> Home > Programs > Conferences & Meetings > Previous Symposia > Antibodies 1998 > Carlos F. Barbas III

Carlos F. Barbas III, Christoph Rader, David Segal, Roger Beerli, Birgit Dreier
The Scripps Research Institute, La Jolla, CA

Targeting Cancer: Proteins and Genes

Proteins…
Since the development of the hybridoma approach, a large number of rodent monoclonal antibodies (mAbs) with specificity for antigens of therapeutic interest have been generated and characterized. The fact that rodent antibodies are highly immonogenic in humans, however, severely limits their clinical applications, especially when repeated administration is required for therapy. Two routes in antibody engineering have been taken to overcome the immunogenicity of mAbs, either the humanization of rodent mAbs or the direct generation of human mAbs. The latter route has recently gained importance with the development of new human mAb sources from immune, naive, and synthetic human antibody libraries displayed on phage as well as from transgenic mice. Twenty years of mAb generation by the classical hybridoma technology, however, has yielded a number of promising pharmaceutical candidates and their humanization compares well to the de novo generation and characterization of human mAb for accessing clinical applications in the coming years. Ideally, antibody humanization must not diminish specificity and affinity towards the antigen while immunogenicity must be completely eliminated. It has become apparent that the accomplishment of both aims is usually a time-consuming and costly undertaking with even the most current humanization strategies. Here we report the development of a new humanization strategy that combines rational design with combinatorial selections using phage display. We demonstrate that this approach provides a rapid route to antibody humanization and demonstrate its application to the humanization of mouse mAb LM609 that is directed to the human integrin avb3. We chose LM609 as a model antibody for our humanization strategy because of its clinical potential. Recent findings by Brooks et al. in a chorioallantoic membrane model and a SCID mouse/human skin chimeric model have shown that the intravenous administration of LM609 is able to reduce growth and metastasis of human tumors due to the inhibition of angiogenesis induced by the tumors. These findings have suggested that integrin avb3 may be a target and LM609 a tool for cancer therapy.

…and Genes
From the simplest of organisms to the most complex, transcriptional regulation is achieved primarily by proteins that bind nucleic acids. The advent of genomic sequencing and the availability of the complete sequences of several genomes provides new opportunities to study biology and to develop therapeutic strategies through specific modulation of the transcription of target genes. Of the DNA-binding motifs which have been manipulated by design or selection, the TFIIIA-related Cys2His2 zinc finger proteins have demonstrated the greatest potential for manipulation into general and specific transcription factors. Each Cys2His2 zinc finger domain consists of approximately 30 amino acids and typically binds 3 base pairs of double-stranded DNA sequence. Given a 3 base pair recognition sequence per zif domain, a collection of 64 domains recognizing all 64 ‘codons’ would allow any sequence to be targeted. Using phage display we have made significant progress towards selecting this alphabet of zinc finger domains.

For specific gene control, single zinc finger domains are insufficient. Specific delivery of a DNA binding protein to a single site within a genome as complex as that found in humans, 3.5 billion bp, requires an address of at least 16bp. Statistically assuming random base distribution, a unique 16 bp sequence will occur only once in 416 or 4.3 billion nucleotides, roughly the same size of a human genome (3.5 x 109 bp). An 18 bp address would be specific within 68 billion base pairs of sequence. The 18 bp address could be specified by a protein containing six zinc fingers if the periodicity of the protein domains could match that of the DNA over this extended sequence. An address of this length would be sufficient to uniquely specify any locus within all known genomes. While natural proteins containing long polydactyl arrays of zinc finger domains have been inferred from sequence, no zinc finger proteins have been demonstrated to bind such a long contiguous DNA sequence. Structural studies of the five-finger GLI-DNA complex and biochemical studies of the nine-finger protein TFIIIA have demonstrated that DNA binding in these polydactyl proteins is dominated by the interactions of a select few fingers. Sequence specific binding of more than three contiguous zinc finger domains within polydactyl proteins has yet to be observed. We have now demonstrated that zinc finger proteins containing 6 selected zinc finger domains and binding 18bp of DNA can be constructed. Expression of these proteins as fusions to activation or repression domains allows transcription to be specifically up or down modulated within human cells. Polydactyl zinc finger proteins should be broadly applicable as genome-specific transcriptional switches in gene therapy strategies and as novel tools for the study of cancer.