SENP1 Decreases RNF168 Phase Separation to Promote DNA Damage Repair and Drug Resistance in Colon Cancer

In response to various types of DNA damage, cells use a set of signaling cascades, known as the DNA damage response (DDR), to sense and repair lesions to maintain genomic integrity and cellular homeostasis. A DNA double-strand break (DSB) is considered the most harmful DNA lesion (1). Two dedicated pathways, namely nonhomologous end joining (NHEJ) and homologous recombination (HR), are responsible for DSB repair (2–4). Following the occurrence of a DSB, a cascade of proteins is recruited to the damaged sites, among which, the RING finger protein 168 (RNF168), an E3 ubiquitin ligase of H2A (5, 6), has been identified as a necessary factor for the recruitment of repair proteins to damaged chromosomes, including 53BP1, the most critical effector in the NHEJ repair pathway (7).

Mounting evidence indicates that protein posttranslational modifications (PTM) play important roles in the DDR (8, 9). Similar to phosphorylation and ubiquitination, reversible SUMOylation, modulated by E1 activating enzyme, E2 conjugating enzyme, E3 ligating enzyme, and sentrin/SUMO-specific proteases (SENP), fine-tunes the recruitment of repair proteins to the damaged sites (10–12). For example, the SUMOylation E3 ligases PIAS1 and PIAS4 have been shown to promote DSB repair (10). SUMOylation of MDC1 is required for its degradation and removal from damaged sites (13). TIP60 SUMOylation attenuates its interaction with DNA-PKcs, thus facilitating HR (14). SENP2 regulates NEMO and NF-κB activation induced by DNA damage (15). However, whether SUMOylation/deSUMOylation regulates the function of RNF168 in the DDR remains unknown.

Recent studies have revealed that liquid–liquid phase separation (LLPS) acts as a governing mechanism for biomolecular condensation and membraneless organelle formation, which serves to regulate multiple biological processes including transcription, autophagy, intracellular signaling, and genomic integrity (16–20). For example, poly-ubiquitin chain-induced phase separation of adaptor protein p62 regulates protein quality by segregating the autophagosome marker LC3 (21). Pol II–mediated transcription takes place in nuclear condensates that assemble at both the promoter and enhancer loci (22–24). PTMs serve critical functions in governing LLPS (25–27); for example, SUMOylation regulates the disassembly and formation of stress granules (18), while ubiquitin binding disrupts UBQLN2 LLPS (28). PARylation has also been reported to promote the phase separation of hnRNP A1 (29). In addition, DDR proteins have been shown to undergo LLPS; for instance, DNA damage-induced 53BP1 phase separation coordinates damage detection with gene expression and checkpoint activation (30, 31). Moreover, LLPS regulates cellular processes involved in cancer cell pathology, and the dysregulation of LLPS is increasingly implicated as a previously hidden driver of oncogenic activity (19, 32). Therefore, it is of great importance to identify new PTM regulators that impact cancer progression through LLPS.

Here, we demonstrate that SUMOylated RNF168 induces LLPS, which sequesters 53BP1 into RNF168-formed nuclear puncta, subsequently reducing the recruitment of RNF168 to the DNA damage site and ultimately impairing the efficiency of the NHEJ repair pathway. In response to DSB, SENP1 specifically deSUMOylates RNF168 and disrupts the formation of nuclear condensates, which promotes NHEJ and cell survival. Moreover, SENP1 is expressed at a higher level in colon adenocarcinoma (COAD) and is related to a poor prognosis, whereas disrupting SENP1 leads to genomic instability and increases the sensitivity of colon cancer cells to chemotherapeutic drugs. Our findings provide novel insight into the molecular mechanism underlying the safeguarding of genomic integrity and the regulation of drug resistance by the SENP1-RNF168–53BP1 axis.

Rabbit anti-SUMO3 mAbs were purchased from Proteintech (catalog no. 11251–1-AP; RRID: AB_2198405) and Abcam (catalog no. 2970–1; RRID: AB_2198534). Mouse anti-Flag (catalog no. F3165; RRID: AB_259529) and anti-HA (catalog no. H9658; RRID: AB_260092) mAbs were bought from Sigma-Aldrich. Mouse anti-Myc (catalog no. M047–3; RRID: AB_591112), anti-His (catalog no. D291–3; RRID: AB_10597733), anti-α-tubulin (catalog no. M175–3; RRID: AB_10697817) mAbs and rabbit anti-GFP mAb (catalog no. 598; RRID: AB_591819) were procured from MBL. Rabbit anti-RNF168 and anti-SENP1 mAbs were purchased from Proteintech (catalog no. 21393–1-AP; RRID: AB_10733883) and Abcam (catalog no. ab108981; RRID: AB_10862449), respectively. Mouse anti-γH2AX (catalog no. DR1016–100UG; RRID: AB_437862) mAb was purchased from Millipore. Rabbit anti-H2A mAb was bought from Cell Signaling Technology (catalog no. 12349; RRID: AB_2687875). Rabbit anti-β-actin mAb was procured from Abclonal (catalog no. AC004, RRID: AB_2737399). Mouse anti-53BP1 mAb (catalog no. sc-517281; RRID: AB_2921289), rabbit anti-53BP1 mAb (catalog no. 88439; RRID: AB_10694558), and rabbit anti-BRCA1 pAb (catalog no. BS1036; RRID: AB_1662959) were purchased from Santa Cruz, Cell Signaling Technology, and Bioworld, respectively.

Plasmids expressing human SENP1, His-SUMO1, and His-SUMO3 were kindly provided by Prof. Jinke Cheng, Shanghai Jiao Tong University School of Medicine. Human wild-type RNF168 and the K210R mutant were inserted into the pcDNA-3Myc, pCMV-HA, pcDNA-3Flag, pEGFP-C1, and pET-28a vectors. Human RNF168 K173R and K203R mutants in the pcDNA-3Myc vector were generated by site-directed mutagenesis. The cDNAs of PIAS family genes were cloned into the pCMV-HA vector. All short hairpin RNA (shRNA) sequences were cloned into the pLKO.1 vector. SENP1 was inserted into the pEGFP-N3 vector. H2A was inserted into the pcDNA-3Flag vector. With a view to mimicking constitutive SUMOylation of RNF168, RNF168 was fused to SUMO3 in tandem by insertion mutagenesis, and then cloned into the pcDNA-3Myc and pEGFP-C1 vectors. All constructs were confirmed by DNA sequencing.

HEK293T, SW480, HeLa, MEF, NCM460, HT29, FHC, LOVO, and RKO cells were cultured in DMEM supplemented with 10% FBS (Gibco). HCT116 cells were maintained in RPMI1640 (Gibco). MEF SENP1 knockout cells and controlled wild-type cells were kind gifts from Jinke Cheng (Shanghai Jiao Tong University, Shanghai, China). The identities of all cell lines were authenticated by short tandem repeat analysis and verified by PCR to ensure the absence of Mycoplasma contamination.

Experiments using mice were performed following the ‘Principles for the Utilization and Care of Vertebrate Animals’ and the ‘Guide for the Care and Use of Laboratory Animals’. Animal studies were approved by the Institutional Animal Care and Use Committee (IACUC) of the Center for Experimental Animal Research (China) and Peking University Laboratory Animal Center (IACUC no. LSC-ZhengX-2–1). Fifteen female BALB/c nude mice (4 weeks of age) were bought from Beijing Vital River Laboratory Animal Technology and randomly grouped into five categories. A total of 4 × 10⁶ HCT116 cells were injected subcutaneously into the armpits, with 2 × 10⁶ cells on each side. Tumor size was measured every 2 days from day 7 using a caliper, and the volume of each tumor was calculated using the following formula: volume = length × width²/2. Mice were all sacrificed on day 12 postinjection.

HEK293T cells were transfected with His-SUMO and the indicated plasmids for 48 hours, then harvested and lysed in His-pulldown lysis buffer. The lysate was incubated with 40 μL Ni²⁺ beads for 4 hours at room temperature. After washing 4 times with different wash buffers, the beads were denatured in 2× SDS loading buffer, and the level of SUMOylation was assessed using the indicated antibodies.

The chromatin extraction assays were conducted as previously described (33). Cells were collected and lysed in chromatin extraction buffer A (10 mmol/L PIPES pH 6.8, 100 mmol/L NaCl, 300 mmol/L sucrose, 3 mmol/L MgCl₂, 1 mmol/L EGTA, 0.2% Triton X-100) on ice for 30 minutes and then centrifuged at 3,000 × g, 4°C for 5 minutes. The supernatant was collected as soluble element, and cell pellets were lysed in chromatin extraction buffer B (3 mmol/L EDTA, 0.2 mmol/L EGTA, 1 mmol/L DTT) and centrifuged at 3,000 × g, 4°C for 5 minutes. The supernatant was completely removed, and the sediment was resuspended in buffer C (50 mmol/L Tris pH 8.0, 150 mmol/L NaCl, 1 mmol/L EDTA, 0.1% SDS, 1% Triton X-100) as the chromatin element. Both elements were denatured in 2× SDS loading buffer and tested by immunoblotting.

Cells were transfected with the indicated plasmids using PEI reagent. At 48 hours post-transfection, cells were collected and fixed in precooled methanol for 10 minutes at −20°C according to a previously described procedure (33). Nuclear DNA was stained with 4′,6-diamidino-2-phenylindole (DAPI; 1 g/mL). Images were obtained with a confocal microscope (Zeiss LSM-710 NLO & DuoScan, Germany; Zeiss LSM-800 NLO & DuoScan; Zeiss LSM-880 NLO & DuoScan; DeltaVision OMX SR) using the 63× oil objective lens. Quantitation was performed using the Imaris software (Bitplane).

Laser microirradiation was carried out following previously described procedures (34). HeLa cells were cultured in glass-bottomed Petri dishes and irradiated with an ultraviolet laser (365-nm pulsed nitrogen laser, 16 Hz pulse, 60% output). Images were taken using a Nikon A1 confocal imaging system, and time-lapse images were captured at the indicated times. ImageJ software were used to calculate the signal intensity of each cell.

Expression of recombinant His-tagged proteins in E. coli strain BL21 or Rosetta was induced with 0.5 mmol/L isopropyl-β-D-thiogalactopyranoside when the bacterial cultures reached an OD₆₀₀ value of 0.6. After expressing at 16°C for 20 hours, fusion proteins were purified on an Ni-Sepharose column according to the manufacturer's instructions (GE HealthCare).

Cells were transfected with the indicated plasmids using PEI reagent. At 48 hours post-transfection, cells were collected and washed with PBS buffer, lysed in modified RIPA buffer (50 mmol/L Tris-HCl pH 7.4, 150 mmol/L NaCl, 1 mmol/L EDTA, protease inhibitors), and sonicated using a JY92-II sonicator (on for 5 seconds, off for 5 seconds; 10 cycles). After pre-clearing with Protein G-Sepharose beads for 1 hour, the supernatant was incubated with the indicated antibodies or control IgG antibodies overnight, and then gently mixed with 50 μL Protein G Sepharose beads for a further 4 hours. Finally, the samples were denatured, separated by SDS-PAGE, and transferred to nitrocellulose membranes (GE HealthCare). Membranes were blocked by 5% skim milk and then probed with specific primary and secondary antibodies. The protein signals were detected and quantitated using the Odyssey Infrared Imaging System and Odyssey V3.0 software (LI-COR Bio-sciences).

HEK293T cells were transfected with Flag-RNF168 and Flag-pcDNA plasmids for 48 hours, and then lysed in RIPA buffer, sonicated using a JY92-II sonicator, and pre-cleared with Protein G Sepharose beads. The supernatants were first incubated overnight with an anti-Flag antibody and then with 50 μL Protein G Sepharose beads for a further 4 hours. Following centrifugation at 3,000 rpm, 4°C for 5 minutes, the supernatants were completely removed and Flag-peptide was added to competitively bind to the Protein G Sepharose beads. RNF168 and its associated proteins were released, precipitated using TCA, and subjected to mass spectrometry (MS) analysis. The proteomics data have been deposited in the ProteomeXchange Consortium via the PRIDE partner repository, with the dataset identifier PXD036815.

Two guide RNAs were designed against sequences conserved across the SENP1 or RNF168 locus using the Brunello human CRISPR-Cas9 screening library. The following guides were selected:

SENP1-gRNA1: aagatagggaatatacactgtgg

SENP1-gRNA2: gtgtaaaggaaaatgtgtggtgg

RNF168-gRNA1: cagcgtgtggttacacgggaggg

RNF168-gRNA2: agaaattctctcgtcaacgtgg

In brief, lentiviral vector particles were generated by transient transfection of HEK293T cells with gRNA (cloned into the LentiCRISPR V2 vector) and the packaging plasmids (PMD2.G and PSPAX2) using Lipofectamine 3000 (Invitrogen). Lentiviral vector supernatants were collected on days 2 and 3 post-transfection and used for knockout in HEK293T/HCT116 cells. SENP1- and RNF168-knockout clones were grown from single cells obtained by flow cytometry, and the knockout efficiency of each positive clone was verified by immunoblotting.

Seven shRNA sequences were synthesized and cloned into the pLKO.1 vector:

#1: ccggccgaaagacctcaagtggattctcgagaatccacttgaggtctttcggtttttg

#2: ccggtactggaactaagacatcgagctcgagctcgatgtcttagttccagtatttttg

#3: ccggaagacatcgagacaggtgtaaaggactcgagtcctttacacctgtctcgatgtctttttttg

#4: ccggctcgatgtcttagttccagtactcgagtactggaactaagacatcgagttttt

#5: ccggcgagaaagattgcgccagattctcgagaatctggcgcaatctttctcgttttt

#6: ccggtgaccattacacgcaaagatactcgagtatctttgcgtgtaatggtcattttt

shRNA168: ccggcgacactttctccacagatatctcgagatatctgtggagaaagtgtcgtttttg

Lentiviral vector particles were generated by transient transfection of HEK293T cells with the lentiviral packaging components (pLP1, pLP2, pVSVG). Supernatants were collected on days 2 and 3 and used to infect target cells. Knockdown cells were screened using puromycin.

Neutral comet assays were performed to detect cell genome stability using the CometAssay Kit (Trevigen). Cells were harvested and resuspended in ice-cold PBS buffer. The neutral single-cell agarose gel was placed in specific running buffer to let DNA helix uncoiling and the level of DNA damage was visualized by electrophoresis under alkaline conditions. Images were obtained using Olympus Ix73 fluorescence microscope with a 10× objective lens. Quantitation was conducted using the Casp Lab software v1.2.2 (University of Wroclaw, Wroclaw, Poland). Approximately 80 cells were analyzed in each group using the Comet Assay IV software (Perceptive Instruments).

The DNA repair efficiency assays were conducted as previously described (34). For the NHEJ assays, cells were transfected with the linearized pEGFP-Pem1-Ad2 plasmids and DsRed plasmid for 48 hours, and then harvested and washed with 1× PBS buffer. For the HR assays, cells were cotransfected with DR-GFP, an I-SceI expression vector, and a DsRed plasmid, and then harvested at 48 hours post-transfection and washed with 1× PBS buffer.

The intensity of green (EGFP) and red (DsRed) fluorescence signals was measured by FACS using a LSRFortessa or FACSVerse instrument (BD Biosciences). The percentage of EGFP/DsRed double-positive cells to DsRed-positive cells was taken as the NHEJ or HR repair efficiency. The results were normalized to those of the RNF168 wild-type cells. All the samples were analyzed using the FlowJo software to determine the number of GFP-positive cells relative to that of cells expressing DsRed.

Paraffin-embedded human colon normal and cancer samples were kindly provided by Xiuqin Zhang (Peking University). Slides were stained with rabbit anti-SENP1 (1:100) at 4°C overnight and then with secondary antibodies. Visualization was carried out using DAB (3′3′-diaminobenzidine tetrahydrochloride) as the substrate. The intensity of SENP1 was classified as follows: 1, no staining; 2, weak reactivity; 3, moderate reactivity; 4, strong reactivity; 5, very strong reactivity. A Leica DM IRE2 microscope was used to obtain the images.

The apoptosis assay was performed using the Annexin V-FITC Apoptosis Detection kit (Beyotime Biotechnology). Cells were collected and stained with Annexin V and propidium iodide (PI), and the apoptosis rate was detected using a LSRFortessa or FACSVerse instrument (BD Biosciences). The Q3 quadrant represented Annexin V–positive cells, which are in the early stage of apoptosis. The Q2 quadrant showed cells positive for both Annexin V and PI, which are in the late stage of apoptosis. A total of 1 × 10⁴ cells were counted for each group.

Cells were seeded onto Transwell inserts (Corning) containing Matrigel-coated (BD Biosciences) porous membranes and incubated for 24 hours. Using cotton swabs to remove the cells remaining on the insert, while using methanol to fix cells adhering to the lower side of the membrane. Then crystal violet staining was conducted, and observed by microscopy.

A Duolink In Situ PLA Kit (DUO92101, Sigma-Aldrich) was used to perform the proximity ligation assay (PLA) assay according to the manufacturer's instructions as previously described (34). In brief, cells were fixed using 4% paraformaldehyde for 20 minutes at room temperature, permeabilized with 0.5% Triton X-100, and blocked with 1× blocking solution buffer. Subsequently, cells were incubated at 4°C overnight with primary antibodies targeting RNF168 and SUMO3 and then with PLA PLUS and MINUS PLA probes at 37°C for 1 hour. After washing three times, the ligation-ligase solution was added, and cells were incubated at 37°C for 30 minutes, followed by incubating with amplification polymerase solution at 37°C in the dark for 100 minutes, after which, cells were stained with Mounting Medium containing DAPI. Fluorescence images were obtained under a confocal laser scanning microscope (Zeiss LSM 880) using a 63× oil objective lens.

HEK293T cells were transfected with HA-tagged SUMOylation E3 ligase PIAS1, which was subsequently enriched using an anti-HA antibody. The in vitro SUMOylation assay was conducted following previously described procedures (34). Briefly, His-tagged wild-type or mutant RNF168 proteins were diluted in in vitro SUMOylation buffer, immunoprecipitated with Mg-ATP, SUMO E1, SUMO E2, SUMO3, and Protein G Sepharose bound to E3 ligase. The mixture was incubated for 2 hours at 37°C following the manufacturer's manual from Abcam.

Fluorescence recovery after photobleaching (FRAP) experiments were performed using a Zeiss LSM 880 microscope with a 63× oil immersion objective. EGFP-RNF168-SUMO3 bodies (in vivo) were bleached until the fluorescence intensity was reduced to 30% of the initial value. After photobleaching, recovery time-lapse images were captured for the indicated times. Following a previously described method (29), the pre-bleaching fluorescence intensity was set to 100% and the fluorescence intensity at each time point was normalized (I_t) and used to calculate the fluorescence recovery according to the following formula: FR(t) = I_t/I_{pre-bleaching}. Image J was used for quantitation, and GraphPad Prism was used to plot and analyze FRAP data.

Proteins at the indicated concentration were incubated in droplet formation buffer (150 mmol/L NaCl, 50 mmol/L Tris-HCl pH 7.5, 10% glycerol, 1 mmol/L DTT, 10% PEG8000) for 2 minutes as reported previously (16). The in vitro droplet formation assay was carried out on glass-bottomed Petri dishes and observed using a Zeiss LSM 880 microscope with a 63× oil immersion objective.

Cells transiently expressing the indicated plasmids were lysed in NP-40 lysis buffer, incubated at 4°C for 30 minutes, and centrifuged at 18,000 × g for a further 30 minutes. The supernatant was removed and neutralized with 1 mol/L Tris-HCl (pH 8.0) as the soluble fraction. The pellet was washed with ddH₂O and resuspended in 0.2 mol/L HCl as the nuclear fraction. Samples were denatured in 6× SDS loading buffer, subjected to SDS-PAGE, and detected by immunoblotting.

HeLa cells were cultured on glass-bottomed Petri dishes and transfected with the indicated plasmids. The Micropoint system (Andor) was directly coupled to the epifluorescence path of the Nikon A1 confocal imaging system, and time-lapse images were captured at the indicated times.

MEF cells were treated with 1 μmol/L nocodazole for 12 hours to arrest mitosis at G₂–M phase. Cells were collected and resuspended in 0.06 mol/L KCl, incubated at 37°C for 30 minutes, and then centrifuged at 1,000 rpm for 5 minutes. The supernatant was removed and the pellet was fixed in freshly prepared fixation buffer (methyl alcohol: glacial acetic acid = 3:1) for 10 minutes. The cell suspension was extracted, dropped onto a slide, and dried at 60°C. Images were taken using a Zeiss LSM 880 microscope with a 63× oil immersion objective.

A CCK-8 kit (Dojindo Molecular Technologies Inc) was used to perform the cell proliferation assay according to the manufacturer's protocol. Cell suspensions (100 μL/well) were seeded onto 96-well plates and allowed to adhere. During the process of cell growth, 10 μL CCK-8 solution was added to each well at different time points and the absorbance was measured at 450 nm using a microplate reader.

Cells were cultured to 100% confluence and a physical gap was then created using a p10 tip. The process of cell migration into the gap was monitored by photography at the indicated times.

Cells were cultured on 6-well plates at the indicated density and treated with different concentrations of chemotherapeutic drugs. After culture for 14 days, colony formation was observed by crystal violet staining and quantitated using the Image-J software.

Expression variance and survival rate curves were analyzed and constructed using the UALCAN portal and GEPIA2 website, respectively, with COAD-related data screened out from the database. SUMOylation sites were predicted by the SUMOplot Analysis Program, GPS-SUMO, JASSA v4, and the Ron Hay lab group websites. Sequence alignment was performed using the ClustalW website and depicted using the ENDscript/ESPript website.

Statistical results were acquired from at least three independent biological replicates. Figure legends explained each detailed n values. Almost all data included are presented as the mean ± SEM, unless otherwise specified. GraphPad Prism 8.0 software was employed to obtain P values using Student t test (two-tailed), Mann–Whitney test (for two group comparisons), or one-way ANOVA. *, P < 0.05; **, P < 0.01; ***, P < 0.001.

The data supporting this study are available from the corresponding author upon reasonable request. The MS proteomics data have been deposited in the ProteomeXchange Consortium, with the dataset identifier PXD036815. Source data are provided with this paper. The data analyzed in this study were obtained from The Cancer Genome Atlas (TCGA) database (Colon Adenocarcinoma TCGA data) at https://ualcan.path.uab.edu/index.html.

SUMOylation plays important roles in the regulation of MDC1, 53BP1, and BRCA1 functions in the DDR pathway (10, 13). A previous study suggested that RNF168 is a substrate of SUMOylation (35); however, the modification site and functional significance of RNF168 SUMOylation remain to be elucidated. To further characterize RNF168 SUMOylation, a His-SUMO pulldown assay was performed to evaluate the type of SUMOylation, which revealed that RNF168 was modified by both SUMO1 and SUMO3 (Fig. 1A). Because SUMO2/3 proteins are more abundant and responsible for environmental stress than SUMO1 (36, 37), we chose SUMO3 for further study. The SUMOylation of RNF168 by SUMO3 was also confirmed using denaturing immunoprecipitation (IP) assays (Supplementary Fig. S1A). More importantly, the SUMOylation of endogenous RNF168 was validated in HEK293T, HCT116, and SW480 cells using IP assays (Fig. 1B). Using His-SUMO pulldown, we further confirmed that RNF168 is mono-SUMOylated, with Myc-tagged chimeric plasmids containing one SUMO3, a homodimer of SUMO3, or a heterodimer of SUMO3 and SUMO1, and endogenous RNF168 as the controls (Supplementary Fig. S1B).

His-SUMO pulldown screening of the SUMO E3 ligases of RNF168 showed that PIAS1, PIAS3, and PIASy promoted RNF168 SUMOylation (Fig. 1C), and coimmunoprecipitation (co-IP) confirmed that these E3 ligases interacted with RNF168 (Fig. 1D). Among them, PIAS1 displayed the strongest effect; therefore, we used this E3 ligase in subsequent experiments. Furthermore, the in vitro SUMOylation assay demonstrated that RNF168 was SUMOylated by SUMO3 in the presence of PIAS1 (Supplementary Fig. S1C–S1E).

Lysine residues undergoing SUMOylation are typically found within the SUMO modification consensus motif ψKXE, in which ψ is a large hydrophobic residue and X is any residue. Prediction using SUMOplot (https://www.abcepta.com/sumoplot) and GPS-SUMO (http://sumosp.biocuckoo.org/) revealed three lysine residues of RNF168, K173, K203, and K210, as potential SUMOylation sites. Accordingly, we constructed three RNF168 mutants, K173R, K203R, and K210R, and examined their effects on RNF168 SUMOylation using IP assays. In comparison with wild-type RNF168, only the K210R mutation notably reduced the SUMOylation level of RNF168 (Fig. 1E). Similarly, an in vitro SUMOylation assay using purified proteins (Supplementary Fig. S1E) showed that the SUMOylation level of the RNF168 K210R mutant is markedly decreased as compared with that of wild-type RNF168 (Fig. 1F). Furthermore, sequence alignment revealed that K210 is conserved among diverse species (Supplementary Fig. S1F). Collectively, these results demonstrate that RNF168 is mono-SUMOylated predominantly at K210.

To explore the impact of SUMOylation on RNF168, following previous studies that have used a chimera to simulate stable modifications (38–40), we constructed a GFP-tagged chimera RNF168-SUMO3 to mimic the SUMOylated form of RNF168 and conducted immunofluorescence (IF) assays using HeLa cells expressing different plasmids (Fig. 2A). Strikingly, microscopic visualization of fluorescently tagged RNF168-SUMO3 chimeric proteins revealed that SUMOylation induced RNF168 condensation, while the SUMOylation-defective K210R mutant did not yield puncta (Fig. 2A; Supplementary Fig. S2A). Because intermolecular interactions can be weakened or enhanced by the modification of key residues (41), we hypothesized that SUMOylation may play a key role in regulating the formation of RNF168 nuclear condensates through LLPS. To test this hypothesis, we characterized the puncta using 3D Z-stack scanning and 3D shape rendering and calculated the volume and surface area using the sphericity formula. The results showed that the puncta of SUMOylated RNF168 had a spherical shape, with a diameter of approximately 1.86 μm (Fig. 2B). Moreover, we conducted a FRAP assay in RNF168-SUMO3-expressing HeLa cells to investigate whether SUMOylation also affects the dynamics of RNF168 puncta in vivo. The fluorescence intensity of the SUMOylation-mimetic RNF168-SUMO3 protein was almost completely recovered within 1 minute post-bleaching, reflecting that SUMOylated RNF168 underwent rapid exchange with the surrounding environment (Fig. 2C and D). Subsequent tracing of the fusion and fission processes by live cell imaging revealed the occurrence of liquid-like properties (Fig. 2E; Supplementary Video 1, 2).

To confirm the LLPS of SUMOylated RNF168, we performed an in vitro droplet formation assay using purified SUMOylated and unSUMOylated RNF168 proteins according to the procedure shown in Fig. 2F. The droplets were only observed in the presence of SUMOylated RNF168 and not within the RNF168 and SUMO3 protein mixture (Fig. 2G). Consistent with the observations in Fig. 2A, the droplet number of the RNF168 K210R SUMOylation-deficient mutant was largely decreased relative to that of the SUMOylated wild-type RNF168 (Fig. 2H), suggesting that SUMOylation of RNF168 is essential for LLPS formation. In addition, because proteins that undergo LLPS frequently contain intrinsically disordered regions (IDR; ref. 27), we employed AlphaFold and the IUPred2 website to identify a long IDR sequence in RNF168 that includes the SUMOylation site Lys²¹⁰ (Supplementary Fig. S2B and S2C). Taken together, these results indicate that SUMOylated RNF168 undergoes LLPS.

RNF168 is an early response protein in the DSB repair pathway, and RNF168-catalyzed H2A K13/15 ubiquitination (6) is critical for 53BP1 accumulation (7). To evaluate the effect of SUMOylation on RNF168 function, laser microirradiation was used to generate localized DNA damage in HeLa cells expressing GFP-RNF168 or GFP-RNF168-SUMO3, and the recruitment of RNF168 to the damaged site was monitored. Notably, wild-type RNF168 was quickly recruited to the damaged site, whereas the SUMOylated form was not (Fig. 3A). Next, we constructed RNF168^−/− HEK293T and HCT116 cells using the CRISPR/Cas9 system (Supplementary Fig. S3A and S3B) and re-expressed wild-type RNF168, RNF168-SUMO3, and RNF168 K210R to evaluate their interactions with H2A in response to etoposide treatment-induced DNA damage. As expected, relative to wild-type RNF168, the SUMOylated RNF168 showed a weaker interaction with H2A, while the SUMOylation-deficient mutant K210R showed a stronger H2A interaction (Fig. 3B and C; Supplementary Fig. S3C and S3D). Consistently, the ubiquitin pulldown and acid chromatin fractionation assays detected no H2A ubiquitination in cells without etoposide treatment, and the level of H2A ubiquitination decreased in cells expressing RNF168-SUMO3 and increased in those expressing the K210R mutant in response to DNA damage (Fig. 3D and E). These results indicate that SUMOylation inhibits RNF168 function in the DDR by reducing its recruitment to the damaged site and attenuating its function as an E3 ubiquitin ligase of H2A.

It is well known that 53BP1 and BRCA1 play pivotal roles in NHEJ and HR repair; therefore, we used IF to elucidate whether SUMOylation of RNF168 further affects the downstream repair proteins 53BP1 and BRCA1. Notably, endogenous 53BP1, but not BRCA1, colocalized with the RNF168-SUMO3 condensates (Fig. 3F; Supplementary Fig. S3E). In addition, super-resolution microscopy confirmed the colocalization of RNF168-SUMO3 with 53BP1 (Fig. 3G; Supplementary Fig. S3F). These observations suggest that SUMOylated RNF168 restrains 53BP1 within the phase-separated puncta. Moreover, we found that RNF168 knockdown HCT116 cells have a significantly reduced repair efficiency of NHEJ and HR. The NHEJ repair efficiency could be rescued by the reintroduction of wild-type RNF168 and was even promoted by the K210R mutant but not by RNF168-SUMO3 (Fig. 3H; Supplementary Fig. S3G), while the HR repair efficiency was similarly recovered by wild-type RNF168, the K210R mutant, and RNF168-SUMO3 (Supplementary Fig. S3H), suggesting that RNF168 SUMOylation has no significant impact on HR repair. In accordance with these data, the recruitment of 53BP1 to chromatin was notably reduced in etoposide-treated RNF168^−/− HEK293T cells expressing the RNF168-SUMO3 chimera (Fig. 3I). Furthermore, we examined whether RNF168 SUMOylation affected the colocalization of 53BP1 with γH2AX (a damage marker) using IF assays, which revealed that SUMOylated RNF168-induced puncta are larger than the DNA damage-induced foci containing wild-type RNF168 (Supplementary Fig. S3I). In etoposide-treated RNF168-SUMO3-expressing cells, SUMOylation-induced nuclear condensates of RNF168 compartmentalized 53BP1, resulting in reduced colocalization of 53BP1 with γH2AX; however, in cells harboring RNF168, 53BP1 was not restrained in the RNF168 puncta and colocalized with γH2AX (Fig. 3J). Taken together, these data demonstrate that SUMOylation drives RNF168 phase separation, which reduces RNF168 recruitment to the damaged site, inhibits H2A ubiquitination, and sequesters 53BP1 in the droplets, ultimately downregulating NHEJ repair efficiency.

Because our data indicate that SUMOylation negatively regulates the function of RNF168 in the DDR, we subsequently explored the regulatory mechanism underlying RNF168 SUMOylation. Considering that SENPs are deSUMOylating enzymes that reverse SUMOylation and are critical for the regulation of many protein functions, we sought to identify the specific SENP responsible for the regulation of RNF168 SUMOylation. IP and subsequent MS were conducted, which identified SENP1 and SENP3 as potential partners of RNF168 (Fig. 4A). Co-IP assays demonstrated an association between endogenous and exogenous RNF168 and SENP1 (Fig. 4B; Supplementary Fig. S4A), while an interaction between RNF168 and SENP3 was not detected even following overexpression (Supplementary Fig. S4B). Furthermore, the His-SUMO pulldown assay showed that SENP1 clearly reduced RNF168 SUMOylation (Fig. 4C), while SENP1 knockdown in HEK293T, HCT116, and SW480 cells increased RNF168 SUMOylation (Fig. 4D–F; Supplementary Fig. S4C and S4D). To clarify whether SENP1 decreases RNF168 SUMOylation in a deSUMOylase activity-dependent manner, we constructed a SENP1 C603A mutant in which the catalytic activity was abolished. Indeed, the level of RNF168 SUMOylation was substantially reduced by wild-type SENP1; however, the C603A mutant had significantly less effect (Fig. 4D). In addition, SENP1 attenuated the interaction between RNF168 and its E3 ligase PIAS1 (Fig. 4G). These findings indicate that the ability of SENP1 to reduce RNF168 SUMOylation relies not only on its catalytic activity but also on its ability to impede the interaction between RNF168 and PIAS1. Taken together, these data suggest that SENP1 is the specific enzyme responsible for RNF168 deSUMOylation (Fig. 4H).

SUMOylation is a reversible and dynamic PTM that plays critical roles in cellular processes. Considering that SENP1 specifically deSUMOylates RNF168 and that SUMOylation negatively regulates the damage repair function of RNF168, we investigated whether SENP1 responds to DNA damage to dynamically regulate the function of RNF168 in the DDR. We performed a laser microirradiation assay using EGFP-SENP1-expressing cells to examine whether SENP1 is recruited to the damaged site. Notably, SENP1 was rapidly recruited to the damaged sites (within 10 sec after laser treatment; Fig. 5A), and IF assays revealed that SENP1 formed foci in the nucleus in response to bleomycin, ionizing radiation, or etoposide treatment (Supplementary Fig. S5A).

To identify the upstream regulator of SENP1 recruitment, we treated cells with inhibitors of ATM (KU55933), DNA-PK (NU7026), ATR (VE821), and PARP (olaparib) and detected SENP1 recruitment by laser microirradiation. The results showed that KU55933 treatment blocked the accumulation of SENP1, which indicates that ATM plays a critical role in the regulation of SENP1 recruitment (Fig. 5B). Because ATM is a kinase, we next explored whether SENP1 interacts with ATM and undergoes phosphorylation. Indeed, both the interaction between SENP1 and ATM and the phosphorylation level of SENP1 were markedly increased in cells treated with etoposide in comparison with untreated cells (Supplementary Fig. S5B), and the level of SENP1 phosphorylation reduced in cells treated with ATM inhibitor (KU55933; Supplementary Fig. S5C), suggesting that SENP1 is rapidly recruited to DNA damage sites in an ATM-dependent manner. Consistently, we found that etoposide treatment promoted the association between SENP1 and RNF168 (Fig. 5C), and the level of endogenous RNF168 SUMOylation decreased after etoposide treatment (Fig. 5D).

Because our results showed that SUMOylation promotes the LLPS of RNF168 and that SENP1 deSUMOylates RNF168, we compared the puncta of endogenous RNF168 in wild-type cells with those in SENP1^−/− cells using IF to further characterize the physiological effect of SENP1 on the LLPS formation of endogenous RNF168. Significant endogenous RNF168 LLPS was observed in both HEK293T and HCT116 SENP1 knocked out but not wild-type cells (Fig. 5E and F). Consistently, knockdown of SENP1 reduced the NHEJ repair efficiency (Fig. 5G). Moreover, we assessed the effects of SENP1 depletion on genomic integrity by performing neutral comet and metaphase spread assays. Relative to wild-type SENP1, SENP1 knockdown or knockout induced longer comet tails and a higher frequency of chromosomal breakage and fusion events (Fig. 5H and I). These data suggest that SENP1 is involved in the DNA damage repair pathway by interacting with RNF168 and downregulating RNF168 SUMOylation, thereby playing an essential role in the maintenance of genomic integrity.

Because our results indicate that SENP1 is critical for genomic stability maintenance, we next explored the regulatory functions of SENP1 in cancer cells. We examined the effect of SENP1 on cell growth by generating both SENP1 knockout and knockdown HCT116 cells and verifying SENP1 depletion (Fig. 6A). SENP1 knockdown HCT116 cells showed slower growth rates than those of wild-type cells (Fig. 6B). Furthermore, we investigated the effect of SENP1 depletion on the sensitivity of cancer cells to chemotherapy drugs using SENP1 wild-type and SENP1^−/− HCT116 cells treated with etoposide. Knockout of SENP1 resulted in a higher apoptosis rate than that of wild-type cells (Fig. 6C). Consistent with these observations, relative to wild-type HCT116 cells, SENP1 knockdown inhibited the migration and invasion abilities of cancer cells (Fig. 6D and E). Together, these results indicate that SENP1 plays important roles in regulating cell proliferation and migration, and dysfunction of SENP1 renders cancer cells sensitive to DNA damaging agents.

SENP1 plays important roles in maintaining genomic integrity and regulating cell status when exposed to chemotherapeutic drugs. To explore the biological significance of SENP1, we first investigated the relationship between SENP1 and human cancers. Bioinformatics analyses revealed that SENP1 is highly expressed in tumor tissue of COAD patients (Fig. 7A). Consistent with these analyses, IHC staining confirmed the higher abundance of SENP1 in COAD tissues than that in adjacent normal tissues (Fig. 7B). And higher expression of SENP1 were observed in the human COAD cell lines HCT116, SW480, LOVO, HT29, and RKO, than in the normal colon cell lines NCM460 and FHC (Fig. 7C). We further validated the effect of SENP1 level on phase separation of endogenous RNF168 by IF assays using cancer cell HCT116 and normal cell FHC that displayed different expression level of endogenous SENP1. As expected, condensation of endogenous RNF168 was observed obviously in FHC cells with very low level of SENP1 and HCT116 cells with knocked out SENP1, but not in wild-type HCT116 that expressing high level of SENP1 (Fig. 7D), suggesting that the level of endogenous SENP1 plays a critical role in regulation of the formation of RNF168 LLPS under physiological condition. In addition, 53BP1 was also colocalized with endogenous RNF168 condensates in FHC cells (Supplementary Fig. S6A). Moreover, PLA experiments were performed to compare the levels of endogenous RNF168 SUMOylation in cancer tissues and adjacent normal tissues. As expected, COAD tissues exhibited a weaker association between RNF168 and SUMO3 relative to adjacent normal tissues (Fig. 7E), indicating that the overall RNF168 SUMOylation level decreased in colon cancer. In addition, in situ PLA assays showed that SENP1 knockdown resulted in condensation of SUMOylated RNF168 (Fig. 7F), demonstrating that SENP1 knockdown increased SUMOylation of RNF168.

The clinical effectiveness of chemotherapy is limited by drug resistance, which is the major cause of treatment failure and mortality in patients with colon cancer (42). Because deSUMOylation of RNF168 plays a protective role by promoting NHEJ repair in the cellular response to DNA damage, we therefore speculated that the decreased RNF168 SUMOylation level caused by high SENP1 levels in cancer tissues might contribute to drug resistance. To verify this hypothesis, HCT116 RNF168 knockout cells transfected with HA-RNF168 or HA-RNF168-K210R were treated with the clinical COAD medicine oxaliplatin and subjected to apoptosis analyses. RNF168 knockout promoted the rate of apoptosis; reintroduction of the SUMOylation-defective K210R mutant led to a lower rate of apoptosis relative to RNF168^−/− cells rescued with RNF168 wild-type (Fig. 7G). In keeping with these results, we generated HCT116 RNF168^−/− cells stably expressing the lentivirus transfer vector pLV04-RNF168 or pLV04-RNF168 K210R (Fig. 7H) and performed a colony formation assay. After treatment with different concentrations of SN38 (a metabolite of the COAD chemotherapeutic drug irinotecan) or oxaliplatin, HCT116 RNF168^−/− cells expressing the SUMOylation-defective K210R mutant displayed a higher survival rate than those reintroduced with wild-type RNF168 (Fig. 7I; Supplementary Fig. S6B and S6C). These results suggest that cells with lower levels of RNF168 SUMOylation are more resistant to chemotherapy drugs.

To elucidate the physiological effect of RNF168 SUMOylation in mice, HCT116 wild-type cells, HCT116 RNF168^−/−cells, HCT116 SENP1^−/− cells, and RNF168^−/− HCT116 cells with reintroduced SUMOylation-defective K210R or wild-type RNF168 were subcutaneously injected into the armpits of 4-week-old female BALB/c nude mice. After inoculation, we observed tumor volume every two days until the tumor sizes approached approximately 1,000 mm³. As expected, RNF168^−/− and SENP1^−/−, which strongly disturb genomic integrity, inhibited tumor growth as well; the tumor size in the SENP1^−/− group was notably smaller than the wild-type group. Surprisingly, the RNF168^−/− group failed to form tumors, but this failure was rescued by reintroduction of RNF168 wild-type and the K210R mutant. As expected, rescue with the SUMOylation-deficient mutant K210R, which promotes NHEJ repair, showed larger tumors than the wild-type RNF168 group. In addition, the levels of DNA damage marker γH2AX were assessed in the tumor tissues, which showed that smaller tumor correlated with more extensive damage (Fig. 7J).

Consequently, analyses of the correlation between SENP1 and COAD prognosis using TCGA databases revealed that COAD patients with higher SENP1 expression levels had lower survival (Supplementary Fig. S6D). Furthermore, other types of cancer such as mesothelioma (MESO), liver hepatocellular carcinoma (LIHC), kidney renal papillary cell carcinoma (KIRP), and adrenocortical carcinoma (ACC) showed a similar survival tendency as the COAD patients (Supplementary Fig. S6E). Collectively, these data suggest that a high level of SENP1 in cancer tissues affects chemotherapy drug efficacy and cancer prognosis through downregulation of RNF168 SUMOylation.

The DDR is fine-tuned by many types of posttranslational protein modifications. In the present study, we identified that PIAS1 promotes SUMOylation and SENP1 deSUMOylates RNF168 at Lys²¹⁰. More importantly, SUMOylated RNF168 undergoes LLPS in the nucleus, which blocks the recruitment of RNF168 to the damaged site, restrains the RNF168-H2A interaction and H2A ubiquitination, compartmentalizes 53BP1 in the condensates, and ultimately downregulates NHEJ repair efficiency. In response to DNA damage, SENP1 is quickly recruited to the damaged sites where it decreases RNF168 SUMOylation, which disrupts RNF168 LLPS and facilitates damage repair. SENP1 is upregulated in COAD, which renders cancer cells insensitive to chemotherapy drugs through its inhibition of RNF168 SUMOylation-mediated LLPS, thus causing drug resistance.

Previous studies have reported the crucial role of RNF168 in the DSB repair pathway and in human diseases; dysfunction of RNF168 leads to radiosensitivity, immunodeficiency, dysmorphic features, and learning difficulties, known as the RIDDLE syndrome (1, 43). In the DDR pathway, RNF168 ubiquitinates histone H2A at K13/15 (6), and ubiquitinated H2A serve as a scaffold for recruitment of downstream repair proteins such as 53BP1 (7). In addition, RNF168 itself is a substrate of ubiquitination and SUMOylation (5, 8, 35). SUMO1 has been suggested to modify RNF168 (35), and RNF168 has been shown to associate with proteins modified by SUMO2 and/or SUMO3 (44). However, the functional effects of RNF168 SUMOylation remain largely unknown. Here, we confirmed that RNF168 is mono-SUMOylated and demonstrated that RNF168 SUMOylation had a significant function in the regulation of DNA damage repair by undergoing LLPS.

Because the germline P granule has been reported to be the first phase-separated condensate, many LLPS-formed membraneless organelles, including nucleoli, Cajal bodies, centromeres, promyelocytic leukemia nuclear bodies, and stress granules, have been reported to play important roles in regulating many biological processes (17, 21, 45). In the DDR pathway, many proteins undergo LLPS as well, which regulate the efficiency of DNA repair. For example, phase separation of the RNA binding protein FUS is sufficient for repair initiation (46), and TopBP1 assembles biomolecule condensates and triggers the checkpoint in response to DNA replication impediments (47). Many factors, especially PTMs, are involved in the regulation of LLPS under different conditions. Poly-ubiquitination-modulated p62 phase separation drives the segregation of autophagic cargo (21). Dephosphorylation and SUMOylation are required for the formation of stress-induced NELF condensates (48). In addition, SUMOylation also coordinates the disassembly of the stress granules that contributes to amyotrophic lateral sclerosis pathogenesis (18). Here, we demonstrate that RNF168 undergoes LLPS, which is upregulated by SUMOylation. More importantly, we identified 53BP1 as a component of RNF168 condensates (Fig. 3F and G, Supplementary Fig. S3E and S3F), which is consistent with previous observations that RNF168 is associated with 53BP1 (7) and that 53BP1 shows phase separation properties (30, 31). It is noteworthy that RNF168 LLPS downregulates DNA damage repair by blocking the recruitment of RNF168 to damaged sites and compartmentalizing the downstream key repair protein 53BP1 as well, elucidating a novel mechanism underlying PTM-mediated LLPS in the DDR pathway.

DeSUMOylation enzyme SENPs are well known to balance protein levels between their SUMOylated and un-SUMOylated forms. The aberrant expression of SENPs is correlated with cancer development and progression. In response to fasting, SENP1 regulates mitochondrial metabolism by deSUMOylating Sirt3 (49). SENP2 inhibits NF-κB activity by reducing NEMO SUMOylation (15). Furthermore, mitotic SENP3 activation promotes the innate immune response in tumor cells (50). In our study, we demonstrate that in response to DNA damage, SENP1 is recruited to DNA damage sites in an ATM-dependent manner to promote NHEJ by deSUMOylating RNF168; however, whether SENP1 is directly phosphorylated by ATM and the identity of the phosphorylation site require further investigation. More importantly, the level of endogenous SENP1 plays critical role in regulation of the LLPS of RNF168 and NHEJ. We found that SENP1 is highly expressed in COAD patients who showed a low level of RNF168 SUMOylation, reduced RNF168 LLPS, promoted NHEJ efficiency, and experienced chemotherapeutic drug resistance (Fig. 7). Consistently, we demonstrated that COAD cells with depleted SENP1 were more sensitive to chemotherapy drugs and showed decreased abilities of proliferation and migration, thus leading to effective treatment. Using cell-derived xenograft (CDX) nude mice models, we further proved that knockout of SENP1 and RNF168 significantly inhibited tumor growth, which could be rescued by introduction of the RNF168 SUMOylation-deficient mutant K210R. Meanwhile, analysis of the TCGA database revealed that the expression level of SENP1 is negatively correlated with survival rate of many types of cancers including MESO, LIHC, KIRP, and ACC, suggesting that SENP1 affects cancer treatment efficacy by modulating RNF168 SUMOylation.

In summary, we identified that RNF168 SUMOylation impeded RNF168 function in DNA damage repair through LLPS. The formation of reversible functional micrometer-sized RNF168 condensates by SENP1-modulated SUMOylation in response to DNA damage provides insight into a novel molecular mechanism underlying DDR regulation. Depletion of SENP1 renders cancer cells sensitive to DNA damaging drugs, suggesting a potential strategy for enhancing the efficacy of chemotherapy.

No disclosures were reported.

M. Wei: Conceptualization, data curation, formal analysis, investigation, visualization, methodology, writing–original draft. X. Huang: Investigation. L. Liao: Investigation. Y. Tian: Investigation. X. Zheng: Conceptualization, resources, supervision, funding acquisition, writing–original draft, project administration, writing–review and editing.

The authors sincerely thank Prof. Jinke Cheng for providing the human SUMO1, SUMO2/3, and SENP1 plasmids. The authors are grateful to the National Center for Protein Sciences at Peking University, particularly Guilan Li, Yinghua Guo, and Liying Du, for technical help. They also appreciate the assistance of Ye Liang from the Core Facility of the State Key Laboratory of Membrane Biology at Peking University and Siying Qin, Dong Liu, and Liqin Fu from the Core Facilities of Life Sciences at Peking University for their assistance with the protein MS analysis and microscopy imaging. This work was supported by the National Key R&D Program of China (2022YFA1302803), the National Natural Science Foundation of China (82130081, 32270756), and Natural Science Foundation of Beijing Municipality (5212008).

The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.