Fusogenic domains in herpes simplex virus type 1 glycoprotein H.

Infection of eukaryotic cells by enveloped viruses requires fusion between the viral envelope and the cellular plasma or endosomal membrane. The actual merging of the two membranes is mediated by viral envelope glycoproteins, which generally contain a highly hydrophobic region termed the fusion peptide. The entry of herpesviruses is mediated by three conserved proteins: glycoproteins B, H (gH), and L. However, how fusion is executed remains unknown. Herpes simplex virus type 1 gH exhibits features typical of viral fusion glycoproteins, and its ectodomain seems to contain a putative internal fusion peptide. Here, we have identified additional internal segments able to interact with membranes and to induce membrane fusion of large unilamellar vesicles. We have applied the hydrophobicity-at-interface scale proposed by Wimley and White (Wimley, W. C., and White, S. H. (1996) Nat. Struct. Biol. 3, 842-848) to identify six hydrophobic stretches within gH with a tendency to partition into the membrane interface, and four of them were able to induce membrane fusion. Experiments in which equimolar mixtures of gH peptides were used indicated that different fusogenic regions may act in a synergistic way. The functional and structural characterization of these segments suggests that herpes simplex virus type 1 gH possesses several fusogenic internal peptides that could participate in the actual fusion event.

Membrane fusion occurs in a wide variety of biological processes, including the infection of cells by enveloped viruses; and as a result of fusion, the viral capsids are transferred to the cytoplasm of the newly infected cells (1,2). Energy barriers hinder spontaneous fusion between two apposing membranes; however, during viral infections, these barriers are overcome by specific viral "fusion proteins" present in the viral envelope. Despite the lack of sequence homology between fusion proteins from distantly related viruses, most of them share a common motif termed the "fusion peptide," which is composed mostly of hydrophobic residues and which is thought to interact directly with the target membrane and to trigger events that lead to membrane fusion (3,4). At least two classes of viral fusion proteins have been identified; in both classes, tightly regulated conformational changes are involved in membrane fusion (5). Class I fusion proteins are present in many unrelated virus families such as paramyxoviruses, orthomyxoviruses, retroviruses, and filoviruses (6 -11). Class II fusion proteins are found in members of the Flaviviridae family, which includes tickborne encephalitis virus (12,13), dengue virus (14,15), yellow fever virus, West Nile virus, and hepatitis C virus, and in members of the Togaviridae family, of which the best characterized are Semliki Forest virus and rubella virus (16 -18).
Class I fusion proteins (4,19) are composed of three identical subunits, whose functional form is generated from a precursor that is cleaved into two fragments; the membrane-anchored fragment contains the fusion peptide at or proximal to its N terminus. During fusion, the three fusion peptides become exposed and are inserted into the target cell membrane, generating an intermediate that is anchored both to the cellular and viral membranes; thus, the refolding of the protein into a trimeric coiled coil (N-and C-terminal heptad repeats) relocates the fusion peptides and the transmembrane anchor domains to the same end of the coiled coil, bringing viral and cellular membranes together.
Class II fusion proteins do not form spiky projections perpendicular to the viral membrane, but lie parallel to the membrane and possess an internal fusion peptide (13,15,20). In particular, they are predominantly non-helical (having a ␤-sheet structure); they are not cleaved; and the fusion peptide corresponds to an internal hydrophobic fusion loop. The proteins are composed of three domains, and the fusion peptide is contained in domain II; they are homodimers with the two subunits arranged in a headto-tail orientation. During fusion, they undergo dramatic irreversible structural changes that result in a homotrimeric form lifting up from the virus surface and projecting the internal fusion loop toward the target membrane.
The presence of a fusion peptide is thus a feature of both Class I and II viral fusion proteins (21)(22)(23); these hydrophobic sequences are involved in driving the initial partitioning of the fusion protein into the target membrane, making the viral envelope glycoprotein an integral component of both viral and cellular membranes. The real interplay between fusion peptides and the membrane is still unknown. Fusion peptides are believed to facilitate local dehydration; to help overcome the energetic barriers associated with the fusion process; and to serve as membrane anchors that facilitate partitioning of other regions of the viral envelope proteins to the membrane, which can subsequently participate in membrane merging. Fusion peptides are hydrophobic stretches of 10 -30 amino acid resi-dues, typically rich in alanine and glycine residues that may form sided helices, with bulkier and more hydrophobic residues on one side associating with the membrane and small residues on the other. Although there is conflicting evidence on the active secondary structures of fusion peptides, several studies have shown that fusion peptides can flip between different conformations depending upon their environment. Because membrane fusion is a dynamic process, structural plasticity appears to be crucial for the fusion process, and the orientation of the fusion peptide within lipids is also important (24,25).
Although it was initially thought that viral fusion glycoproteins contained a single fusogenic region responsible for the actual merging of the membranes, over the last few years, a more complex view has emerged (26). This is illustrated by recent studies on Sendai virus (27,28). In addition to the Nand C-terminal heptad repeats, similar to those found in Class I fusion proteins, an extra leucine zipper is located in the interior of the fusion protein; this leucine zipper, together with the N-terminal fusion peptide, is involved in membrane fusion. The importance of the second fusion peptide in Sendai virusinduced membrane fusion was highlighted by the finding that a synthetic peptide corresponding to the internal fusion peptide inhibits the fusion between the virus and red blood cells (27). Moreover, the two heptad repeats do not induce membrane fusion by themselves, but the N-terminal heptad repeat was shown to assist the N-terminal fusion peptide in the fusion process. Thus, the actual fusion step of Sendai virus is not mediated solely by the action of an N-terminal fusion peptide, but rather is the consequence of the concerted action of several regions of the fusion glycoprotein.
Studies using synthetic peptides that correspond to the sequences of viral fusion peptides have in many cases shed light on the molecular mechanisms involved in virus-mediated fusion. Moreover, synthetic peptides appear to be the most useful for studies on the minimum and precise molecular and structural requirements for membrane destabilization and represent an alternative method that is not hampered by the difficulties of analyzing mutations that could affect the overall folding of the protein. A possible way to elucidate the molecular mechanism of membrane fusion is to study the interaction of synthetic peptides corresponding to specific domains of viral fusion proteins with model membranes. In fact, synthetic peptides corresponding to the N-terminal fusion peptide of several viruses, including influenza virus (29), simian immunodeficiency virus (30), and human immunodeficiency virus type 1 (HIV-1) 1 (24,25,(31)(32)(33), have been shown to interact with the lipid bilayer and to promote fusion of lipid vesicles.
Herpesviruses are structurally complex enveloped viruses that have at least 12 glycoproteins on their surface. Unlike orthomyxoviruses, paramyxoviruses, filoviruses, and retroviruses, which use a single glycoprotein for membrane fusion, herpesviruses employ multicomponent membrane fusion machines that comprise at least three proteins, glycoprotein B (gB), glycoprotein H (gH), and glycoprotein L (gL), all of which are conserved in all members of the Herpesviridae family. In addition to this core requirement for fusion (as exemplified by a number of reports on different members of the herpesvirus family (34,35)), other glycoproteins may also be involved. For example, herpes simplex virus type 1 (HSV-1)-mediated fusion requires gB, gH, gL, and glycoprotein D (36 -38), whereas fusion of the ␥-herpesvirus (Epstein-Barr virus) is mediated by gB, gH, gL, and gp42 (39 -41).
Although much progress has been made in understanding how membrane fusion is promoted by single-component fusion proteins, little is known about how multiple components mediate fusion. Although it seems likely that multicomponent fusion machines require cooperation among the fusion proteins, it remains unclear if and how herpesvirus glycoproteins interact with one another, and the molecular details of the fusion process are not understood.
None of the required four HSV-1 glycoproteins has yet been found to possess all the characteristics of Class I or II fusion proteins. Nevertheless, a number of studies have implicated key roles for both gH and gB in the fusion process (42)(43)(44). Mutations in the extracellular, transmembrane, and cytoplasmic domains of HSV-1 gH have been shown to influence membrane fusion (45,46), as have mutations in the cytoplasmic domain of gB. Furthermore, a recent study (47) identified a potential ␣-helical region in the ectodomain of HSV-1 gH that showed some attributes of an internal fusion peptide. Because it is now becoming increasingly clear that the fusion mechanism employed by some viruses involves more than simply the interaction of an N-terminal fusion peptide with membranes, we attempted to identify regions of HSV-1 gH that may have this property. We therefore screened the amino acid sequence of gH for regions of highly interfacial hydrophobicity that show homology to other known viral fusion peptides, and synthetic peptides corresponding to those regions were tested for their ability to induce the fusion of large unilamellar vesicles.
Proteomics Computational Methods-Domains with significant propensity to form transmembrane helices were identified with TMpred (ExPASy proteomics server, Swiss Institute of Bioinformatics) and MPEx (Membrane Protein Explorer, available at blanco.biomol.uci.edu/mpex). TMpred is based on a statistical analysis of TMbase, a data base of naturally occurring transmembrane glycoproteins (48), whereas MPEx detection of membrane-spanning sequences is based on experimentally determined hydrophobicity scales (49,50). Sequences with a propensity to partition into the lipid bilayer were also identified with MPEx using interfacial settings, with mean values for a window of 11 amino acids. In particular, hydropathy plots corresponding to the pre-transmembrane domains were obtained using the hydropathy index of Kyte and Doolittle (51) and the interfacial hydrophobicity scales of Wimley and White (49) for individual residues. Secondary structure predictions were performed using ANTHEPROT software. Alignments were performed using BLAST (52) and ClustalW (53). The gH sequence used was taken from the Swiss Protein Database (accession number P08356).
Peptide Synthesis-Peptides were synthesized using the standard solid-phase Fmoc method on a Shimadzu PSSM8 multispecific peptide synthesizer. The NovaSyn®TGA resin (substitution of 0.25 mmol/g) was used as the solid-phase support, and syntheses were performed on a scale of 100 mol. The sequences are shown in Table I.
Lipid Mixing Assay-Membrane lipid mixing was monitored using the resonance energy transfer assay reported by Struck et al. (56). The assay is based on dilution of the NBD-PE (donor) and Rho-PE (acceptor) groups. Dilution due to membrane mixing results in an increase in NBD-PE fluorescence. Thus, we monitored the change in donor emission as aliquots of peptides were added to vesicles. Vesicles containing 0.6 mol % of each probe were mixed with unlabeled vesicles at a 1:4 ratio (final lipid concentration of 0.1 mM). Small volumes of peptides in Me 2 SO were added; the final concentration of Me 2 SO was no higher than 2%. The NBD emission at 530 nm was followed with the excitation wavelength set at 465 nm. A cutoff filter at 515 nm was used between the sample and the emission monochromator to avoid scattering interference. The fluorescence scale was calibrated such that the zero level corresponded to the initial residual fluorescence of the labeled vesicles, and the 100% value corresponding to complete mixing of all lipids in the system was set by the fluorescence intensity of vesicles upon the addition of 0.05% (v/v) Triton X-100 at the same total lipid concentrations of the fusion assay. All fluorescence measurements were conducted in five different sets of LUVs: PC, PC/Chol (1:1), PC/PE (1:1), PC/PS/Chol (1:1:1), and PC/PE/SM/Chol (1:1:1:3). Lipid mixing experiments were repeated at least three times, and the results were averaged. Control experiments were performed using scrambled peptides and Me 2 SO.
Tryptophan Fluorescence Measurements-Emission spectra of the peptides containing at least one tryptophan residue (gH-(220 -226), gH-(579 -597), and gH-(626 -644)) in the absence or presence of target vesicles (1:1 PC/Chol) were recorded between 300 and 400 nm with an excitation wavelength of 295 nm. Trp fluorescence measurements were done in the absence and presence of iodide, which acts as an aqueous collisional quencher.
Circular Dichroism Measurements-CD spectra were recorded using a Jasco J-715 spectropolarimeter in a 1.0-cm quartz cell at room temperature. The spectra are an average of three consecutive scans from 260 to 195 nm, recorded with a bandwidth of 3 nm, a time constant of 16 s, and a scan rate of 10 nm/min. Spectra were recorded and corrected for the blank. Mean residues ellipticities were calculated using the equation obsd/lcn, where obsd is the ellipticities measured in millidegrees, l is the length of the cell in centimeters, c is the peptide concentration in moles/liter, and n is the number of amino acid residues in the peptide. The percentage of helical content was calculated from measurements of helical mean residue ellipticity at 222 nm (57). We used [] 222 values of 0 and Ϫ40,000 (1 Ϫ 2.5/n) degrees cm 2 dmol Ϫ1 per amino acid residue for 0% and 100% helicity; n is the number of amino acid residues. Solutions of peptides (0.2 M) were prepared in water and at various percentages of trifluoroethanol (TFE). Peptide samples in SDS were prepared using the following protocol (58). All peptides were first dissolved in TFE. Immediately after preparation, the peptide solution was added to an equal volume of an aqueous solution containing the appropriate SDS concentration, and water was added to yield a 16:1 ratio of water to TFE by volume. The samples were vortexed and lyophilized overnight. The dry samples were rehydrated with deionized water to yield final SDS concentrations of 3, 6 mM, and 10 mM. Peptide samples in lipids were prepared using the following protocol (59). All peptides were first dissolved in TFE. Immediately after preparation, the peptide solution was added to an equal volume of a chloroform solution containing the appropriate lipid concentration. Solutions were dried with a nitrogen gas stream and lyophilized overnight. The dry samples were rehydrated with deionized water to yield final lipid concentrations of 0.05 and 0.9 mM. Small unilamellar vesicles of lyso-PC were prepared from LUVs by sonication.

Interfacial Hydrophobicity Analysis, Helical Propensity, and
Identification of Peptide Sequences-To identify hydrophobic stretches in the sequence of gH with the potential to interact with target membranes, we applied the hydrophobicity-at-interface scale proposed by Wimley and White (49). The hydrophobicity-at-interface scale together with Kyte-Doolittle analysis (an estimate of hydrophobicity based on bulk phase partitioning of side chain hydrophobicity alone) (51) have been successfully used to detect putative regions involved in partitioning within sequences of several fusogenic viral proteins (11). As shown in Fig. 1A, the combined use of both plots revealed several interesting features in the sequence of gH. The first peak at the N terminus corresponds to the signal peptide sequence. Then, we identified two hydrophobic regions corresponding to gH-(220 -262) and gH-(381-420) that seem to combine an overall hydrophobic character arising from side chains (Kyte-Doolittle scale) with a high ability to partition at the interface. Toward the C terminus, four additional significant peaks corresponding to gH-(468 -486), gH-(493-537), gH-(579 -597), and gH-(626 -644) were detected, even though they showed a lower tendency to partition at the interface compared with gH-(220 -262) and gH-(381-420). Finally, at the C terminus, we identified a large peak corresponding to the pre-transmembrane and transmembrane regions of the glycoprotein.
We also screened the extracellular domain of gH for regions with homology to known fusion peptides. We selected a set of

HSV-1 Fusion
well characterized fusion peptides from Class I fusion proteins of different viruses (orthomyxoviruses, paramyxoviruses, and retroviruses) and aligned them with the sequence of HSV-1 gH. As shown in Fig. 2A, we identified two sequences (gH-(402-420) and gH-(495-513) that showed some similarity to other fusion peptides. Both sequences were contained in peptides found from the analysis of the hydrophobicity-at-interface plots, viz. gH-(381-420) and gH-(493-537), indicating that both our strategies are useful to identify membrane-interacting regions. Moreover, we identified gH-(626 -644), which aligns with the N-terminal coil of HIV-1 gp41. When gH-(579 -597) and gH-(626 -644) are in the helical conformation (Fig. 2B), the polar residues are distributed on both sides of the helices, but they concentrate in the portion closer to the C-terminal end, resulting in an extremely hydrophobic N terminus and a polar C terminus, giving the peptides an amphiphilic character. The dependence of both the extent and kinetics of lipid mixing on the peptide/lipid molar ratio was analyzed. In separate experiments, increasing amounts of gH-(220 -262), gH-(381-420), gH-(468 -486), gH-(493-537), gH-(579 -597), and gH-(626 -644) were added to fixed amounts of vesicles, and the percentage of lipid mixing as a function of the peptide/lipid molar ratio was calculated. Fig. 3A shows the results of lipid mixing assays using PC/Chol-containing vesicles. gH-(468 -486) was unable to induce lipid mixing under these conditions. However, we observed significant vesicle fusion in the presence of gH-(220 -262), gH-(381-420), gH-(579 -597), and gH-(626 -644). This latter peptide was the most effective in inducing lipid mixing. Scrambled peptides were also tested in this assay, and no fusion was observed. gH-(626 -644) was able to induce lipid mixing at lower peptide/ lipid ratios compared with the other peptides we tested, and we detected fusion with this peptide (at a peptide/lipid ratio of 0.4) of ϳ50%. At the same peptide/lipid ratio, the other peptides reached ϳ20% lipid mixing.
Some viruses show a specific requirement for particular lipids in the target membrane. To determine whether this is also the case for HSV-1, we generated liposomes of different lipid compositions and measured the ability of the gH peptides to induce lipid mixing. The results are shown in Fig. 4. We found a common trend for the six peptides we tested and noted a significant increase in fusion of liposomes containing Chol. In particular, fusion was highly efficient with liposomes containing PC and Chol at a 1:1 ratio. Although fusion still occurred in the absence of Chol, it was much less efficient, and all the peptides analyzed were less effective in inducing fusion. Inclusion of Chol in the target membrane, although not absolutely essential, appears to facilitate fusion of HSV-1.
All the peptides showed the highest percentage of fusion in PC/Chol except for gH-(493-537), which seemed to fuse better in PC/PE and PC/PE/SM/Chol. The outer leaflet of human cellular membranes is composed mainly of PC, PE, SM, and Chol; therefore, the testing of peptide-induced fusion of PC/PE/ SM/Chol liposomes is more relevant to the physiology of viral infection of human cells. The percentages of lipid mixing obtained for all the peptides were comparable with those obtained with the other liposome preparations, except for PC, which always supported the lowest lipid mixing percentage. gH-(626 -644) showed the highest percentage of lipid mixing in PC/PE/ SM/Chol and PC/Chol.
Assuming that all the sequences would be present at an equimolar ratio within a potential fusogenic complex during gH-mediated fusion, we also analyzed the membrane fusion abilities of combinations of the most active peptides (Fig. 3B). gH-(220 -262), gH-(381-420), gH-(493-537), and gH-(626 -644) were combined together at equimolar amounts, and following incubation with liposomes, a clear synergistic effect was observed. In fact, low doses (peptide/lipid ratio of 0.2) that alone induced up to 5-15% fusion of liposomes caused ϳ75% fusion. We performed this experiment first with the three most active peptides (gH-(220 -262), gH-(381-420), and gH-(626 -644)), and we obtained 53% fusion at a peptide/lipid ratio of 0.2. The addition of gH-(579 -597) resulted in no significant increase in fusion, whereas the addition of gH-(493-537) induced an increase in fusion of up to 75%. Although the results of fusion experiments performed using single peptides indicated that gH-(579 -597) induced a higher degree of fusion compared with gH- (493-537), the experiments performed with the combinations showed that gH-(493-537) was more effective when these peptides were used together to induce fusion. This result

HSV-1 Fusion
may be correlated with the higher conformational flexibility of gH-(493-537), which, as shown by CD analysis, was able to assume different conformations in different environments.
Tryptophan Fluorescence Emission-Because the results of the lipid mixing experiments strongly suggested that the gH peptides we have identified can interact with membranes, we compared the fluorescence emission spectra of gH-(220 -226), gH-(579 -597), and gH-(626 -644) in the presence of PC/Cholcontaining vesicles with those in phosphate-buffered saline. The fluorescence emission of tryptophan residues increases when the amino acid enters a more hydrophobic environment, and together with an increase in quantum yield, the maximum spectral position will be shifted toward shorter wavelengths (blue shift). Fig. 5 shows the fluorescence emission spectra of gH-(220 -262), gH-(579 -597), and gH-(626 -644) upon interaction with PC/Chol-containing vesicles. In all cases, changes in the spectral properties of the three peptides were observed, suggesting that the single tryptophan residue of gH-(579 -597) and gH-(626 -644) and the three tryptophan residues of gH-(220 -262) are located in a less polar environment upon interaction with lipids. Emission intensity was enhanced, and the maxima shifted to lower wavelength. These results suggest that all three peptides are capable of penetrating a lipid bilayer. We observed that the levels of Trp fluorescence of both peptides in solution decreased with time (data not shown), an effect that may be related to self-aggregation due to their hydrophobic character. This effect may compete with membrane association and therefore preclude correct determination of real partition coefficients.
Secondary Structures of Synthetic Peptides-Because structural conformations have been shown to be important for the fusogenic activity of fusion peptides, the secondary structures of peptides in aqueous solution and in membranes were analyzed by CD spectroscopy as measured in water, TFE, and SDS and in the presence of lyso-PC-and lyso-PE-containing small unilamellar vesicles. Under all conditions tested, the spectra were not reliable below 200 nm because of light scattering and therefore are not shown.
The CD spectrum in buffer solution indicated a random coil conformation for gH-(626 -644). A decrease in peptide environmental polarity occurs when the peptide is transferred from water to membrane interfaces; the effect of polarity on peptide conformation can be studied using aqueous mixtures of TFE ( Fig. 6). In the presence of 20% TFE, gH-(626 -644) already showed two negative bands at ϳ208 and 222 nm, suggesting the adoption of an ␣-helical conformation. Calculations of helical content as described previously (57) corresponded to 27% helix; increasing concentrations of TFE induced additional helical stabilization (ϳ35%). This result suggests that the peptide contains a particular sequence of amino acids that, in a nonpolar environment, can adopt an ␣-helical conformation. gH-(220 -262), gH-(381-420), and gH-(493-537) showed spectra in buffer that indicated the presence of an extended structure with a minimum at ϳ218 nm. For these three peptides, the same ␤-type conformation was found in 20% TFE; however, the spectra were consistent with the peptides becoming more structured in low polarity solvent. In fact, increasing amounts of TFE induced stabilization of ␣-helical structures, with 27% helical content for gH-(220 -262), 57% helical content for gH-(381-420), and 31% helical content for gH-(493-537) at 80% TFE. Thus, CD analysis for gH-(220 -262), gH-(381-420), and gH-(493-537) would be consistent with the presence of ␤-sheet structures in buffer that are preserved in low polarity environments. gH-(468 -486) and gH-(579 -597) showed spectra typical of a random coil conformation, with no considerable increase in ␣-helical content upon the addition of TFE.
At low percentages of TFE, a ␤-form can exist if a segment has ␤-forming potential, but excess TFE usually disrupts the ␤-form and may convert it into a helix if the segment also has helix-forming potential. As shown in Fig. 6, gH-(220 -262), gH-(381-420), and gH-(493-537) showed a strong potential to give ␤-structures at low percentages of TFE, whereas at higher percentages, they converted to ␣-helices. Only gH-(626 -644) showed a strong potential to give ␣-helices also at low percentages of TFE.
Therefore, it is possible that the secondary structures of gH-(220 -262), gH-(381-420), gH-(493-537), and gH-(626 -644) change from a random coil to an ␣-helix upon membrane binding. To determine whether this is the case, CD spectra of the four peptides in the presence and absence of liposomes and in the presence or absence of SDS were determined. As shown in Fig. 7 gH-(220 -262) was a random coil in buffer solution as well as in SDS (3 and 10 mM); however, when the peptide was incubated in the presence of lyso-PC-containing liposomes, at high lipid/peptide ratios, the ␣-helical content slightly increased. gH-(381-420) showed a high helical content (ϳ32%) in SDS (3 and 10 mM) as well as in lyso-PC-containing liposomes at the highest lipid/peptide ratio tested in our experiments. gH-(493-537) showed a high helical content (ϳ28%), but lower percentages were observed at lower lipid/peptide ratios. gH-(626 -644) showed the highest helical content in lyso-PC (ϳ38%). DISCUSSION Viral fusion proteins interact simultaneously with two membranes during the fusion process; the fusion peptide, a conserved stretch of hydrophobic amino acids, inserts into the target membrane and triggers fusion (2). Hydropathy analysis based on the hydrophobicity-at-interface scale proposed by Wimley and White (49) enabled us to detect six regions of HSV gH that may be involved in the interaction of the viral envelope and host cell membranes. The Wimley-White hydropathy analysis has been reported to be superior to other methodologies for detecting potential membrane-interacting sequences within viral fusion protein ectodomains (11); the main advantage is that it takes into account the effect of membrane interface on partitioning. The interface is composed of a complex mixture of water and chemically heterogeneous phospholipid groups in which significant changes in polarity occur at short range; thus, aromatic residues appear to be the most hydrophobic ones when located at interfaces. According to our results, gH-(220 -262) and gH-(381-420) showed the highest probability to localize at the bilayer interface, but the other four gH peptides identified also showed a certain hydrophobicity that is compatible with membrane-interacting sequences.
Here, we have demonstrated that HSV-1 gH-(220 -262), gH-(381-420), gH-(579 -597), and gH-(626 -644) are able to induce rapid membrane fusion and to act in a synergistic way. Furthermore, these peptides showed characteristics of membraneinteracting regions as measured by tryptophan fluorescence emission analysis and by their tendency to assume particular secondary structures by CD analysis.
Liposomes have been widely used as model systems to understand the molecular mechanism of viral membrane fusion (60,61). In particular, membrane fusion can be divided into three steps. 1) The two membranes that are supposed to fuse must approach each other closely; 2) there is a transient disruption of stable bilayer structures; and 3) the two membranes must mix their components and fuse into one membrane. One widely accepted view is that the fusion peptide is responsible for the fusion of the two membranes inserting into the target membrane and somehow destabilizing the lipid bilayer. It has been reported recently that other domains of the fusion glycoproteins are involved in the process and that these domains do not insert into the bilayer, but may interact closely with the target membrane, thereby promoting the formation of fusion intermediates (26 -28).
Although much information on the correlation of the fusogenic activity of viral synthetic fusion peptides with peptide secondary structure is available, detailed studies on the nature of specific interactions of fusion peptides with different phospholipid vesicles are still lacking. An interesting feature of membrane fusion is its requirement for specific lipids in the target membrane (62). The zwitterionic phospholipids (mainly PC and PE) together compose the majority of the membrane phospholipids of eukaryotic cells, Gram-negative bacteria, and many Gram-positive bacteria (62). PC and PE have been treated as interchangeable in many experimental designs. Because of the more desirable physical properties of PC in forming vesicles and defined structures in solution, PC has been preferentially used in many in vitro studies. However, there are significant and important differences in the chemistry and properties of these two lipids, and it is now clear that they are not functionally exchangeable in supporting biological processes. Therefore, experimental design involving lipids must be governed by the lipid specificity of the system.
Chol is almost entirely non-polar, with a single OH group attached to the ring system. Thus, when incorporated into the bilayer, this very small head group is not enough to shield the non-polar part of Chol from interfacial water, so Chol must use the head groups of the adjacent lipids as umbrellas to shield itself from interfacial water. The main consequence of this is that it enhances Chol-lipid lateral interactions and precludes Chol-Chol interactions within the bilayer, which would increase Chol exposure to water. PE has a smaller head group cross-sectional area than PC and is able to accommodate only one molecule of Chol/molecule of PE, whereas PC can accommodate two molecules of Chol/molecule of PC. It is interesting that the outer leaflet of the plasma membranes of most mammalian cells contains a significant fraction of glycosphingolipids in which the head group cross-sectional area could be larger than that in phospholipids, thus potentially providing an umbrella not only over a single Chol molecule, but also over Chol multimers (63). One feature of the viral fusion mechanism is the varying requirement for specific lipids in the target cell membrane. For example, for simian immunodeficiency virus and HIV fusion peptides, lipid mixing occurs provided there is PE in the lipid bilayer (32, 64); vesicular stomatitis virus prefers PS (65), and influenza virus and Sendai virus peptides interact preferen-tially with PC membranes (27,29,66). Chol is a major component of mammalian membranes in vivo, and several studies have focused on the effect of incorporating Chol into liposomes on fusion. Some studies (67) report that inclusion of Chol reduces fusion, whereas others report increased activity. The increased activity is often coupled with the inclusion of other Studies with several viruses suggest that plasma membrane microdomains (or lipid rafts) that are highly enriched in Chol and sphingolipids play a crucial role in virus entry. Bender et al. (68) reported that lipid rafts serve as a platform for HSV entry and cell signaling. In fact, treatment of virions with a Chol-sequestering drug inhibited entry; therefore, the presence of Chol could be a necessary aspect of the membrane to fuse (68). Our results further support this model; in fact, the highest levels of fusion for all the peptides were obtained using vesicles containing Chol.

HSV-1 Fusion 28641
(493-537), gH-(579 -597), and gH-(626 -644) promoted fusion of liposomes of different composition, the varied levels of fusion observed for the various liposomes suggest a different mode of interaction. This possibility may result in a different structure for the peptide-vesicle complex. The best results were obtained in PC/Chol for all the peptides except gH-(493-537), and our experiments showed that omission of Chol from liposome preparations strongly reduced the extent of lipid mixing. Liposome fusion experiments showed that gH-(468 -486) was unable to induce lipid mixing and that gH-(493-537) was able to induce only low percentages of lipid mixing, whereas the highest levels of fusion were obtained using gH-(220 -262), gH-(381-420), gH-(579 -597), and gH-(626 -644).
We investigated the binding of gH-(220 -262), gH-(579 -597), and gH-(626 -644) to phospholipid vesicles containing PC/Chol using fluorescence techniques. The Trp residue of these three peptides provides a suitable chromophore that can be used to monitor the peptide-lipid interactions. The fluorescence spectra show that the tryptophan residues would be located in a less polar environment upon interaction with the lipids on the basis of the blue shift observed in the tryptophan fluorescence emission maximum and the quantum yields.
A decrease in peptide environmental polarity occurs when the peptide is transferred from water to membrane interfaces, resulting in increased ␣-helical content. Thus, the effect of polarity on peptide conformation was studied using TFE, SDS, and vesicles containing lyso-PC. CD studies showed that gH-(220 -262), gH-(381-420), gH-(493-537), and gH-(626 -644) were able to adopt different conformations when challenged in different environments, i.e. TFE, aqueous solution, SDS, and phospholipid vesicles.
Beyond the ability to interact with lipid membranes, one of the most important properties of fusion peptides is their flexibility as well as their ability to adopt different secondary structures (64, 69 -72). There is good evidence that influenza hemagglutinin adopts random coil, ␣-helical, and ␤-sheet conformations in different environments (71). It may be unstructured in solution on its way to the target membrane; the helical form may prevail at low concentrations in membranes, but self-associated ␤-sheets may be induced at higher concentrations in the membrane interfaces. Analogous conformational transitions have been observed with the fusion peptide of HIV gp41 (64). Because fusion is an extremely rapid and multistep process, structural flexibility (rather than the rigid adoption of a particular secondary structure) may be a key property of fusion peptides. It is possible that a critical feature of fusion peptides is their ability to flip between different secondary structures extremely rapidly, rather than the adoption of any single, well defined secondary structure. In most fusion peptide studies to date, the fusion peptides appear to have considerable structural plasticity and can exist as either ␣-helices or ␤-strands (64, 69 -72).
Thus, as reported in the literature for other fusion peptides (24,25), the gH peptides we have examined seemed to change from an ␣-helical structure at high lipid/peptide ratios toward a ␤-strand conformation at low lipid/peptide ratios, and the same behavior was observed when moving from high to low percentages of TFE. In particular, at low percentages of TFE, we found ␤-structures for gH-(220 -262), gH-(381-420), and gH-(493-537), whereas in 20% TFE, they converted to ␣-helices.
As is true for influenza virus (29), the fusion activity can be modulated by the action of distinct target membrane lipids, suggesting that distinct lipids in target membranes can act as a reversible secondary switch for fusion peptides, providing an elegant means for regulating membrane fusion activity. gH-(381-420) partially corresponds to the peptide that has been shown recently to possess some features of an internal fusion peptide in HSV gH (47); it is indeed able to induce a considerable level of fusion in model membranes and exhibits a highly conformational flexibility, forming both ␣and ␤-structures according to different environmental conditions. gH-(381-420) may correspond to the canonical fusion peptide as also shown from alignments with the fusion peptides of several other viruses (Fig. 1B). gH-(626 -644) is characterized by a strong propensity to form helices despite the conditions used, but this peptide induced the highest level of liposome fusion. It is possible that the two regions have to cooperate to induce complete fusion. Unable to give any defined structure, gH-(468 -486) is not involved in forming a fusion pore, whereas the remaining hydrophobic peptides may be involved to a lesser extent, mainly by interfering either in the early stages of fusion pore formation or at a later stage in stabilizing the fusion pore. It is of interest that gH-(468 -486) overlaps a domain of gH that has been shown recently to possess the characteristics of a heptad repeat sequence (heptad repeat-1) and that is effective in inhibiting viral fusion at a concentration of 500 M (73). gH-(468 -486) is also able to reduce viral infectivity, 2 supporting the view that this hydrophobic sequence of gH may be directly involved in membrane fusion.
The biophysical observation that peptides derived from gH are capable of interacting with membranes is also consistent with mutagenesis studies because partial deletions of gH-(381-420) or substitutions (L382P and GLL384-386WPP) result in decreased fusion activity (47). A previous report characterized a large set of linker insertion mutants of HSV gH and noted that insertions at some sites (i.e. amino acids 226, 381, and 591) result in gross misfolding of the protein and are defective in cell fusion and infectivity complementation activity (45). Nevertheless, additional mutagenesis studies will be required to better define key residues of gH that are involved in membrane destabilization.
In this study, we have also demonstrated that the combined use of the most effective peptides has a cooperative effect, achieving maximum fusion activity when they are used together, even at low concentration. We have demonstrated that regions of HSV-1 gH can induce membrane fusion and that gH appears to contain several membrane-interacting segments that may play a role in the fusion process. In particular, we propose that gH-(381-420) may be the canonical fusion peptide. gH-(626 -644) also plays a fundamental role in the fusion process, whereas gH-(220 -262), gH-(493-537), and gH-(579 -597) may interact with the membrane in later stages of the fusion process. Our results lend some support to a model of fusion in which several domains of the fusion glycoprotein have a high tendency to locate at the interface between the viral and target membranes. As recently reported by Gianni et al. (47), HSV gH presents a fusion peptide exposed on a fusion loop, made by cysteines 2 and 4 (74), rather than at the N terminus, but it seems that several different fusogenic domains are needed to execute complete fusion.
These results may shed some light on the complex mechanism of HSV entry into host cells. This mechanism involves other viral and cellular components, and considering the virion structural complexity, it remains plausible that gH is not the only glycoprotein to contain a fusion peptide.
Our findings do not demonstrate that gH behaves as a Class I or II fusion protein. In fact, the real mechanism could be somewhere in between the two canonical classes, and it is likely also that gB is directly involved in the fusion mechanism by direct insertion into the membrane bilayer. The main conclusion emerging from our studies is that the interaction of glyco-protein D with one of its receptors may induce conformational modifications of gH, leading to the exposure of several regions of the glycoprotein that interact with target membranes and subsequent virus entry. Further analysis of these regions of gH, both in biochemical assays and in the context of viral infection, will be of value in determining in more detail the mechanism of HSV-1-mediated membrane fusion.