RAD18 localizes to sites of DNA double-strand breaks
RAD18 is well-known for its function in DNA damage bypass and postreplication repair in yeast and vertebrates through promoting monoubiquitination of proliferating cell nuclear antigen (PCNA) at stalled replication forks 22–25
. Although RAD18 has also been implied in HR repair 26,27
, exactly how RAD18 participates in this process remains elusive. Many proteins involved in DSB-induced checkpoint control and DNA repair physically localize to DNA damage sites. As shown in , RAD18 IRIF was readily detected following ionizing radiation or camptothecin (CPT) treatment, indicating a possible role of RAD18 in DSB response.
RAD18 forms DNA double-strand break-induced foci
Generally, the damage-induced focus formation reflects the assembly of proteins at the vicinity of DNA breaks and hence these proteins become chromatin bound. Biochemical fractionation experiments show that a significant portion of RAD18 shifted from the low salt extractable fraction (soluble fraction) to the acid extracted fraction (chromatin fraction) following IR (). Moreover, the chromatin fraction of RAD18 can be easily released upon nuclease treatment (). These data suggest that RAD18 accumulates onto chromatin after DNA damage.
Damage-induced RAD18 foci formation requires RNF8 E3 ubiquitin ligase
To determine where RAD18 fits in the established DNA damage-signaling cascade, we examined its IRIF formation in an exhaustive panel of cell lines with known genetic defects in DNA damage checkpoint components. In sharp contrast to their respective wild-type counterparts, we failed to detect IR-induced RAD18 focus formation in H2AX, MDC1 or RNF8 deficient cells (). However, RAD18 relocalization to γ-H2AX containing foci was not noticeably affected in cells with BRCA1, 53BP1, NBS1 or Rap80 deficiency (). These data suggest that RAD18 acts downstream of H2AX, MDC1 and RNF8 in the known DNA damage signal transduction pathway.
To further explore the role of RNF8 in targeting RAD18 to IRIF, we established mouse RNF8−/− MEFs stably expressing wild-type, the FHA or RING domain deletion mutants of human RNF8. Work from our lab and others have demonstrated while the FHA domain of RNF8 is required for its own recruitment via its interaction with MDC1, its E3 ligase activity is required for the accumulation of subsequent checkpoint and repair proteins to DSBs 10,12,13
. Indeed, only reconstitution with wild-type RNF8 restored RAD18 IRIF (), suggesting that RAD18 also requires RNF8 E3 ligase activity to be recruited to DSB sites. Likewise, UBC13, the E2 ubiquitin conjugating enzyme that works with RNF8, was also critical for RAD18 foci formation (). Corroborating with its role in targeting RAD18 to DNA damage-induced foci, RNF8 depletion also significantly reduced the accumulation of RAD18 to chromatin fraction after IR ().
RNF8/UBC13 is required for DSB-induced RAD18 recruitment
RNF8 is critical for the formation of IR-induced ubiquitin conjugates at DSBs 10–14,28
, which can be detected by the use of anti-ubiquitin antibody FK2 29,30
. These FK2 foci can be impaired by the depletion of free nuclear ubiquitin achieved by a short treatment of cells with proteasome inhibitor MG132 before irradiation. Consistently, the accumulation of RAD18 at DNA damage sites was abrogated when cell were pretreated with MG132 (Supplementary information, Fig. S1
). Given that RAD18 also exhibits ubiquitin ligase activity, we further tested whether RAD18 itself would contribute to FK2 foci formation. In contrast to RNF8, RAD18 was not required for IR-induced FK2 foci formation and H2AX ubiquitination (). Furthermore, RAD18 was not required for Rap80 or BRCA1 foci formation ().
The Zinc Finger domain of RAD18 targets RAD18 to the sites of DNA breaks
Next, we sought to identify the region(s) within RAD18 that are important for its translocation to IRIF. As shown in , only the Zinc Finger (ZNF) domain deletion mutant (ΔZNF) of RAD18 totally lost its focus forming ability, while wild-type, the RING domain point mutant (C28F mutant; Cys28
changed to Phe), the RING domain deletion mutant (ΔRING), the SAP (S
cinus, and P
IAS) domain deletion mutant (ΔSAP) and the RAD6 binding-domain deletion mutant (ΔR6) 25
all could form IRIF. These observations suggest the RAD18 ZNF domain, but not its E3 ligase activity, is essential for targeting RAD18 to IRIF.
The ZNF domain of RAD18 is required for RAD18 localization to the sites of DNA damage
To further clarify whether the ZNF domain alone is sufficient to target RAD18 to sites of DNA damage, we employed a RAD18 ZNF domain (residues 186–240) that harbors an N-terminal NLS (NLS-ZNF) and a ZNF point-mutation that would disrupt the zinc finger domain structure (NLS-ZNF-C207F) (mutation of Cys207 to Phe207). As shown in , while the NLS-ZNF fusion protein could form foci, the NLS-ZNF-C207F mutant and the SAP domain of RAD18 that also harbors an N-terminal NLS (NLS-SAP; residues 241–290) failed to do so, suggesting that the ZNF domain alone fulfills the role of targeting RAD18 to DSBs. Moreover, similar to the full-length protein, the foci formation of the NLS-ZNF fusion protein also depends on RNF8, MDC1 and H2AX ().
The ZNF domain of RAD18 binds directly to ubiquitin in vitro
RNF8 and UBC13 are known to generate polyubiquitin chains that further recruit ubiquitin-binding proteins to DNA damage sites 31–34
. Recent studies have revealed that some ZNF domains are capable of binding to ubiquitin and have been renamed as ubiquitin binding ZNF (UBZ) domains. Here, we examined whether the ZNF domain of RAD18 would bind to ubiquitin in vitro
. Using an ubiquitin-glutathione S-transferase fusion protein (Ubi-GST), we showed that Ubi-GST specifically bound to wild-type and the SAP domain deletion mutant (ΔSAP) of RAD18, but not to a RAD18 mutant that lacks the ZNF domain (ΔZNF) (). In addition, Ubi-GST pulled down the NLS-ZNF fusion protein, but not NLS-ZNF-C207F mutant in vitro
(). This ubiquitin-binding activity of RAD18 in vitro
is entirely consistent with its ability to localize to damage-induced foci in vivo
, suggesting that RAD18, similar to Rap80, might associate with certain RNF8/UBC13-catalysed ubiquitylated protein(s) at DSBs.
Notably, the RAD18 ZNF domain, but not the Rap80 ZNF domain, has an affinity for different polyubiquitin chains in vitro
(Supplementary Fig. S2a
), implying the ability to bind ubiquitin is specific for the RAD18 ZNF domain. Moreover, GST-RAD18 ZNF domain binds to K48 or K63-linked ubiquitin chains with similar affinities, 36 ± 10 nM for the K63-linked chains, and 17 ± 4 nM for the K48-linked chains (Supplementary Fig. S2b, c
). The significance of this finding is not yet clear.
RAD18 promotes homologous recombination in RNF8-dependent manner
Since RAD18 localization appears to be regulated in response to DSBs, we determined whether RAD18 is required for cell survival following this type of DNA damage. We obtained mouse RAD18−/− MEF cells 23
and established derivative cell lines stably expressing either wild-type or the various deletion/point mutants of human RAD18. Cells deficient in RAD18 were sensitivity to IR or CPT treatment (). Furthermore, cells reconstituted with wild-type or the ΔSAP mutant of RAD18, but not those with the RAD18 ZNF or its RING domain deletion mutants, restored cellular resistance to DNA breaks (). Unexpectedly, the E3 ligase inactivating mutants C28F and ΔR6 still could restore cell survival following IR or CPT treatment (). These results suggest that RAD18 is required for cell survival following DNA double-strand breaks and that both the ZNF and RING domains, but not its E3 ligase activity, are critical for this function of RAD18.
RAD18 promotes homologous recombination
Since CPT-induced replication-associated DSBs are usually repaired by HR repair 35
, the observed requirement of RAD18 in DSB repair may suggest that RAD18 is involved in HR. Indeed, this possibility was raised earlier by studies using chicken DT40 cells 26,27,36,37
. To confirm the role of RAD18 in HR, we performed a gene conversion assay to examine HR efficiency using the DR-GFP reporter system 38
. Significantly, re-introduction of wild-type RAD18 or its SAP domain deletion mutant restored HR repair in RAD18−/− cells, whereas the ZNF or RING domain deletion mutants are defective in this assay (, Supplementary Fig. S3a
). The E3 ligase inactivating mutants C28F and ΔR6 could also restore HR repair (, Supplementary Fig. S3a
), indicating that the E3 ligase activity of RAD18 is not required for its HR function. Consistently, depletion of RAD6A has no effect on gene conversion (Supplementary Fig. S4
RAD18 deficient chicken DT40 cells are hypersensitive to CPT and this hypersensitivity can be reversed by additional inactivation of NHEJ 27
, indicating that RAD18 in chicken cells may regulate NHEJ pathway and thus influence CPT sensitivity. To address whether this is the case in mammalian cells, we performed clonogenic survival assays using RAD18+/+ and RAD18−/− cells exposed to IR or camptothecin alone or in combination with the NHEJ inhibitor NU7441. NU7441 treatment increased the sensitivity in both RAD18 wild-type and deficient MEFs (Supplementary Fig. S5a
), suggesting that unlike the situation in DT40 cells, RAD18 does not appear to participate in NHEJ and it may directly facilitate HR repair in mammalian cells. As a control, we showed that wild-type and RAD18 deficient cells have comparable cell cycle profiles (Supplementary Fig. S6a
). The fact that RAD18 expression increases in S and G2/M phases (see Supplementary Fig. S6b
) is also consistent with its role in HR repair since HR pathway mainly operates in S and G2 phases of the cell cycle.
The recombination protein RAD51 is the key component of the homologous recombination repair machinery and the formation of Rad51 foci can be used as an indicator of HR repair. Indeed, IR-induced Rad51 foci formation was reduced in RAD18−/− MEF cells (, Supplementary Fig. S3b
). We also obtained similar data using human HCT116 and HCT116-RAD18−/− cell lines (data not shown). In agreement with the results from our clonogenic and gene conversion assays, RAD51 foci formation can be restored by reconstitution of RAD18−/− cells with wild-type, the SAP deletion mutant and E3 ligase inactivating mutants C28F and ΔR6, but not with the ZNF or RING domain deletion mutants ().
The striking correlation between the ZNF domain of RAD18 required for its recruitment to DSBs and also for HR repair led us to postulate that RNF8 might be the possible upstream signaling molecule that facilitates RAD18 function in HR. As expected, HR repair and IR-induced Rad51 foci formation were noticeably impaired in RNF8 −/− MEFs (, Supplementary Fig. S3c
). Furthermore, while wild-type RNF8 could restore this HR deficiency, the FHA domain deletion or the RING domain deletion mutants of RNF8 failed to do so (, Supplementary Fig. S3c
RNF8 participates in homologous recombination
Previously report suggests that Chk1 promotes HR repair through directly interacting with and phosphorylating RAD51 39
, raising the possibility that RAD18 might regulate HR repair by influencing Chk1 activation. As shown in , RAD18 was not required for Chk1 activation following ionizing radiation, thus indicating that RAD18 may control HR repair via a novel mechanism.
RAD18 interacts with RAD51C
To explore how RAD18 participates in HR repair, we generated a 293T derivative cell line stably expressing a triple-tagged RAD18 for the identification of potential RAD18-interacting proteins. Following a tandem affinity purification (TAP) scheme, proteins associated with RAD18 were identified by mass spectrometry analysis (Supplementary Table 1
). Results revealed a number of known RAD18-associated proteins, including PCNA, ubiquitin and RAD6A/UBE2A. Interestingly, one of these RAD18-associated proteins is RAD51C, a key component that exists in both Rad51 paralog complexes in human cells 20,21,40–44
We first confirmed the interaction between RAD18 and RAD51C. As shown in , RAD18 interacted with RAD51C, weakly with RAD51D. In addition, RAD51C specifically interacted with GST-RAD18 but not GST alone (). This interaction was also detected in vivo between endogenous proteins before or after irradiation ().
Next we sought to identify the region on RAD51C responsible for its interaction with RAD18 (). While neither the ATPase Walker A/Walker B motifs nor the Linker domain in between were required for this interaction, we found that the very N-terminus of RAD51C was necessary for its binding to RAD18 (). Interestingly, the N-terminus of RAD51C is highly conserved throughout evolution (), suggesting that it may carry out an important function of RAD51C.
Conversely, we analyzed a series of internal deletion/point mutations of RAD18 and found that the RING domain of RAD18 is critical for its interaction with RAD51C (). Moreover, although the interaction between RAD51C and RAD18 depends on the RING domain of RAD18, the conserved Cysteine mutants C28F, C25F and C46F retained the ability to interact with RAD51C, indicating that the RAD18 E3 ligase activity and the intact structure of its RING domain may not be required for this interaction (, Supplementary Fig. S7
). More importantly, the requirement of RAD18 RING domain, but not its E3 ligase activity, for the retention of RAD51C to damaged chromatin () further indicates that RAD18 facilitates the accumulation of RAD51C via a direct protein-protein interaction. Together, these data suggest that a physical interaction between RAD18 and RAD51C may play a significant role in regulating HR repair and explain why the RING domain, but not its endowed E3 ligase activity, is required for its function in HR repair and cell survival following DSBs.
The ability of RAD51C to function in HR correlates with its association with RAD18
RAD51C-deficient cells display HR defects and exhibit hypersensitivity to agents that would induce DNA double-strand breaks 45–47
. Previous studies have also reported a significant reduction of RAD51 foci formation in RAD51C-deficient or depleted cells 48,49
. To explore the physiological relevance of the highly conserved N-terminus of RAD51C, which is required for its binding to RAD18, we used the RAD51C-D1 mutant that lacks this conserved region. Clonogenic assays indicated that only reconstitution with wild-type RAD51C, but not with the RAD51C-D1 mutant, in RAD51C-deficient IRS3 cells restored cell survival following DSBs (). In addition, re-introduction of wild-type RAD51C restored HR repair efficiency to a level comparable with those observed in wild-type CHO-V79 cells, whereas the RAD51C-D1 mutant was defective in this assay (, Supplementary Fig. S3d
). Moreover, we also observed a significant reduction of RAD51 foci formation in RAD51C-D1 reconstituted IRS3 cells when compared with cells reconstituted with wild-type RAD51C ().
RAD18-binding is critical for RAD51C function in homologous recombination
To rule out the possibility that the phenotypes observed in RAD51C-D1 reconstituted IRS3 cells may be due to a failure of forming RAD51 paralog complexes, we performed co-IP experiments. As shown in , the interaction between RAD51B or XRCC3 with RAD51C-D1 is similar to that with wild-type RAD51C. These results suggest that the specific interaction between RAD18 and the N-terminus of RAD51C is likely to be important for RAD51C function in HR repair.
RAD18 participates in two independent DNA damage repair pathways
The known function of RAD18 in vivo is to facilitate PCNA monoubiquitination following DNA damage, especially in response to UV-induced lesions. We first confirmed the role of RAD18 in promoting UV-induced PCNA monoubiquitination (). Since RNF8 is required for RAD18 relocalization following DSBs, we tested whether RNF8 would also be required for RAD18 function following UV damage. Interestingly, UV-induced PCNA monoubiquitination was readily detected in RNF8 depleted HeLa cells or RNF8 deficient MEFs (), indicating that RNF8 is exclusively involved in RAD18 function following DNA double-strand breaks. Indeed, while PCNA monoubiquitination was easily detected upon UV irradiation, the monoubiquitination of PCNA was either absent or very weak in cells treated with CPT or ionizing radiation (). These results indicate that RAD18 may participate in two independent repair processes. We further examined which domains of RAD18 would be required for its function in promoting PCNA monoubiquitination. Intriguingly, while UV-induced PCNA monoubiquitination was largely abrogated in RAD18 deficient cells ectopically expressing either the RAD18 E3 ligase inactivating mutants including ΔRING, C28F and ΔR6 or the SAP domain deletion mutant, PCNA monoubiquitination was readily observed in cells expressing wild-type or the RAD18 ZNF domain deletion mutant (). These results suggest that both the SAP domain and a functional E3 ligase activity, but not the ZNF domain of RAD18 are required for PCNA monoubiquitination. In accordance with tolerance to UV-induced DNA damage involves damage-induced PCNA monoubiquitination, the UV sensitivity of RAD18−/− cells stably expressing RAD18 ZNF deletion mutant was similar to that observed in cells expressing wild-type RAD18. In contrast, RAD18−/− cells expressing the E3 ligase inactivating mutants (ΔRING, C28F or ΔR6) or the SAP domain deletion mutants were more sensitive to UV damage (). Thus, different domains of RAD18 are required for cell survival following UV damage or DNA double-strand breaks, indicating that RAD18 participates in these repair pathways via distinct mechanisms ().
RAD18 participates in two distinct DNA repair pathways