Antigenic cartography of H1N1 influenza viruses using sequence-based antigenic distance calculation

Abstract

Background: The ease at which influenza virus sequence data can be used to estimate antigenic relationships between strains and the existence of databases containing sequence data for hundreds of thousands influenza strains make sequence-based antigenic distance estimates an attractive approach to researchers. Antigenic mismatch between circulating strains and vaccine strains results in significantly decreased vaccine effectiveness. Furthermore, antigenic relatedness between the vaccine strain and the strains an individual was originally primed with can affect the cross-reactivity of the antibody response. Thus, understanding the antigenic relationships between influenza viruses that have circulated is important to both vaccinologists and immunologists.

Results: Here we develop a method of mapping antigenic relationships between influenza virus stains using a sequence-based antigenic distance approach (SBM). We used a modified version of the p-all-epitope sequence-based antigenic distance calculation, which determines the antigenic relatedness between strains using influenza hemagglutinin (HA) genetic coding sequence data and provide experimental validation of the p-all-epitope calculation. We calculated the antigenic distance between 4838 H1N1 viruses isolated from infected humans between 1918 and 2016. We demonstrate, for the first time, that sequence-based antigenic distances of H1N1 Influenza viruses can be accurately represented in 2-dimenstional antigenic cartography using classic multidimensional scaling. Additionally, the model correctly predicted decreases in cross-reactive antibody levels with 87% accuracy and was highly reproducible with even when small numbers of sequences were used.

Conclusion: This work provides a highly accurate and precise bioinformatics tool that can be used to assess immune risk as well as design optimized vaccination strategies. SBM accurately estimated the antigenic relationship between strains using HA sequence data. Antigenic maps of H1N1 virus strains reveal that strains cluster antigenically similar to what has been reported for H3N2 viruses. Furthermore, we demonstrated that genetic variation differs across antigenic sites and discuss the implications.

Keywords: Antigenic cartography; Antigenic distance; H1N1; Hamming distance; Hemagglutinin; Influenza.

Fig. 1

Comparison of Experimental Data with Sequence-based Antigenic Distance. a Relative HAI log titer of ferret antiserum. b ELISA absorbances of ferret antiserum to rHA. c Amount of mouse derived monoclonal antibodies bound to rHA by ELISA

Fig. 2

GOF After Dimension Reduction. Goodness-of-fit calculations after dimension reduction into k dimensions (1–10). Two methods were used to calculate the GOF depicted by open or closed boxes

Fig. 3

Sequence-based Map of H1N1 HA Proteins. Antigenic cartography of sequence-based antigenic distance calculations for HA protein sequences on 4838 H1N1 viruses. Each point on the map represents a HA protein antigen. The distance between two HA protein antigens on the map represents the antigenic distance between the two antigens. Historical Strains are labeled. Points are colored based on categorical hierarchical clustering

Fig. 4

Precision and Accuracy. a Precision of the distances between antigens on the map was determined using random sampling of HA sequences (0–4000 sequences) to construct the map. Each sampling was performed 50 times. Precision is given as the mean percent difference between distances of randomly chosen sequences and distances in original map. Error bars represent the standard deviation between results of sequence samplings. b ROC-AUC measurements of accuracy for each of assay type and criteria, HAI-A and HAI B represent 2-fold and 1:40-dilution criteria, respectively. c ROC curves over the range of antigenic distances cutoffs

Fig. 5

Epitopic Distances of H1 HA Epitopes. a-e Antigenic maps for each antigenic site (Sa, Sb, Ca1, Ca2, Cb, respectively). Strains were colored as in Fig. 2. f Average antigenic distances between all strains for each antigenic site