Protection of mice against Human respiratory syncytial virus by wild-type and aglycosyl mouse-human chimaeric IgG antibodies to subgroup-conserved epitopes on the G glycoprotein

Human respiratory syncytial virus (HRSV) subgroups A and B remain the most important, potentially preventable causes of morbidity and mortality in infants in developed countries (Collins et al., 2001). The viruses infect the majority of babies during their first year of life and approximately 2 % are admitted to hospital, generally with bronchiolitis or pneumonia. Protection against infection correlates with antiviral antibodies measured in cord blood or maternal blood perinatally (Ogilvie et al., 1981; Ward et al., 1983), and infants of mothers with high antibody levels experience their first infection with HRSV at a later age (Glezen et al., 1981). Animal studies indicate that the two major surface glycoproteins of the virus, the fusion protein (F) and the putative attachment protein (G), are the major targets for antibody-mediated immunity, and passively administered antibodies to both F and G are protective against experimental infection in rodent models of HRSV infection (Routledge et al., 1988; Stott et al., 1986; Taylor et al., 1984; Walsh et al., 1984).

The F glycoprotein is responsible for the fusion of viral and host-cell membranes, but may also play a role in virus attachment (Feldman et al., 2000). A relatively high proportion of monoclonal antibodies (mAbs) to F neutralize infectivity in vitro and block the fusion of virus-infected cells into multinucleated syncytia. Protection in animal models is Fc-independent and correlates with fusion inhibition rather than neutralization in vitro (Taylor, 1994). As F is largely conserved, both between virus subgroups A and B and among virus isolates within a subgroup, it is a favourable target for prophylactic antibody development. A fusion-inhibiting murine mAb to a conserved epitope, Palivizumab, has been humanized and reduces hospitalization rates (IMpact-RSV Study Group, 1998) and virus loads (DeVincenzo et al., 2003) when administered passively to at-risk human infants.

In contrast, the role of the G glycoprotein in infection has not been characterized fully. Early studies indicated a role in virus attachment and the protein carries both a heparin-binding site (Feldman et al., 1999) and a fractalkine receptor-binding site (Tripp et al., 2001), both of which may contribute to virus binding. However, recent studies indicate that mutant viruses lacking the G glycoprotein retain infectivity in some cell cultures, although they are attenuated in animal models, suggesting that G has important functions other than cell attachment (Teng et al., 2001).

Unlike the F glycoprotein, G is highly variable (Melero et al., 1997). Variation appears to be driven by immune pressure (Woelk & Holmes, 2001), but it is not clear whether G is under greater pressure to evolve than F or whether its structure is more forgiving of sequence variation, allowing immune escape. The molecule adopts a hairpin-loop structure, anchored into the membrane via its N-terminal signal sequence. Amino acid sequence variation is concentrated largely in the highly glycosylated arms of the hairpin, and epitopes located in these regions tend to be strain specific. The bend in the hairpin is more conserved, forming a distinctive cystine noose overlapped by a putative C3XC chemokine receptor and heparin-binding sites (Doreleijers et al., 1996; Langedijk et al., 1996). Antibody epitopes in this region tend to be subgroup specific or conserved between subgroups.

Using a series of C-terminally truncated mutants, Olmsted et al. (1989) demonstrated that the C-terminal 68 aa and the N-terminal 180 aa of the G glycoprotein, which include the variable arms of the hairpin structure, are not necessary for protection in rodents. Peptides containing aa 173 and 186, the central more-conserved region, induce protective immunity, dependent, at least for subgroup A, upon the integrity of intramolecular disulfide linkages between Cys¹⁷⁶ and Cys¹⁸² (Simard et al., 1997; Trudel et al., 1991). The amino acids between the cysteines vary between the subgroups, but are highly conserved within each subgroup and the protection conferred was subgroup specific. Antibodies to this region are generally low or undetectable in convalescent sera of infants and children (Åkerlind-Stopner et al., 1995; Palomo et al., 2000). Passive supplementation with mAbs of this specificity may augment passive immunity conferred by anti-F mAbs.

Although antibodies recognizing subgroup-specific epitopes on the G glycoprotein confer resistance in vivo (Bastien et al., 1997; Plotnicky-Gilquin et al., 1999; Trudel et al., 1991), these antibodies do not neutralize the virus in vitro. Their mechanism of protection is, therefore, unclear. In studies by Corbeil et al. (1996), the protection afforded by a passively administered subgroup A-specific mAb Fab fragment or intact mAb given to mice depleted of complement by treatment with cobra venom was reduced, suggesting that protection by these antibodies involves the complement system. Here, we sought to test this hypothesis further by isolating and cloning the variable region genes of a subgroup A-specific anti-HRSV G mAb, 1C2, and constructing a pair of plasmids encoding chimaeric mousehuman immunoglobulins containing 1C2 variable regions, one of which carried a mutation in the C_H2 domain rendering the antibody defective in complement activation and FcγR binding. The protective efficacy of mutant and wild-type chimaeric antibodies was compared in mice.

Virus and antibodies.
HRSV strain A2 was kindly provided by Dr A. J. Stott (Institute for Animal Research, Compton, Berkshire, UK) and grown in HeLa cells. High-titre virus stocks for mouse challenge studies were prepared as described previously and infectivity titres were assessed by fluorescence focus assays and expressed as focus-forming units (f.f.u.) ml¹ (Hayes et al., 1994). Anti-HRSV G glycoprotein 1C2 mAb (IgG2a_κ) was produced by conventional hybridoma technology (Morgan et al., 1987) and purified on a protein ASepharose column. Anti-HRSV F glycoprotein mAb 1E3 (West et al., 1994) was used as unpurified mouse ascitic fluid. mAb Campath 1H (Alemtuzumab) is a human IgG1_κ antibody specific for the human CD52 antigen (Hale et al., 1988). mAb YTH 12.5 is a humanized IgG1_λ antibody specific for human CD3 antigen (Routledge et al., 1991).

Cloning mAb 1C2 variable-region genes and expression of chimaeric antibodies.
The 1C2 V_L and V_H genes were isolated from 1C2 hybridoma cells by RT-PCR using total cellular RNA as the template. Oligonucleotide primers designed to anneal to mouse κ and γ2a constant region mRNA (primers EJK194 and EJK193, respectively) were used to prime the reverse transcriptase step (Table 1). The V_H and V_L forward primers (CM195 and CM197, respectively) and the V_H reverse primer (CM196) used for the subsequent PCR step were based on those described by Orlandi et al. (1989). The V_L reverse PCR primer (CM215) was designed specifically for amplification of mouse V_κII genes. The PCR primers incorporated restriction enzyme sites (NcoI and XhoI for V_H, ApaLI and NotI for V_L), which were used to clone the PCR products into the phage antibody expression vector pHEN2 (MRC Centre for Protein Engineering, Cambridge, UK). Phage antibody clones were screened for anti-G protein-binding activity by ELISA, using BSA conjugated to a peptide corresponding to aa 172187 of the HRSV G protein as the capture antigen.

Table 1. Oligonucleotide primers

The V_L and V_H genes from a positive clone (p530) were used to construct one chimaeric mouse V_Lhuman C_L cDNA gene and two chimaeric mouse V_Hhuman C_H cDNA genes, suitable for expression in mammalian cells (Fig. 1). The light-chain gene consisted of a mouse V_H leader sequence (without intron) derived from the vector M13VKPCR1 (Orlandi et al., 1989) linked in frame to the 1C2 (p530) V_κ gene, which was in turn linked in frame to a human C_κ gene (Hieter et al., 1980). A Kozak sequence for optimum initiation of translation (Kozak, 1987) was placed upstream of the start codon. The first chimaeric heavy-chain gene consisted of the same Kozak and leader sequence as used for the light chain, linked in frame to the 1C2 (p530) V_H gene, followed by a cDNA derivative of the human G1m(1, 17) γ1 constant-region gene (Takahashi et al., 1982). The second chimaeric heavy-chain gene had exactly the same structure as the first, except that the γ1 constant region carried the aglycosyl Asn²⁹⁷→Ala substitution mutation (Isaacs et al., 1992) in its C_H2 domain and thus lacked the constant-region N-linked carbohydrate-attachment site.

(10K): Fig. 1. Chimaeric 1C2 light- and heavy-chain cDNA gene structures. The structure of the 1C2 mouse Vκhuman Cκ light-chain gene (a), the 1C2 mouse VHhuman Cγ1 heavy-chain gene (b) and its aglycosyl derivative (c) are shown. The position of the Kozak sequence (K), leader sequence (L) and restriction endonuclease sites that played a role in the assembly of the genes or their insertion into the GS expression vector system are indicated. H, HindIII; A, ApaLI; M, MaeII; E, EcoRI; P, PstI; B,BstEII; B with a hash indicates deletion of BstEII.

Two expression plasmids, one containing the light-chain gene plus the γ1 heavy-chain gene (p534) and the second containing the light-chain gene plus the aglycosyl γ1 heavy-chain gene (p535), were produced using the glutamine synthetase (GS) expression vector system (Lonza Biologics). In each case, the heavy and light genes were ligated into the vectors pEE6.hCMV-BglII and pEE12, respectively (between the HindIII and EcoRI sites of the vectors), and then combined into a single H+L plasmid following the protocol recommended by Lonza Biologics. The expression vectors, linearized by digestion with SalI, were transfected separately into NS0 cells by electroporation. NS0 clones secreting chimaeric antibody were identified by ELISA of culture supernatants using goat anti-human IgG Fc (Sigma-Aldrich) as capture antibody and biotinylated mouse anti-human κ light chain (BD Biosciences Pharmingen) as detector antibody, followed by extravidin peroxidase (Sigma-Aldrich). Antibody-secreting transfectants were recloned in soft agar and cultivated in 24-well plates. The culture supernatants were tested for anti-HRSV G protein-binding activity using HRSV-infected HeLa cell antigen by ELISA. The 1C2 γ1 and aglycosyl γ1 chimaeric antibodies were purified from culture supernatant using protein ASepharose CL-4B (Pharmacia) column chromatography as described by Harlow & Lane (1988). Both heavy and light chains of the purified antibodies were identified by Western blotting, as described by Samson et al. (1986). The possibility that traces of bovine anti-RSV might have co-purified from the fetal calf serum (FCS) used in the NS0 culture medium was tested by immunofluorescent staining of HRSV strain A2-infected HeLa cells using a fluorescein isothiocyanate-conjugated rabbit anti-bovine IgG secondary antibody. No staining was seen with either 1C2 chimaeric antibody preparation or the FCS used in culture of the NS0 cell line, whilst a positive control bovine anti-HRSV serum gave a strong reaction.

Antibody assays.
Antibodies were titrated in infectious focus reduction neutralization assays as described previously (Routledge et al., 1988). When neutralizing activity was being tested in the presence of complement, an equal volume of freshly reconstituted rabbit serum (Sigma-Aldrich) was added to each serum dilution. Relative binding avidities of the 1C2 chimaeric antibodies were analysed in a competitive binding ELISA. Dilutions of each protein A-purified chimaeric antibody were mixed with an equal volume containing 10 µg mouse 1C2 mAb ml¹ diluted in PBS containing 0·05 % Tween 20 and incubated at 37 °C for 90 min in the wells of an ELISA plate coated with HRSV-infected HeLa cell antigen. The humanized anti-human mAb Campath-1H was used as a negative-control competitive antibody. The relative amount of mouse 1C2 mAb binding to the antigen was determined by an additional incubation with 1 : 2000-diluted goat anti-mouse IgG peroxidase conjugate (DakoCytomation) at 37 °C for 1 h, followed by ortho-phenylenediamine substrate solution. Competitive binding activities were expressed as the minimum antibody concentration that gave 50 % reduction in the level of binding activity.

Mouse protection study.
Groups of six female BALB/c mice (6 weeks old) were injected intravenously with 200 µl PBS containing 1 mg purified mAb and challenged 24 h later by intranasal inoculation of 1·4x10⁷ f.f.u. HRSV strain A2 under fluorothane anaesthesia. Control mice were given either PBS or the humanized mAb YTH 12.5. Four days post-inoculation, mice were killed and the lungs were removed and homogenized in RPMI 1640 (Gibco-BRL) supplemented with 10 % FCS. Homogenates were assayed for the presence of HRSV and the data were analysed using the MannWhitney Wilcoxon rank test.

Expression of wild-type and aglycosyl γ1 1C2 chimaeric antibody in NS0 cells
The V_H and V_L genes of mAb 1C2 were isolated by RT-PCR, cloned and assembled with human κ light chain and human γ1 heavy constant-region genes to form a mousehuman chimaera in a pEE vector, p534, carrying a GS selectable marker and under the control of the human cytomegalovirus early promoter. In parallel, a mutant version of the chimaera was generated, p535, in which an Asn²⁹⁷→Ala mutation was introduced into the human γ1 constant region, thus encoding an aglycosyl version of the antibody lacking the C_H2-domain N-linked glycosylation site.

NS0 cells transfected with either p534 or p535 were cloned and the clones were selected for high-level immunoglobulin secretion, as assessed by ELISA. Culture supernatants from high-yielding clones were purified on protein ASepharose and analysed by polyacrylamide gel electrophoresis and Western blotting (Fig. 2). Heavy and light chains of the 1C2 γ1 chimaeric antibody were approximately 50 and 25 kDa, respectively, whilst the heavy chain of the 1C2 aglycosyl γ1 chimaeric antibody was approximately 46 kDa, smaller than that of the wild type as expected, as it lacks a carbohydrate side chain.

(73K): Fig. 2. Comparison of the original mouse 1C2 mAb and the two human chimaeric 1C2 mAbs by Western blotting. Replicate blots were loaded with biotinylated molecular mass markers (M), original mouse 1C2 mAb (Mo), human γ1 chimaeric mAb (γ1) or human aglycosyl γ1 chimaeric mAb (Aγ1) and either notstained with detector antibody (X) or stained with goat anti-mouse peroxidase conjugate (Y) or sheep anti-human IgGbiotin (Z). All three blots were then stained with extravidinperoxidase conjugate.

Specificity and relative avidities of the 1C2 chimaeric antibodies
The anti-HRSV specificity of the NS0-derived chimaeric antibodies was confirmed by ELISA on HRSV-infected HeLa cell lysates and control uninfected cell lysates (data not shown). The relative binding avidities of the two 1C2 chimaeric antibodies and the original mouse 1C2 mAb were compared in a competitive binding ELISA, using HRSV-infected HeLa cell lysate as the capture antigen. As shown in Fig. 3, the two antibodies were approximately equivalent with 12·5 and 25 µg ml¹ of the chimaeric γ1 and aglycosyl γ1 mAbs, respectively, required to give 50 % reduction in the level of binding produced by 10 µg mouse 1C2 mAb ml¹.

(9K): Fig. 3. Competitive inhibition of the binding of mouse 1C2 mAb to HRSV-infected HeLa cell antigen using 1C2 chimaeric γ1 and aglycosyl γ1 mAbs as the competitor antibodies. , Campath 1H mAb; , chimaeric aglycosyl γ1 1C2 mAb; •, chimaeric γ1 1C2 mAb.

In vitro neutralization activity of anti-G mAbs
The mouse 1C2 mAb and its two chimaeric versions were tested for their ability to neutralize infectious HRSV in vitro in an infectious focus reduction assay (Table 2).

Table 2. Minimum mAb concentration (µg ml1) required to give 50 % virus neutralization Numbers represent the antibody concentration at which a 50 % reduction in infectious foci was achieved, read from mean curves generated from titration of antibodies in triplicate.

As expected, Campath 1H, which is not specific for HRSV and was used as a negative control, gave no virus neutralization in either the presence or absence of complement. The neutralization positive-control anti-F mAb 1E3 gave good levels of neutralization in both the presence and absence of complement, achieving a 91±7 and 80±8 % reduction in infectious foci, respectively, and the 50 % reduction end points were very similar. Neither murine 1C2 mAb nor either of its γ1 chimaeric versions was able to produce significant neutralization of HRSV in the absence of complement. However, in the presence of complement, 0·33 µg murine 1C2 mAb ml¹ and 1·56 µg wild-type γ1 chimaeric mAb ml¹ produced a 50 % reduction in infectious foci, although neutralization was incomplete, even at higher mAb concentrations (62±4 and 50±20 % reduction, respectively, for 100 µg antibody ml¹). In contrast, addition of complement produced no effect on the neutralization activity of the aglycosyl γ1 chimaeric mAb.

In vivo immunoprophylaxis of HRSV infection using the 1C2 γ1 and aglycosyl γ1 human chimaeric mAbs
Groups of six mice were injected intravenously with 1 mg murine 1C2 mAb, wild-type γ1 chimaeric mAb or aglycosyl γ1 chimaeric mAb and challenged with infectious HRSV 24 h later by intranasal inoculation. Control mice were treated with either PBS or the humanized mAb YTH 12.5 (specific for the human CD3 antigen) as a human IgG1 isotype control, prior to challenge. Four days post-inoculation, mice were killed and their lungs were macerated and assayed for the presence of infectious HRSV. All three groups of mice that had received either the original murine or the chimaeric antibodies had significantly less HRSV in their lungs than the mice given either PBS or the control CD3 mAb (P<0·005, Fig. 4). The protective effect produced by the original murine 1C2 mAb and the chimaeric γ1 mAb was significantly greater than that produced by the chimaeric aglycosyl γ1 mAb (P<0·005). No virus was recovered from the lungs of these first two groups of mice, whereas a mean of 1·95x10³ f.f.u. HRSV (g tissue)¹ was found in the lungs of mice treated with the aglycoslyl γ1 mAb. These results indicate an important role for the Fc region in the protective effect of the 1C2 anti-G attachment mAb against HRSV infection.

(7K): Fig. 4. Immunoprophylactic effect of mouse and chimaeric 1C2 anti-G mAbs on HRSV infection in BALB/c mice. The graph shows the geometric mean quantity of HRSV recovered from the lungs of groups of six mice 4 days after intranasal challenge with virus. Twenty-four hours prior to challenge, the mice were given a single intravenous dose of the indicated anti-G, control antibody or PBS. , Mouse 1C2 mAb; , chimaeric γ1 1C2 mAb; , chimaeric aglycosyl γ1 1C2 mAb; CD3 mAb; •, PBS.

mAb 1C2 is cross-reactive with all isolates of HRSV subgroup A tested but unreactive with subgroup B (Morgan et al., 1987). Similar antibodies have been shown to bind to the cystine noose region of the G glycoprotein and 1C2 binds strongly to a peptide corresponding to aa 172187 of the HRSV A2 strain, which incorporates all four cysteines of the cystine noose. The antibody is protective against infection with HRSV strain A2 when given prophylactically in rodents and may have potential as a prophylactic agent in human infants.

In common with other antibodies binding to the cystine noose region (Bastien et al., 1997; Plotnicky-Gilquin et al., 1999; Trudel et al., 1991), 1C2 does not neutralize the virus in vitro. However, as G-deficient mutants of HRSV retain infectivity for cell cultures (Teng et al., 2001) and the role of the G glycoprotein in vivo is unclear, it is not possible to conclude that direct interaction of antibody and G glycoprotein does not mediate protection in vivo. Alternatively, the antibody may inhibit virus replication indirectly by targeting Fc-activated defence mechanisms against virus or virus-infected cells.

In an attempt to resolve the mechanism of protection, Corbeil et al. (1996), working with a similar antibody, 18A2B2, found that papain-generated Fab fragments of the antibody failed to protect. Furthermore, whilst treatment of mice with cobra venom, which blocks complement components C3 and C5, partially reduced protection, DBA/2J mice, which are deficient in C5 only, remained protected. They concluded from these experiments that protection was mediated by Fc-dependent pathways and that the early stages of the complement cascade were involved. However, this approach has potential flaws, as Fab fragments, as well as lacking the ability to activate Fc-mediated defences, lack other attributes of intact immunoglobulin molecules. Most notably, the avidity of the antibody is reduced, as is its ability to cross-link antigen, and the half life of Fab in serum is much shorter than that of intact molecules (Brown et al., 1987). The effect of cobra venom on HRSV infection in mice is also poorly characterized and it has not been demonstrated that the attenuation of the protective effect observed was directly attributable to the depletion of complement.

The isolation here of the variable-region genes and the creation and expression of mousehuman antibody chimaeras has provided an alternative approach to this problem. The introduction of an Asn²⁹⁷→Ala mutation in the γ1 heavy chain removes an N-linked carbohydrate attachment site, creating an aglycosyl form of the antibody. Aglycosylation does not affect antibody avidity for antigen (Bolt et al., 1993), but it does cause a subtle conformational change in the Fc region (Morrison et al., 1993) that inhibits complement activation (Duncan & Winter, 1988; Tao & Morrison, 1989). Aglycosylation also eliminates the binding of human IgG1 antibodies to FcγRI and -II, and although it may not prevent binding to FcγRIII, it seems to prevent subsequent activation of the FcγRIII-bearing cell (Morrison et al., 1993). Thus, aglycosylation of IgG1 blocks both its ability to activate the complement pathway and its ability to recruit Fc receptor-bearing cells capable of mediating antibody-dependent cell-mediated cytotoxicity or phagocytosis. However, aglycosylation does not alter the serum half-life of IgG1 significantly (Tao & Morrison, 1989), nor does it greatly affect the overall size of the IgG molecule, which might alter its ability to interfere sterically with antigen function. Although size is also an important factor in the tissue penetration properties of an antibody, it remains unclear whether the aglycosyl mutation affects the release of serum antibody onto the mucosal surface. It has not been possible to compare directly the tissue penetration of the two antibodies in these experiments, as recovery of antibody from the mucosa will be affected both by the degree of inflammation and by the antigen loads, both of which will vary with the ability of the antibody to block virus infection.

Whilst the original mouse 1C2 mAb and the wild-type chimaera reduced virus replication in the lungs of mice inoculated with HRSV to undetectable levels, virus was recovered from the lungs of mice given the aglycosyl chimaera, albeit at lower mean levels than the placebo-treated controls. This clearly confirms the role of Fc functions in the protective efficacy of 1C2. Whether Fc functions via complement, via the recruitment of Fc receptor-bearing effector cells or via release of antibody onto the mucosal surface remains undetermined. Although 1C2 exhibited complement-enhanced neutralization, mAb 18A2B2 did not (Corbeil et al., 1996). As both antibodies are protective in mice, it seems unlikely that protection is mediated via complement-mediated virus-induced lysis and that, if complement is involved, it acts via mechanisms not observable in cell culture.

Despite the significant reduction in its protective capacity, aglycosyl antibody achieved a significant reduction in virus titre compared with placebo-treated controls. This argues that mechanisms of antibody protection are complex and that both Fc- and non-Fc-mediated interactions with the G glycoprotein play a role. An understanding of non-Fc mechanisms depends on elucidating the role of the G glycoprotein in vivo and the aglycosyl antibody may be useful in designing experiments in this area.

Footnotes

^†, G. L. Toms and E. G. Routledge †Present address: Department of Microbiology, Faculty of Medicine, Srinakharinwirot University, Sukhumvit 23, Bangkok 10100, Thailand. ‡Present address: Angel Biotechnology, 44 Colbourne Crescent, Nelson Park Industrial Estate, Cramlington, Northumberland NE23 1WB, UK.