1. Introduction
Respiratory syncytial virus (RSV) is a significant human pathogen but, despite decades of effort, no licensed vaccines exist. RSV infections can result in severe respiratory disease in the very young, the elderly, and immunocompromised populations. This virus is a common cause of severe acute lower respiratory track disease in infants and young children worldwide [
1]. Infections of this population in the US frequently result in hospitalization and in developing countries, the infections cause significant mortality [
1,
2]. In the elderly, the importance of this virus rivals the impact of influenza infections [
3,
4,
5,
6]. It is estimated that the virus results in 11,000 to 17,000 elderly deaths per year in the US and ten times that number of RSV associated hospitalizations [
7]. The world population over the age of 60 is forecast to reach 2.1 billion, more than 20% of the population, by 2050 (World Population Aging Report 2015, United Nations). Thus, RSV infections will result in a greatly increased global public health burden in the next few decades. Mortality due to RSV infection in stem cell transplant patients is estimated to be between 6–80% [
8,
9]. Furthermore, RSV infections also result in significant morbidity in normal adult populations [
10].
Development of an RSV vaccine has been attempted since the late 1960s, without success. One major factor contributing to the failures was the lack of appreciation for the role that the conformation of the RSV F protein plays in the stimulation of protective antibody responses. Like many viral fusion proteins, the RSV F protein is folded into a metastable, pre-fusion conformation which, upon fusion activation, refolds into a structurally very different post-fusion conformation [
11,
12,
13,
14,
15]. Recently, the pre-fusion form of the F protein has been shown to be most effective at inducing optimally neutralizing antibodies (NAbs) [
15,
16]. However, because of the instability of the pre-fusion conformation of the F protein, most vaccine candidates, until recently, have only contained the post-fusion form of F protein [
17]. Indeed, recent clinical trials using the post-fusion form of the F protein have failed [
18,
19].
We have developed novel virus-like particle (VLP) vaccine candidates for RSV [
20,
21,
22,
23,
24]. VLPs are ideal platforms for vaccines targeting many pathogens. In contrast to soluble proteins, VLPs robustly stimulate immune responses without the addition of adjuvant [
25]. Because the production of VLPs does not require viral replication, multiple antigens and different conformational forms of antigens can be assembled into VLPs, in contrast to attenuated viruses, which must remain infectious. VLPs are also safer as vaccines for many populations, such as the very young or the elderly with compromised health, compared to infectious, attenuated, or vector viruses, since they do not contain a genome and do not produce a spreading infection.
McLellan et al. have identified mutations in the RSV F protein (DS-Cav1 mutant) that stabilize the pre-fusion conformation [
16]. We have constructed VLPs containing either this stabilized pre-fusion RSV F protein or a stabilized post-fusion F protein together with the RSV G protein and have established the superiority of the pre-fusion F containing VLPs [
22,
23,
24,
26] over post-F VLPs in inducing neutralizing antibodies (NAbs) in both mice and cotton rats. However, three reports [
27,
28,
29] indicated that soluble DS-Cav1 pre-F is still somewhat unstable and converts to the post-F form upon storage. The reported instability of DS-Cav1 F could negatively impact immune responses to the F protein, so we characterized immune responses to four alternative mutation stabilized pre-fusion F proteins. We examined their expression and assembly into VLPs, the stability of the pre-fusion conformation in VLPs, the reactivity of VLP associated pre-F proteins with anti-F monoclonal antibodies, and their induction of neutralizing antibodies following the immunization of mice. Results were compared with VLPs containing the DS-Cav1 pre-F protein. In addition, we quantified the blocking of the binding of prototype monoclonal antibodies by antibodies induced following immunization with each VLP using two different soluble pre-fusion F proteins as targets. Our results indicate that VLPs expressing different versions of stabilized pre-fusion RSV F proteins can differentially impact immunogenicity.
2. Materials and Methods
2.1. Cells, Plasmids, and Viruses
ELL-O, Vero cells, and Hep2 cells were obtained from the American Type Culture Collection and grown in DMEM (Invitrogen. Waltham, MA, USA) supplemented with penicillin, streptomycin (Invitrogen), and 5% (Vero cells) or 10% fetal bovine serum (ELL-O, Hep2 cells) (Invitrogen). Expi293F cells, obtained from ThermoFisher/Invitrogen, were grown in Expi293 media (ThermoFisher/Gibco/Invitrogen Waltham, MA, USA). RSV, A2 strain, was obtained from Dr. Robert Finberg. Virus stocks were prepared from infected Hep2 cells as previously described [
20].
VLPs containing the RSV F and G protein ectodomains (from RSV stain A2) were assembled, as chimera proteins, with the Newcastle disease virus (NDV) core proteins NP and M, as previously described [
20,
24]. The construction, expression, and incorporation of the chimera protein NDVHN/RSVG (H/G) into VLPs have been previously described [
21]. The construction, expression, and incorporation into VLPs of the stabilized pre-fusion DS-Cav1 F/F protein to generate VLP-H/G+DS-Cav1 F/F (abbreviated DS-Cav1 VLPs), and the stabilized post-fusion F protein to create VLP-H/G+post-F/F (abbreviated post-F VLPs), have been previously described [
24]. Chimera proteins containing alternative versions of the pre-fusion F protein were constructed by introducing mutations into the wild-type F/F chimera. PR-DM F/F and PR-TM F/F contained mutations N67I, S215P, or N67I, S215P, and D486N, respectively. SC-DM F/F and SC-TM F/F both had deletions of the p27 sequence, including the two cleavage sites combined with the insertion of a linker sequence GSGSGRS, as diagramed in
Figure 1. In addition, SC-DM F/F and SC-TM F/F had two (N67I, S215P) or three (N67I, S215P, D486N) amino acid substitutions, respectively.
The constructions of genes encoding the soluble pre-F protein and the soluble post-F protein have been previously described [
24,
26].
2.2. Antibodies
RSV F monoclonal antibody clone 131-2A (Millipore MAB8599) was used in RSV plaque assays. Murine monoclonal antibodies mAb1112 and mAb1243 (generous gifts of Dr. J. Beeler), and human mAb D25, mAb AM14, and mAb motavizumab (generous gifts of Dr. J. McLellan), were used to verify F protein conformation in ELISA analysis of VLPs and soluble F proteins, and for antibody blocking experiments. Palivizumab used for antibody blocking experiments was the generous gift of Dr. Jorge Blanco. The anti-RSV F protein HR2 antibody and anti-NDV F-tail antibody used for Western Blots are polyclonal antibodies specific to the HR2 domain of the RSV F protein or the cytoplasmic tail of NDV F protein [
20]. The anti-RSV G protein antibody is a polyclonal antibody raised against a peptide containing G protein amino acids 180–198 (ThermoFisher Waltham, MA, USA). Secondary antibodies against goat, mouse, and rabbit IgG were purchased from Sigma. A secondary antibody against human IgG was purchased from Southern Biotech.
2.3. VLP Preparation, Purification, and Characterization
The conformation of the F protein in the VLP preparations was verified by reactivity to mAbs. The characterization of purified preparations of Pre-F/F VLPs and Post-F/F VLPs has been previously published [
24,
26]. For preparations of VLPs to be used as immunogens (abbreviated as DS-Cav1 VLPs, post-F VLPs, PR-DM VLPs, PR-TM VLPs, SC-DM VLPs, SC-TM VLPs), ELL-0 cells growing in T-150 flasks were transfected with cDNAs encoding the NDV M protein, the NDV NP, the chimera protein H/G, and one of the five Pre-F/F proteins or the Post-F/F protein, as previously described [
20]. At 24 h post-transfection, heparin (Sigma) was added to the cells at a final concentration of 10 /mL to inhibit the rebinding of released VLPs to cells. At 72, 96, and 120 h post-transfection, cell supernatants were collected and VLPs purified by sequential pelleting and sucrose gradient fractionation, as previously described [
30]. Briefly, cell debris from the supernatant was removed by centrifugation at 5000 rpm (Sorvall GSA SLA-1500 rotor), and VLPs in the supernatant were pelleted by centrifugation in a Type 19 Rotor (Beckman) at 18,000 rpm for 12 h. The resulting pellet was resuspended in TNE buffer (25 mM Tris-HCl, pH 7.4, 150 mM NaCl, and 5 mM EDTA), dounce homogenized, and layered on top of a discontinuous sucrose gradient (2 mL 65% sucrose and 4 mL 20% sucrose). The gradients were centrifuged in an SW 28 rotor (Beckman Brea, CA, USA) at 24,000 rpm for 6 h and the fluffy layer at the 20–65% sucrose interface, containing the VLPs, was collected, mixed with two volumes of 80% sucrose, placed on top of a 1 mL layer of 80% sucrose in a SW41 Beckman centrifuge tube, and then over layered with 3.5 mL of 50% sucrose and 2 mL of 10% sucrose. The gradients were centrifuged to equilibrium for 18 h at 38,000 rpm. The VLPs, all of which floated up into the sucrose to the same density, were collected and concentrated by centrifugation in an SW50.1 rotor for 16 h at 38,000 rpm. All sucrose solutions were w/v and dissolved in TNE buffer and all centrifugations were conducted at 4 °C.
The characterization of purified preparations of all VLPs was completed as previously described for Pre-F VLPs and Post-F VLPs [
24,
26]. The conformation of the F protein in the VLP preparations was verified by reactivity to mAbs (as in Figure 3 and previously described [
24,
26]). Protein concentrations of VLP associated F proteins were calculated from a standard curve generated with a parallel western blot of a purified soluble F protein of a known concentration.
2.4. Preparation of Soluble F Proteins
Expi293F cells were transfected with cDNAs encoding the soluble DS-Cav1 pre-F protein, the soluble SC-TM pre-F protein, and the soluble post-F protein. At six days post transfection, total cell supernatants were collected and cell debris removed by centrifugation. Soluble polypeptides were then purified on columns using the His tag and then the strep tag, as previously described [
24,
26]. Our soluble DS-Cav1 pre-F protein and soluble SC-TM pre-F protein efficiently bind to pre-fusion specific mAbs AM14 and D25. The soluble post-F does not bind AM14 or D25, but does bind motavizumab, a site II antibody. Validation of these soluble proteins is described in Blanco, et al. [
26].
2.5. Detection of Cell Surface Protein by Surface Biotinylation
ELL-0 monolayers were grown in 35 mm plates and transfected with cDNAs encoding the F/F proteins or F/F and H/G proteins. After 48 h, the monolayers were washed three times with PBS-CM (PBS with 0.1 mM CaCl2 and 1 mM MgCl2). PBS-CM containing 0.5 mg/mL sulfo-NHS-SS-biotin (Pierce Biotechnology, Waltham, MA, USA) was added and cells were incubated for 40 minutes at 4 °C. Unbound biotin was absorbed with 2 mL DMEM containing fetal calf serum (10%) and cells were washed three times with PBS and lysed with RSB lysis buffer (0.01 M Tris-HCl [pH 7.4], 0.01 M NaCl, 1.5 mM MgCl2) containing 1% Triton X-100, 0.5% sodium deoxycholate, 2.5 mg of N-ethyl maleimide per mL, and 0.2 mg of DNase per mL. Lysates were incubated for 1 h at room temperature or overnight at 4 °C with neutravidin-agarose (Pierce), containing 0.3% SDS, that had been washed with PBS containing 0.5% tween-20 and 5 mg/mL BSA and then with PBS containing 0.5% Tween-20 and 1 mg/mL BSA. Precipitates containing biotinylated proteins were recovered by centrifugation, washed three times with PBS containing 0.5% Tween-20 and 0.4% SDS, resuspended in gel sample buffer (125 mM Tris-HCl, pH 6.8, 2% SDS and 10% glycerol) with 0.7M β-mercaptoethanol, and resolved by polyacrylamide gel electrophoresis. F proteins in the precipitate were detected by Western analysis using an anti-NDV F tail antibody.
2.6. Measures of Relative Binding of Mab to Purified VLPs
VLPs containing equivalent amounts of F protein (determined by Western blots) were added to microtiter wells and incubated for 2–4 h at room temperature. Different dilutions of different mAb were added to the wells, incubated for 2 h, and removed, and the wells were washed with PBS. The mAbs were then removed, and the plate was washed in PBS and incubated with goat anti-human IgG coupled to HRP for 2 h at room temperature. Bound HRP was detected using TMB (3,3’5,5’-tetramethylbenzidin, ThermoFisher34028, Waltham, MA, USA) and the reaction was stopped with 2N sulfuric acid. Color was read in the SpectraMax Plus Plate Reader (Molecular Devices) using SoftMax Pro software. Results are expressed as optical density (OD).
2.7. Determination of Stability of Pre-Fusion F Conformation
For determination of the stability of pre-fusion F conformation in VLPs, VLPs with equivalent amounts of F protein were incubated at different temperatures, different pHs, or different salt concentrations for one hour. The VLPs were then bound to microtiter wells overnight at 4 °C. The wells were incubated with PBS-2% BSA and then incubated with mAb D25 for one hour, and the binding of mAb was detected using anti-human IgG coupled to HRP. Bound HRP was detected as described above.
2.8. Determination of Total Anti-F Protein IgG in Sera
For the determination of anti-pre-F protein or post-F protein IgG antibody levels, wells of microtiter plates (ThermoFisher/Costar, Waltham, MA, USA) were coated with either purified soluble DS-Cav1 F protein or soluble post-fusion F protein (30 ng/well) and incubated overnight at 4 °C, before being blocked with 2% BSA for 16 h. Different dilutions of sera, in PBS-2% BSA and 0.05% Tween, were added to each well and incubated for 2 h at room temperature. Wells were then washed with PBS, incubated with sheep anti-mouse antibody coupled to HRP (Sigma A5906, St. Louis, MO, USA), and incubated for 1.5 h at room temperature. Bound HRP was detected using TMB (3,3’5,5’-tetramethylbenzidin, ThermoFisher34028) and the reaction was stopped with 2N sulfuric acid. Color was read in the SpectraMax Plus Plate Reader (Molecular Devices, San Jose, CA, USA) using SoftMax Pro software. Amounts of anti-pre-F or anti-post-F IgG (ng/mL) in each dilution were calculated using a standard curve generated in parallel using defined amounts of purified murine IgG.
2.9. RSV Plaque Assays, Antibody Neutralization, and Antibody Blocking
RSV was grown in Hep2 cells, and RSV plaque assays were accomplished on Vero cells as previously described [
24,
26]. Antibody neutralization assays in a plaque reduction assay have been previously described [
24]. Neutralization titer was defined as the reciprocal of the dilution of serum that reduced the virus titer by 50%.
To measure the ability of polyclonal sera to block the binding of mAbs, different dilutions of sera were diluted in PBS-1% BSA, and then incubated for one hour at room temperature in wells of Ni-coated microtiter plates (Pierce/ThermoFisher) containing pre-bound 50 ng soluble DS-Cav1 pre-F protein or soluble SC-TM pre-F protein. Ni-coated plates were used in order to bind the soluble pre-F proteins via the histidine tag at the carboxyl terminus of the protein and thus orient the protein in the well with the apex of the molecule projecting upwards, as in virus particles. After removal of the serum, the wells were incubated with 200 ng/mL of purified mAb diluted in PBS-1% BSA for 10 min at room temperature. The mAb was then removed, and the plate washed in PBS and incubated with goat anti-human IgG coupled to HRP. After incubation for one hour at room temperature, the bound HRP was detected as in ELISA assays. The total anti-pre-F IgG in the different serum dilutions used for mAb blocking was determined using a standard curve of purified murine IgG (Southern Biotech. Pittsburg, PA, USA) in order to measure the ng of the serum anti-pre-F antibody that blocked binding of the mAb.
2.10. Animals, Animal Immunization, and RSV Challenge
Mice, four-week-old BALB/c, from Taconic laboratories, were housed (groups of five) under pathogen-free conditions in micro isolator cages at the University of Massachusetts Medical Center animal quarters. Protocols requiring open cages were accomplished in biosafety cabinets. BALB/c mice, in groups of five animals, were immunized by intramuscular (IM) inoculation of VLPs containing 7 μg F protein in 0.05 mL of TNE (50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 5 mM EDTA) containing 10% sucrose. Boosts contained 3 µg of VLP F protein. For infections with RSV, the animals were lightly anesthetized with isoflurane and then infected by intranasal (IN) inoculation of 50 μL of virus (1 × 106 pfu). All animal procedures and infections were performed in accordance with the University of Massachusetts Medical School IACUC approved protocols; IACUC docket # A1982-17, approved 9-13-2017 to 9-12-2020.
2.11. Statistical Analysis
Statistical analyses (student t test and one-way ANOVA) of data were accomplished using Graph Pad Prism 7 software.
4. Discussion
Following the groundbreaking studies of McLellan, et al, [
16] who identified mutations in the RSV F protein (DS-Cav1 mutant protein) that stabilized the pre-fusion conformation, numerous laboratories and companies have reported the generation of alternative stabilized pre-fusion F proteins [
27,
34,
35,
36,
37,
38,
39]. There have been only a few reports of systematic comparisons, other than total neutralizing antibody titers, of any differences between alternative versions of stabilized pre-fusion F proteins with respect to the properties of the protective antibodies they induce. Because the reported instability of the pre-fusion conformation of the soluble DS-Cav1 pre-fusion F protein could negatively impact the immune response to the protein, we explored the properties of alternative pre-fusion RSV F proteins. Krarup, et al. [
27] have described a number of different mutations of the RSV F protein reported to stabilize the RSV pre-fusion F protein. For our study, four of these mutants were selected for the characterization of protein expression, the efficiency of assembly into VLPs, the stability of the pre-fusion conformation, the mAb reactivity of VLP associated F protein, and the induction of neutralizing antibody responses in mice, comparing the results with the VLP associated DS-Cav1 pre-F protein. In addition, the ability to block the binding of prototype monoclonal antibodies to protein targets by serum antibodies induced by each VLP was quantified. Taken together, the results are consistent with the conclusion that not all mutation stabilized RSV pre-F proteins have the same conformation or induce the same antibody responses. They do not induce similar levels of neutralizing antibodies or induce serum antibodies with similar antibody specificities.
A significant difference between the pre-F proteins was their levels of expression. The PR and SC mutants were expressed, on cell surfaces, at significantly higher levels than the DS-Cav1 F mutant and the post-F mutant. This finding suggested that the synthesis, folding, or intracellular transport of the PR and SC mutant proteins are more efficient than that of the DS-Cav1 F protein. This difference likely accounts for the different ratios of F and H/G and F and NP in the different VLPs.
The reactivity of the different VLPs to the pre-fusion specific anti-F protein monoclonal antibodies was surprisingly different. The PR-TM and SC-TM VLPs bound both mAbs D25 and AM14 at higher levels than DS-Cav1, indicating differences from DS-Cav1 VLPs. These differences may be due to altered accessibility of the mAb binding sites, by different affinities of the mAb to the different F proteins in VLPs, or differences in the conformation of the VLP associated F protein pre-fusion epitopes. By contrast, PR-DM bound both antibodies quite poorly and at much lower levels than DS-Cav1. These results suggest that the PR-DM mutant protein may be predominantly in a post-fusion conformation, or more likely in a conformation intermediate between the pre-F and post-F proteins. Monoclonal antibodies to sites in common with both pre- and post-fusion F, motavizamab, a site II antibody, and mAb1112 (site I), bound to all five pre-F VLPs at similar levels. However, there were some differences in the binding of the mAb 1243, a site IV antibody, again suggesting F protein conformational differences between the VLPs.
There are three reports of the instability of soluble DS-Cav1 pre-F conformation [
27,
28,
29] and, indeed, we have observed the loss of reactivity of this protein to mAb D25 upon storage. However, this protein, assembled into VLPs, was stable during incubation at high temperatures, high and low salt concentrations, high and low pH, and multiple cycles of freeze thaw, or upon prolonged storage. The pre-fusion conformation of the other four VLP associated pre-F proteins was also stable. Possibly, anchoring of the proteins in VLP membranes helps to stabilize the pre-fusion conformation. Additionally, the fusion of the ectodomains of these proteins to the NDV TM and CT domains and the inclusion of the foldon sequence at the carboxyl terminus of the F protein ectodomains may serve to stabilize the pre-fusion F protein conformation.
Measures of neutralizing antibodies (NAbs) in sera showed that the PR-TM and PR-DM VLPs stimulated levels quite similar to those stimulated by the DS-Cav1 VLPs. Interestingly, the PR-DM VLPs were as effective as the DS-Cav1 and PR-TM VLPs in terms of the induction of NAbs in spite of the finding that the PR-DM VLPs bound pre-fusion specific mAbs D25 and AM 14 very poorly. We have previously reported that Post-F VLPs after both a prime and boost in mice stimulated about 1.5 to two-fold lower neutralization titers than DS-Cav1 VLPs [
24]. Thus, the titers after PR-DM VLP immunization may reflect a mix of pre-F and post-F content in these VLPs. Alternatively, the PR-DM may be in a conformation intermediate between pre- and post-F, but a conformation that stimulates neutralizing antibodies. By contrast, SC-TM VLPs stimulated NAbs titers three-fold higher than DS-Cav1 VLPs, indicating that this version of the pre-fusion F protein more effectively stimulated NAbs in mice, a result consistent with the increased binding of mAb AM14 and D25 to these VLPs. SC-DM VLPs stimulated NAb levels twice those of DS-Cav1, consistent with the binding of D25 mAb to this VLP. This VLP may stimulate other pre-fusion specific antibodies not tested here.
There were no significant differences in the levels of total anti-pre-F or post-F-binding IgGs stimulated by the five different VLPs. This result suggests that different NAbs titers may be due to the different populations of specific antibodies in each serum. With the goal of defining potential differences in the populations of antibodies induced by the different VLPs, we quantified the amounts of anti-pre-F binding IgG required to block the binding of a given amount of monoclonal antibody to a pre-fusion target protein. Our results showed that the five-different pre-F VLPs induced quite different amounts of antibodies that blocked the binding of D25, AM14, or palivizumab to the target F protein.
Complicating this analysis was the finding that the measured concentration of antibody that blocked mAb binding varied with the target F protein used. For example, the measured ng/mL of D25-blocking antibodies and AM14-blocking antibodies induced by DS-Cav1 VLPs was quite different when using soluble DS-Cav1 as the target compared to the values obtained using the soluble SC-TM protein target. These results further support the idea that alternative pre-F proteins induce different populations of anti-F antibodies.
There are potentially several reasons for the ability of sera to block the binding of an mAb to a target protein. The polyclonal sera may have antibodies that bind directly to the epitope recognized by the mAb and thus they will directly block the binding of that mAb. However, mAb blocking may also involve relative affinities of the antibodies to the specific mAb binding site. It is likely that polyclonal antibodies with a lower avidity will not block binding as effectively as antibodies that have undergone affinity maturation. Indeed, at least for mAb D25, polyclonal antibodies from four and seven weeks post-boost in general blocked mAb binding better than antibodies at two weeks post-boost (
Figure S3, Supplementary Materials). Differences could also relate to the affinities of the mAb to the target. For example, motavizumab is reported to have a higher affinity for site II than palivizumab [
40]. This differential affinity could account for the observation that approximately 700 ng/mL of DS-Cav1 VLP sera was required to block the binding of motavizumab (
Figure 7, panel C) to the DS-Cav1 target, while only 350 ng/mL of the same sera was required to block palivizumab binding (
Figure 8, panel C). It is also possible that polyclonal antibodies in sera will be directed, not to the specific epitope recognized by the mAb, but to regions of the molecule in the vicinity of the epitope. Binding of these antibodies to off-site targets may block mAb binding to its site by masking the epitope, a concept described by Mousa, et al., for antibodies to site II [
41]. Results of competition of any of the sera with the mAbs could be due to any or all of these possible mechanisms.
Based on the combined results, particularly of mAb binding to VLPs, neutralization titers of sera in mice, and competition for binding of mAbs by sera induced by the five pre-F VLPs, the best antigen for inclusion in a vaccine is likely the SC pre-fusion F proteins, particularly SC-TM. The ultimate selection will depend upon the results of protection from RSV challenge. These studies are not informative in mice since even a single RSV infection results in the complete protection of mice from RSV challenge due, at least in part, to the limited permissiveness of mice to RSV replication. Thus, as expected, there was no detectable RSV in the lungs of any of the immunized mice four days after RSV challenge at day 147. Challenge studies, as well as assays for lung pathology after challenge, are better accomplished in cotton rats, which are quite permissive to RSV. Indeed, these studies, which are ongoing, clearly suggest significant differences in the protection provided by immunization with the different pre-F VLPs (in preparation).