Structure of a Membrane-binding Domain from a Non-enveloped Animal Virus

The γ1-peptide is a 21-residue lipid-binding domain from the non-enveloped Flock House virus (FHV). Unlike enveloped viruses, the entry of non-enveloped viruses into cells is believed to occur without membrane fusion. In this study, we performed NMR experiments to establish the solution structure of a membrane-binding peptide from a small non-enveloped icosahedral virus. The three-dimensional structure of the FHV γ1-domain was determined at pH 6.5 and 4.0 in a hydrophobic environment. The secondary and tertiary structures were evaluated in the context of the capacity of the peptide for permeabilizing membrane vesicles of different lipid composition, as measured by fluorescence assays. At both pH values, the peptide has a kinked structure, similar to the fusion domain from the enveloped viruses. The secondary structure was similar in three different hydrophobic environments as follows: water/trifluoroethanol, SDS, and membrane vesicles of different compositions. The ability of the peptide to induce vesicle leakage was highly dependent on the membrane composition. Although the γ-peptide shares some structural properties to fusion domains of enveloped viruses, it did not induce membrane fusion. Our results suggest that small protein components such as the γ-peptide in nodaviruses (such as FHV) and VP4 in picornaviruses have a crucial role in conducting nucleic acids through cellular membranes and that their structures resemble the fusion domains of membrane proteins from enveloped viruses.

Animal viruses require a built in membrane penetration mechanism. The strategies of infection adopted by enveloped and non-enveloped viruses are different, depending on the way the virus enters and leaves a cell (1). The cellular entry by enveloped viruses, such as influenza virus and retroviruses, is the most thoroughly studied (2,3). In general, infection by an enveloped virus starts with the fusion of viral and cellular membranes, mediated by viral envelope glycoproteins that contain well defined fusion domains. Fusion appears to be triggered by an optimum pH (such as the low pH at lysosomes) or by binding to a specific receptor (2,3). Recent studies have suggested that fusion peptides tend to present a kinked region, regardless of the overall secondary structure proposed (2,4,5). The kink confers to the domain a boomerang-like structure, which can act as anchor of the virus particle to the membrane.
Non-enveloped viruses do not require membrane fusion for entry into cells, but a membrane-binding motif is present in some of them (6 -9). In Flock House virus (FHV), 2 a non-enveloped RNA insect nodavirus, ␥ 1 -peptide has a membrane binding activity. FHV has been used as a model for the investigation of animal virus assembly, maturation, structure, and evolution (7, 8, 10 -14). The ␥ 1 -peptide contains the 21 N-terminal residues of the 44-residue ␥-peptide, a cleavage product of the coat precursor protein ␣. The cleavage of ␣ occurs after assembly of FHV particles and is required for acquisition of virion infectivity, which also results in a significant increase in particle stability (15). It has been proposed that this lipophilic domain could be a membrane-permeabilizing agent in the viral RNA translocation process (8). The ␥ 1 -domain has an amphipathic character and an amino acid sequence containing several residues with short side chains, which are a common feature of fusion peptides (2,16,17). Although there are few structures of fusion peptides from enveloped viruses, no solution structure is available for membrane-binding domains from small icosahedral viruses. Previous studies by NMR and biophysical methods with Flock House virus-like particles indicated that the cleaved wild type virus-like particles has a greater mobility than the uncleaved mutant because of the sharp lines of the cleaved ␥-peptide (12).
NMR studies of fusion peptide sequences from enveloped * This work was supported by grants from Conselho Nacional de Desenvolvi- viruses have provided valuable information about the relationship between structure and membrane binding and fusion activity, such as herpesvirus (18), influenza hemagglutinin (4,19), and sea urchin fertilization protein (20). However, no data are available for icosahedral viruses. In this study, we perform CD and NMR experiments to establish for the first time the solution structure of a lipid-binding domain from a small nonenveloped virus. The three-dimensional structures of ␥ 1 -peptide at pH 6.5 and pH 4.0 in a hydrophobic environment are reported. We evaluate conformational variability and membrane-permeabilizing activity by fluorescence spectroscopy at different pH values and different membrane compositions. Our results shed light into the mechanisms of how small non-enveloped particles bind to membranes and insert their genomes into the host cells.
CD Spectroscopy-CD data were performed using a Jasco J-715 spectrophotometer with a 2-mm path length cuvette. The spectra were recorded from 190 to 250 nm at a scanning rate of 50 nm/min with a wavelength step of 1.0 nm. Spectra were acquired for 200 M solutions of ␥ 1 -peptide at pH 7.0, diluted with 10 mM sodium phosphate buffer/water, in the presence of trifluoroethanol (10, 20, 30, and 50%) or 60 mM SDS or for 702.8 M solutions of ␥ 1 -peptide in the presence of lipid vesicles.
NMR Spectroscopy-NMR measurements were carried out at 20°C on a Bruker Avance DRX 600 or DRX 400 spectrometer operating at 600.04 MHz. Lyophilized synthetic ␥ 1 -peptide was prepared in 5 mM sodium phosphate buffer at pH 7.0 and pH 4.0 containing d 3 -trifluoroethanol/water (1:1, v/v) and 10% D 2 O. Final samples contained 4.2 mM peptide at pH 6.5 and 3.6 mM peptide at pH 4.0. Resonances of the sample at pH 6.5 were assigned from TOCSY (spin lock time of 70 ms) using the MLEV-17 pulse sequence (21), COSY-GP, and NOESY (mixing time of 150 ms) spectra (supplemental Figs. S1-S3) (22,23). The NOESY spectra were collected with 200 data points in F1 and 4096 data points in F2. The TOCSY spectra were collected with 400 data points in F1 and 4096 data points in F2. Resonances of the sample at pH 4.0 were assigned from TOCSY (MLEV-17 pulse sequence, spin lock time of 70 ms) and NOESY (mixing time 160 ms). The NOESY spectra were collected with 512 data points in F1 and 2048 data points in F2. The TOCSY spectra were collected with 300 data points in F1 and 2048 data points in F2. NMR data sets were also collected for ␥ 1 -peptide incorporated into SDS and DPC micelles. Samples were prepared by dissolving 1 mM of ␥ 1 -peptide in buffered water (pH 7.0 and 4.0) containing 160 and 300 mM SDS micelles and 2.5 mM of ␥ 1 -peptide in buffered water (pH 7.0 and 5.5) containing 120, 200, and 300 mM DPC micelles, and to each sample was added 10% D 2 O. For the spectra in SDS and DPC, TOCSY experiments were performed using the MLEV-17 spin-locking pulse with a mixing time of 70 ms (21), and NOESY spectra were performed using a mixing time of 120 ms (22,23). The NOESY spectra were collected with 512 data points in F1 and 2048 data points in F2 (for SDS) and with 512 data points in F1 and 4096 data points in F2 (for DPC). Water suppression for the samples was achieved by a pre-saturation pulse at the water frequency (24). All spectra were recorded in time-proportioned phase increment mode (25). The NMR data were processed by NMRPIPE (26). All NMR spectra were analyzed using NMRVIEW software package version 5.0.3 (27).
Structure Calculation-Distance restraints were derived from the 150-ms NOESY spectrum of the peptide at pH 6.5 and the 160 ms NOESY spectrum at pH 4.0. NOE cross-peak intensities were measured and converted into distances. Structures were calculated with the program CNS-solve version 1.1. One hundred structures were calculated for each sample, using simulated annealing protocol, applying cartesian-cartesian angle molecular dynamics. The stereochemical quality of the lowest energy structures was analyzed by PROCHECK-NMR (28). The display, analysis, and manipulation of the three-dimensional structures were performed with the program MOLMOL (29).
ANTS/DPX Release Assay-Peptide-induced release of aqueous vesicle content was measured by using the ANTS/DPX assay (30). Large unilamellar vesicles (LUV) were prepared according to the extrusion method of Hope et al. (31) in 10 mM sodium phosphate, 100 mM NaCl (pH 7.0). LUVs containing 12.5 mM ANTS, 45 mM DPX, 100 mM NaCl, and 10 mM sodium phosphate were obtained by separating the unencapsulated material by gel filtration in a GPC300 column (SynChropak, Micra Scientific, Inc.) eluted with 10 mM sodium phosphate, in 100 mM NaCl (pH 7.0). Fluorescence measurements were performed by setting the ANTS emission at 523 nm and the excitation at 353 nm with a 500 nm cut-off filter in the emission beam. The absence of leakage (0%) corresponded to fluorescence of the vesicles at time 0; 100% leakage was taken as the fluorescence value obtained after addition of 1% (v/v) Triton X-100. The degree of permeabilization was then inferred from Equation 1, where F is the fluorescence intensity after the addition of protein; F 0 is the initial fluorescence of the intact LUV suspension; and F T is the fluorescence after the addition of Triton X-100.
Carboxyfluorescein Release Assay-For fluorescent probe release experiments, a self-quenching solution (100 mM) of carboxyfluorescein (2Ј,7Ј-bis-(2-carboxyethyl)-5-(and-6)-carboxyfluorescein; Molecular Probes), 10 mM sodium phosphate buffer (pH 7.0) was entrapped in L-␣-phosphatidylcholine LUVs. The vesicles obtained (0.2 M diameter) were separated from the non-incorporated probe by gel filtration using a G-75 Sephadex column (Amersham Biosciences). Measurements were performed in sodium phosphate buffer at pH 6.5 and 7.0. The fluorescence measurements were performed in 1 ml of buffer consisting of 10 mM sodium phosphate at pH 7.0, in a quartz cuvette with stirring. Fluorescence was recorded as a function of time using an excitation wavelength of 490 nm and an emission wavelength of 518 nm with 2.5 nm bandwidth slits. Release was initiated by the addition of peptide and monitored by the fluorescence intensity increase after the addition of the peptide as described for the ANTS/DPX assay.

RESULTS
␥ 1 -Peptide-binding Domain Structure-In water (10 mM sodium phosphate buffer at pH 7.0), the CD spectrum of ␥ 1 -peptide was that of a typical "random coil" with only one band at ϳ200 nm ( Fig. 1). A change from random coil to an ␣-helical conformation was observed when increasing amounts of TFE (up to 50%) were added to the peptide in phosphate buffer (pH 7.0) (Fig. 1). Double minimum bands at about 222 and 208 -210 nm and a maximum band at 191-193 nm, which are characteristic of an ␣-helix, were observed in concentrations of 20% TFE and above. An ␣-helix pattern was also observed with the sample in 60 mM of SDS ( Fig. 2A) or when inserted in membrane vesicles of different compositions (Fig. 2B). The high similarity of the spectra indicates that the peptide is assuming a very similar conformation in the three different media.
NMR studies were conducted in three different media as follows: TFE/water, SDS, and DPC. For the TFE/water condition, the NMR experiments were performed with ␥ 1 -peptide in 50% TFE/buffered water solution at pH 6.5 and pH 4.0. pH 4.0 is close to the values of the acidic lysosomal vesicular compartments. The sequence-specific assignments of ␥ 1 -peptide at pH 6.5 and pH 4.0 were carried out using NOESY, TOCSY, and COSY spectra ( Fig. 3; supplemental Figs. 1S, 2S, and 3S), as proposed by Wuthrich (32). Despite the complexity of the spectra obtained, all residues of each peptide could be sequentially assigned via d␣N(i ϩ 1), d␣N(i ϩ 1), and dNN (i ϩ 1) NOEs. The ␣-proton (␣H) resonances were unambiguously identified on the basis of NOESY spectra at pH 6.5 (Fig. 4A). To locate the elements of secondary structure, chemical shifts of the ␣H were compared with statistical chemical shift values typical of a random coil conformation to calculate the chemical shift index deviation. The differences between the measured chemical shifts of ␣H and the standard values for a random coil polypeptide reveal an upfield shift of the residues Met-3 to Ile-10 and Ile-11 to Leu-15 (Fig. 5), suggesting two helical regions connected by a flexible hinge. Secondary structure content was also evaluated by medium and long range backbone 1 H-1 H distances ( Table 1).
The NMR data set (TOCSY and NOESY) obtained from ␥ 1 -peptide incorporated into DPC (supplemental Figs. 4S-7S) and SDS (Figs. 8S and 9S) micelles displayed peaks with sharp lines but a weak dispersion of signals, not allowing complete sequence-specific assignment. In Fig. 6, the NOESY spectra of the ␥ 1 -peptide in the presence of DPC micelles at two different pH values are superimposed. Although we were able to identify some of the peaks, there was less dispersion than in TFE/water. Particularly at pH 7.0, the lines were less sharp, which could be explained by protein association (Fig. 6).
To try to elucidate whether the increased peptide line widths in micelles are because of oligomerization of the peptide rather than the result of intermediate exchange between different con-formations, we used different temperatures and two detergent concentrations (supplemental Figs. 6S and 7S). The data show no significant changes in the line widths at different temperatures, which would be expected in case the increased line widths were because of intermediate exchange between different conformations. The increase in detergent concentration did not result in any considerable change in line width as well.
One hundred structures were calculated for each pH utilizing 179 (pH 6.5) and 208 (pH 4.0) interproton distance restraints ( Table 1). The chemical shift index data and NOE connectivities showed a tendency toward helix formation in the N-terminal segment followed by a break and a second helix in the C-terminal half, at both pH values. The backbone structures of the 20 lowest energy conformers in 50% TFE at pH 6.5 and 4.0 are shown (Fig. 7). At both pH values, the peptides present a kinked structure, similar to the fusion domain from influenza hemagglutinin (4). A helical content in the N-terminal portion and a sharp bend in the middle of the domain occur at both pH values. However, the C-terminal half has fewer observable d(ii ϩ 3) NOE interactions, suggesting a less regular structure than the N-terminal half. There were sig- nificant differences between the structure observed at pH 6.5 and that seen at pH 4.0 with respect to the extent and position of the elements of secondary structure. At pH 6.5 the NOE connectivities indicate two distinct helical regions, from Trp-4 to Arg-6 and from Lys-12 to Leu-15. The first region shows a tendency to form a short 3 10 -helix, which is stabilized by hydrogen bonds from the amides of Val-7 and Glu-5 to the carbonyls of Trp-4 and Met-3, respectively. In the second region, the presence of NOE connectivities between ␣N(ii ϩ 3) of Ile-10 and Ser-13 and of Lys-12 and Leu-15 and between ␣N(ii ϩ 4) of Ile-11 and Leu-15 strongly supports the observation of an ␣-helix in the 20 structures of lowest energy. However, we did not find hydrogen bonds that could stabilize a regular structure in the C-terminal portion. The residues in the center of the peptide have a slight tendency toward a helical or turn conformation based on the NOEs ␣N(ii ϩ 3) between Val-7 and Ile-10, Ile-10, and Ser-13 and ␣␤,␣␥,␣␦ (ii ϩ 3) between Lys-8 and Ile-11. Indeed, this region exhibits a bend, or kink, previously found in the fusogenic fertilization peptide from sea urchin fertilization protein binding (20) and in the fusion domain of the influenza hemagglutinin (4,19). At pH 4.0, the N-terminal domain displays a 3 10 -helix that extends from Glu-5 to Ser-9. This motif is stabilized by hydrogen bonds from the NHs of Ser-9, Val-7, and Met-3 to the carbonyls of Arg-6, Trp-4, and the hydroxyl of Ser-2, respectively. The NOEs ␣␤(ii ϩ 3) between Ile-11 and Ser-14, ␣␤(ii ϩ 4) between Lys-8 and Lys-12, and a long range connectivity between Ile-10 and Ala-16 and between Ile-11 and Leu-15 were observed exclusively at pH 4.0.
The structure of the C-terminal portion of the peptide at pH 4.0 differs substantially from that proposed for pH 6.5, because there was no helical content in any of the 20 lowest energy structures. The structural difference between pH 6.5 and pH 4.0 can be attributed to neutralization of Glu-5 (protonation at low pH). The number of observed NOEs involving Glu-5 in ␥ 1 -peptide at pH 4.0 is greater than at pH 6.5 ( Fig. 3; supplemental Fig.  2S). The presence of extra NOEs for Glu-5 and for other residues at pH 4.0 provides the strong evidence for the conformational differences between the two pH values. It is quite likely that negatively charged Glu-5 limits the conformational flexibility of the C-terminal portion that is more positively charged. Lipophilicity surface potential maps of the peptides at pH 6.5 and 4.0 showed the amphipathic nature of the two structures   ( Fig. 7). Fig. 7 also shows the arrangement of the hydrophobic residues based on the lowest energy structures at each pH value (Fig. 7, c and d). At pH 6.5, the more hydrophobic residues were observed in two major regions on each face of the molecule. Residues Trp-4, Ile-11, and Leu-15 appear on one face and residues Met-3, Trp-4, Val-7, and Ile-10 appear on the other (Fig.  7). At low pH, the polar and apolar regions are segregated so that the more hydrophobic residues mentioned above all lie on one side of the molecule (Fig. 7, right) with the more hydrophilic residues on the opposite side. Interestingly, the conformational change between pH 6.5 and pH 4.0 did not seem to affect the interaction between side chains from Lys-8 and Lys-12; they face each other at both pH values despite the differences of the chemical environment (Fig.  6). Lysine 8 and lysine 12 are of particular interest, because x-ray crystallography data from Flock House virus (33) show that both side chains from these lysines contact the phosphodiester backbone of the packaged viral RNA, and this interaction may be required to position the RNA correctly between the capsid subunits. It seems that this contact confers a certain organization on the RNA within the virion and may also be critical for formation of the protein subunit interactions that are established during assembly of the virion (15).
Effect of pH on ␥ 1 -Peptide-mediated Leakage of Dyes from Liposomes-In order to obtain more information about the interaction of ␥-peptide with biological membranes, the leakage of encapsulated DPX/ANTS from liposomes was monitored. The fluorescent dye leakage assay is a well established method for the study of pore formation in membranes (34,35). Fig. 8 shows the release of DPX/ANTS from liposomes of different composition, induced by addition of 50 M of ␥ 1 -peptide. The effects of ␥ 1 -peptide were highly dependent on lipid composition. A much greater release was observed with vesicles composed of PC/PE/SPM/Cho (1:1: 1:0.2) when compared with vesicles enriched with cholesterol (PC/PE/ SPM/Cho (1:1:1:1.5)) ( Fig. 8). For the two conditions (low and high cholesterol), there was no significant difference between pH values 4.0 and 6.5. Therefore, cholesterol shows a great importance for the interaction of FHV ␥ 1 -peptide with the target membrane and the induction of permeabilization. However, in both conditions, ␥ 1 -peptide binds to the membrane (Fig. 2B). This is relevant because the plasma membrane of insect cells contains around 10 times less cholesterol than those isolated from mammalian cells (36). Membranes with less cholesterol are more fluid, which may favor lateral association between the peptides. The critical dependence on the peptide concentration for the leakage (Table 2) corroborates the hypothesis of oligomerization in the liposomes with low cholesterol. On the other hand, the membrane destabilization cannot be associated with membrane fusion, because of the lack of mixing of the internal aqueous content of vesicle-entrapped DPX and ANTS (data not shown). The same results of leakage were obtained when another assay (carboxyfluorescein) was used (data not shown).

DISCUSSION
Here we show the solution NMR structure of a membranebinding peptide contained in the protein shell of a small, nonenveloped icosahedral virus. The three-dimensional structure of FHV ␥ 1 -domain was determined at pH 6.5 and 4.0 in a hydrophobic environment. The three-dimensional structure determined by x-ray crystallography for Flock House virus (11,33) showed that the first 15 residues of the 21 residues of ␥ 1 -peptide (residues 364 -385 of precursor protein ␥) are in an amphipathic ␣-helical conformation, whereas the remaining 28 C-terminal residues (385-407) of ␥-peptide were not visible in the x-ray electron density map. Although the atomic structure of the C-terminal portion of the ␥-peptide is still unknown, it plays a significant role in recognition of FHV RNA during assembly (37). The NMR solution structures of ␥ 1 -peptide also show an amphipathic helix, but it is less extensive and structur-ally different from the one determined by crystallography. The structural difference observed in the x-ray and NMR analyses are not surprising given that ␥-peptides associated with virus particles interact with the packaged viral genome. Specifically, the side chains of lysines 8 and 12 contact the negatively charged phosphodiester backbone of the RNA. During cell entry, the ␥-peptides are likely to be released from the virion, and their structure could change to that observed in our NMR analysis before insertion into a cellular membrane.
The observation of kinked NMR structures found in ␥ 1 -peptide in solution are remarkably similar to the structures determined for class I fusion peptides such as influenza hemagglutinin (4,19), human immunodeficiency virus, type 1, gp41 (38), and the sea urchin fertilization protein bindin "B18" (20). On the other hand, they differ from the NMR structure reported for the reptilian reovirus p14 fusion peptide, which has a loop structure found in class II fusion peptides (39). However, no fusion activity was detected for the ␥ 1 -peptide, although a high leakage activity was observed. The structure similarity indicates a common mechanism for the interaction between leakage-inducing domains of non-enveloped viruses and fusion domains of enveloped viruses. Indeed, Lai et al. (40) have recently demonstrated that a specific angle in the boomerang structure is required for membrane fusion. Fusion peptides have only to insert into the outer leaflet of the lipid bilayer, whereas the leakage activity requires formation of a channel across it.
The circular dichroism data demonstrated that the peptide adopts similar secondary structures in TFE/H 2 O, SDS, and in lipid vesicles. The NMR experiments in SDS and DPC micelles also point in the same direction, even though we were not able to solve the atomic structure in the micelles. Altogether, the NMR and CD data do show that the TFE/water structure can mimic what is seen with lipids.
As the pH is decreased from 6.5 to 4.0, ␥ 1 -peptide undergoes a conformational change, which leads to a less ordered struc-  5 (left panel) and pH 4.0 (right panel). a, view of the backbone superimposition of the 20 lowest energy structures between residues Met-3 and Ala-18 at pH 6.5 and pH 4.0. b, ribbon diagrams of the representative conformer at pH 6.5 and pH 4.0. Side chains for Lys-8 and Lys-12 are shown in blue. c and d show the surface plot in two views, following the axis. Surface potential maps are colored according to a lipophilicity scale, where brown is more lipophilic and blue is more hydrophilic.
ture at the C terminus. The NOE patterns observed for the N terminus of both peptides indicate the presence of a 3 10 -helix, which has been proposed to be an intermediate structure in the folding/unfolding of ␣-helices (41,42). Based on helix-coil theory, 3 10 -helix should be populated in the helix-coil transition (43). The pH-dependent structural change of ␥ 1 -peptide may reflect a conformational equilibrium between 3 10 -helix and random coil structures. The increase of pH favors the hydrogen bonding interaction between residues that stabilize 3 10 -and ␣-helical structures.
The combination of NMR and fluorescence results at both pH values shows that a change in pH promotes alteration in protein conformation but does not affect the ability of ␥ 1 -peptide to perturb mimetic membranes. In both pH values, Lys-8 and Lys-12 are spatially related, which may be a prerequisite for the interaction with the sugar-phosphate backbone during transfer of the encapsidated RNA. Based on these data we hypothesize that the interaction of ␥ 1 -peptide with mimetic membranes may not require a specific pH. On the other hand, we did find a high dependence on the lipid composition of the membrane (Fig. 8). Thus, the hydrophobic environment and pH were necessary conditions for the membrane binding, but lipid composition was the determining factor for leakage. An enticing hypothesis is that lipid constitution of the vesicles would affect oligomerization of the peptide, which would open the channel. The oligomerization may explain why we had much more difficulty in assigning the spectra in DPC micelles.
Altogether, the structural and fluorescence data indicate that the insertion of a protein subunit from an icosahedral virus into a target membrane is carried out by a structural motif similar to that found in enveloped viruses. However, it induces vesicle leakage and no fusion. Picornaviruses are closely related p ϭ 3 icosahedral viruses that cause diseases in humans and other mammals (44 -47). They contain a small protein (VP4) derived from maturation-proteolytic cleavage of VP0 that seems to assume a conformation similar to that of ␥-peptide in nodaviruses (44,46). Our results strongly suggest that small protein components such as ␣ protein in nodaviruses and VP4 in picornaviruses have crucial role in disrupting membranes during viral entry.