The Evolutionary Pathway to Virulence of an RNA Virus

Abstract

Paralytic polio once afflicted almost half a million children each year. The attenuated oral polio vaccine (OPV) has enabled world-wide vaccination efforts, which resulted in nearly complete control of the disease. However, poliovirus eradication is hampered globally by epidemics of vaccine-derived polio. Here, we describe a combined theoretical and experimental strategy that describes the molecular events leading from OPV to virulent strains. We discover that similar evolutionary events occur in most epidemics. The mutations and the evolutionary trajectories driving these epidemics are replicated using a simple cell-based experimental setup where the rate of evolution is intentionally accelerated. Furthermore, mutations accumulating during epidemics increase the replication fitness of the virus in cell culture and increase virulence in an animal model. Our study uncovers the evolutionary strategies by which vaccine strains become pathogenic and provides a powerful framework for rational design of safer vaccine strains and for forecasting virulence of viruses. VIDEO ABSTRACT.

Figure 1. Schematic of analyses and experiments performed in the manuscript

Experiments and data are illustrated on the left and computational approaches for data analysis on the right. (A) A phylogenetic analysis of cVDPV consensus sequences from epidemics across the globe was undertaken with the aim of detecting parallel substitutions under selection (in red) as compared to the ancestral OPV2 state (in black). A novel Markov model denoted ParaSel was developed, which assumes that selection will lead to an increase in the rate of substitutions into a certain nucleotide, and decrease in the rate of loss of this nucleotide (exemplified as the “G”). The model then allows calculating the likelihood of the data and performing model choice, followed by inference of specific sites where selection led to parallel substitutions. (B) An experimental evolution approach was used to monitor the emergence of mutations in OPV2 conferring an evolutionary fitness advantage during growth in cell culture at elevated body temperature. (C) Direct assessment of the effect of mutations inferred in (A) and (B) on viral virulence was performed using (i) competition assays between selected mutants and OPV2 at elevated body temperature, and (ii) a mouse model of infection that allowed comparing the virulence of selected mutants versus OPV2. An ABC approach (Methods) was used to infer whether an increase in the frequency of a mutation over time represents significant adaptive evolution. The method compares theoretical simulations of allele frequencies to the empirical data. Accordingly, deleterious alleles are expected to be present at low frequencies throughout the experiment due to purifying selection, whereas adaptive alleles will increase rapidly in frequency. Neutral alleles will accumulate at a rate equal to the mutation rate.

Figure 2. Phylogenetic analysis of cVDPV2 sequences. See also Fig. S1 and Table S1

(A) Maximum-likelihood phylogenetic tree for 424 cVDPV sequences, based on synonymous polymorphisms in the non-recombinant capsid region. The x-axis represents substitutions per site, where 0.011 substitutions are equivalent to one year of viral circulation based on the PV molecular clock (Jorba et al., 2008). Clades were collapsed for visualization. (B) The distribution of the number of independently occurring substitutions inferred across the phylogeny. Transitions are colored in black and transversions are in grey. Notably the transition/transversion ratio is much higher in events mapped two or more times. (C) The ratio of the number of non-synonymous to synonymous substitutions across time shows that strong adaptation occurs early on in the epidemics, which is relaxed later on, and that incomplete purifying selection likely occurs during recent evolution. (A) through (C) all refer to the non-recombinant capsid region.

Figure 3. Results of ParaSel phylogenetic model allow reconstructing the regain of virulence of cVDPVs from independent epidemics across the globe. See also Fig. S2 and Table S3

Seven substitutions and two recombination events are inferred to be under parallel positive selection. (A) The timeline shows the approximate inferred timing of each event, illustrating a sequential process of increase in fitness. Three “waves” of events are observed, inferred based on the estimated fixation time of each event. The relative timing of each event within each wave could not be inferred consistently and is shown based on one of the epidemics (B) A projection of each of the nine events (bottom) onto the phylogenetic tree of the cVDPVs (top). A thin line colored in red/grey/green, corresponding to the colors in panel A, represents a substitution or recombination event under parallel selection present in the sequence corresponding to the leaf of the tree on the top. Blue represents the ancestral OPV2 state, whereas white represents a third alternative. (C) A table summarizing the properties of the substitutions detected by ParaSel, with colors as in (A). Ts refers to transition, whereas Tv refers to transversion. Additional information specifies the location of the substitution on the accepted RNA structure of the 5’ UTR (Andino et al., 1990; Pilipenko et al., 1989), or the amino-acid replacement and its location in the capsid forming genes VP1 through VP4.

Figure 4. Results of in vitro experimental evolution at 33°C and 39.5°C. See also Figs. S3 and S5 and Table S5-S7

Viruses were serially passaged in HeLa cells for seven passages (corresponding to 14 generations) at both temperatures. Passages were sequenced using highly accurate CirSeq sequencing (Acevedo et al., 2014). (A) Four mutations predicted with ParaSel were validated using the experimental evolution approach. Time-series trajectories are shown in solid lines (39.5°C) and dashed lines (33°C), with colors corresponding to those in Fig. 3. For U398C, lack of coverage precluded inferring reliable allele frequencies at 33°C. The grey line in each box represents the neutral allele behavior over time, based on the mean behavior of synonymous mutations in the relevant class of mutation, excluding CpG and UpA sites. (C) Distributions of genome-wide fitness values obtained at 33°C (blue) and 39.5°C (red) show that the 39.5°C distribution is shifted to the right, indicating more adaptation at elevated temperature.

Figure 5. Direct assessment of virulence of mutations. See also Fig. S6

(A) Results of competition assay of the three gate-keeper mutations reveal that all three mutations outcompete the OPV2 strain, on their own or in combination. Mutant frequencies varied between 0.3 and 0.6 at the first passage and were normalized for purpose of presentation to begin at 0.5. (B) Survival analysis of susceptible mice infected with OPV2 gate-keeper mutants. The analysis reveals that the A481G mutation leads to increased virulence as compared to OPV2, with an even stronger effect observed for the combination of all three mutants. Potential epistatic interactions are observed in (A) and (C) for combinations of mutations (see text).

Figure 6. A model summarizing the proposed path/s to virulence of OPV2. See also Table S4

Illustrative fitness landscape of OPV2, which starts off as poorly adapted to replication in human cells at high temperature and is thus illustrated at the bottom of the landscape. Colored arrows (corresponding to Fig. 3) represent substitutions that increase the fitness of the virus, while the grey arrows represent recombination with a HEV-C sequence. Substitutions disrupting CpG/UpA are illustrated by dashed arrows, with potentially several different options leading to the same fitness altitude in the landscape.