High-resolution mapping of evolutionary trajectories in a phage

Abstract

Experimental evolution in rapidly reproducing viruses offers a robust means to infer substitution trajectories during evolution. But with conventional approaches, this inference is limited by how many individual genotypes can be sampled from the population at a time. Low-frequency changes are difficult to detect, potentially rendering early stages of adaptation unobservable. Here we circumvent this using short-read sequencing technology in a fine-grained analysis of polymorphism dynamics in the sentinel organism: a single-stranded DNA phage PhiX174. Nucleotide differences were educed from noise with binomial filtering methods that harnessed quality scores and separate data from brief phage amplifications. Remarkably, a significant degree of variation was observed in all samples including those grown in brief 2-h cultures. Sites previously reported as subject to high-frequency polymorphisms over a course of weeks exhibited monotonic increases in polymorphism frequency within hours in this study. Additionally, even with limitations imposed by the short length of sequencing reads, we were able to observe statistically significant linkage among polymorphic sites in evolved lineages. Additional parallels between replicate lineages were apparent in the sharing of polymorphic sites and in correlated polymorphism frequencies. Missense mutations were more likely to occur than silent mutations. This study offers the first glimpse into "real-time" substitution dynamics and offers a robust conceptual framework for future viral resequencing studies.

Keywords: experimental evolution; next-generation sequencing; substitution analysis.

FIG. 1.—

Analytical schema depicting filters employed (blue) and statistical analyses undertaken (green). All analyses were nonparametric. TP0: time point zero; DRP: deviant read proportion.

FIG. 2.—

Histograms showing the frequencies of genome positions with different deviant read proportions across time point zero samples (A) and experimental samples (B). Gray bars indicate frequencies from significantly polymorphic sites only, and blue bars or portions of bars indicate the additional numbers of positions removed by the binomial filter (see text). Figure 2 is continued on the following page.

FIG. 2.—

Histograms showing the frequencies of genome positions with different deviant read proportions across time point zero samples (A) and experimental samples (B). Gray bars indicate frequencies from significantly polymorphic sites only, and blue bars or portions of bars indicate the additional numbers of positions removed by the binomial filter (see text). Figure 2 is continued on the following page.

FIG. 3.—

Circular histograms of deviant read proportions across the genome in lineages B and C. The height of each spoke represents deviant read proportions at each genome position from 0% (at the inner ring) to the threshold value (for highly polymorphic sites) of ∼1.51% (at the outer ring; for exact frequencies of highly polymorphic sites, see table 1). Colors represent time points with red = time point 1, green = time point 3, and blue = time point 4.

FIG. 4.—

Scatterplots showing the deviant read proportions of polymorphic sites in both lineages at given time points. (A) Highly polymorphic sites detected in one or the other or both samples are shown; (B): significantly polymorphic sites detected only in both samples are shown.

FIG. 5.—

Box and whisker plots showing ranked deviant read proportions associated with different classes of polymorphism in each experimental sample. (A) Transitions and transversions contrasted; (B): missense and silent mutations contrasted. Ranking is carried out within each polymorphism class but between samples.