INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

Respiratory syncytial virus (RSV) remains the most important cause of serious viral respiratory illness in infants and young children. Although previous contact with the virus provides partial immune protection from subsequent reinfections, the development of an effective vaccine has proven extremely difficult (17). Recently, promising attenuated live virus vaccine candidates have been identified during studies with experimental primates and phase I clinical trials (16, 22, 25). These vaccine candidates must strike a fine balance between attenuation and immunogenicity and be suitable for use in both seronegative children over 6 months of age and very young infants with maternally derived RSV-specific antibodies (15, 16). Subunit vaccine preparations, while not as immunogenic as live virus, are currently being evaluated as a means of boosting the immunity of elderly populations (18) and children suffering from cystic fibrosis (31). However, because of the association between vaccination with nonreplicating virus antigens and enhancement of clinical disease, subunit formulations are not suitable for young seronegative infants (26).

An important factor in the assessment of virus vaccine candidates is their ability to elicit neutralizing antibodies. This is especially true for viruses such as RSV, since neutralizing antibodies have been shown to play a major role in resistance to disease in humans (28) as well as in protection from infection in experimental animals (14, 33-35, 39, 40). Both RSV infection and RSV vaccines elicit neutralizing and nonneutralizing antibodies reactive with envelope glycoproteins. It is unclear what distinguishes these classes of antibodies and how they are elicited in humans. Both issues are important for vaccine design.

We have approached these issues by cloning a set of human antibodies elicited to the RSV envelope by natural infection. Analysis of human antibody responses has been greatly hindered in the past by the difficulties of obtaining human monoclonal antibodies (MAbs) representative of these responses. Phage library technology provides a possible strategy for solving this problem. The technique does involve random recombination of antibody heavy and light chains, which was initially thought to exclude the study of antibody responses; however, detailed investigations of a number of antibody responses to pathogens and autoantigens by the library approach have suggested that, notwithstanding this limitation, the cloned antibodies reflect broad aspects of natural responses (1, 7, 9, 19, 32, 41). In particular, epitope specificities found in the polyclonal serum are usually rescued in the corresponding library. The reasons are not fully understood but likely include some reforming of in vivo heavy- light chain combinations and domination of the binding specificity by one chain so that the partner is less important.

In this report, we have examined human antibody responses, both neutralizing and nonneutralizing, in RSV infection by using the library approach. We have restricted our analysis to human antibodies that develop against the F glycoprotein of the virus during the course of natural infection. The F glycoprotein is well defined as a major antigenic target of RSV-neutralizing antibodies. However, several lines of evidence suggest that the epitopes eliciting this protective response are conformational and highly sensitive to perturbations of the tertiary structure of the protein. For example, earlier vaccine trials with formalin-inactivated virus or affinity-purified F protein induced an imbalanced, predominantly nonneutralizing antibody response (27, 31). Moreover, many attempts to elicit a protective antibody response by immunizing animals with large synthetic peptides containing a portion of the F glycoprotein sequence failed, presumably because conformational determinants present in the native protein were not faithfully reproduced in the peptides (5, 23, 36).

These observations should be interpreted in light of the cellular processing of the F protein and its assembly into virion spikes. First produced as an inactive glycosylated precursor, F₀, the protein is subsequently modified by the addition of N-linked carbohydrate and then assembled into homo-oligomers in the rough endoplasmic reticulum. F₀ is subsequently cleaved endoproteolytically in the Golgi apparatus, yielding two subunits (F₁ and F₂) that are covalently linked to each other by disulfide bonds. It is believed that four F₁---F₂ molecules interact via the F₁ subunit to produce the viral envelope tetramer spike. Thus, the F glycoprotein is present during infection in a number of immature forms as well as the mature form found on virions and the surface of infected cells. Any or all of these forms may stimulate the humoral immune system. It should also be noted that the mature form of the F glycoprotein, once assembled, could be dissociated to generate further forms of the protein; for the sake of simplicity, these are included here under the general grouping of immature forms of F.

In this study, we report that a number of human MAbs representing major anti-F protein specificities react strongly with immature forms of the F glycoprotein but only very weakly with the mature form expressed on RSV-infected cells. These antibodies are nonneutralizing and were presumably elicited by immature forms of the F glycoprotein. In contrast, one human MAb reacts almost exclusively with the mature form of the F glycoprotein on RSV-infected cells and virions and essentially not at all with the immature forms. It is potently neutralizing and was presumably elicited by the mature envelope on virions or infected cells. We suggest that this envelope glycoprotein dichotomy may be an important strategy for evading host immunity, since infection tends to produce, through lysis of infected cells, relatively large amounts of the immature envelope, which will elicit a strong humoral response (29, 30). A conformationally altered mature envelope molecule on the virus will be less susceptible to this antibody response. An important corollary of this hypothesis is that when recombinant or affinity-purified envelope molecules bear a closer structural similarity to the immature than to the mature envelope, as is often the case, these molecules may not be particularly good vaccines. In other words, this vaccine strategy may be handicapped because viruses have already evolved a mechanism for avoiding the humoral response to immunodominant epitopes presented on immature envelope molecules.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Antibody reagents. (i) Phage library construction, phage selection, and purification of soluble recombinant Fab fragments. Four antibody Fab ( chain of immunoglobulin G1 [IgG1]) libraries expressed on the surface of filamentous phage were prepared from bone marrow RNA of human immunodeficiency virus type 1 (HIV-1)-seropositive donors previously infected with RSV as described elsewhere (8). Phage bearing RSV-specific Fab fragments were selected from the libraries by a panning procedure (8) with recombinant FG protein (kindly provided by M. Wathen, Upjohn) or affinity-purified F protein (kindly supplied by E. Walsh) bound to enzyme-linked immunosorbent assay (ELISA) wells. Phage taken from the fourth or fifth panning round were converted to a soluble Fab expression system, and the clones were analyzed for RSV reactivity in an ELISA (41). The light-chain and heavy-chain variable-region sequences of RSV-reactive antibody clones were determined by dideoxy sequencing as described elsewhere (41). Selected clones were grown and purified by use of a polyclonal goat anti-human IgG Fab fragment (Pierce) bound to a protein G matrix (Pharmacia).

(ii) Polyclonal antibodies. Serum samples were obtained from two healthy volunteers and pooled and from an HIV-1-seropositive donor. Polyclonal IgG was purified from the sera by use of a protein A matrix (Pharmacia).

Characterization of RSV-specific antibodies. Purified recombinant Fab fragments and polyclonal IgG were measured for their ability to bind to RSV antigens by ELISA, surface plasmon resonance, and flow cytometry analyses.

(i) Surface plasmon resonance. The kinetic constants for the binding of recombinant Fab fragments to affinity-purified F protein were determined by use of a BIAcore instrument (Pharmacia). F protein was coupled to the (carboxymethyl)dextran matrix of a CM5 sensor chip (Biosensor) by use of N-hydroxysuccinimide and N-ethyl-N'-[(dimethylamino)propyl]-carbodiimide hydrochloride (20). Following coupling, unreacted N-hydroxysuccinimide ester groups were inactivated with 1 M ethanolamine (pH 8.0). Typically, 10,000 resonance units were immobilized. Binding measurements were performed with binding buffer (Pharmacia). The association rate constant (k_on) of each Fab fragment was measured over a range of concentrations (62 to 1,000 nM) at a flow rate of 10 µl/min. The dissociation rate constant (k_off) was determined by increasing the flow rate to 50 µl/min after the association phase. The binding surface was regenerated after each measurement with 10 mM HCl-1 mM NaCl (pH 2.0). Values for k_on and k_off were calculated by use of BIAcore kinetics evaluation software. The equilibrium dissociation constants (K_d) were determined from the rate constants by use of average k_on and k_off values obtained from four independent measurements.

(ii) Flow cytometry. Antibody reactivity with F protein on the surface of RSV-infected cells was examined by flow cytometry. Monolayers of HEp-2 or Vero cells were infected at a multiplicity of infection of 3. At 24 h postinfection, the cells (2 × 10⁶ per sample) were detached from the culture flasks by incubation with 5 mM EDTA in phosphate-buffered saline (PBS), washed twice in PBS, and resuspended in FACS buffer (PBS containing 1% fetal calf serum and 0.02% azide). The resuspended cells were incubated for 30 min on ice with 100 µl of a range of concentrations of recombinant Fab fragments (0.04 to 10 µg/ml) or polyclonal IgG (7.8 to 250 µg/ml) and then washed three times in FACS buffer. Cell surface-bound antibody was detected by incubation for 30 min on ice with a 1:100 dilution of goat anti-human IgG Fab fragment-fluorescein isothiocyanate conjugate (Jackson ImmunoResearch). The cells were washed three times in PBS, resuspended in 1 ml of PBS, and analyzed by use of a FACScan (Becton Dickinson).

(iii) Competitive ELISA. To determine if the recombinant antibodies were able to compete with each other for binding to affinity-purified F protein, one Fab fragment (CM68) was biotinylated with a biotin labeling kit (Boehringer Mannheim Biochemicals) in accordance with the manufacturer's instructions. ELISA wells were coated with 100 ng of F protein, which was then reacted with 10 µg of unlabeled Fab fragment per ml for 1 h at 37°C. Biotin-labeled CM68 was added to a final concentration of 1 µg/ml, and the mixture was incubated for an additional 15 min at 37°C before being washed 10 times with PBS containing 0.05% Tween 20. Antigen-bound CM68 was detected by use of goat anti-biotin-alkaline phosphatase conjugate (Sigma). Binding titers were expressed as a percentage of the signal given by biotinylated CM68 in the absence of competing nonbiotinylated Fab fragment, i.e., {100  [experimental optical density at 490 nm/control optical density at 405 nm (in the absence of competing Fab fragment)]} × 100.

Viruses and cell culture. HEp-2 cells were cultured in RPMI 1640 supplemented with 2 mM L-glutamine and 5% fetal calf serum. Vero cells were cultured in Dulbecco modified Eagle medium supplemented with 5% fetal calf serum. RSV strains A2 and Long were used to infect HEp-2 cells. RSV strain cp-52, a cold-passage mutant of a subgroup B virus that lacks the G and SH glycoproteins (21) (kindly provided by B. Murphy and S. Whitehead), was used to infect Vero cells. Virus titers were determined by a plaque assay as described elsewhere (10).

Plaque reduction assay. Antibody neutralization activity was measured by a plaque reduction assay in the absence of complement with HEp-2 cell cultures and strain A2 (subgroup A virus) (10). The neutralizing antibody titer was calculated as the highest dilution of purified Fab fragment or IgG that reduced plaque numbers by 50%.

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Recovery of RSV-specific recombinant Fab fragments and assay of their neutralizing activity. Three IgG1 antibody Fab libraries, prepared from bone marrow RNA of HIV-1-positive donors possessing high levels of serum antibody neutralizing activities, were displayed on the surface of filamentous phage. The libraries were selected against affinity-purified F protein bound to ELISA wells. Antibody clones taken from the final round of panning were screened in an ELISA for reactivity with F protein. DNA sequence analysis of the antigen-reactive clones identified five novel RSV F protein-specific recombinant Fab fragments (CM2, CM6, CM68, DM2, and DM1) with distinct sequences (Table 1). An additional antibody, Fab fragment RSV19, was generated by panning one of the libraries against recombinant FG glycoprotein, a baculovirus-expressed chimera of the F and G glycoproteins. RSV19, described previously, exhibits high neutralizing activity against a wide range of virus strains in vitro (2, 14). In contrast, none of the antibodies recovered in panning experiments with affinity-purified F protein were able to neutralize RSV in a plaque reduction assay at concentrations of up to 50 µg/ml (data not shown).

Epitope mapping. The ability of the recombinant Fab fragments to compete either with each other or with mouse MAbs for binding to affinity-purified F protein was examined by an inhibition ELISA. Fab fragment CM68 was biotin labeled and used to compete with the remaining unlabeled Fab fragments. As shown in Table 2, with the exception of RSV19, all of the F protein-specific Fab fragments appeared to compete with CM68, indicating that all of the recombinant antibodies, except for RSV19, are directed to a single antigenic site on the F protein. We next used CM68 to compete against mouse MAbs directed to four antigenic regions (A, AB, B, and C) of the F protein thought to be important targets for virus neutralization (3). The results are also presented in Table 2. Two mouse MAbs directed against the B site competed efficiently with the Fab fragments. In addition, a MAb recognizing the C site also consistently inhibited CM68 binding, albeit to a lesser extent than the B site-specific MAbs. Conversely, inhibition was not observed when CM68 was used to compete against MAbs binding to the A or AB antigenic site of the F protein or against RSV19.

The epitope of a potent neutralizing antibody is not present in purified F glycoprotein or in lysates of virus-infected cells. In order to begin to better understand the properties that distinguish neutralizing from nonneutralizing antibodies, we measured the binding constants of each of the F protein-specific Fab fragments by using the BIAcore instrument. The binding kinetics were measured over a range of antibody concentrations against immobilized affinity-purified F protein. Kinetic analyses of the resulting sensograms yielded k_on and k_off values, and K_d values were determined as k_on/k_off (Table 3). K_d values obtained for the nonneutralizing Fab fragments ranged between 3.7 and 17 nM, suggesting that all of these antibodies bound to the affinity-purified antigen relatively tightly. To our surprise, however, Fab RSV19 bound to this antigen only very weakly, with a calculated K_d value of 6 µM. The efficiency with which this antibody neutralized virus in vitro (50% plaque reduction at approximately 0.2 µg/ml or 4 nM), via a mechanism which must proceed via Fab binding to F protein assembled into virion spikes, suggests that its overall binding constant for the F protein in this mature form is considerably higher (see below).

View larger version (18K): [in this window] [in a new window]   FIG. 1.   ELISA reactivity of recombinant Fab fragments RSV19 and CM68 with RSV-infected cell lysates (lysate) and affinity-purified F glycoprotein (APF).

A potent neutralizing antibody binds tightly to F protein on the surface of infected cells, whereas nonneutralizing antibodies bind weakly. We next sought to determine how well the antibodies recognize the F protein as it occurs on the surface of virus-infected cells. Here the protein is likely to be organized into a mature, oligomeric structure, ready for incorporation into the envelope of budding virions. The ability of the recombinant antibodies to bind to RSV-infected HEp-2 cells was measured in a flow cytometry assay. Antibody binding profiles were produced by measuring the mean fluorescence of cell populations incubated with each antibody over a range of concentrations (Fig. 2). The data suggest that the relative affinities of the recombinant antibodies for mature F protein differ dramatically from those calculated for affinity-purified F protein. Whereas neutralizing Fab fragment RSV19 bound weakly to affinity-purified F protein, it bound considerably more tightly to F protein on the infected cell surface than did nonneutralizing Fab fragments (Fig. 2).

View larger version (16K): [in this window] [in a new window]   FIG. 2.   Reactivity of recombinant Fab fragments with F protein on the surface of RSV-infected cells, as measured by flow cytometry. The MF of infected cells relative to that measured under the same conditions for uninfected cells was calculated following incubation with each antibody over a range of concentrations.

Quantitative correlation between antibody binding to viral envelope glycoproteins on the surface of infected cells and neutralizing activity. The observations described above suggest that a key difference between neutralizing and nonneutralizing antibodies may be their relative capacities to recognize a mature envelope protein, such as that encountered on the surfaces of RSV-infected cells and virions. Support for this interpretation was sought by comparing Fab fragment RSV19 and polyclonal immune IgG for neutralizing activity and capacity to bind to the surface of RSV-infected cells. Flow cytometric analyses performed with RSV19 and polyclonal IgG preparations were used to calculate the antibody concentration required to produce 50% maximum binding to the infected cell surface. These values were then compared with the concentration of the same antibody preparation required to produce a 50% reduction of RSV plaque formation in an in vitro neutralization assay. Since the role of complement in clearing RSV infection has been documented previously (13), neutralization assays were performed without the addition of complement. The results are presented in Table 4. RSV19 neutralized virus efficiently (average neutralizing concentration, 0.23 µg/ml). The same antibody produced 50% maximum binding to RSV-infected cells at a similar concentration (0.58 µg/ml). There was a similar relationship between neutralization and binding for one of the polyclonal IgG preparations, and for the other preparation, neutralization was somewhat more efficient.

Relative proportions of antibodies reactive with purified F protein and mature oligomeric F protein in human antibody responses to RSV infection. To obtain an estimate of the relative proportions of human antibodies induced to different envelope presentations of F protein as a result of RSV infection, a purified pooled polyclonal IgG preparation was titrated against affinity-purified F protein in an ELISA and against mature oligomeric F protein on infected cells by flow cytometry. Although saturation for binding to infected cells was not achieved, the data suggest approximately equivalent levels of binding to the two envelope presentations (Fig. 3). Only a portion of the observed binding to the two F glycoprotein presentations corresponded to the same antibody fraction of the polyclonal IgG preparation because excess affinity-purified F protein was able to inhibit only about 20% of the binding of IgG to infected cells (Fig. 4).

View larger version (16K): [in this window] [in a new window]   FIG. 3.   Comparative binding of purified pooled human IgG to different antigenic presentations of F protein. The binding of human IgG (IgGp1) over a range of concentrations to affinity-purified F protein in an ELISA and to F protein as it occurs on the surface of RSV-infected cells was measured (the level of binding to G and SH proteins on the surface of infected cells is much lower than that to F protein; see the text).

View larger version (16K): [in this window] [in a new window]   FIG. 4.   Competition between human polyclonal IgG and affinity-purified F protein for binding to infected cells. Human IgG was incubated with a range of concentrations of affinity-purified F protein. Binding to cells infected with either A2, B1, or cp-52 virus was measured by flow cytometry. Results are expressed as the percentage of IgG bound in the absence of a competitor.

View larger version (10K): [in this window] [in a new window]   FIG. 5.   Competition between human polyclonal IgG and Fab fragment RSV19 for binding to infected cells. IgG binding to RSV-infected cells preincubated with a range of concentrations of RSV19 was determined by flow cytometry. Results are expressed as the percentage of IgG bound in the absence of the competing Fab.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

In this study, we cloned a number of different F protein-specific antibodies by panning phage display libraries against purified and recombinant F antigens. The majority of the antibodies that were recovered were nonneutralizing, although their affinities for the antigens against which they were panned were in the nanomolar range. Using a competition ELISA, we observed that the nonneutralizing epitope region to which the bulk of the recombinant antibodies were directed was highly immunogenic and accounted for about one half of the total serum IgG reactivity against purified F protein (Table 2). These results are consistent with those of other studies (37, 38) in which a large panel of nonneutralizing antibodies was recovered by panning antibody phage libraries prepared from immune donors against a recombinant preparation of the extracellular portion of the F protein. In addition, Beeler and van Wyke Coelingh (3) also found that the ability of mouse MAbs to bind to F protein produced in a vaccinia virus recombinant was not predictive of their neutralizing activity.

In this study, the reason for the lack of neutralizing activity of many of the antibodies was shown to be their low affinity for mature oligomeric F protein present on the surface of RSV-infected cells and presumably also virions. In contrast, an antibody (RSV19 [2]) that exhibited a high level of neutralizing activity for RSV bound with a high affinity to mature oligomeric F protein but did not bind to purified F protein. This antibody was originally obtained by selection of a phage Fab library on an FG fusion protein preparation. It can only be hypothesized that this preparation must contain at least a fraction of the protein in a conformation akin to that of mature oligomeric F protein. Unfortunately, we have been unable to obtain more of this protein to test this hypothesis.

In any case, the results describe antibodies reactive with purified F protein and viral lysates and not with oligomeric F protein and vice versa. The simplest interpretation of the data is that two different immunogens are responsible. One immunogen appears to be immature F protein, as present in RSV-infected cell lysates and affinity-purified F protein prepared from such lysates. It is likely that a similar immature F protein is made available to the immune system following lysis of infected cells during natural infection. The other immunogen appears to be mature oligomeric F protein, as found on the surface of infected cells and virions. Studies with a polyclonal IgG from pooled human sera suggest that the two responses are of approximately equal magnitudes and that the degree of overlap of the two responses is relatively limited. Only 20 to 30% of the IgG response to mature oligomeric F protein could be inhibited by affinity-purified F protein, suggesting either that there is limited sharing of epitopes between mature and immature forms of F protein (up to 20 to 30% overlap of responses to mature and immature forms) or that some immature F protein is expressed on the surface of RSV-infected cells (no overlap).

A modification of the antigenic makeup of the envelope as it is assembled for presentation on the virion surface has potential evolutionary advantages. Immature forms of the envelope will be released from infected cells by cell lysis, possibly in relatively large amounts, during viral infection. This release will stimulate an antibody response that will be neutralizing if the antibodies react with the virion surface. Antigenic modification during envelope spike assembly can reduce the functional efficacy of this antibody response. The effect will be most noticeable if the mature envelope spike has a relatively low immunogenicity, as is the case for HIV-1 (6, 24). For RSV, however, there does appear to be a significant human antibody response to the mature envelope. In any case, in general, this factor may contribute to the limited success of subunit vaccines in eliciting efficient protective neutralizing antibody responses (4, 12, 31, 36), since viruses may have evolved to minimize antibody responses to the immature envelope molecules that are often used as the basis for subunit vaccines.