Dissertation committee:
Αντώνης Κολοκούρης, Καθηγητής, Τμήματος Φαρμακευτικής, ΕΚΠΑ.
Graham Ladds, Professor, Department of Pharmacology, University of Cambridge.
Εμμανουήλ Μικρός, Καθηγητής Τμήματος Φαρμακευτικής, ΕΚΠΑ.
Αθανάσιος Παπακυριακού, Iνστιτούτο Βιοεπιστημών και Εφαρμογών, Εθνικό Κέντρο Έρευνας Φυσικών
Επιστημών, Δημόκριτος.
Νικόλαος Λουγιάκης, Eπικ. Καθηγητής Τμήματος Φαρμακευτικής, ΕΚΠΑ.
Μίνος Ματσούκας, Eπικ. Καθηγητής Τμήματος Μηχανικών Βιοϊατρικών Επιστημών, ΠΑΔΑ.
Θωμάς Μαυρομούστακος, Kαθηγητής Τμήματος Χημείας, ΕΚΠΑ.
Summary:
In the pursuit of developing effective therapeutics, structure-based drug design has emerged as a powerful approach, leveraging our understanding of molecular structures to design molecules with enhanced binding and functional properties. The search for effective therapeutics targeting adenosine receptors (ARs), members of the G protein-coupled receptors (GPCRs) family, has gained substantial significance due to their involvement in various pathological conditions.
The aim of this thesis is to explore the field of structure-based drug design, focusing on the development of antagonists targeting adenosine receptors A1 and A3. In pursuit of this objective, advanced computational methods like alchemical free energy perturbation and kinetic binding calculations are employed. In Chapter 1, there is an introduction on GPCRs with emphasis on ARs. Agonists, antagonists, and dual antagonists are mentioned that act on the orthosteric and allosteric sites. Chapter 2 describes the principles of the methodologies applied through the present thesis.
In Chapter 3, the synthesized derivatives of 7-aryl or alkylamino-pyrazolo[3,4-d]pyridazine provided a novel scaffold for developing ligands against ARs. We have pharmacologically characterized these compounds using functional cAMP assays and fluorescent ligand displacement binding studies, expanding our study to the antiproliferative potential of these agents as well. The introduction of a 3-phenyl group, together with a 7-benzylamino and 1-methyl group at the pyrazolopyridazine scaffold, generated the antagonist compound 10b which displayed 26 nM affinity and a residence time (RT) 60 min for the human A1R, 7.4 nM affinity and RT = 73 min for the human A3R and low μΜ affinity for the human A2BR while not be toxic against the normal cell line. The site of the N-methyl substitution on the pyrazole ring had a remarkable effect on the bioactivity, since the corresponding 2-methyl-3-phenyl derivative (15b) had no significant affinity, while when the 3-phenylgroup of 10b was replaced by an isopropyl group, the resulting derivative 10a possessed considerably reduced affinity. We compared the binding interactions of the regio-isomers 10b and 15b with molecular dynamics (MD) simulations and the results suggested that the 2-methyl group in 15b hinders the formation of hydrogen bonding interactions with N6.55 which are considered critical for the stabilization inside the orthosteric binding cavity. Mutagenesis experiments for 10b against A1R provided results that complement the observations from MD simulations. We showed that L2506.51A mutation resulted in only a slight reduction of binding affinity concerning 10b while the Y2717.46A mutation caused a 10-fold reduction in binding affinity of this compound. Mutation to alanine of residues T913.36, H2516.52 or S2677.42, which are deep in the orthosteric binding affinity, did not affect binding affinity.
In Chapter 4, we report the identification of 7- (phenylamino)-pyrazolo[3,4-c]pyridines L2−L10, A15, and A17 as low-micromolar to low-nanomolar A1R/A3R dual antagonists, with 3-phenyl-5-cyano-7-(trimethoxyphenylamino)-pyrazolo[3,4-c]pyridine (A17) displaying the highest affinity at both receptors with a long residence time of binding, as determined using a NanoBRET based assay. Two binding orientations of A17 produce stable complexes inside the orthosteric binding area of A1R in MD simulations, and we selected the most plausible orientation based on the agreement with alanine mutagenesis supported by affinity experiments. Interestingly, for drug design purposes, the mutation of L2506.51 to alanine increased the binding affinity of A17 at A1R. We explored the structure−activity relationships against A1R using alchemical binding free energy calculations with the thermodynamic integration coupled with the MD simulation (TI/MD) method, applied on the whole GPCR−membrane system, which showed a good agreement (r = 0.73) between calculated and experimental relative binding free energies.
In Chapter 5, we sought to develop a computational model of inactive adenosine A3 receptor (A3R), not yet resolved experimentally, for drug design purposes. We tested five homology models of inactive human A3R (hA3R) that are either publicly available or available from a web-resource. After merging 3 homology models by similarity, we came up with homology Models 1 and 2 and the AlphaFold2-based Model 3. We observed that these models showed good agreement in the orthosteric binding area except in upper region where Models 1, 2 differed from Model 3 in the orientation of side chains of R1735.34, M1725.33 and M1745.35 located in the extracellular loop 2 (EL2). We compared Models 1-3 regarding predictions of the experimentally determined thermodynamic and kinetic stability for the pyrazolo[3,4-d]pyridazine antagonists. The protein Models 1-3 in TI/MD calculations performed with good agreement (r = 0.74, 0.62 and 0.67, respectively) between the calculated and experimental relative binding free energies. The τ-Random Acceleration Molecular Dynamics (τRAMD) simulations effectively distinguished between compounds with short and long RT within the receptor only with Models 1, 2, since in Model 3 the orientation of R1735.34 located at the top of ligands’ exit route affected compound dissociation. By optimizing the orientation of side chains of residues M1725.33, R1735.34, M1745.35 in Model 3 the optimized Model 3 was generated. τRAMD simulations using the optimized model 3 correctly ranked ligands according to their residence time inside binding site. Furthermore, the performance of TI/MD calculations with the optimized Model 3 was improved such as the Pearson correlation coefficient was increased from r = 0.67 to 0.84 while the mean assigned error was reduced from 0.81 kcal mol-1 to 0.56 kcal mol-1.
