[HTML][HTML] Insufficient evidence for rare activation of latent HIV in the absence of reservoir-reducing interventions

AL Hill, DIS Rosenbloom, JD Siliciano… - PLoS …, 2016 - journals.plos.org
PLoS Pathogens, 2016journals.plos.org
A priority for HIV cure research is measuring latent infection that can fuel viral recrudescence
if patients cease antiretroviral therapy (ART). One important quantity that is difficult to
measure is the rate at which latently infected cells activate, giving rise to spreading infection
[1]. Pinkevych et al. recently estimated this rate by analyzing several clinical cohorts and
concluded that one cell activates every 6 days [2]. This rate is 24-fold lower than our
previous estimate of 4 cells/day [1], resulting in far more optimistic predictions for the …
A priority for HIV cure research is measuring latent infection that can fuel viral recrudescence if patients cease antiretroviral therapy (ART). One important quantity that is difficult to measure is the rate at which latently infected cells activate, giving rise to spreading infection [1]. Pinkevych et al. recently estimated this rate by analyzing several clinical cohorts and concluded that one cell activates every 6 days [2]. This rate is 24-fold lower than our previous estimate of 4 cells/day [1], resulting in far more optimistic predictions for the prospects of reservoir-reducing therapy. We question their estimation approach and suggest that a higher rate is likely for most infected individuals.
The main problem with the analysis is that it treats all participants as identical, with a homogeneous virus population that activates and grows at the same rate. The only variation between participants considered is the (random) timing of viral activation. After treatment interruption, it is common to see one person rebound after two weeks, and another after three weeks. By their model, the first person's virus must have activated one week before the second person's virus did. If the activation rate were high, this would be extremely unlikely. The authors conclude that viral activation occurs about once per week, not multiple times per day as previously proposed [1, 3–5]. Below, we offer an alternative explanation. We show that interperson variation in viral activation or growth rates can explain the observed variation in rebound times. While Pinkevych et al. discuss the issue of interperson variation, they hypothesize that their fitted values properly estimate cohort averages. Our analysis below challenges this hypothesis: interperson variation does not merely lead to variation around a central estimate in activation rate but rather skews the estimate downward. We conclude by arguing that high activation rates are suggested by the weak correlation between preinterruption latent reservoir sizes and postinterruption viral rebound times. In the Pinkevych et al. model, latently infected cells reactivate at a constant rate (k). Reactivated virus cannot spread until drug “washes out” following ART interruption. Viral activation occurs at a random, exponentially distributed time (average value 1/k) after washout. Plasma virus levels then grow exponentially from V0 at rate r. The proportion of individuals with viral rebound to detectable levels by a particular time can be calculated and compared to participant data to estimate parameters.
PLOS