Human U2OS and HEK293 cells were cultured in Dulbecco’s modified eagle’s medium (DMEM, Gibco), containing 10% fetal bovine serum (FBS, Capricorn), 100 U/ml penicillin-streptomycin and 2 mM glutamine (Gibco) at 37 °C and 5% CO₂. Acetic acid (Sigma), etoposide (Sigma) and CX-5461 (Selleckchem) were used as indicated. Small interfering (si)RNA (siControl scrambled, D-001810-01-05; siNONO, L-007756-01-0005, Dharmacon) were transfected with Lipofectamine 2000 (Invitrogen) and serum-reduced medium (OptiMEM, Gibco) according to the manufacturer’s protocol. For proximity labeling cells were incubated with 0.5 mM biotin-phenol (Iris Biotech) for 30 min at 37 °C and 1 mM hydrogen peroxide (Sigma). Labeling was quenched by 10 mM sodium ascorbate (Sigma), 5 mM trolox (Sigma) and 10 mM sodium azide (Sigma).

Coding sequences were PCR-amplified from pcDNA3.1-APEX2-NES (Addgene) and pmCherry-NONO (Ling-Ling Chen, Shanghai Institute of Biochemistry and Cell Biology). To create pRRL-APEX2-HA, amplified APEX2-HA and the pRRL-puro vector (Elmar Wolf, Biocenter Würzburg) were restricted with PacI/SpeI (4 h, 37 °C) and ligated with T4 DNA ligase (16 °C, overnight). To create pRRL-NONO-APEX2-HA, amplified NONO and pRRL-APEX2-HA were digested with PacI/MluI (4 h, 37 °C) and ligated with T4 DNA ligase (16°C, overnight). pRRL-NONO-APEX2-HA was transformed into E. coli and purified using a Miniprep Kit (NEB) according to the manufacturer’s protocol. The plasmid was incubated with PacI/SpeI or buffer (4 h, 37°C), separated on a 0.9% agarose 1x TBE gel (20 min, 150 V), and stained with ethidium bromide (0.2 μg/ml, Sigma) under UV light. All enzymes were purchased from NEB. For primers see Supplementary Table S1.

Plasmids (10 µg pRRL-NONO-APEX2-HA, 10 µg pPAX2, 2.5 µg pMD2.G) and 30 µL polyethylenimine (PEI, Calbiochem) were diluted in 500 µL OptiMEM, vortexed, incubated (25 min, RT) and added dropwise to HEK293 cells, which were preincubated in 5 ml DMEM/2% FBS, and transfected for 8 h. Virus was harvested 3x every 12 h, sterile filtered and frozen (-80°C). For infection, U2OS cells were cultured in viral mixture [(1.5 ml DMEM, 1.5 ml pRRL-NONO-APEX2-HA virus, 6 µL polybrene (Invitrogen)] and incubated (24 h, 37°C). The mixture was replaced by DMEM containing 2 μg/ml puromycin (Invivogen) for 10 days of polyclonal selection.

Cells were seeded at different densities and incubated for 72 h, washed in PBS, stained in 0.5% crystal violet solution/20% methanol (10 min, RT) and washed (3 × 20 min) in PBS. Stainings were scanned (Epson) and quantified (ImageJ, NIH).

Cells were washed in PBS, trypsinised, resuspended in DMEM and centrifuged (1200 rpm, 3 min, 4°C). Pellets were washed in PBS, centrifuged (1200 rpm, 3 min, 4°C), resuspended in 1 ml PBS and fixed in 4 ml 100% ethanol (−20°C, overnight). Cells were pelleted (1500 rpm, 10 min) and resuspended in 1 ml PBS. 1 x 10⁶ cells were stained with 54 µM propidium iodide (Sigma) in the presence of 24 μg/ml RNase A (Sigma) (30 min, RT, dark), sorted and analysed by a FACSDiva 9.0.1 flow cytometer and software (BD Biosciences).

Cells were lysed, boiled and sonicated in sample buffer (250 mM Tris-HCl pH6.8, 8% SDS, 40% glycerol, 8% β-mercaptoethanol, 0.02% bromophenol blue). Extracts were separated by SDS-PAGE, transferred onto nitrocellulose membranes (GE Healthcare), blocked in PBS/0.1% Triton-X-100/5% milk (PBST) (1 h, RT) and probed with primary antibodies (4°C, overnight). Membranes were washed in PBST, incubated with secondary antibodies (1 h, RT) and washed in PBST without milk. Signals were visualised with an ECL kit (GE Healthcare) and an imaging station (Fuji). Membranes were stained with ponceau S (0.5% ponceau S, 1% acetic acid) prior to blocking. For immunoprecipitation (IP), cells were trypsinised, washed with cold PBS and centrifuged (1200 rpm, 3 min). Pellets were lysed in five volumes IP buffer (200 mM NaCl, 0.5 mM EDTA, 20 mM HEPES, 0.2% NP-40, 10% glycerol, 400U RNase inhibitor, 1 x protease/phosphatase inhibitor) for 10 min on ice. Lysates were centrifuged (13,000 rpm, 10 min) and supernatants were incubated (3 h, 4°C) with 2 µg primary antibodies pre-conjugated with 25 µL Dynabeads (protein G or streptavidin C1, Invitrogen). Immunocomplexes were immobilised on a magnet (Invitrogen), washed with 800 µL IP buffer (3 × 10 min, 4 °C) and eluted with sample buffer (5 min, 95 °C). Gels were silver stained with a kit (Invitrogen) according to the manufacturer’s protocol. For antibodies see Supplementary Table S2.

Cells were harvested, washed in cold PBS, incubated in five volumes IP buffer (10 min on ice) and centrifuged (10 min, 13,000 rpm). Total RNA from 25% of lysate (input) was resuspended in TRIzol (Invitrogen). Remaining supernatant was split into 25% aliquots and incubated with 4 µg antibodies (3 h, 4°C). Antibodies were pre-conjugated with 25 µL Dynabeads overnight (protein G, Invitrogen). Immune complexes were washed 4 x in 800 µL IP buffer. Total and immunoselected RNA samples were purified using TRIzol according to the manufacturer’s protocol (Invitrogen). For quantitative analysis, samples were reverse-transcribed with Superscript reverse transcriptase III (Invitrogen) using region-specific primers. cDNA was quantified by real-time quantitative PCR using the SensiFAST SYBR No-ROX Kit (Bioline) and a StepOne Real Time PCR thermal cycler (Applied Biosystems). For the calculation of % of input the average cT values measured for input, IP and IgG controls were transformed into absolute replicative concentrations. Formula: conc_abs = 1/2^cT. Absolute replicative concentrations of IP and IgG samples were then transformed to relative replicative concentrations by normalising to inputs values, with input values set to 1. Formula: conc_relIP = conc_absIP/conc_absinput and conc_relIgG = conc_absIgG/conc_absinput. The % of input values were calculated by subtraction of relative IgG values from relative IP values, division by 1 (i.e. relative input values) and multiplication by 100. Formula: [(conc_relIP-conc_relIgG)/1]*100. For primers see Supplementary Table S1.

Cells were grown on coverslips, washed with PBS, fixed with 3% para-formaldehyde (Sigma) for 8 min at RT, washed with PBS, permeabilised with PBST (4 min, RT) and blocked in PBS/10% FBS (2 h, 4°C). Primary antibodies were incubated overnight at 4°C in PBS/0.15% FBS. Cells were washed in PBST, incubated with secondary antibodies in PBS/0.15% FBS (1 h, RT) in a dark humidified chamber and washed in PBST. Nuclei were stained with 4‘,6-diamidino-2-phenylindol (DAPI)-containing mounting medium (Vectashield), imaged (CLSM-Leica-SP2) and processed (ImageJ, NIH). Channels were acquired sequentially, between frames, and with equal exposure times. >100 cells per condition were quantified. Colocalisation was assessed by using RGB profiler (ImageJ) and by calculation of the Pearson’s correlation coefficient with JACoP (ImageJ). Parameters were set as described (Bolte and Cordelières, 2006). For antibodies see Supplementary Table S2.

We aimed to establish a system that allows the unbiased and stringent detection of NONO interactors by APEX2-mediated in vivo proximity labeling. Using PCR cloning, we generated inserts encoding NONO and HA-tagged APEX2, and ligated them into the pRRL expression vector (Figure 1A). An analytic digest displayed the NONO-APEX2-HA insert at the predicted size of 2.2 kb (Figure 1B). Next, we transduced the pRRL-NONO-APEX2-HA construct into U2OS cells and selected a polyclonal U2OS:NONO-APEX2-HA cell line. We used an HA antibody in confocal imaging to determine how many cells express NONO-APEX2-HA and reproducibly scored around 80% HA-positive nuclei in a total number of 280 analysed cells (Figure 1C). We assessed the growth and cell cycle of U2OS:NONO-APEX2-HA cells and observed no significant differences compared to parental cells (Supplementary Figures S1A, S1B). We conclude that the stable expression of NONO-APEX2-HA is non-toxic in U2OS cells.

To analyse the expression and subcellular localisation of NONO, we acquired a specific antibody and first validated its applicability in immune assays by using siRNA technology (Supplementary Figures S2A, S2B). We observed a significant reduction in NONO signals on immunoblots and in confocal imaging upon transfection of NONO-targeting siRNA. NONO accumulates in the nucleolus of DNA-damaged cells (Moore et al., 2011). We incubated U2OS cells with the DNA topoisomerase-II inhibitor etoposide and assessed the localisation of NONO by confocal imaging. We observed pan-nuclear signals for NONO in a subset of cells upon incubation with etoposide, but no other treatments (Figure 2A). We confirmed the DNA damage-induced nucleolar accumulation of NONO by costaining with the nucleolar exosomal subunit EXOSC10 and quantified 30–50% of cells with pan-nuclear NONO staining (Figure 2B and Supplementary Figure S2C), which was further confirmed by RGB profiler measurements (Supplementary Figure S2D). Etoposide also induced partial colocalisation of NONO with nucleolar Nucleophosmin 1 (NPM1) (Supplementary Figure S2E). A fraction of EXOSC10 translocates to the nucleoplasm in response to DNA damage (Domingo-Prim et al., 2019). However, we could not detect prominent colocalisation of EXOSC10 with γH2A.X foci upon incubation with etoposide (Supplementary Figure S3A). Instead, the bulk of EXOSC10 remained nucleolar (Supplementary Figure S3B). We conclude that short-term incubation with etoposide triggers the nucleolar accumulation of NONO and does not impair nucleolar integrity. In response to UV irradiation, the ubiquitin-ligase ring finger protein 8 (RNF8) targets NONO for proteasomal degradation (Deshar et al., 2019). Thus, we monitored the NONO protein level by immunoblotting. Incubation with etoposide strongly induced the serine-139-phosphorylated H2A.X variant (γH2A.X), a marker for DNA damage, but did not alter NONO levels (Figure 2C). NONO binds the I-SceI cleavage site of a DSB reporter (Krietsch et al., 2012). To test if NONO accumulates at endogenous DSBs, we costained NONO with the p53-binding protein 1 (53BP1) or γH2A.X (Supplementary Figure S4). Whilst 53BP1 and γH2A.X clearly colocalised, no obvious colocalisation of NONO with the DSB markers could be observed upon etoposide treatment. We conclude that the accumulation of NONO at DSB foci is modest.

Next, we tested the expression and localisation of NONO-APEX2-HA in response to DNA damage. We first monitored the size and expression level of NONO-APEX2-HA by immunoblotting (Figure 2D). Samples from U2OS:NONO-APEX2-HA, but not wild type cells, displayed HA antibody reactivity at 86 kDa irrespective of DNA damage, which was confirmed by 6-8-fold stronger γH2A.X signals. The NONO antibody generated a 55 kDa band in all conditions and an additional 86 kDa band in samples from U2OS:NONO-APEX2-HA cells. To assess the NONO-APEX2-HA localisation, we used the HA antibody in confocal microscopy in combination with NPM1 staining. The HA antibody stained nucleoplasmic in unperturbed cells, but displayed additional nucleolar reactivity in 50–60% of DNA-damaged cells (Figure 2E). RGB profiler measurements revealed elevated colocalisation of HA and NPM1 signals in the presence of etoposide (Figure 2F). We conclude that NONO-APEX2-HA expresses irrespective of DNA damage and localises comparable to endogenous NONO.

NONO forms complexes with the splicing factor proline and glutamine rich (SFPQ) and the paraspeckle component 1 (PSPC1) (Knott et al., 2016). Thus, we asked if NONO-APEX2-HA associates with paraspeckle proteins. We immunoprecipitated NONO from wild type U2OS cells and detected comparable amounts of NONO, SFPQ and PSPC1, irrespective of DNA damage (Figure 3A). Next, we immunoprecipitated NONO-APEX2-HA using NONO or HA antibodies (Figure 3B). Again, we coenriched paraspeckle components irrespective of DNA damage. Strikingly, we detected an additional 86 kDa band that was clearly enriched after immunoprecipitation with the HA antibody. Further, we immunoselected SFPQ from U2OS wild type and U2OS:NONO-APEX2-HA lysates (Figure 3C). The SFPQ antibody coenriched paraspeckles proteins from both lysates, but NONO-APEX2-HA from the U2OS:NONO-APEX2-HA lysate only. We also assessed if NONO-APEX2-HA associates with the nuclear enriched assembly transcript 1 (NEAT1). The long non-coding RNA NEAT1 is essential for the formation of paraspeckles and a cognate substrate of NONO (Shav-Tal and Zipori, 2002; Chen and Carmichael, 2009; Clemson et al., 2009; Sasaki et al., 2009; Knott et al., 2016). We immunoprecipitated NONO-APEX2-HA with the HA antibody and quantified the level of coenriched NEAT1 transcripts by RT-qPCR and could detect significantly enriched levels of NEAT1 upon HA immunoselection from lysates of U2OS:NONO-APEX2-HA, but not wild type cells (Figure 3D, left). Next, we repeated the immunoprecipitation with the NONO antibody. Strikingly, we coenriched similar amounts of NEAT1 from wild type and U2OS:NONO-APEX2-HA cells (Figure 3D, right). Further, we compared the subcellular localisation of NONO-APEX2-HA with paraspeckle proteins and DSB markers (Figure 3E and Supplementary Figure S5). Indeed, cells that stained positive for NONO-APEX2-HA displayed a strong colocalisation with NONO, SFPQ and PSPC1, but neither with 53BP1 nor γH2A.X. We conclude that NONO-APEX2-HA associates with paraspeckle components and the lncRNA NEAT1.

Next, we wished to test if NONO-APEX2-HA is catalytically active. The human genome encodes naturally biotinylated enzymes (Knowles, 1989). We first asked if these are detectable in wild type U2OS cells. Using a horse radish peroxidase-conjugated streptavidin probe (Strep-HRP) on immunoblots, we observed four bands that migrated at predicted sizes, but were not detectable upon NONO immunoselection (Supplementary Figure S6A). APEX2 uses hydrogen peroxide (H₂O₂) to generate reactive biotin-phenol radicals that conjugate with proximal amino acids and transcripts (Fazal et al., 2019; Hung et al., 2016; Lam et al., 2015) (Figure 4A). To test labeling kinetics, we incubated U2OS:NONO-APEX2-HA cells with 1 mM H₂O₂ (Supplementary Figure S6B). We observed a smear of Strep-HRP signals in the inputs and upon biotin immunoprecipitation in the presence of H₂O₂, but only naturally biotinylated proteins without H₂O₂. Reassuringly, the HA antibody displayed a single 86 kDa band upon biotin immunoprecipitation, most clearly after a brief incubation with H₂O₂. We repeated the labeling and included wild type cells as control (Figure 4B). Again, Strep-HRP displayed naturally biotinylated proteins in all input conditions, whereas additional biotinylated proteins were only observed in samples from U2OS:NONO-APEX2-HA cells that were incubated with H₂O₂. We reprobed the Strep-HRP blot with the NONO antibody and detected NONO at 55 kDa in all conditions and NONO-APEX2-HA at 86 kDa specifically in samples from U2OS:NONO-APEX2-HA cells irrespective of H₂O₂ treatment. Upon biotin immunoprecipitation, however, the HA antibody stained NONO-APEX2-HA only in the presence of H₂O₂. Next, we assessed biotinylation in vivo. We tested if biotin could be traced as pan-nuclear signal and costained NONO-APEX2-HA with biotin using a NeutrAvidin-568 probe (Figure 4C). Indeed, NeutrAvidin-568 displayed strong colocalisation with HA in the nucleoplasm, but not the cytoplasm of unperturbed cells. In the presence of etoposide, pan-nuclear NeutrAvidin-568 staining could be observed. Biotin-phenol and H₂O₂ raise some concerns about toxicity (Qin et al., 2021; Weissinger et al., 2021). Thus, we monitored the onset of DNA damage by APEX2 substrates in U2OS:NONO-APEX2-HA cells. We assessed the levels of the tumour suppressor protein p53, γH2A.X, serine-1981-phosphorylated ataxia telangiectasia mutated (pATM) kinase and its phospho-substrates by immunoblotting (Supplementary Figure S6C). We observed little induction of DNA damage markers by the APEX2 substrates, but strong induction of DNA damage by etoposide, which was not further elevated by the combination of APEX2 substrates and etoposide. This suggests that the bulk of DNA damage is triggered by etoposide, but not by APEX2 substrates. Finally, we tested if NONO-APEX2-HA biotinylates distinct proteins. We performed proximity labeling and probed for paraspeckle components (NONO, SFPQ, PSPC1), nucleolar proteins (NPM1, Fibrillarin), and cytoplasmic Vinculin on immunoblots (Figure 4D). Indeed, we detected the nuclear and nucleolar proteins, but not Vinculin, upon biotin immunoprecipitation irrespective of DNA damage and dependent on the presence of H₂O₂ (Supplementary Figure S6D). We conclude that NONO-APEX2-HA biotinylates proteins in vivo.