anti-GCN4 scFV

Ribosome display was used to select from a murine library antibody single-chain fragments directed against a helix-destabilized variant (7Pro,14Pro) of the GCN4 leucine zipper peptide. The selection procedure yielded a family of closely related scFVs which deviated by 1-5 mutations from the consensus sequence. Allthough different mutants were present in different clones, the mutated clones showed significantly higher binding affinities than the putative ancestral sequence. It appears likely that the different mutations were selected from errors introduced in vitro by the reverse translation and PCR amplification steps of the ribosome display procedure (J.Hanes et.al.). The best scFV showed a dissociation constant of 4±1 x 10^-11M. The mutations distinguishing this clone from the consensus sequence could not be detected in the original library.

To assess the possible effects of the different mutants and to decide which of the mutations would merit further study, a homology model of the anti-GCN4 scFV was built. By combining the best mutations and further rounds of ribosome display with stringent off-rate selection, the K_D could be further decreased to xx M

Sequence comparison of the different anti-GCN4 clones with their VL and VH consensus sequences: VL-domain, VH-domain, CHIME-display of the anti-GCN4 model showing the location of the mutated residues

Comparison of the anti-GCN4 with the murine germline sequences: VL-subclass: VLl-1, VH-subclass: VH-2

Identification of potential template structures in the PDB for homology modelling: VL-templates, VH-templates, VH CDR H3s of the same length.

CHIME-display of the anti-GCN4 model, the 4D5/ab48 hybrid model, the GCN4-4D5/ab48 hybrid lambda graft model and the GCN4-4D5/ab48 hybrid kappa graft model.

Comparison of the anti-GCN4 model with the experimental structure of the anti-GCN4 scFV-peptide complex

Comparison of the 4D5/ab48 hybrid model with the experimental structure of the hybrid scFv

anti-GCN4 / 4D5 / ab48 Loop Grafts

The anti-GCN4 scFvs preferentially bind monomeric forms of the GCN4 leucin zipper peptide (C.Berger et al.), but cross-react with the wild-type GCN4 leucin-zipper sequence, prevent its dimerization and inhibit binding of GCN4 to DNA. When expressed in the cytoplasm of yeast cells, this scFV fragment inhibits GCN4-regulated gene expression to a moderate but significant extent (A.Wörn et al.). In the reducing environment of the cytoplasm, the disulfide bonds in the VL and VH domains of the fragment cannot be formed. The limited thermodynamic stability of the reduced scFV fragment limited its binding activity and its ability to interfere with GCN4-DNA interaction.

Although in some scFV fragments replacement of the reduced cysteines by a Val-Ala pair yielded a slightly more stable scFV fragment, the Cys-free anti-GCN4 scFV was found to be somewhat less stable than the reduced form of the wild-type scFV. To obtain a disulfide-free anti-GCN4 scFV with sufficent stability and activity for intracellular application, the antigen-binding specificity of the wild-type anti-GCN4 antibody was grafted onto our most stable scFV, a hybrid framework constructed from the VL domain of the anti-EGFR antibody 4D5 and the VH domain derived from of the anti-levan antibody ab48 by the introduction of several stabilizing mutations.

The anti-GCN4 loop graft combines very divergent antibody domains: For the VL domain, the CDRs of a murine VLl-1 domain was grafted onto a framework representing the human VLk-1 consensus sequence (sequence alignment), while for the VH domain, the CDRs of a murine VH-2 domain were grafted to a modified murine VH-4 framework (sequence alignment). An experimental structure existed only for the 4D5 VL domain, all other models were derived by homology modelling. Since the VL/VH dimer interface residues of the murine VLl domain differ significantly from the corresponding residues in a VLk domain, and these difference could potentially affect the relative orientation of the two domains, two grafts were designed: A kappa-graft retaining the kappa interface, which was predicted to be more stable but might potentially interfere with antigen binding, and a lambda-graft retaining the typical murine VLl-1 dimer interface residues of anti-GCN4

Functionality and thermodynamic stability of the anti-GCN4 scFV constructs