Networks of genomic co-occurrence capture characteristics of human influenza A (H3N2) evolution - PubMed (original) (raw)
Networks of genomic co-occurrence capture characteristics of human influenza A (H3N2) evolution
Xiangjun Du et al. Genome Res. 2008 Jan.
Abstract
The recent availability of full genomic sequence data for a large number of human influenza A (H3N2) virus isolates over many years provides us an opportunity to analyze human influenza virus evolution by considering all gene segments simultaneously. However, such analysis requires development of new computational models that can capture the complex evolutionary features over the entire genome. By analyzing nucleotide co-occurrence over the entire genome of human H3N2 viruses, we have developed a network model to describe H3N2 virus evolutionary patterns and dynamics. The network model effectively captures the evolutionary antigenic features of H3N2 virus at the whole-genome level and accurately describes the complex evolutionary patterns between individual gene segments. Our analyses show that the co-occurring nucleotide modules apparently underpin the dynamics of human H3N2 evolution and that amino acid substitutions corresponding to nucleotide co-changes cluster preferentially in known antigenic regions of the viral HA. Therefore, our study demonstrates that nucleotide co-occurrence networks represent a powerful method for tracking influenza A virus evolution and that cooperative genomic interaction is a major force underlying influenza virus evolution.
Figures
Figure 1.
Overview of network construction. (A) Flow diagram showing the computational processes of analyzing H3N2 virus genomic co-occurrence network. (B) Mathematical framework of Step 4, construction of the viral genomic co-occurrence network. See Methods for details.
Figure 2.
Evolving network connectivity of human influenza virus strains. (A) The average vertex connectivities (K) for each of the 1032 H3N2 nucleotide co-occurrence networks were calculated and plotted by flu season. Each marker represents one strain’s nucleotide co-occurrence network isolated from New York (circles), Memphis (asterisks), Hong Kong (squares), or Canterbury (crosses). The connectivity clusters for predominant strains (see Methods for detailed explanation of calculations) are colored according to their best-matched antigenic clusters in Smith et al’s analysis (Smith et al. 2004). The title denoted “Antigenicity” indicates the color-coded antigenic clusters in Smith’s work. (B) Phylogenetic tree derived from the concatenated viral genomes of H3N2 strains isolated between 1968 and 2006. Bootstrap values are shown for key nodes. The strains within a connectivity cluster in A are mapped to the phylogenetic tree and are shaded with the same color as the connectivity clusters. The clades are labeled using the season from which the majority of the strains were isolated. (C) Season-by-season analysis of whole-genome sequence evolution of 1032 strains. Genetic (nucleotide substitution) distances relative to A/Hong Kong/1/68 were calculated from the phylogenetic tree in B for each of the 1032 strains. The color and symbol representation are the same as in A. (D) Standardized vertex connectivities (K) for the simulated sampling of nucleotide co-occurrence networks are plotted by season. The blue line indicates the average K for all strains sampled from that season.
Figure 3.
Evolutionary patterns of individual influenza gene segments. (A,B) Average standardized connectivity K of H3N2 nucleotide co-occurrence networks for gene segments (A) HA, NA, MP, and NS; and (B) NA, NP, PA, PB1, and PB2 from 1998–2006. The largest changes in connectivity K in HA, NA, and NS before 2004 are labeled in red and indicated by arrows. (C,D) Phylogenetic trees of (C) HA and (D) NA nucleotide sequences from influenza A viruses sampled from 1997 to 2006. Bootstrap values are shown for the key nodes. The clades for strains during 2000–2006 were labeled using the season from which the majority of the strains were isolated.
Figure 4.
Quantification of correlated changes between H3N2 gene segments between 1998 and 2006. (A) Heat map of the rates (R) of season-to-season intragenic (left panel) and intergenic (right panel) connection changes for the simulated sampling of H3N2 strains from 1998 to 2006. Significant intragenic HA change is labeled with an arrow. See Methods for detailed calculations. (B) The extent of cooperative changes (C) between gene segments for the simulated sampling of H3N2 strains from 1998 to 2006. See Methods for detailed calculations.
Figure 5.
Characteristics of co-occurring nucleotide modules. (A) Heat map of the distribution of transient and transition modules from 1998 to 2006 subject to hierarchical clustering based on the strain-module profile. Rows correspond to transient and transition modules, and columns correspond to the indicated season. If a transient module is present in a majority of strains in a season, it is blue; otherwise it is gray. Transition modules are labeled according to which of the two member modules is present in the majority of strains in a season (first member, blue; second member, red). (B) Module-based amino acid changes corresponding to the nucleotide changes in transient and transition modules from 1998 to 2006. Each row represents a single amino acid position in a viral protein (labeled on the right). Amino acids (single-letter abbreviations) are also color-coded, so that mutations can be seen as changes both in amino acid identity and color. See Supplemental Figure S3 for all module-based amino acid changes from 1968 to 2006.
Figure 6.
Module-based H3N2 amino acid substitutions mapped onto the HA structure. Amino acid substitutions from 1968 to 2006 are mapped onto the (left panel) exposed and (right panel) buried surface of the HA monomer. Residues are colored by module-based amino acid changes (red), non-module-based changes (yellow), and intact residues (purple). The five antibody epitopes (Wiley et al. 1981) and the receptor-binding site (Skehel and Wiley 2000) are circled. The module-based amino acid substitutions carrying two or more residues in 1998–2006 are labeled.
References
- Bhat N., Wright J.G., Broder K.R., Murray E.L., Greenberg M.E., Glover M.J., Likos A.M., Posey D.L., Klimov A., Lindstrom S.E., Wright J.G., Broder K.R., Murray E.L., Greenberg M.E., Glover M.J., Likos A.M., Posey D.L., Klimov A., Lindstrom S.E., Broder K.R., Murray E.L., Greenberg M.E., Glover M.J., Likos A.M., Posey D.L., Klimov A., Lindstrom S.E., Murray E.L., Greenberg M.E., Glover M.J., Likos A.M., Posey D.L., Klimov A., Lindstrom S.E., Greenberg M.E., Glover M.J., Likos A.M., Posey D.L., Klimov A., Lindstrom S.E., Glover M.J., Likos A.M., Posey D.L., Klimov A., Lindstrom S.E., Likos A.M., Posey D.L., Klimov A., Lindstrom S.E., Posey D.L., Klimov A., Lindstrom S.E., Klimov A., Lindstrom S.E., Lindstrom S.E., et al. Influenza-associated deaths among children in the United States, 2003-2004. N. Engl. J. Med. 2005;353:2559–2567. - PubMed
- Bush R.M., Bender C.A., Subbarao K., Cox N.J., Fitch W.M., Bender C.A., Subbarao K., Cox N.J., Fitch W.M., Subbarao K., Cox N.J., Fitch W.M., Cox N.J., Fitch W.M., Fitch W.M. Predicting the evolution of human influenza A. Science. 1999;286:1921–1925. - PubMed
- Chen H., Smith G.J., Zhang S.Y., Qin K., Wang J., Li K.S., Webster R.G., Peiris J.S., Guan Y., Smith G.J., Zhang S.Y., Qin K., Wang J., Li K.S., Webster R.G., Peiris J.S., Guan Y., Zhang S.Y., Qin K., Wang J., Li K.S., Webster R.G., Peiris J.S., Guan Y., Qin K., Wang J., Li K.S., Webster R.G., Peiris J.S., Guan Y., Wang J., Li K.S., Webster R.G., Peiris J.S., Guan Y., Li K.S., Webster R.G., Peiris J.S., Guan Y., Webster R.G., Peiris J.S., Guan Y., Peiris J.S., Guan Y., Guan Y. Avian flu: H5N1 virus outbreak in migratory waterfowl. Nature. 2005;436:191–192. - PubMed
MeSH terms
Substances
LinkOut - more resources
Full Text Sources