RNA polymerase II collision interrupts convergent transcription

Abstract

Antisense noncoding transcripts, genes-within-genes, and convergent gene pairs are prevalent among eukaryotes. The existence of such transcription units raises the question of what happens when RNA polymerase II (RNAPII) molecules collide head-to-head. Here we use a combination of biochemical and genetic approaches in yeast to show that polymerases transcribing opposite DNA strands cannot bypass each other. RNAPII stops but does not dissociate upon head-to-head collision in vitro, suggesting that opposing polymerases represent insurmountable obstacles for each other. Head-to-head collision in vivo also results in RNAPII stopping, and removal of collided RNAPII from the DNA template can be achieved via ubiquitylation-directed proteolysis. Indeed, in cells lacking efficient RNAPII polyubiquitylation, the half-life of collided polymerases increases, so that they can be detected between convergent genes. These results provide insight into fundamental mechanisms of gene traffic control and point to an unexplored effect of antisense transcription on gene regulation via polymerase collision.

Graphical abstract

Figure 1

RNAPII Collision Is a Block to Transcript Elongation In Vivo (A) Left: Nucleic acid-centric view of convergent transcription. Right: Protein-centric view of elongation, showing a crystallographic model of convergently transcribing RNAPII elongation complexes (Kornberg, 2007) (DNA in green/blue; RNA in red). (B) GAL10-GAL7 and GAL10-GAL7ΔTATA constructs, with short and long G-less cassettes indicated. (C) Autoradiograph of RNA isolated from GAL10-GAL7ΔTATA (“−,” convergent transcription, lanes 1 and 3) or GAL10-GAL7 (“+,” 2 and 4) incubated for 75 min in the presence of glucose (Glc, lanes 1 and 2) or galactose (Gal, lanes 3 and 4). Mean percent distal cassette transcribed is indicated below, after normalizing to proximal G-less cassette levels (distal/proximal set to 100 in GAL10-GAL7ΔTATA). The standard error calculated from two biological replicates is presented below. (D) Schematic showing experimental layout and results.

Figure 2

Reconstitution of Convergent Transcription In Vitro (A) Mono-EC (left) and di-EC (right) schematics showing EC constituents and predicted transcript lengths. ³²P end-labeled transcript is indicated by black sphere. (B) Upper left: Schematic showing the position of epitope-tagged RNAPII molecules in di-ECs. Lower left: experimental strategy. Right: Autoradiograph of transcripts from mono- (M) and di-ECs (D) after adding all four NTPs (lanes 1 and 2), or leaving out GTP (lanes 3 and 4), respectively. Transcript length indicated on right. (C) A time course of collision. Di-ECs were incubated with all NTPs for the times indicted. (D) Schematic showing ECs before (upper) and after (lower) transcription. RNAPII “footprint” is based on crystal structures (Gnatt et al., 2001; Kettenberger et al., 2004; Wang et al., 2006).

Figure 3

Elongation Complexes Remain on DNA Following Collision (A) Experimental schematic (left) and autoradiograph showing native agarose electrophoresis of mono- (M) and di-ECs (D), before (lanes 1 and 2) or after (lanes 3 and 4) transcription. (B) Autoradiograph of transcripts isolated from gel slices containing mono- (lane 3 in 3A) or di- (lane 4 in 3A) ECs, respectively, resolved by denaturing PAGE.

Figure 4

RNAPII Is Rapidly Removed from Convergent Genes in a Ubiquitylation-Dependent Manner Upper: Collision gene constructs with position of qPCR product indicated below (black line). Lower: RNAPII ChIP in WT and elc1Δ, in the absence (black) and presence of collision (white). ChIP values were divided by the input, telomere, and IgG control values. Density in the absence of collision was set to one, and the value in the presence of collision was expressed relative to that. Error bars show standard error (from three biological replicates).

Figure 5

Slow Clearance of RNAPII from the Deactivated Gene Due to Collision (A) RNAPII density (ChIP) in the endogenous GAL1 gene in WT and elc1Δ cells as genes are switched off (x axis shows time in minutes after glucose addition). Error bars show standard error (from three biological replicates). ChIP values were divided by the input and telomere signal. Values obtained at time = 0 in each case were set to 100, and other values are shown relative to that. (B) As in (A), but with the GAL10-GAL7 collision construct.

Figure 6

Evidence for Collision between Natural Convergent Genes (A) Average RNAPII density between the 1,478 convergent (left) and 1,492 divergent (right) genes in WT (blue lines) and elc1Δ (red lines), as measured by ChIP-Seq. Log2 ChIP signal divided by IgG control signal is shown relative to the midpoint (0) between the genes. Colored lines represent the average of two independent experiments. A schematic is shown above graphs, with arrows indicating the direction of transcription. Note that the distance from the end (convergent), or beginning (divergent), of an ORF to the midpoint obviously differs from gene to gene (see Figure S4). The compared gene pairs are listed in Table S1. (B) RNA levels, measured by nascent RNA-Seq, downstream of convergent genes. 0 indicates end of ORF. Lines represent the average of three independent experiments.