KDM4C (GASC1) lysine demethylase is associated with mitotic chromatin and regulates chromosome segregation during mitosis

Various types of human cancers exhibit amplification or deletion of KDM4A-D members, which selectively demethylate H3K9 and H3K36, thus implicating their activity in promoting carcinogenesis. On this basis, it was hypothesized that dysregulated expression of KDM4A-D family promotes chromosomal instabilities by largely unknown mechanisms. Here, we show that unlike KDM4A-B, KDM4C is associated with chromatin during mitosis. This association is accompanied by a decrease in the mitotic levels of H3K9me3. We also show that the C-terminal region, containing the Tudor domains of KDM4C, is essential for its association with mitotic chromatin. More specifically, we show that R919 residue on the proximal Tudor domain of KDM4C is critical for its association with chromatin during mitosis. Interestingly, we demonstrate that depletion or overexpression of KDM4C, but not KDM4B, leads to over 3-fold increase in the frequency of abnormal mitotic cells showing either misaligned chromosomes at metaphase, anaphase–telophase lagging chromosomes or anaphase–telophase bridges. Furthermore, overexpression of KDM4C demethylase-dead mutant has no detectable effect on mitotic chromosome segregation. Altogether, our findings implicate KDM4C demethylase activity in regulating the fidelity of mitotic chromosome segregation by a yet unknown mechanism.


INTRODUCTION
Nucleus of the eukaryotic cells is composed of DNA and proteins that are organized in higher-order structure termed chromatin (1). One main function of chromatin is to allow sophisticated packaging of the DNA into the eukary-otic nucleus. The second function is to provide a dynamic platform that regulates the execution of diverse processes such as replication, gene expression, DNA repair and recombination (2)(3)(4)(5)(6). Dysregulation of these processes or mutations affecting chromatin-remodeling complexes has been linked to many multi-system disorders and cancer development (3,7,8).
Two families of lysine demethylases (KDM) have been identified, confirming that histone methylation is a reversible and dynamically regulated process (32)(33)(34). One bryonic stem cells (59), and male lifespan in Drosophila (60). Interestingly, various types of human cancer show misregulated expression of KDM4A-D members (61)(62)(63). For example, KDM4C (also known as GASC1) is amplified in esophageal squamous carcinomas, medulloblastomas and breast cancer. The depletion of KDM4C inhibits proliferation of several tumor cell lines, while overexpression of the protein induces transformed phenotypes in mammary epithelial cells (24,50,(64)(65)(66)(67)(68)(69). Likewise, KDM4B overexpression promotes gastric tumorigenesis and its depletion leads to cell-cycle arrest and apoptosis of gastric cancer cell lines. Additionally, KDM4B depletion causes a decrease in colony formation and cell proliferation of estrogen-receptor-positive breast cancer cells and impairs therefore the development of normal breast tissues in vivo (70)(71)(72). Altogether, these observations suggest that overactivity of KDM4 proteins have a causative role in tumorigenesis, and therefore understanding the mechanisms by which KDM4 proteins promote chromosomal instability becomes critical.
Here, we determined KDM4A-C subcellular localization during mitosis. We reveal a previously unrecognized differential mitotic localization of KDM4A-C members.
While KDM4C is associated with mitotic chromatin, KDM4A-B proteins are excluded from chromatin throughout prometaphase-telophase. Also, we show that KDM4C Tudor domains are essential for its association with mitotic chromatin, thus characterizing a novel function of the Tudor domains in recruiting protein to mitotic chromatin. In addition, we demonstrate that dysregulation of KDM4C demethylase activity, but not KDM4B, promotes chromosome instability (CIN) by impairing the fidelity of mitotic chromosome segregation. These observations suggest that CIN, found in cancer cells driven by KDM4C dysregulation, could result from mitotic chromosome missegregation.

Plasmid construction and domain swapping
The constructions of all plasmids used in this study are described in Table 1. All point mutations were introduced using the QuickChange site-directed mutagenesis kit (Stratagene). A complete list of all primers and their sequences is described in Table 2. Domain swapping was performed as described in Table 1. All constructs used in this study were verified by nucleotide sequencing or restriction digestion.

Generation of stable cell lines
U2OS-TetON stable cell lines that conditionally express the fusions EGFP-KDM4B and EGFP-KDM4C were established as previously described (73). Cell line expressing EGFP-KDM4A was generated as follows. A fragment including EGFP-KDM4A was subcloned into pTRE2-puro (Clontech). The resulting pTRE2-puro-Hs-KDM4A vector was transfected into U2OS-Tet-ON cells (Clontech). Puromycin-resistant clones (0.6 g/ml Puromycin) were selected and tested for doxycycline-induced expression of EGFP using fluorescence microscopy. Clones that showed EGFP expression only after the addition of 1 M doxycycline (Sigma, D9891) were selected for further characterizations.

Transient transfection
Cell transfections with plasmid DNA or siRNA were performed using Poly Jet (Bio-Consult) and Lipofectamine 2000 (Invitrogen), respectively, following the manufacturer's instructions. siRNAs used in this study include Stealth KDM4B-C siRNA (Invitrogen) and Stealth RNAi negative control. All constructs and siRNA sequences are available upon request.

Immunofluorescence and Microscopy
Cells were grown on coverslips for 24-48 h before fixation and then washed twice with PBSX1, fixed with 4% paraformaldehyde for 10 min, permeabilized with 0.15% Triton-X-100 and 0.15% Tween-20 in PBSx1 for 10 min, blocked with 3% BSA, 0.2% Tween-20 and 0.2% Triton-X-100 for 1 h at RT, stained with the appropriate primary antibodies (see Table 3 for a complete list of all antibodies used in this study) for 3 h at 37 • C, washed three times with wash buffer (0.2% Tween-20 and 0.2% Triton-X-100 in PBSx1), stained with AlexaFluor488, Alex-aFluor568 (Molecular Probes; 1:500) or DyLight 649 (Jackson ImmunoResearch; 1:500) secondary antibodies for 1 h at RT in dark and then washed as above. Slides were then mounted using VECTASHIELD mounting medium with DAPI (VECTOR) and photographed using an inverted microscope Confocal Zeiss LSM 700 with 40X oil EC Plan Neofluar objective.

KDM4C, but not KDM4A-B, protein is associated with mitotic chromatin
Revealing the subcellular localization of proteins is vital for understanding their biological function(s). We sought therefore to determine the localization of KDM4A-C members, which share common domain architecture consisting of JmjN, JmjC, two PHD and two Tudor domains (Figure 1A). Toward this, we established U2OS-TetON cell lines expressing comparable protein levels of EGFP-KDM4A-C fusions upon the addition of doxycycline (Dox). As shown in Figure 1B, addition of Dox induces the expression of EGFP-KDM4A-C fusions, which leads to a severe reduction in H3K9me3 and H3K36me3 levels. On the other hand, the H3K4me3 levels were not affected by overexpression of KDM4A-C proteins.
To determine the localization of EGFP-KDM4A-C fusions during the different stages of mitosis, U2OS-TetON-EGFP-KDM4A-C cells were treated with Dox for 36 h, fixed and stained with DAPI to visualize mitotic cells. Results show that KDM4C protein is localized to mitotic chromatin from prometaphase to telophase. In striking contrast to KDM4C, EGFP-KDM4A-B fusions are excluded from mitotic chromatin ( Figure 1C). Interestingly, the levels of H3K9me3 on mitotic chromatin are severely reduced in cells overexpressing EGFP-KDM4C comparing to cell overexpressing EGFP-KDM4A fusion ( Figure 1D).
Next, to assess whether the mitotic localization of the endogenous KDM4A-C proteins is similar to their overexpressed EGFP-fused forms, we first tested the suitability of commercial KDM4A-C antibodies to detect the native forms of KDM4A-C proteins by immunofluorescence analysis. U2OS cells were transfected with expression constructs expressing EGFP-KDM4A-C fusions (green) and immunostained with KDM4A-C antibodies (red). Results show that the intensity of the red signal in cells expressing the EGFP-KDM4A-C fusions is much higher than the untransfected cells (Supplementary Figure S1). This result confirms that these antibodies detect KDM4A-C proteins in cells and can be used for immunofluorescencebased studies. MCF7 cells were then immunostained using KDM4A-C antibodies to detect their mitotic localization. Results show that, similar to the localization of EGFP-KDM4A-C fusions, the endogenous KDM4A-B are excluded from mitotic chromatin, while KDM4C protein is associated with chromatin during the different mitotic stages ( Figure 1E). Altogether, these observations demonstrate for the first time that, unlike KDM4A and KDM4B, KDM4C is associated with mitotic chromatin and triggers the demethylation of H3K9me3 mark.

The Tudor domains of KDM4C mediate its localization to mitotic chromatin
To map KDM4C region that mediates its localization to mitotic chromatin, we performed domain-swapping analysis between KDM4A and KDM4C proteins. First, we swapped the regions containing the two PHD and the two Tudor domains between KDM4A and KDM4C proteins. As a result, two chimeras were produced (Table 1): chimera1 encodes the first 708 amino acids of KDM4C protein, which includes the JmjN and JmjC domain, fused to the last 357 amino acids of the KDM4A containing the two PHD and the two Tudor domains. Chimera2 is the reciprocal chimera, which encodes the first 714 amino acids of KDM4A protein and the last 361 amino acids of the KDM4C protein containing the two PHD and the two Tudor domains. U2OS cells were transfected with expression vectors encoding chimera1 and 2 and the mitotic localization was determined as described in Figure 1B. Results show that while chimera1 shows chromatin-excluded localization ( Figure  2A), chimera2 exhibits chromatin-bound localization during mitosis ( Figure 2B). Together, these results suggest that the C-terminal region containing the two PHD and the two show that the C-terminus of KDM4C, containing the two Tudor domains, is essential and sufficient for its association with mitotic chromatin. Panels (E)-(F) show that the distal Tudor domain is essential but not sufficient for the localization of KDM4C at mitotic chromatin. In panels (A)-(F), U2OS cells were transfected with expression constructs encoding the indicated chimeras fused to EGFP (green). DNA is stained with DAPI (blue). Results are typical of 2-3 different experiments and each image represents at least 10 different cells. Image acquisition and scoring were performed by a student who was blind of the experimental condition.
Tudor domains mediate the distinct mitotic localization of KDM4A and KDM4C proteins.
To identify the domain that mediates the association of KDM4C with mitotic chromatin, we repeated the domainswapping analysis and exchanged the regions containing only the two Tudor domains between KDM4A and KDM4C proteins. This analysis produced chimera3 and chimera4 (Table 1). Chimera3 encodes the first 865 amino acids of KDM4C protein and the last 178 amino acids of KDM4A that contain the two Tudor domains. Chimera4 encodes the first 885 amino acids of KDM4A protein and the last 190 of KDM4C protein. We observed that chimera3 shows chromatin-excluded localization ( Figure 2C) and chimera4 has chromatin-bound localization during mito-sis ( Figure 2D). These results show that the replacement of KDM4C Tudor domains with those of KDM4A protein leads to its exclusion from mitotic chromatin. In addition, KDM4C C-terminal region, consisting of the two Tudor domains, leads to the association of KDM4A with mitotic chromatin. Collectively, we concluded that the KDM4C Cterminus, containing the two Tudor domains, is essential and sufficient for its association with mitotic chromatin.
To determine whether both Tudor domains are required for KDM4C mitotic localization, we swapped the C-terminus region containing the distal Tudor domain between KDM4C and KDM4A (Table 1). Chimera5, which encodes the first 934 amino acids of KDM4C fused with the last 129 amino acid containing the distal Tudor domain of KDM4A, is excluded from mitotic chromatin (Figure 2E). On the other hand, chimera6 that encodes the first 954 amino acids of KDM4A fused to 101 amino acids of KDM4C, which includes its distal Tudor domain, remains excluded from chromatin ( Figure 2F). We concluded therefore that the C-terminus of KDM4C containing the distal Tudor domain is essential but not sufficient for its mitotic chromatin localization.

Mapping candidate residues within KDM4C Tudor domains that regulate its localization to mitotic chromatin
Domain-swapping analyses suggest that the localization of KDM4C at mitotic chromatin is mediated by its Tudor domains ( Figure 2). To map residues within the Tudor domains of KDM4C that regulate its association with mitotic chromatin, we performed sequence alignment of KDM4A-C proteins using MUSCLE software and searched for residues that are conserved between KDM4A and KDM4B proteins but not in KDM4C ( Figure 3A). Noticeably, comparison of the amino acid sequences shows that the proximal Tudor domain of KDM4A and KDM4B contains four identical amino acids, DNLY, which appear as RDTF in KDM4C protein (corresponds to 919-922 amino acids). On this basis, we speculated that these four residues might be implicated in regulating KDM4C localization at mitotic chromatin. Site-directed mutagenesis was used to substitute RDTF residues of KDM4C with DNLY, and the mitotic localization of KDM4C RDTF/DNLY mutant was determined. Results show that EGFP-KDM4C RDTF/DNLY mutant is excluded from mitotic chromatin ( Figure 3B). We concluded therefore that the RDTF residues are critical for KDM4C association with mitotic chromatin. Next, we sought to address whether RDTF residues are sufficient for KDM4C localization at mitotic chromatin. To do so, we substituted the DNLY (corresponds to 939-942 amino acids) of KDM4A with RDTF. Results show that, similar to the wild-type KDM4A, KDM4A DNLY/RDTF mutant remains excluded from mitotic chromatin ( Figure 3C). Altogether, these observations suggest that RDTF residues are required, but not sufficient, for the localization of KDM4C at mitotic chromatin.
Interestingly, it was recently shown that KDM4A D939 residue is essential for the binding of the KDM4A Tudor domain with methylated H4K20me2 (56). Therefore, we sought to address whether this residue is also involved in regulating KDM4A and KDM4C association with mitotic chromatin. Toward this, we generated KDM4C R919D . Results show that this mutant is excluded from mitotic chromatin ( Figure 3D). This observation suggests that, similar to the RDTF residues, KDM4C R919 residue is essential for KDM4C localization at mitotic chromatin. Collectively, our data strongly suggest that sequences at the two Tudor domains are required for the distinct localization of KDM4C at mitotic chromatin.

Dysregulation of KDM4C expression promotes mitotic chromosome missegregation
The localization of KDM4C on mitotic chromatin raises a possibility that it might be implicated in regulating chromosome segregation. To assess this possibility, we looked at four abnormal mitotic phenotypes in cells overexpressing or depleted of either KDM4B or KDM4C. The abnormal phenotypes include misaligned chromosomes during metaphase ( Figure 4A), lagging chromosomes, anaphasetelophase bridges ( Figure 4B and C) and multiple centrosomes ( Figure 4D). To deplete KDM4B-C, U2OS cells were transfected with KDM4B-C siRNA sequences (KDM4C siRNA #46, #58 and #59; KDM4B #06). Western blot reveals that all siRNA sequences targeting KDM4B-C show severe reduction in the protein levels compared to control siRNA ( Figure 4E).
KDM4C-depleted cells were subjected to immunofluorescence analysis using ␣-tubulin and ␣-Pericentrin antibodies and DNA was stained with DAPI to allow the identification of mitotic cells. Metaphase cells with misaligned chromosomes were counted for each KDM4C siRNA sequences and divided by the total number of metaphase cells (n = 448 total metaphases). Results show that KDM4C depletion using either one of the three siRNA sequences leads to over 3-fold increase in percentage of metaphase cells with misaligned chromosomes compared to cell transfected with control siRNA (n = 107 metaphase cells) (Figure 4F). Similar increase was also obtained in anaphasetelophase cells with lagging chromosomes or anaphasetelophase bridges (n = 272 anaphase-telophase cells) compared to control cells (n = 101) ( Figure 4G). Next, we sought to address whether KDM4C depletion affects centrosome number. Mitotic cells were analyzed based on ␣tubulin and Pericentrin staining. Results show no detectable changes in the percentage of mitotic cells with multiple spindle poles between control and KDM4C-depleted cells (multipolar spindle pole formation was present in 0.5±0.4% of KDM4C-depleted cells and 0.4±0.3% of control cells). Interestingly, KDM4B-depleted cells show no significant increase in the percentage of abnormal mitotic cells ( Figure  4F and G). Altogether, these observations show for the first time that depletion of KDM4C, but not KDM4B, affects the fidelity of mitotic chromosome segregation, therefore suggesting that CIN in cancers lacking KDM4C can result in part from mitotic chromosome missegregation.

The demethylase activity and the mitotic localization of KDM4C influence the integrity of mitotic chromosome segregation
To gain further insights into KDM4C mitotic function, we sought to address whether its demethylase activity is implicated in regulating chromosome segregation. Toward this, we generated KDM4C 'demethylase-dead' mutant. As we have previously reported (76), Ser198Met mutation, within the JmjC of KDM4C, is expected to abolish an existing hydrogen-bond network, disrupting the coordination of ␣-KG within the catalytic site and consequently abrogating the demethylase activity. Indeed, western blot and immunofluorescence analysis show that overexpression of KDM4C-S198M in U2OS cells has no effect on H3K9me3 levels, whereas overexpression of wild-type KDM4C leads to a sever decrease in the levels of H3K9me3 mark ( Figure  6A and B). Next, we looked at abnormal mitosis in cells overexpressing EGFP-KDM4C-S198M demethylase-dead mutant. Results show no detectable effect on the percentage of cells showing abnormal chromosome segregation ( Figure  6C and D). This observation suggests that dysregulation of KDM4C demethylase activity disrupt the fidelity of mitotic chromosome segregation.
Next, we sought to address whether the defective chromosome segregation in cells overexpressing KDM4C is due to its localization on mitotic chromatin. To this end, we overexpressed EGFP-KDM4C-R919D mutant, which is not associated with mitotic chromatin (Figure 3D), and determined the percentage of abnormal mitotic cells. Results show that EGFP-KDM4C-R919D overexpression has no significant increase in the percentage of metaphases with misaligned chromosomes ( Figure 6C). On the other hand, it led to 1.8-fold increase in the percentage of anaphasetelophase cells showing lagging chromosomes or anaphasetelophase bridge ( Figure 6D). Altogether, we made two interesting conclusions. The first is that the defective mitotic chromosome segregation in cells overexpressing or depleted of KDM4C is due to alteration in KDM4C demethylase activity. The second is that part of the abnormal mitotic phenotype (lagging chromosomes and anaphase-telophase bridges) is independent of its localization on mitotic chromatin.

DISCUSSION
This study shows a unique localization of KDM4C on mitotic chromatin that differs from the closely related members, KDM4A and KDM4B, which are excluded from chromatin during mitosis (Figure 1). This differential mitotic localization of KDM4A-C suggests that they might have distinct functions during mitosis. Similar to our observations, it was reported that the three different isoforms of the heterochromatin protein (HP1␣, HP1␤ and HP1␥ ), which share the same domain architecture, are localized differently during interphase and mitosis. HP1␣ is the only isoform that remains associated with chromatin during mitosis, while HP1␤ and HP1␥ showed different mitotic distribution compared to their nuclear localization during interphase (77,78). In agreement with this, HP1␣ isoform plays a central role in mitotic chromosome segregation by regulating Aurora B activation and dissociation from chromosome arms (79). Furthermore, protein phosphatase-1 (PP-1) isoforms ␣, ␥ 1 and ␦, which share nearly identical catalytic domains, localize differently during interphase and mitosis suggesting unique roles for each of the PP-1 isoforms during the different cell-cycle stages (80).
Our data suggest that the two Tudor domains are involved in the regulation of KDM4C localization during mitosis. Interestingly, the Tudor domain was shown to mediate the subcellular localization of proteins. For example, the Tudor domain of TDRD3 is both required and sufficient for its localization to stress granules, which are cytoplasmic structures involved in RNA metabolism (81). Similarly, Tudor domain-containing protein, Yb, which is required for the primary processing of piRNAs and transposon repression, localizes to a cytoplasmic structure called the Yb body via its Tudor domain (49). Also, the Tudor domain mediates the recruitment of various proteins to chromatin. For instance, it was found that foci formation of 53BP1 protein after DNA damage is mediated by the binding of its Tudor domain to H4K20me2 mark (56,82). Moreover, it was shown that KDM4A is guided by its Tudor domain to H3K4me3 and H4K20me3 regions to demethylate H3K9me3 and H3K36me3 methyl marks (83,84). Here, we further expand the function of the Tudor domain by characterizing a previously unrecognized role of the Tudor domain in recruiting KDM4C to mitotic chromatin. Future work will be required to identify the mechanism by which the Tudor domains regulate KDM4C association with chromatin during mitosis.
In addition, we reveal a novel role of KDM4C in regulating mitotic chromosome segregation. Our results show that the levels of KDM4C protein are critical for the proper chromosome segregation. KDM4C, but not KDM4B, knockdown or overexpression increases the frequency of abnormal mitotic cells showing misaligned chromosomes during metaphase, anaphase bridge and chromosome lagging. Further, overexpression of KDM4C-S198M demethylase-dead mutant has no detectable effect on the fidelity of chromosome segregation. These results imply that CIN can result from loss or gain of KDM4C demethylase activity. In accord with this, recent analysis of the Cancer Genome Atlas revealed that KDM4C is lost in some cancer types and overexpressed in others (62). It should be noted however that we cannot exclude the existence of additional yet unknown mechanisms that contribute to CIN found in cancer driven by either lack or overactivity of KDM4C lysine demethylase. In this regard, it was recently shown that KDM4A overexpression induces copy number gains of specific genomic regions which are known to contain oncogenes (62).
Determining the cellular localization of protein is often a crucial step toward understanding its biological functions (85). Our data showing that KDM4C is localized at mitotic chromatin and promotes chromosome segregation is in accordance with the mitotic localization pattern of other proteins, which are associated with mitotic chromatin and play a role in chromosome condensation and sister chromatid separation. For example, the Tudor-domain protein EKL-1 is localized at mitotic chromatin and promotes chromosome segregation (86). Likewise, Aurora B kinase, which is associated with mitotic chromosomes, is involved in chromosome segregation and cytokinesis (87,88).
How does KDM4C misregulation disrupt the fidelity of mitotic chromosome segregation? There are three main possibilities which are not necessarily mutually exclusive. First, KDM4C serves as a scaffold protein for recruit- that overexpression of EGFP-KDM4C-S198M has no detectable effect on the levels of H3K9me3. Protein extracts were prepared from U2OS-TetON cells expressing either EGFP-KDM4C-WT or EGFP-KDM4C-S198M and immunoblotted using the indicated antibodies. (B) Immunofluorescence analysis of U2OS-TetON expressing either EGFP-KDM4C-WT (bottom) or EGFP-KDM4C-S198M (top). Cells were stained for H3K9me3 (red). DNA is stained with DAPI (blue), and the EGFP-KDM4C is in green. (C) and (D) Histograms showing the percentage of metaphases with misaligned chromosomes (B) and anaphase-telophase cells that exhibit lagging chromosomes or anaphase-telophase bridges (C). U2OS-TetON cells expressing EGFP-KDM4C-S198M or EGFP-KDM4C-R919D were subjected to immunofluorescence and mitotic cells were acquired using confocal microscope. n, number of mitotic cells counted. Error bars represent standard deviation from two independent experiments. ing other proteins that are required for proper chromosome segregation. In accord with this, cells overexpressing KDM4C-R919D mutant (which does not localize to mitotic chromatin) show no increase in the percentage of metaphases with misaligned chromosome ( Figure 6C). Second, KDM4C regulates the activity of non-histone proteins, which are involved in regulation of chromosome segregation, through demethylating their lysine residues. In support of this, it was shown that the polycomb protein, Pc2, is a substrate of KDM4C. Interestingly, Pc2 is SUMO E3 ligase that promotes sumoylation of multiple proteins (89), a modification which is essential for proper chromosomes segregation (90). Third, by demethylating KDM4C histone substrates such as methylated H3K9 or H3K36 residues. In agreement with this, it was shown that the level of H3K9me3 mark decreases as cells exit mitosis, during the period between anaphase and cytokinesis. This decrease is essential for chromosome congression and segregation. It was also found that H3K9me3-deficient cells exhibit a wide range of abnormal mitotic phenotypes such as an increase in misaligned and lagging chromosomes, which leads to aneuploidy, nondisjunction and the appearance of micronuclei at cytokinesis or early G1 (91). Likewise, it was previously shown that loss of H3K9 methyltransferase, Suv39h, or overexpression of H3K9 demethylase KDM4B leads to CIN (18,92). Collectively, our data suggest that definite methylation levels of KDM4C substrates might be required to ensure proper chromosome segregation. Nonetheless, further studies will be required to address how alterations in KDM4C levels promote chromosome missegregation.