The E3 ubiquitin ligase RNF8 transduces the DNA damage signal via an ubiquitin-dependent signaling pathway (original) (raw)

Cell. Author manuscript; available in PMC 2008 Nov 30.

Published in final edited form as:

PMCID: PMC2149842

NIHMSID: NIHMS35444

Michael S.Y. Huen

1 Department of Therapeutic Radiology, Yale University School of Medicine, New Haven, CT 06520

Robert Grant

2 Center for Cancer Research, Depts. of Biology and Biological Engineering, Massachusetts Institute of Technology, Cambridge, MA 021329

Isaac Manke

2 Center for Cancer Research, Depts. of Biology and Biological Engineering, Massachusetts Institute of Technology, Cambridge, MA 021329

Kay Minn

3 Department of Oncology Research, Mayo Clinic College of Medicine, Rochester, MN 55905, USA

Xiaochun Yu

4 Division of Molecular Medicine and Genetics, Department of Internal Medicine, University of Michigan Medical School, Ann Arbor, MI 48109, USA

Michael B Yaffe

2 Center for Cancer Research, Depts. of Biology and Biological Engineering, Massachusetts Institute of Technology, Cambridge, MA 021329

Junjie Chen

1 Department of Therapeutic Radiology, Yale University School of Medicine, New Haven, CT 06520

1 Department of Therapeutic Radiology, Yale University School of Medicine, New Haven, CT 06520

2 Center for Cancer Research, Depts. of Biology and Biological Engineering, Massachusetts Institute of Technology, Cambridge, MA 021329

3 Department of Oncology Research, Mayo Clinic College of Medicine, Rochester, MN 55905, USA

4 Division of Molecular Medicine and Genetics, Department of Internal Medicine, University of Michigan Medical School, Ann Arbor, MI 48109, USA

corresponding authorCorresponding author.

*These authors contributed equally to this study.

Summary

DNA damage signaling utilizes a multitude of post-translational modifiers as molecular switches to regulate cell cycle checkpoints, DNA repair, cellular senescence and apoptosis. Here we show that RNF8, a FHA/RING domain-containing protein, plays a critical role in the early DNA damage response. We have solved the X-ray crystal structure of the FHA domain structure at 1.35Å. We have shown that RNF8 facilitates the accumulation of checkpoint mediator proteins BRCA1 and 53BP1 to the damaged chromatin, on one hand through the phospho-dependent FHA domain-mediated binding of RNF8 to MDC1, on the other hand via its role in ubiquitylating H2AX and possibly other substrates at damage sites. Moreover, RNF8-depleted cells displayed a defective G2/M checkpoint and increased IR sensitivity. Together, our study implicates RNF8 as a novel DNA damage responsive protein that integrates protein phosphorylation and ubiquitylation signaling, and plays a critical role in the cellular response to genotoxic stress.

Introduction

Faithful duplication and segregation of DNA is essential to maintain genomic integrity during cell division. DNA lesions elicit a DNA damage response, which collectively includes DNA repair, activation of cell cycle checkpoints, chromatin remodeling, cellular senescence and apoptosis. Mutations in a variety of components involved in these cellular processes directly contribute to tumorigenesis (Bartkova et al., 2005; Gorgoulis et al., 2005), highlighting the importance of these damage-induced signaling cascades in tumor suppression. Accumulating evidence suggests that the ATM/ATR-dependent phosphorylation of histone variant H2AX to create γH2AX is the initial signal for subsequent accumulation of various mediators/repair proteins to DNA lesions (Bassing et al., 2003; Celeste et al., 2003). A positive feedback loop has been proposed in which ATM/ATR concentrates at γH2AX-containing double strand breaks via MDC1 to further phosphorylate adjacent H2AX molecules and amplify the DNA damage signal (Lou et al., 2006; Stucki et al., 2005). Through this signal amplification step, a number of mediator/repair proteins, including BRCA1 and 53BP1, concentrate to sites of DNA damage to facilitate downstream checkpoint activation.

We and others have previously demonstrated that tandem BRCT domains serve as phospho-peptide binding motifs that mediate protein-protein interactions (Manke et al., 2003; Yu et al., 2003). Specifically, a number of DNA damage response proteins, including BRCA1 and MDC1 (Fernandez-Capetillo et al., 2002; Goldberg et al., 2003; Lou et al., 2003a; Lou et al., 2003b; Stewart et al., 2003), harbor BRCT domains that mediate binding to their respective partners in a phosphorylation-dependent manner (Yu and Chen, 2004; Stucki et al., 2005). In addition to tandem BRCT domains, the FHA domain constitutes a separate class of phospho-peptide binding modules (Durocher et al., 2000). Many FHA domain-containing proteins have been reported to play a role in DNA repair, cell cycle arrest, and pre-mRNA processing (Sun et al., 1998; Li et al., 2000). The reversibility and sequence selectivity of ligand binding afforded by these and other phospho-peptide binding domain-containing proteins allows individual protein-protein interactions that control downstream responses to be tightly regulated in a stimulus-dependent manner.

Recent studies have provided additional insight into the phosphorylation-dependent regulation of the DNA damage signaling network. However, the detailed mechanisms by which the initial γH2AX signal at DNA lesions becomes propagated, amplified, and modified to concentrate checkpoint mediator proteins to these sites remains obscure. Here we report our study of an FHA and Ring domain-containing protein, RNF8, which serves as the molecular linker for communication between the protein phosphorylation and protein ubiquitylation pathways that are crucial for the activation and maintenance of the DNA damage response.

Results

RNF8 is a DNA damage responsive protein

We have previously studied the role of the FHA domain and Ring domain-containing protein Chfr in mitosis (Yu et al., 2005). In the course of these studies we used a protein named RNF8 as a control because it is the only other known mammalian protein that shares a similar domain organization with Chfr (Supplementary Figure 1A). RNF8 was initially reported to interact with Class III human ubiquitin-conjugating enzymes (E2s) through its RING domain (Ito et al., 2001). RNF8 was later shown to bind to the Retinoid X Receptor and regulate its transcriptional activity (Takano et al., 2004). Because several FHA domain-containing proteins are known to play a role in DNA damage signaling, we investigated whether RNF8 or Chfr might participate in the DNA damage response. Cells stably expressing tagged-RNF8 or Chfr were irradiated. Interestingly and surprisingly, RNF8 foci can be readily observed after DNA damage, and these foci co-localized with the DNA damage marker γ-H2AX (Figure 1A). Despite the resemblance of RNF8 and Chfr (Supplementary Figure 1A), we did not observe any Chfr focus formation following DNA damage (Figure 1A), indicating that these two related proteins have distinct cellular functions.

An external file that holds a picture, illustration, etc. Object name is nihms35444f1.jpg

RNF8 is involved in mammalian DNA damage response

A) Localization of tagged RNF8 and Chfr in response to IR. Cells expressing Myc-tagged RNF8 or Chfr were irradiated and immunostained with anti-Myc and anti-pH2AX antibodies. B) Localization of endogenous RNF8 before and after IR treatment in 293T cells. Immunostaining experiments were performed using anti-RNF8 and anti-pH2AX antibodies. C, D) RNF8 relocalizes to chromatin fraction after IR (C), which is reversible following micrococcal nuclease treatment (D). Procedures were carried out as described in Methods and immunoblotting experiments were conducted using indicated antibodies. E) Genetic dependence of RNF8 relocalization following DNA damage. Deficient cells and their respective wild-type counterparts were infected with retrovirus expressing Flag-tagged RNF8. Immunostaining experiments were performed using anti-Flag and anti-γH2AX antibodies. F) The FHA domain, but not the RING domain, of RNF8 targets its localization to DNA damage foci. Cells expressing Flag-tagged wild-type or mutants of RNF8 were mock treated or irradiated and immunostaining were carried out using indicated antibodies.

To confirm the observed IR-induced focus localization of RNF8, we generated a polyclonal antibody specifically recognizing RNF8 (Supplementary Figure 1B). IR-induced foci (IRIF) of endogenous RNF8 can be readily visualized (Figure 1B). The fact that RNF8 foci overlap with those of γH2AX prompted us to speculate that RNF8 might function in the DNA damage response. We therefore examined the localization of RNF8 with several proteins known to be involved in this damage-induced signaling cascade. As expected, RNF8 colocalizes with MDC1, NBS1, 53BP1, BRCA1, pATM and MCPH1, further lending credence to the potential role of RNF8 in the DNA damage response (Supplementary Figure 1C).

The DNA damage-induced focus formation of checkpoint proteins reflects their localization to chromatin structures at the vicinity of DNA breaks. Indeed, increased amount of RNF8 accumulated in the acid extractable fraction after IR treatment (Figure 1C). Moreover, the less-soluble fraction of RNF8 can be released by nuclease treatment (Figure 1D), suggesting that RNF8 accumulates at the chromatin upon DNA damage. Together, our studies suggest that RNF8 is a novel DNA damage responsive protein.

RNF8 acts downstream of H2AX and MDC1 following DNA damage

It is generally accepted that the phosphorylation of histone variant H2AX is the initial signal upon DNA lesion detection. γH2AX is required for the sustained localization of a number of DNA damage mediator/repair factors at chromatin regions at or near the sites of DNA damage (Paull et al., 2000). To delineate where RNF8 fits in the established DNA damage signaling cascade, we examined IRIF formation of RNF8 in a number of human or mouse cells with deficiencies in various DNA damage checkpoint proteins. Our anti-RNF8 antibody could not detect endogenous RNF8 in mouse embryonic fibroblasts (MEFs), so we used retroviral particles containing a RNF8 expression construct to infect these cells (Figure 1E). In sharp contrast to the control wild-type MEFs, no IR-induced RNF8 focus formation was observed in H2AX deficient MEFs or those reconstituted with the S139A phospho-mutant (Figure 1E and Supplementary Figure 1D). Likewise, RNF8 focus formation was also abrogated in MDC1 deficient cells. On the other hand, RNF8 relocalization to γ-H2AX containing foci is not noticeably affected in cells with BRCA1, 53BP1 or NBS1 deficiency (Figure 1E). These data suggest that RNF8 acts downstream of H2AX and MDC1 in the DNA damage responsive pathway.

The FHA, but not its RING domain, targets RNF8 to DNA breaks

The FHA domain is a phospho-protein binding module (Durocher et al., 2000; Li et al., 2000). Figure 1F shows that tagged wild-type RNF8 formed foci that co-localize with γH2AX following IR treatment. Similarly, foci formation can also be observed for the delRING mutant. On the other hand, the FHA deletion mutant (i.e. delFHA) failed to localize to γH2AX containing foci, suggesting that the FHA domain of RNF8 is important for targeting RNF8 to IR-induced DNA damage sites (Supplementary Figure 1E).

The FHA domain of RNF8 selects phospho-motifs similar to those recognized by tandem BRCT domains

FHA domains, like tandem BRCT domains, recognize amino acid sequences extending 3-4 residues around a central phosphorylated amino acid, with selection determined primarily by residues in the third C-terminal (+3) position (Durocher et al., 2000). However, in contrast to BRCT domains which recognize both pSer and pThr-containing sequences, FHA domains appear only to recognize pThr-containing motifs. We determined the optimal phosphopeptide motifs recognized by RNF8 FHA domain using pThr-oriented peptide library screening (Figure 2A). Intriguingly, the RNF8 FHA domain showed strong selection for Tyr and Phe in the +3 position. This selection for aromatic amino acids differs substantially from the acidic and aliphatic residue selection in the +3 position shown by all other FHA domains for which X-ray crystal structures are available (Durocher et al., 2000; Li et al., 2002). Instead, this selection for aromatic amino acids at +3 position closely resembles the optimal phosphopeptide motifs recognized by the tandem BRCT domains of BRCA1 and MDC1 (Manke et al., 2003; Stucki et al., 2005; Yu et al., 2003).

An external file that holds a picture, illustration, etc. Object name is nihms35444f2.jpg

Structural basis for phosphorylation-dependent binding by RNF8 FHA domain

A) Amino acid selectivity values for the RNF8 FHA domain determined using the phosphothreonine-oriented degenerate peptide library MAXXXX-pT-XXXXAKKK, where X indicates all amino acids except Cys. Values ≥ 1.4 indicate moderate selection; values ≥ 2.0 indicate strong selection. B) Cartoon representation of the RNF8 FHA domain bound to the optimal phosphopeptide ELKpTERY. C) stereo view of the phosphopeptide-binding surface. D) Close up of the phosphate binding pocket, with 2Fo-Fc density map contoured at 2σ. A bound water molecule is evident in the upper center. E–G) Molecular interaction surfaces of the RNF8:phosphopeptide complex, the MDC1 tandem BRCT domain:γ-H2AX phosphopeptide complex, and the Rad53 FHA1:LEVpTEAD phosphopeptide complex. Peptide surfaces are contoured in salmon, protein surfaces are contoured in lime. In the RNF8 FHA domain (E), selection for Tyr over Phe in the +3 position likely results from a water-mediated contact between the Tyr hydroxyl and the backbone nitrogen of Leu-57. In the Rad53 FHA1 structure (G), an Arg residue from the FHA domain occupies the equivalent position as the peptide +3 Tyr in the RNF8 structure (dashed line). (H) Divergence in the phospho-amino acid +3 binding surfaces of the FHA domains of RNF8 and Rad53. The Cα traces of the FHA domains of the RNF8 FHA domain:phosphopeptide complex and the Rad53FHA1 domain:phosphopeptide complex were optimally aligned. The phosphopeptide +3 interacting region is shown in cartoon representation, with the RNF8 FHA domain shaded blue, and its bound phosphopeptide shaded cyan, while the Rad53 FHA1 domain is shaded yellow and its bound phosphopeptide is shaded green. The +3 Tyr residue in the RNF8 optimal phosphopeptide, and the +3 Asp in the Rad53 FHA1 optimal phosphopeptide are shown in stick representation. Note that the +3 Tyr binding site in RNF8 is occluded in the Rad53 FHA domain by an Arg residue that mediates selection for Asp in the +3 position.

To investigate the structural basis for this unusual motif selection, we used limited proteolysis to map the boundaries of the FHA domain, and solved the high-resolution structure of the RNF8 FHA:optimal phosphopeptide complex by X-ray crystallography at 1.35Å (Supplementary Table 1). The global fold of the RNF8 FHA domain is a 9 stranded β-sandwich with the phosphopeptide-binding surface comprised of residues in loops that connect the β-strands at one end of the sandwich (Figure 2B), similar to what has been previously observed in the crystal structures of the 11-stranded Rad53 and Chk2 FHA domains (Durocher et al., 2000; Li et al., 2002). The bound phospho-peptide is in an extended conformation with extensive contacts between the peptide backbone and side-chain and main chain atoms from the RNF8 FHA domain (Figure 2C).

Three structural features observed in the RNF8 FHA:phosphopeptide complex are distinct from other FHA domains: First, the RNF8 FHA domain contains two divergent loops and a C-terminal α-helical extension that cluster along one face of the domain, well removed from the phosphopeptide-interacting surface (Figure 2B, shaded red). This region is likely involved in phospho-independent interactions with potential substrates or with additional portions of the full-length RNF8 molecule. Second, the RNF8 FHA domain makes extensive direct contacts to the phosphate group, including a novel bidentate interaction involving Arg-61 that is not observed in any other FHA domain:phosphopeptide crystal structure (Figure 2D). Third, the selectivity for Tyr/Phe in the +3 position results from its interaction with a flat, mostly non-polar surface relatively enriched in sulfur-containing amino acids (Cys-55, Met-58, Val-110 and Leu-112). Interestingly, the general character of the interaction between the +3 Tyr and the surface of the RNF8 FHA domain is strikingly similar to that observed between the +3 Tyr residue of a γH2AX pSer-containing phosphopeptide and the surface of the tandem BRCT domains of MDC1 critical for MDC1 foci formation (Figure 2E and F). On the other hand, the RNF8 FHA domain +3 interacting surface bears little resemblance to the analogous surfaces in other FHA domains (Figure 2G–H). Thus, it appears that the pThr-binding FHA domain of RNF8 has evolved to bind to similar motifs as those recognized by the BRCT domains of the foci-forming proteins BRCA1 and MDC1.

Phosphopeptide binding by the FHA domain is required for RNF8 foci formation

We directly investigated whether phospho-dependent binding was critical for IRIF formation of RNF8. We found that mutation of Arg-61 to Gln reduced FHA domain:phosphopeptide binding by over 160-fold (Supplementary Figure 2A and B). When the full-length RNF8 R61Q mutant protein was introduced into cells, no R61Q foci were observed after radiation damage (Supplementary Figure 2C), indicating that phospho-dependent binding is required for interaction between the RNF8 FHA domain and its upstream binding partner.

In an experiment with cell lysates, wild-type RNF8 could be pulled down with a phospho-Ser-containing peptide derived from γH2AX (Supplementary Figure 2D) but not with the control unphosphorylated peptide. This interaction was totally abolished in experiments with the delFHA or R61Q mutant proteins. Furthermore, consistent with the previous observation that Chfr does not form IRIF, Chfr did not bind to the phosphorylated H2AX peptide (Supplementary Figure 2E). Although RNF8 bound to phospho-H2AX peptides in a pulldown experiment, we failed to detect any direct binding between the RNF8 FHA domain and a phospho-H2AX peptide by isothermal titration calorimetry (data not shown), raising the possibility that the RNF8:γH2AX interaction observed was indirect. Because the BRCT domains of MDC1 mediate its direct binding to phospho-H2AX, and MDC1 is required for RNF8 IRIF, we examined whether RNF8 interacts with MDC1. Intriguingly, the optimal motif for phospho-peptide binding to the RNF8 FHA domain is pTXXY/F (Figure 2A), and inspection of the MDC1 sequence reveals four potential ATM/ATR phophorylation sites that match this motif (T699QCF, T719QAF, T752QPF and T765QPF). One of these TQPF sites was recently reported to be phosphorylated following DNA damage in a large-scale proteomic study (Matsuoka et al., 2007). We therefore synthesized peptides containing each of these four putative phosphorylation sites and measured their binding to the RNF8 FHA domain. Three of the four bound with affinities comparable to the optimal peptide (Supplementary Figure 2F–I), while the fourth bound more weakly. We next generated a deletion mutant spanning residues 698-768 (Del) of MDC1 (Figure 3A) and showed that MDC1, but not Del, specifically bound to purified GST-RNF8 (Figure 3B and 3C). In addition, RNF8 co-precipitated with wild-type but not Del mutant of MDC1 in vivo (Figure 3D), further implicating these putative phosphorylation sites are required for the interaction between RNF8 and MDC1. A control experiment using the delFHA mutant of RNF8 confirms that this interaction also requires the FHA domain of RNF8 (Figure 3E). Similar results were obtained in reciprocal immunoprecipitation assays. Consistent with the role of MDC1 in mediating RNF8 accumulation at DNA damage sites, an increased amount of MDC1 bound to RNF8 after IR, which was abolished with prior phosphatase treatment (Figure 3F). Likewise, wild-type but not Del mutant of MDC1 restored RNF8 IRIF in MDC1-depleted HeLa cells (Supplementary Figure 3 A–B). Collectively, these in vitro and in vivo results support a possible direct role of MDC1 in facilitating RNF8 localization, via a phospho-specific interaction conferred by the RNF8 FHA domain, to the chromatin following DNA damage.

An external file that holds a picture, illustration, etc. Object name is nihms35444f3.jpg

RNF8 is localized to the sites of DNA damage via a FHA-dependent interaction with MDC1

A) Schematic diagram showing full-length MDC1 (WT) and an internal deletion mutant (Del) of MDC1 that abolishes all four putative phosphorylation sites. B) Commassie staining of purified bacterially-expressed GST-RNF8 protein. C) Full length MDC1 but not the deletion mutant (Del) interacts with RNF8 in a pull-down assay. Lysates from 293T cells over-expressing Flag-tagged MDC1 or its deletion mutant were incubated with GST-RNF8 fusion protein immobilized on the glutathione agarose beads for 2 hours before washing and subsequent analysis by Western blotting with anti-Flag antibody. D) MDC1 but not Del mutant of MDC1 co-immunoprecipitates with RNF8. 293T cells were co-transfected with plasmids encoding myc-tagged RNF8 and plasmids encoding SBP-Flag-MDC1 or its deletion mutant. Lysates were incubated with streptavidin beads for 2 hr at 4°C. Thereafter beads were washed three times with NETN, isolates were separated by SDS-PAGE and analyzed by Western blotting using indicated antibodies. E) RNF8 interacts with MDC1 via its FHA domain. Experiments were conducted similar to that described in D) and immunoprecipitation and immunoblotting were carried out as indicated. F) 293T cells were irradiated (10 Gy; 1Gy=100 Rads) or left untreated and cell extract (NETN + 500 mM NaCl) was treated with or without lambda phosphatase prior to diluting and incubating with bacterially expressed 10 μg of GST-RNF8 protein for 2 hr at 4°C. The GST-RNF8 complex was separated by SDS-PAGE to evaluate the amount of endogenous MDC1 that bound specifically to RNF8.

Both the RNF8 FHA and RING domains are required for BRCA1 and 53BP1 IRIF

To further probe the role of RNF8 at DNA damage sites in vivo, we depleted RNF8 using two different siRNAs and tested whether the IRIF of a number of mediator/repair proteins RNF8 dependent. RNF8 knockdown did not affect γH2AX or MDC1 foci formation at DNA damage sites (Figure 4A and Supplementary Figure 4A–C), however, localization of 53BP1 and BRCA1 to foci was abrogated (Figure 4A), suggesting that RNF8 lies upstream of these DNA damage signaling mediator/effectors and facilitates the accumulation of these proteins to the sites of DNA damage.

An external file that holds a picture, illustration, etc. Object name is nihms35444f4.jpg

RNF8 is required for accumulation of BRCA1 and 53BP1 at the sites of DNA damage

A) HeLa cells were transfected twice with either RNF8 siRNAs or a non-targeting control siRNA. 48 hr after the second transfection, cells were treated with 10 Gy IR and recovered for 6 hours before they were fixed and permeabilized. Immunostaining experiments were performed as described in the Experimental Procedures. B) 53BP1 and BRCA1 IRIF formation are restored in MDC1 deficient cells reconstituted with full-length MDC1 but not with the deletion mutant of MDC1. Expression constructs encoding HA-tagged MDC1 (WT) or its deletion mutant (Del) were transiently transfected into MDC1 deficient MEFs. 24 hours post-transfection, cells were irradiated (10 Gy) and immuno-stained with indicated antibodies. C) HeLa cells depleted of endogenous RNF8 using siRNA#2 were infected with viruses encoding siRNA-resistant wild-type, delFHA or delRING mutant of RNF8. Infected cells were then irradiated and processed as described above to visualize protein localization as indicated.

The MDC1/RNF8 interaction experiments presented in Figure 3 imply that MDC1 may interact with RNF8 and regulate RNF8-dependent events following DNA damage. MDC1 was previously shown to be required for IRIF formation of these checkpoint mediator proteins (Goldberg et al., 2003; Lou et al., 2003a; Stewart et al., 2003; also see Supplemental Figure 4D). Here, we examined whether the RNF8/MDC1 interaction is crucial for these events in vivo. As expected, ectopically expressed MDC1 restored the accumulation of BRCA1 and 53BP1 in response to IR in MDC1−/− MEFs. The MDC1 deletion mutation, which abolishes its interaction with RNF8, did not affect its own focus localization following IR but failed to restore the RNF8-dependent concentration of BRCA1 and 53BP1 at the foci (Figure 4B), suggesting that the MDC1/RNF8 interaction is likely to be required for RNF8-dependent functions following DNA damage.

In order to further explore roles of the RNF8 FHA and RING domains in targeting 53BP1 and BRCA1 to foci, we knocked down RNF8 in HeLa cells using siRNF8#2 (Supplementary Figure 4E) and reintroduced full-length RNF8, delFHA or delRING using constructs containing silent mutations within the RNF8 coding sequence which rendered the reintroduced constructs resistant to the siRNA. Unlike the deletion mutants, reconstitution with full-length RNF8 restored 53BP1 and BRCA1 IRIF in cells depleted of endogenous RNF8 (Figure 4C and Supplemental Figure 4F). Thus, both the phosphopeptide-binding and the ubiquitin ligase activity of RNF8 are required for its function in mediating the accumulation of DNA damage checkpoint proteins at the sites of DNA damage.

RNF8 mediates IR-induced damage-associated ubiquitin conjugates

Our observation that IRIF formation of BRCA1 and 53BP1 requires the RNF8 RING domain suggests that the accumulation of these checkpoint proteins is dependent on protein ubiquitylation at the site of the damaged chromatin. The finding that DNA damage-associated ubiquitin conjugates can be visualized using the anti-Ubiquitin FK2 antibody (Morris and Solomon, 2004; Polanowska et al., 2006) is consistent with this hypothesis. If RNF8 is involved in the ubiquitylation of proteins at the damaged chromatin, H2AX and MDC1 deficiencies, which abrogate RNF8 accumulation at IRIF, might be expected to disrupt damage-dependent FK2 focus formation. This is indeed the case (see Supplementary Figure 5A–C). Recently, the E2 ubiquitin conjugating enzyme UBC13 was also implicated in the ubiquitylation of protein(s) at the chromatin following DNA damage (Zhao et al., 2007). However, the E3 ligase, which provides substrate specificity, has yet to be identified. That RNF8 was demonstrated to interact with UBC13 for substrate ubiquitylation (Plans et al., 2006) prompted us to speculate that RNF8 and UBC13 might act in concert in the DNA damage-signaling cascade. In support of this speculation, we found that damage-associated ubiquitin conjugates were absent in either RNF8-depleted or UBC13-depleted cells (Figure 5A and Supplementary Figure 5D). UBC13 depletion also suppressed the accumulation of 53BP1 and BRCA1 at IRIF, but does not affect focus formation of phospho-H2AX and MDC1 (Supplementary Figure 5E). In addition, RNF8 IRIF can be readily visualized in UBC13-depleted cells, indicating that the damage-dependent RNF8 localisation precedes UBC13 function in the DNA damage response. These data, together with previous reports, imply that RNF8 acts with UBC13 to exert a mediator role in the DNA damage-signaling cascade.

An external file that holds a picture, illustration, etc. Object name is nihms35444f5.jpg

RNF8 functions in concert with UBC13 and is important for IR-induced DNA damage-associated ubiquitin conjugates

A) HeLa cells depleted of endogenous RNF8 or UBC13 were irradiated (10Gy) and immunostained with FK2 and γH2AX antibodies. B) IRIF of UIM-containing protein Rap80 is dependent on RNF8 and UBC13. C) IRIF of damage-associated ubiquitin and D) Rap80 foci formation requires RNF8 FHA and RING domains. HeLa cells infected with virus expressing siRNA-resistant full-length RNF8, delFHA or delRING were transiently transfected with siRNF8#2 to deplete endogenous RNF8. 48 hours after the second transfection, cells were fixed and immuno-stained with indicated antibodies.

RNF8 mediates accumulation of the UIM-containing protein Rap80 at sites of DNA breaks

The ubiquitin-interacting motif (UIM) containing protein Rap80 has recently been shown to relocalise to γH2AX-containing foci in a UIM-dependent manner (Kim et al., 2007; Sobhian et al., 2007; Wang et al., 2007). We found that Rap80 focus formation was abolished in RNF8-depleted and UBC13-depleted cells (Figure 5B), and that its localization requires both the RING and FHA domains of RNF8 (Figure 5C–D). Together these results strongly suggest that the ubiquitin conjugates, whose appearance at foci we have shown to be dependent on both RNF8 and UBC13, might serve as a signal for Rap80 accumulation at the sites of DNA damage.

RNF8 is required for H2AX ubiquitylation following DNA damage

Because H2AX phosphorylation is essential for sustained accumulation of MDC1 and RNF8, which in turn is required for localization of other checkpoint proteins including the Rap80-BRCA1 complex and 53BP1 at DNA damage sites, we speculate that H2AX might be ubiquitylated in a RNF8-dependent manner. We first examined whether H2AX can be ubiquitylated in vivo. By overexpressing tagged H2AX and ubiquitin in the cell, we found that some of the tagged H2AX molecules were ubiquitylated (Figure 6A). Moreover, in HeLa cells stably expressing HA-tagged H2AX, we showed that H2AX ubiquitylation is regulated in an IR-dependent manner in vivo, and that depletion of RNF8 abolished the IR-induced H2AX ubiquitylation. Because of the pivotal role phosphorylated H2AX plays in the localization of checkpoint proteins at IRIF, we used an antibody against γH2AX to examine the state of H2AX ubiquitylation specifically in the phosphorylated form. While RNF8 deficiency does not affect IR-induced H2AX phosphorylation, slower migrating bands corresponding to ubiquitylated endogenous γH2AX species were observed in irradiated cells expressing RNF8 but were compromised in the RNF8 knockdown cells (Figure 6C). While RNF8 is clearly responsible for γH2AX di-ubiquitylation, RNF8 knockdown also appears to lower the levels of γH2AX mono-ubiquitylation. Downregulation of UBC13 also significantly reduced damage-induced ubiquitylation of γH2AX (Supplementary Figure 6C), supporting our hypothesis that RNF8 acts with UBC13 in promoting protein ubiquitylation at or near DNA damage sites. In control experiments we have shown that our anti-phospho-H2AX antibody specifically recognizes phosphorylated H2AX species following DNA damage (Figure 6D and 6E). Together, these results point to a role for RNF8 in regulating γH2AX ubiquitylation in response to DNA damage.

An external file that holds a picture, illustration, etc. Object name is nihms35444f6.jpg

RNF8 is required for H2AX ubiquitylation following DNA damage

A) H2AX is ubiquitylated in vivo. 293T cells were transiently transfected with plasmids encoding myc-tagged ubiquitin with or without plasmids encoding SBP-Flag-H2AX. Immunoprecipitation and immunoblotting were carried out using indicated antibodies. Black arrow indicates doubly ubiquitylated species of H2AX, while grey arrow indicates mono-ubiquitinated H2AX. Multiple-ubiquitinated H2AX species are also pointed out. B) HeLa cells stably expressing HA-tagged H2AX were transfected with control siRNA or RNF8 siRNA were treated with 10 Gy or left untreated. Cells were harvested 1 hr post-irradiation. Cell lysates were prepared, separated by SDS-PAGE and blotted with indicated antibodies. C) HeLa cells transfected with control siRNA or RNF8 siRNA were treated as described (B) and immunoblotting experiments were carried out using indicated antibodies. D) IR-induced H2AX ubiquitylation in H2AX+/+ and H2AX−/− MEFs. Cell lysates prepared from wild-type or H2AX−/− cells before and after irradiation were immunoblotted with anti-H2AX and anti-pH2AX antibodies. E) IR-induced H2AX ubiquitylation requires H2AX phosphorylation. H2AX deficient MEFs stably expressing HA-tagged H2AX or S139A mutant of H2AX were treated with 0 Gy or 10 Gy and immunoblotting was performed using indicated antibodies. F) IR-induced H2AX ubiquitylation requires RNF8 FHA and RING domains. Experiments were carried out as that described in Fig. 5g/5h. Immunoblotting experiments were conducted with antibodies as indicated. Arrow indicates ubiquitylated species of H2AX that only appear after radiation in cells expressing wild-type RNF8.

Because H2AX phosphorylation is critical for RNF8 localization to sites of DNA breaks (Supplementary Figure 1D), we asked whether H2AX ubiquitylation is similarly compromised in cells expressing a non-phosphorylatable S139A H2AX mutant protein. Indeed, only H2AX deficient MEFs reconstituted with wild-type H2AX, but not those reconstituted with the H2AX S139A mutant, supported damage-dependent H2AX ubiquitylation (Figure 6E). Similarly, the RNF8-dependent H2AX ubiquitylation is also compromised in MDC1 deficient cells (Supplementary Figure 6A).

We also reconstituted H2AX−/− cells with H2AX K119/120R mutant. While this H2AX mutant abolished the constitutive mono-ubiquitylation of H2AX, cells expressing this mutant still showed damage-induced ubiquitin conjugates formation and IR-induced H2AX ubiquitylation (Supplementary Figure 6D–E), suggesting that IR-induced H2AX ubiquitylation is distinct from those constitutively mono-ubiquitylated H2AX species.

To further evaluate whether the chromatin-associated RNF8 phosphopeptide-binding and E3 ligase activities are required for H2AX ubiquitylation in response to IR, RNF8-depleted HeLa cells expressing siRNA-resistant full-length, delFHA or delRING versions of RNF8 were examined. As shown in Figure 6F, IR-induced H2AX ubiquitylation required both the FHA and RING domains of RNF8.

RNF8 participates in the G2/M checkpoint and loss of RNF8 renders cells sensitive to ionizing radiation

DNA damage checkpoints have evolved to maintain genomic stability by preventing cells with damaged DNA from entering mitosis. Because RNF8 enables the accumulation of multiple checkpoint proteins at the sites of DNA breaks, we tested the effect of RNF8 in IR-induced cell cycle arrest. While cells transfected with control siRNA displayed normal checkpoint activation, RNF8-depleted cells, like those depleted of BRCA1, failed to properly arrest at the G2/M checkpoint upon IR (Figure 7A and Supplementary Figure 7A). Moreover, restoration of the G2/M checkpoint required reintroduction of full-length RNF8 in RNF8-depleted cells (Figure 7B and Supplementary Figure 7B). Finally, depletion of RNF8 also resulted in a notable increase in IR sensitivity (Figure 7B and Supplementary Figure 7C), further supporting a role of RNF8 in cellular response to DNA damage.

An external file that holds a picture, illustration, etc. Object name is nihms35444f7.jpg

RNF8 is required for G2/M checkpoint control and cell survival following ionizing radiation

A) IR-induced G2/M checkpoint is defective in cells with RNF8 depletion and requires both the RNF8 FHA and RING domains. Summary of the percentages of cells stained positive with phospho-H3 antibody before and after IR treatment from three individual experiments. Error bars indicate standard deviation. HeLa cells were transfected with indicated siRNAs and percentages of mitotic cells before and after radiation were determined by FACS analysis as described in Experimental procedures. B) RNF8-depleted cells display increased radiation sensitivity as determined by colony formation assay. Figure represents value obtained from three separate experiments, each performed in triplicate. Error bars indicate standard deviation. C) A proposed model of the DNA damage responsive pathway involving RNF8. The relocalization of the Rap80-BRCA1 complex and 53BP1 requires RNF8-dependent protein ubiquitylation at the chromatin, whereas the accumulation of NBS1 at DNA damage sites is independent of RNF8.

Discussion

In this study, we have uncovered a novel role of RNF8 in the DNA damage response. Our data strongly suggests that RNF8 channels the initial phosphorylation-dependent marks at DNA lesions to promote the accumulation of multiple checkpoint proteins, including Rap80, BRCA1 and 53BP1, which in turn, contribute to its putative role in maintaining genome integrity.

In eukaryotic cells where DNA is tightly packed, the chromatin structure impedes many of the activities that require access to the genetic materials. Despite our understanding of the roles of H2AX phosphorylation in the DNA damage response, it has only become evident recently that chromatin remodeling and other histone modifications also play important functions in this cellular process. Specifically, a role of the ATP-dependent chromatin-remodeling complexes, such as the INO80, SWR1, RSC, and SWI/SNF, and histone acetyltransferase complexes including Trrap-Tip60 complex have been implicated in DNA repair (Murr et al., 2006; Tsukuda et al., 2005). However, unlike those catalysed by the Trrap-Tip60 complex, the impaired loading of DNA checkpoint proteins observed in RNF8-depleted cells could not be counteracted by pre-incubation with sodium butyrate, chloroquine or hypotonic solution (unpublished results). We hypothesize that RNF8 may be required for the accumulation of checkpoint/repair proteins via a different mechanism, i.e. ubiquitylating proteins at the sites of DNA breaks. Indeed, RNF8 is pivotal for the accumulation of ubiquitin conjugates at the damaged chromatin, which depended on its FHA and RING domains. In support of the role of RNF8 in protein ubiquitylation at sites of DNA breaks, we show that the phosphorylation-dependent RNF8 accumulation at γ-H2AX sites is responsible for the IR-induced H2AX di-ubiquitylation. The E3 ligase complex BRCA1-BARD1 has also been implicated to catalyse ubiquitin polymers at the damaged chromatin (Morris and Solomon, 2004; Polanowska et al., 2006; Yu et al., 2006). However, unlike RNF8 and UBC13, BRCA1-depletion did not compromise the IR-induced H2AX ubiquitylation (Supplementary Figure 6B), suggesting that BRCA1 accumulates at the sites of DNA breaks subsequent to the RNF8/UBC13-dependent H2AX ubiquitylation following DNA damage. Finally, we found that ubiquitylation events catalysed by RNF8 is independent of H2AX K119 and K120, illustrating that these damage-induced H2AX ubiquitination are RNF8-dependent but distinct from the constitutive mono-ubiquitylation of H2AX (Supplementary Figure 6D–E).

A scenario is now emerging in which ubiquitination serves as a post-translational modifier in the DNA damage-dependent signaling cascade (Huang and D’Andrea, 2006). Recent studies suggest a role of histone ubiquitination as a means to remodel the chromatin in order to facilitate the accumulation of DNA repair protein (Kapetanaki et al., 2006; Wang et al., 2006). In addition, mono-ubiquitylation of a number of proteins has been demonstrated to be responsible for protein-protein interactions (Garcia-Higuera et al., 2001; Matsushita et al., 2005; Pavri et al., 2006; Stelter and Ulrich, 2003; Taniguchi et al., 2002). Multiple ubiquitin moieties covalently attached via lysine 63 have also been reported to regulate and promote the accumulation of proteins involved in DNA repair (Hoege et al., 2002; Hofmann and Pickart, 1999). More recently, the ubiquitin interacting motif (UIM)-containing protein Rap80 was discovered to play a role in the BRCA1-dependent DNA damage response, serving as an adaptor for BRCA1 accumulation at sites of DNA breaks (Kim et al., 2007; Sobhian et al., 2007; Wang et al., 2007). Our observation that RNF8 mediates both the IRIF of conjugated ubiquitin and Rap80 at sites of DNA breaks implicates that Rap80 might associate with the damaged chromatin by tethering to certain FK2-reacting ubiquitylated protein(s). The fact that the IR-induced H2AX ubiquitylation is similarly compromised in RNF8-depleted cells not only implicates H2AX as a RNF8 substrate, but also suggests that ubiquitin chains on H2AX and other RNF8 substrates might serve as important docking sites during the transduction of the DNA damage signal.

While the ubiquitin-interacting domain (UIM)-containing protein RAP80 is required for BRCA1 localization, RAP80 is dispensable for 53BP1 localization. It would be interesting to test whether any additional ubiquitin-binding protein would serve to tether 53BP1 to DNA damage sites via damage-associated ubiquitin conjugates. Given the role of histone ubiquitylation in chromatin remodeling, it is also plausible that the RNF8-dependent ubiquitylation events at the vicinity of DNA breaks may enhance the accessibility of 53BP1 to modified histones (Huyen et al., 2004; Sanders et al., 2004; Botuyan et al., 2006). In any case, based on the requirement of RNF8 RING domain, we propose that RNF8 facilitates the transduction of the initial phosphorylation-dependent DNA lesion signal by regulating H2AX ubiquitylation, and possibly other substrates, and thus control the localization of the Rap80-BRCA1 complex and other checkpoint proteins.

In summary, our study implicates a critical role of the E3 ubiquitin ligase RNF8 in supporting genome integrity by licensing the assembly of multiple checkpoint/DNA repair proteins at DNA lesions. The link between protein phosphorylation and ubiquitylation revealed in this study highlights the importance of post-translational modifiers as molecular switches that govern, amongst many cellular processes, the DNA damage response pathway in a stimulus-inducible manner. RNF8 is a key player involved in the cross-talk between different protein modifications, with its FHA domain required for its localization to DNA damage foci through a phosphorylation-dependent interaction, and its E3 ligase catalytic domain required for the further accumulation of Rap80, BRCA1 and 53BP1 (Figure 7F). The interplay between these two protein modification cascades may play an essential role in ensuring the proper execution of cellular response to DNA damage.

Experimental Procedures

Antibodies

The RNF8 polyclonal antibody was raised against a GST-RNF8 fusion protein and was affinity purified using an MBP-RNF8 column. Antibodies against the myc epitope, H2AX, γH2AX, BRCA1, MDC1, 53BP1, were previously described (Lou et al., 2003b; Yu et al., 2006). The anti-H3, anti-FK2 and anti-ORC2 antibodies were obtained from Upstate Cell Signaling. Anti-pATM (Ser1981), anti-GAPDH, and anti-HA, anti-UBC13 antibodies were purchased from Calbiochem, Chemicon, Covance and Anaspec respectively. Anti-actin and anti-Flag (M2) were obtained from Sigma.

Cell Culture and Transfection

293T cells were cultured in RPMI 1640 supplemented with 5% fetal calf serum (FCS), 5% bovine serum and 100 U/ml penicillin, and 100 μg/ml streptomycin and maintained in 5% CO2 at 37°C. Cell transfection was performed using Lipofectamine 2000 (Invitrogen) following manufacturer’s protocol.

Immunostaining procedure

To visualize IRIF, cells cultured on coverslips were treated with 10 Gy IR (1 Gy=100 Rads) followed by recovery for 6 hrs. Cells were then washed with PBS, incubated in 3% paraformaldehyde for 12 minutes and permeabilized in 0.5% triton solution for 5 minutes at room temperature. Samples were blocked with 5% goat serum and incubated with primary antibody for 60 minutes. Samples were washed and incubated with secondary antibody for 60 minutes. Cells were then stained with DAPI to visualize nuclear DNA. The coverslips were mounted onto glass slides with anti-fade solution and visualized using a Nikon ECLIPSE E800 fluorescence microscope.

IR sensitivity, G2/M checkpoint assays and Chromatin fractionation

IR sensitivity and G2/M checkpoint assays were performed as described previously (Lou et al., 2003b). Preparation of chromatin fractions were described previously with modifications (Yu et al., 2006). Briefly, cells were harvested at indicated times after treatment and washed once with PBS. Cell pellets were subsequently resuspended in low salt permeabilization buffer (10mM HEPES pH7.4, 10mM KCl, 0.05% NP-40 and protease inhibitors) and incubated on ice for 10 min. Thereafter, nuclei were recovered and resuspended in 0.2M HCl. The soluble fraction was neutralized with 1 M Tris-HCl pH 8.0 for further analysis. For microccocal nuclease (MNase) treatment, nuclei recovered after low salt extraction was washed and resuspended in nuclease reaction buffer (10mM HEPES pH 7.4, 10 mM KCl, 0.5 mM MgCl2, 2 mM CaCl2). 10 U of nuclease was added and incubated for 30 min on ice. Thereafter, the insoluble fraction was treated essential as above to isolate the chromatin-bound proteins.

Supplementary Material

01

Acknowledgments

We thank Duaa Mohammad and J. Vey for technical help, Dr. Andre Nussenzweig for providing valuable reagent, and Dr Jiri Lukas for the MDC1 construct. Portions of this research were carried out at the Stanford Synchrotron Radiation Laboratory, a national user facility operated by Stanford University on behalf of the U.S. Department of Energy, Office of Basic Energy Sciences. This work was supported by grants from the National Institutes of Health (CA89239, CA92312 and CA100109 to J.C. and GM 60594 to M.B.Y.). J.C is a recipient of an Era of Hope Scholar award from the Department of Defence and a member of the Mayo Clinic Breast SPORE program (P50 CA116201). X.Y. is supported by the Developmental fund from the University of Michigan Cancer Center.

Footnotes

Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

References