Dynamic transcription pre-initiation complex assembly governs initiation efficiency

Transcription initiation is a highly regulated process that determines gene expression outcomes ^1,2 , yet the dynamics of initiation and the mechanisms governing efficiency remain poorly understood. Here, we combine endogenous tagging of human RNA polymerase II (Pol II) and TFIID with simultaneous live-cell, multi-color single-molecule imaging to quantitatively map Pol II behavior during transcription initiation and early elongation. Using GRID (Genuine Rate Identification) analysis, we resolved four distinct kinetic populations of chromatin-bound Pol II. The dynamics of Pol II populations reveal that initiation is highly inefficient, with over 94% of Pol II molecules dissociating within tens of seconds. Kinetic partitioning of Pol II dwell times enables quantification of proximal pausing, which is globally sensitive to CDK9 inhibition. Single-cell analysis uncovers substantial heterogeneity in initiation efficiency and pausing across individual cells. Colocalization of Pol II with TFIID is associated with higher initiation efficiency and reduced promoter-proximal pausing compared to global Pol II. Further dissection of Pol II-TFIID assembly pathways reveals that canonical assembly, where TFIID binds first, is linked to inefficient initiation and frequent pausing. In contrast, non-canonical assembly, where Pol II binds first followed by TFIID, supports more efficient initiation with lower pausing. Together, these findings establish that transcription initiation efficiency is shaped by both the kinetic stability of Pol II engagement and the temporal order of pre-initiation complex assembly, providing a new framework for understanding dynamic gene regulation in vivo.