Multi-omics and biochemical reconstitution reveal CDK7-dependent mechanisms controlling RNA polymerase II function at gene 5'- and 3' ends

Abstract

CDK7 regulates RNA polymerase II (RNAPII) initiation, elongation, and termination through incompletely understood mechanisms. Because contaminating kinases prevent reliable CDK7 analysis with nuclear extracts, we reconstitute RNAPII transcription with purified factors. We show that CDK7 inhibition slows and/or pauses RNAPII promoter-proximal transcription and suppresses re-initiation, and these effects are Mediator and TFIID dependent. Similarly in human cells, CDK7 inhibition reduces transcriptional output by suppressing RNAPII initiation and/or re-initiation. Moreover, widespread 3' end readthrough transcription occurs in CDK7-inhibited cells; mechanistically, this results from rapid nuclear depletion of RNAPII elongation and termination factors (e.g., DSIF, Integrator, NELF, SPT6, PPP1R10/PNUTS, and SCAF8), including high-confidence CDK7 kinase targets. Collectively, these results define how CDK7 governs RNAPII function at gene 5' ends and 3' ends and reveal that nuclear abundance of elongation and termination factors is kinase dependent. Because 3'-readthrough transcription is commonly induced during stress, our results further suggest that regulated suppression of CDK7 activity enables this transcriptional response.

Keywords: 3’-readthrough transcription; CDK7; CP: Molecular biology; Mediator; PRO-seq; RNA polymerase II; TFIID; in vitro transcription; promoter escape; promoter-proximal pausing; proteomics; re-initiation.

Figure 1.. Contaminating kinases in nuclear extracts preclude analysis of CDK7-dependent functions

(A) Overview of immobilized template experiments. (B) Western blots showing all PIC factors bound to the HSPA1B promoter template, as expected. Antibodies targeted RPB1 (RNAPII), MED23, XPB (TFIIH), GTF2F1, GTF2E1, GTF2A1, TBP, and GTF2B. (C) RNAPII western blots show shifted band following NTP addition, consistent with RPB1 CTD phosphorylation. Technical replicates are shown side-by-side for each condition, and data from 2 different biological replicates are shown (experiment 1 & experiment 2). Experiment 2 used a covalent CDK7 inhibitor, SY-351. (D) The shifted RPB1 band was observed after incubation with ATP only or NTPs; thus, it is not exclusively transcription dependent. Shifted RPB1 band was phosphorylated based upon incubation with λ-phosphatase vs. no phosphatase control. An additional wash step followed by incubation in phosphatase buffer caused the decrease in RPB1 band intensity. (E) Summary of kinases bound to the native HSPA1B promoter after washing, based upon mass spectrometry analysis. Although CDK7 was a top hit, many other kinases were detected. The control (CTRL) experiment used beads only. (F) Purified CAK module (CDK Activating Kinase: CDK7, CCNC, and MNAT1) phosphorylates full-length mammalian glutathione S-transferase (GST)-CTD (mw = 100 kDa) in vitro; phosphorylation is blocked by 50 nM SY-5609. Some phosphorylation occurred at 30 min, a time frame longer than in vitro transcription reactions; ATP concentration was 20,000-fold higher than SY-5609 in these experiments (1 mM vs. 50 nM). (G) Purified reconstituted PICs assembled on the HSPA1B promoter contain only one kinase, CDK7. These PICs phosphorylate the RNAPII CTD, but phosphorylation is blocked by SY-5609, although some phosphorylation can be detected after 30 min.

Figure 2.. CDK7 inhibition blocks promoter escape and pause release and reduces overall transcription output

(A) Schematic of the human pre-initiation complex (PIC). (B) Overview of the in vitro transcription pausing assays. SY-5609 (or vehicle control) was incubated with PIC factors prior to PIC assembly. (C) Representative gel images of pausing assays, with pre-escape, paused, and runoff/elongated regions indicated. Each region was independently contrast-enhanced to aid visualization, and each region was adjusted equally across comparisons (e.g., CTRL vs. SY-5609). Quantitation was completed with uniform exposure across the entire lane; gel lanes with uniform exposure are shown in Figure S1B. (D) CDK7 inhibition decreases and slows promoter escape. Each point (n = 6) represents the ratio of pre-escape transcripts (5–15 nt length) to runoff products (150–216 nt) from reactions treated with 50 nM SY-5609, normalized to no-inhibitor controls. Bars represent mean ± SEM. (E) CDK7 inhibition increases pausing. Plot shows the pause index of reactions (n = 6) treated with SY-5609, normalized to no-inhibitor controls. Bars represent mean ± SEM. (F) Overview of in vitro transcription runoff assays. (G) CDK7 inhibition reduces overall RNAPII transcriptional output (n = 7). Bars represent mean ± SEM. Inset: representative gel images of runoff transcription ±SY-5609. See also Figure S1C.

Figure 3.. CDK7 kinase increases RNAPII re-initiation; Mediator and TFIID cooperate with CDK7 to enhance promoter escape and pause release, respectively

(A) Representative smTIRF data, with images of RNA produced from immobilized HSPA1B promoter (scale bar 1 μm), and corresponding intensity trace below. In this case, two transcripts were generated from this single promoter. Green dashed line represents signal filter that distinguishes RNA transcripts from background noise with high confidence. (B) Total number of runoff transcripts at the HSPA1B promoter. Bars represent mean ± SEM from biological replicate experiments. (C) Number of RNAPII re-initiation events ±SY-5609; bars represent mean ± SEM from biological replicate experiments. (D) Number of HSPA1B templates that generated a runoff transcript ±SY-5609. Bars represent mean ± SEM from biological replicate experiments. (E) Percentage of templates that re-initiated ±SY-5609. That is, among templates that produced a runoff transcript, the percentage that generated a second runoff transcript is plotted. Bars represent mean ± SEM from biological replicate experiments. (F) Quantitation of runoff transcription +SY-5609, normalized to no-inhibitor controls (dashed line), comparing PICs ±Mediator (n = 6). Bars represent mean ± SEM. Representative gel lanes in Figure S1F. (G) TFIID affects runoff transcription as a function of CDK7 kinase activity. Data from PICs ± TFIID shown, +SY-5609, normalized to no inhibitor controls (dashed line). No IID = TBP instead of TFIID (n = 7). PIC data (left column, dark purple dots) are the same as in (F). Bars represent mean ± SEM. Representative gel lanes in Figure S1G. (H) Representative gel images from TBP PIC pausing assays, with pre-escape, paused, and runoff/elongated regions indicated. Each region was independently contrast-enhanced to aid visualization, and each region was adjusted equally across comparisons (e.g., CTRL vs. SY-5609). With TBP PICs, paused region transcripts were nearly undetectable, and this region was maximally contrast-enhanced for visualization ±SY-5609. Quantitation completed with uniform exposure across the entire lane; gel lanes with uniform exposure are shown in Figure S1H. (I) In contrast to TFIID PICs, pausing does not occur with TBP PICs, and CDK7-dependent effect on pausing is lost (n = 7). Purple points are the same as those in Figure 2E, shown here for comparison. Bars represent mean ± SEM.

Figure 4.. CDK7 inhibition reduces RNAPII transcription genome-wide, including at promoter-proximal regions, but increases promoter-proximal RNAPII occupancy

(A) Violin plots of normalized PRO-seq count data. Interquartile range (IQR) counts are plotted (i.e., the middle 50% of data), showing a reduction in SY-5609-treated cells. (B) CDK7 inhibition increases RNAPII pause index (n = 2,400 genes; p = 1.26E⁻¹¹). (C) Metagene plot of normalized PRO-seq reads at highly expressed genes (n = 453), focused on promoter-proximal region ±SY-5609. (D) Metagene analysis of ChIP-seq data (normalized to read depth), focused on promoter-proximal region ±SY-5609 at high-occupancy RNAPII genes (n = 185). Elevated RNAPII occupancy +SY-5609 coupled with reduced PRO-seq reads +SY-5609 (C) suggests increased levels of inactive RNAPII around the TSS. As expected, levels of RNAPII CTD ser5-P and ser7-P decrease at the TSS in SY-5609-treated cells.

Figure 5.. Readthrough defects and depletion of RNAPII elongation and termination factors in CDK7-inhibited cells

(A) Representative PRO-seq gene traces demonstrating increased 3′-end readthrough transcription +SY-5609. (B) Metagene analysis at long genes (>120 kb, n = 684) shows extensive 3′ end readthrough transcription with SY-5609 treatment. Position 0 indicates the polyA site. (C) Heatmap summarizing quantitative TMT-MS data from OV90 and HCT116 cells treated with 50 nM SY-5609 for 120 min (vs. DMSO controls). Proteins shown were identified as significantly changing in both OV90 and HCT116 cells (p < 0.01). Asterisk: RNAPII elongation/termination factors. (D) GSEA results (GO Biological Processes set), based on TMT-MS data ±SY-5609, plotted by normalized enrichment score (NES) for OV90 and HCT116 cells. Significantly upregulated or downregulated pathways (false discovery rate [FDR] q < 0.05) are highlighted in color, and a representative subset of pathways are listed.

Figure 6.. CDK7 inhibition (CDK7AS line) depletes RNAPII elongation and termination factors, many of which are phosphorylated by CDK7

(A) Heatmap showing significantly changing (p < 0.01) protein abundance from quantitative TMT-MS experiments in CDK7AS OV90 cells ±3MB-PP1 (10 μM, t = 4 h) or OV90 cells ±SY-5609 (50 nM, t = 2 h). Proteins associated with transcription elongation and termination are starred. (B) GSEA results (GO Biological Processes set), based on TMT-MS data ±SY-5609 (y axis; OV90) or ±3MB-PP1 (x axis; CDK7AS OV90), plotted by their normalized enrichment score (NES). Significantly upregulated or downregulated pathways (false discovery rate [FDR] q < 0.05) are highlighted in color, and a subset of pathways are listed at right. (C) Volcano plot showing proteins with phosphorylation changes ±SY-5609 (50 nM, 1 h) in HCT116 cells. Only proteins that were also nuclear depleted +SY-5609 are included here. Among all proteins depleted with statistical confidence, about 20% showed phospho-site changes (93/483; Venn diagram inset).

Figure 7.. Model: CDK7 regulates RNAPII transcription at gene 5′ ends and 3′ ends through multiple mechanisms

At 5′ ends, CDK7 inhibition slows or stalls RNAPII in the promoter escape and promoter-proximal pause regions, causing a steric block for re-initiation that may contribute to RNAPII inactivation at promoters in CDK7-inhibited cells. CDK7 regulation of promoter escape is Mediator dependent whereas its regulation of pausing is TFIID dependent. Separately, CDK7 inhibition causes nuclear depletion of RNAPII elongation and termination factors, which is sufficient to trigger 3′-readthrough transcription. CDK7 inhibition also causes phospho-site changes in many elongation and termination factors, which may independently contribute to 3′-readthrough defects. PIC factors are represented as globular in the promoter escape and pause regions based upon cryo-EM data that indicate PIC structural dynamics post-initiation.^,,