Neutralizing antibodies against the preactive form of respiratory syncytial virus fusion protein offer unique possibilities for clinical intervention

Abstract

Human respiratory syncytial virus (hRSV) is the most important viral agent of pediatric respiratory infections worldwide. The only specific treatment available today is a humanized monoclonal antibody (Palivizumab) directed against the F glycoprotein, administered prophylactically to children at very high risk of severe hRSV infections. Palivizumab, as most anti-F antibodies so far described, recognizes an epitope that is shared by the two conformations in which hRSV_F can fold, the metastable prefusion form and the highly stable postfusion conformation. We now describe a unique class of antibodies specific for the prefusion form of this protein that account for most of the neutralizing activity of either a rabbit serum raised against a vaccinia virus recombinant expressing hRSV_F or a human Ig preparation (Respigam), which was used for prophylaxis before Palivizumab. These antibodies therefore offer unique possibilities for immune intervention against hRSV, and their production should be assessed in trials of hRSV vaccines.

Human respiratory syncytial virus (hRSV) is the most important viral cause of severe respiratory tract disease in the pediatric population worldwide (1) and is also of considerable importance in the elderly and immunocompromised adults (2). There is no vaccine available against this virus. A trial conducted with formalin-inactivated hRSV in the 1960s did not confer protection and was associated with enhanced disease in infants upon natural infection with the virus (3).

hRSV belongs to the Pneumovirus genus of the Paramyxoviridae family. The viral genome consists of a single-stranded RNA molecule of negative polarity that encodes 11 proteins (4). Two of these proteins are the major surface glycoproteins of the virion, namely: (i) the attachment protein (G) that mediates binding of the virus to the cell surface and (ii) the fusion protein (F) that promotes fusion of the virus and cell membranes during virus entry and also the fusion of the membranes of infected cells with surrounding cells to form syncytia.

Neutralizing antibodies play a major role in protection against hRSV infections. F and G proteins are the only viral antigens able to induce both neutralizing antibodies and relatively long-lived protection in animal models (5, 6). Passive transfer of these antibodies protects mice and cotton rats against a hRSV challenge (7, 8). In human, high titers of serum neutralizing antibodies correlate with protection of adult volunteers against a hRSV challenge (9), and lower risk of hRSV infection in children (10) and the elderly (11). These observations prompted the prophylactic use of an Ig preparation (Respigam) containing high titers of neutralizing antibodies to prevent hRSV infections in high risk infants (12). Respigam has been replaced by a humanized monoclonal antibody (Palivizumab) directed against the hRSV_F glycoprotein (13), which is the only available medicinal product specific for hRSV.

hRSV_F is a type I glycoprotein that assembles as a homotrimer. Each monomer is synthesized as an inactive precursor (F0) that needs to be cleaved at two polybasic sites (I and II) to become fusion competent (14). Site II is equivalent to the single cleavage site of other paramyxovirus F proteins, which precedes a hydrophobic fusion peptide that is inserted into the target membrane during fusion. It has been postulated that the paramyxovirus F proteins remain in a metastable prefusion conformation in the virus particle until the virus binds to the target membrane. The F protein is then activated to initiate a series of conformational changes so that fusion occurs “at the right time and in the right place” (15). After fusion, F acquires a highly stable postfusion conformation determined mainly by the formation of a six-helix bundle (6HB), composed of sequences of two heptad repeats (HRA and HRB) from each monomer (16). The free energy released during the transition of the F protein from the prefusion to the postfusion structure drives the process of membrane fusion.

Palivizumab and most α-F monoclonal antibodies (MAbs) described so far recognizes the postfusion conformation of hRSV_F, represented in an anchorless form of the F protein, referred to as F_TM- (17). This artificial protein, engineered to remove the transmembrane region and the cytoplasmic tail of F, is secreted to the culture medium and it folds spontaneously into the postfusion conformation. However, neutralizing MAbs that recognize F_TM- inhibit infectivity if preincubated with the virus before being used to infect cells (18); i.e., before the F protein is activated for fusion. It is likely therefore that most α-F MAbs hitherto described recognize epitopes shared by both the pre- and postfusion forms of hRSV_F.

We have published that human Ig (Ig) preparations contain neutralizing α-F antibodies that could be purified by binding to F_TM- covalently linked to Sepharose beads (19). However, the antibodies that did not bind to F_TM- retained most of the neutralizing activity present in the Ig preparations (even after depletion of α-G antibodies) and were able to bind to cells infected with either hRSV or a recombinant vaccinia virus that expressed hRSV_F. It appears therefore that certain human neutralizing antibodies recognize structures of the F protein, other than those represented in F_TM-. We now report that this type of antibodies can be obtained from rabbits immunized with recombinant vaccinia viruses expressing a full-length hRSV_F. In addition, both rabbit and human neutralizing α-F antibodies, depleted of those reacting with F_TM-, bind to a prefusion form of hRSV_F, stabilized with intermonomeric disulfide bonds. These results therefore offer possibilities to develop unique α-hRSV neutralizing antibodies with potential applications in the clinic.

Results

Induction of Neutralizing Antibodies in Rabbits Inoculated with Recombinant Vaccinia Viruses Expressing Either hRSV_F or hRSV_F_TM-.

New Zealand White rabbits were inoculated intramuscularly (i.m.) with recombinant vaccinia viruses expressing either a full-length copy of hRSV_F (Vac/Fc) or an anchorless mutant of the same protein (Vac/F_TM-). Representative results of each group of rabbits, shown in Fig. 1, demonstrated that both sera contained antibodies able to bind to purified F_TM- in an ELISA (Fig. 1A) and to inhibit hRSV infectivity in a microneutralization test (Fig. 1B). However, the sera of rabbits inoculated with Vac/Fc (named α-Fc, see Methods for the system used to label the antibodies) neutralized hRSV significantly better than the sera of rabbits inoculated with Vac/F_TM- (α-F_TM-).

Fig. 1.

Induction of binding and neutralizing antibodies in rabbits immunized with recombinant vaccinia viruses expressing different forms of hRSV_F. Serial dilutions of sera from rabbits inoculated with either Vac/Fc (α-Fc) or Vac/F_TM- (α-F_TM-) were tested for binding to purified F_TM- adsorbed to ELISA plates (A) or hRSV neutralization (B). Control serum from a preimmune rabbit was included as control. Total α-Fc antibodies were purified by protein A-Sepharose chromatography from the sera of rabbits immunized with Vac/Fc. The α-Fc antibodies were loaded onto a column of F_TM-Sepharose, and the unbound antibodies (α-Fc/ΔF_TM-) were collected and saved. After washing, the antibodies bound to the column (α-Fc/F_TM-) were eluted. The α-Fc, α-Fc/ΔF_TM-, and α-Fc/F_TM- antibodies were then tested for binding to F_TM- (C) and for hRSV neutralization (D). Total α-F_TM- antibodies from rabbits inoculated with Vac/F_TM- were processed similarly to yield the α-F_TM-/ΔF_TM- antibodies that did not bind to the F_TM-Sepharose column and the α-F_TM-/F_TM- antibodies eluted from this column. α-F_TM-, α-F_TM-/ΔF_TM-, and α-F_TM-/F_TM- antibodies were tested for binding to F_TM- (E) and hRSV neutralization (F). Results are presented as means ± SEM of three independent experiments.

The α-Fc antibodies were purified with protein A-Sepharose and subsequently loaded onto a column of F_TM- protein covalently linked to Sepharose beads. As expected, the antibodies that did not bind to the column (α-Fc/ΔF_TM-) failed to react with the F_TM- protein in an ELISA (Fig. 1C), whereas the α-Fc/F_TM- antibodies eluted from the column showed a higher level of binding to F_TM- than the starting antibodies (α-Fc). Remarkably, although the α-Fc/F_TM- antibodies inhibited hRSV infectivity in a neutralization test, the unbound α-Fc/ΔF_TM- antibodies retained most of the neutralizing activity of the original α-Fc antibodies (Fig. 1D). Considering that the antibodies eluted from the affinity column (α-Fc/F_TM-) should be specific only for the F protein whereas α-Fc and α-Fc/ΔF_TM- antibodies should contain multiple specificities present in a rabbit serum, the results of Fig. 1D underscore the neutralizing capacity of α-Fc/ΔF_TM-, although direct comparison of specific activities between α-Fc/F_TM- and α-Fc/ΔF_TM- would require estimation of the percentage of F-specific antibodies in each antibody preparation.

The antibodies from rabbits inoculated with Vac/F_TM- (referred to as α-F_TM-) were processed similarly to the α-Fc antibodies. Again, the antibodies not retained in the column of F_TM- (α-F_TM-/ΔF_TM-) were unable to bind to this protein in an ELISA, whereas the antibodies eluted from the column (α-F_TM-/F_TM-) showed a higher level of binding to F_TM- than the starting material (Fig. 1E). However, in this case, the antibodies unbound to the column of F_TM- were unable to neutralize hRSV infectivity (Fig. 1F), in clear distinction with the equivalent antibodies obtained from rabbits inoculated with Vac/Fc (compare Fig. 1 D and F). In addition, both α-F_TM- and α-F_TM-/F_TM- antibodies were significantly less neutralizing than the corresponding antibodies obtained from rabbits inoculated with Vac/Fc (compare the x axis of Fig. 1 D and F).

Characterization of the Neutralizing Activity of the α-Fc/ΔF_TM- Antibodies.

Several tests were performed to confirm that the neutralizing activity of the α-Fc/ΔF_TM- antibodies was due to antibodies specific against hRSV_F and not to spurious cross-reactions.

First, the reactivity of α-Fc, α-Fc/ΔF_TM-, and α-Fc/F_TM- antibodies was tested by flow cytometry with HEp-2 cells infected with hRSV. As shown in Fig. 2A, the α-Fc/ΔF_TM- antibodies could bind specifically to the surface of hRSV-infected cells, although the level of fluorescence was lower than that achieved with α-Fc and α-Fc/F_TM-.

Fig. 2.

Rabbit α-Fc antibodies recognize HRSV_F expressed at the cell surface even if depleted of antibodies able to bind to F_TM- (α-Fc/ΔF_TM-). HEp2 cells were infected with hRSV (Long strain, multiplicity of infection 5 pfu per cell) and tested 48 h later by flow cytometry for cell surface labeling with α-Fc, α-Fc/ΔF_TM-, and α-Fc/F_TM- antibodies (A). Alternatively, HEp-2 cells were infected with recombinant vaccinia virus, either hRSV_P (Vac/P) or hRSV_Fc (Vac/Fc), and tested 24 h later by flow cytometry for cell surface labeling with α-Fc, α-Fc/F_TM-, and α-Fc/ΔF_TM- antibodies before (B) or after adsorption of the antibodies to cells infected with Vac/P (C) or Vac/Fc (D). Mock infected cells (dashed lines) were included always as control. Note the reactivity of α-Fc/F_TM- antibodies with cells infected with Vac/P (B Right), although lower than the reactivity of α-Fc and α-Fc/ΔF_TM- probably because anti-vaccinia antibodies that bound unspecifically to the F_TM- column. Results are representative of at least three independent experiments.

To exclude illegitimate cross-reactions of α-vaccinia virus antibodies, the α-Fc, α-Fc/ΔF_TM-, and α-Fc/F_TM- preparations were adsorbed to HEp-2 cells infected with a vaccinia recombinant expressing the phosphoprotein (P) of hRSV (Vac/P). Before adsorption, the three antibody preparations reacted positively by flow cytometry with cells infected with either Vac/Fc or Vac/P (Fig. 2B). However, after adsorption to HEp-2, cells infected with Vac/P the reactivity with cells infected with this virus was lost while the reactivity with cells infected with Vac/Fc remained essentially unchanged (Fig. 2C). Finally, the three antibody preparations were adsorbed to HEp-2 cells infected with Vac/Fc. In this case, the three adsorbed antibodies lost the reactivity with cells infected with either Vac/P or Vac/Fc (Fig. 2D).

Antibody preparations were further analyzed for binding to F_TM- and hRSV neutralization before and after the adsorptions of the previous paragraph. Adsorption to cells infected with Vac/P (ΔVac) had no impact on either antibody binding to purified F_TM- (Fig. 3 A–C) or hRSV neutralization (Fig. 3 D–F). In contrast, adsorption of the three antibody preparations to cells infected with Vac/Fc reduced the reactivity of α-Fc (Fig. 3A) and α-Fc/F_TM- (Fig. 3C) to background levels, whereas the reactivity of α-Fc/ΔF_TM- remained negligible (Fig. 3B). The most significant result was that the neutralizing activity of the three antibody preparations was abrogated after adsorption to cells infected with Vac/Fc (Fig. 3 D–F).

Fig. 3.

α-Fc, α-Fc/ΔF_TM-, and α-Fc/F_TM- antibodies lost their neutralizing activity after incubation with cells infected with Vac/Fc. The three antibody preparations of Fig. 2 were depleted of antibodies binding to cells infected with either Vac/P (ΔVac) or Vac/Fc (ΔFc) and then tested either before or after the different adsorptions for binding to F_TM- (A–C) and for hRSV neutralization (D–F). Although the antibodies incubated with cells infected with Vac/Fc were also depleted of anti-vaccinia antibodies (i.e., they were also ΔVac), they were labeled ΔFc to distinguish them of those incubated with cells infected with Vac/P (labeled ΔVac in this figure). Results are presented as means ± SEM of three independent experiments.

In summary, the neutralizing activity of α-Fc antibodies depleted of those binding to F_TM- (α-Fc/ΔF_TM-) was removed by adsorption to cells infected with Vac/Fc (Fig. 3E). At this point, we postulated that the α-Fc/ΔF_TM- antibodies recognize a conformation of hRSV_F, which the protein adopts when inserted into membranes but not when expressed as a soluble anchorless molecule. The possibility that this conformation may correspond to the prefusion form of hRSV_F is supported by the results of Fig. S1. Thus, when hRSV was preincubated with saturating amounts of α-Fc, α-Fc/ΔF_TM-, or α-Fc/F_TM- and then pelleted to eliminate any free remaining antibodies before being used to infect new cells, virus infectivity was already inhibited. Therefore, the three antibody preparations (and, in particular, α-Fc/ΔF_TM-) could bind and neutralize hRSV before the F protein was activated for fusion. The results of Fig. S1 also suggest that neutralizing α-Fc/F_TM- antibodies may recognize epitopes shared by the pre- and postfusion forms of hRSV_F (see below under Discussion).

Stabilization of the Prefusion Form of hRSV_F and Reactivity with Rabbit and Human Antibodies.

To test whether α-Fc/ΔF_TM- antibodies could bind to the prefusion form of hRSV_F removed from a membrane environment, we attempted to purify this protein in the prefusion conformation. Previous efforts to solubilize the prefusion form of hRSV_F have failed, most likely because of the metastable state of this conformation and its spontaneous refolding to the postfusion form upon manipulation. Therefore, stabilization of the prefusion form of hRSV_F was attempted via the introduction of intermonomeric disulfide bridges that would prevent refolding after solubilization (Fig. 4A). Several artificial disulfide bonds were designed, based on a model of the prefusion hRSV_F (Fig. 4B), built with the atomic coordinates of the structure solved for the homologous parainfluenza virus 5 (PIV5) F protein (20).The best results were obtained with a protein referred to as FcN_2C-C, in which four amino acids were mutated to cysteines (shown schematically in Fig. 4 A and B). The basic residues at the two polybasic cleavage sites of FcN_2C-C and the wild-type hRSV_F (referred as FcN) were replaced by asparagine residues to avoid cleavage, as done with PIV5 (20). Finally, a His-tag was added to the C terminus of each protein to facilitate purification in Ni(2+) columns, and both constructs were incorporated into vaccinia virus recombinants. The location of introduced cysteine residues is shown in homology models of both the pre- and postfusion structures in Fig. 4 B and C. It is worth stressing that the extra disulfide bonds can only be formed intermonomerically and only if the F protein is in the prefusion conformation (Fig. 4B), but not if refolded to the postfusion structure (Fig. 4C).

Fig. 4.

Stabilization of the prefusion form of hRSV_F with intermonomeric disulfide bonds. (A) Scheme of the hRSV_F primary structure, denoting the three main hydrophobic regions: signal peptide (SP), fusion peptide (FP), and transmembrane region (TM). Also indicated are the two furin cleavage sites (arrows) that generate the F2 and F1 chains and the cysteines (gray dots and numbers) that participate in the intramonomeric disulfide bonds observed in the pos-fusion structure of hRSV. HRA and HRB represent the heptad repeat sequences of the F1 chain. Partial sequences, shown below the scheme, highlight the asparagines (N, boldface) introduced to ablate furin cleavage in FcN and FcN_2C-C constructs and the cysteines introduced in the latter construct (shown in orange) to make new intermonomeric disulfide bonds. The His-tag sequence added at the C terminus is shown above the scheme. (B) 3D model of the prefusion conformation of hRSV_F built with the coordinates of the PIV5_F protein (20) (PDB ID code 2B9B). Each monomer is colored differently. Cysteine residues (C) replacing residues 481, 489, 509, and 510 in FcN_2C-C are represented by golden balls alternating in two of the monomers (magenta and blue). (C) Postfusion structure of hRSV_F, reported by MacLellan et al. (26) (PDB ID code 3RRR). Colors and locations of engineered cysteine residues are presented as in B. Purified FcN, FcN_2C-C, and F_TM- proteins (Methods) were resolved by SDS/PAGE and either stained with Coomassie blue (D) or electrotransferred to nylon membranes and stained by Western blot with α-F_TM- antibodies (E). Electron microscopy of the three purified proteins (F–H) negatively stained with uranyl acetate. (Scale bar: 50 nm.) Two-times-magnified selected molecules are shown in Insets.

SDS/PAGE of boiled samples under nonreducing conditions showed the presence after Coomassie blue staining of a high molecular band (>250 kDa) in purified FcN_2C-C—compatible with a F protein trimer—which was absent in FcN (Fig. 4D). Under these conditions, the FcN lane contained a major band of ≈80 kDa, corresponding to the F monomer (F0) and a minor band of ≈190 kDa, which might be an F protein dimer, although the nature of this band has not been investigated further. These two bands were seen as minor components of the unreduced FcN_2C-C, indicating that formation of the extra disulfide bonds was not 100% efficient. Purified F_TM- (also modified to change the cleavage sites as in FcN and FcN_2C-C; SI Methods) yielded a single band migrating slightly faster than F0, in agreement with its lower molecular mass. The relatedness of all these bands with hRSV_F was demonstrated by Western blotting with α-F_TM- antibodies (Fig. 4E).

Electron microscopy of negatively stained FcN revealed cone-shaped molecules with a uniform contour and size (Fig. 4G) that resembled the structure of F_TM- in the postfusion conformation (Fig. 4F), except for a smaller globular density at one end of FcN that may represent detergent bound to the transmembrane region of this molecule and that is absent in F_TM-. In contrast, FcN_2C-C molecules were morphologically more heterogeneous but with a significant proportion of rounded shape structures (Fig. 4H), resembling the heterogeneity and shape reported for the prefusion form of PIV5_F (21). Recently, Chaiwatpongsakorn et al. (22) have published electron micrographs of hRSV_F in the prefusion conformation (obtained by alternative methods to those used here) that also resemble the images of FcN_2C-C molecules shown in Fig. 4H.

The preparations of rabbit antibodies described in previous sections were tested by ELISA for binding to purified F_TM-, FcN, and FcN_2C-C proteins. The results of Fig. 5A demonstrate that α-Fc and α-F_TM- antibodies were able to bind similarly to the three proteins, whereas α-Fc/ΔF_TM- antibodies were able to bind to FcN_2C-C but failed to bind to F_TM- and FcN. These results strongly support the conclusion that α-Fc/ΔF_TM- antibodies are specific for the prefusion form of hRSV_F, represented in the FcN_2C-C and, therefore, do not require a membrane environment for binding. The conclusion that the FcN_2C-C protein is in the prefusion conformation is further supported by the lack of binding of antibodies specific for the 6-helix bundle (a structure unique of the postfusion form), whereas these antibodies bound efficiently to the F_TM- and FcN proteins (Fig. 5A, Bottom Right).

Fig. 5.

The stabilized prefusion hRSV_F protein is bound by rabbit and human neutralizing antibodies depleted of those that recognize F_TM-. (A) The three proteins, F_TM-, FcN, and FcN_2C-C of Fig. 4 were bound to ELISA plates via a bridge made with MAb 101F and then tested for binding to the antibodies indicated at the bottom of each image. α-Fc, α-F_TM-, and α-Fc/ΔF_TM- have been described in Fig. 1. The α-6HB antibodies (Methods) were raised in rabbits inoculated with a bacterially expressed construct that mimics the six-helix bundle present in the postfusion form of hRSV_F (28). RG is the Ig preparation (Respigam) used some years ago in the clinic (a generous gift of MedImmune). A sample of this preparation was also depleted of antibodies binding to F_TM-Sepharose (RG/ΔF_TM-). (B) RG was depleted sequentially of antibodies binding to F_TM-Sepharose (RG/ΔF_TM-), G-Sepharose (RG/ΔF_TM-/ΔG), and cells infected with either Vac/Fc (RG/ΔF_TM-/ΔG/ΔFc) or Vac/FcN_2C-C (RG/ΔF_TM-/ΔG/ΔFcN_2C-C), and all these antibody preparations were used in a microneutralization test with hRSV. C- is a preimmune rabbit antibody. (C) F_TM-, FcN, and FcN_2C-C proteins were adsorbed directly to microtiter plates and used in an ELISA with the four MAbs indicated in each image. 2F, 47F, and 101F are murine MAbs that recognize nonoverlapping epitopes in hRSV_F (29) Palivizumab (a generous gift of Abbott España) is a humanized MAb in use as a prophylactic treatment for hRSV infections. Escape mutants selected with Palivizumab partially overlap those selected with MAb 47F (30), and both antibodies bind to similar peptides of hRSV_F (31, 32).

We have reported that antibodies present in a human Ig preparation (Respigam, RG) were capable of neutralizing hRSV infectivity, even if depleted of those binding to F_TM-(19). To clarify the specificities of these antibodies, they were tested in a sandwich ELISA with the F_TM-, FcN, and FcN_2C-C proteins. As shown in Fig. 5A, Bottom Left, RG antibodies bound to the three proteins; however, after depletion of the antibodies binding to F_TM-Sepharose (RG/ΔF_TM-; Fig. 5A, Bottom Right), they failed to bind to F_TM- and FcN although continued to react with FcN_2C-C, mimicking the reactivity of rabbit α-Fc/ΔF_TM- antibodies. Furthermore, as reported (19), RG and RG/ΔF_TM- antibodies neutralized efficiently hRSV even if the later antibodies were depleted of α-G antibodies (RG/ΔF_TM-/ΔG) but not if depleted additionally of antibodies that bound to cells infected with Vac/Fc (RG/ΔF_TM-/ΔG/ΔFc) or Vac/FcN_2C-C (RG/ΔF_TM-/ΔG/ΔFcN_2C-C) (Fig. 5B). Thus, a large proportion of the RG neutralizing activity, raised presumably after natural infections with hRSV, is afforded by antibodies able to bind specifically to the prefusion form of hRSV_F and present at the surface of cells infected with either Vac/Fc or Vac/FcN_2C-C.

Remarkably, despite significant structural differences between FcN_2C-C and FcN (i.e., pre- and postfusion forms of hRSV_F; Fig. 4 B and C), both proteins (and F_TM-) bound similarly four α-F MAbs specific to different epitopes, including palivizumab used in the clinic (Fig. 5C).

Discussion

Most α-hRSV_F antibodies described so far in the literature recognize epitopes present in the postfusion form of the F protein. Commonly, reactivity with purified F_TM- or Fc or immunoprecipitation of hRSV-infected cell extracts have been used to assign antibody specificities. It is expected that under all of those conditions the F protein will be folded in the postfusion conformation, such as for the FcN protein of Fig. 4. Even virus preparations may contain a sizeable amount of F in the postfusion form, as shown for PIV5_F (23). It is also known that preparations of purified hRSV contain a large proportion of noninfectious particles in which the F protein could be in the postfusion form. Therefore, it was only after depletion of antibodies that bound to F_TM- that those specific for the prefusion form of hRSV_F could be uncovered. These antibodies account for most of the neutralizing activity found in human Ig preparations (not only Respigam but also another Ig preparation, Flebogamma; ref. 24) and in sera of rabbits inoculated with Vac/Fc. These antibodies, however, were absent from sera of rabbits inoculated with Vac/F_TM-, illustrating the importance of the F protein structure presented to the immune system for the antibody outcome.

Previously we have shown that antibody affinity is crucial for protection of mice from hRSV infection (25). It is likely that a major proportion of the F protein present in the formalin or UV inactivated virus or the purified F protein that failed to induce protection in those studies was in the postfusion structure. Taking into consideration our new results showing that antibodies specific for the prefusion form of hRSV_F account for most of the neutralizing activity of an immune response, it is possible that absence of this form may have contributed to a failure to produce a protective immune response.

The structure of hRSV_F in the postfusion conformation has been solved recently by X-ray crystallography (26, 27). Modeling of the prefusion form and comparison with the actual structure of the postfusion form confirmed the preservation of local structures in both conformations that contained epitopes recognized by certain MAbs. These results are in full agreement with previous electron microscopy studies of immune complexes of F_TM- with MAbs (17) and with the results of Fig. 5C, which showed cross-reactivity of four MAbs with the pre- and postfusion forms of hRSV_F. These results also explain the frequent isolation of MAbs that recognize both conformations and the capacity of MAbs or antibodies such as α-Fc/F_TM- (Fig. S1) to inhibit virus infectivity even before activation of the F protein for fusion (18).

At least a proportion of FcN and FcN_2C-C proteins should be in the prefusion form when expressed at the cell surface as inferred from their capacity to deplete the α-Fc/ΔF_TM- and RG/ΔF_TM- antibodies of their respective neutralizing activities (Fig. S2 and Fig. 5B). However, it is worth pointing out that both proteins were engineered to ablate the two cleavage sites required for activity and, consequently, remain uncleaved. Therefore, subtle structural differences between those two proteins (while inserted in the cell membrane) and the actual prefusion protein may have escaped the antigenicity tests and electron microscopy studies, carried out in the present work. When solubilized, however, FcN apparently refolded into the postfusion form as seen by electron microscopy. In contrast, FcN_2C-C (for which the yield is 10–20 times lower than for FcN) was inhomogeneous but contained a proportion of rounded molecules that resemble the reported prefusion shapes of PIV-5 (21) and hRSV_F proteins (22). FcN_2C-C preparations also contain a small proportion of molecules in which the additional cysteine residues were not disulfide bonded, as seen by SDS/PAGE (Fig. 4 D and E). This heterogeneity was reproducibly found in several preparations of FcN_2C-C and was maintained at similar levels during storage. Thus, additional efforts are needed to obtain homogeneous preparations of prefusion hRSV_F amenable for structural studies. Furthermore, FcN_2C-C (or other alternative constructs) may require further improvements for efficient immunogenicity, because preliminary results indicate that Vac/FcN_2C-C is able to induce neutralizing antibodies in rabbits but at lower titers than Fc, most likely due to the lower expression level (Fig. S3).

The actual mechanism of neutralization by the antibodies specific for the prefusion form of hRSV_F remains to be elucidated. For instance, it is unknown whether the high neutralizing capacity of the α-Fc/ΔF_TM- or RG/ΔF_TM- relies on synergistic effects of antibodies binding simultaneously to several epitopes. In this sense, it would be important to evaluate the neutralizing potency of individual antibodies (e.g., MAbs) specific for the prefusion form of hRSV_F in comparison with antibodies that recognize epitopes shared by the prefusion and postfusion conformations, such as those of Fig. 5C. The availability of stabilized prefusion forms of hRSV_F should facilitate the search of MAbs or other type of molecules that bind specifically to this conformation and interfere with its activation. This approach could be used to develop a new generation of more effective prophylactic antibodies for treatment of infants at high risk of hRSV infections or even extended to other paramyxoviruses for development of novel prophylactic and/or therapeutic agents.

Methods

Preparation and Testing of Polyclonal Antibodies Directed Against Different Forms of hRSV_F.

New Zealand White rabbits were inoculated intramuscularly (i.m.) with the vaccinia virus recombinants indicated in Fig. 1 legend on days 0 and 21 and bled 3 wk later. Serum antibodies were purified by using a protein A-Sepharose column and further processed by affinity chromatography to F_TM- protein covalently linked to Sepharose beads. The unretained fraction (named ΔF_TM-; see SI Methods for antibody nomenclature) was saved, and the bound antibodies were eluted with acidic buffer. The different antibody preparations were tested for ELISA binding to F_TM- and virus neutralization. Depletion of certain specific antibodies was also achieved after incubation with cells infected with either hRSV or vaccinia virus recombinants expressing different forms of the F protein. Human antibodies present in Respigam were processed similarly to rabbit antibodies.

Stabilization of the Prefusion Form of hRSV_F.

Vaccinia virus recombinant expressing full-length F (Vac/Fc) has been described (17). This recombinant was modified by changing the basic residues at the two cleavage sites of hRSV_F to Asparagines as indicated in Fig. 4A to generate Vac/FcN. Additionally, the residues Leu481, Asp489, Ser509, and Asp510 of hRSV_F were substituted by Cysteines to generate Vac/FcN_2C-C. Finally, a His tag was added to the C terminus of FcN and FcN_2C-C for purification purposes. Additional experimental details are provided in SI Methods.