Capturing Glimpses of an Elusive HIV Gp41 Pre-Hairpin Fusion Intermediate

Summary

NMR studies on an HIV gp41 construct reported in this issue have revealed the conformational dynamics of a possible trimeric pre-hairpin fusion intermediate and its interactions with lipids. Two alternative fusion pathways are compared based on this work and a previous NMR structure of a monomeric lipid-bound gp41 ectodomain construct.

Membrane enveloped viruses including HIV enter cells by membrane fusion. The fusion reactions are catalyzed by viral envelope glycoproteins, which function as receptor binding and as fusion proteins. In the case of type I viral fusion proteins like the HIV gp41, the homotrimeric proteins refold from three-helix bundles on the surface of a circulating virus (Bartesaghi et al. 2013; Julien et al. 2013; Lyumkis et al. 2013) to a six-helix bundle consisting of trimers of helical hairpins upon completion of membrane fusion (Chan et al. 1997; Tan et al. 1997; Weissenhorn et al. 1997). The energy gained from this highly exothermic refolding reaction is generally believed to fuel the endothermic merger of the viral and cellular membranes. But, exactly how protein refolding and membrane fusion are mechanistically coupled has remained elusive despite significant progress of many groups that have worked on this important problem for many years. One approach to tackle the problem is to search for structural intermediates between the initial prefusion and postfusion states. Since these intermediates are intrinsically unstable and involve dynamic rearrangements of lipids and the lipid-interacting domains of the fusion proteins, it is perhaps impossible to capture these intermediates by protein crystallography. However, NMR offers unique possibilities to search for such dynamic intermediates in the presence of lipids or detergents and the lipid-interacting domains of the fusion proteins.

In an article published in this issue, Lakomek et al. (2014) come a significant step closer to a structural characterization of the so-called pre-hairpin intermediate of the HIV envelope fusion protein gp41. In this work, they have used a construct encompassing the entire gp41 sequence, except for the fusion peptide (FP) and its linker region (FPPR, combined residues 1–26) and a cytoplasmic C-terminal domain (CD, residues 195–345). This construct (residues 27–194) containing the N-terminal heptad repeat (NHR), the cysteine-linked immunodominant loop (IL), the C-terminal heptad repeat (CHR), the membrane proximal external region (MPER) and the transmembrane domain (TMD) was structurally and dynamically analyzed in significant molecular detail by solution NMR in dodecylphosphocholine (DPC) micelles. As expected from the crystal structures of the pre- and postfusion forms of gp41, the NHR and CHR domains formed helices (with minor irregularities at heptad-repeat intervals), and the NHR domain retained its homo-trimeric structure, which was also verified by analytical ultracentrifugation. The CHR domains underwent free motions on ns and µs to ms timescales that were independent of similar motions of and within the trimeric NHR domains. Portions of both heptad repeats and the intervening immunodominant loop interacted transiently with lipid micelles suggesting that lipid interactions are not only confined to the fusion peptides and TMDs of gp41. The results lead to a model of the prehairpin fusion intermediate that is depicted schematically in the top path of Figure 1.

Figure 1. Two competing models of membrane-interacting pre-hairpin fusion intermediates of HIV gp41 based on solution NMR studies of various gp41 constructs in lipid micelles.

The top path features pre-hairpin intermediates with trimeric NHR domains that may be interacting with lipids only transiently through their C-terminal end and IL domains according to Lakomek et al. (2014). The bottom path features pre-hairpin intermediates with dissociated NHR and CHR domains that both interact with opposing lipid bilayers according to Roche et al. (2014). Both pre-hairpin intermediate models fluctuate with ns dynamics and are in equilibrium (on a µs to ms dynamical scale) with early extended pre-fusion and late presumably lipid-mixing fusion intermediates. See text for abbreviations of gp41 sequence domain elements.

Two earlier NMR studies on slightly different gp41 constructs in association with DPC micelles nicely complement the work of Lakomek et al. In one of these, some of the same authors studied a longer construct that also contained the FP and linker region, i.e. residues 1–194 (Lakomek et al. 2013). Although this construct contains both major membrane-interacting ends, i.e. the FP and TM domains, and therefore potentially is a better physiological model to study the prehairpin intermediate and perhaps the yet uncharacterized membrane-anchored postfusion state of gp41, the construct posed severe limitations on the resolution of NMR spectra collected in DPC micelles. The resonances of the entire C-terminal half of the molecule comprising the CHR, MPER, and TMD regions were exchange-broadened beyond detection. However, the FP domains were observable and demonstrated that the FP regions moved independently from the NHR three-helix bundle.

A construct just comprising the NHR and CHR linked by a six-amino-acid linker that replaced the IL region of gp41 was structurally characterized in the other study that also employed solution NMR in DPC micelles (Roche et al. 2014). These substantial truncations permitted a full resonance assignment and a complete structure to be determined by NMR, which was not possible with the longer constructs. Quite interestingly and perhaps surprisingly, the NHR trimer of this construct was dissociated into monomers by interaction with DPC. This was not due to the lack of the IL domain because similar results were obtained in its presence. Apparently, anchoring the polypeptide into a membrane mimetic by either the FP or TM domains is required to keep the NHR trimeric. The authors of this latter paper concluded that the unzippered helices may lead to a quite different pre-hairpin intermediate, in which the two heptad repeat helices interact with the two opposing membranes before reassociating into the six-helix bundle for membrane fusion (bottom path of Figure 1). Future experiments, probably requiring structural techniques that work in lipid bilayer membranes, will be needed to distinguish between these two interesting possibilities.

Work that has been conducted to study potential structural intermediates in another membrane fusion system, namely that involving SNARE proteins that catalyze the opening of exocytotic fusion pores at neuronal synapses may provide some clues. The dynamic prefusion structures of the vesicle SNARE synaptobrevin (Ellena et al. 2009) and the target membrane SNARE syntaxin (Liang et al. 2013) have been solved by solution NMR in lipid micelles. These structures were also more folded than their soluble counterparts in the absence of lipids and therefore suggested lipid-assisted folding (and perhaps fusion) pathways that are further modulated by membrane curvature (Liang et al. 2014). Eager spectators may be in for many more revealing surprises before we fully understand the folding-fusion coupling in the intracellular and viral membrane fusion field!