Lowered pH Leads to Fusion Peptide Release and a Highly Dynamic Intermediate of Influenza Hemagglutinin

Abstract

Hemagglutinin (HA), the membrane-bound fusion protein of the influenza virus, enables the entry of virus into host cells via a structural rearrangement. There is strong evidence that the primary trigger for this rearrangement is the low pH environment of a late endosome. To understand the structural basis and the dynamic consequences of the pH trigger, we employed explicit-solvent molecular dynamics simulations to investigate the initial stages of the HA transition. Our results indicate that lowered pH destabilizes HA and speeds up the dissociation of the fusion peptides (FPs). A buried salt bridge between the N-terminus and Asp1122 of HA stem domain locks the FPs and may act as one of the pH sensors. In line with recent observations from simplified protein models, we find that, after the dissociation of FPs, a structural order-disorder transition in a loop connecting the central coiled-coil to the C-terminal domains produces a highly mobile HA. This motion suggests the existence of a long-lived asymmetric or "symmetry-broken" intermediate during the HA conformational change. This intermediate conformation is consistent with models of hemifusion, and its early formation during the conformational change has implications for the aggregation seen in HA activity.

Figure 1

An electrostatic environment induced by lowered pH increases the rate of FP release. (A) Schematic diagram showing the change of secondary structure of a monomer between the pre- and post-fusion crystal structures. The corresponding crystal structures are shown in the supplemental information. The sequence is partitioned into sections differentiated by their structural change in the HA₂ rearrangement, with FP beginning at the N-terminus and S5 ending at the C-terminus. Not shown is an additional C-terminal trans-membrane domain. FP (residue ID 1–20) and TBS (two beta-strands, residue ID 21 to 37) are hydrophobic domains at the N-terminus. They are missing in the post-fusion crystal structure. S1 (residue ID 38 to 54, yellow) and S3 (residue ID 76 to 104, red) maintain their secondary structure. S2 (residue ID 55 to 75, blue) changes from a loop to helix. S4 (residue ID 105 to 129, green) and S5 (residue ID 130 to 175, orange) have a partial secondary structure shift. Loop3-4 (residue ID 107 to 112) is a region of low helical propensity at S3–S4 interface that undergoes a helix-loop transition. (B) Pre-fusion HA₂ crystal structure (PDB ID: 2HMG) shows the FPs to be buried in a hydrophobic pocket within S4. 3 N-terminal peptides of HA₁ (one for each monomer, residue ID 10 to 18 of HA₁, violet) are disulfide bonded to S5. The inset magnifies a view of the FPs in the crystal and after dissociation during the simulation, with N-termini of FPs and ASP₁₁₂ shown as spheres. The left inset defines the geometric center of the hydrophobic pocket as the center of mass of the three ASP₁₁₂ in the crystal. (C) The distance between the N-terminus of each FP and the center of the hydrophobic pocket. The curves are averaged over a 0.5 ns window for clarity. (D) The survival probability for the release of the first FP at high temperature (T = 398K) plotted as a function of time for the cases with neutral (red filled, 10 runs) and negative (green filled, 7 runs) ASP₁₁₂. Also shown are control simulations (grey filled, 6 runs) with neutral ASP₁₁₂ at pH 7.

Figure 2

Helical order of domains S3/S4 during different stages of the conformational rearrangement. Representative snapshots highlighting the structure of regions contacting Loop3-4 are shown for HA₂ (A) before FP release, (B) after FP release, and (C) after the dissociation of S1/S2. (D) The helical order of a residue is determined by averaging over the three monomers the probability that the STRIDE algorithm determines an alpha helical secondary structure. The black curve is averaged over a 700 ns simulation at pH 4.5 started from the crystal structure, before dissociation of the FP and corresponding to (A). The red curve is averaged for a 2 μs at pH 5 where the FPs are already ejected, corresponding to (B). The green curve is 6 μs of simulation after the deletion of S1/S2. Loop3-4 shows substantial disorder compared with the other parts of S3/S4.

Figure 3

Disordering of Loop3-4 upon breaking of the S1–S4 interface leads to large fluctuations in HA₂. (A) The coordinate θ_S3 estimates the deviation of S3 from a perpendicular orientation to the viral surface by measuring the angle the coiled-coil makes relative to S5. S5 is anchored to the viral membrane by a C-terminal transmembrane region. (B) Two possible structures for an SBI taken from explicit-solvent simulations and dual-basin SBS (see 2 Basins C, E and F in ref. for explanation of the SBS results). (C) θ_S3 as a function of time during the final 5 μs of a constant temperature (310K) simulation of HA₂ with S1 and S2 removed. (D) Effective diffusion coefficient is obtained from the mean square displacement of the angle 〈δθ_S3(τ)²〉 in the diffusive regime. Linear fitting is drawn as a black dashed line.

Figure 4

A cartoon illustration of possible fusogenic routes for HA₂ from a symmetry-broken intermediate (SBI). (A) A “sequential route” where three FPs of each HA₂ are inserted into host membranes. This route is called sequential because the arms fold into their post-fusion coiled-coil before the membranes are brought together. This resembles the “spring-loaded” model.^, (B) A “cooperative route” where the FPs are inserted into both the host and viral membranes. The SBI facilitates this route by bringing the FPs closer to viral membrane. This route is termed cooperative because the coiled-coil forms only as the membranes are brought close together. In both of these routes, S5 will eventually melt and wrap around HA₂, presumably helping to bring two membranes together. If S5 breaks before at least one FP is inserted into the host membrane, likely all the FPs are pinned at the viral surface leading to a inactive conformational transition.^,