A bundling of viral fusion mechanisms

Membrane fusion is a fundamental process in neurotransmission, vesicle trafficking, and infection by enveloped viruses. Mutational studies of class I viral fusion proteins have shown that simply pulling the two membranes together is not sufficient (1–3) to catalyze fusion in a physiological context: in these systems, a specific activity of the membrane-inserted peptides is required for full fusion. The precise mechanism of membrane activity has remained elusive. Although a number of hypotheses have been put forward (4), definitively establishing which combination of these contribute to fusion has been challenging. A combined structural and computational study published in PNAS (5) helps reinvigorate this debate. This article by Donald et al. (5) presents important conformational data on membrane-inserted fusion domains. Perhaps more significantly, it represents an emerging paradigm whereby computational studies can be used to help interpret mechanistic complexity in experimental data.

New Data on Parainfluenza Virus Fusion Assemblies

In their study, Donald et al. (5) present analytical ultracentrifugation results showing that the parainfluenza virus 5 (PIV5) fusion peptide assembles into hexamers in dodecylphosphocholine micelles. Minimal evidence was found for other multimeric states. The C-terminal transmembrane domain did not self-associate but did associate with the fusion peptide when mixed at a 1:1 ratio. These findings for the transmembrane domain are consistent with previous structural and biochemical findings of a three-helix bundle prefusion conformation and a postfusion state that is structurally a six-helix heterohexameric bundle of transmembrane domains and fusion peptides (6, 7). CD and IR spectra of the fusion peptide in micelles were consistent with an α-helical secondary structure and orientation oblique or normal to the bilayer. Based on these data, a model of a membrane-inserted six-helix bundle was developed and evaluated by using molecular mechanics and molecular dynamics simulation. FTIR dichroism ratios calculated from the most stable structure were in good agreement with the experimental data.

These results raise a number of interesting questions compared with data on influenza and HIV fusion peptides. If a single unifying mechanism exists to explain viral fusion peptide activity, one would expect to find similar oligomeric and conformational features in these other well studied class I viral fusion proteins. The FTIR insertion angles calculated for PIV5 are within the error of previous results on influenza (8). However, the

Donald et al. propose a mixture of mechanisms between the protein pore and the lipidic stalk models.

NMR structures of influenza fusion peptides in micelles show a kinked helix (9) or a helical hairpin (10). Whether these conformations change upon formation of a peptide assembly is, of course, an open question. FRET data suggest a loose association of influenza peptides in lipid bilayers (11), although thermodynamic data on membrane binding by influenza fusion peptides suggest an equilibrium between an interfacial β-sheet aggregate and a helical monomer (12). HIV gp41 fusion peptides display an apparent equilibrium between helical and β-sheet forms that can be modulated by a number of factors, including lipid composition (13–16); it is not well established which of these forms is physiologically active. The trimeric structure of HIV, influenza, and PIV5 ectodomains is also an important factor, as it greatly increases the local concentration of transmembrane domains and fusion peptides, respectively. Whether these represent fundamental differences between the fusion peptides of the two viruses or can be attributed to differences in constructs and experimental methodology is yet unclear.

Mechanistic Heterogeneity in Membrane Fusion and Beyond

Substantial progress has been made in elucidating potential mechanisms for formation of a lipidic stalk or, alternately, a protein-lined fusion pore. However, explaining the mechanism responsible for terminal hemifusion mutants has been much more elusive. In our opinion, this is a critical outstanding question in explaining fusion. In their study, Donald et al. (5) propose a mixture of mechanisms between the protein pore and the lipidic stalk models. This is a powerful idea, and one that could be expanded to a mixed model that could explain the hemifusion mutants in the following manner: a primary lipidic mechanism forms a hemifusion stalk, allowing for close association of C-terminal TM and N-terminal fusion peptide domains. This close association might then drive formation of a mixed six-helix bundle and opening of a fusion pore. This speculative model would then clearly explain how terminal hemifusion mutants could arise, particularly ones in the C-terminal TM domain. Further experiments are clearly necessary to elucidate such a mechanism. Whether the aforementioned model or something else holds, the approach that Donald et al. (5) use—biophysical experiments coupled to computational models to help interpret a spectrum of mechanisms—will be of important general utility in helping to resolve outstanding questions of fusion mechanism.

The proposal of Donald et al. (5) that one could understand fusion as a unified spectrum of possible mechanisms is potentially of broad and general importance. This idea is particularly appealing because the biophysics of lipids may allow for a wide variety of mechanisms, but a particular biological realization—e.g., protein sequence, lipid composition—may be at one particular end of the continuum. The generality of this idea of a spectrum of mechanisms is supported by recent examples appearing in other contexts, such as protein folding (17) and lipidic fusion itself (18). Furthermore, this idea of a unifying spectrum of mechanisms suggests natural roles for biophysicists and biologists to, respectively, identify which possible mechanisms are physically achievable as opposed to which particular realizations have arisen evolutionarily.