Developing Modular Protein Therapeutics as Alternatives to Monoclonal Antibodies for Cancer Immunotherapy

Abstract

Monoclonal antibodies have been successfully developed as PD-L1 antagonists, showing unprecedented anti-cancer immune response and efficacy with their high affinity and exquisite specificity. Despite its substantial success, the development of antibody drugs is approaching an asymptote because their structural inflexibility limits the ability to tune their valency, receptor accessibility, and blood circulation duration. Their efficacy in treating solid tumors is also limited by low tumor penetration due to their large size and structural inflexibility. These inherent challenges to utilizing antibodies as therapeutics confine their further improvements. To address these limitations, we have developed a modular protein therapeutic with rationally tunable valency, affinity, pharmacokinetics, and tumor penetration. In modular protein therapeutics, we can independently tune their affinity and valency for any given target, as well as modulate their pharmacokinetics and tumor penetration. To create a modular PD-L1 antagonist, we chose the human tenth fibronectin type III domain (FN3) as the “affinity module” because it is a small (~10 kDa), structurally robust protein domain that has six disordered loops that are similar to the complementarity-determining regions (CDRs) of antibodies. We then oligomerized the affinity module to enhance its binding to PDL1 via the avidity. To optimize pharmacokinetics, we fused the oligomerized affinity module with a “half-life module”: 1) an elastin-like polypeptide (ELP) that is injectable as a solution at room temperature but forms a gel-like depot at 37 °C and provides sustained, the zero-order release of the fusion; or 2) an albumin-binding domain that binds to and exploits the endogenous albumin to significantly extend the plasma half-life. The fusion with a “half-life module” would enable our modular protein therapeutics to rival the pharmacokinetics of antibodies. In this thesis, we discovered PD-L1-binding FN3 proteins (aPDL1-FN3) using phage display and modulated their valency and affinity for their equilibrium dissociation constants (KD) in the picomolar range. Unlike bivalent antibodies, the multivalency of FN3 is not restricted, and tetra-valency was chosen because (aPDL1-FN3)4 reached a plateau in terms of binding ability to PD-L1, measured by surface plasmon resonance (SPR), in vitro PD-L1 neutralization assay, and flow cytometry. To overcome glomerular filtration cutoff (~50 kDa) and improve pharmacokinetics, we genetically fused the (aPDL1-FN3)4 protein with either elastin-like polypeptides (ELPs) or albumin-binding proteins (ABDs). Using these fusions, we studied pharmacokinetics, biodistribution, and tumor uptake as compared to anti-PD-L1 antibodies. Also, we validated in vivo preclinical efficacy using three different immunocompetent mice models: 1) B16.F10 melanoma model; 2) CT26 colon cancer model; and 3) MC38 colon cancer model. These results demonstrated that our modular protein therapeutics successfully mimic antibodies as alternative therapeutics and have the potential to outperform antibodies regarding multivalency and cellular and tumor penetration. We believe that this research project serves as a proof-of-concept for modular protein therapeutics where tunable efficacy and pharmacokinetics can lead to a clinical utility that can eventually overcome the hurdles of traditional antibody-based therapy.