We constructed a single-chain Fv antibody library that permits human complementarity-determining region (CDR) gene fragments of any germline to be incorporated combinatorially into the appropriate positions of the variable-region frameworks VH-DP47 and VL-DPL3. A library of 2 x 109 independent transformants was screened against haptens, peptides, carbohydrates, and proteins, and the selected antibody fragments exhibited dissociation constants in the subnanomolar range. The antibody genes in this library were built on a single master framework into which diverse CDRs were allowed to recombine. These CDRs were sampled from in vivo-processed gene sequences, thus potentially optimizing the levels of correctly folded and functional molecules, and resulting in a molecule exhibiting a lower computed immunogenicity compared to naive immunoglobulins. Using the modularized assembly process to incorporate foreign sequences into an immunoglobulin scaffold, it is possible to vary as many as six CDRs at the same time, creating genetic and functional variation in antibody molecules.

Immunoassays for haptens depend on competitive hapten-anti-hapten reactions, and consequently their sensitivities are significantly influenced by the affinities of anti-hapten antibodies. Thus, genetically engineered antibodies, which have much higher affinities than native antibodies, should increase assay sensitivities. Here, we created a mutated single-chain Fv fragment (scFv) against estradiol-17beta (E(2)) that allowed immunoassays with a much improved sensitivity. Two steps of affinity maturation were performed on a "wild-type" scFv (scFv#E4-4) composed of V(H) and V(L) domains from a mouse anti-E(2) antibody (Ab#E4-4). First, we conducted complementarity-determining region (CDR)-targeted mutagenesis by "CDR-shuffling". Gene fragments encoding CDRs H2, H3, L1, and L3, each of which contained random point mutations, were combined by "shuffling" into the gene encoding the scFv#E4-4 scaffold. After phage display and repeated panning, we isolated a mutated scFv clone [scFv#m1-e7; Ile(L29)Val] that had 5-fold higher affinity (K(a) = 2.6 x 10(8) M(-1)) compared to the Ab#E4-4 Fab fragment (Fab#E4-4). Next, the entire V(H) and V(L) of this clone were randomly mutated by error-prone polymerase chain reaction (PCR). From this library, we found an improved clone, scFv#m2-c4 (K(a) = 6.3 x 10(8) M(-1); Lys(H19)Arg, Tyr(H56)Phe, Ser(H84)Pro, Glu(H85)Gly, Gln(L27)Arg, Leu(L36)Met, Ser(L63)Gly, and Ser(L77)Gly). ScFv#m2-c4 had more than 10-fold higher sensitivity (the midpoint of its dose-response curve was 0.56 ng) than Fab#E4-4 (midpoint 9.0 ng/assay) in a competitive E(2) radioimmunoassay, and even higher sensitivity [midpoint 21 pg/assay, and a limit of detection of 0.47 pg (1.7 fmol)/assay] in a competitive enzyme-linked immunosorbent assay. Cross-reactivity with selected E(2)-related endogenous steroids strongly suggested that scFv#m2-c4 has improved specificity compared to conventional antibodies.

No abstract