Mapping major SARS-CoV-2 drug targets and assessment of druggability using computational fragment screening: Identification of an allosteric small-molecule binding site on the Nsp13 helicase

Abstract

The 2019 emergence of, SARS-CoV-2 has tragically taken an immense toll on human life and far reaching impacts on society. There is a need to identify effective antivirals with diverse mechanisms of action in order to accelerate preclinical development. This study focused on five of the most established drug target proteins for direct acting small molecule antivirals: Nsp5 Main Protease, Nsp12 RNA-dependent RNA polymerase, Nsp13 Helicase, Nsp16 2'-O methyltransferase and the S2 subunit of the Spike protein. A workflow of solvent mapping and free energy calculations was used to identify and characterize favorable small-molecule binding sites for an aromatic pharmacophore (benzene). After identifying the most favorable sites, calculated ligand efficiencies were compared utilizing computational fragment screening. The most favorable sites overall were located on Nsp12 and Nsp16, whereas the most favorable sites for Nsp13 and S2 Spike had comparatively lower ligand efficiencies relative to Nsp12 and Nsp16. Utilizing fragment screening on numerous possible sites on Nsp13 helicase, we identified a favorable allosteric site on the N-terminal zinc binding domain (ZBD) that may be amenable to virtual or biophysical fragment screening efforts. Recent structural studies of the Nsp12:Nsp13 replication-transcription complex experimentally corroborates ligand binding at this site, which is revealed to be a functional Nsp8:Nsp13 protein-protein interaction site in the complex. Detailed structural analysis of Nsp13 ZBD conformations show the role of induced-fit flexibility in this ligand binding site and identify which conformational states are associated with efficient ligand binding. We hope that this map of over 200 possible small-molecule binding sites for these drug targets may be of use for ongoing discovery, design, and drug repurposing efforts. This information may be used to prioritize screening efforts or aid in the process of deciphering how a screening hit may bind to a specific target protein.

Fig 1. Major SARS-CoV-2 drug target proteins analyzed using pharmacophore mapping and computational fragment screening.

(A) Nsp5 Mprot (B) Nsp16 2’-O MT, (C) Nsp12 RdRp (D) Nsp13 Helicase (E) S2 Spike. Protein structures are shown using ribbon diagram with RGB (red-green-blue) rainbow coloring from the N-terminus (blue) to the C-terminus (red).

Fig 2. Overview of pharmacophore mapping and computational fragment screening.

(A) Nsp13 helicase solvent mapping simulations with benzene solvent. (B) TOP50 sites are identified from analysis (hydrophobic contacts & geometric clustering). (C) CHARMM-based free energy calculations are utilized to calculate the free energy of the TOP50 sites and re-rank the list of aromatic pharmacophore sites by ΔG. Site 01 for the Nsp13 helicase is shown (red). (D) Fragment library screening is performed using the Site 01 aromatic pharmacophore a center-of-mass reference. Ligand efficiencies are then calculated from fragment screening data.

Fig 3. Aromatic heterocycle fragment replacement library.

Additional details in supplementary information.

Fig 4. Calculated ligand efficiency comparisons of aromatic pharmacophore sites.

The data is shown plotted focusing on the rank order of the first 15 most favorable sites (A) or over all of the sites characterized (B).

Fig 5. Ligand efficiency comparisons from computational fragment screening.

Data is shown on the same Y-axis scale for free energy (ΔG_bind) and ligand efficiency comparing numerous sites on different target proteins using the FRAG100 library as a benchmark: (A) Nsp5 Mpro (B) Nsp16 2’-O MT (C) Nsp12 RdRp (D) Nsp13 Helicase (E) S2 Spike.

Fig 6. Pharmacophore mapping successfully identifies Nsp5 Mpro inhibitor substrate recognition site pharmacophores.

The crystal structure of Nsp5 Mpro (6W63.pdb) bound to a broad spectrum non-covalent inhibitor is shown using ribbon diagram with rainbow coloring (A) and is shown with a transparent gray surface in (B). The three protease P1, P2, and P3 substrate recognition binding pockets were identified with favorable aromatic pharmacophores as shown for Site 01, Site 05 and Site 04 respectively (C). Several crystal structures of other fragment ligands (6YNQ.pdb, 5RGZ.pdb, 5RF3.pdb, 6YNQ.pdb, 5R81.pdb) independently confirm the position of the Site 01 (P1) and Site 05 (P2) pharmacophores shown in (D) thru (I).

Fig 7. Experimental structures from fragment screening confirm other Nsp5 Mpro “minor” binding sites successfully predicted.

The crystal structure of Nsp5 Mpro (6W63.pdb) is shown illustrating the reverse or “minor” binding surface using ribbon diagram with rainbow coloring (A) and is shown with a transparent gray surface in (B). Several crystal structures of other fragment ligands (5RFC.pdb, 5REE.pdb, 5REG.pdb) independently confirm the positions of the “minor” binding sites: Site 02, Site 11 and Site 23 pharmacophores shown in (C) and (D).

Fig 8. Pharmacophore mapping successfully identifies the Nsp16 sinefungin binding site and other PPI sites.

The most favorable two sites identified are shown on a ribbon diagram in (A) and on a surface representation in (B) where Site 01 (red) was the most favorable site identified that was proximal to Site 02 (magenta), which superimposes with the most favorable pharmacophore site in the bound sinefungin structure. Two protein-protein interaction (PPI) sites on the Nsp16 binding surface were identified as favorable aromatic pharmacophore binding sites shown in (C) and (D). Site 03 (red) and Site 06 (magenta) represent Nsp16 / Nsp10 heterodimer interaction sites where the Nsp10 ribbon is shown (E) in rainbow binding to the complementary Nsp16 surface shown in gray.

Fig 9. Favorable aromatic pharmacophore binding sites identified on the Nsp12 RdRp.

Favorable sites are shown in red or magenta on the rainbow ribbon structure shown in (A) and on the gray binding surface shown in (B). Several protein-protein interaction (PPI) sites on the Nsp12 binding surface were identified as favorable aromatic pharmacophore binding sites shown in (B). Site 01 (red), Site 08 and Site 20 (magenta) represent Nsp12 / Nsp8 heterodimer interaction sites where the Nsp8 ribbon is shown (B) in rainbow binding to the complementary Nsp12 surface shown in gray. Zoom in surface views are shown to illustrate corresponding Nsp8 hydrophobic residues side chains that participate in PPI on the Nsp12 surface.

Fig 10. Domain structure and favorable aromatic pharmacophore sites on the Nsp13 helicase.

Nsp13 structure is shown in a rainbow ribbon diagram in (A) where the N-terminal zinc-binding-domain (ZBD) is colored blue and the C-terminal REC2A domain is colored red. Favorable binding sites Site 01, Site 19 and Site 07 are highlighted in magenta and labeled. The location of RNA and ADP binding are modeled onto the structure via a structural alignment of another eukaryotic helicase (2xzl.pdb) bound to RNA and ADP. In (B) the root-mean-squared-deviation (RMSD) calculated between two recent Nsp13 helicase structures captured in different conformational states in a CryoEM structure of a complete SARS-CoV-2 replicase complex is shown in (blue). A similar RMSD as a function of residue plot is shown for 500 ps of standard molecular dynamics (MD) simulation of the (6jyt.pdb) structure at 310 K in all-atom solvent (black). Favorable binding sites Site 01, Site 19 and Site 07 are highlighted in magenta and shown bound to the Nsp13 helicase surface in (C) and (D) where Site 01 is the most favorable site on the entire Nsp13 helicase found on the N-terminal ZBD (C), where Site 07 was the most favorable site identified in the vicinity of the ATPase active site (D).

Fig 11. Ligand efficiency for several favorable binding sites on Nsp5 Mpro and Nsp13 helicase from computational fragment screening.

Data is shown for ΔG_bind (A) and ligand efficiency (B) comparing numerous sites using the FRAG100 library as a benchmark. Ligand efficiency data is also shown as a function of fragment molecular weight (C) and calculated LogP (D).

Fig 12. Ligand efficiency for favorable binding sites on Nsp5 Mpro and Nsp13 helicase from computational fragment screening.

Data is shown for ligand efficiency (A) and (B) comparing several sites using the entire FRAG3700 library. Ligand efficiency data is also shown as a function of fragment molecular weight (C) and calculated LogP (D).

Fig 13. Structure of the Nsp12: Nsp13 replication transcription complex reveals that the Nsp13 zinc-binding-domain (ZBD) mediates protein-protein interactions.

(A) A ribbon diagram of the structure of the complex (6xez.pdb) [61] shows how two subunits of the helicase form protein-protein interactions with both Nsp8 (red) and Nsp12 in the active complex through interactions with the helicase ZBD domain (blue). (B) A zoom in view shows that the previously predicted Site 01 on the ZBD (magenta) is proximal (within 5–6 Å) to the Nsp8:Nsp13 protein-protein interaction site in the replication complex.

Fig 14. Nsp13 helicase allosteric zinc-binding-domain (ZBD) small-molecule binding site.

The ZBD structures of the flexible residues forming Site 01 are shown on a ribbon diagram in (A) and the residues involved in Site 01 and CHAPSO binding are labeled. A surface diagram (B) shows the complementary surface of the ZBD (6xezE.pdb) conformational state captured binding to CHAPSO. The most favorable aromatic pharmacophore groups identified on this surface were labeled “A” (yellow) and “B” (blue). A surface diagram is shown in (C) showing the complementary surface of the ZBD on the new crystal structure (6zslA.pdb) where Site 01 is also found to be extremely favorable, ligand efficient (LE = 0.35) and clashing with the CHAPSO binding site. Cross-docking analysis calculating ligand efficiencies for these pharmacophore sites (D) show that previously identified Site 01 is the most favorable of all of these at the CHAPSO binding site. Site 01 exhibits the highest ligand efficiency in three independent structures [01_6jytA, 04_6xezF, and 05_6zslA]. Ligand efficiency from computational fragment screening (FRAG1000) shown in (E) demonstrates that Site 01 is more favorable for ligand binding in the 04_6xezF and 05_6zslA ZBD conformational states, compared to the CHAPSO 03_6xezE “B” (blue) pharmacophore site.

Fig 15. Nsp13 helicase ZBD allosteric Site 01 is more favorable than ZBD protein-protein interaction (PPI) sites.

Pharmacophore Site 01 (red), Site 14 (cyan) and Site 49 (green) are shown on a surface diagram (A) of the ZBD. PPI1 (Site 14) overlaps with Nsp12-Nsp13 (B) and Nsp8-Nsp13 (C) PPI sites, where PPI2 (Site 49) overlaps with residue M62 of Nsp8 on the Nsp8-Nsp13 complex of the new replicase complex. Ligand efficiency from computational fragment screening (FRAG100) shown in (D) demonstrates that Site 01 (red) is more favorable than both of the PPI sites (cyan) and (green). Site 01 (red) was found to be most favorable for ligand binding in two new independently determined SARS-CoV-2 [04_6xezF] and [05_6zslA] Nsp13 experimental structures.

Fig 16. Favorable aromatic pharmacophore binding sites identified on the S2 spike protein.

Favorable Sites 01, 02, 03, 05 and 07 are shown in magenta sites on a rainbow ribbon diagram in (A) and on a surface representation in (B). Important hydrophobic residues forming these sites are shown for Site 01 (C) Site 02 (D) Site 03 (E) Site 05 (F) and Site 07 (G).

Fig 17. Ligand efficiencies from pharmacophore mapping of the S2 spike protein as a function of residue.

All residues identified as having a high druggability and sequence-conservation by [120] are shown as small gray or black squares at the top of the figure. Black indicates a higher sequence conservation compared to gray between more distantly related coronaviruses (SARS, MERs, etc). Ligand efficiencies of the TOP50 sites (according to residues provided S1 Table in S1 File) are shown as either blue or red dots as a function of residue. Blue dots are pharmacophore sites (and corresponding residues) not identified in the analysis of Trigueiro-Louro et al., [120] whereas red dots indicate binding site residues identified as a Union of both datasets. The residues that define pharmacophore Sites 01, 02, 03, 05, and 07 exhibit high relative ligand efficiency to other sites on the S2 Spike protein and were also identified by Trigueiro-Louro et al., [120].