Determinants of reservoir contraction and expansion in vivo, ex vivo, and in vitro

The advent of antiviral therapy (ART) led to decreased morbidity and mortality among individuals with HIV by reducing viral loads to undetectable levels. It also revealed a treatment resistant reservoir. Infected CD4+ T cells are the major contributor to that reservoir and are a source of recurrent viremia when ART is stopped. Our long-term goal is to dissect the contraction and expansion forces exerted on the persistent HIV reservoir by performing parallel sequencing experiments in resting CD4+ T cells in vivo, ex vivo, and in vitro. We were motivated to monitor the HIV reservoir by proviral sequencing 1-6 to overcome the erroneous HIV DNA estimates that are dominated by defective proviruses. We first monitored reservoir dynamics in individuals treated during chronic infection using a sequencing approach that estimates reservoir size by multiplying the level of total HIV DNA by the frequency of intact proviruses 5. To capture a spectrum of immune responses we expanded our cohort to include those treated during acute infection or those treated while demonstrating elite control. An important advantage of utilizing sequencing to estimate reservoirs over other approaches is the ability to identify large proviral clones (or repeated sequences). By removing large proviral clones from our analysis we could provide a better estimate of inherent decay. Moreover, analyzing the repeated sequences alone provides an estimate of clonal expansion. While many in the field are studying the role of antigen and homeostatic cytokines on clonal expansion 7-9 we focus on changes in the genetic makeup of the provirus as we noticed specific viral elements were enriched in proviral clones (Fig. 3, blue D1-D4+,4). This observation led to the development of a machine learning approach that analyzes viral elements systematically overtime, without bias, to define genetic elements that promote reservoir contraction or expansion. Similarly, we will identify interactions between elements. In Fig. 12, we show the results of our machine learning analysis which revealed a role for D1 in reservoir decay, Nef for immune protection and Tat for clonal expansion 10.