Evolution of HIV-1 within untreated individuals and at the population scale in Uganda

Abstract

HIV-1 undergoes multiple rounds of error-prone replication between transmission events, resulting in diverse viral populations within and among individuals. In addition, the virus experiences different selective pressures at multiple levels: during the course of infection, at transmission, and among individuals. Disentangling how these evolutionary forces shape the evolution of the virus at the population scale is important for understanding pathogenesis, how drug- and immune-escape variants are likely to spread in populations, and the development of preventive vaccines. To address this, we deep-sequenced two regions of the HIV-1 genome (p24 and gp41) from 34 longitudinally-sampled untreated individuals from Rakai District in Uganda, infected with subtypes A, D, and inter-subtype recombinants. This dataset substantially increases the availability of HIV-1 sequence data that spans multiple years of untreated infection, in particular for different geographical regions and viral subtypes. In line with previous studies, we estimated an approximately five-fold faster rate of evolution at the within-host compared to the population scale for both synonymous and nonsynonymous substitutions, and for all subtypes. We determined the extent to which this mismatch in evolutionary rates can be explained by the evolution of the virus towards population-level consensus, or the transmission of viruses similar to those that establish infection within individuals. Our findings indicate that both processes are likely to be important.

Linking influenza virus evolution within and between human hosts

“…Under neutral evolution, we expect within-host and global evolutionary rates to be identical, and deviations from these neutral expectations shed light on how selection acts across evolutionary scales. For instance, HIV and hepatitis C virus (HCV) both evolve more rapidly within than between hosts, probably because viruses acquire adaptations to specific hosts that often revert after transmission (Alizon and Fraser, 2013;Gray et al, 2011;Herbeck et al, 2006;Lemey et al, 2006;Lythgoe and Fraser, 2012;Raghwani et al, 2018;Zanini et al, 2015).…”

“…Influenza's global evolutionary rate is easily estimated from phylogenies of patient consensus sequences, which represent transmitted strains (Rambaut et al, 2008), but evolutionary rates during acute infections are more challenging to calculate. Several studies have estimated within-host evolutionary rates for chronic viruses like HIV and HCV by sequencing longitudinal viral samples (Alizon and Fraser, 2013;Gray et al, 2011;Lemey et al, 2006;Raghwani et al, 2018). However, longitudinal samples are difficult to collect for viruses like influenza that cause acute infections, and acute infections also provide limited time for genetic diversity to accumulate.…”

A de novo approach to inferring within-host fitness effects during untreated HIV-1 infection

“…We applied an evolutionary inference method to deep-sequencing data spanning multiple years of infection from 34 untreated individuals living in Rakai, Uganda, enabling us to infer positive selection acting on part of the gp41 region of env (324 base pairs) and the p24 region of gag (387 base pairs) [31].…”

“…We adopt a parsimonious approach, assigning selection to the smallest set of variants required to explain the observed multi-locus sequence data under a likelihood model. Applied to targeted sequence data from a substantial cohort of 34 untreated individuals living in Uganda [31], we determine how selection drives viral evolution. In the presence of pervasive interference between alleles in linkage disequilibrium with one another, our consideration of data from a large number of individuals is fundamental in providing statistical confidence in the assignment of selection.…”

“…Since mutations that enable the virus to evade host immune responses are often costly in terms of viral replication [69,70], sites harbouring escape mutations are likely to evolve towards the population-level consensus when transmitted to a new host since the selective pressure of a specific immune response is removed, but costs associated with the mutation remain [2,31,[71][72][73][74]. This hypothesis is supported by the observation that allelic substitutions during the course of untreated infection occur towards population consensus much more frequently than expected by chance [2,31]. If analysis is restricted to sites where substitutions are observed, this bias is confined mainly to nonsynonymous substitutions [31], which is expected if immune escape mutations are generally nonsynonymous.…”