The Haemagglutination Inhibition AssayAntigenic differences between influenza viruses are routinely measured using the haemagglutination inhibition (HI) assay. The HI assay is a binding assay based on the ability of haemagglutinin (HA), the surface glycoprotein of the influenza virus, to agglutinate red blood cells, and the complementary ability of animal antisera raised against the same or related influenza strains to block this agglutination. Thus an HI titre gives information about the affinity of an antiserum for a virus strain. One can interpret a titre value as a rough measure of distance between the antiserum and the virus. We use this data in an algorithmic way to aid understanding of the evolutionary dynamics of influenza.
Although the HI assay has provided essential antigenic difference information on influenza viruses for over 60 years, HI data are difficult to interpret quantitatively. Assays of the same strain and antiserum will occasionally produce different values, even in the same laboratory. There have been attempts to quantify and visualise such binding assay data, but none have been used widely or persistently, nor have they provided an underlying theory that resolves paradoxes in the data. As a result, these data have almost universally been interpreted by eye, and have been considered reliable enough for judging only large antigenic differences. In the case of influenza virus, this means differences of sufficient magnitude to necessitate an update of the vaccine strain (an extensive and expensive procedure).
Antigenic cartography is the process of applying mathematical, computational, and statistical techniques to challenging antigenic binding assay data to create antigenic maps. These are not maps in the geographical sense, but in a biological sense: they provide a spatial layout of assay components (virus strains and antisera in the case of influenza), allowing the precise measurement of distances and directions amongst components. This gives a visualisation of the underlying data and, more importantly, provides a concrete mathematical foundation for the quantitative analysis of antigenic data.
In time, we expect that antigenic maps will become as ubiquitous in the analysis of antigenic data as phylogenetic trees are in the analysis of genetic data. However, importantly, antigenic maps differ from genetic analysis because they reflect antigenic properties of pathogens. While genetic analysis has provided major insights, these insights are of greater impact when related to phenotypic traits. Although antigenic maps themselves are of direct inherent utility—providing intuitive visualisations of assay data, as used to inform the WHO influenza vaccine strain selection process—they also provide a general characterisation that can inform and direct genetic analysis, and make such analysis more immediately applicable. For example, there is a close relationship between genetic and antigenic change in human influenza A(H3N2) virus, but genetic distance alone is sometimes an unreliable predictor of antigenic distance. Single amino acid changes have been shown to cause large changes in the binding properties of virus strains. Antigenic maps can show precisely when large movements in the antigenic space result from minimal genetic change. Antigenic cartography therefore offers an improved understanding of genetic and antigenic evolution.
Above. The first antigenic map of the influenza A(H3N2) virus from 1968 to 2003. Each strain is coloured to represent the antigenic cluster to which that strain belongs, while the antisera used in the HI assays are shown as uncoloured open shapes. Clusters are named with two letters referring to the location of the first isolation and two digits refer to year of isolation. Both the vertical and horizontal axes represent antigenic distance; this is similar to the way both axes on a geographical map represent geographic distance.