Detecting HIV-1 superinfection by pol gene population sequencing among untreated HIV-1-infected men who experience sudden rises in plasma HIV-1 RNA load

HIV superinfection is defined as infection with a second strain of HIV at least 1 month after primary infection [1]. At the population level, superinfection drives HIV genetic diversification through recombination. At the patient level, it reveals ongoing transmission risk and may result in accelerated disease progression and acquisition of drug-resistant strains [2,3]. Treatment guidelines recommend that all newly diagnosed patients undergo pol population sequencing in order to detect transmitted drug resistance [4,5]. Although it is recognized that a risk of superinfection exists among untreated patients who engage in high-risk behavior, the factors that should prompt repeat sequencing prior to starting antiretroviral therapy (ART) in order to identify a change in the drug susceptibility profiles remain poorly defined [4]. The issue is further compounded by uncertainties about the true incidence of superinfection, reflecting technical difficulties in detection [1]. We investigated the yield of pol sequencing for detecting superinfection and changes in the drug susceptibility profiles among ART-naive homosexual men who, during routine monitoring, experienced a sudden increase in plasma viral load. The viral load was quantified with the Abbott RealTime HIV-1 assay. Eligible patients had at least three viral load measurements taken as part of routine care while ART-naive. During monitoring, they showed a viral load increase at least 0.5 log10 copies/ml relative to the mean of two previous measurements differing by less than 0.45 log10 copies/ml (viral load event). There were 573 eligible patients who over median 23 [interquartile range (IQR) 9–44] months underwent median 7 (IQR 4–12) viral load measurements, on average every 3 months (SD = 3). Overall, 182 of 573 (31.8%) patients experienced 226 viral load events, with median (range) viral load increase of 0.7 (0.5–1.5) log10 copies/ml. Pol (protease aa 1-99; reverse transcriptase (RT) aa 1-335) population sequences were obtained from plasma collected at two time points before (baseline) and two time points after (follow-up) the viral load event. Among 47 viral load events in 46 patients, two (4.3%) events in two (4.4%) patients showed a divergence of baseline and follow-up pol sequences by phylogenetic analysis. The divergence was confirmed by gag and integrase population sequencing and clonal pol sequencing (Fig. 1). The first patient had initially presented with an acute seroconversion illness characterized by severe occipital headache. Five months later, he experienced a viral load event (+0.6 log10 copies/ml) coinciding with a CD4 cell count decline (423 to 247 cells/μl) and recurrence of the same symptoms. The baseline strains had the RT mutation V179D [resistance effects for the nonnucleoside RT inhibitor (NNRTIs)], and the secondary protease mutations L10I and L63T; the follow-up strains carried no drug-resistance mutations in RT and the secondary mutations L63P, A71T and I93L in protease. The viral load remained above 5.0 log10 copies/ml after the event, whereas the CD4 cell count declined further leading to initiation of ART. The second patient experienced a viral load event (+1.9 log10 copies/ml) 18 months after the initial HIV diagnosis, having previously shown a viral load below 3.0 log10 copies/ml and a CD4 cell count above 1000 cells/μl. The viral load event occurred in the absence of symptoms or observed CD4 cell count decline. The baseline strains had the RT mutations M41L and T215S (resistance effects for multiple nucleos(t)ide RT inhibitors (NRTIs); the follow-up strains showed no drug-resistance mutations. Over the following 2 years, the viral load returned stably to below 3.5 log10 copies/ml, whereas the CD4 cell count remained above 1000 cells/μl in the absence of ART. These findings are consistent with previous observations that superinfection may occur in both early and established HIV infection and in the face of relatively effective immune responses [6].Fig. 1: Phylogenetic analysis of HIV-1 pol clonal sequences from two patients with divergence of pol, integrase and gag population sequences observed between samples collected before (baseline) and after (follow-up) a sudden increase in plasma HIV-1 RNA load. Maximum-likelihood trees were constructed under the General Time Reversible model of nucleotide substitution with proportion of invariable sites and gamma-distributed rate heterogeneity. Bootstrap support values at least 95 are indicated (1000 replicates). Divergent sequences were analyzed for the presence of recombination using SimPlot distance plotting, Recombination Detection Program, version 3.41, and tree construction using sequence fragments. There was no evidence of recombination between sequences. All sequences were subtype B.Whether pol population sequencing is sufficiently sensitive to detect all cases of superinfection is doubtful. From a clinical perspective, however, pol population sequencing is widely available in clinical practice to guide the composition of ART. Previous reports have shown that superinfection can result in both acquisition and loss of drug-resistance mutations [3,7]. We observed a significant change in the drug susceptibility profiles of the two patients with superinfection, with loss of major drug-resistance mutations. However, archiving of the previously detected mutants should be considered likely even when the incoming variant appears to replace the initial strain as the dominant virus in peripheral blood. The ability of the superinfecting virus to outgrow the initial strain has been described [1]. In fact, previous studies have suggested that cases in which the superinfecting virus becomes dominant are those in which viral load increases are observed [8]. In summary, sudden viral load increases occurred relatively commonly among untreated patients; however, using pol population sequencing as a screening tool, sequence divergence and a change in the drug susceptibility profiles were found in only 4% of viral load events. In clinical practice, yield may be enhanced by taking into account risk factors for superinfection, whereas changes in CD4 cell counts may not be detected at the time of the viral load event. Acknowledgements The study was supported by Research and Development Fund of the Royal Free Hampstead NHS Trust. There is no commercial funding to declare. The authors have no conflict of interest to declare in relation to this study.