Presence, persistence and effects of pre-treatment HIV-1 drug resistance variants detected using next generation sequencing: A Retrospective longitudinal study from rural coastal Kenya

Abstract

Background: The epidemiology of HIV-1 drug resistance (HIVDR) determined by Sanger capillary sequencing, has been widely studied. However, much less is known about HIVDR detected using next generation sequencing (NGS) methods. We aimed to determine the presence, persistence and effect of pre-treatment HIVDR variants detected using NGS in HIV-1 infected antiretroviral treatment (ART) naïve participants from rural Coastal Kenya.

Methods: In a retrospective longitudinal study, samples from HIV-1 infected participants collected prior [n = 2 time-points] and after [n = 1 time-point] ART initiation were considered. An ultra-deep amplicon-based NGS assay, calling for nucleotide variants at >2.0% frequency of viral population, was used. Suspected virologic failure (sVF) was defined as a one-off HIV-1 viral load of >1000 copies/ml whilst on ART.

Results: Of the 50 eligible participants, 12 (24.0% [95% CI: 13.1-38.2]) had at least one detectable pre-treatment HIVDR variant against Protease Inhibitors (PIs, n = 6 [12%]), Nucleoside Reverse Transcriptase Inhibitors (NRTIs, n = 4 [8.0%]) and Non-NRTIs (n = 3 [6.0%]). Overall, 15 pre-treatment resistance variants were detected (frequency, range: 2.3-92.0%). A positive correlation was observed between mutation frequency and absolute load for NRTI and/or NNRTI variants (r = 0.761 [p = 0.028]), but not for PI variants (r = -0.117 [p = 0.803]). Participants with pre-treatment NRTI and/or NNRTI resistance had increased odds of sVF (OR = 6.0; 95% CI = 1.0-36.9; p = 0.054).

Conclusions: Using NGS, pre-treatment resistance variants were common, though observed PI variants were unlikely transmitted, but rather probably generated de novo. Even when detected from a low frequency, pre-treatment NRTI and/or NNRTI resistance variants may adversely affect treatment outcomes.

Fig 1

Bar graphs showing overall distribution of [a] HIV-1 drug resistance (HIVDR) variants by time points and drug classes, and [b] genotypic susceptibility of pre-treatment HIVDR variants detected by next generation sequencing from HIV infected antiretroviral-naïve individuals in a rural clinic in Coastal Kenya (n = 50)^$.

Fig 2. Phylogenetic tree of 83 pre-treatment HIV-1 consensus sequences collected at the first and/or second time points from HAART naïve participants (N = 50).

Phylogenetic tree constructed from consensus fasta files by Maximum-likelihood estimation using the GTR model of nucleotide substitution with gamma distributed rate heterogeneity and approximate likelihood ratio test Shimodaira-hasegawa (aLRT-SH) branch support assessment. Local ‘reference’ sequences (taxa not labelled, n = 78) included. Branches leading to nodes with aLRT-SH support of >0.90 and >0.95 are coloured orange and red respectively. Individual sequences are labeled in bold letters ending with 1 (first time point) or 2 (second time point). Sequences with at least one surveillance HIV drug resistance mutation detected at >2% frequency by next generation sequencing are coloured in green. Sequence pairs from the same participant shaded in grey.

Fig 3

Scatter graphs illustrating the distribution and relationship between HIV-1 low frequency pre-treatment drug resistance mutation frequency and mutation load amongst [a] overall number of mutations, and [b] drug-class specific mutations observed from a HIV clinic in rural Kenya (N = 15)*.