Dissertation committee:
Antonios Kolocouris - Professor, Section of Pharmaceutical Chemistry, School of Pharmacy, National and Kapodistrian University of Athens
Emmanuel Mikros - Professor, Section of Pharmaceutical Chemistry, School of Pharmacy, National and Kapodistrian University of Athens
Andrew Tsotinis - Professor, Section of Pharmaceutical Chemistry, School of Pharmacy, National and Kapodistrian University of Athens
Angeliki Kourounakis - Professor, Section of Pharmaceutical Chemistry, School of Pharmacy, National and Kapodistrian University of Athens
Graham Ladds - Lecturer, Department of Pharmacology, University of Cambridge
Vassilis Myrianthopoulos - Assistant Professor, Section of Pharmaceutical Chemistry, School of Pharmacy, National and Kapodistrian University of Athens
Thanassis Papakyriakou - Senior Researcher at the Institute of Biosciences and Applications, National Centre for Scientific Research “Demokritos”
Summary:
Pharmaceutical companies and academic research laboratories are involved in intense efforts to identify antagonists with selectivity for each adenosine receptor (AR) subtype as potential clinical candidates for "soft" treatment of different diseases. All AR sub-types play distinct roles throughout the body. A2AR antagonists can be useful for treating cancer, central nervous system (CNS) disorders; A1R antagonists can provide kidney-protective agents, anti-asthmatic and CNS agents; A3R antagonists are promising for therapeutic applications in asthma, glaucoma and A2BR antagonists for diabetes, asthma and chronic obstructive pulmonary disease. The reported crystal structures of A2AR in complex with agonists or antagonists and of A1R with an antagonist, along with other advances attributed to the progress of GPCR crystallography have made structure-based approaches an attractive strategy for drug design against adenosine receptors which are pharmaceutically important targets. The A2AR is one of the best studied receptors of all class A GPCRs. Additionally, among the 688 known GPCRs, class A is the 7th more intensely investigated. The application of virtual screening and medicinal chemistry studies for a few decades now has resulted in a high number of bioactive compounds (~ 11000) against A2AR as was retrieved from ChEMBL20. An introduction to GPCRs and ARs is the subject of first Chapter.
In the second Chapter of the thesis, is presented the virtual screening (VS) results of the small Maybridge HitFinder library of 14,400 compounds against A2AR, using its crystallographic structure in complex with the antagonist ZM241385, through a combination of structure-based and ligand-based procedures. This is one of the few VS against ARs reported in the literature which however use the ZINC library of millions compounds . The docking poses were re-scored by applying energy minimization using CHARMM software with CHARMM19 ff of the ligand inside rigid receptor and consideration of desolvation energy electrostatics using the Poisson-Boltzmann equation, i.e., using Molecular Mechanics-Poisson Boltzmann Surface Area (MM-PBSA) method reduced to include only energy minimization. Out of the eight selected and tested compounds, three showed micromolar affinity for the A2A and A3Rs and two were low micromolar binders only to the A3R receptor using radiolabelled assays. Thus, although initially targeting the A2AR, the project resulted in the following percent of successful binder hits: 25% for A2A and 63% for A3R. Of particular interest for futher exploration as novel chemical probes, are the synthetically feasible Κ1 and Κ5 with a 2-amino-thiophene-3-carboxamide and a carbonyloxycarboximidamide structure, respectively. Given the similarity between ARs orthosteric binding sites, obtaining highly selective receptor antagonists is a challenging but critical task.
In a second step, described in Chapter 3 of the thesis, based on the structure of mainly two promising active hits, possessing a 2-amino -3-carboxamide-thiophene and a carbonyloxycarboximidamide chemotype respectively, 19 more compounds were selected by similarity for testing. For this second series of 19 compounds, 17 were found to bind to the ARs using radiolabelled assays. Eight of those revealed A3-selective affinity with Ki values in the micromolar to low micromolar regime. Along with 2-amino-thiophene-3-carboxamide and carbonyloxycarboximidamide derivatives we identified a new class of ligands, the 3-acylamino-5-aryl-thiophene-2-carboxamides.
These three classes of compounds have novel chemotypes with low Tc values (< 0.17) compared to the known ARs ligands have been identified: (a) The 2-amino thiophene-3-carboxamide (2-NH2 and 3-CONHR; class A) thiophenes with low micromolar affibity to A2AR and A3R. (b) The 3-acylamino-5-aryl-thiophene-2-carboxamides (class B) including the new substitution pattern (2-CONH2 and 3-NHCOR) of the thiophene ring, which -compared to 2-NH2 and 3-CONHR' substitution pattern- enhances the affinity for A1R and A3R. (c) The carbonyloxycarboximidamide derivatives (class D), many having selective A3R affinity. The selective A3R ligands with micromolar affinity and novel chemotypes found here, may contribute to the treatment of the A3R-related human pathologies.
Compound K18 (O4-{[3-(2,6-dichlorophenyl)-5-methylisoxazol-4-yl]carbonyl}-2-methyl-1,3-thiazole-4-carbohydroximamide) with the carbonyloxycarboximidamide chemotype have the lowest micromolar binding affinity to A3R (Kd=0.898 μΜ) between the discovered lead compounds. We focused on the selective A3R ligand K18 with the lowest micromolar binding affinity to A3R and purchased and measured the binding affinity of 12 new carbonyloxycarboximidamide analogs including mainly compounds that bear a biphenyl instead of 3-phenyl-isoxazole for additional structure-activity relationship (SAR) studies.
The 39 tested molecules resulted in similar docking poses against A1, A2A, A2B or A3Rs. The experimental structures of A1, A2ARs, after completion of missing loops, were used for the simulations. Since A2B or A3Rs are unsolved, homology models were applied. Using the docking poses of the ligands as starting structures, the performance of hundreds of 20ns-molecular dynamics (MD) simulations, using Desmond software with OPLS2005 force field (ff), allow the differentiation of stable and unstable docking poses based on the RMSD values for the displacement of the ligand from its starting docking pose inside the orthosteric binding area. Generally, stable or unstable docking poses agree with the experimental results of radiolabelled values of binding affinity. The stability of the stable complexes were further tested using 100ns-MD simulations using Desmond software with OPLS2005ff and Amber software with amberff14sb provided the basic features of the binding interactions with A1, A2A, and A3Rs for compounds exhibiting affinity.
The MD simulations show the basic features of the binding interactions with A1, A2A, and A3Rs for compounds exhibiting affinity. The complexes with A2A and A1Rs were stabilized through hydrogen bonding interactions between an amino or amido group of the ligand and N(6.55) of the AR. E(5.30) can be involved also in hydrogen bonding interactions with the bound ligand. A2AR ligands include a lipophilic bulky substituent which was oriented towards the extracellular area, close to EL2 and TM7, and a smaller lipophilic group which was fitted deep in the binding region, close to L(6.51) and H(6.52). Similar interactions have been described in the X-ray structures between antagonists and A2AR. Interestingly, for the A1R ligands the ligand covers a larger space between TM5/TM6 and TM1/TM2, as shown in the recent X-ray structure between an antagonist and A1R. (Glukhova et al. 2017c) Many of the ligands studied in this report, i.e. K1, K2 K5-K7, K9-K14, K16, K18, K20-K27, K31, K32, K35, bind to A3R. We suggest that selectivity against A3R is boosted by increasing the size and lipophilicity of a suitable substituent reflecting a better fit with V(5.30). Compounds K6, K7, K10, K12-K15, K17, K18, K25, K27, K31, K32 are selective binders to A3R. These findings are in line with previously published results from our group on the description of the orthosteric binding area of highly selective A3R agonists with a bulky group in a compatible position, like the 3-iodo-benzyl group in N6-position in N6-(3-iodobenzyl)-adenosine-5′-N-methyluronamide (IB-MECA) which has increased binding affinity for A3R. In partcular we applied MD simulations and Molecular Mechanics-Generalized Born (MM-GBSA) in combination with mutagenesis data.
A3R antagonists have been described as potential treatments for numerous diseases including asthma. In Chapter 4, it is described that the 39 potential A3R antagonists were screened using agonist-induced inhibition of cAMP. Positive hits were assessed for AR subtype selectivity through cAMP accumulation assays. The antagonist affinity was determined using Schild analysis (pA2 values) and fluorescent ligand binding using the bioluminescence resonance energy transfer (BRET) method. Further, a likely binding pose of the most potent antagonist K18 was determined through MD simulations using an homology model of A3R, combined with mutagenesis studies.
Eventually it was suggested that K18, which contains a 3-(dichlorophenyl)-isoxazole group connected through carbonyloxycarboximidamide fragment with a 1,3-thiazole ring, is a specific A3R (<1 µM) competitive antagonist. Structure-activity relationship investigations revealed that loss of the 3-(dichlorophenyl)-isoxazole group significantly attenuated K18 antagonistic potency. Mutagenic studies supported by MD simulations identified the residues important for binding in the A3R orthosteric site.
We also introduce a model that enables estimates of the equilibrium binding affinity for rapidly dissociating compounds from real-time fluorescent ligand-binding studies. These results demonstrate the pharmacological characterization of a selective competitive A3R antagonist and the description of its orthosteric binding mode.
In Chapter 5, the binding profile of the selective K18 inside the orthosteric binding site of A3R was further investigated and a computational model was also suggested for A3R in complex with antagonists by applying detailed simulations.
The A3R is currently an important drug target,(Liang and Jacobson 2002; Okamura et al. 2004) and there is a lack of available structures. In this work using experimental pA2 values from mutagenesis experiments, a computational model for the description of a specific antagonist binding with orthosteric binding area of A3R is approved. In particular, we generated a computational model based on: (a) An homology model of A3R in complex with K18 and the most likely binding conformation of K18 inside WT A3R orthosteric binding area which was investigated using, MD simulations with amber99sb, and MM-PBSA and MM-GBSA calculations. (b) The effect of point-mutations of residues in the orthosteric binding area to K18 activity.
We first tested if the amber ff99sb can describe the conformational change from active to inactive form of A2AR when the active form is complexed with ZM213485 in hydrated POPE bilayers. Since, we observed the characteristic reduction in distance between TM3 and TM6 from ca 11 to 7.5 Ǻ we used ff99sb as appropriate for the MD simulations of the complexes between K18 and WT or mutant A3Rs.
In a previous study, it was found experimentally and confirmed computationally using the same model that critical interactions for IB-MECA activity to A3R include residues at the TM5, TM6 and EL2. These are F1685.29, L2466.51, V1695.30, N2506.55 forming direct interactions with agonist and M1775.38, L903.32 at the bottom of the orthosteric binding area which include indirect interactions. Other critical direct interactions for IB-MECA activity include the additional residues at the bottom of the binding area, T943.36, S2717.42, H2727.43 and I2687.39.
Three likely different docking poses of K18 and its congeners K5, K17 differing in conformation and orientation inside the binding area were examined by molecular dynamics (MD) simulations with amber and MM-PBSA calculations. Two of them have equal energies with thiazole ring oriented deep in the receptor and dichlorophenyl of K18 oriented towards either TM5, TM6 or TM1, TM2. The significance of these conformations was investigated using the site-directed mutagenesis experiments and biological activities results of mutant A3Rs in complex with K18 which suggested that the dichlorophenyl ring of K18 is oriented towards TM5, TM6.
Thus, according to our computational model the competitive antagonist K18 is stabilized inside the A3R orthosteric binding area through an "up TM5, TM6" conformer which interacts directly with some common residues with the agonist. It forms a π-π interaction with F1685.29, van der Waals interactions with L903.32, V1695.30, L2466.51, and hydrogen bond interactions with N2506.55. In the middle region of the A3R, K18 makes contacts with residues M1775.38, I2496.54 which are not in contact with IB-MECA. To add further contrast, IB-MECA contacts residue W1855.46 whereas K18 does not. From these residues M1775.38 causes a negation of both agonists and antagonist potency when mutated to alanine. L903.32 is a residue in contact with K18 but not in contact with the agonists suggesting that K18 sits higher in the orthosteric binding region. L903.32A mutation causes correspondingly an increase in the potency of K18 and a reduction in the potency of agonists. Our calculations describe why the majority of mutated residues to alanine, which are in contact with K18 antagonist in the WT receptor, reduce or eliminate potency, i.e. correspondingly V1695.30, M1775.38 or L2466.51, F1685.29, N2506.55. Additionally, the computational model shows that the selectivity of K18 is not only due to direct interactions with the binding area residues. Remote residues which are positioned at the edges of the binding area in EL2, TM5 and TM6, like M1745.35 at 4 Å may act by modulating the structure of the pocket. Residue M1745.35 is important for NECA and K18 activity since its mutation to alanine reduce potency. The results produced experimental pA2 values which were used as experimental probes for MD simulations and binding free energy MM-GBSA calculations for of K18 in complex with 14 mutant A3Rs. Using the MM-GBSA calculated ΔGeff values it was possible to distinguish three sets of mutant receptors, i.e. those that reduce or negate K18 potency at the A3R, those that bind stably and maintain potency and those that increase potency compared to WT A3R. The calculated ΔGeff values for K18 and experimentally determined pA2 values displayed very good correlation, with r = -0.81. In our previous work investigating IB-MECA and NECA agonists binding to A3R, the correlation between calculated ΔGeff values and experimental pIC50 values was also fair (correspondingly r = -0.69 and r = -0.76).
The characterization of the area TM6-EL2-TM5 in A3R which includes lipophilic residues is very important for structure-based drug design of selective ligands. Although this area is considered to be occupied from the lipophilic groups of selective ligands, like the iodo-benzyl group in IB-MECA, the experimental results show and the computational model supports that the mutation V1695.30E causes an increase in IB-MECA and NECA activity, rather than the expected reduction, and that I2536.58 is not an important residue of this region. We also show here that I2536.58 and V1695.30E maintains K18 antagonistic potency. It is also interesting that the potency of K18 is enhanced by the mutations of L903.32A in the low region or L2647.35A in the middle/upper region which are directly interacting residues with K18, suggesting an empty space in the orthosteric area available for increasing antagonist potency. These findings could have significant impact on the design of potent and selective ligands targeting A3R.
