Antibody humanization by framework shuffling

We report here the humanization of a mouse monoclonal antibody (mAb B233) using a new technique which we call framework shuffling. rnAb B233 was raised against the human receptor tyrosine kinase EphA2 which is selectively up-regulated in many cancer cell lines and as such constitutes an attractive target for cancer therapy. The six CDRs of B233 were fused in-frame to pools of corresponding individual human frameworks. These human frameworks encompassed all known heavy and light (K) chain human germline genes. The resulting Fab combinatorial libraries were then screened for binding to the antigen. A two-step selection process, in which the light and heavy chains of the parental mAb were successively humanized, resulted in the identification of several humanized variants that retained binding to EphA2. More precisely, after conversion to human IgG1, the dissociation constants of three select fully humanized variants ranged from 3 to 48 nM. This brings the best framework-shuffled, humanized binder within 5-fold of the avidity of parental mAb B233. Importantly, these humanized IgGs also possessed biochemical activities similar to those of parental mAb B233 as judged by induction of EphA2 phosphorylation. Thus, without requiring any rational design or structural information, this new humanization approach allows to rapidly identify various human framework combinations able to support the structural feature(s) of the CDRs which are essential for binding and functional activity. (c) 2005 Elsevier Inc. All rights reserved.