Differential neutralization efficiency of hemagglutinin epitopes, antibody interference, and the design of influenza vaccines

Abstract

It is generally assumed that amino acid mutations in the surface protein, hemagglutinin (HA), of influenza viruses allow these viruses to circumvent neutralization by antibodies induced during infection. However, empirical data on circulating influenza viruses show that certain amino acid changes to HA actually increase the efficiency of neutralization of the mutated virus by antibodies raised against the parent virus. Here, we suggest that this surprising increase in neutralization efficiency after HA mutation could reflect steric interference between antibodies. Specifically, if there is a steric competition for binding to HA by antibodies with different neutralization efficiencies, then a mutation that reduces the binding of antibodies with low neutralization efficiencies could increase overall viral neutralization. We use a mathematical model of virus-antibody interaction to elucidate the conditions under which amino acid mutations to HA could lead to an increase in viral neutralization. Using insights gained from the model, together with genetic and structural data, we predict that amino acid mutations to epitopes C and E of the HA of influenza A/H3N2 viruses could lead on average to an increase in the neutralization of the mutated viruses. We present data supporting this prediction and discuss the implications for the design of more effective vaccines against influenza viruses and other pathogens.

Fig. 1.

HA and antibody interference. (A) Globular head of a monomer of HA (Protein Data Bank ID code 1hgf), showing five antibody-binding sites (or epitopes) and the receptor-binding site. The figure was drawn by using the PyMOL molecular graphics system. (B) Interference between antibodies that bind to two different HA epitopes. Illustrated are cross-sections of an IgG molecule and an HA trimer. The molecules were drawn approximately to scale. IgG is a Y-shaped molecule that can be separated into three fragments (two Fab fragments and one Fc fragment) of approximately the same size. A Fab fragment has approximate dimensions 80 × 50 × 40 Å (7). In comparison, an HA trimer has a length of ≈135 Å and a diameter of ≈55 Å (8), approximately equal to the width of the 40- × 50-Å distal surface of a Fab fragment, which contains the Fab-binding pocket.

Fig. 2.

Effects of antibody interference on viral neutralization. (A) Wild-type HA [denoted HA^(a)] containing two epitopes, a high-neutralization efficiency epitope (denoted E₁) located close to the receptor-binding site, and a low-neutralization efficiency epitope (denoted E₂) located farther from the receptor-binding site. A variant of HA^(a) [denoted HA^(b)] contains mutations to E₁, whereas another variant [denoted HA^(c)] also contains mutations to E₂. (B) Antibodies raised against HA^(a) bind to E₁ of HA^(c) much more readily than they bind to E₁ of HA^(b) because the additional mutations to E₂ remove antibody interference. Viruses carrying the more mutated HA^(c) are therefore neutralized more efficiently than viruses carrying the less mutated HA^(b) by wild-type antibodies. Lines emanating from the receptor-binding site indicate that the site is not completely occluded.

Fig. 3.

Deleterious effects of antibody interference on the host and proposed strategy for influenza vaccine design. (A) If viral HA contained only epitopes with high-neutralization efficiences, then only viruses with large epitopic changes could escape from antibodies. (B) Antibody interference from low-neutralization efficiency epitopes enables viruses with small epitopic changes also to escape from antibodies. (C) Proposed low-interference vaccine strain is genetically modified from viral target at low-neutralization efficiency epitopes of HA. Vaccine-induced antibodies only recognize high-neutralization efficiency epitopes of target. (D) Antibodies induced by low-interference vaccine strain have low affinity for low-neutralization efficiency epitopes of the target and therefore do not interfere with antibodies to high-neutralization efficiency epitopes, implying better neutralization. (E) Without antibody interference the target virus cannot escape from vaccine-induced antibodies via small epitopic changes. Lines emanating from the receptor-binding site indicate that the site is not completely occluded.

Fig. 4.

Antigenic effects of amino acid changes to individual epitopes of influenza virus HA. Antigenic effects (regression coefficients) were estimated as described in Materials and Methods. The 95% confidence limits for each estimated regression coefficient, obtained by bootstrap resampling, are shown. A, B, C, D, and E denote the five epitopes of HA. O are epitopic sites that do not belong to any of the five epitopes, and N are sites not known to be bound by antibodies (see Table S1 for additional details).

Fig. 5.

Correlation between amino acid differences at individual HA epitopes and transitions between 10 pairs of temporally adjacent influenza A/H3N2 viral antigenic clusters. The correlation between amino acid differences and the transition from antigenic cluster K to K′ was quantified as described in Materials and Methods. In the figure, a positive regression coefficient indicates that an increase in the number of amino acid differences at the associated epitope increases the probability that a given virus belongs to K′. Error bars indicate standard errors of estimated regression coefficients, and they were computed by assuming that the number of viruses belonging to either cluster K or K′ is binomially distributed. A, B, C, D, E, O, and N are defined in the legend of Fig. 4.