Because of their high lethality, BoNT have been classified by the Centers for Disease Control and Prevention into category A, which includes substances that could be used as biological weapons. However, because botulinum toxins have been utilized in recent years for treating dystonia, practical use and laboratory handling of them have become relatively commonplace. Therefore, prevention of biohazards caused by them has become increasingly important, as have techniques for adequate emergency treatment. In some countries, vaccination with botulinum toxoid has been implemented and found to be effective in prevention of infection (18, 19). However, inoculation of toxoid vaccines has been limited to individuals at particularly high risk, such as certain military personnel. Serum therapy has been successfully used for post-exposure treatment of botulism, mainly with equine antitoxin immunoglobulin (20). However, equine antitoxin can cause anaphylactic shock in humans (21).
Some disadvantages of using mouse monoclonal antibodies as immunotherapeutic agents for humans have been reported (7, 8). It is naturally essential to limit or avoid possible side effects linked to the murine origin of such antibodies. One of the approaches to achieve this is to replace most of the mouse sequences with their human counterparts. In this study, as a first step towards achieving this, we prepared and characterized mouse-human chimeric antibodies against BoNT/A.
We obtained six clones of mouse mAb that exhibited a neutralizing effect against BoNT/A. Among these clones, three clones (2–2, 4–3 and 9–4) recognized a heavy chain in BoNT/A. The other clones (2–4, 2–5 and B-1) were not detected by immunoblotting analysis (Fig. 1), suggesting that their clones may recognize conformational epitopes. We selected the four clones that exhibited the highest neutralizing activities against BoNT/A and constructed mouse-human chimeric antibodies on templates of these clones. Every chimeric antibody retained the individual properties of its murine counterpart (Table 2), showing that we had succeeded in transferring both antigen-binding and neutralizing activities from mouse mAbs to chimeric antibodies. In competitive ELISA, binding of AC24 and AC25 was partially disrupted by mouse mAbs from 2–5 and 2–4, respectively (Table 3), suggesting that AC24 and AC25 bind to a neighboring, or the same, site. After we had shuffled VH and VL of AC24 and AC25, their binding and neutralizing activities were almost identical and their levels remained the same as those of these clones (AC24 and AC25) before shuffling (Table 4). These results may be attributable to a difference between AC24 and AC25 of only one amino acid in CDR3 of the VL region (Fig. 2). The results of competitive ELISA also reflect their highly homologous sequences in the VL region. Binding and neutralizing activities of AC9424, which consists of VH of AC94 and VL of AC24, were almost identical with those of AC24 and AC94. These results indicate that we succeeded in transferring both antigen-binding and neutralizing activities from chimeric antibodies to shuffling antibodies. In AC2494, which consists of VH of AC24 and VL of AC94, the binding activity to BoNT/A was almost unchanged, but the neutralizing activity had surprisingly increased to more than tenfold those of AC24 and AC94 (Table 4). Because AC2494 (1.25 to 2.5 μg/mouse) perfectly neutralizes 10 LD50/mouse of BoNT/A, it is probable that this antibody neutralizes more than 4000 LD50/mg antibody. These data also indicate that the shuffling antibody has strong neutralizing activity that is comparable to that of BabyBIG (3000 LD50/mg antibody), which was developed as an orphan drug consisting of human-derived botulism antitoxin antibodies (22). Previous studies have reported a correlation between binding effect and neutralizing activity against bacterial toxin (23, 24). It is probable that the neutralizing activity of an antibody against toxin correlates closely with antigen-binding activity. However, in spite of the increased neutralizing activity of AC2494, its antigen-binding activity was almost identical to those of AC24 and AC94. This may be attributable to a difference between VH and VL domains in the original clones. VH and VL domains are packed together and the hypervariable loops on each domain contribute to binding antigen (25). The relative importance of VH and VL domains in creating antigen-binding sites is under much discussion. Many kinds of antigen-specific single VH domains can be isolated using a phage-displayed library (26, 27). Although there has been limited isolation of VL domains with binding activity, VL domains also exhibit antigen-specific binding activity (28). Among particular antibodies, the relative importance of VH and VL domains may vary. In our chimeric antibodies, it is not clear whether their binding activity to BoNT/A depends on VH or VL domains. However, the combination of VH and VL from AC24 and AC94, respectively, may vigorously restrict structural changes of BoNT/A for binding to surface nerve-cell receptors and translocation across the cell membrane. It is not easy to prepare hybridomas that produce high titer neutralizing monoclonal antibodies against BoNT/A. VH and VL domains have a restricted structural spectrum of antigens that they can recognize. There would be very few VH or VL domain repertoires that could perfectly neutralize BoNT/A. In this study, we succeeded in constructing a high-potency neutralizing chimeric antibody, AC2494, in vitro using DNA shuffling of VH and VL domains from another clone. The combination of VH and VL domains, as in AC2494, might not exist in natural antibodies produced by B cells. Researchers have previously designed many shuffling antibodies with strong antigen affinity from immunoglobulin libraries using phage display systems (29, 30). By shuffling 20 nonbinding clones against the hapten nitrophenyl phosphonamidate, 240 clones of antigen binding clones were generated (29). Human shuffling antibody in which the light chains binding to TNF-α were paired with a repertoire of heavy chains displayed on phage was found to have a binding affinity similar to a mouse mAb (30). These data indicate that it is possible to build more neutralizing recombinant antibodies by selecting and shuffling VH and VL domains from a variety of repertoires against BoNT/A.
When researchers tested chimeric antibodies in humans, they found that humanized rodent antibodies were less immunogenic and that this increased their half-life and improved their efficacy (31). Because of its neutralizing efficiency, our shuffling recombinant chimeric antibody (AC2494) seems to be a good candidate for passive human immunotherapy. However, as part of this antibody is still derived from mice, reshaping of a completely human antibody would ensure safer usage.