Codon bias and frequency-dependent selection on the hemagglutinin epitopes of influenza A virus

Abstract

Although the surface proteins of human influenza A virus evolve rapidly and continually produce antigenic variants, the internal viral genes acquire mutations very gradually. In this paper, we analyze the sequence evolution of three influenza A genes over the past two decades. We study codon usage as a discriminating signature of gene- and even residue-specific diversifying and purifying selection. Nonrandom codon choice can increase or decrease the effective local substitution rate. We demonstrate that the codons of hemagglutinin, particularly those in the antibody-combining regions, are significantly biased toward substitutional point mutations relative to the codons of other influenza virus genes. We discuss the evolutionary interpretation and implications of these biases for hemagglutinin's antigenic evolution. We also introduce information-theoretic methods that use sequence data to detect regions of recent positive selection and potential protein conformational changes.

Fig. 1.

The relationship between the volatility of a codon and the average volatility of its one-point neighbors. Stop codons are not counted as neighbors. Codon bias is preserved on mutation under both the Miyata (Upper; r² = 0.73, P < 10^-5) and the Hamming (Lower; r² = 0.38, P < 10^-7) metrics. As a result, codon bias is heritable along a mutating viral lineage.

Fig. 2.

The relationship between codon and amino acid diversity at each of the 329 residues in a 525-sequence alignment of the HA1 gene. The lower scatterplot shows an enlarged portion of the upper scatterplot. Diversity is quantified by using the modified Simpson index. Points are colored according to the epitope in which each residue lies: red (A), orange (B), green (C), light blue (D), and dark blue (E). The nonepitopic residues are shown in black. Epitopic residues generally lie along the diagonal x = y, whereas nonepitopic residues generally lie along the x axis. Exceptions to this pattern are discussed in the text. The top 25 diagonal residues are identified in Table 3.

Fig. 3.

The solved structure (12) of a 1968 strain of HA, consisting of two chains 329 and 175 residues long, shown here in its monomer form. The commonly used (15) definitions of the five antibody-combining regions are shown in red (A), orange (B), green (C), light blue (D), and dark blue (E). Residues 271-Asp, 220-Arg, 112-Val, 31-Asp, 5-Gly, 3-Leu, and 2-Asp are shown in yellow. These seven residues have not previously been characterized as positively selected but nevertheless show the same genomic pattern of codon variation as many epitopic residues (Fig. 2). The residues in yellow may represent sites that, in the current HA, are directly involved in antibody combination, comutate with epitopic residues, or determine the conformation of epitopes.