Development of a QSP model for the prediction of mRNA influenza vaccine immunogenicity

Objective: Globally, seasonal influenza is estimated to cause up to 650,000 respiratory deaths annually, with vaccination being the most effective preventative measure [ 1]. Influenza vaccination is a crucial public health strategy. However, the vaccine effectiveness of currently licensed seasonal influenza vaccines have suboptimal and inconsistent effectiveness from season to season, in part due to antigenic differences between vaccine strains and circulating viruses [ 2] [ 3]. Consequently, there has been an increased focus on the application of mRNA-based technologies for influenza vaccines which have been shown to elicit a robust, multi-faceted immune response, inducing both B- and T-cell immunity [ 4]. Preclinical characterization of vaccine candidates across a combination of different formulations, antigen constructs, and doses provides valuable qualitative insights. However, uncertainties remain in translating from preclinical assessments to clinical outcomes remain given the complexities of prior immunity and, host factors that impact immunity such as the impact of aging on immunogenicity in humans. We developed a novel quantitative systems pharmacology (QSP) platform model with an ultimate goal to predict clinical vaccine immunogenicity from preclinical in vivo immunogenicity data to inform candidate selection and de-risk clinical trial design decisions. As an initial step, we use the model to recapitulate influenza vaccine responses in preclinical animal species including mice and non-human primates. Methods: We adapted a previously developed multiscale mathematical model for vaccine-induced immunogenicity prediction [ 5– 7]. This model comprises key biological immune mechanisms including antigen presentation, activation, proliferation, and differentiation of immune cells [ 8]. The model is modified to represent the immunogenicity responses of a monovalent influenza mRNA vaccine (mIRV) encoding an H1 hemagglutinin (HA) in mice and non-human primates. Results: The QSP model was parameterized to match the immune response to mIRV vaccination in mice and non-human primates. For each preclinical species, the model adequately predicted the dynamics of the immune response, specifically hemagglutination inhibition (HAI) antibody titers, following two doses of mIRV administered 28 days apart. Specifically, the model described the peak HAI titer response for the two-dose regimen and the durability of vaccine induced HAI titers up to 5 months post-vaccination [ 2]. Conclusion: We developed a QSP model of immunogenicity that captures key features of the immune response to mIRV vaccination in preclinical animal species. This QSP framework will be updated with emerging data from the corresponding clinical studies of this vaccine (NCT05052697 [ 9], NCT05540522 [ 10], NCT05572450 [ 11]), to investigate key preclinical predictors of clinical immunogenicity. Furthermore, future efforts will focus on leveraging the model to predict clinical vaccine immunogenicity based on preclinical assessments and will account for pre-existing immunological memory observed clinically. In the future, this can help streamline candidate selection and de-risk vaccine development decisions.