Site 4 is found on the interface of S1 and S2 as shown in (
Figure 2) and numerous entry inhibitors have been modeled there by various groups. Nelfinavir was one of the first (May, 2020) small molecules that was reported to bind at this site according to molecular docking [
70]. Subsequently (Sept, 2020) other entry inhibitors such as toremifene have also been proposed to bind at Site 4 according to molecular docking [
71]. Interestingly, prior to the pandemic it was known that some natural product saponin derivatives (glycyrrhizin, aescin, α-hederin) exhibited entry inhibitor activity [
72], and several new series of saponin derivative natural products have been shown to be SARS CoV-2 inhibitors [
73,
74]. As discussed previously, OA saponin derivative
12f was found to bind to the S2 subunit by SPR [
65]. Similar in structure ursolic acid derivatives, such as
UA-30 have also been reported to bind at the Site 4 S1/S2 interface site [
73]. In a new series of 3-O-b-chacotriosyl ursolic acid derivatives, the derivative
UA-18 was shown to bind to Spike by SPR and is shown to be best modeled as binding to this S1 and S2 interface site [
73]. In this study, the double mutants (N764A/R765A), (Q957A/K964A), and quadruple mutant (N764A/R765A/Q957A/K964A) [
73] were used to confirm the
UA-18 binding site. Interestingly, in another study of betulonic acid saponin derivatives,
BA-4 is a representative molecule that shows strong binding to S2 by SPR and does not bind to S1 [
74]. Shown in
Figure 3 is an example model of OA saponins binding this site, very similar to published models [
73,
74]. However, molecular modeling studies from our laboratory has also concluded that at some specific OA saponins such as
12a,
12f, are predicted to bind more favorably at the arbidol Site 5 rather than Site 4 [
63]. In summary, toremifene, nelfinavir and a variety of natural product saponin derivatives are proposed to bind at Site 4 and presumably small molecule binding at this site acts to stabilize the prefusion conformation.
Moving our attention to another proposed binding site, as shown in
Figure 1, Site 6 is another proposed binding site on the S2 segment of the SARS CoV-2 Spike is analogous to the proposed camphecene binding site on influenza HA2 [
75] confirmed by mutagenesis and resistance mutations [
76,
77]. For a novel series of borneol ester derivatives, Yarovaya et al., identified that their derivatives
11 and
21 with the greatest activity in pseudovirus infections in both original Wuhan strain and Delta (B.1.617.2) strains [
78]. The authors conclude that these derivatives
11 and
21 are well-modeled binding to an allosteric site that is analogous to the camphecene binding site on influenza HA [
75]. On the SARS CoV-2 Spike protein, this binding site for borneol esters
11 and
12 is in close proximity to the fusion peptide helical conformation binding site in the prefusion conformation of Spike [
78]. Key protein-ligand interactions that define this site using Wuhan strain numbering include S2 residues that are just C-terminal of the fusion peptide (F823, L826 and F833) and residues (P1057, H1058) from the S2 connector domain (CD) (res:986-1035). The borneol ester derivative binding site on the SARS CoV-2 Spike is supported from sequence analysis, extensive molecular docking, MD simulations and supported by the rationale of how camphecene derivatives are proposed to bind to influenza HA [
75,
76,
77].
Last, we would also like to highlight Site 7, an additional putative binding site on the S2 segment for both peptides and small molecules. This Site 7 on the S2 subunit has been reported to be thermodynamically favorable for binding small molecules in several reports [
15,
63,
79,
80]. In this binding site, a key hydrophobic contact from residue W886 is an important structural element of a long helix-turn-helix (res:867-890) which is N-terminal to the HR1 and undergoes major conformational changes upon the transition from the prefusion to the postfusion conformation. In the prefusion conformation, this site is formed from W886, four residues from the N-terminus of the CD (Q1036, K1038, V1040, Y1047), and residues R1107 and N1108 which are N-terminal to the SH domain also undergo major refolding in the prefusion to postfusion transition. To our knowledge, the thermodynamic favorability for this site was first reported from a virtual screening study that was focused on the S2 segment that identified chitosan to bind in proximity to this site [
79]. In our study using a pharmacophore mapping, docking and free energy approach, we identified this specific site as one of the most thermodynamically favorable binding sites on the S2 segment in the prefusion conformation [
15,
63] as well as the arbidol site (discussed above). In another study, a fusion inhibitor tripeptide VFI was also proposed to dock to this site, based on docking the VFI tripeptide to the entire Spike protein [
80]. While this site has not yet been validated by mutagenesis, our laboratory and others will aim to see if point mutants effect fusion or small molecule binding. With recent advances in CryoEM characterization of Spike conformational states, it may be possible to characterize small molecule complexes of representative fusion inhibitors bound to Spike. As with other viruses in the past such as Influenza and HIV, prior to small molecule binding sites being experimentally confirmed by either mutagenesis or X-ray crystallography, it is important to remember that there is an important role for molecular modeling approaches to predict likely binding sites and to guide experimental approaches towards confirming them.