Detection of Low-Frequency HIV Type 1 Reverse Transcriptase Drug Resistance Mutations by Ultradeep Sequencing in Naive HIV Type 1-Infected individuals

Antiretroviral therapy (ART) has considerably reduced the morbidity and the mortality of HIV-infected people. ¹ The improvement of clinical management has been made possible by the development of well-tolerated antiretroviral drugs targeting different steps of the HIV life cycle but also by the molecular characterization of the viral population at the individual level.

Guidelines recommend HIV drug resistance testing before initiation of therapy and at treatment failure. ² The genotypic testing for drug-resistant HIV variants guides the choice of the appropriate ART since the presence of drug-resistant mutations (DRMs) can impair virological outcome. This is particularly challenging before the prescription of a first-line therapy. Epidemiological studies have reported that in most industrialized countries, the prevalence of transmitted drug-resistant viruses was around 10%, representing a nonnegligible proportion of patients presenting a risk of preexisting antiretroviral (ARV) resistance. ^3,4 However, this prevalence may be underestimated. In these studies, transmitted DRMs were detected using a standard approach to genotypic resistance testing, i.e., population-based sequencing that has a detection limit of 20% of the virus quasispecies. ⁵ Below this threshold, low-frequency variants carrying additional DRMs were not detected and, thus, could imply a bias in genotypic resistance test interpretation.

Recently, more sensitive methods have been developed to reveal the presence of these low-DRMs such as ultradeep pyrosequencing. This technology is used for research purposes only since the clinical relevance of these low-DRMs remains controversial. However, several studies agreed that the presence of low-frequency nonnucleoside reverse transcriptase inhibitor (NNRTI) DRMs significantly increased the risk of virological failure with NNRTI-based regimens. ⁶

The aim of our study was a survey of low-DRMs in HIV-1-infected individuals eligible for first-line therapy at the University Hospital of Bordeaux, France, in 2011–2012. We used ultradeep pyrosequencing (UDPS) to estimate the prevalence of reverse transcriptase (RT) DRMs.

Participants included in this study were HIV-1-infected individuals followed up at the Bordeaux University Hospital. All were adults with HIV-1 infection confirmed by Western blot testing and naive to ART. Forty-seven, for whom a genotypic resistance test was requested by the physician in the context of the discovery of HIV-1 infection or before ART prescription, were selected for UDPS analysis. Since there are, of course, no recommendations for treatment initiation using UDPS data, the therapeutic decision was based on the presence of DRMs above the threshold of 20%. Thus, our work must be considered as reflecting the situation in a French hospital in 2011–2012. Thirty-one patients (65.9%) were infected with B subtype viruses and 16 with non-B subtypes including eight CRF02_AG, one G, one CFR01_AE, one F1, and six with more complex recombinant forms. Median baseline CD4 cell counts and median viral load (VL) were 264 (range: 3–699) CD4 cells/mm³ and 5.03 (range: 2.78–6.32) log₁₀ copies/ml, respectively.

Two independent strategies were used to obtain amplicons encompassing the RT region (amplicon range size: 300–400 bp). In both cases, HIV RNA was extracted from 1 ml of plasma using the High Pure Viral RNA kit (Roche Diagnostics) and was used as a template for cDNA generation. The first strategy was a homebrew protocol (M-A. Vandenhende et al., unpublished) and allowed the generation of three amplicons, RT1, RT2, and RT3. In the second strategy, three amplicons (B, C, and D) were generated according to the protocol provided with the second generation of the Roche prototype primer plates allowing for the processing of 10 samples per polymerase chain reaction (PCR) plate. All amplicons harbored Roche 454 Life Sciences specific sequencing adaptors A and B and specific MIDs. PCR products were purified using magnetic beads (Agencourt AMPure Kit, Beckman Coulter, Berried, Germany) and quantified by fluorometry (Quant-iT PicoGreen dsDNA assay kit, Invitrogen, Carlsbad, CA). An emulsion-based clonal amplification (emPCR amplification) using a GS Junior Titanium emPCR Kit (Lib-A) was then performed using equimolar concentrations of amplicons. UDPS was performed in both the forward and reverse directions on a Roche 454 Life Science GS Junior Instrument. Amplicon Variant Analyzer (AVA 2.7) software (Roche 454 Life Sciences) was used to align all amplicon reads and variant frequencies at each nucleotide position relative to the sequence of HIV-1 reference strain HxB2 were calculated. Raw data were submitted to GenBank under accession number SRP026411. An average of 1,640 reads per amplicon was obtained independent of the protocol used for amplicon generation.

Drug resistance mutations detected by UDPS are listed in Table 1. We focused our analysis on NRTI and NNRTI mutations listed for transmitted drug resistance surveillance in 2009. ⁷ DRMs were accepted as significant variants when present at a frequency ≥1% among the total number of reads and if they were present in both sequence directions. This threshold took into account data from the literature ⁸ and previous results from our group ⁹ (M-A. Vandenhende et al., unpublished). No surveillance transmitted drug mutation was detected for 39 of the 47 individuals. This high prevalence of individuals without any mutations was in accordance with the results obtained in Netherlands by Pingen and colleagues. ¹⁰ Low-abundance NNRTI resistance mutations were rarely observed in ARV-naive individuals by several others groups. ^11,12

OPEN IN VIEWER

Table 1.

List of Drug Resistance Mutations Detected in 12/47 Individuals

		1% <DRM <20%		DRM >20%
Individual number	Subtype	UDPS mutations	Mutational VL (cp/ml)	UDPS mutations	Mutational VL (cp/ml)	Initial ART	Follow-up
1	CFR02_AG	M184V (3.62%) K103N (2.91%)	25,608 20,584	—	—	TDF/FTC NVP	VF at M6: M184V (94.4%) and K103N (96.4%)
2	B	M230L (2.04%)	88	—	—	Not treated	—
3	B	K103N (11.76%)	20,565	E138A (94.66%)	165,539	Not treated	—
4	URF	M184V (16.82%)	64,370	—	—	3TC/ZDV NVP	VL <50 at M6
5	G	—	—	M184I (92.81%) Y181C (93.13%)	87,500 87,811	TDF/FTC EFV	Switch at M2 to TDF/FTC/RAL
6	B	—	—	T215D (99.85) T210W (100.00%)	4,619 4,628	TDF/FTC ATZ/r	VL <50 at M6
7	F1	—	—	K103N (99.82%);	111,698	Not treated	—
8	CRF22_01A1	K70E (19.40%) T215I (3.85%)	3,371 669	M184V (89.00%) K103N (83.41%) Y188L (88.19%)	15,467 14,496 15,326	Not treated	—
9	CRF02_AG	—	—	E138A (99.18%)	281,572	TDF/FTC NVP	VL <50 at M6
10	B	—	—	E138A (89.41%)	1,884,637	TDF/FTC RAL	NA
11	B	E138A (1.52%) E138G (1.71%)	240 270	—	—	Not treated	—
12	B	E138A (11.98%) E138K (1.25%)	19371 2021	—	—	TDF/FTC DRV/r	NA
13 to 47	25 B 10 non-B	—	—	—	—	19 not treated  16 treated  1 TDF/FTC  NVP  12 TDF/FTC  DRV/r  2 TDF/FTC  ATZ/r  1 TDF/FTC  RPV	—

RT UDPS was performed on HIV RNA obtained from a plasma sample collected during the first visit at the hospital. NRTI and NNRTI mutations listed for transmitted drug resistance surveillance as well as rilpivirine resistance mutations defined by ANRS were reported. The mutational viral load was the absolute number of mutated viral RNA copies. First-line treatment prescribed as well as the follow-up were also indicated. M2 and M6 correspond to month 2 and 6 after ART initiation. NA, not available.

UDPS, ultradeep pyrosequencing; VL, viral load; ART, antiretroviral treatment; RT, reverse transcriptase; TDF, tenofovir; FTC, emtricitabine; NVP, nevirapine; 3TC, lamivudine; ZDV, zidovudine; EFV, efavirenz; ATZ/r, atazanavir/ritonavir; RAL, raltegravir; DRV/r, darunavir/ritonavir; RPV, rilpivirine.

High-abundance mutations (>20%), i.e., mutations that could have been detected by population sequencing, were observed in four individuals. Individual 5 was infected with the G subtype virus harboring the NRTI mutation M184I (92.81%) and the NNRTI mutation Y181C (93.13%). Deep sequencing of plasma viral RNA from individual 6 revealed high levels of the TAMs T215D and T210W. The prevalence of NNRTI mutation K103N was 99.89% in plasma from individual 7, infected with the F1 virus. One individual (8) was infected with a recombinant form of HIV-1-carrying DRMs both at high and low abundance: M184V (89.00%), K103N (83.41%), Y188L (88.19%), K70E (19.40%), and T215I (3.85%). The size and location of the designed amplicons allowed us to argue that the M184V and K103N mutations were linked on the same genome and that the K70E mutation was also present in 20% of these HIV variants. This notion of linkage could be important since some combinations of DRM, such as K65R and M184V, impaired replication capacity ¹³ or are needed to confer high resistance to etravirine or rilpivirine. ^14,15 By contrast with allele-specific PCR, UDPS technology has the ability to sequence amplicons that can cover the region encompassing both mutations and, then, conclude with the presence of DRMs on the same genome.

We could hypothesize that these individuals, at least those who carried viruses with a high level of M184I/V and coming from Africa, have already been treated (5 and 8). Nevertheless, the prevalence of individuals presenting high-level mutations leading to resistance to one RT drug was 8.5%, which is comparable to the prevalence observed in France in patients who are ART therapy naive. ⁴

Four individuals (8.5%) were infected with viruses harboring DRMs detected by UDPS only. We observed the presence of the M184V mutation in two non-B subtype viruses at the 3.62% and 16.82% level. The low-abundance NNRTI K103N mutation was also detected in two samples (2.91% and 11.76%) and 2.04% of M230L was quantified in one individual infected with a B subtype HIV-1. These low-frequency variants could arise due to the highly error-prone replication of HIV or direct transmission of resistant variants. The proportion of variants possessing mutations was high in three of the four individuals harboring only low-frequency DRMs, with a mutational load higher than 20,000 copies per ml (cp/ml) (Table 1). Previous work has demonstrated a strong correlation between the presence of the K103N mutation and a viral load of 2,000 cp/ml and virological failure. ¹⁶ Among our group of individuals who received treatment, we observed virological failure for individual 1. This individual was treated with tenofovir, emtricitabine, and nevirapine since initial population sequencing did not show any DRMs. Virological failure occurred 6 months later, explained by the emergence of K103N and M184V at >90%. Retrospective UDPS analysis of a sample collected before ART initiation revealed low-frequency mutations K103N (20,584 mutated cp/ml) and M184V (25,608 mutated cp/ml) suggesting the presence of preexisting DRMs that emerged under ART selective pressure.

In addition to the first generation NNRTI mutations listed for transmitted drug resistance surveillance, we looked for the mutations in HIV-1 RT that have been observed from rilpivirine (RPV)-treated patients with virological failure. ¹⁷ We noticed the presence of RPV-resistant mutations in nine (19.1%) individuals (Table 1). High-abundance RPV resistance mutations were found in five individuals. In two of the five cases, the DRMs were common to all the NNRTI drugs (cross-resistance). A high level of the rilpivirine E138A mutation was observed in three individuals. The same E138A mutation was detected at low abundance, i.e., 1.52% and 11.98%, in two other samples (data not shown). The E138K and E138G mutations were also found at 1.25% and 1.71%, respectively, after deep sequencing analysis of the viral population. With the commercialization of Eviplera (Gilead), which is recommended for patients who have not received anti-HIV treatment before, it is important to estimate the prevalence of minority RPV resistance mutations in primary HIV-1-infected patients. Nicot et al. have shown that major RPV-associated resistant mutations were detected in 15% of their population study. ¹² Although their impact on virological outcome has not been investigated yet, they may increase the risk of virological failure for a regimen containing RPV, as has been suggested for efavirenz-associated and nevirapine-associated resistant mutations. ⁶

Is there an added value for performing deep-sequencing analysis compared to a conventional population sequencing in clinical routine practice? Our results showed that data on low-frequency DRMs have to be taken into consideration before a first -line ART. The case of the individual with preexisting M184V and K103N minority mutations who failed on the NRTI+NNRTI regimen illustrated the importance of these low-frequency mutations for treatment response. Further studies are still needed to evaluate the clinical significance of each DRM, but we demonstrated that UDPS in clinical routine practice could be useful to guide a first-line ART prescription.