Neddylation Promotes Ubiquitylation and Release of Ku from DNA-Damage Sites

Summary The activities of many DNA-repair proteins are controlled through reversible covalent modification by ubiquitin and ubiquitin-like molecules. Nonhomologous end-joining (NHEJ) is the predominant DNA double-strand break (DSB) repair pathway in mammalian cells and is initiated by DSB ends being recognized by the Ku70/Ku80 (Ku) heterodimer. By using MLN4924, an anti-cancer drug in clinical trials that specifically inhibits conjugation of the ubiquitin-like protein, NEDD8, to target proteins, we demonstrate that NEDD8 accumulation at DNA-damage sites is a highly dynamic process. In addition, we show that depleting cells of the NEDD8 E2-conjugating enzyme, UBE2M, yields ionizing radiation hypersensitivity and reduced cell survival following NHEJ. Finally, we demonstrate that neddylation promotes Ku ubiquitylation after DNA damage and release of Ku and Ku-associated proteins from damage sites following repair. These studies provide insights into how the NHEJ core complex dissociates from repair sites and highlight its importance for cell survival following DSB induction.


INTRODUCTION
The DNA-damage response (DDR), comprising the sensing, signaling, and repair of damaged DNA, requires recruitment and post-translational modification (PTM) of many proteins at DNA-damage sites (Polo and Jackson, 2011). Effective DSB repair is essential for genomic stability, with hereditary DSB repair defects causing cancer predisposition, immunodeficiency, developmental defects, and hypersensitivity to DNA damaging agents (Jackson and Bartek, 2009;Ciccia and Elledge, 2010). DSB repair mainly occurs through two pathways: homologous recombination (HR) and nonhomologous endjoining (NHEJ). Classical NHEJ requires binding of the Ku70/ Ku80 heterodimer to DNA ends, with ensuing recruitment of DNA-PKcs, PAXX, and end-processing factors leading to repair by the DNA ligase IV/XRCC4/XLF complex (Davis and Chen, 2013;Grundy et al., 2014;Wang and Lees-Miller, 2013;Ochi et al., 2015;Xing et al., 2015). While the main NHEJ proteins have been characterized, it is not yet clear how their recruitment to, and dissociation from, DSBs is regulated.

Neddylation Occurs at DSB Sites
To determine whether NEDD8 is present at DNA-damage sites, we used laser microirradiation to generate DSBs in cells presensitized with bromodeoxyuridine (BrdU;Lukas et al., 2003). This revealed that both stably expressed GFP-tagged ( Figure 1B) and endogenous (Figure S1A) NEDD8 were detectable at DNAdamage sites within minutes, co-localizing with Ser-139 phosphorylated histone H2AX (gH2AX), an established DSB marker (Rogakou et al., 1998). Pre-incubating cells for 1 hr with MLN4924 at a dose that effectively inhibits NEDD8 conjugation in cells ( Figure S1B) blocked NEDD8 recruitment to sites of laser microirradiation ( Figures 1B and S1A), indicating that DNA-damage-induced NEDD8 accrual requires an active neddylation pathway.
The ubiquitin machinery, particularly the ubiquitin E1 UBE1, can utilize highly overexpressed NEDD8, causing ''false'' neddylation of substrates (Hjerpe et al., 2012). Importantly, GFP-NEDD8 conjugation detected by immunoblotting of extracts from our stable cell line was blocked by MLN4924, but not by depleting UBE1 ( Figure S1C). This indicated that overexpressed GFP-NEDD8 in this cell line was not substantially used by the ubiquitin system. However, depletion of UBE1 did reduce GFP-NEDD8 recruitment to DNA-damage sites ( Figure S1D), although to a lesser extent than MLN4924 treatment (Figures 1B and S1D), demonstrating that NEDD8 accumulation is at least partially dependent on ubiquitylation-mediated events. Of note, NEDD8 accumulation was only observed in BrdU pretreated cells ( Figure S1E), implying that in our system, NEDD8 accrual was largely promoted by DSBs rather than other forms of damage (Lukas et al., 2003). NEDD8 recruitment did not require the activity of PARP or the DDR kinases ATM, ATR, and DNA-PK ( Figure S1F). Indeed, impairing DNA repair by inhibiting these kinases actually increased NEDD8 accumulation at laser sites (Figure S1F), supporting a role for neddylation in DSB-dependent events.
Although it was reported recently by Ma et al. (2013) that neddylation promotes ubiquitylation at sites of DNA damage, we found that, in our system, robust inhibition of neddylation by MLN4924 did not decrease ubiquitylation at DNA-damage sites as detected by the FK2 antibody ( Figure 1B). In the Ma et al. (2013) study, neddylation was inhibited by depleting RNF111/Arkadia, which they reported to be a NEDD8 E3 ligase. However, RNF111 is also a well-established ubiquitin E3 ligase with a role in the DDR (Poulsen et al., 2013) and it was not determined by Ma et al. whether the effects they observed on ubiquitylation and other aspects of the DDR were due to the ubiquitin E3 activity, rather than the reported NEDD8 E3 ligase activity of RNF111.
Through assessing GFP-NEDD8 recruitment kinetics in live cells, we found that NEDD8 accumulated at damaged sites as early as 5 min after microirradiation and persisted until 40 min in most cells ( Figure S1G, left). To further investigate neddylation dynamics, we treated cells with MLN4924 immediately before laser microirradiation. In these cells, NEDD8 was initially detected at damaged sites (5 min) and then rapidly disappeared, being undetectable by 15 min (Figure S1G, right). The initial accumulation of NEDD8 in this instance most likely represents the time taken for neddylation to be completely inhibited in cells by MLN4924 (which occurs within 5 min; Brownell et al., 2010). These data therefore suggested that neddylation is a dynamic modification that occurs and turns over at DSBs, although we cannot exclude the possibility that some pre-neddylated proteins accumulate at damaged sites then disperse. To corroborate our findings, we tested for DNA-damage-dependent recruitment of neddylation-pathway components. Crucially, this revealed that the NEDD8-conjugating E2 enzymes UBE2M and UBE2F ( Figure 1C), and the deneddylating-complex catalytic subunit, CSN5 ( Figure 1D), were recruited to DNA-damage sites with kinetics similar to that of GFP-NEDD8. Furthermore, CSN5 recruitment was blocked by MLN4924 ( Figure 1D), implying that neddylation is required for CSN5 recruitment. Collectively, these data strongly supported a model in which neddylation and deneddylation actively occur at DSB sites.

Neddylation Promotes Cell Survival after NHEJ
In light of the above findings and because inhibiting neddylation can sensitize cells to DNA-damaging agents (Kee et al., 2012;Wei et al., 2012;Yang et al., 2012;Garcia et al., 2014), we hypothesized that neddylation promotes DSB repair. To investigate this, we tested the effects of depleting UBE2M or UBE2F by small interfering RNAs (siRNAs) on clonogenic cell survival following ionizing radiation (IR) treatment. Notably, while both UBE2M and UBE2F were recruited to DNA-damage sites (Figure 1C), UBE2M but not UBE2F depletion significantly sensitized cells to IR ( Figures 1E and S1H). We speculate that functional compensation by UBE2M (and potentially lower levels of UBE2F compared to UBE2M in the cells we tested) may explain why UBE2F was recruited to laser lines but its depletion did not sensitize cells to IR.
Although DSB repair by HR is restricted to S and G2 cells and can take several hours to complete (Shibata et al., 2011), NHEJ occurs in all cell-cycle stages, with most simple breaks being repaired within minutes (Wang et al., 2001;DiBiase et al., 2000). Because NEDD8 accrual at DNA-damage sites was rapid and occurred in most cells, we speculated that neddylation might regulate NHEJ. In accord with this, depleting UBE2M with two independent siRNAs significantly reduced the number of cell colonies arising in an assay for random plasmid integration ( Figure 1F), which is mediated by NHEJ as well as alternative DNA end-joining processes.

Neddylation Promotes Ku Release from DNA Damage Sites
To explore the impact of neddylation on NHEJ, we used highresolution microscopy together with an RNase A-based extraction method to study formation and dissolution of Ku IR-induced foci (IRIF). In agreement with published findings , Ku foci in control cells were formed within 8 min following IR and then decayed over time, returning to near baseline levels by 1 hr (Figures 2A and 2B). Strikingly, while not impairing Ku IRIF formation, MLN4924 treatment significantly delayed their dissolution, with high numbers of Ku foci remaining even after 2 hr (Figures 2A and 2B). This effect was not through MLN4924 itself causing DNA damage because parallel treatments of non-irradiated cells with MLN4924 did not induce Ku IRIF or gH2AX ( Figure S2A; MLN4924 treatment for longer than 6 hr did cause DNA damage as previously described by Soucy et al., 2009). Of note, while Feng and Chen (2012) published that RNF8 depletion caused Ku80 retention at laser microirradiation sites, we were unable to detect any effect of RNF8 depletion on the resolution of Ku IRIF (data not shown).
To test whether the effect of MLN4924 on Ku removal was indeed via UBA3 inhibition, we generated U2OS cell lines stably expressing wild-type UBA3 or a UBA3 Ala-171 to Thr mutant (UBA3-A171T) that confers MLN4924 resistance Milhollen et al., 2012). As expected, NEDD8 conjugation was abolished by MLN4924 in cells expressing wild-type UBA3 but not UBA3 A171T ( Figures 2C and S2B). Importantly, while both cell lines showed comparable Ku IRIF kinetics under control conditions ( Figure 2D), MLN4924 caused persistent Ku IRIF only in cells expressing wild-type UBA3 ( Figure 2D), thus indicating that the effect of MLN4924 on Ku was via UBA3 inhibition. Although we initially considered the possibility that Ku IRIF persistence reflected defective DSB repair, this did not appear to be the case because MLN4924 did not impair the time-dependent reduction of IR-induced gH2AX, a well-established readout of DSB repair Lö brich et al., 2010), detected either by immunoblotting ( Figure S2C) or immunofluorescence microscopy (Figures 2A and 2E; note in Figure 2E that gH2AX did persist following DNA-PK inhibition). During these analyses, we found that the size and intensity of Ku foci were unaffected by MLN4924, indicating that MLN4924 does not lead to more Ku molecules being loaded onto each DSB (data not shown). Collectively, these findings suggested that blocking neddylation does not affect Ku loading but rather impairs Ku removal from damage sites after repair has occurred.
To assess the above model by a different approach, we used immunoblotting to monitor the accumulation of Ku and other NHEJ factors in RNase A-resistant chromatin fractions after treating cells with the radiomimetic compound phleomycin. In accord with our immunofluorescence data, inhibiting neddylation with MLN4924 caused Ku80 and Ku70 persistence on chromatin after treating cells with a pulse of phleomycin ( Figure 3A). Similarly, MLN4924 caused persistence of the NHEJ factors XRCC4, LIG4, and XLF, suggesting that they are recruited and subsequently released concomitantly with Ku ( Figure 3A; as shown on the right, total levels of Ku80, XRCC4, LIG4, and XLF were unaltered by DNA damage and/or MLN4924). These data supported a model in which neddylation promotes removal of Ku and other NHEJ factors from DNA-damage sites.

Proteomics Identifies Neddylation-Dependent Ku Interactors
To identify factors that might associate with Ku in a NEDD8pathway-dependent manner, we used human RPE-1 cells stably expressing GFP or RPE-GFP-Ku70 cells expressing endogenously tagged GFP-Ku70, wherein the GFP-tag was fused to one of the XRCC6 chromosomal alleles , in SILAC (stable isotope labeling of amino acids in cell culture) studies followed by liquid chromatography-tandem mass spectrometry (LC-MS/MS; Figure 3B; Table S1). Applying a cutoff of R2-fold enrichment for Ku-specific binding, we identified several known Ku interactors as well as various other proteins, including CUL4A ( Figure 3B; a highly related cullin CUL4B had a ratio of 1.8). Subsequent reciprocal co-immunoprecipitation studies confirmed CUL4A as a Ku interactor (Figures 3C and S3A). Notably, we found that depletion of either CUL4A or CUL4B significantly reduced NEDD8 accrual at DNA-damage sites ( Figure S3B), and stably expressed GFP-CUL4A and GFP-CUL4B were both recruited to DSB sites ( Figure S3C). To investigate their potential functional roles in Ku release from chromatin following DNA repair, we depleted CUL4A/CUL4B by siRNA and established cell lines stably expressing inducible dominant-negative CUL4A or CUL4B. However, by neither of these approaches were we able to demonstrate consistently strong effects on Ku removal (data not shown). Nevertheless, we also noted that neither approach inhibited CRL4 ubiquitylation activity to a level comparable to MLN4924 treatment, as monitored by protein levels of the CRL4 substrates CDT1, p27, and p21 ( Figures S3D and S3E). We therefore concluded that residual CRL enzymatic activity and/or functional redundancy between CUL4A, CUL4B, and probably other cullins likely precluded us from observing effects in these studies. Indeed, in vitro studies have implicated CUL1 in the removal of Ku from DNA in cell-free Xenopus laevis egg extracts (Postow et al., (B) MLN4924 blocks NEDD8, but not ubiquitin recruitment to DNA-damage sites. U2OS-GFP-NEDD8 cells were pre-treated for 1 hr with DMSO or 3 mM MLN4924 and laser microirradiated. Cells were fixed after 20 min and visualized by immunofluorescence as indicated. Graph shows average intensity of GFP-NEDD8 at the laser line from three experiments ±SD. White bar represents 10 mM. Asterisks indicate statistically significant difference to control (*p < 0.05; **p < 0.01; ***p < 0.001; ****p % 0.0001). (C) GFP-UBE2F and GFP-UBE2M are recruited to DNA-damage sites. U2OS cells stably expressing GFP-UBE2F or GFP-UBE2M were laser microirradiated, fixed, and visualized as in (B). Graph shows average percentage of gH2AX positive cells with detectable GFP-UBE2M or GFP-UBE2F recruitment from five independent experiments ±SD. White bar represents 10 mM. (D) GFP-CSN5 recruitment to DNA-damage sites is blocked by MLN4924. U2OS cells stably expressing GFP-CSN5 were treated as in (B). Images were acquired by live cell imaging. Laser tracks are indicated by dashed white lines. White bar represents 10 mM (E) UBE2M depletion causes hypersensitivity to IR. Clonogenic U2OS cell survivals were performed after transfection with indicated siRNAs and doses of IR. Each point represents an average of at least three independent experiments (except UBE2M-2 which was repeated twice). Error bars correspond to SDs, and asterisks are as in (B). (F) UBE2M depletion causes an NHEJ defect. Random plasmid integration assay was performed in U2OS cells transfected with indicated siRNAs. Error bars correspond to SD of at least three independent experiments (asterisks as in B). See also Figure S1.  Figure 1B). White bar represents 10 mM. (C) U2OS-A171T UBA3 are resistant to MLN4924. U2OS cells stably expressing WT UBA3 or A171T UBA3 were treated with DMSO or 3 mM MLN4924 for 1 hr and analyzed by immunoblotting with indicated antibodies. Endogenous UBA3 was depleted with a siRNA to the 3 0 UTR ( Figure S2B). Neddylated conjugates are detected with a NEDD8-specific antibody.
(legend continued on next page) 2008; Postow and Funabiki, 2013), thus indirectly supporting the idea that Ku could be a shared substrate of CUL1 and CUL4A/B in human cells. In regard to the above, we found that depletion of RBX1-which functions together with UBE2M (Huang et al., 2009) and is the NEDD8 and ubiquitin E3 ligase for cullins 1, 2, 3, 4A, and 4B-increased the number of Ku IRIF at all time points tested and caused persistence of Ku and NHEJ factors on chromatin after phleomycin treatment ( Figures S3F-S3H). These effects on NHEJ-factor kinetics were less marked than with MLN4924 treatment, however, and also the kinetics of Ku release following RBX1 depletion differed from that seen with MLN4924 treatment. These differences might reflect the potency (D) MLN4924 effects on Ku80 foci retention are through UBA3 inhibition. U2OS cells stably expressing WT UBA3 or A171T UBA3 were processed as in (A), and results were quantified as in (B). (E) MLN4924 does not affect gH2AX recovery after IR. Quantification of total gH2AX intensity per nucleus in cells treated with 10 Gy IR then harvested at indicated times. Samples were prepared as in (A). Pre-treatment for 1 hr with 3 mM DNA-PK inhibitor (DNA-PKi) used as a positive control. Statistical analysis as in (B). AU, arbitrary units. See also Figure S2.

Figure 3. Proteomics Identifies Neddylation-Dependent Ku Interactors
(A) MLN4924 causes retention of NHEJ factors on the chromatin. U2OS cells were pretreated with DMSO or 3 mM MLN4924 for 1 hr and then treated with 500 mM phleomycin (Phleo) for 1 hr. Cells were left to recover in the presence of MLN4924 or DMSO following phleomycin removal and then collected at the indicated times. Cells were preextracted with CSK buffer + RNase A prior to lysis (chromatin; left) or lysed as whole cell extracts (right) and immunoblotted with indicated antibodies. Black arrow indicates XRCC4. (B) RPE-1 cells stably expressing GFP or Ku70 endogenously tagged with GFP were labeled with light, medium, or heavy isotopes and treated as indicated. Cell lysates were subjected to GFP retrieval. Enriched proteins were resolved by SDS-PAGE and proteolysed in gel with trypsin, and peptides were analyzed by LC-MS/MS. The scatterplot shows the logarithmized SILAC ratio of GFP-KU70/GFP control and GFP-KU70 + MLN4924/GFP-KU70. The known Ku interactors and CUL4A (ratio 2.23) are labeled in black font and open blue circles. In pink are interactions enhanced upon MLN4924. In green are interactions decreased upon MLN4924 (see also  Table S1). (C) Experiment repeated as in (B) without isotope labeling of cells. Following GFP IP, cell lysates were immunoblotted with indicated antibodies. Note that CUL1 (Postow and Funabiki, 2013) and CUL4B were not detected in Ku immunoprecipitates. See also Figure S3. of MLN4924 treatment compared to incomplete RBX1 depletion ( Figure S3H) and/or could be influenced by prolonged RBX1 depletion over 72 hr compared to 1 hr exposure to MLN4924. Collectively, these data were consistent with CRL activity promoting Ku removal, although we acknowledge that other factors might also be involved.
In support of a model in which neddylation promotes dissociation of the NHEJ apparatus ( Figure 3A), our proteomics data and subsequent co-immunoprecipitation studies revealed that the interaction between Ku and DNA ligase 4/XRCC4, as well as the recently identified NHEJ complex component PAXX (Ochi et al., 2015;Xing et al., 2015), was significantly enhanced when neddylation was blocked by MLN4924 (Figures 3B and  3C; Table S1). Interestingly, the interaction between Ku and several other proteins, including topoisomerase 2A (TOP2A), was also enhanced when neddylation was blocked by  Table S1). This could occur because such factors directly interact with Ku on the chromatin and are therefore released with Ku, or, alternatively, these factors might interact with chromatin in other ways, in a manner that is regulated by neddylation. Strikingly, almost all the proteins whose association with Ku was diminished upon MLN4924 treatment comprised factors associated with the 26S proteasome, as well as the segregase/unfoldase VCP (valosin containing protein; also known as p97) that targets ubiquitylated proteins to dissociate them from molecular assemblies, frequently promoting their proteasomal degradation (Meerang et al., 2011). Notably, MLN4924 treatment did not affect the levels of VCP or proteasome components ( Figure S3I). Collectively, these findings suggested that VCP and proteasomal components recognize DNA-damage-dependent, NEDD8-mediated ubiquitylation of Ku, and/or other NHEJ components.

Neddylation Promotes Ku Ubiquitylation following DNA-Damage Induction
In light of the above findings, we tested whether neddylation might promote ubiquitylation of Ku. Thus, we immunoprecipitated, under stringent conditions (1 M NaCl; see Experimental Procedures), endogenous Ku70 from RPE-GFP-Ku70 cells . Subsequent immunoblotting revealed that Ku ubiquitylation was markedly increased by phleomycin treatment ( Figure 4A, lanes 2 and 3), whereas no ubiquitylated species were detected in cells expressing GFP alone (lane 1). Furthermore, this ubiquitylation did not occur with DNAdamaging agents that do not directly yield DSBs ( Figure S4A). Crucially, inhibiting neddylation with MLN4924 strongly reduced Ku ubiquitylation induced by phleomycin treatment ( Figure 4A, lanes 4 and 5; note in Figure S4B that overall ubiquitylation in cells was not affected by MLN4924). Blocking NHEJ-mediated DSB repair with a DNA-PK inhibitor, which has been shown to inhibit Ku release from DNA-damage sites , also impaired phleomycin induced Ku ubiquitylation ( Figure 4A, lanes 6 and 7), as did siRNA depletion of DNA ligase 4 (data not shown), suggesting that Ku ubiquitylation occurs as a consequence of DSB repair. Although Ma et al. (2013) suggested that RNF168 recruitment to DNA-damage sites is neddylation dependent, we found that RNF168 depletion did not affect Ku ubiquitylation ( Figure S4C), nor did it enhance Ku, XRCC4, or XLF persistence on chromatin after treating cells with phleomycin ( Figure S4D; see Figure S4E for RNF168 depletion).
To establish whether the ubiquitylation we observed above was occurring specifically on Ku, we analyzed ubiquitylation in GFP-Ku70 immunoprecipitates by SILAC-based LC-MS/MS. This identified three sites (K195, K265, and K481) on Ku80 and one (K114) on Ku70, upon which ubiquitylation was increased following phleomycin treatment (SILAC ratio M/L) and was blocked by MLN4924 pretreatment (SILAC ratio H/M; Figure 4B; Table S2; the mass spectrum of K481 is shown in Figure S4F as an example). We investigated whether the sites identified on Ku80 were required for damage-dependent ubiquitylation of Ku and release of Ku from DNA-damage sites. This established that Ku ubiquitylation in the context of an exogenously expressed Ku80 mutant, with lysines 195, 265, and 481 (3K-R) mutated to arginine, still occurred following DNA-damage (data not shown). In addition, we found that the mutant form of Ku was still recruited and released from DNA damaged chromatin with kinetics similar to those of the wild-type protein (data not shown). These data suggested that there may be functional redundancy between the mapped sites and further, as yet unidentified ubiquitylation sites on Ku80 and or Ku70. To try to address this issue, we generated a U2OS cell line expressing an inducible mutant of Ku80 with all but one lysine mutated to arginine (we excluded K265, which has been shown to make direct contact with the DNA and is therefore likely to be important for DNA binding; Walker et al., 2001). Unfortunately, this all-butone lysine mutant protein was not recruited to DNA-damage sites (data not shown) and could not therefore be used for further experiments. We did not attempt to generate cell lines expressing inducible mutants of Ku80 in combination with lysineto-arginine mutants of Ku70, and it is possible that dimerization with endogenous Ku70 is sufficient to enable ubiquitylation and release of Ku complexes containing the Ku80 3K-R protein.

DISCUSSION
We have shown that neddylation is a dynamic modification at DNA-damage sites and that neddylation promotes cell survival after DSB induction. Furthermore, we have established that neddylation promotes the ubiquitylation of Ku upon DNA repair, and that this is associated with the release of Ku and other NHEJ factors from repair sites. Significantly, our work has identified DNA-damage induced, neddylation-dependent ubiquitylation of K195, K265, and K481 in Ku80 and K114 in Ku70. Interestingly, K265 and K481 lie within the core DNA binding domain of Ku80, with K265 directly making contact with DNA (Walker  (Grundy et al., 2013) of Ku70 and Ku80, respectively, that are thought to mediate protein-protein interactions ( Figures 4C and 4D). The locations of these ubiquitylation sites suggest how Ku ubiquitylation on these and other sites could both trigger the dissociation of Ku from other NHEJ proteins as well as being associated with its release from DNA.
Collectively, the available data suggest the following model ( Figure 4E): first, following DSB induction, Ku, PAXX, DNA-PK, XRCC4, LIG4, and XLF assemble at the DNA-damage site (Davis and Chen, 2013;Ochi et al., 2015;Xing et al., 2015); next, following DNA repair, Ku is ubiquitylated in a DNA-damageand neddylation-dependent manner to promote the release of Ku and other NHEJ factors from the site of repair. Significantly, our proteomics analyses revealed that VCP and various proteasome subunits interact with Ku in a DNA-damage induced and neddylation-dependent manner. It is known that VCP can unfold ubiquitylated proteins and is important for extracting certain DNA-repair proteins from chromatin (Dantuma et al., 2014), and, while a role for VCP in removing Ku from DNA has been proposed (Postow, 2011), it has not yet been demonstrated in the literature. Because it has been reported that Ku removal from damage sites is not affected by proteasome inhibition (Postow et al., 2008), we suggest that in addition to disrupting interactions between Ku, DNA, and other NHEJ components, Ku ubiquitylation likely promotes targeting by VCP, leading to extraction of Ku from chromatin, perhaps then followed by proteasome-dependent Ku degradation ( Figure 4E). As the Ku70/Ku80 heterodimer forms a highly stable ring structure encircling DNA ends (Walker et al., 2001), if Ku remains DNA bound once a DSB has been repaired, it would likely interfere with various processes, particularly transcription and replication (Frit et al., 2000;Ono et al., 1996). By serving as a barrier to complete genome replication and/or segregation, the persistence of Ku and other NHEJ factors on repaired DNA could thus account for the decreased cell survival we have observed when neddylation is abrogated after DSB induction as well as reduced NHEJ-dependent cell colony formation in plasmid integration assays. We recognize, however, that neddylation might also regulate several DSB repair pathways and that there are likely to be multiple mechanisms accounting for IR hypersensitivity upon MLN4924 treatment. Finally, we note that because NEDD8 pathway components are overexpressed or mutated in many human cancers, NEDD8-pathway inhibition is a promising anti-cancer strategy (Watson et al., 2011). Accordingly, our findings highlight opportunities for combining MLN4924 with DSB-inducing agents and for exploring cancer-genetic backgrounds where this combination might be particularly effective.

EXPERIMENTAL PROCEDURES
For more details on experimental procedures, please refer to the Supplemental Experimental Procedures.

DNA Damage and Drug Treatments
Cells were preincubated with inhibitors for 1 hr prior to genotoxic treatments, and MLN4924 (Active Biochem) was used at 3 mM unless otherwise indicated. ATMi (KU55933) and DNA-PKi (NU7441; Tocris Bioscience) were used at 10 and 3 mM, respectively. PARPi (olaparib; Stratech Scientific) was used at 10 mM. ATRi (ATR-45; OSUCCC Medicinal Chemistry, Ohio State University) was used at 1 mM. X-ray irradiation was performed with a calibrated irradiation system (Cell Rad Faxitron) fitted with an aluminum filter for soft X-rays. Cells were irradiated in culture medium at room temperature, and standard, 10-Gy irradiation required an exposure time of 3 min 36 s. Phleomycin (Melford Labs) was used at 500 mM for 1 hr, Carboplatin (Sigma-Aldrich) was used at 100 mM for 1 hr, and Camptothecin (Sigma-Aldrich) was used at 1 mM for 1 hr. Cells were UV-irradiated with 10 J/m 2 and analyzed 1 hr after.

Detection of Ku
For detection of Ku on chromatin and Ku IRIF, cells were processed as described previously .

Laser Microirradiation and Immunofluorescence
Laser microirradiation of cells and immunofluorescence were as previously described .
In Vivo Ubiquitylation of Ku RPE-1 cells endogenously tagged with GFP-Ku70 at one chromosomal allele  grown in 10-cm plates were pretreated with DMSO or MLN4924 (3 mM, 1 hr) and then treated with Phleomycin (500 mM, 1 hr) and lysed in a lysis buffer (20 mM Tris [pH 7.5], 40 mM NaCl, 2 mM MgCl 2 , 10% glycerol, 0.5% NP-40) containing benzonase 18 U (Novagen) and supplemented with EDTA-free protease inhibitors (Roche) at room temperature. Lysates were adjusted to 0.5 M NaCl and incubated on ice for 30 min and cleared of debris by centrifugation at 4 C at 21130 relative centrifugal force (rcf). IP of GFP-tagged Ku70 was carried out with GFP-trap agarose beads (ChromoTek) for 2 hr at 4 C. Beads were washed four times in lysis buffer containing 1 M NaCl and subjected to SDS-PAGE and immunoblotting.

Clonogenic Survival Assay
Cells were seeded at low density, in triplicate, at two dilutions, in 6-well plates and treated with IR after 24 hr. Cells were left to recover at 37 C for 10-14 days to allow colony formation. Cells were stained with 0.5% crystal violet/20% ethanol and counted. Results were normalized to plating efficiencies.

Random Plasmid Integration Assay
Assays were performed as previously described .

Statistical Analysis
When required, an unpaired Student's t test was calculated using GraphPad software (www.graphpad.com). Quantifications are based on at least three independent experiments unless otherwise specified. In all figures, significant differences between specified pairs of conditions, as judged by the t test, are highlighted by asterisks (*p < 0.05; **p < 0.01; ***p < 0.001; ****p % 0.0001).

ACKNOWLEDGMENTS
We thank Thimo Kurz (University of Dundee, UK) for providing MLN4924 and Kate Dry, Rimma Berlotserkovskaya (S.P.J.'s laboratory), and Eric Lightcap (Takeda Pharmaceuticals) for critical reading of the manuscript. We thank Sylvie Urbe and Michael Clague (University of Liverpool, UK) for providing the GFP-CSN5 plasmid, the Division of Signal Transduction Therapy (University of Dundee, UK) for providing UBE2M and UBE2F plasmids, Matthew Petroski (Sanford-Burnham Medical Research Institute, US) for providing FLAG-UBA3 wild-type (WT) and FLAG-UBA3-A171T constructs, and Nico Dantuma (Karolinska Institute, Sweden) and Changshun Shao (Rutgers University) for providing CUL4A and CUL4B plasmids, respectively. We also thank Nicola Lawrence, Alex Sossick, and Richard Butler (Gurdon Institute, Cambridge, UK) for help with microscopy, Volocity   Quantification shows fold change in GFP--NEDD8 intensity at the laser line compared to background of a representative experiment. Graph shows median intensity values with inter--quartile range. P--values were calculated using the non--parametric Mann--Whitney test (asterisks as Figure 1B). White bar = 10 µM.
(G) GFP--NEDD8 recruitment to sites of damage is a dynamic process. Live cell imaging of U2OS cells stably expressing GFP--NEDD8 pre--treated for 5 min with DMSO or 3 µM MLN4924 and laser microirradiated. White arrows mark sites of GFP--NEDD8 recruitment. Graph shows fold change of GFP--NEDD8 intensity at laser lines from 5 min time point. Error bars represent standard deviation between three independent experiments (asterisks as Figure 1B). White bar = cell survivals were performed in U2OS cells transfected with the indicated siRNA and subjected to increasing doses of IR. Immunoblot shows depletion of UBE2M and UBE2F. As UBE2F and UBE2M co--depletion was highly toxic, we were unable to test the effects of their combined depletion on cell survival after IR.
Each point represents an average of at least three independent experiments.
(E) siRNA depletion of CUL4A or CUL4B fails to efficiently inhibit CRL4 activity.
CUL4A and CUL4B were depleted using the indicated siRNAs (siCUL4A--2 or siCUL4B--1) either alone or in combination. Where indicated, 1 µM MLN4924 was added 16 h prior to cell lysis. Cell lysates were subject to immunoblotting with the indicated antibodies. MLN4924 treatment caused marked stabilization of CDT1 and p27 representing effective CRL4 inhibition (Emanuele et al., 2011;Higa et al., 2006). siRNA depletion of CUL4A had marginal effects on CDT1 and p27 stabilization compared to MLN4924.
(F) RBX1 depletion increases Ku foci numbers following DNA damage. U2OS cells were transfected with siRNA against RBX1 or control (siCTRL) and then treated with 10 Gy IR. Cells were collected and analyzed as Figure 2A+B. The higher number of Ku foci detected at 8 min in the siRBX1 sample likely reflects a failure to release Ku at this early time--point compared to control cells.
(G) RBX1 depletion increases retention of Ku and NHEJ factors on the chromatin following DNA damage. U2OS cells were transfected with two independent siRNAs against RBX1 or control (siCTRL) and then treated with 500 µM Phleomycin for 1 h. Samples were collected as in Figure  3A. Intensity of bands was quantified using Fiji software and intensities for Ku80, XRCC4 and XLF were normalized to intensity of H2AX. Graphs show fold change from the untreated sample.
(I) MLN4924 treatment does not affect levels of VCP or PSMD14. RPE1 cells expressing GFP or endogenously tagged with GFP--Ku70 were treated with 3 µM MLN4924 or DMSO for 1 h, followed by 500 µM Phleomycin for 1 h as indicated.
Whole cell extracts were immunoblotted with the indicated antibodies. Black arrow indicates VCP. pre--treatment (SILAC heavy). The peptide sequence, mass/charge (m/z) and charge state are indicated. Table S3. Antibodies used in this study.    Table  S4. A sequence against firefly luciferase was used as siCTRL.

Plasmids.
For a list of all primers used in this study, see Table S5. All constructs were fully sequenced and are mutation free. NEDD8 was PCR--amplified from a human fetal HA--Ubiquitin has been described previously . To generate all other stable cell lines; 48 h after transfection, cells were seeded at low density, in the presence of 0.5 mg/ml G418 (GIBCO, Life Technologies).
Individual clones were isolated, grown in the presence of selection media and again, expression of full--length protein was validated by immunoblotting.
For whole cell extracts cells were lysed in an SDS lysis buffer (4% SDS, 20% Glycerol, 125 mM Tris--HCl pH 6.8) and protein concentration was measured.
Lysates were then diluted to equal concentration and supplemented with 10% β--Mercaptoethanol and 0.005% Bromophenol blue. Samples were resolved on 4--12% gradient Bis--Tris NuPAGE gels (Novex, Life Technologies) and transferred onto nitrocellulose membranes (GE Life Sciences). Membranes were blocked in 5% milk in TBS containing 0.1% Tween--20 and incubated with the corresponding primary antibody (Table S3) followed by an appropriate secondary antibody coupled to horseradish peroxidase (Fisher Scientific).
Detection was performed with ECL reagent (GE Healthcare). Quantification of the immunoblots was done using Fiji software (http://fiji.sc/Fiji).

Immunofluorescence
For laser microirradiation experiments, cells were washed three times with PBS and then fixed in 2% PFA (15 min). Cells were permeabilised with 0.2% Triton X--100/PBS (5 min) and then blocked with 5% BSA in PBS/0.1% Tween 20 (PBS--T) for 10 min. Cells were stained for 1 h at room temperature with the indicated primary antibodies in 5% BSA/PBS--T, washed with PBS--T and then stained for 1 h at room temperature with the appropriate goat secondary antibodies coupled to Alexa Fluor 488 or 568 fluorophores in 5% BSA/PBS--T.

Detection of Ku on chromatin
For the detection of Ku on chromatin ( Figure  3A) cells were processed similarly to as described previously . Briefly, cells were pre--extracted twice for 3 min with CSK buffer (10 mM Pipes, pH 7.0, 100 mM NaCl, 300 mM sucrose, 3 mM MgCl2 and 0.7% Triton X--100) supplemented with 0.3mg/ml RNase A for 3 min, then washed 3 times with PBS and collected in a SDS lysis buffer (all at room temperature) and processed for immunoblotting as described above.

Detection of Ku foci
Coverslips were prepared as described previously . Briefly, cells were seeded onto 160 µm coverslips (VWR International) 24 h prior to the experiment. Cells were washed three times with PBS and then pre--extracted by incubating twice for 3 min in CSK buffer (see Immunoblotting) supplemented with 0.3 mg/ml RNase A and washed three times in between with PBS (all at room temperature). Cells were then fixed in 2% PFA for 15 min. Before staining, cells were treated with PBS/0.2% Triton X--100 for 5 min, washed with PBS--T and then blocked with 5% BSA/PBS--T for 10 min. Cells were incubated in primary antibodies in 5% BSA/PBS--T for 1 h at room temperature, washed in PBS--T and then incubated for 1 h at room temperature with goat secondary antibodies coupled to Alexa Fluor 488 or 594 fluorophores (Life Technologies) in 5% BSA/PBS--T. Cells were mounted onto glass slides using Vectashield (Vector Laboratories).

High resolution microscopy and deconvolution
As described previously for the visualization of Ku foci , high--resolution images were acquired on a Deltavision PersonalDV (Applied Precision/GE Healthcare) equipped with a 1,024×1,024 CCD camera (CoolSNAP HQ2; Photometrics), and a 100× U Plan S Apochromat/1.40 NA oil objective (Olympus) and controlled with SoftWoRx software x5.5 (Applied Precision/GE Healthcare). Z stacks were taken at 0.2--µm intervals and the fluorescent channels were acquired sequentially. Deconvolution was then performed within SoftWoRx in conservative mode. Brightness and contrast were adjusted and images were cropped using Photoshop CS5 (Adobe).

Quantification of Ku foci
Deconvoluted images of >10 cells per sample were submitted to automatic focus detection using Volocity 6.3 (PerkinElmer). Cells were selected randomly based on DAPI staining. As the DNA content and therefore the number of double strand break induced foci doubles during the S--and G2--phases of the cell cycle, DAPI volume was used as a surrogate marker of DNA content. The number of foci detected per nucleus were therefore adjusted to the mean DAPI volume. Of note, a 1 h, 3 µM treatment of MLN4924 did not significantly alter the mean DAPI volume.

Laser microirradiation
Cells grown on glass--bottom dishes (Willco Wells) in phenol red--free complete medium (Invitrogen) were treated with 10 µM bromodeoxyuridine (BrdU) for inverted microscope (Olympus) equipped with a 37°C heating stage (Ibidi) and a 405 nm laser diode (6 mW) focused through a 60× UPlanSApo/1.35 oil objective to yield a spot size of 0.5--1 mm. The laser settings 0.40 mW output, 50 frames, unless otherwise indicated, were chosen to generate a detectable damage response in a manner dependent on BrdU presensitization and without noticeable cytotoxicity. Cells were analyzed by immunofluorescence using a confocal microscope (FluoView 1000; Olympus).

Random Plasmid Integration assay
Between 2--rounds of siRNA transfection, U2OS cells were transfected with BamHI-XhoI--linearized pEGFP--C1 (Clontech). 24 h after DNA transfection, cells were collected, counted and plated at low confluency onto: a 15 cm dish and 6 cm dish in complete medium, and a 15 cm dish in complete medium with 0.5 mg/ml G418 (GIBCO, Life technologies). The following day, transfection efficiency was calculated by determining the proportion of GFP--positive cells on the 6 cm dish. Cells on the 15 cm dishes were incubated at 37 °C for 10--14 days until colonies had formed. Colonies were stained with 0.5% crystal violet/20% ethanol and counted. Random plasmid integration events were normalized to transfection and plating efficiencies. The P--value was calculated using an un-paired Student's t--test.

RPE--1 cells stably expressing GFP and RPE--1 cells expressing Ku70
endogenously tagged with GFP were grown in SILAC medium as described above for the identification of Ku ubiquitylation sites. RPE--1--GFP were grown in Light and RPE1--GFP--Ku70 were grown in Medium or Heavy media for two weeks prior to the experiment. A full 15 cm plate of cells was treated with Phleomycin (500 µM, 1 h; Light RPE--GFP and Medium RPE--GFP--Ku70) or pre--treated with