Widespread interspecific phylogenetic tree incongruence between mosquito-borne and insect-specific flaviviruses at hotspots originally identified in Zika virus

Abstract

Intraspecies (homologous) phylogenetic incongruence, or 'tree conflict' between different loci within the same genome of mosquito-borne flaviviruses (MBFV), was first identified in dengue virus (DENV) and subsequently in Japanese encephalitis virus (JEV), St Louis encephalitis virus, and Zika virus (ZIKV). Recently, the first evidence of phylogenetic incongruence between interspecific members of the MBFV was reported in ZIKV and its close relative, Spondweni virus. Uniquely, these hybrid proteomes were derived from four incongruent trees involving an Aedes-associated DENV node (1 tree) and three different Culex-associated flavivirus nodes (3 trees). This analysis has now been extended across a wider spectrum of viruses within the MBFV lineage targeting the breakpoints between phylogenetic incongruent loci originally identified in ZIKV. Interspecies phylogenetic incongruence at these breakpoints was identified in 10 of 50 viruses within the MBFV lineage, representing emergent Aedes and Culex-associated viruses including JEV, West Nile virus, yellow fever virus, and insect-specific viruses. Thus, interspecies phylogenetic incongruence is widespread amongst the flaviviruses and is robustly associated with the specific breakpoints that coincide with the interspecific phylogenetic incongruence previously identified, inferring they are 'hotspots'. The incongruence amongst the emergent MBFV group was restricted to viruses within their respective associated epidemiological boundaries. This MBFV group was RY-coded at the third codon position ('wobble codon') to remove transition saturation. The resulting 'wobble codon' trees presented a single topology for the entire genome that lacked any robust evidence of phylogenetic incongruence between loci. Phylogenetic interspecific incongruence was therefore observed for exactly the same loci between amino acid and the RY-coded 'wobble codon' alignments and this incongruence represented either a major part, or the entire genomes. Maximum likelihood codon analysis revealed positive selection for the incongruent lineages. Positive selection could result in the same locus producing two opposing trees. These analyses for the clinically important MBFV suggest that robust interspecific phylogenetic incongruence resulted from amino acid selection. Convergent or parallel evolutions are evolutionary processes that would explain the observation, whilst interspecific recombination is unlikely.

Keywords: Mosquito-borne viruses; RY-coding; flavivirus; phylogenetic incongruence; recombination; selection.

Figure 1.

Schematic representation of the recombination breakpoints and resulting seven protein loci across the MBFV amino acid genome, viz. prM, E, NS1–NS3 HELICc, NS2A–NS3 DEAD, C-terminal NS3–NS4A, NS4B and C–NS5, identified in the interspecific recombination events between ZIKV–SPOV (DENV Aedes-associated clade) with the Culex I and II-associated MBFV resulting in a chimeric genome. (Supplementary Information 1). Combined locus pairs (NS1–NS3 HELICc and C–NS5) were identical in topology using the SOWHAT test prior amalgamation. The predominant immunological associations are indicated in the diagram below after the recombination breakpoints were robustly established for both contiguous and combined loci. Note, C-protein and NS5-protein are juxtaposed during replication due to genome circularisation.

Figure 2.

The analytical approach to genome-wide interspecific recombination using ‘data mining’ via non-parametric bootstrap analysis and then the SOWHAT test followed by Bayesian analysis between the MBFV described for the seven loci described in Fig. 1. The’data mining’ method and SOWHAT amino acid alignment datasets, differ in the number of taxa in the analysis, but not in interspecific genetic diversity. Note, the trees used for SOWHAT analysis are zero constrained.

Figure 3.

Amino acid maximum likelihood network consensus phylogeny (Dendroscope 3) between the E-protein and C-NS5 protein phylogenies where differences in tree branching order (incongruence) between the E-protein and C-NS5 trees are represented by curved vertical lines that support ‘8 floating horizontal lines’ representing the MBFV members of interest to this study. ‘Trees 1–8’ denote the eight interspecific recombination events, better defined in Table 1.

Figure 4.

An extensive programme of phylogenetic analyses, involving 50 recognized MBFV lineage members was instigated to look for evidence of widespread interspecific phylogenetic incongruence amongst the MBFV. The figures present amino acid maximum likelihood analyses and posterior probabilities of phylogenetic incongruence for all non-ZIKV MBFV and ISFV using the breakpoints for ZIKV described recently (Gaunt et al. 2020). Numbers above each node are percentage bootstrap values which in most cases are above 75 per cent. The order of the bootstraps and posterior probabilities along each internal branch represents the separate protein trees that identify the protein loci defined immediately above each tree. For example, in Fig 4a i, the left-hand tree represents the E-protein, whilst the right-hand mirror tree depicts separate protein alignments for NS1–NS3 HELICc (bootstrap ‘a’), NS2A–NS3 DEAD (bootstrap ‘b’) and NS4A–NS3 HELICc (bootstrap ‘c’) and the respective bootstrap scores a, b,c:, are indicated above the major branches followed by the posterior probabilities a; b;c. This protocol is adopted for all trees presented in Fig 4. The ‘Bat MBFV’ in Fig. 4c refers to bat-borne flaviviruses within the MBFV lineage.

Figure 5.

Schematic diagram of the MBFV lineage genome showing phylogenetic incongruence using the SOWHAT test (Tables 2 and 3), amino acid Bayesian posterior probabilities (models 1–6) and first and second codon trees for SLEV and YFV group alignments (Methods). NS3 HEL§ refers to NS3 HELICc and NS4A* refers to the C-terminal NS3 HELICc as well as NS4A. The line below each genome denotes the circularisation during replication connecting C to NS5. All non-parametric bootstraps and posterior probabilities are available on request.

Figure 6.

RY-coded wobble codon maximum likelihood trees from concatenated loci and their comparison to single locus amino acid maximum likelihood and Bayesian trees. Note the Bayesian MCMC would not converge for the RY-codon ‘wobble codon data’ but did achieve convergence for non-RY-coded, nucleotide data (Fig. 5).