The latent reservoir for HIV consists of provirus stably integrated into long-lived lymphocytes and represents a barrier to a cure. This reservoir decays with a half-life of ~44 months during long-term suppressive ART, but the relative role of intrinsic T cell longevity vs proliferation in this persistence is unknown. Recent characterization of integration sites and full-genome sequences has suggested that clonal proliferation contributes to reservoir stability, but this rate has yet to be quantified. Estimating the value of this turnover rate is necessary to understand whether anti-proliferative therapy is a viable treatment option.
Here we develop a method to infer the underlying dynamics of cells in the latent reservoir from clone size distributions in sampled cells. We created a dynamic, stochastic mathematical model for cells in the latent reservoir, and fit this model to data using a Bayesian Markov Chain Monte Carlo estimation procedure that includes the size of the latent reservoir before ART initiation, the time on ART, and the later sampling of infected cells. The inference algorithm was applied to two sources of data: HIV integration site frequencies and intact provirus determined by full-genome sequencing.
Using HIV DNA integration site data and a simple homogeneous model of latent cell division and death, we estimated that on average latent cells divide around 10 times/yr [95% CI 5-20/yr], a much higher turnover than predicted by the total decay rate of 0.2/yr (e.g. half-life 44 months). For individual patients, the best-estimated turnover rate varied by up to an order of magnitude [2-30/yr]. Results were similar using intact virus only, though more uncertain due to smaller sample sizes. We used simulated populations to confirm our inference method was unbiased and required ~500 samples from each patient to reduce the uncertainty in turnover rate to +/-40% (95% CI). We found that an augmented model which also allowed for rare, burst-like proliferation could explain the clone size distribution better than the simple model. These findings suggest that therapy which reduced proliferation by 50% could reduce reservoir half-life to a few months, raising the potential for eradication with a few years of ART.
Our findings suggest that proliferation of latently-infected cells is a major contributor to the stability of the total and intact DNA reservoir and that reducing this proliferation may have potential as a curative intervention.