Differential recognition of computationally optimized H3 hemagglutinin influenza vaccine candidates by human antibodies

Among circulating influenza viruses in humans, H3N2 viruses typically evolve faster than other subtypes and have caused severe illness and deaths in millions of people since emerging in 1968. Computationally optimized broadly reactive antigen (COBRA) technology is one strategy to broaden vaccine-elicited antibody responses among influenza subtypes. In this study, we determined the structural integrity of an H3N2 COBRA HA, TJ5, and, as nearly all humans have pre-existing immunity to H3N2 influenza viruses, we probed the antigenic profile of several H3N2 COBRA HAs by assessing recognition of these immunogens by human B cells and monoclonal antibodies (mAbs). Of three recently described COBRA H3 HA antigens (TJ5, NG2, and J4), we determined that TJ5 and J4 HA proteins recognize pre-existing B cells (from the 2017-2018 vaccine season) more effectively than NG2 HA and a wild type Hong Kong/4801/2014 protein. H3 HA-specific human mAbs recognize wild type and COBRA HA proteins, and have functional activity against a broad panel of H3N2 viruses. mAb TJ5-5 recognizes TJ5 and J4 HA proteins, but has poor recognition of NG2 HA, similar to the global B cell analysis. To probe these recognition differences and to verify the structural integrity of the TJ5 HA protein, we determined a 3.4 Å structure via cryo-electron microscopy of TJ5-5 complexed with the TJ5 HA, which revealed residues important to the differential binding. Overall, these studies determined that COBRA H3 HA proteins have correct antigenic and structural features, and are recognized by B cells and mAbs isolated from seasonally vaccinated humans. Importance Vaccine development for circulating influenza viruses, particularly for the H3N2 subtype, remains challenging due to consistent antigenic drift. Computationally optimized broadly reactive antigen (COBRA) technology has proven effective for broadening influenza hemagglutinin (HA) elicited antibody responses compared to wild type immunogens. Here we determined the structural features and antigenic profiles of H3 COBRA HA proteins. Two H3 COBRA HA proteins, TJ5 and J4, are better recognized by pre-existing B cells and monoclonal antibodies from the 2017-2018 vaccine season compared to COBRA NG2 and a wild type A/Hong Kong/2014 HA protein. We determined a cryo-EM structure of one mAb that poorly recognizes NG2, mAb TJ5-5, in complex with the TJ5 COBRA HA protein and identified residues critical to mAb recognition. As NG2 is more effective than TJ5 for a recent Hong Kong/2019 virus, these data provide insights into the diminished effectiveness of influenza vaccines across vaccine seasons.