Whilst it is possible to analyse the interaction of antibodies with effector functions in in-vitro systems it is still very difficult to understand why some antibodies are effective in-vivo and others are not. Unfortunately, in many circumstances attempts to correlate the human in-vitro and in-vivo data are neither practical nor ethical. As a result the information gained about a set of antibodies, such as the CAMPATH-1 antibodies, is often incomplete or contradictory. For example, the original rat IgM antibody CAMPATH-1M (clone YTH66.9) [21] was exceptionally good at lysing cells with autologous complement in-vitro and found use in the removal of lymphocytes from bone marrow prior to allografting [23]. However despite being very efficient in activating complement in-vivo, the antibody proved very poor in lysing lymphoid leukaemia cells [24],[25]. Similarly a rat IgG2a antibody with similar specificity (Clone YTH34.5) was good at fixing complement in-vivo but failed to clear significant numbers of cells [24],[25]. However, a class-switch variant of YTH34.5 from IgG2a to rat IgG2b (CAMPATH-1G)[32] proved very effective in the in-vivo treatment of lymphoid leukaemia [24],[25]. The in-vitro correlate is that rat IgG2b seems to be better than rat IgM or IgG2a at interacting with Fc receptors and triggering ADCC [9],[32], and that complement activation alone cannot account for in-vivo effectiveness. This does not help in our understanding of what facet of ADCC, or which Fc receptors on which cells, or what other effector mechanisms, including complement might contribute to cell killing by antibody in-vivo [4],[33],[34],[35],[53],[54]. In fact it is not even certain that in-vitro assays such as ADCC represent any in-vivo mechanism for cell-killing.
Animal models might help in our understanding of the requirements for cell killing through antibody providing that the data is interpreted cautiously. For example it has been observed that some (but not all) monoclonal antibodies to some mouse cell surface antigens are able to effectively deplete those cell populations in-vivo. The depletion appears to be dependant upon both the antibody specificity and on the antibody isotype and therefore this closely resembles the human situation. Similarly the mouse possesses a complement system and a series of Fc receptors expressed on homologous populations of effector cells. To a certain degree these effector systems and immunoglobulin molecules interact with and share conserved structural features with the human systems [53],[54]. That the conservation is not complete is supported by comparisons of the known sequences and by the functional differences which account for phenomena such as homologous restriction of complement activation [6] and for the isotype specificity of binding to Fc gamma receptors [53]. The important correlation is that different immunoglobulin isotypes interact with different effector mechanisms in a given species [34],[35],[53],[54]. For example mouse IgG3 binds very well to human Fc gamma RI but only poorly to the mouse equivalent whereas mouse IgG2b binds poorly to human Fc gamma RI and well to the mouse equivalent [53].
By observing which effector mechanisms operate in-vivo in the mouse it should be possible to derive a set of rules applicable to most species. In a recent series of experiments a matched set of recombinant antibodies to the mouse CD8 antigen was constructed [55]. All four human IgG subclasses as well as rat IgG2b but not other human isotypes seemed to work effectively in depleting cells in-vivo [55]. The next step was to introduce mutations into the antibodies which would interfere with their abilities to activate the murine effector systems. Thus an aglycosyl mutant of the IgG1 antibody was no longer able to deplete cells even though it coated cells in-vivo and remained for a reasonable period of time. As discussed above aglycosyl IgG is unable to interact effectively with several effector systems [34],[35],[49],[50]. Further work in this model using mutations which only interfere with one effector system should help resolve which functions are critical.