> Home > Programs > Conferences & Meetings > Previous Symposia > Antibodies 1998 > Leonard G. Presta

Leonard G. Presta
Genentech, Inc.
S. San Francisco, CA

Humanized Antibodies for Cancer Therapy

The use of non-human antibodies as therapeutic agents in humans is attended primarily by two problems. When challenged with the presence of non-human antibody, the human immune system mounts a response that can reduce the therapeutic value of the non-human antibody. Therapeutic efficacy may also be reduced due to weak recruitment of effector functions (e.g. antibody-dependent cytotoxicity) that may be necessary for bioactivity of the antibody. There are two categories of engineered antibodies that can replace non-human antibodies of therapeutic utility. A chimeric antibody entails substitution of an entire, intact, variable domain of a human antibody with that from a non-human one. In contrast, a humanized antibody includes only substitution of the six antigen-binding loops (CDRs) of the human antibody with those from a non-human one; exchanging some non-CDR residues may also be required to attain binding of the humanized antibody that is similar to that of the parent non-human antibody.

Several chimeric and humanized antibodies directed to cancer therapy have been developed at Genentech: RituxanÔ (chimeric anti-human CD20) has been approved for low-grade non-Hodgkin’s lymphoma; Phase III trials of an anti-pl85HER2 antibody for breast cancer have been completed; and Phase I/11 trials of an anti-VEGF antibody for solid tumor therapy are underway.

Designing the humanized antibody is not always straightforward, and each of the Genentech humanized antibodies has presented unique problems. The anti- pl85HER2 antibody required inclusion of a very unusual arginine framework residue at position 66 in the light chain; this residue was unusual in that in humans this position is conserved as glycine in kappa light chains and lysine in lambda light chains. Recently, an anti-human Protein C antibody with possible utility in controlling tumor vascularization was humanized. While the parent murine antibody exhibited a 2-fold calcium dependence for binding to Protein C zymogen, the humanized antibody exhibited an increased 5-fold calcium dependence. Investigation of the calcium dependence showed that two framework residues in the antibody light chain were involved in the calcium dependence. For humanization of an anti-human VEGF antibody with possible utility for solid tumor therapy, a monovalent phage display method was attempted in addition to the method used at Genentech for other antibodies. Finally, the sequence of the antigen-binding loops in the parent non-human (and consequently in the humanized) antibody may be problematic. An example is a humanized anti-human IgE antibody in which an aspartic acid-glycine sequence in one of the binding loops undergoes a spontaneous rearrangement that results in loss of binding capability; removing the problematic asp-gly sequence required redesigning the antigen-binding loops to counterbalance the loss of binding.