The role of mutation rate in HIV-1 diversification and adaptation in vivo

Abstract

A characteristic feature of HIV-1 is its ability to develop a diverse viral population that can adapt to hostile environments. The source of this diversity is attributed to the high mutation rate of the virus, often estimated to be around one mutation per 20,000 base pairs per round of replication. However, this estimate has been generated from assays that either use lab-adapted viruses in low-throughput in vitro assays or from deep sequencing of patient viruses, which are subject to fitness cost biases. Using a humanized mice model system, we have previously identified two HIV-1 strains, JR-CSF and REJO.c, that respond differently to treatment with AAV-delivered VRC07 (a broadly neutralizing antibody), with JR-CSF always escaping the antibody and REJO.c only escaping about half of the treatments. By deep sequencing the HIV env genes, we found that JR-CSF diversifies much more than REJO.c, generating more mutations per round of replication across the genome. Because this difference in diversity was not explained by the difference in growth rates between the two viruses, we investigated whether the viruses had different mutation rates. We developed a new method called ERR-Seq (Error Rate of Replication Sequencing), which measures the in vivo mutation rate and mutational profiles of any HIV-1 GagPol on any template sequence, following a single round of lentiviral integration into the host genome. This approach controls for differences in viral replication fitness and detects all possible types of mutations, irrespective of fitness cost. Using ERR-Seq, we profiled 3 HIV-1 strains: NL4-3, JR-CSF, and REJO.c. With over 61 million sequenced bases and 1500 observed mutations in HIV-1 Env, we generated the highest resolution mutation rate profile to date. We found that both NL4-3 and JR-CSF have a greater than 8-fold higher intrinsic mutation rate than REJO.c. We also found that both GagPol genotype and template sequence independently contribute to significant differences in mutation rate between the three viral strains. Finally, we found that the mutation rate profiles of JR-CSF and REJO.c are predictive of the synonymous diversity of replicating JR-CSF and REJO.c in humanized mice, suggesting that the ERR-Seq data is biologically relevant.