Abstract
The stages of transcription initiation and elongation are critical in the regulation of HIV-1 gene expression. Recent single-molecule imaging in living cells has shown that HIV-1 transcription occurs across multiple time scales and plays a key role in the control of latency. However, the molecular mechanisms of HIV-1 transcription remain poorly understood due to the lack of a unified modeling framework and advanced computational methods for analyzing HIV-1 imaging data. Here, we present a general stochastic model that characterizes HIV-1 transcription dynamics and computes the distributions of initiation times and nascent RNA counts. Our results show that coordination between initiation and elongation modulates transcription dynamics and that leveraging initiation-time data enhances model identification. Meanwhile, we develop a statistical inference method that integrates initiation-time data and nascent RNA data. Our results show that incorporating initiation-time data allows for accurate inference of the initiation rate and elongation time, with these parameter estimates being independent of the models used. When applied to HIV-1 transcription data in living cells, our theory and inference methods confirm the dual role of Tat in HIV-1 transcriptional regulation. In addition, the optimal predictive model indicates that Tat induces viral reactivation and latency exit by altering the number of silent states of the promoter. Our approach may provide the potential to improve current HIV-1 cure strategies.