Histone Demethylase KDM4B Promotes DNA Damage by Activating Long Interspersed Nuclear Element-1

Epigenetic abnormalities along with genetic alterations are thought to represent one of the major driving forces of cancer initiation and progression (1). Histone-modifying enzymes, including histone demethylases, play important roles in establishing eukaryotic cell chromatin epigenetic landscapes by removing methyl groups from the lysine residues of histone tails, thereby regulating the transcriptional activity of target genes, and are always dysregulated and mutated in various types of cancer (2–4).

H3K9me3 (trimethylation of Lys 9 of histone 3) is primarily catalyzed by histone methyl transferase Suv39h1 and can be erased by KDM4 histone demethylase family proteins. By interacting with HP1α and H3K9 methyltransferases, H3K9me3 maintains a silent, compact heterochromatin conformation (5). Moreover, transient formation of repressive chromatin by H3K9me3 represses gene expression in euchromatic regions. For example, the tumor suppressor retinoblastoma protein (pRb) recruits the Suv39h1/HP1 methylation system to the cell-cycle–regulatory gene cyclin E promoter to repress its expression (6). In addition to its role in gene regulation, H3K9me3 has been shown to safeguard genome stability because cells lacking Suv39h1 display decreased double-strand break (DSB) repair ability and increased radiosensitivity (7). These observations implied that abnormal H3K9me3 might be an important etiology of cancer. Interestingly, H3K9me3 has also been shown to be enriched in several families of retrotransposons, including long interspersed nuclear element-1 (LINE-1) and long terminal repeat (LTR; refs. 8, 9), but the relevant biological significance is still elusive.

KDM4B, which was first identified in 2004, is a member of the histone demethylase KMD4 family. KDM4B specifically recognizes tri- and dimethylated lysine 9 (H3K9me3/2) and tri- and dimethylated lysine 36 (H3K36me3/2) on histone H3, reducing both modifications to the monomethylated state (10). KDM4B might serve as an oncogene because its overexpression has been frequently identified in a number of malignancies, such as breast, colorectal, ovarian, lung, gastric, and prostate cancer (11–14). KDM4B was originally shown to antagonize H3K9me3 in pericentric heterochromatin in mammalian cells (15). Recent studies have revealed that KDM4B also functions in gene expression regulation. In concert with estrogen receptor (ER) signaling, KDM4B demethylates H3K9me3 in regulatory regions of ER-responsive genes, including MYB, MYC, and CCND1, thereby releasing the transcriptional block, promoting breast cancer cell proliferation and forming a positive-acting regulatory loop with estrogen receptor (ER; refs. 16, 17). In addition, in ovarian cancer cells, hypoxia induces KDM4B overexpression and reduces H3K9me3 at promoter regions of LOXL2, LCN2, and PDGFB, thus facilitating seeding of metastases in the viscera and peritoneal wall (11). Recent studies have also indicated that KDM4B overexpression mostly abolishes pericentromeric H3K9me3 foci, resulting in a significant increase in the chromosome mis-segregation rate and promotion of genome instability (18). While all these studies indicated that the primary oncogenic function of KDM4B is demethylating H3K9me3 in different tumors, a genome-wide analysis of the effect of KDM4B on global or local H3K9me3 levels has not been reported.

In this study, we systemically analyzed the whole-genome H3K9me3 distribution in cancer cells and found that H3K9me3 was largely enriched in LINE-1. Our chromatin immunoprecipitation (ChIP)-seq analysis revealed that ectopic KDM4B or KDM4B depletion induced H3K9me3 alterations in evolutionarily young LINE-1 elements. We further show that KMD4B overexpression led to elevated LINE-1 mRNA and protein levels, as well as increased retrotransposition activity, copy number, and DNA damage, most likely because the ectopic KDM4B mediated changes in H3K9me3 levels. Interestingly, breast cancer cells expressed higher KDM4B protein and exhibited enhanced LINE-1 expression and copy number compared with cell lines with lower KDM4B expression. Strikingly, pharmacologic inhibition of KDM4B clearly reduced LINE-1 expression and DNA damage. Therefore, our study suggests an unexpected role of KDM4B overexpression in tumorigenesis, namely by stimulating LINE-1 expression, transposition, and the associated DNA damage through its histone demethylase activity.

MCF7, T47D, MDA-MB-231, MCF10A, and ZR-75-30 cell lines were purchased from ATCC between 2012 and 2014. 293T, HCT116, and U2OS cell lines were obtained from Institute of Cellular Resources, Shanghai Institutes for Biological Sciences, Chinese Academy of Sciences, (Beijing, China) in 2011 and 2015. All cell lines involved in our experiments were reauthenticated by short tandem repeat analysis every 6 months after resuscitation in our laboratory and routinely tested to verify the absence of Mycoplasma with LookOut Mycoplasma PCR Detection Kit (Sigma, MP0035). The cells revived from frozen stocks were used within 10 to 20 passages or for no longer than 2 months in total for any experiment.

HCT116, 293T, MDA-MB-231, and ZR-75-30 cells were cultured in DMEM (HyClone, SH30022.01) supplemented with 10% FBS (FBS; Gibco, 10099141). MCF7 cells were cultured in Eagle minimum essential medium (Genom, 11700) with 10% FBS, 0.01 mg/mL human recombinant insulin (Sigma, I9278), and nonessential amino acids (Invitrogen, 11140050); MCF10A cells were cultured in DMEM/F12 (Invitrogen, 1130-032) with 5% horse serum (Invitrogen, 16050-122), 20 ng/mL recombinant human EGF (Protech, AF-100-15), 0.5 μg/mL cholera toxin (Sigma, C8052), and 0.01 mg/mL human recombinant insulin (Sigma, I9278). T47D and U2OS cells were maintained in RPMI 1640 medium (HyClone, SH30809.01) with 10% FBS, and the addition of 0.01 mg/mL human recombinant insulin (Sigma, I9278) was required for T47D cells. Stable vector- and KDM4B-expressing cell lines were generated via lentivirus transduction and selected with puromycin. Briefly, lentivirus was generated in 293T cells transfected with the helper plasmids pMD2.G and psPAX2 and vector or KDM4B plasmid using Lipofectamine 2000 (Invitrogen, 11668). Supernatants containing the packaged lentivirus were collected after 48 hours, passed through a 0.45-μm filter and added to target cells together with 1 μg/mL polybrene (Sigma, H9268). Infected cells were selected in medium containing puromycin (Cayman Chemicals, 13884). The cells were cultured at 37°C in a humidified atmosphere containing 5% CO₂.

Primary antibodies targeting the following molecules were purchased: KDM4B (Bethyl Laboratories, A301-478A), GAPDH (ProteinTech, 60004-1-Ig), α-Tubulin (Santa Cruz Biotechnology, sc-8035), ORF1p (Cell Signaling Technology, 88701), ORF2p (Santa Cruz Biotechnology, H-110), γH2AX (Cell Signaling Technology, 9718), GFP-Tag (Abmart, M20004), HA-Tag (Abmart, M20003), H3K9me3 (ABclonal, A2360), and H3 (Cell Signaling Technology, 4499S). Secondary antibodies included horseradish peroxidase–conjugated goat anti-mouse IgG (Servicebio, GB23301) and goat anti-rabbit IgG (Servicebio, GB23303). The inhibitor NSC 636819 was purchased from TOCRIS.

Total RNA was extracted from cells using TRIzol (Invitrogen, 15596018) according to the manufacturer's instructions. For cDNA preparation, 1 μg of total RNA was reverse-transcribed for RT-PCR using All-in-One cDNA synthesis Super Mix (Biotool, B24403). Quantitation of all gene transcripts was performed by quantitative RT-PCR using SYBR Green Real time PCR Master Mix (Biotool, B21802) and the Bio-Rad sequence detection system with GAPDH expression as the internal control. All reactions were carried out in triplicate. The data were analyzed using the 2^−ΔΔC_t method. The primers used are listed in Supplementary Table S1.

Transfection was performed using FuGENE 6 (Promega, E2691) according to the manufacturer's instructions, and the cells were harvested 48 hours posttransfection. Short hairpin RNA (shRNA) lentiviral constructs against human KDM4B and shRNA control hairpin against GFP in the Lv-shRNA-GP backbone were purchased from YiLe Inc. The target sequence in KDM4B mRNA is 5′-GTGGAAGCTGAAATGCGTGTA-3′ for shRNA. For lentiviral stock preparation, 293T cells were cotransfected with Lv-shRNA-GP–based vector and three packaging plasmids (pGag-Pol, pRev, and pVSVG). Supernatants containing packaged lentivirus were collected after 48 hours, passed through a 0.45-μm filter and added to target cells along with 1 μg/mL polybrene (Sigma, H9268). Infected cells were selected in puromycin-containing medium.

Subconfluent cells were transfected with small interfering RNAs (siRNA) using Lipofectamine RNAiMAX (Invitrogen, 13778150) with specific siRNA at a final concentration of 50 nmol/L according to the manufacturer's instructions. The siRNAs for KDM4B, ORF1 and ORF2 and the nontargeted control were purchased from GenePharma Inc. (Shanghai, China). The sequences of the siRNAs target KDM4B, ORF1, and ORF2 were as follows: siKDM4B-1, 5′-GCGCAGAAUCUACCAACUU-3′, siKDM4B-2, 5′-CAAAUACGUGGCCUACAUA-3′; siORF1-1, 5′-CAAUGGAAGAUGAAAUGAATT-3′, siORF1-2, 5′-GGGAGGACAUUCAAACCAATT-3′; siORF2-1, 5′-GACUCCCACACAUUAAUAATT-3′, siORF2-2, 5′-GCAUCAUUCUGAUACCAAATT-3′.

Cells were transfected with the LINE-1 retrotransposition cassette EF06R (Addgene plasmid # 42940) using FuGENE 6 Transfection Reagent (Promega, E2691) according to the manufacturer's instructions. On day 5 posttransfection, cells were harvested, washed in PBS, and kept on ice prior to FACS analysis. The FACS analysis was performed on a FACSAria III (BD Biosciences) using red-versus-green fluorescence plots. Transfection controls were performed in parallel with pEGFP-N1 to calculate the transfection efficiency. The retrotransposition frequency was determined by gating on GFP⁺ cells. The data were analyzed using SD FACSDiva7.0 software.

MCF7 cells were plated on coverslips in 6-well plates and transfected with plasmids expressing KDM4B (phage-GFP-KDM4B) using FuGENE 6 Transfection Reagent according to the manufacturer's instructions. At 48 hours posttransfection, cells were fixed in 4% paraformaldehyde solution, permeabilized with 0.2% Triton X-100, and blocked in 5% BSA solution. The coverslips were incubated with relevant primary antibodies overnight at 4°C followed by the corresponding secondary antibodies in the dark at room temperature for 1 hour and then DAPI for 10 minutes. Slides were then washed and mounted with a Prolong Anti-fade Kit (Invitrogen) and visualized by confocal microscopy (Leica TCS SP8 STED). The primary antibodies used in this experiment targeted KDM4B (Bethyl Laboratories, A301-478A), ORF1p (Cell Signaling Technology, 88701) and ORF2p (Santa Cruz Biotechnology, H-110). CY-3 conjugated goat anti-rabbit IgG (Boster, BA1032) and FITC-conjugated goat anti-mouse IgG (Boster, BA1101) were used as secondary antibodies. ImageJ software was used to quantify the intensity of the red fluorescent signal from ORF1p.

MCF7-WT, MCF7-KDM4B, MCF7-shKDM4B cells were plated in 10-cm dishes until confluency and fixed with 37% fresh formaldehyde to a final concentration of 1% for 10 minutes, and the reaction was quenched with 0.125 mmol/L glycine for 5 minutes. The cells were washed three times with ice-cold PBS and then harvested in ChIP digestion buffer (50 mmol/L Tris-HCl, pH 7.6, 1 mmol/L CaCl2, 0.2% Triton X-100, and protease inhibitor cocktail). After digestion by MNase (NEB, M0247) at 37°C for 20 minutes, 0.5 mol/L EDTA was added to terminate the digestion. Finally, DNA was digested to 1–2 nucleosomes. Four volumes of ChIP dilution buffer (20 mmol/L Tris-HCl, pH 8.0, 150 mmol/L NaCl, 2 mmol/L EDTA, 1% Triton X-100, 0.1%SDS) was added to the supernatant. The resulting lysate was then incubated with Protein G Sepharose 4 Fast Flow (GE Healthcare, 17061805), which was blocked with salmon sperm DNA for 1 hour and incubated with antibodies at 4°C overnight. Bound complexes were washed once with each of the following buffers: low salt wash buffer (0.1% SDS, 1% Triton X-100, 2 mmol/L EDTA, 20 mmol/L Tris pH 8.1, 150 mmol/L NaCl), high-salt wash buffer (0.1% SDS, 1% Triton X-100, 2 mmol/L EDTA, 20 mmol/L Tris pH 8.1, 500 mmol/L NaCl), LiCl wash buffer (0.25 mol/L LiCl, 1% deoxycholic acid, 1 mmol/L EDTA, 10 mmol/L Tris pH 8.1) and TE buffer (1 mol/L Tris pH 8.1, 0.5 mol/L EDTA). DNA was eluted from the beads using chIP elution buffer (0.1 mol/L NaHCO₃, 1% SDS, 20 μg/mL proteinase K). The elution was incubated at 65°C overnight, and DNA was extracted using a DNA purification kit (Tiangen Biotech, DP214). The purified DNA was assayed by quantitative PCR with a CFX Connect Real-Time PCR system. Assays were repeated at least three times. The data are shown as the average ± SD values of representative experiments. The sequences of the primers are presented in Supplementary Table S2.

Validated DNA samples were prepared for next-generation sequencing on a HiSeq X Ten system for 150-bp paired-end sequencing. Quality control of the ChIP-seq data was performed using Fastqc, and, low-quality bases and adaptor contamination were then deleted using Trim galore. After quality control and data filtering, the data were mapped to hg38 using the BWA algorithm.

SICER v1.1 (19) was used to identify broad peak distributions, under the setting of false discovery rate (FDR) < 0.01, and the results from this peak calling were stored as BED files. The SICER-df function of SICER was used to detect differential binding peaks (FDR<0.01) among the three types of cell lines. Homer v4.10 (20) was used to determine the genomic annotations of the H3K9me3 peaks. For data visualization, DeepTools v3.0.2 (21) was applied to generate heatmaps of the three major H3k9me3-enriched chromatin domains, each row represents one scaled H3K9me3 peaks that includes ± 1 kb of flanking regions. The average profiles were produced by ngs.plot v2.61 (22), reads per million mapped reads (RPM) was used to calculate the H3k9me3 enrichment level in IP versus Input. The success probability of the foreground and background were used to calculate the enrichment score, which was defined as the fold-change of these two values. The P value was analyzed using a one-sided binomial test. BigWigs files were plotted using the IGV browser. The annotation of LINE-1 elements, LTR, SINE and other repeats was obtained from RepeatMasker (assembly of December 2013) at http://genome.ucsc.edu/cgi-bin/hgTables.

DNA damage in MCF7 cells was measured using the alkaline single-cell gel electrophoresis technique (comet assay). The comet assay was performed using an OxiSelect Comet Assay Kit (Cell Biolabs, STA-350). Prior to the assay, OxiSelect comet agarose (Cell Biolabs, 235002) was melted at 90°C for 20 minutes and then cooled at 37°C for 20 minutes. Cells were harvested at 1 ×10⁵ cells/mL in PBS and combined with comet agarose at a 1:10 ratio (v/v). Next, 75 μL/well was immediately pipetted onto an OxiSelect 3-well comet slide (Cell Biolabs, STA-352), and the slides were transferred to 4°C for 15 minutes. The samples were then lysed in lysis buffer [250 mmol/L NaCl, 20% EDTA solution (Cell Biolabs, 235004), 10% DMSO, 10% 10 × lysis solution (Cell Biolabs, 235005), pH ∼ 10.0] at 4°C for 1 hour. Next, the slides were carefully transferred to chilled alkaline solution (300 mmol/L NaOH, 1 mmol/L EDTA) and immersed for 30 minutes at 4°C. Subsequently, the slides were transferred to a horizontal electrophoresis chamber filled with cold alkaline electrophoresis solution (300 mmol/L NaOH, 1 mmol/L EDTA) and electrophoresed at 1 V/cm, 300 mA for 25 minutes, after which, they were rinsed briefly twice in prechilled DI water and once in 70% ethanol. After the slides were treated with agarose and completely dried and stained with 1 × Vista Green DNA Dye (Cell Biolabs, 235003) for 20 minutes, they were examined using a fluorescence microscope with a FITC filter. Comet tail length was analyzed by Comet Assay Software Project (CASP) software. All steps after agarose treatment were conducted in the dark to prevent additional DNA damage.

ChIP-seq data have been deposited in the Gene Expression Omnibus (GEO) repository under accession number GSE121642.

KDM4B specifically catalyzes the removal of trimethylated lysine 9 on histone H3. To better understand the oncogenic role of KDM4B in tumorigenesis, we first downloaded H3K9me3 ChIP-seq data from Gene Expression Omnibus (GEO) at the National Center for Biotechnology Information (NCBI) and analyzed the genome-wide H3K9me3 distribution in various types of cancer cells using HOMER software. As shown in Fig. 1A, in MCF7 breast cancer cells (GSM946852), we identified 8206 H3K9me3 peaks in LINE-1 in total, which represented the highest number of H3K9me3 peaks among all the DNA elements. In addition, the data also indicated that the H3K9me3 signal was abundant in noncoding regions, such as introns and LTR elements, which is consistent with a previous study (23–25). In contrast, the H3K9me3 peak numbers were much smaller in exons compared with other DNA components, reinforcing the idea that the H3K9me3 signal represents transcriptional repression. Similarly, although the H3K9me3 peaks number varied slightly in four additional cancer cell lines, all of which were also heavily annotated into the LINE-1 elements (Supplementary Table S3). To further confirm the conclusion, we performed ChIP-Seq with H3K9me3 in MCF7 cells, and our results showed that the pattern of genomic H3K9me3 distribution was highly similar to previous observations (Fig. 1B). Because LINE-1 transcriptional activation is functionally associated with cancer, we focused on the effect of KDM4B on LINE-1 biogenesis.

Stable KDM4B knockdown and overexpression MCF7 cell lines were generated and verified by Western blot analysis. As shown in Fig. 1C, both KDM4B depletion and overexpression failed to result in a global H3K9me3 change, which is consistent with a previous study (26). To examine whether KDM4B could regulate local H3K9me3 levels, we performed H3K9me3 ChIP-seq with wild-type (WT), shKDM4B, and ectopic KDM4B MCF7 cell lines. Indeed, the resulting heatmap did not reveal a visible difference in the global trimethylated H3K9 mark caused by altered KDM4B expression (Fig. 1D).

We defined the overlap of shKDM4B-induced H3K9me3 increased peaks and ectopic KDM4B-mediated decreased peaks as the 430 KDM4B-dependent H3K9me3 peaks. Upon KDM4B loss, those local regions exhibited an obvious increase in H3K9me3 peak values, while KDM4B overexpression resulted in a lower abundance (Fig. 1E). These results suggested that KDM4B overexpression in cancer cells might lead to local rather than global H3K9me3-level alteration. We then questioned which LINE-1 elements with the highest H3K9me3 levels were involved in KDM4B-dependent regulation.

Among the KDM4B-dependent H3K9me3 peaks, the annotation of peaks indicated that 30% of the KDM4B-dependent H3K9me3 peaks contained the LINE-1 DNA element, suggesting that a significant proportion of the LINE-1 H3K9me3 level might be regulated by KDM4B (Fig. 2A; Table 1). Subsequently, by overlapping the 430 KDM4B-dependent H3K9me3 peaks with total LINE-1s, we identified approximately 34,408 LINE-1 elements (3.60% of the total LINE-1 elements) were tightly correlated with KDM4B-dependent H3K9me3 peaks. To evaluate the H3K9me3 enrichment in KDM4B-dependent H3K9me3 peaks containing LINE-1, RPM (reads per million mapped reads) was used to calculate the H3K9me3 enrichment level in IP versus input (Fig. 2B). Enrichment profile plots of the KDM4B-dependent H3K9me3 peaks overlapped SINE, LTR, gene body, and intergenic in three cell types were shown in Supplementary Fig. S1. Because the full-length LINE-1 is able to maintain the competency of further retrotransposition (27), we then overlapped the full-length LINE-1 elements with the 430 KDM4B-dependent H3K9me3 peaks and identified 466 full-length LINE-1 elements. Because it has been reported that LINE-1 elements have been replicating and evolving in mammals (27), we performed the sequence alignment of full-length LINE-1s base on the phylogenetic tree of LINE-1 (Supplementary Fig. S2A). Interestingly, the L1PA1-7 species, which represents the evolutionarily young LINE-1 elements, showed greater H3K9me3 enrichment than elements from the older L1M family (27, 28). The number of full-length LINE-1 species overlapped KDM4B-dependent H3K9me3 peaks, the total number of each LINE-1 species together with the enrichment score and their p-value are listed in the Table 2. Importantly, in modern humans, only the young LINE-1 elements have retained the ability to jump autonomously (29).

The KDM4B-dependent H3K9me3 alterations on full-length LINE-1 elements were verified by the two representative examples of full-length LINE-1 elements (L1PA4, L1PA6), which were involved in the young LINE-1 elements shown in Genome Browser tracks and were readily confirmed by ChIP-qPCR analysis in MCF7 and 293T cells, respectively (Fig. 2C; Supplementary Fig. S2B). Moreover, we knocked down KDM4B with two distinct siRNAs in another breast cancer cell line T47D. As shown in Supplementary Fig. S2C, the result showed that KDM4B depletion promoted the H3K9me3 accumulation in the selected LINE-1 elements in T47D. Because H3K9me3 is a marker of transcription repression, our results implied that KDM4B might promote LINE-1 expression by demethylating H3K9me3.

Full-length LINE-1 contains two well-recognized open reading frames (ORF1 and ORF2). ORF1 encodes a 40-kDa protein (ORF1p) with RNA binding and nucleic acid chaperone activities, while ORF2 encodes a 150-kDa protein (ORF2p) with endonuclease and reverse-transcriptase activities required for LINE-1 autonomous retrotransposition (30).

To investigate the role of KDM4B in LINE-1 transcription regulation, U2OS or 293T cell lines stably overexpressing KDM4B (KDM4B) and 293T cells with KDM4B knockdown (shKDM4B) were generated. As shown in Fig. 3A and Supplementary Fig. S3A, ectopic KDM4B increased the ORF1 protein level and both the ORF1p and ORF2 mRNA levels. Consistently, KDM4B depletion led to reduced ORF1 protein as well as LINE-1 transcription (Fig. 3B). To verify whether this regulation was dependent on KDM4B histone demethylase activity, we generated a mutant, E191Q, in which most of the KDM4B enzymatic activity was eliminated (15). As shown in Fig. 3C and Supplementary Fig. S3B, transient transfection with WT KDM4B rather than the mutant significantly promoted LINE-1 gene expression. We also expressed phage-EGFP-KDM4B in HCT116 cells and performed an indirect immunofluorescence assay. As shown in Supplementary Fig. S3C, the ORF1 protein (red) was clearly increased in cells expressing EGFP-KDM4B (green) relative to cells without KDM4B. In addition to KDM4B, another two KDM4 family members, KDM4A and KDM4C, also possess H3K9 demethylase activity. Thus, we then overexpressed all three KDM4 proteins in 293T cells, and the Western blot results indicated that similar to KDM4B, KDM4A and KDM4C also promoted LINE-1 expression, indicating that the KDM4 family probably serves as important LINE-1 regulators (Supplementary Fig. S3D). These results suggest that ectopic KDM4B promotes LINE-1–encoded gene expression and might potentiate LINE-1–autonomous retrotransposition.

We used an established system to examine whether KDM4B could affect LINE-1 retrotransposition activity as described previously (31). Briefly, the retrotransposition cassette (Supplementary Fig. S4A) contained a full-length human LINE-1 retrotransposon (driven by its native promoter) with antisense EGFP expression at its 3′ UTR. The EGFP gene was interrupted by a γ-globin intron in the same orientation as the LINE-1 transcript. This arrangement ensured that functional EGFP expression occurred only after a LINE-1 retrotransposition event or insertional mutagenesis, that is, following LINE-1 expression, γ-globin intron splicing, reverse transcription and insertion of a copy of LINE-1 into the genomic DNA of the host cell. We transfected 293T-KDM4B or 293T-vector cells with the engineered human LINE-1 retrotransposition cassette (EF06R) to detect LINE-1 retrotransposition events. EGFP-N1 was employed to monitor the transfection efficiency. The expression of EGFP in the EF06R-transfected cells could be observed under a fluorescence microscope, and the proportion of cells with fluorescence were measured by flow cytometry. On day 5 posttransfection, an EGFP signal could easily be observed in 293T-overexpressing KDM4B but not in 293T-vector cells (Fig. 4A). For quantitative analysis, the cells were harvested and screened by flow cytometry on red-versus-green fluorescence plots. After the transfection efficiency was calibrated by EGFP-N1 (Supplementary Fig. S4B), we found that KDM4B overexpression resulted in a 5-fold increase in LINE-1 retrotransposition frequency (Fig. 4B). Thus, KDM4B overexpression resulted in an increase in LINE-1 retrotransposition.

To assess whether the enhanced LINE-1 retrotransposition activity caused by ectopic KDM4B could influence the LINE-1 copy number, we then quantified the LINE-1 copy number of genomic DNA derived from 293T-stable cell lines using quantitative RT-PCR with primers spanning three different regions of LINE-1 including 5′ UTR, ORF1, and ORF2. We observed an obvious increase of DNA content in all of the three regions in the KDM4B-overexpressing cells compared with the vector-expressing cells in two different internal controls, which were designed for human endogenous retrovirus (HERVH) and alpha-satellite (SATA; Fig. 4C).

Previous studies have shown that LINE-1 genes are highly expressed in breast cancer (32, 33); however, whether their overexpression is correlated with the KDM4B protein level is not known. To test this hypothesis, we measured LINE-1–encoded ORF1 protein expression in normal breast epithelial cells (MCF10A) and breast cancer cells (MDA-MB-231, MCF7, ZR-75-30, and T47D) via Western blot analysis (Fig. 5A; Supplementary Fig. S5). The results showed that T47D cells, which have been reported to have the highest level of KDM4B protein among tested breast cancer cells (34), also exhibited the highest ORF1 protein levels. Moreover, a recent study has also shown that LINE-1 mRNA expression in T47D cells is higher than in other breast cancer cells (32). In addition, both MCF7 and ZR-75-30 cells moderately express KDM4B, and both showed clear ORF1p expression. In contrast, neither KDM4B nor ORF1p could be detected in MCF10A normal breast epithelial cells or MDA-MB-231 breast cancer cells, both of which contain barely detectable LINE-1 protein levels. To investigate the effect of different levels of KDM4B content on LINE-1 copy number variations, we quantified the LINE-1 copy number of genomic DNA derived from MCF10A, MDA-MB-231, MCF7, T47D and ZR-75-30 cells as performed in 293T cells, as shown in Fig. 5B, higher LINE-1 copy numbers were revealed in T47D, MCF7 and ZR-75-30 cell lines, all of which contain higher KDM4B protein level compared with MCF10A and MDA-MB-231. At the same time, the LINE-1 retrotransposition activity of various breast cancer cells and normal breast epithelial cell, including MCA10A, MCF7, T47D cell lines were assessed by a reporter system that described in Supplementary Fig. S4A. As shown in Supplementary Fig. S6A, in consensus with the western result in Fig. 5A, the breast cancer cell T47D exhibited the highest retrotransposition activity and MCF7 cell also showed more plots than MCF10A. We also expressed phage-EGFP-KDM4B in MCF7 cells and performed an indirect immunofluorescence assay. As shown in Fig. 5C, the ORF1 protein (red) level was clearly increased in cells expressing EGFP-KDM4B (green) relative to cells without KDM4B. These results suggest that KDM4B overexpression might be an indicator of the LINE-1 protein level and associated with retrotransposition activity in cancer cells.

To further confirm such a correlation in breast cancer cells, we knocked down KDM4B in MCF7 cells and found that KDM4B depletion resulted in reduced ORF1 protein levels (Fig. 5D). We then overexpressed KDM4B in stable MCF7 shKDM4B cells, and as shown in Fig. 5E, the ORF1 protein level was partially restored upon ectopic KDM4B expression. The effect of KDM4B on LINE-1 copy number in MCF7 breast cancer cells was evaluated by quantifying the LINE-1 copy number of genomic DNA derived from MCF7 stable cell lines. Correspondingly, we observed an obvious increase in LINE-1 content in KDM4B-overexpressing breast cancer cells compared with the control in two different internal controls. In addition, MCF7-stable cell lines were also subjected to detect the LINE-1 retroposition activity by the FACS Aria III (BD Biosciences) to count the red-versus-green fluorescence plots, similar to the observation in 293T, the ectopic KDM4B in MCF7 also result in enhanced retrotransposition activity (Supplementary Fig. S6B).

Previous reports have demonstrated that overexpression of LINE-1 or ORF2p alone in cancer cells leads to DSBs because of increased retrotransposition activity (35–37). Consistently, our results showed that overexpression of ORF2p in 293T and HCT116 cells caused an increase in γH2AX, which is a hallmark of DNA damage (Supplementary Fig. S7A). Because our data indicated that KDM4B overexpression could promote LINE-1 expression and the associated retrotransposition, we reasoned that ectopic KDM4B would be able to cause DNA damage. To test this hypothesis, we transiently transfected 293T cells with WT or the histone demethylase activity-dead mutant KDM4B, and the results showed that WT KMD4B but not the mutant led to enhanced γH2AX, suggesting that KDM4B promotes DNA damage (Fig. 6A; Supplementary Fig. S7B). Consistently, KDM4B overexpression increased DNA damage while KDM4B depletion reduced DNA damage in MCF7 cells (Fig. 6B). Indeed, the immunofluorescence results showed that KDM4B overexpression resulted in an increase in γH2AX foci (Fig. 6C). Similarly, knockdown of either KDM4B or ORF2 using two different siRNAs in T47D cells resulted in clearly reduced γH2AX levels compared with the control (Fig. 6D and E). To further confirm the conclusion that ectopic KDM4B promotes DNA DSBs by activating LINE-1, we knocked down ORF2 in MCF7 cells overexpressing KDM4B, and the result showed that KDM4B overexpression enhanced DNA damage, while the ORF2 depletion in KDM4B-overexpressing cells partially restored the KDM4B-mediated DNA damage (Fig. 6F).

KDMs are increasingly recognized as therapeutic targets for cancer and other diseases (38–40). Chia-Han Chu performed a structure-based, virtual screening for KDM inhibitors and identified a compound (NSC636819) that possesses KDM4A/4B-selective inhibition activity (41). We treated MCF7 cells with different concentrations of this inhibitor for 48 hours, and LINE-1 expression was determined by Western blotting with ORF1p antibody. The expression of ORF1p decreased gradually in an inhibitor titration-dependent manner (Fig. 7A). We then treated MCF7, T47D, and MDA-MB-231 breast cancer cells with 10 μmol/L inhibitor, and the results showed that MCF7 and T47D cells but not MDA-MB-231 cells displayed attenuated DNA damage (Fig. 7B). Because MDA-MB-231 cells express little KDM4B, this result implies that NSC636819 treatment to reduce DNA damage in cancer cells is KDM4B dependent. To further confirm the therapeutic effect of the KDM4B inhibitor, an alkaline comet assay was used. We found that a comet tail could be observed in most cells overexpressing KDM4B but not in control cells, suggesting that ectopic KDM4B promoted DNA damage. Strikingly, when MCF7 cells stably overexpressing KDM4B were treated with 20 μmol/L inhibitor, the comet tail nearly disappeared, indicating alleviation of DNA damage, the comet tail was quantified with the Comet Assay Software Project (CASP; Fig. 7C). These data strongly suggest that KDM4B might serve as a cancer therapeutic target to reduce DNA damage and safeguard genome integrity.

Oncogene derepression plays a critical role in all the hallmarks of cancer. An extensive number of studies have indicated that by demethylating tri- or dimethylation on H3K9, KDM4B broadly enhances oncogene expression in various types of cancer. Another important driving force underlying cancer progression in almost all malignancies is genome instability; however, the correlation between KDM4B overexpression and genome instability is not fully understood. Here, we show that KDM4B is a novel LINE-1 regulator. KDM4B-mediated H3K9me3 alteration is prominently enriched in evolutionarily young human LINE-1 elements. KMD4B overexpression resulted in increased LINE-1 gene expression and clearly enhanced its retrotransposon activity concomitant with DNA damage. Thus, a correlation between KDM4B and genome instability was established, revealing a novel etiology of KDM4B overexpression in tumors.

LINE-1 is the only known autonomous non-long terminal repeat retrotransposon. LINE-1 integrates into the genome by target-primed reverse transcription (TPRT), and its insertion can alter the mammalian genome in many ways upon retrotransposition and therefore lead to genomic instability, which may contribute to genetic disorders and cancer (42). There are approximately 1000,000 LINE-1 copies in the human genome, accounting for approximately 17% of the genome (43, 44). However, our study revealed that more than 35% of global H3K9me3 is enriched in LINE-1, a ratio much higher than 17%, suggesting that LINE-1 bears abundant H3K9me3 and H3K9me3 alterations in LINE-1 might be crucial for LINE-1 activation. Notably, LINE-1 encoded proteins are frequently overexpressed in breast cancer cells or tissues (32, 33, 45). Thus, LINE-1 regulators may also be associated with breast cancer. Indeed, a number of factors have been identified to regulate LINE-1 activity, including the H3K9 histone methyltransferases Suv39H1 and SETDB1 (46, 47). However, whether an eraser of H3K9 methylation is involved in LINE-1 expression regulation is unknown. Our study provides the first evidence that KDM4B effectively promotes LINE-1 expression and contributes to genome instability. In fact, only a few LINE-1 elements possess retrotransposition competency due to 5′ truncations, inversions, and point mutations within the LINE-1-encoded ORFps. A recent study indicated that in humans only evolutionarily young LINE-1 retains retrotransposition activity (29). Coincidentally, KDM4B mediated the H3K9me3 alteration preferentially in young LINE-1s, again reinforcing the idea that KDM4B is an important LINE-1 activity regulator and serves as an oncogene.

KDM4B has been implicated in the process of DNA repair and may act as a DNA damage response (DDR) protein because it can be recruited rapidly to DNA damage sites and may facilitate the repair process (48). Another study showed that KDM4B could be directly induced by P53 through promoter binding and reduce the level of H3K9 trimethylation at pericentric heterochromatin, thereby promoting heterochromatin relaxation and increasing accessibility for DNA repair factors after DNA damage (49). Furthermore, KDM4B regulates P53 transcriptional targets genes, such as p21, PIG3, and PUMA (26). However, KDM4B overexpression leads to genome instability by demethylating pericentromeric H3K9me3, which results in a significant increase in the chromosome mis-segregation rate and promotes genome instability (18). Our study identified a distinct mechanism by which ectopic KDM4B induces DNA damage and genome instability. Thus, KDM4B overexpression may play a dual role in tumorigenesis by either generating genome instability or contributing to failure of anticancer therapy that relies on induction of DNA damage for enhanced DNA repair efficiency.

Accumulation of DNA damage is one of the major driving forces in breast cancer. A previous study provided evidence that endogenous estrogen oxidative metabolites in the mammary gland were sufficient to induce DNA DSBs and genomic instability, which could be attenuated by antioxidant treatments (50). However, additional origins of DNA damage and other strategies that target the breast cancer cell endogenous DNA damage have not been reported. Our results suggest that KDM4B overexpression is an unexpected inducer of endogenous DNA damage, and specific pharmacologic inhibition of KDM4B might reduce DNA damage in breast cancer cells with high levels of KDM4 family proteins. Given that genome instability underlies cancer initiation as well as cancer progression, our study may help in developing new strategies for cancer prevention and cancer therapy.

No potential conflicts of interest were disclosed.

Conception and design: Y. Xiang, F. Li

Development of methodology: Y. Xiang, Q. Zheng, H. Ke, J. Cheng, W. Xiong, X. Shi, F. Li

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): Y. Xiang, L. Wei, M. Zhao, F. Li

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): Y. Xiang, K. Yan, X. Lu, F. Li

Writing, review, and/or revision of the manuscript: Y. Xiang, K. Yan, X. Lu, X. Lu, F. Li

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): Y. Xiang, F. Yang, P. Wang, L. Fu, F. Li

Study supervision: X. Lu, F. Li

This work was supported in part by grants from the Strategic Priority Research Program of the Chinese Academy of Sciences XDB13000000; The National Natural Science Foundation of China (81472628, 91631107, 91531305 and 31771416); The Key Research Program of the Chinese Academy of Sciences; National Key R&D program of China (2017YFA0503900); Hubei Province Health and Family Planning Scientific Research Project WJ2017Q001; The Natural Science Foundation of Hubei Province of China (2017CFA017); and The Innovation Program of the Beijing institute of genomics.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.