DOI
https://doi.org/10.25772/X077-TH73
Author ORCID Identifier
https://orcid.org/0000-0002-1126-5886
Defense Date
2024
Document Type
Thesis
Degree Name
Doctor of Philosophy
Department
Pharmaceutical Sciences
First Advisor
Martin Safo
Second Advisor
Qingguo Xu
Third Advisor
Aaron May
Fourth Advisor
Glen Kellogg
Fifth Advisor
Moustafa Dawoud
Abstract
Viruses are submicroscopic infectious agents causing immense global disease burdens. Propagation of viral particles relies on proteolytic cleavage of polyprotein precursors by host or virally-encoded proteases to liberate functional components necessary for replication and infection cycles. These processing events present vulnerable intervention points for antiviral targeting. This work focused on two indispensable viral proteases - the SARS-CoV-2 main protease and the NS3 protease domain from hepatitis C virus.
SARS-CoV-2 relies on the main protease (Mpro) for critical processing of virally-encoded replicase polyproteins through specific cleavages to generate mature non-structural proteins that coordinate viral maturation and assembly pathways. Owing to this indispensable role, Mpro presents an attractive target for therapeutic intervention to disrupt coronavirus pathogenesis. Furthermore, the near-identical substrate recognition features and catalytic domain homology with the SARS-CoV main protease enabled rapid transfer of structural knowledge to inform inhibitor design efforts using prior lead compounds. In this study, an ensemble of small molecule discovery platform including computational screening, synthetic chemistry, enzymology and biophysical characterization was constructed to systematically identify inhibitors against this important drug target. Three virtual screening protocols using complementary in silico techniques were utilized– ligand-based 3D pharmacophore searches, protein structure-centric molecular docking, and artificial intelligence models employing deep neural networks. This triaged computational workflow resulted in prospective hit candidates. In parallel, quantitative structure-activity examinations of a small, focused library of 168 synthetically derived α-ketoamide compounds revealed a reactive Michael acceptor warhead amenable for covalently targeting a key catalytic cysteine residue. Fluorescence resonance energy transfer (FRET) enzyme assay confirmed dose-dependent SARS CoV-2 Mpro inhibition for 10 ligands – 7 from virtual screening pipelines (MA1, MA2, MA3, MA4, MA5, MA6 and MA7) and 3 α-ketoamide derivatives (MCP212, MCP221, AND MCP256) – with low micromolar half maximal inhibitory concentrations between 1.7-55 μM, with MA5 showing the most potent inhibitory activity. Direct binding quantification of MA1, MA2, MA3 MA4, MA5, MA6 and M7 using microscale thermophoresis (MST) resulted in moderate affinity of 4.93-55.28 μM with MA4 showing the highest affinity. Attempts to determine the crystal structures of Mpro in complex with ligands failed, resulting in the non-liganded bound Mpro structure, likely due to poor aqueous solubility. Nevertheless, the 1.8Å resolution unliganded Mpro structure provided insight into plasticity elements lining the substrate binding cleft. Microsecond timescale explicit-solvent molecular dynamics simulations tracked long-term dynamic stabilities of inhibitor-bound complexes, corroborated through rigorously computed binding free energy predictions. Moreover, MD studies indicated stable complexes between the MA compounds and the SARS-CoV-2 main protease, evidenced by consistent protein and ligand root mean square deviation values over the trajectories. Lastly, enrichment factor and hit success rate were calculated to assess performance of the virtual screening approaches. The AI-powered workflow demonstrated the highest performance with high EFs of 12.5-16. This observation is consistent with the fact that AI virtual screening approach attained the highest hit success rate of 30% compared to the traditional structure-based and docking-based workflows approaches.
The hepatitis C virus (HCV) nonstructural protein 3 (NS3) possesses two essential enzymatic activities - an N-terminal serine protease domain followed by a C-terminal RNA helicase module. The protease domain mediates critical downstream viral polyprotein processing through binding to an NS4A cofactor peptide, which induces a catalytically competent configuration. In isolation, the NS3 protease domain exhibits negligible intrinsic catalytic activity. However, upon association with the hydrophobic 54-amino acid NS4A cofactor, the protease domain attains sufficient enzymatic activity required for mediating viral polyprotein cleavages linked to HCV replicative competency. Binding of NS4A induces key structural rearrangements in the otherwise natively disordered apo-NS3 protease, enabling organization of the catalytic triad residues into a functional conformation optimized for substrate turnover. This activation paradigm presented possibilities for allosteric modulation of the NS3 protease through engineered NS4A variants designed to retain binding affinity but subtly distort functional geometries via strategic mutations. Results from our study validated this hypothesis, revealing a designed nanomolar-binding NS4A variant, PEP15 with a single cyclohexylglycine (xG) mutation (V23xG) that bind with NS3 but eliminated enzymatic activity. Microscale thermophoresis quantifications revealed PEP15 associated with the NS3 protease domain with remarkable nanomolar binding affinity (Kd of 22.23 ± 0.297 nM), approximately two orders of magnitude stronger than the native NS4A cofactor peptide (Kd of 2.595 ± 0.0015 μM). This significantly improved binding affinity despite a single residue substitution substantiates the significant energetics contributions of the engineered glycine mutation and validates the allosteric targeting rationale underlying the inhibitor design. Differential scanning fluorimetry indicated unexpected reductions in thermal stability relative to the native complex or isolated protein controls. The reduced temperatures could be due to misaligned active site elements due to allosteric or strain effects which secondarily distort global stability. Molecular dynamic simulations provided insights into the unexpected biophysical findings by modeling dynamics and stability of the PEP15-NS3 complex. The trajectories revealed favorable occupying of the deep hydrophobic environment lining the NS3 allosteric pocket by the engineered glycine substitution. Notably, the modelling also captured shifting of the key SER139 hydroxyl moiety away from the organized catalytic triad geometric center. Displacement of this nucleophilic residue plausibly misaligns other proximal components due to intricate hydrogen bonding networks. Structural rearrangement of the active site elements likely contributes to the abolished enzymatic activity despite high affinity binding of the PEP15 peptide.
In conclusion, this work exemplified productive integration of computational predictions and experimental characterization to identify micromolar potency SARS-CoV-2 Mpro inhibitors, while also developing robust machine learning frameworks that may assist in expediting data-driven responses against future microbiological threats. Additionally, this work established a framework for designing non-peptidic small molecule inhibitors targeting the allosteric NS4A binding site on the HCV NS3 protease, involving strategic substitution of the validated glycine mutation with non-peptidic isosteres to retain crucial contacts while introducing adjacent chemical elaboration to enable potent disruption of the NS3-NS4A protein-protein interface. Overall, the pragmatic fusion of simulations with benchtop assays constituted a versatile platform poised for swift activation against pressing health challenges by progressing customized inhibitors modulating unexploited sites on pathogen proteins.
Rights
© Mohammed AlAwadh
Is Part Of
VCU University Archives
Is Part Of
VCU Theses and Dissertations
Date of Submission
5-10-2024