Discovery and Characterization of Allosteric Antibodies

Antibodies are the fastest-growing class of treatments, with a majority of them being approved in the oncology area. Proteins are inherently flexible, and this flexibility can be exploited to modulate ligand-binding and target activity, a phenomenon referred to as allostery. This doctoral thesis focuses on the discovery and characterization of allosteric antibodies for cancer immunotherapy, specifically targeting the epidermal growth factor receptor (EGFR). EGFR, a transmembrane protein belonging to the receptor tyrosine kinase family, can be dysregulated and overexpressed, leading to cancer proliferation. As a widely pursued target, four therapeutic antibodies are approved for the treatment of EGFR+ cancers. These antibodies all compete with the natural ligand EGF for binding. Here, alternative strategies were explored for the discovery of allosteric antibodies to leverage EGFR biology and develop the next generation of anti-EGFR antibodies. The first aim of this doctoral thesis was to develop a robust strategy for the direct identification of non-ligand-competitive antibody fragments from yeast surface display (YSD) libraries. Traditionally, the identification of allosteric binders is performed using in vitro assays or by structural studies. Conventional screening of YSD libraries typically employs fluorescence-activated cell sorting (FACS) in the presence of the target protein using a two-color staining strategy. By leveraging the precise control over several parameters of FACS, an additional criterion was introduced during the sorting: the labeled ligand. Two YSD libraries, one displaying fragment-antigen binding (Fab) fragments and the other variable heavy domain of heavy chain only antibodies (VHHs), previously enriched for EGFR-binders, were utilized to develop and optimize this approach. These libraries were incubated with various concentrations of either biotinylated EGF or EGF Fc fusion protein, followed by secondary detection using either streptavidin-fluorescence or anti Fc antibody fluorescence conjugates. These libraries were screened using FACS with a three-color gating strategy, to simultaneously detect functional display, EGFR binding, and EGF presence. Conditions were meticulously compared to ensure the identification of the most promising candidates. By comparing sequencing results with the traditional sorting method, a greater number of hit candidates were obtained, demonstrating that this platform can significantly streamline the discovery process for candidates with the desired properties. The second study focused on the discovery and characterization of allosteric antibodies that inhibit EGFR allosterically. To this end, an immune camelid yeast surface display library was generated and screened by FACS in the presence of EGF Fc fusion protein, based on the optimized sorting conditions established in the first study. After two rounds of sorting, eleven binders were identified and reformatted as Fc fusion proteins. Binding to four different allosteric sites was confirmed through competition assays via biolayer interferometry. The first aspect of allosteric modulation was examined using a fluorescence polarization assay to evaluate the impact of antibodies on the affinity of EGFR for EGF, revealing various modulations. The second aspect of allosteric modulations involved the inhibition of EGFR-mediated signaling pathways. The impact on the MAPK/ERK pathway was assessed using a cell line reporter, which demonstrated that all the antibodies inhibited this pathway, with inhibitory concentrations ranging from two-digit picomolar to two-digit nanomolar. Subsequent homogeneous time resolved fluorescence (HTRF) assays determined the impact of antibodies on the PI3K/AKT pathway, revealing that one binder was able to inhibit this critical pathway. This binder exhibited greater inhibitory properties for both pathways compared to the commercially approved EGF-competitive antibody cetuximab. This investigation demonstrated that EGFR can be allosterically inhibited by an antibody more effectively than by EGF-competitive antibodies. Additionally, as none of these antibodies bind to paralogue proteins, they suggest targeting allosteric sites of EGFR presents a better selective profile and may represent a potential new mechanism of action. Bispecific natural killer cell engagers (NKCEs) represent an alternative approach for tumor cell killing by bringing into close proximity NK cells with tumor cells. Previous studies demonstrated the successful engagement of NKp30 on NK cells with tumor cells expressing EGFR. These studies were conducted on tumor cells that did not express one of the natural ligands of NKp30: B7-H6, striking out natural ligand recognition. In the third study presented here, NKp30-based NKCEs were engineered to avoid the competition with B7-H6. First, an immune camelid yeast surface display library was generated. This library was incubated simultaneously with NKp30 and B7-H6 to obtain VHHs that would bind to allosteric sites of NKp30, based on the method established in the first section. Eight candidates were identified, and BLI assays confirmed binding to two different allosteric sites of NKp30. Intriguingly, antibodies that did not cross-react with the rat ortholog also exhibited poor affinity to the endogenous NKp30 on NK92 cells. Subsequently, rat cross-reactive antibodies were reformatted as DuoBody with either cetuximab (anti-EGFR moiety) or trastuzumab (anti-HER2 moiety). These NKCEs were tested for their killing capacities on HeLa cells (B7-H6+/EGFR+) or SK-BR-3 cells (B7-H6+/HER2+). These constructs demonstrated a significant enhancement of killing capacities correlated with B7-H6 expression. An enhanced-targeting approach was proposed for allosteric NKCEs, wherein the natural recognition of B7-H6 complements the induced close proximity facilitated by NKCEs. The last part of this work presents a literature review of targets modulated allosterically by antibodies. This review highlights the growing interest for antibodies in targeting allosteric sites, which opens up novel avenues for druggability. Additionally, allosteric antibodies have great potential as tools in the drug discovery process, such as structural studies or to facilitate high-throughput screening. Finally, we intend to illuminate a roadmap for their design using contemporary computational approaches.