|Home | About | Journals | Submit | Contact Us | Français|
The clinical relevance of detecting minority drug-resistant HIV-1 variants is uncertain.
To determine the effect of pre-existing minority non-nucleoside reverse transcriptase inhibitor (NNRTI)-resistant variants on the risk of virologic failure (VF), we reanalyzed a case-cohort substudy of efavirenz recipients in ACTG A5095. Minority K103N or Y181C populations were determined by allele-specific PCR (ASPCR) in subjects without NNRTI resistance by population sequencing. Weighted Cox proportional hazards models adjusted for recent adherence estimated the relative risk of VF in the presence of NNRTI-resistant minority variants.
The evaluable case-cohort sample included 195 subjects from the randomly selected subcohort (51 with VF, 144 without failure [NF]), plus 127 of the remaining subjects with VF. Presence of minority K103N or Y181C mutations, or both, was detected in 8 (4.4%), 54 (29.5%) and 11 (6%), respectively, of 183 evaluable subjects in the random subcohort. Detection of minority Y181C mutants was associated with an increased risk of VF in the setting of recent adherence (HR=3.45, CI=1.90, 6.26), but not in non-adherent subjects (HR=1.39, CI=0.58, 3.29). Of note, 70% of subjects with minority Y181C achieved long-term viral suppression.
In adherent patients, pre-existing minority Y181C mutants more than tripled the risk of VF of first-line efavirenz-based ART.
Antiretroviral treatment guidelines recommend using the non-nucleoside reverse transcriptase inhibitor (NNRTI) efavirenz or a ritonavir-boosted protease inhibitor (PI), plus a fixed-dose combination of nucleoside or nucleotide reverse transcriptase inhibitors (NRTIs), for initial anti-HIV therapy [1, 2]. The efficacy of NNRTI-based regimens, however, is threatened in the presence of drug-resistant HIV-1 among newly infected or newly diagnosed persons [3, 4]. We previously showed that presence of pre-existing NNRTI-resistant mutants detected by population sequencing was associated with a 2.3-fold increased risk of virologic failure (VF) to first-line efavirenz-based antiretroviral therapy (ART) .
Antiretroviral drug resistance testing is recommended in HIV-1-infected subjects before starting ART to guide the selection of appropriate first-line regimens [2, 3]. Studies show that antiretroviral drug resistance testing is cost-effective , and improves the virologic, immunologic and clinical outcomes of ART [5, 7–10]. Current genotypic resistance assays, however, do not detect resistant viruses present in less than 15–20% of the viral population [11, 12]. New assays such as allele-specific real-time PCR (ASPCR) enable detection of low-abundance mutants with greater sensitivity [13–20]. Through preferential amplification of different allelic variants in real-time PCR conditions, ASPCR consistently detects mutants present in less than 0.1% of the virus population .
Relative to population sequencing of plasma viruses, ASPCR testing increases the detection of particular antiretroviral drug-resistance resistance mutations by 1.5- to 3-fold in different clinical settings [14, 15, 21, 22]. Whether drug-resistant mutants present at such low levels are associated with an increased risk of virologic failure of ART remains unresolved. We sought to address this question using ASPCR to detect selected NNRTI resistance mutations in pre-treatment plasma specimens from subjects in a case-cohort study of patients with and without VF in the efavirenz arms of AIDS Clinical Trials Group (ACTG) protocol A5095 [23, 24], a randomized trial of initial ART.
The ACTG A5095 study (clinicaltrials.gov identifier: NCT00013520) was a randomized, controlled trial that compared the efficacy of efavirenz plus a fixed-dose combination of two or three nucleoside reverse transcriptase inhibitors (NRTIs) with that of a fixed-dose triple-nucleoside regimen in previously untreated HIV-1-infected subjects with plasma HIV-1 RNA level of 400 copies/mL or greater (Amplicor or UltraSensitive HIV-1 Monitor Assay version 1.0; Roche Molecular Systems, Branchburg, NJ) [23, 24]. For subjects meeting the criteria for virologic failure (2 consecutive measurements of HIV-1 RNA level ≥200 copies/mL, with the first measurement at least 16 weeks after study entry), population sequencing (TruGene; Siemens, Norwood, MA) of plasma viruses was performed from samples stored at the time of first virologic failure and at baseline.
A case-cohort study of ACTG A5095 was performed to determine the prevalence of NNRTI resistance and its impact on treatment outcome in the efavirenz-containing arms of this trial [5, 25]. The case-cohort sample consisted of a random sample (subcohort) stratified by, and drawn from, the efavirenz-containing arms of A5095, plus additional cases (virologic failures) that were not selected to be in the subcohort. The current report presents further analyses of the existing case-cohort study; the results of the relationship between mutations as detected by population sequencing and virologic failure for this cohort have been previously published. 
To identify pre-treatment minority K103N and Y181C variants, blinded pre-treatment plasma samples with no NNRTI resistance detected by population sequencing, were reanalyzed using ASPCR. The primary outcome measure for the case-cohort study was the occurrence of virologic failure; the primary variable of interest was presence or absence of minority K103N and/or Y181C variants in the pre-treatment samples. Minority variants were defined as variants detected by ASPCR, but not by population sequencing. Data on recent adherence, defined as not missing any doses over the past four days , were captured as part of A5095 while the subject was on randomized treatment at weeks 4, 12, and 24 and then every 24 weeks using a self-administered adherence questionnaire . Given that non-adherence was associated with an increased risk of virologic failure in the main A5095 study , as-treated analyses in the current study were adjusted for recent self-reported adherence.
Viral RNA was extracted from one milliliter of plasma (QIAamp Viral RNA Mini Kit, Valencia, CA) after centrifugation at 24000 x g for 1 hour at 4°C, and PCR-amplified. (Supplementary material) PCR reactions proceeded as previously published [14, 19]. Clinical specimens were analyzed in the same batch with serially diluted standards (range, 102 to 107 standard DNA copies). The percentage of HIV-1 sequences containing each mutation was calculated as: percent mutated sequences = 100 × (quantity of mutant sequences)/(quantity of total HIV-1 sequences). In addition to the sensitivity threshold for each ASPCR assay, we calculated a specific detection threshold for each sample, defined as the minimum proportion of variants that could be detected based on the subject’s plasma HIV-1 RNA level (pVL), the volume of plasma used in the RNA extraction (V), the fraction of the RNA elution volume used for cDNA synthesis (fe), and the assumed efficiencies of the RNA extraction (ERNAX) and cDNA synthesis (EcDNA).
Based on the random subcohort, the prevalence of baseline minority K103N and/or Y181C mutants was estimated; the prevalence of each minority variant was compared between virologic failures and non-failures using the Fisher’s exact test. Using an exact test for homogeneity of odds ratios, the prevalence of Y181C mutants was compared between virologic failures and non-failures across the following subgroups: subjects with or without the K103N mutation, 4-drug or 3-drug EFV-based treatment, and screening HIV-1 RNA level. Summary statistics of the demographics of subjects in the random subcohort by pre-existing minority K103N and/or Y181C mutants, population resistance, or no NNRTI resistance are described, as well as for additional subjects with virologic failure. Weighted Cox proportional hazards models  were used to estimate the risk of virologic failure in the presence and absence of minority K103N and/or Y181C mutants at baseline among subjects without NNRTI resistance mutations by population sequencing. Unadjusted intent-to-treat and as-treated analyses showed similar results. Further as-treated analyses were adjusted for recent self-reported adherence; the presence of an interaction between recent self-reported adherence and baseline NNRTI resistance was examined. Such an interaction would imply a different impact of the presence of NNRTI resistance mutations at baseline dependent on recent adherence. Subjects without ASPCR results for either codon 103 or 181 were counted as missing unless otherwise specified. In a post-hoc analysis, all subjects with low-level Y181C mutants at baseline and experiencing virologic failure were evaluated for the presence of resistance by population sequencing at time of failure. A post-hoc analysis in the random cohort compared the mean change in viral load at day 14 on study from baseline in subjects with no NNRTI resistance mutations to subjects with low-abundant Y181C and to subjects with bulk resistance using the Wilcoxon rank sum test. All P-values and confidence intervals presented are nominal, unadjusted for multiple comparisons.
Of the 220 randomly sampled subjects, 57 (26%) were cases (virologic failures) and 163 (74%) were controls (non-failures) (Figure 1). Eleven controls had less than 16 weeks of follow-up and therefore were not evaluable for the protocol-defined criteria for virologic failure; these subjects were excluded from analyses of virologic failure and were not assayed by ASPCR. Reasons for premature study discontinuation included loss to follow-up (5), unable to get to clinic (4), toxicity (1) and clinical event (1). The fully evaluable random cohort sample included 195 subjects with at least 16 weeks of follow-up of whom 12 were considered NNRTI-resistant by population sequencing and 183 had complete ASPCR results. Note, an additional 4 subjects had only ASPCR data for K103N or Y181C, but not both (2 and 2, respectively). One hundred twenty-seven additional failures were added to the random subcohort (7 NNRTI-resistant by population sequencing and 120 with complete ASPCR results). Overall, the total case-cohort sample included 322 subjects (178 failures and 144 non-failures).
The detection threshold of the ASPCR was defined as more than 3 standard deviations above the mean of 20 repeated assays using the wild-type RT from pNL4-3 as a control target. Detection thresholds were: K103N (AAC) = 0.003%, K103N (AAT) = 0.001% and Y181C = 0.03%. The difference in real-time PCR threshold cycle values (ΔCt) between mutant and wild-type DNA equivalents was always > 17 cycles. Proportion measurements were linear down to at least 0.1% in all cases.
As previously reported , the prevalence of pre-treatment NNRTI resistance by population sequencing in the randomly sampled subcohort was 5%; this included 6 subjects with K103N alone, 2 with K103N together with a second NNRTI resistance mutation (other than Y181C), 0 with Y181C alone, and 1 with both K103N and Y181C by population sequencing at baseline. Of the 183 subjects assayed for the presence of pre-existing low-abundance K103N and/or Y181C mutants using ASPCR, variants carrying the K103N, Y181C, or both were detected in 8 (4.4%), 54 (29.5%) and 11 (6%) subjects, respectively. Table 1 summarizes baseline demographics of the random subcohort by presence or absence of preexisting minority NNRTI resistance mutations, and for the additional subjects with virologic failure.
Among subjects in the random subcohort in whom minority NNRTI-resistant variants were detected, the median (interquartile range) levels of mutants were: K103N, AAC allele= 0.012% (0.008%–0.116%); K103N, AAT allele= 0.013% (0.005%–0.053%); and Y181C=0.060% (0.048%–0.089%). Of note, the levels of Y181C and K103N mutants detected in individual samples were all below 1% (Figure 2 and data not shown).
Pre-existing low-abundance K103N mutants were detected less often than Y181C variants. K103N minority mutants were as frequent in non-failures as failures in the random subcohort with respect to Y181C minority mutants (Figure 3). Of the 185 subjects in the random subcohort, 58% of virologic failures compared to 29% of non-failures had low-abundance Y181C mutants at baseline (P=0.001). The relative prevalence of Y181C mutants in subjects with virologic failure compared to subjects without virologic failure was similar across subgroups defined by presence or absence of the K103N mutation, assignment to the 4-drug or 3-drug arm, and screening HIV-1 RNA level (Figure 3). In exploratory analyses, we were unable to define a threshold level of Y181C mutants that distinguished failures and non-failures with high sensitivity and specificity (not shown). A post-hoc analysis of the baseline characteristics of subjects with low-abundance Y181C mutants in the randomly sampled subcohort showed no differences between virologic failures and non-failures regarding screening HIV-1 RNA levels, CD4+ T-cell counts, or race/ethnicity (not shown). Note, these analyses include all subjects with results for the respective minority variant.
An as-treated weighted Cox proportional hazards model adjusted for recent adherence showed a significantly increased risk of virologic failure for subjects with an NNRTI-resistant virus by population sequencing compared with those with wild-type virus by population sequencing and ASPCR (hazard ratio [HR]= 4.00, 95% confidence interval [CI]= 1.72, 9.09). Among subjects with wild-type HIV-1 by population sequencing, detection of low-abundance Y181C mutants by ASPCR was associated with an increased risk of virologic failure (HR=2.54, 95% CI= 1.53, 4.20). A significant association with the detection of minority K103N mutants and an increased risk of virologic failure was not detected (P=0.22), but the direction of the effect was similar (HR=1.58, 95% CI=0.76, 3.28). As seen in the study overall , subjects with recent non-adherence also had an increased risk of virologic failure compared with adherent subjects (HR=2.30, 95% CI=1.40, 3.78).
Further modeling suggested an interaction between baseline presence of low-abundance Y181C mutants and recent adherence (P=0.08), showing that in the presence of recent non-adherence, the effect of minority Y181C was diminished (Figure 4). Among adherent subjects, the presence of minority Y181C by ASPCR had an increased risk of virologic failure compared to those that were sensitive by both population sequencing and ASPCR (HR=3.45, 95% CI= 1.90, 6.26; P<0.001); among non-adherent subjects, the presence of minority Y181C did not show a significantly increased risk of virologic failure (HR=1.39, 95% CI= 0.58, 3.29, P=0.46). Similar results were obtained when repeating this analysis using the presence of any minority variant (either K103N or Y181C) (data not shown).
Sixty-five subjects with low-abundance Y181C mutants at baseline experienced virologic failure and had a viral genotype (by population sequencing) available at the time of virologic failure. No resistance mutations were detected in 27 (41.5%), K103N was detected in 25 (38.5%), Y181C in 5 (7.7%) and K101E in 4 (6.2%); 2 of these 4 [3.1%] also had the K103N mutation (Table 2).
Following these observations we chose to examine whether low-abundant Y181C mutants impacted initial declines (to day 14) in HIV-1 RNA level upon treatment initiation. To avoid bias from over-sampling of virologic failures, these analyses were limited to the random cohort. Among subjects in the random cohort with available sequencing and day 14 viral load data, the mean change in viral load at day 14 on study from baseline was not significantly different between subjects with low-abundant Y181C and subjects with no NNRTI resistance mutations (P=0.97); a significantly smaller mean change in viral load at day 14 from baseline was detected in subjects with bulk resistance (mean change=−1.53 log 10copies/mL) as compared to subjects with no NNRTI resistance mutations (mean change=−2.06 log 10copies/mL) (difference: 0.52, 95% CI=0.13,0.91 copies/mL, P=0.01).
Detection of pre-existing minority Y181C mutants encoding NNRTI resistance was associated with a more than 3-fold increased risk of virological failure to initial ART with efavirenz-based regimens in ART-naive HIV-1-infected subjects in the presence of perfect adherence. The increased risk persisted across subjects with diverse baseline characteristics, including those with plasma HIV-1 RNA levels greater than or equal to, or less than 100,000 copies/mL; the risk magnitude was considerable and clinically relevant. Importantly, the impact of the presence of low-abundance Y181C mutants on the risk of virologic failure was diminished among non-adherent subjects. These findings confirm the importance of pre-existing resistant viruses present as minority members of the viral quasispecies in determining the virologic outcome of ART, particularly in the case of drugs with a low genetic barrier to resistance. They also underscore the clinical need for improving the sensitivity of genotypic drug resistance assays.
Mutations Y181C and K103N were chosen for the ASPCR analysis because they are the most frequent NNRTI resistance mutations found after virological failure to nevirapine and efavirenz. Minority Y181C and K103N mutants were detected by ASPCR in nearly 40% of subjects with wildtype virus by standard genotypic testing. This prevalence represented an almost 9-fold increase in the detection of primary NNRTI resistance when the results of ASPCR plus population sequencing (44%) were compared to population sequencing alone (5%).
In the current study, we did not detect an association with the presence of low-abundance K103N mutants and increased risk of virologic failure. This observation contrasts with previous studies, including our own finding in the same study population of a significantly increased risk of virologic failure when K103N was detected by population sequencing. However, this discrepancy may be attributable to the relatively small number of subjects with low-abundance K103N mutants identified by ASPCR.
It is noteworthy that all of the mutants identified by ASPCR in our study were present at levels below 1%. Although these low levels could represent underestimation due to polymorphisms at the primer binding sites in the target sequences, it would be surprising if this were the case in every subject tested. A more likely explanation is that Y181C and K103N mutants present at higher levels had already been identified by population sequencing, since samples from those subjects were not retested by ASPCR. This interpretation is consistent with data generated by ultradeep pyrosequencing , which found that NNRTI-resistant mutants were either present at relatively high levels (>20%, and thus detectable by population sequencing) or at low levels (generally below 1%–5%). These findings suggest that ultrasensitive resistance assays should have sufficient sensitivity to detect variants present at less than 1%–5% of the plasma virus population.
Our study extends the findings of two earlier studies. A retrospective case-control analysis from the U.S. Centers for Diseases Control and Prevention (CDC) applied a modified ASPCR technique to baseline samples drawn from two clinical trials of efavirenz-containing first-line regimens . Presence of minority mutations at RT codons 103, 181 or 184 was associated with an 11-fold increased odds of virological failure, but these mutations were detected in only a small number of subjects (7/95 with virologic failure and 2/221 with virologic suppression). The contribution of each individual mutation to the risk of virologic failure could not be assessed.
Analysis of baseline resistance by ultradeep pyrosequencing in the Flexible Initial Retrovirus Suppressive Therapies (FIRST) study, which compared initial ART strategies including an NNRTI, PI or both  found that pre-existing minority NNRTI-resistant variants more than tripled the hazard of virological failure in ART-naive subjects starting NNRTI-based therapy . Similarly, all 4 subjects in the PI arm in whom PI-resistant minority variants were detected experienced virologic failure, but the numbers were too small to show a statistically significant increase in the risk of virologic failure.
The clinical application of ASPCR or any other resistance assay requires a precise refinement of thresholds that identify subjects at greatest risk of virologic failure. We were unable to define a threshold level of mutants that distinguished between subjects with virologic failure and subjects without virologic failure with high sensitivity and specificity. Although subjects with minority Y181C variants were at greater risk of virologic failure, 70% of these subjects nevertheless achieved long-term viral suppression on their initial efavirenz-based regimen. In post-hoc exploratory analyses, we were unable to identify factors that explained this difference. The high sensitivity of ASPCR may capture natural fluctuations within the quasispecies over time that are not necessarily clinically significant. Conversely, the other two studies addressing the clinical relevance of minority variants used higher thresholds for detecting minor variants. Because the modified ASPCR method used in the CDC study  was designed to detect mutant viruses above the natural quasispecies frequency of each mutation, the actual threshold for detecting the K103N and Y181C mutants was 0.9% and 1.0%, respectively, which is at least two orders of magnitude higher than with our approach. Indeed, the clonal frequencies of the resistant variants in the CDC study ranged between 0.6% and 12.5%, suggesting a lower sensitivity of the ASPCR method used by the CDC, relative to ours. Similarly, due to the error rate of pyrosequencing, the cut-off for detecting minority variants in the FIRST study  was established at 1%. Whereas the CDC and the FIRST studies could have missed clinically relevant minority mutants, a number of minority mutants detected in our study did not contribute to virologic failure during the study period. Determining the optimum threshold to maximize sensitivity and specificity requires analysis of a larger number of samples than available in studies performed to date.
Although the presence of pre-existing low-abundance Y181C mutants was associated with a greater risk of virologic failure, other EFV resistance mutations were more commonly found at the time of virologic failure. Similar results were obtained in the FIRST study . It is possible that presence of the Y181C mutants was a marker for presence of other, undetected NNRTI mutants that emerged under efavirenz selection. Alternatively, the low-level EFV resistance conferred by Y181C could have allowed ongoing virus replication that led, in turn, to the later accumulation of other NNRTI resistance mutations such as K103N or G190S. Persistence of Y181C might have been selected against by the coadministration of zidovudine, since Y181C increases HIV-1 susceptibility to that drug . Because we did not attempt to detect minority NRTI-resistant variants at baseline, we cannot fully rule out that preexistence of some of such variants (e.g. M184V) could have influenced the risk of virological failure.
In conclusion, low-abundance NNRTI-resistant variants significantly increased the risk of virologic failure to initial antiretroviral therapy with efavirenz among adherent subjects; these minority mutants did not add to the risk of failure with non-adherence. More sensitive resistance assays could improve the clinical management of HIV-infected subjects. The clinical application of such assays, however, will require further technical developments, a better understanding of the role of low-abundance resistant variants in different clinical scenarios and, refinement of assay thresholds that identify patients at greatest risk of virologic failure.
We thank the 33 participating AIDS Clinical Trials Units, the A5095 study team members and the study subjects for their contributions to this work. We also thank Danielle Smith (Massachusetts General Hospital), Russell Young (University of Colorado Health Sciences Center), Lorraine Sutton (Vanderbilt University), J. Darrell Darren Hazelwood (University of Alabama, Birmingham) and Leslie Petch (University of North Carolina at Chapel Hill) performed HIV-1 genotyping assays for ACTG A5095.
Funding/Support: This work was supported by NIH grants AI38858 and AI068636 (AIDS Clinical Trials Group Central Grant), AI069419, AI051966, AI069472, AI069452, RR024996; subcontracts from grants AI38858 and AI06836 with the Virology Support Laboratories at Massachusetts General Hospital, the University of Alabama, the University of North Carolina, and Vanderbilt University; the Birmingham Veterans Affairs Medical Center; and the Harvard University and University of Alabama at Birmingham and the University of North Carolina at Chapel Hill Centers for AIDS Research (AI060354, AI027767, and AI50410). Dr. Paredes was awarded the “La Caixa” Grant for Post-Graduate Studies, Caixa de Pensions de Barcelona, “La Caixa”, Spain. Bristol-Myers Squibb and GlaxoSmithKline provided drug for this study as well as financial support for plasma HIV-1 RNA determinations.
Clinical Trials Registration: clinicaltrials.gov Identifier: NCT00013520
Dr. Paredes reports having received research grants (awarded to the irsiCaixa Foundation) from Boehringer-Ingelheim, Monogram, Pfizer, and Merck; and received speaker honoraria from Siemens Medical Solutions. Dr. Shikuma reports having had affiliations with or financial involvement with Boehringer-Ingelheim, Bristol-Myers Squibb, Gilead, and GlaxoSmithKline. Dr. Johnson reports serving as a consultant to and/or having received grant support from Bayer, Bristol-Myers Squibb and GlaxoSmithKline. Dr. Fiscus reports having received speaker honoraria from Gen-Probe and Abbott Molecular and receiving kits from Gen-Probe, Abbott Molecular, and Perkin-Elmer. Dr. D’Aquila reports having received grant support from Bristol-Myers Squibb and being a consultant to Boehringer-Ingleheim and GlaxoSmithKline. Dr. Gulick reports having received research grants (awarded to Cornell University) from Merck, Pfizer, Schering and Tibotec; served as an ad hoc consultant to Boehringer-Ingelheim, Bristol-Myers Squibb, Gilead, GlaxoSmithKline, Merck, Pathway, Pfizer, Progenics, Schering, Tibotec, and Virostatics; and serving as DSMB Chair for Koronis. Dr. Kuritzkes reports having served as a consultant for and received speaker's fees and/or research support from Boehringer-Ingelheim, Bristol-Myers Squibb, and GlaxoSmithKline; and served as a consultant for and received research support from Bayer and Siemens. Dr. Ribaudo served on the DSMB for Koronis. Ms. Lalama and Dr. Schackman had no financial disclosures to report.
Previous Presentation: Presented in part at the 15th Conference on Retroviruses and Opportunistic Infections, February 3–6, 2008, Boston, MA, Abstract 83.