Antiviral Therapy 2011; 16:925 929 (doi: 10.3851/IMP1851) Short communication Prevalence of low-level HIV-1 variants with reverse transcriptase mutation K65R and the effect of antiretroviral drug exposure on variant levels Michael J Kozal 1 *, Jennifer Chiarella 1, Elizabeth P St John 2, Elizabeth A Moreno 2, Birgitte B Simen 2, Todd E Arnold 2, Max Lataillade 1,3 1 Yale University School of Medicine and Veterans Affairs Connecticut Healthcare System, New Haven, CT, USA 2 454 Life Sciences A Roche Company, Branford, CT, USA 3 Research and Development, Bristol Myers Squibb, Wallingford, CT, USA *Corresponding author e-mail: Michael.kozal@yale.edu Background: It has been reported that treatment-naive individuals infected with HIV-1 subtype C may be more likely to harbour viral variants possessing a K65R reverse transcriptase gene mutation. The objectives of this study were to determine the prevalence of low-level K65R variants within different HIV-1 subtypes and to assess the effects of antiretroviral exposure on K65R variant levels. Methods: Treatment-naive individuals infected with different HIV-1 subtypes were genotyped by ultra-deep sequencing. Samples were evaluated for low-level variants to 0.4% or 1% levels depending upon viral load. Estimated mutational load was calculated by multiplying the percentage of the variant by the plasma viral load. Results: A total of 411 treatment-naive individuals were evaluated by ultra-deep sequencing to 1% levels; 4 subjects (0.97%) had K65R variants at 1% or had a very high mutation load. All four subjects had variants with linked drug resistance mutations suggesting transmitted resistant variants. 147 ARV-naive subjects were sequenced to 0.4% levels; 8.8% (13/147) had K65R low-level variants identified: 2.2% (2/92) in subtype B, 35.7% (10/28) in subtype C (P<0.001 for B versus C) and 3.7% (1/27) in non-b/c subtypes. The 13 ARV-naive subjects with K65R variants at <1% received tenofovir plus emtricitabine plus a ritonavir-boosted protease inhibitor (TDF+FTC+PI/r) and 5 subsequently experienced virological failure. There was no enhancement in K65R levels by percentage or mutational load compared to pre-therapy levels. Conclusions: Low-level K65R variants were more frequently identified in subtype C. K65R variants at >1% levels likely represent transmitted resistant variants. The clinical implication of low-level K65R variants below 1% in treatment-naive subjects who receive TDF+FTC+PI/r remains to be determined as the majority are very lowlevel and did not increase after antiretroviral exposure. Introduction The codon K65R mutation in HIV-1 reverse transcriptase gene (RT) is a major resistance mutation conferring intermediate resistance to the antiretroviral agents tenofovir (TDF), emtricitabine (FTC), abacavir (ABC) and lamivudine (3TC), and low level resistance to stavudine (d4t) [1 3]. Previous research has found that HIV-1 subtype C virus may be more likely than other subtypes to harbour K65R variants, possibly because of template-dependent mutagenesis [4]. Studies have shown an increased frequency of K65R in subtype C antiretroviral therapy failures (notably with two nucleoside reverse-transcriptase inhibitor [NRTI] regimens and WHO recommended regimens containing two NRTIs and a non-nucleoside reverse-transcriptase inhibitor [NNRTI]) [5 8]. Several researchers have recently reported increased rates of RT K65R in HIV subtype C and investigated the possible mechanisms [9 11] contributing to this phenomenon. These reports suggest that the HIV RT codon 64 66 region may be a hot spot [9] for mutations with template-dependent mutagenesis [11], decreased transcription RT fidelity [10], frame shifts [9], polymerization assay and pyrosequencing errors [10] all possibly contributing to the increased detection of K65R in subtype C. These authors raised an important question about the clinical significance of low-level K65R variants in subtype C and identified a need for better clinical data 2011 International Medical Press 1359-6535 (print) 2040-2058 (online) 925
MJ Kozal et al. with longitudinal results to assess the persistence and impact of these variants on ART outcomes. Given our large repository of ultra-deep sequencing data from well-characterized patient cohorts, we set out to determine the prevalence of K65R in different HIV-1 subtypes, evaluating level of detection by ultradeep sequencing, and the effects of ART selection pressure on K65R variant levels to help determine if K65R detection is dependent upon subtype, assay or both, and if low-level K65R variants are clinically significant. Methods To determine the prevalence of low-level K65R variants within different HIV-1 subtypes infecting ARVnaive subjects and the effects of ARV-exposure on K65R variant levels, 411 unique subjects from multiple studies [12 14] were evaluated by ultra-deep sequencing. Baseline specimens prior to the initiation of therapy from 411 ART-naive subjects with longitudinal data and 60 virological failure specimens from ARV-experienced subjects were included. Ultra-deep sequencing was performed as described in previous studies [12 14]. Samples were further evaluated for low-level variants to 0.4% depending upon HIV viral load. HIV viral load is proportional to the sensitivity of the sequencing results. Ultra-deep sequencing results to 0.4% were accepted for a sample if the viral load was greater than 30,000 HIV RNA copies/ml and if the number of sequencing reads through RT codon 65 were greater than 700. The vast majority of samples had HIV viral loads >100,000 copies/ml (mean 230,465 copies/ml; IQR 42,300 378,500) and >1,000 clonal sequencing reads through RT codon 65. Sampling of viral RNA molecules from an extraction solution that follows Poisson distribution is subjected to the stochastic effects of sampling variation [13,15,16]. These effects are lessened when the number of template RNA molecules increases (typically >1,000) [13,15]. Since current HIV RNA extraction techniques may only recover 10% of viral genomes from a plasma sample [13,16], specimens with starting HIV RNA copy numbers >10,000 copies/ml typically have sufficient extracted RNA to allow for ultra-deep sequencing to be attempted, but as the viral load decreases the ability to obtain a fully representative sample of HIV RNA templates decreases significantly [13 16]. One can still perform ultra-deep sequencing of low HIV RNA copy number samples; however, the levels of mutations identified represent the proportion of sequenced PCR amplicons containing the mutation and may or may not represent the actual proportion in the plasma sample. Ultra-deep sequencing results were analysed by the depth of sequencing and by viral load. An estimated Table 1. Treatment-naive HIV-1-infected individuals with a K65R variant at >1% or with a high mutational load K65R variant levels (estimated variant HIV-1 subtype mutational load) a Additional drug resistance mutations identified in sample C 3.11% (23,325) NRTI: K65R (3.11%), K65R (0.69%), M184V (13.79%), L210W (9.27%), 0.69% (5,175) bc T215Y (9.1%), K219Q (4.66%) NNRTI: G190A (4.57%) PI: L10I (3.22%), L10I (77.13%), I13V (72.43%), L33F (2.39%), M36I (82.5%), L63P (1.06%), H69K (94.02%), I93L (40.99%), I93L (4.6%) C 1.22% (684) NRTI: K65R (0.49%), K65R (1.22%) 0.49% (275) b PI: L10F (1.37%), I13V (1.36%), G16E(1.47%), K20R (96.19%), K20V (1.76%), V32I (1.72%), M36I (98.29%), M46I (2.33%), H69K (88.89%), I93L (97.44%) B 0.9% (15,481) c NRTI: K65R (0.9%), D67E (2.6%) K70R (67%), K70S (14%), K70T (10%), V75I (97%), F77L (98%), F116Y (99%), Q151M (98%) NNRTI: V108I (3.1%) BF 7.85% (693) NRTI: M41L (7.47%), K65R (7.85%), K65R (0.31%), D67N (33.78%), 0.31% (27) b K70R (6.34%), Y118I (1.65%), L210W (16.74%), T215C (20.98%), K219N (22.46%) NNRTI: K103N (16.36%), K103R (10.27%), V179D (73.28%), Y181C (2.34%) PI: L10R (2.51%), M36I (91.67%), M46I (3.3%), D60E (2.51%), L63P (3.28%), I64L (92.88%), V77I (1.79%), V82I (3.58%), V82I (89.09%), I93L (90.53%) ARV-naive subjects (n=411) were evaluated by ultra-deep sequencing for K65R to 1% levels; four subjects (0.97%) had K65R variants at 1% or had a very high mutation load. All four subjects with K65R variants at high levels had sequences with linked drug resistance mutations suggesting transmitted resistant variants (for example, multiple linked nucleoside reverse transcriptase inhibitor [NRTI], non-nucleoside reverse transcriptase inhibitor [NNRTI] or protease inhibitor [PI] mutations). a An estimated variant mutational load was calculated by percent of variant detected sample plasma HIV viral load. b Additional K65R variant with different codon change detected on analysis (for example, AGA and AGG codon changes detected). c Experienced virological failure with same K65R triplet detected. 926 2011 International Medical Press
Low level K65R variants and ART mutational load was calculated as the variant frequency multiplied by the sample viral load. Results Among 411 ART-naive individuals, ultra-deep sequencing detected K65R at levels 1% in only four subjects (0.97%), including two patients with subtype C, one with subtype B and one with subtype BF. In these four subjects with a K65R there were other resistance mutations (NRTI, NNRTI or protease inhibitor [PI]) that were linked with other mutations in the template sequences suggesting that these were transmitted resistance mutations (Table 1). Of the 411 treatment-naive subjects, 147 (36%) were sequenced to 0.4% to <1% depths. The average viral load for these samples was 230,465 copies/ ml (IQR 42,300 378,500). Of these 147, 13 (8.8%) had K65R low-level variants identified. These 13 samples with a K65R variant between 0.4 to 1% had a mean HIV viral load of 212,215 copies/ml (IQR 52,100 341,000) with an average number of ultradeep sequencing reads through K65R of 1,563 clonal sequences (IQR 1,032 1,911). The average percent of sequence reads with K65R was 0.58% (range 0.41 0.76%) with an average estimated mutational load of 1,299 (range 147 4,510). Among the different HIV-1 Figure 1. Prevalence of low level HIV-1 variants with a K65R mutation in the reverse transcriptase gene at <1% levels among different HIV-1 subtypes 90 80 70 60 50 40 30 20 10 0 Percentage of samples with K65R at <1% levels100 Subtype B P<0.001 K65R varient present at >0.4 to <1% level Non-subtype B/C Of the 147 subjects with ultra-deep sequencing performed to levels between 0.4 to 1%, 13 (8.8%) had a reverse transcriptase K65R low-level variant identified. K65R variants were identified in 2.2% (2/92) of subtype B, 35.7% (10/28) in subtype C (P<0.001 for B versus C) and in 3.7% (1/27) in non-b/c subtypes (BF, A, AE and F1). subtypes, subjects with a low-level K65R variant <1% were identified in 10 of 28 (35.7%) with subtype C compared to 2 of 92 (2.2%) with subtype B (P<0.001; Figure 1) and 1 of 27 (3.7%) in non B or C subtypes. The effects of ARV-exposure on the 13 subjects with a K65R at <1% level are presented in Figure 2. No enhancement of the K65R variants were seen in subjects treated with TDF+FTC+ ritonavir-boosted protease inhibitor (PI/r) in two years of follow up. Discussion In this study, low-level K65R variants were more frequently identified in subjects infected with HIV-1 subtype C. K65R variants at >1% levels likely represented transmitted resistant variants given that multiple resistance mutations were identified in each subject and some of the resistance mutations were linked within the same variant. This finding is similar to the recent report by Lipscomb et al. [17] who, using a different sensitive genotyping method, reported the identification of transmitted low-level K65R variants possessing additional resistance mutations. Two of the four subjects in our study with K65R >1% experienced virological failure on TDF+FTC+PI/r. The subjects that experienced virological failure had high estimated K65R mutational loads and other drug resistance mutations present in the infecting quasispecies. However, the implication of low-level K65R variants below 1% in treatment-naive subjects who receive TDF+FTC+PI/r remains to be determined as the majority of variants were at very low level (mean 0.58%), had low estimated K65R mutational loads and did not increase after ARV exposure with drugs known to select for K65R. The lack of K65R variant enhancement suggests that other active components of the regimen may be able to prevent selection and/ or alternatively their identification may be a result of viral or assay polymerization errors especially in subtype C as described by others [9,18]. The finding that K65R is identified more frequently in subtype C confirms the previous findings described by others using different [9 11] and similar sensitive genotyping methods [18]. As with other sensitive genotyping methods it is known that ultra deep pyrosequencing has important limitations [16]. HIV RNA extraction and complementary DNA synthesis limits, assay polymerization and pyrosequencing errors may all lead to an incorrect low level variant detection. For example, the typical pyrosequencing error is a single base over or under call of homopolymer runs [18]. There are two A-stretches within HIV-1 pol. Due to the nucleotide flow order, pyrosequencing error may occur only in one direction (forward or reverse) or look different in the two directions; if found in both directions, Antiviral Therapy 16.6 927
MJ Kozal et al. Figure 2. Antiretroviral treatment outcome and K65R mutational load for 13 patients with a K65R at <1% levels who were treated with TDF+FTC+PI/r 13 ARV-naive subjects with K65R variants at <1% levels Initiated TDF+FTC+PI/r (10C, 2B and 1AE) No virological failure in 8 subjects after 2 years (7C, 1AE) 5 Virological failures; 4 had UDS performed One subtype B at VF VL 50 No UDS performed BL: VL 88,200; K65R: 0.62% (ML 547) VF: VL 4,520; K65R: 0.56% (ML 25) Subtype B BL: VL 750,000; K65R: 0.47% (ML 3,525) VF: VL 3,780; No K65R on UDS BL: VL 52,100; K65R: 0.55% (ML 287) VF: VL 97,400; K65R: 0.76% (ML 740) BL: VL 204,000; K65R: 0.76% (ML 1,550) VF: VL 180,000; K65R: 1.28% (ML 2,304) 0.52% (ML 936) The figure summarizes the baseline (BL) and virological failure (VF) K65R variant quantities in individuals with a variant present at <1% levels at baseline prior to initiating tenofovir plus emtricitabine plus ritonavir-boosted protease inhibitor (TDF+FTC+PI/r). In total, 8 of the 13 subjects with K65R (7 with subtype C) did not have virological failure after 2 years of therapy. The remaining five subjects experienced virological failure, including one patient with subtype B whose virus was not analysed by deep sequencing at failure due to a very low HIV viral load (VL [copies/ml]; 50 copies/ml). Of the remaining four subjects with virological failure, one had subtype B and three had subtype C, all had K65R variants below 2% frequency by ultra-deep sequencing (UDS) and the variants were not detected by standard genotyping. ML, mutational load. errors are more likely due to PCR error or alignment. When calling K65R codons in subtype C, the highest confidence can be given to occurrences of the AGG codon (two concurrent nucleotide changes from wild type), and variants that are detected in both directions and above the expected PCR error rate. We had chosen 0.4% as the lowest detection level as it is more than 3 times greater than PCR error, but error occurring in early rounds of amplification may still significantly affect results [18]. Another limitation of this study is that the estimated mutational load was calculated by multiplying the results from two separate molecular assays - HIV viral load and mutant variant frequency by ultra-deep sequencing. Different PCR primer sets are used in the two assays and the HIV RNA templates may or may not have been similar, thus the mutational load values should be viewed only as an estimate. The data from our study suggest that K65R variants at levels below 1% did not enhance after exposure to TDF+FTC+PI/r. However, the lack of K65R enhancement after treatment exposure may not be the same for other antiretroviral treatment regimens with lower genetic barriers as seen with WHO-recommended first line regimens (two NRTIs plus NNRTI). Boosted-PIbased therapy has been reported to protect against the development of K65R and virological failure, having lower rates than with older NRTI- and NNRTI-based regimens [19 21]. Thus, clinically relevant K65R levels for specific antiretroviral treatment regimens will require individual study. The impact of low level resistant variants on therapy is likely a multifactorial process with contributions from variant mutational load, mutation linkage and the genetic barrier of the regimen all contributing. Acknowledgements These data were presented in part at the International HIV & Hepatitis Virus Drug Resistance Conference 8 12 June 2010, Dubrovnik, Croatia (abstract number 26). Disclosure statement Yale University receives grant support from Merck, Pfizer, Gilead, Abbott and Bristol Myers Squibb for studies that MJK serves as the primary investigator. MJK receives royalties from a patent owned by Stanford University for some HIV diagnostic tests. ML is an 928 2011 International Medical Press
Low level K65R variants and ART employee of Bristol Myers Squibb. EPSJ, EAM, TEA and BBS are employees of 454 Life Sciences. JC declares no competing interests. References 1. Rhee SY, Gonzales MJ, Kantor R, Betts BJ, Ravela J, Shafer RW. Human immunodeficiency virus reverse transcriptase and protease sequence database. Nucleic Acids Res 2003; 31:298 303. 2. Lanier R, Moffatt A, Stone C, et al. Comparison of traditional plasmid-based clonal sequencing to 454 ultra-deep sequencing for HIV clinical isolates with reverse transcriptase (RT) mutations at K65R and L74V. Antivir Ther 2007; 12 Suppl 1:S150. 3. Underwood MR, Ross LL, Irlbeck DM, et al. Sensitivity of phenotypic susceptibility analyses for non thymidine nucleoside analogues conferred by K65R or M184V in mixtures with wild-type HIV-1. J Infect Dis 2009; 199:84 88. 4. Coutsinos D, Invernizzi CF, Xu H, et al. Template usage is responsible for the preferential acquisition of the K65R reverse transcriptase mutation in subtype C variants of human immunodeficiency virus type 1. J Virol 2009; 83:2029 2033. 5. Doualla-Bell F, Avalos A, Brenner B, et al. High prevalence of the K65R mutation in human immunodeficiency virus type 1 subtype C isolates from infected patients in Botswana treated with didanosine-based regimens. Antimicrob Agents Chemother 2006; 50:4182 4185. 6. Brenner BG, Oliveira M, Doualla-Bell F, et al. HIV-1 subtype C viruses rapidly develop K65R resistance to tenofovir in cell culture. AIDS 2006; 20:F9 F13. 7. Hosseinipour MC, van Oosterhout JJ, Weigel R, et al. The public health approach to identify antiretroviral therapy failure: high-level nucleoside reverse transcriptase inhibitor resistance among Malawians failing first-line antiretroviral therapy. AIDS 2009; 23:1127 1134. 8. Sungkanuparph S, Manosuthi W, Kiertiburanakul S, Saekang N, Pairoj W, Chantratita W. Prevalence and risk factors for developing K65R mutations among HIV-1 infected patients who fail an initial regimen of fixed-dose combination of stavudine, lamivudine, and nevirapine. J Clin Virol 2008; 41:310 313. 9. Li J-F, Lipscomb J, Wei X, et al. Detection of minority K65R variants in NRTI-naive subtype B and C HIV-1- infected individuals. J Infect Dis 2011; 203:798 802. 10. Varghese V, Wang E, Babrz F, et al. Nucleic acid template and the risk of a PCR-induced HIV-1 drug resistance mutation. PLoS ONE 2010; 5:e10922. 11. Coutsinos D, Invernizzi CF, Moisi D, et al. A templatedependent dislocation mechanism potentiates K65R reverse transcriptase mutation development in subtype C variants of HIV-1. PLoS ONE 2011; 6:e20208. 12. Simen BB, Huppler Hullsiek K, Novak RM, et al. Low abundance drug resistant viral variants in chronically HIVinfected antiretroviral-naive patients significantly impact treatment. J Infect Dis 2009; 199:693 701. 13. Le T, Chiarella J, Simen BB, et al. Low-abundance HIV drug-resistant viral variants in treatment-experienced persons correlate with historical antiretroviral use. PLoS ONE 2009; 4:e6079. 14. Lataillade M, Chiarella J, Yang R, et al. Prevalence and clinical significance of HIV drug resistance mutations by ultra-deep sequencing in antiretroviral-naive subjects in the CASTLE study. PLoS ONE 2010; 5:e10952. 15. Stenman J, Lintula S, Rissanen O, et al. Quantitative detection of low-copy-number mrnas differing at single nucleotide positions. Biotechniques 2003; 34:172 177. 16. Shafer RW. Low-abundance drug-resistant HIV-1 variants: finding significance in an era of abundant diagnostic and therapeutic options. J Infect Dis 2009; 199:610 612. 17. Lipscomb JT, Owen SM, Johnson JA. Dynamic expression of HIV-1 drug resistance mutations during acute infection. Antivir Ther 2010; 15 Suppl 2:A47. 18. Varghese V, Wang E, Babrzadeh F, et al. Nucleic acid template and the risk of a PCR-induced HIV-1 drug resistance mutation. PLoS ONE 2010; 5:e10992. 19. Waters L, Nelson M, Mandalia S, et al. The risks and incidence of K65R and L74V mutations and subsequent virologic responses. Clin Infect Dis 2008; 46:96 100. 20. von Wyl V, Yerly S, Böni J, et al. Factors associated with the emergence of K65R in patients with HIV-1 infection treated with combination antiretroviral therapy containing tenofovir. Clin Infect Dis 2008; 46:1299 1309. 21. Grant PM, Taylor J, Nevins AB, et al. International cohort analysis of the antiviral activities of zidovudine and tenofovir in the presence of the K65R mutation in reverse transcriptase. Antimicrob Agents Chemother 2010; 54:1520 1525. Accepted 13 January 2011; published online 13 July 2011 Antiviral Therapy 16.6 929