Histone demethylase KDM5B is a key regulator of genome stability

Significance

DNA double-strand breaks are generally repaired in the context of highly organized chromatin. However, how epigenetic mechanisms are involved in the maintenance of the genetic fidelity remains poorly understood. Here we report that lysine-specific histone demethylase 5B (KDM5B), a well-defined transcriptional repressor, promotes double-strand break signaling and is required for efficient DNA repairs. We demonstrated that KDM5B, in doing so, functions to orchestrate checkpoint activation and cell survival after DNA damage. Our results provide evidence to indicate that KDM5B is an important genome caretaker and a critical regulator of genome stability.

Abstract

Maintenance of genomic stability is essential for normal organismal development and is vital to prevent diseases such as cancer. As genetic information is packaged into chromatin, it has become increasingly clear that the chromatin environment plays an important role in DNA damage response. However, how DNA repair is controlled by epigenetic mechanisms is not fully understood. Here we report the identification and characterization of lysine-specific histone demethylase 5B (KDM5B), a member of the JmjC domain-containing histone demethylases, as an important player in multiple aspects of DNA double-strand break (DSB) response in human cells. We found that KDM5B becomes enriched in DNA-damage sites after ironizing radiation and endonuclease treatment in a poly(ADP ribose) polymerase 1- and histone variant macroH2A1.1-dependent manner. We showed that KDM5B is required for efficient DSB repair and for the recruitment of Ku70 and BRCA1, the essential component of nonhomologous end-joining and homologous recombination, respectively. Significantly, KDM5B deficiency disengages the DNA repair process, promotes spontaneous DNA damage, activates p53 signaling, and sensitizes cells to genotoxic insults. Our results suggest that KDM5B is a bona fide DNA damage response protein and indicate that KDM5B is an important genome caretaker and a critical regulator of genome stability, adding to the understanding of the roles of epigenetics in the maintenance of genetic fidelity.

The ability of cells to maintain genome integrity is vital for cellular homeostasis. Defects in the maintenance of genome stability underlie a number of developmental disorders and human diseases including cancer (1–3). Compared with other types of DNA lesions, DNA double-strand breaks (DSBs) are particularly dangerous to cells because failure to repair these kinds of damage in an appropriate manner can cause cell death, and aberrant repair can lead to gross chromosomal abnormalities that may eventually lead to tumorigenesis (1, 2, 4).

Upon detection of DSBs, cells activate local and global DNA damage response (DDR) events that promote cell-cycle checkpoint activation and DNA repair signaling (5, 6). For example, in response to DSBs, phosphorylation of the histone variant H2AX (γH2AX) by DDR protein kinases such as ataxia telangiectasia mutated (ATM) (7) creates an extensive modified chromatin environment that allows spatiotemporal redistribution and accumulation of checkpoint and repair factors, including DNA-damage check point-1 (MDC1) and breast cancer susceptibility gene 1 (BRCA1), into repair centers, forming microscopically visible nuclear aggregates known as foci (4, 8). The two extensively studied DSB repair pathways are homologous recombination (HR) and nonhomologous end-joining (NHEJ) (2, 9). In NHEJ, the DSB ends are blocked from 5′ end resection and held in a close proximity by DSB end-binding protein complex, the Ku70–Ku80 heterodimer (10). NHEJ promotes direct ligation of the DSB ends in an error-prone manner (2, 8). In contrast, HR is largely error free and is initiated when the DSB is resected by nucleases and helicases, generating ssDNA overhangs. This structure can invade homologous duplex DNA, which is used as a template for DNA synthesis to restore the original genetic information (11, 12). Meanwhile, ssDNA generates a structural platform for another signaling module triggered by assembly and activation of the ataxia telangiectasia and Rad3-related (ATR) kinase (10). Eventually, ATM and ATR amplify the signals generated at DSBs by phosphorylating several regulatory proteins, including CHK1, CHK2, and p53, that coordinate cell cycle progression or induce cell apoptosis (5). All of these occurrences are essential for timely initiation, amplification, and transmission of the DNA damage signaling.

Because nuclear DNA is packaged into chromatin, accumulating evidence suggests that DNA repair occur both temporally and spatially in the context of highly structured chromatin surrounding the breaks (6, 10), an environment that enables repair factors to detect DNA lesions, assemble, and function properly and promptly. Consistent with this notion, a number of epigenetic regulators that physically or chemically modify chromatin structures have been linked to DSB repair (13), the outcome of which is predominantly determined by chromatin remodeling events as well as histone modification profiles around the breaks (6, 14). These modifications include but are not limited to phosphorylation, acetylation, ubiquitination, and methylation. DSBs might not only induce the formation of specific histone modifications, but also entail alterations of the constitutive modification patterns in a dynamic manner. For example, similar to histone acetylation, recent reports suggest that deacetylation also plays a critical role in DSB response and processing (15, 16). Although histone methylation, another reversible modification, has been linked to the initial phase of repair (17, 18), whether and how histone demethylation contributes to DSB repair are currently unknown.

Here we report on the identification and characterization of lysine-specific histone demethylase 5B (KDM5B), a member of the JmjC domain-containing histone demethylases (19), as an important contributor to DNA repair and signaling pathway in human cells.

Results and Discussion

Histone Demethylase KDM5B Is Enriched in DNA Damage Sites.

Previously, we showed that KDM5B (also known as JARID1B or PLU-1) co-opts the LSD1/NuRD complex to suppress the angiogenic/metastatic potential of breast cancer by repressing chemokine CCL14 (20). To further explore the biological function of KDM5B and to investigate the importance of the regulation of histone methylation in the maintenance of genome stability, we first examined the nuclear redistribution of KDM5B after DNA damage. To this end, nuclear proteins were extracted from untreated or ionizing radiated (IR) U2OS cells and divided into a chromatin-free fraction and chromatin-bound proteins. It was noted that KDM5B was preferentially accumulated in the chromatin fraction after DNA damage, as revealed by Western blotting with KDM5B antibody (Fig. 1A). Accordingly, nuclear KDM5B in the soluble fraction decreased upon DNA damage (Fig. 1A). Similar results were obtained in nuclear fractions from MCF-7 cells and HeLa cells (Fig. 1A). Meanwhile, IR triggered dramatic phosphorylation of γH2AX, one of the immediate targets of ATM on chromatin, whereas the enrichment of histone H3 remained unchanged (Fig. 1A). These observations indicate that DNA damage induces cellular mobilization of KDM5B, suggesting that KDM5B might be involved in DNA damage repair, a notion that is consistent with the current understanding of the majority of DDR-associated chromatin factors that are recruited to damaged DNA sites from preresident chromatins and/or the soluble nuclear pool (21, 22).

Fig. 1.

KDM5B accumulates at sites of DSBs. (A) KDM5B is associated with chromatin in response to DNA damage. U2OS, MCF-7, or HeLa cells were treated with 6 Gy of IR. Nuclear proteins were extracted and fractioned 1 h after IR and subjected to Western blotting analysis with the indicated antibodies. (B) KDM5B occupancy at chromatin flanking DSB generated by endonuclease I-SceI. DR-GFP U2OS cells were transfected with control or pooled KDM5B siRNA in the absence or presence of I-SceI expression. Soluble chromatin from these cells was then immunoprecipitated with anti-KDM5B or anti-γH2AX. The final DNA extractions were amplified by quantitative real-time PCR using primer that covers the DNA sequences flanking the I-SceI site. Data are normalized to control (no I-SecI) samples and each bar represents the mean ± SD for triplicate experiments. (C) Western blotting of KDM5B and HA-I-SceI. (D) KDM5B occupancy at chromatin flanking DSB generated by endonuclease AsiSI. U2OS cells were transfected with HA-ER-AsiSI in the absence or presence of 0.5 μM 4OHT. ChIP experiments were performed using anti-KDM5B or anti-γH2AX with primers that cover the DNA sequences flanking the AsiSI site and the break distal region. Each bar represents the mean ± SD for triplicate experiments. (E) 4OHT-induced AsiSI activation and DNA damage was monitored by Western blotting analysis with the indicated antibodies.

To test the hypothesis that KDM5B is recruited to DSB sites and participates in the DNA repair process, we used a cell-based system (DR-GFP) in which a defective GFP cassette containing the recognition site for endonuclease I-SceI is stably incorporated into genome and a DSB could be generated by transient transfection with a vector encoding for I-SceI (Fig. 1B) (23). This system has been widely used to identify the DSB-binding pattern of a number of DNA repair enzymes with ChIP assay (21, 24). Concomitant with the incorporation of γH2AX, KDM5B was detected to bind to the DSB proximal site after transfection of I-SceI (Fig. 1B). However, KDM5B knockdown by its specific siRNA resulted in diminished enrichment of KDM5B at the defined DNA break site (Fig. 1B). The knockdown effect of KDM5B and the overexpression of the endonuclease I-SceI were verified by Western blotting (Fig. 1C).

To further investigate the involvement of KDM5B in DSBs, we used another endonuclease AsiSI-based system in which sequence-specific DSBs could be generated endogenously (25). We found that KDM5B was recruited and γH2AX was incorporated to the break proximal site, but not the distal region about 2 Mb away from the break site upon AsiSI activation by 4-hydroxyl tamoxifen (4OHT) treatment (Fig. 1D). The working efficiency of the AsiSI system was validated by marked induction of AsiSI and γH2AX in response to 4OHT, as evidenced by Western blotting analysis (Fig. 1E). Collectively, the data support an argument that KDM5B is mobilized and recruited to DSB sites upon DNA damage.

KDM5B Enrichment in DNA Damage Sites Is Poly(ADP Ribose) Polymerase 1-Dependent.

Similar to DNA single-strand breaks and DNA nicks, DSBs also activate poly(ADP ribose) polymerase (PARP) enzymes that modify target proteins with poly(ADP ribose) (PAR) chains at DNA damage sites, thereby stimulating the recruitment and/or activity of repair factors in the early steps of the DDR (26, 27). Interestingly, it was recently reported that KDM5B is physically associated with and chemically modified by PARP1 in transcription regulation (28). To further support the argument that KDM5B is mobilized and recruited to DSB sites and to investigate the molecular insight into the recruitment of KDM5B to DNA damage sites, we hypothesized that the accumulation of KDM5B to DNA damage sites is dependent on PARP1. To test this, we first performed coimmunoprecipitation experiments with MCF-7 cell extracts. Immunoprecipitation (IP) with antibodies against KDM5B followed by immunoblotting (IB) with antibodies against PARP1 demonstrated that KDM5B was efficiently coimmunoprecipitated with PARP1 (Fig. 2A), as previously reported (28). Remarkably, we found that the interaction between PARP1 and KDM5B was enhanced when MCF-7 cells were exposed to the radiomimetic DNA damage agent neocarzinostatin (NCS) (Fig. 2A). To investigate whether DNA damage influences KDM5B chemical modification by PARP1, FLAG-tagged KDM5B was transiently transfected into MCF-7 cells and the cells were treated with NCS before cellular lysates were collected. IP with antibodies against FLAG followed by IB with antibodies against PAR demonstrated that KDM5B was highly PARylated when cells were exposed to NCS (Fig. 2B). Collectively, these experiments point to a role for PARP1/KDM5B in DNA damage response.

Fig. 2.

PARP1 regulates KDM5B recruitment to sites of DNA damage. (A) Genotoxic insult enhanced the association between PARP1 and KDM5B. MCF-7 cells were treated with 0.1 μg/mL NCS for 1.5 h. Whole-cell lysates were immunoprecipitated with antibodies against KDM5B followed by immunoblotting with antibodies against the indicated proteins. (B) PARylation level of KDM5B increased upon NCS treatment. MCF-7 cells transfected with FLAG-tagged KDM5B were treated with NCS for 1.5 h. Then whole-cell lysates were immunoprecipitated with antibodies against FLAG followed by immunoblotting with antibodies against the indicated proteins. (C) PARP1 depletion impaired KDM5B association with damaged chromatin. MCF-7 cells were transfected with control or pooled PARP1 siRNA in the absence or presence of NCS. The chromatin fractions of nuclear proteins (Upper) and whole-cell lysates (WCL, Lower) were collected for Western blotting analysis with the indicated antibodies. (D) PARP1 inhibitor PJ-34 impaired KDM5B relocalization to damaged chromatin. MCF-7 cells pretreated with vehicle or PJ-34 (5 μM) were exposed to NCS (Left) or IR (Right). The chromatin fractions of nuclear proteins were collected for Western blotting analysis with the indicated antibodies (Upper). Meanwhile, the whole-cell lysates from cells with the same treatment were immunoprecipitated (IP) with antibodies against PARP1 followed by immunoblotting with the indicated antibodies (Lower). (E) PARP1 depletion impaired KDM5B recruitment to I-SceI–induced DSB. DR-GFP U2OS cells were transfected with control or pooled PARP1 siRNA in the absence or presence of I-SceI expression. Soluble chromatin was prepared for qChIP assays with antibodies against KDM5B. Each bar represents the mean ± SD for triplicate experiments. (F) PARP1 inhibitor PJ-34 impaired KDM5B recruitment to I-SceI–induced DSB. DR-GFP U2OS cells pretreated with vehicle or PJ-34 were transfected with control vector or I-SceI. Soluble chromatin was prepared for qChIP assays with antibodies against KDM5B. Each bar represents the mean ± SD for triplicate experiments.

To further substantiate the argument that KDM5B mobilization to DNA breaks is dependent on PARP1, chemically synthesized control siRNA and PARP1-specific siRNA were transfected into MCF-7 cells. Treatment of PARP1-depleted MCF-7 cells with NCS revealed that loss of function of PARP1 led to a diminished chromatin binding of KDM5B (Fig. 2C), but not that of histone H3 (Fig. 2C). The knockdown effect of PARP1 and the expression of the whole cellular KDM5B were examined by Western blotting (Fig. 2C). Analogously, reduced level of KDM5B binding to chromatin was detected when cells were treated with PARP1 enzymatic inhibitor PJ-34 (Fig. 2D), indicating that the catalytic activity of PARP1 is required for KDM5B accumulation on damaged chromatin. In support of this, IP with antibodies against PARP1 followed by IB with anti-PAR and KDM5B demonstrated that the PARP1 enzymatic activity as well as the physical association between KDM5B and PARP1 were augmented by DNA damage and blocked by PJ-34 (Fig. 2D).

Next, we investigated the impact of loss of function of PARP1 on KDM5B recruitment to site-specific DSB. To this end, DR-GFP U2OS cells were cotransfected with endonuclease I-SceI and with specific siRNA molecules to knock down the expression of PARP1. Quantitative ChIP (qChIP) showed that depletion of PARP1 was associated with a diminished recruitment of KDM5B to the break site upon I-SceI cutting (Fig. 2E). The same is true when cells were treated with PJ-34 (Fig. 2F). Collectively, these results indicate that the redistribution of KDM5B to damaged chromatin is dependent on PARP1 and requires the catalytic activity of PARP1.

PARylated KDM5B Is Recognized and Recruited by MacroH2A1.1 at Breaks.

It is reported that PARP-1 leads to a permissive chromatin environment by preventing demethylation of H3K4me3 through PARylation, inhibition, and exclusion of KDM5B in gene transcription (28). However, recent studies indicate PAR-dependent accumulation of gene repression-associated chromatin regulators such as histone variant macroH2A1.1 (29), the NuRD complex, or the polycomb complex (30, 31), modulating chromatin modification as well as structure locally at sites of DNA breaks to facilitate signaling and/or repair of DNA damage. Thus, it seems that PARP1 functions in a context-dependent manner to influence the chromatin environment.

Because the macrodomain of macroH2A1.1 binds PAR and senses PARP1 activation upon DNA damage (29), we propose macroH2A1.1 is a potential effector protein to read PARylated KDM5B. To test this hypothesis we performed in vitro nucleosome binding experiments (Fig. S1A), and the results showed that PARylation impairs the interaction of KDM5B with H2A nucleosomes (Fig. S1B) and subsequently inhibits KDM5B enzymatic activity (Fig. S1C), which is consistent with previous findings (28). In contrast, PARylation led to an increased KDM5B binding to macroH2A1.1 nucleosomes, but not to that of macroH2A1.2 lacking PAR binding activity (29, 32) (Fig. S1B). Accordingly, the demethylation activity of PARylated KDM5B was enhanced when incubating with macroH2A1.1 nucleosomes (Fig. S1C). Moreover, in vivo nucleosome binding experiments demonstrated that the interaction between KDM5B and macroH2A1.1 nucleosomes was intensified by IR exposure but impaired by PJ-34 (Fig. S1D). These results suggest that the affinity of PARylated KDM5B to chromatin is determined, at least in part, by nucleosome content, and that macroH2A1.1 macrodomain is a reader module for PAR-conjugated KDM5B.

To further support the notion that PAR binding activity of macroH2A1.1 is required for KDM5B recruitment to DSBs, DR-GFP U2OS cells were treated with specific siRNA against macroH2A1.1 followed by I-SceI transfection. qChIP results showed that macroH2A1.1 depletion led to an impaired recruitment of KDM5B to DSB (Fig. S1E). In addition, the impaired KDM5B recruitment that resulted from macroH2A depletion could only be restored by macroH2A1.1 but not the other alternative splicing variant macroH2A1.2 (Fig. S1F), although these two variants exhibited comparable levels of incorporation at break sites (Fig. S1F). The knockdown efficiency and specificity were determined by quantitative RT-PCR (qRT-PCR) (Fig. S1E) and Western blotting (Fig. S1F). These results suggest that macroH2A1.1 is an important regulator in recognizing and recruiting PARylated KDM5B to damaged chromatin. A model that depicts different outcomes of PARylated KDM5B in transcription versus DSBs is illustrated in Fig. S1G.

KDM5B Is Required for Efficient HR and NHEJ.

To address the functional significance of the recruitment of KDM5B in DSB sites, we next examined the effect of KDM5B on the repair efficiency of two major DSB repair pathways, NHEJ and HR. For NHEJ, HEK293 cells carrying a stably integrated DNA fragment with two recognition sites for the I-SceI endonuclease were subjected to DSB induction by transient expression of I-SceI (Fig. 3A). Restriction of the two I-SceI sites followed by NHEJ excision eliminates the translation start codon of the otherwise nonsense transcript and enables the reading frame shift and subsequently expression of the GFP gene driven by an upstream CMV promoter (Fig. 3A). NHEJ repair efficiency is monitored and expressed as the percentage of cells expressing GFP protein. The experiments indicate that depletion of KDM5B expression was associated with a reduced percentage of GFP-positive cells to nearly 40–50% of that of control cells (Fig. 3B), which was comparable to the effect caused by depletion of KU70, an essential component of NHEJ repair (Fig. 3B). These data suggest that KDM5B is required for efficient NHEJ. The knockdown effects of KDM5B and Ku70 were monitored by Western blotting (Fig. 3C).

Fig. 3.

KDM5B depletion impairs HR and NHEJ efficiency. (A) Assays for NHEJ of chromosomal DSBs in HEK293 cells. (B) GFP-positive cell fraction in NHEJ HEK293 cells. NHEJ efficiency was determined by FACS in Ku70- or KDM5B-deficient cells. Each bar represents the mean ± SD for triplicate experiments. (C) Knockdown effects of Ku70 and KDM5B were examined by Western blotting analysis with antibodies as indicated. (D) Assays for HR efficiency of chromosomal DSB in DR-GFP U2OS cells. (E) GFP-positive cell fraction in DR-GFP U2OS cells. HR efficiency was determined by FACS in BRCA1- or KDM5B-deficient cells. Each bar represents the mean ± SD for triplicate experiments. (F) Knockdown effects of BRCA1 and KDM5B were examined by Western blotting analysis with antibodies as indicated.

The effect of KDM5B on HR repair of a DSB generated in chromosomal DNA was evaluated in U2OS cells using the DR-GFP system (Fig. 3D) in which two incomplete copies of GFP genes are integrated into chromosomal DNA. Cleavage of the I-SceI sites leads to the restoration of the GFP gene through homologous recombination (Fig. 3D). HR repair efficiency is monitored and expressed as the percentage of cells expressing GFP protein. The results showed that depletion of KDM5B resulted in a significantly reduced percentage of GFP-positive cells, an effect comparable to that of knockdown of BRCA1, a key regulator in DSB signaling and repair in HR pathway (Fig. 3E). The knockdown effects of KDM5B and BRCA1 were monitored by Western blotting (Fig. 3F). Together, our experiments indicate that KDM5B is required for efficient repair of DSBs.

The Catalytic Activity of KDM5B Is Required for Efficient DSB Repair.

Along with their recruitment to DNA damage sites, another hallmark of DDR-related chromatin modifiers is their catalytic activities toward break-flanking chromatins, which would provide a platform for subsequent assembly of repair factors (8). Therefore, we asked whether the histone H3 lysine 4 trimethyl (H3K4me3) demethylase activity of KDM5B is required for its involvement in DNA lesion repair. To this end, endonuclease I-SceI and wild-type mouse Kdm5b (Kdm5b-wt) or demethylase activity-defective mouse Kdm5b (H499A, Kdm5b-mt) (33) were cotransfected into KDM5B-depleted NHEJ HEK293 cells. FACS analysis revealed that Kdm5b-wt was able to rescue the decreased NHEJ repair efficiency induced by endogenous KDM5B depletion, whereas Kdm5b-mt failed to do so (Fig. 4A). Similarly, FACS analysis of KDM5B-depleted DR-GFP U2OS cells cotransfected with I-SceI and Kdm5b-wt or Kdm5b-mt demonstrated that Kdm5b-wt expression was associated with an elevated HR repair efficiency in KDM5B-depleted cells, whereas Kdm5b-mt was not (Fig. 4B), although qChIP experiments revealed that both Kdm5b-wt and Kdm5b-mt were effectively recruited to the break region even with a stronger binding capacity than that of endogenous KDM5B (Fig. 4C), excluding the possibility that the loss of catalytic activity might impair the nuclear redistribution of Kdm5b.

Fig. 4.

Demethylase activity of KDM5B is required for its role in DSB repair. (A) Mouse Kdm5b rescued NHEJ deficiency caused by KDM5B depletion. KDM5B-depleted NHEJ HEK293 cells were cotransfected with I-SceI and Kdm5b-wt or Kdm5b-mt followed by FACS analysis. Each bar represents the mean ± SD for triplicate experiments and the P values between the indicated treatments are shown. (B) Mouse Kdm5b rescued HR deficiency induced by KDM5B depletion. KDM5B-depleted DR-GFP cells were cotransfected with I-SceI and Kdm5b-wt or Kdm5b-mt followed by FACS analysis. Each bar represents the mean ± SD for triplicate experiments and the P values between the indicated treatments are shown. (C) Mouse Kdm5b was recruited to DSB region generated by I-SceI. KDM5B-depleted DR-GFP cells were cotransfected with I-SceI and Kdm5b-wt or Kdm5b-mt followed by qChIP assays with antibodies against KDM5B. Each bar represents the mean ± SD for triplicate experiments. (D) H3K4me3 level at DSB was regulated by KDM5B. KDM5B-depleted DR-GFP cells were cotransfected with I-SceI and Kdm5b-wt or Kdm5b-mt for qChIP experiments with antibodies against H3K4me3. Each bar represents the mean ± SD for triplicate experiments and the P value between the indicated treatments is shown. (E) Western blotting analysis of the efficiency of knockdown and overexpression of the indicated proteins. (F) Pooled KDM5B siRNA transfection could efficiently knock down FLAG-tagged human KDM5B expression, but not that of mouse Kdm5b-wt or Kdm5b-mt. The Myc-tagged GATA3 and endogenous β-actin were used as loading controls.

To further consolidate the argument that the catalytic activity of KDM5B is required for its DSB repair activity, qChIP assays were performed to measure H3K4me3 level at sites of DNA damage. In DR-GFP cells, the level of H3K4me3 around the break region was markedly decreased upon I-SceI transfection (Fig. 4D). However, when KDM5B was depleted, the level of H3K4me3 was only slightly changed upon I-SceI treatment (Fig. 4D). In addition, ectopic expression of Kdm5b-wt, but not Kdm5b-mt, could fully restore the down-regulated level of H3K4me3 (Fig. 4D). The efficiency of KDM5B knockdown and of I-SceI/Kdm5b overexpression was monitored by Western blotting (Fig. 4E). The specificity of KDM5B knockdown by pooled siRNA was further examined (Fig. 4F). Collectively, these results support the notion that the H3K4 trimethyl demethylase activity of KDM5B is required for its activity in DSB repair.

KDM5B Is Required for the Recruitment of Ku70 and BRCA1 to Breaks.

To gain more mechanistic insights into the role of KDM5B in DSB repair, we first analyzed IR-induced foci (IRIF) formation of γH2AX, a histone mark spreading to megabases scale and acting as a docking site for MDC1 at DSBs (4). We found that KDM5B knockdown did not affect the formation of γH2AX IRIF or that of MDC1 IRIF upon IR treatment (Fig. 5A).

Fig. 5.

KDM5B promotes Ku70 and BRCA1 accumulation at DSBs. (A) KDM5B depletion did not alter γH2AX and MDC1 IRIF formation. Cells were transfected with control or pooled KDM5B siRNA, exposed to 6 Gy of IR, and immunostained for γH2AX and MDC1. (Scale bars, 10 μm.) (B) KDM5B regulates GFP-Ku70 recruitment to DSBs. KDM5B-depleted cells transiently transfected with GFP-Ku70 were subjected to laser microirradiation (Upper). U2OS cells stably expressed Kdm5b-wt or Kdm5b-mt were cotransfected with pooled KDM5B siRNA and GFP-Ku70. These cells were then subjected to microirradiation (Lower). Representative fields were acquired after 5 min of microirradiation. (Scale bar, 10 μm.) (C) Quantification of relative fluorescence intensities in the microirradiated areas to the background fluorescence in the undamaged region. More than 20 nuclei of cells were scored in triplicate experiments and the P values between the indicated treatments are shown (Left). The efficiency of knockdown and overexpression was examined by Western blotting using the indicated antibodies (Right). (D) KDM5B depletion resulted in changes of BRCA1 IRIF formation, but not that of conjugated ubiquitin (FK2 antibody). Cells were transfected with control or pooled KDM5B siRNA, exposed to 6 Gy of IR, and immunostained with BRCA1 or FK2 antibodies. (Scale bars, 10 μm.) (E) Kdm5b-wt restored the impaired IRIF formation of BRCA1 caused by KDM5B depletion. U2OS cells stably expressing Kdm5-Wt or Kdm5b-mt were transfected with control or pooled KDM5B siRNA, exposed to 6 Gy of IR, and immunostained with BRCA1 antibody. (Scale bar, 10 μm.) (F) Quantitative analysis of BRCA1 IRIF formation. More than 100 nuclei of cells were scored in triplicate experiments. P values between the indicated treatments are shown. (G) Streptavidin beads-coupled native and H3K4me3 peptides were incubated with FLAG-BRCA1– or FLAG-JMJD2A–expressing cell lysates. After washing with lysis buffer, the beads were boiled and subjected to Western blotting analysis.

We then asked whether KDM5B could influence the accumulation of other key DDR proteins in break compartment. For this purpose, we analyzed the recruitment of Ku70, BRCA1, RNF8, and RNF168, the essential components of the DNA damage signaling and repair pathway. Because Ku70 could not form detectable foci after cell exposure to IR or genotoxic agents, laser microirradiation experiment was used to treat GFP-Ku70-transfected U2OS cells in which Ku70 accumulation in the damaged region was manifested by enhanced GFP signals. Notably, KDM5B depletion was associated with an impaired recruitment of Ku70, and, consistent with the observation that KDM5B promotes DSB repair via its catalytic activity, stable integration of Kdm5b-wt, but not Kdm5b-mt, could rescue the recruitment of Ku70 (Fig. 5 B and C). The effect of KDM5B knockdown and the expression of mouse Kdm5b were monitored by Western blotting (Fig. 5C).

Similarly, the accumulation of BRCA1 was impaired in KDM5B-depleted cells (Fig. 5D). Because BRCA1 requires binding to conjugated ubiquitin for its accumulation at DSBs, we examined the impact of KDM5B knockdown on the formation of ubiquitin conjugates (12, 34). The results indicate that the level of ubiquitin conjugates at DSBs was not affected by KDM5B expression status (Fig. 5D). Likewise, RNF8 and RNF168, two key ubiquitin ligases involved in this process (12, 34), remained robustly accumulated at DSBs upon KDM5B depletion (Fig. S2). These data suggest that the recruitment of BRCA1 to DSB-induced chromatin is determined not only by histone ubiquitination, but also by other chromatin-associated events. We reasoned that the recruitment of KDM5B to the DSB sites, particularly the ensuing chromatin modification, is required for efficient BRCA1 binding. Indeed, we found that the expression of the catalytically defective form of Kdm5b (Kdm5b-mt) could not rescue BRCA1 IRIF formation as efficiently as that of the wild-type Kdm5b (Kdm5b-wt) (Fig. 5 E and F), suggesting that H3K4me3 demethylation by KDM5B triggers BRCA1 recruitment. In support of this notion, in vitro peptide pull-down with FLAG-BRCA1–containing cell lysate and K4 trimethylated H3 or unmodified H3 N-terminal peptides showed that BRCA1 bound to native peptides much more efficiently than to H3K4me3 peptides (Fig. 5G), whereas JMJD2A acted otherwise, as previously reported (35). Although the interplay between H3K4me3 and ubiquitin conjugates and how these elements contribute to coordinated DSB repair remains to be investigated, these data support a role for KDM5B in recruitment of Ku70 and BRCA1 and thus in efficient DSB repair through the NHEJ or HR pathway.

KDM5B Regulates Cellular Response to DNA Damage.

To further understand the scope of the function of KDM5B in DDR, we analyzed the effects of its depletion, focusing initially on ATM/ATR-dependent phosphorylation of DDR components in response to IR. Knockdown of KDM5B did not impair IR-induced γH2AX formation (Fig. 5A). Rather, KDM5B-depleted cells exhibited hyperactivation and persistence of IR-induced H2AX phosphorylation (Fig. S3A). Consistently, KDM5B depletion resulted in hyperactivation and delayed removal of IR-induced phosphorylation of CHK1 and CHK2, the downstream effectors of ATM/ATR-dependent pathway (Fig. S3A). Similar results were obtained when cells were exposed to the radiomimetic agent NCS (Fig. S3B). These results are consistent with our above observation that KDM5B-depleted cells are defective in DSB repair.

Multicolor competition assays (MCA) were then performed in which short hairpins targeting KDM5B or firefly luciferase were designed and stably integrated in U2OS cells. The short hairpins were delivered via a lentiviral vector that also contains a GFP cassette, allowing successful knockdown of KDM5B being monitored by FACS. In the experiments, GFP-expressing KDM5B-deficient U2OS cells were cocultured with an equal number of native U2OS cells. The cell mixtures were then left untreated or exposed to IR and analyzed by FACS after 10 d. The sensitivity to IR treatment is determined by monitoring the number of GFP-stained KDM5B-depleted cells relative to the uncolored control cells, after normalization to the relative cell numbers in untreated control mixture. U2OS cells with depletion of known DDR genes such as ATM were highly sensitive to IR, and, significantly, depletion of KDM5B by two independent shRNAs also rendered U2OS cells sensitive to IR (Fig. S3C). Consistently, KDM5B depletion was associated with a significantly compromised cell survival after exposure to NCS (Fig. S3C). The same was true when MCA assays were carried out in MCF-7 cells (Fig. S3C). These results indicate that KDM5B is required for cells to cope with IR and NCS, supporting the argument that KDM5B is an important player in DSB repair and cell survival after genotoxic insults.

Because DNA repair and cell-cycle pathway are generally coordinated to promote cell viability, we thus investigated whether loss of function of KDM5B affected cell-cycle progression. It was evident that KDM5B depletion resulted in a significant delay of the G₁/S transition (Fig. S3D). Combined irradiation and KDM5B depletion exacerbated the situation (Fig. S3D). In light of the importance of the p53–p21 pathway in G₁/S transition (5), we examined the functional status of this pathway in KDM5B-depleted cells. Our experiments revealed that the levels of p53, phosphorylated p53 (S15p), and the p53 effector p21 were significantly elevated in the absence of KDM5B in p53^+/+ cells, but not in p53^−/− cells (Fig. S3E). Interestingly, the level of γH2AX was constantly up-regulated in KDM5B-depleted cells, regardless of p53 status (Fig. S3E), suggesting that H2AX phosphorylation is probably an upstream event of the p53 signaling pathway. Consistently, KDM5B depletion-induced delay of G₁/S transition was not observed in p53-deficient cells (Fig. S3F), suggesting that this effect is dependent on p53 activation.

We were interested in whether cell cycle arrest induced by loss of KDM5B might be associated with spontaneous occurrence of DNA lesions. Indeed, increased levels of γH2AX were detected in KDM5B-depleted cells as early as 2 d after siRNA transfection (Fig. S3 A, B, and E). In addition, it was repeatedly observed that the phosphorylation of CHK1 and CHK2 increased in KDM5B-deficient cells even without genotoxic exposure (Fig. S3 A and B). Therefore, it seems that loss of KDM5B was associated with an increased susceptibility of cells to spontaneous DNA breaks, thereby activating the p53–p21 pathway and eventually leading to cell-cycle arrest. Collectively, these results indicate that KDM5B plays important roles in preserving genome stability and orchestrating cell-cycle progression. Finally, we demonstrated that KDM5B is a potential substrate of ATM/ATR-like protein kinase (Fig. S4 A–D) and KDM5B recruitment to DSB was not detected by laser microirradiation (Fig. S5), possibly due to the fact that the amount or intensity of accumulated KDM5B is beyond the resolution of the laser-generated microscopy.

Materials and Methods

Differential Extraction of Nuclear Proteins.

Differential protein extraction was performed using the Subcellular Protein Fractionation Kit for Cultured Cells (78840; Thermo Scientific) according to the manufacturer’s instructions.

Immunofluorescence.

Cells on glass coverslips (BD) were fixed with 2% (vol/vol) paraformaldehyde and permeabilized with 0.2% Triton X-100 in PBS. Samples were then blocked in 5% (g/100 mL) donkey serum and stained with the appropriate primary and secondary antibodies coupled to AlexaFluor 488 or 594 (Invitrogen). Confocal images were captured on FluoView1000 Olympus using a 60× oil objective.

Sequence Information.

The sequences of siRNAs, shRNAs and primers are provided in Tables S1–S4.