Determination of Ligand Binding Modes in Hydrated Viral Ion Channels to Foster Drug Design and Repositioning

Target-based design and repositioning are mainstream strategies of drug discovery. Numerous drug design and repositioning projects have been launched to fight the ongoing COVID-19 pandemic. The resulting drug candidates have often failed due to the misprediction of their target-bound structures. The determination of water positions of such structures is particularly challenging due to the large number of possible drugs and the diversity of their hydration patterns. To answer this challenge and help correct predictions, we introduce a new protocol HydroDock, which can build hydrated drug–target complexes from scratch. HydroDock requires only the dry target and drug structures and produces their complexes with appropriately positioned water molecules. As a test application of the protocol, we built the structures of amantadine derivatives in complex with the influenza M2 transmembrane ion channel. The repositioning of amantadine derivatives from this influenza target to the SARS-CoV-2 envelope protein was also investigated. Excellent agreement was observed between experiments and the structures determined by HydroDock. The atomic resolution complex structures showed that water plays a similar role in the binding of amphipathic amantadine derivatives to transmembrane ion channels of both influenza A and SARS-CoV-2. While the hydrophobic regions of the channels capture the bulky hydrocarbon group of the ligand, the surrounding waters direct its orientation parallel with the axes of the channels via bridging interactions with the ionic ligand head. As HydroDock supplied otherwise undetermined structural details, it can be recommended to improve the reliability of future design and repositioning of antiviral drug candidates and many other ligands with an influence of water structure on their mechanism of action.


Figure S1
The wet docking result of AA to M2A after Refinement S. To include water molecules in docking experiments, H atoms were added and a two-step minimization was applied (Refinement S). After refinement, the H atoms pointed towards the experimental position of ligand AA repelling the positive charge of the aminium moiety of the ligand, resulting in a misdocked AA. Thus, in order to refine the orientation of the H atoms, the two-step minimization in Refinement S was not enough, involvement of a short MD simulation (100 ps) was also necessary (Refinement R). M2A is shown as grey cartoon, A30 is shown as sticks, water molecules are shown as red and white thick lines and AA is shown as teal sticks.

Figure S2
Wet docking of AA to M2A with Gasteiger (A) and TIP3P (B) partial charges on the water molecules. A) Although the H-bonding network of the water molecules was established properly after Refinement R, ligand AA did not get closer to its experimental binding position during docking experiments. AA coordinated to the carbonyl oxygen of V27 with its aminium moiety, instead of the water oxygens (RMSD= 3.12 Å). In this case, water molecules were equipped with Gasteiger-Marsili partial charges which was found an improper partial charge model for water molecules B) Using TIP3P partial charges on water molecules, ligand AA found its binding conformation close to the experimental position (Table 3). TIP3P water partial charges are two times greater in absolute value, than Gasteiger-Marsili ones (Table  below), resulting in a larger attractive Coulomb interaction energy with the waters in the TIP3P case, which explains the differences between the docking results. Although using a uniform charge system is a standard protocol in docking calculations, in the case of water molecules, the use of TIP3P charges can be recommended even if the solutes have a different charge system like Gasteiger-Marsili. M2A is shown as grey cartoon, refined water molecules are shown as red and white thick lines, V27 is shown as sticks in A figure and AA is shown as teal sticks.

Figure S3
The match between crystallographic reference water positions (red spheres) and the predicted water positions (blue spheres) of system 6bkk in the M2A (grey cartoon) ion channel (values of deviations are listed in Table 2). An overall excellent match was found, with an exception of D:w109. This mismatch is not surprising, as the O-O distance of D:w109 and B:w208 is 2.4 Å and that of D:w109 and D:w105 is 2.6 Å indicating (non-reproducible) close contacts in the experimental structure. Water numbering follows that of PDB 6BKK.

Figure S4
The movement of RA (teal sticks) during MD simulation of the 4 th step of HydroDock in the M2A channel (apo). 4 frames are shown of the MD simulation: A) the starting (1 st step of HydroDock, dry docked ligand binding mode), B) the 1 st , C) the 100 th and D) the 354 th , representative binding mode after HydroDock ( Table 4). The rotational movement of RA is seen from an initially head-to-tail binding mode to the final representative binding mode of HydroDock. During the rotation the H-bonds of the aminium moiety of RA rearrange, which is mediated by water molecules from the 2 nd step of HydroDock.

Figure S5
The orientation of ligand AA (binding mode AA3) after dry docking (A, Supp. Table 5) and HydroDock (B, Supp.  (Fig. 6b). EC2 is shown as grey cartoon, interacting amino acids are shown as all atom representation grey sticks, and AA is shown as teal sticks. Water molecules are shown as red and white sticks.

Figure S6
RA1 and RA2 dry docked binding modes converged to a common position in the EC2 channel (grey cartoon) during MD simulations in Step 4 of HydroDock. A) The two drydocked binding modes of RA (grey sticks) provided two distinct starting points for the MD simulations. B) The two representative binding modes of RA converged during MD simulations, and from two distinct binding modes took up final binding modes that are close to each other (teal sticks). Waters are not shown.

Table S1
Comparison of holo (6BKK) and apo (3LBW) water structures. Protein superposition RMSD: 0.4 Å. Water molecules were considered identical within a match threshold of 1.0 Å, exceptions are marked. In the holo system there were 10 water oxygen positions within 5.0 Å of the ligand, when compared to the 6 water oxygens of the apo systems, which was the reason to use the holo system as reference.
Holo water # Apo water # Distance (Å) A:w102 A:w100 0.5 C:w208 a The line connecting water molecules D:w122 and B:w112 in the apo structure is turned by 90º around the vertical axis of M2A, when compared to that of the water molecules of the holo structure, which results in a slightly elevated distance.

Table S4
The statistics of generation of the ligand conformation pool of Table 4 (step 5 of  HydroDock). The conformational pools resulted after the MD simulations (step 4 of HydroDock) contain different number of ligands each. The length of the MD simulations varies from 40 to 100 ns, ligand positions are written every 0.1 ns. Using HydroDock method we suggest a minimum of 40 ns of MD simulations (for a system of the same size as 6BKK), as after reaching this limit, the length of the MD simulation has no significant improving effect on the results in the case of drug-like ligands. For RA somewhat longer simulations were carried out compared to AA and SA, to see if the head-to-tail conformation would eventually turn over in the case of the holo system.     a Dissociated after 23.6 ns, the representative structure was selected from the first 236 MD snapshots. b Dissociated after 31.8 ns, the representative structure was selected from the first 318 MD snapshots.