Background: The relevance of low-frequency resistance variants is unclear. Here, we present the largest detailed analysis of baseline PR, RT, and IN deep sequencing data from antiretroviral-naïve participants treated with guideline recommended INSTI+2 NRTI regimens, and investigate the impact of low-frequency variants on treatment outcome.
Methods: Baseline deep sequencing of PR, RT, and IN was conducted for all participants in two randomized trials of antiretroviral-naïve participants initiating therapy on B/F/TAF, DTG/ABC/3TC, or DTG+F/TAF (Studies 1489+1490). The deepType HIV assay (Seq-IT, Germany) and Gilead analysis pipeline were used for sequencing. Virologic outcomes were assessed at Week 96 by FDA Snapshot.
Results: Baseline deep sequencing data for PR, RT, and IN were available for 1270/1274 participants. Deep sequencing results at a ≥15% cutoff were similar to population Sanger sequencing. To establish appropriate cutoffs for resistant variant calling, deep sequencing was performed on HIV plasmid and virus controls, and analyses of mutation counts using different cutoffs of percent frequency, mutational viral load, and number of reads were performed. High confidence in calling mutations present at frequencies ≥2% was observed. Table 1 shows detailed results for M184V. Between 2%-15%, additional primary resistance mutations not detected by population sequencing were detected: 3.7% NRTI-R, 3.7% INSTI-R, 4.2% NNRTI-R, and 5.8% PI-R. Eleven participants had the exclusion mutation M184V/I at baseline (2.1% to 12.9%), and all 11 of these participants had HIV-1 RNA < 50 copies/mL at Week 96 (n=10) or study drug discontinuation (n=1). Overall, participants with low-frequency resistance mutations had HIV-1 RNA < 50 copies/mL at Week 96 at rates similar to the overall population of the study.
Conclusions: In Studies 1489+1490, primary resistance mutations were detected at low frequencies between 2%-15% in an additional 16% (204/1270) of ART-naïve participants. These resistance variants did not affect viral suppression or virologic failure for B/F/TAF or other INSTI+2 NRTI regimens.

M184V Detection by Deep Sequencing Using Different Methods
Method 1: Mutation Frequency (percent of reads, %)
# of Participants with M184V01459321124
Method 2: Mutational Viral Load (baseline viral load x percent of reads with mutation, copies/mL), independent of frequency cutoff
# of Participants with M184V04153677414646
Method 3: Number of Reads, independent of frequency cutoff
# of Participants with M184V025181245861105
[Table 1. Detection of the RT M184V Mutation by Deep Sequencing in Baseline Samples from Studies 1489 and 1490 (n=1270)]