Most display methods are based on antibody fragments binding activity. However, the preferred format for therapeutic applications is the IgG whose binding properties are affected by reformatting. We thus developed a display vector and an engineered mammalian cell line that allow both phage display and direct generation of cells stably secreting a monoclonal human IgG for functional screening. This system allows the screening of phage-derived antibody fragments as full-length IgG. In addition, using optimized selection procedures and synthetic library design, we show it is possible to directly select antibodies modulated by the pH for improved therapeutic targeting.

We developed a high-diversity single-chain variable fragment phage display library from naïve mice that allowed us to identify a suite of fragments that bound the serine protease Fibroblast Activation Protein (FAP). Expressed on the surface of both cancer-associated fibroblasts and myofibroblasts involved in wound healing, FAP is a target for cancer and fibrosis. Using a humanized version of our lead fragment in an IgG format, we were able to imaging FAP expression by positron emission tomography in murine models of prostate cancer and idiopathic pulmonary fibrosis with high accuracy and precision.   

A robust, alternative peptide-scaffolds platform, which consists of disulfide bond-constrained peptides that are resistant to thermal, proteolytic, and chemical degradation, has been developed with phage display. Ensembles of tens of libraries with overall diversity of 10^11 have been constructed. Binders against a target of interest have been identified from the libraries and further optimized with display technology and chemical synthesis to highly potent and stable leads with ideal pharmacological properties.

We have generated a series of conditionally masked therapeutic antibodies that are unmasked in the tumor microenvironment by acid pH and/or by proteolytic removal of the masking entity. Using anti-idiotype shark-derived vNARs, pH-responsive paratope binders were generated by yeast display screening. As an additional broadly applicable approach to modulate effector functions of IgG1 therapeutic antibodies, a protease-cleavable Fc-masking scFv was developed, efficiently blocking Fc-gamma receptor binding conditionally.

A subset of antibodies found in cattle comprises ultralong CDR-H3 regions of up to 70 amino acids. These peculiar entities display a huge structural diversity that makes them attractive for biomedical applications. We have generated a robust platform process to specifically harness this subset after immunization. Moreover, we have engineered bispecific common-light-chain antibodies based on ultralong CDR-H3 diversities and demonstrated that these molecules are versatile for conditionally activating NK cells.

We used a phenotypic discovery approach comprising differential antibody display screening of primary leukaemic B cells and healthy donor PBMCs to identify antibodies and targets associated with direct and immune effector cell-mediated killing of leukaemic B cells.  The inhibitory FcγRIIB was identified as a promising target. Anti-FcγRIIB overcame rituximab resistance in vivo and is currently being trialed in clinical Phase I/II trials with relapsed/refractory NHL patients.

Chromosomal translocation-proteins and mutant RAS are among hard-to-drug proteins in cancer. Intracellular antibodies are a starting point as inhibitors via design to block protein-protein interactions or to carry effector functions. Fusion with E3 ligases creates bio-degrader intracellular antibodies to eliminate target proteins. The effect of bio-degraders targeting LMO2 and mutant RAS on cancer cells will be discussed and approaches to establish delivery of intracellular antibodies as drugs per se.

New nucleolin-targeting nanobodies with cytotoxic activity against nucleolin-overexpressing breast cancer cells were developed. The nanobody-Fc antibody presented nucleolin-mediated ADCC capacity. Antibodies were optimized by random mutagenesis in E. coli and selected by endocytic properties by flow cytometry. The best nanobody candidates presented a single mutation in CDR3 and showed increased potency. Data will be presented on the optimization and biologic activities of selected anti-nucleolin antibodies.

Current antibody discovery and optimization methods reliably produce strong binders to the immunodominant epitopes of most targets by enriching for a single property: target affinity. However, therapeutic antibodies require multi-variable optimization for binding to target(s), animal orthologues, paralogues, related proteins, and other human proteins – as well as optimization for developability. Here, we will demonstrate how multi-dimensional antibody-antigen protein interaction datasets generated using the AlphaSeq platform enable the discovery and optimization of antibodies to diverse target epitopes with desired specificity and cross-reactivity profiles. In addition to direct selection of antibodies with desirable binding properties, AlphaSeq data is used to train machine learning models to improve the speed and quality of subsequent antibody designs.​