Loss of ubiquitin E2 Ube2w rescues hypersensitivity of Rnf4 mutant cells to DNA damage

SUMO and ubiquitin play important roles in the response of cells to DNA damage. These pathways are linked by the SUMO Targeted ubiquitin Ligase Rnf4 that catalyses transfer of ubiquitin from a ubiquitin loaded E2 conjugating enzyme to a polySUMO modified substrate. Rnf4 can functionally interact with multiple E2s, including Ube2w, in vitro. Chicken cells lacking Rnf4 are hypersensitive to hyroxyurea, DNA alkylating drugs and DNA crosslinking agents, but this sensitivity is suppressed by simultaneous depletion of Ube2w. Cells depleted of Ube2w alone are not hypersensitive to the same DNA damaging agents. Similar results were also obtained in human cells. These data indicate that Ube2w does not have an essential role in the DNA damage response, but is deleterious in the absence of Rnf4. Thus, although Rnf4 and Ube2w functionally interact in vitro, our genetic experiments indicate that in response to DNA damage Ube2w and Rnf4 function in distinct pathways.

humans, there are three SUMO paralogues, SUMO1, SUMO2 and SUMO3 that can be linked to the side chains of lysine residues in target proteins. The SUMOylation consensus site Ψ Kx(D,E) (Ψ : hydrophobic amino acid, X: any aminoacid) is found in SUMO2/3 but not SUMO1 which allows the formation of chains of SUMO2/3 (polySUMO). SUMOylation is involved in for many cellular processes including protein stability, protein interaction, activity and localization 20 . Mutation or altered regulation of enzymes essential for protein SUMOylation have been implicated in cancer, heart disease, virus infection and diabetes [21][22][23] . Rnf4 is a RING type E3 ubiquitin ligase which contains four SUMO interaction motifs in its N-terminal region that allow the protein to interact with SUMO chains and drive their ubiquitination 24 . Recent work revealed that Rnf4 is essential for DNA damage repair of DSB by homologous recombination 25,26 . Together these observations suggest that Rnf4 is a key protein which allows the ubiquitination of SUMOylated protein targets after DNA damage and that ubiquitination and SUMOylation modification pathways are closely interconnected by Rnf4 to repair damaged DNA.
To promote ubiquitination upon DNA damage, the ubiquitin E3 ligase Rnf4 must associate with its cognate ubiquitin E2 conjugating enzyme. Ube2w is an unusual E2 enzyme that catalyses N-terminal monoubiquitination of target proteins in vitro 27,28 . Recent work revealed that Ube2w can functionally interact with Rnf4 to promote N-terminal mono-ubiquitination of SUMO chains in vitro 27 . Monoubiquitinated SUMO chains can be further modified by the heterodimeric E2 conjugating enzyme Ubc13/Uev1 to form K63 linked ubiquitin chains 27 . K63-ubiquitin chains formed by the Ubc13 complex are important for DNA damage repair by homologous recombination. In DNA damaged cells that are depleted of Rnf4, the levels of K63 ubiquitin chains are reduced 25 . Ube2w has also been shown to allow anchoring of K63 ubiquitin chains on internal lysine of Trim5α during restriction of HIV reverse transcription 29 . Together these observations suggest that Ube2w and Rnf4 may work together to promotes K63 ubiquitin chain formation upon DNA damage.
Here we show that cells lacking Rnf4 but not Ube2w are hypersensitive to replication stress, inter strand DNA crosslinking agents and DSB caused by γ -irradiation. In contrast, cells lacking Ube2w are not hypersensitive to the DNA damaging agents tested, but inactivation of Ube2w in Δ RNF4 mutant cells rescues the DNA damage sensitivity. These data indicate that Ube2w does not have an essential role in the DNA damage response, but is deleterious in the absence of Rnf4. Thus, while Rnf4 and Ube2w functionally interact in vitro, these genetic experiments suggest that in response to DNA damage Ube2w and Rnf4 function in distinct pathways.

Results
Ube2w inactivation suppresses ΔRNF4 hypersensitivity to DNA damage inducing drugs. In human cells, the role of Rnf4 in DSB repair by HR, NHEJ and ICL repair has been previously described 25,26,30 . In DT40 cells, a role for Rnf4 in DNA damage repair was observed in replication associated stress 25 . As an ubiquitin E3 ligase Rnf4 must work in DNA repair with the help of an ubiquitin E2. The ubiquitin E2 Ube2w has been shown to promote monoubiquitination of SUMO chains in an Rnf4 dependent manner in vitro 27 .
To study the role of Ube2w and Rnf4 in DNA damage repair at the genetic level, we created chicken DT40 cell lines deleted for RNF4 (Δ RNF4); UBE2W (Δ UBE2W) and the double deletion Δ RNF4 Δ UBE2W (Fig. 1a). The three cell lines were viable and proliferated at similar rates as the wild type cell line. We subjected each cell line to DNA damage inducing drugs and measured the cell survival rate by colony formation assay. As observed before, Δ RNF4 cells were hypersensitive to replication stress caused by hydroxyurea (HU) exposure (Fig. 1b). Δ UBE2W (Δ RNF4, Δ UBE2W) were analysed by western blotting using the indicated antibodies. 3 ng of recombinantly expressed rat Rnf4 and Human Ube2w isoform1 protein were used as a control. Molecular weight markers (kDa) are indicated on the right. (b) Wild type cells and cells deficient for Δ RNF4; Δ UBE2W and Δ RNF4, Δ UBE2W were subjected to replication stress by HU. The concentration of HU is indicated on the X-axis (μ M). Effect on each cell line is indicated by the percentage of colony formation on the Y-axis (logarithmic scale). Error bars represent two standard deviations from the mean (2SD).
Scientific RepoRts | 6:26178 | DOI: 10.1038/srep26178 mutants showed no sensitivity to HU induced replication stress (Fig. 1b). Surprisingly, we observed that the double deletion cell line of Δ RNF4 Δ UBE2W was not hypersensitive to HU induced replication stress (Fig. 1b). These results showed that inactivation of Ube2w in a Δ RNF4 background rescued the HU-induced replication stress hypersensitivity observed in Δ RNF4 mutants, thus suggesting that Ube2w is deleterious in the absence of Rnf4 and that Ube2w and Rnf4 function in distinct pathways of DNA damage repair.
To confirm that the hypersensitivity to HU induced replication stress was due to Rnf4 deficiency and not caused by secondary mutation we expressed mCherry-Rnf4 in the Δ RNF4 cell line. Δ RNF4 hypersensitivity to HU was entirely complemented by the expression of chicken mCherry-Rnf4 ( Fig. 2a and supplementary Fig. S1a). Similarly, to confirm the specific role of Ube2w in Δ RNF4 Δ UBE2W double mutant cells, we ectopically expressed human E2 Ube2w isoform 2 in these cells ( Fig. 2b and supplementary Fig. S1c) and demonstrated that they regained the HU hypersensitivity of DT40 Δ RNF4 cells. Δ UBE2W mutant cells are not sensitive to HU, but overexpression of human E2 Ube2w in these cells induced a mild level of sensitivity to HU induced replication stress (supplementary Fig. S1b). We noted that tagged forms of Ube2w were unable to restore the HU hypersensitivity of Δ RNF4 Δ UBE2W cells suggesting that tagging Ube2w renders the protein inactive in vivo (supplementary Fig. S1g,h).
We next tested the sensitivity of Δ RNF4 mutant cells to other types of DNA damage. In addition to HU, Δ RNF4 cells are hypersensitive to ionising radiation, MMC and cisplatin (Fig. 3) and display mildly increased sensitivity to the DNA alkylating agent methyl methanesulfonate (MMS) (supplementary Fig. S2a). In each case the observed hypersensitivity is lost in Δ RNF4 Δ UBE2W cells (Fig. 3). Although Δ RNF4 cells are hypersensitive to the DNA polymerase inhibitor aphidicolin, this hypersensitivity was not rescued in Δ RNF4 Δ UBE2W cells (supplementary Fig. S2a). As observed for HU ( Fig. 1) Δ UBE2W mutant cells are not hypersensitive to any of these agents (Fig. 3

Prolonged FancD2/I monoubiquitination induced by MMC in the absence of by Rnf4 is rescued by UBE2W inactivation.
Cisplatin and MMC generate interstrand ICLs in DNA that are repaired by the ICL repair pathway. The Δ RNF4 mutant cell line is hypersensitive to the ICL inducing drugs MMC and cisplatin: at 50 ng/ml MMC, Δ RNF4 cells are greater than 200 fold more sensitive than wild type cells (Fig. 3). This extreme sensitivity to DNA cross linking agents is also observed in DT40 cells with mutations in FA pathway components 5 . Thus to establish if the FA pathway is functional in Δ RNF4 cells we determined the levels of monoubiquitinated FancI and FancD2 upon MMC treatment.
In response to a 60 minutes MMC treatment, the levels of both monoubiquinated FancI and FancD2 increase in all cell lines, reaching peak levels around 8-24 H after treatment (Fig. 4a). This indicates that the FA pathway is activated in all of the mutant cell lines. To compare the efficiency and timing of ubiquitin modification of FancI and FancD2 in response to MMC in wild type and Δ RNF4, Δ UBE2W and Δ RNF4 Δ UBE2W cells, extracts were analysed by quantitative Western blotting (Fig. 4b,c and supplementary Fig. S3). It is known that 24-48 H after treatment the decreased levels of monoubiquitinated FancI and FancD2 are a reflection of active ICL repair. Consistent with recent observations in human cells 31 the monoubiquitinated form of FancI accumulates to a higher level in response to MMC treatment in Δ RNF4 cells compared to wild type cells (Fig. 4b,c and supplementary Fig. S3). Levels of monoubiquitinated FancI were similar in wild type, Δ UBE2W and Δ RNF4 Δ UBE2W cells, indicating that the increased accumulation of monoubiquitinated FancI in Δ RNF4 cells was abrogated by codepletion of Ube2w (Fig. 4b,c and supplementary Fig. S3). In untreated cells the levels of monoubiquitinated FancD2 were similar in all cell types as was the rate of accumulation of monoubiquitinated FancD2 in response to MMC (Fig. 4b,c and supplementary Fig. S3).

Prolonged DNA damage induced foci formation in ΔRNF4 cells is suppressed by UBE2W inactivation.
Sustained monoubiquitination of FancI in response to MMC in Δ RNF4 cells compared to wild type levels of monoubiquitinated FancI in Δ RNF4 Δ UBE2W cells suggests that a late step in the repair pathway is defective in the absence of Rnf4 when Ube2w is present, but this pathway is restored when Rnf4 and Ube2w are codepleted. Following activation of the FA pathway, the assembly and disassembly of Rad51 filaments is a key step in the repair of ICL DNA damage. We therefore assessed the formation of Rad51 foci as readout for a late step in the repair of damaged DNA. As an early marker of DNA damage we co-stained with γ -H2ax. Cells were either untreated or treated with MMC for 60 minutes and then fixed and stained after 4 or 16H. Quantification of the number of γ -H2ax and Rad51 foci per nucleus over time provides a read out of DNA repair. Untreated wild type and mutant cells exhibit only a few γ -H2ax and Rad51 foci (Fig. 4a-c), but after MMC treatment, all cells lines displayed an increased number of γ -H2ax and Rad51 foci per nucleus. 16H after DNA damage, more than a third of the wild type, Δ UBE2W and Δ RNF4 Δ UBE2W cells had no γ -H2ax or Rad51 foci suggesting that the DNA damage is repaired in these cells (Fig. 5a-c and supplementary Fig. S4). However, in Δ RNF4 cells 16H after recovery, more than 90% of cells show a large number of large γ -H2ax and Rad51 foci, suggesting that ICL DNA repair foci are not resolved efficiently in Δ RNF4 cells. Δ RNF4 Δ UBE2W cells have a similar number of γ -H2ax and Rad51 foci per nucleus as wild type cells indicating that the repair defect in Δ RNF4 cells is rescued by The deleterious effect of Ube2w in the absence of Rnf4 is conserved in human cells. To analyse the genetic interactions between Ube2w and Rnf4 in the DNA damage response of human cells CRISPR/Cas9 technology was used to generate Δ RNF4, Δ UBE2W and Δ RNF4 Δ UBE2W cells (Fig. 6a) These cell lines were treated with DNA damaging drugs and cell survival determined by clonogenic survival assays. Δ RNF4 cells were hypersensitive to hydroxyurea (Fig. 6b), although the defects observed were less pronounced than in chicken cells. As observed in chicken cells, the defect caused by depletion of Rnf4 is rescued by codepletion of Ube2w. Δ UBE2W cells were not sensitised to DNA damage (Fig. 6b). These results suggest that the deleterious effect of Ube2w in the absence of Rnf4 is conserved in human cells and that Ube2w and Rnf4 function in distinct pathways of DNA repair.

Discussion
The role of the ubiquitin E2 conjugating enzyme Ube2w has been rather elusive. While the ubiquitin E2 conjugating enzyme Ube2T has been shown to functionally interacts with the FancL E3 ligase to mono-ubiquitinate FancI and FancD2 in response to ICL 32,33 ,Ube2w was reported to promote monoubiquitination of FancD2 upon UV damage 34 . An in vitro study showed that Ube2w can promote monoubiquitination on FancD2 also in the absence of E3 ligase FancL 11 . Critically, in vivo studies of Ube2w/Ubc16 in several organisms failed to identify a role in DNA damage [34][35][36] . In agreement with these observations we have found that Δ UBE2W cells are not sensitised to any of the DNA damaging agents tested. This is also consistent with data from a high throughput siRNA screen assessing the role of all known E2 conjugating enzymes in the DNA damage response. Knockdown of Ube2w did not lead to a major defect in repair efficiency and did not alter the number of foci per cell staining with 53BP1 and γ -H2ax, although there was a 50% reduction in the number of foci per cell that stained with the anti-ubiquitin antibody FK2 37 . Using a genetic approach in chicken DT40 cells we could show that in the absence of Rnf4 the presence of Ube2w has a deleterious effect on cell survival in response to a range of DNA damaging agents. The precise molecular explanation for this observation has not been established but one possibility is that in combination with an E3 ligase other than Rnf4, Ube2w monoubiquitinates a substrate or substrates in response to DNA damage that acts to prime the substrate for the Rnf4 mediated synthesis of K63 ubiquitin chains that may create a platform for the action of effector molecules. In the absence of Rnf4 the normally transient, monoubiquitinated intermediate created by Ube2w would accumulate and its failure to be further ubiquitinated might block a downstream step in DNA repair. As Δ UBE2W cells are not hypersensitive to DNA damaging agents the monoubiquitination mediated by Ube2w must be redundant and in the absence of Ube2w this modification could be carried out by another E2 ubiquitin conjugating enzyme. In fact it is known that Ube2w can work with multiple E3 ligases and could generate monoubiquitinated substrates in combination with Brca1 38 . Likewise Rnf4 has been shown to functionally interact with multiple ubiquitin E2 conjugating enzymes including Ube2w, Ubc13/Uev1 and UbcH5 27 . Identification of targets for Ube2w mediated monoubiquitination in cells lacking Rnf4 is challenging but would represent a major step forward in understanding the deleterious role of Ube2w in the absence of Rnf4. Hypersensitivity to cisplatin in Fanconi anemia mutant ceIls caused by a failure to repair interstrand crosslinks can be suppressed by inactivating the NHEJ pathway 39,40 . Therefore the suppression of DNA damage hypersensitivity in Δ RNF4 by the inactivation of Ube2w could be explained if Ube2w would be a component of NHEJ pathway. SUMO modification of FancI and FancD2 has been shown to allow recruitment of Rnf4, leading to polyubiquitination of FancI and FancD2 and their removal by the Dvc1-p97 segregase 31 . Early work suggested that Ube2w can promote monoubiquitination of FancD2 upon UV treatment and Ube2w interact with FancL at DNA damage repair foci 34 . However, these experiments have been carried out by using tagged versions of Ube2w which in our hands were non functional in vivo 34 . We observed an increase in monoubiquitination of FancI/D2 upon MMC treatment in Δ RNF4 cells which is suppressed by the inactivation of Δ UBE2W. The role of Ube2w in (a,c) Whole cells extracts of Human HCT116 wild type cells and cells deficient for Rnf4 (Δ RNF4), Ube2w (Δ UBE2W) and Rnf4 and Ube2w (Δ RNF4 Δ UBE2W) were analysed by Western blotting using the indicated antibodies. 3 ng of recombinantly expressed rat Rnf4 and Human Ube2w isoform1 protein were used as a control. Molecular weight marker is indicated on the right inside (kDa). (b) Wild type cells and cells deficient for Δ RNF4; Δ UBE2W and Δ RNF4, Δ UBE2W were subjected to cisplatin, mitomycin C, γ -irradiation and replication stress by HU. The concentration of HU is indicated on the X-axis. Effect on each cell line is indicated of the percentage of colony formation on the Y-axis (logarithmic scale). Data represented as indicated: Wild type (Blue losange), Δ RNF4 (red square), Δ UBE2W (brown circle), Δ RNF4 Δ UBE2W (green triangle). Error bars represent 2 SD.
Scientific RepoRts | 6:26178 | DOI: 10.1038/srep26178 this process is unclear but taken together these observations suggest that the control of post translational modification of FancI/D2 is a critical step controlled by Rnf4 and Ube2w.
It is possible that the increase in monoubiquitination of FancI/D2 observed in Δ RNF4 is due to the lack of efficient DNA repair in these cells. Indeed, we could observe in Δ RNF4 cells upon MMC treatment an increased level of Rad51 foci formation which is an indication of repair delay or failure after ICL damage. However, Rnf4 inactivation in human cells by siRNA leads to a reduction of Rad51 recruitment after DNA double strand break 25,26 . This discrepancy in observations can be explained by differences in the regulation of DNA repair pathways in different cell lines. Chicken DT40 cells have a rapid cell cycle with an extended S-phase that favours homologous recombination, whereas the predominant repair pathway in most human cells is NHEJ. Thus Rnf4 may act on several targets at different stages on the same pathway as suggested by Yin et al. 25 . We observed an increased recruitment of Rad51 after ICL damage which is a different type of DNA damage from that studied in Yin et al. 25 . It is possible that Rad51 has a different role and regulation in these two types of DNA damage. A recent study of a Rad51 dominant mutant revealed its function in ICL repair independently of HR repair 41 . This suggests that Rad51 and its regulators are potential targets for Rnf4 and Ube2w to control DNA repair.
Ube2w is ubiquitously expressed in tissues and cells at variable levels 35,36 . The highest expression level of Ube2w is found in prostate, breast and lung cell lines 35,36 . This enzyme has been identified only in higher eukaryotes 34 . In human and mouse, three isoforms have been described with variable N-terminal and C-terminal extensions due to alternative splicing (NCBI). A recent study has shown that the flexible C-terminus of Ube2w is involved in dimerisation and substrate specificty 42,43 . It is thus possible that different isoforms of Ube2w have different activities although this has not been tested. Ube2w isoform one mediates N-terminal monoubiquitination in vitro 27,28 . However, in cells, the N-terminus of most proteins is processed and post translationally acetylated. As this would block N-terminal ubiquitination by Ube2w, substrates for N-terminal ubiquitination by Ube2w would have to be created by proteolitic cleavage or de novo protein synthesis. A recent study shows that in the case of Trim5α, Ube2w modifies a lysine side chain to prime the synthesis of K63-linked ubiquitin chain in vivo rather than promoting N-terminal mono-ubiquitination 29 . Identifying the target and each Ube2w isoform activity upon DNA damage treatment is a crucial but challenging step in order to understand the molecular mode of action of Ube2w.
In summary this work reveals for the first time that depletion of Rnf4 unmasks a deleterious action of Ube2w in DNA damage repair pathway after ICL, alkylated DNA damage and replication stress and suggests that Ube2w and Rnf4 function in distinct pathways of DNA repair.

Materials and Methods
Generation of RNF4 −/− and UBE2W-/-/-cell lines. The scheme for the generation of the UBE2W gene disruption is outlined in (supplementary Fig. S5). Note, in DT40 cells the UBE2W locus is on the trisomic chromosome II. The targeting construct for UBE2W disruption was generated by amplifying the 5′ homology arm using DT40 genomic DNA as a template and the PCR primer pair 5′ GCAAAATGAT CCACCTCCCG GAATGAC and 5′ GGAATATTGT CACCAGTAAA CATGACC, and cloned into pCR2.1 using the TOPO TA cloning kit (Invitrogen). The 5′ arm was recovered as a NotI/BamHI fragment and cloned into a pBluescript vector to generate pBS-UBE2W5′ arm. The 3′ homology arm was amplified using the primer pair 5′ GTAAGGCAGG ATGGGAGGGA CAGAGTTAG and 5′ GCAGCAAGTT CAGTTATATC ACTGCCATCC, cloned into pCR2.1, recovered as BamHI/EcoRV fragment and cloned into pBS-UBE2W5′ arm. The puromycin (first allele), blasticidin (second allele) and histidinol (third allele) resistance cassettes were inserted into the BamHI restriction site. Targeted integrations were detected by Southern blot analysis of BamHI/KpnI-digested genomic DNA. Generation of RNF4 knock out has been described previously 25 . CRISPR/Cas9 knock out. Gene knock out in human cell using CRISPR/Cas9 technology has been described previously 44,45 . Shortly, humanized Cas9 wild type (Addgene 42229/ pX260) or nickase Cas9-D10A (Addgene 42333/ pX334) combined with gRNA targetting vector were transfected into cells using GeneJuice (Millipore). Each targeting protospacer for human RNF4 and UBE2W were introduced by direct PCR mutagenesis into pA608 (pA608 is modified version of Addgene 41824 guide RNA cloning vector). The sequence used to target Human RNF4 gene is "gacgctttctctgagtagca" (pA622) and "gctactcagagaaagcgtcg" (pA624). The sequence used to target Human UBE2W gene is "gttccatcatggcgtcaatgc" (pA629).
Scientific RepoRts | 6:26178 | DOI: 10.1038/srep26178 Antibody affinity purification. The procedure was described previously 24 . Briefly, 30mg of dialysed recombinant protein untagged HsUbe2w (or ggRnf4) were covalently bound to activated NHS sepharose beads. 5ml of each sheep serum pass onto the affinity column. Bound antibodies were eluted with 0.1M glycine.