Transcriptional Repression of Histone Deacetylase 3 by the Histone Demethylase KDM2A Is Coupled to Tumorigenicity of Lung Cancer Cells*

Background: Overexpression of the epigenetic repressor KDM2A promotes lung tumorigenesis. Results: Transcriptional inhibition of HDAC3 expression by KDM2A releases cell cycle and proinvasive genes from HDAC3-mediated repression and positively regulates cell proliferation and invasiveness. Conclusion: KDM2A-mediated repression of HDAC3 is linked to KDM2A-promoted tumorigenicity. Significance: Our findings provide a novel epigenetic insight into how KDM2A promotes lung tumorigenesis and have implications for therapeutic intervention. Dysregulated expression of histone methyltransferases and demethylases is an emerging epigenetic mechanism underlying cancer development and metastasis. We recently showed that the histone H3 lysine 36 (H3K36) demethylase KDM2A (also called FBXL11 and JHDM1A) is necessary for tumorigenic and metastatic capabilities of KDM2A-overexpressing non-small cell lung cancer (NSCLC) cells. Here, we report that KDM2A transcriptionally represses the histone deacetylase 3 (HDAC3) gene by removing methyl groups from dimethylated H3K36 at the HDAC3 promoter in KDM2A-overexpressing NSCLC cells. KDM2A depletion reduced expression levels of cell cycle-associated genes (e.g. CDK6) and cell invasion-related genes (e.g. NANOS1); these levels were rescued by ectopic expression of KDM2A but not its catalytic mutant. These genes were occupied and down-regulated by HDAC3. HDAC3 knockdown significantly recovered the proliferation and invasiveness of KDM2A-depleted NSCLC cells as well as the levels of CDK6 and NANOS1 expression in these cells. Similar to their previously reported functions in other cell types, CDK6 and NANOS1 were required for the proliferation and invasion, respectively, of KDM2A-overexpressing NSCLC cells. In a mouse xenograft model, HDAC3 depletion substantially restored the tumorigenic ability of KDM2A knockdown cells. These findings reveal a novel cancer-epigenetic pathway in which the antagonistic effect of KDM2A on HDAC3 expression releases cell cycle-associated genes and cell invasion-related genes from HDAC3 repression and indicate the importance of this pathway for tumorigenicity and invasiveness of KDM2A-overexpressing NSCLC cells.

Histone lysine methylation is considered a key chromatin mark that mediates epigenetic and transcriptional regulation of gene expression (1,2). Unlike histone acetylation, which is exclusively associated with gene activation, this modification is linked to activation or silencing of gene expression (3,4). The methylation effect depends on the lysine residues involved, among which are histone H3 lysine 4 (H3K4), 3 H3K9, H3K27, H3K36, H3K79, and H4K20 (5). For example, methylated H3K36 is connected to gene activation, whereas methylated H3K27 is related to gene silencing. Like histone acetylation, which can be removed by histone deacetylases (HDACs), histone lysine methylation is reversibly regulated: it is catalyzed by histone lysine methyltransferases (1,2) and can be removed by histone lysine demethylases (KDMs) (6,7). Notably, this modification exists at three different states (i.e. mono-, di-, and trimethylation).
Lung cancer is the most prevalent cause of cancer-related mortality in the United States and worldwide (8). Non-small cell lung cancer (NSCLC) is responsible for up to 85% of lung cancer and includes adenocarcinomas, squamous cell carcinomas, and large cell carcinomas. In NSCLC, genetic mutations and abnormalities in kinase signaling pathway members have been well documented (9). For instance, in lung adenocarcinomas, activating mutations for oncogenes frequently occur in K-RAS and epidermal growth factor receptor gene, whereas mutations in tumor suppressor genes, such as LKB1, are also detected (9,10). Thus, much research focus for NSCLC has been put on understanding the kinase signaling pathway.
Recently, it has been increasingly evident that dysregulation of histone methylation and its modifiers may be an important factor in lung tumorigenesis and metastasis (11,12). Alterations in histone lysine methylation are associated with clinical prognosis of lung cancer (13,14). For example, global levels of trimethylated H4K20, a gene-repressive mark, are decreased in squamous cell carcinomas, and its low levels correlate with poor prognosis (14). In addition to methylation marks, some histone methylation modifiers are known to have altered expression levels in lung cancer. For instance, the H3K27 methyltransferase EZH2 and the H3K36 methyltransferase MMSET (also called WHSC1 and NSD2) are significantly overexpressed in lung tumors (15). Recently, we showed that the H3K36 demethylase KDM2A (also known as FBXL11 and JHDM1A) is frequently up-regulated in NSCLC tumor samples and promotes the tumor growth and invasive abilities of NSCLC cells (16). Consistent with the notion that KDM2A is associated with transcriptional repression by erasing methyl groups from the gene activation mark dimethylated H3K36 (H3K36me2) (17)(18)(19), we also reported that KDM2A increases cellular levels of phospho-ERK1/2, a type of oncogenic signal, by down-regulating the expression of dual specificity phosphatase 3 (DUSP3) whose protein dephosphorylates ERK1/2 (16).
To gain greater insights into how KDM2A regulates cell proliferation and invasion, we sought to identify more KDM2A target genes. Here, we provide evidence that HDAC3 is an important target gene of KDM2A. Transcriptional repression of the HDAC3 gene by KDM2A-catalyzed H3K36 demethylation up-regulates HDAC3 target genes, including the cell cycle-associated gene CDK6 and the cell invasion-related gene NANOS1 in two KDM2A-overexpressing NSCLC cell lines. Furthermore, our results suggest that epigenetic repression of HDAC3 expression by KDM2A is needed for the tumorigenic and invasive abilities of KDM2A-overexpressing NSCLC cells.
In Vitro Gene Silencing Using siRNA-For knockdown experiments, siRNAs against KDM2A, HDAC3, CDK6, and NANOS1 were purchased from Dharmacon or Integrated DNA Technologies, Inc. (IDT). The siRNA sequences are listed in Table 1. As controls, siRNA against luciferase GL3 RNA (siLuc) and siControl were used. Cells (5 ϫ 10 4 ) in a 6-well plate were transfected with siRNAs at a final concentration of 100 nM using Lipofectamine RNAiMAX (Invitrogen). Following 72-96 h of incubation, cells were harvested for mRNA and protein analysis or used for cell proliferation and invasion assays.
Quantitative RT-PCR-Total RNA was isolated using RNeasy kits (Qiagen) according to the manufacturer's instructions. Then cDNA was synthesized using the iScript cDNA synthesis kit (Bio-Rad) according to the manufacturer's instructions. Quantitative PCR was performed in triplicate using SYBR Green and gene-specific primers (Table 1). Messenger RNA levels were quantified using CFX Manager software (Bio-Rad) and were normalized to ␤-actin mRNA levels. The relative mRNA levels represent the -fold change compared with the levels in control cells.
Cell Proliferation and Invasion Assays-To assess cell proliferation, 1 ϫ 10 4 cells/well in 24-well plates were seeded in triplicate, and cell number were counted at the indicated time points. For the cell invasion assay, the Boyden chamber assay with a modification was performed. In brief, cells (1 ϫ 10 5 ) were seeded on the Matrigel-coated membrane in the inserts. After 18 h of incubation, cells that had invaded the Matrigel and migrated to the other side of the membrane were stained and counted.
Cell Cycle Analysis and BrdU Incorporation Assays-For cell cycle analysis, cells were fixed in cold 75% ethanol for 30 min at 4°C and washed twice in PBS. Then cells were incubated at 37°C for 30 min in a buffer containing 50 g/ml propidium iodine, 5 mM MgCl 2 , 10 mM Tris-HCl (pH 7.0), and 25 g/ml RNase A. DNA contents were analyzed using flow cytometry (BD FACSCanto II, BD Biosciences).
For BrdU incorporation assays, 1-2 ϫ 10 5 cells were seeded in 96-well plates in triplicate. BrdU reagent was added to each well, and the plate was incubated for 24 h. BrdU incorporation was measured at 450 nm absorbance according to the manufacturer's protocol (Millipore).
Rescue Experiments by Ectopic Expression-To test the effects of ectopic expression of KDM2A in H1975 and H1792 cells, the cDNAs encoding KDM2A, its catalytic mutant mKDM2A, and GFP were individually cloned into the mammalian expression vector pFLAG-CMV2 as described previously (16). The resulting expression plasmids (2 g) were transfected into either siControl-or siKDM2A-treated cells (2-5 ϫ 10 5 cells) in 60-mm dishes using 10 l of Lipofectamine 2000 (Invitrogen). These cells were harvested after 72 h of incubation for further analysis, such as quantitative RT-PCR.
Quantitative ChIP Assay-A ChIP assay was performed as described previously (20). After chromatin immunoprecipitation using specific antibodies, DNA was purified using the Qiagen PCR purification kit. Purified DNAs for KDM2A, HDAC3, and the histone marks of interest were amplified by quantitative PCR (qPCR) and normalized to input. Relative occupancy indicates the change over the control value. Enrichment represents PCR values from specific antibody relative to those from IgG (e.g. anti-KDM2A/IgG and anti-HDAC3/IgG).
Mouse Xenograft Study-To determine whether the effect of KMD2A knockdown on tumorigenesis is dependent on HDAC3, three groups of cells (shControl-treated, KDM2A-depleted, and KDM2A/HDAC3-depleted H1792 cells) were compared for their tumorigenicity in a subcutaneous xenograft model. KDM2A-depleted cells were generated using shRNA against KDM2A as described previously (21). KDM2A/ HDAC3-depleted cells were generated by treating KDM2A-depleted cells with 50 nM siHDAC3-9 using Lipofectamine RNAiMAX. For comparison, the other two groups of cells were also transfected with 50 nM siScramble (a control siRNA). After a 24-h incubation, all three groups of cells were retransfected with the same amounts of siRNAs at the same concentrations.
An additional 72 h later, cells were harvested and suspended in RPMI 1640 medium without serum. Cells (1.5 ϫ 10 6 ) were subcutaneously injected into the dorsal flanks of male nude mice (8 weeks old). At least five mice were injected for each group and observed for 10 weeks for tumor formation. The ellipsoid volume formula (1/2 ϫ L ϫ W ϫ H) was used to calculate the tumor volume. Ten weeks after the injection, the mice were killed, and their tumors and lungs were collected. For pathological analysis, hematoxylin and eosin (H&E) and immunochemical staining was performed.
Statistical Analysis-For statistical analysis, each experiment was performed in triplicate and repeated at least three times. Data are presented as the mean Ϯ S.E. (error bars). Statistical significance was tested by the two-tailed Student's t test. * (p Ͻ 0.05), ** (p Ͻ 0.01), and *** (p Ͻ 0.001) indicate statistically significant differences. GraphPad Prism software was used for all statistical analyses.

KDM2A Indirectly Up-regulates Expression of Cell Cycle-associated Genes and Cell Invasion-related Genes in KDM2A-
overexpressing NSCLC Cells-In our effort to better understand the mechanisms by which KDM2A may regulate the proliferation and invasion of NSCLC cells, we revisited our recent whole genome mRNA expression data in which a number of genes were commonly modulated by two different siKDM2As in two KDM2A-overexpressing NSCLC cell lines, H1975 and H1792.

Gene
Application Because our recent study showed that KDM2A knockdown decreased the S phase percentages and invasive abilities of H1975 and H1792 cells (16), we paid our attention to cell cycleassociated genes and cell invasion-related genes. In particular, the cell cycle-associated genes CDK6 and NEK7 and the invasion-related genes NANOS1 and RAPH1 were of interest because these genes were down-regulated by KDM2A knockdown (16). CDK6 was demonstrated to be often overexpressed in lung tumors (22). CDK6 activity promotes expression of cell cycle regulators and assembly of the prereplication complex (23), and pharmacologic inhibition of CDK6 induces cell cycle arrest (24). NEK7, a centrosomal kinase, regulates proper spindle assembly and mitotic progression (25). NANOS1 is overexpressed in some human invasive lung carcinomas and increases the production of MT1-matrix metalloproteinase to induce matrix metalloproteinase-dependent invasion of carcinoma cells (26). RAPH1 is known to promote cell proliferation and to play an important role in cell motility (27). To confirm our previous microarray results, we compared expression levels of the CDK6, NEK7, NANOS1, and RAPH1 genes between KDM2A-depleted cells and control siRNA-treated cells using quantitative RT-PCR. Two KDM2A-overexpressing NSCLC cell lines H1975 and H1792 were examined. Our results ensured that these genes were down-regulated by KDM2A knockdown (Fig. 1, A and B). Next, we examined whether ectopic expression of KDM2A restores expression of the KDM2A-regulated genes CDK6, NEK7, NANOS1, and RAPH1 in KDM2A-depleted cells. Wild-type KDM2A and its catalytic mutant mKDM2A were transiently expressed as described previously (16). Exogenous expression of wild-type KDM2A, but not mKDM2A, significantly rescued expression of these genes in H1975 and H1792 cells (Fig. 1, C-F, showing data for CDK6 and NANOS1; similar data are not shown for NEK7 and RAPH1). These results indicate that KDM2A and its demethy-lase activity may be requisite for expression of the CDK6, NEK7, NANOS1, and RAPH1 genes.
KDM2A is linked to transcriptional repression of its target genes (17)(18)(19). Therefore, KDM2A may indirectly up-regulate the CDK6, NEK7, NANOS1, and RAPH1 genes whose expression levels were decreased by KDM2A knockdown. Thus, we determined whether these genes are indirectly or directly regulated by KDM2A using quantitative ChIP assays. Because it has been demonstrated that KDM2A is located at regions spanning transcription start sites (16,17), we analyzed KDM2A occupancy near those sites in these genes. ChIP results showed that the regions near the transcription start sites in the CDK6, NEK7, NANOS1, and RAPH1 genes were not occupied by KDM2A. In contrast, DOCK7 and ZNF652 whose expression levels were increased by KDM2A knockdown were KDM2A target genes (Fig. 1, G and H). In line with its role in repressing transcription, these results indicate that KDM2A indirectly upregulates expression of the CDK6, NEK7, NANOS1, and RAPH1 genes while directly down-regulating expression of the DOCK7 and ZNF652 genes.
Expression of the HDAC3 Gene Is Repressed by KDM2A-catalyzed H3K36 Demethylation-Because CDK6, NEK7, NANOS1, and RAPH1 were not direct KDM2A target genes, we hypothesized that these genes may be down-regulated by a transcriptional co-repressor encoded by a KDM2A-repressed gene. Specifically, we reasoned that HDAC3 might be involved in the repression of these genes as our previous microarray results indicated that HDAC3 is a KDM2A-repressed gene (16). HDAC3 is a well known transcriptional co-repressor (28 -30). Similar to the microarray results, our quantitative RT-PCR data showed that HDAC3 expression levels were up-regulated in KDM2A knockdown cells ( Fig. 2A). In addition, our ectopic experiments demonstrated that expression of wild-type KDM2A, but not its catalytic mutant mKDM2A, significantly repressed expression levels of HDAC3 in KDM2A-depleted cells (Fig. 2, B and C). These results indicated that KDM2A and its enzymatic activity are critical for KDM2A-mediated repression of the HDAC3 gene.
To determine whether KDM2A is recruited to the HDAC3 gene, we examined enrichment levels of KDM2A over control IgG at a proximal promoter region of the HDAC3 gene using quantitative ChIP assays (Fig. 2D). Our results showed that KDM2A occupied the proximal promoter near the transcription start site in the HDAC3 gene, indicating that HDAC3 is a KDM2A target gene (Fig. 2E). These results are in agreement with our previous study showing that KDM2A is localized near the transcription start site of the DUSP3 gene (16). Because KDM2A removes methyl groups from H3K36me2, we determined the effect of KDM2A knockdown on H3K36me2 levels at the proximal promoter of the HDAC3 gene. Quantitative ChIP results demonstrated that KDM2A knockdown resulted in increased H3K36me2 levels at the proximal promoter of the HDAC3 gene in H1975 and H1792 cells (Fig. 2, F and G). These results indicate that KDM2A-catalyzed demethylation of H3K36me2 at the HDAC3 promoter plays an important role in KDM2A-mediated repression of the HDAC3 gene.
HDAC3 Directly Represses Expression of the CDK6, NEK7, NANOS1, and RAPH1 Genes-To examine the possibility that HDAC3 may down-regulate expression levels of CDK6, NEK7, NANOS1, and RAPH1 genes, we depleted HDAC3 in H1792 and H1975 cells using two different siRNAs against HDAC3 (Fig. 3, A and B). Quantitative RT-PCR results showed that HDAC3 depletion by two independent siHDAC3s increased expression levels of these genes (Fig. 3, C and D). Next, we examined whether HDAC3 is recruited to CDK6, NEK7, NANOS1, and RAPH1 genes using quantitative ChIP. Our results showed that HDAC3 occupied the regions spanning the transcription start sites at these genes (Fig. 4, A-D), consistent with the previous report showing that chromatin peaks of HDAC3 cover the transcription start sites (31). These results indicate that HDAC3 directly represses CDK6, NEK7, NANOS1, and RAPH1 genes and support the idea that KDM2A up-regulates expression of these genes by repressing HDAC3 expression at transcriptional levels.
To assess whether HDAC3 represses its target genes by deacetylation, we examined the effect of HDAC3 knockdown on histone acetylation levels at the HDAC3 target genes, such as CDK6 and NANOS1. Consistent with the deacetylase activity of HDAC3, our ChIP results demonstrated that HDAC3 depletion increased acetylation levels in histones H3 and H4 at these genes in H1792 and H1975 cells (Fig. 5, A-D). Similar to no obvious effect of KDM2A knockdown on cellular levels of H3K36me2, HDAC3 depletion did not affect total cellular levels of H3K9 acetylation and H4 acetylation, suggesting that HDAC3 erases histone acetylation in a gene-specific manner (Fig. 5E).
In KDM2A-overexpressing NSCLC Cells, CDK6 Positively Regulates Cell Proliferation, whereas NANOS1 Is Required for Cellular Invasiveness-As mentioned earlier, it has been shown that CDK6 regulates cell cycle progression in non-lung cancer cell lines (e.g. U2OS), whereas NANOS1 modulates cellular invasion in the breast cancer cell lines Hs578T and BT549 and the transformed lung epithelial cell line BZR. To assess whether CDK6 and NANOS1 play similar roles in KDM2A-overexpressing NSCLC cells, we assessed the effect of CDK6 or NANOS1 knockdown on the proliferation and invasiveness of H1792 cells. CDK6 depletion strongly inhibited cell proliferation and BrdU incorporation into the replicated DNA of S phase cells but only weakly impeded cellular invasiveness (Fig.  6, A-D). On the contrary, NANOS1 knockdown only marginally inhibited cell proliferation and BrdU incorporation but markedly abrogated cellular invasiveness (Fig. 6, E-H). These results suggest that in KDM2A-overexpressing NSCLC cells CDK6 is associated largely with cell proliferation, whereas NANOS1 is related mainly to cellular invasiveness.

KDM2A-mediated Repression of HDAC3 Expression Is Important for in Vitro Cell Proliferation and Invasion as Well as
Xenograft Tumor Formation-To determine whether KDM2A regulates the CDK6 and NANOS1 genes in an HDAC3-dependent manner, we reduced HDAC3 in KDM2A-depleted H1975 and H1792 cells using siHDAC3 (Fig. 7, A and B). HDAC3 knockdown significantly restored CDK6 and NANOS1 mRNA levels in KDM2A-depleted H1975 and H1792 cells (Fig. 7, C-F). In addition, HDAC3 depletion significantly rescued deficient proliferation of KDM2A-depleted cells by restoring the S phase population (Fig. 8, A-C). Consistent with this, HDAC3 knockdown increased BrdU incorporation in KDM2A-depleted cells (Fig. 8D). Interestingly, HDAC3 depletion did not affect the G 2 /M phase population that was increased by KDM2A knockdown (Fig. 8C). Similar to its effects on cell proliferation, HDAC3 depletion significantly recovered cellular invasiveness (Fig. 9, A and B). These results indicate that transcriptional repression of HDAC3 by KDM2A may contribute to cell proliferation and invasiveness by releasing cell cycle-associated genes and invasiveness-related genes from HDAC3 repression.
To determine the effect of KDM2A-mediated repression of HDAC3 on the tumorigenic ability of NSCLC cells in vivo, we depleted HDAC3 in stably KDM2A-depleted H1792 cells using siHDAC3, subcutaneously injected double knockdown cells into mice, and monitored tumor growth. Similar to siRNA-based transient knockdown of KDM2A, stable knockdown of KDM2A increased HDAC3 expression (Fig. 10, A and B). In agreement with our previous report (16), most mice (n ϭ 4 of 5) that were injected with shLuciferase-treated H1792 cells had tumors within 3 months (Fig. 10C and Table 2). In sharp contrast, no mice injected with KDM2A-depleted cells had detectable tumors (n ϭ 0 of 5) (Fig. 10C and Table 2). Interestingly, HDAC3 depletion greatly restored defective tumorigenic abil-ities of KDM2A-depleted H1792 cells (Fig. 10, C and D, and Table 2). Tumors were confirmed by H&E staining, and CDK6 and NANOS1 levels in a control tumor were comparable with those in a tumor that was originated from double knockdown cells (Fig. 10, E and F). Results obtained from this subcutaneous xenograft experiment indicate that in vivo formation of tumors by KDM2A-overexpressing NSCLC cells may require KDM2Amediated repression of HDAC3 expression.

DISCUSSION
In this study, our results provided evidence that KDM2A represses expression of the HDAC3 gene by demethylating H3K36me2 at the HDAC3 promoter. Consistent with our previous study showing that KDM2A positively regulates the S phase cell population and cellular invasiveness (16), the results presented here indicate that KDM2A enhances expression of the cell cycle-associated genes CDK6 and NEK7 and the cell invasion-related genes NANOS1 and RAPH1 in KDM2A-overexpressing NSCLC cells. Interestingly, expression of these genes was directly repressed by the transcriptional co-repressor HDAC3. Therefore, our current study uncovered the molecular mechanism in which KDM2A-mediated repression of HDAC3 expression may antagonize the transcriptional repression of cell cycle-associated genes and invasiveness-related genes by HDAC3 in KDM2A-overexpressing lung cancer cells (i.e. KDM2A ٜ HDAC3 ٜ cell cycle/invasiveness genes (e.g. CDK6 and NANOS1)) (Fig. 10G). In addition, KDM2A-mediated regulation of HDAC3 expression may be largely dependent on the catalytic activity of KDM2A because the KDM2A catalytic mutant did not repress HDAC3 expression.
It is known that KDM2A acts as a positive regulator for somatic cell reprogramming and cellular antisenescence (32,33). Our previous data suggest that KDM2A overexpression promotes NSCLC tumorigenesis and metastasis by increasing ERK1/2 signaling (16). Data reported here show that HDAC3 depletion significantly revived the tumorigenic and invasive abilities of KDM2A-depleted cells and restored expression of the cell cycle-associated gene CDK6 and the invasion-related gene NANOS1 in KDM2A-depleted cells. Thus, this work supports the notion that KDM2A-mediated repression of HDAC3 expression contributes to NSCLC tumorigenesis and invasion at least in part by up-regulating certain cell cycle-associated and invasion-related genes. On the basis of our previous and current studies, it is possible that KDM2A promotes tumorigenicity and invasiveness of KDM2A-overexpressing NSCLC cells by controlling two pathways for cell cycle and cell invasion: 1) KDM2A ٜ DUSP3 ٜ ERK1/2 3 cell cycle/invasiveness and 2) KDM2A ٜ HDAC3 ٜ cell cycle/invasiveness genes. Interestingly, treatment of KDM2A-overexpressing cells with the MEK1/2 inhibitor U0126 (10 M) that consequently impedes ERK1/2 phosphorylation did not show any obvious effect on HDAC3 levels, suggesting that these two pathways may be independent (data not shown).
Certain histone modifiers regulate gene expression in a cooperative mode. The H3K9 methyltransferase G9a and the H3K4 demethylase JARID1a co-repress gene expression (34), and the H3K4 demethylase LSD1 and HDAC1 and -2 cooperatively repress their co-target genes (35). The H3K27 demethylase UTX and H3K4 methyltransferase MLL4 (also known as ALR and KMT2D) may co-activate their co-target genes (36). In contrast, some histone modifiers antagonize the function of others on the same genes. For example, polycomb repressive complex 2 containing the H3K27 methyltransferase EZH2 may enable cancer cells to resist cellular senescence by inhibiting expression of the cellular senescence genes p16 INK4A and p14 ARF (37), whereas the H3K27 demethylase JMJD3 may neu- tralize EZH2 function by activating the same genes (38). Interestingly, the results reported here indicate that the transcriptional co-repressor KDM2A counteracts the gene-repressive effects of HDAC3 on cell cycle-regulatory genes and cell invasion-associated genes by transcriptionally down-regulating the HDAC3 gene in a catalytic activity-dependent manner. To our knowledge, the epigenetic repression of HDAC3 function by KDM2A is a distinct mode of antagonistic regulation in which a histone lysine demethylase opposes the function of a histone deacetylase via transcriptional repression.
As many as 70% of eukaryotic genes contain CpG islands, which represent genomic regions that have higher than average genome-wide levels of CpG dinucleotides (39). Blackledge et al. (17) reported that via its CXXC domain KDM2A interacts with many transcription start site-associated CpG islands. Interestingly, the HDAC3 gene has a CpG island spanning the predicted transcription start site (data not shown). Thus, it is possible that KDM2A is recruited to the HDAC3 promoter through its association with the CpG-dense region. However, the transcriptional regulation of HDAC3 expression seems more complex. Although our analysis of the tumor database The Cancer Genome Atlas showed a trend for an inverse correlation between KDM2A and HDAC3 mRNA levels in NSCLC tumors, there was no significant correlation between their levels in a range of NSCLC cell lines (data not shown). These results indicate that KDM2A is a key control factor for HDAC3 expression  in certain types of lung cancer cell lines and lung tumors but that other gene-regulatory factors may also be involved in modulating the HDAC3 promoter. Further studies are required for a better understanding of how HDAC3 expression is regulated.
HDAC inhibitors have been recently developed as anticancer agents. Multiple mechanisms underlying the mode of action of HDAC inhibitors have been described, including induced expression of tumor suppressor p21 and the proapoptotic shKDM2A-1, and shKDM2A-1 ϩ siHDAC3-9) were subcutaneously injected into five mice per group. Representative tumors (10 weeks after subcutaneous injection) are shown in dotted circles and indicated by black arrows (C). Tumor volumes were monitored and plotted for 10 weeks (D). E, representative images of H&E staining of xenograft tumor samples (shControl group and shKDM2A-1 ϩ siHDAC3-9 group) and normal skin tissues (shKDM2A-1 group) obtained from the injection sites (10 weeks after subcutaneous injection). Scale bars are indicated (upper panels, 200 m; lower panels, 100 m). F, representative images of immunohistochemical staining of NANOS1 and CDK6 in xenograft tumor samples. Immunohistochemical staining was performed only using tumor tissues from the shControl group and shKDM2A-1 ϩ siHDAC3-9 group. Normal mouse skin tissues from the shKDM2A-1 group were not used for staining because they were not a proper control. CDK6 and NANOS1 levels were quantified using the Chromavision Automated Cellular Imaging System (ACIS-III) from Dako and compared between the shControl group and shKDM2A-1 ϩ siHDAC3-9 group: CDK6, 1 versus 0.73; NANOS, 1 versus 0.54. G, in this hypothetical model, KDM2A transcriptionally represses the HDAC3 gene. Subsequently, cell cycle-associated genes (e.g. CDK6) and invasiveness-related genes (e.g. NANOS1) are released from HDAC3-mediated repression. Data are presented as the mean Ϯ S.E. (error bars). **, p Ͻ 0.01; ***, p Ͻ 0.001.
ligand TRAIL (40 -42). Recently, Chen et al. (43) reported that the HDAC inhibitor trichostatin A may increase the expression of the potential tumor suppressor cluster Dleu2/miR-15a/16-1 via HDAC3 inhibition in two NSCLC cell lines, A549 and H1299. In contrast, our results suggest that HDAC3 is involved in tumor suppression in the NSCLC cell lines H1975 and H1792. These seemingly contradictory results might result from a cell line-dependent function of HDAC3. Thus, it is possible that HDAC3 may act as an oncogenic factor by repressing tumor suppressor genes in A549 and H1299, whereas it may play a tumor-suppressive role in H1975 and H1792 by downregulating expression of cell proliferation-related genes and proinvasive genes.
In addition, the pathologic role of HDAC3 in cancer development might be tissue-dependent. In gastric, prostatic, and colorectal cancer samples, HDAC3 overexpression was significantly associated with poor prognosis (44 -46). In agreement with these reports, it has been shown that HDAC3 appeared to be up-regulated and to repress the tumor suppressor gene p21 in colorectal cancer cells (47). In contrast, other studies reported that liver-specific HDAC3 knock-out mice develop hepatoma (48) and that HDAC3 maintains genomic stability (48) and potentiates apoptosis by down-regulating the protooncogene c-Jun (28), indicating an antitumor function for HDAC3. In support of the tumor-suppressive role of HDAC3, the results reported here showed that HDAC3 knockdown restored defective proliferation and invasiveness of KDM2Adepleted cells. Our additional results indicate that HDAC3 represses cell cycle-associated genes and invasiveness-related genes. Thus, our findings suggest that the inhibition of antitumor function of HDAC3 by KDM2A at transcriptional levels may be requisite for the growth and invasion of KDM2A-overexpressing NSCLC tumors. a Stably KDM2A-depleted (shKDM2A-1) cells, double knockdown (shKDM2A-1 ϩ siHDAC3) cells, and shControl-treated cells were subcutaneously implanted into mice. At the 10th week after injection, numbers of tumor-bearing mice were counted.