Added Value of Next Generation over Sanger Sequencing in Kenyan Youth with Extensive HIV-1 Drug Resistance

Abstract

HIV-1 drug resistance testing in children and adolescents in low-resource settings is both important and challenging. New (more sensitive) drug resistance testing technologies may improve clinical care, but evaluation of their added value is limited. We assessed the potential added value of using next-generation sequencing (NGS) over Sanger sequencing for detecting nucleoside reverse transcriptase inhibitor (NRTI) and nonnucleoside reverse transcriptase inhibitor (NNRTI) drug resistance mutations (DRMs). Participants included 132 treatment-experienced Kenyan children and adolescents with diverse HIV-1 subtypes and with already high levels of drug resistance detected by Sanger sequencing. We examined overall and DRM-specific resistance and its predicted impact on antiretroviral therapy and evaluated the discrepancy between Sanger sequencing and six NGS thresholds (1%, 2%, 5%, 10%, 15%, and 20%). Depending on the NGS threshold, agreement between the two technologies was 62% to 88% for any DRM, 83% to 92% for NRTI DRMs, and 73% to 94% for NNRTI DRMs, with more DRMs detected at low NGS thresholds. NGS identified 96% to 100% of DRMs detected by Sanger sequencing, while Sanger identified 83% to 99% of DRMs detected by NGS. Higher discrepancy between technologies was associated with higher DRM prevalence. Even in this resistance-saturated cohort, 12% of participants had higher, potentially clinically relevant predicted resistance detected only by NGS. These findings, in a young, vulnerable Kenyan population with diverse HIV-1 subtypes and already high resistance levels, suggest potential benefits of more sensitive NGS over existing technology. Good agreement between technologies at high NGS thresholds supports their interchangeable use; however, the significance of DRMs identified at lower thresholds to patient care should be explored further. IMPORTANCE HIV-1 drug resistance in children and adolescents remains a significant problem in countries facing the highest burden of the HIV epidemic. Surveillance of HIV-1 drug resistance in children and adolescents is an important public health strategy, particularly in resource-limited settings, and yet, it is limited due mostly to cost and infrastructure constraints. Whether newer and more sensitive next-generation sequencing (NGS) adds substantial value beyond traditional Sanger sequencing in detecting HIV-1 drug resistance in real life settings remains an open and debatable question. In this paper, we attempt to address this issue by performing a comprehensive comparison of drug resistance identified by Sanger sequencing and six NGS thresholds. We conducted this study in a well-characterized, vulnerable cohort of children and adolescents living with diverse HIV-1 subtypes in Kenya and, importantly, failing antiretroviral therapy (ART) with already extensive drug resistance. Our findings suggest a potential added value of NGS over Sanger even in this unique cohort.

Keywords: HIV-1; Sanger sequencing; drug resistance testing; next-generation sequencing.

FIG 1

Agreement and disagreement between DRMs identified by Sanger and NGS. The figure presents proportions (y axes) of any DRMs (A), NRTI DRMs (B), and NNRTI DRMs (C) that were detected for each participant by Sanger and the six examined NGS thresholds (x axes). Identical DRM profiles identified by both sequencing technologies were considered an agreement. Proportions of individuals with identical and not identical profiles are represented by the colored squares according to the legend. A single binary outcome was assigned to every participant for each NGS threshold. Error bars show lower and higher 95% confidence intervals.

FIG 2

Prevalence of NRTI and NNRTI DRMs identified by Sanger and NGS. The stacked bar figure presents prevalence (y axes) of specific NRTI (A) and NNRTI (B) DRMs that were detected in this study, according to which sequencing technology and NGS threshold they were detected by, as indicated in the legend. Negative values indicate cases identified by Sanger alone.

FIG 3

Discrepancy in detecting specific NRTI DRMs. The heatmap shows specific NRTI DRMs and their proportions (in parenthesis; y axis) according to Sanger sequencing and each of the six NGS thresholds (x axis). The left half of the heatmap indicates DRMs that were detected by NGS among those not detected by Sanger, shown as “NGS+ Sanger–.” The right half of the heatmap indicates DRMs that were not detected by Sanger among those detected by NGS, shown as “Sanger– NGS+.” The intensity of the heatmap colors ranges from 0 to 0.25 according to the scale at the right of the graph. See also Table S3 in the supplemental material.

FIG 4

Discrepancy in detecting specific NNRTI DRMs. The heatmap shows specific NNRTI DRMs and their proportions (in parenthesis; y axis) according to Sanger sequencing and each of the six NGS thresholds (x axis). The left half of the heatmap indicates DRMs that were detected by NGS among those not detected by Sanger, shown as “NGS+ Sanger–.” The right half of the heatmap indicates DRMs that were not detected by Sanger among those detected by NGS, shown as “Sanger– NGS+.” The intensity of the heatmap colors ranges from 0 to 0.50 according to the scale at the right of the graph. See also Table S3.

FIG 5

Discrepancy in the detection of specific DRMs. The figure presents estimates and associated Agresti-Coull 95% confidence intervals of the proportion (x axis) that each NRTI (A) and NNRTI (B) DRM (y axis) was detected by NGS at the 1% threshold, among the Sanger genotypes that did not detect that DRM. The proportion of Sanger genotypes where the DRM was not detected is provided in parentheses. Mutations without a vertical lower bound visible have lower bound coinciding with the estimate of 0. The figure is sorted by the proportion in which the mutation was detected by Sanger (see Fig. 3 and 4).

FIG 6

Predicted resistance scores to future treatment options for Sanger versus NGS. Scatterplots delineate the Sanger (y axes) and the six NGS thresholds (x axes) predicted resistance scores to TDF (top; range, 0 to 100), ETR (middle; range, 0 to 100), and RPV (bottom; range, 0 to 150). The boundaries of susceptible-potential low versus low-intermediate-high predicted resistance scores are highlighted by the green lines within each graph. Agreements in estimating resistance levels between technologies are shown by blue dots, and discrepancies are shown by red dots.

FIG 7

Distribution of study participants according to highest sequence similarly to Sanger by NGS threshold. The plot presents the percentage of participants (y axis) and their 95% bootstrapped confidence intervals that had their highest sequence similarity between Sanger and each of the six examined NGS thresholds (x axis).