Roles of Conserved Tryptophans in Trimerization of HIV-1 Membrane-Proximal External Regions: Implications for Virucidal Design via Alchemical Free-Energy Molecular Simulations

Abstract

The Dual-Action Virolytic Entry Inhibitors, or “DAVEI’s,” are a class of recombinant fusions of a lectin, a linker polypeptide, and a 15-residue fragment from the membrane-proximal external region (MPER) of HIV-1 gp41. DAVEI’s trigger rupture of HIV-1 virions, and the interaction site between DAVEI MPER and HIV-1 lies in the gp41 component of the envelope glycoprotein Env. Here we explore the hypothesis that DAVEI MPER engages Env gp41 in a mode structurally similar to a crystallographic MPER trimer. We used alchemical free-energy perturbation to assess the thermodynamic roles of each of the four conserved tryptophan residues on each protomer of MPER₃. We found that a W666A mutation had a large positive ΔΔG for all three protomers, while W672A had a large positive ΔΔG for only two of the three protomers, with the other tryptophans remaining unimportant contributors to MPER₃ stability. The protomer for which W672 is not important is unique in the placement of its W666 sidechain between the other two protomers. We show that the unique orientation of this W666 sidechain azimuthally rotates its protomer away from the orientation it would have if the trimer were symmetric, resulting in the diminished interaction of this W672 with the rest of MPER₃. Our findings are consistent with our previous experimental study of W-to-A mutants of DAVEI. This suggests that DAVEI MPER may engage HIV-1 Env to form a mixed trimer state in which one DAVEI MPER forms a trimer by displacing a more weakly interacting protomer of the endogenous Env MPER trimer.

Introduction

The envelope glycoprotein (Env) is the only viral protein displayed on the surface of HIV-1 virions. Env is responsible for targeting HIV-1 to CD4+ T-cells and for mediating the membrane fusion and entry process.¹ Env comprises the non-covalently linked subunits gp120 and gp41 that form a heterotrimeric “spike”.^1–4 The gp120 subunits are completely outside the viral membrane and consist of five conserved regions and five variable loops.^1,5,6 The gp120 subunit is the region that interacts with CD4 and coreceptors, leading to conformational changes in both gp41 and gp120.^1,2,5–7 The gp41 subunit anchors the spike to the viral membrane and is thought to contain the machinery needed for fusion of viral and host cell membranes.^1,5,7 Each gp41 subunit includes two helical segments and a membrane proximal external region (MPER) that is likely associated with the membrane of the virus in native Env.^3,8 After gp120 makes contact with CD4 and the coreceptor, gp41 undergoes a conformational change to form the six-helix bundle.^7–10 This refolding is thought to bring the cell and virus membranes together while releasing energy to overcome the kinetic barrier to fusion.^8–10

The dual-acting virucidal entry inhibitors (DAVEI’s) were designed to bind specifically to native HIV-1 Env and trigger the Env conformational cascade without a cellular membrane present, resulting in release of the viral contents.³ The first-generation DAVEI was an engineered protein chimera that had a gp120 glycan binder (cyanovirin-N) linked to the HIV-1 MPER sequence by a flexible polypeptide linker.³ We recently showed that DAVEI requires access to Env MPER, since gp41 MPER-specific antibodies (4E10 and 10E8) outcompete DAVEI function.⁴ We also recently showed that the DAVEI MPER could be substantially minimized by removing the CRAC domain and that only one of the four tryptophan residues on DAVEI MPER was essential for DAVEI function.⁴ The results argued that the DAVEI MPER interacts directly with Env MPER. However, the exact mechanism by which the exogenous MPER of DAVEI interacts and induces Env-mediated cell-free viral lysis has remained unknown.

HIV-1 MPER can be induced to trimerize ex situ as a coiled-coil, as demonstrated by the 3G9R PDB structure, and which could have an important role as possibly one of several intermediate states in the membrane fusion of HIV-1.¹¹ (It should be pointed out that the 3G9R model does not seem to present epitopes for MPER-reactive mAbs,¹² which suggests it is not part of the Env conformational state that recognizes those antibodies.) This led us to ask whether or not a similar trimerization among one DAVEI MPER and two Env MPER’s might be part of the DAVEI-induced lysis mechanism. The 3G9R MPER₃ crystal structure is asymmetric, resulting in different orientations for the tryptophans on each chain. The purpose of the currently reported work was to determine whether or not the thermodynamic stability of this structure is consistent with the recent DAVEI variant study, and in particular the relative importance of different tryptophans on DAVEI. We approached this question using molecular dynamics simulations and free-energy perturbation calculations to determine the change in trimerization free-energy upon mutation of each individual tryptophan residue in the 3G9R structure. We correlated our findings to the asymmetry of the crystal structure, and showed that only one of the three chains in the trimer has a thermodynamic signature consistent with mutational experiments.

Methods

System Setup

The starting point for our simulations was an all-atom, explicit-water system based on the 3G9R structure.¹¹ We extracted from this structure only the MPER fragments (up to residue 23 on each of the three chains A, B, and C; the four tryptophans on each are labeled W666, W670, W672, and W678, consistent with an HXBc2 numbering for HIV-1) which resulted in the following sequence: 662-ELDKWASLWNWFNITNWLWYIK-683. Chains D, E and F in the structure were ignored in our study since in the region tested these chains’ backbones align well with A, B, and C, respectively, and W666 in both chains C and F are similarly inserted between the partner chains. The psfgen utility of VMD¹³ was then used to prepare the system with CHARMM36 parameters and TIP3P waters. We used standard neutralized N- and C-termini. These patches generated bond, angle, and torsion interaction types not in the latest CHARMM36m¹⁴ parameter set, so we used parameters from CHARMM22¹⁵ for these. A complete set of scripts and input files for generating this system is publicly accessible at github.com/cameronabrams/psfgen. In addition to a trimer system, we generated independent systems for each protomer. Three independent systems for each case were generated, resulting in twelve total simulation systems.

Each system was equilibrated for 100 ns using MD simulation via NAMD.¹⁶ All MD simulations were run using a Langevin thermostat set at 310 K with a damping coefficient of 1 ps⁻¹, a Langevin piston barostat with a period of 100 fs and decay of 50 fs, RATTLE bond-length constraints, particle-mesh Ewald electrostatics with a 1 Å mesh spacing, a 12 Å nonbonded cutoff, and a 2-fs time step. After equilibration, mutant systems required for free-energy perturbation (FEP) calculations were generated using the mutator plugin of VMD.¹³

Free Energy Perturbation Calculations

Our basic task was the measurement of the change in free energy of binding of an α-helical MPER fragment to the complementary dimer to form the 3G9R structure upon mutation of one of the four tryptophans on the monomer to an alanine. Hence, we analyzed three independent trimerization reactions:

$${\text{MPER}}_{\mathrm{A}}+{\text{MPER}}_{2(\text{BC})}\to {\text{MPER}}_3$$ (1)

$${\text{MPER}}_{\mathrm{B}}+{\text{MPER}}_{2(\text{CA})}\to {\text{MPER}}_3$$ (2)

$${\text{MPER}}_{\mathrm{C}}+{\text{MPER}}_{2(\text{AB})}\to {\text{MPER}}_3,$$ (3)

where A, B, and C are the chain designations in the PDB entry, MPER_A is the monomer of chain A, and MPER_2(AB) refers to the dimer of chains A and B. For each of these reactions, we computed the ΔΔG associated with mutating each of W666, W670, W672, and W678 of the monomer to alanine. However, binding free energies for such complicated systems were prohibitively difficult to compute directly, so we instead employed the thermodynamic cycle depicted in Fig. 1 to compute each ΔΔG. That is, we needed only to compute free-energy changes upon mutation in the contexts of the monomer and the trimer.

Figure 1.

Thermodynamic cycle used for computation of a ΔΔG associated with mutation of any one particular tryptophan (W666, W670, W672, or W678) on a particular MPER chain (A, B, or C). ΔG₁ and ΔG₂ are free energy changes upon association of a monomer with a dimer to form a trimer in the cases where the monomer is wild-type (black) or a mutant (red), respectively. The sequence (α, β, γ) represents the three cyclic permutations of the chain labels A, B, and C. ΔG₃ and ΔG₄ are the free energy changes upon mutation of a single monomer in the absence of the dimer (grey) or in the presence of the bound dimer, respectively. The effect of the mutation on trimer stability is defined as ΔΔG = ΔG₂ − ΔG₁ but is computed via FEP as the equivalent ΔG₄ − ΔG₃.

With reference to Fig. 1, we directly computed free energies of mutation for a monomer and a trimer, ΔG₃ and ΔG₄, respectively, using free-energy perturbation (FEP) calculations. ΔΔG was defined as the difference in binding free energies for the wild-type and mutant cases, ΔG₁ and ΔG₂, respectively, but due to the cycle, this was equivalent to the difference between ΔG₄ and ΔG₃:

$$\mathrm{\Delta }\mathrm{\Delta }G\equiv \mathrm{\Delta }{G}_2-\mathrm{\Delta }{G}_1=\mathrm{\Delta }{G}_4-\mathrm{\Delta }{G}_3$$ (4)

FEP refers to a technique whereby the free energy associated with changing some order parameter (here designated λ on domain [0,1]) is computed via several independent MD simulations at fixed order parameter values.^17–20 In this approach, the order parameter λ represents the weighting of each pure-state Hamiltonian in the dual-topology Hamiltonian:

$$H(\lambda )=H_0+\lambda H_a+(1-\lambda )H_b$$ (5)

Here, H₀ represents interactions independent of the mutation under consideration, while H_a represents interactions in the mutated system and H_b interactions in the wild-type system. The domain of λ is divided into equally-spaced windows of with 0.05; i.e., λ₀ =0.0, λ₁ =0.05, …, λ₁₉ =0.95, and λ₂₀ =1.0. Free-energy changes across each window, ΔG_i→i+1 ≡ G(λ_i+1) − G(λ_i) are computed using the SOS equation:²¹

$$exp(-\beta \mathrm{\Delta }{G}_{i\to i+1})=\frac{{\langle \text{exp}\left\{-\frac{\beta }{2}[H(x,p_x;{\lambda }_{i+1})-H(x,p_x;{\lambda }_i)]\right\}\rangle }_i}{{\langle \text{exp}\left\{-\frac{\beta }{2}[H(x,p_x;{\lambda }_i)-H(x,p_x;{\lambda }_{i+1})]\right\}\rangle }_{i+1}}$$ (6)

Here, β ≡ 1/k_BT where k_B is Boltzmann’s constant, 〈·〉_i indicates an ensemble average computed from an MD simulation run on the Hamiltonian H(λ_i). Each such MD simulation was run for at least 1.2 million time-steps, or until convergence of the Boltzmann factors were observed.

The overall ΔG for a particular mutation was then computed by summing all of the window-specific ΔG’s:

$$\mathrm{\Delta }G=G(\lambda =1)-G(\lambda =0)=\sum _{i=0}^{N_{\lambda }-1}\mathrm{\Delta }{G}_{i\to i+1}$$ (7)

Here, N_λ is the number of distinct λ values considered; in this case, N_λ is 20, meaning that we ran 21 independent simulations for each replica of each mutation. We performed all MD simulations under constant NVT conditions, which strictly means we were computing differences in Helmholtz free energy, A. However, since ΔG = ΔA + Δ(PV) and we take Δ(PV) to be zero since V is invariant, we can take ΔG as ΔA.

Results and Discussion

We determined ΔΔG’s across all chains and all tryptophan-to-alanine mutations, along with standard deviations from three independent sets of FEP calculation. The results are given in Table 1. In this analysis, W666 was the strongest stabilizer of the homotrimer, with ΔΔG’s in excess of 1.5 kcal/mol for all three chains. At the other extreme, W678 was the weakest. W670 energy calculation was intermediate, with ΔΔG’s close to thermal energy under physiological conditions (0.6 kcal/mol at 310 K). Strikingly, however, the stability conferred by W672 depended on which chain it belongs to: for chains A and B, it was energetically strong, while it had less of an impact for chain C.

Table 1.

ΔΔG’s, in kcal/mol, for each mutation performed on the MPER complex and the respective MPER monomer. Errors are standard deviations from three independent sets of FEP simulations, except those denoted by * which are six independent sets of FEP simulations.

	ΔΔG (kcal/mol)
Mutation	Chain A	Chain B	Chain C
W666A	1.62 ± 0.72	1.63 ± 1.13*	1.50 ± 0.77
W670A	0.87 ± 0.75*	0.97 ± 0.63*	0.70 ± 0.55
W672A	1.30 ± 0.73*	1.46 ± 0.35	0.21 ± 0.29
W678A	0.27 ± 0.17	0.07 ± 0.14	0.15 ± 0.14

We surmised that, for any ΔΔG to be significant, one might expect that the corresponding tryptophan side chain is interacting with atoms from one or both of the other monomers in the trimer, and solvent-exposed in the standalone monomer. We determined values for solvent-exposed surface area (SASA) for each tryptophan side-chain from three independent MD simulations at λ = 0 (wild-type) in Table 2, along with values computed from the static crystal structure. There was a fairly good correspondence between those tryptophans with high SASA and low ΔΔG and vice-versa. W666’s on all three chains participate in many solvent-excluding intermonomer interactions, with W666_C in particular the most buried. W678 on all three chains were very solvent-exposed in the trimer. W672 on chain C was very solvent-exposed while W672 on chains A and B were more shielded from solvent. There was also a fairly good qualitative correspondence between the SASA’s computed from MD and those from the static crystal structure.

Table 2.

Solvent-accessible surface area (SASA) in Ų for each tryptophan side-chain measured from three independent MD simulations except those denoted by * which have six independent MD simulations. Values in parentheses are measured on the 3G9R crystal structure.

	SASA (Å2)
Tryptophan	Chain A	Chain B	Chain C
666	83± 5 (59)	103±16* (105)	79±35 (21)
670	99±8* (84)	102±16* (107)	95±2 (112)
672	87±16* (64)	99±6 (106)	117±17 (131)
678	161±2 (146)	149±11 (168)	151±3 (139)

Although the correspondence between ΔΔG and SASA was strong on average, the large errors associated with a few of the side-chains shown in Tables 1 and 2 did not clearly show whether this correspondence was strong for any single simulation. The correlation in Fig. 2 shows ΔΔG values computed for independent replicas vs. the corresponding SASA values computed from that replica’s λ = 0 MD simulation. There is good agreement among all chains and all replicas for W678 being both high in SASA and low in ΔΔG. It is also clear that residue 672 on chains A and B are with one exception low in SASA and high in ΔΔG, while for chain C, there is agreement among the replicas that its SASA is around 100 Ų or higher with a low ΔΔG, showing the expected relationship between the two. The SASA’s are fairly tightly clustered around 100 Ų for residue 670 in all chains, with a low-to-medium range spread on ΔΔG. Although there was a large spread of values of both SASA and ΔΔG for residue 666 in all chains, the average among these clearly shows low-SASA, high-ΔΔG.

Figure 2.

ΔΔG for trimer formation upon tryptophan-to-alanine mutation from FEP calculations vs in-trimer tryptophan SASA from MD simulations. For each chain A, B, and C, and each tryptophan position 666, 670, 672, and 678, three independent replicas are plotted.

The origin of the special case of low ΔΔG for chain C’s W672 can be at least partially explained based on the lack of exact three-fold symmetry in the trimer crystal structure. The most obvious evidence for this lack of symmetry is the fact that the side chain of W666_C is inserted between chains A and B, while its partners on chain A and B are not inserted. The link between W666_C’s unique conformation and the exceptionally low ΔΔG of W672_C can be rationalized by considering where W672_C would be located if the trimer were perfectly symmetric. First, in a coordinate system in which the polar axis lies along the pseudo-threefold axis of the trimer, consider an operation that transforms the atomic coordinates of chain A onto those of chain B (essentially a rotation around the polar axis of 120°). Now, if we apply this operation twice, we can compare the actual coordinates of chain C to those we would expect chain C to have if it were perfectly symmetry-related to chains A and B. We depict these comparisons in Fig. 3.

Figure 3.

(A) View along the triple-helix symmetry axis of the MPER₃ structure from PDB 3G9R. Chains A, B, and C are labeled, and arrows indicate rotation of chain A to produce the overlaid structures in parts (B) and (C). (B) View along the triple-helix symmetry axis of the symmetry alignment of chain B (pink) onto chain A (purple) (upper), along with a side view (lower). Only backbone heavy atoms and tryptophan heavy atoms are shown. Tryptophans 666, 670, 672, and 678 (HXBc2 numbering) are labeled. Black cylinder is the triple-helix pseudo-symmetry axis. (C) Same as (B), except showing chain C (green) aligned onto chain A.

Fig. 3 shows that the relative coordinates of chain A and B closely reflect what one would expect for two α-helices in a C_3v-symmetric triple helix, with the exception of the rotomeric state of W672; however, as we have already shown, both of these tryptophans appear to be more or less equally important in inter-protomer interactions. Fig. 3B shows in contrast that chains A and C less closely reflect the symmetric expectation. In fact, the backbone of chain C is rotated counterclockwise around the C→N α-helical axis by several degrees relative to that of chain A, as indicated by the green arrow. This is enough of a displacement that 672 on chain C is shifted away from the triple helix relative to 672 on chains A and B.

Taken together, these results suggest that, of the four tryptophans in the MPER segment, W666 is essential for triple-helix stability, and only two of the three helices contribute additional stability via W672, while W670 and W678 are not important for this model’s stability. We considered only the tryptophans in this study because we hypothesized in previous work that at least some of MPER tryptophans were essential for DAVEI function.⁴ In fact, we found that DAVEI variants retained function with all but the N-terminal-most tryptophan (sequence-analogous to W666) mutated to alanine.⁴ We found that EC₅₀ for DAVEI-induced release of viral p24 are 29.8 ± 1.0 nM, 46.8 ± 5.9 nM, 49.7 ± 5.5 nM, 43.1 ± 3.6 nM, and 122.2 ± 11.6 nM for wild type, single, double, triple, and quadruple mutant DAVEI respectively.⁴ We also found that MPER-specific antibodies block DAVEI function, suggesting that some component of DAVEI might require association with one or more viral MPER segments.⁴ The fact that MPER-reactive mAb epitopes¹² are not fully exposed on 3G9R also suggests that mAb binding would prevent MPER multimerization. These findings suggest that the MPER₃ model examined here might represent a mode in which DAVEI-MPER interacts with Env MPER’s if the DAVEI-MPER corresponds to chain C in the model. Further tests of this hypothesis will necessarily include testing other residues that are potentially essential for the stability of this model and by mutation of native Env residues to test potency of DAVEI, which is currently ongoing.

Conclusions

Using all-atom molecular dynamics simulations and free-energy perturbation calculations based on the 3G9R MPER triple helix, we examined the roles of the four MPER tryptophans on triple-helix stability. Computed ΔΔG’s shows W666 is the most important residue in all three protomers for MPER₃ stability, while W672 is important only on chains A and B, and remaining tryptophans are relatively unimportant. The asymmetry of chain C in the MPER₃ crystal structure, most prominently demonstrated by the deep insertion of W666 inside the triple-helix, also manifests as a symmetry-displacement W672 on chain C relative to A and B due to a counterclockwise azimuthal rotation in this chain compared to the other chains. The importance of W666 and unimportance of W672 in our model is consistent with the results of a previous experimental study that showed diminished activity of our dual-acting virolytic entry inhibitors in which their MPER-sequence tryptophans were systematically mutated to alanines. If one assumes that MPER trimerization in Env is an important intermediate in fusion, this work suggests that mutations in Env that disfavor MPER trimerization might compromise both fusion and susceptibility to DAVEI. Taken together, this further suggests that the type of MPER₃ triple-helix displayed by 3G9R might also be formed by the DAVEI-MPER interaction with the native HIV-1 spike that causes virolysis. How an [HIV-1 Env MPER₂]•[DAVEI MPER] complex would disrupt the native Env MPER₃ to cause virus membrane disruption remains unknown.