Rimantadine Binds to and Inhibits the Influenza A M2 Proton Channel without Enantiomeric Specificity

Abstract

The influenza A M2 wild-type (WT) proton channel is the target of the anti-influenza drug rimantadine. Rimantadine has two enantiomers, though most investigations into drug binding and inhibition have used a racemic mixture. Solid-state NMR experiments using the full length-M2 WT have shown significant spectral differences that were interpreted to indicate tighter binding for (R)- vs (S)-rimantadine. However, it was unclear if this correlates with a functional difference in drug binding and inhibition. Using X-ray crystallography, we have determined that both (R)- and (S)-rimantadine bind to the M2 WT pore with slight differences in the hydration of each enantiomer. However, this does not result in a difference in potency or binding kinetics, as shown by similar values for k_on, k_off, and K_d in electrophysiological assays and for EC₅₀ values in cellular assays. We concluded that the slight differences in hydration for the (R)- and (S)-rimantadine enantiomers are not relevant to drug binding or channel inhibition. To further explore the effect of the hydration of the M2 pore on binding affinity, the water structure was evaluated by grand canonical ensemble molecular dynamics simulations as a function of the chemical potential of the water. Initially, the two layers of ordered water molecules between the bound drug and the channel’s gating His37 residues mask the drug’s chirality. As the chemical potential becomes more unfavorable, the drug translocates down to the lower water layer, and the interaction becomes more sensitive to chirality. These studies suggest the feasibility of displacing the upper water layer and specifically recognizing the lower water layers in novel drugs.

Graphical Abstract

The influenza A matrix 2 (M2) protein is the most extensively studied viroporin and is currently considered the most well-validated viroporin antiviral drug target.¹ It is an attractive target for anti-influenza medications because it mutates less frequently than other proteins within the virus.² However, though the M2 channel mutates infrequently, drug-resistant mutations have become prevalent since the early 2000s, and currently, amantadine and rimantadine, which block M2 wild-type (WT) ptoton function, are no longer recommended for treatment of influenza infections.^3,4 The rise of adamantane-resistant influenza necessitates an in-depth understanding of the stereochemical requirements for M2 channel blockers so that new drugs can be developed.

The M2 protein is a homotetrameric proton channel^5–8 whose minimally functional transmembrane core consists of residues 22–46. ^9,10 The gating residues involved in proton conductance are His37 and Trp41.^11,12 After an influenza virus particle is endocytosed by the host cell, proton transport through the M2 channel lowers the pH within the viral envelope and allows viral ribonucleoproteins (RNPs) to unpack from M1.¹³ Blocking proton transport through the M2 WT channel using adamantanes^14,15 prevents this dissociation from happening and thus prevents viral replication from occurring.^16–18 The proton transport function of M2 plays a second role in the viral life cycle; M2 inhibits the acidification of the trans Golgi compartment, thereby preventing premature activation of the acid-sensitive hemagglutinin protein.^19,20

Structural biology studies of M2 WT-drug complexes have been performed using racemic rimantadine,^21–23 but in a solid-state NMR (ssNMR) spectra using deuterium-labeled (R)- and (S)-rimantadine, the authors showed differences in chemical shifts when the two enantiomers bind to full-length M2 WT.²⁴ These spectral differences were considered to arise from differences in potency and affinity of the isomers. However, electrophysiological (EP), isothermal titration calorimetry (ITC), and antiviral assays carried out using enantiomerically pure rimantadine indicated that the two enantiomers have equal potency against M2 WT.^16,25 Additionally, in vivo experiments in mice have shown similar antiviral activity for both enantiomers.²⁶ One possible explanation for the difference in ssNMR chemical shifts for the two rimantadine enantiomers might relate to differences in the kinetics of binding, which are particularly important for the potency of M2 inhibitors. In the earlier EP study examining rimantadine enantiomer binding kinetics to M2 WT, ¹⁶ the time traces were conducted under kinetic control, which did allow sufficient time to determine accurate measurement of k_off (K_diss is the ratio of k_on to k_off). Here we performed a more stringent kinetic study by recording both the binding and washing curves so that we measured more accurately k_off. Using this improved methodology, we did not observe any difference in binding kinetics between the (R)- and (S)- enantiomers, as shown by similar values for k_on, k_off, and K_d in EP assays. We also tested the potency of the two enantiomers against an amantadine-sensitive strain of influenza in an antiviral plaque assay and found that the EC₅₀ values are indistinguishable.

We also reported the crystal structure of M2(22–46) WT in complex with enantiomerically pure (R)- and (S)-rimantadine. Earlier, a crystal structure had been solved for the complex of M2 with racemic rimantadine, but the density could be interpreted equally well for models involving both enantiomers (as would be expected if they bound with equal affinity) or a single high-affinity enantiomer in multiple conformations. We now showed that, although the crystals of each rimantadine enantiomer complex had different space groups and unit cell dimensions, we observed that the electron density corresponding to the bound drug is similar for the two enantiomers. The density indicates a mechanism of binding in which the rimantadine enantiomer ethylamine can rotate to four positions to form hydrogen bonds with ordered waters in the M2 pore. We observed slight differences between the hydration of the two enantiomers within the pore.

Finally, we employed grand canonical ensemble molecular dynamics simulations (GCMC/MD) to provide further insight into the water network present in the rimantadine binding site. GCMC^27–29 can be used in combination with MD simulations in order to bypass the kinetic barriers associated with the binding and unbinding of occluded water sites. This is done by simulating the system at a constant chemical potential, while inserting and deleting waters from a region of interest (defined as the rimantadine enantiomer binding site, in this case). Additionally, carrying out these simulations at a range of chemical potential values, referred to as GCMC titration, allowed the binding free energy of the water network to be rigorously calculated. It is of interest that the titration analysis indicates that a chiral difference is apparent when the upper water layer of the binding site is displaced compared to the two layers, which are sufficiently flexible to eliminate a chiral preference. To validate our GCMC/MD simulations, we first calculated the relative binding free energy of the (R)- and (S) enantiomers to the M2 WT pore.

RESULTS AND DISCUSSION

X-ray Crystal Structures.

We have solved two X-ray crystal structures of (R)-rimantadine and (S)-rimantadine bound to M2(22–46) WT through cocrystallization experiments (Table S1). Crystals of (R)-rimantadine bound to M2(22–46) WT (PDB ID 6US9) formed at pH 8.5 in the P2₁ space group with unit cell dimensions a, b, c (Å) = 48.18, 48.70, 71.67 and α, β, γ (deg) = 90, 90, 90, and the diffraction was limited to 2.00 Å. Crystals of (S)-rimantadine bound to M2(22–46) (PDB ID 6US8) formed at pH 7.5 in the P2₁2₁2₁ space group with unit cell dimensions a, b, c (Å) = 49.39, 76.09, 98.63 and α, β, γ (deg) = 90, 90, 90. These crystals diffracted to a resolution of 1.70 Å, making this the highest-resolution structure of M2 bound to a drug or inhibitor.^30–32

As in previously solved structures of M2(22–46) WT bound to drugs and inhibitors,^31,32 M2(22–46) WT is in the Inward_closed conformation, and we observed electron density corresponding to the bound drug at the channel’s N-terminus near Ser31, with two layers of ordered water molecules between the bound drug and the channel’s gating His37 residues (Figures 1a–d). These waters form hydrogen bonds to pore-facing carbonyl groups from residues Ala30 and Gly34 and also form a vertical hydrogen bonding network leading down to His37. The X-ray crystal structures remained unchanged during 300 ns MD simulations in hydrated POPC bilayers (Figures 1c,d).

Figure 1.

(a, b) Polder maps are shown here to a contour of 3 σ (dark green mesh) for the X-ray crystal structures of (R)-rimantadine (a) and (S)-rimantadine (b) bound to M2(22–46) WT. Front and back monomer helices have been removed to show the contents of the pore. For both structures, the density corresponding to the rimantadine enantiomer ethylammonium group has been modeled as the average of four rotational conformers. We observed a network of ordered waters between the bound rimantadine enantiomer and the gating His37 residues; the waters form hydrogen bonds with pore-facing carbonyl groups and also form a vertical hydrogen-bond network leading down to the gating His37 residues. (c, d), The radial distribution function (RDF) g(r) for (R)-rimantadine (c) and (S)-rimantadine (d) between the ammonium group and water oxygen atoms from the 300 ns MD simulation shows a strong peak at 2.7 Å, which corresponds to the upper waters layer. The cumulative integrated intensity ∫g(r), shown in the right axis, at 2.7 Å indicates ca. 4 waters which form hydrogen bonds with the ammonium group of rimantadine enantiomer and the carbonyl groups of Ala30. The second broad peak at 4.8 Å corresponds to ca. 5–6 waters of the lower layer, which form hydrogen bonds with the His37 imidazole groups and the carbonyl groups of Gly34.

Though enantiomerically pure (R)- or (S)-rimantadine was used in these crystallization trials, the chirality of the drug is not obvious from the electron density in either structure. This is presumably a result of averaging more than one binding position within the crystal lattice. Thus, we have modeled the bound drug for each of the structures as a superposition of four rotational, by rotation of C1_Ad–C_Ethyl bond, conformers, with the drug ammonium group forming hydrogen bonds to two of the four top layer waters in each position. In our previous crystal structure of racemic rimantadine bound to M2(22–46) WT (PDB ID 6BKL),³² we observed this same ambiguity regarding the position of the rimantadine ethylammonium group, which we interpreted as an effect of cocrystallizing using racemic rimantadine.

The occupancies of the four rotational conformers of (R)- or (S)-rimantadine were allowed to float during refinement; alternate conformers were removed from the model if their occupancy was refined to zero. Each structure contains four tetramers per asymmetric unit; the refined occupancies for the four rimantadine rotational conformers for each tetramer are shown in Table S2. To summarize, for the structure of (R)-rimantadine bound to M2(22–46) WT, three conformers were observed per tetramer with some variations in the occupancy of each (R)-rimantadine conformer depending on the tetramer. In the structure of (S)-rimantadine bound to M2(22–46) WT, three of the tetramers have relatively high (>0.30) occupancy for two of the (S)-rimantadine conformers, with the other two conformers having either zero or <0.05 occupancy. The fourth tetramer has low occupancy (0.01 and 0.11) for two of the conformers and relatively high occupancy (>0.40) for the other two. These data suggest that the binding mechanism remains the same for both enantiomers with the observed slight differences likely due to the hydration of each enantiomer as shown in Figure 1. We subsequently explored the hydration of each enantiomer inside the M2 WT pore using GCMC/MD simulations.

In addition to the density associated with (S)-rimantadine in the pore, we unexpectedly observed electron density that we provisionally assigned to rimantadine in the structure of the complex at 1.7 Å. The additional (S)-rimantadine lies at locations that stabilize the packing of individual tetramers in the crystal lattice (Figures S1 and S2). The two tetramers have an antiparallel orientation, and the rimantadine enantiomer makes contact with residues near the N-terminus of one tetramer and residues near the C-terminus of the other tetramer. The location of the ethylamonnium group is not clear from the electron density, so we have only modeled the adamantyl group. The C-terminal contacts are similar to those proposed for a weak, exterior site observed in solution and ssNMR studies that indicate binding of the adamantanes to the exterior of the channel’s C-terminus at high concentrations of drug (including Leu 40, Ile42, Asp44, Arg45, and Leu46).^33,34 However, these previous spectroscopic studies did not identify contacts near the N-terminus (Leu26, Val28, Ala29, Ile32). Thus, it is likely that the exterior site seen here is a crystallographic artifact.

Electrophysiology and Antiviral Plaque Assays.

To study whether chirality has an effect on the functional activity against the M2 WT channel, we determined the association constant (k_on), the dissociation constant (k_off), and the binding affinity (K_d) for (R)- and (S)-rimantadine in a two-electrode voltage clamp electrophysiological (TEVC) assay using kinetic studies.^35–37

Drugs bind slowly to the channel, and dissociate very slowly, so special procedures are required to determine the on and off rates. We first applied relatively high concentrations (100 μM) of the test compound to fully inhibit the channel. The second-order rate constant for inhibition (k_on) was determined from the time course of inhibition. The drug is then washed out, and the time required to recover from inhibition was measured to determine k_off. Full-length M2 WT from the amantadine-sensitive Udorn strain was expressed in oocytes, and a low pH 5.5 solution was applied to activate the M2 channel. Next, a pH 5.5 solution containing the testing compound was applied to inhibit the M2 channel. Once the current reached the steady state, compound dissociation was initiated by changing the oocyte bathing solution to pH 5.5 without the drug. During the washout, a few pH 8.5 pulses were applied to make sure the current went to baseline to ensure the oocyte quality. The recording traces are shown in Figure 2.

Figure 2.

Rimantadine enantiomers and amantadine binding kinetics against Udorn M2 WT were determined using a combined application and washout procedure TEVC assay. (A) Racemic rimantadine, (B) (R)-rimantadine, (C) (S)-rimantadine, or (D) amantadine was applied to oocytes for 5–7 min after the inward current reached its maximum; then a washout protocol was applied to the oocytes. During the washout, pH 8.5 pulses were applied to make sure the current went to baseline to ensure the oocyte quality. The blue bar above the recording trace indicates the period in which the pH 5.5 Barth solution was applied; the red bar indicates the period in which compounds in pH 5.5 Barth solution was applied. Representative recording traces are shown on the left side of each figure. Data extracted from the recording traces were plotted with an association then dissociation equation in GraphPad Prism 5 as shown on the right side of each figure. The best-fit values are shown in Table 1.

Fitting the binding and washing curves with association and dissociation equations yielded the k_on, k_off, and K_d values for (R)-, (S)-, and racemic rimantadine.³⁷ As shown in Table 1, the binding kinetics parameter (k_on, k_off, and K_d) values for (R)-, (S)- and racemic rimantadine are not significantly different, which suggests that both the (R)- and (S)-enantiomers of rimantadine bind to the M2 channel with equal potency. We also included amantadine as a control: the k_on of amantadine is comparable with that of rimantadine; however, the calculated k_off of amantadine is more than 10-fold faster than that of (R)- and (S)-enantiomers of rimantadine. Therefore, the binding affinity of amantadine to Udorn M2 WT is about 10-fold weaker than rimantadine to M2 WT. Overall, there is no significant difference among the K_d values for (R)- and (S)-rimantadine as well as the racemic rimantadine. We note that a previous study¹⁶ reported 100-fold higher k_off values, resulting in significantly higher K_d values for (R)- and (S)-enantiomers and the racemic mixture of rimantadine with 3.2, 3.9, and 7 μM, respectively, as opposed to the values of 41, 39, and 46 nM, respectively, reported here. In the previous study,¹⁶ only the binding curve was recorded and used for data fitting to derive both k_on and k_off, which might result in significant variations of k_off values (personal communication with Dr. David Busath, the lead researcher of the EP assay in the previous study); the time traces were not recorded for sufficient time to allow accurate measurement of k_off, and thus, the previous values are an upper limit of the true K_d value. In contrast, we performed more stringent kinetic studies by recording both the binding and washing curves in the current study. As such, the k_off values were accurately quantified from the washing curve, which was then used to fit the binding curve to determine k_on. The validity of this methodology in determining the K_d values of M2-S31N and M2-V27A inhibitors has been demonstrated in several studies, which showed a positive correlation between the K_d values and the cellular antiviral activity.^35–37

Table 1.

Summary of the Binding Affinity of Amantadine and Rimantadine Enantiomers against Udorn M2 WT

	rimantadine (racemic)	(R)-rimantadine	(S)-rimantadine	amantadine
concentration tested	50 μM	50 μM	50 μM	100 μM
kon (min−1 M−1)	19600 ± 300	20800 ± 700	22500 ± 300	20500 ± 300
koff (min−1)	(9.1 ± 0.8) × 10−4	(9 ± 2) × 10−4	(8.8 ± 0.8) × 10−4	(119 ± 2) × 10−4
Kd = koff/kon (nM)	46 ± 4	41 ± 9	39 ± 4	580 ± 20

To further confirm that the (R)- and (S)-enantiomers of rimantadine have indistinguishable potency in blocking the M2 channel, we tested these compounds against the amantadine-sensitive A/Soloman Island/3/2006 (H1N1) strain (Figure 3). As expected,²⁷ both (R)- and (S)-rimantadine inhibit viral replication with EC₅₀ values of 19.62 and 24.44 nM, respectively. Taken together, results from both the electro-physiological assay and the antiviral plaque assay showed that both (R)- and (S)-rimantadine have equal potency in blocking the M2 WT channel.

Figure 3.

Cellular antiviral assay results of (R)- and (S)-rimantadine against the amantadine-sensitive A/Soloman Island/3/2006 (H1N1) strain. The antiviral potency was determined in a plaque assay. The EC₅₀ values are the mean ± standard deviation of two independent repeats.

Binding Free Energy Calculations.

To validate our experimental results, the relative binding free energies of binding to M2(22-46) WT and hydration of the (R)- and (S)-rimantadine were calculated using the free energy perturbation method^38,39 (see Figure S3) coupled with MD simulations (FEP/MD) and the multistate Bennett acceptance ratio (MBAR) method for processing the free energy data.^40,41 The free energy perturbations carried out showed that the hydration-free energy difference between the two rimantadine enantiomers (where a positive value would indicate that the (R)-rimantadine is favored, and vice versa) is −0.06 ± 0.03 kcal mol⁻¹, when starting the calculation from the (R)-rimantadine, and 0.000 ± 0.005 kcal mol⁻¹, when starting from the (S)-rimantadine (the uncertainties given represent the standard errors over the three runs). Given that bulk water is an achiral environment, the true value of this hydration free energy difference is zero, and these values, therefore, appeared correct. This result, therefore, supports the view that the free energy protocol used (see Supporting Information) is suitable for this perturbation.

The relative binding free energy difference between the two rimantadine enantiomers was calculated as +0.29 ± 0.04 kcal mol⁻¹ and +0.33 ± 0.06 kcal mol⁻¹, when starting the calculations from the (R)- or (S)-enantiomer, respectively, and the accuracy of the free energy calculations from molecular simulations is typically considered to be around 1 kcal mol⁻¹.⁴² These values show also that the two rimantadine enantiomers bind equally to M2TM WT without any chiral preference. These FEP/MD calculation results with the MBAR method are also in agreement with previous measurements of enantiomers binding affinities using ITC and calculation of their relative free energy of hydration⁴³ and relative binding free energy to M2(22-46) WT using FEP/MD simulations with the BAR method and different alchemical intermediates.^16,44

GCMC Titration.

In the experimental structures, the number of waters in the upper layer (i.e., close to Ala30) is 4 and in the lower layer (i.e., close to Gly34) is 5 (PDB ID 6US9) or 6 (PDB ID 6US8).³²

The GCMC titration calculations involve progressively reducing the chemical potential applied to the water molecules in the ligand-binding site. As the chemical potential is reduced, the more weakly bound water molecules are removed from the binding site, until eventually an entirely dry site results (Figure 4). In addition, the binding free energy of the water network may be calculated from the water occupancy as a function of chemical potential data. The resulting free energy data as a function of water occupancy may be used to determine the equilibrium hydration level, which corresponds to the minimum on the curve of the free energy profile (Figure 5).

Figure 4.

Comparison of the two titration plots for M2 WT in complex with rimantadine enantiomer with the average number of waters observed plotted against the B value (water-binding free energy profiles). The curve was fitted as the mean from 1000 bootstraps of a sum of four sigmoid curves. Results for rimantadine enantiomers are shown as inset images added to show representative structures from different points on the curve. The (R)-rimantadine is shown in cyan, and the (S)-rimantadine is shown in pink. The m = 4 in the legend refers to the fact that a sum of four sigmoid functions was fitted to the data.

Figure 5.

Water binding free energy plots for both rimantadine enantiomers.

For both of the rimantadine enantiomers, the free energy profile (Figure 5) showed a minimum for N = 9, indicating that this is the most favorable number of water molecules in the binding site, in agreement with our crystal structures. Also, in agreement with the experiments is that there was no significant difference in the calculated binding energy for the (R)-enantiomer(−30 ± 1 kcal mol⁻¹), versus the (S)-enantiomer (−29 ± 1 kcal mol⁻¹). Interestingly, the free energy change associated with adding a tenth water to the binding site for the (R)- and (S)-enantiomers are +0.1 ± 0.1 and +0.3 ± 0.2 kcal mol⁻¹, respectively, indicating that the N = 10 state is only slightly less stable at equilibrium.

We next asked whether there might be a difference if the waters were virtually “squeezed out” at a low potential. This can be observed by a selection of simulation snapshots shown in Figure 6 and the violin plots shown in Figure 7, which show the distribution of the number of waters observed (via their z-coordinate, given that this is the normal to the membrane), as well as that of the ligand position within the channel.

Figure 6.

A selection of representative frames from the simulations performed, showing the structural trends as the waters are removed from the GCMC region. The (R)-rimantadine is shown in cyan, and the (S)-rimantadine is shown in pink.

Figure 7.

Violin plots depicting (a) the water distributions observed at each Adams value during the titration calculations; each water point is considered as its z-coordinate (the z-axis is perpendicular to the membrane), relative to the mean z-coordinate of the C_α atoms of the four His37 residues. (b) The ligand positions observed at each Adams value during the titration calculations; each point is considered as the z-coordinate (the z-axis is perpendicular to the membrane) of the nitrogen atom of the ligand, relative to the mean z-coordinate of the C_α atoms of the four His37 residues. The data for the (R)- and (S)-enantiomer are shown in blue and red, respectively. The plots are normalized such that all violins have the same width.

The general trend observed was that, as the chemical potential or B value decreases and waters are removed from the GCMC region, waters from the upper layer, i.e., close to Ala30, are removed first (see Figure 7a), and the ligand gradually descends further into the protein to interact directly with the lower layer, i.e., close to Gly34 (Figure 6b,f), and then the His37 residues (Figures 6a,e and 7b). It becomes clear that the (R)-rimantadine is more resistant to dehydration (see Figures 4, 6b,f, and 7a) and descended further into the channel than the (S)-rimantadine (see Figures 6a,e and 7b), and the data shown in Figure S5 showed that the (R)-enantiomer appears to show a greater propensity for hydrogen bonding with the His37 residues (see also additional discussion in the Supporting Information). In parallel, when less water is in the binding site, the difference in the computed water network binding free energies between the two enantiomers becomes more significant. For example, a network of 4 water molecules that form the lower water layer appeared to be 2.6 ± 0.6 kcal mol⁻¹ more stable in the presence of the (R)-enantiomer (N = 4 in Table S4). Thus, there is a hypothetical stereochemical difference when the upper water layer is removed from the binding site.

The free energy results obtained also imply that it would be very difficult for a ligand to gain a significant affinity increase by displacing the lower water layer. Given that the binding free energy of the network of 9–10 waters is in the vicinity of ca. 30 kcal mol⁻¹, displacement of the upper layer would result in a network of 4 waters, with a binding free energy in the vicinity of 20 kcal mol⁻¹ (Table S4). This suggests a significant thermodynamic cost associated with the displacement of the lower water layer, in addition to the cost of displacing the upper layer. This is consistent with previous studies of drugs such as spiro-adamantyl amines, which are more extended than rimantadine; these compounds, which bind with high affinity, displace the upper, but not the lower layer of waters in the lumen of the channel.^32,45

At lower levels of hydration, due to the chirality of the protein, there is a difference in terms of the computed stability of the water network, which can differentiate the binding of a chiral ligand. Thus, chirality might be an important factor in designing new, novel drugs, which displace the upper layer of waters and tightly engage to tightly bound the lower water layer.

CONCLUSIONS

The environment within the M2 pore is chiral: l-amino acids form right-handed α-helical monomers, which come together to form a left-handed tetrameric helical bundle. Since this chiral environment has the potential to create enantiospecific binding, it was important to investigate whether the chirality of drugs had an effect on drug binding and inhibition. Though the chirality of the M2 WT protein creates the possibility of preferential binding of rimantadine enantiomers, we did not observe this in X-ray crystal structures of M2(22–46) WT bound to (R)- and (S)-rimantadine, in EP kinetics assays using the full-length channel, or in antiviral plaque assays. Instead, the crystal structures showed that both rimantadine enantiomers bind to the M2 pore and that the ethylammonium headgroup of both rimantadine enantiomers averages over four orientations. We observed slight differences in the hydration of each rimantadine enantiomer in the crystal structures. However, these slight differences in hydration did not have an effect on drug binding or channel inhibition. We hypothesize that the two-layer water network, the upper layer adjacent to Ala30 and the lower layer to Gly34, is sufficiently flexible to mask the chirality of the binding site. The GCMC/MD simulation analysis indicated that, as the upper layer of waters is removed, a more substantial chiral difference is observed in the stability of the water networks; i.e., two layers of water are needed to lose this chiral templating effect. This potential enantiomeric selectivity may be of interest for compounds that are able to displace the upper water layer.

MATERIALS AND METHODS

Sample Preparation, Crystallization, Diffraction, and Structure Solving.

Solid-phase peptide synthesis using FMOC chemistry and sample preparation and peptide-drug sample reconstitution into the lipid cubic phase (LCP) were carried out using previously described methods^32,46,47 and the Supporting Information. Data collection was carried out at 100 K at the Advanced Light Source (ALS) beam 8.3.1. Data processing was carried out in Mosflm⁴⁸, and data were scaled in the CCP4 suite.⁴⁹ The structure was solved using molecular replacement in Phaser-MR⁵⁰ with structure PDB ID 6BKL³² as a search model. See the Supporting Information for detailed methods regarding crystallization and data collection.

TEVC Assay.

mRNA syntheses expressing the WT of A/Udorn/72 full M2 protein, oocyte culture, microinjection of oocytes, and electrophysiological TEVC recordings were carried out as previously described.^35,37,51 The K_d measurement and curve fitting were carried out as previously described.^36,37 Briefly, the percentage of current during the application of rimantadine enantiomers and washout protocol was plotted with an association then dissociation equation in GraphPad Prism 5.

Plaque Reduction Assay.

Plaque reduction assays were performed in MDCK cells with the A/Soloman Island/3/2006 (H1N1) virus as previously described.^36,51–53 A/Soloman Island/3/2006 (H1N1) M2 contains the same sequence of M2 WT as A/Udorn/72. Briefly, confluent cells were washed with phosphate buffered saline (PBS) and infected with virus diluted in Dulbecco’s Modified Eagle’s Medium (DMEM) medium supplemented with 0.5% BSA for a final concentration of approximately 100 plaque forming units (PFU) per well. Viral infection was synchronized for 30 min at 4 °C and then incubated for 1 h at 37 °C in a 5% CO₂ atmosphere. The inoculum was aspirated, and cells were washed and incubated in a DMEM overlay media containing different concentrations of compound, 2 μg/mL N-acetyl trypsin, and 1.2% Avicel microcrystalline cellulose (FMC BioPolymer, Philadelphia, PA) at 37 °C in a 5% CO₂ atmosphere. Cells were stained 2 days post infection with 0.2% crystal violet dye. EC₅₀ values were calculated by plotting the plaque area per well against the rimantadine concentration applied with a dose–response function in Prism 5.

Molecular Dynamics Simulations.

The complex of M2(22-46) WT with (S)-rimantadine (PDB ID 6US8) or (R)-rimantadine (PDB ID 6US9) was embedded in a POPC lipid bilayer extending 30 Å beyond the solutes, and the system was placed in an orthorhombic box (90 × 90 × 105 ų). The number of lipids added was ca. 200. The bilayer was then solvated by a 30 Å thick layer of water. Na⁺ and Cl⁻ ions were placed in the water phase to neutralize the systems and to reach the experimental salt concentration of 0.150 M NaCl. The total number of atoms was ca. 80 000. Membrane generation and system solvation was carried out using the “System Builder” utility of Maestro-Desmond 202054, and periodic boundary conditions were applied. We performed 300 ns MD simulations using the amber99sb force field for the protein, CHARMM36 force field to lipids, while the generalized Amber force field (GAFF) was used for the ligand parameterization with the antechamber module of AMBER18 software; intermolecular interactions were calculated with amber99sb force field. The TIP3P force field was used for the water. The MD simulations were performef with the Desmond 2020^55,56 software for each of the two rimantadine enantiomers in complex with M2(22-46) WT (in the X-ray crystal structures PDB ID 6US8 or PDB ID 6US9) embedded in a hydrated lipid bilayer to test the stability of the experimental structures. See the Supporting Information for a detailed description of the simulation conditions, the protocol and related references.

GCMC Titration Calculations.

GCMC method was used to allow the number of waters within a simulation to vary according to the chemical potential, in order to simulate the grand canonical (μVT) ensemble. This allows the waters within the M2 pore to exchange with bulk water much more rapidly, bypassing the kinetic barriers which can limit water sampling in conventional MD simulations (see Supporting Information and Figure S4). GCMC is adequate for the simulation of the hydration of M2 pore at equilibrium with bulk water using the chemical potential reflected by the B_equil value. Description of the theory upon which GCMC is based and a detailed description of the simulation conditions can be found in the Supporting Information. The GCMC simulations were carried out using version 7.3.1 of OpenMM.⁵⁷ Version 1.0.0 of the grand Python module⁵⁸ was used with OpenMM to carry out the GCMC sampling, with the MD sampling carried out using the native OpenMM functionality.

For the free energy calculations reported here, the AMBER ff14sb force field was used to describe the protein and intermolecular interactions, the lipid¹⁷ force field for the membrane, TIP3P for the water and the Joung-Cheatham parameters were used for the ions. The ligands were described using the GAFF with AM1-BCC charges, generated using antechamber. All non-bonded interactions within 12 Å were calculated directly, Lennard-Jones interactions were switched to zero from 10–12 Å, and long-range electrostatic interactions were calculated using the Particle Mesh Ewald method (see the Supporting Information for details and references).

The region in which GCMC moves were carried out was defined as a sphere with a radius of 6.0 Å (for which B_equil = −6.820), centered on the average coordinate of the C_α atoms of the Gly34 residues. This was chosen to ensure that the two water layers of interest are covered by the GCMC sampling. Version 1.0.0 of the grand Python module⁵⁸ was used alongside OpenMM to carry out the GCMC sampling, with the MD sampling carried out using the native OpenMM functionality.

For each enantiomer, an initial equilibration procedure was applied, beginning with a very short minimization, following which the waters present in the GCMC region were deleted. This was followed by GCMC/MD equilibration at B_equil, to allow the waters in the GCMC region to be repopulated without bias, and then constant pressure equilibration to allow the volume of the system to equilibrate. The GCMC/MD stage was carried out in three phases to ensure rigorous equilibration of the waters before significant movement of the protein complex: first 1 ps of MD, with 5k moves attempted every 100 fs; then 10 ps of MD, with 100 moves every 200 fs; and finally, 50 ps of MD, with 500 moves every 250 fs. The constant pressure equilibration was carried out for 500 ps. The resulting equilibrated structure for each enantiomer was used as a starting point for each set of independent repeats in order to ensure that they all had the same system volume and, therefore, sampled the same ensemble.

For each independent set of repeats, for each enantiomer, a further GCMC/MD equilibration was carried out at B_equil for 500 ps, with 50 GCMC moves executed every 250 fs. From this, multiple simulations were run at 21 equally spaced B values, from −24.820 to −4.820. Each of these simulations was equilibrated for 500 ps, with 50 GCMC moves every 250 fs, to allow the number of waters to adjust to the new B value. This was followed by a production run of 2.5 ns, with 20 GCMC moves per 500 fs, with simulation frames written out every 5 ps. The average number of waters observed in the GCMC region, 〈N〉, was calculated over these production runs. Three sets of these simulations were carried out for each enantiomer, generating three values of 〈N〉 per B value, per enantiomer (Table S3).

Free Energy Calculations.

In order to determine the relative free energies of both hydration and binding for the two enantiomers, free energy calculations were performed in OpenMM, using version 0.7.1 of the perses Python module,⁵⁹ using the force fields and non bonded cutoff values described previously. This involved perturbing each of the enantiomers into an achiral intermediate compound, in which the methyl group of rimantadine is replaced by a hydrogen atom. Full details of these calculations and references are provided in the Supporting Information.