Charged residues are involved in membrane fusion mediated by a hydrophilic peptide located in vesicular stomatitis virus G protein

Abstract

Membrane fusion is an essential step of the internalization process of the enveloped animal viruses. Vesicular stomatitis virus (VSV) infection is mediated by virus spike glycoprotein G, which induces membrane fusion at the acidic environment of the endosomal compartment. In a previous work, we identified a specific sequence in VSV G protein, comprising the residues 145 to 164, directly involved in membrane interaction and fusion. Unlike fusion peptides from other viruses, this sequence is very hydrophilic, containing six charged residues, but it was as efficient as the virus in catalyzing membrane fusion at pH 6.0. Using a carboxyl-modifying agent, dicyclohexylcarbodiimide (DCCD), and several synthetic mutant peptides, we demonstrated that the negative charges of peptide acidic residues, especially Asp¹⁵³ and Glu¹⁵⁸, participate in the formation of a hydrophobic domain at pH 6.0, which is necessary to the peptide-induced membrane fusion. The formation of the hydrophobic region and the membrane fusion itself were dependent on peptide concentration in a higher than linear fashion, suggesting the involvement of peptide oligomerization. His¹⁴⁸ was also necessary to hydrophobicity and fusion, suggesting that peptide oligomerization occurs through intermolecular electrostatic interactions between the positively-charged His and a negatively-charged acidic residue of two peptide molecules. Oligomerization of hydrophilic peptides creates a hydrophobic region that is essential for the interaction with the membrane that results in fusion.

• Membrane fusion
• fusion peptide
• vesicular stomatitis virus
• dicyclohexylcarbodiimide
• hydrophobicity

Introduction

Virus replication depends on the transfer of viral genome and accessory proteins to the cytosol or to the nucleus of a host cell. In the case of enveloped virus, this entry process involves the fusion of the virus envelope with the plasma or the endosomal membranes of the host cell [1]. The membrane fusion reaction is catalyzed by viral surface glycoproteins, which undergo conformational changes triggered by either their interaction with a cellular receptor or by the acidification of the endosomal pH. Viral fusion glycoproteins contain a short sequence directly involved in the interaction with the target membrane during the fusion reaction, known as the fusion peptide. Fusion peptides from several viruses have been identified by mutagenesis experiments, in which a single amino acid change abolished the fusion activity of the glycoprotein. The sequence of the fusion peptides is generally conserved within the viral family, but not among different families.

Based on structural differences, the viral fusion proteins were classified into two groups. Class I fusion proteins form trimeric spikes predominantly folded as α-helices with a hydrophobic fusion peptide located at the N-terminal end of the protein [2]. After binding to a cellular receptor or on exposure to low pH, the protein forms an extended conformation and the fusion peptide inserts into the target membrane. The post-fusion conformation is a hairpin-like structure in which the fusion peptide and the membrane anchor are at the same end [2]. In class II fusion proteins three domains folded largely on β-sheets are arranged in a continuous protein lattice formed by dimers [3]. The fusion peptide is an internal loop between two β-strands, buried in the dimer interface. The determination of the post-fusion structure of class II fusion proteins revealed a surprising convergence of the class I and class II fusion mechanisms [4]. The acidic pH of the endosome induces a disassembly of envelope proteins dimers, which rearrange in trimers with the fusion peptide loops clustered at one end of an elongated molecule.

The viruses that belong to the Rhabdoviridae family are widely distributed in nature and their hosts range from vertebrates and invertebrates animals to many species of plants. All the rhabdoviruses present a bullet-shaped structure that is formed by two major components: the nucleocapsid and the envelope. The envelope is a lipid bilayer derived from the host cell containing trimeric transmembrane spikes composed by the viral surface glycoprotein G. Vesicular stomatitis virus (VSV) is the prototype of the Rhabdoviridae family. VSV G protein is involved both in the cell recognition and in the membrane fusion reaction, which occurs in the acidic environment of the endosome after virus internalization. Mutagenesis experiments have shown that substitution of conserved Gly, Pro, or Asp located in the region between amino acids 117 and 137 either abolished the fusion ability of G protein or shifted the optimum pH of fusion [5–7]. This led the authors to propose that this segment would be the VSV G protein putative fusion peptide, although direct evidence that this particular region interacts with the target membrane is still lacking. Studying the requirement for PS in the target membrane [8] and the crucial role of G protein His residues for VSV fusion [9], we found another candidate to be the VSV fusion peptide, the PS binding site of the rhabdoviruses G protein [10]. This segment was firstly characterized for viral hemorrhagic septicemia virus (VHSV), a rhabdovirus of salmonids [11], [12], and then was found among all rhabdoviruses [10]. For VSV, it corresponds to the sequence between amino acid 145 and 164 (sequence VTPHHVLVDEYTGEWVDSQF). We have demonstrated that a synthetic peptide corresponding to this sequence was as efficient as the whole virus in catalyzing fusion, whereas the putative fusion peptide failed to induce fusion [9]. Moreover, as found for VSV-induced membrane fusion, the fusion induced by the peptide was dependent on pH and on the presence of PS in the target membrane.

An interesting feature of VSV peptide^145–164 is that it contains four acidic amino acid residues, two aspartic and two glutamic acids, which would be negatively charged at the fusion pH. Using the carboxyl-modifying agent dicyclohexylcarbodiimide (DCCD) and several mutant peptides, we showed in this work that the acidic residues, especially Asp¹⁵³ and Glu¹⁵⁸, participate in the formation of a hydrophobic domain at pH 6.0, which is necessary to the peptide-induced membrane fusion. In addition, the formation of this hydrophobic region as well as the membrane fusion itself, are highly dependent on peptide concentration, suggesting the involvement of peptide oligomerization.

Materials and methods

Chemicals

Phosphatidylserine (PS) and phosphotidylcholine (PC) from bovine brain, and dicyclohexylcarbodiimide (DCCD) were purchased from Sigma Chemical Co., St Louis, MO, USA. N-(lissamine Rhodamine B sulfonyl) phosphatidylethanolamine (Rh-PE), N-(7- nitro-2,1,3-benzoxadiazol-4-yl) phosphatidylethanolamine (NBD-PE) and 8-anilinonaphthalene-1-sulfonate (ANS) were purchased from Molecular Probes Inc., Eugene, OR, USA. All other reagents were of analytical grade.

Peptides synthesis

VSV peptide^145–164 (sequence VTPHHVLVDEYTGEWVDSQF) and histidine and acid residues mutants were synthesized by solid phase using the Fmoc methodology and all protected amino acids were purchased from Calbiochem-Novabiochem (San Diego, CA) or from Neosystem (Strasbourg, France). The syntheses were done in an automated bench-top simultaneous multiple solid-phase peptide synthesizer (PSSM 8 system from Shimadzu, Tokyo, Japan). The final deprotected peptides were purified by semipreparative HPLC using an Econosil C-18 column (10 µm, 22.5×250 mm) and a two-solvent system: (A) trifluoroacetic acid/H₂O (1:1000, v/v) and (B) trifluoroacetic acid/acetonitrile/H₂O (1:900:100, v/v/v). The column was eluted at a flow rate of 5 ml.min⁻¹ with a 10 or 30 to 50 or 60% gradient of solvent B over 30 or 45 min. Analytical HPLC was performed using a binary HPLC system from Shimadzu with a SPD-10AV Shimadzu UV/vis detector, coupled to an Ultrasphere C-18 column (5 µm, 4.6×150 mm), which was eluted with solvent systems A1 (H₃PO₄/H₂O, 1:1000, v/v) and B1 (acetonitrile/H₂O/H₃PO₄, 900:100:1, v/v/v) at a flow rate of 1.7 ml.min⁻¹ and a 10–80% gradient of B1 over 15 min. The HPLC column eluted materials were monitored by their absorbance at 220 nm. The molecular mass and purity of synthesized peptides were checked by MALDI-TOF mass spectrometry (TofSpec-E, Micromass) and/or peptide sequencing using a protein sequencer PPSQ-23 (Shimadzu Tokyo, Japan).

Peptide modification with DCCD

A solution of dicyclohexylcarbodiimide (DCCD) was freshly prepared by dilution of the reagent in ethanol. Peptide^145–164 was diluted in 20 mM MES, 30 mM Tris buffer, pH 6.0 and incubated for 1 h at room temperature with DCCD, in a molar ratio of DCCD/peptide of 40.

Preparation of liposomes

PC and PS at a molar ratio of 1:3 were dissolved in chloroform and evaporated under nitrogen. The lipid film formed was resuspended in 20 mM MES, 30 mM Tris buffer (pH indicated in the Figure legends) at a final concentration of 1 mM. The suspension was vortexed vigorously for 5 min. Small unilamellar vesicles (SUV) were obtained by sonicating the turbid suspension using a Branson Sonifier (Sonic Power Company, Danbury, CT) equipped with a titanium microtip probe. Sonication was performed in an ice bath, alternating cycles of 30 sec at 20% full power, with 60-sec resting intervals until a transparent solution was obtained (approx. 10 cycles). For fusion assays, 1 mol% of each Rh-PE and NBD-PE was incorporated in the lipid films.

Liposome fusion assay

Liposomes composed of PC-PS (1:3) containing equal amounts of unlabeled vesicles and vesicles labeled with Rh-PE and NBD-PE were prepared in 20 mM MES, 30 mM Tris buffer (pH indicated in the Figure legends), at a final phospholipid concentration of 0.1 mM. The fusion reaction was initiated by addition of the peptide. Fusion was followed by the resonance energy transfer assay as described in Struck et al. (1981) [13]. The samples were excited at 470 nm and the fluorescence intensity was collected at 530 nm, using a Hitachi F-4500 Fluorescence Spectrophotometer.

Mass spectrometry

A stock solution of VSV peptide^145–164 in ethanol was diluted at different concentrations (as indicated in the Figure legends) in 20 mM MES, 30 mM Tris buffer pH 6.0, and incubated with DCCD for 1 h at room temperature. Before analysis by mass spectrometry, the buffer was removed by loading the peptide on a ZipTip C18 (Millipore, Billerica, USA). The peptide was washed with TFA 0.1% (v/v) and eluted in 5 µl acetonitrile 50%/formic acid 1% (v/v). The samples were loaded into a nanoflow capillary (Proxeon, Odense, Denmark). ESI mass spectra were acquired on a quadrupole time-of-flight instrument (Q-Tof Ultima – Micromass/Waters, Manchester, UK) operating in the positive ion mode, equipped with a Z-spray nanoelectrospray source. Capillary voltages of 1.1–1.5 kV and cone voltage of 50 V typically were used. The source temperature was held at 80°C. The spectra represent the average of 1 sec scans. Data acquisition was performed with a MassLynx 4.0 system. The exact mass of the peptide was determined after processing of the spectra by the software Transform (Micromass/Waters, Manchester, UK). For MS/MS studies, the quadrupole was used to select the charged parent ion, which was subsequently fragmented in a hexapole collision cell using argon as collision gas and an appropriate collision energy. MS/MS data were processed by a maximum entropy data enhancement program, MaxEnt 3 (Micromass/Waters, Manchester, UK). Amino acid sequence was semi-automatically deduced with the peptide sequencing program, PepSeq.

Infrared spectroscopy

ATR-FTIR (Attenuated Total Reflection Fourier Transform infrared spectroscopy) spectra were recorded on a Bruker IFS-55 FTIR spectrophotometer (Bruker, Karlsruhe, Germany) equipped with a liquid nitrogen-cooled mercury-cadmium-telluride detector [14–16]. The spectrophotometer was continuously purged with dried air. The internal reflection element (ATR) was a germanium plate (50×20×2 mm) with an aperture angle of 45°, yielding 25 internal reflections. Samples were deposited on the germanium element. Films were formed by slowly evaporating the sample on one side of the ATR plate under a stream of nitrogen. Samples were rehydrated by flushing D₂O-saturated N₂ for 30 min at room temperature.

Results

Effects of Glu and/or Asp modification on peptide-induced membrane fusion

We have previously shown that fusion induced by peptide^145–164 depends on the protonation of its His¹⁴⁸ and His¹⁴⁹ residues, which confer positive charges to the peptide at pH below 6.0 [12]. However, the peptide also contains four negatively charged amino acid residues (Asp¹⁵³, Glu¹⁵⁴, Glu¹⁵⁸ and Asp¹⁶¹) and no other positively charged residue. To investigate the role of these negative charges in peptide-induced fusion, we used DCCD, a compound that labels Asp and Glu residues located in hydrophobic environments [17]. Peptide-induced membrane fusion was quantified by NBD-PE/Rh-PE energy transfer assay (Figure 1a). DCCD labeling completely abolished the ability of the peptide to mediate membrane fusion. To confirm the importance of the negative charges of these residues, fusion was assayed at pH 4.0, a pH close to the pKa of the carboxyl groups of lateral chain of the acidic residues (Figure 1b). Fusion did not occur at pH 4.0, even when the peptide was not incubated with DCCD, confirming the importance of the negative charges. No lipid mixing was observed at pH 7.5, confirming the role of the positively charged histidines.

Figure 1.  Membrane fusion induced by peptide^145–164 depends on the negative charges of its acidic residues. (A) Effect of the modification of the acidic residues with DCCD. Equal amounts of unlabeled vesicles and vesicles labeled with Rh-PE and NBD-PE were incubated with the peptide^145–164 or the peptide pre-incubated with DCCD. The vesicles were composed of PC:PS (1:3) and were prepared in 20 mM MES, 30 mM Tris buffer, pH 6.0, in a final phospholipid concentration of 0.1 mM. Peptide-induced membrane fusion activity was measured by the increase in the NBD-PE fluorescence. NBD-PE was excited at 470 nm, and the intensity was collected at 530 nm, during 10 min. The final peptide concentration was 20 µg/ml. (B) Effect of pH. Fusion reaction was measured as described above except that the pH was adjusted to 7.5, 6.0 or 4.0, as indicated in the Figure.

Mass spectrometry analysis revealed the presence of different populations containing up to three modified residues (Figure 2). The computed monoisotopic mass of each population agrees with the addition of one, two or three DCCD (monoisotopic mass = 206.17 Da) molecules: 2356.13 (+0), 2562.26 (+1), 2768.44 (+2) and 2974.62 (+3), respectively.

Figure 2.  Labeling of the peptide^145–164 with DCCD. ESI-MS spectra were recorded to monitor the addition of DCCD groups to the Asp and Glu residues of the peptide^145–164. The peptide (20 µg/ml) was incubated for 1 h with DCCD in 20 mM MES, 30 mM Tris buffer, pH 6.0. Before analysis, the buffer was removed on ZipTip C18 and the peptide was solubilized in 50% acetonitrile/ 1% formic acid (v/v). The different populations observed (arrows) correspond to the addition of 1, 2 and 3 DCCD molecules, respectively. The accuracy of mass measurement was in the order of 8 ppm.

Acidic residues are involved in the formation of a hydrophobic region necessary for peptide-induced fusion

It is well established that negatively charged phospholipids, especially PS, are required for VSV binding to membranes and fusion [8], [18]. This suggests that the acidic amino acids of the peptide^145–164 might not be involved directly in the interaction with the membrane. Since the carboxyl modification by DCCD is highly favored in hydrophobic environment [17], one possibility is that the negative charges could participate in the formation of a hydrophobic fusion-active structure maintained either by intra-molecule or by inter-molecule interactions. To investigate the formation of hydrophobic regions, we used the fluorescent probe ANS, whose binding to non-polar segments in proteins is accompanied by a large increase in its fluorescence quantum yield. Figure 3a shows that the peptide binds ANS in a pH-dependent manner. ANS binding was maximal at pH 6.0, showing that the formation of the hydrophobic domain was maximal at the fusogenic pH. This result is quite similar to those previously obtained by us for the whole virus and purified G protein [19]. The hydrophobic domain is lost at lower pHs, suggesting that the protonation of negatively charged residues impairs the formation of this domain. Indeed, peptide modification with DCCD led to a great decrease in ANS binding (Figure 3b), confirming that the acidic amino acids are important for the formation of the hydrophobic domain.

Figure 3.  Acidic residues are involved in the formation of a hydrophobic region. (A) ANS binding as a function of pH. Peptide^145–164 was diluted to a final concentration of 20 µg/ml in 20 mM MES, 30 mM Tris buffer, pH as indicated in the Figure, and incubated with 1 µM ANS. ANS was excited at 360 nm and the emission was collected at 492 nm. (B) Peptide^145–164 without modification (—) or modified with DCCD (…) were incubated with 1 µM ANS in 20 mM MES, 30 mM Tris buffer, pH 6.0, and the ANS fluorescence spectra were collected after excitation at 360 nm.

To investigate which were the residues directly involved in peptide hydrophobicity and fusogenic activity, we synthesized a number of mutants (Figure 4a). Substitution of His¹⁴⁹, Glu¹⁵⁴ and Asp¹⁶¹ did not affect the exposure of the hydrophobic domain, while His¹⁴⁸, Asp¹⁵³ and Glu¹⁵⁸ were essential for the hydrophobicity (Figure 4b). Peptide hydrophobicity greatly correlates to peptide fusogenic activity, except for the His¹⁴⁹ mutant, which binds ANS but is not fusogenic (Figure 4c). At pH 6.0, in which the ANS binding is maximal, the His might be protonated, and positively-charged, whereas the Glu and Asp are probably non-protonated, and negatively-charged. One possibility is that electrostatic interactions between one of the acid residues from one peptide and one of the His residues of other peptide drive the formation of an oligomeric hydrophobic structure at pH 6.0.

Figure 4.  His¹⁴⁸, Asp¹⁵³ and Glu¹⁵⁸ are important for peptide hydrophobicity and fusion. (A) Amino acid sequences of the peptides used in this study. Peptide wt used in this study corresponds to VSV G protein residues between 145 and 164. Six different mutants were also synthesized: His¹⁴⁸, His¹⁴⁹, Glu¹⁵⁴ or Glu¹⁵⁸ were substituted by Gln residues, and Asp¹⁵³ or Asp¹⁶¹ were substituted by Asn residues. The dots represent wt residues. (B) Peptide^145–164 or the mutant peptides were incubated with 1 µM ANS in 20 mM MES, 30 mM Tris buffer, pH 6.0, and the ANS fluorescence spectra were collected after excitation at 360 nm. (C) Kinetics of membrane fusion by peptide^145–164 or mutant peptides. Equal amounts of unlabeled vesicles and vesicles labeled with Rh-PE and NBD-PE were incubated with each of the peptides and the membrane fusion was measured monitoring the increase in NBD-PE fluorescence. The vesicles were composed of PC:PS (1:3) and were prepared in 20 mM MES, 30 mM Tris buffer, pH 6.0, in a final phospholipid concentration of 0.1 mM. NBD-PE was excited at 470 nm, and fluorescence intensity was collected at 530 nm during 10 min.

Peptide oligomerization

To evaluate the requirement of peptide oligomerization for hydrophobicity and fusion, we checked the dependence on peptide concentration (Figure 5). The extent of lipid mixing varied with the peptide concentration in a higher than linear fashion (Figure 5a and 5b), supporting the hypothesis that peptide promotes membrane fusion as oligomers. A drastic increase in ANS binding occurred at the peptide concentration range required for fusion (Figure 5c), suggesting that peptide hydrophobicity is a consequence of its oligomerization. The threshold of aggregation was neither altered for the fusogenic mutants E154Q and D161N nor for the non-fusogenic H149Q mutant (Figure 5c).

Figure 5.  Peptide oligomerization is required for hydrophobicity and fusion. (A) Kinetics of membrane fusion at different peptide concentration. Equal amounts of unlabeled vesicles and vesicles labeled with Rh-PE and NBD-PE were incubated with the peptide^145–164 in a final concentration of 3, 6, 12, 18, 30 and 100 µg/ml, and the membrane fusion was measured monitoring the increase in NBD-PE fluorescence. The vesicles were composed of PC:PS (1:3) and were prepared in 20 mM MES, 30 mM Tris buffer, pH 6.0, in a final phospholipid concentration of 0.1 mM. NBD-PE was excited at 470 nm, and the fluorescence intensity was collected at 530 nm during 10 min. (B) Percentage of fusion after 10 min as a function of peptide concentration. (C) ANS binding as a function of peptide concentration. Peptide^145–164 (•) and the mutants H148Q (○), H149Q (▪), D153N (▾), E154Q (▵), E158Q (▾), and D161N (▿) were diluted in 20 mM MES, 30 mM Tris buffer, pH 6.0, to a final concentration as indicated in the Figure, and incubated with 1 µM ANS. ANS was excited at 360 nm and the emission was collected at 492 nm.

Analyses by mass spectrometry indicated that DCCD labels the peptide only at concentrations that favor oligomerization (Figure 6). Since DCCD labeling occurs in hydrophobic environments, this is another evidence that oligomerization contributes to the formation of a hydrophobic region that is probably necessary for the interaction with the membrane and consequently, for fusion.

Figure 6.  DCCD labeling is dependent on peptide concentration. Peptide^145–164 diluted to a final concentration of 3 µg/ml (A) or 100 µg/ml (B) was modified for 1 h with DCCD at pH 6.0 and ESI-MS spectra were recorded. No labeling was observed with 3 µg/ml and a labeling comparable to the one described in Figure 3 was observed for 100 µg/ml.

We analysed peptide secondary structure at pH 7.5 and 6.0 by infrared spectroscopy (Figure 7). The 1600–1700 cm⁻¹ region (amide I) corresponding to the C = O stretching vibration is the most sensitive to the secondary structure of the proteins and each secondary structure absorbs at different wavelengths. The peak at 1650 cm⁻¹ is characteristic of a peptide folded in an α-structure [14]. More importantly, the fold is not affected by pH (Figure 7a), suggesting that the hydrophobicity is a consequence of reorganization of the oligomeric state of the peptides rather than due to changes in their secondary structure. IR spectra of the non-fusogenic mutants did not reveal any structural changes as compared with wild type spectrum (Figure 7b), excluding the possibility that the mutations disrupt the structure and not only the generation of the hydrophobicity.

Figure 7.  Secondary structure measurements. (A) ATR-FTIR spectra were recorded from thin films obtained by slowly evaporating a sample containing 10 µg of peptide^145–164 (at pH 7.5 or 6.0, as indicated in the Figure) on an attenuated total reflection element. The samples were rehydrated by flushing D₂O-saturated N₂ for 30 min at room temperature. (B) The wild-type spectrum was compared with the spectra obtained, at pH 6.0, for the mutants H148Q, H149Q, D153N and E158Q, as indicated in the Figure.

Discussion

All enveloped viruses have transmembrane glycoproteins that mediate fusion between the virus membrane and host cell membrane, initiating the viral replication cycle. Viral fusion proteins vary in their mode of activation and their structural features, and based on these differences, they were grouped in two classes. It has been proposed that both class I and class II fusion proteins are synthesized in a metastable state and the native state is prevented from achieving the lower-energy fusogenic conformation by a kinetic barrier imposed during the folding and/or maturation [20]. In the case of Influenza virus, for example, the hemagglutinin (HA) folds within the cell as the fusion-incompetent precursor that undergoes proteolytic cleavage to generate the mature, two-chain native state [2]. This metastability allows the coupling of an energetically expensive membrane-fusion reaction to an energetically favorable conformational change, what could drive the reaction toward complete membrane fusion [20]. However, several results suggest that the glycoproteins of the rhabdoviruses catalyze fusion through a different mechanism.

A striking difference between VSV fusion and other viruses-induced membrane fusion is the reversibility of G protein conformational changes induced by low pH, even after the virus interaction with the membrane [21–23]. This suggests that the metastability is not absolutely required for viral membrane fusion. Another finding revealing a different mechanism of membrane recognition by VSV was the demonstration of an essential role of electrostatic interactions in VSV binding to membranes [8]. The electrostatic nature of VSV interaction with membranes is an interesting observation considering that most of the viral fusion proteins studied so far bind to membrane through hydrophobic interactions [24–26]. Indeed, it is believed that, in a certain stage of fusion process, the fusion peptide is exposed and inserted into the membrane of the target cell [27] and several studies using isolated fusion peptides have evaluated peptide insertion and orientation into the lipid bilayer [28–32]. However, in the case of the peptide^145–164, a linear peptide used in this study, the mode of interaction with the lipid bilayer might probably be different since this peptide is very hydrophilic. This is clearly shown in Figure 8, which compares the plots of average hydropathies of the amino acid residues between VSV peptide and HIV-1 fusion peptide using the hydropathy scale of Kyte and Doolittle [33]. While HIV-1 fusion peptide is very hydrophobic, presenting most of the hydropathies above zero, VSV peptide hydropathies are mostly below zero, what characterizes a hydrophilic peptide.

Figure 8.  Kyte-Doolittle plots of fusion peptide hydropathy. Hydrophobic profile of VSV peptide[145–164] and HIV-1 fusion peptide (sequence VGIGALFLGFLGAAGSTHGA). The hydropathy of these peptides was plotted from the amino terminus to the carboxy terminus by averaging hydropathy values over a window of 5 residues. More positive values are assigned to more hydrophobic residues.

An important question raised from these data is: how could this very hydrophilic peptide mediate membrane fusion? We believe that the answer to this question was found when we showed that the oligomerization of the peptide confers hydrophobicity to the oligomer. The increase in hydrophobicity correlates to the peptide fusogenic activity. Both are maximal at pH 6.0 and decrease as pH is increased to 7.5 or decreased to 4.0. Thus, we propose that the self-association of the peptides at pH 6.0 promotes the formation of a hydrophobic region important for the interaction with the target membrane.

It has already been shown that other fusion peptides associate at the membrane surface. For example, comparing the membrane interaction properties of a synthetic coiled-coil Influenza hemagglutinin fusion peptide with the monomeric peptide, Lau et al. [34] showed that the trimerization of the peptide increased lipid mixing, liposome leakage and membrane destabilization, suggesting an important role for the oligomerization of fusogenic peptides in the fusion mechanism. Studies using different peptide constructs suggested that the oligomers are formed as a result of membrane association and that small oligomers may be more fusogenic than monomers or large aggregates [35]. For HIV-1 fusion peptide, the ability of the peptide to form aggregates was also correlated to its ability to induce membrane fusion [28].

In most cases, the oligomerization of the fusion peptides is driven by changes in their secondary structure. Influenza hemagglutinin fusion peptide exists in at least two interconvertible forms at membrane surface: monomeric α-helical peptides that insert into the bilayer, and self-associated peptides adopting a β-sheet structure [36]. The equilibrium between these two forms is dependent on the pH and on the ionic strength. HIV-1 fusion peptides form oligomeric β-strand structures when associated to membranes [37]. GALA, a synthetic fusogenic peptide, undergoes a conformational change to an amphipathic helix when the pH is reduced [38]. IR spectroscopic data demonstrated that the peptide inserted deeply into the lipid bilayer, oriented parallel with respect to the lipid acyl chains [39]. In our case, however, we demonstrated by infra-red spectroscopy that there is no change in peptide secondary structure induced by lowering the pH from 7.5–6.0. Thus, rather than favored by conformational changes, the oligomerization of peptide^145–164 seems to be triggered by the protonation of His residues at the pH range of fusion, which creates positive charges in the peptides probably involved in electrostatic interactions with the negatively-charged residues. This explanation is in agreement with our previous results showing the importance of the electrostatic interactions for VSV fusion [8], and adds to our previous findings the idea that these electrostatic interactions are responsible for the formation of a hydrophobic region directed involved in VSV interaction with the membrane during fusion. This mechanism is different from that observed for WAE, an amphipathic negatively charged peptide that induces fusion of liposomal phosphatidylcholine membranes. Pécheur et al. (1999) provided evidence that it is the peptide penetration rather than peptide oligomerization that modulates peptide-induced fusion [40].

Using peptide mutants we demonstrated that His¹⁴⁸, Asp¹⁵³ and Glu¹⁵⁸ are essential for the hydrophobicity and fusion. Asp¹⁵³ is conserved among at least 13 different animal rhabdoviruses representing four recognized genera [41]. His¹⁴⁸ is also conserved except that it is substituted in the Novirhabdovirus genus for a lysine, which should also be protonated at the fusion pH (pK 10.53), or for a serine in the Lyssavirus genus. Glu¹⁵⁸ probably has a specific role in VSV fusion since it is not conserved among the rhabdovirus G proteins. We found that His¹⁴⁹ is important for fusion but not for hydrophobicity, suggesting that this residue is probably involved in the direct binding to PS in the target membrane.

Taken together, our data suggest that the negative charges of peptide acidic residues participate in intermolecular electrostatic interactions with positively-charged His residues, leading to the formation of a hydrophobic domain at pH 6.0, which is necessary to the peptide-induced membrane fusion. This suggests a mechanism of membrane interaction and destabilization resembling that one promoted by the antimicrobial peptides acting through a carpet model [42]. This model predicts an initial interaction between the hydrophilic portion of the peptide and negatively-charged membranes through electrostatic interactions. Then a reorientation of the hydrophobic residues of the peptide toward the hydrophobic core of the membrane occurs, causing membrane disintegration by disrupting the bilayer curvature.

Secondary structure prediction using the program PSIPRED Protein Structure Prediction Server (us.expasy.org) suggests that a loop is formed between the residues Glu¹⁵⁴ and Val¹⁶⁰. The formation of this loop is supported by the presence of residues Tyr¹⁵⁵ and Trp¹⁵⁹, which have large side chains, and by the Gly¹⁵⁷ in the middle, allowing the structure to bend at this point. A similar loop was found in the fusion peptide of dengue virus E glycoprotein between residues 99 and 105 [43], suggesting that some similarity might occur between VSV and class II fusion peptides. One important question to be posed is whether the data obtained using the peptide^145–164 could be correlated to the role of the peptide when taking part of the whole G glycoprotein. The peptide^145–164 is located within the domain shown to interact with the target membrane at the fusogenic pH [44]. In addition, its His residues (His¹⁴⁸ and His¹⁴⁹) were shown to be modified after treatment of the whole virus with DEPC, leading to the inhibition of virus-induced fusion [9]. These observations support the hypothesis that the peptide^145-164 is also active within the whole protein.

We would like to thank Adriana S. de Melo for technical assistance, Dr Erik Goormaghtigh for helpful assistance with ATR-FTIR experiments and Dr Fabio C. Almeida for helpful suggestions. This work was supported by grants from Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq), Centro Argentino-Brasileiro de Biotecnologia (CABBIO), Fundação Carlos Chagas Filho de Amparo à Pesquisa do Estado do Rio de Janeiro (FAPERJ) and Fundação de Amparo à Pesquisa do Estado de São Paulo (FAPESP). F. A. C. was recipient of PDEE fellowship from Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES). G.V. thanks Action de Recherches Concertées (ARC) for financial support.