Isolation and engineering of a high affinity antibody against P-selectin glycoprotein ligand-1 (PSGL-1)

Abstract

We aim to develop novel protein antagonists of P-selectin adhesion as anti- inflammatory therapeutics. Blocking P-selectin adhesion is particularly attractive because this adhesion mediates leukocyte rolling which occurs early in the inflammatory cascade before extensive tissue damage caused by the amplification of inflammation by proinflammatory cytokines. Currently, no subnanomolar antagonists of selectin adhesion are available. The low affinity of current antagonists results in the need for frequent administration and large doses in order to obtain inhibition. High affinity antagonists are desirable because they can be administered in smaller amounts thus reducing the risk of harmful side effects and reducing production costs. Our approach for developing high affinity antagonists is to combine error prone PCR and in vivo homologous recombination to mimic in yeast the broad spectrum of mutagenic strategies exploited by B cells such as somatic hypermutation, receptor revision (... CDR replacement), receptor editing (chain shuffling), and amino acid insertions and deletions. Together with yeast surface display and flow cytometric screening (FACS), this approach has been used to effect at least a five order of magnitude affinity improvement in a single chain antibody (scFv) directed against the N-terminal 19 amino acids of P-selectin glycoprotein ligand- 1 (PSGL- 1). Three rounds of engineering were performed after an initial pool of binders was isolated from a non-immune scFv library. Chain shuffling was found to be important for generating an improved mutant in the first round of engineering.

(Cont.) For the final round of engineering, four different libraries were generated: one with random mutations, one with preferential replacement of the ... CDR1, one with preferential replacement of the ... CDR1 and the ... CDR2, and one with preferential replacement of the light chain. All of these methods produced two order of magnitude affinity improvements except the light chain exchange library. However, the CDR exchange libraries gave equivalent affinity improvement despite the fact that they were 77 fold smaller than the random mutagenic library. In addition, an insertion in CDR2 of the VH was isolated in the best binder from both of the CDR exchange libraries and this mutation could not have been found through random mutagenesis. These results suggest that chain shuffling is best used when the affinity of the antibody to be matured is weak (> 1 [mu] M). In addition, receptor revision is an equally robust method as random mutagensis for the generation of ultra-high affinity binders. The best antibody from the library with preferential replacement of ... CDR1 and ... CDR2 was converted to an IgG and characterized. It was found to better inhibit P-selectin binding to PSGL-1 than the commercially available antibody KPLI in a static adhesion assay and an in vitro rolling assay. Our integrated approach, made possible by in vivo homologous recombination in yeast, decreases the likelihood of convergence upon a single high affinity solution and increases the probability of obtaining an antibody with desired secretory properties and therapeutic potential. This facile method for combining all the mutational strategies used in nature should prove as a valuable tool in the antibody engineering field.