Antibody Humanization

 The VH and VL CDRs comprise the antigen binding site and contain the majority of amino acids that make direct contact with antigen. Jones et al. hypothesized that the CDRs from a murine antibody could be ‘grafted’ on to human frameworks to make a so called ‘humanized’ antibody. Grafting the CDRs from the V_H domain of a murine antibody to an hapten antigen on to human frameworks resulted in an antibody that bound the hapten with comparable affinity when combined with the murine light chain variable domain. Such CDR grafting has been applied to an anti-lymphocyte mAb and to an antibody to respiratory syncytial virus (RSV), but is not always successful due to the critical role that framework residues may play in both supporting the proper conformation of the CDRs and contributing antigen contacting amino acids.
With the determination of a large number of antibody X-ray crystallographic structures, many now take a more directed approach to humanization. After the murine V_H and V_L genes have been sequenced, human frameworks are selected that are the most homologous to the murine V genes. This is straightforward given the large number of antibody sequence databases, including those of Kabat (http://www.bioinf.org.uk/abs/seqtest.html) and IMGT (http://imgt.cines.fr/). Modeling the human frameworks and mouse CDRs on to a homologous antibody structure allows the identification of murine residues that may either contact antigen or directly or indirectly influence the conformation of the CDRs. Finally, amino acids in the human frameworks that are ‘rare’ or ‘unusual’ are replaced with residues more typical of those positions. Approaches such as this tend to retain more murine residues in the framework regions than with CDR grafting and have been used to humanize antibodies to the interleukin receptor and interferon-gamma. A number of variations to the above approach have been developed. These include replacing murine CDR residues that are outside the antigen binding loops with human residues, as done by Presta and colleagues, the use of human germline V genes as the acceptor frameworks and grafting the murine CDRs into human frameworks with the most homologous CDRs, as opposed to most homologous V genes or frameworks (superhumanization). Alternatively, library approaches, as described below, can be used to humanize murine antibodies.

 In these approaches, one of the murine V domains, for example the V_H domain, is paired with a library of human light chains and a chimeric antibody containing murine V_H and human V_L is selected. The murine V_H is then replaced with a library of human V_Hs paired to the new human V_L and a fully human antibody is selected. With any of these methodologies, it can be anticipated that the binding constants of the humanized antibodies may be less than those of the
parental antibodies. If so, additional constructs can be generated and evaluated or library approaches can be used to return affinity to that of the parental antibody .
More than half of the antibody therapeutics licensed by the FDA are humanized antibodies. These include such ‘blockbuster’ antibodies as Herceptin (trastuzmab) for the treatment of breast cancer, Avastin (bavacizumab) for treatment of colon cancer and Synagis (palivizumab) for prevention of respiratory syncytial virus (RSV) infection. Humanized antibodies are clearly less immunogenic than murine antibodies.

They also appear to be less immunogenic in some instances than chimeric antibodies. Humanized antibodies are not completely free of immunogenicity and anti-humanized antibody responses (HAHA) can be detected in some patients. Such responses may be related to the
number of non-human amino acids retained in the humanized antibodies, in addition to the dose, the immunocompetence of the individual and the specific target of the antibody.