Various proteins self-organize via liquid–liquid phase separation (LLPS) into nucleolar subdomains, which are needed for highly stereotyped ribosome assembly1,2. At fibrillar centres in the heart of mammalian nucleoli, the major rRNA molecules needed to assemble ribosomes are generated by Pol-I-dependent transcription of rRNA genes within ribosomal DNA (rDNA) repeats1,3. Within rDNA, rRNA genes are separated by large intergenic spacers (IGSs) (Extended Data Fig. 1a). At nucleolar rRNA genes, Pol I synthesizes precursor rRNAs (pre-rRNAs) that are processed into mature 28S, 18S and 5.8S rRNA molecules as they migrate to the granular component at the nucleolar periphery. Outside nucleoli, Pol III synthesizes 5S rRNA molecules that are targeted to nucleoli for processing. Mature rRNAs are packaged into 40S and 60S ribosomal subunits for export to the cytoplasm. Traditionally, the nucleolar Pol I and nucleoplasmic Pol III are viewed as the sole mammalian RNA polymerases that directly mediate housekeeping ribosome biogenesis. Interestingly, in the budding yeast Saccharomyces cerevisiae, Pol II is physically enriched at rDNA IGSs, but this phenomenon is deleterious because it drives ageing without affecting rRNA expression3,4,5. It is unclear whether nucleolar Pol II exists in higher organisms or directly promotes ribosome biogenesis in any species.

Active Pol II at rDNA IGSs

To investigate whether Pol II exists within human nucleoli, we first used immunofluorescence coupled to super-resolution microscopy. Within nucleoli, which were outlined by nucleophosmin (NPM), we observed foci corresponding to active Pol II phosphorylated on serine 2 (pS2) (Fig. 1a and Extended Data Fig. 1b, c). Chromatin immunoprecipitation (ChIP) showed that pS2 and another active form of Pol II, phosphorylated on serine 5 (pS5), were enriched across rDNA, with the highest levels—at IGS28 and IGS38—being comparable to those at known Pol-II-transcribed loci (Fig. 1b and Extended Data Fig. 1a, d–f). The Pol II activator cyclin-dependent kinase 9 (CDK9) was similarly enriched across IGSs (Extended Data Fig. 1g). pS2 and CDK9 were also enriched across the IGSs of IMR90 fibroblasts, indicating that enrichments are not limited to tumorigenic cells (Extended Data Fig. 1h, i). Unlike Pol II and CDK9, Pol I and its initiation factor, upstream binding factor 1 (UBF, also known as UBF1), localized primarily to rRNA genes, although low Pol I levels existed across IGSs (Extended Data Fig. 1j, k). Notably, Pol II was overrepresented relative to Pol I only within IGSs (Extended Data Fig. 1l). These data suggest that rDNA loci are cohabited by Pol I and Pol II.

Fig. 1: Pol I and Pol II localize to rDNA IGSs and compete to modulate IGS ncRNA levels.
figure 1

a, Representative immunofluorescence and super-resolution microscopy images showing the localization of pS2 Pol II within NPM-delineated nucleoli. Scale bar, 5 μm. b, Enrichment of pS2 Pol II across rDNA as revealed by chromatin immunoprecipitation (ChIP). Enrichments, accounting for typical background fluctuations across repetitive DNA loci, were calculated as (percentage of input/IgG) = (percentage of input for protein immunoprecipitation)/(percentage of input for mock IgG immunoprecipitation). c, Effect of a 3-hour Pol II inhibition (iPol II) using flavopiridol (FP) or α-amanitin (AMN) on rRNA biogenesis as measured in live single-cell pulse-chase assays using 5-fluorouracil (FU)-labelled RNA. d, e, Cell-population-based RNA pulse-chase assays were used to assess pre-rRNA synthesis (d) and processing (e) following a 3-hour inhibition of Pol I or Pol II (iPol I/II; low-dose actinomycin-D, LAD). f, Pol I promotes, and Pol II represses, IGS ncRNAs, as shown by reverse transcription with quantitative polymerase chain reaction (RT–qPCR). af, Experiments carried out with HEK293T cells; data shown as means ± s.d.; data in b and Extended Data Fig. 1d–f, j–l were from large experimental sets sharing immunoglobulin G (IgG) controls; n = 3 biologically independent experiments (bf); two-tailed t-test (b); one-way analysis of variance (ANOVA) with Dunnett’s multiple comparisons test (ce); image in a is representative of two independent experiments.

Source data

To determine whether rRNA biogenesis is rapidly affected following Pol II perturbation, we conducted a three-hour treatment using the Pol II inhibitors α-amanitin (AMN) or flavopiridol in pulse-chase experiments. Pol II inhibition perturbed global ribosome biogenesis (Fig. 1c). Specifically, unlike Pol I inhibition by low-dose actinomycin-D (LAD), Pol II inhibition almost fully abolished pre-rRNA processing (Fig. 1d, e and Extended Data Fig. 2a–c), indicating a distinct mechanism of ribosome biogenesis arrest. Cell viability and global protein levels were unchanged following Pol II inhibition, arguing against indirect effects (Extended Data Fig. 2d, e). Furthermore, a 30-min Pol II inhibition was sufficient to strongly disrupt rRNA processing, suggesting a direct function for Pol II through its enrichment at rDNA (Extended Data Fig. 2f; Pol II inhibition hereafter was for three hours unless otherwise indicated). These data suggest that Pol II might directly support nucleolar rRNA expression through its association with IGSs.

In different cell types, we detected IGS noncoding RNAs (ncRNAs) that decreased in abundance following Pol I inhibition (Fig. 1f and Extended Data Fig. 2g, h). Strikingly, IGS ncRNAs were markedly induced and found to be de novo transcribed upon Pol II inhibition (Fig. 1f and Extended Data Fig. 2h, i). Simultaneous inhibition of Pol I abolished the induction of IGS ncRNAs by Pol II inhibition (Extended Data Fig. 2j, k). Thus, Pol II counters Pol-I-dependent synthesis of IGS ncRNAs. Strand-specific transcript analysis of IGSs identified sense intergenic ncRNAs (sincRNAs) and antisense intergenic ncRNAs (asincRNAs) that were transcribed by Pol I and Pol II, respectively (Extended Data Fig. 2l–n). The sincRNA/asincRNA ratio paralleled Pol I/Pol II enrichment across IGSs (Extended Data Fig. 2m, o). The data so far indicate that Pol II operates directly across the IGSs, where it generates asincRNAs and limits the spurious synthesis of sincRNAs by Pol I.

Pol II maintains nucleoli via sincRNA control

Given that nucleolar organization is essential for rRNA synthesis and processing, we characterized disordered proteins at the nucleolar subdomains that are essential for these functions (Extended Data Fig. 3a, b). NPM delineates the granular component of the nucleolus, the LLPS of which is required for rRNA processing2,6,7. Pol II inhibition abrogated the phase separation of NPM, which was quickly reorganized into ruffled bodies before undergoing complete mixing with the nucleoplasm (Fig. 2a and Extended Data Fig. 3c–e). At nucleolar fibrillar centres, UBF (which is enriched at the promoters of rRNA genes) forms small foci6. Pol II inhibition resulted in UBF relocation to the nucleolar periphery, where UBF formed large spheres, rings or crescent-shaped bodies exhibiting wetting behaviour (Fig. 2b and Extended Data Fig. 3f–h). Changes in NPM and UBF coincided with global nucleolar disorganization (Extended Data Fig. 3i) and matched sincRNA induction kinetics (Fig. 1f). UBF bodies generated upon Pol II inhibition exhibited greater fluorescence recovery after photobleaching (FRAP; Extended Data Fig. 3j)6, suggesting decreased UBF–rDNA interactions or rDNA relocation to less viscous environments. Notably, the former nucleolar space that became surrounded with UBF signals following Pol II inhibition showed positive staining with Congo red (Extended Data Fig. 3k), indicating the presence of stress-induced, solid-like nucleolar amyloid bodies8,9. The data suggest that Pol II inhibition partly and strongly disrupts the organization of rRNA synthesis and processing sites, respectively. Under these conditions, aberrant liquid-to-solid phase transitions occur within the remnant nucleolar space.

Fig. 2: Pol II represses sincRNAs to maintain nucleolar structure and function.
figure 2

a, b, Effects of a 3-hour Pol II inhibition on NPM (a) and UBF (b) localization, as shown by immunofluorescence microscopy. Examples of normal and defective phenotypes are respectively marked by magenta and white arrowheads. DAPI, 4′,6-diamidino-2-phenylindole. c, Low-complexity sincRNA, not high-complexity control RNA, promoted the formation of liquid droplets in the presence of amyloid-converting motif (ACM) peptides in vitro. Shown on the images is the concentration of ACM peptides incubated with 1 μM of the indicated RNA. df, In cells subjected to Pol II inhibition (FP), nucleolar organization was restored by coinhibition of Pol I (LAD; d), removal of FP (wash; d), or treatment with sincRNA-repressing ASOs (e), which also restored rRNA biogenesis as indicated by live single-cell FU-RNA pulse-chase assays (f). Percentages indicating phenotypic rescue relative to FP-treated cells are shown on graphs as applicable. af, Experiments with HEK293T cells; data are shown as means ± s.d.; one-way ANOVA with Dunnett’s multiple comparisons test (d, f); two-tailed t-test (e); n = 5 biologically independent experiments (d); n = 3 biologically independent experiments (e, f); images in ac are representative of two independent experiments; scale bars, 5 μm (yellow) and 1 μm (white).

Source data

Nucleolar amyloid bodies usually emerge following environmental stresses such as heat shock6,8,9. Specifically, heat shock causes proteins with the amyloid-converting motif (ACM) to form nucleolar liquid droplets, which undergo phase transition into solid-like amyloid bodies (Extended Data Fig. 4a, b). Knockdown of different sincRNAs prevented heat-shock-induced formation of ACM-containing nucleolar liquid droplets in vivo (Extended Data Fig. 4c). In a cell-free in vitro system, incubating ACM peptides with a sincRNA segment induced liquid droplet formation (Fig. 2c and Extended Data Fig. 4d)9. Moreover, strand-specific RNA sequencing (ss-RNA-seq) revealed that heat shock induced sincRNA and repressed asincRNA levels at IGSs (Extended Data Fig. 4e). Thus, environmental stress represses asincRNA levels and promotes sincRNA-dependent nucleolar remodelling. The results also show that sincRNAs induce liquid droplets in vitro and promote liquid droplets and consequent solid-like amyloid bodies in vivo.

Next, we assessed whether sincRNA repression restores nucleolar organization and function in live cells subjected to Pol II inhibition. Nucleolar organization was restored after Pol II inhibitor wash-off, Pol I co-inhibition, or direct repression of sincRNA levels with antisense oligonucleotides (ASOs) (Fig. 2d, e and Extended Data Fig. 5a). ASOs also partly restored rRNA biogenesis (Fig. 2f). An overexpressed sincRNA localized to nucleoli without decreasing rRNA biogenesis (Extended Data Fig. 5b–d), indicating that nucleolar disruption may depend on specific combinations of sincRNAs or that endogenous sincRNAs have distinctive modifications or interactors. However, cell types with naturally elevated sincRNA levels exhibited more NPM-marked nucleoli (Extended Data Figs. 2g, 5e). Of note, long-term Pol II inhibition may compromise nucleoli indirectly, by limiting the ability of Pol II to synthesize the U8 small nucleolar RNA (snoRNA) or Alu RNA molecules10,11,12. However, following our short-term Pol II inhibition, nucleolar disruption coincided with sincRNA induction in the absence of changes in U8 or Alu levels (Extended Data Fig. 6a, b). Additionally, in contrast with Pol II inhibitors, pharmacological agents2,12 disrupting nucleolar organization or global protein translation failed to induce sincRNA levels (Extended Data Fig. 6c–e). Thus, sincRNA accumulation drives nucleolar disorganization, and not vice versa. Together, these results show that Pol II constitutively represses different Pol-I-dependent sincRNAs to prevent unscheduled stress-mimicking nucleolar phase transitions, and to maintain endogenous nucleolar condensates that are essential for rRNA biogenesis.

Pol II sets an R-loop shield for Pol I

Nucleoli are naturally enriched in R-loops, which are triplex nucleic acid structures harbouring a DNA–RNA hybrid and single-stranded DNA13. Therefore, we postulated that baseline R-loop levels across IGSs may have beneficial effects through the modulation of Pol I–Pol II crosstalk. DNA–RNA hybrid immunofluorescence (DRIF) revealed nucleolar R-loops that were partly repressed by Pol II inhibition (Fig. 3a) or the recombinant DNA–RNA hybrid repressor RNase H1 (Extended Data Fig. 7a–c). DNA–RNA hybrid immunoprecipitation (DRIP) revealed that several IGS sites exhibited R-loop signals, which peaked at the junctions between rRNA genes and IGSs and were sensitive to RNase H1 (Fig. 3b and Extended Data Fig. 7d)14. Despite markedly higher transcription of rRNA genes relative to IGSs, negative GC skews may be one of several different factors favouring antisense IGS R-loops (Extended Data Fig. 7e)15. Notably, R-loop repression by RNase H1 overexpression partly mimicked Pol II inhibition, increasing sincRNA expression at most IGS sites tested (Extended Data Fig. 7f, g). Together, these findings suggest that R-loops are important molecular mediators of sincRNA repression by Pol II.

Fig. 3: Repression of an IGS R-loop shield disrupts nucleoli.
figure 3

a, Pol II inhibition repressed nucleolar R-loops. b, DRIP analysis shows RNase H1-sensitive R-loop peaks at rDNA. c, The RED–LasRR system created to achieve inducible locus-associated R-loop repression. d, The short guide RNA for IGS28 (sgIGS28) enriched the tetracycline (Tet+)-induced RED or dRED at IGS28 in anti-GFP ChIP, using IgG as control. Enrichments are normalized to a non-targeting control (sgNT). RED and dRED data were from different experiments but are shown on the same graph as a space-saving measure. e, Using RED or dRED together with sgIGS28 respectively decreased or increased R-loop levels at IGS18. f, g, RED sgIGS28 induced ncRNA levels (f) and disrupted NPM localization (g). The percentages of cells exhibiting ruffled NPM localization are indicated on the images (g). ag, HEK293T cells; data are shown as means ± s.d.; two-tailed Mann–Whitney U-test, n = 100 cells (a); or two-tailed t-test, n = 3 biologically independent experiments (b, df); scale bar, 5 μm. Percentage changes relative to respective sgNT samples are indicated above or on bars (e, f).

Source data

RNase H1 overexpression remains the gold-standard method by which to interrogate R-loop function13. However, with this approach, RNase H1 is often not enriched at the studied loci, where the observed phenotypic changes may also be due to R-loop repression elsewhere. To specifically interrogate the function of IGS-associated R-loops, we created a tetracycline-inducible RNase H1–eGFP–dCas9 (RED) fusion protein to achieve locus-associated R-loop repression (a process that we abbreviate as ‘RED–LasRR’, or ‘red laser’; Fig. 3c and Extended Data Fig. 7h; eGFP is enhanced green fluorescent protein). As a control, this system uses a similar chimaeric protein that comprises catalytically dead RNase H1 (denoted dRED).

Similar to the RNase H1 protein16, RED and dRED displayed nucleolar and nucleoplasmic localization in the absence of short guide RNAs (sgRNAs) (Extended Data Fig. 7i, j). Within the IGS, constitutive chromatin looping juxtaposes the IGS27/28 sites with IGS16/18 sites17,18. Therefore, we investigated whether a pool of three sgRNAs targeting IGS28 (sgIGS28) can enrich RED at IGS28 and repress the strong R-loop peaks at IGS16/18. ChIP confirmed successful targeting and similar enrichment of RED and dRED at the IGS28 site upon coexpression of sgIGS28 (Fig. 3d). Targeting RED, but not dRED, to IGS28 repressed only the strong R-loop peak at IGS18, while inducing a subset of sincRNAs across the IGSs (Fig. 3e–f and Extended Data Fig. 7k). Using RED with sgIGS38, which is spatially distal to the IGS18 site17,18, failed to alter R-loop or ncRNA levels at IGS18 (Extended Data Fig. 7l, m). Targeting dRED to IGS28 stabilized R-loops without decreasing sincRNA levels at IGS18, suggesting that maximal function of IGS18 R-loops is already achieved endogenously (Fig. 3e, f). Of note, ncRNA levels were similarly decreased at the IGS28 site to which RED or dRED was targeted without affecting Pol II enrichments (Fig. 3f and Extended Data Fig. 7n), and the RED–LasRR system can be used to target the fusion proteins to a single-copy locus outside of rDNA (Extended Data Fig. 7o). Using the guide RNAs (gRNAs) targeting RED to IGS28, individually, failed to achieve R-loop repression at the IGSs (Extended Data Fig. 7p). This argues against the possibility that targeting of RED to non-rDNA sites via any single gRNA or the RNase H1 moiety of the fusion protein indirectly represses IGS R-loops. Although the RED/sgIGS28-dependent lowering of R-loops only partially induces sincRNAs, this still mimicked early Pol II inhibition, as shown by the perturbation of NPM architecture into indistinct, ruffled bodies (Fig. 3f, g). This highlights the disruptive impact that even small increases in sincRNA levels can exert on nucleoli. The data show that asincRNAs generated by Pol II form an antisense R-loop shield that limits the synthesis of Pol-I-dependent sincRNAs, which can abrogate nucleolar organization and function. The RED–LasRR system will support studies on the numerous roles of R-loops in genome expression and stability.

Senataxin supports the R-loop shield

We next set out to identify additional factors that may regulate nucleolar Pol II. Senataxin (SETX) is a human neurodegeneration-linked helicase19. SETX and its yeast orthologue Sen1 have several transcription-modulatory roles, including Pol II loading and R-loop repression20,21,22. Sen1 associates with rDNA IGSs to promote Pol I transcription termination and to silence lifespan-shortening IGS ncRNAs16,23,24. We found that SETX was enriched across human IGSs, especially at IGS28, and exhibited nucleolar localization (Extended Data Fig. 8a, b). The IGS28 SETX peak overlapped one Pol II peak and the intergenic promoter marks H3K27ac, H3K9ac and H3K4me3 in ENCODE ChIP–seq data (Extended Data Fig. 8c). Sequential ChIP revealed that SETX was preferentially coenriched with Pol II compared with Pol I at IGS28 (Fig. 4a). Thus, SETX is coenriched with Pol II at IGSs, especially at a putative intergenic promoter at IGS28.

Fig. 4: Nucleolar Pol II reinforcement by SETX and nucleolus-disrupting sincRNAs in cancer.
figure 4

a, Sequential immunoprecipitations (IPs) revealed preferential coenrichment of SETX with Pol II at IGSs. Signals for Pol I/SETX and Pol II/SETX immunoprecipitations are normalized to signals from Pol I/IgG and Pol II/IgG, respectively. b, c, SETX knockout in two clones decreased R-loops (b) and induced IGS ncRNAs (c). d, e, Single-cell analysis of SETX-knockout cells showed that ASO-mediated repression of Pol-I-dependent sincRNAs partly rescues nucleolar organization (d) and rRNA biogenesis (e). Percentages indicating the magnitude of ASO-mediated phenotypic rescue are shown above graph bars where applicable. f, The patient-derived Ewing sarcoma cell line EWS502 and U2OS osteosarcoma cells with siRNA-mediated depletion of EWS breakpoint region 1 (EWSR1) showed disrupted nucleoli by electron microscopy. g, RNA-seq data indicate increased ncRNA levels at the IGSs of EWS502 and TC32 cells, as compared with IMR90 control cells. kbp, kilobasepairs. h, i, Single-cell analysis showed that sincRNA knockdown partly restores nucleolar organization (h) and rRNA biogenesis (i) in EWS502 cells. j, Model showing how a Pol-II-dependent R-loop shield limits Pol-I-dependent sincRNAs, which compromise nucleolar organization and function. ACM, amyloid-converting motif; NE, nuclear envelope. ai, Cells were HEK293T (ae) or as indicated (fi); data are shown as means ± s.d.; one-way ANOVA with Dunnett’s multiple comparisons test (a, d, e, h, i) and one-way ANOVA with Tukey’s multiple comparisons test (c); n = 4 biologically independent experiments (a), n = 2 biologically independent experiments (b, duplicates for each of wild-type, knockout 1 and knockout 2), n = 6 biologically independent experiments (c, triplicates for each knockout), n = 3 biologically independent experiments (d, e, h, i); images in f, g are representatives of two independent experiments; scale bar, 1 μm.

Source data

Notably, SETX knockout decreased the intergenic enrichment of Pol II and its R-loops (Fig. 4b and Extended Data Fig. 8d, e). This change was accompanied by increased intergenic Pol I enrichment (Extended Data Fig. 8f), elevated sincRNA synthesis (Fig. 4c and Extended Data Fig. 8g), and decreased Pol I localization at rRNA genes (Extended Data Fig. 8h). Unlike SETX knockout, the forced release of Pol I from rRNA gene promoters through knockdown of transcription initiation factor 1A (TIF1A) decreased pre-rRNA levels without inducing sincRNA levels (Extended Data Fig. 8i, j). This suggests that SETX loss prevents Pol II from shielding the IGSs from de novo Pol I loading. In addition, northern blotting did not show increases in pre-rRNA length upon Pol II or SETX disruption, arguing against rRNA gene read-through as the basis for increased IGS transcription by Pol I (Extended Data Fig. 8k). Thus, IGS R-loops act more as a shield that prevents Pol I recruitment, rather than a barrier that limits read-through transcription. Increases in sincRNA levels in SETX-knockout cells were associated with nucleolar disorganization and pre-rRNA processing defects, which were partly countered by sincRNA knockdown (Fig. 4d, e and Extended Data Fig. 8l–n). That SETX loss partly mimicked Pol II inhibition probably reflects the partial coenrichment of SETX and Pol II at IGSs. Additionally, SETX knockout did not lower IGS epigenetic silencing marks (Extended Data Fig. 8o), suggesting that SETX loss does not promote sincRNA levels by abrogating epigenetic silencing. In fact, SETX knockout slightly increased silencing marks, possibly reflecting epigenetic compensation constraining the magnitude of sincRNA induction. The data indicate that SETX is coenriched with IGS Pol II and supports it in repressing a subset of Pol-I-dependent sincRNAs that can disrupt nucleolar organization and function. SETX may achieve this effect by promoting the efficient loading and release of Pol II at an IGS28 intergenic promoter.

sincRNAs can disrupt nucleoli in cancer

We then aimed to identify a setting in which naturally elevated sincRNA levels may compromise nucleolar structure and function. Nucleolar organization, which is intimately related to cellular growth and viability, may be an adjunct in the diagnosis and treatment of some cancers25. In fact, nucleolar disruption upon Pol II dysregulation is similar to the constitutive disorganization of nucleoli in human Ewing sarcoma (EWS) tumours, related patient-derived EWS502 or TC32 cells, and U2OS osteosarcoma cells with depletion of EWS breakpoint region 1 (EWSR1) (Fig. 4f and Extended Data Fig. 9a, b)26. To determine whether alterations in sincRNA levels could underlie this phenotype, we reanalysed RNA-seq and DRIP–seq data from EWS and healthy IMR90 control cells to include rDNA17,26. EWS cells exhibited increased ncRNA and R-loop levels across IGSs (Fig. 4g and Extended Data Fig. 9c–e). Strikingly, in EWS cells, nucleolar disorganization and an rRNA biogenesis defect were countered by sincRNA knockdown (Fig. 4h, i and Extended Data Fig. 9f). These findings suggest that natural increases in sincRNA levels can explain aberrant nucleolar morphologies that are commonly observed in cancer25. R-loop increases in this setting may reflect selection for cells that have compensated for the increased sincRNA levels.


Our findings indicate that, in mammalian cells, antisense transcription by nucleolar Pol II generates an R-loop shield at rDNA IGSs to block Pol-I-dependent sense intergenic transcripts, which can compromise nucleolar condensates underlying rRNA expression (Fig. 4j and Extended Data Fig. 10). Processes that restrain R-loops at human IGSs probably exist, as unrestricted IGS R-loops destabilize yeast rDNA4,16,27. However, our findings differ from those in yeast, where IGS transcription does not regulate rRNA28,29 and Sen1 limits deleterious IGS ncRNAs by enforcing epigenetic silencing and transcript turnover21,23. At the IGSs of human cells under stress8, protective sense RNAs are likely to be induced through local repression of antisense RNA and R-loops. Nucleolar Pol II at IGSs may also mediate crosstalk with cellular differentiation, which is partly driven by promoter-associated transcripts that are dependent on Pol I or Pol II30,31,32,33. Future work should explore the potential use of sincRNAs and nucleolar disorganization as cancer biomarkers, and whether tumours exhibiting such features are hypersensitive to Pol-II-inhibiting drugs34. Overall, we identify nucleolar Pol II as a new master regulator of ribosome biogenesis, with broad implications for health and disease.


No statistical methods were used to predetermine sample size. The experiments were not randomized. The investigators were not blinded to allocation during experiments and outcome assessment, except for the quantification of microscopy images.

Cell culture and general materials

Human HEK293T, HeLa, HAP1 and osteosarcoma (U2OS) cells were cultured in Dulbecco’s modified Eagle medium (DMEM, Wisent Bioproducts) with 10% fetal bovine serum (FBS, Wisent). HEK293T T-REx cells (ThermoFisher Scientific) were cultured in DMEM supplemented with 10% tetracycline-free FBS and 1% penicillin/streptomycin. EWS502 and IMR90 cells were cultured in Roswell Park Memorial Institute (RPMI) medium supplemented with 10% FBS. All cell lines were cultured in the presence of 1% (v/v) penicillin/streptomycin (Wisent) at 37 °C in a humidified atmosphere with 5% CO2. Transfection of cultured cells was achieved using Lipofectamine3000 (Invitrogen, catalogue number L3000008), Lipofectamine RNAiMAX (Invitrogen, catalogue number 13778150) and Polyjet DNA transfection reagent (SignaGen Laboratories, SL100688). For transfections with plasmids encoding GFP–UBF1 or RNaseH1, 70% confluent cells were transfected with 1–3 μg of plasmid per well of a six-well plate; pcDNA3 served as control for RNaseH1 overexpression. For Pol II inhibition, cells were treated either with the reversible inhibitor flavopiridol (2 μM, inhibits Pol II pS2; Santa Cruz catalogue number sc-202157) or with the irreversible inhibitor α-amanitin (AMN, 50 μg ml−1, inhibits translocation; Abcam catalogue number ab144512). Other drug treatments were LAD (50 ng ml−1), MG132 (10 μM), doxorubicin (Dox, 300 nM), camptothecin (CPT, 10 μM), cycloheximide (CHX, 100 μM) or 1,6-hexanediol (HEX, 0.1% v/v). Antibodies, primers, guide RNAs and northern probes are listed in the Supplementary Information (Supplementary Tables 14). For Ewing sarcoma analyses, the Ewing sarcoma cell line TC32 was procured from the Children’s Oncology group ( and EWS502 was a kind gift from S. Lessnick (Nationwide Children’s Hospital, OH). Both cell lines were grown in RPMI (Corning). The control cell lines IMR90 (a primary fibroblast cell line) and U2OS (a human osteosarcoma cell line) were purchased from the American Type Culture Collection (ATCC) and grown in DMEM (Corning). Media were supplemented with 10% heat-inactivated FBS (Atlanta Biologicals). Cells were maintained at 37 °C in a humidified atmosphere with 5% CO2, confirmed using short tandem repeat (STR) profiling and tested for mycoplasma contamination. All siRNA transfections were conducted using Lipofectamine RNAimax (ThermoFisher Scientific) according to the manufacturer’s protocols.

Chromatin immunoprecipitation

Cells were grown to 80% confluence in 15 cm plates and crosslinked by adding 1% (v/v) formaldehyde at room temperature for 10 min. The reaction was quenched with 125 mM glycine for 5 min at room temperature. Cells were washed twice with cold phosphate-buffered saline (PBS), lysed with 10 ml lysis buffer (5 mM PIPES, 85 mM KCl, 0.5% (v/v) NP-40, complete protease-inhibitor cocktail (Roche)), scraped into tubes, and incubated for 10 min on ice. Cells were then pelleted at 1,000 r.p.m. for 10 min at 4 °C and resuspended in 500 μl of nuclear lysis buffer (50 mM Tris-HCl, 10 mM EDTA, 10% (w/v) SDS, complete protease-inhibitor cocktail) and incubated on ice for 10 min. Lysates were sonicated eight times for 20 s each at 40% amplitude at 4 °C with intermittent incubations on ice for 2 min. Centrifugation at 12,500g for 10 min at 4 °C clarified lysates. We set aside 10 μl of sheared chromatin for each sample as input. We diluted 50 μl of chromatin at a 1/10 ratio in immunoprecipitation dilution buffer (16.7 mM Tris-HCl pH 8.0, 0.01% (w/v) SDS, 167 mM NaCl, 1.2 mM EDTA, 1.1% (v/v) Triton-X100, complete protease inhibitor) and incubated with 5 μg of antibody on a rotator overnight at 4 °C. Samples were then incubated at constant rotation with 25 μl of prewashed Dynabeads protein G (Life Technology, catalogue number 10004D) for 2 h at 4 °C. Beads were washed once with a low-salt wash buffer (20 mM Tris-HCl, 0.1% (w/v) SDS, 1% (v/v) Triton X-100, 2 mM EDTA, 150 mM NaCl), once with high-salt wash buffer (20 mM Tris-HCl, 0.1% SDS, 1% Triton X-100, 2 mM EDTA, 500 mM NaCl), once with LiCl wash buffer (10 mM Tris-HCl, 1% (v/v) NP-40, 1% (w/v) sodium deoxycholate, 1 mM EDTA, 250 mM LiCl), and twice with TE buffer (10 mM Tris-HCl pH 8.0, 1 mM EDTA) before two rounds of incubation with 100 μl of elution buffer (1% SDS, 100 mM NaHCO3) for 15 min at room temperature. The eluates were incubated with 8 μl of 5 M NaCl on a rotator at 65 °C overnight. We added 3 μl of 10 mg ml−1 RNase A (ThermoFisher Scientific, catalogue number EN0531) and incubated samples first at room temperature for 30 min, and then with 4 μl of 0.5 M EDTA, 8 μl of 1 M Tris-HCl and 1 μl proteinase K (Roche, catalogue number 03115887001) at 45 °C for 2 h. DNA was purified using gel/PCR DNA-fragment extraction (Geneaid, catalogue number DF300) and diluted with 150 μl of TE buffer. Primers are listed in the reagents table included with the Supplementary Information. Following ChIP–qPCR analysis, ChIP enrichments, accounting for typical background fluctuations across repetitive DNA loci, were calculated as (Percentage of input/IgG) = (Percentage of input for protein immunoprecipitation)/(Percentage of input for mock IgG immunoprecipitation). The mean IgG background is also shown on ChIP graphs (Fig. 1b and Extended Data Figs. 1d–k, 8a).

Sequential chromatin immunoprecipitation

Similar to regular ChIP, for sequential chromatin immunoprecipitation (ChIP–reChIP), cells were grown to 80% confluence, crosslinked and lysed. For the first round of immunoprecipitation, samples were diluted 1/10 in immunoprecipitation dilution buffer (100 μl chromatin plus 900 μl immunoprecipitation dilution buffer) and incubated with 5 μg of anti-Pol-I (anti-RPA135 subunit) or anti-Pol-II (anti-C-terminal domain (CTD)) on a rotator overnight at 4 °C. Samples were then incubated at constant rotation with 25 μl of pre-washed Dynabeads for 2 h at 4 °C. Similar to regular ChIP (see above), beads were washed once with low-salt wash buffer, once with high-salt wash buffer, once with LiCl wash buffer, and twice with TE buffer before one 30-min incubation with 50 μl elution buffer containing 10 mM DTT. Eluates from each of the first immunoprecipitation tubes corresponding to the same antibody were combined, diluted 20-fold in cold immunoprecipitation dilution buffer and incubated overnight at 4 °C with 5 μg of anti-senataxin (anti-SETX) antibody. Once again, beads were incubated at constant rotation with 25 μl of pre-washed Dynabeads for 2 h at 4 °C, washed once with low-salt wash buffer, once with high-salt wash buffer, once with LiCl wash buffer, and twice with TE buffer before two rounds of incubation with 100 μl of elution buffer for 15 min at room temperature, and overnight incubation at 65 °C with 8 μl of 5 M NaCl. Similar to regular ChIP, samples were treated with RNase A/proteinase K (Roche, catalogue number 03115887001) and purified; qPCR was then performed.

Quantitative PCR

Quantitative real-time PCR was performed using a Bio-Rad CFX Connect Real-Time. Ten microlitres of qPCR reactions each containing SensiFAST SYBR No-ROX kit (FroggaBio, catalogue number BIO-98050), 200 nM of each of the forward and reverse primers, and 1 μl of diluted complementary DNA, diluted input, diluted immunoprecipitation ChIP or diluted DRIP DNA depending on the experiment. PCR comprised one cycle of 95 °C for 5 min and 60 °C for 30 s, followed by 39 cycles of 95 °C for 5 s and 60 °C for 30 s, and a final melt curve of 65 °C to 95 °C in 0.5 °C steps at 5 s per step.

RNA extraction

Cells grown to 70–80% confluence were washed with RNase-free PBS before RNA isolation using a Qiagen RNeasy mini Kit (catalogue number 74104).

Reverse transcription

For regular reverse transcription, 1 μg of total RNA was treated with 1 μl of 10× DNase-I reaction buffer and 1 μl of DNase I Amp grade (1 U μl−1; ThermoFisher, catalogue number 18068015), and then incubated for 15 min at room temperature. The reaction was quenched with 1 μl of 25 mM EDTA and incubated for 10 min at 65 °C. We carried out 10 μl reverse-transcription reactions using 10 mM deoxynucleoside trisphophate (dNTPs), 50 μM random nonamers (Sigma, catalogue number R7647), 500 ng total RNA, 5× first-strand buffer, 100 mM dithiothreitol (DTT), 40 U μl−11 RNaseOUT (Invitrogen, catalogue number 10777019) and 200 U μl−1 M-MLV reverse transcriptase (Invitrogen, catalogue number 28025013) at 25 °C for 10 min, 37 °C for 60 min and 70 °C for 15 min. For pre-rRNA pulse chase, an additional step comprising 5 min at 85 °C was added to release the RNA from beads. The reverse-transcription reaction was diluted 1:5, and 4 μl were used in qPCR amplification. For strand-specific (ss)RT–qPCR, 30 μg of total RNA was treated with DNase I (10 U DNase I in a 100 μl reaction) for 30 min at 37 °C. The reaction was stopped by adding 2 μl of 250 mM EDTA, pH 8.0, and incubating at 75 °C for 10 min. RNA was precipitated with 25 μl RNA precipitation solution (0.8 M trisodium citrate, 1.2 M NaCl) and 50 μl isopropanol. Samples were incubated for 10 min at room temperature and 20 min at −20 °C, and then centrifuged at 7,500g for 20 min at 4 °C. Supernatants were aspirated and pellets were air-dried for 10 min. Pellets were resuspended in 30 μl deionized, diethylpyrocarbonate-treated (ddDEPC) H2O and incubated at 65 °C for 5 min. Concentrations of purified RNA were measured using NanoDrop.

We designed strand-specific primers to allow the detection of sense and antisense transcripts at the same locus as described previously35. Briefly, a primer of roughly 18 bp was designed to recognize the strand of interest (for example, a reverse primer to detect sense transcripts, a forward primer to detect antisense transcripts). A nonsense sequence (CGAGGATCATGGTGGCGAATAA) was added to tag the 5′-end of each strand-specific IGS primer. As a control within each reverse-transcription reaction, we generated a reverse primer to detect 7SK sense transcripts (we added a T7 sequence to the 5′ end of this primer to distinguish it from IGS primers). Separate reverse-transcription reactions were carried out for each transcript of interest. Each 10 μl reverse-transcription reaction contained 200 ng purified RNA, 5 μM strand-specific tagged primer (comprising roughly 18 bp specific to the transcript of interest, with the nonsense sequence CGAGGATCATGGTGGCGAATAA added to the 5′-end), 5 μM control sense primer (for example, 7SK), 1 mM dNTPs, 1× first-strand buffer, 10 mM DTT, 40 U RNaseOUT and 200 U of M-MLV reverse transcriptase. False-prime reactions were also carried out for each RNA sample and were conducted by replacing the transcript-of-interest primers with DEPC ddH2O. Reactions were incubated at 25 °C for 10 min, 37 °C for 60 min, and 70 °C for 15 min. Resulting cDNA was diluted 1 in 10. Each cDNA sample represents one strand-specific transcript of interest and 7SK sense transcripts as a control. Each cDNA sample was amplified using primers directed at the strand-specific transcript of interest (using ss_Tag and hIGS_forward primers for sense transcripts or ss_Tag and hIGS_reverse primers for antisense transcripts), as well as 7SK (using T7 and 7SK forward primers). False-primed cDNA was amplified using all primer sets. qPCR reactions were performed at 95 °C for 5 min and 60 °C for 30 s, followed by 39 cycles of 95 °C for 5 s and 60 °C for 30 s. Results were analysed using the following formula: ΔΔCt = 2^-(ΔCtMutant − ΔCtWT), where ΔCt = Ct(transcript of interest) − Ct(control), and Ct is the cycle threshold. Values were normalized to those of false-prime reactions.

Population-level pre-rRNA pulse-chase

Click-iT Nascent RNA capture (Invitrogen, catalogue number C10365) was used. Cells were seeded in six-well dishes at 500,000 cells per well and allowed to grow to 40–50% confluence. Twenty-four hours later, cells were incubated with 0.15 mM ethyl uridine (EU) for 1 h, then with EU-free media for 2.5 h. Total RNA was extracted using Qiagen RNeasy kit (Qiagen, catalogue number 74104), and 1 μg of extracted RNA was incubated with 25 μl Click-iT EU buffer, 4 μl CuSO4, 1.25 μl biotin azide, 1.25 μL Click-iT reaction buffer additive 1 for 3 min before addition of 1.25 μl Click-iT reaction buffer additive 2 and incubation for 30 min. The reaction mix was then incubated with 1 μl of UltraPure Glycogen (Roche, catalogue number 10901393001), 50 μl of 7.5 M ammonium acetate, and 700 μl of chilled 100% ethanol at −80 °C overnight. RNA was then pelleted using centrifugation at 13,000g for 20 min at 4 °C and two rounds of washes with 700 μl of 75% ethanol. We then treated 1 μg of the RNA with 31 μl Click-iT RNA binding buffer and 2 μl RNaseOUT before incubation for 5 min at 68–70 °C. The heated RNA-binding reaction mix was incubated with 12 μl of washed bead suspension at room temperature for 30 min. The beads were washed five times with Click-iT reaction wash buffer 1 and five times with Click-iT reaction wash buffer 2. The beads were then resuspended with 12 μl of Click-iT reaction wash buffer 2 and incubated at 68–70 °C for 5 min before proceeding with reverse transcription and qPCR. Processing was measured by qPCR assessment of the levels of unprocessed pre-rRNA containing the 5′ external transcribed spacer (ETS) compared with the total levels of mature rRNA.

Single-cell rRNA biogenesis assay

On the day before the assay, cells from different experimental conditions were harvested and seeded onto poly-l-lysine (PLL)-coated coverslips in 24-well plates. On the day of the assay, live cells were pulse-labelled with 1 mM 5-fluorouracil (5-FU; Sigma, catalogue number F5130) for 15 min, gently washed with unlabelled media, and chased for 30 min. Cells were then fixed and immunostained as described in the Methods section ‘Endogenous protein immunofluorescence’. Double immunofluorescence labelling of nucleolar fibrillar centres or 5-FU-labelled RNA was performed using an anti-ATXN2 or anti-BrdU antibody, respectively. Random single cells were imaged captured at 100× using a Nikon C2+ confocal microscope coupled to NIS-Elements AR software (Nikon). Images were equally and evenly contrasted and ribosome biogenesis was measured as the ratio of ATXN2-marked nucleolar fibrillar centres with surrounding rRNA rings over the total number of nuclear ATXN2 foci.

Nuclear run-on

Click-iT Nascent RNA capture (Invitrogen, catalogue number C10365) was used for nuclear run-on (NRO). The setup was similar to that in the ‘Population-level pre-RNA pulse-chase’ section above, except that the total RNA was extracted after a 1 h incubation with 0.15 mM EU. Similar to pulse-chase labelling, the extracted RNA was biotinylated, precipitated, washed using Dynabeads, and reverse transcribed; qPCR was performed to measure the synthesis of nascent sincRNAs.

DNA–RNA hybrid immunoprecipitation

For DNA–RNA hybrid immunoprecipitation (DRIP) experiments, cells were first seeded in 60 mm plates at 2.5 × 106 cells per ml and allowed to grow to 70% confluence. Cells were then washed twice with ice-cold PBS, scraped, and centrifuged at 253g for 5 min. Cell pellets were resuspended in 1.6 ml TE buffer and incubated with 41.5 μl of 20% SDS and 5 μl of proteinase K overnight at 37 °C. Then, 1.6 ml of phenol-chloroform was added to cells before centrifugation at 466g for 5 min at room temperature. The aqueous layer was transferred and the addition of phenol-chloroform was repeated. The DNA was then precipitated by adding a 1/10 volume of 3 M NaOAc, pH 5.2, and 2.4 volumes of 100% ethanol to the aqueous layer. The DNA fibre was washed five times with 70% ethanol, resuspended in TE buffer and incubated with 3.5 μl spermidine (Bioshop, catalogue number SPR070), 35 μl buffer 2.1 (NEB), 5 μl HindIII (NEB, R01045), 10 μl EcoRI (Thermo Fisher, ERO271), 10 μL BsrGI (NEB, R05755), 5 μl XbaI (NEB, R01455) and 2 μl SspI (NEB, R0132). We then added 40 μl of 3 M NaOAc, pH 5.2, and one volume of phenol-chloroform to the digested DNA, which was then centrifuged at maximum speed for 5 min. The aqueous layer was transferred, and addition of phenol-chloroform was repeated.

To precipitate the DNA, 2.4 volumes of cold 100% ethanol were added to the aqueous layer, incubated at −20 °C for 15 min, and centrifuged at maximum speed for 30 min at 4 °C. The DNA pellet was washed with 70% ethanol and spun at maximum speed for 5 min at 4 °C. The dry pellet was resuspended in 50 μl TE buffer, and 4.4 μg of the DNA was incubated with 350 μl TE buffer, 50 μl 10× binding buffer (100 mM NaPO4 pH 7.0, 1.4 M NaCl, 0.5% (v/v) Triton X-100) and 10 μg of either mouse IgG or S9.6 antibody at 4 °C overnight. Immunoprecipitation samples were incubated with previously washed Dynabeads for 2 h at 4 °C. Samples were then washed three times with 1× binding buffer and eluted off the beads by incubation with DRIP elution buffer (50 mM Tris-HCl, pH 8.0, 10 mM EDTA, 0.5% (w/v) SDS) and proteinase K for 45 min at 55 °C. The DNA was then purified using gel/PCR DNA fragment extraction (Geneaid, catalogue number DF300) and qPCR of purified DNA was performed. The specificity of the S9.6 antibody for RNA–DNA hybrids was confirmed by in vitro treatment with RNase H1 in all experiments. We also screen all antibodies for specificity by ensuring that signals do not exhibit any statistically significant changes following treatment with RNase III (NEB, catalogue number M0245S). Following ChIP–qPCR analysis, background IgG mock signal was subtracted from S9.6 immunoprecipitation signal to generate a DRIP signal, which was then plotted as a raw DRIP signal or as a relative DRIP signal when normalized to a given site or condition.

Fluorescence recovery after photobleaching

HEK293T cells were transfected with a GFP–UBF1 plasmid 24 h before cell passaging to 2 cm glass-bottomed live microscopy dishes. Next day, the roughly 75% confluent cells were treated with either flavopiridol to a final concentration of 2 μM or dimethylsulfoxide (DMSO) as a control. Cells were incubated for 3.5 h and subjected to fluorescence recovery after photobleaching (FRAP) microscopy. Confocal microscopy was executed using a ×100 oil-immersion lens (numerical aperture 1.47) on a Leica DMi8 motorized inverted microscope (Leica Microsystems) coupled to a VT-iSIM multipoint scanner (VisiTech International) and detected with a Flash 4.0 v3 sCMOS camera (Hamamatsu). FRAP was performed using the iLas FRAP system (Gattaca Systems). Design of the acquisition journals and system integration were by Quorum Technologies. Images were acquired with a 488-nm excitation wavelength laser at 15% intensity. Cells were initially imaged 20 times, and the point of interest was subsequently bleached with a 405-nm laser for 36 ms at a laser intensity of 26%. Cells were then imaged repeatedly for approximately 1 min post-bleach to capture recovery. Signal intensity was measured using MetaMorph analysis software. For analysis, the intensity of the region of interest was normalized to a nucleoplasmic background region at every time point. These background-adjusted values were then normalized to the intensity value from the first time point. The bleach time points (6–6.3 s) display saturated fluorescence as the bleached region of interest and were therefore not included in any quantification. Post-bleach control focus intensity values of greater than 1 are a result of bleach-induced decreases in nucleoplasmic background.

Creation and use of RED–LasRR system

Full-length human RNase H1 was fused to eGFP and the deactivated Streptococcus pyogenes Cas9 (with D10A and H840A mutations). The 5,844-nucleotide RNaseH1–SV40NLS–eGFP–SV40NLS–dCas9 gene was synthesized and cloned into the pcDNA4/TO plasmid (Invitrogen) using NotI and XbaI restriction sites (here, NLS is a nuclear localization sequence, and SV40 is simian virus 40). To ensure protein flexibility, a (GGGS)4 linker was inserted between RNase H1 and the first SV40NLS, and another between eGFP and the second SV40NLS. GGS linkers were also inserted between the first SV40NLS and eGFP, and between the second SV40NLS and dCas9. dRNaseH1-eGFP-dCas9 was generated by introducing the point mutation D210N to RNase H1 using the Q5 site-directed mutagenesis kit (NEB, catalogue number E055450) according to the manufacturer’s protocol, with a modification of a 15-min instead of 5-min incubation with the KLD enzyme mix at room temperature. The oligonucleotide sequences for PCR amplification were 5′-gttctgtatacaaacagtatgttt -3′ and 5′-cagtttattgatgttttgagtctt -3′. The resulting RNaseH1–SV40NLS–eGFP–SV40NLS–dCas9 (RED) or its RNase H1-dead version (dRED) was then integrated into the T-REx (ThermoFisher Scientific) tetracycline-controlled expression system. Inducible expression of the fusion proteins is thus based on the binding of tetracycline to the Tet repressor, thereby derepressing the promoter controlling the expression of the RED and dRED fusion protein. To achieve locus-specific RED and dRED–LasRR enrichment, cells were allowed to reach 70% confluence over a period of roughly 24 h. For the inducible condition, cells were incubated with medium containing tetracycline (1 μg ml−1); for the uninduced condition, cells were incubated with tetracycline-free medium. All cells were transfected with 3 μg of RNH1–eGFP–dCas9 and dRED–LasRR plasmid per 60-mm plate by using Lipofectamine3000 (ThermoFisher) as per the manufacturer’s instructions. Induced cells were then cotransfected using RNAiMAX (ThermoFisher) as per the manufacturer’s instructions with either 4.5 μl of 10 pmol μl−1 nontargeting sgRNA, 1.5 μl of 10 pmol μl−1 of each of three sgRNAs for IGS18, 1.5 μl of 10 pmol μl−1 for each of three sgRNAs for IGS28, 1.5 μl of 10 pmol μl−1 for each of three sgRNAs for IGS38, or 1.5 μl of 10 pmol μl−1 for each of three sgRNAs for the β-actin 5′ pause element. The cells were incubated for 36 h before further experiments were performed.

CRISPR-mediated genome editing

For CRISPR-mediated gene knockout of SETX, CRISPR/Cas9 plasmids (pCMV–Cas9–GFP) were purchased from Sigma-Aldrich to express with the scrambled guide RNA and guide RNA for SETX (first intron). The transfections of the plasmids into the Flp-In 293 T-REx cell lines were performed with FuGENE transfection reagent (Roche, catalogue number E269A). We transfected 2 μg of the plasmid into HEK293T cells; one day after transfection, we sorted cells by BD FACSAriaTM flow cytometry (Donnelly Centre, Univ. Toronto), and plated single GFP-positive cells into 96-well plates. To confirm SETX knockout, the expression levels of SETX in each clone were detected by qPCR.

Northern blotting

RNA was prepared as described in the ‘RNA extraction’ and ‘Reverse transcription’ sections above. We then electrophorized 3.5 μg of RNA, and digoxygenin (DIG)-labelled the DNA probe for northern blotting using the DIG-high prime DNA labelling and detection starter kit I as per the manufacturer’s protocol (Roche, catalogue number 11745832910). Northern blots were performed using the DIG northern starter kit as per the manufacturer’s protocol (Roche, catalogue number 12039672910), with the following modification: electrophoresis was conducted at 15 V for 24 h at room temperature; RNA was UV-crosslinked (2,400 kJ for 1 min) to a positively charged nylon membrane; gels were blotted by capillary transfer with 20× SSC buffer (3 M NaCl, 0.3 M sodium citrate) overnight; a hybridization temperature of 50 °C was used; blots were hybridized overnight; and 200 μl of NBT/BCIP solution in 10 ml of detection buffer (0.1 M Tris-HCl pH 9.5, 0.1 M NaCl), was used for blot development.

DNA–RNA hybrid immunofluorescence

We seeded 60,000–80,000 cells per PLL-coated coverslip and allowed them to adhere for 24–36 h. Cells were fixed using 1% (v/v) formaldehyde for 15 min at room temperature, washed three times with 1× PBS for 5 min each, permeabilized with 500 μl of 0.3% (v/v) Triton-X100 for 5 min at room temperature, and washed again three times with PBS. Coverslips were blocked using 500 μl of 5% bovine serum albumin (BSA) for 1 h at room temperature, transferred to humidified chambers and incubated with 60 μl of primary antibody (1:500 of S9.6 antibody, 1% (w/v) BSA, 1× PBS) for 1 h at room temperature. After washing with PBS, cells were incubated with 60 μl of secondary antibody (1% BSA, 1:250 of goat anti-mouse 488 or 1:250 of goat anti-mouse 568) for 1 h at room temperature. The cells were washed again with PBS and incubated with 100 μl of DAPI (0.5 μl of DAPI per ml of PBS) for 2–4 min. The coverslips were then mounted onto microscope slides using DAKO mounting reagent, sealed with nail polish, and allowed to dry for 30 min. Images were acquired using a C2+ confocal microscope with a Plan-Apochromat TIRF ×100 oil objective (numerical aperture 1.45) and NIS-Elements AR software (Nikon). The specificity of the S9.6 antibody for RNA–DNA hybrids was confirmed by in vitro treatment with RNase H1 (NEB, catalogue number M0297S) under the same experimental conditions. Signals were also confirmed to differ from those yielded by immunofluorescence using J2, an antibody against double-stranded (ds)RNA.

Amyloid-body staining with Congo red

We seeded 40,000 cells on PLL-coated coverslips and allowed them to adhere for 24–36 h. Cells were fixed using 1% (v/v) formaldehyde and incubated for 15 min at room temperature, washed three times with 1× PBS for 5 min each, and permeabilized with 500 μl of 0.3% (v/v) Triton-X100 for 5 min at room temperature. The coverslips were then immersed in 250 μl of 0.05% (v/v) Congo red (Sigma, catalogue number C6277) solution for 15 min, followed by four cycles of 2 min rinsing with 500 μl of double-distilled H2O. The coverslips were then transferred to humidified chambers and nuclear counterstained with 100 μl DAPI, incubated for 4 min, and mounted on microscope slides using DAKO mounting reagent. Images were acquired using a C2+ confocal microscope with a Plan-Apochromat TIRF ×100 oil objective (numerical aperture 1.45) and NIS-Elements AR software (Nikon).

Endogenous protein immunofluorescence

We seeded 40,000 cells onto PLL-coated coverslips. Cells were fixed using 1% formaldehyde for 1 min at room temperature, washed with 1× PBS three times (5 min each wash), permeabilized with 500 μl of 0.3% Triton-X100 for 5 min at room temperature, and washed again three times with 1× PBS. Coverslips were blocked using 500 μl of 5% BSA for 1 h at room temperature, transferred to humidified chambers and incubated with 60 μl of primary antibody (1% BSA and anti-UBF or anti-NPM antibodies) for 1 h at room temperature. After washing with PBS, cells were incubated with 60 μl of secondary antibody (1% BSA, 1:250 of goat anti-mouse 488, 1:250 of goat anti-rabbit 568) for 1 h at room temperature. Coverslips were washed again with PBS, incubated with 100 μl DAPI for 2 min, mounted onto microscope slides using DAKO mounting reagent, and allowed to dry for 30 min. Images were captured at ×100 or ×60 using a Nikon C2+ confocal microscope coupled to NIS-Elements AR software (Nikon). For methanol/acetone-fixation-based immunofluorescence, the protocol was similar except that cells were fixed using ice-cold methanol for 15 min at room temperature, washed once with cold acetone, and washed with 1× PBS (3 × 5 min) before blocking with 5% BSA. Super-resolution microscopy was captured with a Leica DMI6000 SP8 LIGHTNING microscope using the HC PL APO CS2 x93 objective (numerical aperture 1.3, pinhole 110.5 µm). Images were deconvolved using Leica LIGHTNING deconvolution software and processed with Leica LAS software.

Stress-induced droplets and amyloid bodies

The ACM-containing VHL protein was transfected as pFLAG–VHL–GFP using Lipofectamine3000 according to the manufacturer’s protocol and treated/visualized 24 h post-transfection. For siRNAs (100 pmol), cells were transfected using RNAiMAX (ThermoFisher Scientific) at 90% confluency, split next day into fresh plates at 70% confluency to allow for subsequent GFP transfection with Lipofectamine3000, and treated/harvested 48 h post-transfection. The siRNAs (ThermoFischer Scientific) used were siControl (catalogue number 4390843), si-sincRNA16 (catalogue number 4399666) and si-sincRNA22 (catalogue number 4390828). For live-cell imaging of the ACM-containing and GFP-tagged VHL protein, cells were seeded in 145-μm-thick, 35-mm glass-bottom plates. Live-cell images were captured by confocal microscopy (Leica TCS SP5; Leica Microsystems, Mannheim, Germany), fitted with a variable temperature and 5% CO2 environmental chamber (Okolab), using a ×63 oil-immersion Plan-Apochromat objective (numerical aperture 1.4). Images were uniformly adjusted to increase brightness/contrast in Photoshop (Adobe).

In vitro droplet formation

Peptides were custom synthesized by GenScript (New Jersey, USA) at more than 95% purity. Peptide stock solutions were kept at 50 mM in nuclease-free-water. 5′-carboxyfluorescein (FAM)-labelled RNAs were synthesized by Integrated DNA Technologies (IDT, Coralville, IA) and resuspended in 50 mM NaCl to 100 μM. We mixed 1 μM low- or high-complexity ncRNA with the indicated peptide concentrations in 150 mM NaCl. Droplets were placed on a 1.5 coverslip and imaged after a 10-min incubation on a Zeiss AxioObserver D1 microscope using a ×63 Plan-Apochromat objective (numerical aperture 1.4).

Locked nucleic acid ASO knockdown of sincRNAs

Custom-designed locked nucleic acid (LNA) ASO GapmeRs were ordered from Qiagen. Sequences of 975–1,000 bp corresponding to IGS regions were entered into Qiagen’s custom antisense LNA GapmeR design page. The top-ranked ASOs based on Qiagen’s optimal design score were selected for each of IGS18, IGS20, IGS22 and IGS24, with standard desalting purification, phosphorothioate backbone modifications, and no-label/ready-to-label design specifications. ASO transfection was performed using RNAiMAX (ThermoFisher Scientific) according to the manufacturer’s protocols. ASOs were as follows: antisense LNA GapmeR control negative control B (catalogue number 339515, LG0000001-DDA; gctcccttcaatccaa), IGS18 (LG00210930-DDA; agtgtgctctgtgaac), IGS20 (LG00210936-DDA; acgcaagaaaggaaga), IGS22 (LG00210956-DDA; acgtgaccgagagaaa) and IGS24 (LG00210966-DDA; gtgacgtgtagagatt).

Subcellular fractionation by sucrose gradient

Cells were trypsinized, centrifuged at 1,000 r.p.m. for 4 min at 4 °C, washed with PBS and recentrifuged. The pellet was resuspended in osmotic buffer (10 mM HEPES pH 7.9, 1.5 mM MgCl2, 10 mM KCl, 0.5 mM DTT). The cells were then lysed using ten strokes of a tight pestle in a dounce homogenizer. Dounced cells were centrifuged at 1,000 r.p.m. for 5 min at 4 °C. The supernatant was retained as the cytosolic fraction. The pellet was resuspended in a 0.25 M sucrose plus 10 mM MgCl2 solution, and deposited over a 0.35 M sucrose plus 0.5 mM MgCl2 layer. The sample was centrifuged at 1,000 r.p.m. for 5 min at 4 °C. The sample was then resuspended in a 0.25 M sucrose plus 10 mM MgCl2 solution and sonicated at 25% power six times for 10 s with intermittent periods of 10 s rest on ice. The sample was deposited over a 0.88 M sucrose plus 0.5 mM MgCl2 layer and centrifuged at 3,500 r.p.m. for 10 min at 4 °C. The supernatant was retained as the nucleoplasmic fraction. The pellet was resuspended in a 0.35 M sucrose plus 0.5 mM MgCl2 solution and centrifuged at 3,500 r.p.m. for 5 min at 4 °C. The pellet was the nucleolar fraction. GAPDH transcripts, which are most abundant in the cytosolic fraction and are depleted from the nucleolar fraction, served as the control.

Aligning sequencing reads to human rDNA IGS

First, we used the Bowtie package to build a version of the human genome assembly hg19 with rDNA sequence; the newly built assembly is ‘hg19_plus_rDNA’. The human rDNA sequence FASTA file was obtained as is from the National Center for Biotechnology Information (NCBI; under GenBank accession code U13369.1, which refers to the ‘Human ribosomal DNA complete repeating unit’. This FASTA file, along with those for chromosomes 1–22, X, Y and M from hg19, obtained from the University of California at Santa Cruz (UCSC) genome browser (, was used to build the new assembly. Next, for testing, we aligned Pol II reads from HeLa cells to this new genome assembly using the Bowtie package aligner. The reads from two replicates were obtained from ENCODE and concatenated. Duplicate reads were removed with the package BBmap and its clumpify tool. Then, the alignment was performed with the parameter ‘-m 1’, which instructs bowtie to refrain from reporting any alignments for reads having more than one reportable alignment. This ensures that only uniquely aligning reads are reported. The alignments were processed further with Samtools to retain only those reads aligning to the rDNA sequence, and to compute the depth/number of reads at each position in the rDNA coordinates. These depths were plotted with an R script.

Calculation of GC skew

Using the roughly 43-kbp rDNA sequence obtained from the rDNA sequence FASTA file, we assessed GC skew, GC observed/expected ratio and GC percentage using (1-bp-at-a-time) sliding windows of size 50, 500 or 1,000 bp. Definitions are as follows: GC skew = (number of Gs − number of Cs)/(number of Gs + number of Cs); CG observed/expected ratio = sliding-window length × number of CpGs/(number of Cs × number of Gs); GC percentage = 100 × (number of Gs + number of Cs)/sliding-window length. To obtain an overall value/quantification and statistic with which to compare coding and IGS regions, we obtained the mean GC skews for the coding and IGS regions with window size 1,000. In the coding region the mean GC skew is 0.02346459; and in IGS regions, it is −0.1541796. Applying a Welch’s t-test to the GC skews from these two regions gives a P-value of less than 2.2 × 10−16. The script for all above analyses is called getGCskewEtc_rRDNA.R and is available upon request.


For Ewing- and osteosarcoma-related analyses, sample preparation and sequencing were carried out as described26. RNA-seq and DRIP–seq data sets have been deposited at Gene Expression Omnibus (GEO; under accession code GSE68847. Identification of rDNA IGS peaks from RNA-seq and DRIP–seq was conducted as per the section ‘Aligning sequencing reads to human rDNA IGS’ above, including a normalization of called peaks to the total number of reads per sample. For assessment of signals at non-rDNA loci, aligned .bam files were depth normalized and binned using bamCoverage from deepTools36. The resulting bigWig files were loaded into Integrated Genome Viewer (IGV)37 and the depicted regions were saved. For RNA-seq with/without heat shock, sequencing was performed on a cDNA library of total RNA (non-rRNA-depleted) using stranded paired-end reads. After discarding reads mapped to the rRNA gene (including the 5′-ETS, inverted repeat sequence (IRS)1/2 and 3′-ETS), we mapped the remaining reads to GRCh38. BAM files were separated into forward and reverse strand files (bash script). The remaining reads aligned to supercontig GL000220.1; this is within the latest human genome assembly, which contains a 43-kb rDNA cassette. Signals were normalized to an internal non-stress responsive control site at IGS35. To calculate changes in sincRNA and asincRNA levels following heat shock, we binned the IGSs into 5,000-bp bins, and calculated the change in absolute read counts for each bin. The average of these changes was calculated to obtain a global percentage change across the entire IGS region. The sequencing data are available at GEO under accession code GSE115731. ChIP–seq enrichments were generated by the ENCODE Project Broad Institute for H3K27ac, H3K9ac, H3K4me3 and H3K36me3, and by ENCODE Project SYDH for RNA pol II ChIP–seq. Briefly, bedGraph files previously generated38 by mapping ChIP–seq and input data from ENCODE Project Consortium 2012 to the human rDNA sequence from BAC clone GL000220.1 were used to generate IGV genome tracks. We note that qPCR and sequencing analyses of repetitive DNA loci reveal an average profile for the studied repeats and should not be interpreted as an absolute enrichment for any given unit within the repeats.

Transmission electron microscopy

Cell pellets were fixed in phosphate-buffered 4% formaldehyde plus 1% glutaraldehyde fixative for at least 2 h. Samples were subsequently rinsed in 0.1 M phosphate buffer for 5 min and then fixed in 1% Zetterqvist’s buffered osmium tetroxide for 1 h. After a short rinse in Zetterqvist’s buffer for 1 min, the samples were dehydrated in increasing concentrations of alcohol (70%, 95%, 100%) for 10 min followed by propylene oxide. Finally, pellets were embedded in epoxy resin. Ultrathin sections were contrasted with uranyl acetate and Reynold’s lead citrate and observed with a JEOL 1230 TEM equipped with an Advanced Microscopy Techniques (AMT) camera system.

Images of human tumour sections

Images of tumour sections stained with haematoxylin and eosin were obtained through the Sinai Health System (Toronto) without any identifiable personal health information and without personal information, following Institutional Research Ethics Board approval (Sinai Heath Systems, 17-0103-E).

Statistical analysis

GraphPad Prism-based calculations of P-values were carried out via t-test, one-way ANOVA (with Dunnett’s or Tukey’s multiple comparison test), or Mann–Whitney U-test. Unless otherwise indicated, replicate information is as follows. All data from pulldowns, reverse transcription and viability markers were generated using the indicated number of biological replicates. For blots, images are representative of data obtained from two independent biological replicates. For microscopy, images are representative of phenotypes observed in at least two independent biological replicates, and quantifications are based on at least 100 cells from two technical replicate cultures.

Reporting summary

Further information on research design is available in the Nature Research Reporting Summary linked to this paper.