RBBP6 loss disrupts PAS-dependent termination of sense transcription.

A. Western blot demonstrating the depletion of dTAG-RBBP6 over a time course of dTAGv-1 addition. SPT5 serves as a loading control.

B. Genome browser track of NEDD1 in POINT-seq data from dTAG-RBBP6 cells treated or not (2hr) with dTAGv-1. RBBP6 depletion induces a termination defect in the protein-coding direction (downstream of the indicated PAS) but not the upstream antisense direction. The Y-axis shows Reads Per Kilobase per Million mapped reads (RPKM).

C. Metaplot of POINT-seq data from dTAG-RBBP6 cells treated (RBBP6 dep) or not (Ctrl) (2hr) with dTAGv-1. This shows 1316 protein-coding genes selected as separated from any expressed transcription unit by ≥10kb. Signals above and below the x-axis are sense and antisense reads, respectively. The Y-axis scale is RPKM. TSS=transcription start site; TES=transcription end site (this marks the PAS position).

D. Heatmap representation of the data in C, which displays signal as a log2 fold change (log2FC) in RBBP6 depleted versus un-depleted conditions.

INTS11 loss disrupts the termination of antisense transcription.

A. Western blot demonstrating homozygous tagging of INTS11 with dTAG and the depletion of INTS11-dTAG after 1.5 hr treatment with dTAGv-1. INTS8 serves as a loading control.

B. Genome browser track showing POINT-seq signal over RNU5A-1 and RNU5B-1 in INTS11-dTAG cells treated or not (1.5 hr) with dTAGv-1. Note, that INTS11 depletion induces a termination defect in each case. Y-axis shows RPKM following spike-in normalisation.

C. Genome browser track showing POINT-seq signal over PGK1 and its upstream antisense region in INTS11-dTAG cells treated or not (1.5 hr) with dTAGv-1. Y-axis shows RPKM following spike-in normalisation.

D. Metaplot of POINT-seq data from INTS11-dTAG cells treated or not (1.5 hr) with dTAGv-1. This shows 1316 expressed protein-coding genes that are separated from any expressed transcription unit by ≥10kb. Signals above and below the x-axis are sense and antisense reads, respectively. Y-axis shows RPKM following spike-in normalisation.

E. Heatmap representation of the data in D, which displays signal as a log2 fold change (log2FC) in INTS11 depleted versus undepleted conditions over a region 3kb upstream and downstream of annotated TSSs.

Transcription initiation is more efficient and focused in the sense direction.

A. Schematic of sPOINT-seq protocol. The POINT-seq protocol is followed, in which chromatin is isolated and engaged RNAPII is immunoprecipitated. Short transcripts are preferentially amplified during library preparation (see Experimental Procedures for full details).

B. Comparison of POINT-(top trace) and sPOINT-seq (lower trace) on PGK1. Y-axis units are RPKM.

C. Metaplot comparison of POINT-(top plot) and sPOINT-seq (lower plot) profiles across the 684 highest expressed protein-coding that are separated from expressed transcription units by ≥10kb.. Signals above and below the x-axis are sense and antisense reads, respectively. Y-axis shows RPKM following spike-in normalisation.

D. Top metaplot shows full read coverage for sPOINT-seq performed in INTS11-dTAG cells treated or not (1.5hr) with dTAGv-1. The lower metaplot is the same data but only the 5’ end of each read is plotted. Both plots display a region +/-1kb from the TSS of the top 20% (based on sPOINT signal) promoters. Y-axis signals are RPKM following spike-in normalisation.

E. Metaplot zoom of the antisense TSS signals deriving from the lower plot in D. This makes clear the dispersed sites of initiation. The Y-axis scale is RPKM following spike-in normalisation.

F. Genome browser track of PGK1 promoter region in sPOINT-seq. This showcases the focused sense TSS (black arrows) and the dispersed antisense reads (brackets) generalised in the metaplots in D and E. Note the higher y-axis scale (RPKM) for sense vs. antisense.

Promoters lose their directionality when CDK9 is inhibited.

A. Genome browser track of CDCA7 in POINT-seq data derived from INTS11-dTAG cells either untreated, dTAG treated, NVP-2 treated or dTAG and NVP-2 treated (2.5 hr). Signals above and below the Y-axis are sense and antisense reads, respectively. The Y-axis scale shows RPKM following spike-in normalisation.

B. Metaplot of POINT-seq analysis of POINT-seq in INTS11-dTAG cells depleted or not of INTS11 and treated or not with NVP-2 (2.5 hr). This shows 1316 protein-coding genes selected as separated from any expressed transcription unit by ≥10kb. The regions 3kb upstream and downstream of genes are included. Y-axis units are RPKM following spike-in normalisation.

C. Metaplot of the same CDK9i + and -dTAG data shown in B but zoomed into the region 3kb upstream and downstream of the TSS.

D. Heatmap representation of the data in C, which displays signal as a log2 fold change (log2FC) in INTS11 depleted versus un-depleted conditions covering a region 3kb upstream and downstream of the TSS.

E. qRT-PCR analysis of INTS11-dTAG cells transfected with the HIV reporter construct with or without TAT then depleted or not of INTS11. Quantitation shows signals relative to those obtained in the presence of INTS11 and the absence of TAT after normalising to MALAT1 RNA. n=3. Error bars show standard deviation. ** denoted p=0.01. Note that INTS11 depletion was performed concurrently with transection (14hr in total).

F. Model for promoter directionality depicting higher levels of focused transcriptional initiation in the sense direction together with opposing gradients of CDK9 and INTS11 activity that peak in sense and antisense directions, respectively.

A. Western blot showing three separate dTAG-RBBP6 cell clones. In each case, homozygous tagging is demonstrated by the size shift versus endogenous RBBP6 (HCT116 lanes). Tagged RBBP6 is completely depleted by 2hr treatment with dTAGv-1 whereas endogenous RBBP6 is unaffected. SPT5 serves as a loading control. Clone number 2 (red asterisk) was selected for the experiments in Figure 1.

B. Graph plotting the number of AAUAAA sequences in antisense transcripts derived from a 3kb window upstream of TSSs. Y-axis is the number of transcripts, and the x-axis shows the AAUAAA count per transcript.

C. Heatmap of control or RBBP6-depleted POINT-seq showing the RBBP6 effect on antisense transcripts without an AAUAAA or those that contain at least one AAUAAA. Most are unaffected by RBBP6 loss. The Y-axis is RPKM.

A. Genome browser track showing POINT-seq signal over the first 7kb of GLCCI1 in INTS11-dTAG cells treated or not (1.5 hr) with dTAGv-1. This is an example of where INTS11 loss affects protein-coding transcription as previously described17. Y-axis shows RPKM following spike-in normalisation.

B. Metaplot of POINT-seq data from INTS11-dTAG cells treated or not (1.5hr) with dTAGv-1. These are protein-coding genes arranged head-to-head. Signals above and below the x-axis are sense and antisense reads, respectively. Y-axis shows RPKM following spike-in normalisation.

C. Metaplot of POINT-seq data from INTS11-dTAG cells treated or not (1.5hr) with dTAGv-1. This shows RNAs derived from enhancer clusters, which generally initiate transcription bidirectionally. Signals above and below the x-axis are sense and antisense reads, respectively. Y-axis shows RPKM following spike-in normalisation.

A. Plot showing the size distribution of fragments mapped by sPOINT-seq. This demonstrates its selectivity toward transcripts <150nts.

B. Violin plot of the sum sPOINT 5’end spike-in normalised RPKM signal across a 500bp window from the protein-coding TSSs shown in Figure 3D (n=3060). These samples are untreated with dTAG to quantitate the normal levels of sense vs. antisense initiation/pausing in a window -/+ 500bp from the TSS in respective directions for antisense and sense. This shows more initiation/pausing in sense vs. antisense directions.

C. Genome browser track of ACTB promoter region in sPOINT-seq from INTS11-dTAG cells treated or not with dTAGv-1 (1.5 hr). This showcases the focused sense TSS (black arrows) and the dispersed antisense reads (brackets). Note the different y-axis scales (RPKM) for sense vs. antisense.

D. Genome track of RNU5A-1 and RNU5B-1 in RNAPII ChIP-seq performed on INTS11-dTAG cells treated or not with dTAGv-1 (2hr). INTS11 depletion causes RNAPII build-up beyond both genes indicating a transcription termination defect. The Y-axis scale is RPKM.

E. Metaplot of RNAPII occupancy of protein-coding TSS regions (the same gene set employed for the sPOINT analysis in Figure 3) derived from RNAPII ChIP-seq performed on INTS11-dTAG cells treated or not with dTAGv-1 (2hr). The Y-axis scale is log10 qvalue of peak pileup values normalised to spike in control.

A. Genome browser track of TARDBP in POINT-seq data derived from INTS11-dTAG cells either untreated, dTAG treated, NVP-2 treated or dTAG and NVP-2 treated (2.5 hr). The Y-axis scale shows RPKM.

B. As for A, but for NEDD1.

C. qRT-PCR analysis of PGK1, ENY2, PLEKHF2, and NUDCD1 pre-mRNAs in INTS11-dTAG cells treated or not with dTAGv-1 and at the same time exposed or not to DRB (all 2.5 hr). To enrich nascent transcripts, primers detect intronic RNA. Quantitation shows fold change versus spliced actin relative to samples untreated with dTAGv-1 or DRB. DRB treatment substantially reduces signal, which is restored when INTS11 is co-depleted. Error bars show standard deviation. n=3, * denotes p≤0.05.

D. Analysis of published18 INTS11 and CDK9 ChIP-seq data showing their respective occupancy within 1kb of the TSS. Based on this data, they are slightly enriched upstream (INTS11) and downstream (CDK9) of the TSS. The Y-axis units are RPKM.