Eliminating a Region of Respiratory Syncytial Virus Attachment Protein Allows Induction of Protective Immunity without Vaccine-enhanced Lung Eosinophilia

RSV and rVVs were grown on HEp-2 cells and titered as previously described (8). Frameshift mutants were isolated by selecting escape variants in vitro with anti-G monoclonal antibody (63G) (14). All stocks were free of mycoplasma contamination as determined by DNA hybridization (Genprobe Inc., San Diego, CA).

pGEM4 derived plasmids encoding the G proteins of frameshift mutants R63/1/2/3 and R63/2/4/8 have been described previously (14). The plasmid encoding the double frameshift G protein mutant R63/2/4/1 was derived from R63/2/4/8; in this case, an extra adenosine was inserted after nucleotide 648 by PCR-based site-directed mutagenesis as described (15). The G protein inserts of the above plasmids were isolated by digestion with HindIII and EcoRI, blunt-ended with Klenow polymerase, and ligated to SmaI linearized pSC11 vector. CV-1 cells were infected with the WR strain of VV and immediately transfected with the pSC11-derived plasmids. Progeny virus was plaque purified three times on HuTK-124B cells in the presence of 15 μg/ml of bromodeoxyuridine to select for thymidine kinase–negative recombinants. The rVVG27 (expressing amino acids 124–203) and rVVTer (empty vector) vaccinia viruses were a gift of Dr. M. Trudel (Centre de Recherche en Virologie, Institut Armand-Frappier, Québec, Canada) and have been described previously (16). rVVG contains the insert for the whole G protein and rVVβgal contains the LacZ gene inserted into the vaccinia vector as a negative control in all experiments (17, 18).

Anesthetized 8–10-wk-old female BALB/c mice (Harlan Olac, Bicester, UK) were scarified with 2 × 10⁶ PFU recombinant vaccinia at the base of the tail. After 2 wk, mice were intranasally inoculated with 1.5 × 10⁶ PFU/mouse of Long and RSV mutants or 3 × 10⁶ PFU/mouse of RSV A2 strain.

7 d after RSV infection, bronchoalveolar lavage (BAL) fluid was collected as previously described (19). In brief, individual mice were terminally anesthetized with pentobarbitone and bled via the femoral artery. Lungs were perfused six times with 1 ml of lignocaine in Eagle's media (Sigma Chemical Co., Poole, UK). 200 μl of BAL fluid was cytocentrifuged onto glass slides and stained with Giemsa's reagent for cytological analysis. Eosinophils were counted by flow cytometry as the proportion of the granulocytes compared with total cells and confirmed with microscopic examination of cytospin slides to distinguish between polymorphonuclear cells and eosinophils.

Mice were scarified with recombinant vaccinias 14 d before challenge with whole RSV. 4 d after challenge with RSV, whole lungs were disrupted using glass homogenizers (Jencons, Leighton Buzzard, UK) in 1.3 ml RPMI (Sigma Chemical Co.) supplemented with 2 mM glutamine (GIBCO BRL, Paisley, Scotland), 100 U/ml penicillin, and 100 μg/ml streptomycin on ice. Homogenates were clarified at 10,000 g for 1 min. 50 μl of supernatant and twofold serial dilutions thereof were titrated on HEp-2 monolayers in 96-well plates and plaques were assayed as previously described (8). The theoretical limit of detection for this assay was 5 PFU/lung.

Kruskal-Wallis tests were used to test for effects between groups and Mann-Whitney U tests were used to perform comparisons between the experimental and control groups. Data analysis was performed using SPSS statistical software.

A series of RSV escape mutants were generated using a monoclonal antibody against the G protein (14). These mutants contain frameshift mutations generating G proteins with truncations and/or alterations in the COOH terminus of the protein (Fig. 1). Mice were scarified with rVVG or rVVβgal (control construct) and 2 wk were later intranasally challenged with the parental Long strain or one of the different mutant viruses. 7 d after infection, BAL fluid was collected and the percentage of eosinophils in the BAL was assessed. As in primary infection, mice scarified with rVVβgal and challenged with either Long or mutant viruses showed no eosinophilia. However, mice scarified with rVVG and challenged with whole RSV showed marked pulmonary eosinophilia except for one mutant. Mutant 63/1/2/3 failed to induce eosinophilia. Mice scarified with rVVG and challenged with mutant 63/1/2/3 generated a low level of eosinophilia similar to mice scarified with rVVβgal (P ≥0.86) (Fig. 2). The low level of eosinophilia observed in mice scarified with rVVG after intranasal challenge with 63/1/2/3 was significantly different from mice scarified with rVVG followed by either Long, 63/2/4/1, or 63/2/ 4/8 (P ≤0.01, 0.01, and 0.01, respectively). The differences between mutant 63/1/2/3 and 63/2/4/8 lie between amino acids 193 and 205 (Fig. 1). From these data, the portion of the G protein responsible for eosinophil induction can be localized to this region.

A possible explanation for this finding is that the large deletion in 63/1/2/3 altered viral infectivity and subsequent eosinophilia generation. Viral lung titers on days 4 and 7 after infection showed no differences between the parental Long strain virus and viral mutants (data not shown). Despite the large alteration of the COOH terminus of the G protein, all mutant viruses infected the lungs efficiently and replicated as well as Long strain virus and all were cleared by day 7.

To complement the studies described above and ensure that the eosinophilia was due only to sensitization to the attachment protein G, rVVs expressing mutant G proteins were used to sensitize mice. Mice were scarified with recombinant vaccinia and challenged intranasally with wild-type RSV (A2 strain) after 14 d. The G protein of the A2 strain is 95% homologous to Long strain and is identical within region 193– 205. Mirroring the first set of studies, no eosinophilia was observed in the BAL of mice scarified with rVV63/1/2/3 (Fig. 3,A). The level of eosinophilia in rVV63/1/2/3-vaccinated mice was not significantly different from rVVβgal-primed mice (P ≥0.556) and was significantly reduced compared with that seen in mice primed with rVVG, rVV63/ 2/4/1, or rVV63/2/4/8 (P ≤0.02, 0.02, and 0.05, respectively). To confirm that the lack of eosinophilia observed in mutant rVV63/1/2/3 mice was not due to a loss of stability or secondary structures, mice were scarified with recombinant vaccinia (rVVG27) expressing 80 amino acids of the G protein (124–203) (Fig. 1). After intranasal RSV challenge, marked eosinophilia was observed, similar to that of mice primed with rVVG (P ≥0.29) and significantly greater than in control mice primed with rVVTer (the control construct) (P ≤0.03) (Fig. 3 B). This confirms that the eosinophilic region lies within this 80-amino acid portion of the G protein. Together, these data using mutant viruses and rVV with inserted mutant G proteins or portions of the G protein, suggest that amino acids 193–203 contain the critical region responsible for lung eosinophilia.

The level of eosinophilia in the lungs is only one measure of vaccine-enhanced illness; another is weight loss. Each mouse was weighed daily after intranasal challenge with RSV. Sensitization with the G protein from mutant virus 63/1/2/3 did not significantly enhance weight loss during RSV challenge compared with control vaccination with rVVβgal, whereas rVVG or any of the other mutants did enhance weight loss (Fig. 4,A). To confirm that this region contains an epitope that is responsible for vaccine-enhanced weight loss, recombinant vaccinia expressing amino acids 124–203 of the G protein (rVVG27) was used to prime mice followed by intranasal challenge with live RSV (Fig. 4 B). Mice sensitized with this portion of the G protein suffered marked weight loss equal to that caused by wild-type G. These data illustrate that the portion of the G protein responsible for pulmonary eosinophilia is also responsible for vaccine-enhanced weight loss.

To examine whether rVV63/1/2/3 could still provide protection even though it does not induce eosinophilia, mice were scarified with recombinant vaccinias and challenged intranasally with wild-type RSV. All mice vaccinated with either of the frameshift mutants (rVV63/2/4/8 or rVV63/1/2/3) or with wild-type G (rVVG) were protected from RSV infection at day 4 (Fig. 5) and no virus was recovered from the lungs of any mice at day 7 (data not shown). These results confirm that it is possible to protect mice from subsequent RSV infection without inducing lung eosinophilia.

This is the first study to show that a single region within the G protein is responsible for enhanced illness and lung eosinophilia and that in its absence it is possible to induce protective immunity without pathology. The critical importance of this region for lung eosinophilia was evident both at the priming (rVV63/1/2/3) and challenge (63/1/ 2/3) phases of the immune response. Mapping eosinophilia induction with the mutant G proteins and rVVG27 locates the critical region to be amino acids 193–203. There are two possible explanations for this finding. First, the mutation in 63/1/2/3 could have eliminated all CD4⁺ T cell recognition of the G protein. Since G-induced RSV disease is mainly a T cell–mediated, immunopathological process, a reduction in the CD4⁺ T cell response would lead to less cellular infiltration into the lungs, less eosinophilia, no RSV-specific antibody response, and less weight loss. A second possibility is that the mutation introduced into 63/ 1/2/3 does not eliminate the entire CD4⁺ response but rather a vigorous, eosinophil-inducing component thereof. We believe this to be the case since priming with rVV63/ 1/2/3 generates an appreciable antibody response and a pulmonary lymphocytic infiltrate similar to that seen in wild-type G-sensitized mice, but with a significant decrease in the ratio of CD4⁺:CD8⁺ cells (data not shown). This suggests that a strong CD4⁺ T cell response is abolished by the alteration of residues 193–203 and that this response is essential for induction of lung eosinophilia. Recently, using overlapping peptides, a T cell epitope that induces a Th2 response has been mapped to amino acids 184–198 (19a). This peptide includes six residues from the critical region mapped in our study (KPGKKT). Using a predictive algorithm for class II MHC binding peptides (strong and weak anchor positions: 1, 4, 9: I, K or R, K, respectively) (20), we have identified two possible I-E^d epitopes, both of which would be disrupted in the 63/1/2/3 mutant. These potential epitopes are located at amino acids 185-193 (ICKRIPNKK) and 189–197 (IPNKKPGKK) (bold type indicates anchor and subanchor residues).

In other viral systems, single epitopes have been shown to be responsible for Th2-like CD4⁺ responses. Proteins from mouse hepatitis virus (JHM; reference 21), hepatitis B virus (22), and retroviruses (23) have been shown to generate Th2 responses and have been mapped using peptides and T cell clones and lines. Unlike our study, in none of these systems has it been possible to examine the effects of altering these epitopes upon the host immune response and induction of pathology. Interestingly, the critical epitope in JHM virus is also I-E^d–restricted (ILNKPRQKR) and contains at least four identical or similar residues to the proposed critical motifs of RSV G. It is possible that these homologous epitopes have similar binding affinities for MHC class II molecules, thus skewing towards a Th2-type response (24).

In this and in previous studies, induction of eosinophilia is indicative of a Th2-type response (10, 11, 25). The recruitment of eosinophils into the alveolar space could be due to IL-5 production (26) or the induction of chemokines such as eotaxin (27) or RANTES (regulated on activation, T cell expressed and secreted; reference 28). In our study, this same region is also responsible for the increased weight loss that is observed upon challenge. Weight loss, however, is not a specific indicator of Th2-type, eosinophilic pathology since mice lose weight late in a primary infection and may show disease enhancement due to sensitization to other RSV proteins (10). Hussell et al. have shown that IL-12 treatment of mice primed with rVVG induces a Th1-type response and diminishes eosinophilia but weight loss is unaffected or in some cases enhanced (29). Our data indicate that region 193–203 of the G protein not only induces eosinophilia (probably via the Th2 pathway), but also induces an immune response (such as TNF-α; reference 30) that is responsible for increased weight loss.

These findings provide a framework which may allow the development of an RSV vaccine without vaccine-enhanced illness. In the mouse model of RSV, vaccine-enhanced illness presents two possible pathways for eosinophil induction: first, a CD4⁺ response to eosinophilic epitopes within the primary structure of the G protein such as described here, or second, the elimination of the CD8⁺ response that controls the Th2 response (31). We have recently shown that it is possible to generate eosinophilia to the F protein of RSV when CD8⁺ cells are deleted. Normally, priming with the F protein of RSV generates a Th1-type response and virus-specific CTLs that are an abundant source of IFN-γ. When CD8⁺ T cells are deleted, the response is switched to a Th2-type response that generates lung eosinophilia. It appears that in the F response, CD8⁺ cells are responsible for controlling the Th2-type response, most likely through the production of IFN-γ (13). Similarly, the failure of the formalin-inactivated RSV vaccine trials could have been due to the lack of CD8⁺ T cell induction since soluble antigen is processed and presented via the MHC class II pathway. Both aspects of the immune response need to be considered for the development of future RSV vaccines.

Our results suggest that it is possible to induce a protective immune response with the G protein without inducing a detrimental eosinophilic response. Since the G protein does not induce a CTL response in BALB/c mice (32), eosinophil induction could in part be due to the lack of a CD8⁺ response. In spite of this, we have shown that it is possible to induce protective immunity with the G protein if the eosinophilogenic portion has been altered. Vaccines need to induce a CD4⁺ response to promote B cell production of protective antibody, but IFN-γ production by CD8⁺ cells may also be essential to protect against augmentation of illness. However, induction of a strong and unopposed CD4⁺ response can also be detrimental. Sensitization with the M2 protein alone leads to enhanced illness, and transfer of isolated CD4⁺ T cells causes augmented disease as well (10, 33). An ideal vaccine should strike a balance between inducing CD4⁺ and CD8⁺ T cell responses. Vaccination that stimulates either a CD4⁺ or CD8⁺ T cell response alone might lead to pathological consequences (34). Whether these tenets for vaccine development will hold true in an outbred human population remains to be tested.

We would like to thank Dr. M. Trudel for providing vaccinia recombinants rVVG27 and rVVTer. We are also grateful to Dr. B.A. Askonas for critical review of this manuscript, Dr. E.T. Bergman for statistical analysis of this data, and Teresa Delgado for technical assistance with the generation of vaccinia recombinants.

This work was supported by the Wellcome Trust and J.A. Melero was supported by grant PM96-0025 from Dirección General de Enseñanza Superior, MEC.