The post‐translational modification of DNA repair and checkpoint proteins by ubiquitin and small ubiquitin‐like modifier (SUMO) critically orchestrates the DNA damage response (DDR). The ubiquitin ligase RNF4 integrates signaling by SUMO and ubiquitin, through its selective recognition and ubiquitination of SUMO‐modified proteins. Here, we define a key new determinant for target discrimination by RNF4, in addition to interaction with SUMO. We identify a nucleosome‐targeting motif within the RNF4 RING domain that can bind DNA and thereby enables RNF4 to selectively ubiquitinate nucleosomal histones. Furthermore, RNF4 nucleosome‐targeting is crucially required for the repair of TRF2‐depleted dysfunctional telomeres by 53BP1‐mediated non‐homologous end joining.
RNF4 is an important ubiquitin ligase in the DNA damage response (DDR) that targets SUMOylated proteins. This study shows that it also contains a nucleosome‐targeting motif that crucially supports the DDR genome‐wide.
RNF4 promotes ATM‐dependent 53BP1 recruitment and repair at dysfunctional telomeres.
RNF4 RING domain contains an RNF168 and RING1b‐related nucleosome‐targeting motif.
RNF4 recognizes both SUMO and nucleosomes to support DNA repair
Post‐translational protein modifications by ubiquitin and small ubiquitin‐like modifier (SUMO) execute critical functions in orchestrating the DNA damage response (DDR) pathways at DNA double‐strand breaks (DSBs) , . For example, the ubiquitin E3 ligases RNF8 and RNF168 polyubiquitinate histones H2A and H2AX , , which like the sumoylation of DSB‐associated substrates such as BRCA1 by the E3 SUMO ligases PIAS1 and PIAS4 , , is a crucial component of the DDR to DSBs.
As at DSBs, ubiquitin signaling also plays a critical role in the cellular response to dysfunctional telomeres. Telomere function is ensured by a protein complex, termed shelterin, bound at telomeric DNA repeats , , , , . A key shelterin constituent is the double‐stranded DNA repeat‐binding protein TRF2 . Depletion of TRF2 activates an ATM‐dependent DDR that leads to phosphorylation of the histone H2AX (γH2AX) and the serial recruitment of MDC1, RNF8, RNF168, 53BP1 and RIF1 , , , , , , . Telomeres depleted for TRF2 become a substrate for the non‐homologous end joining (NHEJ)‐pathway leading to end‐to‐end chromosomal fusions . In the absence of RNF8 or RNF168, 53BP1 cannot be recruited to telomeres, and NHEJ‐mediated telomere fusions are severely impaired , . Although sumoylation by PIAS1/4 has not yet been implicated in the processing of dysfunctional telomeres, the Nse2 (MMS21) E3 SUMO ligase is required for the maintenance of telomere length by the “alternative lengthening of telomeres” (ALT) pathway in telomerase‐deficient cancer cells .
How signaling through the SUMO and ubiquitin pathways is integrated to promote telomere repair is not known. The SUMO‐targeted E3 ubiquitin ligase (STUbL) family of proteins mediates crosstalk between ubiquitin and SUMO. STUbL‐dependent ubiquitination of SUMO conjugated target proteins, coupled or not with target degradation at the proteasome, drives key processes in the maintenance of genome stability , , , . The human STUbL RNF4 was recently demonstrated to act directly in DSB repair, by targeting multiple sumoylated proteins including MDC1, 53BP1 and BRCA1 , , , .
Herein, we demonstrate that RNF4 critically supports the ATM‐ and 53BP1‐dependent NHEJ pathway that promotes the fusion of dysfunctional telomeres. Furthermore, we determine that in addition to its SUMO‐interacting motifs (SIMs), RNF4 requires a previously undefined DNA binding and/or nucleosome‐targeting motif within its RING domain to execute its DNA repair functions.
Results and Discussion
DNA binding by the RNF4 RING domain
Sequence analysis revealed a cluster of basic residues, R177‐K178‐K179, adjacent to the human RNF4 RING domain C‐terminal zinc‐binding site, which is well conserved amongst homologs (Supplementary Fig S1). Like hRNF4, RNF168 and RING1b also contain this three residue basic cluster adjacent to their zinc‐binding site (Fig 1A), while other RING‐type E3 ubiquitin ligases do not (Supplementary Fig S1). Interestingly, this basic cluster supports nucleosome‐specific ubiquitination of H2A by both RNF168 and RING1b, and additionally, DNA binding by RING1b , .
We structurally determined human RNF4120–190 RING domain by both in‐solution small angle X‐ray scattering (SAXS) and by protein crystallography to 2.0 Å resolution (4PPE.pdb; Supplementary Table S1). The latter provided a structure consistent to that previously determined (3NG2.pdb) and an indication of the quality of our recombinant material. Superimposition of the hRNF4 RING domain crystal structure with that of RING1b and RNF168 (PDB codes: 2CKL & 3L11) revealed a well‐conserved architecture at the basic cluster region (Fig 1B). Rigid body‐based docking analyses using DOT ,  of DNA oligonucleotides against both the entire hRNF4 crystal structure and against the entire RNF4:UbcH5:ubiquitin complex structure (PDB code: 4AP4; ) suggested a localized DNA binding site. These predicted interactions are centered at the RKK basic cluster, in addition to a neighboring set of basic residues, His182‐Lys183‐Arg184, occurring in hRNF4 but not RING1b and RNF168 (Fig 1C and Supplementary Fig S2A). Using in‐solution SAXS , we determined that hRNF4 RING domain forms a dimer in solution. Interestingly, the RNF4 structures with the C‐terminus in a folded conformation, either apo‐protein (PDB code 2XEU) or as observed in a complex with UbcH5:ubiquitin (4AP4), are a closer fit to the in‐solution SAXS experimental data than RNF4 structures with the C‐terminus in the extended conformation (Supplementary Fig S2). The in‐solution folded C‐terminus conformation of hRNF4 is consistent with previous studies indicating that this conformation is required for RNF4 activity . We used the SAXS‐based method MONSA that employs a multiphase modeling to fit simultaneously multiple curves, and thus applies a contrast method to define interaction of protein and DNA constituents within the larger complex , . We collected SAXS data on apo‐hRNF4 (Supplementary Fig S3), hRNF4 bound to a 12 bp double‐stranded (ds) DNA oligonucleotide, and the 12 bp oligonucleotide alone (Supplementary Fig S4). The experimental SAXS curve of the hRNF4 RING domain in complex with 12 bp DNA versus the theoretical best linear combination of free DNA and free protein, revealed a distinct difference between these two plots (Supplementary Fig S4), indicating an interaction between the DNA and protein. MONSA provided a further clear indication that the protein and DNA components form an overall complex in solution that is in agreement with DOT‐based predictions (Fig 1D).
We verified the DNA binding property of the hRNF4 RING in vitro, using a biotin‐labeled DNA duplex immobilized on streptavidin beads. Compared to the Glutathione S‐transferase (GST) control, hRNF4 RING bound well to the DNA duplex (Fig 2A). Our molecular docking analyses suggested that K179 within the RKK basic cluster directly interacts with DNA (Fig 1C). In agreement, both our hRNF4 structure and that of the RNF4:UbcH5:ubiquitin  complex show that K179 is solvent exposed and not involved in protein–protein interactions (Fig 1B). Strikingly, a point mutation, hRNF4‐K179D, abolished the interaction between hRNF4 RING and DNA, which has a Kd of approximately 1 μM (Fig 2A, Supplementary Fig S5).
We next compared the activity of wild‐type versus K179D mutant hRNF4 RING in an in vitro E3 ubiquitin ligase assay. Wild‐type and K179D hRNF4 support similar kinetics of diubiquitin and polyubiquitin species formation, indicating that hRNF4‐K179D retains full E3 ubiquitin ligase activity (Fig 2B and D, Supplementary Fig S5). Titration up to 150‐fold molar excess of DNA over hRNF4 into the ubiquitination reaction demonstrated that DNA binding is neither activating nor inhibitory for hRNF4 (Fig 2C).
Together, the foregoing data show that the hRNF4 RING domain contains a non‐specific DNA binding element related to that in the RING1b RING domain . However, we note that DNA binding per se may not be the key function of the hRNF4 basic cluster, as it is structurally related to motifs in the RNF168 and RING1b RING domains that support nucleosome‐targeting , . This structural homology has interesting implications for hRNF4 function in the DDR.
Nucleosome targeting supported by hRNF4 RING basic cluster
As both RNF168 and RING1b utilize their basic clusters to selectively modify histones within the nucleosomal context , , we tested for similar functionality of hRNF4. Assayed against free histone, both wild‐type and K179D hRNF4 strongly promoted H3 ubiquitination (Fig 2D and E). Notably however, whereas wild‐type hRNF4 potently ubiquitinated H3 within assembled nucleosomes, the K179D mutant was strongly defective in this context (Fig 2D and E). Thus, the K179D mutation specifically attenuates the nucleosome‐directed ubiquitin ligase activity of hRNF4. Importantly, the ability of hRNF4 to ubiquitinate nucleosomal H3 is mirrored by the ability of wild‐type, but not K179D, to interact with nucleosomes in a gel shift assay (Supplementary Fig S5). As these nucleosomes lack linker DNA, the hRNF4 RING likely binds nucleosomal DNA and may optimally position hRNF4 to ubiquitinate H3. Overall, these results echo those for RNF168 and RING1b , , indicating the existence of a small subfamily of RING‐type E3 ubiquitin ligases that share a nucleosome‐targeting motif.
A key role for RNF4 at dysfunctional telomeres
We next determined whether RNF4 plays a critical role at dysfunctional telomeres. To obtain complete and synchronous telomere deprotection, we used TRF2 conditional knockout mouse embryonic fibroblasts (MEFs)  that contain an inducible Cre recombinase (TRF2 Flox/Flox, Rosa26‐Cre‐ER; ). Following 4‐hydroxytamoxifen (OHT) treatment, CRE‐mediated recombination eliminates TRF2, thereby deprotecting chromosome ends.
To test whether RNF4 plays a role in the DDR at TRF2‐depleted telomeres, we reduced RNF4 expression by more than 90% using two independent lentiviral shRNA constructs (Supplementary Fig S6). Strikingly, metaphase spread analysis revealed that RNF4 is required for the formation of telomere fusions in TRF2‐depleted cells (Fig 3A, Supplementary Table S2). This was confirmed in terminal restriction fragment analysis by the absence of slower migrating species that correspond to telomere fusions (Fig 3B and C). Given the pivotal role of 53BP1 in telomere fusion , , we tested whether 53BP1 recruitment to dysfunctional telomeres was inhibited in RNF4‐depleted cells. Indeed, attenuating RNF4 activity caused a significant reduction in 53BP1 co‐localization with γH2AX in telomere dysfunction‐induced foci (TIFs; Fig 3D and E, Supplementary Fig S7). These defects in telomere fusion and 53BP1 localization are not explained by changes in cell cycle profile in TRF2‐ and RNF4‐depleted cells (Fig 3F); or a failure to activate the ATM pathway as assessed by CHK2 phosphorylation (Fig 3G). Moreover, the levels of γH2AX, MDC1, RNF168 and bulk ubiquitin conjugates (FK2) in TIFs at TRF2‐depleted telomeres were similar in the presence or absence of RNF4 activity (Supplementary Fig S8). Interestingly, we detected a concurrent increase in the recruitment of BRCA1 into TIFs when RNF4 activity was compromised (Supplementary Fig S7). This result is consistent with the recent finding that 53BP1‐binding protein RIF1 antagonizes recruitment of BRCA1 into TIFs .
hRNF4 nucleosome‐targeting ability is critical to the telomeric DDR
We next tested whether RNF4's nucleosome‐targeting activity is required for the DDR at dysfunctional telomeres by expressing sh‐resistant RNF4 constructs in TRF2‐ and RNF4‐depleted cells (Fig 4A). Cells expressing the RNF4‐K179D mutant allele showed a strong defect in 53BP1 recruitment at telomeres (Fig 4B) and consistent with this, impaired NHEJ‐mediated chromosome fusion (Fig 4C, Supplementary Table S2). A similar phenotype was observed in cells complemented with an empty vector or an RNF4 allele lacking its SUMO‐interacting motifs (SIMs; Fig 4B and C). In contrast, expression of an sh‐resistant wild‐type RNF4 construct fully restored the ability of cells to process dysfunctional telomeres (Fig 4B and C). Together, these data indicate that both hRNF4 nucleosome‐targeting ability and SIM‐dependent SUMO recognition are required for the telomeric DDR.
RNF4 functions in the ATM‐dependent DDR
The critical role for RNF4 at dysfunctional telomeres contrasts with its minor role at DNA DSBs induced by either ionizing radiation (IR) or laser irradiation , , . Distinct from IR‐induced DSBs that first evoke an ATM‐dependent, and then rapidly following 5′–3′ DNA resection, an ATR‐dependent response , TRF2 depletion activates the ATM‐ but not the ATR‐dependent pathway . We therefore speculated that RNF4 plays a crucial role in the ATM‐dependent DDR, which we tested by depleting the telomere‐associated protein TPP1 to trigger ATR but not ATM activation . Interestingly, RNF4 activity was dispensable for the recruitment of 53BP1 at TPP1‐depleted telomeres (Fig 5A and C, Supplementary Fig S9). These data indicate that RNF4 plays an important role in 53BP1 recruitment downstream of ATM, but is dispensable once ATR is activated.
We next determined whether the differential requirement for RNF4 in 53BP1 recruitment extended to global DSB repair, as opposed to reflecting a telomere specific role. To this end, we monitored the formation of 53BP1 DDR foci following IR at both early (ATM‐dependent) and late (ATR‐dependent) time points. RNF4 depletion caused a significant reduction in 53BP1 recruitment at early time points (~10 min), which was not apparent at later time points (Fig 5B and D). Thus, the ATM to ATR transition in the DDR likely explains the somewhat delayed but ultimately normal 53BP1 recruitment following IR treatment of RNF4‐depleted cells (Fig 5B and D ).
The kinetic defect observed in 53BP1 recruitment to DSBs in RNF4‐depleted cells provided an assay for any role of the RNF4 basic cluster in the DDR at DSBs. We tested whether RNF4‐K179D was able to support the recruitment of 53BP1 into early IR‐induced foci, by complementing RNF4‐depleted MEFs with sh‐resistant RNF4 constructs. Whereas wild‐type RNF4 largely complemented the defect in 53BP1 recruitment into DDR foci, both the SIM and RNF4‐K179D mutants were similarly deficient in this response (Fig 5E). Thus, the key role of the hRNF4 RING basic cluster in the telomeric DDR extends to its genome‐wide DDR function.
Here, we have identified RNF4 as a key new player in the telomeric DDR, acting downstream of ATM to promote the recruitment of 53BP1 and subsequent NHEJ‐mediated fusion of chromosomes. Furthermore, we have defined a basic cluster motif in the hRNF4 RING domain that is crucial for this DDR activity. The structural similarity of the hRNF4, RNF168 and RING1b basic clusters that enable selective ubiquitination of nucleosomal histones suggests intriguing functional parallels for these E3 ubiquitin ligases.
Our analyses support a minimally bipartite model for target discrimination by RNF4 in the DDR. The recognition of chromatin bound SUMO conjugates via its intrinsically disordered amino terminus acts as a primary “latch” for RNF4 recruitment , . Subsequently, we propose that the RNF4 RING domain can bind proximal chromatin to ubiquitinate nucleosomal histones and other chromatin‐associated proteins to promote DNA repair (Supplementary Fig S9).
Materials and Methods
Cell culture, telomere analysis, Fluorescence‐activated cell sorting
4‐OHT treatment of TRF2F/F, R26‐CreER MEFs as well as lentiviral and retroviral infections were as previously described . Cells were harvested 60 h post‐addition of 4‐OHT. Telomere Blots, fluorescence in situ hybridization, metaphase analysis and immunofluorescence (IF) were performed using existing protocols . For quantification, at least 200 cells were counted following IF analysis. Cell cycle was determined by propidium iodide staining and acquired on a BD FACScalibur.
SAXS and crystallographic data collection/analysis
Small angle X‐ray scattering data were collected at the SIBYLS beamline . Data set on 15 μl sample volumes of purified hRNF4 RING domain. Initial SAXS data analyses were conducted with ScÅtter 2.0 software (https://bl1231.als.lbl.gov/scatter), and subsequently with the ATSAS software suite . Standard data collection and processing protocols were used for the MONSA analysis, with data being collected on apo‐RNF4, DNA oligonucleotide and RNF4 with DNA oligonucleotide, and shape reconstructions were conducted over multiple independent runs.
Docking calculations were performed using the global, systematic search program DOT, which is part of the DOT2 Suite distributed at: http://www.sdsc.edu/CCMS , , . dsDNA of 10, 15 or 20 nt in length were used in the docking calculations. The 2000 placements with the most favorable interaction energies, calculated as sum of electrostatics and van der Waals intermolecular energy terms, were kept, and the top 20 ranked hits were further analyzed in detail, with the aid of computer graphics. Further details including antibodies and primer sequences are available in Supplementary Materials and Methods.
LMG conceived experiments and carried out the cell biology analyses, and JP performed the ubiquitination assays. MNB, ELD, JJPP and JAT conceived study, analyzed data and wrote manuscript. JJPP, ASA and CH performed protein structural and biophysical analyses. DM and AC performed docking studies. MN executed ubiquitination assays and ITC analyses. TW and KL provided nucleosomal substrates and advice on their use. RPR contributed to biophysical and SAXS data analyses.
Conflict of interest
The authors declare that they have no conflict of interest.
Supplementary Figure S1
Supplementary Figure S2
Supplementary Figure S3
Supplementary Figure S4
Supplementary Figure S5
Supplementary Figure S6
Supplementary Figure S7
Supplementary Figure S8
Supplementary Figure S9
Supplementary Table S1
Supplementary Table S2
Supplementary Material and Methods
M.N.B. is supported by a Scholar Award from the Leukemia & Lymphoma Society. This study was funded by NIH Grants GM068608 and GM081840 awarded to M.N.B., GM088409 to K.L., AR059968 to J.J.P.P., a Pew Scholars Award, the Novartis Advanced Discovery Institute and NIH grant AG038677 to E.L.D. X‐ray scattering and diffraction technologies at the SIBYLS beamline (BL12.3.1) are funded by the IDAT program, supported by DOE Office of Biological and Environmental Research plus NIH Grants GM105404 and CA092584.
FundingNIH GM068608 GM081840 GM088409 AR059968 GM105404 CA092584
- © 2014 The Authors