Within-Host Viral Diversity: A Window into Viral Evolution

Abstract

The evolutionary dynamics of a virus can differ within hosts and across populations. Studies of within-host evolution provide an important link between experimental studies of virus evolution and large-scale phylodynamic analyses. They can determine the extent to which global processes are recapitulated on local scales and how accurately experimental infections model natural ones. They may also inform epidemiologic models of disease spread and reveal how host-level dynamics contribute to a virus's evolution at a larger scale. Over the last decade, advances in viral sequencing have enabled detailed studies of viral genetic diversity within hosts. I review how within-host diversity is sampled, measured, and expressed, and how comparative studies of viral diversity can be leveraged to elucidate a virus's evolutionary dynamics. These concepts are illustrated with detailed reviews of recent research on the within-host evolution of influenza virus, dengue virus, and cytomegalovirus.

Keywords: diversity; evolution; models; quasispecies; sequencing.

Figure 1

Within-host diversity and virus evolution. (a) Advances in sequencing technology have revolutionized the study of within-host viral diversity. The type and frequency of mutations (red letters) in a population are easily obtained from NGS (b) Viral diversity can be measured and compared using a variety of metrics. Ideally, these metrics capture the varying impact of mutations present at high (blue bars) vs. low (yellow bars) frequency. (c) Differences in diversity across tissues (top, different colored viruses) or changes over time (bottom) can be used to model within-host viral dynamics due to selection and genetic drift. (d) Studies of within-host viral diversity have provided insights into the evolution of influenza virus, Dengue virus, and cytomegalovirus and the extent to which within-host virus populations can be accurately described as “quasispecies” or “mutant clouds.”

Figure 2

Viral sequencing and diversity metrics. (a) Due to the negative impact of most mutations, the vast majority of sequence variants are relatively rare. The sensitivity and specificity of NGS for rare variant detection is highly dependent on the number of genomes sequenced and largely independent of read depth. Sensitivity drops at lower inputs, because rare variants can only be detected if the population is completely sampled. Smaller populations require more amplification, which propagates RT-PCR error. As a result, variants identified in low input populations are more likely to be false positives than true positives. (b) The diversity of two populations (different color viruses) expressed as richness, the number of variants or genotypes; evenness, the relative abundances of each variant in the population; and the site frequency spectrum, the numbers of different mutants and their respective frequencies. Both have a richness of 10 genotypes. The even population has equal numbers (n=4) of each genotype, and the 10 genotypes are present at a frequency of 0.1 (grey bars). The uneven population has the same 10 genotypes, but one is present at a frequency of 0.5 (black), one at 0.1 (red), and one at 0.05 (blue). The rest are singletons (summed as grey bars). (c) Shannon entropy at sites across a hypothetical 10 kilobase viral genome for two populations. In the genome, the noncoding regions are represented as lines and two different reading frames on the coding region are represented as boxes. The Shannon entropy at polymorphic sites for the two populations are shown as red and blue bars. Nonpolymorphic sites have an entropy of zero.

Figure 3

Using within-host data to elucidate evolutionary dynamics. (a) The frequency of a mutation can increase or decrease due to selective and nonselective processes. Positive selection will increase the frequency of a beneficial mutation (e.g., a mutation that leads to antibody escape) and will decrease the frequency of a detrimental mutation (e.g., a surface protein mutant that can no longer bind its receptor). A neutral mutation (blue) can increase in frequency if it is linked to a beneficial mutation on the same genome (hitchhiking) or decrease in frequency if it is linked to a detrimental one (background selection). Genetic drift is a change in a mutation’s frequency due to stochastic processes. It typically occurs in small populations due to random sampling. Genetic drift also occurs during population bottlenecks or splits followed by expansions. (b) Criteria for identifying within-host variants potentially under positive selection. From left to right: enrichment of high frequency variants in viral proteins or certain protein domains, shown here as nonsynonymous single nucleotide variants (SNV) in antigenic sites (red) modeled on the structure of the influenza hemagglutinin trimer; identification of a variant already known to mediate a certain phenotype, such as antibody escape or receptor binding; identification of the same mutation in multiple individual hosts not linked by transmission; identification of a within-host variant that is subsequently seen in different host populations or at different evolutionary scales.

Figure 4

Quasispecies diversity and virulence. (a) Diverse RNA virus populations are often referred to as quasispecies, mutant swarms, or mutant clouds. According to viral quasispecies theory, diversity is a determinant of viral phenotype and virulence. Here, high mutation rates lead to a diverse population, and the mutant spectrum is optimized by natural selection. Maintenance of a diverse population within hosts enables rapid adaptation and ultimately virulence. (b) An alternative model. Recent work from my laboratory suggests that the evolution of RNA dependent RNA polymerases is shaped by a speed-fidelity trade-off. Selection favors faster polymerases and faster polymerases are inherently more error-prone. Faster replication also leads to more rapid within-host spread and increased virulence. Empiric data across viral systems demonstrates that within-host diversity is rapidly lost or partitioned due to purifying selection, bottlenecks, and population splits. Thus, within-host diversity is neither maintained nor optimized by selection.