Many microbial populations rapidly adapt to changing environments with multiple variants competing for survival. To quantify such complex evolutionary dynamics in vivo, time resolved and genome wide data including rare variants are essential. We performed whole-genome deep sequencing of HIV-1 populations in 9 untreated patients, with 6-12 longitudinal samples per patient spanning 5-8 years of infection. The data can be accessed and explored via an interactive web application. We show that patterns of minor diversity are reproducible between patients and mirror global HIV-1 diversity, suggesting a universal landscape of fitness costs that control diversity. Reversions towards the ancestral HIV-1 sequence are observed throughout infection and account for almost one third of all sequence changes. Reversion rates depend strongly on conservation. Frequent recombination limits linkage disequilibrium to about 100bp in most of the genome, but strong hitch-hiking due to short range linkage limits diversity.DOI: http://dx.doi.org/10.7554/eLife.11282.001

HIV-1 infection cannot be cured because the virus persists as integrated proviral DNA in long-lived cells despite years of suppressive antiretroviral therapy (ART). In a previous paper (Zanini et al, 2015) we documented HIV-1 evolution in 10 untreated patients. Here we characterize establishment, turnover, and evolution of viral DNA reservoirs in the same patients after 3–18 years of suppressive ART. A median of 14% (range 0–42%) of the DNA sequences were defective due to G-to-A hypermutation. Remaining DNA sequences showed no evidence of evolution over years of suppressive ART. Most sequences from the DNA reservoirs were very similar to viruses actively replicating in plasma (RNA sequences) shortly before start of ART. The results do not support persistent HIV-1 replication as a mechanism to maintain the HIV-1 reservoir during suppressive therapy. Rather, the data indicate that DNA variants are turning over as long as patients are untreated and that suppressive ART halts this turnover.DOI: http://dx.doi.org/10.7554/eLife.18889.001

Mutation rates and fitness costs of deleterious mutations are difficult to measure in vivo but essential for a quantitative understanding of evolution. Using whole genome deep sequencing data from longitudinal samples during untreated HIV-1 infection, we estimated mutation rates and fitness costs in HIV-1 from the dynamics of genetic variation. At approximately neutral sites, mutations accumulate with a rate of 1.2 × 10−5 per site per day, in agreement with the rate measured in cell cultures. We estimated the rate from G to A to be the largest, followed by the other transitions C to T, T to C, and A to G, while transversions are less frequent. At other sites, mutations tend to reduce virus replication. We estimated the fitness cost of mutations at every site in the HIV-1 genome using a model of mutation selection balance. About half of all non-synonymous mutations have large fitness costs (>10 percent), while most synonymous mutations have costs <1 percent. The cost of synonymous mutations is especially low in most of pol where we could not detect measurable costs for the majority of synonymous mutations. In contrast, we find high costs for synonymous mutations in important RNA structures and regulatory regions. The intra-patient fitness cost estimates are consistent across multiple patients, indicating that the deleterious part of the fitness landscape is universal and explains a large fraction of global HIV-1 group M diversity.

BackgroundUltra-deep pyrosequencing (UDPS) is used to identify rare sequence variants. The sequence depth is influenced by several factors including the error frequency of PCR and UDPS. This study investigated the characteristics and source of errors in raw and cleaned UDPS data.ResultsUDPS of a 167-nucleotide fragment of the HIV-1 SG3Δenv plasmid was performed on the Roche/454 platform. The plasmid was diluted to one copy, PCR amplified and subjected to bidirectional UDPS on three occasions. The dataset consisted of 47,693 UDPS reads. Raw UDPS data had an average error frequency of 0.30% per nucleotide site. Most errors were insertions and deletions in homopolymeric regions. We used a cleaning strategy that removed almost all indel errors, but had little effect on substitution errors, which reduced the error frequency to 0.056% per nucleotide. In cleaned data the error frequency was similar in homopolymeric and non-homopolymeric regions, but varied considerably across sites. These site-specific error frequencies were moderately, but still significantly, correlated between runs (r = 0.15–0.65) and between forward and reverse sequencing directions within runs (r = 0.33–0.65). Furthermore, transition errors were 48-times more common than transversion errors (0.052% vs. 0.001%; p<0.0001). Collectively the results indicate that a considerable proportion of the sequencing errors that remained after data cleaning were generated during the PCR that preceded UDPS.ConclusionsA majority of the sequencing errors that remained after data cleaning were introduced by PCR prior to sequencing, which means that they will be independent of platform used for next-generation sequencing. The transition vs. transversion error bias in cleaned UDPS data will influence the detection limits of rare mutations and sequence variants.

Deep sequencing is a powerful and cost-effective tool to characterize the genetic diversity and evolution of virus populations. While modern sequencing instruments readily cover viral genomes many thousand fold and very rare variants can in principle be detected, sequencing errors, amplification biases, and other artifacts can limit sensitivity and complicate data interpretation. For this reason, the number of studies using whole genome deep sequencing to characterize viral quasi-species in clinical samples is still limited. We have previously undertaken a large scale whole genome deep sequencing study of HIV-1 populations. Here we discuss the challenges, error profiles, control experiments, and computational test we developed to quantify the accuracy of variant frequency estimation.