A detailed characterization of drug resistance during darunavir/ritonavir monotherapy highlights a high barrier to the emergence of resistance mutations in protease but identifies alternative pathways of resistance

Abstract Background Maintenance monotherapy with ritonavir-boosted darunavir has yielded variable outcomes and is not recommended. Trial samples offer valuable opportunities for detailed studies. We analysed samples from a 48 week trial in Cameroon to obtain a detailed characterization of drug resistance. Methods Following failure of NNRTI-based therapy and virological suppression on PI-based therapy, participants were randomized to ritonavir-boosted darunavir (n = 81) or tenofovir disoproxil fumarate/lamivudine +ritonavir-boosted lopinavir (n = 39). At study entry, PBMC-derived HIV-1 DNA underwent bulk Protease and Reverse Transcriptase (RT) sequencing. At virological rebound (confirmed or last available HIV-1 RNA ≥ 60 copies/mL), plasma HIV-1 RNA underwent ultradeep Protease and RT sequencing and bulk Gag-Protease sequencing. The site-directed mutant T375A (p2/p7) was characterized phenotypically using a single-cycle assay. Results NRTI and NNRTI resistance-associated mutations (RAMs) were detected in 52/90 (57.8%) and 53/90 (58.9%) HIV-1 DNA samples, respectively. Prevalence in rebound HIV-1 RNA (ritonavir-boosted darunavir, n = 21; ritonavir-boosted lopinavir, n = 2) was 9/23 (39.1%) and 10/23 (43.5%), respectively, with most RAMs detected at frequencies ≥15%. The resistance patterns of paired HIV-1 DNA and RNA sequences were partially consistent. No darunavir RAMs were found. Among eight participants experiencing virological rebound on ritonavir-boosted darunavir (n = 12 samples), all had Gag mutations associated with PI exposure, including T375N, T375A (p2/p7), K436R (p7/p1) and substitutions in p17, p24, p2 and p6. T375A conferred 10-fold darunavir resistance and increased replication capacity. Conclusions The study highlights the high resistance barrier of ritonavir-boosted darunavir while identifying alternative pathways of resistance through Gag substitutions. During virological suppression, resistance patterns in HIV-1 DNA reflect treatment history, but due to technical and biological considerations, cautious interpretation is warranted.


Introduction
HIV drug-resistant variants acquired at the time of infection or enriched under selective drug pressure can integrate into the DNA of memory CD4 T-cells as provirus and become part of the HIV DNA archive within the host cell. 1,2Archived resistant variants can re-emerge if virus production resumes, thus potentially retaining life-long clinical significance.While the available evidence is not completely consistent, studies have demonstrated that detecting drug resistance-associated mutations (RAMs) in the HIV-1 DNA of virologically suppressed patients predicts virological outcomes when switching to a different treatment regimen. 2,3Interestingly, some studies have produced surprising findings.We previously reported that detecting archived RAMs during suppressive ART with two NRTIs plus a ritonavir-boosted PI was associated with a reduced likelihood of virological rebound after switching to ritonavir-boosted darunavir monotherapy. 4aintenance monotherapy with ritonavir-boosted darunavir has been studied in Western Europe among virologically suppressed patients without previous treatment failure. 5A multicentre study conducted in Burkina Faso, Cameroon and Senegal investigated ritonavir-boosted darunavir monotherapy among 50 patients who had achieved virological suppression on ritonavir-boosted PI-based triple ART after failure of two NRTIs plus an NNRTI. 6These studies uniformly reported an increased risk of viraemia when switching to ritonavir-boosted darunavir monotherapy, and also consistently demonstrated a low risk of treatment-emergent darunavir RAMs.
We conducted a randomized clinical trial in Cameroon comparing ritonavir-boosted darunavir maintenance monotherapy with standard of care (SOC) triple ART with two NRTIs plus ritonavirboosted lopinavir in adults living with HIV.Similar to the study from Burkina Faso, Cameroon and Senegal, 6 participants were virologically suppressed on ritonavir-boosted PI-based triple ART having experienced virological failure of first-line NNRTI-based therapy.We previously reported the virological outcomes of the ritonavir-boosted darunavir arm in relation to the detection of RAMs in cellular HIV-1 DNA at study entry. 4The aim of this analysis was to take advantage of unique trial samples and conduct a detailed characterization of the virological outcomes of the entire trial population.We performed a sensitive assessment of treatmentemergent resistance using ultradeep sequencing (UDS), including the evaluation of Gag mutations, and compared the resistance patterns detected in cellular HIV-1 DNA at study entry, when participants showed virological suppression, with those observed in plasma HIV-1 RNA during subsequent virological rebound.

Study population
Monotherapy in Africa, New Evaluations of Treatment (MANET) was a randomized, open-label trial based at Hôpital Central Yaoundé in Cameroon (NCT02155101).Between August 2014 and July 2015, 120 adults receiving virologically suppressive ART with two NRTIs plus a ritonavir-boosted PI (typically ritonavir-boosted lopinavir) were randomized to either ritonavir-boosted darunavir monotherapy (800/100 mg once daily) for 48 weeks (n = 81) or SOC with tenofovir disoproxil fumarate coformulated with lamivudine plus ritonavir-boosted lopinavir for 24 weeks (n = 39).Eligibility criteria comprised having received two NRTIs plus a ritonavir-boosted PI for ≥12 weeks, CD4 count > 100 cells/mm 3 , plasma HIV-1 RNA < 60 copies/mL in two screening measurements taken 4-12 weeks apart, negative hepatitis B surface antigen (HBsAg) and absence of significant disease or laboratory abnormalities.Pregnancy or planning to become pregnant were exclusion criteria.Following randomization, participants of both arms attended scheduled study visits at Weeks 4, 12 and 24; the ritonavir-boosted darunavir arm also attended scheduled study visits at Weeks 36 and 48.The study was approved by the University of Liverpool Ethics Committee (RETH000605) and the Cameroon National Ethics Committee (2013/07/347) and overseen by an independent trial safety board.

Statistical analysis
The characteristics of the population at study entry, summarized as categorical and continuous variables, were compared by chi-squared test, Fisher's exact test or Wilcoxon Mann-Whitney test, as appropriate.The trial primary endpoint was the proportion of participants with plasma HIV-1 RNA < 400 copies/mL at Week 24 (FDA snapshot).Secondary virological endpoints measured through to Week 24 in both arms and through to Week 48 in the ritonavir-boosted darunavir arm comprised proportions with plasma HIV-1 RNA < 60 copies/mL and emergence of RAMs in participants with virological rebound, defined as confirmed (or last available) HIV-1 RNA ≥ 60 copies/mL.Virological failure was defined as confirmed (or last available) HIV-1 RNA ≥ 400 copies/mL.

Laboratory tests
At the Centre Pasteur of Cameroon in Yaoundé, safety parameters and CD4 cell counts were measured using freshly collected samples.Plasma was separated from whole venous blood in EDTA within 2 h of collection and stored at −80°C.HIV-1 RNA was quantified with the Biocentric assay (Bandol, France; lower limit of quantification 60 copies/mL).At the Chantal Biya International Reference Centre for Research on HIV/AIDS Prevention & Management in Yaoundé, PBMCs were isolated by Ficoll-Hypaque gradient centrifugation and stored at −80°C.Sanger sequencing of Protease (amino acids, aa 1-99) and Reverse Transcriptase (RT, aa 1-335) was performed as described. 7Sample aliquots were shipped frozen to the UK for HIV-1 RNA sequencing (see below) and for the quantification of total HIV-1 DNA in PBMCs by real-time PCR as described. 8

HIV-1 RNA sequencing
Plasma samples from participants experiencing virological rebound underwent UDS as previously described. 9,10Briefly, samples with HIV-1 RNA < 10 000 copies/mL were enriched by ultra-centrifugation at 35 000 rpm for 20 min at 4°C; following extraction with the QIAamp viral RNA kit (QIAGEN, UK), a 1300 bp amplicon was generated covering Protease (aa 1-99) and RT (aa 1-335), purified with Agencourt AMPure XP magnetic beads (Beckman Coulter, UK) and quantified by the Qubit dsDNA High Sensitivity Assay Kit on the Qubit 3.0 fluorometer (Invitrogen, UK).A DNA library was prepared with the Nextera XT DNA Sample Prep Kit (Illumina, USA), followed by sequencing with the Illumina MiSeq Reagent Kit v2.After checking for quality, reads were analysed applying a frequency threshold of 1% and described as low frequency variants (representing 1%-14% of the variants in a sample) and high-frequency variants (≥15%).Using plasma samples from participants experiencing virological rebound on ritonavir-boosted darunavir, after extraction with the QIAamp Viral RNA kit (QIAGEN), viral RNA was reverse transcribed using the Superscript III One-Step RT PCR Kit with Platinum ® Taq High Fidelity enzyme followed by amplification using Platinum ® PCR SuperMix High Fidelity (Invitrogen).A 2200 bp amplicon spanning gag and protease was generated by nested PCR as detailed in Table S1 (available as Supplementary data at JAC Online), followed by Sanger sequencing.
Abdullahi et al.

Phenotypic characterization of the gag cleavage site mutation T375A
T375A was inserted by site-directed mutagenesis into the WT vector P8.9NSX, which contains the protease and RT sequences of the NL4-3 strain of HIV-1, 19 using the QuikChange Multi Site-Directed Mutagenesis Kit (Stratagene, UK).The amplified DNA was enriched by digestion of the parental DNA with DpnI and XL1-blue and supercompetent cells were transformed with the digested DNA.The plasmidic DNA was isolated using QIAprep Spin Miniprep Kit (QIAGEN) and screened for the presence of T375A by Sanger sequencing.Susceptibility to darunavir, atazanavir, lopinavir, indinavir and saquinavir, and replication capacity were determined by a single-cycle assay in HEK293T cells as described. 19Virus replication in the presence of drug was determined by luciferase quantification 48 h postinfection relative to no-drug controls.The mean 50% inhibitory concentration (IC 50 ) from three separate experiments was calculated and results expressed as fold-changes (FC) in IC 50 compared with WT P8.9NSX.Viral replicative capacity was determined by luciferase quantification in the absence of drug, calculating the mean luciferase activity from ≥4 values within the linear range, and expressed as percentage relative to WT P8.9NSX.

Study population
At study entry, the 120 participants had received ART for a median of 7.5 years, including prior first-line NNRTI-based ART for a median of 3.0 years and current second-line ritonavir-boosted PI-based ART for a median of 3.1 years (Table 1).Most (106/120; 88.3%) were receiving once-daily tenofovir disoproxil fumarate/lamivudine plus twice-daily ritonavir-boosted lopinavir.The characteristics of the two arms were similar overall, although the ritonavir-boosted darunavir arm showed higher levels of total HIV-1 DNA and higher prevalence of RAMs in HIV-1 DNA (Table 1).As per the eligibility criteria, prior to randomization all participants had shown a suppressed viral load (< 60 copies/mL) in two screening measurements, which were separated by a median of 7 weeks (range 6-12).No prior viral load measurements were available in the medical records.

Resistance patterns in plasma HIV-1 RNA at virological rebound
Sequences were obtained by UDS in 23 participants who experienced virological rebound (ritonavir-boosted darunavir arm n = 21; SOC arm n = 2).HIV-1 RNA levels at the time of testing were median 3.0 log 10 copies/mL (range 2.0-4.1).NRTI and NNRTI RAMs were found in 9/23 (39.1%) and 10/23 (43.5%) samples, respectively and 8/23 (34.7%) samples had RAMs for both classes; most RAMs occurred at frequency ≥15% (Table 2).No participant had darunavir RAMs at either high or low frequency.Two participants, both in the ritonavir-boosted darunavir arm, showed the nelfinavir RAM D30N at low frequency (3%-4%); before starting ritonavir-boosted darunavir, they had received ritonavir-boosted lopinavir for >4 years and had no prior nelfinavir exposure.No stop codons or hypermutation were found in plasma sequences.

Comparison of resistance patterns in HIV-1 DNA and HIV-1 RNA
There were 18 participants with paired cellular HIV-1 DNA and plasma HIV-1 RNA sequences (Table 3).Phylogenetic analyses confirmed clustering of the pol sequences (Figure 1).With 6/18 (33.3%) sample pairs, the resistance patterns were fully consistent, comprising 5 with no RAMs in either sample and 1 with the NNRTI HIV-1 drug resistance in cellular HIV DNA and plasma HIV RNA RAM K103N in both samples (Table 3).A further 6/18 (33.3%) sample pairs showed partial consistency, comprising 5 with fewer RAMs in HIV-1 DNA and 1 with fewer RAMs in HIV-1 RNA.With 5/18 (27.8%) sample pairs, RAMs were detected only in HIV-1 DNA; 3 HIV-1 DNA sequences showed in-frame stop codons and 1 with the NNRTI RAM M230I also showed hypermutation.One sample pair showed M230I in HIV-1 RNA only (frequency 1%).There was no consistency in the detection of the Protease RAM D30N when comparing sample pairs, and no linkage was detected between pol sequences from the five individuals with D30N (Figure S1).

Gag mutations
Among participants who experienced virological rebound on ritonavir-boosted darunavir, eight underwent gag sequencing, including four who were tested longitudinally (Table 4).None of the sequences contained RAMs in Protease.Screening for Gag mutations significantly associated with PI exposure revealed four sequences with five cleavage site mutations.One participant with longitudinal samples showed emergence of K436R in p7/p1 between Week 24 and Week 36, when K436R replaced T375N in Abdullahi et al.
p2/p7.Two sequences showed T375A in p2/p7, in one case occurring with K436R.The phenotypic effects of T375A were analysed by site-directed mutagenesis.The mutation reduced PI susceptibility by approximately 5-FC for atazanavir, indinavir and lopinavir and 10-FC for darunavir and saquinavir (Figure 2).The mutation also increased replicative capacity to 161% (±5.0) relative to WT control.Furthermore, all sequences showed mutations within other Gag regions.Among participants with longitudinal results,    Abdullahi et al.

Discussion
We studied a population in Cameroon that following failure of first-line NNRTI-based ART started two NRTIs plus a ritonavirboosted PI (typically ritonavir-boosted lopinavir) in the absence of virological monitoring.After confirmation of virological suppression, participants were assigned to either maintenance monotherapy with ritonavir-boosted darunavir or tenofovir disoproxil fumarate/lamivudine plus ritonavir-boosted lopinavir.
Although interpretation is limited by the small number of participants, the findings align with published data from a similar population, 6 indicating a heightened risk of viraemia among individuals on ritonavir-boosted darunavir monotherapy.Over 48 weeks, we observed several instances of virological rebound (confirmed or last available HIV-1 RNA ≥ 60 copies/mL) and virological failure (HIV-1 RNA ≥ 400 copies/mL) in this group, with nearly 30% of participants experiencing rebound viraemia.Of note, the study adopted an HIV-1 RNA threshold of 400 copies/ mL to define virological failure in line with similarly designed clinical trials conducted in the region. 6At virological rebound, the resistance patterns in plasma HIV-1 RNA were partially reflective of those detected in cellular HIV-1 DNA at study entry.UDS confirmed the absence of treatment-emergent darunavir RAMs in Protease.However, we detected mutations in Gag and demonstrated an effect on darunavir susceptibility in the absence of RAMs in Protease.
In cell culture, two pathways of darunavir resistance have been characterized anchored around the Protease RAMs I50V or I84V. 20Darunavir RAMs have also been shown to emerge in HIV-1 drug resistance in cellular HIV DNA and plasma HIV RNA PI-experienced people with pre-existing Protease RAMs but have rarely occurred when starting ritonavir-boosted darunavir de novo; the high resistance barrier is thought to reflect tight binding of darunavir to the Protease enzyme and high plasma concentrations achieved through pharmacological boosting. 20Monotherapy studies have also reported a negligible risk of darunavir RAMs. 5 Most studies employed Sanger sequencing for detecting resistance, but two studies applied more sensitive techniques.One study using single-genome sequencing in five participants with HIV-1 RNA > 400 copies/mL detected darunavir RAMs (V32I, I47V, I50V) in one participant lacking the mutations by Sanger sequencing. 21A second study using UDS in 14 participants with HIV-1 RNA > 1000 copies/mL found I54T in the Protease of a participant lacking Protease RAMs by Sanger sequencing; 22 however, I54T is not considered a darunavir RAM. 11We complement these data by reporting UDS results from 21 participants and by extending the analysis to all cases of confirmed or last available plasma HIV-1 RNA ≥ 60 copies/mL, including 12 with viral load < 1000 copies/mL.No darunavir RAMs were detected in the 21 participants.
Two participants on ritonavir-boosted darunavir showed the Protease RAM D30N in rebound HIV-1 RNA but not in HIV-1 DNA at study entry.The significance is doubtful.D30N is a non-polymorphic substrate-cleft mutation that is selected by and causes high-level resistance to nelfinavir. 23The aspartate-to-asparagine substitution alters the hydrogen bond interaction with the aniline NH 2 group of darunavir, causing some loss of binding affinity. 24However, the mutation occurred only at low frequency (<5%) in rebound HIV-1 RNA.Furthermore, there is no evidence of an association with darunavir resistance. 11At study entry, D30N was also detected in the HIV-1 DNA of three ritonavir-boosted lopinavir-experienced participants, but there is similarly no evidence that ritonavir-boosted lopinavir selects for D30N. 23PCR-induced error or APOBEC3G-mediated hypermutation, an innate defence mechanism that aims to impair virus functionality, may also explain the unexpected detection of D30N. 25 Substitutions that have been related to hypermutation include D30N and M46I in Protease and E138K, M184I, G190E and M230I in RT. [25][26][27] We found hypermutation in several HIV-1 DNA sequences showing these mutations.
9][30][31][32][33][34] Mutations associated with PI exposure typically involve Gag cleavage sites, although mutations in other regions have also been implicated. 18,33,34We had sufficient sample for full-length gag and protease sequencing in a subset of participants experiencing virological rebound on ritonavir-boosted darunavir.All had Gag mutations previously associated with PI exposure, including mutations at the cleavage sites p2/p7 and p7/p1 and within p17, p24, p2 and p6.In one case, we were able to demonstrate emergence of K436R in p7/p1 while HIV-1 RNA levels ranged between 13 381 and 30 762 copies/mL on ritonavir-boosted darunavir.K436R was associated with PI exposure and resistance in previous studies and was shown to occur during ritonavir-boosted darunavir monotherapy. 28,33We also detected T375A in p2/p7 in two participants.T375A was previously detected in an individual receiving ritonavir-boosted darunavir monotherapy in combination with S451N in p1/p6 and several additional Gag mutations. 33Given Abdullahi et al.
the limited data available, we produced a site-directed mutant with T375A and observed a 10-FC reduction in darunavir susceptibility, alongside increased viral replication capacity.Outside of cleavage sites, we saw emergence of D121G in p17, T190I in p24 and P478Q in p6.T190I is an HLA-B*81-associated mutation with a compensatory role in viral fitness. 35Regrettably, we were unable to expand the gag analyses due limited sample volumes.Of note, we focused the analysis of Gag sequences on mutations previously found to be associated with PI exposure, using strict criteria to define the association.This approach may not account for all possible Gag mutations that may contribute to darunavir resistance.In fact, significant mutations may not necessarily be confined to consistent changes at a few sites. 36Importantly, all participants with Gag mutations regained virological suppression after receiving adherence support and returning to ritonavir-boosted PI-based triple ART.
There is growing interest in the potential clinical utility of sequencing cellular HIV-1 DNA to guide treatment decisions during virological suppression. 2As we previously reported on a smaller subset, 4 the resistance patterns detected in HIV-1 DNA at study entry were consistent with the previous failure of NNRTI-based ART.The resistance patterns detected in rebound HIV-1 RNA on ritonavirboosted darunavir were generally in agreement with those found at study entry, but discrepancies were common.This is not unexpected.Several HIV-1 DNA sequences containing RAMs were defective and could not be expected to sustain virus production. 37A further consideration is that in the setting of virological suppression the HIV-1 DNA input into the sequencing test is small and detection of RAMs can become stochastic.Due to limited infrastructure, we did not perform UDS on PBMCs, which may have increased the detection of RAMs in HIV-1 DNA, 38 although even UDS cannot be expected to overcome issues of input size and provide a full representation of the resistance archive. 2 Furthermore, the clinical significance of detecting low-frequency RAMs, particularly in the context of regimens with a high barrier to resistance, remains doubtful.
In summary, the resistance patterns of cellular HIV-1 DNA sequenced during virological suppression, whilst largely reflective of treatment history, are only partially consistent with the resistance patterns of rebound viraemia.Both biological and technical factors may account for the discrepancies and should be taken into consideration when applying the test clinically.We confirm the high resistance barrier of ritonavir-boosted darunavir but demonstrate that Gag substitutions, including T375A in in p2/ p7, provide an alternative pathway of darunavir resistance.HIV-1 drug resistance in cellular HIV DNA and plasma HIV RNA

cTable 3 .
Comprising T215S, T215C, T215I and T215V.d D30N was the only RAM detected in Protease.HIV-1 drug resistance in cellular HIV DNA and plasma HIV RNA Paired cellular HIV-1 DNA sequences obtained at study entry and plasma HIV-1 RNA sequences obtained at virological rebound (n = 18 participants) disoproxil fumarate; TA, thymidine analogues (zidovudine or stavudine).a HIV-1 DNA sequences were obtained from PBMCs by Sanger sequencing; HIV-1 RNA sequences were obtained from plasma by UDS (Illumina MiSeq).b Total exposure at study entry.c The frequency of the mutant is indicated in parentheses.d HIV-1 DNA sequences with in-frame stop codons.e HIV-1 DNA sequences showing hypermutation.f Last available HIV-1 RNA measurement.g First HIV-1 RNA measurement of confirmed virological failure.h Second HIV-1 RNA measurement of confirmed virological failure.i First HIV-1 RNA measurement of confirmed virological rebound.j Sample also underwent gag sequencing.

Figure 1 .
Figure 1.Maximum-likelihood phylogenetic tree of paired pol sequences of HIV-1 DNA obtained from PBMCs at study entry (blue) and HIV-1 RNA obtained from plasma at virological rebound (red) (n = 18 pairs).Control sequences were obtained from Genbank (duplicate sequences excluded).The tree was inferred using 1000 bootstrap replicates.This figure appears in colour in the online version of JAC and in black and white in the print version of JAC.

Table 4 .a
Gag mutations in plasma HIV-1 RNA of participants who experienced virological rebound during darunavir/ritonavir monotherapy a ID b Gag mutations (relative to HIV-1 HXB2) significantly associated with PI exposure are shown, as pre-defined by comparing full-length Gag-Protease sequences from 200 PI-naive and 191 PI-experienced individuals; Gag cleavage sites are in italics; there were no RAMs in Protease.b In 4 participants, two longitudinal samples were tested.c Total exposure at study entry.

Table 1 .
Characteristics of the population at study entry (n = 120)

Table 2 .
Major RAMs in cellular HIV-1 DNA at study entry (n = 90) and in plasma HIV-1 RNA at virological rebound (n = 23) a