Pyrazolo[4,3-e][1,2,4]triazolo[1,5-c]pyrimidines to develop functionalized ligands to target adenosine receptors: fluorescent ligands as an example

Abstract

A series of adenosine receptor antagonists bearing a reactive linker was developed. Functionalization of these derivatives is useful to easily obtain multi-target ligands, receptor probes, drug delivery systems, and diagnostic or theranostic systems. The pyrazolo[4,3-e][1,2,4]triazolo[1,5-c]pyrimidine scaffold was chosen as a pharmacophore for the adenosine receptors. It was substituted at the 5 position with reactive linkers of different lengths. Then, these compounds were used to synthesise probes for the adenosine receptors by functionalization with a fluorescent moiety. Both series of compounds were evaluated for their binding at the four adenosine receptor subtypes. Different affinity and selectivity profiles were observed towards hA₁, hA_2A and hA₃ adenosine receptors. In particular, fluorescent compounds behave as dual hA_2A/hA₃ ligands. Computational studies suggested different binding modes for developed compounds at the three receptors. Both molecular docking and supervised molecular dynamics (SuMD) simulations confirmed that the preferred binding mode at the single receptor was driven by the substitution present at the 5 position. Obtained results rationalized the compounds' binding profile at the adenosine receptors and pave the way for the development of more potent conjugable and conjugated ligands targeting these membrane receptors.

---

Introduction

Functionalization of a ligand is a strategy useful in the design of receptor probes,^1,2 drug delivery systems^3–5 and multitarget ligands.⁶ In particular, a pharmacophore is conjugated to a function bearing moiety by means of a linker, which is incorporated at an insensitive site on the pharmacophore structure.⁷ For example, receptor probes could be obtained by introducing a fluorophore,⁸ or a lipid could be used to gain specific characteristics useful in delivering the active molecule at a target tissue.⁹

Adenosine receptors (ARs) are family A G protein-coupled receptors (GPCRs) involved in a number of physiological and pathological conditions that make them promising therapeutic targets for several diseases. In particular, four distinct AR subtypes have been identified so far: A₁, A_2A, A_2B and A₃ ARs.¹⁰ Each subtype has a specific distribution in the human body, which confers to their ligands particular potential therapeutic applications, such as in: pain conditions (A₁ and A₃ agonists),¹¹ myocardial perfusion imaging (A_2A agonists),¹² cardiac and renal conditions (A₁ antagonists),¹³ neurodegenerative diseases (A_2A antagonists),¹⁴ cancer immunotherapy (A_2A antagonists),¹⁴ inflammatory bowel disease (A_2B antagonists)¹⁵ and glaucoma (A₃ antagonists).¹⁶ It is clear that a pharmacophore, with tunable affinity towards AR subtypes might be useful in order to easily obtain theranostic agents, drug delivery systems or multivalent ligands for these receptors. The pyrazolo[4,3-e][1,2,4]triazolo[1,5-c]pyrimidine (PTP) is a well-known scaffold (Fig. 1) from which many potent AR antagonists have been derived. Among these, preladenant (1) is an A_2AAR antagonist under clinical development to treat neurodegeneration and other pathological conditions.¹⁷ As depicted in Fig. 1, the PTP scaffold could be simultaneously substituted at different positions, which determines a desired specific affinity and selectivity profiles for the targets. In addition, we have extensive experience on this scaffold.^18–21 For these reasons, in this work we have developed conjugable AR antagonists bearing a PTP scaffold (2–19). In order to understand the structural requirements crucial to retain the affinity and selectivity towards ARs, we have investigated these antagonists both from a pharmacological and computational perspective. In a previous study, these compounds were functionalized with a fluorophore leading to fluorescent receptor probes (20–25), which we have reported here, with their full characterization, as an example of application of our strategy.^22,23

Fig. 1 General structure of the pyrazolo[4,3-e][1,2,4]triazolo[1,5-c]pyrimidine (PTP) nucleus and structure of the hA_2AAR antagonist preladenant (1).

Results and discussion

Development of conjugable adenosine receptor ligands

In order to obtain a molecule that is easily functionalized with the desired function-bearing moiety, we decided to use the PTP scaffold and to introduce the linker at the 5 position. In fact, in our previous AR docking studies, we obtained data supporting that alkylamino substituents at the 5 position of PTP point away from the binding pocket of the ARs, thus becoming a suitable molecular site for modification.²⁰ The rationale is that the PTP scaffold lies inside the binding cleft of the receptor to provide high affinity and selectivity towards the target, while the linker could bridge between the pharmacophore and a function-bearing moiety in the extracellular side of the membrane.

In order to investigate the effect on the AR affinity, we decided to design nine homologous derivatives bearing diamino linkers of different lengths, from a hydrazino linker to longer diaminopolyglycolic chains. These chains are versatile because one amino group could be reacted with the AR ligand and the other with the detection moiety. In fact, there are several groups that could react with amines, such as carboxylic acids or their activated derivatives, isocyanates and isothiocyanates.

Chemistry. As reported in Scheme 1, for the synthesis of desired compounds a straightforward procedure was applied starting from the 5-chloro-pyrazolo[4,3-e][1,2,4]triazolo[1,5-c]pyrimidine intermediate 26.²² Specifically, it involved a nucleophilic substitution of chlorine with the mono-tert-butyloxycarbonyl (Boc)-protected diaminoalkyl chain in the presence of a base, which afforded 2–10, followed by a standard acid-mediated Boc deprotection to obtain derivatives 11–19.

Scheme 1 Synthesis of diamino-linked PTP 2–19. Reagents and conditions: a) appropriated mono-Boc-protected-diamino linker, TEA, ethanol, 110 °C, sealed tube, 2 h; b) tert-butyl piperazine-1-carboxylate, TEA, ethanol, 110 °C, sealed tube, 2 h; c) ethyl acetate-HCl, rt, 1 h.

Pharmacology and structure–activity relationship. Synthesized compounds were evaluated for their binding at the four human ARs (hARs). Both Boc-protected (2–10) and free amino conjugable (11–19) derivatives were assayed. As shown in Table 1, few compounds displayed selectivity towards a single AR. In addition, none of the compounds showed potency below 3 μM at the hA_2BAR.

Table 1 Binding profile at the four hAR subtypes for compounds 2–19

Compd	X	R	hA1^a (Ki nM)	hA2A^b (Ki nM)	hA2B^c (Ki nM)	hA3^d (Ki nM)
a Displacement of specific [^3H]-CCPA binding at hA1ARs expressed in CHO cells (n = 3–6). b Displacement of specific [^3H]-NECA binding at hA2AARs expressed in CHO cells (n = 3–6). c Inhibition of the NECA-stimulated adenylyl cyclase in CHO cells (n = 3–6) expressing hA2BARs. d Displacement of specific [^3H]-HEMADO binding at hA3ARs expressed in CHO cells (n = 3–6). Data are expressed as geometric means with 95% confidence limits.
2, WS86	—	Boc	7820 (6510–9380)	3900 (3630–4200)	>10 000	173 (124–241)
3	–CH2CH2–	Boc	2550 (1740–3760)	624 (467–834)	>10 000	553 (470–651)
4, WS88	–CH2CH2CH2–	Boc	1770 (1430–2200)	195 (129–295)	>3000	32.1 (21–49.1)
5	–CH2CH2CH2CH2–	Boc	1750 (1310–2340)	313 (256–383)	11 500 (8590–15 300)	84.4 (79.5–89.7)
6	–CH2CH2CH2CH2CH2–	Boc	2580 (1690–3940)	346 (299–401)	3000	129 (98.1–171)
7, WS91	–CH2CH2OCH2CH2OCH2CH2–	Boc	12 200 (10 700–13 800)	1260 (995–1590)	>10 000	585 (438–781)
8	–CH2CH2CH2O(CH2CH2O)2CH2CH2CH2–	Boc	8480 (5960–12 060)	674 (568–799)	>10 000	264 (136–515)
9	—	Boc	4070 (3470–4770)	2170 (1980–2370)	∼10 000	856 (696–1053)
10	—	Boc	>100 000	21 400 (18 900–24 300)	>10 000	737 (505–1077)
11	—	H × HCl	7100 (4870–10 300)	233 (128–422)	>10 000	174 (126–241)
12	–CH2CH2–	H × HCl	2500 (1480–4230)	756 (454–1260)	>10 000	2400 (1938–2971)
13	–CH2CH2CH2–	H × HCl	1440 (1290–1610)	253 (186–344)	>10 000	3784 (3031–4723)
14	–CH2CH2CH2CH2–	H × HCl	1220 (1080–1370)	169 (108–263)	>3000	4205 (3715–4760)
15, WS98	–CH2CH2CH2CH2CH2–	H × HCl	778 (608–994)	52.6 (45.1–61.4)	>3000	2246 (2139–2359)
16	–CH2CH2OCH2CH2OCH2CH2–	H × HCl	5110 (4830–5400)	314 (256–387)	>10 000	2662 (2461–2880)
17	–CH2CH2CH2O(CH2CH2O)2CH2CH2CH2–	H × HCl	1660 (1350–2050)	348 (269–450)	>10 000	312 (264–368)
18	—	H × HCl	5520 (3530–8620)	682 (610–762)	>10 000	2914 (1714–4955)
19	—	H × HCl	>100 000	12 900 (10 400–15 900)	>10 000	8646 (7293–10 250)

---

In particular, in the Boc-protected series (2–10) we observed low affinity towards hA₁AR: long or bulky linkers (7–10) were not tolerated at this receptor, while simple alkyl chains lead to Ki values in the range of 1.7–2.6 μM (i.e.7, hA₁AR Ki = 12 200 nM vs.4, hA₁AR Ki = 1770 nM). Better results were obtained at the hA_2AAR, where the affinity follows the same trend observed at the hA₁AR. Nevertheless, most of the compounds displayed affinities in the submicromolar range, with a propyl linker providing the best Ki value (4, hA_2AAR Ki = 195 nM). On the other hand, Boc-protected derivatives were quite good A₃AR antagonists (2–10, hA₃AR Ki = 32.1–856 nM). In particular, a clear hA₃AR affinity enhancement was observed comparing derivatives 3 and 4 (WS88), which bear an ethyl and a propyl chain, respectively (3, hA₃AR Ki = 553 nM vs.4, hA₃AR Ki = 32.1 nM). Elongation of the linker caused a progressive decrease of hA₃AR affinity (i.e.5, hA₃AR Ki = 84.4 nM, 7, hA₃AR Ki = 585 nM) as also observed when constrained in a piperazine ring (10, hA₃AR Ki = 737 nM). With a Ki value of 32.1 nM, compound 4 was the most potent derivative towards the hA₃AR, and it showed 55- and 6-fold selectivity against hA₁ and hA_2A ARs, respectively. Thus, it was the analogue that displayed the best balance between affinity and selectivity. The most selective compound for the hA₃AR was the piperazine derivative 10, which displayed a >136- and 29-fold higher affinity value at the hA₃AR, with respect to hA₁ and hA_2A ARs, respectively, but it showed an affinity in the high nanomolar range towards hA₃AR (Ki = 737 nM). The other compounds of this subseries displayed a common selectivity profile at the ARs that was very low against hA_2AAR (hA_2A/hA₃ = 1.1–3.7) and higher towards hA₁AR (hA₁/hA₃ = 20–55). A different behaviour was observed when the shorter hydrazine linker was present at the 5 position of PTP. In fact, first of all, with compound 2 (WS86) we gained hA₃AR affinity with respect to the ethyl compound 3 (3, hA₃AR Ki = 553 nM vs.2, hA₃AR Ki = 173 nM). Moreover, this compound exhibited one of the best selectivity profiles against both hA_2AAR (22.5-fold) and hA₁AR (45-fold).

However, when the Boc group was removed (leading to the free amino series 11–19) affinity at the hA₁AR was not particularly affected (e.g.9, hA₁AR Ki = 4070 nM vs.18, hA₁AR Ki = 5520 nM). Conversely, at the hA₃AR we observed a pronounced detrimental effect (e.g.9, hA₃AR Ki = 856 nM vs.18, hA₃AR Ki = 2914 nM) and a simultaneous increase of hA_2AAR affinity (e.g.9, hA_2AAR Ki = 2170 nM vs.18, hA_2AAR Ki = 682 nM). As a consequence, in general, free amino compounds could be considered hA_2AAR antagonists. Among these, an exception was seen with compounds 11 and 17, which showed similar Ki values towards hA_2A and hA₃ ARs (11, hA_2AAR Ki = 233 nM, hA₃AR Ki = 174 nM; 17, hA_2AAR Ki = 348 nM, hA₃AR Ki = 312 nM). Thus, they could be considered as dual hA_2A–hA₃ AR antagonists. Interestingly, it seems that for hA₃AR, with the hydrazine linker compounds (2, 11) the presence of the Boc group did not influence the affinity at this receptor (2, hA₃AR Ki = 173 nM vs.11, hA₃AR Ki = 174 nM), while it had a strong effect on the hA_2AAR (2, hA_2AAR Ki = 3900 nM vs.11, hA_2AAR Ki = 233 nM). Notably, observing data for compounds bearing oxygen-containing chains (16, 17), the longer {[(propoxy)-ethoxy]-ethoxy}-propyl linker conferred higher hA₃AR affinity than the diethoxyethyl linker (17, hA₃AR Ki = 312 nM vs.16, hA₃AR Ki = 2662 nM). A similar behaviour was observed, although to a less extent, in Boc-protected compounds (8, hA₃AR Ki = 264 nM vs.7, hA₃AR Ki = 585 nM). Concerning the structure–activity relationship (SAR) at the hA_2AAR, an affinity enhancement was observed upon chain elongation from 2 to 5 methylene units that decreased when longer or bulkier linkers were present at the 5 position. Compound 15 displayed the best hA_2AAR affinity of the series, with a Ki value of 52.6 nM and a 15- and 43-fold selectivity against hA₁ and hA₃ ARs, respectively.

Longer linkers, which should extend into the extracellular part of the receptor, gave especially low affinity compounds and variable selectivity profiles. This observation suggested that the linker had a very crucial role in the recognition process. In order to better understand this behaviour, we performed computational studies on the ligand–receptor recognition process.

Conjugation with a fluorescent moiety

In order to prove the effectiveness of our strategy, we have reported here an example of application where amino-linked compounds 11, 13–17 were conjugated with fluorescein isothiocyanate to afford fluorescent conjugates 20–25.²² Fluorescent ligands could be used to probe the ARs serving as pharmacological tools. As an alternative to radioligands, such fluorescent probes could be useful both in the drug discovery process and in physiological or pathophysiological studies at cellular and tissue levels.²⁴

Chemistry. The free amino derivatives (11, 13–17) were conjugated with fluorescein isothiocyanate (FITC) in the presence of a base, affording fluorescent thioureidic probes 20–25 (Scheme 2). Spectroscopic characterization of derivatives 20–25 showed an absorption maximum wavelength at 512–525 nm and a maximum emission wavelength at 537–541 nm. Thus, the obtained compounds showed a less than optimal Stokes shift, as could be observed in the spectrum of compound 21 (Fig. 2).

Scheme 2 Synthesis of fluorescent probes 20–25.²² Reagents and conditions: a) FITC, TEA, DMF, rt, 48 h.

Fig. 2 Superimposition of absorbance spectra of compound 13 (10⁻³ M) and of its fluorescent derivative 21 (10⁻⁴ M) and emission spectra of compound 21 (10⁻⁵ M).

Pharmacology and structure–activity relationship. As reported in Table 2, fluorescent derivatives 20–25 were tested for binding affinity at hA₁, hA_2A and hA₃ ARs. However, given the precursors' inactivity towards the hA_2BAR subtype, this receptor was not considered a target for the synthesized compounds. In addition, results at hA₁AR showed that all compounds were unable to bind the receptor at low micromolar concentrations. Instead, concerning hA_2A and hA₃ ARs a SAR could not be deduced from the data reported in Table 2. Nevertheless, compounds displayed affinities in the micromolar range that were similar towards the two receptor subtypes, and thus they could be considered as poor hA_2A/hA₃ AR antagonists. The most potent compounds of the series were those bearing the propyl (21) and the pentyl (23) linkers (21, hA_2AAR Ki = 460 nM, hA₃AR Ki = 720 nM; 23, hA_2AAR Ki = 920 nM, hA₃AR Ki = 780 nM;), while longer chains led to a severe detrimental effect (i.e.25, hA_2AAR Ki = 6010 nM, hA₃AR Ki = 7090 nM).

Table 2 Binding profile at the four hAR subtypes for compounds 20–25^a

Compd	X	hA1^b (Ki nM)	hA2A^c (Ki nM)	hA3^d (Ki nM)
a All data are from ref. 22. b Displacement of specific [^3H]-RPIA binding at hA1ARs expressed in CHO cells (n = 3–5). c Displacement of specific [^3H]-CGS21680 binding at hA2AARs expressed in HEK293 cells (n = 3–5). d Displacement of specific [^125I]I-AB-MECA binding at hA3AR expressed in CHO cells (n = 3–5). Duplicate determination. Data are expressed as geometric means with 95% confidence limits.
20	—	>10 000	2480 (1640–3320)	2390 (1070–3710)
21	–CH2CH2CH2–	>10 000	460 (340–580)	720 (580–860)
22	–CH2CH2CH2CH2–	>10 000	1750 (810–2690)	5660 (4440–6880)
23	–CH2CH2CH2CH2CH25–	9500^d	920 (700–1140)	780^d
24	–CH2CH2OCH2CH2OCH2CH2–	>10 000	>10 000	9390 (5930–12 850)
25	–CH2CH2CH2O(CH2CH2O)2CH2CH2CH2–	>10 000	6010 (5170–6850)	7090 (6350–7830)

---

The low affinity of these compounds highlights again the crucial role of the linker in the recognition process, suggesting that the PTP core alone is not sufficient to provide the desired affinity and selectivity at the hARs. The introduction between the PTP core and the linker of a substituent able to achieve a specific affinity for one of the receptors could enhance the affinity and selectivity of the conjugated compounds.

Computational studies

Molecular docking simulations were performed with the aim to rationalize the affinity data and SAR, as described in detail in the Experimental section. According to our recently published work,²⁵ all molecular docking studies have been carried out with a sodium ion and its first hydration shell in all AR structures.

As previously reported, also in this case, regardless of the nature of the receptor subtype considered, docking simulations suggested, for all ligands belonging to the PTP class, three different and alternative binding modes (BM) in their orthosteric binding sites.²⁰ With a specific reference to the position assumed by the furyl moiety, the binding modes differ for the orientation of the ring along the main axis of the binding cavity (see details in Fig. 3): in particular, BM1 and BM3 orient the furyl ring down compared to the extracellular environment (for BM1, alternative up orientation is also retrieved), while BM2 orients it towards the TM2. Interestingly, the three binding modes guarantee the stabilizing interaction with Asn (6.55) and with other hydrophobic residues, such as Phe (EL2). The relative BM sampling frequency strongly depends critically on the nature of the substituent at the 5-position of the PTP scaffold, suggesting that the specific substituents there can modulate the receptor potency and/or receptor subtype selectivity. Considering this specific PTP series, docking studies suggested that BM2 is the most frequently sampled for any receptor, even if BM3 is its most relevant alternative on the hA₃AR. In Fig. 4–6 are shown the interaction energy fingerprint (IEF) maps of both electrostatic and hydrophobic contributions, calculated for each sampled pose on all three receptor subtypes.

Fig. 3 The binding modes observed for the PTP scaffold during molecular docking simulations.

Fig. 4 A) IE map of the electrostatic interaction between each compound (y-axis) and each residue (x-axis) of hA₁AR. The strength of the interaction is represented by a colorimetric scale going from blue to red for negative to positive values. B) IE map representing the hydrophobic interaction between each compound (y-axis) and each residue (x-axis) of hA₁AR. The strength of the interaction is represented by a colorimetric scale going from white to dark green for low to high values.

Fig. 5 A) IE map of the electrostatic interaction between each compound (y-axis) and each residue (x-axis) of hA_2AAR. The strength of the interaction is represented by a colorimetric scale going from blue to red for negative to positive values. B) IE map representing the hydrophobic interaction between each compound (y-axis) and each residue (x-axis) of hA_2AAR. The strength of the interaction is represented by a colorimetric scale going from white to dark green for low to high values.

Fig. 6 A) IE map of the electrostatic interaction between each compound (y-axis) and each residue (x-axis) of hA₃AR. The strength of the interaction is represented by a colorimetric scale going from blue to red for negative to positive values. B) IE map representing the hydrophobic interaction between each compound (y-axis) and each residue (x-axis) of hA₃AR. The strength of the interaction is represented by a colorimetric scale going from white to dark green for low to high values.

Interestingly, regardless of the nature of the receptor subtype considered, almost any analyzed PTP analogue can be accommodated in the orthosteric site by the canonical binding pattern, already annotated for other AR antagonists: these include the hydrogen bonds and the aromatic hydrophobic π–π interactions with Asn (6.55) and Phe (EL2), respectively. Results can be related to the structural and sequence diversities between receptors. Indeed, favorable electrostatics are detected along the EL1 and TM2 for ligands docked to the hA_2AAR, while, for the hA₁AR, some repulsion is achieved at the TM3, i.e. non-conserved residues play a major role.

To analyze in more detail the interactive scheme of some representative ligands belonging to this class of antagonists, per-residue interaction energy profiles (electrostatic, van der Waals and hydrophobic contributions) were calculated for compounds 4 (WS88), 7 (WS91), 15 (WS98) and 23, and shown in Fig. 7–10. Conserved and non-conserved residues were selected for each receptor subtype to better appreciate their contributions in the stabilizing ligand–receptor binding state. Moreover, docking poses are also collected in Videos S1–S3 (ESI†), together with electrostatic and hydrophobic contributions of peculiar residues.

Fig. 7 Interaction energy histograms of compound 4 (WS88) for conserved and non-conserved residues of any receptor subtype.

Fig. 8 Interaction energy histograms of compound 7 (WS91) for conserved and non-conserved residues of any receptor subtype.

Fig. 9 Interaction energy histograms of compound 15 (WS98) for conserved and non-conserved residues of any receptor subtype.

Fig. 10 Interaction energy histograms of compound 23 for conserved and non-conserved residues of any receptor subtype.

The first analyzed antagonist was derivative 4, the most potent compound of the series towards the hA₃AR (Ki = 32.1 nM), despite its modest selectivity (hA₁/hA₃ = 55.3 and hA_2A/hA₃ = 6.1). Interestingly, derivative 4 preferably recognizes the orthosteric binding site of the hA₁AR using the interactive scheme BM1up, the hA_2A subtype using the interactive scheme BM3, and the hA₃ using the interactive scheme BM1down. As shown in Fig. 7, these three different binding modes are characterized by three different interaction energy profiles and the analysis shows that the most favorable interactive scheme is for subtype hA₃ engaging strong hydrophobic contacts with the conserved Phe168 (EL2), with the non-conserved Leu90 (3.32), and an electrostatically favorable interaction with Glu19 (1.39), without showing relevant repulsive van der Waals contributions. Conversely, an unfavorable van der Waals term is detected when derivative 4 is docked to both hA₁ and hA_2A ARs, mediated by Asn254 and Asn253, respectively (6.55).

The second analyzed antagonist is derivative 7, a moderately selective compound for the hA₃AR, especially when compared to the hA₁AR (hA₁/hA₃ = 135.7 and hA_2A/hA₃ = 2.1). This antagonist preferably recognizes the orthosteric binding site of the hA₁AR using the interactive scheme BM2, and the hA_2A and hA₃ using the interactive scheme BM3. Interestingly, unless for Glu172 (EL2) and Val87 (3.32), the analysis of the interaction energy profiles indicates a very modest stabilizing interaction network compared to the hA₃AR. At the hA_2AAR, the ligand interacts unfavorably with His250 (6.52). The most favorable profile is observed at the hA₃AR.

Derivative 15 is the highest affinity compound of the series at the hA_2AAR (Ki = 52.6 nM). Nevertheless, it shows affinity constants in high nanomolar (Ki = 778 nM) and micromolar ranges (Ki = 2246 nM), for the hA₁ and hA₃ ARs, respectively. Derivative 15 preferably recognizes the orthosteric binding site of the hA₁ and hA_2AARs using the interactive scheme BM2, but interactive scheme BM3 is used at the hA₃AR. For this ligand, at the hA₁AR, several stabilizing interactions were detectable, in particular with Phe171 (EL2), Glu172 (EL2) and Asn254 (6.55), as shown in Fig. 9. This stabilizing network of interactions is also present at the hA_2A AR, even if the intensity of each interaction energy term is lower than that measured at the hA₁AR. To appropriately interpret the lower affinity determined experimentally against the hA₃ subtype, it is possible to invoke the crucial repulsive interactions detected for Val169 (EL2) and Phe168 (EL2), together with highly unfavorable van der Waals contributions by Asn250 (6.55). Involvement of these residues is likely to be critical for affinity.

The final analyzed antagonist is derivative 23, the most representative compound among the fluorescent ligands of the series, and the most selective towards hA₃AR. As expected, BM2 is the only interactive scheme sampled for each receptor subtypes, allowing a better comparison (Fig. 10). As already described, this specific binding mode allows us to orient the substituent that carries the fluorescent probe towards the extracellular environment. Unfortunately, in this case, considering the relevant volumetric properties of these ligands and the fact that their receptor recognition involves not only the orthosteric binding sites but also relevant portions of the extracellular loops (in particular, the EL2), the analysis of the interaction energy profiles is not very helpful to interpret the experimental selectivity profile. However, it is emphasized that all the most crucial interactions with the conserved Phe (EL2) and Asn (6.55) are preserved. Moreover, the amino acid sequences and the architectures of EL2 play an important role in controlling both the selectivity profiles and ligand affinities of the fluorescent analogues against all receptor subtypes.

Finally, to support or negate the hypothesis suggested by docking studies concerning the presence of different and alternative ligand binding modes (BM) inside the receptor orthosteric binding site, we decided to carry out a supervised molecular dynamics (SuMD) study with the aim to explore the possible ligand–receptor recognition pathways (from an unbound state to the orthosteric bound state) and to verify if the final states reached by supervised trajectories converge to the most populated poses obtained by molecular docking. For this specific scope, we decided to use the smaller PTP analogue of this series, derivative 2 (WS86), which showed a moderate affinity against the hA₃AR (Ki = 173 nM) and a reasonable selective profile (hA₁/hA₃ = 45.2 and hA_2A/hA₃ = 22.5). Docking simulations suggested that derivative 2 preferably recognizes the orthosteric binding site of the hA₁AR using the interactive scheme BM1up; the hA_2AAR used the interactive scheme BM3, whereas the hA₃ used two distinct interactive schemes BM1down and BM2. As already reported for all previously analyzed antagonists, these four different binding modes are characterized by four different interaction energy profiles and the analysis shows that the most favorable interactive scheme for subtype hA₃AR is the one in which the ligand preferably adopts the interactive scheme BM1down, establishing the already described stabilizing interaction with the conserved Phe168 (EL2) and Asn250 (6.55) (Fig. 11).

Fig. 11 Interaction energy histograms of compound 2 (WS86) for conserved and non-conserved residues of any receptor subtype. BM1down and BM2 are reported for the hA₃AR subtype.

Considering the appreciable hA₃AR affinity, all SuMD simulations have been performed using this specific receptor subtype. Five independent supervised simulations have been carried out starting from an unbound state in which the ligand has been randomly positioned 40 Å far away from the center of the orthosteric hA₃AR binding cavity.

As expected, and already reported in previous studies,²⁶ each SuMD trajectory describes a specific ligand–receptor recognition pathway (see Fig. 12, Panel A) and allows the ligand to sit in different regions of the orthosteric binding site (bound state) and adopt different binding poses (see Fig. 12, Panel B). Even if the different bound states obtained by SuMD simulations are not perfectly superimposable to those suggested by docking experiments, we emphasize that the possibility to have a multiple binding mode for a single ligand is clearly shown in the results of the SuMD trajectory analysis. Particularly interesting is the result extrapolated from Replica 5, in which the observed bound state is close to the one described as BM2, as shown in Video S4.† As previously reported, also for derivative 2, the different SuMD trajectories described the presence of several meta-binding sites (energetically stable binding sites spatially distinct from the orthosteric site) with an energy profile (Fig. 12, Panel A) that smoothly decreases from an average value 〈E₁〉 of −30 kcal mol⁻¹ (for distances >10 Å) to an average value 〈E₂〉 of −50 kcal mol⁻¹ (〈E₂〉 ≈ 2 〈E₁〉), in a distance window between 2 and 10 Å. This behavior of the interaction energy profile is typical of a ligand–receptor recognition process characterized by a reasonably high k_on value.

Fig. 12 A) Total compound 2 (WS86)-protein interaction energy vs. distance from the binding site, for each SuMD simulation, represented by different colors. Average energies are reported for particular distance ranges. B) Visual indicative inspection of final states as vdW surfaces within the binding site. Peculiar regions of the transmembrane receptor are reported.

As already anticipated, different bound states obtained by SuMD simulations are not perfectly superimposable to the poses suggested by docking experiments, even if in Replica 5 the SuMD bound state is close to the one described as BM2. Starting from the Replica 5 final bound state, a non-biased molecular dynamics simulation has been performed to analyse the time-evolution (stability) of this specific bound state. Surprisingly, after 250 ns of simulation, the ligand flips its original position assuming a new configuration very similar to the BM3 detected in docking simulations, as shown in Fig. 13, and more energetically favored. This evidence indicates that the accommodation of the furyl ring down, compared to the extracellular side, is the most favored one.

Fig. 13 Compound 2 (WS86)-protein interaction energy vs. simulation time of the SuMD complex (binding mode 2). Total energy is decomposed in electrostatic (Elec, blue) and van der Waals (vdW, red) energy. Chemical structures of BM2 and BM3 are also reported.

The analysis of both supervised and non-biased MD trajectories showed that the adaptive positioning of the ligand inside the orthosteric binding site is strictly related to the simultaneous conformational variation of the side chains of the two amino acids Phe168 (EL2) and Trp243 (6.48) or, in other words: the final ligand pose depends on the reshaping of the binding cavity (induce fit) caused by the side chains of the two amino acids Phe168 (EL2) and Trp243 (6.48). Moreover, considering that all SuMD simulations have been carried out with a sodium ion and its first hydration shell in the hA₃AR structure, it is interesting to note that both the sodium ion and its coordinated water molecules are also involved in this transition (Video S4†).

In conclusion, the combination of molecular docking and dynamics (supervised and non-biased) simulations can be considered an interesting computational strategy to improve the descriptive capacity of the interaction modality of this new series of AR antagonists. In particular, SuMD can facilitate the interpretation at a molecular level of the possible ligand–receptor recognition pathways bridging the stability of the canonical orthosteric binding (thermodynamic binding propensity) with the ease of accessing the orthostatic site (kinetic binding propensity).

Experimental

Chemistry

Synthesis of compound 26 is reported in the literature.²⁰ Mono-Boc-protected-diamines were synthesized following the procedure reported by Onajole et al.²⁷

General chemistry. Reactions were routinely monitored by thin-layer chromatography (TLC) on silica gel (Macherey-Nagel, SIL G UV254). Flash chromatography was performed using Macherey-Nagel, silica 60, 230–400 mesh. Light petroleum ether refers to the fractions boiling at 40–60 °C. Melting points were determined on a Buchi-Tottoli instrument and are uncorrected. ¹H-NMR were determined in CDCl₃, DMSO-d₆ or D₂O solutions with a Varian Gemini 200, Varian 400 or a Varian 500 spectrometer of the University of Trieste, peaks positions are given in parts per million (δ) downfield from tetramethylsilane as an internal standard, and J values are given in Hz. MS-ESI analysis was performed using an ESI Bruker 4000 Esquire spectrometer and compounds were dissolved in methanol, accurate mass spectra were recorded on a micrOTOF-Q-Bruker and compounds were dissolved in methanol. Absorption experiments were performed on a Varian Cary 5000 spectrophotometer. Fluorescence measurements were recorded on a Varian Cary Eclipse fluorescence spectrophotometer, with an excitation filter at 361 nm and an emission filter at 390–600 nm.

General procedure for synthesis of 5-substituted-2-furan-2-yl-8-methyl-8H-pyrazolo[4,3-e][1,2,4]triazolo[1,5-c]pyrimidines 2–10. 5-Chloro-derivative 26 (100 mg, 0.364 mmol) was dissolved in 5 mL of ethanol in a tube. Triethylamine (51 μL, 0.364 mmol) and the appropriate amine (1.092 mmol) were added to the mixture, and the tube was sealed and heated at 110 °C for 2 h. The reaction was monitored by TLC (ethyl acetate 9 : methanol 1) to assess the disappearance of the starting material, then the solvent was removed under reduced pressure and the residue was purified by flash chromatography (ethyl acetate 9.5 : MeOH 0.5) affording the desired compound.
N′-(2-Furan-2-yl-8-methyl-8H-pyrazolo[4,3-e][1,2,4]triazolo[1,5-c]pyrimidin-5-yl)-hydrazine carboxylic acid tert-butyl ester (2). Yield 90%; white solid; mp 160 °C; ¹H-NMR (200 MHz, CDCl₃) δ: 1.49 (9H, s), 4.13 (3H, s), 6.59 (1H, dd, J = 2, J = 4), 6.81 (1H, bs), 7.21 (1H, d, J = 4), 7.62 (1H, d, J = 2), 7.90 (1H, bs), 8.16 (1H, s). HRMS (ESI-QTOF): m/z [M + Na]⁺ calcd for C₁₆H₁₈N₈O₃: 393.1394, found: 393.1398.
[2-(2-Furan-2-yl-8-methyl-8H-pyrazolo[4,3-e][1,2,4]triazolo[1,5-c]pyrimidin-5-ylamino)-ethyl]-carbamic acid tert-butyl ester (3). Yield 85%; white solid; mp 217 °C; ¹H-NMR (200 MHz, CDCl₃) δ: 1.43 (9H, s), 3.50–3.55 (2H, m), 3.76–3 (2H, m), 4.11 (3H, s), 4.96 (1H, bs), 6.58–6.65 (2H, m), 7.20 (1H, d, J = 4), 7.62 (1H, d, J = 2), 8.10 (1H, s). HRMS (ESI-QTOF): m/z [M + Na]⁺ calcd for C₁₈H₂₂N₈O₃: 421.1707, found: 421.1709.
[3-(2-Furan-2-yl-8-methyl-8H-pyrazolo[4,3-e][1,2,4]triazolo[1,5-c]pyrimidin-5-ylamino) propyl]-carbamic acid tert-butyl ester (4). Yield 66%; white solid; mp 217 °C; ¹H-NMR (200 MHz, CDCl₃) δ: 1.46 (9H, s), 1.93 (2H, quint, J = 6), 3.27 (2H, bs), 3.75 (2H, q, J = 6), 4.11 (3H, s), 5.11 (1H, bs), 6.53–6.59 (2H, m), 7.21 (1H, d, J = 4), 7.62 (1H, d, J = 2), 8.08 (1H, s). ¹³C NMR (101 MHz, CDCl₃) δ 192.40, 156.64, 155.89, 147.23, 145.65, 144.67, 144.48, 143.07, 126.90, 114.68, 112.46, 109.11, 79.71, 77.48, 77.16, 76.84, 40.73, 38.80, 37.84, 29.99, 28.55. HRMS (ESI-QTOF): m/z [M + Na]⁺ calcd for C₁₉H₂₄N₈O₃: 435.1864, found: 435.1862.
[4-(2-Furan-2-yl-8-methyl-8H-pyrazolo[4,3-e][1,2,4]triazolo[1,5-c]pyrimidin-5-ylamino)-butyl]-carbamic acid tert-butyl ester (5). Yield 79%; white solid; mp 167 °C; ¹H-NMR (200 MHz, CDCl₃) δ: 1.44 (9H, s), 1.60–1.84 (4H, m), 3.14–3.26 (2H, m), 3.48–3.80 (2H, m), 4.10 (3H, s), 4.59 (1H, bs), 6.31 (1H, bs), 6.59 (1H, dd, J = 2, J = 4), 7.21 (1H, d, J = 4), 7.62 (1H, d, J = 2), 8.08 (1H, s). ES-MS (methanol): m/z 449.47 [M + Na]⁺. HRMS (ESI-QTOF): m/z [M + Na]⁺ calcd for C₂₀H₂₆N₈O₃: 449.2020, found: 449.2021.
[5-(2-Furan-2-yl-8-methyl-8H-pyrazolo[4,3-e][1,2,4]triazolo[1,5-c]pyrimidin-5-ylamino)-pentyl]-carbamic acid tert-butyl ester (6). Yield 82%; white solid; mp 119 °C; ¹H-NMR (200 MHz, CDCl₃) δ: 1.43–1.82 (15H, m), 3.08–3.18 (2H, m), 3.69 (2H, dd, J = 7, J = 14), 4.11 (3H, s), 4.54 (1H, bs), 6.27 (1H, bs), 6.60 (1H, dd, J = 2, J = 4), 7.22 (1H, d, J = 4), 7.63 (1H, d, J = 2), 8.09 (1H, s). ES-MS (methanol): m/z 463.2 [M + Na]⁺. HRMS (ESI-QTOF): m/z [M + Na]⁺ calcd for C₂₁H₂₈N₈O₃: 463.2177, found: 463.2179.
(2-{2-[2-(2-Furan-2-yl-8-methyl-8H-pyrazolo[4,3-e][1,2,4]triazolo[1,5-c]pyrimidin-5-ylamino)-ethoxy]-ethoxy}-ethyl)-carbamic acid tert-butyl ester (7). Yield 55%; sticky foam; ¹H-NMR (200 MHz, CDCl₃) δ: 1.41 (9H, s), 3.36 (2H, bs), 3.56–3.66 (6H, m), 3.80–3.92 (4H, m), 4.11 (3H, s), 5.14 (1H, bs), 6.59 (1H, dd, J = 2, J = 4), 6.62 (1H, bs), 7.23 (1H, d, J = 4), 7.62 (1H, d, J = 2), 8.11 (1H, s). HRMS (ESI-QTOF): m/z [M + Na]⁺ calcd for C₂₂H₃₀N₈O₅: 509.2231, found: 509.2233.
[3-(2-{2-[3-(2-Furan-2-yl-8-methyl-8H-pyrazolo[4,3-e][1,2,4]triazolo[1,5-c]pyrimidin-5-ylamino)-propoxy]-ethoxy}-ethoxy)-propyl]-carbamic acid tert-butyl ester (8). Yield 67%; pale pink solid; mp 69 °C; ¹H-NMR (200 MHz, CDCl₃) δ: 1.44 (9H, s), 1.68–1.82 (2H, m), 1.96–2.15 (2H, m), 3.24 (2H, bs), 3.51–3.86 (14H, m), 4.12 (3H, s), 5.00 (1H, bs), 6.60 (1H, dd, J = 2, J = 4), 6.84 (1H, bs), 7.22 (1H, d, J = 4), 7.62 (1H, d, J = 2), 8.08 (1H, s). HRMS (ESI-QTOF): m/z [M + Na]⁺ calcd for C₂₆H₃₈N₈O₆: 581.2807, found: 581.2805.
{4-[(2-Furan-2-yl-8-methyl-8H-pyrazolo[4,3-e][1,2,4]triazolo[1,5-c]pyrimidin-5-ylamino)-methyl]-cyclohexylmethyl}-carbamic acid tert-butyl ester (9). Yield 79%; pale yellow solid; mp 65 °C; ¹H-NMR (200 MHz, CDCl₃) δ: 1.43–1.80 (19H, m), 2.95–3.08 (4H, m), 4.10 (3H, s), 4.56 (1H, bs), 6.35 (1H, bs), 6.59 (1H, dd, J = 2, J = 4), 7.20 (1H, d, J = 4), 7.62 (1H, d, J = 2), 8.07 (1H, s). C₂₄H₃₂N₈O₃. M.W.: 480.56. HRMS (ESI-QTOF): m/z [M + Na]⁺ calcd for C₂₄H₃₂N₈O₃: 503.2490, found: 503.2494.
4-(2-Furan-2-yl-8-methyl-8H-pyrazolo[4,3-e][1,2,4]triazolo[1,5-c]pyrimidin-5-yl)-piperazine-1-carboxylic acid tert-butyl ester (10). Yield 63%; pale yellow solid; mp 112 °C; ¹H-NMR (200 MHz, CDCl₃) δ: 1.52 (9H, s), 3.68–3.70 (4H, m), 3.85–3.99 (4H, m), 4.15 (3H, s), 6.59 (1H, dd, J = 2, J = 4), 7.23 (1H, d, J = 4), 7.64 (1H, d, J = 2), 8.19 (1H, s). HRMS (ESI-QTOF): m/z [M + Na]⁺ calcd for C₂₀H₂₄N₈O₃: 447.1864, found: 447.1866.

General procedure for the synthesis of amino derivatives 11–19. tert-Butyloxycarbonyl derivatives 2–10 (100 mg) were dissolved in a saturated solution of hydrogen chloride in ethyl acetate (5 mL) and stirred at room temperature for 1 hour. The reaction was monitored by TLC (ethyl acetate 9 : methanol 1, the desired compound doesn't move). The solvent was then removed under reduced pressure and the desired compounds (11–19) were filtered off.
N′-(2-Furan-2-yl-8-methyl-8H-pyrazolo[4,3-e][1,2,4]triazolo[1,5-c]pyrimidin-5-yl)-hydrazinium; chloride (11). Yield 65%; yellow solid; mp 192 °C; ¹H-NMR (200 MHz, CDCl₃–D₂O) δ: 4.13 (3H, s), 4.77–4.83 (2H, bm), 6.59 (1H, dd, J = 2, J = 4), 7.21 (1H, d, J = 4), 7.63 (1H, d, J = 2), 8.13 (1H, s). ¹³C NMR (101 MHz, DMSO-d₆) δ 155.71, 153.57, 149.37, 145.53, 144.97, 143.31, 126.62, 112.81, 112.28, 98.34, 40.22. HRMS (ESI-QTOF): m/z [M + H]⁺ calcd for C₁₁H₁₀N₈O: 271.1050, found: 271.1045.
2-(2-Furan-2-yl-8-methyl-8H-pyrazolo[4,3-e][1,2,4]triazolo[1,5-c]pyrimidin-5-ylamino)-ethyl-ammonium; chloride (12). Yield 62%; pale yellow solid; mp 212 °C; ¹H-NMR (200 MHz, D₂O) δ: 3.31–3.40 (2H, m), 3.79–3.83 (2H, m), 3.92 (3H, s), 6.56 (1H, bs), 6.86 (1H, bs), 7.59 (1H, bs), 7.97 (1H, s). HRMS (ESI-QTOF): m/z [M + H]⁺ calcd for C₁₃H₁₄N₈O: 299.1363, found: 299.1367.
3-(2-Furan-2-yl-8-methyl-8H-pyrazolo[4,3-e][1,2,4]triazolo[1,5-c]pyrimidin-5-ylamino)-propyl-ammonium; chloride (13). Yield 75%; pale yellow solid; mp 198 °C; ¹H-NMR (200 MHz, D₂O) δ: 2.06 (2H, m), 3.12 (2H, t, J = 7), 3.56 (2H, t, J = 7), 3.93 (3H, s), 6.57 (1H, dd, J = 2, J = 4), 6.94 (1H, d, J = 4), 7.59 (1H, d, J = 2), 8.00 (1H, s). HRMS (ESI-QTOF): m/z [M + H]⁺ calcd for C₁₄H₁₆N₈O: 313.1520, found: 313.1521.
4-(2-Furan-2-yl-8-methyl-8H-pyrazolo[4,3-e][1,2,4]triazolo[1,5-c]pyrimidin-5-ylamino)-butyl-ammonium; chloride (14). Yield 69%; pale brown solid; mp 224 °C; ¹H-NMR (200 MHz, D₂O) δ: 1.72–1.88 (4H, m), 3.05 (2H, t, J = 7), 3.43 (2H, t, J = 7), 3.92 (3H, s), 6.59 (1H, dd, J = 2, J = 4), 6.92 (1H, d, J = 4), 7.62 (1H, d, J = 2), 8.01 (1H, s). HRMS (ESI-QTOF): m/z [M + H]⁺ calcd for C₁₅H₁₈N₈O: 327.1676, found: 327.1674.
5-(2-Furan-2-yl-8-methyl-8H-pyrazolo[4,3-e][1,2,4]triazolo[1,5-c]pyrimidin-5-ylamino)-pentyl-ammonium; chloride (15). Yield 73%; pale yellow solid; mp 126 °C; ¹H-NMR (200 MHz, D₂O) δ: 1.41–1.56 (2H, m), 1.70–1.74 (4H, m), 3.03 (2H, t, J = 7), 3.37 (2H, t, J = 7), 3.91 (3H, s), 6.56 (1H, dd, J = 2, J = 4), 6.88 (1H, d, J = 4), 7.58 (1H, d, J = 2), 7.95 (1H, s). HRMS (ESI-QTOF): m/z [M + H]⁺ calcd for C₁₆H₂₀N₈O: 341.1833, found: 341.1833.
2-{2-[2-(2-Furan-2-yl-8-methyl-8H-pyrazolo[4,3-e][1,2,4]triazolo[1,5-c]pyrimidin-5-ylamino)-ethoxy]-ethoxy}-ethyl-ammonium; chloride (16). Yield 61%; sticky foam; ¹H-NMR (200 MHz, D₂O) δ: 3.16 (2H, bs), 3.74–3.87 (10H, m), 3.96 (3H, s), 6.61 (1H, dd, J = 2, J = 4), 6.96 (1H, d, J = 4), 7.64 (1H, d, J = 2), 8.07 (1H, s). HRMS (ESI-QTOF): m/z [M + H]⁺ calcd for C₁₇H₂₂N₈O₃: 387.1888, found: 387.1885.
3-(2-{2-[3-(2-Furan-2-yl-8-methyl-8H-pyrazolo[4,3-e][1,2,4]triazolo[1,5-c]pyrimidin-5-ylamino)-propoxy]-ethoxy}-ethoxy)-propyl-ammonium; chloride (17). Yield 56%; sticky foam; ¹H-NMR (200 MHz, D₂O) δ: 1.87–1.99 (4H, m), 3.05 (2H, t, J = 7), 3.55–3.75 (14H, m), 3.97 (3H, s), 6.62 (1H, dd, J = 2, J = 4), 6.99 (1H, d, J = 4), 7.65 (1H, d, J = 2), 8.08 (1H, s). HRMS (ESI-QTOF): m/z [M + H]⁺ calcd for C₂₁H₃₀N₈O₄: 459.2463, found: 459.2464.
N-((4-(Aminomethyl)cyclohexyl)methyl)-2-(furan-2-yl)-8-methyl-8H-pyrazolo[4,3-e][1,2,4]triazolo[1,5-c]pyrimidin-5-ammonium; chloride (18). Yield 66%; sticky foam; ¹H-NMR (200 MHz, D₂O) δ: 1.78–1.82 (10H, m), 2.84–3.18 (4H, m), 3.78 (3H, s), 6.34 (1H dd, J = 2, J = 4), 6.58 (1H, d, J = 4), 7.36 (1H, d, J = 2), 7.73 (1H, s). HRMS (ESI-QTOF): m/z [M + H]⁺ calcd for C₁₉H₂₄N₈O: 381.2146, found: 381.2148.
4-(2-Furan-2-yl-8-methyl-8H-pyrazolo[4,3-e][1,2,4]triazolo[1,5-c]pyrimidin-5-yl)-piperazin-1-ium; chloride (19). Yield 67%; white solid; mp > 300 °C; ¹H-NMR (200 MHz, D₂O) δ: 3.46–3.54 (4H, m), 3.98–4.14 (7H, m), 6.63 (1H, dd, J = 2, J = 4), 7.06 (1H, d, J = 4), 7.68 (1H, d, J = 2), 8.21 (1H, s). ¹³C NMR (101 MHz, DMSO-d₆) δ 154.35, 154.14, 150.41, 145.74, 145.27, 145.22, 126.02, 112.37, 112.16, 98.86, 44.59, 41.98, 40.09. HRMS (ESI-QTOF): m/z [M + H]⁺ calcd for C₁₅H₁₆N₈O: 325.1520, found: 325.1521.

General procedure for the preparation of fluorescein isothiocyanate-conjugated pyrazolo[4,3-e][1,2,4]triazolo[1,5-c]pyrimidine derivatives 20–25. 0.214 mmol of compounds 11, 13–17 were dissolved in 5 mL of dry DMF, then 30 μL of TEA (0.214 mmol) were added and the mixture was stirred for a few minutes. At the reaction were added 83 mg of FITC (0.214 mmol) and it was stirred for 48 h. The reaction was monitored with TLC using 9 ethyl acetate : 1 methanol as an eluent. The solvent was removed under reduced pressure and the residue was purified by preparative TLC in the same solvent dissolving the residue in a little amount of methanol and the final compounds (20–25) were obtained as orange solids.
5-(2-(2-(Furan-2-yl)-8-methyl-8H-pyrazolo[4,3-e][1,2,4]triazolo[1,5-c]pyrimidin-5-yl)hydrazinecarbothioamido)-2-(6-hydroxy-3-oxo-3H-xanthen-9-yl)benzoic acid (20). Yield 63%, orange solid, mp 180 °C (methanol); ¹H NMR (500 MHz, DMSO-d₆) δ: 3.78 (3H, s), 6.44 (4H, bs), 6.57 (1H, s), 6.75 (2H, bs), 6.85 (1H, s), 7.32 (1H, d, J = 7.3), 7.72 (2H, m), 8.05 (2H, m), 8.11 (1H, s), 10.83 (1H, bs), 12.45 (1H, bs). UV: λ_max 525. Fluorescence: λ_em 541. HRMS (ESI-QTOF): m/z [M + Na]⁺ calcd for C₃₂H₂₁N₉O₆S: 682.1228, found: 682.1224.
5-(3-(3-((2-(Furan-2-yl)-8-methyl-8H-pyrazolo[4,3-e][1,2,4]triazolo[1,5-c]pyrimidin-5-yl)amino)propyl)thioureido)-2-(6-hydroxy-3-oxo-3H-xanthen-9-yl)benzoic acid (21). Yield 54%, orange solid, mp > 250 °C (methanol); ¹H NMR (500 MHz, DMSO-d₆) δ: 1.97 (m, 2H), 3.51–3.70 (4H, m), 3.99 (3H, s), 6.30 (4H, m), 6.68 (2H, d, J = 8.3), 6.71 (1H, s), 7.10 (1H, d, J = 7.6), 7.19 (1H, s), 7.67 (1H, s), 7.92 (1H, s), 8.06 (1H, s), 8.18 (1H, s), 8.45–8.54 (2H, m), 10.17 (1H, bs). UV: λ_max 518. Fluorescence: λ_em 541. HRMS (ESI-QTOF): m/z [M + Na]⁺ calcd for C₃₅H₂₇N₉O₆S: 724.1697, found: 724.1699.
5-(3-(4-((2-(Furan-2-yl)-8-methyl-8H-pyrazolo[4,3-e][1,2,4]triazolo[1,5-c]pyrimidin-5-yl)amino)butyl)thioureido)-2-(6-hydroxy-3-oxo-3H-xanthen-9-yl)benzoic acid (22). Yield 56%, orange solid, mp > 250 °C (methanol); ¹H NMR (500 MHz, DMSO-d₆) δ: 1.67 (m, 2H), 1.70–1,77 (2H, m), 3.48–3.66 (4H, m), 4.00 (3H, s), 6.53 (2H, d, J = 8.3), 6.57–6.66 (4H, m), 6.72 (1H, s), 7.13 (1H, d, J = 8.3), 7.20 (1H, d, J = 3.2), 7.77 (1H, d, J = 8.2), 7.92 (1H, s), 8.02 (1H, t, J = 5.6), 8.27 (1H, s), 8.55 (1H, s), 8.59 (1H, s), 10.36 (1H, bs). UV: λ_max 521. Fluorescence: λ_em 537. HRMS (ESI-QTOF): m/z [M + Na]⁺ calcd for C₃₆H₂₉N₉O₆S: 738.1854, found: 738.1854.
5-(3-(5-((2-(Furan-2-yl)-8-methyl-8H-pyrazolo[4,3-e][1,2,4]triazolo[1,5-c]pyrimidin-5-yl)amino)pentyl)thioureido)-2-(6-hydroxy-3-oxo-3H-xanthen-9-yl)benzoic acid (23). Yield 49%, orange solid, mp 220 °C (methanol); ¹H NMR (500 MHz, DMSO-d₆) δ: 1.37–1.47 (2H, m), 1.65–1.76 (2H, m), 3.49 (4H, m), 4.00 (3H, s), 6.37–6.54 (4H, m), 6.64 (2H, d, J = 9), 6.71 (1H, dd, J = 4, J = 2), 7.06 (1H, d, J = 8), 7.19 (1H, d, J = 4), 7.75 (1H, d, J = 8), 7.92 (1H, s), 7.98 (1H, t, J = 7), 8.24 (1H, s), 8.45–8.68 (2H, m), 10.50 (1H, bs). UV: λ_max 522. Fluorescence: λ_em 541. HRMS (ESI-QTOF): m/z [M + Na]⁺ calcd for C₃₇H₃₁N₉O₆S: 752.2010, found: 752.2011.
5-(3-(2-(2-(2-((2-(Furan-2-yl)-8-methyl-8H-pyrazolo[4,3-e][1,2,4]triazolo[1,5-c]pyrimidin-5-yl)amino)ethoxy)ethoxy)ethyl)thioureido)-2-(6-hydroxy-3-oxo-3H-xanthen-9-yl)benzoic acid (24). Yield 43%, orange solid, mp 216 °C (methanol); ¹H NMR (500 MHz, DMSO-d₆) δ: 3.30 (m, 2H), 3.60 (6H, m), 3.67 (4H, m), 4.00 (3H, s), 6.37–6.54 (4H, m), 6.63 (2H, d, J = 7.7), 6.71 (1H, s), 7.12 (1H, d, J = 8.1), 7.19 (1H, d, J = 2.9), 7.72 (1H, s), 7.83 (1H, s), 7.92 (1H, s), 8.17 (1H, s), 8.25 (1H, s), 8.55 (1H, s), 10.11 (1H, bs). UV: λ_max 522. Fluorescence: λ_em 538. HRMS (ESI-QTOF): m/z [M + Na]⁺ calcd for C₃₈H₃₃N₉O₈S: 798.2065, found: 798.2068.
5-(3-(3-(2-(2-(3-((2-(Furan-2-yl)-8-methyl-8H-pyrazolo[4,3-e][1,2,4]triazolo[1,5-c]pyrimidin-5-yl)amino)propoxy)ethoxy)ethoxy)propyl)thioureido)-2-(6-hydroxy-3-oxo-3H-xanthen-9-yl)benzoic acid (25). Yield 53%, orange solid, mp 245 °C (methanol); ¹H NMR (500 MHz, DMSO-d₆) δ: 1.74 (2H, m), 1.92 (2H, m), 2.89–3.30 (m, 16H), 4.00 (3H, s), 6.27 (4H, m), 6.57–6.88 (3H, m), 7.04 (1H, s), 7.19 (1H, s), 7.68 (1H, s), 7.92 (1H, s), 8.14 (1H, s), 8.50 (1H, s), 10.31 (1H, bs). UV: λ_max 512. Fluorescence: λ_em 540. HRMS (ESI-QTOF): m/z [M + Na]⁺ calcd for C₄₂H₄₁N₉O₉S: 870.2640, found: 870.2643.

Pharmacology

All pharmacological methods followed the procedures as described earlier.^22,28–32 In particular, two slight different protocols were used for compounds 2–19 (ref. 28–31) and for compounds 20–25.^22,32 For radioligand binding at the hA₁, hA_2A and hA₃ ARs, [³H]CCPA or [³H]R-PIA, [³H]NECA or [³H]CGS21680 and [³H]HEMADO or [¹²⁵I]I-AB-MECA were used, respectively. Radioligand binding at hA_2BARs is problematic as no high-affinity radioligand agonist is commercially available for this subtype.³³ Therefore, inhibition of NECA-stimulated adenylyl cyclase activity was determined as a measurement of potency of compounds, following a well-known procedure.²⁸

Computational studies

The MOE suite³⁴ was exploited in order to perform general molecular modeling operations. All preliminary computations were carried out on a 12 CPU (Intel® Xeon® CPU E5-1650 3.80 GHz) Linux workstation. SuMD and MD simulations were performed exploiting the ACEMD engine³⁵ on NVIDIA drivers: NVIDIA GTX 980Ti and NVIDIA GTX 980. The CHARMM36/CHARMM³⁶ general force field (CGenFF) combination was adopted. The GOLD³⁷ suite and goldscore scoring function were used as the docking program. 3D structures of the ligands were built by the MOE-builder tool. Ionization states were predicted using the MOE-protonate 3D tool. Tautomers and atom hybridization were checked. Structures were minimized by the MMFF94x until the root mean square (RMS) gradient fell below 0.05 kcal mol⁻¹ A⁻¹. For each ligand, 20 docking simulation runs were performed on each receptor subtype, searching on a sphere of 20 Å radius, centered on backbone nitrogen of the conserved Asn (6.55). Ligand and protein partial charges were calculated by means of the MMFF94 and AMBER14:EHT³⁸ force fields, respectively.

The hA_2A and hA₁ AR coordinates were retrieved from the Protein Data Bank³⁹ using the crystal structures coded as 3PWH⁴⁰ and 5UEN,⁴¹ respectively. Given the lack of any crystallographic structures for hA₃AR, a homology model was used. The model was built using 5UEN as a template, as reported in our recently submitted work,⁴² after a docking/structural based model assessment. According to our recently published work,²⁵ all molecular docking studies have been carried out with a sodium ion and its first hydration shell in all adenosine receptor structures.

The Ballesteros–Weinstein⁴³ numbering system is used sometimes to indicate conserved residues. IE electrostatic and van der Waals energy contributions to the binding energy were calculated by MOE, together with per residue electrostatic and hydrophobic interactions. Per-residue information was reported in the so-called “Interaction Energy Fingerprints”: they consist of heat maps reporting the strength of the interaction of each residue (x-axis) and each ligand (y-axis) according to a colorimetric scale going from blue to red for negative to positive values in the case of electrostatic contributions and from white to dark green for low to high values for hydrophobic contributions. Interactions of most relevant residues were also reported by histograms, whose height is proportional to the strength of the interaction. Plots were generated using Gnuplot 4.6.⁴⁴ 2D chemical structures in figures were drawn with Marvin Sketch.⁴⁵ Molecular graphics were performed with the UCSF Chimera package,⁴⁶ except Fig. 12B, which was obtained using the VMD program (version 1.9.3).⁴⁷ The in-house MMsDocking video maker tool was exploited to produce videos showing the docking poses, per residue hydrophobic and electrostatic contributions for selected residues, experimental binding data, and scoring values. Representations of docking poses were produced using the UCSF Chimera package, 2D depictions were constructed by the cheminformatics toolkit RDKit,⁴⁸ and the heat maps were obtained by Gnuplot 4.6; in the end, videos were mounted using MEncoder.

Ligand 3 force field parameters for MD simulations were initially retrieved from the Paramchem web service and then deeply optimized in concordance with CGenFF, at the MP2/6-31G* level of the theory⁴⁹ by using Gaussian 09 (ref. 50) and RESP partial charges. Systems were embedded in a 1-palmitoyl-2oleyl-sn-glycerol-3-phospho-choline (POPC) lipid bilayer, according to the pre-orientation provided by the Orientations Proteins in Membrane (OPM) database⁵¹ and by using the VMD membrane builder plugin. Lipids within 0.4 Å from the protein were removed and TIP3P⁵² model water molecules were added to solvate the system by means of Solvate1.0.⁵³ System charge neutrality was reached by adding Na⁺/Cl⁻ counterions to a final concentration of 0.154 M. Equilibration was performed through a three-step procedure. In the first one, 1500 conjugate-gradient minimization steps were applied in order to reduce the clashes between protein and lipids. Then, a 5 ns long MD simulation was performed in the NPT ensemble, with a positional constraint of 1 kcal mol⁻¹ Å⁻² on ligand, protein, and lipid phosphorus atoms. During the second stage, 10 ns MD simulation in the NPT ensemble was performed constraining all the protein and ligand atoms but leaving the POPC residues free to diffuse in the bilayer. In the last equilibration stage, positional constraints were applied only to the ligand and protein backbone alpha carbons for further 5 ns MD simulation. All the MD simulations were performed using: (1) an integration time step of 2 fs; (2) the Berendsen barostat⁵⁴ in order to maintain the system pressure at 1 atm; (3) the Langevin thermostat⁵⁵ to maintain the temperature at 310 K with a low dumping of 1 ps⁻¹; (4) the M-SHAKE algorithm⁵⁶ to constrain the bond lengths involving hydrogen atoms; and (5) a long-range cutoff of 10 Å. According to the SuMD approach,^57,58 the timescale needed to reproduce complete intermolecular complex formations is in the range of nanoseconds, instead of hundreds of nanoseconds or microseconds usually necessary with unsupervised MD. Sampling is gained by applying a tabu-like algorithm to monitor the distance between the centers of mass of the ligand and the binding site, during short unbiased MD simulations. SuMD considers the ligand atoms and the atoms of user-defined protein residues to monitor the distance between the centers of mass of the binder and the binding site. A series of 600 ps unbiased MD simulations are performed and after each simulation, the distance points collected at regular time intervals are fitted into a linear function. If the resulting slope is negative the next simulation step starts from the last set of coordinates and velocities produced, otherwise the simulation is restarted by randomly assigning the atomic velocities. Short simulations are perpetuated under the supervision until the distance between the ligand and the binding site goes below 5 Å, then, the supervision is disabled and a classical MD simulation is performed. For the orthosteric center of mass, we considered hA₃AR residues N250, F168, H272, and S247.

NAMD energy Plugin 1.4 (ref. 59) was used to calculate ligand–protein interaction energies.

Conclusions

A series of pyrazolo[4,3-e][1,2,4]triazolo[1,5-c]pyrimidines bearing diaminoalkyl chains at the 5 position (2–19) was developed to target ARs after conjugation with the desired function-bearing moiety. Compounds 2–19 showed different binding profiles at hA₁, hA_2A and hA₃ ARs. In particular, compound 4 and compound 15 are the most potent compound towards hA₃ and hA_2A ARs, respectively. Docking simulations suggested four different binding modes for synthesised compounds at the considered ARs. The preference for one of these binding modes is regulated by the substitution at the 5 position. SuMD analysis on hA₃AR confirmed the existence of multiple binding modes and that sodium and its hydration molecules are involved in the transition from one pose to another. An example of application of these compounds was reported. In particular, conjugation with fluorescein isothiocyanate led to dual, but weak, hA_2A and hA₃ AR fluorescent ligands. Docking revealed that the insertion of the bulky fluorescent moiety on the PTP scaffold forced the ligand to adopt the BM2, in which the linker oriented the fluorophore to the extracellular side of the receptor, and could be the reason for the weak AR interaction. Notably, combination of molecular docking with SuMD techniques helped to better understand the SAR of these derivatives. This can facilitate the future development of a second generation of more potent and selective conjugable compounds with different linkers, and, in addition, the inclusion of an alternative scaffold to PTP is not excluded.

Conflicts of interest

There are no conflicts to declare.

Acknowledgements

KAJ thanks the NIDDK Intramural Res. Program (ZIA DK3117) for support. SF, SM and GS thank the Italian Ministry for University and Research for support.

Notes and references

1. L. A. Stoddart, L. E. Kilpatrick, S. J. Briddon and S. J. Hill, Neuropharmacology, 2015, 98, 48–57.
2. D. E. Sosnovik and R. Weissleder, Curr. Opin. Biotechnol., 2007, 18, 4–10.
3. B. Nabi, S. Rehman, S. Khan, S. Baboota and J. Ali, Brain Res. Bull., 2018, 142, 384–393.
4. M. Srinivasarao and P. S. Low, Chem. Rev., 2017, 117, 12133–12164.
5. R. Haag and F. Kratz, Angew. Chem., Int. Ed., 2006, 45, 1198–1215.
6. R. Morphy and Z. Rankovic, J. Med. Chem., 2005, 48, 6523–6543.
7. D. K. Tosh, L. S. Yoo, M. Chinn, K. Hong, S. M. Kilbey, M. O. Barrett, I. P. Fricks, T. K. Harden, Z. G. Gao and K. A. Jacobson, Bioconjugate Chem., 2010, 21, 372–384.
8. F. Ciruela, V. Fernández-Dueñas and K. A. Jacobson, Neuropharmacology, 2015, 98, 58–67.
9. M. Markovic, S. Ben-Shabat, S. Keinan, A. Aponick, E. M. Zimmermann and A. Dahan, Pharmaceutics, 2018, 10, 1–17.
10. J.-F. Chen, H. K. Eltzschig and B. B. Fredholm, Nat. Rev. Drug Discovery, 2013, 12, 265–286.
11. J. Sawynok, Neuroscience, 2016, 338, 1–18.
12. A. Guerrero, Curr. Med. Chem., 2018, 25, 3597–3612.
13. B. Hocher, Kidney Int., 2010, 78, 438–445.
14. B. Cacciari, G. Spalluto and S. Federico, Mini-Rev. Med. Chem., 2018, 18, 1168–1174.
15. G. Haskó, B. Csóka, Z. H. Németh, E. S. Vizi and P. Pacher, Trends Immunol., 2009, 30, 263–270.
16. S. L. Cheong, S. Federico, G. Venkatesan, A. L. Mandel, Y.-M. Shao, S. Moro, G. Spalluto and G. Pastorin, Med. Res. Rev., 2013, 33, 235–335.
17. S. Redenti, A. Ciancetta, G. Pastorin, B. Cacciari, S. Moro, G. Spalluto and S. Federico, Curr. Top. Med. Chem., 2016, 16, 3224–3257.
18. B. Cacciari, C. Bolcato, G. Spalluto, K. Klotz, M. Bacilieri, F. Deflorian and S. Moro, Purinergic Signalling, 2007, 3, 183–193.
19. S. L. Cheong, A. Dolzhenko, S. Kachler, S. Paoletta, S. Federico, B. Cacciari, A. Dolzhenko, K. Klotz, S. Moro, G. Spalluto and G. Pastorin, J. Med. Chem., 2010, 53, 3361–3375.
20. S. Federico, A. Ciancetta, D. Sabbadin, S. Paoletta, G. Pastorin, B. Cacciari, K. N. Klotz, S. Moro and G. Spalluto, J. Med. Chem., 2012, 55, 9654–9668.
21. S. Federico, S. Redenti, M. Sturlese, A. Ciancetta, S. Kachler, K.-N. Klotz, B. Cacciari, S. Moro and G. Spalluto, PLoS One, 2015, 10, e0143504.
22. E. Kozma, T. S. Kumar, S. Federico, K. Phan, R. Balasubramanian, Z.-G. Gao, S. Paoletta, S. Moro, G. Spalluto and K. A. Jacobson, Biochem. Pharmacol., 2012, 83, 1552–1561.
23. K. A. Jacobson, S. K. Thatikonda, E. E. Kozma, G. Spalluot, S. Moro and S. Federico, US2013/0190335A1, 2013, p. 32.
24. E. Kozma, P. S. Jayasekara, L. Squarcialupi, S. Paoletta, S. Moro, S. Federico, G. Spalluto and K. A. Jacobson, Biochem. Pharmacol., 2013, 23, 26–36.
25. E. Margiotta, G. Deganutti and S. Moro, J. Comput.-Aided Mol. Des., 2018, 32, 1337–1346.
26. A. Ciancetta, D. Sabbadin, S. Federico, G. Spalluto and S. Moro, Trends Pharmacol. Sci., 2015, 36, 878–890.
27. O. K. Onajole, P. Govender, T. Govender, G. E. M. Maguire and H. G. Kruger, Struct. Chem., 2009, 20, 1067–1076.
28. K.-N. Klotz, J. Hessling, J. Hegler, C. Owman, B. Kull, B. B. Fredholm and M. J. Lohse, Naunyn-Schmiedeberg's Arch. Pharmacol., 1998, 357, 1–9.
29. A. De Lean, A. A. Hancock and R. J. Lefkowitz, Mol. Pharmacol., 1982, 21, 5–16.
30. K. N. Klotz, G. Cristalli and M. Grifantini, J. Biol. Chem., 1985, 260, 14659–14664.
31. S. Federico, S. Redenti, M. Sturlese, A. Ciancetta, S. Kachler, K.-N. Klotz, B. Cacciari, S. Moro and G. Spalluto, PLoS One, 2015, 10, e0143504.
32. Z. G. Gao, J. B. Blaustein, A. S. Gross, N. Melman and K. A. Jacobson, Biochem. Pharmacol., 2003, 65, 1675–1684.
33. S. Hinz, W. M. Alnouri, U. Pleiss and C. E. Müller, Purinergic Signalling, 2018, 14, 223–233.
34. Molecular Operating Environment (MOE), 2013.08; Chemical Computing Group ULC, 1010 Sherbooke St. West, Suite #910, 6Montreal, QC, H3A 2R7 (Canada), 2017.
35. M. J. Harvey, G. Giupponi and G. D. Fabritiis, J. Chem. Theory Comput., 2009, 5, 1632–1639.
36. A. D. MacKerell, N. Banavali and N. Foloppe, Biopolymers, 2000–2001, 56, 257–265.
37. GOLD - The Cambridge Crystallographic Data Centre (CCDC). https ://www.ccdc.cam.ac.uk/solutions/csd-disco very/compo nents/gold/. Accessed 27 Dec 2017 (Genetic Optimization for Ligand Docking, version 5.4.2).
38. J. Wang, R. M. Wolf, J. W. Caldwell, P. A. Kollman and D. A. Case, J. Comput. Chem., 2004, 25, 1157–1174.
39. H. M. Berman, J. Westbrook, Z. Feng, G. Gilliland, T. N. Bhat, H. Weissig, I. N. Shindyalov and P. E. Bourne, Nucleic Acids Res., 2000, 28, 235–242.
40. A. S. Doré, N. Robertson, J. C. Errey, I. Ng, K. Hollenstein, B. Tehan, E. Hurrell, K. Bennett, M. Congreve, F. Magnani, C. G. Tate, M. Weir and F. H. Marshall, Structure, 2011, 19, 1283–1293.
41. A. Glukhova, D. M. Thal, A. T. Nguyen, E. A. Vecchio, M. Jörg, P. J. Scammells, L. T. May, P. M. Sexton and A. Christopoulos, Cell, 2017, 168, 867–877.
42. E. Margiotta and S. Moro, Appl. Sci., 2019, 9, 821.
43. J. A. Ballesteros and H. Weinstein, Methods Neurosci., 1995, 25, 366–428.
44. Gnuplot homepage. http://gnuplot.sourceforge.net/, Accessed 27 Dec 2018.
45. MarvinSketch. https://docs.chemaxon.com/displ ay/docs/MarvinSketch+Home. Accessed 27 Dec 2018.
46. E. F. Pettersen, T. D. Goddard, C. C. Huang, G. S. Couch, D. M. Greenblatt, E. C. Meng and T. E. Ferrin, J. Comput. Chem., 2004, 25, 1605–1612.
47. W. Humphrey, A. Dalke and K. Schulten, J. Mol. Graphics, 1996, 14, 33–38 , 27–28.
48. RDKit: Cheminformatics and Machine Learning Software, Available from: http://www.rdkit.org.
49. M. Head-Gordon, J. A. Pople and M. J. Frisch, Chem. Phys. Lett., 1988, 153, 503–506.
50. M. J. Frisch, G. W. Trucks, H. B. Schlegel, G. E. Scuseria, M. A. Robb, J. R. Cheeseman, G. Scalmani, V. Barone, G. A. Petersson, H. Nakatsuji, X. Li, M. Caricato, A. Marenich, J. Bloino, B. G. Janesko, R. Gomperts, B. Mennucci, H. P. Hratchian, J. V. Ortiz, A. F. Izmaylov, J. L. Sonnenberg, D. Williams-Young, F. Ding, F. Lipparini, F. Egidi, J. Goings, B. Peng, A. Petrone, T. Henderson, D. Ranasinghe, V. G. Zakrzewski, J. Gao, N. Rega, G. Zheng, W. Liang, M. Hada, M. Ehara, K. Toyota, R. Fukuda, J. Hasegawa, M. Ishida, T. Nakajima, Y. Honda, O. Kitao, H. Nakai, T. Vreven, K. Throssell, J. A. Montgomery Jr., J. E. Peralta, F. Ogliaro, M. Bearpark, J. J. Heyd, E. Brothers, K. N. Kudin, V. N. Staroverov, T. Keith, R. Kobayashi, J. Normand, K. Raghavachari, A. Rendell, J. C. Burant, S. S. Iyengar, J. Tomasi, M. Cossi, J. M. Millam, M. Klene, C. Adamo, R. Cammi, J. W. Ochterski, R. L. Martin, K. Morokuma, O. Farkas, J. B. Foresman and D. J. Fox, Gaussian 16, Revision B.01, Gaussian, Inc., Wallingford CT, 2016.
51. M. A. Lomize, A. L. Lomize, I. D. Pogozheva and H. I. Mosberg, Bioinformatics, 2006, 22, 623–625.
52. W. L. Jorgensen, J. Chandrasekhar, J. D. Madura, R. W. Impey and M. L. Klein, J. Chem. Phys., 1983, 79, 926–935.
53. H. Grubmuller and V. Groll, Available online: http://www.mpibpc.mpg.de/grubmueller/solvate.
54. H. J. C. Berendsen, J. P. M. Postma, W. F. van Gunsteren, A. DiNola and J. R. Haak, J. Chem. Phys., 1984, 81, 3684–3690.
55. R. J. Loncharich, B. R. Brooks and R. W. Pastor, Biopolymers, 1992, 32, 523–535.
56. U. Essmann, L. Perera, M. L. Berkowitz, T. Darden, H. Lee and L. G. A. Pedersen, J. Chem. Phys., 1995, 103, 8577–8593.
57. D. Sabbadin and S. Moro, J. Chem. Inf. Model., 2014, 54, 372–376.
58. A. Cuzzolin, M. Sturlese, G. Deganutti, V. Salmaso, D. Sabbadin, A. Ciancetta and S. Moro, J. Chem. Inf. Model., 2016, 56, 687–705.
59. J. C. Phillips, R. Braun, W. Wang, J. Gumbart, E. Tajkhorshid, E. Villa, C. Chipot, R. D. Skeel, L. Kale and K. Schulten, J. Comput. Chem., 2005, 26, 1781–1802.

---

Footnotes

† Electronic supplementary information (ESI) available: Videos concerning docking poses, electrostatic and hydrophobic contributions of peculiar residues and results of SuMD for synthesised compounds. See DOI: 10.1039/c9md00014c

‡ Current affiliation: Sygnature Discovery, Nottingham, UK.

---