Influence of prior NNRTI experience on virologic response to therapy
As previously reported [13
], virologic failure in ACTG study 398 was strongly associated with prior NNRTI exposure either to nevirapine or delavirdine (efavirenz exposure was not permitted). By study week 24, protocol-defined virologic failure had occurred in 83% (176 of 212) of NNRTI experienced patients versus 58% (156 of 269) of NNRTI naive participants (ITT analysis, P <0.001) [13
]. Standard genotype analysis was performed on baseline samples from 452 of the 481 study participants. NNRTI resistance mutations were detected in 165 participants (9 of 246 NNRTI-naïve and 156 of 206 NNRTI-experienced patients) and the presence of resistance was associated with virologic failure by week 24 (P <0.001). In the current study, we found that a history of NNRTI experience was significantly associated with virologic failure at week 24 after controlling for baseline NNRTI resistance (OR 2.07 95% CI 1.1–3.88; P = 0.024, Wald test of logistic regression). To explore this observation further, we examined the change in plasma HIV-1 RNA from baseline to week 24 among three subgroups: 1) NNRTI-naïve with baseline standard genotype negative for NNRTI resistance (N = 235); 2) NNRTI-experienced with NNRTI resistance detected at baseline by standard genotype (N = 156); and 3) NNRTI-experienced with baseline standard genotype negative for NNRTI resistance (N = 48). The fourth group of NNRTI-naïve subjects with NNRTI resistance at baseline (N = 9) was too small to include in this analysis.
shows that the change in HIV-1 RNA from entry to week 24 in the NNRTI-experienced subgroup without baseline NNRTI resistance by standard genotype resembled that of the genotype-positive NNRTI-experienced subgroup, which was inferior to that of the NNRTI-naïve subgroup. This observation led to the hypothesis that minor populations of NNRTI-resistant variants compromised the response to efavirenz-containing multidrug therapy in NNRTI-experienced patients with negative standard genotypes for NNRTI resistance.
Median change in HIV-1 RNA (log10 copies/ml) by NNRTI experience and baseline NNRTI resistance
To test this hypothesis, we used two more sensitive methods, single-genome sequencing and allele-specific PCR, to detect minor NNRTI-resistant variants in plasma samples from study entry and to assess their impact on virologic response.
NNRTI-resistance mutations detected by single-genome sequencing
For single-genome sequencing analyses, 27 baseline and virologic failure sample pairs were randomly selected as described in Methods. For NNRTI-experienced patients, the median duration after discontinuation of prior NNRTI therapy was 366 days (range 0–555 days). Sequences were generated and analyzed at entry and failure time points except for participant 11E for whom no sample was available at virologic failure.
Single-genome sequencing results are summarized in . A total of 1566 sequences from study entry were analyzed (an average of 46 sequences/sample), allowing the detection of variants present at a frequency of ≥ 5% with 90% confidence. Only ~12 sequences/sample were analyzed at virologic failure because NNRTI-resistant variants were dominant as determined by standard genotype analysis.
NNRTI resistance detected by single-genome sequencing
The number and types of NNRTI-resistance mutations detected at study entry were significantly different between NNRTI-naïve and -experienced groups (). Specifically, one or more sequences encoding an NNRTI resistance mutation were detected in 8 of 12 NNRTI-experienced patients compared with 3 of 15 NNRTI-naïve patients (P=0.022). Furthermore, the fraction of sequences encoding NNRTI-resistance mutations was higher in the NNRTI-experienced group (31 of 468 sequences) than in the NNRTI naïve group (3 of 773 sequences; P<0.001).
In this small subset of 27 participants analyzed by single-genome sequencing, the HIV-1 RNA response from entry to virologic failure was inferior in the NNRTI-experienced group (+0.13 ± 0.19 [SE] log10 copies/ml) compared with the NNRTI-naïve group (−0.35 ± 0.13 [SE] log10 copies/ml; P=0.032).
Phylogenetic analyses were performed to assess the relatedness between sequences with NNRTI resistance mutations at entry and virologic failure. Close clustering between baseline and failure sequences containing NNRTI resistance mutations, as evidenced by high bootstrap values, was observed in 1 of 3 NNRTI-naïve subjects and 1 of 8 NNRTI-experienced subjects, with the suggestion of clustering in a second NNRTI-experienced subject. illustrates these phylogenetic relationships. In subject 7N from the NNRTI-naive group (), a baseline sequence containing K103N clustered closely (bootstrap value 96) with sequences at failure containing K103N linked to Y188C or M230L. By contrast, baseline sequences from NNRTI-naïve subject 5N and 12N containing P225H and L100I, respectively, did not cluster with the predominant resistant variants at failure ().
Phylogenetic analysis comparing relatedness between single genome sequences at baseline and virologic failure
NNRTI-experienced subject 5E had two distinct populations of sequences at entry, one wildtype and one containing several linked NNRTI-resistance mutations (K101E/Y181C/G190A). The mutant sequences at entry intermingled in multiple clusters (bootstrap values >60) with the mutant sequences at failure (). For NNRTI-experienced subject 3E, entry sequences containing K101E showed a trend toward clustering (bootstrap value 46) with failure sequences containing L100I/K101E/Y188L or L100I/K101E/G190A (). The observed clustering of entry and failure mutant sequences for these subjects was not altered by changing drug-resistance mutations to wildtype (not shown).
The NNRTI-resistant mutants detected at entry for the 5 other NNRTI-experienced subject () did not cluster with the mutant sequences at failure. To 13 illustrate, patient 8E had sequences at entry and failure containing K103N, but these were not closely related ().
NNRTI-resistance mutations detected by allele-specific PCR
To further assess if minor populations of NNRTI-resistant variants were more frequent in the NNRTI-experienced patients and associated with virologic failure, baseline samples from 103 subjects (76 NNRTI-naive and 27 NNRTI-experienced) with negative standard genotype for NNRTI resistance were tested using allele-specific PCR for mutants K103N and Y181C. Of the 103 subjects selected, 5 subjects (2 NNRTI-naïve and 3 NNRTI-experienced) were excluded since follow-up HIV-1 RNA measurements were not available. In addition, allele-specific PCR could not be obtained in 4 subjects for Y181C and 3 for K103N, leaving 95 subjects evaluated for K103N and 94 for Y181C (). A significant association was found between NNRTI-experience and Y181C variants at mutant cutoff frequencies >0.5% (5/22 versus 3/72, P=0.016) and >1% (5/22 versus 3/72, P=0.016). A trend for association was found between NNRTI-experience and K103N at frequencies >0.5% (4/24 versus 4/71, P = 0.11) and >1% (3/24 versus 2/7, P = 0.10).
Impact of K103N and Y181C Variants on HIV-1 RNA Response
K103N at frequencies >1.0% was strongly associated with inferior HIV-1 RNA response from baseline to week 24 by intent-to-treat (ITT) analysis (P <0.001; ). Specifically, K103N >1.0% was associated with a rise in HIV-1 RNA of 0.5 log10 copies/ml, whereas K103N ≤1.0% was associated with a decrease in HIV-1 RNA of 1.1 log10 copies/ml. K103N at >0.5% frequency was also significantly associated with inferior HIV-1 RNA response, although the strength of association was lower (P = 0.006). K103N at frequencies >1% was also associated with protocol-defined virologic failure in both on-treatment (P = 0.036) and ITT analyses (P = 0.053). By contrast, Y181C variants were not significantly associated with either protocol-defined virologic failure or change in HIV-1 RNA from entry to week 24 (), although the average reduction in HIV-1 was lower for those with >1% Y181C (−0.4 log10 copies/ml) versus those with ≤1.0% mutant (−1.1 log10 copies/ml).