Project 3: Antigen design

PROJECT SUMMARY - PROJECT 3: ANTIGEN DESIGN Project 3 focuses on developing generalizable approaches to the modeling, design, and evaluation of native-like antigens for arenaviruses, phenuiviruses, and paramyxoviruses. Powerful machine-learning (ML) methods enable stabilization of antigens in desired conformations and oligomeric states by structure prediction and sequence redesign. In Aim 1, we will develop methods for accurate structure prediction of viral glycoproteins, high-throughput ranking of constructs, and ML-based and deep mutational scanning (DMS)-guided design of stabilized antigens. These methods will be used to develop generalizable design strategies for phenuivirus Gn antigens with improved production yields and thermostability while preserving antigenicity. These methods will also be used to generate design strategies for single-chain and heteromeric antigens that stabilize native-like conformations of Gn-Gc heterodimers across Phenuiviridae. In Aim 2, we will develop similarly generalizable ML-based and DMS-guided approaches to stabilize oligomeric arenavirus and paramyxovirus antigens in particular conformations that elicit potent neutralizing antibodies. To design trimeric, prefusion-stabilized arenavirus glycoprotein complex (GPC) and paramyxovirus fusion (F) proteins, we will use sequence redesign, refine antigenic backbone, and generate de novo backbone. We will develop ML-based sequence design and de novo backbone generation strategies that enable design of thermostable, monomeric paramyxovirus receptor binding proteins (RBPs) and soluble, native-like tetrameric RBPs. In Aim 3, we will use biochemical, structural, and immunological characterization of designed antigens to evaluate the accuracy and generalizability of our design approaches and down-select to the most promising constructs. We will build upon the Institute for Protein Design's extensive infrastructure for protein production and characterization, using measurements of yield, thermostability, and antigenicity to map the strengths and limitations of different computational design approaches and refine design strategies. Serological and structural characterization of sera from immunized mice at the UW, Fred Hutchinson Cancer Center, and UTMB will identify constructs that elicit potent neutralizing antibody responses. To rigorously evaluate these constructs and distinguish lead candidates, we will leverage small animal challenge models being developed by UTMB to identify constructs that confer effective protection.