Drug resistance and tropism as markers of the dynamics of HIV-1 DNA quasispecies in blood cells of heavily pretreated patients who achieved sustained virological suppression

Abstract

Objectives

The objective of this study was to address the dynamics of archived resistant quasispecies in cell-associated HIV-1 DNA over time in heavily ART-experienced patients with currently suppressed plasma HIV-1 RNA.

Methods

Longitudinal ultra-deep sequencing (UDS) analysis of reverse transcriptase, protease and V3 Env regions was performed on blood-cell-associated HIV-1 DNA samples. Drug-resistance-associated mutations (DRAMs) and tropism were interpreted using the ANRS and Geno2Pheno algorithms. We analysed frozen blood cells from patients enrolled in the INNOVE and ANRS 123 ETOILE studies who achieved sustained viral suppression after salvage optimized ART (SOT).

Results

Samples were available at baseline and 6 and ≥12 months after SOT initiation in 10 patients. V3 loop sequences displayed wide intra-individual dynamics over time. Viral variants harbouring DRAMs exhibited three non-exclusive scenarios. First, when SOT exerted the same selective pressure as previous failing regimens, some viral quasispecies still harboured the same DRAMs at the same level as at the time of virological failure. Thus, as DRAMs were mostly associated with the same viral variant, variants with a complete resistance pattern remained archived. Second, some viral variants harbouring DRAMs were no longer detected over time when SOT consisted of new antiretroviral classes or had resistance profiles distinct from those of previous failing regimens. Third, variants with new DRAMs associated with SOT emerged in blood cells during follow-up despite sustained virological control.

Conclusions

Using longitudinal UDS analysis and focusing on DRAMs and tropism as markers, we demonstrated that, despite sustained virological control, archived HIV-1 DNA quasispecies continued to evolve.

Introduction

HIV-1 ART is a life-long treatment. While virological success is achieved in most patients, ART often needs to be modified due to adverse events, drug–drug interactions, simplification to improve adherence, or treatment failure. Treatment modification is complicated in heavily ART-experienced patients harbouring viruses with past drug-resistance-associated mutations (DRAMs) who are currently controlled under effective treatment. In these patients, in whom plasma HIV-1 RNA is suppressed and therefore cannot be amplified, genotypic resistance testing may be performed on blood-cell-associated HIV-1 DNA. Population HIV-1 DNA genotyping does not recapitulate exhaustively all the accumulated DRAMs previously reported from genotypes detected in plasma samples.^1,2 Therefore, ART history and cumulative past plasma HIV-1 RNA genotypes are more informative than HIV-1 DNA genotypes to guide treatment modifications.

It has been described that all circulating HIV-1 variants are archived in blood cells and other reservoirs, such as gut or semen, for an unknown duration.^3–6 Thus, archived resistant variants can persist for a lengthy period of time, as they are archived notably in central memory CD4 T cells^7,8 with a high proliferative potential and a long half-life.⁹

While the Sanger sequencing sensitivity to detect resistant variants within quasispecies of WT viruses is around 20%, ultra-deep sequencing (UDS) assay sensitivity allows for a sensitivity threshold as low as 1%.^10–13 Approaches based on UDS have been used to evaluate the impact of minority drug-resistant variants of plasma HIV-1 RNA on the outcome of a first-line ART combination. Baseline minority resistant variants have been associated with increased rates of virological failure with NNRTI-based regimens, but not with regimens containing ritonavir-boosted PIs or integrase inhibitors.^14,15 A clinically relevant cut-off of 1% minority NNRTI-resistant variants has been suggested to have significant impact on the NNRTI-based first-line regimen outcome.¹⁶ To date, all studies using UDS to evaluate DRAMs were cross-sectional and performed on HIV-1 RNA from plasma samples of naive patients or plasma samples collected at the time of virological failure.^14–16

To better understand the evolution of archived resistant viral reservoirs over time in heavily ART-experienced patients with currently suppressed plasma HIV-1 RNA, we undertook a longitudinal UDS analysis on blood-cell-associated HIV-1 DNA samples. We selected and included patients with previous virological failures, who were all exposed to three classes of drugs—NRTIs, NNRTIs and PIs—and who subsequently achieved sustained virological control on a salvage optimized ART (SOT) regimen.

The main objective was to address the dynamics and persistence of resistant viruses harbouring DRAMs archived in cell-associated HIV-1 DNA, using UDS, after lengthy periods of suppression of HIV replication in blood plasma.

Methods

Study population

Patients were selected from two randomized clinical trials with an enfuvirtide-based optimized regimen arm: (i) the INNOVE study,¹⁷ which enrolled ART-experienced patients with virological failure who were harbouring in plasma multiresistant viruses that were still genotypically susceptible to at least two active compounds; and (ii) the ANRS 123 ETOILE study,¹⁸ which enrolled ART-experienced patients with virological failure and a genotypic score showing two or fewer active drugs, and who received an SOT including or not including enfuvirtide. Nantes Hospital Ethics Committee approved the INNOVE study and institutional ethics review boards approved the ANRS 123 ETOILE study, and all patients gave written informed consent.

Our current analysis of HIV-1 DNA genotyping is a substudy focusing on patients for whom frozen blood cells were available at different timepoints and who achieved virological suppression as of month 3 after the initiation of an SOT. A total of 29 whole-blood samples were available for 10 patients. There were six patients from the INNOVE study with PBMCs available at baseline, at month 6 and at a third, distant timepoint (4–6 years) and four patients from the ETOILE study with whole-blood samples frozen at baseline and months 6 and 12.

Total HIV-1 DNA quantification

Total cell-associated DNA was extracted from 200 μL of whole blood or pelleted PBMCs with NucleoSpin Blood kit (Macherey-Nagel, Hoerdt, France) or QIAmp DNA Mini kit (Qiagen, Courtaboeuf, France) respectively, with a final elution volume of 100 μL. Total HIV-1 DNA was quantified by a real-time PCR assay with the Generic HIV DNA Cell assay (Biocentric, Bandol, France) as described previously.¹⁹

Amplicon generation

UDS was performed on the longitudinal DNA extracts for reverse transcriptase (RT), protease (PR) and V3 loop regions.^20,21

For the V3 Env region amplification, a first-round PCR was performed using the following primers (in 5′–3′ orientation): 1V3f, GGCTTAGGCATCTCCTATGGCAGGAAGAAG (5954–5983; this and all following positions were based on HIV-1 HXB2, GenBank accession number K03455) and 1V3r, GGTCTTAAAGGTACCTGAGGTCTGACTGGA (9000–9029).²¹ The first-round PCR was performed as follows: 2 min at 98°C; 10 cycles of 30 s at 98°C, 30 s at 57°C, 4 min at 68°C; 40 cycles of 30 s at 98°C, 30 s at 63°C, 4 min + 5 s/cycle at 68°C; 10 min at 68°C followed by cooling to 4°C. Then, a fragment encompassing the V3 region was generated by nested PCR with the following primers (in 5′–3′ orientation): Env7, AATGGCAGTCTAGCAGAAG (7008–7026) and ED33, TTACAGTAGAAAAATTCCCCTC (7360–7381). The nested PCR was performed under the following conditions: 2 min at 98°C; 40 cycles of 30 s at 98°C, 30 s at 55°C, 90 s at 72°C; 10 min at 72°C followed by cooling to 4°C. First-round and nested PCRs were performed according to the manufacturer recommendations with Phusion High Fidelity PCR Master Mix with HF Buffer (Thermo Scientific, Vilnius, Lithuania).

For the PR/RT pol region amplification, a first-round PCR was performed using the following primers (in 5′–3′ orientation): 5′prB, GAAGCAGGAGCCGATAGACA (1794–1812) and MJ4, CTGTTAGTGCTTTGGTTCCTCT (3399–3420). The first-round PCR was performed as follows: 2 min at 98°C; 10 cycles of 30 s at 98°C, 30 s at 57°C, 2 min at 68°C; 40 cycles of 30 s at 98°C, 30 s at 63°C, 2 min + 5 s/cycle at 68°C; 10 min at 68°C followed by cooling to 4°C. First-round PCR was performed according to the manufacturer recommendations with Phusion High Fidelity PCR Master Mix with HF Buffer (Thermo Scientific, Vilnius, Lithuania). The nested PCR was adapted from the 454/Roche HIV-1 Genotyping Kit v2.^14,20 Our first-round PCR products were subsequently amplified using the microtitre plate of this kit with dried-down primer pairs, to generate four amplicons encompassing PR/RT genes: A (PR), B, C and D (RT).

Each V3 or PR/RT nested PCR primers consisted of a 454 sequencing adaptor (for either forward or reverse sequencing), sample-specific multiplex identifier sequence, and a specific viral sequence for PCR priming.

HIV drug resistance testing by UDS

Each amplicon was purified with Agencourt AMPure XP magnetic beads (Beckman Coulter, Krefeld, Germany). Amplicons were analysed using an Agilent DNA 1000 Bioanalyser (Agilent Technologies, Santa Clara, CA, USA) to verify their quality and purity. The purified amplicons were quantified with a fluorescence assay Quant-it Picogreen (Invitrogen, Eugene, OR, USA) following the manufacturer's instructions. Purified amplicons were pooled in equimolar amounts. Clonal amplification on beads (emulsion PCR), bead isolation (breaking) and sequencing was performed according to the manufacturer's protocol, on a GS Junior (Roche/454, Meylan, France).

DRAM and variant analysis

The sequence reads were analysed with GS amplicon variant analyser (AVA) software (Roche), which assigns each read to the proper amplicon and patient sample using the multiplex identifier. This software also aligns the generated sequence reads with the HIV WT HXB2 sequence and identifies quantitatively drug resistance mutations according to a pre-defined DRAMs list. UDS results were interpreted using the ANRS resistance algorithm V24 (www.hivfrenchresistance.org). The major NRTI, NNRTI and PI resistance mutations were identified using the update 2015 Stanford drug resistance mutation list (http://hivdb.stanford.edu). DRAMs were used as markers to characterize HIV-1 DNA blood reservoir dynamics. Genetic signatures of the four amplicons encompassing PR/RT genes were used to reconstruct the whole variants. Resistant HIV DNA load was calculated at each timepoint based on the percentage of quasispecies harbouring DRAMs by UDS and the total cell-associated HIV-1 DNA load.

The tropism of each variant was inferred from the V3 amino acid sequence by geno2pheno available at http://454.geno2pheno.org/index.php.²¹ V3 loop quasispecies diversity was defined by Shannon entropy.²² This measure was calculated as −[Σp_i × log₂(p_i)]/log₂(N), where p_i is the relative frequency of each variant i (i = 1;…;n where n is number of variants by sample) and N is the total number of sequences analysed by sample. Entropy results can vary from 0 (single variant) to 1 (each sequence represents a single variant).

Intra-patient and inter-sample quasispecies comparisons were assessed using the Wilcoxon signed-rank test and Mann–Whitney test, respectively (http://marne.u707.jussieu.fr/biostatgv/).

Definition of minority resistant variants

The detection limit of drug resistance mutations was fixed at 1%. This conservative cut-off was chosen based on previous reports indicating combined PCR and sequencing error rates of 0.1%–0.3% using 454 amplicon sequencing of HIV-derived clonal sequences.^21,23,24 Variants present in more than 20% of the quasispecies were considered to be major resistant variants, and variants present at a proportion between 1% and 20% were considered to be minor variants.

Results

Patients and UDS

Patients' characteristics at the time of initiation of SOT (baseline) are described in Table 1. At the time of enrolment, patients were in virological failure with a median HIV-1 RNA load of 4.7 log₁₀ copies/mL (range, 3.2–5.2 log₁₀ copies/mL). Virological success was achieved as soon as month 3 with SOT and sustained during follow-up for all patients. The complete characteristics of study patients are described elsewhere.^17,18,25 Cell-associated HIV-1 DNA load decreased significantly between baseline and month 6 on successful SOT (median, 3.31 log₁₀ copies/10⁶ PBMCs at baseline versus 2.95 log₁₀ copies/10⁶ PBMCs at month 6; P = 0.004). Globally, cell-associated HIV-1 DNA remained high, with a median value of 3.0 log₁₀ copies/10⁶ PBMCs (range, 1.9–4.2 log₁₀ copies/10⁶ PBMCs).

Among the 29 samples, UDS results were available for 29 (100%), 28 (97%), 26 (90%), 26 (90%) and 27 (93%) of the amplicons V3, A, B, C and D, respectively. UDS yielded a median of 3553 reads per V3 amplicon (range, 2403–10 196 reads) and a median of 2277 reads per PR/RT amplicon (range, 825–6391 reads).

Longitudinal viral genotropism in blood cells and V3 variants by UDS

Overall, we found 125 variants, including 67% of minority variants (Table 1). At baseline, UDS indicated the presence of R5 quasispecies (R5 variants represented >95% of the quasispecies) in six patients (patients 1, 3, 4, 6, 7 and 9) and R5X4 dual-mixed variant populations in the remaining four (patients 2, 5, 8 and 10). At baseline, minority X4 variants were found in patient 2 (n = 1: 1.1%), patient 5 (n = 3: 12.4%, 10.6%, 1.7%), patient 8 (n = 1: 12.4%) and patient 10 (n = 6; 5.8%, 2.6%, 1.4%, 1.3%, 1.2%, 1.0%).

During follow-up, three out of six patients who harboured a major R5-tropic variant displayed the emergence of dual R5X4-tropic quasispecies at month 6 (patients 1, 4 and 7). None of these patients harboured minority X4 variants at baseline. Patient 4's quasispecies reverted to R5 at month 12, and the X4 variants became dominant (>80% of the quasispecies) in the third sample for the remaining two (patients 1 and 7).

Only one patient (patient 5) with R5X4 dual-mixed quasispecies in PBMCs at baseline was treated with a CCR5 antagonist. For this patient, the proportion of X4 variants increased from 25.8% at baseline to 99.5% after 4 years on a maraviroc-containing SOT (patient 5; Figure 1a).

We observed two distinct evolution patterns for V3 loop sequences on efficient SOT. First, six patients (patients 1, 3, 5, 7, 9 and 10) harboured conserved variants over time (e.g. patients 10 and 3; Figure 1b and c) with small variations in their relative frequencies. Second, the remaining four patients (patients 2, 4, 6 and 8) had quasispecies with no identical sequence between baseline and the third sample, depicting pronounced fluctuations of variants (e.g. patient 6; Figure 1d).

The median Shannon entropy was 0.093 (range, 0.004–0.209) and 0.087 (range, 0.005–0.180) at baseline and on the last visit sample, respectively, showing no significant difference in diversity level over time (P > 0.05). Also, there was no significant difference in diversity level between R5 and dual-mixed R5X4 quasispecies (P > 0.05), with median entropy of 0.059 (range, 0.004–0.209) and 0.087 (range, 0.005–0.265), respectively.

Longitudinal DRAMs in the PR/RT sequence archived in blood cells and variants by UDS

A median of two variants was detected in each sample, with 27% of minority variants. The number of variants per sample was significantly lower in PR/RT sequences than in V3 sequences; this observation was true when considering all variants (P < 0.001) or only minority variants (P = 0.002). Archived DRAMs were detected in all 29 samples, conferring resistance to the three major drug classes (NRTI, NNRTI and PI). PI DRAMs are shown in Table 2, with NRTI and NNRTI DRAMs in Table 3, and the dynamics between major DRAMs and the three drug classes are depicted in Figure 2. In all patients' quasispecies and at each timepoint, DRAMs to the three drug classes were linked on the same viral variant.

Concerning the dynamics of DRAMs archived in PBMCs, three distinct evolutionary patterns were observed. Resistant viral DNA load varied according to the size of the total cell-associated HIV-1 DNA level (Figure 2b). Most emerging and persisting DRAMs were detected with a high resistant viral DNA load (>2 log₁₀ copies/10⁶ PBMCs; patients 1, 2, 3, 4, 5, 7 and 10). Conversely, variants no longer detected at the last visit were mostly present at a low resistant load at baseline (<2 log₁₀ copies/10⁶ PBMCs; patients 6, 8, 9 and 10).

• First, some baseline DRAMs were still detectable during follow-up and were all linked on the same variant (patients 1, 2, 4 and 5). These DRAMs were generally related to persistent selective pressure by one or more drugs of the SOT. Indeed, patients 1, 2 and 4 had baseline viral DNA quasispecies considered resistant to at least one drug of the SOT. For example, patient 2 harboured at baseline a unique viral variant with PI, NRTI, and NNRTI DRAMs conferring resistance to lopinavir and intermediate resistance to tenofovir. Both lopinavir and tenofovir were part of the SOT, and the viral variant was still major after 1 year of successful treatment. Although no NNRTI was used as part of SOT in this patient, NNRTI DRAMs (i.e. Y181C) were still detected throughout follow-up as they were linked on the same variant harbouring lopinavir and tenofovir DRAMs. However, this patient had an important HIV-1 DNA load decrease over time (−0.88 log₁₀ copies/10⁶ PBMCs within a year on SOT), which resulted also in a reduced resistant viral DNA load (3.34 log₁₀ copies/10⁶ PBMCs at baseline versus 2.36 log₁₀ copies/10⁶ PBMCs at month 12).

• Second, some baseline variants with linked DRAMs were no longer detectable through follow-up on efficient SOT (patients 3, 6, 8, 9 and 10). In these cases, DRAMs were not related to drugs in the SOT. In patients 3, 6, 8 and 9, no variant harboured DRAMs to any of the drugs in the SOT, according to baseline UDS viral DNA genotypes. For example, patient 8 had a major variant (approximately 60% of the archived quasispecies) harbouring DRAMs to the three drug classes (i.e. NRTI: D67N, K70R, K219E; NNRTI: K103N, Y181C; PI: L24I, M46I, I54V, V82A). This variant was no longer detectable as of month 6 of follow-up. These DRAMs were not related to the SOT. The other variants found in this patient harboured no DRAM. Also, resistant viral DNA load at baseline was significantly higher (P = 0.05) for patients harbouring quasispecies in which viral DNA DRAMs were still detectable over the follow-up period (patients 1, 2, 4 and 5; median, 3.21 log₁₀ copies/10⁶ PBMCs; range, 2.39–3.52 log₁₀ copies/10⁶ PBMCs) than for those with no detectable viral DNA DRAMs over time (patients 6, 8, 9 and 10; median, 1.32 log₁₀ copies/10⁶ PBMCs; range, 0.17–2.98 log₁₀ copies/10⁶ PBMCs).

• Third, variants with new DRAMs (not detected at baseline) were detected throughout follow-up, despite sustained suppression of HIV-1 RNA in plasma (patients 3, 5, 7 and 10). Drugs in the SOT may have selected for these variants harbouring these additional DRAMs. In patients 3, 5, 7 and 10, the new resistance patterns were present on major variants of the archived quasispecies. For example, patient 7 harboured a major L100I and L76V viral variant after 6 years on an etravirine- and darunavir-containing SOT.

Discussion

The presence of HIV drug-resistant variants poses a significant challenge to effective therapeutic intervention at any stage of infection, including that in patients under suppressive ART. Additionally, the role of past DRAMs evidenced in plasma in heavily ART-experienced patients must be taken into account for any treatment modification. Using UDS techniques on longitudinal PBMC samples of patients who were enrolled in the INNOVE and ANRS 123 ETOILE studies, who harboured multiresistant quasispecies and who achieved virological control on SOT, we showed that archived DRAMs can still be detectable in the cellular reservoir for lengthy periods of time, i.e. up to 6 years. Compared with Sanger sequencing, UDS provided a better sensitivity (1%), a relative abundance of quasispecies harbouring DRAMs and an analysis of linked mutations on the same viral variant. So, focusing on viral variants, we showed that the persistence of DRAMs depends on their association with the same viral variant. Moreover, new resistant variants with additional resistance mutations in the cellular reservoir that were not detected at baseline can be detected during follow-up despite fully and prolongedly effective ART. Combined tropism and DRAMs data confirmed that HIV-1 DNA archived quasispecies in PBMCs continue to evolve despite sustained virological control.

V3 loop sequences showed wide intra-individual variability over time, with multiple scenarios for the evolution of archived viral quasispecies. Three patients who harboured a major R5-tropic variant displayed the emergence of dual R5X4-tropic quasispecies. This scenario may be due to a heavy selective pressure from immune system recovery on effective ART²⁶ and also to the SOT regimen composition.²⁷ Indeed, in one patient with dual-mixed R5X4 DNA quasispecies at baseline, X4 variants were selected after 4 years on a maraviroc-containing regimen, despite virological control.

We used resistance mutations in PR and RT genes to assess the dynamics of viruses archived in blood cells. Despite the usual dynamic processes affecting the pool of infected cells, some HIV-1 DNA samples isolated from PBMCs far from SOT initiation and under long-term virological control still exhibited the same DRAMs as at the time of virological failure. This finding indicates that resistant variants that entered the cellular reservoir during past virological failures persisted and were not replaced.²⁸ In this case, persistence of resistant variants was associated with ongoing selective pressure due to the composition of SOT on at least some linked mutations. We hypothesized that the mutated viral variant was able to replicate at low levels and so remained archived over time, while the remaining active drugs were able to achieve plasma virological control. Since DRAMs were mostly associated with the same viral variant, the complete resistance profile was maintained. The long-time persistence of these archived variants can be explained by the characteristics of target cells, particularly the central memory CD4 T cells with high proliferative capacity and long half-life.^8,9 Conversely, some viral variants harbouring major DRAMs were no longer detected in blood cells over time when the SOT regimen consisted of drugs from new classes or with a distinct resistance profile. Such a loss of detection of archived resistant variants may be due to the detection limit of very rare variants or to a dilution effect over time: with the death of cells harbouring those variants, the viral burden of these variants will probably be lower than that in cases with persistency. Immunological pressure in the absence of the selective pressure of the corresponding drug might also play a role in this (apparent) loss of viral variants. Whether such a phenomenon could allow reuse of drugs with resistance history remains very elusive.

Interestingly, variants with new DRAMs associated with SOT appeared during follow-up in patients who had achieved sustained virological control as early as 12 weeks after SOT initiation. This finding suggests that new viral quasispecies may be archived in blood cells despite effective ART. These new variants have most probably been selected directly in the blood compartment: indeed, routine assays for HIV-1 RNA measurement have a quantification limit of 20–50 copies/mL, but almost half of the patients under suppressive ART have a residual viraemia between 1 and 50 copies/mL that can persist for a long time.^29,30 This residual replication, in combination with drug pressure, might have selected for viruses with new DRAMs.³¹ Unfortunately, we did not have enough available plasma to perform an ultrasensitive HIV-1 RNA measurement technique in the present study. Besides, viral variants harbouring these new DRAMs might pre-exist at very low levels or might have been selected in other reservoirs such as lymph nodes, the gut or the genital tract,^4–6 because of insufficient drug penetration in these compartments.³² Some of these new DRAMs were G-to-A mutations (i.e. E138K and M230I in RT, patient 10), implicating APOBEC3 editing in the emergence of archived DRAMs in blood cells.³³ Although no stop codons were detected, indicating viability of these viral variants, the new DRAMs did not contribute to subsequent virological failure. The clinical relevance of appearing DNA DRAMs on successful ART has to be investigated.

Moreover, resistant viral DNA load varied also according to the size of the HIV reservoir. Given that most patients experienced an important decrease in cell-associated HIV-1 DNA level over time, the resistant viral DNA load decreased as well. Besides, the resistant viral DNA load might play a role in the blood viral quasispecies dynamics: high levels led to variants' persistence whereas low levels led to variants' lack of detection after lengthy periods of controlled viraemia. These resistant HIV-1 DNA loads could influence the efficacy of new ART with cross-resistance or reused ART.

In conclusion, this study is the first to use the UDS technique to assess the longitudinal dynamics of viral resistant quasispecies archived in blood cells of heavily pretreated patients harbouring multiresistant virus and achieving sustained virological control for a prolonged period of time. Using DRAMs and tropism as markers, we demonstrated that, despite virological control, the diversity of the quasispecies continued to evolve. We showed that archived virus with DRAMs can persist for lengthy periods of time, which should be taken into account when modifying ART, even after many years of virological suppression.

Funding

The ANRS 123 ETOILE study was supported by the ANRS (Agence Nationale de Recherche sur le SIDA et les Hépatites Virales, Paris, France), and the INNOVE study was supported by Roche Laboratories, France. For the present substudy, we received a grant from Roche Laboratories, France. The funders had no role in the study design, data collection and analysis, decision to publish or preparation of the manuscript.

Transparency declarations

None to declare.

References

Author notes