Interfacial pre-transmembrane domains in viral proteins promoting membrane fusion and fission.

Membrane fusion and fission underlie two limiting steps of enveloped virus replication cycle: access to the interior of the host-cell (entry) and dissemination of viral progeny after replication (budding), respectively. These dynamic processes proceed mediated by specialized proteins that disrupt and bend the lipid bilayer organization transiently and locally. We introduced Wimley-White membrane-water partitioning free energies of the amino acids as an algorithm for predicting functional domains that may transmit protein conformational energy into membranes. It was found that many viral products possess unusually extended, aromatic-rich pre-transmembrane stretches predicted to stably reside at the membrane interface. Here, we review structure-function studies, as well as data reported on the interaction of representative peptides with model membranes, all of which sustain a functional role for these domains in viral fusion and fission. Since pre-transmembrane sequences also constitute antigenic determinants in a membrane-bound state, we also describe some recent results on their recognition and blocking at membrane interface by neutralizing antibodies.

Introduction

Membrane fusion and fission can be considered as energetically unfavorable reactions that involve hundreds of lipid molecules with no change in covalent bonding. These complex physiological reactions proceed coupled to the basic processes of local rupture (exposure of the hydrophobic interior) and deformation (bending) of lipid bilayers (reviewed in Refs. [1], [2]). Enveloped animal viruses rely on the fusogenic activity of membrane integral surface glycoproteins to enter and infect their host-cells [3], [4], [5], [6], [7], while fission is required for pinching-off of newly assembled virions [1] . Membrane fission is also operational intracellularly all along virus infectious cycles [8] . For example trafficking of viral proteins during assembly at the plasma or intracellular membranes and perhaps during intracellular translocation of the capsid complexes is dependent on this process [9] .

Fusion protein machinery delivers energy in order to perform a series of jobs, including bringing the membranes into close contact and the sequential formation of only-lipid (“stalks”) and lipid-aqueous connections (“fusion pores”) [1], [7], [10], [11], [12], [13], [14]. Although these proteins share little sequence homology even between members of the same family, evolutionary convergence has produced a number of common features [1,7,12] : functional priming at the cell surface through proteolytic processing, oligomeric organization (homotrimers of heterodimers in most cases), activity localized into ecto and transmembrane domains of the integral subunit, and the presence of a fusion peptide (FP), a highly conserved, hydrophobic domain, usually located at or close to the free amino-terminus of the integral subunit.

Viral fusogenic function, on the other hand, has been assembled within two structurally divergent scaffolds [4], [12], [15]. Class I fusion proteins are all characterized by their capacity to fold into a highly stable 6-helix bundle (6-HB). In this “hairpin-like” trimeric structure three helices pack in the reverse direction against hydrophobic grooves outside a triple-stranded coiled-coil, so that the ectodomain amino- and carboxy-termini are placed at the same end of the molecule. It is generally thought that amino-terminal FP sequences insert into the target cell membrane in a “pre-hairpin” stage [16], [17], [18]. Subsequent production of the 6-HB would therefore induce close apposition of the viral and cell membranes anchored through the transmembrane domain (TMD) and the FP, respectively. In contrast, class II proteins are composed of antiparallel β-sheet structures and possess internal FPs [4], [12], [15], [19]. Despite these differences, class II proteins have evolved into structures capable of triggering fusion following a mechanism essentially similar to that of class I proteins [15], [19], namely, FP insertion into the target membrane and folding into a low-energy trimer that brings into contact TMD- and FP-anchored membranes.

Thus, protein conformational energy, such as that released upon 6-HB formation, must be coupled with membrane merger by means of membrane-inserting specialized domains. In this review we will not address the FPs, since these force-transmitting domains, together with additional assistant elements scattered along the ectodomain sequences, have been recently subject of thorough revision in these series [17], [18], [20]. Here, we will specifically focus on the domains that precede transmembrane anchors in viral proteins promoting fusion–fission. These sequences have been defined as distinct domains according to their tendency to partition from water into the membrane interface (designated as “pre-transmembrane” domains, PreTM), and also attending to its implication in fusion and viral neutralization processes (designated as “membrane-proximal external regions”, MPER).

Since a hydrophilic, flexible, structurally irregular PreTM linker might keep the 6-HB far from the close contacting dehydrated areas required for membrane fusion initiation (as far as 70Å apart from the viral membrane in the case of HIV), it has been argued that force transmission requires the establishment of contacts between this element and the viral membrane [21] . Moreover, the proximal location of PreTM covalently linked to TMD ensures the high effective concentrations favoring its insertion into membranes. Accordingly, the capacity of PreTM regions for inserting into membranes has been extensively studied in model vesicular systems using representative synthetic peptides. These studies further suggest a putative role of membrane-inserted PreTMs as promoters of lipid bilayer destabilization. In the following sections we review experimental results, obtained in different viral systems, all of which sustain the functional activity of viral PreTMs in fusion–fission, either by transmitting protein conformational energy into membranes, and/or by perturbing lipid bilayer integrity.

Hydrophobicity-at-interface: a tool for unraveling viral PreTM sequences

In previously published review work we described the application of Wimley–White (WW) scales to the detection of viral sequences that show tendency to partition into membrane interfaces [22] . The analysis was performed with the premise that glycoprotein globular ectodomains behave as nonconstitutive membrane proteins, i.e., initially soluble species capable of inserting spontaneously into uncharged cell target membranes in the course of the fusion process. Wimley and White calculated the water-to-membrane interface transfer free energies for each amino acid in the context of unfolded peptide chains (ΔG wiu-s) [23], [24]. These free energy values were postulated as a hydrophobicity scale that compiles the energetic components dictating initial partitioning of sequences into membranes, specifically when this occurs in the absence of repulsive or attractive electrostatic interactions. As compared to classical hydropathy indexes based on bulk-phase partitioning of side chain alone, WW “interfacial hydrophobicity scale” takes into account the effect of the membrane interface on partitioning and includes the energy cost of peptide bond partitioning into it. This originates important differences in residue hydrophobicities, aromatic residues, namely Phe, Tyr and Trp, appearing to be the most hydrophobic ones when located at membrane interfaces.

Hydrophobicity-at-interface WW scale was first applied to HIV-1 and Ebola FPs with the aim of confirming their tendency to spontaneously partition from water into the bilayer interface [25], [26], [27]. Features of HIV-1 and Ebola FP sequences as potential membrane anchors were readily evident upon Kyte–Doolittle (KD) hydropathy analysis [28] . Thus, KD and WW hydropathy plots displayed positive peaks coincident with the FP sequences. However, a further analysis combining both scales disclosed segregated peaks at the membrane-proximal region of HIV-1 gp41 ectodomain ( Fig. 1 A and Refs. [29], [30]). When compared to the KD peak spanning gp41 TMD, the positive maximum detected by WW was shifted 15–20 residues towards the N-terminus (continuous red plot in top panel of Fig. 1 A). The combined WW–KD analysis thus suggested the existence of a wide nonpolar region (approximately 40–50 residues long) at the C-terminus of the gp41 ectodomain ( Table 1 ). This region would be segmented into two domains: one, positioned somewhat farther from the protein C-terminus (residues 664–683 in HIV-1), which would remain stably inserted within the membrane interface (designated as PreTM [29] ), and another one located closer to the C-terminus that constitutes the TMD.

Fig. 1

Interfacial PreTM domains in viral fusion proteins as predicted by the combined WW–KD hydrophobicity analysis. A) Membrane-proximal external and transmembrane regions of HIV-1 gp41. Hydropathy (mean values for a window of 11 amino acids) was plotted according to the Kyte–Doolittle hydropathy index (black lines), Wimley–White interfacial hydrophobicity (top panel, red-continuous trace) or Wimley–White interfacial moment (bottom panel, red-dotted line). B) Predicted location of gp41 AIS-PreTM (red cylinder) at the viral membrane interface. The cryoET density map contour of SIV envelope spike is indicated by the gray surface [36] . C) Analyses of membrane-proximal external and transmembrane regions in selected examples.

Table 1

AIS-PreTM-TMD motifs in selected examples of viral fusion proteins

Protein a	Sequence b
HIV-1 Env	658-QELLELDKWASLWNWFNITNWLWYIKLFIMIVGGLVGLRIVFAVLSVVNR-707
SARS Co-S	1181-DLQELGKYEQYIKWPWYVWLGFIAGLIAIVMVTILLCCMTSCCSCL-1226
Ebola GP	639-GDNDNWWTGWRQWIPAGIGVTGVIIAVIALFCICKFVF-676
FIV Env	760-GKTGIQQLQKWEDWVRWIGNIPQYLKGLLGGILGIGLGVLLLILCLPTLVD-810
VSV-GP	449-NPIDFVEGWFSSWKSSIASFFFIIGLIIGLFLVLRVGIYLYIKLK-493

Numbering based on the HIV-1 HXB2, Ebola Zaire, FIV Petuma, SARS-CoV Urbani and VSV Indiana, strains/serotypes.

Sequences in boxes and bold span, amphipathic-at-interface and hydrophobic-at-interface domains, respectively, as determined by WW algorithm. Underlined sequences in italics denote hydrophobic transmembrane sequences estimated according to KD algorithm. The range of each sequence has been inferred from plots displayed in Fig. 1 .

Further development of the WW analysis included the calculation of the hydrophobic-at-interface moment [21], [31]. Hydrophobic moments measure the periodicity of residue distribution along the secondary structure elements [32], [33]. Preferential orientation of the hydrophobic residues towards one face of the structure element has been proposed to favor hydrophobic interactions between protein structural components and between membranes and protein sequences. Computation of the hydrophobic moment using the membrane interface-to-water transfer free energies (ΔG iwu) for each amino acid [23] suggested that gp41 PreTM can be elongated at the N-terminus to include a stretch whose face shows strong affinity for membranes, if folded into an α-helix ( Fig. 1 A red-dotted plot in bottom panel). It must be noted that this amphipathic-at-interface stretch could not be detected when KD or Eisenberg scales were used [21] . Moreover, the moment peak shown in Fig. 1 A (bottom panel) was present in different HIV-1 isolates and in other lentiviruses thereby suggesting that the N-terminal segment of gp41 PreTM consists in a conserved amphipathic-at-interface sequence, recently designated as AIS [34] .

In summary, the hydrophobicity analysis in Fig. 1 A predicts the existence of a gp41 region anchored to the viral membrane through an interfacial sequence, amphipathic at its N-terminus, which is followed by the fully hydrophobic transmembrane helix ( Table 1 and cartoon in Fig. 1 B). The physiological relevance of gp41 AIS-PreTM-TMD motif has recently received support from structural studies in intact SIV and HIV virions [35], [36]. Cryoelectron microscopy tomography has disclosed stalk regions that project as “legs” from the trimeric Env complex, with their “feet” just above the plane of the envelope (tripod-like structure, contour in Fig. 1 B) [36], [37]. This observation would be consistent with native Env AIS-PreTM monomers being located at the viral surface, and inserted in parallel to the external membrane monolayer plane.

Extension of WW analysis to envelope spikes of several virus families suggests that an interfacial location of membrane-proximal external regions might be a common structural feature [29], [38]. Fig. 1 C and Table 1 provide several examples that illustrate different outcomes of the combined WW–KD hydrophobicity analysis. The presence of a hydrophobic-at-interface, aromatic-rich sequence within the fusogenic S2 subunit of SARS coronavirus spike S was already reported [39], [40]. The analysis in Fig. 1 C adds to this finding the existence of a pattern similar to that previously described for HIV-1 gp41 (panel A). The plots thus indicate that the fully hydrophobic-at-interface SARS S2 PreTM sequence can be elongated at the N-terminus to include an amphipathic-at-interface stretch (see also Table 1 ). A comparable but different situation occurs in the case of the Ebola fusion protein. Ebola GP contains within the membrane-proximal region of the GP2 ectodomain an approximately 10aa-long sequence showing a high tendency to partition into membrane interfaces (Ref. [31] , see also Table 1 ). However, computation of the hydrophobic-at-interface moment for this element reflects the existence of a longer amphipathic-at-interface sequence that would span over ca. 17 amino acids ( Table 1 ).

Finally, the retroviral FIV Env and vesiculoviral VSV-GP spike sequences serve to illustrate an additional case. Both PreTMs are characterized by negligible mean interfacial hydrophobicity. The presence of several conserved Trp-s within fusogenic FIV gp36 subunit's membrane-proximal sequence was previously noted [41], [42]. These hydrophobic-at-interface residues are inter-dispersed with helical periodicity by charged residues that bear the lowest tendency to partitioning into membrane interfaces ( Table 1 ). Similarly, aromatic residues at the VSV G PreTM are intervened by stretches with low propensity for partitioning ( Table 1 and Ref. [43] ). As a result, in both cases the mean hydrophobicity is low and interfacial moment high, reflecting a propensity for interfacial insertion, but only upon PreTM folding as an α-helix.

In what follows we summarize a series of experimental results that sustain the involvement of the predicted interfacial PreTM domains in membrane-rearrangements leading to viral fusion–fission.

Structure–function studies of HIV-1 gp41 PreTM

Mutational analyses

Mutational analysis of gp41 MPER by Salzwedel and co-workers [44] provided compelling evidence to support PreTM involvement in HIV-1 fusion. Deletion of the entire PreTM (Δ665–682) abrogated the ability of the envelope glycoprotein to mediate cell–cell fusion without affecting the normal maturation, transport, or CD4-binding ability of the protein. This fusion phenotype was also demonstrated for W(1–5)A, a PreTM mutant devoid of interfacial hydrophobicity by substituting alanine residues for the five tryptophan residues within this sequence. Further analysis using a three-color cell–cell fusion assay [45] confirmed that gp41 lacking PreTM is incapable of mediating both membrane-and-volume mixing. Interestingly, the fusion induced by W(1–5)A mutant was blocked at the stage of non-expanding small fusion pores. This activity is similar to that exhibited by GPI-linked hemagglutinin (GPI-HA) of influenza virus [46] .

We applied interfacial hydrophobicity to the analysis of the numerous mutations generated by Salzwedel and co-workers ( Fig. 2 and Ref. [21] ). Hydropathy plots of mean values calculated for a sliding window of 5 amino acids displayed two defined positive peaks, consistent with a slightly longer PreTM that encompasses two segregated interfacial subdomains ( Fig. 2 ). The N-terminal peak was found to span the conserved amphipathic-at-interface helix or AIS domain [34] . We observed that a number of blocking deletions invariably correlated with the ablation and/or merging of interfacial subdomains [21] . An extension of this analysis is displayed in Fig. 2 B. Overall free energy of partitioning (left) and interfacial hydrophobicity distribution (right) was plotted vs. fusogenicity of N-terminal subdomain Ala mutants. Free energy of partitioning, computed as the sum of the free energies of the constituent amino acids, correlated linearly with activity of most substitution mutants, but displayed significant activity for ΔG iwu = 0, indicating that favorable PreTM partitioning into membranes contributes to, but is not essential for gp41-mediated fusion. In addition, partitioning free energy cannot account for the fusion-inhibitory effects of scrambled SC7 and W(1–3)A mutations (hollow symbols). The effect of these mutations on function seems to better correlate with distribution of hydrophobicity into two subdomains (right panel). When hydrophobicity-at-interface peak ratios were plotted vs. fusogenicity, a positive correlation was observed for these and other mutants. This observation supports the functional relevance of distributing interfacial hydrophobicity into two segregated subdomains.

Fig. 2

Hydrophobicity-at-interface analysis of gp41 PreTM mutants. A) Gp41 PreTM sequence. Mutated Trp residues are in bold. Bars on top span the two interfacial domains disclosed when mean hydrophobicity values are calculated for a window of 5 amino acids (see panel C). B) Correlation between PreTM overall free energy of partitioning (left) or peak intensity ratio (right) and fusogenicity (average number of syncytia in the field as reported by Salzwedel et al. [44] ). Plotted values in right panel correspond to peak 1/peak 2 intensity ratios as obtained from hydropathy-at-interface plots similar to that displayed in panel C. In both panels hollow symbols indicate values not correlating with the represented linear regressions. C) Effect WW hydrophobicity distribution on PreTM function. Left: Hydropathy-at-interface plots (window of 5 amino acids) for the PreTM-TMD regions (amino acids 660–720 in Env precursor) derived from the wild-type and mutants W(1,3,4)A and W(1–3)A. Fusion index as in previous panel. Right: Membrane fusion induced in a vesicular system by DKWASLWNWFNITNWLWYIK (empty circles), DKAASLWNAFNITNALWYIK (black circles) and DKAASLANAFNITNWLWYIK black squares) peptides, representing PreTM regions of gp160 wt, W(1,3,4)A and W(1–3)A mutants, respectively. Final extents (percentages after 30 min) of intervesicular mixing of lipids were determined as a function of the peptide-to-lipid mol ratios in 1-palmitoyl-2-oleoylphosphatidylcholine:cholesterol (1.5:1) vesicles.

A remarkable example of interfacial hydrophobicity distribution effect is that of W(1–3)A and W(1,3,4)A substitutions, only differing in the position of one Trp residue ( Fig. 2 C). While the former mutation abolished gp41-induced fusion, the latter maintained ca. 70% of wild-type activity. The hydropathy plots revealed the effect of these mutations on hydrophobicity distribution (left panel). Whereas W(1,3,4)A mutant retained two hydrophobic-at-interface segments, W(1–3)A showed a single interfacial domain (left panel). Panel C compares the fusogenic activities of peptides representing these sequences (PreTMp-s) in a lipid vesicle system. The peptide corresponding to the fusion-competent W(1,3,4)A phenotype was as efficient as the wild-type peptide in fusing lipid vesicles. In comparison, the peptide representing the fusion-defective W(1–3)A phenotype displayed lower activity. Thus, fusion measured in a vesicular system suggests that the observed fusion W(1–3)A phenotype might reflect this mutant's incapacity to induce membrane-restructuring.

Structural studies

In accordance with the TMD paradigm, PreTM is postulated to fold within the low-polarity membrane interface environment adopting a helical conformation [24], [47], but also self-aggregate therein [48] . Structural studies support both possibilities ( Fig. 3 ). NMR spectroscopy revealed the formation of a well-defined helical structure for KWASLWNWFNITNWLWYIK peptide monomers in dodecylphosphocholine micelles (panel A and Ref. [49] ). Four of the five tryptophan residues, as well as the tyrosine residue, formed a “collar” of aromatic residues along the axial length of the helix. Nuclear Overhauser effects to the headgroup and interfacial protons of protonated dodecylphosphocholine confirmed that the aromatic residues are positioned within the membrane-water interface of a phospholipid bilayer sustaining a “Velcro-like” interaction. The bulk of the polar residues is positioned on one face of this structure, with the hydrophobic phenylalanine side chain on the opposing face, forming an amphipathic structure.

Fig. 3

Structural features of gp41 PreTM. A) NMR structure of KWASLWNWFNITNWLWYIK in DPC micelles [49] (PDB entry: 1JAU). Aromatic side chains are displayed. B) X-ray structure of KGWNWFDITNWGK peptide bound to Fab 4E10 [50] (PDB entry: 1TZG), 2 peptides displaying aromatic side chains have been oriented to show the interaction in the crystal unit cell. C) CD spectra of PreTMp (DKWASLWNWFNITNWLWYIK) in DPC (5 mM) micelles (dotted) and POPC vesicles (continuous trace). Lipid-alone signal was subtracted from these samples. D) Fourier self-deconvolved infrared spectra of PreTMp incorporated into POPC membranes. The peptide was added to vesicles at 1:50 and 1:1000 peptide-to-lipid mol ratios, top and bottom spectrum, respectively. Lipid–peptide complexes were subsequently isolated from the floating fractions after ultracentrifugation in D2O buffer. The spectra are not drawn to the same scale to better show the 1680–1620 cm− 1 amide I region. Structures in panels A and B were rendered with the Swiss-PDB-viewer program.

An additional PreTM atomic structure was obtained from the resolution of Fab 4E10 (see below) in complex with KGWNWFDITNWGK peptide (panel B and Ref. [50] ). The bound peptide also adopts a helical conformation in which residues contacting the paratope, W672, F673, I675, and T676, map to one face of the helix. Interaction of two peptide chains in the unit cell shows the close interdigitation of W670 and W678 indole side chains. Dimers perhaps mimic a low-energy conformation in the intact gp41 oligomer or reflect a biologically important interaction of the peptide Trp residues with the viral membrane [50] .

Low-resolution structural techniques provide evidence for the existence of monomeric and multimeric PreTM states in a membrane environment (panels C and D). CD helical spectrum of monomeric PreTMp in DPC micelles displayed a [θ]222/[θ]208 ratio < 1 (dotted trace in panel C, see also Ref. [49] ). By comparison, CD spectrum of PreTMp incorporated into PC vesicles at a 1:33 peptide-to-lipid mol ratio (continuous trace) resulted in an increase of the intensity of the positive peak together with a [θ]222/[θ]208 ellipticity ratio > 1, which supports the existence of helix–helix interactions such as those described for heterodimeric coiled-coils and/or transmembrane helical bundles [51], [52], [53]. Fluorescence data indeed indicated that rhodamine-labeled PreTMp in lipid vesicles self-associates at the doses used to carry out these measurements [21], [54] (see also Fig. 4 ).

Fig. 4

HIV-1 gp41 PreTMp surface aggregation and correlation with induced membrane-restructuring in lipid vesicles of different composition. The percentage of Rho-PreTMp quenching (triangles and dotted lines) and PreTMp-induced leakage (circles and continuous lines) and lipid-mixing (squares and dashed lines) after 30 min of incubation with vesicles was plotted as a function of the lipid-to-peptide mol ratio. Final lipid concentration was 100 μM in all cases.

IR spectroscopy confirmed the presence of helical structures as both, monomers or multimers in PC membranes (panel D and Ref. [21] ). At the peptide-to-lipid ratio of 1:1000, where the peptide is most likely monomeric according to the rhodamine fluorescence data, PreTMp spectrum showed a symmetrical band at 1653cm− 1 consistent with a canonical α-helix ( Fig. 5 B). At the 1:50 ratio where self-association is prominent, the α-helical band splitted into two components, absorbing at 1656 and 1649 cm− 1, besides the presence of a band at 1639 cm− 1, as was postulated previously for coiled-coils [55] . Thus, combined CD and IR structural data are compatible with the existence of a monomeric helical structure, prevalent in membranes at low doses, which would self-associate with increasing membrane loads.

Fig. 5

Inhibition of HIV-1 gp41 PreTMp-induced membrane-restructuring by Nab 4E10. A) Model for epitope recognition at the membrane interface by Fab 4E10 (Solid ribbons). Space-filled atoms in blue denote hydrophobic-at-interface residues exposed to solvent in CDR H2 and H3 loops. Atoms in green belong to bound KGWNWFDITNWGK peptide. Structure, derived from 1JAU PDB entry, was rendered with the Swiss-PDB-viewer program. B) Inhibition of ongoing ANTS leakage with Mab4E10. POPC:Chol (2:1) LUV suspensions (100 μM lipid) were treated with 1 μM PreTMp (arrow 1) and, further supplemented (arrow 2) with 25 μg/ml of Mab4E10 (black traces). Controls in absence of antibody are in red. Blue traces correspond to dansyl fluorescence increase (arbitrary units) upon PreTMp addition. Adapted from Fig. 3 in Ref. [103] . C) Recognition of PreTMp bound to ELISA plaques (circles), PC:Chol vesicles (squares) and PC:SPM vesicles (triangles). Inhibition percentages (Leakage) or percentages of maximal binding (ELISA) were plotted as a function of Mab4E10 binding site-to-PreTMp mol ratio (i.e., 2 binding-sites per Ma4E10 molecule).

Finally, it is important to note that the mutagenesis studies by Salzwedel et al. did not support the requirement of helical conformations in the function of this region [44] . The potential role of an α-helical PreTM structure in fusion was tested by deleting two central nonconserved residues, or substituting W672 for with proline. These mutations that would destabilize an α-helical conformation were found to have no significant effect on gp41 function, therefore arguing against a topological requirement for a helical structure. However, the reduction of the fusogenicity of the glycoprotein observed after scrambling PreTM central residues (mutant SC7) suggests that the clustering of hydrophobic residues in an α-helical secondary structure may indeed be important, but not essential, for the function of this region.

Membrane interactions of representative peptides

Early definition of HIV-1 gp41 PreTM as a membrane-inserting domain prompted the comparison of synthetic DKWASLWNWFNITNWLWYIK (PreTMp) and AVGIGALFLGFLGAAGSTMGARS (FPp) in their capacities for perturbing membranes [29], [30]. These studies were conducted under the assumption that PreTM might represent a second FP, i.e., an additional domain participating in the induction of the viral membrane perturbations required for fusion. Sustaining this notion, lower doses of PreTMp were required to induce fusion and permeabilization of liposomes [29], [30]. Moreover, PreTMp-induced restructuring effects were modulated by the main lipids existing in the viral envelope ( Fig. 4 and Ref. [54] ). Aloia and co-workers [56], [57] found in HIV-1 membranes increased Chol-to-phospholipid molar ratios and high levels of sphingomyelin (SPM), ca. 2–3 times that of the host-cell surface membranes. These findings were subsequently confirmed using a lipidomic approach [58] . It is therefore assumed that the HIV-1 membrane is enriched in Chol and SPM, an observation that correlates with the virion selectivity for specific segregated membrane regions through which virions emerge during maturation [59] .

Results displayed in Fig. 4 illustrate the effect of these lipids on PreTMp-induced leakage and fusion of liposomes. PreTMp induced leakage from pure PC liposomes at high peptide-to-lipid ratios (ca. > 1:100 mol:mol) concomitant to an increase in its self-aggregation, as revealed by an increase in the quenching level of Rhodamine-labeled species from approximately 40 to 80% (PC labeled panel). No significant fusion could be detected even at the highest doses tested, namely 1:10. Under conditions that did not significantly affect partitioning, inclusion of SPM or Chol in POPC membranes had dissimilar effects. SPM sustained the formation of lytic pores even when Rhodamine indicated quenching levels below 60% (PC:SPM panel). In contrast, Chol promoted fusion activity under self-aggregation levels (60–80% rhodamine quenching) also inducing leakage (PC:Chol panel). A similar Chol-dependence for PreTMp-induced vesicle–vesicle fusion was found in experiments comparing PreTMp with longer derivatives [60] .

When the three lipids were combined (PC:SPM:Chol panel) to emulate conditions putatively existing at the external membrane monolayer of HI virion, high levels of self-aggregation (rhodamine quenching > 80%), concomitant to an enhancement in fusion and lytic activities were observed at low doses (in the order of 1:1000 peptide-to-lipid ratio). Conditions described to disturb occurrence of lateral separation of s0–l0 lipid phases in these systems, reproduced the high peptide-dose requirements for leakage as found in pure POPC vesicles and inhibited fusion [54] . In summary, the quenching efficiency data as a function of the peptide-dose in LUV suggests that indeed peptide-aggregates are involved in membrane perturbations, and that s0–l0 phase-coexistence promotes PreTM clustering at the mol ratios existing in the HI virion envelope. Thus, these results support the notion that the PreTM domain aids in the clustering of gp41 within the HIV-1 envelope, and in destabilization of the bilayer architecture at the loci of fusion.

The existence of a direct interaction between Chol and PreTM domain was already posited by Salzwedel et al. [44] . PreTM carboxy-terminal LWYIK sequence was later identified as a potential “cholesterol recognition/interaction amino acid consensus” (CRAC) by measuring the binding to cholesterol of a series of gp41-derived sequences fused to maltose binding protein [61] . Recently reported experimental evidence and modeling studies provide the physicochemical grounds for a direct interaction of PreTM-CRAC with Chol [62], [63], [64]. PreTM–Chol interaction would be based on two capacities of CRAC sequence, namely, wrapping and blocking of the interfacial cholesterol OH group by H-bond interactions, and stacking of aromatic side chains with A ring of cholesterol. This interaction model has been proven through site-directed mutagenesis [63] . A conservative L679I substitution, which otherwise reduces the interaction of CRAC-representative peptides with Chol, resulted in a significant attenuation of Env fusogenicity measured in a cell–cell fusion assay. Thus, these results established a direct link between capacity to bind Chol-rich domains and biological activity.

In conjunction, vesicle perturbation assays and calorimetry determinations suggest that a PreTM–Chol complex might form at the interface of the external viral membrane monolayer with the potential of inducing membrane perturbations upon self-assembly [54], [62], [63]. However, as discussed by Epand [64] , the fact that L679I mutation did not completely abolished gp41 function posits the possibility that a fraction of the fusion activity measured in cells is not dependent on PreTM–Chol interactions. Sustaining the notion that Chol-dependent membrane perturbations induced by PreTM are not absolutely required for gp41-induced cell–cell fusion, an Env construct bearing FP blocking mutation and functional PreTM did not destabilize the plasma membrane upon receptor/co-receptor engagement [65] . Conversely, activated Env with functional FP but deleted PreTM was able to permeabilize the plasma membrane of expressing cells.

Studies on viral PreTM domains involved in membrane fusion

A PreTM domain detected within gp36 fusogenic subunit has been suggested to play a functional role in FIV entry [41], [42], [66]. A synthetic peptide modeled on such motif inhibited FIV replication very efficiently, most likely by blocking entry [66] . In a later report, 11 mutated clones having the Trp-rich motif scrambled or variously deleted or substituted were produced and evaluated in their capacity of infecting susceptible cells [41] . The functional effects of conserved Trp mutations were analyzed on assembled virions. The viral particles generated by the constructs having the Trp motif deleted, scrambled, or Ala substituted at two or three Trp positions were indeed replication-incompetent. In the latter mutants infection was restored by providing an independent source of intact FIV envelope glycoproteins or by addition of the fusing agent polyethylene glycol, thus indicating that their defect resided primarily at the level of cell entry. The fact that FIV PreTM was functionally active in fusion was supported by results of another site-directed mutagenesis study [42] . In this case, the W(1–3)A Env protein showed similar levels of expression and processing as the wild-type, but was completely defective in forming syncytia. It was concluded that this region must be critical for fusion by FIV Env and that the presence of tryptophans is required for this property.

The high degree of conservation observed upon sequence alignment of envelope glycoprotein (G) MPERs from several different vesiculoviruses suggested that PreTM might be also important for VSV G function. This was confirmed by Whitt and co-workers [43], [67], [68] who showed the involvement of this domain in VSV fusion and budding (see next section ). Deletion of the 13 stem amino acids (N449 to W461) that span the VSV AIS domain ( Table 1 ) dramatically reduced cell–cell fusion activity and reduced virus infectivity approximately 100-fold [43] . However, mutation of conserved aromatic residues responsible for amphipathicity (W457, F458, and W461) either singly or together had only modest effects on cell–cell fusion activity. Notably, PreTM-TMD-cytoplasmic tail constructs (GS) expressed in cells could also potentiate the membrane fusion activity of heterologous viral fusion proteins [68] . For some fusion proteins, there was as much as a 40-fold increase in syncytium formation when GS was coexpressed compared to that seen when the fusion protein was expressed alone. Fusion potentiation by GS was not protein specific, since it occurred with both pH-dependent as well as pH-independent fusion proteins. Truncation mutations indicated that the complete VSV AIS was required for fusion potentiation. Together, these data support the hypothesis that the PreTM domain contributes to VSV G-mediated membrane fusion activity and functions as a fusogenic unit, yet the conserved aromatic residues seem to be not essential for membrane fusion or infectivity in the context of the assembled virus.

The membrane interactions of Ebola PreTMp-s, studied in a vesicular system, supported the involvement of Ebola G PreTM in membrane fusion ( Fig. 1 and Ref. [31] ).Thus, DNDNWWTGWRQWIPAGIG and DWGDAPNWRWNWGIIGTQ, representing the Ebola glycoprotein PreTM or a “scrambled” control with a different hydrophobic-at-interface moment, respectively, were characterized in a study that combined the assessment of insertion into lipid monolayers, changes in intrinsic fluorescence and infrared spectroscopy [31] . Only the former bound the membrane interface under equilibrium conditions and adopted therein the predicted α-helical conformation. Following a pattern similar to that previously described for HIV PreTMp, the presence of the raft-associated lipid sphingomyelin did not affect membrane insertion, but it stimulated highly the membrane-destabilizing capacity of Ebola PreTMp. The SPM effect seemed to not arise from an intrinsic instability of the vesicles, since the scrambled control sequence could not induce leakage at any tested dose even though it interacted with liposomes. The outcome of this study underscored the importance of residue distribution within the Ebola PreTM sequence, specifically for the induction of membrane perturbations.

WW hydrophobicity analyses also put forward the existence of FP and PreTM-like membrane-transferring domains in SARS-CoV S sequence [38], [39], [40]. Membrane interactions of S2 PreTM-representative KYEQYIKWPWYVW peptide were subsequently characterized by means of CD, as well as in lipid vesicle permeabilization assays [39] . Even though the structural analyses indicated a low propensity for a defined secondary structure, S2 PreTMp strongly partitioned into lipid membranes and induced lipid vesicle permeabilization at peptide-to-lipid mol ratios of 1:100. It was found that a scrambled YEWKWIYWYPVKQ peptide was severely impaired in its capacity to induce vesicle permeabilization. Thus, partitioning of the wild-type peptides into the lipid interface was sufficient to specifically disorganize membrane integrity. A mechanism was proposed according to which CoV S PreTM, aligned with the FP and TMD during membrane apposition, would function to perturb the target cell membrane, and also to provide a continuous track of hydrophobic surface. Assembly of this structure would result in lipid-membrane fusion and subsequent viral nucleocapsid entry.

A combined KD and WW analysis was also employed to unravel Herpes simplex virus type-1 (HSV-1) glycoprotein H (gH) sequences showing propensity to interact with membrane interfaces [69], [70]. A putative helical membrane-interacting domain was identified just preceding the gH TMD [70] . A collection of HSV-1 gH PreTMp-s was synthesized and characterized. It was found that STALLLFPNGTVIHLLAFDTQPVAAIA (designated as pTM5) induced fusion of PC/cholesterol-containing phospholipid vesicles, while SHVLTAPALTFNLTDFVPILALAGIQA, a scrambled version with same amino acid composition, was defective in that activity. Interestingly, fusion activity in this case was related to the capacity of the PreTMp wild-type to adopt a β structure in contact with membranes. It was concluded that, similar to amino-terminal FPs, structural flexibility, rather than the rigid adoption of a particular secondary structure, might be a key property of the HSV-1 PreTM domain as a fusogen.

Interfacial domains in viral membrane fission

Completion of the newly assembled virions budding at cell membranes, known as “pinching-off” step, requires membrane fission [1], [2], [71], [72], [73]. Conformational energy release occurs in this case accompanied by multimerization of capsid proteins into spherical surfaces that grow in contact with the internal membrane monolayer. As one of the best-studied examples, Gag polyprotein is the main responsible for budding of immature HIV-1 particles, a process involving a complex interplay between viral and cell-host factors [72], [74]. Gag-membrane association proceeds via the N-terminal myristoyl moiety of matrix (MA) domain coupled to surface oligomerization [75], [76], [77]. Thus, Gag-membrane association and assembly therein are likely to promote each other, thereby driving the budding process.

Similarly, the minimal components needed to initiate, drive, and complete the Rhabdovirus budding process are the condensed RNP core in association with the matrix (M) protein [78] . However, in this case, in the absence of virally encoded glycoproteins the number of virus particles released from infected cells is significantly less than that made by WT virus. Therefore, the envelope glycoprotein contributes to virus egress. Involvement of VSV G PreTM in this process was studied using a series of recombinant ΔG-VSVs expressing chimeric glycoproteins with truncated stem sequences [67] . The recombinant viruses having chimeras with 12 or more PreTM residues, and including the G protein transmembrane-cytoplasmic tail domains, produced near-wild-type levels of particles. In contrast, viruses encoding chimeras with shorter or no PreTM sequences produced ca. 10- to 20-fold less. In line with its putative association with the membrane interface, these authors suggested that the VSV G PreTM might promote virus release by inducing positive membrane curvature at sites where virus budding occurs.

Alphavirus 6K protein provides a further opportunity for the study of interfacial domain activity in fission: firstly, this small protein's sequence comprises a single PreTM-TMD motif [79] , and, secondly, 6K is not required for alphavirus budding, but catalyzes the process [80] , which allows recovery of budding defective, but still infectious particles, upon cell transfection with alphaviral replicon.

The budding of alphavirus particles is essentially promoted by interaction of assembled nucleocapsids with the carboxy domain of envelope glycoprotein E2 [73], [81]. Despite the association of the 6K protein with the plasma membrane and its interaction with E1–E2 glycoproteins, very little 6K is incorporated into the budded virus particles. The N-terminal 6K ectodomain encompasses two hydrophobic-at-interface segments that can mediate association of this sequence with the external membrane monolayer [79] . Conservation of the interfacial 6K segments among divergent members of the Alphavirus genus suggested a functional role for these motifs. Accordingly, Sindbis virus (SV) 6K variants that contained substitutions interfering with the capacity of the N-terminus to partition into membranes without affecting KD hydrophobicity were obtained and characterized. Electron microscopy of cells transfected with these mutants disclosed accumulation at the plasma membrane of otherwise morphologically normal virus particles, suggesting that mutations introduced in the interfacial 6K sequence resulted in impaired virus budding, likely due to inefficient pinching-off of the assembled virus particles.

PreTM domains as targets for viral fusion inhibition

Since entry inhibition may block early stages of viral infection, the large conformational changes that are the energy source for membrane fusion constitute an important target in development of antiviral therapies. A paradigmatic case is that of the human immunodeficiency virus, (HIV). The HIV fusion inhibitor T-20 (enfuvirtide) is currently in use within the context of rescue therapies active against multidrug resistant viruses [82] . This peptidic compound dominantly competes with gp41 helical sequences for binding to hydrophobic grooves transiently exposed at the ectodomain surface [83], [84]. Thus, it is postulated that T-20 blocks close contact between cell target and viral membranes by preventing 6-HB formation.

In principle the functional domains that transmit 6-HB formation energy into membranes, such as FP or PreTM, may constitute alternative targets for antiviral development. Compelling evidence that sustains suitability of FP function as a target for inhibitor development has been again obtained in the case of HIV-1. Early studies showed that oligopeptides homologous in sequence to the FP were indeed capable of inhibiting gp41-induced fusion [85], [86], [87], [88]. Also supporting this view we have recently identified gp41 inhibitors from a d-amino acid hexapeptide library, based on their capacity to block synthetic FPp in vitro [89] , and Munch et al. [90] have isolated a naturally produced oligopeptide targeting FP that was able to block HIV-1 infection. Thus, force-transmitting sequences emerge as alternative therapeutic targets [90], [91].

The ability displayed by HIV-1 PreTMp to self-associate in solution and membranes raised the possibility of PreTM homo-oligomer formation at some stage during gp41-induced fusion process [21], [54]. It was inferred that PreTM homo-oligomer formation might represent a target for potential inhibitors. This hypothesis was tested analyzing PreTMp ability to interfere with cell–cell fusion induced by HIV-1 envelope [21] . HIV-1 PreTMp inhibited gp41-induced fusion in a dose-dependent manner with an apparent IC50 value of 577 nM. A W(1–3)A mutant-like peptide unable to form homo-oligomers was used as negative control and shown to be devoid of inhibitory effect. In control experiments, neither of these two peptides inhibited influenza hemagglutinin-mediated fusion implying that PreTMp inhibitory effect was through its interaction with HIV envelope protein, and not through its partitioning into cell membranes and causing non-specific perturbations.

As mentioned before, a cognate 20-mer peptide derived from FIV transmembrane glycoprotein PreTM, potently inhibited viral infectivity in tissue culture (IC50  ≈ 4 nM) [66], [92], [93]. Antiviral activity of the leading PreTM sequence was found to map to a short segment containing three conserved Trp residues. Thus, further analyses focused on a derivative of eight amino acids (770WEDWVRWI777), designated C8, which still displayed potent antiviral activity (IC50  ≈ 40 nM). C8 activity was found to be dependent on conservation of the Trp motif, most likely because these residues are responsible for its conformational propensity. Furthermore, the retroinverso analogue of this peptide (riC8) was shown to be effective in vivo, in experiments demonstrating its ability to decrease mean viral load when administered to chronically FIV-infected cats [92] .

Albeit with an apparent lower affinity, peptides analogous to the S2 subunit PreTM domain inhibited SARS-CoV plaque formation (inhibition by 40–70% at concentrations of 15–30 μM) [94] . In line with these observations, HSV gH PreTM-derived pTM4 and pTM5 peptides, modified by addition of a tail of lysine residues, were able to induce almost 100% inhibition of HSV-1 infectivity at concentrations in the range of 100 μM [70] . Underscoring the specificity of the observed effect, the scrambled version of pTM5 failed to inhibit infectivity under similar experimental conditions.

PreTM domains as targets for viral neutralization

Only two broadly neutralizing anti-HIV-1 antibodies isolated from peripheral blood mononuclear cells of HIV infected non-symptomatic patients have been described to react with gp41 [95], [96]. These antibodies, designated as 2F5 and 4E10, recognize epitopes within the conserved PreTM domain (reviewed in Refs. [97], [98], [99], see also Fig. 1 ). 2F5 (662ELDKWAS668) and 4E10 (671NWFNIT676) core epitopes roughly span AIS and the junction between the two PreTM interfacial subdomains, respectively. Both of them have developed paratopes capable of inserting into membrane interfaces, which putatively enables them for recognition of membrane-inserted epitope sequences [ [50], [99]– 102 ]. Recent experimental work by our group sustains specific recognition and blocking at membrane interfaces by both antibodies [100], [103], although following different molecular mechanisms.

2F5 epitope appears to be an intrinsically flexible sequence with access to different conformations depending on the molecular context [34], [104], [105], [106], [107], [108], [109]. Crystallographic analysis of the 656NEQELLELDKWASLWN671 sequence in complex with Fab 2F5 displays an extended conformation at the amino-terminus, followed by two consecutive type I β-turns, which is assumed to represent a defined, native-like gp41 structure [101] . Thus, it is inferred that Mab2F5 will recognize with lower affinity the α-helical conformation adopted by AIS upon partitioning into the membrane interface [100] . In contrast, 4E10 antibody binds PreTM epitope in the helical conformation that is also stable at membrane interfaces [50], [110], [111] ( Fig. 3 ).

MAb4E10 binding to the membrane-inserted epitope is reflected by its capacity of blocking PreTMp-induced membrane-restructuring effects ( Fig. 5 and Ref. [103] ). The fact that membrane-restructuring blocking was dependent on the presence of Chol suggests an adaptation of 4E10 antibody for recognition of PreTM species residing at the virion external membrane interface [103] . In contrast, 4E10 did not greatly affect PreTMp-induced perturbations in PC:SPM vesicles, most likely because PreTMp surface gets occluded in the transmembrane pores responsible for permeabilization of this type of vesicles [54], [103]. Recognition at membrane interfaces was also made evident by fluorescence microscopy of lipid monolayers transferred onto glass coverslips [103] . In these experiments, addition of Mab4E10 was shown to induce surface clustering of PreTMp sequence in presence of Chol, but not in its absence. It was concluded that a putative mechanism of gp41 blocking might be based on Mab4E10 capacity for rendering cross-linked spikes incompetent for fusion.

In summary, adopted conformation, aggregation state and topology as affected by the envelope lipids may condition Mab4E10 recognition of membrane-inserted PreTM epitope.

Concluding remarks

In principle we cannot exclude the possibility that PreTM folds as a component of globular ectodomains, either during synthesis or at some stage before protein activation [112] . More likely, PreTM is recruited into the 6-HB during progression of the reaction mediated by class I fusion proteins. The observation that 2F5 and 4E10 epitopes disappear concomitant to fusion [113] , i.e., coupled to 6-HB completion [114] , would sustain this notion in the case of HIV-1 gp41. In fact, Mab2F5 recognizes its epitope with lower affinity in the context of synthetic 6-HB versions [115] . Thus, the absence of activity observed in cell–cell fusion assays upon complete gp41 PreTM ablation probably originates from the incapacity for coupling 6-HB formation to membrane merger [45], [65].

However, as disclosed by WW analyses, membrane-inserted PreTM states exist with high probability ( Fig. 1 ), a prediction that seems to fulfill for native HIV Env glycoproteins incorporated into virions [36], [37]. In the context of cells expressing PreTM motifs, membrane-inserted PreTM states appear not to be absolutely required for initiation and progression of fusion or fission, their functional relevance being circumscribed to an assisting effect on those processes. In support of this view, a HIV-1 gp41 devoid of PreTM interfacial hydrophobicity is still competent in opening small fusion pores [45] (see also Fig. 2 ), ΔG-VSV is capable of inducing hemifusion on its own and also of functioning as a potentiator of fusion induced by other viral glycolproteins [68] , and SV 6K catalyzes budding by promoting the pinching-off step [79] .

Differential surface increase of the external membrane monolayer upon insertion might provide a common molecular mechanism to explain these catalytic effects [1], [2]. Monolayer surface imbalance might contribute to membrane deformation and bending (positive curvature) at the points of virus budding. It is inferred that imparting positive curvature to distal trans-monolayers engaged in pinching-off will promote completion of the process [1], [116]. More difficult to envisage is how positive curvature applied to contacting cis-monolayers might promote fusion. A possibility is that fusion is actually restrained by PreTM-induced positive curvature (committed state) until the process is activated. In such a case, extracting PreTM out of the interface, for instance by recruiting it into the growing 6-HB structure, might differentially reduce the surface of contacting monolayers, thereby stimulating fusion.

More specific structural and functional roles that might explain the high degree of HIV-1 PreTM sequence conservation, can be related to assembly, maturation and fusion activation of virions. These roles are likely to depend on its folding as an interfacial α-helix, which binds Chol, self-assembles in a raft environment, and is endowed with the capacity of restructuring this type of membranes. Indeed, most if not all PreTM mutants described by Salzwedel et al. dramatically affected incorporation of glycoprotein into virions [44] . These authors were the first to propose that one possible role for the conserved tryptophan residues in enhancing the incorporation of glycoprotein into virions could involve an interaction with cholesterol. Subsequently reported experimental evidence supports this notion [61], [62], [63], [64]. Such interactions could facilitate the enrichment of glycoprotein within plasma membrane raft-domains through which the virus might bud [59] , or even be employed in generating Chol-rich domains [62], [117]. Thus, PreTM might function as a raft-sensor to pilot Env protein to budding sites during viral assembly. Furthermore, specific folding and self-association within a Chol-enriched raft environment might be crucial for gp41 function [56], [118], [119], [120], [121].

The low spike density existing in the virion envelope (≈ 10–20 spikes and ≈ 300,000 lipids, see Refs. [35], [58]) might in part account for HIV-1 PreTM sequence conservation. It is tempting to speculate that clusters of inserted PreTMs might be initially responsible for the induction of local positive curvature within laterally segregated areas of the external viral membrane monolayer. As a matter of fact, electron tomography of intact virions reveals spikes distributed forming clusters of at least seven members [36] , and clusters of rod-shaped proteins at the focal points where the actual membrane merger develops [122] . Within these distinct areas, monolayer intrinsic curvatures could hypothetically change their sign upon PreTM desorption, thus facilitating fusion pore opening.

An alternative not exclusive option is that PreTM acts as an FP-like fusogenic sequence in laterally segregated PC-Chol domains, by promoting and/or stabilizing fusion pore architecture [29], [30], [123]. Sustaining this notion HIV-1 PreTMp partitions into POPC or POPC:SPM membranes but is only fusogenic when Chol is added to the lipid composition ( Fig. 4 and Ref. [54] ). Chol requirement actually supports that the membrane-restructuring processes observed in liposomes are indeed bona fide reflection of a gp41 PreTM physiological role [118], [121] ( Fig. 4 ). Moreover, PreTMp membrane activity is affected by blocking mutations ( Fig. 2 ), occurs at membrane doses relevant for spike density ( Fig. 4 ) and is inhibited by neutralizing antibodies ( Fig. 5 ). Attending to data reported in the literature, such an intrinsic membrane-perturbing activity could be a general feature of viral PreTMs [21], [29], [30], [31], [39], [40], [41], [54], [60], [70], [79], [124], [125], [126]. However, structure–function requirements may vary depending on the virus family, and adoption of an interfacial helical conformation seems not to be a prerequisite for membrane-disruption induced by the different PreTMp-s [30], [39], [70], [124].

The close proximity to the membrane surface of TMD-linked PreTM implies an effective concentration high enough as to push partitioning equilibrium towards membrane-inserted species. Even for sequences that are not specially enriched in aromatic residues, a mild degree of hydrophobicity together with access to a defined secondary structure upon partitioning, is likely to suffice for stable insertion into the membrane interface. Thus, there arises the question of long aromatic-rich PreTM domains as representing an adaptive improvement in the context of viral infection strategies. Again, HIV-1 provides an example of the way this adaptation might operate.

Most surface of native gp41 in virions remains inaccessible to neutralizing antibodies due to steric occlusion by surface gp120 subunit and/or homo-oligomerization [97], [98], [99]. Just the MPER stem region is potentially exposed to solvent [112], [127]. Thus, the probability of neutralization by binding to MPER epitopes might be reduced by two complementary mechanisms: i) limiting the envelope spike number (i.e., reduction of binding-sites and concomitant avidity decrease), and ii) occluding MPER epitopes into the interface of the viral external membrane monolayer (affinity decrease). Long aromatic-rich PreTM might play a central role in driving both effects. In one hand, as discussed before, its capacity to self-associate at the raft-like membrane surface would buttress the reduction of the number of spikes at the viral surface required for fusion. On the other, its high degree of hydrophobicity-at-interface ensures embedding of stem epitope sequences into the membrane, thereby occluding key residues for recognition and/or inducing membrane-specific secondary structure.

Indeed, insertion into membrane interfaces, and even phospholipid recognition therein, have been postulated to correlate with the broadly neutralizing capacity of anti-gp41 2F5 and 4E10 antibodies [100], [102], [128], [129], [130], [131], [132]. Mab4E10 is capable of binding and blocking PreTMp in membranes at peptide-to-lipid ratios relevant for the low spike virion density [100,102,103] . We have proposed a mechanism of neutralization by 4E10 based on its capacity to cross-link interfacial PreTM sequences [103] . According to that mechanism 4E10 would disorganize the spike clusters required for fusion initiation. Further studies of the mechanisms governing membrane-inserted epitope recognition will lead to new insights regarding the know-how to design effective anti-HIV-1 vaccines.

Note added to proofs

The structure recently reported by Sun et al. (Immunity 28 (2008), 52-63) supports the bipartite nature of HIV-1 AIS-PreTM inserted at the membrane interface.