Mining Druggable Sites in Influenza A Hemagglutinin: Binding of the Pinanamine-Based Inhibitor M090

Abstract

Assessing the binding mode of drug-like compounds is key in structure-based drug design. However, this may be challenged by factors such as the structural flexibility of the target protein. In this case, state-of-the-art computational methods can be valuable to explore the linkages between structural and pharmacological data. Following this strategy, extended molecular dynamics simulations and thermodynamic integration calculations are used to examine the binding of the potent antiviral inhibitor M090 and related pinanamine-based analogues, covering a 250-fold difference in inhibitory potency to the influenza A hemagglutinin, which is essential for virus entry and membrane fusion. This analysis has disclosed the hydrophobic shielding effect played by the 3-cyclopropylthiophene moiety in M090. Furthermore, the results support the negative effect of the resistance-induced E74₂ → D mutation, which should weaken the binding by increasing the structural flexibility of the L2-BS loop. The results pave the way to exploration of the antiviral activity of novel compounds.

Figure 1

Structural representation of HA. The monomeric chains HA₁ (gray) and HA₂ (green) formed after proteolytic cleavage assemble, forming a trimer where the fusion peptide (FP) (magenta) remains buried until acidification in the endosome. The putative binding site of M090 (BM-II), which is shaped by an α-helix (H-BS) and two loops (L1-BS and L2-BS), is displayed as an orange surface that encloses M090 (shown as sticks).

Figure 2

Representation of the three initial binding modes of M090 investigated in this study: (A1) the pose with the restrained HB between the amine nitrogen and T309₁; (B1) the unrestrained arrangement in the interior of the pocket; (C1) the reversed pose where the thiophene and pinanamine units have been exchanged relative to A1 and B1. Representative snapshots of the poses obtained at the end of the MD simulations with distances to D85₂ and T309₁ are shown in the central column (A2, B2, and C2). Finally, superpositions of the ligands taken every 100 ns along the last 500 ns of the MD simulation are displayed in the right column (A3, B3, C3). Note that only two pockets containing overlapped ligands are shown in C3 due to the release of M090 from the third binding pocket. Values denote the distance averaged for the three distinct M090 molecules bound to HA.

Figure 3

(A) Representation of the free energy changes determined from TI calculations for the alchemical (forward and backward) transformations between M090-related derivatives. The numbering of the derivatives follows the notation used by Zhao et al. in ref (32). Green values denote the sum of the free energy changes for the processes implicated in the closure of the thermodynamic cycle. (B) Comparison between the differences in binding affinity estimated from the experimental EC₅₀ data and the relative free energy differences predicted from TI calculations. Bars denote the errors of calculated and experimental relative free energy changes. Dotted and dashed lines denote the 95% confidence and prediction intervals. All values (kcal mol^–1) are relative to M090. MUE and RMSE stand for mean unsigned error and root-mean-square error, respectively.

Figure 4

Representative snapshots of the binding pose obtained for compounds (A) M15, (B) M27, (C) M19, and (D) M30 (shown as sticks) at the end of the alchemical transformations. (A) The n-butyl chain attached to position 3 of the thiophene unit adopts a folded conformation due to the steric clash with loop L2-BS. (B) Superposition of the complexes with n-butyl derivatives M15 (bound to position 3) and M27 (attached to position 4). The steric clash observed for M15 with L2-BS is alleviated in M27, where the n-butyl chain adopts an extended conformation and forms van der Waals contacts with P307₁, I308₁, and the side chain of K82₂. (C, D) The cyclopentyl unit in M19 occludes the access of water molecules, thus creating hydrophobic shielding of the electrostatic interaction between D85₂ and K83₂ in the neighboring protomer, mimicking the protecting effect of the cyclopropyl unit in M090. This effect is lost in compounds M04, M06, and M07, which have unsubstituted furan and thiophene rings.

Figure 5

Overlaid structures of (A) M090 and (B) M24 (orange sticks) with N-[(5-bromothiophen-2-yl)methyl]adamantan-1-amine (gray sticks) and superposition of the overlaid compounds onto the X-ray crystallographic structure of the S31N mutated variant of the M2 proton channel (PDB ID 2MUV).

Figure 6

(A) Representation of the two main essential motions of the backbone atoms that shape the binding pocket for the (left) apo and (right) holo species of wild type and E74₂ → D mutated HA. (B) Projections of the snapshots sampled for the apo wild type HA (red) and its mutated variant (yellow). (C) Centroids of the clusters observed for the apo wild type and mutated HA (backbone of the binding pocket in red and yellow, respectively).