In this study, six individuals were enrolled. All were Japanese men who have sex with men (MSM) and were infected with subtype B. Thus, the patients were from the major demographic of HIV-infected people in Japan (Table 1) [29]. All cases were on-ART at the time this of their enrollment to the study (S2–S7 Tables). A total of 180 plasma samples were collected from the six patients from November 1996 to July 2007. All individuals had been taking ARVs for more than 5 years, during which the treatment regimens were switched multiple times because of the difficulties in controlling viremia (S2–S7 Tables). Multiple samples were obtained from each case, 30 sampling points with three- to four-month intervals. As all cases were in virological failure, and most of the VL at the time of collection above 1000 copies/ml. However, we also attempted to amplify lower VL samples as well. There are only eighteen sampling points with longer intervals, six months to a year, in case 6 (S7 Table). Treatment regimens and period of treatments are summarized in Table 1.

Specimens were anonymous and residual diagnostic material from subjects who provided written consent for HIV testing. The Ethical Committee for Biomedical Science of the NIID determined that this testing did not involve identifiable human subjects and has approved the study. All testing was performed at the National Institute of Infectious Diseases (NIID), AIDS Research Center in Tokyo, Japan.

HIV RNA copy number of each sample was measured using the Amplitaq Monitor v 1.5 (Roche Diagnostics, Indianapolis, IN). The detection limit of viral load was set to <50 copies / mL plasma in this study. CD4-positive T cell count data was obtained from clinical information supplied from hospitals where the individuals were treated. HIV RNA was extracted from patient plasma as described previously [30]. Briefly, HIV RNA was extracted from 200 μL of patient plasma and eluted into 100 μL of elution buffer. Extracted RNA samples were stored in -80°C until use. The HIV PR-RT region was amplified by one-step RT-PCR (TAKARA One Step RNA PCR kit) with a forward primer (DRPRO5: AGA CAG GYT AAT TTT TTA GGG A) and a reverse primer (DRRT34: GCT ATT AAG TCT TTT GAT GGG TCA TA). RT-PCR amplification conditions were 55°C for 40 minutes for the RT reaction followed by 40 cycles of 95°C for 10 seconds, 52°C for 5 seconds, and 72°C for 90 seconds. In the individual that the amplification of RT-PCR did not generate sufficient template, nested-PCR was performed using a forward primer (PROFWD1F: CAG ATC ACT CTT TGG CAA CGA CC) and a reverse primer (GEN4R: ATC CCT GGG TAA ATC TGA CTT GC) [30]. Nested-PCR amplification conditions were 94°C for 1 minute and 30 cycles of 94°C for 10 seconds, 55°C for 4 seconds and 74°C for 15 seconds.

All samples were analyzed by bulk sequencing employing our in-house protocol as reported elsewhere [30]. In brief, protease-RT regions (1144bp) were RT-PCR amplified followed by nested-PCR. Sequence reaction was performed by BigDye Terminator v3.1 Cycle Sequencing Kit, and analyzed by ABI-3100 auto sequencer. Drug resistance mutations were defined according to the mutation list proposed by Stanford HIV Drug Resistance Database [31].

To detect low-frequency subpopulations with drug resistance mutations, we used highly sensitive AS-PCR validated for subtype B HIV as described previously [14]. Briefly, mutation-specific primers were designed for seven reverse transcriptase inhibitor resistance mutations, M41L, K65R, K70R, K103N, Y181C, M184V, and T215F/Y (S1 Table). According to the specificity of the primers, T215F primer cross reacted with T215L/I/ L/V, and T215Y primer with T215D/N /Y (S1 Table). The HIV-1 total copy primers, ComFWD and ComREV, that span nucleotides 258–420 in RT were used with the common probes, Com1P and Com2P (S1 Table) [14]. For multiple mutation screening, several resistance mutation-specific reactions can be performed simultaneously. The cycle number at which the fluorescence emission exceeds the background fluorescence threshold is the threshold cycle (CT) and is the unit of measure for comparing the differences in amplification signals (ΔCT) between the total copy and mutation-specific reactions. All samples were tested in duplicate with the means of the total copy and mutation-specific CTs used for the determination of the ΔCT. Each ΔCT cutoff value for interpreting the presence of drug resistance mutations was determined previously [14] and were between 8.5 to 10.5 amplification cycles, for validated assay cut-offs ranging from 0.3% to 2.0% mutant, depending on the assay. AS-PCR for detection of each drug-resistant mutation was performed following the conditions previously described [30].

To evaluate the genetic relationship between minority-level K103N variants detected by AS-PCR and those detected by bulk sequencing, amplicons from AS-PCR reactions that yielded positive results were sequenced and phylogenetic trees were constructed. To increase the genetic sequence information for phylogenetic analyses, we extended the K103N-specific amplicon to the 181-184.REV reverse primer, which allowed amplification of RT codons 103 to 220 (336 bp) (S1 Table). The T215I-positive amplicons were similarly analyzed with amplicons spanning codon 215 to 101. To avoid possible bias by K103N and T215I mutation in the phylogenetic analysis, K103N and T215I codon positions were removed from the sequences before analysis. The sequences were aligned by the Clustal W program with a set of reference sequences recommended by the Los Alamos sequence database (http://www.hiv.lanl.gov/content/index). Maximum likelihood phylogeny was inferred by MEGA6 [32] based on the General Time Reversible model with gamma distribution (5 categories) and invariable sites [33].

Drug resistance mutations were detected by bulk sequencing in all 6 individuals who continued to experience insufficient viral suppression despite receiving salvage therapies in attempts to control viremia (Fig 1A–1F), (S2–S7 Tables). AS-PCR detected drug resistance mutations in 5 of 6 individuals in time points that had previously undetectable mutations by bulk sequencing as shown in S2–S7 Tables.

K70R, K103N, M184V and/or T215I were detected at minority levels in 4 of 6 individuals several years prior to detection by bulk sequencing (Individual 2, 3, 4 and 5). In individual 2, K70R was detected as minority population by AS-PCR prior to bulk sequencing and continued to persist for >1000 days. (Fig 1B and S3 Table). In individual 3, K70R, K103N and T215I were detected as minority populations by AS-PCR, though not continuously, at least once before these mutations became detectable by bulk sequencing (Fig 1C and S4 Table). K103N and T215I were detected by bulk sequencing soon after the initiation of regimen which included EFV, followed by slight decrease of plasma VL (Fig 1C). Evidence of K103N and T215I minority-level populations were identified in specimens collected 30 and 16 months, respectively, prior to the time point at which resistance was detected by direct sequencing. Similarly, in individual 4, K103N was detected as a minority subpopulation 14 months prior to becoming detectable by bulk sequencing. (Fig 1D). In individual 5, M184V was detected by AS-PCR nearly 6 months prior to detection by bulk sequencing (Fig 1E). In individual 6, drug resistance mutations were detected as major population throughout the observation period, and no minority drug resistance mutations were detected (Fig 1F).

Persistence of drug-resistance variants after withdrawal of ARV drug-selective pressures were observed by AS-PCR in 4 of 6 individuals (Individual 1, 2, 3, and 4). In individual 1, K103N was selected during nevirapine treatment and continuously detected for 8 months. The mutation became undetectable by bulk sequencing 6 months after the treatment was switched to a protease inhibitor containing regimen. However, a K103N-possessing variant persisted at low frequency for nearly 2 years without nevirapine administration and was re-selected when EFV was used (Fig 1A). Similar patterns of minority resistance persistence were observed in other individuals; such as Y181C and M184V. In individual 2, K103N was detected twice by AS-PCR but never became major population (Fig 1B). In individual 4, M184V was detected sporadically as a low-frequency variant after 3TC was withdrawn from their regimen, suggesting a 3TC-selected minority population had persisted around the threshold of AS-PCR detection limit. M41L was also found as minority population in individual 3 by AS-PCR. It persisted for more than 2 years after it became undetectable by bulk sequencing (Fig 1C).

To clarify whether minority-level drug-resistant variants retrospectively detected by AS-PCR were the origins of majority resistant variants possessing the same mutations that later emerged, phylogenetic relatedness was assessed. In the phylogenetic tree for individual 3, the ID9 (day 957) AS-PCR amplicon (K103N positive) clustered with the K103N bulk sequence at ID 22–24 (days 2046–2117), suggesting that the K103N minority population at ID 9 had high genetic identity to the K103N-positive population that subsequently arose at ID22-24 (Fig 2A, S4 Table). A similar finding was observed for T215I positive populations in Individual 3. Two T215I-positive AS-PCR sequences on ID 17(day 1731) and 21 (day 1930) were clustered with ID 25–27 and 31(days 2215–2303, 2552), suggesting their close genetic relationships (Fig 2B, S4 Table). In Individual 4, the sequence of the K103N-positive amplicon detected at ID 2 (day 102) was within the cluster of K103N-positive sequences collected at later period ID 9–26 (days >917) (Fig 2C, S5 Table). The results of the three phylogenetic analyses suggested that pre-existing minority drug resistances may affect subsequent treatment outcome, supporting clinical importance of detecting minority-levels of drug-resistant variants prior to treatment changes.