https://orcid.org/0000-0002-5570-1010
Figures
Abstract
HIV-1 drug resistance remains a major challenge to treatment and control efforts, particularly in sub-Saharan Africa (sSA). However, standard resistance genotyping does not adequately capture linked drug resistance mutations within the viral quasispecies that may influence virologic failure (VF). We used a next-generation sequencing primer ID (NGS-Primer ID) assay to characterize linked HIV-1 drug resistance mutations in plasma from participants in the Resistance Testing to Improve Management of Virologic Failure (REVAMP) study who had detectable viremia on first-line non-nucleoside reverse transcriptase inhibitor (NNRTI)-based antiretroviral therapy (ART) and were maintained on NNRTI-based regimens. For each participant, we calculated a weighted genotypic susceptibility score (wGSS) based on the GSS of each reported pattern and its sequence-supported frequency within the sample. Plasma specimens from 108 participants were sequenced. Sanger sequencing showed a median GSS of 1.0 (IQR, 1.0-2.0), whereas NGS-Primer ID identified a median of 10 distinct resistance patterns per participant (IQR, 5–17), with 67% (IQR, 8–92) of reported DRM-pattern frequency within participants corresponding to linked dual-class mutations. A broad spectrum of drug-susceptible and resistant- DRM patterns were observed, with the median difference between the highest and lowest pattern-specific GSS being 2.0 (2.0-2.75). Within each participant, a median of 10% of reported patterns showed less resistance, while 19% showed more resistance compared to that predicted by Sanger sequencing. Individuals with persistent VF had significantly lower wGSS at study entry than those who later achieved virologic re-suppression (median, 1.3 vs 2.1; p < 0.001). NGS-Primer ID revealed substantial intra-host diversity in linked HIV-1 drug resistance patterns that was not captured by conventional Sanger sequencing. Incorporating linked resistance patterns into susceptibility assessment may improve prediction of subsequent virologic failure.
Author summary
HIV-1 drug resistance remains a major challenge to treatment, especially in sSA, where treatment options may be limited. The clinical impact of linked resistance mutations on the same viral genome remains poorly understood because standard sequencing methods cannot fully resolve them. We used an NGS-Primer ID assay to detect linked HIV drug resistance mutations in people with HIV-1 subtype C who were failing a first-line non-nucleoside reverse transcriptase inhibitor (NNRTI)-based regimen in KwaZulu-Natal, South Africa. We found extensive diversity in drug resistance patterns both within and across individuals that was not captured by Sanger sequencing. We also found that a weighted genotypic susceptibility score based on linked variants at study entry differed significantly between individuals who later achieved virologic suppression and those with persistent virologic failure.
Citation: Pillay M, Choudhary MC, Deo R, Naumenko S, Tamura TJ, Etemad B, et al. (2026) Extensive diversity and impact of drug-resistant HIV-1 variants in individuals with prior virologic failure. PLoS Pathog 22(5): e1014118. https://doi.org/10.1371/journal.ppat.1014118
Editor: Mary F. Kearney, National Cancer Institute, UNITED STATES OF AMERICA
Received: August 11, 2025; Accepted: March 25, 2026; Published: May 12, 2026
Copyright: © 2026 Pillay et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: Sequencing data generated in this study have been deposited in the NCBI Sequence Read Archive (SRA) under BioProject accession number PRJNA1433141. The bioinformatics pipeline used for sequence processing and drug resistance analysis is available on GitHub at https://github.com/rinkideo/hivdrm.
Funding: MJS received funding from the National Institutes of Health (R01AI124718). JZL received funding from the National Institutes of Health (AI138801 and AI060354). VM received support from the Emory Center for AIDS Research through the National Institutes of Health (P30AI050409) for work related to this manuscript. MCC received salary support from the National Institutes of Health through grant AI138801. Funder website: https://www.nih.gov/. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: I have read the journal’s policy and the authors of this manuscript have the following competing interests: JZL has consulted for AbbVie, Merck, and IMUNON. VM has received investigator-initiated research grants to the institution and research support from Eli Lilly, Bayer, Gilead Sciences, Merck, and ViiV Healthcare. SMC has received research grants from Merck and ViiV Healthcare for investigator-initiated studies. The other authors have declared that no competing interests exist.
Introduction
The World Health Organization estimated that more than 40.8 million people were living with HIV in 2024, with nearly two-thirds residing in sub-Saharan Africa (sSA) [1,2]. South Africa remains heavily affected by the HIV epidemic, particularly in KwaZulu-Natal (KZN), which has one of the highest adult HIV prevalences in the country [3,4]. Although South Africa has made major gains in ART scale-up, with over 5.9 million individuals receiving treatment by the end of 2023, ongoing virologic failure and increasing drug resistance continue to threaten these gains [2,5–7]. In response to rising NNRTI resistance, South Africa transitioned in 2019 to tenofovir/lamivudine/dolutegravir (TLD) as the preferred first- and second-line regimen [8].
HIV-1 drug resistance remains a major barrier to durable virologic suppression [9–12]. Because HIV replicates rapidly and with low polymerase fidelity, it exists within an individual as a genetically diverse viral quasispecies [13–15]. This within-host diversity may influence treatment outcome, particularly when drug-resistant minority variants (DRMVs), defined here as variants comprising <20% of the viral population, are present [16]. However, the clinical significance of DRMVs has been inconsistent across studies, with data from sub-Saharan Africa differing from findings in the United States and Europe [17–20]. Differences in viral subtype, adherence, and treatment context may contribute to these discrepancies [21,22].
Beyond the presence of individual drug resistance mutations, the genomic linkage of resistance mutations may also be clinically important. Prior studies suggest that virologic failure is more likely when mutations conferring resistance to more than one drug class are linked on the same viral genome than when they are present on separate variants [23–27]. However, data on linked dual-class resistance and its clinical relevance remain limited, particularly in African populations.
Standard Sanger sequencing does not reliably detect low-frequency variants and cannot determine whether resistance mutations are linked on the same viral genome, while conventional next-generation sequencing improves sensitivity but does not fully resolve linkage [28,29]. To address this limitation, we used an NGS-Primer ID assay, which labels individual cDNA molecules prior to amplification and reduces PCR-associated error and recombination [23,25,27]. We applied this approach to participants in the REVAMP trial with virologic failure on first-line non-nucleoside reverse transcriptase inhibitor (NNRTI)-based ART to characterize linked HIV-1 drug resistance patterns and assess their association with subsequent treatment outcome.
Results
Participant characteristics
A total of 108 participants failing a first-line NNRTI-based regimen consisting of TDF, FTC and EFV were included in this analysis (Table 1). Of the 108, 52% were female (52%), and the median age was 38 years (interquartile range [IQR] 31–44). The median baseline HIV-1 viral load at the time of VF was 4.1 log10 RNA copies/mL (IQR 3.6-4.8), with a median Sanger sequencing genotypic susceptibility score (GSS) of 1.0 (IQR 1.0-2.0). All participants had subtype C virus as determined by the Stanford HIV Drug Resistance Database, with phylogenetic analysis providing supportive evidence (S1 Fig).
NGS-Primer ID uncovers remarkable diversity of HIV drug-resistance patterns
In this study, NGS-Primer ID was used to characterize the diversity of reported DRM patterns within each sample, which provides a far more detailed view of HIV drug resistance mutations than Sanger sequencing or traditional NGS, as illustrated in S2 Fig. NGS-Primer ID revealed a remarkable diversity of HIV-1 drug-resistance patterns, with a broad array of linked resistance mutations (Figs 1A; S3). The multiple colors within each stacked bar plot represent the relative proportions of reported DRM-containing and non-DRM patterns in each participant, including patterns harboring multiple linked resistance mutations. Reported DRM patterns present at frequencies ≤5% of total input sequences within a sample are grouped and shown in dark grey. Approximately one-third of participants harbored extreme diversity, with very low-frequency reported DRM patterns (each present at <5% of total input sequences within a sample) accounting for a substantial proportion of the reported patterns in some participants. A diverse range of mutational patterns as observed in Fig 1A were also observed across participants (S3 Fig). In this study, participants had a median of 10 [5, 17] different reported patterns, and 67% [8%, 92%] of the reported DRM-pattern frequency within an individual corresponded to dual-class resistance mutations (Fig 1B). Reported patterns with dual-class resistance mutations were present at significantly higher frequencies than reported DRM patterns harboring single-class resistance mutations (median 67.4% [IQR 7.5–92.0] vs 9.55% [IQR 1.9–37.9]; Wilcoxon rank-sum test p < 0.001; Fig 1B).
(A) A representative subgroup of participants is shown. Each stacked bar plot corresponds to an individual participant, and the colors represent specific mutational patterns. Pattern frequencies were calculated from the number of sequences assigned to each pattern relative to the total number of input sequences in the sample. Reported DRM patterns present at frequencies ≤5% are shown in dark grey. (B) Box plots showing the proportion of reported DRM-pattern frequency within participants with either dual (NRTI & NNRTI) or single drug class resistance mutations. Each dot represents a participant, the horizontal line within a box represents the median, the box represents the interquartile range, and the whiskers represent the data range excluding outliers. VF = virologic failure, VS = virologic suppression.
We calculated the genotypic susceptibility score (GSS) for each reported pattern and uncovered a diverse range of GSS values within each individual. This significant intra-host variation was not detected by conventional Sanger sequencing, highlighting the increased sensitivity of our single-genome resolution approach (Fig 2). We observed co-occurrence of both resistant and susceptible reported patterns within all 108 participants, as illustrated in Fig 2.
Each row represents a unique study participant (y-axis) labeled with participant identification numbers (PIDs). The columns represent the cumulative genotypic susceptibility (GSS) values (x-axis), with lighter colors indicating the proportion of sequences assigned to patterns closer to 0%, and darker red shades represents closer to 100%. Sanger Sequencing GSS is denoted by dotted boxes. The first two heatmaps (A & B) show participants with virologic failure (VF) and the third heatmap (C) shows participants with virologic suppression (VS).
Sanger sequencing can underestimate or overestimate the level of drug resistance present within the viral quasispecies
We next examined variants for which GSS values differed between Sanger sequencing and NGS–Primer ID. Nearly 30% of all reported patterns showed discordant GSS values when assessed by NGS–Primer ID (Fig 3). At the participant level, NGS–Primer ID identified both increased and decreased levels of resistance compared with Sanger sequencing. Within each participant, a median of 10.2% of reported patterns exhibited higher GSS values (indicating greater drug susceptibility) than predicted by Sanger sequencing, whereas a median of 19.1% of reported patterns showed lower GSS values (indicating greater resistance). Across participants, there was substantial within-host heterogeneity, with a median range of pattern-specific GSS values (difference between the highest and lowest GSS) of 2.0 (IQR 2.0–2.75).
Each column (x-axis) represents a different participant as shown by unique participant identification numbers (PIDs), and the y-axis shows the proportion of sequences assigned to reported patterns by NGS-Primer ID that had a higher GSS (Green lollipops showing GSS > 0, indicating less resistance;) or lower GSS (Blue lollipops showing GSS < 0, indicating greater resistance) when compared to Sanger sequencing. The dashed lines represent the median proportion of variants identified by NGS-Primer ID with higher GSS than Sanger or lower GSS than Sanger.
Representative data from two participants exemplify the diverse mutational patterns observed across the entire 108-participant cohort. Detailed analysis of these select individuals provides further evidence that Sanger sequencing often misrepresents the level of resistance, either underestimating (Fig 4) or overestimating (Fig 5) the actual viral susceptibility present within a clinical specimen. In PID 219, Sanger sequencing identified a genotypic susceptibility score (GSS) of 3, implying a fully active ART regimen (Fig 4A). In contrast, the NGS-Primer ID platform identified 14 distinct mutational patterns, including multiple variants possessing drug resistance (Figs 4B; S4). This high-resolution analysis confirmed that while 69.5% of variants remained fully susceptible with a GSS of 3, 30.4% of the viral variants demonstrated measurable drug resistance. This subset included 21.2% of variants with a GSS of 2 and 9.2% of variants harboring both K103N and M184V mutations (Fig 4B; 4C), resulting in a GSS of 1 for those specific viral genomes (Fig 4C). These findings illustrate the limitations of Sanger sequencing in resolving the true diversity and extent of resistant variants within a viral quasispecies. Furthermore, our data suggest that Sanger sequencing can misrepresent the population by either underestimating or overestimating the resistance burden present in the quasispecies.
(A) Heatmap showing variant-level genotypic susceptibility scores (GSS) identified by NGS-Primer ID. The non-dashed boxes indicate the distribution of NGS-Primer ID-derived GSS values across variants in this participant. (B) Pie chart showing the distribution of mutational patterns detected in participant 219. (C) Histogram showing the distribution of reported patterns across GSS categories.
(A) Heatmap showing the distribution of variant-level genotypic susceptibility scores (GSS) identified by NGS-Primer ID, together with the corresponding Sanger sequencing GSS. (B) Pie chart showing the distribution of distinct mutational patterns detected in participant 387. (C) Histogram showing the distribution of reported patterns across GSS categories.
For PID 387, Sanger sequencing yielded high-level resistance against all ARVs, leading to a GSS = 0.25 (Fig 5A). However, NGS-Primer ID identified extremely diverse mutational patterns (i.e., 18 different mutational patterns) present within the quasispecies of this participant (Figs 5B; S5), with only 3.2% of variants actually harboring resistance to all ARVs, i.e., GSS = 0.25 (Fig 5C). Approximately 83% of variants had a GSS ≥ 1 and 48% had a GSS > 2. Together, these results demonstrate that NGS-Primer ID provides a more granular view of the viral quasispecies and reveals how Sanger sequencing can underestimate or overestimate the levels of drug resistance present within an individual. The mutational patterns and GSS obtained from Sanger Sequencing for all participants are shown in S1 Table.
Weighted genotypic susceptibility score by NGS-Primer ID distinguishes individuals with eventual virologic failure on continued NNRTI-based ART
The weighted GSS (wGSS) was calculated by taking the average GSS across all reported patterns identified by NGS-Primer ID at the time of study entry to predict subsequent virologic suppression with continued NNRTI-based ART. Participants with persistent virologic failure (VF) had lower wGSS at study entry than those who achieved virologic suppression (median 1.3 [IQR 1.0–2.0] vs 2.1 [IQR 1.3–2.8]; Wilcoxon rank-sum test p < 0.001; Fig 6A). In contrast, the use of GSS by Sanger sequencing at the time of study entry did not demonstrate any significant differences between individuals who eventually had VF or re-suppression (Fig 6B).
(A) Box plots comparing weighted genotypic susceptibility scores (wGSS) derived from NGS-Primer ID at study entry between participants with subsequent virologic failure (VF) and those with virologic suppression (VS). (B) Box plots comparing Sanger-derived genotypic susceptibility scores (GSS) at study entry between participants with subsequent virologic failure (VF) and those with virologic suppression (VS). In both panels, the center line indicates the median, the box indicates the interquartile range, and the whiskers indicate the data range excluding outliers. wGSS = weighted GSS; VS = Virologic Suppression; VF = Virologic Failure.
Discussion
In this study, NGS-Primer ID revealed substantial within-host diversity in linked HIV-1 drug resistance patterns among participants with virologic failure on first-line NNRTI-based ART. Compared with Sanger sequencing, this approach identified a much more complex mixture of susceptible and resistant variants within individual participants, including frequent dual-class linked resistance mutations. Variant-specific genotypic susceptibility scores varied widely within participants, and nearly 30% of variants showed GSS values that differed from those predicted by Sanger sequencing. Importantly, a weighted genotypic susceptibility score derived from linked variants at study entry distinguished participants with persistent virologic failure from those who later achieved virologic suppression, whereas Sanger-based GSS did not.
These findings help explain why standard resistance genotyping may not always reflect subsequent clinical outcome in individuals with virologic failure. Sanger sequencing provides a consensus-level view of the viral population and does not reliably detect low-frequency variants or determine whether resistance mutations are linked on the same viral genome. As a result, clinically relevant within-host heterogeneity may be missed, leading to an incomplete picture of overall drug susceptibility within a sample [28,30,31]. This interpretation is consistent with prior work showing that conventional resistance testing does not always align with treatment outcome, particularly in resource-limited settings and in individuals with complex resistance patterns [32–37]. In our cohort, this was evident in both directions, with Sanger sequencing underestimating resistance in some participants and overestimating it in others. The participant-level examples further show that consensus genotyping may not reflect the relative distribution of susceptible and resistant variants within a specimen.
Our results also extend previous work on DRMVs by showing that the genomic context in which resistance mutations occur may be as important as their presence alone. Studies examining the relationship between DRMVs and virologic failure have reported inconsistent findings across geographic and treatment settings [10,17,19–22,31,38,39]. One likely explanation is that assays focused only on the presence or frequency of individual mutations cannot determine whether mutations affecting more than one drug class are linked on the same viral genome. In our cohort, participants had a median of 10 distinct resistance patterns, and dual-class resistance variants were common and often present at higher frequencies than single-class resistant variants. Together, these findings suggest that the structure of the viral quasispecies, rather than simply the presence of isolated mutations, may contribute meaningfully to treatment outcome [24,26].
Our findings are also in line with prior studies showing that linked dual-class resistance mutations are associated with an increased risk of virologic failure [24,26]. By incorporating both variant-specific susceptibility and relative abundance, wGSS provides a more representative measure of the resistance landscape within a participant than consensus genotyping alone. In this study, lower wGSS values at study entry were associated with persistent virologic failure on continued NNRTI-based ART, suggesting that haplotype-informed susceptibility metrics may better identify individuals at risk of poor response when treatment is not changed.
The paper has several limitations. First, this analysis was performed on plasma specimens from participants receiving NNRTI-based regimens, which are no longer the dominant first-line therapy in many settings following the transition to dolutegravir-based treatment [8,40]. Even then, these findings find relevance; NNRTI-associated resistance continues to shape treatment histories in many ART-experienced individuals, rilpivirine remains a component of long-acting treatment strategies, and the broader principle that linked resistance patterns may improve prediction of treatment outcome is likely applicable beyond NNRTI-based regimens [41,42]. Second, this was a retrospective analysis based on stored specimens from a selected subset of REVAMP participants [37,43]. Because our sequencing was restricted to the time of study entry (TP1) to predict subsequent outcomes, we were unable to evaluate longitudinal changes, such as transient or persistent failure at subsequent time points. Third, although NGS-Primer ID reduces PCR-associated error and recombination and enables single-genome resolution, it is technically demanding and remains better suited to research settings than routine clinical implementation at present [23,25,27]. In addition, the ability to detect very low-frequency variants depended on sample-specific sequence support and was therefore not uniform across all specimens. Accordingly, variants near the 1% level should be interpreted cautiously. Finally, adherence and pharmacologic factors were not directly incorporated into this analysis, although both are likely to contribute importantly to subsequent virologic outcome [32,34].
Despite these limitations, this study provides proof-of-principle evidence that resolving linked HIV-1 resistance mutations can yield clinically meaningful information not captured by standard genotyping. In settings where treatment options are limited or where individuals remain on partially active regimens, a more accurate view of the within-host resistance landscape may improve risk stratification and inform regimen selection. More broadly, these findings support further evaluation of linkage-informed resistance approaches in contemporary ART settings, including INSTI-based regimens and long-acting therapies [41,44]. Furthermore, future longitudinal studies are needed to prospectively track how the composition and linkage structure of these complex variants shift under ongoing selective pressure.
In summary, NGS-Primer ID uncovered extensive within-host diversity in HIV-1 drug resistance patterns among participants with prior virologic failure on NNRTI-based ART and showed that a weighted susceptibility measure based on linked resistance patterns may better predict subsequent treatment outcome than Sanger sequencing. These findings underscore the importance of considering both the composition and linkage structure of the viral quasispecies when interpreting HIV drug resistance and provide a foundation for future studies in the setting of modern ART regimens.
Methods
Ethics Statement
This study was approved by the Mass General Brigham Institutional Review Board (IRB) and the Biomedical Research Ethics Committee (BREC) at the University of KwaZulu-Natal (BREC reference number: BREC/00005292/2023 and REVAMP BREC reference number: BFC377/16). This study did not obtain informed consent from individuals, due to use of anonymized stored specimens for genotyping therefore the need for informed consent was waived upon ethics approval.
Study participants
Remnant plasma specimens were obtained from participants in the REVAMP (Resistance Testing Versus Adherence Monitoring for Patients) study conducted in KZN, South Africa. The REVAMP study was a randomized controlled study (2016–2019), aimed at investigating whether resistance testing improves rates of virologic re-suppression after VF, among individuals receiving HIV care in public health programs across sSA (NCT02787499) [37,43]. The REVAMP study had two arms: a standard of care (SOC) arm and a resistance testing (RT) arm. Participants were enrolled and randomized to each arm from four hospitals in the eThekwini District of KZN (Addington, Clairwood, King Dinizulu, and Wentworth). Additional details on the REVAMP study are described in the S1 Appendix. In this analysis, we included South African participants with VF who were continued on an NNRTI-based ART regimen with adherence counselling.
Study specimens
We performed retrospective viral Sanger sequencing on stored frozen plasma specimens from participants enrolled in the REVAMP study [37,43]. Additional details on the selection of specimens for NGS-primer ID testing are described in the Supplementary appendix (S2-S5 Tables). A control library of clonal HIV-1 subtype C sequences was prepared at defined concentrations by mixing clones at concentrations of 68.4%, 25%, 5%, 1%, 0.5% and 0.1% and processed in parallel with the specimens [29]. The PCR amplicons from the control library were gel purified, quantified by Nanodrop spectrophotometry and used to validate our library preparation method, to delineate the error rate and to determine the threshold for detecting DRMVs down to 0.1% of the viral population. The presence of minority variants (MVs) present at nucleotide positions in the control library amplicon was assessed for having the same invariant base [29]. The upper range of the assay error rate quantified as 3 standard deviations (SDs) above the mean MV percentage detected among the invariant bases [29].
Laboratory procedures for Sanger sequencing
Plasma specimens with viral loads ≥1000 copies/mL were retrieved from -80°C storage and equilibrated to room temperature before processing. HIV-1 viral RNA was extracted from 1mL of patient plasma specimens using a NucliSENS easyMAG automated extraction platform (BioMérieux, Marcy l’Etoile, France), according to manufacturer’s instructions. RNA was reverse-transcribed into cDNA using SuperScript III First-Strand Synthesis System (ThermoFisher Scientific, Massachusetts, USA) and HIV-1 subtype C gene-specific primers.
A nested PCR amplification of the protease and reverse transcriptase genes was performed using Southern African Treatment Resistance Network custom primers [45]. Cycle sequencing reactions were performed on purified products using a Big-Dye Terminator v3.1 kit (Applied Biosystems, Foster City, CA, USA). We assessed sequence quality and generated consensus sequence alignments using Geneious Prime software 2021.1.1 (Biomatters Ltd, New Zealand). We detected drug resistance mutations and determined GSS values for efavirenz (EFV), emtricitabine (FTC), and tenofovir disoproxil fumarate (TDF) using the Stanford University HIV Drug Resistance Database v9.0 [46].
Primer ID–based reverse transcription and cDNA generation
HIV-1 viral RNA was extracted from stored remnant plasma specimens (viral loads ≥1000 copies/ml) using the TRIzol (ThermoFisher Scientific, Massachusetts, USA) method, according to manufacturer’s instructions. Extracted RNA was reverse transcribed using the SuperScript IV RT-PCR kit (ThermoFisher Scientific, Waltham, USA) consisting of a high-fidelity reverse transcriptase capable of synthesizing a 600 bp region of HIV-1 pol gene spanning the major primary and secondary resistance mutations for the reverse transcriptase gene (codons 38–239). The cDNA synthesis was performed in a 50μl reaction consisting of a 5μM concentration of HIV-1 subtype C specific primer with Primer ID tags (S6-S8 Tables), 10mM dNTPs (Thermofisher Scientific, USA), 5 × RT Buffer (Thermofisher Scientific, USA), 0.1M DTT, 100mM RNaseOUT (Thermofisher Scientific, USA; 40 U/μl), and SuperScript IV RT (Thermofisher Scientific, USA; 200 U/μl). The primers were degraded, and the RNA/DNA duplex was denatured with Ribonuclease H (RNase H) (5 units; New England Biolabs, USA). cDNA with primer ID-tags were precipitated overnight. S2 Fig and the S1 Appendix provide an overview and hypothetical example of how NGS-Primer ID offers a more detailed view of the diversity of resistance variants beyond that of Sanger or traditional NGS.
PCR amplification, library construction, and next-generation sequencing
Primer ID-tagged cDNA molecules were amplified in 5 replicates using 2 × Kapa Hi-Fi Hot Start Uracil + reaction mix (KAPA Biosystems, Massachusetts, USA). A PCR master mix consisting of a concentration of 10μM forward primer 2589 FC in the first round PCR, 10μM forward primer 2709 FC in the second round, and a 10μM deoxyuridine (dU)-containing reverse primer PrimRegion- R-5Us in both rounds of PCR master mix was prepared as described in the S9-S11 Tables.
The five replicates from each PCR round were combined and purified using the QIAquick PCR Purification Kit (Qiagen, Hilden, Germany), according to manufacturer’s instructions. The PCR products enriched with dU primers were treated with uracil-DNA glycosylase (UDG) followed by cleavage at abasic sites, yielding double stranded DNA (dsDNA) with 17-nucleotide (nt) 3′-overhangs on both ends. NGS linkers were hybridised to the 17-nt 3′ overhangs and the dsDNA was completed with Klenow fragment DNA polymerase and dNTPs. The linker ligated DNA libraries were resolved on a 1.5% agarose gel and the ~ 719 bp fragment was excised and purified using QIAquick Gel Extraction kit (Qiagen, Hilden, Germany), according to manufacturer’s instructions. MiSeq Illumina sequencing was performed on primer ID tagged DNA libraries using the MiSeq v2 kit (Illumina Incorporation, California, USA). Sequence fastq files were demultiplexed and exported for bioinformatics analyses. The sequences were analyzed using the HIV-DRLink tool [47,48]. Additional details on the sequence analysis are described in the S1 Appendix.
HIV Drug-Resistance Analysis Pipeline (hivdrm) and Statistical Analysis
Sequencing reads were quality filtered, demultiplexed, and trimmed to remove adapters, primers, barcodes, and UMI sequences. For UMI-tagged libraries, reads sharing the same UMI were grouped and collapsed into consensus sequences using an 80% nucleotide agreement threshold to reduce PCR and sequencing errors. Consensus sequences were aligned to the study reference sequence corresponding to the targeted genomic region, and linked haplotypes were reconstructed across the amplicon. Drug-resistance mutations were identified and interpreted using the hivdrm pipeline (available at https://github.com/rinkideo/hivdrm) together with SierraPy, a Python interface to the Stanford HIV Drug Resistance Database [49] that translates viral sequences into drug-resistance mutations and predicts susceptibility to individual antiretroviral drugs. The pipeline summarizes reported DRM patterns, the number of sequences supporting each DRM pattern, and the corresponding pattern frequencies within a sample. For each sample, the frequency of a given DRM pattern was calculated as the number of sequences assigned to that pattern divided by the total number of input sequences in the sample. Accordingly, the summed frequency of all reported DRM patterns reflects the proportion of sample sequences containing one or more reported DRMs, whereas the remainder represents sequences without reported DRMs. These summary metrics therefore reflect sequence-level support for reported DRM patterns and should not be interpreted as direct measures of total sample-level template sampling depth or total recovered UMI counts. More details are provided in the S1 Appendix.
We used non-parametric Wilcoxon rank sum test to compare the distribution of the wGSS between individuals with and without viral suppression on maintenance of first-line ART. We used version 1: R software (4.4.3) [50], tidyverse (2.0.0) [51] and ggplot2 (3.5.1) [52] for statistical analysis and figure plotting.
Supporting information
S1 Fig. Maximum likelihood phylogenetic tree.
Maximum-likelihood phylogenetic tree including all sequences from participants in this study together with HIV-1 reference sequences in a rectangular layout. The tree was inferred using the generalized time reversible model with a proportion of invariant sites and gamma-distributed rate variation among sites (GTR + I + G). Branch support was assessed with 1,000 bootstrap replicates. Branch lengths represent the number of nucleotide substitutions per site (scale bar = 0.03).
https://doi.org/10.1371/journal.ppat.1014118.s001
(TIF)
S2 Fig. Schematic illustration of genotypic susceptibility score (GSS) estimation from different sequencing approaches.
The figure shows a hypothetical participant receiving an antiretroviral regimen consisting of efavirenz (EFV), tenofovir (TDF), and emtricitabine (FTC) (1). Example outputs are shown for Sanger sequencing (2), next-generation sequencing (NGS) (3), and NGS-based ultrasensitive single-genome sequencing with primer identifiers (NGS-PrimerID) (4). For each approach, detected drug resistance mutations are used to derive regimen-level GSS values. The schematic also illustrates summary GSS metrics derived from variant-level data, including maximum GSS, minimum GSS, and weighted GSS (5), and a conceptual distribution of GSS values across detected variants (6).
https://doi.org/10.1371/journal.ppat.1014118.s002
(TIF)
S3 Fig. Distribution of viral mutational patterns across the study cohort.
Each stacked bar plot corresponds to an individual participant, and the colors represent specific mutational patterns. Pattern frequencies were calculated from the number of sequences assigned to each pattern relative to the total number of input sequences in the sample. Reported DRM patterns present at frequencies ≤5% are shown in dark grey.
https://doi.org/10.1371/journal.ppat.1014118.s003
(TIF)
S4 Fig. Detailed mapping of resistance patterns in PID 219.
This figure illustrates a specific case where conventional Sanger sequencing underestimated the complexity of the reported resistance-pattern landscape in the sample. The first column identifies distinct mutational patterns, clarifying whether resistance markers are physically linked on individual viral genomes. The second and third columns quantify these patterns by listing the total number of sequences identified and their corresponding percentages within the specimen. Subsequent columns denote the presence of specific drug resistance mutations, which are highlighted with orange boxes. The final row provides the cumulative frequency of each individual mutation relative to the total number of input sequences in the sample.
https://doi.org/10.1371/journal.ppat.1014118.s004
(TIF)
S5 Fig. Detailed mapping of resistance patterns in PID 387.
This figure illustrates a specific case where conventional Sanger sequencing overestimated the complexity of the reported resistance-pattern landscape in the sample. The first column identifies distinct mutational patterns, clarifying whether resistance markers are physically linked on individual viral genomes. The second and third columns quantify these patterns by listing the total number of sequences identified and their corresponding percentages within the specimen. Subsequent columns denote the presence of specific drug resistance mutations, which are highlighted with orange boxes. The final row provides the cumulative frequency of each individual mutation relative to the total number of input sequences in the sample.
https://doi.org/10.1371/journal.ppat.1014118.s005
(TIF)
S1 Table. Sample-level clinical, Sanger genotypic susceptibility, and NGS-Primer ID pattern summary metrics at HIV viral load time point 1.
https://doi.org/10.1371/journal.ppat.1014118.s006
(DOCX)
S2 Table. Genotypic Susceptibility Scores with corresponding susceptibility levels.
https://doi.org/10.1371/journal.ppat.1014118.s007
(DOCX)
S3 Table. Selection Criteria for Plasma specimens GSS ≥ 1.
https://doi.org/10.1371/journal.ppat.1014118.s008
(DOCX)
S4 Table. Selection criteria used for Plasma specimens with GSS < 1.
https://doi.org/10.1371/journal.ppat.1014118.s009
(DOCX)
S5 Table. Participant specimens selected from the REVAMP Study for NGS-Primer ID.
https://doi.org/10.1371/journal.ppat.1014118.s010
(DOCX)
S6 Table. cDNA Primers with Primer ID tags used for reverse transcription.
https://doi.org/10.1371/journal.ppat.1014118.s011
(DOCX)
S7 Table. Complimentary DNA synthesis Master Mix 1.
https://doi.org/10.1371/journal.ppat.1014118.s012
(DOCX)
S8 Table. Complimentary DNA synthesis Master Mix 2.
https://doi.org/10.1371/journal.ppat.1014118.s013
(DOCX)
S9 Table. First-Round Master Mix and Conditions.
https://doi.org/10.1371/journal.ppat.1014118.s014
(DOCX)
S10 Table. Second-Round Master Mix and Conditions.
https://doi.org/10.1371/journal.ppat.1014118.s015
(DOCX)
S11 Table. cDNA Amplification Primers used in First and Second Round PCR.
https://doi.org/10.1371/journal.ppat.1014118.s016
(DOCX)
S1 Appendix. Supplementary Appendix.
Supplementary Methods: REVAMP Study: Participant enrolment and Randomization. REVAMP Study: Inclusion and Exclusion Criteria. Plasma Specimen Selection. Next-generation sequencing with Primer ID (NGS-Primer ID). HIV drug-resistance analysis pipeline (hivdrm). Phylogenetic Analysis. Supplementary Results. Diverse range of mutational patterns observed by NGS-Primer ID across all participants. Characterization of Resistant and Susceptible variants using weighted GSS by NGS-Primer ID.
https://doi.org/10.1371/journal.ppat.1014118.s017
(DOCX)
Acknowledgments
The authors acknowledge support from staff at the Department of Virology, National Health Laboratory Service (NHLS) and staff at Harvard/Brigham Virology Specialty Laboratory, Boston, Massachusetts, USA.
References
- 1.
HIV Statistics, Globally and by WHO Region, 2025. https://cdn.who.int/media/docs/default-source/hq-hiv-hepatitis-and-stis-library/who-ias-hiv-statistics_2025-new.pdf?sfvrsn=5023deae_15 2025 July 28.
- 2. UNAIDS. Fact sheet - Latest global and regional statistics on the status of the AIDS epidemic. 2024. https://www.unaids.org/en/resources/documents/2024/UNAIDS_FactSheet
- 3. UNAIDS. Country factsheets. South Africa 2023. https://www.unaids.org/en/regionscountries/countries/southafrica 2023 October 1.
- 4.
The sixth South African national HIV, behavioural and health survey. 20 years of strategic HIV and public health data. Human Sciences Research Council. https://hsrc.ac.za/wp-content/uploads/2024/07/SABSSM_VI_EXEC_REPORT_2PP.pdf
- 5. Barth RE, van der Loeff MFS, Schuurman R, Hoepelman AIM, Wensing AMJ. Virological follow-up of adult patients in antiretroviral treatment programmes in sub-Saharan Africa: a systematic review. Lancet Infect Dis. 2010;10(3):155–66. pmid:20185094
- 6. Moyo E, Moyo P, Murewanhema G, Mhango M, Chitungo I, Dzinamarira T. Key populations and Sub-Saharan Africa’s HIV response. Front Public Health. 2023;11:1079990. pmid:37261232
- 7. Rhee S-Y, Kassaye SG, Barrow G, Sundaramurthi JC, Jordan MR, Shafer RW. HIV-1 transmitted drug resistance surveillance: shifting trends in study design and prevalence estimates. J Int AIDS Soc. 2020;23(9):e25611. pmid:32936523
- 8. Murphy RA, Sunpath H, Lu Z, Chelin N, Losina E, Gordon M, et al. Outcomes after virologic failure of first-line ART in South Africa. AIDS. 2010;24(7):1007–12. pmid:20397305
- 9. Bertagnolio S, Hermans L, Jordan MR, Avila-Rios S, Iwuji C, Derache A, et al. Clinical Impact of Pretreatment Human Immunodeficiency Virus Drug Resistance in People Initiating Nonnucleoside Reverse Transcriptase Inhibitor-Containing Antiretroviral Therapy: A Systematic Review and Meta-analysis. J Infect Dis. 2021;224(3):377–88. pmid:33202025
- 10. Chimukangara B, Lessells RJ, Rhee S-Y, Giandhari J, Kharsany AB, Naidoo K. Trends in pretreatment HIV-1 drug resistance in antiretroviral therapy-naive adults in South Africa, 2000–2016: a pooled sequence analysis. EClinicalMedicine. 2019;9:26–34.
- 11. Kemp SA, Kamelian K, Cuadros DF, PANGEA Consortium, Vukuzazi Team, Cheng MTK, et al. HIV transmission dynamics and population-wide drug resistance in rural South Africa. Nat Commun. 2024;15(1):3644. pmid:38684655
- 12. Moyo S, Hunt G, Zuma K, Zungu M, Marinda E, Mabaso M, et al. HIV drug resistance profile in South Africa: Findings and implications from the 2017 national HIV household survey. PLoS One. 2020;15(11):e0241071. pmid:33147285
- 13. Carr A, Mackie NE, Paredes R, Ruxrungtham K. HIV drug resistance in the era of contemporary antiretroviral therapy: A clinical perspective. Antivir Ther. 2023;28(5):13596535231201162. pmid:37749751
- 14. Clutter DS, Jordan MR, Bertagnolio S, Shafer RW. HIV-1 drug resistance and resistance testing. Infect Genet Evol. 2016;46:292–307. pmid:27587334
- 15. Iyidogan P, Anderson KS. Current perspectives on HIV-1 antiretroviral drug resistance. Viruses. 2014;6(10):4095–139. pmid:25341668
- 16. Alves BM, Siqueira JD, Garrido MM, Botelho OM, Prellwitz IM, Ribeiro SR, et al. Characterization of HIV-1 Near Full-Length Proviral Genome Quasispecies from Patients with Undetectable Viral Load Undergoing First-Line HAART Therapy. Viruses. 2017;9(12):392. pmid:29257103
- 17. Boltz VF, Bao Y, Lockman S, Halvas EK, Kearney MF, McIntyre JA, et al. Low-frequency nevirapine (NVP)-resistant HIV-1 variants are not associated with failure of antiretroviral therapy in women without prior exposure to single-dose NVP. J Infect Dis. 2014;209(5):703–10. pmid:24443547
- 18. Chimukangara B, Giandhari J, Lessells R, Yende-Zuma N, Sartorius B, Samuel R, et al. Impact of pretreatment low-abundance HIV-1 drug-resistant variants on virological failure among HIV-1/TB-co-infected individuals. J Antimicrob Chemother. 2020;75(11):3319–26. pmid:32772079
- 19. Kyeyune F, Gibson RM, Nankya I, Venner C, Metha S, Akao J, et al. Low-Frequency Drug Resistance in HIV-Infected Ugandans on Antiretroviral Treatment Is Associated with Regimen Failure. Antimicrob Agents Chemother. 2016;60(6):3380–97. pmid:27001818
- 20. Inzaule SC, Hamers RL, Noguera-Julian M, Casadellà M, Parera M, Kityo C, et al. Clinically relevant thresholds for ultrasensitive HIV drug resistance testing: a multi-country nested case-control study. Lancet HIV. 2018;5(11):e638–46. pmid:30282603
- 21. Gianella S, Richman DD. Minority variants of drug-resistant HIV. J Infect Dis. 2010;202(5):657–66. pmid:20649427
- 22. Stella-Ascariz N, Arribas JR, Paredes R, Li JZ. The Role of HIV-1 Drug-Resistant Minority Variants in Treatment Failure. J Infect Dis. 2017;216(suppl_9):S847–50. pmid:29207001
- 23. Boltz VF, Rausch J, Shao W, Hattori J, Luke B, Maldarelli F, et al. Ultrasensitive single-genome sequencing: accurate, targeted, next generation sequencing of HIV-1 RNA. Retrovirology. 2016;13(1):87. pmid:27998286
- 24. Boltz VF, Shao W, Bale MJ, Halvas EK, Luke B, McIntyre JA, et al. Linked dual-class HIV resistance mutations are associated with treatment failure. JCI Insight. 2019;4(19):e130118. pmid:31487271
- 25. Jabara CB, Jones CD, Roach J, Anderson JA, Swanstrom R. Accurate sampling and deep sequencing of the HIV-1 protease gene using a Primer ID. Proc Natl Acad Sci U S A. 2011;108(50):20166–71. pmid:22135472
- 26. Palmer S, Kearney M, Maldarelli F, Halvas EK, Bixby CJ, Bazmi H, et al. Multiple, linked human immunodeficiency virus type 1 drug resistance mutations in treatment-experienced patients are missed by standard genotype analysis. J Clin Microbiol. 2005;43(1):406–13. pmid:15635002
- 27. Zhou S, Jones C, Mieczkowski P, Swanstrom R. Primer ID Validates Template Sampling Depth and Greatly Reduces the Error Rate of Next-Generation Sequencing of HIV-1 Genomic RNA Populations. J Virol. 2015;89(16):8540–55. pmid:26041299
- 28. Lecossier D, Shulman NS, Morand-Joubert L, Shafer RW, Joly V, Zolopa AR, et al. Detection of minority populations of HIV-1 expressing the K103N resistance mutation in patients failing nevirapine. J Acquir Immune Defic Syndr. 2005;38(1):37–42. pmid:15608522
- 29. Li JZ, Stella N, Choudhary MC, Javed A, Rodriguez K, Ribaudo H, et al. Impact of pre-existing drug resistance on risk of virological failure in South Africa. J Antimicrob Chemother. 2021;76(6):1558–63. pmid:33693678
- 30. Johnson JA, Geretti AM. Low-frequency HIV-1 drug resistance mutations can be clinically significant but must be interpreted with caution. J Antimicrob Chemother. 2010;65(7):1322–6. pmid:20466851
- 31. Li JZ, Paredes R, Ribaudo HJ, Svarovskaia ES, Metzner KJ, Kozal MJ, et al. Low-frequency HIV-1 drug resistance mutations and risk of NNRTI-based antiretroviral treatment failure: a systematic review and pooled analysis. JAMA. 2011;305(13):1327–35. pmid:21467286
- 32. Boyd MA, Moore CL, Molina J-M, Wood R, Madero JS, Wolff M, et al. Baseline HIV-1 resistance, virological outcomes, and emergent resistance in the SECOND-LINE trial: an exploratory analysis. Lancet HIV. 2015;2(2):e42-51. pmid:26424460
- 33. Durant J, Clevenbergh P, Halfon P, Delgiudice P, Porsin S, Simonet P, et al. Drug-resistance genotyping in HIV-1 therapy: the VIRADAPT randomised controlled trial. Lancet. 1999;353(9171):2195–9. pmid:10392984
- 34. Johnston V, Cohen K, Wiesner L, Morris L, Ledwaba J, Fielding KL, et al. Viral suppression following switch to second-line antiretroviral therapy: associations with nucleoside reverse transcriptase inhibitor resistance and subtherapeutic drug concentrations prior to switch. J Infect Dis. 2014;209(5):711–20. pmid:23943851
- 35. Kuritzkes DR, Lalama CM, Ribaudo HJ, Marcial M, Meyer WA 3rd, Shikuma C, et al. Preexisting resistance to nonnucleoside reverse-transcriptase inhibitors predicts virologic failure of an efavirenz-based regimen in treatment-naive HIV-1-infected subjects. J Infect Dis. 2008;197(6):867–70. pmid:18269317
- 36. Lorenzi P, Opravil M, Hirschel B, Chave J-P, Furrer H-J, Sax H, et al. Impact of drug resistance mutations on virologic response to salvage therapy. AIDS. 1999;13(2):F17–21.
- 37. Siedner MJ, Moosa M-YS, McCluskey S, Gilbert RF, Pillay S, Aturinda I, et al. Resistance Testing for Management of HIV Virologic Failure in Sub-Saharan Africa : An Unblinded Randomized Controlled Trial. Ann Intern Med. 2021;174(12):1683–92. pmid:34698502
- 38. Cozzi-Lepri A, Noguera-Julian M, Di Giallonardo F, Schuurman R, Däumer M, Aitken S, et al. Low-frequency drug-resistant HIV-1 and risk of virological failure to first-line NNRTI-based ART: a multicohort European case-control study using centralized ultrasensitive 454 pyrosequencing. J Antimicrob Chemother. 2015;70(3):930–40. pmid:25336166
- 39. Mbunkah HA, Bertagnolio S, Hamers RL, Hunt G, Inzaule S, Rinke De Wit TF, et al. Low-Abundance Drug-Resistant HIV-1 Variants in Antiretroviral Drug-Naive Individuals: A Systematic Review of Detection Methods, Prevalence, and Clinical Impact. J Infect Dis. 2020;221(10):1584–97. pmid:31809534
- 40. Marcelin A-G, Soulie C, Wirden M, Barriere G, Durand F, Charpentier C, et al. Emergent resistance-associated mutations at first- or second-line HIV-1 virologic failure with second-generation InSTIs in two- and three-drug regimens: the Virostar-1 study. J Antimicrob Chemother. 2025;80(1):95–101. pmid:39436753
- 41. Perez Navarro A, Nutt CT, Siedner MJ, McCluskey SM, Hill A. Virologic Failure and Emergent Integrase Strand Transfer Inhibitor Drug Resistance With Long-Acting Cabotegravir for HIV Treatment: A Meta-analysis. Clinical Infectious Diseases. 2024;81(2):274–85.
- 42. Uchitsubo K, Masuda J, Akazawa T, Inoue R, Tsukada K, Gatanaga H, et al. Nucleos(t)ide reverse transcriptase inhibitor-sparing regimens in the era of standard 3-drug combination therapies for HIV-1 infection. Glob Health Med. 2020;2(6):384–7. pmid:33409418
- 43. Siedner MJ, Bwana MB, Moosa M-YS, Paul M, Pillay S, McCluskey S, et al. The REVAMP trial to evaluate HIV resistance testing in sub-Saharan Africa: a case study in clinical trial design in resource limited settings to optimize effectiveness and cost effectiveness estimates. HIV Clin Trials. 2017;18(4):149–55. pmid:28720039
- 44. Hill A, Fairhead C, Manalu S, Venter F, Collins S, Landman R, et al. Could reduced dosing maintain more people on antiretrovirals after the sudden cuts in USAID funding? A crisis response. AIDS. 2025;39(7):F1–4. pmid:40265619
- 45. Manasa J, Danaviah S, Pillay S, Padayachee P, Mthiyane H, Mkhize C, et al. An affordable HIV-1 drug resistance monitoring method for resource limited settings. J Vis Exp. 2014;(85):51242. pmid:24747156
- 46. Tang MW, Liu TF, Shafer RW. The HIVdb system for HIV-1 genotypic resistance interpretation. Intervirology. 2012;55(2):98–101. pmid:22286876
- 47. Shao W, Boltz VF, Hattori J, Bale MJ, Maldarelli F, Coffin JM, et al. Short Communication: HIV-DRLink: A Tool for Reporting Linked HIV-1 Drug Resistance Mutations in Large Single-Genome Data Sets Using the Stanford HIV Database. AIDS Res Hum Retroviruses. 2020;36(11):942–7. pmid:32683881
- 48. HIV-DRLink Tool. https://github.com/bcbio/hivdrm 14 November 2024.
- 49. Liu TF, Shafer RW. Web resources for HIV type 1 genotypic-resistance test interpretation. Clin Infect Dis. 2006;42(11):1608–18. pmid:16652319
- 50.
Core R, Team. R: A language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing. 2021.
- 51. Wickham H, Averick M, Bryan J, Chang W, McGowan L, François R, et al. Welcome to the Tidyverse. JOSS. 2019;4(43):1686.
- 52.
Wickham H. ggplot2: Elegant Graphics for Data Analysis. New York: Springer-Verlag. 2016.