Data don’t lose their value after publication and researchers at the Center for Influenza Vaccine Research for High-Risk Populations (CIVR-HRP) demonstrated just that in their 2020 study published in Vaccine. The authors, Amanda Skarlupka, Ph.D., Andreas Handel, Ph.D., and Ted Ross, Ph.D., used immunological and clinical data generated from prior mouse and ferret studies to evaluate if antigenic distance measures (ADM) could predict hemagglutination inhibition (HAI) and infection outcomes.
Vaccination is the primary strategy used to mitigate the effects of annual, seasonal influenza epidemics. Influenza vaccines are formulated to induce immune responses against one representative strain from each of the four influenza virus subtypes endemic in humans, attempting to protect against two dominant influenza A virus (IAV) and influenza B virus (IBV) strains per vaccine. Vaccine-induced immune responses are primarily directed against the influenza virus surface glycoprotein, hemagglutinin (HA) as the most abundant and immunodominant antigen on the surface of influenza. However, the continual accumulation of mutations, particularly in key antibody-binding epitopes, within HA necessitates annual influenza vaccine efficacy evaluation, and often reformulation. Determining whether influenza vaccine strains require an update relies upon HAI results, wherein researchers assess the ability of anti-HA antibodies to react against a variety of circulating influenza and candidate vaccine strains. To assist in the vaccine selection process, computational methods were developed using HA protein sequence of immunodominant epitopes and antigenic similarity coupled with these data.
Skarlupka et al. (2020) used data from studies published in 2017 and 2019 to calculate ADM and determine if these measures correlated with HAI antibody responses and clinical outcomes observed in mice and ferrets. In the prior studies, these animals were vaccinated with HA protein from swine and human influenza viruses, as well as computationally optimized broadly reactive antigen (COBRA) HA antigens. Post-vaccination sera were collected from these animals and assayed via HAI across a panel of human and swine influenza viruses to generate HAI titers. Additionally, the animals were infected with swine and human influenza viruses and monitored for weight loss, as well as sampled for lung viral titer to evaluate infection outcomes. Using these pre-existing data, the researchers generated statistical models to determine if ADM calculated from HA amino acid sequences could predict anti-HA antibody titers, weight loss, and lung viral titer.
Overall, the researchers found that ADM correlated with changes in HAI between influenza strains and these results were consistent across both the mouse and ferret datasets. When the association between ADM and clinical outcomes were evaluated, similar trends were discovered in which ADM correlated with weight loss and lung viral titer in mice following influenza challenge.
However, these associations were not strong enough to predict HAI activity or disease outcomes. In their discussion, the authors explain that the data used in this study were generated from naïve animal models without previous influenza exposures. They theorize that the predictive capability of ADM may be stronger in humans, or repeat exposure animal models, with robust immune histories to influenza. Therefore, while the authors recognize the value of data derived from animal sources, they also caution the use of computational models developed from these datasets and encourage further exploration and optimization of these models.