|Home | About | Journals | Submit | Contact Us | Français|
Heterochromatin plays an essential role in the preservation of epigenetic information, the transcriptional repression of repetitive DNA elements and inactive genes and the proper segregation of chromosomes during mitosis. Here we identify KDM2A, a JmjC-domain containing histone demethylase, as a heterochromatin-associated protein that is required to maintain the heterochromatic state, as determined using both a candidate-based approach and an unbiased siRNA library screen. Moreover, we demonstrate that KDM2A represses transcription of small non-coding RNAs that are encoded by clusters of satellite repeats at the centromere. Finally, we show that KDM2A is required to sustain centromeric integrity and genomic stability, particularly during mitosis. Since the disruption of epigenetic control mechanisms contributes to cellular transformation, these results, together with the low levels of KDM2A found in prostate carcinomas, suggest a role for KDM2A in cancer development.
Heterochromatin is composed of highly repetitive satellite DNA and repressed genes, and this highly compacted and repressive environment is mediated by the presence of heterochromatin-associated proteins and specific epigenetic markers.1, 2 Centromeric heterochromatin contains sequences that produce small non-coding RNA molecules required for maintaining the heterochromatin state and centromeric integrity.3-7
KDM2A is an evolutionarily conserved and ubiquitously expressed member of the KDM2 Lys histone demethylase family.8 Because it contains a JmjC-domain and an F-box motif, this protein is also known as JHDM1A and FBXL11, respectively.9, 10 KDM2A was previously identified as a dimethyl Lys36 histone H3 (H3K36me2) histone demethylase.10 In this study, we characterize KDM2A as an integral heterochromatin-associated protein, and we show that KDM2A plays a crucial role in silencing centromeric satellite repeats and maintaining the heterochromatic state.
To begin assessing the biological function of KDM2A, we investigated its cellular localization in interphase cells by using indirect immunofluorescence analysis. KDM2A localized to the nucleus of cells and showed a punctate staining pattern throughout the nucleoplasm and perinucleolar enrichment (Fig. 1A), as compared to the human paralogue KDM2B/JHDM1B, which was previously identified as a nucleolar-resident protein.11 Epe1, the fission yeast orthologue of KDM2, interacts with Swi6 [the yeast heterochromatin protein 1 (HP1) orthologue] and is distributed across all major heterochromatic domains, including pericentric heterochromatin.12 To determine if the punctate staining pattern of KDM2A reflected heterochromatin association, we performed colocalization experiments with KDM2A and the HP1 variants (HP1-α, HP1-β and HP1-γ). Indirect immunofluorescence demonstrated that KDM2A partially colocalized with heterochromatin foci associated with endogenous HP1 variants (Fig. 1B). Partial colocalization of KDM2A and HP1 proteins was calculated by the Pearson (rp) and Spearman correlation coefficients (rs)13 (Figure 1, B and C). Notably, KDM2A was not observed to preferentially colocalize with any of the HP1 variants, which are known to differ in their heterochromatic localization.1 Moreover, immunoprecipitation of exogenous HP1 variants showed association with endogenous KDM2A, but not KDM2B (Fig. 1D). Interestingly, KDM2A was also independently identified by mass spectrometry analysis as an HP1-α interacting protein in a pool-down experiment using a GST-HP1-α fusion protein incubated with a HeLa nuclear extract (HP and MP, unpublished results). Together, these data demonstrate that KDM2A is a heterochromatin-associated protein that interacts with HP1.
Heterochromatin-associated proteins have been shown to regulate HP1 localization and contribute to heterochromatin formation and maintenance in mammalian cells.1, 2, 14-16 Since KDM2A was found to localize to heterochromatin and associate with HP1, we examined the status of HP1 on chromatin following KDM2A knockdown in mouse NIH-3T3 and human HeLa cells (Figure 2, A and B). Indirect immunofluorescence of HP1-γ in KDM2A knockdown mouse cells displayed a significant loss of HP1 heterochromatin-associated nuclear foci compared to control cells (Fig. 2C). Moreover, we observed a decrease in the amount of HP1-γ associated with chromatin KDM2A siRNA-treated mitotic cells during anaphase (Fig. 2D and data not shown), at a time when HP1-γ is found to be exclusively on DNA.17 These data suggest KDM2A knockdown results in the delocalization of HP1 from chromatin.
Because heterochromatin maintenance requires the presence of HP1 on chromatin, we sought to determine if the silencing of KDM2A compromises the overall heterochromatic state. To this end, we employed the use of epigenetically silent GFP (green fluorescent protein) reporter genes that have been integrated into the genome of HeLa cells.18, 19 In these GFP(-) HeLa reporter cells, reactivation of GFP expression occurs following heterochromatin disruption, such as after HP1-γ knockdown or via inhibition of histone deacetylases by histone deacetylase inhibitors like trichostatin A.18, 19 Using these reporter cells, we found that knockdown of KDM2A, but not KDM2B, resulted in the robust reactivation of GFP compared to cells transfected with a control siRNA (Fig. 2E). These results suggest KDM2A is involved in heterochromatin maintenance similar to Epe1.7, 12, 20, 21 Strikingly, an independent and unbiased whole-genome siRNA library screen utilizing these GFP(-) HeLa reporter cells (performed by AP and RK) identified KDM2A, but not KDM2B, as a gene involved in the heterochromatin maintenance of epigenetic silencing (Fig. 3F). Thus, these observations are consistent with a role of KDM2A as a regulator of the heterochromatic state.
All HP1 variants associate with centromeric heterochromatin.1 Thus, we investigated whether KDM2A was associated with centromeric heterochromatin by examining the localization of KDM2A in comparison with the centromeric protein CENP-A (centromere protein A). In interphase cells, we observed partial KDM2A colocalization with CENP-A, including within the heterochromatin that surrounds the nucleolus (Fig. S1). In areas where colocalization between KDM2A and CENP-A was not observed, we detected KDM2A foci immediately adjacent to centromeric regions (Fig. S1). Partial colocalization of KDM2A and CENPA was calculated by the Pearson (rp) and Spearman correlation coefficients (rs)13 (Fig. S1). Moreover, we investigated whether KDM2A was associated with centromeric heterochromatin by performing chromatin immunoprecipitation (ChIP) assays followed by quantitative real-time polymerase chain reaction (PCR) using primers spanning the pericentric region of a human chromosome (chromosome 4). KDM2A was found to bind to this region of chromosome 4 (Fig. 3A), and the specificity of this binding was confirmed by the fact that a DNA-binding mutant [KDM2A(CXXC)] and KDM2B failed to bind to this locus. The lack of significant enrichment of KDM2A on a housekeeping gene promoter (GAPDH) (data not shown), further validated specificity for pericentric loci. Treatment of these DNA complexes with the methylation-sensitive HpaII (5'-CCGG-3') enzyme abolished the binding of KDM2A to the pericentric region of human chromosome 4 (Fig. 3A), suggesting recognition by KDM2A to unmethylated CpG sequences at this locus. Thus, these data show that KDM2A is enriched at pericentric heterochromatin and binds to CpG islands, as predicted based on the presence of the CXXC DNA-binding motif.22
In fission yeast and in higher eukaryotic cells, small RNA molecules transcribed by RNA polymerase II (Pol II) from within centromeric and pericentromeric regions appear to be necessary to initiate and sustain repressive chromatin modifications.3-7 In human cells, the primary repetitive DNA at centromeres is α-satellite DNA, which consists of a 171-bp monomer that is tandemly arranged into higher-order arrays that extend for 100-kb to several megabases. Given that KDM2B was characterized as a transcriptional repressor of ribosomal RNA genes11, which are also encoded by tandem DNA repeats within the nucleolus, we investigated whether KDM2A contributed to the transcriptional regulation of repetitive α-satellite DNA sequences within centromeric heterochromatin in human HCT-116 cells. Using previously reported primers for analyzing transcription of these repeats, we conducted semi-quantitative RT-PCR on total RNA from KDM2A-depleted HCT-116 cells (Fig. 3B). We observed an increase in the transcription of the α-satellite transcripts of chromosomes 2, 4 and 13/21 in KDM2A-depeleted cells compared to control cells (Fig. 3C). H3K36me2 is a histone modification that is associated with positive regulation of Pol II-mediated transcription23 and is negatively regulated by KDM2A.8 In agreement with an increase in α-satellite transcription, ChIP analysis at the pericentric region of human chromosome 4 in KDM2A siRNA-treated cells displayed a significant increase in the levels of H3K36me2, but not H3K9me3, compared to control (Fig. 3D and Fig. S2), while the levels of H3K36me2 and H3K9me3 at the GAPDH promoter were unaffected (Fig. 3D and Fig. S2). Since KDM2A repressed α-satellite transcription in human cells, we also investigated whether KDM2A contributed to the transcriptional regulation of repetitive DNA sequences within the centromeric heterochromatin of murine cells by using previously characterized primers for analyzing minor and major satellite repeats.3, 16 Major satellite repeats surround the minor satellites, and these two sets of repeats are separated by short stretches of CG-rich sequences24 that may provide an ideal platform for the DNA-binding capabilities of KDM2A.22 Semi-quantitative RT-PCR on total RNA in KDM2A-depleted NIH-3T3 cells displayed a substantial increase in the transcription of the major, but not the minor, satellite repeats compared to control cells (Fig. 3E). Taken together, these data demonstrate that KDM2A is a transcriptional repressor of pericentric α-satellite and major satellite repeats in human and mouse cells, respectively.
KDM2A specifically demethylates H3K36me2,10 and since we observed an increase in H3K36me2 at pericentric chromosome 4 following KDM2A knockdown (Fig. 3D), we determined whether the JmjC-domain of KDM2A is required for satellite silencing in mammalian cells. Constructs encoding human KDM2A and a JmjC-domain mutant [KDM2A(ΔJmjC)] were introduced into NIH-3T3 cells that were previously treated with siRNA oligos against murine KDM2A. KDM2A siRNA-treated NIH-3T3 cells exogenously expressing human (and therefore siRNA insensitive) KDM2A displayed a significant decrease in major satellite RNA transcription compared to KDM2A siRNA-treated NIH-3T3 cells transfected with an empty vector (Fig. 4F). In contrast, KDM2A(ΔJmjC) was unable to rescue this effect due to KDM2A downregulation (Fig. 4F), suggesting a role for histone demethylation in the inhibition of transcription by KDM2A.
Dissecting the relationship between heterochromatin and centromeric RNA transcription is the basis of ongoing studies.25-27 Given the reactivation of GFP expression in GFP(-) HeLa reporter cells detected in KDM2A siRNA treated cells (Figure 2, E and F) and the increase in α-satellite transcription observed following KDM2A knockdown (Figure 3C), we sought to determine if forced accumulation of α-satellite RNA transcripts compromise the heterochromatic state. To that end, we transfected GFP(-) HeLa reporter cells with an expression vector driving transcription of one α-satellite repeat (171-bp monomer) from a human chromosome (chromosome 4) and monitored reactivation of GFP expression (Fig. 4G). Remarkably, we found that cells exogenously expressing these centromeric RNAs exhibited reactivation of GFP compared to cells transfected with an empty vector (Figure 4G), suggesting that the accumulation of α-satellite transcripts observed KDM2A depleted cells (Figure 3C) may be responsible of the heterochromatin disruption due to the downregulation of KDM2A (Figure 2, E and F).
RNA transcripts generated from within pericentric heterochromatin are thought to assist in the maintenance of centromere structure,26 and interactions with single-stranded RNA molecules are required for integrity of the kinetochore structure in mitosis.28 Importantly, while RNA molecules are required for centromeric integrity, the accumulation of satellite transcripts may lead to defects in chromosome segregation and mislocalization of centromere-associated proteins essential for centromere function.6 Based on these reports, we determined whether silencing of KDM2A impacted centromere integrity and chromosome segregation in mitosis. We analyzed HCT-116 cells, which have a relatively stable karyotype,29 transfected with control or KDM2A siRNA oligos and conducted indirect immunofluorescence analysis of CENP-A and α-tubulin to determine the state of centromeres and the mitotic spindle, respectively. While the mitotic spindle appeared unaffected, knockdown of KDM2A resulted in an increase in the misalignment of centromeres along the metaphase plate (36%, 45/125), as visualized with CENP-A, compared to control (12%, 15/125) (Fig. 4A). This phenotype was accompanied by segregation defects that were revealed by the detection of chromosome breaks (Fig. 4A) and a dramatic increase in the amount of chromosome bridges in KDM2A-siRNA treated cells (51%, 89/174) compared to control cells (16%, 38/228) (Fig. 4B). Together, these data suggest that KDM2A-mediated regulation of satellite RNAs and maintenance of heterochromatin may result in the loss of mitotic fidelity.
The loss of centromeric integrity, minor changes in the expression of non-coding RNAs and the deregulation of heterochromatin contribute to cellular transformation and genomic instability.6, 30, 31 As such, a recent study has demonstrated that HP1 variants display decreased expression levels in prostate cancer cell lines and prostate cancer tissues compared to normal tissues.32 In addition, an elevated frequency of chromosomal rearrangements is observed in the centromeric regions of prostate cancer patients,33 which is consistent with the loss of centromeric integrity associated with decreased levels of HP1 and heterochromatin stability. To investigate a potential involvement of KDM2A in prostate cancer, we searched the Oncomine on-line databases for differential KDM2A expression in normal versus tumorigenic prostate tissues. Interestingly, in 4 out of 4 studies, the expression of KDM2A, but not KDM2B, was found significantly decreased in prostate carcinomas compared to normal prostate tissue (Fig. 4C; Fig. S3 and S4). Furthermore, a fifth study showed that the decrease in KDM2A expression was correlated with prostate tumor grade (Fig. S2). Taken together, these results suggest that the decrease in KDM2A levels may contribute to the high incidences of centromeric rearrangements and mitotic aberrations observed in prostate carcinomas.
Abnormal chromosome segregation is a common characteristic of human tumors, and many of the molecular origins of chromosome missegregation derive from defective centromere and kinetochore assembly and function.31 While mutations and/or the altered expression of genes that impinge on kinetochore function may impact chromosome segregation, recent studies have focused on characterizing the function of non-coding RNAs derived from centromeric repeats in the formation and maintenance of heterochromatin in supporting proper centromere function.
Here we show that KDM2A localizes to heterochromatin and binds HP1 proteins. As HP1 interactors provide various possibilities for HP1-binding to sites that may not have intrinsically strong affinity for HP1, the association of KDM2A and HP1 may contribute to the stability of HP1 on chromatin that displays specific epigenetic marks on DNA or nucleosomes. Moreover, we show KDM2A is essential for maintaining the heterochromatic state. We also show that KDM2A provides a fundamental function for silencing the transcription of satellite RNAs that are embedded within centromeric repeats in a JmjC-domain dependent manner. Implicating the regulation of satellite RNA expression in heterochromatin maintenance, we demonstrate that forced accumulation of RNA transcripts encoded by a human α-satellite repeat compromises heterochromatin stability.
Intriguingly, the molecular function of KDM2A uniquely parallels that of the human paralogue KDM2B, which localizes to a particular cellular compartment (the nucleolus) and binds DNA repeats (ribosomal) to modulate expression of ribosomal RNAs.11 Equally noteworthy is that the fission yeast orthologue of KDM2A (Epe1) interacts with the S. pombe HP1 orthologue (Swi6) and plays a central role in maintaining heterochromatin and protecting centromere function and chromosome segregation fidelity.7, 12, 20, 21 Since a natural mutation in the catalytic region of the JmjC-domain of Epe1 is considered to render the protein a catalytically inactive histone demethylase,34 the requirement for the JmjC-domain in Epe1 function is currently not entirely understood. As the disruption of epigenetic control mechanisms contribute to cellular transformation, our study provides critical insight into the role of aberrant transcription of centromeric satellite repeats in heterochromatin maintenance for fidelity in mitosis.
Total RNA was extracted from NIH-3T3 the use of the RNeasy kit (Qiagen), and reverse transcriptions were performed as described previously.11 RT-PCR procedures and the primers used for α-satellites and major and minor satellites were as described previously.3, 28
Cells were transfected by using HiPerFect transfection reagent (Qiagen), in accordance with the manufacturer's protocol, with pool of four siRNA oligonucleotides (5'-CAACAGCGAUCCCAAGUUA-3', 5'-GGACACGGGUUCCAAAUAA-3', 5'-CAACAUCCCUAUGCAGUUA-3' and 5'-GCAAGCAGAUCACUCGAAA-3') targeting human KDM2A (Dharmacon) and with pool of four siRNA oligonucleotides (5'-CAAGAAUUCUGAUGGACUU-3', 5'-GAGAUUGCGUGGUACCAUG-3', 5'-GAGGUACUUCUGUUUGGUA-3' and 5'-GCACAAUGGACACGAUACU-3') targeting mouse KDM2A (Dharmacon). siRNA oligonucleotides used for KDM2B were previously described11. The siRNA oligonucleotide sequences used in the GFP(-) screen were as follows: HDAC1 (5'-CUAAUGAGCUUCCAUACAAUU-3'), KDM2A siRNA #1 (5'-ACCCAUUUCGUUGCUACCCAA-3'), KDM2A siRNA #2 (5'-CUCACUGGAGUUCCUAUAGUA-3'), KDM2B siRNA #1 (5'UAGGAGUGGACUAGAAGUUUA-3'), KDM2B siRNA #2 (5'-AGGAGTGGACTAGAAGTTTAA-3'). A 21-nucleotide siRNA duplex corresponding to a non-relevant gene (lacZ) was used as control.
In collaboration with Invitrogen, a polyclonal antibody against KDM2A was generated by immunizing rabbits with a peptide containing amino-acid residues 747-760 of human KDM2A. A polyclonal antibody against KDM2B was generated and previously characterized 11. Rabbit polyclonal antibodies were as follows: anti-FLAG (F7425; Sigma), anti-di-methyl H3K36 (07-274; Upstate), anti-tri-methyl H3K9 (07-523; Upstate), anti-Alexa Fluor 568 (A11036; Molecular Probes) and anti-phospho-H3 (Ser 10) (06-570; Upstate). Mouse monoclonal antibodies were as follows: anti-M2 FLAG (F3165; Sigma), anti-HA (Covance), anti-HP1-α (MAB3584, Chemicon), anti-HP1-β (07-333, Upstate), anti-HP1-γ (MAB3450, Chemicon), anti-α-tubulin (32-2500; Zymed), anti-KDM2B (H00084678-M09; Abnova) and anti-Alexa Fluor 488 (A21121; Molecular Probes).
Gene expression data was collected and processed from the Oncomine website (www.oncomine.org) as previously described11 for five studies profiling gene expression of KDM2A or KDM2B in prostate cancer.36-40
Chromatin immunoprecipitation (ChIP) was conducted as previously described.11 Primers for the pericentric region of human chromosome 4 were previously described.41 HpaII treatment was conducted as previously described 42.
KDM2A mutants (KDM2A(CXXC); (C571A,C574A,C577A) and KDM2A(JmjC); (Δ148-316)) were generated using the QuickChange Site-directed Mutagenesis kit (Stratagene). To express α-satellite transcripts, an α-satellite repeat (171-bp monomer) from human chromosome 443 was subcloned into the multiple cloning site (MCS) of pECFP-N1 (Clontech) with previously used primers specific for that locus.28 A stop codon was added immediately following the 171-bp insert to prevent the generation of a pECFP-fusion protein.
Indirect immunofluorescence procedures were as described previously.11 Images were acquired with a Nikon Eclipse E800 fluorescence deconvolution microscope.
The PSC colocalization plug-in (http://www.cpib.ac.uk/~afrench/coloc.html) for ImageJ software (NIH) was used to determine Pearson (rp) and Spearman correlation coefficients (rs) as described previously.13 Images were prepared in Adobe Photoshop CS3 and ImageJ.
We thank N. Tenese for suggestions and critically reading the manuscript; F. Bassermann for helping with FACS analysis; and N. Tenese, S. Smith and L. Wong for reagents. D.F. is grateful to A. Nans. M.P. is grateful to T. M. Thor for continuous support. This work was supported by grants from the National Institutes of Health to M.P. M.P. is an Investigator with the Howard Hughes Medical Institute. The authors declare no competing financial interests.