Genetic characterization of human coxsackievirus A6 variants associated with atypical hand, foot and mouth disease: a potential role of recombination in emergence and pathogenicity

Abstract

Human coxsackievirus A6 (CVA6) is an enterically transmitted enterovirus. Until recently, CVA6 infections were considered as being of minor clinical significance, and only rarely aetiologically linked with hand, foot and mouth disease (HFMD) associated with other species A enteroviruses (particularly EV71 and CVA16). From 2008 onwards, however, CVA6 infections have been associated with several outbreaks worldwide of atypical HFMD (aHFMD) accompanied by a varicelliform rash. We recently reported CVA6-associated eczema herpeticum occurring predominantly in children and young adults in Edinburgh in January and February 2014. To investigate genetic determinants of novel clinical phenotypes of CVA6, we genetically characterized and analysed CVA6 variants associated with eczema herpeticum in Edinburgh in 2014 and those with aHFMD in CAV isolates collected from 2008. A total of eight recombinant forms (RFs) have circulated worldwide over the past 10 years, with the particularly recent appearance of RF-H associated with eczema herpeticum cases in Edinburgh in 2014. Comparison of phylogenies and divergence of complete genome sequences of CVA6 identified recombination breakpoints in 2A-2C, within VP3, and between 5' untranslated region and VP1. A Bayesian temporal reconstruction of CVA6 evolution since 2004 provided estimates of dates and the actual recombination events that generated more recently appearing recombination groups (RF-E, -F, -G and -H). Associations were observed between recombination groups and clinical presentations of herpangina, aHFMD and eczema herpeticum, but not with VP1 or other structural genes. These observations provided evidence that NS gene regions may potentially contribute to clinical phenotypes and outcomes of CVA6 infection.

Fig. 1.. Phylogeny of CVA6 variants in different genome regions. Maximum-likelihood tree of (a) VP1 [positions 2430–3267 numbered using the reference CVA6 sequence, Gdula (GenBank accession number AY421764)], (b) 3Dpol (positions 6245–6988), (c) 5′ UTR (positions 24–752) and (d) VP4/partial VP2 (positions 753–1187). Trees were reconstructed using optimal substitution models: Kimura two-parameter (K2P)+invariant sites (I) for VP1; K2P+I+gamma distribution (Γ) for 3Dpol; K2P+I for 5′ UTR and K2P+Γ for VP4/2. Trees were rooted with the prototype sequence, GenBank accession number AY421764 (not shown). Bootstrap resampling (100 replicates) was used to determine robustness of groupings; values of ≥70 % shown. Sequences were coloured according to their recombination group assignments based on 3Dpol phylogeny.

Fig. 2.. Tree positions (TreeOrder scan) of species A serotypes across the genome. (a) Tree positions represented on the x-axis of available complete genome sequences of EV-A, including CVA6 genomes obtained in the current study. Sequences of different EV types are coloured as labelled on the left-hand y-axis. The analysis excluded the more divergent EV76, EV89–91 and other simian-derived types, with the trees rooted using the poliovirus type 3 reference sequence, Leon (GenBank accession number K01392, labelled PV3-OG). Trees were reconstructed from sequential 300-base fragments of the EV-A sequence alignment, incrementing by 30 bases between trees. Groupings are based on clades showing ≥70 % bootstrap support. (b) Correspondence between the phylogeny of sequential fragments through the enterovirus genome with their type assignments, scored from 0 (complete concordance) to 1 (no association) as described previously (Simmonds & Welch, 2006). An enterovirus genome diagram drawn to scale is included to indicate positions where segregation changes occurred. Sequence positions are numbered relative to the poliovirus type 3 Leon sequence outgroup (GenBank accession number K01392).

Fig. 3.. Comparison of pairwise distances in VP1 and 3Dpol between and within recombination groups. Comparison of divergence between VP1 and 3Dpol regions (a) within recombination groups, and (b) between RF-H and other recombination groups. Axes depict nucleotide sequence divergence (uncorrected p distances) in the two genome regions.

Fig. 4.. Inferred timescale for the recent evolution of CVA6. Temporal reconstruction of recombination events in CVA6 using a time-correlated MCMC phylogeny reconstruction of VP1 sequences using the prototype sequence, GenBank accession number AY421764, as an outgroup (not shown). Recombination groups in each lineage are indicated by branch colours. The tree is plotted on a linear timescale, with nodes labelled with inferred dates of lineage splits (in years before present). Grey bars show 95 % highest posterior density intervals for node date estimates.

Fig. 5.. TreeOrder scan of CVA6 complete genome sequences. (a) Tree positions represented on the y-axis of CVA6 genomes with different recombination groups coloured as labelled on the right-hand y-axis. Trees were constructed from sequential 300-base fragments of the CVA6 sequence alignment, incrementing by 30 bases between trees. Groupings are based on clades showing ≥70 % bootstrap support. (b) Segregation scores (see legend to Fig. 2) for sequences classified by recombination group. Sequence positions are numbered relative to the poliovirus outgroup (GenBank accession number K01392).

Fig. 6.. Divergence of RF-A and -G from other recombination groups across the CVA6 genome. Scan of nucleotide sequence divergence (a) between RF-A and descendant recombination groups (RF-E, -F and -H), and (b) between RF-G with RF-A and -H. Plotlines show ratios of divergence (y-axis) in different genome positions (x-axis) to divergence in the VP1 gene (positions 2441–3355). Distances were calculated between sequential 300-base fragments of the CVA6 alignment incrementing by 30 bases between data points.