Development of allosteric inhibitors of flaviviral processing proteinases and Hepatitis C NS3/4A viral proteinase
The following is the summary presentation of already published results. The data are arranged into sequential timeline of research and development flow.
Identification of allosteric inhibitors of West Nile virus NS2B-NS3Pro flavivirus 2-component processing proteinase.
The viral proteinase is composed of NS3pro catalytic domain and NS2B peptide cofactor. The active site of NS3pro catalytic domain is largely conserved in the human and viral serine proteinases. It lacks the structural features which could be readily exploited to achieve both the specificity and potency of the inhibitors.
Luckily, the unique feature of the flaviviral proteinases is the requirement for specific productive interactions of NS2B peptide cofactor with NS3pro catalytic domain for NS3pro domain to be enzymatically active. Thus, allosteric interference with these productive interactions using small molecule ligands could lead to development of potent and selective NS2B-NSpro inhibitors (see publication).
The figure below represents the basic idea of targeting strategy:
Two conformations of NS2B-NS3pro complex are shown: “closed” active – NS2B peptide is part of the active site (green circle); “open” inactive – NS2B cofactor is on the opposite side of the active site. The star on the active complex indicates location of the docking site.
The docking site was selected as indicated above, the NS2B cofactor was removed before the docking simulation.
The Q-MOL docking curve:
The complete NCI DTP SDF library was used as a source of ligands (≈ 275,000 structures). The Q-MOL primary VLS converged to 85 ligands (99.97% enrichment factor), out of which, top 50 predicted binders were ordered from NCI and tested in in vitro cleavage assay. Out of 50 assayed ligands, 3 ligands had IC50 < 1 µM (black filled circles), 2 ligands had IC50 in range 1 µM - 10 µM, and 13 ligands had IC50 > 10 µM (18 active hits in total, 36% success rate). Inset: plot of ligands molecular weight (MW) vs. their rank. The inset is shown to demonstrate that there is no correlation between computed binding energy and the size of a ligand.
The table below summarizes the outstanding success of Q-MOL virtual ligand screening:
The numbers are shown for the inhibitors the IC50 value of which is below 10 mM. “+” indicates that the IC50 value of the inhibitor is in the respective concentration range. The most potent inhibitors are in bold. The IC50 values were determined in the reactions which, in addition to the inhibitor, included the purified proteinase and the Pyr-RTKR-AMC peptide substrate.
Also note that ligands were assayed in parallel on a homologous Dengue virus NS2B-NS3pro proteinase. In line with the targeting strategy, a number of West Nile virus proteinase inhibitors were also active in Dengue NS2B-NS3pro proteinase assay.
Further, it was demonstrated that some of identified inhibitors are potent against another homologous NB2B-NS3pro proteinase from Zika virus (see publication):
Finally, the most potent pan flaviviral proteinase inhibitor NSC86314 was shown in an independent study to be active in vivo, and was co-crystallized with a mutant super-open (inactive) form of Zika proteinase 7M1V (image below).
Crystal structure of Zika virus mutant NS2B/NS3 proteinase in super-open conformation with bound compound NSC86314.
Identification of allosteric inhibitors of Hepatitis C virus NS3/4A-NS4A proteinase.
Hepatitis C NS3/4A serine proteinase requires a peptide cofactor NS4A for its activation. Thus, its mechanism of activation is similar to that of West Nile virus proteinase. Therefore the same in silico drug discovery strategy could be applied as for West Nile proteinase (see publication).
As the exact placement of NS4A cofactor is unknown, the Q-MOL virtual ligand screening was conducted against 3 docking sites on the surface of NS3/4A catalytic domain:
The location of docking sites are displayed with red molecular surface patch, green molecular surface patch corresponds to the active site of the enzyme. Inset: plot of ligands molecular weight (MW) vs. their rank.
Note that the Q-MOL protein-ligand docking was successful in finding biologically active ligands for all 3 targeted allosteric sites. The complete NCI DTP library was used for docking. The primary Q-MOL VLS converged to 84, 87 and 88 top individual hits for sites 1, 2 and 3, respectively. After visual inspection, 14, 37 and 36 ligands were selected for ordering from NCI DTP for sites 1, 2 and 3, respectively. Out of which, 7 (50%), 15 (41%) and 18 (50%) ligands exhibited activity for sites 1, 2 and 3, respectively. The docking Site 3, structurally similar to West Nile virus NS2B-NS3 docking site, produced the largest number of biologically active potent small molecule ligands.
The validated hits obtained for Site 3 after primary Q-MOL VLS were further computationally optimized, and new analogues re-tested (figure below).
As a result, the most potent of identified inhibitors for Site 3, NSC704342 had wild-type HCV NS3/4A IC50 = 183 nM (compare to telaprevir IC50 = 148 nM). NSC704342 also maintained its potency across the panel of most common HCV NS3/4A mutations.
The table below summarizes the results of a single round of Q-MOL computational optimization:
What is Q-MOL computational optimization?
The primary VLS hits are biologically validated ligands, predicted after the very first round of VLS when a complete compound library was used for the docking. Because each of the primary hits targets diverse protein conformations in vicinity of some local minima, further computational optimization is necessary to converge on ligands targeting near apparent native receptor conformations. To this end, the chemical space of each primary hit is "extended" by searching for analogues in SDF compound library. The similarity between structures is measured by computing distance between proprietary Q-MOL chemical fingerprints which are based on protein-ligand docking parameters. For each hit, closest 256 analogues are retrieved, and combined together into a single ligands pool. All of the new ligands are than docked and sorted by relative binding energy. Because these ligands represent a structural subset that targets a conformational ensemble near apparent native state of a receptor, the top (lowest energy) hits are than selected for biological evaluation. See the Q-MOL publication pre-print.
Anton Cheltsov Ph.D.