Transmitted HIV-1 Drug Resistance in a Large International Cohort Using Next-Generation Sequencing

Results From the Strategic Timing of Antiretroviral Treatment (START) Study

JD Baxter; D Dunn; A Tostevin; RL Marvig; M Bennedbæk; A Cozzi-Lepri; S Sharma; MJ Kozal; M Gompels; AN Pinto; J Lundgren


HIV Medicine. 2021;22(5):360-371. 

In This Article


The START trial represents one of the largest global cohorts with NGS characterization of TDR. The prevalence of TDR overall and by drug class was highest using the 2% detection threshold as a significant number of mutations identified were minor variants. At the 2% threshold, the highest TDR prevalence was observed for PI DRMs (11.4%) while at the 20% threshold the highest TDR prevalence occurred with NNRTI DRMs (4.9%). Within drug classes, a greater proportion of participants harboured PI and INSTI minor variants compared with the NRTI- and NNRTI-associated DRMs. As such, Sanger sequencing, which is comparable to NGS at the 20% threshold of detection, would be expected to underestimate TDR prevalence markedly, particularly for the PI and INSTI drug classes.

As described in this study, a higher proportion of PI minor variants identified when using NGS for detection of TDR was also reported in a UK surveillance study in recently infected MSMs.[11] Cunningham et al.[11] found that the majority of low-frequency variants (62%) detected were PI DRMs, despite these mutations rarely being observed by traditional Sanger sequencing in treatment-experienced patients failing therapy in the UK.

Unlike most other studies of global TDR, our data also included NGS for surveillance INSTI DRMs. The INSTI DRMs identified were almost exclusively minor variants which occurred mostly between the 2% and 5% thresholds. The most commonly detected surveillance INSTI DRMs were G143CHR, associated with high-level resistance to raltegravir, but these were only detected as minor variants. The only INSTI DRMs detected above the 20% threshold were G140S and G140A, which are associated with intermediate resistance to raltegravir and elvitegravir.[21,22] Given the time period in which these samples were collected, there would have been limited prior population exposure to integrase inhibitors, particularly in regions such as Africa and Asia, therefore the majority of these mutations probably represent naturally occurring low-level variants rather than actual DRMs selected under drug pressure and then transmitted. However, their significance remains uncertain and it is unknown if these low-level INSTI variants could potentially compromise efficacy of integrase-based regimens. INSTI DRMs have been reported to occur as natural polymorphisms and can be detected in samples from individuals obtained prior to the advent of the integrase inhibitors.[23] Further research in understanding their potential clinical significance is warranted given the widespread use of integrase-based regimens for initial therapy. In a recent analysis of data from INTEGRATE, a retrospective study from nine European HIV cohorts, baseline INSTI DRMs were found in one of 512 (0.2%) antiretroviral-naïve individuals by Sanger sequencing, suggesting that INSTI TDR is uncommon and unlikely to be detected with this methodology.[24]

Variability of TDR by geographic region was observed in the START study population. By drug class, the highest prevalence of NNRTI DRMs was seen in the USA while the highest prevalence of NRTI DRMs occurred in Australia. Previous studies have shown that the prevalence of transmitted NNRTI resistance has historically been higher in North America than in Europe.[3,4,6] The number of samples from participants included in this analysis was more limited from Australia, but the relatively high prevalence of NRTI TDR has previously been observed with Sanger sequencing.[4,13] This is probably due to extensive prior exposure of viruses to NRTIs in the pool of individuals transmitting HIV-1 in this region.

Geographic variability was also seen in the level of DRM variants detected at different thresholds, particularly for NNRTI TDR. Overall NNRTI TDR was lowest in Africa, occurring in 6.6% of study participants; however, unlike the USA and South America, the majority of NNRTI DRMs occurred as minor variants (mostly in 5–20% of the quasispecies). Similarly, the prevalence of NNRTI TDR in Asia was 7.5%, mostly in 2–5% of the quasispecies. This would suggest that most NNRTI DRMs associated with TDR in these regions may be occurring as minor variants and would not be detected by traditional Sanger sequencing.

The clinical relevance of minor HIV-1 drug-resistant variants remains controversial. Transmission of minor variants associated with resistance to different drug classes has been documented in acutely infected individuals, although these are probably rare events.[25] A recent review of 103 studies of low-abundance drug-resistant variants found that it is difficult to evaluate the clinical impact of minor variants on first-line ART regimens given the heterogeneity of study designs and different laboratory methods used.[26] However, multiple studies assessing the impact of NNRTI minor variants have shown that the presence of these low-level DRMs prior to initiation of treatment may reduce the virological response and increase the likelihood of failure with first-line NNRTI-based regimens.[3,27–30]

More recent global TDR surveillance data using Sanger sequencing in treatment-naïve individuals have demonstrated significantly increasing prevalence rates of NNRTI TDR in Africa and Asia, reaching levels of > 10% in some countries.[9] WHO guidelines have suggested that if the prevalence of pre-treatment drug resistance (which includes treatment-naïve people and those with previous ART exposure starting first-line therapy) exceeds 10% in a country, then use of non-NNRTI-based first-line regimens should be considered.[31] Because of this, a number of countries in Africa and Asia have revised national treatment guidelines to INSTI-based regimens as preferred initial therapy.[9] Furthermore, if non-NNRTI alternatives are unaffordable then pre-treatment drug resistance testing is recommended.[32] Assuming low-level NNRTI variants are clinically important, our data would suggest that the prevalence of NNRTI TDR is being underestimated in many regions of the world where previous surveillance efforts have been based on traditional sequencing methods.

In this diverse population of study participants, we found no clear baseline participant demographic predictors for the presence of drug class-specific TDR other than an association between PI DRMs and age. This association appears to have been largely due to the presence of low-level PI variants and was not observed above the 5% detection threshold. There was also a weak trend of increasing NNRTI TDR over the time of study enrolment from 2009 to 2013.

When examining individual mutations by variant threshold, different patterns of DRM detection by drug class were observed. Some DRMs tended to occur almost exclusively or predominantly as low-level variants and this was observed in each drug class. The most commonly detected NRTI DRMs, K219QENR and M184VI, occurred predominately as minor variants while the majority of patients with T215 revertants had the mutation in > 80% of the quasispecies. Of the most frequent NNRTI DRMs, G190ASE occurred predominately as minor variants while K103NS was dominant in the quasispecies. Within the PI-related DRMs, the same picture emerged, with M46IL and D30N predominately seen as minor variants, whereas L90M (associated with saquinavir resistance), if present, dominated the quasispecies. Although smaller numbers of individuals were included in the Cunningham et al. study,[11] a similar pattern was observed, with the M46IL and D30N detected predominately as minor variants while the majority of L90M mutations occurred above the 20% threshold.

The detection of some DRMs predominately at low levels is likely due to impaired viral fitness, which has been described for mutations such as M184VI and D30N.[3,5,33–36] The M184VI may also be linked with other DRMs and tends to wane over time due to overgrowth of more replication competent wild-type virus.[37] However, mutations like the RT K103NS, the T215 revertants and the PI L90M appear to have little effect on viral fitness and may persist for prolonged periods in individuals with TDR.[3,8,28,36,38,39]

Predicted phenotypic susceptibility was based on an expanded list of mutations which had been detected by NGS and results varied by threshold of detection. Given that certain mutations occurred more frequently as minor variants, some agents had a greater degree of predicted phenotypic resistance, such as nelfinavir and raltegravir, when using the 2% threshold of detection. At the 20% threshold, predicted resistance to these agents was significantly diminished and remained mostly low or intermediate. By contrast, efavirenz and lamivudine/emtricitabine had mostly high-level predicted resistance at all thresholds. This is due to single mutations causing high-level resistance to these agents, which were detected at all thresholds, although the 184VI was less frequently detected above the 20% threshold. Overall, the first-generation NNRTIs efavirenz and nevirapine (which has a similar resistance profile) would be predicted as the least active agents in those initiating therapy in this population. Although there was a moderate amount of reduced susceptibility to rilpivirine observed at all thresholds, this was mostly intermediate- and low-level resistance. Rilpivirine is a second-generation NNRTI which usually requires multiple mutations to cause high-level resistance. Of note, there was very little predicted phenotypic resistance to darunavir and dolutegravir. These agents would probably be highly active in this population, as well as the newer second-generation integrase inhibitor bictegravir.

Potential limitations of this study include the threshold of detection for minor variants, as well as the ability to reliably sequence participant samples. NGS is a highly sensitive assay and there is the possibility of low-level variants occurring as laboratory artefacts or sequencing errors, but prior studies have demonstrated that most discrepancies in variant calls occur below the 2% threshold.[11,18,40,41] We were limited by our ability to produce an evaluable result by NGS from all available samples, particularly for RT and IN DRMs. Another limitation of our study was that we did not assess TDR in those individuals with low-level viraemia, as NGS was only attempted in those participants with baseline HIV RNA levels > 1000 copies/mL. Furthermore, although the average time from diagnosis of the participants was 1 year prior to study enrolment and all had early-stage HIV disease with CD4 counts > 500 cells/μL, this cohort probably represents a mix of those with more recent infection and others with a significantly longer duration of infection.

In summary, using NGS in the START population revealed significant geographic diversity in prevalence of TDR. Our study resulted in detection of a large proportion of low-level variants which would not have been detected by traditional Sanger sequencing, the method commonly used to assess individual pre-treatment drug resistance and for population TDR surveillance. Further studies will be needed to assess the potential clinical impact of transmitted minority variants on treatment response. Given that TDR continues to occur in most regions of the world and prevalence has been increasing to certain agents at an alarming rate in some countries, it will be important to continue surveillance efforts and consider the use of more sensitive drug resistance assays.