A genome-wide IR-induced RAD51 foci RNAi screen identifies CDC73 involved in chromatin remodeling for DNA repair

Abstract

To identify new regulators of homologous recombination repair, we carried out a genome-wide short-interfering RNA screen combined with ionizing irradiation using RAD51 foci formation as readout. All candidates were confirmed by independent short-interfering RNAs and validated in secondary assays like recombination repair activity and RPA foci formation. Network analysis of the top modifiers identified gene clusters involved in recombination repair as well as components of the ribosome, the proteasome and the spliceosome, which are known to be required for effective DNA repair. We identified and characterized the RNA polymerase II-associated protein CDC73/Parafibromin as a new player in recombination repair and show that it is critical for genomic stability. CDC73 interacts with components of the SCF/Cullin and INO80/NuA4 chromatin-remodeling complexes to promote Histone ubiquitination. Our findings indicate that CDC73 is involved in local chromatin decondensation at sites of DNA damage to promote DNA repair. This function of CDC73 is related to but independent of its role in transcriptional elongation.

OPEN

ARTICLE

Citation: Cell Discovery (2015) 1, 15034; doi:http://dx.doi.org/10.1038/celldisc.2015.34

Web End =10.1038/celldisc.2015.34 http://www.nature.com/celldisc

Web End =www.nature.com/celldisc

A genome-wide IR-induced RAD51 foci RNAi screen identies CDC73 involved in chromatin remodeling for DNA repair

Patrick Herr1, Cecilia Lundin1, Bastiaan Evers1,6, Daniel Ebner2, Christina Bauerschmidt2, Guy Kingham3, Timea Palmai-Pallag4, Oliver Mortusewicz1, Oliver Frings5, Erik Sonnhammer5, Thomas Helleday1

1Science for Life Laboratory, Division of Translational Medicine and Chemical Biology, Department of Medical Biochemistry and Biophysics, Karolinska Institutet, Stockholm, Sweden; 2Target Discovery Institute, Nufeld Department of Medicine, University of Oxford, Headington, UK; 3CR-UK/MRC Oxford Institute for Radiation Oncology, Department of Oncology, University of Oxford, Oxford, UK; 4School of Life Sciences, University of Lincoln, Brayford Pool, Lincoln, UK; 5Science for Life Laboratory, Bioinformatics Centre Stockholm, Department of Biochemistry and Biophysics, Stockholm University, Stockholm, Sweden

To identify new regulators of homologous recombination repair, we carried out a genome-wide short-interfering RNA screen combined with ionizing irradiation using RAD51 foci formation as readout. All candidates were conrmed by independent short-interfering RNAs and validated in secondary assays like recombination repair activity and RPA foci formation. Network analysis of the top modiers identied gene clusters involved in recombination repair as well as components of the ribosome, the proteasome and the spliceosome, which are known to be required for effective DNA repair. We identied and characterized the RNA polymerase II-associated protein CDC73/Parabromin as a new player in recombination repair and show that it is critical for genomic stability. CDC73 interacts with components of the SCF/Cullin and INO80/NuA4 chromatin-remodeling complexes to promote Histone ubiquitination. Our ndings indicate that CDC73 is involved in local chromatin decondensation at sites of DNA damage to promote DNA repair. This function of CDC73 is related to but independent of its role in transcriptional elongation.

Keywords: genome-wide siRNA screen; RAD51; homologous recombination; tumor suppressor; CDC73; Parabromin; DNA repair; chromatin remodeling

Cell Discovery (2015) 1, 15034; doi:http://dx.doi.org/10.1038/celldisc.2015.34

Web End =10.1038/celldisc.2015.34 ; published online 1 December 2015

Introduction

The DNA damage response is a safeguarding mechanism that ensures maintenance of the genomic integrity of cells. Aberrant DNA repair leads to genomic instability and cancer [1]. Two main pathways have been identied to repair DNA double-strand breaks (DSBs) in the cell. Non-homologous end-joining resolves DSBs by direct ligation of DNA ends

and is therefore more error prone. Homologous recombination repair (HRR) requires an intact sister chromatid as a template [1] enabling error-free repair of DSBs. To nd new regulators involved in HRR, we performed a genome-wide short-interfering RNA (siRNA) screen combined with ionizing radiation (IR). The cellular phenotype we scored by high content imaging was the absence of RAD51 recruitment to sites of DNA damage. RAD51 is a protein specically involved in homology searching and strand pairing [2] and thus is essential in HRR-mediated DSB repair. We thereby aimed to enrich our candidate gene set for HRR-specic factors.

RNA-metabolizing enzymes have previously been identied to be involved in DNA damage response and HRR [3, 4]. Among many known HRR proteins, the

Correspondence: Thomas HelledayTel: [phone]; Fax: [phone]; E-mail: mailto:[email protected]

Web End [email protected]

6Current address: The Netherlands Cancer Institute, 1006 BE Amsterdam, Amsterdam, The Netherlands.

Received 14 September 2015; accepted 16 September 2015

CDC73 in Histone eviction at double-strand breaks 2

top hit of our validation experiments was CDC73, which is encoded by the HRPT2 tumor suppressor gene. CDC73 was shown to be associated with the PAF1/RNA polymerase II transcriptional elongation complex [5].

The chromatin state is pivotal for the ability of transcription factors and polymerases to access DNA. Histone modications such as ubiquitination and methylation have been described to be required for temporary Histone eviction to allow accessibility of the DNA. Monoubiquitination of Lysine 120 on Histone 2B results in temporary displacement of H2B [6] and the ubiquitin ligases RNF20/40 bind CDC73 to mediate H2B monoubiquitination [7]. ATM was shown to directly activate RNF20/40 to mediate monoubiquitination at Lysin 120 in Histone 2B (H2BK120Ub). This direct activation in response to DNA damage is required for fast initiation of repair of DNA lesions [8, 9]. A direct role for RNA polymerase II in the DNA damage response was established through the detection and repair of DNA mismatches[10]. DNA double strand breaks in transcribed regions were shown to be preferentially repaired by HR to ensure precise repair of the lesion [11]. Monoubiquitination of H2B, however, was until recently only linked to ongoing transcription [12] and the mechanisms underlying damage-dependent Histone modication are just beginning to be revealed.

CDC73, as part of the PAF1 complex (PAF1c), is known to be mutated in a variety of cancers like parathyroid carcinoma, breast carcinoma and colorectal carcinoma [13]. Here we describe a novel role for CDC73 in the DNA damage response, which is independent of its transcriptional role with the PAF1c. We found CDC73 to interact with core Histones H2B and H3 and interestingly did we also discover interactions between CDC73 and members of the SCF/Cullin and INO80 chromatin-remodeling complexes like UBA1, CAND1, CUL1, FBXO21 and RUVBL2. These interactions are likely to be required to ensure efcient chromatin remodeling around the DSB to promote accessibility of the DNA for downstream repair events. In live-cell imaging experiments we show that CDC73 is required for efcient displacement of H2B from chromatin. Similar to loss of CDC73 does depletion of SCF/Cullin and INO80 complex components reduce the abundance of H2BK120Ub. We suggest a model in which CDC73 as part of the transcription complex recruits chromatin-remodeling factors of the SCF/Cullin and INO80 complexes and is required to modify the chromatin surrounding the DSB. Our study presents a new

mechanism of action of CDC73, which provides important new insights into the tumor suppressor role of CDC73 in cancer.

Results

Genome-wide siRNA screen for regulators of HRR

HRR is induced by two-ended DSBs (caused by restriction enzymes or IR) or by one-ended DSBs formed at collapsed replication forks, caused, for instance, by the topoisomerase Iinhibitor Camptothecin (CPT) [14], and different enzymes may be involved in the subsequent HRR. Previously, a genome-wide screen using the I-SceI restriction endonuclease to induce HRR between GFP genes in the DRGFP assay was performed [4]. Here, we study the formation of RAD51 foci as readout and compare one-ended and two-ended DSBs. Prior to the genome-wide screen, we carried out a smaller siRNA screen on 600 genes involved in DNA damage response and repair and scored for formation of RAD51 foci after two-ended (using IR) or one-ended (using CPT) DSBs to trigger HRR. To our surprise, we found a high correlation between the genes involved in forming RAD51 foci following CPT and IR (Figure 1a), indicating either that the same proteins are involved in two-ended and one-ended HRR or alternatively that a similar substrate for HRR is formed after both IR- and CPT-induced lesions. We previously demonstrated that IR-induced HRR is predominately triggered by secondary one-ended DSBs formed when replication forks collide into IR-induced lesions [15]. As IR- and CPT-induced HRR are predominately formed by the same one-ended DSBs, we decided to carry out a full genome-wide screen using IR.

For the genome-wide siRNA screen, we used siRNA pools targeting ~ 18 000 genes. Following siRNA knockdown, cells were irradiated, xed after 4 h and stained for RAD51 and DNA content (Figure 1b). The screen was run in duplicates with a good correlation between the runs (R = 0.68) (Supplementary Figure S1A). Negative (scrambled) and positive (RAD51) control siRNAs showed a signicant difference in RAD51 recruitment (Supplementary Figure S1B). Furthermore, knockdown of known HRR proteins like BRCA1, BRCA2 and SHFM1 were found to signicantly reduce levels of RAD51 foci formation, indicating a successful screen (Figure 1c, Supplementary Figure S1C).

HRR is known to be active only in the S/G2 phase of the cell cycle and we therefore excluded cells in the G1 phase from our analysis. Cells in G1

Cell Discovery | http://www.nature.com/celldisc

Web End =www.nature.com/celldisc

Patrick Herr et al.

3

1

Fraction of G2 cells with

RAD51 foci (IR)

0.8

control siRNA

Dharmacon siGenome library

control siRNA

0.6

4 Gy

0.4

68 h

4 h

0.2

0 0.2 0.4 0.6 0.8 1

Reverse transfection of

U2OS cells

IR treatment Fixation and antibody staining

followed by image acquisition Data handling

Fraction of G2 cells with

RAD51 foci (CPT)

10

x104

6

Fraction of G2 cells with RAD51 foci (IR)

6

Amount of cellsMean number of RAD51 foci

4

5

8

5

4

Mean number of RAD51

foci in S/G2 cells

4

3

6

2

3

2

4

2

1

0 1 2 3 4 5 6 7 8 x10

4

2

DAPI signal (IxA)

1

0

S/G2 cells

scramble

RAD51

PALB2

BARD1

BRCA2

RPA1

CHK1

BRCA1

SHFM1

G1 cells

Cell population

Discarded from analysis used in analysis

0

0.5 1 1.5

Relative HR activity

Figure 1 (a) Irradiation (4 Gy) and treatment with CPT (1 M) show very close correlation in RAD51 foci formation. (b) Layout and results of the primary genome-wide siRNA screen. Cells were reverse transfected with a siRNA library and irradiated with 4 Gy 68 h after siRNA transfection. At 72 h after siRNA transfection, cells were xed and imaged with an InCell Analyzer. Final data analysis was performed using MatLab (MathWorks, Natick, MA, USA). (c) Mean RAD51 foci numbers and s.d. for known HR proteins from the two duplicate runs. (d) Cells were sorted according to DNA content by only counting cells in the S/G2 phase. Plot showing mean number of RAD51 foci over a histogram for DAPI intensity. G1 to S-phase transition was set at DAPI intensity where mean number of RAD51 foci is 2 (horizontal blue line). (e) Illustration of late HR proteins in a green shading, which is dened by defective repair of the DRGFP reporter but rather normal recruitment of RAD51.

have no RAD51 foci but as they progress through S-phase the number of RAD51 foci increases (Figure 1d). To only analyze cells in the S/G2 phase, we set our G1 to S-phase cut-off at the DAPI (4,6-diamidino-2-phenylindole) intensity where mean number of RAD51 foci equals 2. Cells in the G1 population were discarded and only data from cells in S/G2 phase were analyzed further. To account for plate-to-plate variation, as well as row and column effects, data were normalized using a three-dimensional (3D)-B score calculation to minimize systematic errors (Supplementary Figure S2AD). As for z-score calculations, the 3D-B score value will be positive or negative depending on raw data being above or below population mean. In our screen, a low 3D-B score represents siRNAs decreasing RAD51 foci formation after irradiation. For the whole screen, the values for 3D-B score fell between 4.3 to +4.4

and almost 600 siRNAs had a 3D-B score lower than 2, that is, 2 s.d. from the mean of the whole screen (Supplementary Figure S1D, Supplementary Table S1).

Hit detection and validation

We compared the results from our screen to the data set from another genome-wide HRR siRNA screen using I-SceI-induced two-ended DSBs and recombination between non-functional GFP genes as readout[4]. Interestingly, we found overall good correlation between the screens (Supplementary Figure S1E), indicating that proteins involved in HRR of two-ended DSBs are likely also involved in HRR of IR-induced lesions. To identify the proteins that suppress both HRR and RAD51 foci formation most effectively, we calculated a HRR score as the product from score values of both screens for each gene. Besides known

Cell Discovery | http://www.nature.com/celldisc

Web End =www.nature.com/celldisc

CDC73 in Histone eviction at double-strand breaks 4

HRR genes, this revealed new potential HRR suppressors (Supplementary Figure S1F). Proteins involved in catalyzing late recombination events are intact in forming RAD51 foci, which persist for a longer time. Hence, the gene products involved late in HR are likely to exhibit high RAD51 foci levels while poorly catalyzing HRR. Interestingly, we nd proteins involved in replication (PCNA, RFC1, POLA1) and Holiday Junction resolution (TOP3A, WRN) producing high levels of RAD51 foci and low HRR activity, in line with these proteins being involved in late HRR steps (Figure 1e).

RNAi screening typically suffers from off-target effects and in particular off-target depletion of RAD51 protein has commonly been observed, resulting in false-positives [4]. Since our readout is RAD51 foci formation after DNA damage induction, our results would be highly affected if RAD51 were depleted in an off-target manner. Therefore, we performed network analysis on our data, assuming that if a gene is part of a RAD51 network and connected to another candidate gene, it is less likely that the effect is due to an off-target activity of the siRNA. First, data from the duplicates were merged by producing a rank list based on 3D-B score for each duplicate screen run and then ranks were summed to get a merged rank list. The number of genes to further analyze was chosen based on their correlation between the duplicate runs. Starting at 50 genes and increasing by 25 for each calculation, we found that a sample size of 200 genes gave the best correlation (Supplementary Figure S3), and therefore the top 200 of the merged rank list were chosen for further functional characterization using pathway analysis (Figure 2a). To characterize the genes in a network context they were mapped to the human FunCoup network (Figure 2b) [16, 17]. The resulting sub-network of 108 genes was further subjected to network cluster analysis using the MGclus method [18]. Using the hypergeometric probability distribution, we found a set of pathways signicantly associated with the 200 genes, notably proteasome, ribosome, HRR and spliceosome (Figure 2c; Supplementary Table S2B). The clustering yielded a list of 7 clusters covering 98 genes, which were annotated using KEGG pathway annotations. This largely agreed with the all-gene analysis, but also revealed which cluster was specically associated with particular pathways (Figure 2c; Supplementary Table S2A). Also, some new pathway associations appeared, such as pathways in cancer. Interestingly, the percentage of gene overlap with previous HRR screens is higher for 108 network connected genes and suggests that ltering our data for

network connections enriches for true-positives and reduces the number of false-positives (Supplementary Tables S3A and S3B).

For experimental validation, we chose 87 candidate genes and used siRNAs from a different vendor than in the primary screen. Selected siRNAs were reverse transfected into U2OS cells using the same schematics as for the primary screen. First, we scored the reproducibility of the decrease in RAD51 foci formation after irradiation. Forty-two of the siRNAs showed a statistically signicant decrease in RAD51 foci formation after IR (Figure 2d; Supplementary Table S4). For further validation of a HRR defect, we evaluated RPA foci formation after IR, as well as HRR activity in the DRGFP assay. SiRNAs for 47 genes resulted in signicant decrease in the number of GFP-positive cells in the HRR assay compared with the control population (Figure 2d; Supplementary Table S4). Thirty-nine of those siRNAs were the same as had scored positive in the RAD51 foci staining (Figure 2d). Finally, we scored RPA foci formation and found that 17 genes had statistically signicant higher level of RPA foci and all of them also showed an HRR defect in the RAD51 foci and DRGFP assays (Figure 2e, Supplementary Table S5). Depleting cells of RAD51 by siRNA lead to RPA foci accumulation (Supplementary Figure S4A) and a strong decrease in GFP+ cells (Supplementary Figure S4B). The top candidate gene from our conrmation experiments was CDC73 and we therefore decided to analyze its function in HRR in more detail.

CDC73 loss generates genomic instability

The top hit from our rescreen CDC73 is encoded by the HRPT2 tumor suppressor gene and has been described as a component of the human PAF1/RNA polymerase II complex [5]. CDC73 is mutated in various types of cancer and absence of staining serves as a diagnostic marker [13]. The mode of action of CDC73 in cancer formation and progression, however, remains largely unknown. As we identied CDC73 as a potential regulator of HRR, we set out to determine the role of CDC73 in the maintenance of genomic integrity. Phosphorylation of ATM (P-ATM) and H2AX (H2AX), both hallmarks of an activated DNA damage response induced on formation of DSBs, was signicantly increased after knockdown of CDC73 (Figure 3a and b). Measuring the tail moment after alkaline single cell electrophoresis (Comet Assay) revealed increased fractionation of DNA in CDC73-depleted cells (Figure 3c), also indicating DSB formation. To rule out an effect on cell cycle

Cell Discovery | http://www.nature.com/celldisc

Web End =www.nature.com/celldisc

Patrick Herr et al.

5

Figure 2 Schematic for how siRNAs were picked for validation. (a) 3D-B scores from the duplicate siRNA screens for RAD51 foci were combined with good correlation (R = 0.76). (b) The top 200 candidate genes were mapped to the human FunCoup network resulting in a subnetwork of 108 genes. (c) Network clustering resulted in 7 sub-clusters from which 87 genes were selected for validation assays. The clusters provide more detailed functional annotation, determined by signicantly enriched KEGG pathways. (d) Forty-two of the siRNAs that were validated showed signicantly lower levels of RAD51 foci compared with control. Of these, 39 also had signicantly lower HR-activity in the DRGFP assay and 17 conrmed an HR-defect in all 3 validation assays (that is, low RAD51 foci levels, increased RPA foci levels and low levels of GFP+ cells in the DRGFP assay). (e) Results of the 17 siRNAs that had signicantly different levels compared with the control in the validation assays. Data showing mean fold values and s.d. from two independent experiments.

Cell Discovery | http://www.nature.com/celldisc

Web End =www.nature.com/celldisc

CDC73 in Histone eviction at double-strand breaks 6

progression of CDC73-depleted cells, we stained cells for PCNA recruitment (Figure 3d) and analyzed incorporation of BrdU (5-bromo-2-deoxyuridine) in combination with propidium iodine by Fluorescence-activated cell sorting (FACS) (Figure 3e). In agreement with the data from our screen, we did not detect an effect on cell cycle progression, nor did we detect increased cleavage of PARP1 as a measure of cell death (Figure 3f). Taken together, depletion of CDC73 leads to increased DSBs and the activation of the DNA damage response.

Previous reports stated effects on cell cycle progression and cell viability upon depletion of CDC73

[19, 20]. The difference to what we observed can be explained by different experimental procedures. As CDC73 is involved in transcriptional elongation, we designed a short-term siRNA-silencing protocol (48 h) at which the CDC73 protein has largely disappeared, as measured by western blot, but secondary effects caused by affecting global transcription are not yet relevant. The use of ve different siRNA sequences allowed us to safely conclude that our results are neither linked to an off-target effect on RAD51 directly nor to depletion of the RAD51 protein due to reduced transcriptional rates (Supplementary Figure S1G).

P=0.027

4 *

H2AX DAPI

ctrlsiRNA

10

average # of H2AX

Foci

3

10

G1 (G): 23.12% S (I): 42.27%

G2 (J): 20.63%

G1 (G): 22.38% S (I): 44.14%

G2 (J): 20.82%

2 1

CDC73siRNA

ctrlsiRNA

CDC73siRNA

ctrlsiRNA

CDC73siRNA

ctrlsiRNA

CDC73siRNA

ctrlsiRNA

FL1

10

ctrlsiRNA

CDC73siRNA

10

*** P=0.0001 P-ATM DAPI

PCNA DAPI

0 200 400 600 800 1000

0 200 400 600 800 1000

FL4

FL4

average # of P-ATM

Foci

12

CDC73siRNA

10

10

8

6

10

FL1

4 2

10

ctrlsiRNA

CDC73siRNA

CDC73siRNA

CDC73

10

*** P=0.0004

60 50 40

average tail moment

30 20 10

CDC73

siRNA:

ctrlsiRNA

35

cells with > 9 PCNA

Foci (%)

30

PARP1

CDC73

Actin

ctrl

25

20

15

10

5

siRNA

ctrlsiRNA

Figure 3 CDC73 silencing leads to genome instability. (a and b) Average number of H2AX and P-ATM foci in U2OS cells depleted for CDC73 is signicantly increased compared with control (*P = 0.027 and ***P = 0.001, respectively). (c) Comet

Assay with siRNA-transfected U2OS cells. CDC73 depletion leads to signicantly increased DNA damage as analyzed by the average tail moment (***P = 0.004). (d) Normal S-phase distribution in CDC73-depleted cells based on PCNA staining. (e) CDC73 knockdown cells do not undergo cell cycle arrest and show a normal rate of BrdU incorporation as analyzed by FACS and propidium iodine counterstain. (f) CDC73 silencing does not lead to apoptosis measured by the amount of cleaved PARP1 on western blot. For all experiments n 2. Average and s.d. are plotted. P-values are calculated with Students t-test and scale bars represent 50 m.

Cell Discovery | http://www.nature.com/celldisc

Web End =www.nature.com/celldisc

Patrick Herr et al.

7

Loss of CDC73 results in delayed repair of DSBs

To study the dynamics of DNA repair in absence of CDC73, we treated U2OS cells with hydroxyurea (HU), which depletes the endogenous pool of dNTPs, subsequently leading to replication stress, stalled replication forks and DSBs [21]. Cells were released after 18-h treatment, and recruitment of H2AX, 53BP1 and RPA was quantied at several time points over 48 h. Foci of all three markers were not resolved after depletion of CDC73 (Figure 4ac). Interestingly, untreated CDC73-decient cells also showed an accumulation of RPA foci over time (Figure 4b). These ndings indicate a failure in resolving DNA damage at the level of or downstream of 5 end resection or alternatively there could be more collision events between transcription and replication and more generation of unprotected single-stranded DNA. Indicative of a role for CDC73 in DNA repair at the level of efcient resection, we did observe a reduction of RPA S33 phosphorylation in HU-treated CDC73-silenced cells (Figure 4d). ATR-mediated RPA phosphorylation serves as a marker for repair factor recruitment [22]. We did not detect any changes in ATR S428 phosphorylation or on another ATR target Chk1 S345, arguing against an effect on the activation of ATR directly (Figure 4d). Alternatively the altered chromatin structure at DNA damage sites in CDC73-depleted cells could hamper phosphorylation of RPA by ATR.

CDC73 is required for efcient H2B monoubiquitination and eviction

Recently the ubiquitin ligases RNF20/40 were implicated in DNA DSB repair by monoubiquitinating H2BK120 [8, 9]. Similar to our observations on the loss of CDC73, loss of RNF20 and RNF40 has been shown to lead to chromosomal instability [23]. Furthermore, CDC73 has been shown to directly bind to RNF20/40[7]. We reasoned that loss of CDC73 might lead to a defect in H2BK120ub and therefore limit the accessibility of repair factors to damaged chromatin (interfere with efcient 5 end resection). Interestingly, we found that CDC73 binds to H2B and H3 in coimmunoprecipitation (co-IP) experiments (Figure 5a), which we conrmed by IPmass spectroscopy (IPMS) (Supplementary Table S6). Furthermore CDC73 silencing lead to a decrease in monoubiquitinated H2B in cells subjected to replication stress with HU (2 mM)

(Figure 5b). This nding correlated with decreased levels of H2BK120Ub also in unstressed cells, which was rescued by re-expression of a siRNA-resistant CDC73 (Figure 5c).

These results indicate that CDC73 may have a role in H2B-mediated chromatin remodeling, and potentially H2B release at sites of DSBs. To study H2B release from chromatin directly after DNA damage induction we performed laser irradiation experiments in H2B-GFP expressing U2OS cells [24]. H2B-GFP dynamics were recorded in live-cells immediately after irradiation. Induction of DNA damage was determined by red uorescent proteinXRCC1 recruitment (Figure 5d). Strikingly, CDC73 silencing signicantly reduced the migration speed of H2B away from the site of damage immediately after irradiation. Expression of a H2BK120 mutant protein resulted in the same decrease promoting our hypothesis (Figure 5e and f). To directly test the release of H2B from the nucleo-some, we did a nucleosome stability assay [25]. At a salt concentration of 0.6M NaCl, we eluted H2B and H3 from the chromatin in control cells. In contrast, we recovered less soluble H2B and H3 in cells depleted for CDC73 (Figure 5g). We conclude that on loss of CDC73, monoubiquitination of H2B and subsequent efcient eviction of H2B from the chromatin is impaired.

CDC73s role in DSB repair is independent of transcriptional elongation

To functionally determine essential domains in CDC73 required to mediate DNA repair, we generated siRNA-resistant constructs of CDC73 harboring deletions in either the CDC73 core domain (C), which is required for transcriptional elongation but does not affect PAF1c assembly [26], or the 227X mutation, a mutation found in patients which codes for a protein unable to bind PAF1c [5] (Figure 6a). All constructs are expressed and in the nucleus (Figure 6b). To study the requirement of these domains specically for HRR, we depleted endogenous CDC73 and reconstituted cells with the rescue constructs. HRR efciency was determined by the DRGFP assay [27]. Not only did the WT construct completely rescue the defect in HRR but also the C mutant fully reconstituted HRR activity (Figure 6c). These ndings indicate that the effect of CDC73 depletion on DSB repair is specic and that the transcriptional function of CDC73 in immediate DNA repair is not important. Supporting this nding, we did not nd other PAF1c components Leo1, Rtf1 or Ctr9 in our screen for RAD51 foci (Supplementary Figure S1E). However PAF1 knockdown does affect recombination efciency as determined in the DRGFP reporter assay (Figure 6d) and affects RAD51 recruitment in our screen (Supplementary Figure S1E). Supporting a necessity of

Cell Discovery | http://www.nature.com/celldisc

Web End =www.nature.com/celldisc

CDC73 in Histone eviction at double-strand breaks 8

a

b

c

d

Figure 4 Loss of CDC73 results in delayed resolution of DNA DSB. Cells were followed over 48 h after being released from HU (2 mM per 12 h) treatment. (ac) 53BP1, RPA and H2AX are not released from sites of damage in CDC73 knockdown cells.

(b) CDC73 knockdown alone steadily increases the accumulation of RPA. (d) RPA phosphorylation is reduced in CDC73 knockdown but ATR expression and activity are not affected (n 2). Average and s.d. are plotted. Scale bars represent 50 m.

CDC73 in the PAF1c or binding to RNA pol II for the repair of H2BK120 does short time (2 h) treatment of cells with the transcription inhibitor DRB, which blocks RNA pol II phosphorylation, lead to the same effect as CDC73 silencing in the H2B eviction experiment (Figure 6e). This nding is furthermore supported by the fact that the CDC73-227X mutant completely fails to rescue the loss of endogenous CDC73 in the DRGFP assay (Figure 6f). Loss of CDC73 in our experiments did not affect the abundance of essential repair proteins like 53BP1, CtIP, PARP1, BRCA1 and RAD51, supporting our observation that the repair defect caused by loss of CDC73 is not due to blocked transcription of essential repair genes (Figure 6g and h). In search for proteins that could account for the HRR defect in the absence of CDC73, we performed Immunoprecipitation of eGFP-tagged CDC73

followed by mass spectrometry. As a measure for the quality of our experiment, we detected all PAF1c components to bind to CDC73-eGFP but not to eGFP alone. Interestingly, we did nd a number of chromatin-remodeling factors belonging to the SCF/Cullin and INO80 complexes. To conrm the hits from the mass spectrometry experiments, we repeated the Immunoprecipitation followed by western blot. Besides the conrmation of the interaction of CDC73 with RNF20, we conrmed interaction with UBA1, RUVBL2, CUL1, CAND1 and FBXO21. RUVBL2 is a component of the INO80 and NuA4 chromatin-remodeling complex, and CUL1, CAND1, and FBXO21 are SCF complex components (Figure 7a). The interactions were not altered after DNA damage (IR, 2 Gy), indicating that these interactions are also involved in canonical transcription (Figure 7b). The

Cell Discovery | http://www.nature.com/celldisc

Web End =www.nature.com/celldisc

Patrick Herr et al.

9

227X mutant that retains only the 227 most N-terminal aminoacids still interacts with the candidate proteins, demonstrating that the PAF1c interaction domain and the chromatin-remodeling domain are separated

(Figure 7b). A direct role for FBXO21, CUL1, CAND1 and RUVBL2 in H2B ubiquitination has not been described yet. We therefore checked the level of H2BK120Ub by western blot and detected a decrease

Cell Discovery | http://www.nature.com/celldisc

Web End =www.nature.com/celldisc

CDC73 in Histone eviction at double-strand breaks 10

Figure 6 The role of CDC73 in HRR is independent of its transcriptional function. (a) Illustration of the siRNA-resistant CDC73, the core mutant (C) and the 227X mutant. (b) Immunouorescence of the eGFP-tagged constructs in U2OS cells in combination with DAPI (blue) and Phalloidin (red). (c) The defect in HRR as measured by the DRGFP assay can be rescued by re-expression of a WT (*P = 0.029) and the C mutant. (d) PAF1 silencing does reduce HRR (*P = 0.009). (e) DRB treatment (50 m, 2 h) has the same effect on H2B mobility as CDC73 silencing (*P = 0.033). (f) The CDC73-227X mutant does not rescue HRR in the

DRGFP assay (**P = 8 10-4). (g) CDC73 knockdown does not affect the expression of a number of essential genes for DNA repair and has no effect on the recruitment of BRCA1 (h). For all experiments (n 2). Average and s.d. are plotted. P-values are calculated with Students t-test and scale bars represent 50 and 10 m.

Figure 5 CDC73 silencing leads to impaired H2B ubiquitination and reduced DNA damage-induced H2B release. (a) CDC73 binds to Histones H2B and H3 in co-immunoprecipitation experiments. PAF1 serves as a positive control. (b) H2BK120Ub is reduced in CDC73 knockdown cells. The effect is enhanced after exposure to HU (2 mM). Calculated ratios (H2Bub/H2B) are indicated below the blot. (c) CDC73 silencing leads to decreased ubiquitination on H2BK120, which is rescued by expression of a siRNA-resistant CDC73 construct (R) (EV, empty vector control transfection). (d) Low energy laser irradiation (405nm) leads to DNA damage-induced recruitment of red uorescent proteinXRCC1. (e) Cells were analyzed for the migration speed of H2B GFP away from the site of DNA damage. Quantication was done using the plot prole tool in ImageJ (inlet) in the indicated area (red square in d) and the distance was measured at a dened gray value for start and end time points (red line). Each experiment was repeated at least 3 times and at least 10 cells were analyzed per experiment (Average and s.d. P-values are calculated with Students t-test and scale bar represents 5 m). CDC73 silencing reduces the speed of H2B migration signicantly (*P = 0.017) and to a similar speed as the H2BK120AGFP mutant (**P = 0.006). (f) Representative images for the quantication in e (tirr,

timepoint of irradiation; t120s, 2 min after irradiation). Images are displayed in pseudocolors for better illustration. (g) H2B and H3 bind tighter to chromatin in absence of CDC73 at a salt concentration of 0.6 mM NaCl. GAPDH and Ku70/86 are shown as soluble controls. Ratios for CDC73 knockdown and H2B and H3 in the soluble fraction are quantied and summarized in the table.

Cell Discovery | http://www.nature.com/celldisc

Web End =www.nature.com/celldisc

Patrick Herr et al.

11

siRNA

siRNA

0.54

siRNA

siRNA

siRNA

FBXO21

siRNA

IP:GFP Input

IP:GFP Input

ctrl

CUL1

ctrl

+ +

CDC73-eGFP

+ +

+ + + + + +

CDC73-227X-eGFP

CAND1

H2B

H2BUb

CAND1

eGFP

+ +

+ +

+ +

IR 2 Gy

CUL1

H2B

UBE1A

Actin

+ +

+ +

CDC73-eGFP

0.65

0.59

H2BUb/H2B

+ +

eGFP

FBXO21

H2B

H2BUb

ctrl

H2BUb

CAND1

UBE1A

Actin

H2B

H2BUb

CUL1 FBXO21

CAND1

CDC73

CUL1

CDC73

Actin

0.83

CDC73

Actin

Actin

RUVBL2

RUVBL2

siRNA

RNF20

GFP

PAF1

siRNA

siRNA

PAF1

siRNA

PAF1

ctrl

RUVBL2

GFP

RUVBL2

H2BUb

ctrl

PAF1

H2B

H2BUb/H2B

CDC73

CDC73 Actin

0.46

Actin

chromatin remodelling complex

HR

chromatin remodeller recruitment

1.2

% GFP + cells

H2B ubiquitination

Ub

CDC73

*

P< 0.05

1.00.80.60.40.2

RPA

Histones

DSB

Histone eviction

Ub

DNA

Ub

- + + + + +

siRNA

ISceI

ctrl

ctrl

FBXO21

CAND1

CUL1

RUVBL2

DNA resection RAD51 recruitment

RAD51

Endonuclease

Figure 7 CDC73 interacts with a number of chromatin-remodeling factors. (a) Co-IP with CDC73 conrmed its interaction with UBE1A, CAND1, CUL1, FBXO21, KU86, RUVBL2 and RNF20. PAF1 serves as a positive control and actin as a negative control. (b) The interactions are unchanged on IR treatment (2 Gy). (c) siRNA-mediated silencing of FBXO21, RUVBL2, CAND1, CUL1 and PAF1 leads to a decrease in H2BK120Ub. Ratios H2BK120Ub:H2B are indicated. (d) siRNA knockdown of CAND1, CUL1 and RUVBL2 leads to a similar decrease in homologous recombination repair as CDC73 knockdown (s.d. of average; more than three repeats). (e) Model describing the mode of action of CDC73 at DSB. CDC73 recruits chromatin-remodeling components that mediate H2BK120Ub and subsequent eviction of H2B from chromatin. The decondensed state of the chromatin allows repair enzymes to engage with DNA and proceed with resection and downstream DNA repair.

on siRNA-mediated silencing similar to the decrease after CDC73 siRNA silencing (Figure 7c), which was also reected in a decrease in the ability to repair DSBs in the HRR reporter cell line (Figure 7d). The effects are very similar to the decrease in H2BK120Ub and HRR activity after CDC73 silencing, and also PAF1 silencing affects H2BK120Ub to a similar extent. The effect on the stability of CDC73 itself is, however, only minor after silencing of CUL1, CAND1 and PAF1 (Figure 7c).

Summarizing our results, we propose a model in which CDC73, as a component of the PAF1c recruits a number of chromatin-remodeling factors which

mediate H2BK120Ub to ensure efcient release of H2B from the chromatin and allow DNA repair factors to access the DNA for the repair of DSBs. This effect is dependent on CDC73 engaging with the PAF1c but is independent of transcription. This mechanism serves as a rapid sensor of DNA damage at transcribed regions and is essential for faithful repair by HR.

Discussion

Here we present a genome-wide siRNA screen for regulators of HRR-mediated DNA repair activated

Cell Discovery | http://www.nature.com/celldisc

Web End =www.nature.com/celldisc

CDC73 in Histone eviction at double-strand breaks 12

after IR. We identied new factors involved in HRR and present the results of our screen as a resource to the scientic community.

In general, HRR activated after IR- and CPT-induced lesions employ similar proteins for the subsequent formation of RAD51 foci, and proteins involved in HRR triggered by restriction endonucleases are similar to those induced by IR. So it appears as no matter how HRR is initiated the subsequent repair is overall similar. Also, IR-induced RAD51 foci form relatively late in S-phase, at a time when a majority of DNA has been replicated (Figure 1d). This is likely a reection of HRR being triggered by secondary DSBs after IR, occurring when replication forks collide with complex (clustered) lesions [15].

We present the characterization of the tumor suppressor CDC73 as an important novel regulator of HRR-mediated DNA repair. Loss of CDC73 leads to increased genomic instability which is likely to be due to paused RNA polymerase, increasing the likelihood of collisions between RNA polymerase and the replication fork. This adds CDC73 to the long list of DSB repair tumor suppressor genes preventing cancer by promoting genome stability.

There is multiple emerging evidence that processes in transcription are closely linked to DNA repair. It is conceivable that decondensation of chromatin and accessibility of the DNA are equally important for both processes. For example, inhibition of proteasomal activity was shown to affect homologous recombination [28], and loss of RNF20 results in genomic instability [23]. Also monoubiquitination on Histone 2B was directly linked to chromosomal stability[29] and it was recently shown that the ubiquitin ligases RNF20 and RNF40 are required for the repair of DSBs [8, 9].

We describe here a novel function for CDC73 in promoting and directing Histone 2B monoubiquitination for the efcient eviction of H2B and DSB repair that is independent of CDC73s transcriptional function. Loss of CDC73 results in decreased levels of H2BK120Ub during DNA repair, less soluble H2B and H3 and a reduced mobility of H2B directly at sites of irradiation (Figure 7e). We discover the interaction of CDC73 with members of the SCF/CUL1 chromatin-remodeling complex and the chromatin remodeler RUVBL2 that is mediated by the N-terminal domain of CDC73. The SCF complex as well as RUVBL2 have been previously implicated in the repair of DNA DSB with RUVBL2 being a direct target of ATM and ATR [3036]. Initial recruitment of repair factors is

not impaired in the absence of CDC73 as neither expression of genes required for resection is affected nor the recruitment of the single-stranded DNA-binding protein RPA and upstream factors. We do, however, observe a defect in the clearance of RPA from single-stranded DNA, a reduction in RPA S33 phosphorylation and subsequently impaired recruitment of RAD51 and repair of DSBs. These results lead us to conclude that on loss of CDC73, the nucleosomes remain in a condensed state presenting a physical barrier for the repair machinery. The nding that the phosphorylation of H2AX and the subsequent recruitment of 53BP1 and RPA to sites of DSBs are not impaired but the resolution of those foci is strongly delayed furthermore supports this statement. We also conclude that the RNF8/RNF168 branch of the DNA damage response[1] is unaffected by loss of CDC73. Instead, the repair machinery is unable to proceed when it encounters the tightly packed chromatin and fails to conclude the repair process (Figure 7e). Another possible scenario could be RUVBL2-mediated nucleosome sliding and CDC73/ PAF1cRNA pol II backtracking to expose damaged DNA as it was recently described for UvrD in Escherichia coli [37].

Taking together, here we present a novel role of CDC73 in DNA repair and genome stability, which is independent of its transcriptional function. We identied a number of chromatin-remodeling factors to interact with CDC73 and that their function on chromatin in this context depends on the recruitment to the PAF1c and thereby to transcribed DNA. We propose that this newly identied function of CDC73 in Histone eviction at sites of DNA damage together with the previously shown role of CDC73 in regulating oncogene expression is important to prevent tumorigenesis.

Materials and Methods

Cell culture

U2OS cells (ATCC, United Kingdom (U.K.), Guernsey, Ireland, Jersey and Liechtenstein #HTB-96) were cultured in DMEM with 10% fetal calf serum, penicillin (60100 g ml1),

streptomycin (100 g ml1) at 37 C containing 5% CO2, in a humidied incubator. Cells were regularly checked for mycoplasma contamination (MycoAlert, Lonza Walkersville, Walkersville, MD, USA).

Antibodies

The following antibodies were used: ATR (sc1887 N19); RAD51 (sc8349); RPA (Cell Signaling Technology, Danvers, MA, USA, 4E4); P-ATM (sc47739); H2B (ab52484); H2BK120Ub

Cell Discovery | http://www.nature.com/celldisc

Web End =www.nature.com/celldisc

Patrick Herr et al.

13

(Cell Signaling Technology, D11); H3 (sc10809); Actin (ab6276); CDC73 (Cell Signaling Technology, D38E12); P-RPA (Bethyl Laboratories NB100-S44); P-ATR (Cell Signaling Technology, S428); P-Chk1 (Cell Signaling Technology, 133D3); Chk1 (Cell Signaling Technology, 261DS); 53BP1 (ab36823); GFP (sc8334); PARP1 (sc7007 F2); KU70 (sc17789 E5); CtIP (Bethyl Laboratories, Inc., Montgomery, TX, USA, A300-488A); BRCA1 (sc642); H2AX (Millipore, Merck Chemicals and Life Science AB, Solna, Sweden, JBW301); PAF1 (ab137519); GAPDH (sc365062); KU86 (sc528); Fibrillarin (ab5821); UBA1 (Cell Signaling, 4890); CAND1 (Bethyl Laboratories, Inc. A302-901A-T); RNF20 (Cell Signaling Technology, 9425); RNF40 (Bethyl Laboratories, A300-719AT); RUVBL2 (Bethyl Laboratories, A302-536A-T); CUL1 (Bethyl Laboratories, A303-373A-T); FBXO21 (ab179818); Flour-555 (A21434, Invitrogen, Life Technologies Europe BV, Stockholm, Sweden); Alexa Flour-488; Phalloidin 594 (Sigma-Aldrich Company Ltd, Dorset, UK, 51927); mouse, rabbit, rat horseradish peroxidase-conjugated ab (Abcam, Cambridge, UK); mouse and rabbit IRDye-conjugated ab (Licor, Lincoln, NE, USA).

Generation of the U2OS-DRGFP and non-homologous end-joining cell lines

The U2OS-DRGFP cell line was generated by transfecting 1 106 U2OS cells with 4 g of the DRGFP vector. Stable cell lines were selected with 3 g ml1 puromycin and single clones were picked, expanded and later checked for GFP expression with or without transfection with an I-SceI endonuclease-expressing vector. A cell line with low (o0.1%) spontaneous expression and high number of GFP-positive cells (43%) after I-SceI-induction was selected. The same protocol was followed for generation of the non-homologous end-joining expressing cell line (Addgene 40026).

FACS analysis

DRGFP: 20 000 Cells were seeded in 12 well plates. 24 h later cells were transfected with a ISceIendonuclease plasmid (150 ng, JetPEI (Polyplus, Bioparc, Illkirch, France)) to induce DSBs. GFP-positive cells were analyzed using FACS (Beckman Coulter Navios, Archimedesvgen, Bromma, Sweden) 48 h after I-SceI transfection as described previously [2, 27]. For rescue experiments, cells were rst transfected with 5 nM siRNA 24 h after seeding. After 8 h, DNA was transfected: CDC73 in combination with I-SceI (150 ng each). Cells were collected for FACS 48 h later. For cell cycle analysis, cells were labeled with BrdU (20) (BD Pharmingen, BD Biosciences, San Jose, CA,

USA, BrdU Flowkit) and counterstained with propidium iodine. Data were analyzed with Kaluza software (Beckman Coulter AB, Bromma, Sweden). More than 20 000 cells were analyzed per experiment.

siRNA screen

The siRNA screen using the Dharmacon (VWR International - Sweden AB, Stockholm, Sweden) human siGENOME siRNA library (druggable, drug targets and genome subsets) was performed using solid-phase transfection plates [5, 38, 39] from CytoPathnder, Tokyo, Japan. These plates contained siRNA (siGENOME SMARTpool, four individual siRNA

sequences per gene, Thermo Scientic), transfection reagent (DharmaFECT 1, #T-2001-02 Thermo Fisher Scientic, Waltham, MA, USA) and accelerator (CytoPathnder) in every well of 384-well plates, except outer columns. U20S cells (1 000 cells per well) were reverse transfected onto 384-well plates at a nal siRNA concentration of 10 nM and incubated for 68 h when plates were irradiated with 4 Gy (1.96 Gy min1, Cs137). At 72 h after siRNA transfection, cells were xed and subsequently stained with RAD51 antibody and nuclear stain (DAPI). All liquid handling steps were performed on a Janus Automated Workstation (Perkin-Elmer Waltham, MA, USA) and automated imaging (four images per channel, two channels per well)) was conducted using an IN Cell Analyzer 1000 (GE Healthcare, Danderyd, Sweden). Images analysis was done using the IN Cell Investigator software (GE Healthcare Life Sciences, Uppasala, Sweden) to score RAD51 foci count as well as DAPI intensity per cell. The genome-wide siRNA screen was run in technical and biological duplicates.

For the rescreen of the 87 selected siRNAs, Ambion siRNA (pool of three individual siRNA sequences per gene) was used from the . For all 3 validation assays, U2OS cells (6000 cells per well on 96-well plate) or U2OS-DRGFP cells (10 000 cells per well) were reverse transfected with siRNA at a concentration of 10 nM. DharmaFect 1 was used as a transfection reagent at a 1:1 000 dilution. Control siRNAs were the same for both primary screen and validation assays, negative controlnon-targeting siRNA from Qiagen, Valencia, CA, USA (#1022076)and positive controlRAD51 siRNA (Thermo Scientic #M--003530-04 and Ambion, Life Technologies Europe BV, Stockholm, Sweden, #4392420). All assays were run in duplicate.

3D-B score normalization

We developed and applied a modication of the B score calculation as described in Malo et al. [40, 41], that we termed 3D-B score normalization. We rst estimate the systematic column, row and plate effects by performing a three-way median polish over the 3D matrix obtained when stacking the matrices obtained from the individual plates on top of each other. This allows for more robust determination of the systematic row and column effects are obtained due to the fact that they are estimated using the whole data set and not on a per-plate basis. Another advantage is that an integral estimation of the plate-to-plate variation. Systematic deviations were substracted from our data set to obtain a 3D matrix with residuals, that is, data that cannot be explained by the systematic effects estimated and therefore must represent true data and noise. Finally, the residuals are scaled to 1.4826 times the median absolute deviation of the median, which in a large sample of a normally distributed population is an estimator for the population s.d.

Immunouorescence

Screen: cells in either 384-well or 96-well plates were xed using 4% paraformaldehyde in PBS-T0.1 (0.1% Triton X-100 in phosphate-buffered saline (PBS)) for 15 min. Cells were washed with PBS twice, then PBS-T0.3 (0.3% Triton X-100) before blocking in 3% bovine serum albumin (in PBS). Incubation with primary antibody overnight at 4 C and with secondary

Cell Discovery | http://www.nature.com/celldisc

Web End =www.nature.com/celldisc

CDC73 in Histone eviction at double-strand breaks 14

antibody for 1 h at room temperature. All washes with PBS, 3 10 min. DNA was stained with DAPI (#D9542, Sigma) or To-Pro-3-Iodide (Invitrogen 642/661) and images were acquired using an IN Cell Analyzer 1000.

For confocal imaging, cells were grown on cover slips. After treatment, cells were washed in PBS, pre-permeabilised in ice cold PBS 0.5% Triton X-100 for 2 min and xed in PBS 4% PFA. Immunostaining was done according to standard protocols and images were acquired on a Zeiss LSM 780 (Zeiss, Jena, Germany). For high content microscopy, cells were grown in 96-well plates (BD Falcon; Fisher Scientic, Pittsburgh, PA, USA) and images were acquired on an Operetta system (Perkin-Elmer). Images were analyzed with Columbus software (Perkin-Elmer).

For live imaging, cells were seeded on imaging plates (Fluoro dish, World Precision Instruments, Sarasota, FL, USA) and transfected with siRNA. After 48 h, cells were incubated in Hoechst (1:1 000) for 5 min and washed in phenol red-free medium. H2BGFP experiment: Laser irradiation was done after 6 frames (stripe width 6 pixel, scan speed 12.65 s per pixel, 20% laser power (405 nm laser), 40 objective with 6 digital zoom) in a heated (37 C, 5% CO2) and humidied chamber on a Zeiss LSM 780 confocal laser scanning microscope equipped with a ultraviolet-transmitting Plan-Apochromat 63/1.40 Oil DIC M27 objective (Zeiss). Images were acquired for 2 min. H2B plasmid was obtained from Addgene (Cambridge, MA, USA; 11680). Data were quantied using the plot prole function in ImageJ (University of Wisconsin-Madison, Madison, WI, USA). More than 10 cells per experiment were analyzed and the experiment was repeated at least 3 times.

Gene enrichment and network analysis

Statistically overrepresented pathway annotations for the initial list of candidate genes (n = 200) were determined using the hypergeometric probability distribution. P-values were adjusted for multiple testing using the strategy described by Benjamini and Hochberg [6, 42]. Pathway annotations were collected from the KEGG database (as of February 2010) [7, 43]. The obtained data set comprised 236 pathways. For further analysis, the candidate genes were mapped to the human FunCoup network (version 1.0) [8, 9, 16, 17] only including links with a condence cut-off 0.5. The network was clustered into local sub-networks using MGclus[10, 18] and functionally annotated as described above.

Comet Assay

U2OS cells were seeded on six well plates (50 k cells per well). siRNA (5 nM) was transfected the day after with Interferin (Polyplus). Cells were collected after 48 h and washed with 1 PBS. Cells were re-suspended in 1 PBS at a concentration of ~ 5 105 cells ml1. About 50 l cell suspension was mixed with 250 l 1.2% low-melting agarose at 37 C. The cell suspension was added to pre-warmed (37 C) agarose-coated fully frosted slides (Thermo Fisher Scientic) and a coverslip was added on top. Slides were kept on ice for 10 min before removing the coverslip and incubated in lysis buffer (10 mM Tris pH 10.0,2.5 M NaCl, 0.1 M EDTA, 10% DMSO and 1% Triton X-100) at 4 C overnight in the dark. The next day, slides were incubated

in alkaline electrophoresis buffer (0.3 N NaOH, 1 mM EDTA) for 30 min. Electrophoresis was run at 300 mA, 25 V for 30 min in electrophoresis buffer using a Comet Assay tank (Thistle Scientic, Glasgow, UK). Slides were washed in neutralization buffer (0.4 M Tris-HCl pH 7.5) and counterstained with 5 M

YOYO-1 dye (Invitrogen). Images were acquired with a 20 or 10 objective in a Zeiss LSM 510 confocal microscope and quantied using CometScore software (TriTek Corp., Sumer-duck, VA, USA). At least 30 comets per sample were analyzed. Tail moment is calculated as percent DNA in the tail multiplied by the tail length.

Molecular cloning

CDC73 ORF was amplied and cloned into pCDNA3.1. C-terminal eGFP tag was introduced by PCR. The CDC73 siRNA-resistant construct as well as the different mutations were cloned by fusion PCR. The CDC73C construct carries a deletion of the C-terminal 188 aminoacids and the CDC73-227X mutant lacks the C-terminal 304 aminoacids. H2BGFP was ordered from Addgene (11680) [24] and red uorescent protein XRCC1 was described earlier [12].

Co-Immunoprecipitation

About 1.5 106 HEK293T cells were seeded in 10 cm2 dishes. GFP-tagged CDC73 constructs were transfected the next day (6 g per plate) according to manufacturer's protocol (JetPEI). After 48 h, cells were collected and lysed in buffer containing DNAseI (1 g l1; 2.5 mM MgCl2). IP was carried out using the GFP trap kit (ChromoTec, Planegg-Martinsried,

Germany). For IPMS, the protein was run on a SDSPAGE and the gel was stained with Coomassie brilliant blue (Bio-Rad). The protein mixture was in gel digested with Trypsin (Promega, Madison, WI, USA) and analyzed on an Orbitrap XL (Thermo Fisher Scientic). Peptides were annotated with MASCOT and peptides with a protein score o31 were discarded.

Western blotting

Western blotting was carried out following standard protocols with Bio-Rad gels and the Trans-Blot Turbo transfer system (Bio-Rad). Cells were lysed in RIPA buffer 20 min on ice followed by sonication with a needle sonicator (Hielscher Ultrasonics GmbH, Teltow, Germany, UP100H; 70% amplitude; 0.7 cycle; 10 cycles). Images were taken at a ChemiDoc XRS system (Bio-Rad) or an LI-COR Odyssey FC (LI-COR, Lincoln, NE, USA). Quantication was done with ImageStudio (LI-COR).

Cell transfections

Transfections were carried out according to manufacturer's protocols. For DNA transfection, jetPEI (Polyplus) was used, and for siRNA transfections interferin was used (Polyplus). siRNA sequences are control; CDC73#1 5-GGATCTCGA

ACACCCATTA-3; CDC73#2 5-CTATCAAGACTGATC TAGA-3; CDC73#3 5-GACCAGTGTTCTTACGGTT-3;

CDC73#4 5-TTCAGTGTCATACCATGGTAGGTTC-3; CDC73#5 5-GGGCACTGCAATTAGTGTTACAGTA-3, PAF1#1 5-ATCACCTGAGCACATGGATTGATCC-3, PAF1#2 5-TAATCATGGCCTGAGACATCATCTC-3.

Cell Discovery | http://www.nature.com/celldisc

Web End =www.nature.com/celldisc

Patrick Herr et al.

15

CDC73 siRNA #4 and #5 and PAF1 siRNAs were previously described in refs [13, 44]. RAD51 siGENOME SMARTpool (Thermo Scientic M-003530-04); RAD51 Silencer Select (Ambion, Invitrogen 4392420); FBXO21 (ON-TARGET plus (23014) SMARTpool, Thermo Scientic); FBXO22 (ON-TARGET plus (26263) SMARTpool); CAND1 (ON-TARGET plus (55832) SMARTpool); CUL1 (ON-TARGET plus (8454) SMARTpool); RUVBL2 (ON-TARGET plus (10856) SMARTpool).

Conict of Interest

The authors declare no conict of interest.

Acknowledgements

We thank T Yoshikawa at CytoPathnder, D Walker and A McCarthy at PharmacoGenomics Innovate Solutions for the solid-phase transfection plates containing the genome-wide siRNA library. M Dahlberg for database construction of RAD51 foci images, PK/KI for mass spectrometry support, K Basler for the CDC73 construct, B Lundgren for siRNA. Members of the Helleday and Altun labs for discussions. The Helleday Laboratory is funded by the Torsten and Ragnar Sderberg Foundation (TH). This project is primarily supported by the European Research Council (TH). Further support from the Swedish Research Council (TH, ES), the Swedish Cancer Society (TH), AXA Research Fund (TH), the Swedish Pain Relief Foundation (TH), the Wenner-Gren Foundation (PH), the Dutch Cancer Society (BE), EMBO LTF (OM) and BBSRC grant BB/K019597/1 (TP-P).

Author contributions

TH and CL designed/developed the bench-top assay. TH, CL and DE carried out the siRNA screens. TH and PH designed and carried out the CDC73 experiments. BE did 3D-B score analysis. PH, CL, OM and TH wrote the manuscript. OF and ES did network analysis.

References

1 Ciccia A, Elledge SJ. The DNA damage response: making it safe to play with knives. Mol Cell 2010; 40: 179204.

2 West SC. Molecular views of recombination proteins and their control. Nat Rev Mol Cell Biol 2003; 4: 435445.3 Paulsen RD, Soni DV, Wollman R et al. A genome-wide siRNA screen reveals diverse cellular processes and pathways that mediate genome stability. Mol Cell 2009; 35: 228239.

4 Adamson B, Smogorzewska A, Sigoillot FD et al. A genome-wide homologous recombination screen identies the RNA-binding protein RBMX as a component of the DNA-damage response. Nat Cell Biol 2012; 14: 318328.

5 Rozenblatt-Rosen O, Hughes CM, Nannepaga SJ et al. The parabromin tumor suppressor protein is part of a human Paf1 complex. Mol Cell Biol 2005; 25: 612620.

6 Pavri R, Zhu B, Li G et al. Histone H2B monoubiquitination functions cooperatively with FACT to regulate elongation by RNA polymerase II. Cell 2006; 125: 703717.

7 Hahn MA, Dickson KA, Jackson S, Clarkson A, Gill AJ, Marsh DJ. The tumor suppressor CDC73 interacts with the ring nger proteins RNF20 and RNF40 and is required for the maintenance of histone 2B monoubiquitination. Hum Mol Genet 2012; 21: 559568.

8 Moyal L, Lerenthal Y, Gana-Weisz M et al. Requirement of ATM-dependent monoubiquitylation of histone H2B for timely repair of DNA double-strand breaks. Mol Cell 2011; 41: 529542.

9 Nakamura K, Kato A, Kobayashi J et al. Regulation of homologous recombination by RNF20-dependent H2B ubiquitination. Mol Cell 2011; 41: 515528.

10 Lindsey-Boltz LA, Sancar A. RNA polymerase: The most specic damage recognition protein in cellular responses to DNA damage? Proc Natl Acad Sci USA 2007; 104: 1321313214.

11 Aymard F, Bugler B, Schmidt CK et al. Transcriptionally active chromatin recruits homologous recombination at DNA double-strand breaks. Nat Struct Mol Biol 2014; 21: 366374.

12 Minsky N, Shema E, Field Y, Schuster M, Segal E, Oren M. Monoubiquitinated H2B is associated with the transcribed region of highly expressed genes in human cells. Nat Cell Biol 2008; 10: 483488.

13 Cui C, Lal P, Master S, Ma Y, Baradet T, Bing Z. Expression of parabromin in major renal cell tumors. Eur J Histochem 2012; 56: e39.

14 Arnaudeau C, Lundin C, Helleday T. DNA double-strand breaks associated with replication forks are predominantly repaired by homologous recombination involving an exchange mechanism in mammalian cells. J Mol Biol 2001; 307: 12351245.

15 Groth P, Orta ML, Elvers I, Majumder MM, Lagerqvist A, Helleday T. Homologous recombination repairs secondary replication induced DNA double-strand breaks after ionizing radiation. Nucleic Acids Res 2012; 40: 65856594.

16 Alexeyenko A, Sonnhammer ELL. Global networks of functional coupling in eukaryotes from comprehensive data integration. Genome Res 2009; 19: 11071116.

17 Alexeyenko A, Schmitt T, Tjrnberg A, Guala D, Frings O, Sonnhammer ELL. Comparative interactomics with Funcoup 2.0. Nucleic Acids Res 2012; 40: D821D828.

18 Frings O, Alexeyenko A, Sonnhammer ELL. MGclus: network clustering employing shared neighbors. Mol Biosyst 2013; 9: 16701675.

19 Wang P, Bowl MR, Bender S et al. Parabromin, a component of the human PAF complex, regulates growth factors and is required for embryonic development and survival in adult mice. Mol Cell Biol 2008; 28: 29302940.

20 Zhang C, Kong D, Tan M-H et al. Parabromin inhibits cancer cell growth and causes G1 phase arrest. Biochem Biophys Res Commun 2006; 350: 1724.

21 Petermann E, Orta ML, Issaeva N, Schultz N, Helleday T. Hydroxyurea-stalled replication forks become

Cell Discovery | http://www.nature.com/celldisc

Web End =www.nature.com/celldisc

CDC73 in Histone eviction at double-strand breaks 16

progressively inactivated and require two different RAD51-mediated pathways for restart and repair. Moi Cell 2010; 37: 492502.22 Vassin VM, Wold MS, Borowiec JA. Replication protein A (RPA) phosphorylation prevents RPA association with replication centers. Mol Cell Biol 2004; 24: 19301943.23 Chernikova SB, Razorenova OV, Higgins JP et al. Deciency in mammalian histone H2B ubiquitin ligase Bre1 (Rnf20/Rnf40) leads to replication stress and chromosomal instability. Cancer Res 2012; 72: 21112119.24 Kanda T, Sullivan KF, Wahl GM. Histone-GFP fusion protein enables sensitive analysis of chromosome dynamics in living mammalian cells. Curr Biol 1998; 8: 377385.25 Xu Y, Sun Y, Jiang X et al. The p400 ATPase regulates nucleosome stability and chromatin ubiquitination during DNA repair. J Cell Biol 2010; 191: 3143.26 Amrich CG, Davis CP, Rogal WP et al. Cdc73 subunit of Paf1 complex contains C-terminal Ras-like domain that promotes association of Paf1 complex with chromatin. J Biol Chem 2012; 287: 1086310875.27 Pierce AJ, Johnson RD, Thompson LH, Jasin M. XRCC3 promotes homology-directed repair of DNA damage in mammalian cells. Genes Dev 1999; 13: 26332638.28 Murakawa Y, Sonoda E, Barber LJ et al. Inhibitors of the proteasome suppress homologous DNA recombination in mammalian cells. Cancer Res 2007; 67: 85368543.29 Sadeghi L, Siggens L, Svensson JP, Ekwall K. Centromeric histone H2B monoubiquitination promotes noncoding transcription and chromatin integrity. Nat Struct Mol Biol 2014, 110.30 Raymond AA, Benhamouche S, Neaud V et al. Reptin regulates DNA double strand breaks repair in human hepatocellular carcinoma. PLoS ONE 2015; 10: e0123333.31 Gospodinov A, Tsaneva I, Anachkova B. RAD51 foci formation in response to DNA damage is modulated by TIP49. Int J Biochem Cell Biol 2009; 41: 925933.32 Matsuoka S, Ballif BA, Smogorzewska A et al. ATM and ATR substrate analysis reveals extensive protein networks responsive to DNA damage. Science 2007; 316: 11601166.33 Moudry P, Lukas C, Macurek L et al. Ubiquitin-activating enzyme UBA1 is required for cellular response to DNA damage. Cell Cycle 2014; 11: 15731582.34 Postow L, Funabiki H. An SCF complex containing Fbxl12 mediates DNA damage-induced Ku80 ubiquitylation. Cell Cycle 2014; 12: 587595.

35 Rosenbaum J, Baek SH, Dutta A et al. The emergence of the conserved AAA+ ATPases pontin and reptin on the signaling landscape. Sci Signal 2013; 6: mr1.

36 Wang H, Zhai L, Xu J et al. Histone H3 and H4 ubiquitylation by the CUL4-DDB-ROC1 ubiquitin ligase facilitates cellular response to DNA damage. Mol Cell 2006; 22: 383394.

37 Epshtein V, Kamarthapu V, McGary K et al. UvrD facilitates DNA repair by pulling RNA polymerase backwards. Nature 2014; 505: 372377.

38 Fujita S, Takano K, Ota E et al. New methods for reverse transfection with siRNA from a solid surface. Methods Mole Biol 2010; 623: 197209.

39 Yoshikawa T, Uchimura E, Kishi M, Funeriu DP, Miyake M, Miyake J. Transfection microarray of human mesenchymal stem cells and on-chip siRNA gene knockdown. J Control Release 2004; 96: 227232.

40 Workman JL. Nucleosome displacement in transcription. Genes Dev 2006; 20: 20092017.

41 Malo N, Hanley JA, Cerquozzi S, Pelletier J, Nadon R. Statistical practice in high-throughput screening data analysis. Nat Biotechnol 2006; 24: 167175.

42 Benjamini Y, Hochberg Y. Controlling the false discovery rate: a practical and powerful approach to multiple testing, J R Stat Soc Ser B 1995; 57: 289300.

43 Ogata H, Goto S, Sato K, Fujibuchi W, Bono H, Kanehisa M. KEGG: kyoto encyclopedia of genes and genomes. Nucleic Acids Res 1999; 27: 2934.

44 Lin L, Zhang J-H, Panicker LM, Simonds WF. The parabromin tumor suppressor protein inhibits cell proliferation by repression of the c-myc proto-oncogene. Proc Natl Acad Sci USA 2008; 105: 1742017425.

(Supplementary information is linked to the online version of the paper on the Cell Discovery website.)

This work is licensed under a Creative Commons Attribution 4.0 International License. The images or other third party material in this article are included in the articles Creative Commons license, unless indicated otherwise in the credit line; if the material is not included under the Creative Commons license, users will need to obtain permission from the license holder to reproduce the material. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/

Web End =http://creativecommons.org/licenses/by/4.0/

Cell Discovery | http://www.nature.com/celldisc

Web End =www.nature.com/celldisc