Fig 2: TAF1 loss impairs Pol II entry into productive elongation.a Immunoblot from SNAP-Pol II TAF1-Halo nuclear extracts showing time-course degradation of TAF1 following treatment with 300 nM HaloPROTAC3 for 0 min, 15 min, 30 min, 1 h, and 2 h. The 2-hour HaloPROTAC3 treatment condition was repeated in two additional independent experiments with similar TAF1 depletion relative to control. b Volcano plot showing proteomic comparison of DMSO (Control, n = 3 biological replicates, 2 h) vs 300 nM HaloPROTAC3 (Treatment, n = 3 biological replicates, 2 h). Differential protein abundance was assessed using limma moderated t-tests, with Benjamini-Hochberg FDR adjustment for multiple hypothesis testing. Proteins with adjusted p-value < 0.05 and |Log2FC | > 0.5 were considered significant. c Live-cell snapshot of SNAP-Pol II and TAF1-Halo labeled cells. The cells were labeled with 6 nM SNAP-Cell TMR-549 (NEB), and 5 nM HaloTag JF646 (Promega) (Single frame, 500 millisecond exposure). Cells treated with either DMSO or 300 nM HaloPROTAC3 for 2 h prior to imaging. Pseudo-coloring was used for display, and a global threshold was applied to each channel to ensure comparable signal range across drug conditions. d Per-cell percentages of Pol II entering productive elongation for DMSO (n = 54 cells, 2 hours) and HaloPROTAC3-treated cells (n = 51 cells, 300 nM, 2 h). A two-sided Mann-Whitney U test was performed to determine statistical significance (P = 2 × 10-5). Source data are provided in the Source Data file.

Fig 3: Establishing a CRISPR knock-in Pol II and TAF1 cell line.a Schematic showing Pol II (PDB: 7NVS) and TAF1 (TFIID, PDB: 6MZL unbound; 6MZM bound) in an unbound and bound state at the super core promoter. SNAP (PDB: 6Y8P) and Halo (PDB: 6U32) tags were inserted into the structured N-terminus of the POLR2A gene and the structured C-terminus of TAF1, respectively. Structural models were assembled using UCSF Chimera and are for illustrative purposes only; they are based on fragments of proteins from published PDB structures and do not represent an experimentally determined full complex. b CRISPR-Cas9 strategy for dual knock-in. SNAP-tagged Pol II was generated first, followed by targeting the HaloTag into the C-terminus of TAF1 in the SNAP-Pol II background. c Western blot analysis confirming expression of tagged proteins. Whole-cell lysates from WT, single knock-in (SNAP-Pol II), and dual knock-in (SNAP-Pol II TAF1-Halo) lines were probed with antibodies against Pol II, TAF1, and vinculin (loading control). Migration shifts indicate successful tagging. Blots are representative of three independent experiments with similar results. d Growth curves showing that SNAP-Pol II and dual knock-in cells exhibit normal proliferation rates compared to WT. Data are plotted as mean ± SEM from three biological replicates, with each biological replicate measured from three technical replicate wells. Curves were fit using a nonlinear Malthusian growth model, and growth rates were compared by an extra sum-of-squares F test (p value = 0.1265; null hypothesis: equal growth rate across all strains). e Live-cell imaging of SNAP-Pol II and TAF1-Halo reveals nuclear localization of both tagged proteins. Cells were labeled with SPY505-DNA (Cytoskeleton), SNAP-Cell 647-SiR (NEB), and HaloTag JFX549 (Promega), and imaged using a single-frame 1-second exposure. Pseudo-coloring was used for display, and a global threshold was applied to each channel to ensure comparable signal range across all cell lines. Source data are provided in the Source Data file.

Fig 4: Single-molecule imaging reveals globally inefficient transcription.a Conceptual model illustrating kinetic loss of Pol II during distinct steps in early transcription. Most Pol II dissociate at or shortly after binding the genome, due to transient binding, unstable early engagement, or promoter-proximal pause instability near the +1 nucleosome (PDB: 7Y8R; outlined in red and orange). Only a small fraction of Pol II can successfully transition into productive elongation (outlined in green). b Live-cell single-molecule imaging of SNAP-Pol II. Top: raw single frame showing sparse fluorescent spots. The orange box outlines a zoomed-in region shown below, where a time-lapse series (0–90 s) captures two individual Pol II binding events. One spot appears at 0 s (green outline), and a second spot appears at 66 s (red outline). c GRID event spectrum of Pol II chromatin binding events in DMSO treated cells (n = 52 cells, 30 min). Fits for a population during a resampling event are represented as open circles. Resampling events with fitted lifetimes below the imaging interval ( < 1 s) were excluded from interpretation. d Immunoblot of nuclear extracts from dual knock-in (SNAP-Pol II TAF1-Halo) cell lines treated with DMSO, 500 nM Triptolide, or 500 nM NVP-2 for 30 min. e Total number of nuclear Pol II binding events per cell in DMSO (n = 52 cells), Triptolide (n = 18 cells), and NVP-2 (n = 59 cells) conditions. Each point represents a measurement from a single nucleus. Kruskal-Wallis with Dunn’s post-hoc was performed to determine statistical significance. f,g GRID state spectrum of Pol II chromatin binding events in DMSO (f, n = 52 cells) and NVP-2 (g, n = 59 cells) treated cells. For NVP-2, populations boxed in green increase relative to DMSO while those boxed in red decrease. Drug concentration and treatment time indicated in (d). Source data are provided in the Source Data file.

Fig 5: Acute TAF1 depletion impairs pause release and decreases SPT5-pThr799.a Per-cell Pol II pausing index for DMSO (n = 54 cells, 2 h) and HaloPROTAC3-treated cells (n = 51 cells, 300 nM, 2 h). A two-sided Mann-Whitney U test was performed to determine statistical significance (P = 5 × 10-8). b GRID state spectrum of Pol II chromatin binding events for the DMSO and HaloPROTAC3 dataset in (a). Abbreviations as in Fig. 3b. For HaloPROTAC3, populations boxed in green increase relative to DMSO while those boxed in red decrease. c Immunoblot in U2OS SNAP-Pol II TAF1-Halo nuclear extracts. Cells were treated with DMSO, NVP-2 (50 nM), or HaloPROTAC3 (300 nM) for 2 h prior to harvesting nuclear extracts. d HaloPROTAC3 washout for TAF1 endogenous rescue immunoblot. DMSO and HaloPROTAC3 treatment and timing is the same as (c). e Per-cell Pol II pausing index for DMSO (n = 111 cells, 2 h), NVP-2 (n = 110 cells, 50 nM, 2 h, adjusted p-value = 4 × 10-6), HaloPROTAC3 (n = 101 cells, 300 nM, 2 h, adjusted p-value = 1 × 10-5), 10-hour (n = 52 cells) and 24-hour (n = 49 cells) HaloPROTAC3 washout conditions. f Total number of Pol II clusters detected across all single cells in DMSO (n = 105 cells, 2 h), NVP-2 (n = 103 cells, 50 nM, 2 h, adjusted p-value = 6 × 10-21), HaloPROTAC3 (n = 94 cells, 300 nM, 2 h, adjusted p-value = 5 × 10-9), 10-hour (n = 52 cells) and 24-hour (n = 44 cells) HaloPROTAC3 washout conditions. g Average number of Pol II binding events detected in clusters across drug conditions used in DMSO (n = 105 cells, 2 h), NVP-2 (n = 99 cells, 50 nM, 2 hours, adjusted p-value = 2 × 10-6), HaloPROTAC3 (n = 97 cells, 300 nM, 2 h, adjusted p-value = 2 × 10-4), 10-hour (n = 52 cells) and 24-hour (n = 46 cells) HaloPROTAC3 washout conditions. A Kruskal-Wallis with Dunn’s post-hoc was performed to determine statistical significance for (e–g). Source data are provided in the Source Data file.