Christiaan H. van Dorp , Jessica M. Conway, Dan H. Barouch, James B. Whitney, Alan S. Perelson
In order to assess the efficacy of novel HIV-1 treatments leading to a functional cure, the time to viral rebound is frequently used as a surrogate endpoint. The longer the time to viral rebound, the more efficacious the therapy. In support of such an approach, mathematical models serve as a connection between the size of the latent reservoir and the time to HIV-1 rebound after treatment interruption. The simplest of such models assumes that a single successful latent cell reactivation event leads to observable viremia after a period of exponential viral growth. Here we consider a generalization developed by Pinkevych et al. and Hill et al. of this simple model in which multiple reactivation events can occur, each contributing to the exponential growth of the viral load. We formalize and improve the previous derivation of the dynamics predicted by this model, and use the model to estimate relevant biological parameters from SIV rebound data. We confirm a previously described effect of very early antiretroviral therapy (ART) initiation on the rate of recrudescence and the viral load growth rate after treatment interruption. We find that every day ART initiation is delayed results in a 39% increase in the recrudescence rate (95% credible interval: [18%, 62%]), and a 11% decrease of the viral growth rate (95% credible interval: [4%, 20%]). We show that when viral rebound occurs early relative to the viral load doubling time, a model with multiple successful reactivation events fits the data better than a model with only a single successful reactivation event.
HIV and SIV are able to persist despite antiretroviral therapy (ART) because of a long-lived reservoir of latently infected CD4+ T cells . Recent studies have shown that the latent reservoir is established very early after infection [2–4], and that the seeding of the reservoir can only be prevented when ART starts extremely early . Other studies have focused on the effect of potentially curative treatment strategies that might extend remission after interruption of ART [6–8].
Materials and methods
The collection of the data is described in detail by Whitney et al. [2, 5]. In short, 36 rhesus macaques were infected with 500 TCID50 of SIVmac251. Combination antiretroviral treatment (a cocktail of tenofovir, emtricitabine, and dolutegravir) was initiated at various times post infection (6 hours, 1, 2, 3, 7, 10, and 14 days). Treatment continued for 24 weeks, and the viral load (VL) was monitored for 16 weeks after treatment interruption, while taking weekly measurements with a limit of detection of 50 RNA copies per mL.
We carefully analyzed a model for SIV and HIV rebound after treatment interruption developed by Pinkevych et al.  and Hill et al.  that takes into account the potential effect of the reactivation of multiple latently infected cells on the rebound time. In doing so, we were able to derive a relatively simple statistical model that can be used for the inference of the rate of recrudescence after treatment cessation, the viral growth rate after recrudescence, and perhaps ultimately the efficacy of novel HIV treatments in delaying viral rebound. Moreover, using our mathematical formulation, the model can be compared to similar models of viral rebound in a statistically rigorous manner. We were able to find strong statistical evidence (ΔWAIC = 11.5) in favor of the multiple-reactivation model over a simple model with only one reactivation event using previously published data from treatment-interruption experiments performed in SIV-infected macaques [2, 5]. We argued that the multiple-reactivation model is most relevant for data sets that contain subjects with early viral rebound, as our SIV data set. This is often the case for human data sets as well. For example in a pooled data set of six ACTG studies , 6–63% of subjects showed detectable viremia within a week, and 21–74% within 2 weeks of ART cessation .
We gratefully acknowledge Garrett T. Nieddu for his technical support.
Citation: van Dorp CH, Conway JM, Barouch DH, Whitney JB, Perelson AS (2020) Models of SIV rebound after treatment interruption that involve multiple reactivation events. PLoS Comput Biol 16(10): e1008241. https://doi.org/10.1371/journal.pcbi.1008241
Editor: Miles P. Davenport, UNSW Australia, AUSTRALIA
Received: February 3, 2020; Accepted: August 12, 2020; Published: October 1, 2020
This is an open access article, free of all copyright, and may be freely reproduced, distributed, transmitted, modified, built upon, or otherwise used by anyone for any lawful purpose. The work is made available under the Creative Commons CC0 public domain dedication.
Data Availability: All scripts and data are available on GitHub (https://github.com/lanl/multiple-reactivation-model).
Funding: Portions of this work were done under the auspices of the U.S. Department of Energy under contract 89233218CNA000001 and supported by National Institutes of Health (www.nih.gov) grants P01-AI131365 (JBW); R01-AI028433 and R01-OD011095 (ASP); AI124377, AI126603, and AI128751 (DHB); and National Science Foundation (www.nsf.gov) grant DMS-1714654 (JMC). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.