Antiviral Activity and Structural Characteristics of the Nonglycosylated Central Subdomain of Human Respiratory Syncytial Virus Attachment (G) Glycoprotein*

Segments of the cystine noose-containing nonglycosylated central subdomain, residues 149–197, of the attachment (G) glycoprotein of human respiratory syncytial virus (HRSV) have been assessed for impact on the cytopathic effect (CPE) of respiratory syncytial virus (RSV). Nα-acetyl residues 149–197-amide (G149–197), G149–189, and G149–177 of the A2 strain of HRSV protected 50% of human epithelial HEp-2 cells from the CPE of the A2 strain at concentrations (IC50) between 5 and 80 μm. Cystine noose-containing peptides G171–197 and G173–197 did not inhibit the CPE even at concentrations above 150 μm. Systematic C- and N-terminal truncations from G149–189 and G149–177 and alanine substitutions within G154–177 demonstrated that residues 166–170 (EVFNF), within a sequence that is conserved in HRSV strains, were critical for inhibition. Concordantly, G154–177 of bovine RSV and of an antibody escape mutant of HRSV with residues 166–170 of QTLPY and EVSNP, respectively, were not inhibitory. Surprisingly, a variant of G154–177 with an E166A substitution had an IC50 of 750 nm. NMR analysis demonstrated that G149–177 adopted a well-defined conformation in solution, clustered around F168 and F170. G154–170, particularly EVFNF, may be important in binding of RSV to host cells. These findings constitute a promising platform for the development of antiviral agents for RSV.

host cells followed by fusion of the lipid bilayers of the viral membrane envelope and the cell plasma membrane (6,7).
All members of the Paramyxoviridae have type I integral membrane glycoproteins with structural and functional homology that mediate membrane fusion (6,7). However, the structural and functional features of the type II membrane glycoproteins that mediate attachment vary considerably between genera (2, 3, 6 -8). Attachment proteins of the Respirovirus (e.g. Sendai virus) and Rubulavirus genera (e.g. mumps virus) have hemagglutinin-neuraminidase attachment proteins that recognize sialylated glycoconjugates on host cell membranes (6,7,9). By comparison, attachment proteins of the Morbillivirus genus, for example, measles virus, hemagglutinate but generally lack neuraminidase activity (6,7). Measles virus apparently utilizes CD46 in the process of cell recognition and attachment (10). RSV also has a type II integral membrane glycoprotein that is believed to mediate attachment (11,12), but it lacks hemagglutinin and neuraminidase activities (13). Pneumovirus attachment proteins are also remarkable because of their high carbohydrate content, which has resulted in them being termed G proteins (2, 3, 8, 14 -17). However, the term RSV attachment glycoprotein will be used herein to distinguish this viral glycoprotein from cellular G proteins.
A variety of observations implicate the nonglycosylated central subdomain as an important epitope for cell recognition and immunological activities of the RSV attachment glycoprotein. Human RSV (HRSV) (18, 19, 28 -31), bovine RSV (BRSV) (20), and ovine RSV (32) attachment glycoproteins contain four closely positioned cysteine residues spaced by the same number of amino acid residues (Fig. 1). The two N-terminally disposed cysteine residues are part of a sequence of six residues that varies at only one position in BRSV and ovine respiratory syncytial virus compared with HRSV (18 -20, 28, 32). Comparison of HRSV subtypes demonstrates that this identity is fur-ther extended in the N-terminal direction for HRSV such that a sequence of 13 residues is identical (18, 19, 28 -31). The glycosylated subdomains, which flank the nonglycosylated central subdomain, exhibit considerable sequence variability, even within subtypes (2, 8, 18 -20, 22, 28 -32). Conservation of sequence identity within the nonglycosylated central subdomain suggests an important biological role for this subdomain, such as a ligand for a cellular receptor (2,3,28). Immunological observations support this contention. For example, neutralizing monoclonal antibodies (33)(34)(35)(36), polyclonal antibodies (33,37), and antibodies from convalescent sera (33,34) map to the nonglycosylated central subdomain. In addition, a protective antigenic response is elicited upon vaccination of mice with synthetic peptides containing a portion of this sequence (38,39), recombinant proteins containing the subdomain (40,41), recombinant bacteriophage displaying a portion of this subdomain (42), and recombinant vaccinia viruses expressing the RSV attachment glycoprotein (43).
Chemical analysis of the central subdomain isolated from HRSV attachment glycoprotein (21) demonstrated that the cysteine residues oxidize to form disulfides with a 1 to 4 and 2 to 3 linkage pattern, described as a cystine noose (44), and the potential glycosylation sites within this subdomain are unoccupied (21). A cystine noose configuration was also indicated for BRSV by analyses of a synthetic peptide containing the bovine sequence with the four cysteine residues oxidized to form disulfides in vitro (45). Conformational detail has either been determined (46,47) or proposed (48) for the entire central subdomain of the BRSV attachment glycoprotein. The structure of the cystine noose in the C-terminal portion of a synthetic form of the BRSV subdomain was determined by NMR (46); however, the N-terminal portion was not structured. A subsequent molecular dynamics simulation study (48) suggested a specific S-shaped loop structure for the N-terminal portion of the BRSV subdomain. This arrangement of an Sshaped N-terminal portion followed by a cystine noose is reminiscent of the structure of the fourth domain of the 55-kDa human tumor necrosis factor receptor (HTNFr) (48,49).
The present report provides direct experimental evidence that the nonglycosylated central subdomain of the RSV attachment glycoprotein functions as a ligand for cellular receptors for RSV. In particular, the HRSV subdomain inhibited the cytopathic effect (CPE) of HRSV on HEp-2 cells. NMR analyses demonstrated that the cystine noose located in the C-terminal portion of the HRSV attachment glycoprotein subdomain has a very similar conformation to the BRSV attachment glycoprotein cystine noose. Surprisingly, the N-terminal portion of the HRSV subdomain was also found to adopt a defined conformation. The capacity to inhibit CPE was mapped to the N-terminal portion of HRSV attachment glycoprotein nonglycosylated central subdomain, and essential residues for this activity have been identified. These core residues serve to stabilize the conformation of the N-terminal portion of the nonglycosylated central subdomain of the HRSV attachment glycoprotein. The present findings represent a promising platform for the development of antiviral agents for RSV.

EXPERIMENTAL PROCEDURES
Synthetic Peptides-Peptides were synthesized using the solid-phase Fmoc (fluorenylmethyloxycarbonyl) approach (50) by Auspep (Melbourne, Australia). Unless otherwise stated, the nomenclature Gx-y indicates a peptide spanning residues x to y of the attachment glycoprotein of the A2 strain of HRSV with N␣-and C␣-terminal acetyl and carboxamide derivatives, respectively.
Molecular weights of purified peptides were determined by electrospray ionization on a PE-Sciex API-100 mass spectrometer and matrixassisted laser desorption/ionization on a Bruker Reflex time-of-flight mass spectrometer (21). Analysis of metastable ions produced by postsource decay was used to confirm the disulfide bond configurations of purified cystine noose-containing peptides (21).
Impact of Synthetic Peptides on the Cytopathic Effect of HRSV-Serial 2-fold dilutions of peptides were prepared in triplicate in 50 l of growth medium (Dulbecco's modified Eagle's medium/F-12; ICN) containing 7% fetal calf serum (CSL), penicillin/streptomycin (Life Technologies, Inc.), and fungizone (Squibb) in 96-well tissue culture plates. Approximately 1-2 ϫ 10 4 HEp-2 cells were added in 50 l of growth medium/well. Peptides and cells were incubated for 1.5 h at room temperature. Approximately 10 TCID 50 (50% tissue culture infectious doses) of the A2 strain of HRSV was then added to two series of the diluted peptides in an additional 10 l of growth medium. Growth medium alone was added to one dilution series of each peptide for toxicity controls. Plates were incubated for 5 days at 37°C, fixed in 1% formalin/saline, and stained with 0.05% neutral red to detect remaining viable cells. The IC 50 cited for a peptide was the concentration of the peptide required to protect 50% of cells from a level of virus that would have been cytopathic to all cells in the absence of an antiviral effector. In the presence of substantial protection, the entire well area stained red, at 50% protection, 50% of the well area stained red, and in the absence of protection, the wells appeared clear. IC 50 values are only presented for concentrations of peptides that did not diminish cell staining in the absence of virus. Ribavirin and G149 -177 were routinely incorporated into assays as a positive antiviral control and an internal standard, respectively.
NMR Spectroscopy-Samples were prepared for NMR by dissolving 5 mg of peptide in 0.55 ml of 10% 2 H 2 O/90% H 2 O. The pH was adjusted with small additions of 0.5 M NaO 2 H or 2 HCl, and reported pH values were measured at room temperature and were not corrected for isotope or solvent effects. 1 H chemical shifts were referenced to 2,2-dimethyl-2-silapentane-5-sulfonate at 0 ppm via the chemical shift of the H 2 O resonance (51) or an impurity at 0.15 ppm.
Spectra were recorded on a Bruker AMX-600 spectrometer. Unless stated otherwise, all spectra were recorded in 90% H 2 O/10% 2 H 2 O at 277 K and pH 5.0, with probe temperatures calibrated according to a previously published method (52). Spectra were recorded, structural constraints were determined, and structure calculations were carried out as described previously (53). Structures were analyzed using MOL-MOL (2.1.0) (54), and structural figures were generated using Insight II and MOLMOL.

Inhibition of the Cytopathic Effect of RSV by the Nonglycosylated Central Subdomain of the RSV Attachment Glycoprotein-
The sequence encompassing residues 152-187 of the HRSV attachment glycoprotein is not glycosylated (21). Furthermore, no sites for N-or O-linked glycosylation are present between residues 149 -151 and 188 -197 at the N and C termini of this sequence. Thus, synthetic G149 -197, with cysteines oxidized into a cystine noose, was the starting sequence for assessment of the impact of the nonglycosylated central subdomain on infectivity of RSV. G149 -197 inhibited the CPE of HRSV A2 strain at 80 M and was toxic to cells above this concentration. Truncated segments and/or variants of G149 -197 were made to determine whether this impact on CPE could be ascribed to specific sequence elements of G149 -197. A summary of the data obtained with these peptides is presented in Fig. 2.
Substantial N-terminal truncation to produce G163-197 did not diminish inhibitory activity. Although it was difficult to clearly distinguish between toxicity and CPE with G171-197, which contained the cystine noose, this peptide did not appear to inhibit CPE at 165 M. Another cystine noose peptide, G173-197, was not effective at inhibiting CPE even at 177 M.
A C-terminally truncated form of G149 -197, G149 -189, exhibited a greater inherent toxicity than G149 -197, but it was also more a more effective inhibitor and inhibited CPE at concentrations where no toxicity was evident in the absence of virus. Truncation of five residues from the N terminus was not detrimental to inhibitory activity, as was evident with G154 -189. However, removal of four more residues from the N terminus reduced inhibitory activity, as was seen with G158 -189.
The N-terminal portion of the subdomain, G149 -177, with two cysteine residues protected with the acetamidomethyl group and a C-terminal glycine amide, was more effective than G154 -189 at inhibiting CPE. Toxicity was only observed with G149 -177 at concentrations of Ͼ70 M. G154 -177 was as potent in inhibiting CPE as G149 -177, but further N-terminal truncation resulted in diminution of activity. G158 -177, G162-177, and G166 -177 were 1 order of magnitude less effective than G154 -177 at inhibiting CPE. G170 -177 was unable to inhibit CPE at 500 M.
The removal of four residues from the C terminus of G149 -177 reduced inhibitory activity by 1 order of magnitude. Removal of the next four residues completely abolished inhibitory activity, and lack of inhibitory activity persisted with removal of the next four residues. Single C-terminal residue truncations, starting with G154 -172, produced a trend wherein in-hibitory activity went from a low level with G154 -172, followed by enhancement of inhibitory activity to the level of 154 -177 with removal of residues 172 and 171. Removal of the next residue (F170) completely abolished inhibitory activity.
G154 -177 of the HRSV B subtype sequence was not as effective against the A2 strain as the homologous A2 peptide. However, the HRSV B subtype peptide was much more effective than homologous A2 peptides with deletions in the block of residues between 166 and 170. G154 -177 of BRSV failed to inhibit CPE caused by the A2 subtype of HRSV at concentrations up to 337 M. Furthermore, a peptide with substitutions of Phe-168 and Phe-170 with Ser and Pro, respectively, based on an antibody escape mutant of the Long A strain of HRSV (35,36), did not inhibit CPE at concentrations up to 358 M.
In summary, (Fig. 2), the above data indicate that inhibition of CPE by the nonglycosylated central subdomain of the RSV attachment glycoprotein 1) required a core sequence of residues 154 -170 for optimal activity, 2) was influenced by residues 154 -157 and 171-176, 3) required F170, and 4) may also have required residues 166 -169.
The influence of individual residues within the stretch of residues 165-176 was assessed by systematic alanine substitutions (Table I). It was notable that the F168A substitution had some negative impact on inhibitory activity of G154 -177, substitution of both F168 and F170 (F168/170A) completely abolished CPE inhibition by G149 -177, and, surprisingly, substitution of E166 enhanced the ability of G154 -177 to inhibit CPE into the submicromolar range.
Solution Structure of the Nonglycosylated Central Subdomain of the RSV Attachment Glycoprotein-Sequence-specific 1 H NMR resonance assignments for peptides corresponding to G149 -189, G149 -177, and G171-197 of the central subdomain of HRSV were made from two-dimensional nuclear Overhauser enhancement spectroscopy and total correlation spectroscopy spectra (55). Due to severe overlap and aggregation, it was not possible to determine the structure of G149 -189 from twodimensional NMR measurements. Plots of the deviation of the C ␣ H and NH chemical shifts for G149 -177 and G171-197 from those of G149 -189 are shown in Fig. 3. The chemical shift differences between the two short peptides and the longer peptide are generally small and limited to the N and C termini. This suggests that the structure of G149 -189 could be approximated by superimposition of the structures of G149 -177 and G171-197 over common residues Val-171 and Pro-172, which displayed similar chemical shifts in G149 -177 and G171-197 compared with G149 -189 (Fig. 3).
The presence of a C ␣ H-C ␣ H NOE between residues Lys-158 and Pro-159 in both G149 -189 and G149 -177 indicated that this peptide bond was in a cis conformation for both peptides, whereas all other peptide bonds involving proline residues had a trans conformation, as evidenced by presence of strong NOEs between the C ␦ H of the Pro residues and the C ␣ H of the preceding residues.
Structures were calculated for G149 -177 using 582 upperbound distance constraints inferred from NOEs, made up of 121 intraresidue, 147 sequential, 218 medium-range (1 Ͻ i-j Յ 4), and 96 long-range NOEs. In addition, 24 backbone dihedral angle constraints based on spin-spin coupling constants were included; no 1 side chain constraints for stereospecifically assigned, nondegenerate, geminal C ␤ H resonances were employed. Structures were initially calculated using DYANA and then refined by simulated annealing in X-PLOR and finally energy-minimized in X-PLOR with the CHARMM force field. A summary of geometric and energetic parameters for these structures is given in Table II. Analysis of the backbone angular order parameters (S) (56, 57) of the final 20 structures of G149 -177 indicated that residues 151-176 were well defined, with S Ͼ 0.8 for both and angles. The backbone RMSD from the mean structure confirmed that the structure was well defined over most of the molecule, except for residues near the C terminus. The overall conformation of G149 -177 is shown in Fig. 4, where the backbone heavy atoms of the 20 best structures (those with the lowest overall energies, excluding the electrostatic term) have been superimposed over residues 151-176.
Structures were calculated for G171-197, which contains the four Cys residues that form two disulfide bridges, using 478 upper-bound distance constraints inferred from NOEs, made up of 92 intraresidue, 132 sequential, 169 medium-range (1 Ͻ i-j Յ 4), and 85 long-range NOEs. In addition, six backbone dihedral angle constraints based on spin-spin coupling constants were included, and one 1 side chain constraint for stereospecifically assigned, nondegenerate, geminal C ␤ H resonance was employed. A summary of geometric and energetic parameters for these structures is given in Table III. Analysis of the backbone angular order parameters (S) (56,57) of the final 20 structures of G171-197 indicated that residues 171-191 were well defined, with S Ͼ 0.8 for both and angles. The resulting structure is a well-defined loop with two short helical segments encompassing residues 173-176 and 180 -185 (Fig. 6). Despite some significant differences in sequence, the structure of the cystine noose fragment of HRSV was very similar to that of BRSV (46). The RMSD between the two families of structures over residues 173-186 (the cystine noose) is 0.65 Ϯ 0.53 Å.

DISCUSSION
Demonstration that peptides from the nonglycosylated central subdomain of the HRSV attachment glycoprotein inhibited the CPE of HRSV provides direct evidence that this subdomain is important for biological function of the RSV attachment glycoprotein. The observation that highly conserved residues 166 -170 within a larger stretch of conserved residues of the human subtypes are least tolerant to substitution and deletions ( Fig. 2; Table I) indicates that this sequence conservation is of particular importance to biological function. The finding that the cystine noose portion of the subdomain did not inhibit CPE may indicate that it has a separate biological function from that mediated by the N-terminal residues of the subdomain. It is unlikely that failure of this cystine noose-containing C-terminal portion to inhibit CPE was due to failure to fold properly because it was found to have the same structure as previously demonstrated for the cystine noose of the BRSV attachment glycoprotein (46). It was somewhat surprising to find that the N-terminal portion of the nonglycosylated central subdomain of the HRSV attachment glycoprotein adopted a stable solution conformation because this region of the BRSV attachment glycoprotein was previously found to lack a defined structure (46).
The finding that the NH and C ␣ H chemical shifts of G149 -177 and G171-197 did not vary significantly from those observed for G149 -189 (Fig. 3) suggested that the structures of the two peptide fragments are retained in the overall structure of the central subdomain of the HRSV attachment glycoprotein and that interactions between the fragments are limited. The structures of G149 -177 and G171-197 were oriented to approximate the structure of the nonglycosylated central subdomain of HRSV by overlapping over residues 171 and 172 (Fig.  6). Val-171 and Pro-172 were overlapped because the similarities of the chemical shifts of these residues when present in G149 -177 or G171-197 compared with when present in G149 -189 indicated that these residues are in the same chemical environments in the three peptides. The structures and biological activities of G149 -177 and G171-197 manifested independently of their relative orientation in G149 -189, thus, the present model of the human subdomain has been compared with a previous model for the central subdomain of the BRSV The best 20 structures after energy minimization in the distance geometry force field were subsequently energy-minimized in the CHARMM force field using a distance-dependent dielectric, as described under "Experimental Procedures." b The numbers of restraints are shown in parentheses. None of the structures had distance violations Ͼ 0.5 Å or dihedral angle violations Ͼ5°.
c Backbone heavy atoms. d All heavy atoms.  The best 20 structures after energy minimization in the distance geometry force field were subsequently energy-minimized in the CHARMM force field using a distance-dependent dielectric, as described under "Experimental Procedures." b The numbers of restraints are shown in parentheses. None of the structures had distance violations Ͼ 0.5 Å or dihedral angle violations Ͼ5°.
c Backbone heavy atoms. d All heavy atoms. attachment glycoprotein (48). Diminution of antiviral activity observed with truncations between residues 171 and 177 of G149 -177 and G154 -177 may have been due to steric occlusion of interactions necessary for antiviral activity. Steric effects could have been caused by altered configurations of the C termini of the truncated peptides.
A similarity in structure of the central subdomain of the BRSV attachment glycoprotein and the C-terminal module of the fourth subdomain of HTNFr has been proposed (48). These molecules have an identical cystine noose motif (i.e. two cystine bridges similarly spaced and linked in a 1-4 plus 2-3 arrangement). Although the N-terminal portion of the nonglycosylated central subdomain of the BRSV attachment glycoprotein had previously been described as unstructured (46), molecular dynamics modeling studies indicated structural homology between the N-terminal part of the nonglycosylated central subdomain of BRSV and the A1 module of HTNFr (48). This module of HTNFr forms a loop that is held together by a cystine bridge (49). Because cysteine residues are not present in the corresponding regions of the BRSV and HRSV attachment glycoproteins, it was proposed that interactions between His-159 and Tyr-170 of the BRSV attachment glycoprotein may provide compensatory stability to the BRSV peptide (48). However, the loop structure determined in the present study for the N-terminal loop of the nonglycosylated central subdomain of the HRSV attachment glycoprotein appears to have little similarity to the A1 module of HTNFr. It was noteworthy that His-159 and Tyr-170 in BRSV are replaced by Pro and Phe, respectively, in HRSV. These residues appear to be involved with other residues in hydrophobic interactions that stabilize the loop (Fig. 5).
One explanation for the inhibition of CPE by N-terminal peptides from the central subdomain of the HRSV attachment glycoprotein is that they interacted directly with a cellular receptor for the corresponding sequences of the viral attachment glycoprotein (2,3,28) and thereby abrogated infectivity. The RSV attachment glycoprotein has been found to be dispensable for infection of cultured cells by RSV (58 -60); syncytium formation occurs with cells expressing the RSV fusion glycoprotein alone (61,62), and the RSV fusion glycoprotein appears to be able to mediate attachment via glycosaminoglycan interaction (26,58,60,63). However, recombinant HRSV passages 10 -20 times more efficiently in cell culture when both the G and fusion glycoproteins are present compared with when only the fusion glycoprotein is expressed (59). Cold passage-derived RSV, with the attachment glycoprotein deleted, was not viable in vivo (58), and syncytium formation by cells expressing the RSV fusion glycoprotein was enhanced by co-expression of the attachment glycoprotein (61, 62). Furthermore, antibodies that map to the nonglycosylated central subdomain of the attach-ment glycoprotein neutralize RSV (33)(34)(35)(36)(37), and antigen preparations containing this portion of the attachment glycoprotein elicit a protective immune response when used as vaccines (38 -43). Thus, interaction of the RSV attachment glycoprotein with receptor(s) would appear to be important to the infectious process of RSV. Although mutations in this region of the RSV attachment glycoprotein result in the ability to evade neutralizing antibodies (35)(36)(37), these mutants may achieve infectivity via fusion glycoprotein attachment alone (26,58,60,63). Interestingly, the peptide used in this study that was based on these mutant viruses had Phe residues at positions 168 and 170 replaced by Ser and Pro, respectively, in the region most critical to CPE inhibition, residues 166 -170. This peptide was totally ineffective at inhibiting CPE, as was the BRSV peptide in which residues 166 -170 are totally variant from HRSV. It may be that this region of the nonglycosylated subdomain, which is conserved in all natural isolates of HRSV but variant in other RSVs, may have a role in determining species susceptibility through stereochemical interactions with receptors of individual species. RSV attachment glycoprotein-mediated attachment also appears to involve interactions with glycosaminoglycans (23)(24)(25)(26)(27) via an epitope encompassed by residues 184 -198 (24). The dependence on residues 166 -170 for inhibition of CPE by the N-terminal portion of the nonglycosylated central subdomain of the attachment glycoprotein and failure of the cystine noose portion to affect CPE inhibition indicate that receptor-mediated attachment of RSV is a multifaceted process. Attachment may involve interactions of attachment (23)(24)(25)(26)(27) and fusion glycoproteins (26,58,60,63) with glycosaminoglycans in addition to interactions involving other portions of the nonglycosylated central subdomain of the attachment glycoprotein.
Another possible role for residues 166 -170 is that they form a self-association domain for the attachment glycoprotein such that other features of the glycoprotein are presented to a cellular receptor for productive viral attachment. A role in selfassociation would be consistent with multimeric models proposed for the maturation of the RSV attachment glycoprotein (22,45,64). Interaction of the synthetic peptides with the viral attachment glycoprotein in a manner that interfered with the self-association could explain their antiviral properties.
Sequences flanking residues 166 -170 of the N-terminal portion of the subdomain may participate in secondary receptor/ associative interactions and/or may provide a scaffold to present residues 166 -170 in the appropriate context for receptor recognition/self-association and/or enhance the solubility of 166 -170. The C-terminal cystine noose portion may also have such a scaffold role in the context of the entire protein structure. Additional biological and biochemical studies will be required to define the inhibitory mechanism of the peptides from the N-terminal portion of the nonglycosylated central subdomain.
The present study has identified G154 -177(166A) as a realistic lead for the development of antiviral agents for RSV. Additional substitution studies should help to delineate the roles of individual residues and may also produce fortuitous enhancement of efficacy such as that seen with G154 -177(E166A). Such knowledge could be exploited for synthesis of more potent peptides and/or peptidomimetics. An understanding of the mechanism of action of the peptides from the Nterminal portion of the nonglycosylated subdomain of the RSV attachment glycoprotein would also guide development of more efficacious antiviral agents for RSV. Peptide-based probes are currently being used to identify receptors for this portion of the attachment glycoprotein and/or delineate the mechanism of their antiviral activity.