity compound, 8, was the least potent among the analyzed phenylic compounds (EC₅₀ = 54.3 nM).

Table 2.

Results of the functional assessment of the tested compounds (cAMP, EC₅₀, nM) and the relationship of cAMP levels and A₃ receptor binding.

Compound	KB	EC50 [nM]	% Emax
7	no antagonist	271	103
8	no antagonist	54.3	102
15	no antagonist	1010	99
16	no antagonist	16.3	102
23	no antagonist	3760	104
24	no antagonist	10.4	93

2.5. Assessment of the affinity of the selected compounds towards the A_2A receptor

Based on our previous observations of the activity of some adenosine and 2′-deoxyadenosine modified with boron clusters as platelet function modulators [12], and the known role of A_2A AR activation in antiplatelet activity, the compound library obtained in the present study was evaluated as an A_2A ligand and in addition to K_i, as shown in Table 1.

The IC₅₀ values for selected boron cluster/phenyl-modified pairs, 7, 8; 15, 16; 17, 18; 23, 24; 39, 40, with preferential affinity for A₃ but also binding to the same extent as A_2A, were compared (Table 3). The results obtained prompted us to test the ability of phenyl and boron cluster-modified adenosine derivative pairs to influence platelet activation in a more complex model of human blood.

Table 3.

IC₅₀ values of radioligand displacement calculated for A_2A receptor.

Compound	IC50 [μM]	Compound	IC50 [μM]
7	>100	23	>100
8	74.1 ± 9.7	24	22.0 ± 4.4
15	>100	35	n.d
16	26.5 ± 5.1	37	>100
17	>100	39	>100
18	14.6 ± 0.6	40	20.3 ± 0.5

IC₅₀ – Compound concentration that inhibits radioligand binding to the receptor by 50%. n.d. – not determined.

2.6. Inhibition of platelet aggregation in human blood

The ability of the synthesized compounds to decrease platelet activation was tested in a stepwise manner. As a first approach, measurements in platelet-rich plasma were conducted. In this method, platelet aggregation was induced by ADP, and the formation of platelet aggregates in the presence of the tested compounds was monitored. Red blood cells (RBCs), which are known to accumulate and metabolize adenosine, are not present in this experimental setup.

The tendency of the tested compounds to be taken up by RBCs was not known; therefore, we decided to utilize simplified conditions in the preliminary tests. If a particular compound was not effective in these RBC-free conditions (platelet-rich plasma), the compound was unlikely to possess inhibitory activity in whole blood, where RBCs are present. These ineffective compounds were not tested in whole blood. As shown in Table 4, compounds 18 and 40 that are modified with a phenyl group significantly decreased platelet activation induced by ADP by approximately 50%, while other compounds did not possess an inhibitory effect or inhibited activation by less than 50%. Compound 24 inhibited platelet aggregation by 20%.

Table 4.

ADP-induced platelet aggregation in platelet-rich plasma in the presence of the compounds 7, 8, 15–18, 23, 24, 35, 37, 39 and 40.

Compound	0	10 μM		50 μM
7	656.2 ± 55.8	686.5 ± 66.1	n.s.	642.8 ± 45.4	n.s.
8	654.9 ± 123.1	695.3 ± 133.4	n.s.	623.4 ± 165.6	n.s.
15	654.8 ± 57.5	698.7 ± 74.4	n.s.	657.1 ± 70.7	n.s.
16	648.5 ± 45.0	662.3 ± 67.0	n.s.	641.1 ± 35.1	n.s.
17	677.7 ± 65.5	646.2 ± 58.6	P < 0.001	639.5 ± 64.7	n.s.
18	708.9 ± 64.5	630.0 ± 80.3	P < 0.01	336.8 ± 108.5	P < 0.001
23	657.1 ± 57.8	715.3 ± 85.8	n.s.	657.0 ± 67.2	n.s.
24	678.9 ± 70.9	649.5 ± 70.2	P < 0.001	548.6 ± 72.2	P < 0.01
35	655.6 ± 38.8	665.7 ± 86.8	n.s.	682.3 ± 17.4	n.s.
37	651.4 ± 59.9	661.1 ± 61.3	n.s.	616.0 ± 24.7	P< 0.05
39	663.0 ± 54.6	678.4 ± 55.7	n.s.	658.0 ± 46.8	n.s.
40	641.1 ± 85.5	566.4 ± 112.9	P < 0.05	289.5 ± 158.9	P < 0.001

Data are presented as the means ± SD, n = 6–9 samples. The results are shown as the area under an aggregation curve measured in arbitrary units and was calculated as the extent of aggregation (a.u.) × time [min]. The significance of the inhibitory effect was tested in a pairwise manner (control vs. compound) using the bootstrap-boosted (10000 iterations) paired one-tailed Student’s t-test. Inhibitory effects greater than 50% are marked in bold. n.s. = nonsignificant.

Additional analyzes of the receptor-ligand binding curves, as well as the percentage changes in platelet activity revealed a clear relationship between the ability of the active compounds to interact with the A_2A receptor and inhibit platelet aggregation.

Analysis of the percentage values of both receptor binding and inhibition of platelet aggregation showed that the only relevant relationship between these two compound features is valid for the A_2A receptor and not for the other receptors (r_{s_A2A} = 0.746, P < 0.01, r_{s_A2B} = 0.378, n.s., r_{s_A1} and r_{s_A3} < 0.1). Of the four compounds interacting with the A_2A receptor, compound 16 did not affect platelet aggregation. Characteristically, the active compounds are N⁶-(3-phenylpropyl) adenosine derivatives 18, 24 and 40 which simultaneously preferentially bind to A₃ AR. In summary, regardless of the receptor preference (mainly in relation to A₃), compounds interacting with the A_2A receptor, modified with a phenyl group, but not a boron cluster, showed antiplatelet activity.

Compounds 18 and 40, which possessed the highest inhibitory activity, were further tested in a whole blood aggregometry assay. In this experiment, platelet aggregation in whole blood was induced with ADP, collagen or arachidonic acid. As shown in Fig. 2, compound 40 significantly decreased ADP-stimulated platelet aggregation at all tested concentrations but did not inhibit aggregation induced by collagen or arachidonic acid (data not shown). In contrast, compound 18 significantly inhibited aggregation induced by either ADP (5 μM) or collagen (1 μg/mL), but it had no effect on platelet aggregation induced by arachidonic acid (data not shown).

Fig. 2.

Effects of compounds 18 and 40 on platelet aggregation in whole blood. The results are presented as the area under an aggregation curve. a) Effect of compound 40 on platelet aggregation induced by ADP. b) Effect of compound 18 on platelet aggregation induced by ADP. c) Effect of compound 18 on platelet aggregation induced by collagen. The significance of differences was determined using a paired Student’s t-test: P < 0.005 for 10 μM compound 40, P < 0.0001 for 30 μM compound 40 and for 50 μM compound 40 ZL-1098; P < 0.0001 for all the tested concentrations of compound 18 (both for ADP- and collagen-induced aggregation), n = 13.

As a final step, the ability of compounds 18 and 40 to inhibit thrombus formation under flow conditions in the presence of collagen was evaluated. Whole blood was perfused over a collagen-coated surface under flow conditions that emulate the conditions in a human artery. The contact of platelets with collagen in combination with shear forces induces the formation of thrombi. As shown in Fig. 3, the thrombi in samples treated with compound 40 were smaller than the thrombi in the untreated samples. Surprisingly, compound 18 did not exert an inhibitory effect under these conditions (data not shown).

Fig. 3.

Effects of compound 40 on the formation of thrombi deposited on collagen under flow conditions. A) Representative images of fluorescently labelled thrombi formed in blood treated with a vehicle (left panel) or 50 μM compound 40 (right panel). B) Comparison of the sizes of thrombi. The sizes of thrombi visualized in the channels were quantified using dedicated software. Each dot represents the summarized area of thrombi larger than 50 μM² formed in the blood of a particular donor treated either with vehicle or compound 40. Significant differences were determined using a paired Student’s t-test, P < 0.01, n = 9.

2.7. Structure-activity relationships (SARs)

The strength of interactions of compounds and adenosine receptors varied depending on the receptor type, and for A₃ AR, the strength of interactions was significantly higher than the strength of interactions for other receptors (ANOVA, p = 0.0001; logK_iA3 vs. logK_iA1, p = 0.0026; logK_iA3 vs. logK_iA2A, p = 0.0003, Tukey’s test; Kruskal–Wallis test, p = 0.0001; K_iA3 vs. K_iA1, p = 0.043; K_iA3 vs. K_iA2A, p = 0.0001). The analysis of categorical data (active versus inactive compounds) showed that the number of active compounds was clearly higher for the A₃ receptor (n = 17) than for the A_2A (n = 6) and A₁ receptors (p = 0.0018, chi-square statistic). Therefore, the receptors were analyzed separately for the SAR model. Among substructure descriptors, the best predictors for A₃ receptor affinity selected using the GRM method (general regression model, GRM, Wald chi-square statistics) were as follows: modifier (carborane or phenyl, p = 0.0008), position of modification (N⁶, C2, or C8; p = 0.028), and sugar type (R and not R; p = 0.018); for A₁ receptor: modifier (carborane or phenyl, p = 0.007), and sugar type (R and not R; p = 0.03); results for A_2A were not statistically significant due to insufficient number of active compounds per group (6 active versus 22 inactive). Structure-activity analysis was performed only for statistically significant classes/groups of descriptors. The regression approach of classification was implemented from the PLS and C&RT modules in Statistica software. The data were divided into a training set (n = 18 compounds) and a test set (n = 10 compounds, 7, 8, 23, 24, 36, 39, 43, 44, 47, 48). The C&RT tree method was used for the prediction of compound classification (Supplementary Material, Figure S129 A and B). The analysis enabled formulation of some general rules. The most important predictor of compound affinity to the A₃ receptor was modification position (relative importance 100%) over modifier type (85%) and sugar type (53%). The analysis showed that active compounds with log K_i ≤ 1 are adenosine modified at the N⁶ position with a phenyl group (n = 3 compounds in the training set); moderately active compounds with log K_i 2–3 are adenosine modified at the C2 and N⁶ positions with boron clusters or at the C2 and C8 positions with phenyl groups; and low active or inactive (log K_i ≥ 4) compounds are C8 adenosine modified with carborane and 2′-deoxyadenosine derivatives. Compounds with arabinoadenosine and 2′-deoxyadenosine scaffolds are less active or inactive. In general, the most active compounds among phenyl derivatives are adenosine modified at the N⁶ position, whereas among carborane derivatives, adenosine is modified at the C2 and N⁶ positions. The boosted tree method gave similar results (importance of modification site 100%, modifier 75%, and sugar type 67%; details of the calculations are shown in Supplementary Material, Figure S129C). Evaluation of the model on numerous newly synthesized compounds in the future may verify the described principles as general for adenosine phenyl and carboranyl derivatives as adenosine receptor ligands, with higher or lower affinity to the receptor.

Additional detailed analysis showed the importance of ligand rigidity in the group of phenyl-modified compounds at both the C2 and C8 positions, which was a statistically significant feature of the compound interaction with the A₃ AR (active compounds 34, 42, 44, and 48 with rigid linkers versus inactive 36, 46, 50, and 52; p = 0.002; GRM). A similar relationship was observed for the corresponding carboranyl analogs at C2 positions, among which the active derivatives contained a rigid linker (compounds 41 and 47 versus 45 and 49).

2.8. Receptor selectivity analysis

Detailed in-depth analysis of the structural similarities shows some specific relationships in receptor selectivity for individual compounds. In the group of N⁶ adenosine derivatives, receptor selectivity was analyzed for individual cases in relation to the linker length and the presence of specific additional groups such as N-ethyl carboxamide (in the C5′ position) and chlorine (Cl in the C2 position of purine ring). Among adenosine derivatives bearing carborane groups at the N⁶ position, linker length decided on receptor selectivity (C₄ – active, A₃ AR selective compounds 7 and 15, C₃ – inactive compound 17 or compound 39 slightly active on A₁ AR), regardless of the presence of the 5′ N-ethyl carboxamide substituent. However, chlorine substitution in the N⁶ carboranyl derivative with a C₃ linker resulted in activity against A₃ AR, but was poor (compound 23). In the group of phenyl N⁶ adenosine derivatives, no compound was found to be selective against one type of receptor; in the group of C₃ linkers, the lowest selectivity was found for unsubstituted phenyl derivatives (40, all receptor types, including A_2A AR). The presence of chlorine at C2 or N-ethyl carboxamide at C5′ made this type of compound more selective for A₃ AR (compounds 18 and 24). Linker extension (C₄ derivative, 8) resulted in a relatively better interaction with A₃ AR than with the other receptor types; however, substitution with N-ethyl carboxamide at C5’ resulted in a loss of selectivity of compound (16). In summary: 1) C₃ linker and 2) long C₄ linker in combination with N-ethyl carboxamide substituent determined lack of receptor selectivity of N⁶-phenyl adenosine derivatives. However, 3) a long C₄ linker and 4) a C₃ linker in combination with N-ethyl carboxamide or chlorine substituents increased the selectivity for A₃ AR.

The group of compounds with phenyl modification at the C8 position with the flexible linker were inactive (compounds 36 and 52), regardless of sugar structure. The rigidity of the linker determined the interaction with adenosine receptors in this group of compounds; however, receptor selectivity depended on the sugar moiety. Ribose (R)-containing compound 34 was A₃ selective, but compound 44 with deoxyribose (dR) was not selective. However, the activity of a group of C8-modified compounds depended on the presence of a phenyl substituent since all analogous derivatives containing the carboranyl modifier were inactive (33, 35, 43, and 51). Interesting structure-selectivity dependencies were observed for C2 derivatives. Whereas the structure of the sugar moiety decided on the receptor selectivity, all these compounds were active, regardless of the linker or modifier structure. R-containing compounds were not selective (41, 42, 45, and 46), and arabinose (Ara)-containing compounds were A₃ selective (compounds 47 and 48).

2.9. Docking results

A homology model of the A₃ receptor was generated from the A_2A receptor (PDB code: 4UG2), as described in the Materials and Methods, and was subsequently used for docking studies. A docking study showed that carborane and phenyl derivatives 15 and 16 are located at analogous positions inside the A₃ AR, i.e., ribose downward and modifier towards the “entrance” of the receptor interior, identified as a classic A₃ AR ligand-binding pocket (Fig. 4 A). This docking position was energetically preferred for compounds 15 and 16 (binding free energy, ΔG, −9.6 ± 0.076 kcal/mol and −10.4 ± 0.11 kcal/mol, respectively, see Table S1 in Supplementary Information). The binding free energy showed a higher affinity of compound 16 to the A₃ AR than the affinity of compound 15, although the calculated K_{i_calc} values for both compounds were in the nanomolar range (26 ± 4.3 nM and 93 ± 8.4 nM, respectively; Fig. 4 A and B; Table S1 in Supplementary Information). The docking results revealed a fourfold lower affinity of compound 15 than the affinity of 16 towards the A₃ receptor. Although qualitatively similar, quantitatively, these results differ from those in vitro in which compound 16 revealed a 47-fold higher affinity towards the A₃ AR than compound 15 (see Table 1).

Fig. 4.

Binding modes of compounds 15 and 16 at A₃ AR binding site. (A and B) Overall look on topography of 15 and 16 preferred docking position: “ribose down” (A), and alternative poses: “ribose up” (B); affinity for A₃ AR was expressed as K_{i calc} values (nM). (C and D) Details of compounds interaction with A₃ AR binding site for preferred docking positions of compounds 15 (C) and 16 (D). Structures were visualized and analyzed using the PyMOL Molecular Graphics System, Version 2.3.4, Schrödinger, LLC (Figures A and B), and The Protein-Ligand Interaction Profiler (PLIP, Figures C and D).

Docking experiments also revealed another interesting property of compounds 15 and 16. The position of ribose directed towards the binding pocket was apparently a preferred docking position for compound 15. A reverse docking position of this compound, i.e., carborane downward, was also observed (regarding the number of docking events), albeit this reverse docking position requires much higher energetic effort (ΔG −9.4 kcal/mol, K_{i_calc} 129 nM; Table S1 in Supplementary Information). Therefore, the carborane moiety seems to stabilize the ribose directed towards the binding pocket. However, unlike derivative 15 bearing a boron cluster, for compound 16, the position of the phenyl residue pointing towards the binding pocket seems possible; the ΔG of such binding was −9.9 kcal/mol (K_{i calc} 53 nM; Table S1 in Supplementary Information). Both docking positions of compounds 15 and 16, ribose directed towards the binding pocket (“ribose down”) and ribose located towards the “entrance” of the receptor (“ribose up”), are stabilized by Phe168 interacting with the purine ring through π-stacking (Table S1 in Supplementary Information). Nevertheless, the position “ribose down” seems preferable, considering the energy cost and the number of interactions with neighboring amino acids (Table S1 in Supplementary Information and Fig. 4C and D). Positions “ribose down” of both compounds were stabilized by hydrogen bonds between ribose and Ser271 and His272 and between the N-ethyl carboxamide group (in the C5 ′position) and Thr94 (Figures C and D). This position is also stabilized by hydrophobic interactions of the carboxamide group and Ile98 or Leu91; in the case of compound 16, an additional hydrophobic interaction of the carboxamide group with Trp185 is formed (Fig. 4 D). Asn250 forms two hydrogen bonds with purine nitrogen atoms (6 N and 7 N), and Phe168 forms π-stacking with purine rings in compounds 15 and 16 (Figures C and D).

Apparently, the nucleoside scaffold of both compounds interacts with similar amino acids of the binding pocket without any steric hindrance resulting from carborane or phenyl groups.

The most interesting parts of the conjugates are carboranyl or phenyl modifiers together with long, fourth-carbon linkers. The flexible linker participates in hydrophobic interactions of compounds 15 and 16 with two (Val169 and Leu264) or three (Val169, Leu264, and Ile268) amino acids, respectively. Significantly, Phe168 forms a parallel interaction with two subfragments of compound 15: π-stacking with purine ring and hydrophobic with a boron vertex of the cluster (Fig. 4C). By these interactions, Phe168 could stabilize the position of compound 15 “in the electrostatic holder” in the binding pocket, including Val169, Leu264, Tyr265 and Ile268 (Fig. 4C). The results obtained partially explain the possible interactions of carboranyl- and phenyl-modified adenosine derivatives with the A₃ receptor at the molecular level.

3. Discussion

Two major types of chemical transformations were used to obtain adenosine derivative pairs containing inorganic boron cluster modifications or alternative organic phenyl modifications:1) Sonogashira-type cross coupling between 8-halogen-adenosine and a boron cluster or phenyl group donor containing a terminal triple bond (compounds 33 and 34), and 2) the substitution of chlorine at carbon 6 of adenosine with an alkylamine containing a boron cluster or a phenyl group (compounds 7, 8, 15–18, 23, 24, 27, and 28).

Catalytic hydrogenation of the triple bond in the rigid alkyne linker (compounds 33 and 34) provided adenosine derivatives 35 and 36 (Table 1) bearing a flexible ethyl tether between the nucleobase and modification. Compound 38 modified at the 2’ position was obtained analogously to the method described for compound 37 using 3-phenyl-1-propanol instead of 3-(1,12-dicarba-closo-dodecaboran-1-yl)-1-propanol via the formation of a formacetal linkage [24].

A modular approach based on the coupling of a modification acceptor (suitably functionalized adenosine component) and modification donor (boron cluster or phenyl group) was applied to synthesize all compounds described in the present study. Commercially available phenyl derivatives, 4-phenylbutylamine (9), 3-phenylpropylamine (14), 3-phenyl-1-propanol and phenyl acetylene (32), were used.

The boron cluster-containing counterparts 1-(4-aminobutyl)-1,12-dicarba-closo-dodecaborane (6), 1-(3-aminopropyl)-1,12-dicarba-closo-dodecaborane (4), 1-(3-hydroxypropyl)-1,12-dicarba-closo-dodecaborane and 2-ethynyl-1,12-dicarba-closo-dodecaborane (31) were synthesized using different chemistries. Methods for preparing azido- and aminoalkylcarboranes and derivatives of isomeric 1,2- and 1,7-dicarba-closo-dodecaborane, as well as their applications in the synthesis of various carborane derivatives have been reported [17–19,21]. Additionally, we previously described the synthesis of azido- and amino-derivatives of 1,12-dicarba-closo-dodecaboranes 2 and 4 [13], although to our knowledge, detailed procedures for their synthesis have not yet been described. Here, we fill the gap (Scheme 1). Three different methods were employed and compared to obtain amines 4 or 6: 1) catalytic hydrogenation of azides, 2) deprotection of Boc-protected amines obtained from suitable bromoalkyl carborane and bis-Bocamine (Scheme 1), or 3) deprotection of phthalimide-protected amines obtained via the alkylation of carborane with N-bromoalkyl phthalimide. Methods 1 and 2 are best suited for the synthesis of 1-(3-aminopropyl)carboranes due to the accessibility of 1-(3-bromopropyl)carboranes prepared conveniently from suitable 1-(3-hydroxypropyl)carboranes available via oxetane ring opening with lithiated carboranes [10]. The phthalimide methodology seems more versatile due to the possible use of alkyl tethers with varying lengths in the N-bromoalkyl phthalimide component. 2-Ethynyl-1,12-dicarba-closo-dodecaborane (31) was synthesized by performing a Pd-catalyzed Kumada-type cross coupling reaction of the corresponding iodinated carborane with Me₃SiCCMgBr, followed by desilylation of the trimethylsilylalkynyl-substituted cluster [37].

The adenosine derivative pairs obtained modified with inorganic (boron cluster) or organic (phenyl) groups at the same location within the nucleoside unit were screened for affinity towards A₁, A_2A, A_2B and A₃ adenosine receptors using a radioligand replacement assay. Pronounced, albeit varying, effects of the modifications on the ligand affinity and selectivity towards the studied receptors were observed. In general, most of the adenosine modifications studied here, regardless of whether a boron cluster or phenyl group was used as the modifying unit, displayed higher affinity for the A₃ AR than for A₁, A_2A or A_2B, which was also confirmed by statistical calculations, including both the classification approach and the factor analysis. An exception is compound 40 with a 3-phenylpropyl modification at N⁶ or, to a lesser extent, compound 39 with a (1,12-dicarba-closo-dodecaboran-1-yl)propyl at the same position, which showed significant interaction with all AR types (A₃, A₁, and A_2A, 40) or tendency to bind to A₁ AR (39).

Interestingly, the transformation of the 5′-OH group in these derivatives into a 5′-N-ethylcarboxamide group (compounds 17 and 18) or an extension of the linker between N⁶ and phenyl or (1,12-dicarba-closo-dodecaboran-1-yl)propyl-group (compounds 7 and 8) switches this preference from A₁/A_2A to A₃ AR (Table 1). Notably, the ability of compounds 18 and 40 to bind to A₁/A_2A AR, even with moderate affinity and preference for A₃, appears to be important for their activity as inhibitors of platelet aggregation in blood (vide infra).

The analogs with the highest selectivity towards A₃ AR include compounds: 16 (A₃ K_i = 9.2 nM): A₁/A₃ = 87, A_2A/A₃ = 772, A_2B/A₃ = 359), 18 (A₃ K_i = 4.5 nM): A₁/A₃ = 49% (at 10 μM) vs. K_i = 4.5 nM, A_2A/A3 = 42% (at 10 μM) vs. K_i 4.5 nM, A_2B/A₃ = 307; 24 (A₃ K_i = 6.4 nM): A₁/A₃ = 519, A_2A/A₃ = 27% (at 10 μM) vs. K_i = 6.4 nM, A_2B/A₃ = 24% (at 10 μM) vs. K_i = 6.4 nM, 40 (A₃ K_i = 7.5 nM): A₁/A₃ = 54% (at 10 μM) vs. K_i = 7.5 nM, A₁/A_2A = 32% (at 10 μM) vs. K_i = 7.5 nM, A₁/A_2B = 264, 41 (A₃ K_i = 206 nM): A₁/A₃ = 14, A_2A/A₃ = 15, 42 (A₃ K_i = 42.3 nM): A₁/A₃ = 15, A_2A/A₃ = 25. Based on the aforementioned observations supported by the=statistical analysis of SAR, the preferential binding to the A₃ AR of the ligands evaluated in the present study seems to be determined by the presence of the modification at N⁶ (or 2-C in the case of compounds 41 and 42), which is consistent with previous reports [5].

Importantly, a rigid ethynyl linker between the modification and nucleobase at position C2 increases the ability of the ligand to bind A₃ tenfold compared with a more flexible ethyl linker for both the boron cluster and phenyl modification: compound 41 A₃ K_i = 206 nM (boron cluster modification attached through an ethynyl linker) vs. compound 45 A₃ K_i = 2780 nM (boron cluster modification attached through an ethyl linker) and compound 42 A₃ K_i = 42.3 nM (phenyl modification attached through an ethynyl linker) vs. compound 46 A₃ K_i = 8610 nM (phenyl modification attached through an ethyl linker). Interestingly, the aforementioned selectivity of the studied adenosine derivatives towards A₃ AR is also valid for carborane modifications, which are better accommodated by A₃ than by A₁, A_2A or A_2B ARs.

Furthermore, for reasons that are currently unclear, regardless of the location of the modification and receptor type, in each pair of counterparts bearing a phenyl group or boron cluster modification the ligand bearing a phenyl group usually showed higher affinity for the specific receptor. According to preliminary findings based on in silico calculations, the observed difference may be due to the 50% higher van der Waals volume of the carborane cage than the rotating phenyl group and different electronic properties of both modifying units [25]. The replacement of the phenyl group with the boron cluster seems to weaken the ligand interactions with the receptor binding pocket.

Analyses of the affinity, selectivity and biological activity (intracellular cAMP production) revealed that the phenyl modifications at the exo-amino group of adenosine (compounds 8, 16 and 24) efficiently interacted with the A₃ receptor. Moreover, the affinity and biological activity of the phenylic compounds were consistent. A comparison of the biological activity of phenylic compounds with their respective boronic analogs (IC₅₀ boronic vs IC₅₀ phenylic) indicated the significance of the phenyl substituent at a particular site of the compound to obtain the best affinity for the A₃ receptor. The replacement of the phenyl group with a boronic moiety resulted in the most dramatic decrease in the biological activity of the receptor treated with compounds 23 and 24. Because a phenylic substitution of 2-chloroadenosine (compound 24) results in an approximately 300-fold increase in affinity for the A₃ receptor compared with 2-chloroadenosine (K_i of 6.4 nM (this work) and 1900 nM [40]), the phenylic base functional group appears to be desirable.

However, higher affinity may not only have a beneficial effect on the compound activity but also, at the expense of selectivity, a negative effect. This problem can be approached by designing selective agonists that, although they activate a given receptor at a relatively high concentration of compound, still activate the remaining receptors to a small extent. One example may be the activity of compounds 15 and 16 towards the A₃ receptor, one boron cluster modified and the other a phenyl group. Analysis of receptor selectivity showed that compound 16, with a higher affinity for the A₃ receptor (K_i = 9.2 nM), binds 100% of the A₁, A_2A and A_2B receptors at micromolar concentrations.

Compound 16 displays relatively low selectivity over A₁ AR (87-fold, A₁ AR K_i value is 803 nM), 771-fold selectivity over A_2A AR (K_i value is 7098 nM), and 362-fold selectivity over A_2B (AR K_i value is 3331 nM). Compound 15 modified with a boron cluster with affinity K_i = 437 nM to A₃ AR activates 100% of the A₃ receptor at micromolar concentrations, but in contrast to 16, Compound 15 does not interact noticeably with A₁, A_2A, or A_2B receptors, showing clearly significant A₃ AR selectivity. An illustrative comparison of the binding of 15 and 16 to the receptors at nanomolar and micromolar concentrations is shown in Fig. 5. Docking results revealed a good dynamic fit of carborane derivative 15 to the interior of the A₃ AR binding pocket without steric hindrance. Experiments in silico have shown that the carborane moiety linked to a nucleoside scaffold via a four-carbon flexible linker significantly interacts with the surrounding hydrophobic amino acids, engaging them to stabilize the compound at the active site. The length and flexibility of the linker may be crucial for the functionality of the carborane as a modifier of A₃ AR ligands based on adenosine scaffolds. An example of the influence of linker structure on A₃ AR affinity can be compound 17, which differs from compound 15 in the length of the linker “only” by one methylene group. This small change, however, turns out to be sufficient for 17 to lose its affinity to this receptor (Table 1). Whether the linker length and its structure could be a paradigm of the effectiveness of adenosine ligands modified with boron clusters requires further research in terms of both chemical synthesis and biological and in silico studies (see Fig. 6).

Fig. 5.

Dose-response curves of binding compounds 15 and 16 to adenosine receptors A₁ (), A_2A (),A_2B (), A₃ (), evaluated in radioligand inhibition assay (details in Experimental Section). Values are mean ± SEM.

Fig. 6.

Biological activity of N⁶ derivatives of adenosine, modified with boron cluster (7, 15, and 23) or phenyl group (8, 16, and 24); median values of K_i are shown. Both groups of derivatives show activity towards A₃ AR. Phenyl compounds interact significantly more strongly with A₃ AR, but also activate A_2A AR, interfering with the activity of blood platelets.

Although ARs are promising therapeutic targets under a wide range of conditions, one difficulty when developing active analogs is that each adenosine receptor affects not just one but a large number of body functions. The development of sufficiently selective AR modulators with the required pharmacological characteristics is an attractive but challenging goal.

The use of a boron cluster modifier instead of ordinary organic modifications such as phenyl groups provides a new perspective and can help to overcome these difficulties, as shown for A₃ AR in the above example. The higher selectivity of some adenosine ligands modified with boron clusters than phenyl-modified derivatives towards the A₃ receptor can be of special interest in light of the A₃ AR-mediated protective response in heart diseases and skeletal muscle injury and other conditions related to the functional role of this receptor [41]. The modifications proposed in this work offer a wide range of specific structural features related to receptor specificity. Among the compounds with lower specificity, there may be those interacting with the A_2A receptor (compounds 18, 24, or 40), an important therapeutic target in the hemostatic system. Even if they interact with the A_2A AR in micromolar concentrations, such a low affinity turns out to be sufficient to influence the activity of components of the hemostatic system.

Antiplatelet therapies, which are administered to patients at risk of myocardial infarction or ischemic stroke, are based on the inhibition of platelet aggregation in the blood. These interventions are always associated with a risk of bleeding. At the same time, some patients respond poorly to antiplatelet therapy. An increase in the dose of the drug in these patients is not the preferred method to bypass this ‘resistance’ due to the risk of bleeding. In turn, combined therapies are applied in which at least two independent pathways regulating platelet activation are targeted simultaneously to limit platelet activation. One potential target is the activation of adenosine receptors on platelets. This approach has not yet been tested in the clinic but has been extensively studied in vitro and in vivo [42–44]. Agonists of adenosine receptors have recently been shown to enhance the inhibitory activity of P2Y₁₂ antagonists – antiplatelet drugs used widely as antiplatelet therapy [45]. Blood platelets express A_2A and A_2B subtypes of adenosine receptors [46,47]. The activation of these receptors increases the intraplatelet concentration of cAMP, which decreases the ability of platelets to aggregate [48]. One of the disadvantages of using AR agonists as antiplatelet agents is their relatively short-term effect following in vivo application [49]. This limitation may be partially attributed to the effective uptake of the drug by red blood cells, which has been shown for adenosine [50]. Thus, novel agonists of ARs with more persistent effects in vivo should be developed.

Two of the newly synthesized compounds described in the present study exerted a significant inhibitory effect on platelet aggregation in whole blood stimulated with ADP or collagen. In the thrombus formation test, only one of the compounds proved to be effective. The latter test emulates pathophysiological phenomena occurring in a diseased artery. The formation of a thrombus under flow conditions involves complex interactions between factors that activate platelets, such as shear forces, thromboxane, ADP or thrombin [51]. Therefore, the thrombus formation test differs from a classical aggregation test in which only one platelet activator is used as a trigger. Thus, the confirmation of the antiplatelet activity of a compound in a classical aggregation test does not unambiguously determine its effectiveness in the thrombus formation test. This finding may also explain why compound 18 effectively inhibited collagen-induced aggregation but did not inhibit thrombus formation on collagen. In contrast, compound 40, which was not effective as an inhibitor of collagen-induced aggregation in the classical test, was a potent inhibitor in the thrombus formation test. Notably, in the thrombus formation test, collagen activated platelets, but thrombus expansion was further driven by the other factors mentioned above. Thus, we postulated that compound 40 is more effective than compound 18 at inhibiting one of these pathways of platelet activation that play a major role in thrombus growth. Since both compounds inhibited ADP-induced aggregation to a similar extent and failed to inhibit arachidonic acid-induced aggregation, another mechanism is likely involved in thrombus formation that is affected by compound 40 to a greater extent than compound 18. Each of the stimuli inducing platelet aggregation activates separate receptors, but the downstream pathways intercept at some points. These points include, among many others, a decrease in intracellular cAMP and an increase in intracellular calcium. Adenosine receptor agonists reverse these effects with varying effectiveness [52]; therefore, their ability to inhibit platelet aggregation induced by these stimuli may also differ between compounds.

The antiaggregation activity of compounds 18 and 40 and their parallel A₃ AR affinity should not be surprising in light of their clear although moderate affinity towards A_2A AR (Tables 1 and 3, Fig. 1). These compounds significantly interact with the A_2A receptor at micromolar concentrations (IC₅₀ for radioligand displacement 14.6 ± 0.6 μM and 20.3 ± 0.5 μM, respectively) (Table 3). However, nanomolar affinity for the A₃ receptor is irrelevant for platelet activation given the absence of this receptor on the platelet surface. Furthermore, aggregation studies were conducted at 10 μM and 50 μM ligand concentrations, which seems to be sufficient for A_2A AR activation, leading to inhibition of aggregation. Compound 40 was an effective inhibitor of thrombus formation under conditions that approximate the rupture of an atherosclerotic plaque. Thus, this compound might represent an interesting candidate for the development of a potential antiplatelet compound. Moreover, compound 40 may be considered a dual-acting A_2A and A₃ AR ligand interacting with a given target depending on the concentration.

Fig. 1.

The effect of compounds 7, 8, 15–18, 23, 24, 35, 37, 39 and 40. (50 μM) on ADP-induced platelet aggregation in platelet-rich plasma in comparison with the effect on the radioligand binding to A_2A receptor. The values are expressed as percentages of maximal control values of radioligand binding and platelet aggregation (AUC parameter). Data points marked with a red line indicate compounds 18, 24 and 40 having both properties: interaction with A_2A receptor (at least 50% of ligand displacement) and effect on the platelet aggregation (at least 20% inhibition). Data points marked with a black line indicate inactive compounds. White triangle - compound 16, that interact with the receptor, but is inactive on blood platelets. Spearman’s correlation coefficient calculated for all data points is r_s = 0.746, P (2-tailed) = 0.0085.

However, developing potent ARs targeting drugs is a difficult and challenging task. One of the factors that must be considered, which is often overlooked, is cross-binding for the new adenosine derivatives with other adenosine-like recognition sites and non-adenosine receptor drug interactions. Discussing details of this issue goes beyond the scope of this publication. Here, we would like to draw attention to two phenomena related only to the first part of that concern. Both compounds 18 and 40, due to the activation of A_2A AR, show desired antiplatelet activity while simultaneously being A₃ AR ligands (Table 1). These properties may at first glance be of concern due to the side effects of A₃ AR activation, but in this case, a side effect may be cardioprotection [41], which for antithrombotics could be an added value.

Another issue is the possible phosphorylation of adenosine ligands possessing free 5′-hydroxy groups. Most adenosine-based A₃ ligands are modified at the 5′-position and are not vulnerable to phosphorylation. However, other derivatives with a free 5′-hydroxyl group, as we showed earlier, can be phosphorylated quite efficiently by adenosine kinase (ADK) [28] and probably also by other kinases. One of the possible paths of metabolic transformation of these derivatives resulting from ADK action, as in the case of cladribine, is conversion in subsequent phosphorylation steps into the corresponding triphosphate. The triphosphates formed interact with DNA/RNA polymerases by competition with natural nucleoside triphosphates, disturbing DNA replication, which in turn may cause cytotoxicity and other adverse effects. Awareness of possible, although not necessary, already recognized activities of potential AR-targeting therapeutics may help minimize side effects, sometimes achieve synergism and optimize therapeutic activities.

4. Conclusions

We described a series of adenosine and 2′-deoxyadenosine pairs modified with a 1,12-dicarba-closo-dodecaborane cluster or alternatively with a phenyl group at the same position and studied their affinity for the A₁, A_2A, A_2B and A₃ adenosine receptors. Regardless of whether a boron cluster or phenyl group was used as the modifying unit and the location of the modification within the nucleobase (C2, C8 or N⁶), most of the studied adenosine derivatives tended to exhibit higher affinity for A₃ AR than for A₁, A_2A or A_2B ARs. This finding expands previous observations of the preferential binding of adenosine ligands modified at N⁶ or C8 to A₃ AR [24,51].

In our study, the highest selectivity for A₃ was observed for adenosine derivatives modified at N⁶, followed by derivatives modified at C2. For adenosine ligands modified at N⁶, the replacement of a hydroxymethyl group at C2’ with an ethylcarbamoyl group or the incorporation of chlorine at C2 increased the selective affinity for A₃ AR. In addition, phenyl modification increased the general affinity of adenosine derivatives for ARs to a greater extent than their boron cluster counterparts. At the same time, the observed higher selectivity of some AR ligands modified with boron clusters compared to the corresponding phenyl derivatives towards the A₃ receptor may open the way for the development of molecules activating A₃ without causing unwanted antiplatelet effects.

All selected adenosine ligands with the highest affinity for A₃ displayed agonist characteristics in the functional study, regardless of whether they were modified with a phenyl group or boron cluster. Finally, the tests of inhibition of blood platelet aggregation allowed us to identify two ligands, 18 and 40, with clear but moderate antiaggregatory properties, most likely due to A_2A AR activation. Both compounds are modified with phenyl groups at the N⁶ position.

Despite the low level of A₃ AR in most cells and tissues, A₃ AR expression is upregulated in blood cells in subjects with some pathological conditions. Indeed, preclinical studies have reported anti-inflammatory, anticancer and cytoprotective effects of A₃ receptor agonists [41,53]. In addition, the cardioprotective effect triggered by A₃ AR activation can be an added value for the identified phenyl-modified compounds with A_2A AR-related antiplatelet activities. The results described here expand the range of uses of adenosine derivatives for further testing in this promising area of the therapeutic application of selective AR ligands.

5. Experimental section

5.1. Synthesis of adenosine derivatives bearing a boron cluster or phenyl group modification

5.1.1. Materials

Commercially available chemicals were of reagent grade and used as received. N-(4-Bromobutyl)phthalimide and 3-phenylpropylamine (14) were purchased from Acros Organics (Geel, Belgium). 4-Phenylbutylamine (9) and phenylacetylene (32) and 10% Pd on activated charcoal were purchased from Sigma-Aldrich (Darmstadt, Germany). 6-Chloroadenosine (5), 8-bromoadenosine (30), 2,6-dichloropurine, 1-O-acetyl-2,3,5-tri-O-benzoyl-β-d-ribofuranose and 2-deoxy-3,5-di-O-(4-toluoyl)-α-d-erythropentofuranosyl chloride (Hoffer’s chlorosugar) were purchased from Carbosynth (Compton - Berkshire, United Kingdom). Silica gel column chromatography was performed on silica gel 60 (230–400 mesh), Sigma-Aldrich (Steinheim, Germany) or E. Merck (Darmstadt, Germany). R_f values refer to analytical TLC performed using pre-coated silica gel 60 F254 plates purchased from Sigma-Aldrich (Steinheim, Germany) or E. Merck (Darmstadt, Germany) and developed in the solvent system indicated. Compounds were visualized with UV light (254 nm). Melting points are uncorrected and were measured on a Büchi (New Castle, DE, USA) Melting Point B-540 apparatus. The yields are not optimized.

Multiplate® test cuvettes for whole blood aggregometry were obtained from Roche Diagnostics GmbH (Mannheim, Germany). Multiplate test cells for turbidimetric aggregometry were purchased from Chronolog Co (Havertown, PA). ADP was purchased from Sigma-Aldrich (Steinheim, Germany). Collagen was from Chronolog Co. (Havertown, PA), arachidonic acid was from Dynabyte (Munich, Germany), Biochips for the assays of thrombi formation were purchased from Cellix (Dublin, Ireland), collagen for chip coating was from Chronolog Co. (Havertown, PA). Anti-CD61 PE-conjugated antibodies for the staining of thrombi were from Becton Dickinson (San Diego, CA).

Melting points are uncorrected and were measured on a Kofler apparatus. The purity of all tested compounds was established by HPLC and was 95–100% with the exception of compound 33, 34, 35 and 51 whose purity was higher than 90%. (for HPLC details please see the experimental section, and for HPLC traces of newly synthesized compounds 7, 8, 15–18, 23, 24, 27, 28, 33–35 see Figures S1–S13 in Supplementary Information).

5.1.2. Methods

Nuclear Magnetic Resonance (NMR).

The ¹H, ¹³C, dept135 and ¹¹B NMR spectra in CDCl₃ or CD₃OD were recorded using a Bruker AVANCE III HD (Billerica, MA, USA) 500 MHz or a BrukerAvance DPX 250 MHz spectrometer; shift values in parts per million are relative to the SiMe₄ internal reference. Multiplets were assigned as s (singlet), bs (broad singlet), d (doublet), t (triplet), m (multiplet), bm (broad multiplet), dd (doublet of doublets), dt (doublet of triplets), td (triplet of doublets), tt (triplet of triplets), ddd (doublet of doublet of doublet). Coupling constants are reported in Hertz (Hz).

Mass spectrometry.

Fast atom bombardment (FAB) mass spectra were recorded on a Finnigan MAT (Bremen, Germany). The m/z was measured in positive and negative modes. The MALDI-TOF spectra were recorded on a Voyager Elite mass spectrometer (PerSeptive Biosystems Inc., Framingham, MA) in linear negative ion mode. The ESI spectra were recorded on a Waters HPLC/MS (Milford, MA, USA). High-resolution mass spectra were recorded on an LTQ Orbitrap Velos instrument (Thermo Scientific, Waltham, MA, USA). The calculated m/z corresponds to calculated values based on the average mass of the elements consisting of natural isotopes.

Ultraviolet spectroscopy measurements (UV).

UV measurements were performed on a GBC Cintra10 UV-VIS spectrometer (Dandenong, Australia). Samples for UV experiments, ca. 0.5 A₂₆₀ OD of each compound were dissolved in 96% C₂H₅OH. The measurement was performed at ambient temperature.

Attenuated total reflectance Fourier transform infrared (ATR–FTIR) spectroscopy measurements.

Infrared absorption spectra were recorded using a Smart iTR diamond attenuated total reflectance (ATR) attachment on a Nicolet 6700 FTIR spectrometer (Thermo Scientific) equipped with an ETC EverGlo* source for the IR range, a Ge-on-KBr beam splitter, and a DLaTGS detector. Samples to be analyzed were placed on a diamond ATR element in the solid form. The Omnic 8.1 software program was used for data acquisition and processing. Alternatively, a Jasco 6200 (Easton, MD, USA) FTIR spectrometer was used, and spectra were run in a KBr pellet.

HPLC analysis.

Analyses were performed on a Hewlett-Packard 1050 system equipped with a UV detector. A Hypersil Gold C18 RP, 250 × 4.6 mm column, 5 μM particle size (Thermo Scientific, Runcorn, UK) was used, and the flow rate was 1 mL/min. All analyses were run at ambient temperature. The gradient elution profile was as follows: 20 min from 0% B to 100% B, 20 min 100% B, and 5 min from 100% B to 0% B. Buffer A contained 0.1 M TEAB (pH 7.0) in a mixture of acetonitrile and water (2:98, v/v), and buffer B contained 0.1 M TEAB (pH 7.0) in a mixture of acetonitrile and water (60:40, v/v). UV detection was performed at λ = 262 nm.

Synthesis of 1-(3-bromopropyl)-1,12-dicarba-closo-dodecaborane (1) was performed as described [18].

Synthesis of 1-(4-aminobutyl)-1,12-dicarba-closo-dodecaborane (6) was performed as described [22].

Synthesis of 2′,3′-O,O-isopropylidene-6-chloro-5′-oxoadenosine (11) was performed according to the literature procedure [32].

2′,3′,5′-tris-O-Benzoyl-2,6-dichloroadenosine (20) [54] 3′,5′-bis-O-(4-methylbenzoyl)-2,6-dichloro-2′-deoxyadenosine (19) [55] were synthesized as described.

2′-O-(1,12-dicarba-closo-dodecaboran-1-yl)propyleneoxymethyl]-adenosine (37) [24] and 2′-O-(3-phenylpropyleneoxymethyl]-adenosine (38) [25] were synthesized as described.

N⁶-[3-(1,12-Dicarba-closo-dodecaboran-1-yl)propyl]adenosine (39) [13] was obtained as described previously.

N⁶-(3-phenylpropyl)adenosine (40) [26] and N⁶-(4-phenylbutyl) adenosine (8) [26] were obtained according to the literature methods.

2-Ethynyl-(1,12-dicarba-closo-dodecaboran-1-yl)adenosine (41) was synthesized as described [13].

2-Phenylethynyladenosine (42) was synthesized as described [27].

Synthesis of 8-[2-(1,12-dicarba-closo-dodecaboran-1-yl)ethynyl]-2′-deoxyadenosine (43) was performed as previously described [12].

Synthesis of 2-ethynyl-(1,12-dicarba-closo-dodecaboran-2-yl) arabinoadenosine (47) [16] and 2-ethyl-(1,12-dicarba-closo-dodecaboran-2-yl)arabinoadenosine (49) [16] was performed as described earlier.

Synthesis of 1-(3-azidopropyl)-1,12-dicarba-closo-dodecaborane (2).

1-(3-Bromopropyl)-1,12-dicarba-closo-dodecaborane (1) (250 mg, 0.94 mmol) was dissolved in anhydrous DMF (3 mL) under argon atmosphere then sodium azide (71.7 mg, 1.1 mmol) was added. The reaction mixture was stirred at room temperature until TLC monitoring (CH₂Cl₂/hexane, 1:1) showed complete conversion of the starting material (ca. 2 h). Then the solvent was evaporated to dryness under vacuum yielding crude product 2 which was purified by silica gel column chromatography using a mixture of CH₂Cl₂/hexane (1:1) as an eluent. Yield: 207 mg, 93%, colourless oil. TLC (CH₂Cl₂/hexane, 1:1): R_f = 0.69; FT-IR (ATR) ν_max, cm⁻¹: 3065, 2960, 2933, 2865 (C–H), 2600 (B–H), 2090 (N₃); ¹H NMR {¹¹B BB} (CDCl₃, 600.25 MHz, 25 °C, TMS) δ, ppm: 1.41–1.47 (2H, m, CH₂CH₂CH₂N₃),1.67–1.71 (2H, m, CH₂CH₂CH₂N₃), 1.67–2.56 (10H, m, BH), 2.65 (1H, m, CH^carborane), 3.13 (2H, t, J = 6.6 Hz, CH₂CH₂CH₂N₃); ¹³C NMR (CDCl₃, 250.13 MHz, 25 °C, TMS) δ, ppm: 28.73, 36.03, 50.72, 58.45,83.63; ¹¹B NMR {¹H BB} (CDCl₃, 80.25 MHz, 25 °C, BF₃/(C₂H₅)₂O) δ, ppm: 12.56 (5B, s), −14.96 (5B, s); ¹¹B NMR (CDCl₃, 80.25 MHz, 25 °C, BF₃/(C₂H₅)₂O) δ, ppm: 14.52 (d, 5B, ²J_BH = 164.8 Hz), −12.15 (d, 5B, ²J_BH = 161.7 Hz); MS (APCI, +Ve): m/z (%) calculated for C₅B₁₀H₁₇N₃: 227.32, found 200.25 [M − N₂]⁺.

Synthesis of N,N-di-tert-butyloxycarbonyl-l-(3-aminopropyl)-1,12-dicarba-closo-dodecaborane (3).

Tetrabutylammonium hydrogensulfate (494.6 mg) was dissolved in 2 M NaOH aqueous solution(1.5 mL), and then a solution of di-tert-butyl iminodicarboxylate (289.7 mg) in CH₂Cl₂ (7.2 mL) was added. To the resultant mixture a solution of 1-(3-bromopropyl)-1,12-dicarba-closo-dodecaborane (1, 340.0 mg, 1.28 mmol) in CH₂Cl₂ (5 mL) was added dropwise. The reaction mixture was stirred under reflux for 3.5 h, then the reaction was quenched with water (16.5 mL). The organic layer was separated, and the aqueous layer was extracted with CH₂Cl₂ (2 × 7 mL). The combined organic layers were dried over anhydrous MgSO₄ and the organic solvent was evaporated to dryness. The resulting crude product was purified by silica gel column chromatography (230–400 mesh) using hexane/CH₂Cl₂ (1:1) as an eluting solvent system. Yield: 463.2 mg (90%); TLC (CH₂Cl₂): R_f = 0.37; FT-IR (ATR): v, cm⁻¹ = 3054 (CH^carborane), 2975 ($\text{C}{\text{H}}_3^{\text{asym}.}$), 2923 ($\text{C}{\text{H}}_2^{\text{asym}.}$), 2872 ($\text{C}{\text{H}}_3^{\text{sym}.}$), 2852 ($\text{C}{\text{H}}_2^{\text{sym}.}$), 2595 (BH), 1768 (C=O), 1466 (CH₂), 1363 (CH₃), 1099 (CO); ¹H NMR (CDCl₃, 500.13 MHz, 25 °C) δ, ppm: 1.39–1.44 (2H, m, CH₂), 1.48 (18H, s, 6 × CH₃), 1.59 (2H, dd, J = 11.9, 5.5 Hz, CH₂), 1.64–2.57 (10H, m, BH), 2.62 (1H, s, CH^carborane), 3.36 (2H, t. J = 7.1 Hz, CH₂); ¹³C NMR (CDCl₃, 125.76 MHz, 25 °C) δ, ppm: 28.17 (CH₃), 28.88 (CH₃), 29.83 (CH₃), 36.21 (CH₂),45.56 (C–N(Boc)₂), 58.20 (C^carborane), 82.50 (O–C(CH₃)₃), 152.49 (C=O); ¹¹B NMR {¹H BB} (CDCl₃, 160.46 MHz, 25 °C) δ, ppm: 15.04 (5B, s), −12.56 (5B, s); ¹¹B NMR (CDCl₃, 160.46 MHz, 25 °C) δ, ppm: 15.54 (5B, d, J = 165.3 Hz), −13.06 (5B, d, J = 162.1 Hz); MS (ESI): m/z (%) calculated for C₁₅H₃₅B₁₀O₄N: 401.3569, found: 424.3475 [M + Na]⁺.

Synthesis of 1-(3-aminopropyl)-1,12-dicarba-closo-dodecaborane (4).

Method A. 1-(3-Azidopropyl)-1,12-dicarba-closo-dodecaborane (2, 150 mg, 0.66 mmol) was suspended in dry methanol (2 mL) under argon atmosphere, next 10% Pd/C (86 mg) was added. Reaction mixture was stirred at room temperature under H₂ atmosphere (balloon reservoir). After 2 h the solvent was evaporated and the residue was resuspended in diethyl ether. Suspension was filtered through a pad of Celite and the filtrate was evaporated to dryness. Oily residue was purified by silica gel column chromatography using a mixture of hexane/Et₂O/Et₃N (1:1:0.02) as an eluent. Yield: 81 mg, 61%, white solid. Method B. To N,N-tert-butyloxycarbonyl-1-(3-aminopropyl)-1,12-dicarba-closo-dodecaborane (3, 475.7 mg, 1.18 mmol) a solution of hydrogen chloride in methanol (9.48 mL) was added. The reaction mixture was stirred for 18 h at room temperature, then solvent was evaporated to dryness under reduced pressure. The resulting crude product was purified by silica gel column chromatography (230–400 mesh) using a linear gradient of CH₃OH in CH₂Cl₂ (5%–14%) as an eluting solvent system. Yield: 214 mg (76%) of free amine containing a trace of hydrochloride salt. TLC (CH₂Cl₂/CH₃OH 95:5): R_f = 0.38 (free amine), (CH₂Cl₂/CH₃OH, 9:1): R_f = 0.09 (hydrochloride); FT-IR (ATR) ν, cm⁻¹: 3250–2800 (${\text{RNH}}_3^{+}$), 3061 (CH^carborane), 2921 ($\text{C}{\text{H}}_2^{\text{asym}.}$), 2851 ($\text{C}{\text{H}}_2^{\text{sym}.}$), 2601 (BH), 1603 (NH); ¹H NMR {¹¹B BB} (CDCl₃, 250.13 MHz, 25 °C, TMS) δ, ppm: 1.26–1.32 (2H, m, CH₂CH₂CH₂NH₂), 1.60–1.63 (2H, m, CH₂CH₂CH₂NH₂), 1.56–2.53 (10H, m, BH), 2.33 (2H, t, J = 7.1 Hz, CH₂CH₂CH₂NH₂), 2.62 (1H, m, CH^carborane); ¹³C NMR (CDCl₃, 250.13 MHz, 25 °C, TMS) δ, ppm:29.84 (CH₃), 36.75 (CH₂-carborane), 49.03 (CH₂–N), 58.23 (C^carborane), 84.50 (C(CH₃)₃); ¹¹B NMR {¹H BB} (CDCl₃, 80.25 MHz, 25 °C, BF₃/(C₂H₅)₂O) δ, ppm: 12.53 (s, 5B), −15.04 (s, 5B), ¹¹B NMR (CDCl₃, 80.25 MHz, 25 °C, BF₃/(C₂H₅)₂O) δ, ppm: 12.53 (d, 5B, ²J_BH = 162.7 Hz), −15.04 (d, 5B, ²J_BH = 165.8 Hz); MS (ESI): m/z calculated for C₅B₁₀H₁₉N: 201.252, found: 202.259 (100) [M+H].

Synthesis of N⁶-[(1,12-dicarba-closo-dodecaboran-1-yl)but-4-yl]adenosine (7).

The synthesis was performed according to a modified literature procedure³⁰.6-Chloroadenosine (5, 50 mg, 0.15 mmol) and 1-(4-aminobutyl)-1,12-dicarba-closo-dodecaborane (6) (39 mg, 0.18 mmol) was dissolved in n-BuOH (0.88 mL) then Et₃N (0.07 mL, 0.52 mmol) was added. The resultant solution was refluxed for 2 h then evaporated to dryness. The crude product was purified by silica gel column chromatography using linear gradient of CH₃OH in CH₂Cl₂ (0–6%) as an eluent. Yield: 56 mg (80%), yellowish solid. TLC (CH₂Cl₂/CH₃OH, 9:1): R_f = 0.40; UV–Vis (95% C₂H₅OH) λ, nm: 228 (min), 267 (max); FT-IR (ATR) ν, cm⁻¹: 3232 (O–H), 2926 (C–H), 2861 (C–H), 2601 (B–H), 1620 (C=N), 1077 (C–O); RP-HPLC, t_R = 23.34 min; ¹H NMR (CD₃OD, 600.13 MHz, 25 °C) δ, ppm:1.25–1.31 (4H, m, butyl CH₂), 1.48–1.53 (2H, m, butyl CH₂),1.70–1.73 (2H, m, CH₂–CH₂–NH), 1.85–2.53 (10H, m, BH), 3.10 (1H, bs, CH^carborane) 3.49 (1H, bs, NH), 3.74 (1H, dd, J = 2.4, 10.2 Hz, 5′_bH), 3.88 (1H, dd, J = 2.4, 10.2 Hz, 5′_aH), 4.17 (1H, d, J = 2.4 Hz, 3′H), 4.32 (1H, d, J = 2.4 Hz, 4′H), 4.73–4,75 (1H, m, 2′H), 5.94 (1H, d, J = 6.6 Hz, 1′H), 8.21 (1H, s, 2-H), 8.24 (1H, s, 8-H); ¹³C NMR (CD₃OD, 150.90 MHz, 25 °C) δ, ppm: 26.38 (CH₂), 28.60 (CH₂), 29.36 (CH₂),38.27 (CH₂), 58.41 (CH^carborane), 62.31 (C5′), 71.32 (C-3′), 74.04 (C-2′), 86.83 (C-4′), 89.91 (C-1′),140.07 (C-8),152.11 (C-2), 154.94 (C-6), ¹¹B NMR {¹H BB} (CD₃OD, 192.55 MHz, 25 °C) δ ppm: 12.65 (s, 5B), −15.12 (s, 5B); MS (ESI, +3 kV, 300 °C): m/z calculated for C₁₆H₃₁B₁₀N₅O₄: 465.56, found: 466.35 [M+H]⁺.

Synthesis of 2′,3′-O,O-isopropylidene-6-chloroadenosine (10).

The synthesis was performed according to the modified literature procedure [30]. 6-Chloroadenosine (5, 1000 mg, 3.49 mmol) was dissolved in dry DMF (13 mL) then p-toluenesulfonic acid monohydrate (130 mg, 0.70 mmol) and 2,2-dimethoxypropane (0.93 mL,10.47 mmol) was added. The resultant solution was stirred overnight at ambient temperature and then evaporated to dryness. The crude product was purified by column chromatography using linear gradient of CH₃OH in CH₂Cl₂ (0–6%). Yield: 1110 mg (97%), white solid. TLC (CH₂Cl₂/CH₃OH, 9:1): R_f = 0.63; UV–Vis (95% C₂H₅OH) λ, nm: 228 (min), 264 (max); FT-IR (ATR) ν, cm⁻¹: 3289 (O–H), 2922 (C–H), 2854 (C–H), 1663 (C=N), 1081 (C–O); ¹H NMR (CDCl₃, 500.13 MHz, 25 °C) δ, ppm: 1.38 (3H, s, CH₃), 1.65 (3H, s, CH₃), 3.82 (1H, d, J = 11 Hz, 5′_aH), 3.98 (1H, d, J = 11 Hz, 5′_bH), 4.56 (1H, s, 4′H), 5.11 (1H, s, 3′H), 5.20 (1H, s, 2′H), 6.00 (1H, s, 1′H), 8.30 (1H, s, 8-H),8.77 (1H, s, 2-H); ¹³C NMR (CDCl₃, 125.76 MHz, 25 °C) δ, ppm: 25.20 (CH₃), 25.36 (CH₃), 63.11 (CH₂) 81.54, 83.21 (C-2′, C-3′), 86.36 (C-4′),94.14 (C-1′), 114.28 (CH₃CCH₃), 133.00 (C-4), 145.04 (C-8), 150.40 (C-2), 151.91 (C-6); MS (ESI, −3 kV, 300 °C): m/z calculated for C₁₂H₁₄ClN₅O₄: 326.74, found: 361.1 [M+Cl]⁻.

Synthesis of 2′,3′-O,O-isopropylidene-5′-ethylamino-5′-oxo-6-chloroadenosine (12).

The synthesis was performed according to a modified literature procedure [33]. 2′,3′-O,O-Isopropylidene-5′-oxo-6-chloroadenosine (11, 475 mg, 1.39 mmol) was dissolved in dry CHCl₃ (7 mL) then dry DMF (0.27 mL, 3.49 mmol) and SOCl₂ (0.51 mL, 6.97 mmol) was added. The resulting solution was stirred under reflux for 4 h then evaporated to dryness. The brown residue was dissolved in dry THF (9.3 mL), cooled to 5 °C then 2 M solution in THF ethylamine was added (1.75 mL, 3.49 mmol). The solution was stirred for 20 min and evaporated to dryness. The crude product was purified by silica gel column chromatography using linear gradient of CH₃OH in CH₂Cl₂ (0–10%) as an eluent. Yield: 293 mg (57%), brown oil. TLC (CH₂Cl₂/CH₃OH, 9:1): R_f = 0.70; UV–Vis (95% C₂H₅OH) λ, nm: 229 (min), 265 (max); FT-IR (ATR) ν, cm⁻¹: 3309 (N–H), 3073 (C–H), 2983 (C–H), 2936 (C–H), 1662 (C=O), 1592 (C=N), 1200 (C–O), 1093 (C–O); ¹H NMR (CD₃OD, 500.13 MHz, 25 °C) δ, ppm: 0.53 (3H, t, J = 7 Hz, CH₃), 1.41 (3H, s, CH₃), 1.59 (3H, s, CH₃), 2,76 (2H, q, J = 7.0 Hz, N–CH₂–C), 4.69 (1H, d, J = 1.5 Hz, 4′H), 5.59 (1H, d, J = 1.5 Hz, 3′H), 5.69 (1H, dd, J = 4.5, 1.5 Hz, 2′H), 6.50 (1H, s, 1′H), 7.65=(1H, bs, NH), 8.69–8.70 (2H,=m, 8-H, 2-H); ¹³C NMR (CD₃OD, 125.76 MHz, 25 °C) δ, ppm: 14.50 (CH₃CH₂N), 25.76, 27.50 (isopropylidene CH₃), 35.13 (CH₃CH₂N) 85.75 (C-3′), 89.90 (C-4′), 93.36 (C-1′), 115.39 (CH₃CCH₃), 133.26 (C-4), 149.06 (C-8), 152.06 (C-2), 153.46 (C-6), 171.85 (C=O); MS (ESI, −3 kV, 300 °C): m/z calculated for C₁₂H₁₄ClN₅O₄: 367.79, found: 402.2 [M+Cl]⁻.

Synthesis of 5′-ethylamino-5′-oxo-6-chloroadenosine adenosine (13).

The synthesis was performed according to the modified literature procedure [33]. 2′,3′-O,O-Isopropylidene-5′-ethylamino-5′-oxo-6-chloroadenosine (12, 150 mg, 0.41 mmol) was dissolved in a mixture of 2 M HCl solution in water and 1,4-dioxane (1:1, 3.5 mL), the resultant solution was stirred for 1 h at 70 °C then was evaporated to dryness. The crude product was purified by reverse phase flash chromatography using a gradient of CH₃OH in H₂O (5–50%) as an eluent. Yield: 54 mg, (41%), yellowish solid. TLC (CH₂Cl₂/CH₃OH, 9:1): R_f = 0.27; UV–Vis (95% C₂H₅OH) λ, nm: 231 (min), 265 (max); FT-IR (ATR) ν, cm⁻¹: 3251 (O–H), 3079 (N–H), 2987 (C–H), 2941 (C–H), 2683 (C–H), 1645 (C=O), 1059 (C–O); ¹H NMR (acetone-d⁶, 600.26 MHz, 25 °C) δ, ppm: 1.16 (3H, t, J = 6 Hz, CH₃), 3.31–3.35 (2H, m, N–CH₂–C), 4.51–4.54 (2H, m, 2′OH,=3′OH),4.82 (1H, d, J = 6 Hz, 4′H), 4.97–5.0 (2H, m, 2′H, 3′H), 6.26 (1H, d, J = 6 Hz, 1′H), 7.99 (1H, bs, NH), 8.81 (1H, s, 8-H), 8.91 (1H, s, 2-H); ¹³C NMR (acetone-d⁶, 150.94 MHz, 25 °C) δ, ppm: 14.22 (CH₃), 33.62 (CH₂–N), 73.12 (C-2′), 73.64 (C-3′), 84.98 (C-4′), 89.26 (C-1′), 132.59 (C-4), 146.23 (C-8), 150.44 (C-2), 151.57 (C-6), 168.82 (C=O); MS (FAB, Gly, +Ve): m/z calculated for C₁₂H₁₄ClN₅O₄: 327.72, found: 328.2 [M+H]⁺.

Synthesis of N⁶-[(1,12-dicarba-closo-dodecaboran-1-yl)but-4-yl]-5′-ethylcarbamoyl-adenosine (15).

The synthesis was performed analogously as described for 6-chloroadenosine adenosine.³⁰ 5′-Ethylamino-5′-oxo-6-chloroadenosine adenosine (13, 50 mg,0.15 mmol) and 1-(4-aminobutyl)-1,12-dicarba-closo-dodecaborane hydrochloride (6) (36 mg, 0.17 mmol) was dissolved in n-BuOH (0.76 mL) then Et₃N (0.07 mL, 0.65 mmol) was added. The resulting solution was refluxed for 2 h then evaporated to dryness. The crude product was purified by silica gel column chromatography using linear gradient of CH₃OH in CH₂Cl₂ (0–6%) as an eluent. Yield: 39 mg, 51%, orange solid. TLC (CH₂Cl₂/CH₃OH, 9:1): R_f = 0.43; UV–Vis (95% C₂H₅OH) λ, nm: 230 (min), 267 (max); FT-IR (ATR) ν, cm⁻¹: 3364 (N–H), 3231 (O–H), 3063 (C-H^carborane), 2928 (C–H), 2855 (C–H), 2602 (B–H), 1647 (C=O), 1614 (C=N), 1049 (C–O); RP-HPLC, t_R = 24.93 min; ¹H NMR (CD₃OD, 600.13 MHz, 25 °C) δ, ppm: 1.21 (3H, t, J = 7.2 Hz, CH₃), 1.24–1.34 (4H, m, CH₂-(CH₂–CH₂)–CH₂), 1.49–1,54 (2H, m, CH₂–CH₂-carborane), 1.74 (2H, m, CH₂–CH₂–NH), 1.98–2.52 (10H, m, BH), 3.10 (1H, s, CH^carborane),3.37 (2H, q, J = 7.2 Hz, N–CH₂–CH₃), 3.50 (1H, bs, NH-Bu), 4.31 (1H, d, J = 4.8 Hz, 3′H), 4.46 (1H, s, 4′H), 4.75–4,77 (1H, m, 2′H), 6.00 (1H, d, J = 7.2 Hz, 1′H), 8.24 (1H, s, 2-H), 8.27 (1H, s, 8-H); ¹³C NMR (CD₃OD, 150.90 MHz, 25 °C) δ, ppm: 13.68 (CH₃), 26.37, 28.62, 29.37,33.67 (CH₂), 38.26 (CH₂–N), 58.41 (C-5′), 71.95 (C-3′), 73.59 (C-2′),85.03 (C-4′), 89.11 (C-1′), 140.68 (C-8), 152.39 (C-2), 154.95 (C-6), 170.66 (C=O); ¹¹B NMR {¹H BB} (CD₃OD, 192.55 MHz, 25 °C) δ, ppm: 12.67 (s, 5B), −15.13 (s, 5B); MS (ESI, +3 kV, 300 °C): m/z calculated for C₁₈H₃₃B₁₀N₆O₄: 506.61, found: 507.37 [M+H]⁺.

Synthesis of 5′-ethylcarbamoyl-N⁶-(4-phenylbutyl)adenosine (16).

The synthesis was performed analogously as described for 6-chloroadenosine [29]. 5′-Ethylamino-5′-oxo-6-chloroadenosine adenosine (13, 96 mg, 0.29 mmol) and 4-phenylbutylamine (9) (0.056 mL, 0.35 mmol) was dissolved in n-BuOH (1.5 mL) then Et₃N (0.15 mL, 1.06 mmol) was added. The resulting solution was refluxed for 2 h then was evaporated to dryness. The crude product was purified by flash chromatography using a gradient of CH₃OH in CH₂Cl₂ (0–10%). Yield: 48 mg, 37%, white solid. TLC (CH₂Cl₂/MeOH, 9:1): R_f = 0.38; UV–Vis (95% C₂H₅OH) λ, nm: 232 (min), 268 (max); FT- IR (ATR) ν, cm⁻¹: 3233 (O–H), 3083 (N–H), 3027 (C–H), 2975 (C–H), 2932 (C–H), 2858 (C–H), 1646 (C=O), 1615 (C=N), 1049 (C–O), 745 (Phe), 698 (Phe); RP-HPLC, t_R = 18.94 min; ¹H NMR (CD₃OD, 500.13 MHz, 25 °C) δ, ppm: 1.20 (3H, t, J = 7.5 Hz, CH₃),1.70–1.75 (4H, m, (CH₂)₂), 2.63–2.68 (2H, m, CH₂-Phe), 3.30–3.38 (4H, m, 2xCH₂NH, overlapped by methanol signal), 3.60 (1H, bs, NH-Bu-Phe), 4.31–4.47 (2H, m, 2′H, 3′H), 4.75 (1H, m, 4′H), 6.01 (1H, d, J = 7.5 Hz, 1′H), 7.11–7.24 (5H, m, Phe-H), 8.23 (1H, s, 2-H), 8.27 (1H, s, 8-H); ¹³C NMR (CD₃OD, 125.76 MHz, 25 °C) δ, ppm:13.56 (CH₃), 30.43, 37,95, 41,84 (3x CH₂), 35.56, 37.05 (2x CH₂N),73.85 (C-3′), 75.49 (C-2′), 86.93 (C-4′), 91.00 (C-1′), 122.07, 127.25, 129.79, 129.93, 142.53, 144.06 (5x Phe), 154.28 (C-8), 156.84 (C-2), 172.56 (C=O); MS (ESI, −3 kV, 300 °C): m/z calculated for C₂₂H₂₈N₆O₄: 440.50, found: 475.3 [M+Cl]⁻.

Synthesis of N⁶-[(1,12-dicarba-closo-dodecaboran-1-yl)prop-3-yl]-5′-ethylcarbamoyl adenosine (17).

The synthesis was performed according to the modified literature procedure [28]. 5′-Ethylamino-5′-oxo-6-chloroadenosine adenosine (13, 59 mg, 0.18 mmol) and 1-(3-aminopropyl)-1,12-dicarba-closo-dodecaborane (4, 43 mg, 0.18 mmol) was dissolved in n-BuOH (0.90 mL) then Et₃N (0.09 mL,0.65 mmol) was added. The resulting solution was refluxed for 2 h, then evaporated to dryness. The crude product was purified by silica gel column chromatography using linear gradient of CH₃OH in CH₂Cl₂ (0–7%) as an eluent. Yield: 52 mg, (59%) white solid. TLC (CH₂Cl₂/CH₃OH, 85:15): R_f = 0.58; UV–Vis (95% C₂H₅OH) λ, nm: 228 (min), 270 (max); FT-IR (ATR) ν, cm⁻¹: 3247 (N–H), 3069 (CH^carborane), 2954 (C–H), 2928 (C–H), 2857 (C–H), 2603 (B–H), 1729 (C=O), 1615 (C=N), 1125 (C–O), 1049 (C–O); RP-HPLC, t_R = 23.94 min; ¹H NMR (CD₃OD, 600.13 MHz, 25 °C) δ, ppm: 1.20 (3H, t, J = 6 Hz, CH₃, overlapped by t-BuOH signal), 1.29 (2H, m, CH₂–CH₂–CH₂), 1.57 (2H, m, CH₂-carborane), 1.78 (2H, m, CH₂–CH₂–NH), 1.81–2.63 (10H, m, BH), 3.13 (1H, s, CH^carborane),3.37 (2H, q, J = 6 Hz, N–CH₂–CH₃), 3.42 (1H, bs, NH-Pr), 4.31 (1H, m, 3′H), 4.46 (1H, m, 4′H), 4.75 (1H, m, 2′H), 6.00 (1H, d, J = 6 Hz, 1′H),8.24 (1H, s, 2-H), 8.27 (1H, s, 8-H); ¹³C NMR (CD₃OD, 150.90 MHz, 25 °C) δ, ppm: 15.51 (CH₃), 30.72, 31.58 (2xCH₂), 35.53, 37.81 (2xCH₂-N), 60.12, 60.55 (2xCH^carborane), 73.46 (C-3′), 75.44 (C-2′), 86.88 (C-4′), 90.96 (C-1′), 142.61 (C-8), 153.91 (C-2), 171.96 (C=O); ¹¹B NMR {¹H BB} CD₃OD, 160.46 MHz, 25 °C) δ, ppm: 12.55 (s, 5B), −14.99 (s, 5B); MS (ESI, −3 kV, 300 °C): m/z calculated for C₁₇H₃₂B₁₀N₆O_4: 492.58, found: 493 [M+H]⁺.

Synthesis of 5′-ethylcarbamoyl-N⁶-(3-phenylpropyl)adenosine (18).

The substitution of the chlorine atom was performed analogously as described for the synthesis of N⁶-(3-(phenylpropyl) adenosine [29]. 5′-Ethylamino-5′-oxo-6-chloroadenosine adenosine (13, 26 mg, 0.08 mmol) was dissolved in n-BuOH (0.4 mL) under argon atmosphere then 3-phenylpropylamine (14) (0.014 mL, 0.10 mmol) and Et₃N (0.04 mL, 0.29 mmol) was added. The resulting solution was refluxed for 1.5 h, then evaporated to dryness. The crude product was purified by reverse phase flash chromatography using gradient of CH₃OH in H₂O (5–70%) as an eluent. Yield: 31 mg, 73%, white solid. TLC (CH₂Cl₂/CH₃OH, 9:1): R_f = 0.48; UV–Vis (95% C₂H₅OH) λ, nm: 231 (min), 267 (max); FT-IR (ATR) ν, cm⁻¹: 3228 (N–H), 3085 (O–H), 3029 (C–H), 2929 (C–H), 2857 (C–H), 1729 (C=O), 1614 (C=N), 1049 (C–O), 744 (Phe), 698 (Phe); RP-HPLC, t_R = 19.36 min; ¹H NMR (CD₃OD, 600.26 MHz, 25 °C) δ, ppm: 1.23 (3H, t, J = 6 Hz, CH₃), 1.31–1.37 (2H, m, CH₂-Phe), 2.04 (2H, kw, J = 6 Hz, C–CH₂–C), 2.77 (2H, t, J = 6 Hz, N–CH₂–C₂H₅-Phe), 3.38 (2H, q, J = 6 Hz, N–CH₂–CH₃), 3.64 (1H, bs, NH-Pr), 4.34–4.35 (1H, m, 4′H), 4.49 (1H, d, J = 6 Hz, 3′H), 4.77–4.81 (1H, m, 2′H, overlapped by water signal), 6.03 (1H, d, J = 6 Hz, 1′H), 7.15–7.18 (1H, m, Phe-H), 7.23–7.28 (4H, m, Phe-H), 8.26 (1H, s, 2-H), 8.30 (1H, s, 8-H); ¹³C NMR (CD₃OD, 150.94 MHz, 25 °C) δ, ppm: 13.62 (CH₃), 32.74, 33.67 (2xCH₂), 72.03 (C-3′), 73.59 (C-2′),85.04 (C-4′), 89.16 (C-1′), 125.48 (Phe), 128.01 (Phe), 140.59 (C-8), 141.59 (Phe), 152.39 (C-2), 155.06 (C-6), 170.64 (C=O); MS (FAB, Gly, +VE): m/z calculated for C₂₁H₂₆N₆O₄: 426.47, found: 427.3 [M+H]⁺.

Synthesis of 2′,3′,5′-tris-O-benzoyl-6-N-[3-(1,12-dicarba-closododecaboran-1-yl)propyl]-2-chloro-2′-adenosine (21) and 2′,3′,5′⁻tris-O-benzoyl-6-N-(3-phenylpropyl)-2-chloro-2′-adenosine (22).

To the 1 eq. of 2′,3′,5′-tris-O-benzoyl-2,6-dichloroadenosine (20) dissolved in dry DMF (10 mL/0.5 mmol of nucleoside), 3 eq. of 3-phenyl-1-propylamine (14) hydrochloride or 1.2 eq. of 3-(1,12-dikarba-closo-dodecaboran-1-yl)propylamine (4) hydrochloride followed by 3 eq. of triethylamine were added and reaction mixture was stirred at room temperature for 20 h. Obtained solution was diluted with ethyl acetate and washed twice with water and once with brine. The organic phase was dried over magnesium sulfate, filtrated and evaporated with silica gel. Compounds 21 and 22 were purified by silica gel column chromatography using a gradient of AcOEt in hexane (20–30%) as an eluent.

Yield 21: 154 mg, 78%, starting from 158 mg of dichloropurine nucleoside; TLC (CH₂Cl₂/CH₃OH 9:1): R_f = 0.88; UV–Vis (95% C₂H₅OH) λ, nm: 227, 250 (min), 230, 271 (max); FT-IR (KBr) ν, cm⁻¹: 3365, 3063, 2927, 2861, 2605, 1968, 1912, 1728, 1620, 1582, 1532, 1477, 1451, 1350, 1314, 1266, 1177, 1120, 1092, 1068, 1025; M.P., °C:94.6–97.1; ¹H NMR (CDCl₃, 500 MHz, 25 °C) δ, ppm: 1.49–1.56 (2H, m, CH₂), 1.60- ~2.80 (10H, bm, BH) – very broad multiplet which superposes with three other peaks within the range (CH-carborane, 2 × CH₂), 1.75–1.69 (2H, m, CH₂), 2.63 (1H, bs, CH-carborane), 3.43 (2H, bs, CH₂(NH)), 4.72 (1H, dd, J = 4.0, 12.0 Hz, 5′_bH), 4.78–4.83 (1H, m, 4′H), 4.89 (1H, dd, J = 3.0, 12.0 Hz, 5′_aH), 5.73 (1H, bs, NH), 6.10–6.17 (2H, m, 2′H + 3′H), 6.43 (1H, d, J = 5.5 Hz, 1′H), 7.35–7.49 (6H, m, Phe-H), 7.52–7.62 (3H, m, Phe-H), 7.86 (1H, s, 8-H),7.93–7.97 (2H, m, Phe-H), 7.99–8.02 (2H, m, Phe-H), 8.07–8.11 (2H, m, Phe-H); ¹³C NMR (CDCl₃, 125 MHz, 25 °C): 29.4 (CH₂), 36.1 (CH₂),40.1 (CH₂), 58.4 (CH-carborane), 63.9 (C-5′), 71.7 (C-2′), 74.5 (C-3′),81.2 (C-4′), 83.8 (C^IV-carborane), 86.3 (C-1′), 119.2 (C^IV–Cl), 128.5 (C^IV), 128.7 (2 × Phe), 128.8 (Phe), 129.4 (C^IV), 129.8 (Phe), 130.0 (Phe), 130.1 (Phe), 133.6 (Phe), 133.9 (Phe), 134.0 (Phe), 138.2 (C₈), 150.0 (C^IV), 155.2 (C^IV), 155.4 (C^IV), 165.3 (C^IV), 165.5 (C^IV), 166.3 (C^IV) (1 (C^IV) carbon is not observed, most probably it is shielded by the intense Phe signal at 128.8 ppm); ¹¹B NMR {¹H BB} (CDCl₃, 160 MHz, 25 °C): 12.95 (s, 5B), −15.40 (s, 5B); HRMS (ESI, + 3 kV, 300 °C): m/z calcd for C₃₆H₄₁ClN₅B₁₀O₇: 798.36922, found: 798.36991 [M+H]⁺.

Yield 22: 529 mg, 92%, starting from 500 mg of dichloropurine nucleoside; TLC (CH₂Cl₂/CH₃OH 9:1): R_f = 0.84; UV–Vis (95% C₂H₅OH) λ, nm: 248 (min), 270 (max); FT-IR (KBr) ν, cm⁻¹: 3368, 3061, 3028, 2935, 2858, 1968, 1914, 1727, 1620, 1581, 1533, 1477, 1451,1350, 1314, 1267,1177,1092,1067,1025; M.P.: 63.2–66.3 °C; ¹H NMR (CDCl₃, 500 MHz, 25 °C): 1.95–2.04 (2H, m, CH₂), 2.72 (2H, t, J = 7.0 Hz, CH₂(Ph)), 3.64 (2H, bs, CH₂(NH)), 4.72 (1H, dd, J = 4.0, 12.0 Hz, 5′_bH), 4.81 (1H, dt, J = 3.0, 4.0 Hz, 4′H), 4.89 (1H, dd, J = 3.0, 12.0 Hz, 5′_aH), 5.96 (1H, bs, NH), 6.12–6.19 (2H, m, 2′H + 3′H),=6.44 (1H, d, J = 5.0 Hz, 1′H), 7.14–7.22 (3H, m, H_Ar), 7.23–7.30 (2H, m, H_Ar), 7.30–7.50 (6H, m, H_Ar), 7.50–7.65 (3H, m, H_Ar), 7.87 (1H, s, 8-H), 7.92–7.98 (2H, m, H_Ar), 7.98–8.04 (2H, m, H_Ar), 8.06–8.14 (2H, m, H_Ar); ¹³C NMR (CDCl₃, 125 MHz, 25 °C): 31.1 (CH₂), 33.1 (CH₂),40.4 (CH₂), 64.0 (C-5′), 71.7 (C-2′), 74.5 (C-3′), 81.1 (C-4′), 86.3 (C-1′), 119.2 (C^IV–Cl), 126.1 (Phe), 128.47 (C^IV), 128.48 (Phe), 128.57 (Phe), 128.65 (Phe), 128.67 (Phe), 128.79 (Phe), 128.80 (C^IV), 129.4 (C^IV), 129.8 (Phe), 129.99 (Phe), 130.05 (Phe), 133.6 (Phe), 133.85 (Phe), 133.93 (Phe), 138.1 (C-8), 141.3 (C^IV), 149.9 (C^IV), 155.2 (C^IV), 155.3 (C^IV), 165.3 (C^IV), 165.5 (C^IV), 166.2 (C^IV); HRMS (ESI, + 3 kV, 300 °C): m/z calcd for C₄₀H₃₅ClN₅O₇: 732.22195, found: 732.22285 [M+H]⁺.

Synthesis of N⁶-[3-(1,12-dicarba-closo-dodecaboran-1-yl)propyl]-2-chloroadenosine (23) and 6-N-(3-phenylpropyl)-2-chloroadenosine (24).

For the removal of benzoyl groups, compounds 21 (154 mg,0.19 mmol) and 22 (448 mg, 0.61 mmol) were treated with 7 N methanolic solution of ammonia (25 mL) at room temperature for 20 h. Next, the silica gel was added directly to the post-reaction mixture, solvent was evaporated under the reduced pressure and the crude products were purified by silica gel column chromatography using a gradient of CH₃OH in CHCl₃ (5–10%) as an eluent.

Yield 23: 78 mg, 82%; TLC (CH₂Cl₂/CH₃OH 9:1): R_f = 0.47; UV–Vis (95% C₂H₅OH) λ, nm: 236 (min), 273 (max); FT-IR=(KBr) ν, cm⁻¹: 3334, 2926, 2605, 1624, 1581, 1534, 1476, 1355, 1309, 1229, 1113, 1078; M.P.: 183.2–184.1 °C; RP-HPLC, t_R = 29.05 min; ¹H NMR (CD₃OD, 500 MHz, 25 °C): 1.55 – ~2.80 (10H, bm, BH) (very broad multiplet which superposes with two CH₂ peaks within the range),1.48–1.58 (2H, m, CH₂), 1.72–1.82 (2H, m, CH₂), 3.10 (1H, bs, CH-carborane), 3.38 (1H, t, J = 7.0 Hz, CH₂(NH)), 3.75 (1H, dd, J = 3.0, 12.5 Hz, 5′_b-H), 3.89 (1H, dd, J = 2.5, 12.5 Hz, 5′_a-H), 5.15 (1H, td, J = 2.5, 3.0 Hz, 4′-H), 4.31 (1H, dd, J = 3.0, 5.0 Hz, 3′-H), 4.66 (1H, dd, J = 5.0, 6.0 Hz, 2′-H), 5.89 (1H, d, J = 6.0 Hz, 1′-H), 8.21 (1H, s, 8-H); ¹³C NMR (CD₃OD, 125 MHz, 25 °C): 30.3 (CH₂), 37.3 (CH₂), 40.8 (CH₂), 59.9 (CH-carborane), 63.3 (C-5′), 72.4 (C-3′), 75.4 (C-2′), 85.3 (C^IV-carborane), 87.9 (C-4′), 91.0 (C-1′), 120.2 (C^IV–Cl), 141.6 (8-C), 150.3 (C^IV), 156.6 (C^IV), 255.3 (C^IV); ¹¹B NMR {¹H BB} (CDCl₃, 160 MHz, 25 °C): 12.92 (5B s), −15.36 (5B s); HRMS (ESI, + 3 kV, 300 °C): m/z calcd for C₁₅H₂₉ClN₅B₁₀O₄: 486.29058, found: 486.29060 [M+H]⁺.

Yield 24: 205 mg, 80%; TLC (CH₂Cl₂/CH₃OH 9:1): R_f = 0.65; UV–Vis (95% C₂H₅OH) λ, nm: 235 (min), 272 (max); FT-IR=(KBr) ν, cm⁻¹: 3316, 2925, 2858, 1624,1581, 1536, 1479, 1453, 1355, 1310, 1226, 1200, 1111, 1078; M.P.: 140.2–140.6 °C; RP-HPLC, t_R = 20.62 min; ¹H NMR (CD₃OD, 500 MHz, 25 °C): 1.98 (2H, tt, J = 7.0, 7.5 Hz, CH₂), 2.71 (2H, t, J = 7.5 Hz, CH₂(Phe)), 3.57 (2H, t, J = 7.0 Hz, CH₂(NH)), 3.72 (1H, dd, J = 3.0, 12.0 Hz, 5′_a-H), 3.89 (1H, dd, J = 2.5, 12.5 Hz, 5′_b-H), 4.15 (1H, td, J = 2.5, 3.0 Hz, 4′-H), 4.32 (1H, dd, J = 3.0, 5.0 Hz, 3′-H), 4.67 (1H, dd, J = 5.0, 6.0 Hz, 2′-H), 5.90 (1H, d, J = 6.0 Hz, 1′-H), 7.10–7.16 (1H, m, Phe-H), 7.17–7.27 (4H, m, Phe-H), 8.21 (1H, s, 8-H); ¹³C NMR (CD₃OD, 125 MHz, 25 °C): 32.1 (CH₂), 34.1 (CH₂), 41.2 (CH₂), 63.4 (C-5′), 72.4 (C-3′), 75.5 (C-2′),87.9, (C-4′), 91.1 (C-1′), 120.3 (C^IV–Cl), 126.9 (Phe), 129.36 (Phe), 129.42 (Phe), 141.5 (C-8), 142.9 (C^IV), 150.2 (C^IV), 155.4 (C^IV), 156.7 (C^IV); HRMS (ESI, +3 kV, 300 °C): m/z calcd for C₁₉H₂₃ClN₅O₄: 420.14331, found: 420.14323 [M+H]⁺.

Synthesis of 3′,5′-bis-O-(4-methylbenzoyl)-N⁶-[3-(1,12-dicarbacloso-dodecaboran-1-yl)propyl]-2-chloro-2′-deoxyadenosine (25) and 3′,5′-bis-O-(4-methylbenzoyl)-N⁶-(3-phenylpropyl)-2-chloro-2′⁻deoxyadenosine (26).

To the 1 eq. of 3′,5′-bis-O-(4-methylbenzoyl)-2,6-dichloro-2′-deoxyadenosine (19), dissolved in dry DMF (10 mL/0.5 mmol of nucleoside), 3 eq. of 3-phenyl-1-propylamine hydrochloride (14) or 1.2 eq. of 3-(1,12-dicarba-closo-dodecaboran-1-yl) propylamine (4) hydrochloride followed by 3 eq. of triethylamine were added and reaction mixture was stirred at room temperature for 20 h. Obtained solution was diluted with ethyl acetate and washed twice with water and once with brine. The organic phase was dried over magnesium sulfate, filtrated and evaporated with silica gel. Compounds 25 and 26 were purified by silica gel column chromatography using a gradient of ethyl acetate in hexane (20–30%) as an eluent.

Yield 25: 140 mg (80%, starting from 135 mg of dichloropurine nucleoside); TLC (CH₂Cl₂/CH₃OH 9:1): R_f = 0.86; UV–Vis (95% C₂H₅OH) λ, nm: 227, 261 (min), 242, 271 (max); FT-IR (KBr) ν, cm⁻¹: 3355, 3063, 2925, 2867, 2608, 1930, 1721, 1615, 1578, 1533, 1503, 1409, 1351, 1309, 1270, 1176, 1099, 1019; M.P.: 80.2–80.9 °C; ¹H NMR (CDCl₃, 500 MHz, 25 °C): 1.48–1.58 (2H, m, CH₂), 1.60 – ~2.80-(10H, bm, BH) – very broad multiplet which superposes with five other peaks within the range (CH-carborane, 2xCH₃, 2xCH₂),1.66–1.78 (2H, m, CH₂), 2.41 (3H, s, CH₃), 2.45 (3H, s, CH₃), 2.64 (1H, bs, CH-carborane), 2.81–2.91 (2H, m, 2′-H), 3.45 (2H, bs, CH₂(NH)),4.61 (1H, td, J = 2.5, 4.0 Hz, 4′-H), 4.66 (1H, dd, J = 4.0, 12.0 Hz, 5′_b-H), 4.72 (1H, dd, J = 4.0, 12.0 Hz, 5′_a-H), 5.65–5.85 (2H, m, 3′-H + NH), 6.51 (1H, dd, J = 6.0, 8.0 Hz, 1′-H), 7.20–7.25 (2H, m, Phe-H), 7.27–7.31 (2H, m, Phe-H), 7.86–7.93 (3H, m, Phe-H + 8-H), 7.93–7.94 (2H, m, Phe-H); ¹³C NMR (CDCl₃, 125 MHz, 25 °C): 21.88 (CH₃), 21.92 (CH₃), 29.5 (CH₂), 36.1 (CH₂), 38.8 (C-2′), 40.1 (CH₂),58.4 (CH-carborane), 64.3 (C-5′), 75.2 (C-3′), 83.2 (C-4′), 83.8 (C^IV-carborane), 84.6 (C-1′), 119.0 (C^IV–Cl), 126.5 (C^IV), 126.8 (C^IV), 129.4 (Phe), 129.5 (Phe), 129.8 (Phe), 130.0 (Phe), 138.0 (C-8), 144.4 (C^IV), 144.7 (C^IV),149.7 (C^IV), 154.9 (C^IV), 155.3 (C^IV), 166.1 (C^IV), 166.3 (C^IV); ¹¹B NMR {¹H BB} (CDCl₃, 160 MHz, 25 °C): 12.94 (5B, s), −15.39 (5B,s); HRMS (ESI): m/z calcd for C₃₁H₄₁ClN₅B₁₀O₅: 706.37939, found: 706.37978 [M+H]⁺.

Yield 26: 276 mg (78%, starting from 300 mg of dichloropurine nucleoside); TLC (CH₂Cl₂/CH₃OH 9:1): R_f = 0.86; UV–Vis (95% C₂H₅OH) λ, nm: 228, 259 (min), 243, 270 (max); FT-IR (KBr) ν, cm⁻¹: 3247, 3108, 3063, 3029, 2933, 2853, 1936, 1721, 1621, 1574, 1498, 1453, 1429, 1408, 1383, 1331, 1279, 1221, 1176, 1140, 1104, 1087, 1048, 1017; M.P.: 156.8–157.5 °C; ¹H NMR (CDCl₃, 500 MHz, 25 °C):1.95–2.04 (2H, m, CH₂), 2.39 (3H, s, CH₃), 2.73 (2H, t, J = 7.0 Hz, CH₂(Phe)), 2.81–2.91 (2H, m, 2′-H), 3.65 (2H, bs, CH₂(NH)), 4.61 (1H, td, J = 2.5, 4.0 Hz, 4′-H), 4.66 (1H, dd, J = 4.0, 12.0 Hz, 5′_a-H), 4.73 (1H, dd, J = 4.0, 12.0 Hz, 5′_b-H), 2.44 (3H, s, CH₃), 5.78–5.72 (1H, m, 3′-H), 5.91 (1H, bs, NH), 6.51 (1H, dd, J = 6.0, 8.0 Hz, 1′-H),7.15–7.25 (5H, m, Phe-H), 7.25–7.31 (4H, m, Phe-H), 7.93–7.87 (3H, m, Phe-H + 8-H), 7.94–7.99 (2H, m, Phe-H); ¹³C NMR (CDCl₃, 125 MHz, 25 °C): 21.8 (CH₃), 21.9 (CH₃), 31.1 (CH₂), 33.1 (CH₂), 38.7 (C-2′), 40.4 (CH₂), 64.3 (C-5′), 75.2 (C-3′), 83.2 (C-4′), 84.5 (C-1′), 119.1 (C^IV–Cl), 126.2 (Phe), 126.5 (C^IV), 126.8 (C^IV), 128.5 (Phe), 128.6 (Phe), 129.40 (Phe), 129.43 (Phe), 129.7 (Phe), 130.0 (Phe), 137.8 (C-8), 141.3 (C^IV), 144.3 (C^IV), 144.7 (C^IV), 149.6 (C^IV), 154.9 (C^IV), 155.5 (C^IV), 166.1 (C^IV), 166.2 (C^IV); HRMS (ESI): m/z calcd for C₃₅H₃₅ClN₅O₅: 640.23223, found: 640.23212 [M+H]⁺.

Synthesis of N⁶-[3-(1,12-dicarba-closo-dodecaboran-1-yl)propyl]-2-chloro-2′-deoxyadenosine (27) and N⁶-(3-phenylpropyl)-2-chloro-2′-deoxyadenosine (28).

For the removal of 4-methylbenzoyl group, compounds 25 (140 mg, 0.20 mmol) or 26 (200 mg, 0.31 mmol) were treated with 7 N methanolic solution of ammonia (25 mL) at room temperature for 20 h. Next, the silica gel was added directly to the reaction mixture, solvent was evaporated under the reduced pressure and crude the products were purified by silica gel column chromatography using a gradient of CH₃OH in CHCl₃ (5–10%) as an eluent. Yield of 27: 69 mg, 74%; TLC (CH₂Cl₂/CH₃OH, 9:1): R_f = 0.49; UV–Vis (95% C₂H₅OH) λ, nm: 235 (min), 271 (max); FT-IR (KBr) ν, cm⁻¹: 3324, 2928, 2605, 1622, 1581, 1535, 1476, 1352, 1308, 1231, 1182, 1094, 1067, 1005; M.P.: 114.1–120.0 °C; RP-HPLC, t_R = 30.56 min; ¹H NMR (CD₃OD, 500 MHz, 25 °C): 1.50- ~2.80 (10H, bm, BH) – very broad multiplet which superposes with three other peaks within the range (H_2’a, H_2’b, 2 × CH₂), 1.49–1.57 (2H, m, CH₂),1.73–1.82 (2H, m, CH₂), 2.40 (1H, ddd, J = 3.0, 6.0, 13.5 Hz, 2′_a-H), 2.77 (1H, ddd, J = 6.0, 7.5, 13.5 Hz, 2′_b-H), 3.11 (1H, bs, CH-carborane), 3.39 (1H, t, J = 7.0 Hz, CH₂(NH)), 3.75 (1H, dd, J = 3.5, 12.0 Hz, 5′_a-H), 3.85 (1H, dd, J = 3.0, 12.0 Hz, 5′_b-H), 4.05 (1H, dt, J = 3.0, 3.5 Hz, 4′-H), 4.57 (1H, dt, J = 3.0, 3.5 Hz, 3′-H), 6.35 (1H, dd, J = 6.0, 7.5 Hz, 1′-H), 8.22 (1H, s, 8-H); ¹³C NMR (CD₃OD, 125 MHz, 25 °C): 30.3 (CH₂), 37.3 (CH₂), 40.8 (CH₂), 41.4 (C-2′), 59.9 (CH-carborane), 63.5 (C-5′), 72.8 (C-3′), 85.3 (C^IV-carborane), 86.9 (C-1′),89.7 (C-4′), 120.0 (C^IV–Cl), 141.2 (C-8), 150.3 (C^IV), 155.3 (C^IV), 156.5 (C^IV); ¹¹B NMR {¹H BB} (CDCl₃, 160 MHz, 25 °C): 12.92 (5B,s), −15.37 (5B, s); HRMS (ESI): m/z calcd for C₁₅H₂₉ClN₅B₁₀O₃: 470.29566, found: 470.29552 [M+H]⁺.

Yield 28: 105 mg, 83%; TLC (CH₂Cl₂/CH₃OH, 9:1): R_f = 49; UV–Vis (95% C₂H₅OH) λ, nm: 232 (min), 271 (max); FT-IR (KBr) ν, cm⁻¹: 3316, 3026, 2925, 2857, 1706, 1623, 1578, 1536, 1477, 1453, 1354, 1309, 1227, 1093, 1065; M.P.: 58.0–58.4 °C; RP-HPLC, t_R = 21.31 min; ¹H NMR (CD₃OD, 500 MHz, 25 °C): 1.98 (2H, tt, J = 7.0, 7.5 Hz, CH₂), 2.41 (1H, ddd, J = 3.0, 6.0, 13.5 Hz 2′_a-H), 2.72 (2H, t, J = 7.5 Hz, CH₂(Ph)), 2.77 (1H, ddd, J = 6.0, 7.5, 13.5 Hz 2′_b-H), 3.57 (2H, t, J = 7.0 Hz, CH₂(NH)), 3.75 (1H, dd, J = 3.5, 12.5 Hz, 5′_b-H), 3.85 (1H, dd, J = 3.0, 12.5 Hz, 5′_a-H), 4.06 (1H, dt, J = 3.0, 3.5 Hz, 4′-H), 4.57 (1H, dt, J = 3.0, 6.0 Hz, 3′-H), 6.36 (1H, dd, J = 6.0, 7.5 Hz, 1′-H), 7.08–7.17 (1H, m, Phe-H), 7.17–7.32 (4H, m, Phe-H), 8.22 (1H, s, 8-H); ¹³C NMR (CD₃OD, 125 MHz, 25 °C): 32.1 (CH₂), 34.0 (CH₂), 41.2 (CH₂), 41.4 (C-2′), 63.6 (C-5′), 72.9 (C-3′), 86.9 (C-1′), 89.7 (C-4′), 120.1 (C^IV–Cl), 126.8 (Phe), 129.3 (Phe), 129.4 (Phe), 141.1 (C-8), 142.9 (C^IV), 150.2 (C^IV), 155.4 (C^IV), 156.6 (C^IV); HRMS (ESI): m/z calcd for C₁₉H₂₃ClN₅O₃: 404.14839, found: 404.14845 [M+H] ⁺.

Synthesis of 8-[2-(1,12-dicarba-closo-dodecaboran-1-yl)ethynyl] adenosine (33) and 8-(2-phenylethynyl)adenosine (34) [23].

The synthesis was performed under anhydrous conditions in an argon atmosphere. To the 1 eq. of 8-bromoadenosine (30), dissolved in dry DMF (10 mL/0.5 mmol of nucleoside), 1.5 eq. of appropriate acetylene derivative, 3 eq. of triethylamine were added, followed by 0.2 eq. of CuI and 0.1 eq. of Pd(PPh₃)₂Cl₂. The reaction mixture was stirred at room temperature for 20 h. Next day, volatiles were evaporated under reduced pressure, obtained residue was redissolved in methanol and evaporated with silica gel, then purified by column chromatography using 10% methanol in chloroform as an eluent.

Yield 33: 71 mg (from 94 mg of 30), 63%; TLC (CH₂Cl₂/CH₃OH, 9:1): R_f = 0.30; UV–Vis (95% C₂H₅OH) λ, nm: 252 (min), 236, 301 (max); FT-IR (KBr) ν, cm⁻¹: 3214, 3047, 2932, 2874, 2614, 2207, 1911, 1648, 1600, 1576, 1477, 1450, 1414, 1370, 1333, 1297, 1238, 1193, 1131, 1078, 1062; M.P.: 201–202 °C; RP-HPLC, t_R = 24.60 min; ¹H NMR (CD₃OD, 500 MHz, 25 °C) δ, ppm: 1.30–3.20 (9H, bm, BH); 3.44 (1H, bs, CH^carborane), 3.70–3.80 (2H, m, CH^carborane +5′_a-H), 3.89 (1H, dd, J = 2.5, 12.5 Hz, 5′_b-H), 4.18 (1H, ddd, J = 2.0, 2.5, 3.0 Hz, 4′-H), 4.38 (1H, dd, J = 2.0, J = 5.5 Hz, 3′-H); 5.01 (1H, dd, J = 5.5, 7.0 Hz, 2′-H); 8.19 (1H, s, 2-H); ¹³C NMR (CD₃OD, 125 MHz, 25 °C) δ, ppm: 64.1 (C-5′); 65.4 (CH^carborane), 68.0 (CH^carborane), 73.1 (C-3′), 74.2 (C-2′), 85.2 (C≡C), 88.8 (C-4′), 91.9 (C-1′), ~98.5 (C≡C, very broad), 120.7 (C^IV), 135.7 (C^IV), 149.5 (C^IV), 154.4 (C-2), 157.5 (C^IV); ¹¹B NMR {¹H BB} (CD₃OD₃, 160 MHz, 25 °C) δ, ppm: 13.53, −14.86; HRMS (ESI): m/z calcd for C₁₄H₂₄B₁₀N₅O₄: 434.28260, found: 434.28284 [M+H]⁺.

Yield 34: 814 mg (from 900 mg of 30), 85%; TLC (CH₂Cl₂/CH₃OH, 9:1): R_f = 0.28; UV–Vis (95% C₂H₅OH) λ, nm: 241, 251, 267 (min), 247, 254, 314 (max); FT-IR (KBr) ν, cm⁻¹: 3333, 2935, 2219, 1649, 1603, 1569, 1503, 1442, 1395, 1332, 1308, 1255, 1227, 1196, 1092, 1045; M.P.: 211–212 °C; RP-HPLC, t_R = 17.16 min; ¹H NMR (DMSO-d₆, 500 MHz, 25 °C) δ, ppm: 3.55 (1H, dd, J = 4.0, 12.0 Hz, 5′_a-H), 3.70 (1H, dd, J = 4.0, 12.0 Hz, 5′_a-H), 3.92–4.08 (1H, m, 4′-H), 4.14–4.34 (1H, m, 3′-H), 5.03 (1H, t, J = 6.0 Hz, 2′-H), 5.06–5.30 (1H, bs, OH), 5.30–5.82 (2H, bm, 2 × OH), 6.06 (1H, d, J = 6.0 Hz, 1′-H),7.47–7.60 (3H, m, Phe-H), 7.60–7.95 (4H, m, Phe-H + NH₂), 8.19 (1H, s, 2-H); ¹³C NMR (DMSO-d₆, 125 MHz, 25 °C) δ, ppm: 62.2 (C-5′), 71.0 (C-3′), 71.8 (C-2′), 78.5 (C≡C), 86.8 (C-4′), 89.5 (C_1’), 94.4 (C≡C), 119.7 (C^IV), 119.9 (C^IV), 129.2 (Phe), 130.5 (Phe), 131.9 (Phe), 133.4 (C^IV), 148.6 (C^IV), 153.4 (C-2), 156.1 (C^IV); HRMS (ESI): m/z calcd for C₁₈H₁₈N₅O₄: 368.13533, found: 368.13530 [M+H]⁺.

Synthesis of 8-[2-(1,12-dicarba-closo-dodecaboran-1-yl)ethyl]-adenosine (35) and 8-(2-phenylethyl)adenosine (36).

Compound 33 (35 mg, 0.08 mmol) or compound 34 (400 mg, 1.09 mmol) were dissolved in methanol (20 mL/1 mmol of nucleoside) then 30 mg of palladium on activated charcoal (10% Pd basis) was added to the solution. The reaction vessel was connected to a balloon filled with hydrogen and the reaction mixture was stirred at room temperature for 20 h. The resultant solution was filtered through Celite, evaporated and the crude product was purified by silica gel column chromatography using 10% methanol in chloroform as an eluent.

Yield 35: 30 mg (85%); TLC (CH₂Cl₂/CH₃OH, 9:1): R_f = 0.29; UV–Vis (95% C₂H₅OH) λ, nm: 231 (min), 262 (max); IR (KBr) ν, cm⁻¹: 3584, 3433, 3340, 3223, 3092, 3050, 2924, 2867, 2604, 1651, 1606, 1578, 1515, 1483, 1449, 1423, 1376, 1334, 1295, 1272, 1234, 1181, 1122, 1070, 1043; M.P.: 204–205 °C; RP-HPLC, t_R = 24.02 min; ¹H NMR (CD₃OD, 500 MHz, 25 °C) δ, ppm: 1.53–1.60 (2H, m, CH₂–CH₂-carborane), 1.62–2.97 (9H, m, BH), 3.00–3.13 (2H, m, CH₂–CH₂-carborane), 3.26 (1H, bs, CH^carborane), 3.46 (1H, bs, CH^carborane), 3.73 (1H, dd, J = 2.5, 12.5 Hz, 5′_a-H), 3.89 (1H, dd, J = 2.0, 12.5 Hz, 5′_b-H), 4.18–4.22 (1H, m, 4′-H), 4.34 (1H, dd, J = 1.5, 5.0 Hz, 3′-H), 5.00 (1H, dd, J = 5.0, 7.5 Hz, 2′-H), 5.92 (1H, d, J = 7.5 Hz, 1′-H), 8.11 (1H, s, 2-H); ¹³C NMR (CD₃OD, 125 MHz, 25 °C) δ, ppm: ~14.1 (CH₂–CH₂-carborane, very broad); 28.1 (CH₂–CH₂-carborane), 64.1 (C-5′), 65.2 (CH^carborane), 67.9 (CH^carborane), 73.3 (C-3′), 74.2 (C-2′), 88.9 (C-4′), 90.6 (C-1′), 119.5 (C^IV), 150.9 (C^IV), 152.5 (C-2), 156.3 (C^IV), 156.8 (C^IV); ¹¹B NMR {¹H BB} (CD₃OD₃, 160 MHz, 25 °C) δ, ppm: 4.80, −13.58, −14.56, −15.93; HRMS (ESI): m/z calcd for C₁₄H₂₈B₁₀N₅O₄: 438.31390, found: 438.31397 [M+H]⁺.

Yield 36: 390 mg (97%); TLC (CH₂Cl₂/CH₃OH, 9:1): R_f = 0.25; UV–Vis (95% C₂H₅OH) λ, nm: 232 (min), 262 (max); IR (KBr) ν, cm⁻¹: 3435, 3348, 3228, 2930, 2894, 1658, 1606, 1547, 1496, 1452, 1386, 1333, 1230,1092, 1044; M.P.:183–184 °C; ¹H NMR (DMSO-d₆, 500 MHz, 25 °C) δ, ppm: 3.02–3.25 (4H, m, $2\times {\text{CH}}_{\text{2}}^{\text{linker}}$); 3.55 (1H, ddd, J = 3.0, 9.5, 12.5 Hz, 5′_a-H), 3.69 (1H, ddd, J = 3.0, 3.5, 12.5 Hz, 5′_b-H), 4.02 (1H, td, J = 2.0, 3.0, 4′-H), 4.15 (1H, ddd, J = 2.0, 4.5, 5.5 Hz, 3′-H), 4.91 (1H, ddd, J = 5.5, 7.0, 7.5 Hz, 2′-H), 5.25 (1H, d, J = 4.5 Hz, 3′-OH), 5.42 (1H, d, J = 7.5 Hz, 2′-OH), 5.85 (1H, d, J = 7.0 Hz, 1′-H), 5.95 (1H, dd, J = 3.5, 9.5 Hz, 5′-OH), 7.15–7.25 (1H, m, Phe-H), 7.25–7.40 (6H, m, Phe-H + NH₂), 8.07 (s, 1H, 2-H); ¹³C NMR (DMSO-d₆, 125 MHz, 25 °C) δ, ppm: 29.3 (CH₂CH₂Phe); 33.1 (CH₂CH₂Phe), 62.3 (C-5′), 71.1 (C-3′), 72.0 (C-2′), 86.9 (C-4′), 88.4 (C-1′), 118.3 (C^IV), 126.2 (Phe), 128.4 (Phe), 140.7 (C^IV), 149.6 (C^IV), 151.4 (C-2), 151.6 (C^IV), 155.6 (C^IV); HRMS (ESI): m/z calcd for C₁₈H₂₂N₅O₄: 372.16663, found: 372.16643 [M+H]⁺.

5.2. Radioligand inhibition assay for compounds 7, 8, 15–18, 23, 24, 27–28

Competition Binding in Human A₁ receptors.

Adenosine A₁ receptor competition binding experiments were carried out in a multiscreen GF/C 96-well plate (Millipore, Madrid, Spain) pre-treated with binding buffer (Hepes 20 mM, NaCl 100 mM, MgCl₂ 10 mM, 2 U/ml adenosine deaminase, pH = 7.4). In each well was incubated 5 μg of membranes from Euroscreen CHO-A₁ cell line and prepared in Plataforma de Screening de Farmacos (USEF) (Lot: A002/13-04-2011, protein concentration = 5864 μg/mL), 1 nM [³H]-DPCPX (120 Ci/mmol, 1 mCi/mL, PerkinElmer NET974001MC) and compounds studied and standard. Non-specific binding was determined in the presence of R-PIA 10 μM (Sigma P4532). The reaction mixture (Vt: 200 μl/well) was incubated at 25 °C for 60 min, after was filtered and washed four times with 250 μl wash buffer (Hepes 20 mM, NaCl 100 mM, MgCl₂ 10 mM pH = 7.4), before measuring in a microplate beta scintillation counter (Microbeta Trilux, PerkinElmer, Madrid, Spain).

Competition Binding in Human A_2A receptors.

Adenosine A_2A receptor competition binding experiments were carried out in a multiscreen GF/C 96-well plate (Millipore, Madrid, Spain) pre-treated with binding buffer (Tris-HCl 50 mM, EDTA 1 mM, MgCl₂ 10 mM, 2 U/ml adenosine deaminase, pH = 7.4). In each well was incubated 5 μg of membranes from HeLa-A_2A cell line and prepared in Plataforma de Screening de Farmacos (USEF) (Lot: A001/18-10-2010, protein concentration = 4288 μg/mL), 3 nM [³H]-ZM241385 (50 Ci/mmol, 1 mCi/mL, ARC-ITISA 0884) and compounds studied and standard. Non-specific binding was determined in the presence of NECA 50 μM (Sigma E2387). The reaction mixture (Vt: 200 μl/well) was incubated at 25 °C for 30 min, after was filtered and washed four times with 250 μl wash buffer (Tris-HCl 50 mM, EDTA 1 mM, MgCl₂ 10 mM, pH = 7.4), before measuring in a microplate beta scintillation counter (Microbeta Trilux, PerkinElmer, Madrid, Spain).

Competition Binding in Human A_2B receptors.

Adenosine A_2B receptor competition binding experiments were carried out in a polypropylene 96-well plate. In each well was incubated 20 μg of membranes from Euroscreen HEK-A_2B cell line and prepared in Plataforma de Screening de Farmacos (USEF) (Lot: A007/22-02-2016, protein concentration = 3211 μg/mL), 25 nM [³H]-DPCPX (164 Ci/mmol, 1 mCi/mL, PerkinElmer NET974001MC) and compounds studied and standard. Non-specific binding was determined in the presence of NECA 1000 μM (Sigma E2397). The reaction mixture (Vt: 250 μl/well) was incubated at 25 °C for 30 min, 200 μL was transferred to GF/C 96-well plate (Millipore, Madrid, Spain) pre-treated with binding buffer (Tris-HCl 50 mM, EDTA 1 mM, MgCl₂ mM, Bacitracin 100 μg/μl, adenosine deaminase 2 U/mL, pH = 6.5), after was filtered and washed four times with 250 μl wash buffer (Tris-HCl 50 mM, EDTA 1 mM, MgCl₂ 5 mM, pH = 6.5), before measuring in a microplate beta scintillation counter (Microbeta Trilux, PerkinElmer, Madrid, Spain).

Competition Binding in Human A₃ receptors.

Adenosine A₃ receptor competition binding experiments were carried out in a multiscreen GF/B 96-well plate (Millipore, Madrid, Spain) pre-treated with binding buffer (Tris-HCl 50 mM, EDTA 1 mM, MgCl₂ 5 mM, 2 U/ml adenosine deaminase, pH = 7.4). In each well was incubated 70 μg of membranes from HeLa-A₃ cell line and prepared in our laboratory (Lot: A003/13-04-2016, protein concentration = 2449 μg/mL), 10 nM [³H]-NECA (27.6 Ci/mmol, 1 mCi/mL, PerkinElmer NET811250UC) and compounds studied and standard. Non-specific binding was determined in the presence of R-PIA 100 μM (Sigma P4532). The reaction mixture (Vt: 200 μl/well) was incubated at 25 °C for 180 min, after was filtered and washed six times with 250 μl wash buffer (Tris-HCl 50 mM pH = 7.4), before measuring in a microplate beta scintillation counter (Microbeta Trilux, PerkinElmer, Madrid, Spain).

Radioligand inhibition assay for compounds 33–55.

Radioligand binding assays for compounds 33–52 were performed using HEK293 cells expressing human A₁, A_2A, or A₃ receptors and standard radioligands, as previously described [39].

5.3. Functional cAMP assay for compounds 7, 8, 15, 16, 23 and 24 in human A₃ receptor

Chinese hamster ovary (CHO) cells transfected with the human A₃ receptor (CHO-A₃, Centro de Investigación CIMUS, Barcelona) were cultured in Dulbecco’s modified eagle’s medium nutrient mixture F-12 ham (Sigma Aldrich; D8062) pH = 7.4, containing 25 mM Hepes and 30 μM rolipram. As the wash buffer Dulbecco’s modified eagle’s medium nutrient mixture F-12 ham pH = 7.4 containing 25 mM Hepes was used. Initially, the CHO-A₃ cells were incubated with the vehicle (for an agonists mode) or with the increasing concentrations of compounds (for an antagonist mode) at 37 °C, 5% CO₂, for 15 min. Next, the increasing concentrations of compounds and 0.1 μM of NECA (for agonists mode), or vehicle and 0.1 μM of NECA (for antagonist mode) were added for the next 15 min (37 °C, 5% CO₂). Subsequently, 10 μМ οf forskolin was added and the incubation was continued for next 5 min (37 °C, 5% CO₂). Determination of the cAMP amount was performed using Amersham cAMP Enzyme immunoassay (Amersham, RPN 225). Absorbance measurement was measured at 450 nm using a Tecan Infinite M1000 Pro Microplate Reader. The MRS1220 and NECA (agonist mode) were used as the standard controls for validation of the assay (antagonist and agonist mode, respectively). The affinity values obtained for standard controls were as follow: NECA, EC₅₀ = 4 nM (described in the literature 20 nM [39]); MRS1220, K_B = 4.0 nM (described in the literature 1.7 nM [56]).

5.4. Patients and biological materials

Blood was drawn from 57 healthy donors including 27 men and 30 women, with the average age of 30 years in the case of men and 33 years in the case of women. The samples were collected into a vacuum tube containing 25 μg/mL hirudin, for whole blood aggregometry, or 0.105 M buffered sodium citrate, for optical aggregation and thrombus formation in flow conditions. The study was performed under the guidelines of the Helsinki Declaration for human research and approved by the Medical University of Lodz committee on the Ethics of Research in Human Experimentation.

5.5. Inhibition of human blood platelets aggregation

Aggregation of blood platelets in a platelet rich plasma.

The blood samples were centrifuged for 12 min at 190×g to obtain platelet-rich plasma (PRP). The platelet count in the PRP was adjusted to 2 × 10⁸ platelets/ml prior to its use in experiments. Five minutes prior to measurements samples were added with a tested compound (dissolved in 0.1% DMSO), control samples were added with 0.1% DMSO. ADP was added to each sample to final concentration of 10 μM to induce platelet aggregation, which was thereafter monitored for 10 min using an optical aggregometer Chrono-Log 490–2D (Chrono-Log,Havertown, PA, USA).The total area under the aggregation curve (AUC) and the maximal value of platelet aggregation (Amax,U) were recorded [57].

Aggregation of blood platelets in whole blood.

Samples were pre-incubated with the aliquot of a tested compound or DMSO as described above. Whole platelet aggregation was monitored thereafter for up to 15 min in response to 6.4 μM ADP, 1 μg/mL collagen or 0.5 mM arachidonic acid using the five-channel Multiplate analyzer (Roche Diagnostics GmbH, Mannheim, Germany). The total area under the aggregation curve (AUC) was recorded [57,58].

Formation of thrombus in flow conditions.

Effects of the tested compounds on the formation of thrombi were assayed with the use of Venaflux platform (Cellix, Dublin, Ireland). The channels of Fluoro + biochips were coated with type I collagen. Citrated blood was incubated with anti-CD61 PE-conjugated antibodies to label blood platelets. Tested compounds or 0.1% DMSO were added 5 min prior to measurements. Samples were recalcified with 1 mM CaCl₂ shortly before running in the system. The samples were then perfused with a shear force of 20 dyn/cm² for 5 min. The thrombi in the channels were imaged with the use of AxioExaminer epifluorescence microscope (Zeiss, Jena, Germany) equipped with a high performance EMCCD camera Rolera em-c² (QImaging, Surrey, Canada). Areas of individual thrombi were estimated with the use of a dedicated ZEN software.

5.6. Structure-activity relationship (SAR)

Current commercial software packages do not have built-in parameters for hexacoordinated boron atoms. Therefore, to enable comparative analysis of carboranyl and phenyl compounds, qualitative structural descriptors were generated. The following substructures/features were selected for analysis: type of sugar (ribose, R; 2′-deoxyribose, dR; and arabinose, Ara), type of modifier (carborane, phenyl), position of substitution (N⁶, C2, and C8), number of carbon atoms in a linker (C₂, C₃, C₄), and linker flexibility (flexible C–C versus rigid C≡C). The activity of the compounds, i.e., receptor affinity, was expressed as log K_i values. Additionally, the receptor affinity values were transformed to categorical variables for classification analysis (active, low active, and inactive) using the k-means clustering method (Statistica software package, version 13.3, TIBCO Software Inc., Palo Alto, CA, USA). The structural features of compounds in relation to A₃, A₁, and A_2A receptor affinity were analyzed separately. For selectivity testing, compounds were categorized into two groups and analyzed as dichotomous categorical variables (A₃-selective and A₃-not selective).

Combining classes/groups of descriptors was performed using feature selection and a variable screening (FSL) module (Data Mining tools in Statistica software). Significant variables for model building were selected using the general regression model module (GRM, backward stepwise and best-subset search procedures, Wald chi-square statistics). Structural features specific only for one group of compounds were subjected to regression analysis separately, e.g., ligand flexibility/rigidity in short C₂-linker compounds.

The relationship of compound activity (log K_i, dependent variable) and previously selected substructural features (categorical predictors) was determined by a regression approach (PLS, or general classification and regression trees, GC&RT module, Statistica, with goodness-of–fit and predictivity evaluation and cross validation, Statistica software).

All synthesized compounds complied with Lipinski’s Rule of Five (online software chemicalize.com, ChemAxon 2020). The average MW values of carborane and phenyl derivatives were 453.86 ± 27.87 and 387.75 ± 27.87, respectively. Although the MW calculated for both derivative groups was significantly different (P < 0.0001, Student’s t-test; n = 15, normal data distribution; Shapiro-Wilk test), this parameter was not related to compound activity (P > 0.05), and it was not included in the analysis.

5.7. Docking study

Homology modeling of the A₃ receptor: The sequence of the A₃ AR (chain A) was collected from the Swiss-Prot Protein Database (P33765). The A₃ AR model was generated by LOMETS (Local Meta-Threading-Server [59]) using a template of the 2.6 Å resolution structure of human A_2A receptor with CGS21680 antagonist bound (Protein Data Bank, PDB code: 4UG2). The sequence identity was 0.43; coverage of the threading alignment was 0.88 (equal to the number of aligned residues divided by the length of the query protein). The loop segments were built with Modeler [60]; the obtained structure was optimized using Maestro Schrodinger software 11.7 (OPLS3e, Schrödinger, Inc., New York, NY, 2013). The A₃ AR model contained one disulfide bridge between the cysteine residue (Cys166) in the second extracellular loop (EL2) and the cysteine residue of TM3, Cys83 (3.25). The pair of carboranyl and phenyl derivatives of NECA were docked to the A₃ AR model (compounds 15 and 16). Ligands were built and initially optimized in 3D using MarvinSketch 20.14.0, 2020, ChemAxon (http://www.chemaxon.com). Partial atomic charges of compounds were calculated using semiempirical CNDO method (HyperChem 7.51 program, HyperCube Inc., Gainesville, FL, USA). The resulting partial negative charges of hydrogen atoms at BH vertices (values from −0.076 to −0.082, depending on the distance from C vertices) allowed us to retain the hydridic character of H atoms in models of carborane derivatives. The compound geometry was finally optimized in an AMBER95 force field (in vacuum, distance dependent dielectric constant 1, electrostatic interactions 0.5, van der Waals interactions 0.5, HyperChem 7.51). Boron atom parameters were implemented in AutoDock 4.2.6, and into AutoDock and AutoGrid parameter files. The following have been added [61]: vdW diameter, Rii 3.96 Å; vdW well depth, epsii 0.034 kcal/mol; atomic solvation volume 32.5281 ų; atomic solvation parameter −0.00143; the remaining parameters were default as for carbon atom. A docking procedure was performed by preserving i) the input of partial atomic charges, and ii) BH hydridic hydrogens. Polar hydrogen atoms were added and Kollman charges were assigned to protein (A₃ AR) atoms. The docking area was centered at the average position of the intra- and extracellular part of the A₃ AR (grid points in xyz, 60 × 60 × 80, grid point spacing 0.375 Å), which broadly covered the entire extracellular part of the receptor, binding pocket and intracellular region. The Lamarckian genetic algorithm (LGA) was used for conformational search of the flexible ligand. Docking parameters were as follows: number of individuals in population 100, GA-LS runs 100, maximum number of energy evaluations 10⁸ (a higher value, 2.5 × 10⁸ did not alter results), maximum number of generations 27000. Validation of the docking method was performed using I-AB-MECA, a high affinity A₃ AR agonist; the calculated I-AB-MECA K_i value was 3.5 nM; binding energy, ΔG, −11.54 kcal/mol (compared to the experimental K_i values: 1.2–2.1 nM; PubChem CID 9958472). The results were analyzed using PLIP (protein-ligand interaction profiler) open source software [62] and LigPlot+ 2.2. [63] Sequence-based GPCR residue numbering was performed according to Ballesteros-Weinstein.