Translatable Affibody Ligand Discovery: Engineered Biopanning, Deep Sequence-Guided Library Design, and Pharmacokinetic Modulation

The affibody is a 58-amino acid three-helix bundle that has been an effective ligand scaffold for clinical molecular imaging probes, targeted therapy, and biotechnology. A key challenge in discovery and development of affibodies, and all protein ligands, is the inconsistency of binders evolved against recombinant target to translate to strong, specific binding to genuine intact cellular target. Direct selection of yeast-displayed antibody libraries on adherent mammalian cell monolayers has been previously used to address this limitation. To enhance the robustness and selective control of cellular panning, we have systematically elucidated the elements that dictate ligand enrichment. Introduction of an extended flexible PAS peptide between the cell-surface tethering protein and the displayed ligand enhances enrichment 44-fold, which enhances discovery of all binders in a library. Moreover, use of mammalian cells with moderate target expression enables modest affinity discrimination to selectively enrich the strongest ligands. We have demonstrated controlled chemical cleavage of displayed ligands to control valency over two orders of magnitude, which empowers strict and broadly titratable affinity discrimination.

Twenty different ligand discovery campaigns were competed, in two scaffolds (affibody and fibronectin domain) and multiple targets, to compare various selection strategies. Direct cell panning enhances discovery of translatable hits four-fold versus campaigns using initial enrichment on magnetic beads coated in recombinant target followed by either flow cytometry with recombinant target or biotinylated cell lysate. Conditions to optimize depletion panning, to remove undesirable binders, will be discussed.

Deep sequencing an epitope-diverse pool (>37,000 unique variants in >5,000 family clusters) of evolved binders identified amino acid motifs consistent with evolutionary efficacy, which revealed a non-spatial gradient of diversity (Shannon entropy: 1.4-4.0) in an optimized paratope library. Implementation of these sitewise amino acid distributions yielded a 50-fold increase in the binder discovery efficiency towards a broad panel of targets. Moreover, affibodies from this constrained library design exhibit dramatically higher thermal stability (24±12ºC higher T_m) than binders from a naïve library.

In parallel, the affibody framework was engineered for reduced charge density while retaining stability, solubility, and binding. A synthetic consensus design strategy – merging directed evolution and bioinformatics of evolutionarily tolerant residues – yielded better variants than directed evolution or informatics alone. Neutralization of three acidic and three basic residues resulted in a 39±10% reduction in non-specific renal retention while increasing tumor targeting 21±5% in murine xenograft models.

Collectively, enhancements in library design, ligand selection, and framework design for biodistribution improve the ability to engineer translatable affibody ligands.