Deep-sequencing of viral genomes from a large and diverse cohort of treatment-naive HIV-infected persons shows associations between intrahost genetic diversity and viral load

Abstract

Background: Infection with human immunodeficiency virus type 1 (HIV) typically results from transmission of a small and genetically uniform viral population. Following transmission, the virus population becomes more diverse because of recombination and acquired mutations through genetic drift and selection. Viral intrahost genetic diversity remains a major obstacle to the cure of HIV; however, the association between intrahost diversity and disease progression markers has not been investigated in large and diverse cohorts for which the majority of the genome has been deep-sequenced. Viral load (VL) is a key progression marker and understanding of its relationship to viral intrahost genetic diversity could help design future strategies for HIV monitoring and treatment.

Methods: We analysed deep-sequenced viral genomes from 2,650 treatment-naive HIV-infected persons to measure the intrahost genetic diversity of 2,447 genomic codon positions as calculated by Shannon entropy. We tested for associations between VL and amino acid (AA) entropy accounting for sex, age, race, duration of infection, and HIV population structure.

Results: We confirmed that the intrahost genetic diversity is highest in the env gene. Furthermore, we showed that mean Shannon entropy is significantly associated with VL, especially in infections of >24 months duration. We identified 16 significant associations between VL (p-value<2.0x10-5) and Shannon entropy at AA positions which in our association analysis explained 13% of the variance in VL. Finally, equivalent analysis based on variation in HIV consensus sequences explained only 2% of VL variance.

Conclusions: Our results elucidate that viral intrahost genetic diversity is associated with VL and could be used as a better disease progression marker than HIV consensus sequence variants, especially in infections of longer duration. We emphasize that viral intrahost diversity should be considered when studying viral genomes and infection outcomes.

Trial registration: Samples included in this study were derived from participants who consented in the clinical trial, START (NCT00867048) (23), run by the International Network for Strategic Initiatives in Global HIV Trials (INSIGHT). All the participant sites are listed here: http://www.insight-trials.org/start/my_phpscript/participating.php?by=site.

Fig 1

Mean Shannon entropy over the HIV genome (top) and number of samples with minimum 500-fold genome sequencing depth over the genome (bottom). A total of 2,447 HIV AA positions are shown on the x-axis. Gag positions 1–239, Pol positions 981–1004, Rev positions 1–6 and Tat positions 1–51 are not shown, as they were not covered by the amplicons. Red horizontal line in the top figure shows the used Shannon entropy threshold. Grey points at the top plot mark the HIV genome positions which were excluded from the analysis (≤800 samples). Grey horizontal line at the bottom plot shows the threshold for the minimum number of samples.

Fig 2. The relationship between sample’s mean Shannon entropy, viral load (VL) and infection duration.

Fig 3. Manhattan plot of the associations between viral load (VL) and Shannon entropy at AA positions projected on the HIV genome.

Red horizontal line marks the Bonferroni p-value threshold (p = 2.0×10⁻⁵). The x-axis denotes gene AA positions. A total of 2,447 HIV AA positions are shown on the x-axis. Gag positions 1–239, Pol positions 981–1004, Rev positions 1–6 and Tat positions 1–51 are not shown, as they were not covered by the amplicons.