Detection of Drug-Resistant Minority Variants of HIV-1 During Virologic Failure of Indinavir, Lamivudine, and Zidovudine

HIV-1 variants emerging early during failure of treatment with indinavir, lamivudine, and zidovudine usually have the lamivudine- resistance mutation M184V but rarely have indinavir-resistance mutations [1, 2]. These observations have raised the possibility of preserving future treatment options by making selected drug substitutions guided by resistance testing.

Genotypic resistance testing is performed by use of DNA sequence analysis. Variability in assay performance occurs most commonly when genetic mixtures are present [3–5]. Few studies have evaluated whether minority variants develop at the time of treatment failure. We determined whether indinavir-resistant minority variants existed in plasma samples from patients for whom an indinavir-containing regimen was failing, using sequence analysis of cloned polymerase chain reaction (PCR) products.

Subjects, materials, and methods. Patients enrolled in AIDS Clinical Trials Group (ACTG) 320 [6] had CD4 cell counts <200 cells/mm3 and prolonged nucleoside experience but were lamivudine- and protease inhibitor (PI)-naive at entry. Informed consent was obtained from all patients included in the present study, and human-experimentation guidelines of the US Department of Health and Human Services were followed in the conduct of this clinical research. Patients were randomized to receive lamivudine and zidovudine (or stavudine), with either indinavir or a matching placebo. Approximately half of the subjects in the indinavir armachieved plasma HIV-1 RNA concentrations <500 copies/mL by week 24 [7]. Stored plasma samples for ACTG 320 were identified by use of these criteria: (1) subjects developed suppression of plasma HIV-1 RNA concentrations, from a baseline level of ⩾10,000 copies/mL to ⩽500 copies/mL, followed by a subsequent rebound to >500 copies/mL; (2) subjects received indinavir; and (3) samples with sufficient volume were stored at the University of Rochester. We chose a baseline value of ⩾10,000 copies/mL to ensure that subjects achieving virus suppression had ⩾ 1 log₁₀ decline in HIV-1 RNA concentration, suggesting adherence to the 3-drug regimen. All samples from the first failure visit (defined as the first visit at which the HIV-1 RNA concentration was >500 copies/mL, after the HIV-1 RNA concentration had previously decreased to ⩽500 copies/mL) meeting these criteria underwent bulk sequence analysis (table 1). Of the 256 subjects in ACTG 320 for whom specimens were stored at the University of Rochester, 30 (12%) had baseline HIV-1 RNA concentrations <10,000 copies/mL, 20 (8%) provided no baseline sample, 51 (20%) provided <2 follow-up samples, 86 (34%) did not develop an HIV-1 RNA concentration <500 copies/mL, and 51 (20%) did not develop an HIV-1 RNA concentration >500 copies/mL after initial suppression. Of the remaining 18 subjects with an appropriate concentration profile, 6 were in the dual-nucleoside arm, and samples from 2 did not have sufficient volume for testing. Samples without evidence of primary indinavir-resistance mutations [8] by bulk sequence analysis underwent clonal analysis.

Nucleotide sequences (protease codons 1-99 and reverse-transcriptase [RT] codons 1-310) were obtained by use of theViroSeq Genotyping Kit (version 2.0; Celera). To prevent degradation of cloned DNA by bacterial uracil-N-glycosylase, dTTP (provided by Applied Biosystems/Celera) was substituted for dUTP during RT-PCR. Duplicate amplifications from each sample were pooled; one half underwent bulk sequence analysis, and the second half was cloned into pCR-XL-TOPO (Invitrogen). Protease sequences from ⩾30 clones/sample were determined (primers A, D, and F). Because the bulk sequences from patients 6 and 9 showed no evidence of M184V at the time of treatment failure, a second set of 30 clones each was isolated from the original PCR, to detect M184V minority variants. All bulk and clonal sequences were submitted to GenBank (AY349971-AY350240).

Analyses of linkage between V82A and M184V for patient 10 were performed on a subset of the same clones that were used to initially determine the proportion of V82A variants in plasma. A total of 9 of 9 V82A clones at week 24 and 6 of 6 V82A clones at week 40 were sequenced, to determine the linked RT sequence. In addition, 6 and 9 clones with 82V were randomly selected at weeks 24 and 40, respectively, for comparison.

A mixture was defined, according to the Celera software (version 2.2), if the peak height of the minority band was ⩾30% of the height of the majority band in ⩾1 primer sequence. Our laboratory also requires that a mixture be present in both sense and antisense primer sequences and that the peak height of the minority band be higher than the neighboring background noise.

Exact confidence intervals (CIs) for the prevalence of minority variant genomes in a subject's plasma samples were constructed on the basis of the assumption that the clonal sequences represent a random sample of plasma viral variants. Fisher's exact test was used to assess changes in the prevalence of minority variants over time within subjects.

To verify the absence of PCR cross-contamination, all clonal and bulk sequences underwent phylogenetic analysis (PHYLIP; version 3.573c; neighbor-joining method). One clone from subject 8 clustered more closely with the sequence of subject 7 than with its own (nucleotide mismatch, 1.9% vs. 5.4%) and was excluded from subsequent analyses because of presumed PCR contamination. All remaining clonal sequences from each patient were closely related to the corresponding bulk sequence and were clustered separately for each subject. This clustering was confirmed by bootstrapping and analysis of polymorphisms specific to each patient's viral sequence.

Results. Bulk sequences were obtained for all 10 patients who met the criteria for testing. The median baseline HIV-1 RNA concentration and CD4 cell count were 99,000 copies/mL (range, 14,000-466,300 copies/mL) and 105 cells/mm³, respectively. The median HIV-1 RNA concentration at the time of treatment failure was 5,000 copies/mL (range, 700-74,800 copies/mL). No baseline sequences from 8 subjects for whom sequences were available showed evidence of lamivudine-resistance or primary indinavir-resistance mutations. All had evidence of nucleoside-resistance mutations in RT, which was consistent with the subjects' extensive previous nucleoside experience. Sequences from 4 of 10 subjects had evidence of a primary indinavir-resistance mutation by bulk sequence analysis at the first failure visit.

We determined whether any of the remaining 6 patients had evidence of indinavir-resistant minority variants that were not evident by bulk sequence analysis, using clonal analysis of PCRamplified products. Sequences from 4 of 6 subjects had evidence of a minority variant occurring at ⩾1 codon associated with resistance to indinavir (table 1); 2 had evidence of unusual variants (M46T and M36V) in only 1 clone each, which could represent PCR-introduced mutation, given their low frequency and the unusual amino acid substitutions. Another had evidence of the secondary resistance mutation M36I in 5 (16%) of 31 clones. The sequence from the fourth patient (patient 10) had evidence of a primary indinavir-resistant minority variant at codon 82 at week 24 (9/30 clones [30%]). V82A was not detected in baseline clones (0/32; CI, 0%-11%; P<.001, vs. week 24).

Patient 10 continued to receive the failing regimen through week 40, 16 weeks after the first treatment failure. Surprisingly, although the V82A minority variant persisted at week 40, it did not increase in prevalence, compared with week 24 (6/36 clones [17%]; CI, 6-33%; P=.25). During this time, the HIV-1 RNA concentration remained >1 log₁₀ below the pretherapy value (700-1200 vs. 41,400 copies/mL).

The bulk sequence for patient 10 at week 24 also had a mixture at RT codon 184. Therefore, we analyzed whether the V82A and M184V drug-resistant variants were genetically linked, by RT sequence analysis, for all V82A-containing clones at weeks 24 and 40. For comparison, we randomly selected an equal number of clones with 82V (table 2). Only 2 of 9 clones at week 24 with V82A in protease were linked with the lamivudine- resistance RT mutation M184V. In contrast, at week 40, 6 of 6 V82A clones had evidence of the mutant M184V codon. At week 24, 4 of 9 clones with 82V had evidence of M184V; this increased to 6 of 6 clones at week 40. Thus, in this genetic background, there clearly was selection for M184V from week 24 to 40, apparently independent of its linkage with V82A. The predominant clonal genotype at week 40 (82V-M184V) also contained the I64V mutation in protease, which was present in 3 (9%) of 32 clones at baseline but was not detected in 30 clones at week 24 (table 1). Two of the original 10 patients (patients 6 and 9) did not have evidence of M184V by bulk sequence analysis at the first failure visit; 7 (23%) of 31 clones and 0 of 30 clones from patients 9 and 6, respectively, had evidence of M184V.

Discussion. We determined whether indinavir-resistant minority variants existed in plasma samples from patients for whom an indinavir-containing regimen failed. Surprisingly, primary PIresistance mutations were detected by bulk sequence analysis at the first failure visit in 4 of 10 patients for whom treatment with indinavir, lamivudine, and zidovudine failed. This high prevalence contrasts with findings of previous studies in which patients for whom this regimen failed did not develop indinavir-resistant HIV-1 variants [1, 2]. A recent large study of patients with advanced HIV-1 infection and limited previous treatment experience found little or no HIV resistance to PIs at failure of an indinavir-containing regimen [9], suggesting that previous nucleoside experience is the most likely explanation for the high frequency of resistance to PIs in the ACTG 320 trial. Also consistent with this hypothesis are the recent observations that, in antiretroviral-naive patients, less-potent PI-dual nucleoside regimens result in higher rates of primary PI-resistance mutations at the time of the first treatment failure [10, 11].

Of the 6 remaining patients without evidence of primary indinavir-resistance mutations by bulk sequence analysis, the primary V82A resistance mutation was present in 30% of clones in 1 patient. A second patient had evidence of the lamivudineresistance M184V mutation present at 23%. These drug-resistant minority variants potentially have significant implications for the design of salvage regimens. The inability to detect these minor variants on routine sequence analysis in these 2 samples was primarily because of the specific base-calling algorithms used to identify mixtures and the less than optimal sequence quality (data not shown). Relaxing our requirements for identifying mixtures would allow detection of these variants but would likely reduce the specificity of detecting mixtures and require extensive validation before implementation.

It is also interesting that sequences from 2 patients that had no evidence of M184V by bulk sequence analysis also had no evidence of PI-resistance mutations. The simplest explanation is poor adherence to the antiretroviral regimen. Yet this does not explain the reduction in virus load to below baseline values at the time of the first treatment failure. Another possibility is that, in these patients, the regimen was highly effective, with infrequent resistance at the first treatment failure, similar to patients in recent trials of potent PI-nucleoside combinations [10, 11].

The clinical significance of the V82A minority variant in patient 10 is unclear, since its frequency of occurrence did not increase from week 24 to 40 and was not associated with an increase in virus load to pretherapy levels. We suspect that the persistently low virus load in this patient could reflect a residual drug effect, since a mutant resistant to all 3 antiretroviral drugs was not selected for. If it is true that there is a limited effective population size of HIV that would prevent rapid and complete circulation of HIV subpopulations, then incomplete circulation of HIV subpopulations could account for the lack of selection for V82A-M184V. Alternatively, in this patient, the V82A-M184V double mutant may have a selective disadvantage relative to the M184V single mutant. We believe that the subsequent outgrowth of the week-40 82V-M184V variant over both the week-40 V82A-M184V and the week-24 82V-M184V variants may be related to the acquisition of the I64V and the loss of the I62V-L63P protease mutations, although we do not have data to directly support this hypothesis.We unfortunately do not have subsequent samples from this patient with which to determine whether further selection for the V82A-M184V double mutant, relative to the 82V-M184V variant, continued after week 40 and whether dominance of the V82A-M184V mutant in the plasma virus population would be associated with increases in virus load to levels closer to pretherapy levels.

The linkage data presented here should be interpreted with caution, since it is possible that PCR-mediated recombination may be contributing to the linkages detected in our clonal analyses. However, we observed consistent linkages among certain codons, at weeks 24 and 40, including protease codons 62, 63, 64, and 82 and RT codons 142, 184, and 207 (table 2). The consistent linkages observed in these 2 independent RT-PCR suggest that PCR-mediated recombination is an unlikely explanation for the dominant linkage patterns seen in patient 10.

In conclusion, current genotyping technologies may not be sensitive enough to detect drug-resistant minority variants, and this limitation needs to be considered when interpreting patterns of HIV drug resistance at the time of treatment failure. The clinical significance of these minority variants is still unclear, and more studies are needed to determine the frequency with which they occur and their association with clinical outcome, given the small numbers of patients evaluated in our study. Studies of linkages between drug-resistance mutations can assist in determining the clinical significance of drug-resistant minority variants by allowing a detailed analysis of the genetic context in which drug-resistance mutations occur. Such studies can lead to a better understanding of which factors limit the selection for multidrug-resistant HIV variants and have the potential to provide important insights into why, in some patients, virus loads at the time of treatment failure remain significantly below pretherapy levels.

Previous SectionNext Section

Figures and Tables

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

Table 1.

Baseline characteristics of patients receiving indinavir (IDV), lamivudine (3TC), and zidovudine (ZDV) and sequence analysis of plasma HIV-1.

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

Table 2.

Linkage of protease V82A and reverse-transcriptase (RT) M184V resistance mutations in patient 10.

• Received June 20, 2003.
• Accepted September 15, 2003.

• © 2004 by the Infectious Diseases Society of America