MDC1 methylation mediated by lysine methyltransferases EHMT1 and EHMT2 regulates active ATM accumulation flanking DNA damage sites

Chromatin dynamics mediated by post-translational modifications play a crucial role in cellular response to genotoxic stress for the maintenance of genome integrity. MDC1 is a pivotal chromatin adaptor in DNA damage response (DDR) and its methylation is essential to recruit repair factors at DNA double-strand break (DSB) sites, yet their precise molecular mechanisms remain elusive. Here we identified euchromatic histone-lysine N-methyltransferase 1 (EHMT1) and EHMT2 as novel regulators of MDC1, which is required for the accumulation of DDR factors e.g. 53BP1 and RAP80, at the DSB sites. MDC1 interacts mainly with EHMT1, which is facilitated by DNA damage-initiated ATM signalling, and EHMT2 dominantly modulates methylation of MDC1 lysine 45. This regulatory modification promotes the interaction between MDC1 and ATM to expand activated ATM on damaged chromatin and dysfunctional telomere. These findings identify EHMT1 and EHMT2 as DDR components, with implications for genome-integrity maintenance through proper dynamic methylation of MDC1.

or KMT1C (lysine methyltransferase 1C)] were identified as mediators for the proper methylation of MDC1. Interestingly, recent studies have shown that EHMT2 7 and its catalytic activity 8,9 promote DNA repair in response to DSB-inducing genotoxic stress. However, the precise mechanism of EHMT2 linked to EHMT1 in DDR remain elusive. Here, we show that EHMT1 and EHMT2-dependent methylation of MDC1 is required for the interaction between activated ATM and MDC1 to amplify the DDR signal, followed by recruitment of repair factors to DNA damage sites including the dysfunctional telomere. These data enhance our understanding of EHMT1 and EHMT2 as DDR components, through proper dynamic methylation of MDC1 to maintain genome integrity.

Results
EHMT1 and EHMT2 are required for recruitment of MDC1-downstream factors to DNA damage sites. In an attempt to identify a potential lysine methyltransferase (KMT) targeting MDC1 within the DDR, we monitored the accumulation of its downstream factors in the DNA damage-induced foci in U2OS cells with small interfering (si)RNA-mediated knockdown of KMTs. We treated cells with neocarzinostatin (NCS), a radiomimetic DNA damage agent, and observed the ubiquitin conjugates immunodetected by the anti-ubiquitin antibody FK2 and the accumulation of 53BP1 at the DNA damage sites (Fig. S1a,b). Among the 9 enzymes tested, the depletion of EHMT1, EHMT2 and SETD8, but not SUV39H1, SETDB1 and MLL1-4, impaired the accumulation of ubiquitin conjugates and 53BP1. When SETD8 was depleted, the distribution of MDC1 was perturbed even in unstressed conditions and impaired the accumulation of MDC1 and ubiquitin conjugates in response to DNA damages (Fig. S2b,c). Supportively, several reports have shown that the depletion of SETD8 leads to genomic instability even under unperturbed conditions 10,11 , and that SETD8 is recruited to DSBs and required for NHEJdirected repair 12,13 . These observations raise the possibility that SETD8 might adjust the physiological structure of chromatin as a fundamental regulator to protect against genomic insults. In contrast, depletion of EHMT1 or EHMT2 had little effect on MDC1 recruitment to damaged chromatin, but it had markedly reduced the accumulation of downstream factors of MDC1 in DDR (Figs S1a and S2a). Therefore, we focused on these two methyltransferases, EHMT1 and EHMT2, which form a heterodimeric complex, as candidates for regulators of MDC1.
To investigate whether EHMT1 and EHMT2 affects MDC1 function as a platform for integrating in or operating the DSB-triggered pathway, we observed local ubiquitylation and subsequent recruitment of repair factors, RAP80 and 53BP1, at the site of damaged chromatin using immunofluorescent analysis. We treated cells with NCS and found that depletion of EHMT1 and/or EHMT2 impaired the accumulation of conjugated ubiquitin, 53BP1 and RAP80 at the DSB sites in all three cell types examined, U2OS, MCF7 and diploid human lung fibroblast cells, using multiple siRNAs targeting EHMT1 or EHMT2 (Figs 1a,b and S3). These results suggest that EHMT1 and EHMT2 are required for the efficient formation of ubiquitin conjugates followed by the recruitment of repair factors at the DNA-damage sites.

EHMT1 interacts with MDC1 and is facilitated by ATM activation in DDR.
To determine whether EHMT1/EHMT2 lysine methyltransferases physically interact with MDC1, lysates of U2OS and MCF7 cells expressing Flag-tagged EHMT1 or EHMT2 were prepared and immunoprecipitation analysis was performed upon the treatment of NCS. We found that the interactions between MDC1 and EHMT1 were detectable even under undamaged conditions, and enhanced by treatment with NCS in these cell lines (Fig. 2a). On the other hand, EHMT2 marginally associated with MDC1 only in U2OS cells, and their interaction was not changed by treatment with NCS. By employing immunoprecipitation of untransfected cells, we confirmed that the endogenous MDC1 interacts with EHMT1, but only a modest MDC1-EHMT2 association was detected (Fig. 2b). From a morphological approach, the in situ proximity ligation assay (PLA) enabled the visualization of the colocalization of MDC1 with EHMT1 or EHMT2 upon exposure to radiomimetic drugs (NCS or Zeocin), in combination with Flag antibody detecting EHMT1 or EHMT2 and MDC1 antibody (Fig. 2c) 14 . The PLA signals of MDC1-EHMT1 and MDC1-EHMT2 were detectable as fine intranuclear foci even under undamaged conditions. Subsequent to DNA damage, the EHMT1-MDC1 PLA signals exhibited a focal accumulation pattern in the nuclei, but EHMT2-MDC1 PLA signals did not change their distributions (Fig. 2c). These results imply that EHMT1 exerts a proactive role in MDC1 regulation by responding to signals induced by DNA damage. To test this notion, we examined whether the ATM activity affects the enhanced interaction of EHMT1 with MDC1 upon DNA damage. As expected, the immunoprecipitation analysis revealed that inhibition of ATM kinase activity by treatment with KU-55933 abolished the increased EHMT1-MDC1 interaction upon exposure to NCS (Fig. 2d). These results indicate that MDC1 interacts mainly with EHMT1 in the absence of DNA damage, and their interaction is facilitated by the DNA damage-initiated ATM signalling.
EHMT2 promotes methylation of MDC1 lysine 45. To assess whether EHMT1 and EHMT2 are responsible for the methylation status of MDC1, we raised an antiserum against the dimethylated lysine 45 by using a synthetic methylated peptide as an antigen (Fig. 3a). Our results from the immunoblotting experiments shown in Fig. 3b revealed that the antibody specifically recognized the WT-MDC1 only, but not the unmethylated mutant, K45A-MDC1 6 , in comparison with the pan-MDC1 ( Fig. 3b; upper) and the preimmune control ( Fig. 3b; middle) antibodies.
To investigate the effects of EHMT1 and EHMT2 on the methylation levels of MDC1, we generated a N-terminal variant of MDC1 containing lysine 45, whose methylation could be detected with a methyl-specific antibody (Fig. 3c), cotransfected with siRNA against EHMT1 and EHMT2 into U2OS cells, and the cell lysates were subjected to immunoblotting experiments (Fig. 3d). As previously reported 15,16 , the levels of H3K9me2 were decreased in either EHMT1 or EHMT2 depleted cells ( Fig. 3d; right). On the other hand, depletion of EHMT2, but not of EHMT1, reduced the methylation levels of MDC1, indicating that EHMT2 is required for modifying MDC1 methylation. Consistently, the catalytic activity of EHMT1 but not EHMT2 is dispensable for their methyltransferase activity in vivo 17 , while the methylation-binding activity of EHMT1 in EHMT1/EHMT2  Immunoprecipitation of endogenous MDC1 (a) with FLAG-tagged EHMT1 or EHMT2 in U2OS and MCF7 cells, or (b) with endogenous EHMT1 or EHMT2 in U2OS cells, exposed or not to NCS (500 ng/ml for 30 min) as indicated. Density ratio of EHMT1 (IP/input) was quantified and indicated at the bottom of panel. (c) In situ proximity ligation assay of U2OS cells (left panels) and MCF7 cells (right panels) transfected with FLAG-tagged mock, EHMT1 or EHMT2, subjected to immunoreaction using MDC1 antibodies combined with γH2AX or Flag antibodies at 1 h after exposure to NCS (50 ng/ml for 15 min) or Zeocin (ZEO 100 μg/ml for 1 h). Scale bar, 10 μm. (d) Immunoprecipitation of endogenous MDC1 and FLAG-tagged EHMT1 in U2OS cells treated with 10 μM of ATM inhibitor for 100 minutes followed by exposure of NCS (500 ng/ml for 30 min). Line marking the multiple bands of MDC1 apparently represent alternatively spliced forms 20   heteromeric formation is essential for gene silencing 18,19 . These observations imply that EHMT1 and EHMT2 have distinct roles, for example, a regulating function through its interactor and catalytic activity, respectively (see Discussion). However, to clarify whether these molecules act through a heterodimeric complex or could function in an independent manner during the DDR process requires further investigation.

Methylation of MDC1 lysine 45 is required for ATM accumulation on damage sites.
Protein methylation frequently mediates protein-protein interaction transducing to downstream biological outcomes. In fact, proper dynamic methylation of MDC1 is essential for assembling the DDR factors, including ubiquitin ligase complexes and repair mediators 6 , although the constitutively unmethylated MDC1 K45A variant itself is still accumulated around the sites of DNA damage (Fig. 3e). Furthermore, MDC1 serves as a platform to iterate the MRN complex-ATM activation loop, thus propagating the DDR signal around the damage sites 5 . Since the FHA domain of MDC1, which is located near the methylation lysine 45 on MDC1 (Fig. 3a), interacts with the MRN complex 20,21 and ATM 22 , it is conceivable that their interaction might be regulated by MDC1 methylation. To test this hypothesis, we performed immunoprecipitation using cells expressing either the MDC1-wild type or unmethylated mutant MDC1-K45A. NBS1 and MRE11 interacted with both the wild type and the unmethylated mutant of MDC1, while ATM coimmunoprecipitated with the wild-type MDC1, but less robustly with K45A (Fig. 4a). Consistent with these results, immunofluorescence analysis also revealed that the accumulation of activated ATM around the damaged sites was impaired by depletion of endogenous MDC1 and was restored by expression of siRNA-resistant sequence encoding HA-tagged wild type MDC1, but not by the MDC1 K45A variant (Fig. 4b). Furthermore, depletion of EHMT1 or EHMT2 abolished the accumulation of activated ATM in response to DNA damage (Fig. 4c), while the total amount of autophosphorylated ATM remained unaffected (Fig. 4d). A previous report supporting this observation showed that the activated ATM failed to accumulate near DSBs in MDC1 −/− cells, even though immunostaining signals of the phospho-ATM were increased upon DNA damage, and that MDC1 with a deleted FHA domain in MDC1 −/− cells failed to restore the phospho-ATM foci formation despite itself forming the MDC1 foci following DNA damage 22 . Collectively, these data suggest that methylation of MDC1 mediated by EHMT1 and EHMT2 is required for the interaction between ATM and MDC1, facilitating the accumulation of activated ATM at the DSB sites, and could conceivably regulate the FHA domain of MDC1.
Furthermore, on uncapped telomeres, ATR/ATM kinases and repair factors including MDC1 are known to be activated and form telomere dysfunction-induced foci (TIFs) in the nuclei 23 . Reportedly, MDC1 promotes the frequency of cells with phosphorylated ATM foci at uncapped telomere lesions, as it does at DSBs 24 . In line with these observations, we examined whether methylation of MDC1 affects the accumulation of activated ATM also at dysfunctional telomeres. Telomere dysfunction was induced by retrovirus introduction of dominant negative TRF2 (TRF2-DN) 23,24 in RPE-1 cells, a human telomerase-immortalized retinal pigment epithelial line, with or without the knockdown of EHMT1 or EHMT2. The TIFs were scored by PLA between MDC1 or activated ATM with another shelterin component TRF1, which is known to be at the telomere dysfunction lesions even in the absence of TRF2 (Fig. 4e). MDC1 itself could localize to the telomere dysfunction lesions irrespective of the presence or absence of these methyltransferases. In contrast, activated ATM did not accumulate at telomere dysfunction lesions in EHMT1-or EHMT2-depleted cells. These data indicate that EHMT1 and EHMT2 are required for the accumulation of activated ATM, but not of MDC1, at telomere dysfunction lesions.
In conclusion, the EHMT1 and EHMT2 interact with MDC1, and interaction between EHMT1-MDC1 is facilitated by ATM activation in DDR. Methylation of MDC1 lysine 45 is modulated mainly by EHMT2 and this methylation is required for the accumulation of activated ATM around the DNA damage sites, including the dysfunctional telomere.

Discussion
Our results suggest the following scenario. EHMT1 is recruited to MDC1 on damaged chromatin in an ATM-dependent manner. The EHMT2 in EHMT1/EHMT2 heterodimer methylates MDC1 to amplify the local activation of ATM to expand the phosphor-signal around the damaged chromatin (Fig. 5). In our finding, unmethylated MDC1 at lysine 45 failed to amplify ATM accumulation on damaged chromatin and to recruit downstream repair factors, while ATM activation itself was observed in EHMT1-or EHMT2-depleted cells (Fig. 4). Consistent with our data, a previous report showed that activated ATM completely failed to accumulate near the DSBs in MDC1 −/− MEF 22 , even though phospho-ATM immuno-staining signals were increased following DNA damage, and that unmethylated MDC1 could form a robust non-chromatin complex with downstream repair factors 6 . Together, these findings suggest that the methylation of MDC1 at lysine 45 might have a crucial role of MDC1 as a platform for the association of activated ATM and downstream factors on the DSB-flanking chromatin.
For the accumulation of methylated MDC1 on damaged chromatin, the ATM signalling initiated by DNA damage primarily facilitates the EHMT1-MDC1 interaction (Fig. 2). In fact, several sites with an S/TQ motif in MDC1 are already reported to be phosphorylated by ATM, thereby propagating the DDR signal on damaged chromatin 4 . On the other hand, the EHMT1 also contain six S/TQ motifs conserved in human, mouse and chicken, which are candidates for ATM phosphorylation, and actually EHMT1 has been detected in a screen for ATM/ATR substrate in DDR 25 . However, individual EHMT1 mutants comprising alanine substitutions of these S/TQ motifs could interact with MDC1 in damage inducible manner (data not shown). Interestingly, a recent report has demonstrated that EHMT2 is phosphorylated by ATM and this phosphorylation is required for EHMT1 localization on damaged chromatin treated by laser micro-irradiation 9 . Possibly, phosphorylation of MDC1, hyperphosphorylation of EHMT1, or another factor mediated by ATM kinase may be important for facilitating the MDC1-EHMT1 association. Further investigation is required to understand the molecular mechanism of this system. As the DDR components, EHMT1 and EHMT2 seem to have distinct roles. For example, EHMT1 exerts a proactive role in MDC1 regulation by responding to signals induced by DNA damage (Fig. 2), whereas EHMT2 mainly contributes to the methylation levels of MDC1 (Fig. 3). Consistently, a previous study demonstrated that the catalytically inactive EHMT1 formed a heterodimer with EHMT2 that could restore the methylation level of histone H3 lysine3 in Ehmt1 −/− MEF, while the catalytically inactive EHMT2 failed to do so in Ehmt2 −/− MEF 17 . Therefore, EHMT2, but not EHMT1, is indispensable for their methyltransferase activity. On the other hand, since methylation-binding activity of EHMT1 in EHMT1/EHMT2 heteromeric formation is essential for gene silencing 18,19 , EHMT1 could exert its biological function by interacting with targets, that is, MDC1 on damaged chromatin.
Furthermore, depletion of EHMT1 or EHMT2 abrogates the recruitment of ATM to the unprotected chromosome ends (Fig. 4e). Even if MDC1 is recruited to a dysfunctional telomere, ATM signal is not sufficiently activated on damaged chromatin without EHMT1 or EHMT2 (Fig. 4c). Given that EHMT1 and EHMT2 are degraded in senescent cells 26 , the prevention of excessive DDR through degradation of EHMT1 and EHMT2 might be required for the survival of replicative senescent cells that possess shorten telomeres. Moreover, the DDR activation on eroded telomeres followed by cell division causes genomic instability through chromosomal breakage-fusion-bridge cycles, which could develop cancer. Indeed, since EHMT2 is often overexpressed and associated with poor prognosis in several cancers 27 , the (re)activation of EHMT2 might contribute to tumorigenesis by leading chromosomal instability caused by persistent DDR.
Taken together, these findings enhance our understanding of the DDR mechanisms through proper dynamic methylation of MDC1, with implications for genome integrity.

Methods
Plasmids and transfection. Total RNA from MRC5 cells was isolated using an RNeasy mini kit (Qiagen) and reverse-transcribed using a SuperScript III (Invitrogen). Human EHMT1 was amplified by PCR using cDNA from MRC5 cells and cloned into the pFLAG-CMV2 or p3XFLAG-CMV7.1 vector (SIGMA). Full-length human MDC1 cDNA in pENTR was obtained from Addgene (plasmid #26427) and was subcloned into pcD-NA-DEST-HA vector in our laboratory stock, using the Gateway cloning technology according to the manufacturer's instructions (Thermo Fisher Scientific). All clone sequences were confirmed. FLAG-EHMT2 was as previously described 26 . The plasmids were transfected into the cells using Xtreme HP reagent (Roche) according to the manufacturer's protocol.

Antibodies.
Immunofluorescence analysis and in situ PLA. Cells grown on grid glass coverslips were fixed in 4% formaldehyde for 10 min and permeabilized in 0.2% Triton X-100/PBS for 5 min at room temperature. Primary antibody with appropriate dilution was added to cells and incubated at 4 °C overnight followed by secondary antibodies coupled to Alexa 488 or 594 (Thermo Fisher Scientific) for 1 hr at room temperature. Coverslips were mounted onto glass slides (Matsunami) with FluorSAVE mounting medium (Millipore). PLA was performed following the manufacturer's instructions using the Duolink anti-Mouse MINUS and anti-Rabbit PLUS In Situ PLA probes and the Duolink In Situ Detection Reagents Green (Olink Bioscience).