The RNA-binding Protein Fused in Sarcoma (FUS) Functions Downstream of Poly(ADP-ribose) Polymerase (PARP) in Response to DNA Damage

Background: FUS has been implicated in the DNA damage response; however, the mechanisms are unknown.

Results: FUS recruitment to DNA lesions is PARP-dependent. Depletion of FUS disrupts DNA repair.

Conclusion: FUS functions downstream of PARP and promotes double-strand break repair.

Significance: This work identifies FUS as a novel factor at DNA lesions and furthers our understanding of RNA-binding proteins in maintaining genomic stability.

Abstract

The list of factors that participate in the DNA damage response to maintain genomic stability has expanded significantly to include a role for proteins involved in RNA processing. Here, we provide evidence that the RNA-binding protein fused in sarcoma/translocated in liposarcoma (FUS) is a novel component of the DNA damage response. We demonstrate that FUS is rapidly recruited to sites of laser-induced DNA double-strand breaks (DSBs) in a manner that requires poly(ADP-ribose) (PAR) polymerase activity, but is independent of ataxia-telangiectasia mutated kinase function. FUS recruitment is mediated by the arginine/glycine-rich domains, which interact directly with PAR. In addition, we identify a role for the prion-like domain in promoting accumulation of FUS at sites of DNA damage. Finally, depletion of FUS diminished DSB repair through both homologous recombination and nonhomologous end-joining, implicating FUS as an upstream participant in both pathways. These results identify FUS as a new factor in the immediate response to DSBs that functions downstream of PAR polymerase to preserve genomic integrity.

Introduction

Exposure to genotoxic agents including ionizing radiation (IR),² H₂O₂, and radiomimetic drugs poses a significant challenge to genomic integrity that is combated by evolutionarily conserved pathways that are collectively referred to as the DNA damage response (DDR) (1, 2). Central to this paradigm is the activation of apical phosphoinositide 3-kinase-like kinases, including ATM, ATM and Rad3-related (ATR), and DNA-dependent protein kinase (DNA-PK) that serve as proximal DNA damage signal transducers. The best characterized of these, ATM, is activated by double-strand breaks (DSBs) to phosphorylate an estimated 700 substrates (3) that impact cell cycle regulation, apoptosis, DNA repair, and many other cellular processes. ATM rapidly accumulates at sites of DSBs in an MRE11-RAD50-NBS1-dependent manner and phosphorylates the histone variant H2AX on serine 139 (γH2AX) (4, 5). γH2AX serves as a scaffold for the recruitment of the mediator protein MDC1 (6–8) and subsequent localization of the E3 ubiquitin ligases, RNF8 and RNF168 (9–13). RNF8 and RNF168 activities are required for the recruitment of the RAP80-ABRA1-BRCA1 complex (12, 13) and p53-binding protein 1 (53BP1) (10, 12–14). The composition of the recruitment complexes may dictate whether DSBs are repaired via NHEJ or HR repair mechanisms (15). Working in parallel to the γH2AX-mediated recruitment pathway is a PARP-dependent pathway that responds primarily to DNA single-strand breaks (SSBs) and participates in SSB repair and alternative-NHEJ (16–19). PARP ADP-ribosylates target proteins including histones and itself at sites of damage, which creates binding sites for proteins harboring PAR binding domains (20). PARP is required for the recruitment of CHD4-NuRD and Polycomb group transcriptional repressor complexes, which mediate histone deacetylation and chromatin compaction near the break site, presumably to reduce interference between transcription and DSB repair (21–24).

FUS is a 526-amino acid member of the FET family of RBPs, which include Ewing sarcoma (EWSR1), Tata-binding protein-associated factor 2N (TAF15), and the Drosophila ortholog of FUS, SARFH/Cabeza (25, 26). FUS was initially identified as a fusion oncogene in myxoid liposarcoma, in which the transcriptional activation domain of FUS is fused to the C/EBP homologous protein (CHOP) (27, 28). In addition, FUS fusion proteins have been identified in a variety of human cancers including acute myeloid leukemia, angiomatoid fibrous histiocytoma, and fibromyxoid sarcoma (29). FUS is composed of N-terminal Gln/Gly/Ser/Tyr-rich and Gly-rich regions that comprise a prion-like domain (PLD) (30), an RNA recognition motif (RRM), arginine/glycine-rich (RGG) domains, and a C-terminal zinc finger domain (ZNF) (see Fig. 2A) (31). FUS binds RNA (32, 33) and ssDNA and dsDNA (33–35) and has been shown to shuttle between the nucleus and cytoplasm (33). In addition to its participation in transcription, FUS has proposed physiological activities involving microRNA processing, splicing, and mRNA transport and maturation (36, 37). Thus, FUS seems to fulfill broad functions in gene expression through transcriptional and posttranscriptional mechanisms. Intriguingly, dominant mutations in FUS cause inherited forms of amyotrophic lateral sclerosis (ALS) and frontotemporal lobar degeneration (38, 39). It is believed that ALS-associated mutations lead to trapping and aggregation of FUS in the cytoplasm; however, other pathogenic mechanisms may be at play.

FIGURE 2.

PARP-dependent recruitment of FUS to sites of DNA damage. A, domain architecture of FUS is shown. NLS, nuclear localization signal. Mutations in the RRM or ZNF of FUS do not disrupt targeting to DNA damage sites. B, U-2 OS cells expressing the indicated GFP-FUS constructs were laser-microirradiated and monitored by live cell imaging. C, ALS-associated mutations in FUS do not abrogate localization to LIDD. U-2 OS cells expressing GFP-Myc-FUS(WT), R521G, or R524S were laser-microirradiated and monitored by live cell imaging. D, FUS localization to sites of DNA damage is PARP-dependent. U-2 OS cells expressing GFP-FUS(WT), S42A, or FUS4A were laser-microirradiated, and recruitment was monitored by live cell imaging. Where indicated, cells expressing GFP-FUS(WT) were pretreated with ATM inhibitor, DNA-PK inhibitor, or PARP inhibitor for 1 h prior to damage. Shown are representative images from at least two independent experiments.

FUS also has an emergent, yet poorly understood participation in the DDR. FUS^−/− mice exhibit defects in B lymphocyte development and activation, male sterility, chromosomal instability, and radiosensitivity, phenotypes that are closely aligned with DSB repair defects (40, 41). Cellular extracts from FUS^−/− testes are unable to promote pairing between homologous DNA sequences in vitro (41), and FUS was shown to promote D-loop formation (34), an essential step in the repair of DSBs through the HR pathway (42). Thus, the available evidence suggests that FUS participates in HR repair, possibly through direct actions at DSBs.

FUS may also regulate the DDR through transcriptional mechanisms. Wang et al. showed that FUS is recruited to the cyclin D1 (CCND1) promoter through an interaction with sense and antisense noncoding CCND1 RNAs. Through inhibition of CREB-binding protein, FUS acts to repress CCND1 expression in response to DNA damage (43). Finally, the finding that FUS is directly phosphorylated by ATM on Ser-42 and possibly neighboring phosphoinositide 3-kinase-like kinase consensus motifs in response to DNA damage provides strong circumstantial support for its role in the DDR (44); however, the functional significance of these posttranslational modifications is uncertain.

Here, we demonstrate that FUS functions in cis to DNA damage via a PARP-dependent mechanism, which is mediated partially, if not completely, through binding of the RGG2 domain to PAR. In addition, we identify a role for the N-terminal PLD in accumulation of FUS at sites of DNA damage. Finally, FUS is required for efficient DNA repair through both HR and NHEJ pathways and for radioresistance. These data establish that FUS is dynamically regulated by DNA damage and functions downstream of PARP to maintain the stability of the genome.

EXPERIMENTAL PROCEDURES

Cell Culture, DNA Damage, and Drugs

U-2 OS and HEK-293T cell lines were obtained from the American Type Culture Collection (ATCC). The HEK-293 cell lines EJ5-GFP and DR-GFP were a kind gift from Dr. Jeremy Stark (Beckman Research Institute of the City of Hope) (45, 46). HEK-293T and HEK-293 cell lines were grown in Dulbecco's modified Eagle's medium (Cellgro). The U-2 OS cell line was grown in McCoy's medium. ATM (KU-55933) and PARP (PJ34) inhibitors (Calbiochem) were used at a final concentration of 20 μm and 1 μm, respectively. DNA-PK inhibitor (NU-7441) (R & D Systems) was used at a final concentration of 5 μm. All of the inhibitors were applied 1 h prior to subsequent analysis. IR was delivered using a JL Shepherd model JL-109 irradiator with a ¹³⁷Cs source at 4.03 grays/min.

Plasmids and Transfections

The GFP-Myc-FUS(WT), GFP-Myc-FUS(R521G) and GFP-Myc-FUS(R524S) plasmids were a kind gift from Dr. Robert Baloh (Cedars-Sinai Medical Center). The GFP-FUS plasmid was a kind gift from Dr. Lawrence Hayward (University of Massachusetts Medical School) and was used for the generation of the following mutants: S42A, S26A/S42A/S61A/S84A (FUS4A), and C428A/C444A/C447A. pCI-NEO FUS(4F-L) was a kind gift from Dr. Udai Pandey (Louisiana State University Health Sciences Center) and was cloned into pCDNA5/FRT/TO/GFP (Addgene plasmid 19444 (47). EWSR1 was cloned into pCDNA5/FRT/TO/GFP from pDEST/EWSR1 (Addgene plasmid 26377) (48). The mCherry and I-SceI plasmids were a kind gift from Dr. Sandra Weller (University of Connecticut Health Center). Point mutants were generated using primers designed by the QuikChange primer design program from Agilent Technologies. N-terminal truncation mutants (Δ1–285, Δ1–374, Δ1–467) were created using primers generating an internal start codon and were cloned into pCDNA5/FRT/TO/GFP. All primers were purchased from Integrated DNA Technologies (IDT). The internal deletion mutant, Δ204–475, was generated through the digestion of a full-length PCR product with BSPHI and EcoRI, which removes the RRM (33). The fragments flanking the RRM were purified and ligated to yield Δ204–475. Mutations were verified by direct sequencing. All transfections were performed using Lipofectamine 2000 (Invitrogen) following the manufacturer's instructions. Cells were analyzed 24–72 h after transfection.

Laser Microirradiation

U-2 OS cells were plated onto 35-mm glass-bottom dishes (MatTek Corporation). Cells were either transfected using the outlined procedure or used directly for microirradiation. Cells were presensitized with 10 μg/ml Hoechst 33342 (Invitrogen) for 20 min at 37 °C. Microirradiation was performed with an A1R confocal microscope (Nikon) equipped with a 37 °C heating stage and 405-nm laser diode focused through a 60× Plan APO VC/1.4 oil objective. All microirradiation was performed using a laser power output of 40%. Live cell imaging was monitored using NIS-Elements viewer software (Nikon). Quantification of images was performed using ImageJ (National Institutes of Health) software.

Lentivirus

The pLKO.1 system was used to package lentiviruses and deliver short hairpin RNA (shRNA). The following shRNA target sequences were designed using the RNAi Consortium online tool (Broad Institute) and were cloned into pLKO.1-TRC (Addgene plasmid 10878) (49), according to the manufacturer's suggestions: FUS Coding Sequence, 5′-ATGAATGCAACCAGTGTAAGG-3′; FUS 3′-UTR, 5′-CAATTCCTGATCACCCAAGGG-3′; CtIP, 5′-CGGCAGCAGAATCTTAAACTT-3′; Lig4, 5′-GCCCGTGAATATGATTGCTAT-3′. Addgene plasmid 1864 containing a nontargeting (NT) shRNA was used as the control (50). Lentiviral particles were produced by transient transfection of HEK-293T cells with pLKO.1, psPAX2 (Addgene plasmid 12260), and pMD2.G (Addgene plasmid 12259) in a ratio of 4:3:1. U-2 OS, HEK-293 EJ5-GFP, and HEK-293 DR-GFP cells were infected with lentivirus and maintained in 1.5 μg/ml puromycin.

Immunofluorescence Microscopy

Cells adhering to glass coverslips were washed with phosphate-buffered saline (PBS), fixed with 4% paraformaldehyde, and permeabilized with 0.2% Triton X-100 in PBS at room temperature. Where indicated, cells were preextracted with CSK buffer (10 mm HEPES, pH 7.4, 300 mm sucrose, 100 mm NaCl, 3 mm MgCl₂, 0.5% Triton X-100) for 2 min on ice. Cells were blocked in 3% BSA for 30 min at room temperature and then incubated overnight at 4 °C with the indicated antibodies including: EWSR1 (Millipore DR1063), γH2AX (Millipore 05636), FUS (Santa Cruz Biotechnology 47711), and MDC1 (Sigma HPA006915). Cells were then incubated with either Alexa Fluor 488 or Alexa Fluor 594 secondary antibodies (Invitrogen) for 1 h at room temperature. Images were captured using an A1R confocal microscope equipped with a 60× Plan APO VC/1.4 oil objective and processed using Image J software.

Western Blotting

U-2 OS, HEK-293 EJ5-GFP, and HEK-293 DR-GFP cells were washed with PBS and treated in ice-cold cell lysis buffer (50 mm Tris, pH 7.5, 300 mm NaCl, 10% glycerol, 0.5% Triton X-100, 2 mm MgCl₂, 3 mm EDTA) supplemented with protease and phosphatase inhibitors. Cell extracts were clarified by centrifugation at 20,000 × g for 10 min. Proteins were denatured and resolved by SDS-PAGE, transferred to polyvinylidene difluoride (PVDF) membranes, and probed with the indicated primary antibodies including: DNA-PKCs (Neomarker MS423), FUS (Santa Cruz 47711), GFP (Santa Cruz 9996), HA (Roche Applied Science 11667475001), HSP90 (Cell Signaling 4877), KU70 (Santa Cruz1487), Lig4 (Abcam 80514), and PAR (Trevigen 4336-BPC-100).

Quantitative Real-time PCR

Total RNA was prepared using TRIzol (Invitrogen) according to the manufacturer's protocol. The cDNA was generated from 2 μg of total RNA by reverse transcription using the iScript cDNA synthesis kit (Bio-Rad). Two percent of the synthesized cDNA reaction was subjected to reaction in a MyiQ real-time PCR (Bio-Rad) with FastStart DNA Master SYBR Green I (Roche Applied Science). CtIP primers used were: forward, 5′-AGACAGTTTCTCCCAAGCAG-3′ and reverse, 5′-CATCACCTTTAAAATACGGCTCC-3′. Primers were designed using SciTools Real-Time PCR Primer Design Tool (IDT).

Colony-forming Assay

Early passage U-2 OS cells expressing NT or FUS shRNA were plated at a low density and allowed to adhere overnight. Cells were exposed to the indicated dose of IR and incubated for an additional 10–14 days with fresh medium supplemented every 5 days. Subsequently, cells were harvested, fixed with methanol, and stained with 0.5% crystal violet. The results were normalized to plating efficiency.

PAR Binding Assays

HA-FUS was transiently expressed in U-2 OS cells and purified with HA-agarose beads (Sigma A2095). Alternatively, endogenous FUS was immunoprecipitated (FUS antibody Abcam 23439) from U-2 OS cells expressing either NT or FUS shRNA. Finally, GFP, GFP-FUS, or GFP-FUS(RGG2) was transiently expressed in U-2 OS cells and purified using antibody against GFP. Immunoprecipitates were vigorously washed in high salt lysis buffer adjusted to 500 mm NaCl and incubated with 25 nm PAR in TBST (0.1% Tween 20) containing 300 mm NaCl for 1 h. The immunoprecipitates were washed in TBST adjusted to 500 mm NaCl and resuspended with Laemmli buffer, boiled, and spotted onto nitrocellulose membrane. Immunoblotting was performed with the indicated antibodies.

DNA Repair Assays

The measurement of NHEJ- and HR-mediated DSB repair has been described previously (45, 46). Briefly, 1.25 × 10⁶ HEK-293 EJ5-GFP or DR-GFP cells expressing shRNA against a NT control, FUS, CtIP, or Lig4 were seeded onto 60-mm plates. The following day cells were transfected with plasmids expressing I-SceI and mCherry using Lipofectamine 2000. After ∼72 h the cells were harvested and analyzed by flow cytometry for GFP and mCherry expression. mCherry served as a transfection control. GFP expression was normalized to transfection efficiency.

RESULTS

FUS and EWSR1 Are Recruited to Sites of DNA Damage

Given that RNA processing factors, including RBMX, PPM1G, THRAP3, hnRNPUL1 and -2, and NONO (51–54) have been shown to participate in the DDR and localize to DNA damage sites, we examined whether this was true for FUS. Human U-2 OS cells were sensitized with Hoechst dye followed by microirradiation with a 405-nm laser. Under these conditions, we observed a robust accumulation of GFP-FUS at DNA damage tracks as marked by staining for γH2AX or MDC1 (Fig. 1A). It should be noted that we did not detect FUS accumulation within microirradiated regions in the absence of Hoechst presensitization. Interestingly, although we observed the recruitment of both FUS and MDC1, very little colocalization within the stripe was observed, suggesting that these proteins occupy spatially distinct sites at damaged chromatin (Fig. 1A). In addition to concentrated GFP-FUS at laser stripes, we observed a subpopulation within structures that resembled nucleoli (Fig. 1A). Finally, we examined the subnuclear localization of TAR DNA-binding protein 43 (TDP-43), an RBP that interacts with FUS and whose mutation is also causal for inherited forms of ALS (55–60); however, GFP-TDP-43 did not accumulate at sites of laser-induced DNA damage (LIDD) under conditions that clearly led to recruitment of GFP-FUS (Fig. 1A). As shown in Fig. 1B, GFP-FUS accumulation at LIDD was detectable as early as 20 s after damage with peak intensity at ∼5 min. The nucleolar subpopulation of GFP-FUS was also observed prior to UV exposure, suggesting that FUS may be a constituent of nucleoli in undamaged cells. Interestingly, upon damage, the intensity of this population decreased significantly (Fig. 1B, compare 20 s and 5 min) but then increased at later time points (Fig. 1B, compare 5 min and 30 min). Finally, although we were unable to detect FUS accumulation at IR-induced foci, we did observe the accumulation of endogenous FUS at LIDD (Fig. 1C). Similar to results observed with GFP-FUS, the majority of FUS did not colocalize with MDC1 (Fig. 1C). These data demonstrate that upon DNA damage, FUS is reorganized within the cell including loss of nucleolar FUS and rapid recruitment to sites of DNA damage.

FIGURE 1.

FUS is recruited to sites of DNA damage. A, detection of GFP-FUS and GFP-TDP-43 at LIDD. U-2 OS cells expressing the indicated GFP-tagged protein were laser-microirradiated, fixed within 10 min, and processed for indirect immunofluorescence with the indicated antibodies. B, time course of GFP-FUS and GFP-EWSR1 occupancy at DNA damage sites. Cells expressing the indicated GFP-tagged protein were laser-microirradiated and monitored by live cell imaging. Shown are representative images from at least two independent experiments. C, detection of endogenous FUS and EWSR1 at LIDD. ∼50 cells were laser-microirradiated within 15 min. Cells were extracted prior to fixation and processed with the indicated antibodies.

We also examined EWSR1, which is a closely related member of the FET family of RBPs (61). Like GFP-FUS, GFP-EWSR1 rapidly and robustly accumulated at LIDD. However, whereas GFP-FUS recruitment persisted for at least 30 min, GFP-EWSR1 recruitment was extinguished within 20 min (Fig. 1B). In addition, endogenous EWSR1 was also detected at LIDD (Fig. 1C). Thus, rapid targeting to DNA damage is an intrinsic property of both FUS and EWSR1; however, FUS is more persistent at LIDD, suggesting that the functions of these proteins may be nonidentical.

Recruitment of FUS to Sites of DNA Damage Is Not Disrupted by Mutations in the RRM or ZNF

FUS contains a single RRM and ZNF (Fig. 2A) domain and has been shown to bind RNA, ssDNA, and dsDNA. To examine a potential role for nucleic acid binding in localizing FUS to DNA damage, we utilized GFP-FUS expression constructs harboring mutations in the RRM or ZNF. Mutation of four Phe (at 305, 341, 359, and 368) to Leu within the RRM (4F-L) severely abrogates RNA binding (62, 63); however, expression of GFP-FUS(4F-L) did not alter recruitment to LIDD (Fig. 2B). Next, we examined a requirement for the ZNF of FUS, which belongs to the RanBP2-type family (64). Based on the participation of Cys-428, -444, and -447 in coordinating zinc (64), we mutated these residues to generate GFP-FUS(C428A/C444A/C447A). Similar to the RRM mutant, disruption of the ZNF did not inhibit FUS localization (Fig. 2B). These data suggest that recruitment of FUS to DNA damage is unlikely to involve direct binding to nucleic acid through the RRM or ZNF domain. In addition, we examined two mutations associated with hereditary ALS, which lie within the predicted nuclear localization signal located at the C terminus of FUS (36). Neither of these mutants abrogated FUS accumulation at LIDD (Fig. 2C).

FUS Localization to Sites of DNA Damage Is PARP-dependent

ATM and PARP signaling represent two important mechanisms for recruitment of DDR factors to DNA lesions (65). Several ATM and DNA-PK phosphorylation sites have been identified within the Gln/Gly/Ser/Tyr-rich region of FUS. Specifically, Ser-42 was shown to be phosphorylated in response to IR in an ATM-dependent manner (44). However, treatment of cells with the ATM inhibitor KU-55933 (ATMi) or the DNA-PK inhibitor NU-7441 (DNA-PKi) did not alter FUS recruitment to LIDD (Fig. 2D). We also mutated Ser-42 to Ala as well as three other Ser residues that have been shown to be phosphorylated by DNA-PK (Ser-26, -61, and -84) to generate GFP-FUS(4A) (44, 66). Neither GFP-FUS(S42A) nor FUS4A altered accumulation at LIDD (Fig. 2D). Taken together, these data demonstrate that phosphoinositide 3-kinase-like kinase signaling does not mediate FUS recruitment to damage sites (Fig. 2D). In contrast, treatment with the PARP inhibitor PJ34 (PARPi) completely blocked FUS accumulation (Fig. 2D).

The kinetics of FUS accumulation were comparable with other factors whose recruitment is PARP-dependent (Fig. 1B) (21, 51). Many of these factors are capable of binding to PAR directly; therefore, we tested the ability of FUS to interact with PAR polymers in vitro. HA-FUS interacted with PAR chains, suggesting that FUS is recruited to damaged chromatin through a direct interaction with PARP ADP-ribosylated substrates (Fig. 3A). The specificity of the FUS-PAR interaction was confirmed using cell lines rendered deficient for FUS through RNAi (Fig. 3B).

FIGURE 3.

FUS interacts with PAR. A, HA-FUS protein was immunopurified (IP) from U-2 OS cells and incubated with 25 nm PAR for 1 h. The interaction was examined by dot-blotting (WB) with anti-PAR antibody. B, endogenous FUS was immunopurified from U-2 OS cells expressing shRNA against either a nontargeting (NT) control sequence or the 3′-UTR of FUS (FUS #2). PAR binding was performed as in A. Shown are representative blots from at least two independent experiments. EV, empty vector.

The RGG2 Domain Is Sufficient for Localization to Sites of DNA damage, whereas the PLD Promotes Robust Accumulation

To begin characterizing the domains responsible for mediating the accumulation of FUS at LIDD, we generated several N-terminal truncation mutants fused to GFP (Fig. 4A). Interestingly, the presence of the RGG2 domain alone was sufficient to mediate localization to DNA damage, albeit with delayed and reduced accumulation relative to full-length FUS (Fig. 4, B and C). GFP-FUS(RGG1/2) displayed slightly increased accumulation, suggesting that all the RGG domains may contribute to PAR binding (Fig. 4C). Although capable of targeting to LIDD, the maximal accumulation of the N-terminal truncation mutants was ∼20% of the full-length FUS protein (Fig. 4C).

FIGURE 4.

The RGG2 domain is sufficient to accumulate at LIDD. A, FUS truncation and internal deletion mutants used in this study are depicted. B, U-2 OS cells expressing the indicated GFP-FUS constructs were laser-microirradiated and monitored by live cell imaging. C, FUS accumulation at sites of DNA damage (purple boxes) at the indicated time points was quantified using ImageJ software. FUS accumulation was normalized by subtracting the fluorescence intensity prior to microirradiation for each cell analyzed. A minimum of five cells were analyzed from two independent experiments. Error bars, S.E. from two independent experiments.

FUS harbors a low complexity PLD spanning amino acids 1–239 (30). We hypothesized that this region amplified FUS accumulation at sites of DNA damage, potentially through oligomerization. Therefore, we generated an internal deletion mutant (Δ204–475), which retains the majority of the PLD fused to the RGG2 domain (PLD/RGG2). The recruitment of GFP-FUS(PLD/RGG2) was much stronger than recruitment of GFP-FUS harboring the RGG2 domain alone but did not reach the level seen for full-length GFP-FUS (Fig. 4, B and C). We conclude that the PLD makes a major contribution to FUS targeting to DNA damage.

The above results suggested that RGG2 specifies FUS targeting to LIDD through an interaction with PAR. Indeed, recruitment of GFP-FUS(RGG2) to LIDD was completely inhibited by pre-treatment with PARP inhibitor (Fig. 5A). Furthermore, immunoprecipitated GFP-FUS(RGG2) interacted with PAR polymers in vitro to a similar if not greater extent than full-length GFP-FUS (Fig. 5B). Finally, cells depleted of endogenous FUS by shRNA displayed a similar recruitment of GFP-FUS(RGG2) to LIDD (data not shown), suggesting that its recruitment was not simply due to an interaction with endogenous FUS protein. Taken together, these data suggest a model whereby the RGG2 domain, possibly in cooperation with the other RGG domains, mediates FUS targeting to DNA damage through interaction with PAR, whereas the PLD amplifies and/or stabilizes FUS accumulation at DNA damage sites.

FIGURE 5.

The minimal FUS RGG2 domain binds PAR. A, U-2 OS cells expressing GFP-FUS(RGG2) were mock- or pretreated with PARP inhibitor for 1 h prior to laser microirradiation. Recruitment was monitored by live cell imaging. Shown are representative images from at least two independent experiments. B, GFP alone, GFP-FUS, or GFP-FUS(RGG2) proteins were immunopurified (IP) from U-2 OS cells and incubated with 25 nm PAR for 1 h. The interaction was examined by dot-blotting (WB) with anti-PAR antibody. Shown are representative blots from at least two independent experiments.

Finally, because microirradiation induces a variety of DNA alterations including SSBs and DSBs as well as base modifications we sought to examine the localization of FUS at defined DSBs using chromatin immunoprecipitation (ChIP). We employed a U-2 OS cell line developed by Legube and colleagues that expresses a 4-hydroxytamoxifen-inducible AsiSI restriction endonuclease (67). Although γH2AX enrichment was observed at each AsiSI test locus following 4-hydroxytamoxifen induction, we failed to detect accumulation of FUS (data not shown). Based on the rapid and transient recruitment of FUS to sites of DNA damage, we may be unable to observe FUS enrichment using this technique.

FUS Promotes Efficient DSB Repair

FUS^−/− murine embryonic fibroblasts display increased chromosomal instability and radiosensitivity (40, 41). Therefore, to address a biological requirement for human FUS in the response to DNA damage, we utilized a lentiviral system to stably express shRNA in U-2 OS cells against a NT control sequence, a coding sequence of FUS mRNA or the 3′-untranslated region. shRNA-mediated knockdown of FUS conferred a modest reduction in clonogenic cell survival in response to a 4-gray dose of IR (Fig. 6); however, these results fell just short of statistical significance using a 95% confidence interval. It is possible that the residual FUS protein is sufficient to promote radioresistance and/or that EWSR1 fulfills some of the same functions in genome protection.

FIGURE 6.

FUS depletion confers modest radiosensitivity in human cells. A, clonogenic survival in response to IR of U-2 OS cells expressing shRNA against a NT control sequence, FUS CDS (FUS #1), or FUS 3′-UTR (FUS #2). Error bars, S.E. from two independent experiments. B, Western blot showing FUS expression in shRNA cell lines.

To address the contribution of FUS at sites of DNA damage, we initially focused on a potential role for FUS in the recruitment of other SSB- and DSB-targeted proteins. Both the SSB repair protein XRCC1 and chromatin remodeling factor CHD4 are recruited to DNA damage in a PARP-dependent manner (21, 68); however, depletion of FUS did not disrupt targeting of either XRCC1 or CHD4 (data not shown). In addition, FUS knockdown had no effect on recruitment of histone deacetylase 2 (HDAC2), which also accumulates at sites of DNA damage (data not shown) (69). Similarly, the phosphorylation of ATM substrates, including 53BP1 and CHK2 (70–73), was comparable between irradiated control and FUS-depleted U-2 OS cells, suggesting that FUS is not required for ATM signaling (data not shown).

We next tested the impact of FUS silencing on DSB, focusing first on NHEJ. We transduced the EJ5-GFP NHEJ reporter cell line developed by Stark and colleagues (46) with FUS or various control shRNAs and measured GFP reactivation upon transient transfection with a plasmid encoding the I-SceI restriction enzyme by flow cytometry. shRNA-mediated knockdown of FUS resulted in a significant (∼30%) reduction in NHEJ compared with cells expressing a NT shRNA control (Fig. 7A). For the purposes of comparison, knockdown of the essential NHEJ factor DNA ligase IV (Lig4) caused a >50% reduction of NHEJ in this assay (Fig. 7A). Importantly, FUS depletion did not alter the expression of critical NHEJ factors KU70 and DNA-PKCs (Fig. 7, B and C). These findings imply a supportive role for FUS in NHEJ.

FIGURE 7.

FUS is required for efficient DSB repair by NHEJ and HR. A and D, HEK-293 EJ5-GFP cells (A) or HEK-293 DR-GFP cells (D) were infected with lentivirus to deliver NT, FUS, Lig4, or CtIP shRNA and incubated for 48 h at which time, cells were transfected with plasmids encoding I-SceI and mCherry and incubated for an additional 72 h. A–E, cells were then harvested and analyzed by flow cytometry (A and D) or Western blotting (B, C, and E). F, CtIP knockdown was by evaluated quantitative PCR. Shown in A and D are repair frequencies relative to the NT shRNA control from at least three independent experiments, except for the Lig4 control, which represents technical replicates. Asterisks denote a statistical difference of p < 0.01 compared with the NT shRNA control.

Previous work has shown that FUS is capable of supporting D-loop formation, which suggests a potential role in HR (34). To test this we used the HEK-293 DR-GFP reporter cell line, which specifically measures repair through HR (46). Knockdown of FUS with either of two shRNAs caused an ∼30% reduction in HR efficiency, whereas knockdown of CtIP, which is required for DSB resection (74), resulted in ∼50% decrease (Fig. 7, D–F). These data suggest that FUS functions in an early step in DSB repair that may be common to both NHEJ and HR.

DISCUSSION

In this work, we demonstrated that the RBP FUS is rapidly recruited to DNA lesions in a PARP-dependent manner and is capable of interacting with PAR chains directly. Although other studies have implicated a role for FUS in maintaining resistance to genotoxic stress, possibly via participation in HR repair (34, 40, 41), the present results indicate that FUS functions are required for optimal DSB repair through both NHEJ and HR. FUS joins a growing list of RNA-processing factors including RBMX, PPM1G, THRAP3, hnRNPUL1 and -2, and NONO, that are recruited to the proximity of DSBs and whose activities are required for genotoxin resistance and DNA repair (51–54).

Substrate PARP ADP-ribosylation provides molecular scaffolding for the localization of factors involved in DNA repair to sites of damage (20, 42). PARP ADP-ribosylation also appears to provide a common mechanism for the recruitment of at least a subset of RNA-processing factors to DNA damage sites including RBMX and NONO (51, 52, 75). NONO is recruited by virtue of an RRM1 domain-dependent interaction with PAR (52). In contrast, the RRM of FUS is dispensable for recruitment to LIDD, although this does not rule out a contribution of the RRM to FUS functions once it accumulates at DNA damage.

Our data indicate that the RGG2 domain interacts with PAR polymers in vitro and mediates the recruitment of FUS to sites of DNA damage. In support of this, FUS was recently identified as a PAR-associated protein through quantitative proteomic approaches (76). The RGG domains of MRE11 as well as the RNA-processing factor hnRNPUL1 mediate recruitment to sites of DNA damage (54, 75, 77), suggesting that this motif may function generally as an atypical PAR binding domain. The fact that FUS targeting to DNA damage was not fully supported by RGG2 suggests that the other RGG repeats may contribute to PAR binding. It is also possible that PAR-independent interactions contribute to the stable association of FUS with DNA damage. In the hnRNPUL1 paradigm, full recruitment required both the MRE11-RAD50-NBS1 complex and PARP-1 (54, 75). We envision a similar scenario for FUS, with the low complexity PLD stabilizing FUS recruitment through heterotypic or homotypic interactions. The FUS PLD has the unusual property of forming hydrogel droplets upon concentration (66, 78), and it was proposed that reversible oligomerization of the PLD plays an important role in the assembly and disassembly of ribonucleoprotein-splicing complexes. It is worth considering that a similar oligomerization-dependent mechanism augments the assembly of FUS complexes at DNA damage. Finally, ChIP experiments failed to reveal FUS recruitment to restriction enzyme-induced DSBs (data not shown), suggesting that FUS does not persist at DSBs for long periods of time. Other transient players in DSB repair, including PARP, are also difficult to detect using this method.³

The details of FUS participation in the immediate response to DNA damage are yet to be determined; however, given that FUS silencing impacted both HR and NHEJ, it may act in a process common to both pathways. FUS could potentially facilitate DSB repair by modulating DSB-proximal transcription and/or through removal of nascently transcribed RNAs. FUS could also deliver noncoding RNAs that are increasingly implicated in DSB repair through unknown mechanisms (78). Nonexclusively, FUS was recently identified as a SUMO E3 ligase for Ebp1 p42 (79), which raises the intriguing possibility that FUS also acts as a SUMO E3 ligase at DSBs. Indeed, a requirement for sumoylation by the E3 ligase enzymes PIAS1 and PIAS4 at DSBs is established (80). Functional analyses of RNA-binding and sumoylation-deficient mutants of FUS will help illuminate the validity of these models. It is also tempting to speculate that oncogenic FUS fusions possess aberrant DSB repair properties. Our results suggest that FUS-CHOP fusion proteins harboring the FUS PLD but lacking RGG repeats are unlikely to support FUS DNA repair activities, and this could be a relevant consideration in FUS-CHOP-positive myxoid liposarcomas, which tend to be radiosensitive (81).

Finally, given that EWSR1 was also recruited to DNA damage, it may function semiredundantly with FUS in DSB repair. Indeed, similar to FUS^−/− mice, EWSR1^−/− mice are highly sensitive to IR and display defects in meiosis and B lymphocyte development (82). Future studies will be required to define the relative contributions of FET family RBPs to DSB repair and genome integrity.