Peptide mimics of the vesicular stomatitis virus G-protein transmembrane segment drive membrane fusion in vitro.

The efficiency of cell-cell fusion mediated by heterologously expressed vesicular stomatitis virus G-protein has previously been shown to be affected by mutating its transmembrane segment. Here, we show that a synthetic peptide modeled after this transmembrane segment drives liposome-liposome fusion. Addition of millimolar Ca(2+) concentrations strongly potentiated the effect of the peptides suggesting that Ca(2+)-mediated liposome aggregation supports the activity of the peptide. Peptide-driven fusion was suppressed by lysolipid, an established inhibitor of natural membrane fusion, and involved inner and outer leaflets of the liposomal bilayer. Thus, transmembrane segment peptide-driven liposome fusion exhibits important hallmarks characteristic of natural membrane fusion. Importantly, the mutations previously shown to attenuate the function of full-length G-protein in cell-cell fusion also attenuated the fusogenicity of the peptide, albeit in a less pronounced fashion. Therefore, the function of the peptide mimic is dependent on its primary structure, similar to full-length G-protein. Together, our data suggest that the G-protein transmembrane segment is an autonomous functional domain. We propose that it acts at a late step in membrane fusion elicited by vesicular stomatitis virus.

The efficiency of cell-cell fusion mediated by heterologously expressed vesicular stomatitis virus G-protein has previously been shown to be affected by mutating its transmembrane segment. Here, we show that a synthetic peptide modeled after this transmembrane segment drives liposome-liposome fusion. Addition of millimolar Ca 2؉ concentrations strongly potentiated the effect of the peptides suggesting that Ca 2؉ -mediated liposome aggregation supports the activity of the peptide. Peptide-driven fusion was suppressed by lysolipid, an established inhibitor of natural membrane fusion, and involved inner and outer leaflets of the liposomal bilayer. Thus, transmembrane segment peptide-driven liposome fusion exhibits important hallmarks characteristic of natural membrane fusion. Importantly, the mutations previously shown to attenuate the function of full-length G-protein in cell-cell fusion also attenuated the fusogenicity of the peptide, albeit in a less pronounced fashion. Therefore, the function of the peptide mimic is dependent on its primary structure, similar to full-length G-protein. Together, our data suggest that the G-protein transmembrane segment is an autonomous functional domain. We propose that it acts at a late step in membrane fusion elicited by vesicular stomatitis virus.
Biological membrane fusion involves a restructuring of lipid bilayers brought into close proximity by membrane-anchored fusion proteins. To date, the most thoroughly characterized fusion proteins are those which mediate fusion of viral envelopes with cellular membranes (1). These viral fusion proteins consist of an ectodomain harboring an amphipathic fusion peptide, a single transmembrane segment (TMS), 1 and a cytoplasmic domain. All these domains appear to cooperate in fusion protein function (2,3). In the case of influenza hemagglutinin (HA), a pH-driven global conformational change of the ectodo-main is thought to eject the fusion peptide toward the target bilayer to establish initial contact between both membranes (4). Synthetic versions of many of these soluble fusion peptides drive liposome-liposome fusion in vitro. Furthermore, the fusogenic function of some synthetic fusion peptides is sensitive to point mutations in a similar fashion as the corresponding fulllength fusion proteins (2,5,6). Therefore, they appear to be partially independent functional domains whose interaction with target membranes is an early event initiating fusion.
A late role in fusion protein function has been ascribed to the TMSs. For example, upon replacement of the TMS by a glycosylphosphatidylinositol membrane anchor (7,8) or by mutating (9) its TMS, it was initially reported that influenza HA looses its ability to mediate complete bilayer fusion but retains hemifusion, i.e. lipid mixing of the contacting monolayers (7). More recently, lipid-anchored HA was found to induce an aqueous continuity between fusion compartments which, however, formed less efficiently and did not enlarge substantially compared with wild-type (wt) HA. Consequently, the HA TMS was suggested to favor pore formation and growth (10). Similar results were obtained with a point mutant (11). In another study, shortening the HA TMS reduced its ability to support the hemifusion to fusion transition whereas point mutations were without effect (12). Furthermore, the function of other fusion proteins derived from the human immunodeficiency virus type 1 envelope glycoprotein (13) or the Moloney murine leukemia virus (14) was compromized by mutations within the TMSs.
The trimeric vesicular stomatitis virus (VSV) G-protein is a fusion protein that is functionally and structurally related to influenza HA. Replacement of the VSV G-protein TMS by a glycosylphosphatidylinositol anchor rendered this protein nonfusogenic, suggesting that its function requires membrane anchoring by a hydrophobic peptide sequence (15). Similarly, a 6-residue deletion or point mutations within the TMS attenuated the fusogenicity of VSV G-protein as indicated by reduced transfer of cytoplasmic contents between fusing cells. However, the mutant proteins were shown to retain hemifusion activity since lipid probes within the external leaflets of the reacting membranes mixed (16). Similar to the case of influenza HA, the TMS of VSV G-protein is therefore thought to act at a late step in membrane fusion.
Here, we demonstrate that incorporation of a synthetic peptide corresponding to the VSV G-protein TMS into liposomal membranes dramatically increases their ability to fuse. Interestingly, those mutations previously shown to affect full-length VSV G-protein function in HeLa cells also attenuated TMSpeptide-induced liposome-liposome fusion. Thus, the TMS-peptide exhibits sequence-specific fusogenicity in the absence of the ectodomains and thus appears to be an independent functional domain.

EXPERIMENTAL PROCEDURES
Peptide Synthesis-Peptides were synthesized by the standard Fmoc (N-(9-fluorenyl)methoxycarbonyl) solid-phase method on an Abimed AMS422 multiple peptide synthesizer. The peptide-resing conjugates were cleaved with trifluoroacetic acid. Upon dilution in formic acid and 10-fold dilution with water, peptides were purified by high pressure liquid chromatography on a YMC ODS-H80 reverse phase column using a 40-95% (w/v) gradient of 80% (v/v) acetonitrile including 1% (v/v) trifluoroacetic acid for elution. The identities of the peptides were confirmed by mass spectrometry using a Finnigan-MAT Vision 2000. Peptide concentrations were determined spectrophotometrically via tryptophan absorbance at 280 nm in a 1:1 (v/v) mixture of trifluoroethanol and dimethyl sulfoxide using an extinction coefficient of 5600 M Ϫ1 cm Ϫ1 .
Preparation of Small Unilamellar Liposomes-Liposomes were prepared from mixtures of egg PC/brain PE/brain PS (Sigma or Avanti Polar Lipids) at a ratio of 6:2:2 (w/w/w) with or without 0.8% (w/w) of NBD-PE and Rh-PE (Molecular Probes). Lipid solutions in chloroform, with or without TMS peptides previously dissolved in trifluoroethanol, were dried under a stream of nitrogen as a thin film, evacuated for at least 3 h, and rehydrated by shaking at 37°C with fusion buffer (25 mM Tris-Cl, pH 7.4, 150 mM NaCl, 0.1 mM EDTA, 5 mM dithiothreitol). Liposomes were formed by sonication (Bransonic Sonifier 250 microtip) under external ice cooling for 2 min and centrifuged at 16,000 ϫ g for 20 min to remove lipid aggregates (generally less than 10% of total lipid as judged from the loss of initial NBD-PE fluorescence). To extinguish NBD fluorescence in the outer leaflet, liposomes were incubated with 20 mM sodium dithionite on ice for 30 min and excess sodium dithionite was removed on Sepharose G-50 spin columns. Where indicated, egg LPC (Sigma) dissolved at 1.1 mg/ml in fusion buffer was slowly pipetted to the liposomes while vortexing.
Determination of P/L Ratios-The ratios of incorporated peptides to liposomal lipids were determined upon separating unbound peptides from proteoliposomes by density gradient centrifugation. 330 l of the liposome preparations were mixed with 670 l of 60% (w/v) sucrose, and overlaid with 3 ml of 30% (w/v) sucrose followed by 0.5 ml of fusion buffer. Upon centrifugation (60,000 rpm, 20 h, 20°C, Beckman SW60 rotor), Ͼ99% of the loaded lipids were found in the top fraction whereas unbound peptide was shown to stay in the bottom. The amounts of lipid-associated peptides were calculated from the tryptophan fluorescence of the top fractions ( ex ϭ 290 nm) using known amounts of added peptides as internal standards. The fluorescence spectra were corrected by subtraction of spectra run with pure control liposomes. Lipid concentrations were quantitated as described (17) and related to peptide concentrations to obtain the P/L ratios given. All P/L ratios were finally corrected for the dependence of tryptophan fluorescence on the hydrophobicity of the environment. To this end, the fluorescence of liposomal peptides and added standards was compared upon detergent lysis where all peptides are expected to reside within detergent micelles.
Fluorescence Spectroscopy-Fluorescence spectra of lipid-bound peptides were recorded from peptide-liposome complexes (P/L ratio ϭ 0.017-0.02) separated from free peptides by sucrose density gradient centrifugation and corrected for spectra recorded from pure liposomes. Solution spectra were recorded from peptides dissolved in dimethyl sulfoxide at a concentration of 8 M. Spectra were recorded using a Shimadzu RF1501 spectrofluorimeter; excitation wavelength was 290 nm, emission wavelengths ranged from 300 to 450 nm, and a slit width of 10 nm was used.
Fusion Assay-Fusion assays were done by a fluorescence dequenching assay (18). Briefly, fluorescently labeled "donor" and un-labeled "acceptor" liposomes (1.35 mg/ml phospholipid) were mixed at a ratio of 1:4 (v/v) on ice, transferred to 96-well microtiter plates (white NUNC non-binding plates with translucent bottoms), heated for 2 min by floating the plates on a 37°C water bath, and NBD fluorescence was immediately assayed at 1-min intervals for 60 min at 37°C using excitation and emission wavelengths of 460 and 530 nm, respectively (BMG Lab Technologies FluoStar). The values were corrected for drifts due to instrument response and liposome binding to the plastic surface as determined by the kinetics of NBD-fluorescence of donor liposomes in separate wells. Where indicated, a calcium chloride solution in fusion buffer was injected after the first minute. For calibration, the initial fluorescence intensity of donor liposomes was taken as 0% fusion and the maximal fluorescence seen upon addition of 0.5% (w/v, final concentration) Triton X-100 was taken as 100% fusion. Initial rates of fusion were calculated by drawing tangents at the steepest (initial) parts of the curves (19). All values were corrected for detergent quenching. The peptide-independent, spontaneous fusion of pure liposomes was routinely determined in parallel and subtracted from the values obtained with peptide-containing liposomes. The mean P/L ratio was 0.018 unless specified otherwise.

A Synthetic VSV G-protein TMS Drives Liposome-Liposome
Fusion-Synthetic peptides harboring the central 15 residues from the predicted TMS of VSV G-protein and a number of mutant sequences were synthesized by solid-phase chemistry. The hydrophobic residues are bordered by 3 lysine residues at both termini to enhance solubility and correct membrane integration (20,21). A tryptophan residue was included for photometric quantitation (Fig. 1). The peptides were successfully incorporated into liposomal membranes by sonication as previously demonstrated for a number of other hydrophobic peptides (21,22). Routinely, a mixture of egg phosphatidylcholine (PC), brain phosphatidylethanolamine (PE), and brain phosphatidylserine (PS) (at a 6:2:2 weight ratio) was used. This lipid composition had previously been found to optimally support the fusogenicity of TMS-peptides derived from SNARE (soluble NSF (N-ethylmaleimide-sensitive factor) proteins while exhibiting low spontaneous fusion in the absence of the peptides. 2 To determine the actual P/L ratios present in the fusion experiments, peptide-liposome complexes were separated from the free peptides, which are more dense, by sucrose density gradient centrifugation. The amounts of co-fractionating peptides and lipids were determined and the P/L ratios calculated (see "Experimental Procedures"). Integration of the peptides into membranes is supported by the fluorescence of the marker tryptophan residue. Depending on the identity of the peptide, fluorescence emission maxima ranged from 335 to 340 nm in the relatively hydrophilic solvent dimethyl sulfoxide and from 317 to 326 nm in peptide-liposome complexes (data not shown). These blue shifts (14 to 22 nm) indicate that the marker tryptophan residues of the peptides, and by implication, their hydrophobic sequence, are embedded within the hydrocarbon phase of the membranes (23).
The ability of the liposomes to fuse to each other was examined upon rapidly shifting the temperature to 37°C by a standard fluorescence dequenching assay (18). This assay is based upon fluorescence resonance energy transfer from N-(7-nitro-2,1,3-benzoxadiazol-4-yl)dioleoylphosphatidylethanolamine (NBD-PE) to N-(lissamine rhodamin B sulfonyl)dioleoyl phosphatidylethanolamine (Rh-PE) being present at quenching concentrations in donor liposomes. Upon fusion of donor liposomes to unlabeled acceptor liposomes, the average distance between the lipid-bound fluorophores increases; this results in an increase of NBD-fluorescence over time which is taken as a measure of lipid mixing.
As shown in Fig. 2, incorporation of the VSV wt TMS-peptide FIG. 1. Amino acid sequences of the peptides used in this study. Dots represent wt residues. The residue numbering system used by Cleverley and Lenard (16) was adopted. strongly increased the ability of liposomal membranes to fuse as compared with spontaneous background fusion observed with pure liposomes. Extents of fusion, seen after 1 h, and initial fusion rates were similar regardless of whether donor and acceptor liposomes contained peptide (VSV/VSV) or whether peptide-containing acceptor liposomes were fused to pure donor liposomes (VSV/pure).
TMS-Peptide-mediated Fusion Is Sensitive to Lysolipid and Involves Both Membrane Leaflets-Biological membrane fusion is thought to proceed through a state termed hemifusion, an intermediate where the outer, but not inner, membrane leaflets mix to form a stalk structure (1,2,24,25). Hemifusion is associated with negative curvature of the outer membrane leaflet. Hence, it is disfavored upon integration of lysolipids into the outer leaflet since lysolipids stabilize positive curvature due to their inverted cone shape (24). Accordingly, lysolipids have been established as potent and reversible inhibitors of different types of natural fusion reactions at a step preceding membrane merger (25). Here, we assessed whether VSV Gprotein TMS-peptide-mediated fusion is inhibited by adding LPC micelles to the preformed liposomes. Indeed, LPC efficiently inhibited TMS-peptide-mediated liposome fusion depending on the applied ratio of diacyl-lipids to LPC (Fig. 3A). This is consistent with transient formation of a hemifusion intermediate in the absence of LPC.
To show that fusion involved both membrane leaflets rather than being arrested at hemifusion, we extinguished the fluorescence of the NBD moiety present in the outer leaflet by dithionite treatment (19). Application of 20 mM dithionite elim-inated an average of 66% of total NBD fluorescence after 30 min on ice and the kinetics of bleaching slowed down considerably beyond this period (Fig. 3B, inset, and data not shown). This suggests preferential bleaching of NBD present in the outer monolayer which accounts for Ϸ64% of total lipid of unilamellar liposomes with mean diameters of Ϸ30 nm previously determined by electron microscopy. 2 Importantly, the kinetics of TMS-peptide-mediated NBD fluorescence dequenching was indistinguishable with or without prior outer leaflet bleaching (Fig. 3B). This indicates that inner and outer membrane leaflets fused with similar efficiency.
Addition of Ca 2ϩ Enhances Peptide-mediated Membrane Fusion-The observation that the TMS-peptides induce liposome fusion is unexpected since membrane fusion mediated by fulllength fusion proteins is regulated by interaction of their ectodomains with the target membranes. This interaction is thought to bring the membranes into close proximity thus lowering the energy barrier that normally prevents fusion. Since our TMS-peptides lack ectodomains, peptide-mediated fusion must depend on random collisions between the liposomes. To examine whether aggregating the liposomes would enhance the fusogenic action of TMS-peptides, we added Ca 2ϩ ions, which are known to aggregate negatively charged liposomes before fusing them (26,27). Upon correction for Ca 2ϩ -

FIG. 2. Fusogenic activity of wt VSV G-protein TMS-peptide.
A, typical fusion kinetics reveal that peptide-containing donor and acceptor liposomes (VSV/VSV) fuse with similar efficiency as pure donor liposomes fuse to peptide-containing acceptor liposomes (VSV/pure). Spontaneous fusion of pure control liposomes (pure/pure) proceeds much less efficiently. B, mean (ϮS.E., n ϭ six to nine independent experiments) extents of fusion, as seen after 1 h, and initial fusion rates are similar regardless of whether the peptide was present in donor and acceptor liposomes or only in acceptor liposomes. Both parameters were corrected for spontaneous fusion.

FIG. 3. Properties of liposome fusion induced by wt TMS-peptide.
A, the inhibitory effect of LPC (given as total LPC concentrations in the assay volume) is similar regardless of whether the peptide was present in donor and acceptor liposomes (squares and circles) or only in acceptor liposomes (diamonds and triangles). For better comparison, mean extents of fusion were normalized (0 mg/ml LPC ϭ 100%). The experiment was done at two different liposome concentrations (diacyllipid concentration was 1.35 mg/ml, squares and diamonds, or 0.45 mg/ml, circles and triangles). Accordingly, the ratio of diacyl-lipids to LPC ranged from 8 to 75. B, inner leaflet mixing after outer leaflet bleaching. Either both, donor and acceptor, liposomes, or only acceptor liposomes contained TMS-peptide. The extents of fusion, as determined by NBD dequenching, were similar with (ϩ) or without (Ϫ) prior dithionite treatment. This indicates fusion of both membrane leaflets. For better comparison, mean extents of fusion were normalized (Ϫ dithionite ϭ 100%). Inset, mean initial NBD fluorescence of donor liposomes was reduced by 66% upon dithionite treatment; this indicates preferential bleaching of the outer leaflet (see text). All values were corrected for spontaneous fusion seen with pure liposomes in parallel experiments. The number of independent experiments was three to four. enhanced background fusion as determined in parallel using pure liposomes, millimolar concentrations of Ca 2ϩ ions were indeed found to strongly increase both initial rate and extent of peptide-induced fusion in a concentration-dependent manner (Fig. 4). Therefore, juxtaposition of the membranes, mediated here by random collisions of the liposomes and potentiated by their Ca 2ϩ -mediated aggregation, appears to be a rate-limiting step in liposome-liposome fusion driven by VSV G-protein TMS-peptide.
Mutations Attenuate the Fusogenicity of the TMS-Peptide-Based on studies using eukaryotic cells, it was previously found that point mutations in the TMS affect fusogenicity of the expressed full-length VSV G-protein. Specifically, mutating the glycine residues individually to alanine reduced cytoplasmic contents mixing by Ϸ40% to 50%, whereas double mutations to alanine or leucine resulted in reductions of Ͼ90% (16).
Here, we tested whether mutant forms of the TMS peptides mimic this sequence specificity of G-protein function. Thus, we compared the fusogenicity of the wt peptide to that of mutant sequences where both glycines were either singly exchanged to alanine or doubly exchanged to alanine, leucine, or valine. Fig.  5A displays the dependence of the mean extent of fusion, as seen after 1 h, on the experimentally determined P/L ratios. In order to better visualize the differences determined at the lower P/L ratios, the mean values obtained for the mutant forms were normalized to the values seen with the wt peptide (ϭ100%). With the single mutations (G6A, G10A) the mean extent of fusion appeared to be only moderately reduced (by 21 to 35%) and only at the lower P/L ratios. For the double mutants (G6A,G10A; G6L,G10L; G6V,G10V) the extent of fusion was decreased by Ϸ50% at the lower P/L ratios and for 13 to 37% at the highest P/L ratios. We also tested mutants where residues other than the glycines were exchanged. Fusion elicited by peptide A3, where 3 residues in the vicinity of the glycines were mutated to alanine (see: Fig. 1), was reduced by 13 to 30%, depending on the P/L ratio. Strong reduction (71 to 91%) was observed for the L5 peptide where 5 exchanges to leucine had been made. A similarly weak fusogenicity was found for an oligoleucine sequence (L16(W)) which was previously shown to function as an artificial TMS (28). The degree of fluorescence dequenching elicited by the mutant peptides was similar with or without prior bleaching of the NBD chromophore of the outer membrane leaflet with dithionite. Thus, fusion elicited by the mutant peptides did not result from hemifusion but involved both membrane leaflets as seen for the wt peptide (see above).
Taken together, the glycine mutations affected the fusogenicity of the TMS-peptides in a fashion similar to that of full-length VSV G-protein (16), albeit less strongly. Thus, our results indicate that the TMS-peptides, at least partially, reproduce the sequence specificity of VSV G-protein function. DISCUSSION Our results demonstrate that incorporation of synthetic peptides modeled after the TMS of VSV G-protein into the membranes of synthetic liposomes strongly increases their ability to fuse. These peptides represent the central hydrophobic part of the predicted TMS which is located between lysine 462 and arginine 483. Since the hydrophobicity of our peptides is rather Extents of fusion were determined for wt TMS-peptide, several mutants (see: Fig. 1), and an oligo-leucine (L16(W)) peptide. A, a plot of the mean values (ϮS.E., n ϭ 2 to 10) against the experimentally determined mean P/L ratios shows that the extents of fusion increase with the P/L ratios. Importantly, most mutant peptides are less fusogenic than the wild-type. The data points were fitted by linear regression analysis. For the sake of clarity, the data for peptides G6A, G10A, and A3 were not fitted. All values were corrected for spontaneous fusion seen with pure liposomes in parallel experiments. Error bars are not shown when smaller than the symbols. B, the mean fusion extents determined for the mutant peptides were normalized to those of the wild-type (ϭ100%). This reveals that the differences in fusogenicity between mutants and wild-type tended to be larger at the lower P/L ratios. The designations of the peptides are given next to the data points obtained at the highest P/L ratio. symmetrically distributed, we assume that they fully insert as ␣-helices into liposomal membranes. This is supported by previous studies where the ability of other lysine-bordered hydrophobic peptides to incorporate into lipid bilayers was tested. Huschilt et al. (20) found that a KKGL 16 KK peptide inserted as an ␣-helix at a perpendicular angle into dipalmitoyl phosphatidylcholine membranes. Furthermore, Webb et al. (29) established that a hydrophobic peptide core of 17 residues fully incorporated into a membrane composed of lipids with C 18 acyl chains. The efficiency of these relatively short peptides to integrate into membranes is ascribed to the flexibility of the hydrophobic part of the bordering lysine side chains. This flexibility allows the charged termini to "snorkel up" toward the lipid head group regions. In a previous model study, the lysine termini bordering 17 hydrophobic residues adopting an ␣-helical conformation have been calculated to extend up to 1.9 nm from the bilayer center (22). By analogy to these model systems, our TMS peptides containing 16 hydrophobic amino acids are likely to span the hydrophobic core of liposomal membranes that are composed of natural lipids which contain mostly C 18 lipid acyl chains.
This in vitro fusion system consisting only of lipids and TMS-peptides appears to display characteristic hallmarks of biological membrane fusion. First, its sensitivity to lysolipid is consistent with the transient existence of a hemifusion state, which is thought to precede complete bilayer merger (2,24). Second, TMS-peptide-driven fusion involves both bilayer leaflets as shown upon bleaching the NBD fluorophore of the outer leaflet by dithionite. Third, both extent and initial rate of fusion are enhanced by Ca 2ϩ -mediated liposome aggregation, thus suggesting that fusion promotion by the peptides is potentiated by membrane proximity.
Importantly, mutating the glycine residues within the TMSpeptide attenuated its fusogenicity to various degrees depending on the number of exchanges made, the type of introduced residue, and the P/L ratio of the proteoliposomes. Since glycine mutations had previously been shown to affect cell-cell fusion mediated by full-length VSV G-protein (16), our liposomebased in vitro approach appears to reflect, at least in part, the sequence specificity of full-length fusion protein function.
In analogy to results obtained with influenza HA TMS mutants (9), the VSV G-protein TMS has previously been shown to function late in the fusion process (16). This late step appears to correspond to the conversion of the hemifusion intermediate to fully fused bilayers and is thought to be facilitated by the TMS. It should be noted that the extent of fluorescence dequenching seen here with pure liposomes, i.e. without added peptide, was reduced to an average of 60% upon outer leaflet bleaching. 3 This suggests that a substantial fraction of peptideindependent spontaneous liposome fusion does not proceed beyond hemifusion. We speculate, therefore, that the hemifusion diaphragm may form spontaneously from randomly colliding liposomes. Its transition to full fusion, however, appears to be facilitated by the TMS-peptides.
What is the characteristic feature of the TMS which drives this transition and which is required for optimal fusion protein function? The TMSs of fusion proteins are expected to be inserted as ␣-helices at orientations close to the membrane normal as demonstrated for the influenza HA TMS (30). A comparison of the amino acid composition of fusogenic and non-fusogenic viral membrane protein TMSs revealed that glycine residues are strongly over-represented in the former (16). It was therefore speculated that the functionally relevant glycines of the VSV G-protein might allow for bending of the TMS helices during the fusion process (16). Indeed, glycine residues have previously been shown to destabilize ␣-helical hydrophobic peptides in membrane-mimetic environments (31). On the other hand, isoleucine was the single other residue type that was also significantly overrepresented in fusogenic TMSs (16). Interestingly, in our A3 or L5 peptide mutants, two or four isoleucines were exchanged for alanine or leucine, respectively, and this resulted in a marked decrease of fusogenicity despite the continued presence of both glycine residues. Isoleucine ranks among those residues with the highest propensities to form ␤-sheet structures (32,33). This property was ascribed to steric interference of its ␤-branched side chain with the local polypeptide backbone which may destabilize the ␣-helical conformation (33). Although isoleucine has been reported not to interfere with ␣-helix formation in hydrophobic environments (23,34), it is tempting to speculate that not only glycine, but also isoleucine or other residues with ␤-branched side chains, impart structural flexibility to fusogenic TMS helices. Thus, local deformation or even transient unfolding of these helices may facilitate the restructuring of the lipid bilayer in fusion. It might be relevant in this context that the ␣-helical synthetic peptide corresponding to the influenza HA TMS increased the acyl chain order of a lipid bilayer (30).
On the other hand, it should be noted that the fusogenic function of VSV G-protein (15), influenza HA (9), or the human immunodeficiency virus type 1 envelope glycoprotein (35) was maintained upon replacement of the respective TMS by unrelated TMSs (9,15). This is an apparent contradiction to the effects of TMS point mutations as discussed above. The data may be reconciled, however, if one assumes that the structural feature(s) rendering a TMS compatible with fusion protein function are shared by certain unrelated TMSs which would be inactive in their normal contexts (2,14).
The fusogenic capacity of synthetic TMSs appears not to be restricted to the VSV G-protein. In another study, 2 we found that peptides modeled after the TMSs of SNARE proteins also drive liposome-liposome fusion in a sequence-specific fashion. As SNARE proteins are essential for membrane fusion in the secretory pathway of eukaryotic cells (36), functional autonomy of TMSs may be a phenomenon relevant for both viral and cellular fusion proteins.