Transcription-replication conflict resolution by nuclear RNA interference

Abstract

Nuclear RNA interference (RNAi) is required for heterochromatin silencing, but Dicer also promotes genome stability by releasing RNA polymerase at sites of replication stress. R-loops are three-stranded DNA:RNA structures that accumulate at transcription-replication (T-R) collisions. We show that in RNase H-deficient cells, which accumulate pathological R-loops, Dcr1 processes R-loops at transcriptional start sites (TSSs) and end sites (TESs), releasing paused RNA polymerase and accounting for small RNAs (sRNAs) resembling DNA-damage-associated sense sRNAs (sdRNAs) found in cancer cells. Genetic evidence implicates nascent transcription-associated R-loops in genome instability in the absence of Dicer, with the helicase domain providing catalytic function reminiscent of related archaeal helicases involved in replication. The RNase H homolog Argonaute (Ago1) promotes genome instability by binding R-loops, and its removal relieves replication stress. Analysis of replication intermediates, DNA and RNA 3' ends, and fork processivity genome wide indicates Dicer resolves head-on T-R collisions, consistent with an ancient origin in DNA replication.

Keywords: Argonaute; Dicer; R-loops; RNA interference; RNA polymerase pausing; RNA-DNA hybrids; RNase H; fission yeast Schizosaccharomyces pombe; replication stress; transcription.

Figure 1.. Dicer prevents unscheduled release of paused RNA Pol II in the presence of R-loops, which results in replication stress

(A) Electrophoretic mobility shift assay (EMSA) of Dcr1 binding to R-loops and D:R hybrids in vitro. (B) Spot growth assays of WT and combinations of rnh1Δ, rnh201Δ, and dcr1Δ using 10-fold dilutions on YES plates without supplement or with indicated doses of the genotoxic agents, hydroxyurea (HU) and camptothecin (CPT). (C and D) Representative images of dcr1Δ (C) and dcr1Δ rnh1Δ rnh201Δ triple mutant (D) cells with DAPI staining of DNA. (E) Doubling time of WT and combinations of rnh1Δ, rnh201Δ, and dcr1Δ. n = 2; error bars represent standard deviation. (F) Heatmaps of R-loop levels of rnh1Δ rnh201Δ, dcr1Δ, and dcr1Δ rnh1Δ rnh201Δ relative to WT. dRNH1 ChIP-seq reads of individual genotypes were first normalized per million reads (RPM), followed by log₂ normalization to input. The track was then subtracted with WT. The heatmap shows all RNA Pol II-transcribed genes (n = 4,346), collapsing gene bodies and with 500 bp upstream of annotated transcription start sites (TSSs) and 500 bp downstream of transcription end sites (TESs). (G) Representative genome track of normalized PRO-seq data of WT (black), dcr1Δ (blue), rnh1Δ rnh201Δ (gray), and triple mutant (red). Note the peak at the TSS of the gene SPAC7D4.03c and the pile-up toward the end of the transcript. (H) Detection of paused RNA polymerase by 3^’-OH RNA ends in asynchronized WT (black), dcr1Δ (blue), rnh1Δ rnh201Δ (gray), and triple mutant (red) cells. PRO-seq reads were normalized per million reads (RPM) and mapped to genome-wide metaplots of collapsed gene bodies plus 500 bp upstream of annotated TSSs and 500 bp downstream of TESs for all RNA Pol II-transcribed genes (n = 4,346). (I) Density plot of transcriptional pausing indices (PIs) after normalization to WT, based on PRO-seq reads from (H). Dotted lines represent the median for each genotype. (J) Density plot of transcriptional termination indices (TIs) of well-isolated genes (see STAR Methods) after normalization to WT based on PRO-seq reads from (H). Dotted lines represent the median for each genotype. (K) Detection of initiating RNA Pol II (top) and elongating RNA Pol II (bottom) via phospho-serine-5 and phosphor-serine-2 ChIP-seq, respectively. Reads were mapped to all RNA Pol II-transcribed genes and normalized as in (H).

Figure 2.. GTFs mediate R-loop-dependent replication stress

(A) Spot growth assays of WT, dcr1Δ, rnh1Δ rnh201Δ, and the triple mutant dcr1Δ rnh1Δ rnh201Δ, with or without the mutant med20–1 allele, spotted on YES plates without supplement or with TBZ or HU. (B and C) Density plot of pausing indices (PIs) (B) and termination indices (TIs) (C) in dcr1Δ, med20–1, and dcr1Δ med20–1 after normalization to WT. Dotted lines represent the median for each genotype. (D) Detection of paused RNA polymerase by sequencing 3’-OH RNA ends using PRO-seq in asynchronized WT (black), dcr1Δ rnh1Δ rnh201Δ (red), med20–1 (orange), and med20–1 dcr1Δ rnh1Δ rnh201Δ (blue) cells. Reads were normalized and mapped as in Figure 1H. (E and F) Density plot of PIs (E) and TIs (F) in dcr1Δ rnh1Δ rnh201Δ, med20–1, and med20–1 dcr1Δ rnh1Δ rnh201Δ after normalization to WT. Dotted lines represent the median for each genotype. (G) Spot growth assay of WT, dcr1Δ, rnh1Δ rnh201Δ, and the triple mutant dcr1Δ rnh1Δ rnh201Δ, with or without the mutant rpb1-T481K allele, spotted on YES plates without supplement or with TBZ or HU. (H and I) Density plot of PIs (H) and TIs (I) in dcr1Δ, rpb1-T481K, and dcr1Δ rpb1-T481K after normalization to WT. Dotted lines represent the median for each genotype.

Figure 3.. The helicase domain of Dicer is required for pausing and genome stability

(A) Domain architecture of S. pombe Dcr1, highlighting the mutants used in this study. Not drawn to scale. (B) Spot growth assays of WT, dcr1-K38R, and dcr1-K38A on YES plates without supplement of with indicated concentrations of CPT and HU. Rad51Δ serves as a positive control. (C) qPCR quantification of relative rDNA copy number (WT F0 = 1) in WT, dcr1-K38R, and dcr1-K38A cells over 4 meiotic generations. (D) qPCR quantification of relative rDNA copy number (WT F0 = 1) in various dcr1 alleles in F0 and F3 generation. (E) PRO-seq detection of paused RNA polymerase by sequencing 3^’-OH RNA ends in asynchronized WT, dcr1Δ, dcr1-K38R, and dcr1-K38A cells. Reads were mapped and normalized as in Figure 1F. (F and G) Density plot of PIs (F) and TIIs (G) of all annotated genes in dcr1Δ, dcr1-K38R, and dcr1-K38A after normalization to WT based on PRO-seq data in (E). Dotted lines represent the median for each genotype.

Figure 4.. Dicer generates sRNAs resembling damage-associated sRNA (sdRNA) from R-loops that mediate replication stress via Ago1

(A) Detection of sRNA (top), chromosome-bound Ago1 (middle), and Ago1-associated sRNA (bottom) in WT (black), dcr1Δ (blue), rnh1Δ rnh201Δ (gray), and the triple mutant (red). sRNA, Ago1 ChIP, and Ago1 RIP sequencing reads were mapped to all annotated genes and normalized as in Figure 1F. (B) Size distribution of total sRNA (left) and Ago1-associated sRNA (right) in WT (black), dcr1Δ (blue), rnh1Δ rnh201Δ (gray), and the triple mutant (red). Sizes are indicated in nucleotides (nt). (C) Spot growth assays of WT, dcr1Δ, and ago1Δ cells with 10-fold dilution on YES plates without supplement or with indicated doses of TBZ, HU, or CPT. (D) Spot growth assays of WT and combinations of rnh1Δ, rnh201Δ, and ago1Δ cells with 10-fold dilution on YES plates without supplement or with indicated doses of TBZ, HU, and CPT. (E) Spot growth assays of WT and combinations of rnh1Δ, rnh201Δ, ago1Δ, and dcr1Δ cells with 10-fold dilution on YES plates without supplement or with indicated doses of TBZ, HU, and CPT. (F) Spot growth assays of WT and combinations of rnh1Δ, rnh201Δ, ago1-D580A, and dcr1Δ cells with 10-fold dilution on YES plates without supplement or with indicated doses of TBZ, HU, and CPT.

Figure 5.. Dicer promotes faithful replication of highly transcribed genes near programmed replication fork blocks

(A) Diagrams of constructs containing the reporter allele ura4-sd20 (red rectangle), associated with the directional replication fork block RTS1-RFB from the rDNA locus on chromosome III. The RFB is integrated 5 kb away from a strong replication origin (ori). After thiamine removal, forks traveling from the centromere toward the telomere are blocked in a polar manner. The ura4-sd20 allele contains a 20 nt duplication flanked by micro-homology. When the ura4-sd20 allele is replicated by a restarted fork, the non-processive DNA synthesis undergoes RS, resulting in the deletion of the duplication and the restoration of a functional ura4⁺ gene. (B) Frequency of RS in indicated strains and constructs. Bars indicate mean values ± 95% confidence interval. Statistical analysis was performed using Student’s t test, compared with WT. n = 10–32. (C) 2D gel electrophoresis (2DGE) analysis of replication intermediates (RIs) in asynchronous WT and dcr1Δ cells using ura4 as a probe. Representative gels are shown for thiamine-treated (RFB off) and untreated (RFB ON) cells undergoing slippage upstream of the RFB (A). The red arrow indicates the “tail” signal, which represents lagging-strand resection (D). Numbers indicate the efficiency of the RFB ± standard deviation (SD). (D) Schematics of RIs observed within the AseI restriction fragment in RFB ON condition. Gray lines indicate secondary signals caused by partial digestion of psoralen-crosslinked RIs. See (G) for annotation. (E) Tail quantification of resected forks from (C). Bars indicate mean values ± SEM. Statistical analysis was performed using Student’s t test, compared with WT. n = 2. (F) Schematic of an rDNA locus on chromosome III. Probes and restriction sites are indicated. H, HindIII; B, BamHI; K, KpnI; ars, autonomously replicating sequence; ETS, external transcribed spacer; RFB, programmed and polar RFB. The orange arrows indicate the transcription unit of rRNA. (G) Illustration representing the expected migration behavior of RI analyzed by 2DGE. The “Y arc” is a series of Y-shaped RI with progressively longer branches, resulting from replication fork progression within the DNA fragment analyzed. The “bubble arc” corresponds to the firing of the replication origin. The vertical X-spike results from X-shaped DNA joint molecules corresponding to RIs undergoing HR in dcr1Δ cells. (H) Representative images of RI analysis by 2DGE within the origin-containing HindIII-KpnI restriction fragment (top) and the RFB-containing BamHI-BamHI restriction fragment (bottom) in WT and dcr1Δ cells. Red arrows indicate fork pausing signals from T-R collisions on both sides of the origin in dcr1Δ cells, while black arrows indicate the position of the programmed RFBs in WT cells, respectively. *p < 0.05, ***p < 0.0005; NS, not significant.

Figure 6.. Dicer rescues stalled forks at R-loop-dependent T-R collisions

(A) Detection of DNA damage, free 3^’-OH ssDNA ends, and free 3^’-OH RNA ends (paused RNA polymerase) in asynchronized WT (black), dcr1Δ (blue), rnh1Δ rnh201Δ (gray), and triple mutant (red) cells. From top to bottom: (upper) DNA damage detected by ChIP-seq of γH2A.X normalized to histone H2A, (middle) free 3^’-OH single-strand DNA ends were detected by GLOE-seq from transcriptional sense and antisense strands, and (lower) 3^’-OH RNA ends were detected by PRO-seq at transcriptional pause sites. GLOE-seq and PRO-seq read counts were mapped and normalized as in Figure 1F. (B) T-R collision at TSS, replication fork reversal at transcriptional pause sites, and rescue by converging forks are consistent with free ssDNA ends (sense peaks in A) arresting downstream of the pause site and leading strand DNA ends from convergent forks arresting at TSS in the triple mutant. Free ssDNA ends (antisense peaks in A) at the pause site in WT and at the TES in dcr1Δ and triple mutant cells correspond to leading strand ends arrested at T-R collisions. (C) Replication fork directionality (RFD) analysis of GLOE-seq data from (A). RFD is defined as the ratio of excess reverse (Crick strand, REV) reads within a region relative to forward (Watson strand, FWD) reads, which is calculated as (REV – FWD)/(REV + FWD). Replication origins were detected by Orc4 ChIP-seq (lower track) and correspond to leading to lagging-strand transitions. (D) Violin plot of the distribution of RFD at all annotated origins (pombase/oriDB), revealing significant replication fork asymmetry genome wide in dcr1Δ and triple mutant cells. p values represent results of one-way ANOVA. ***p < 0.0005; NS, not significant.

Figure 7.. Dicer rescues head-on, but not co-directional, T-R collisions by processing R-loops

(A) Schematic of pulse-chase labeling for replication fork detection by nanopore sequencing. (B) Example of individual nanopore sequencing reads from pulse-labeled WT (FY2317) cells, showing BrdU content after processing using NanoForkSpeed. Diverging forks (replication bubbles) appear as clusters of labeled nucleotides, with fork speed calculated by the slopes of relative signal density at the boundaries. Plots were generated using available dedicated software. (C) Violin plot of the global estimated replication speeds of forks detected in WT and mutant cells. p values represent results of one-way ANOVA. Fork progression is significantly slower in dcr1Δ and triple mutant cells. (D) Violin plot showing the estimated replication speeds for all replication forks mapped to the rDNA loci, separated according to head-on (Head) or co-directional (Co) with rRNA transcription. p values represent results of one-way ANOVA. (E) Violin plot of the estimated replication speed distributions for all replication forks near annotated mRNA transcripts, separated according to head-on or co-transcription to the direction of transcription direction, and by genotype. p values represent results of one-way ANOVA. (F) Model of promoter-proximal cleavage of R-loops by Dicer during T-R collisions. Most of the proteins indicated are required for silencing and/or negatively interact with Dicer during replication stress. *p < 0.05, ***p < 0.0005; NS, not significant.