Lysine-specific Demethylase 2B (KDM2B)-let-7-Enhancer of Zester Homolog 2 (EZH2) Pathway Regulates Cell Cycle Progression and Senescence in Primary Cells*

Sustained expression of the histone demethylase, KDM2B (Ndy1/FBXL10/JHDM1B), bypasses cellular senescence in primary mouse embryonic fibroblasts (MEFs). Here, we show that KDM2B is a conserved regulator of lifespan in multiple primary cell types and defines a program in which this chromatin-modifying enzyme counteracts the senescence-associated down-regulation of the EZH2 histone methyltransferase. Senescence in MEFs epigenetically silences KDM2B and induces the tumor suppressor miRNAs let-7b and miR-101, which target EZH2. Forced expression of KDM2B promotes immortalization by silencing these miRNAs through locus-specific histone H3 K36me2 demethylation, leading to EZH2 up-regulation. Overexpression of let-7b down-regulates EZH2, induces premature senescence, and counteracts immortalization of MEFs driven by KDM2B. The KDM2B-let-7-EZH2 pathway also contributes to the proliferation of immortal Ink4a/Arf null fibroblasts suggesting that, beyond its anti-senescence role in primary cells, this histone-modifying enzyme functions more broadly in the regulation of cellular proliferation.

KDM2B encodes a Jumonji domain family histone H3K36me2 and H3K4me3 demethylase that was identified as a common insertion site in rodent lymphoma insertional mutagenesis screens (1,2), and its ectopic expression is sufficient to transform hematopoietic progenitors (3). The functions of endogenous KDM2B are most clearly defined in the control of the lifespan of mouse embryonic fibroblasts (MEFs). 3 Levels of KDM2B decline upon serial passage of MEFs, and shRNA-mediated depletion of KDM2B results in premature senescence (1,2,4). Reciprocally, KDM2B overexpression immortalizes MEFs in a Jumonji-dependent manner (1,2). This is associated with a reduction in the senescence-associated upregulation of p16 Ink4a as well as suppression of p15 Ink4b (1,2,4). Correspondingly, overexpressed KDM2B can bind directly to the Ink4b/Arf/Ink4a tumor suppressor locus (encoding p15 Ink4b , p19 Arf , and p16 Ink4a ) and demethylate the locus-associated histone H3K36me2 and H3K4me3 (2,4). KDM2B only modestly suppresses p19 Arf expression (2), and thus, because neither ablation of p16 Ink4a alone or in combination with p15 Ink4b results in immortal growth (5,6), KDM2B must have additional downstream mediators in the control of proliferation.
The PRC1 and PRC2 (polycomb repressive complexes 1 and 2) are additional candidate targets for mediating the effects of KDM2B on cellular lifespan. These complexes counteract senescence of primary fibroblasts, in part, through silencing the Ink4a/Arf locus (7,8). Endogenous KDM2B forms a physical complex with Polycomb group (PcG) proteins in both flies and mammals and can facilitate the PRC1-mediated ubiquitylation of H2A, which silences gene expression (2, 9 -11). KDM2B also modulates the activity of PRC2 by up-regulating EZH2, which mediates epigenetic gene silencing by trimethylating histone H3 at lysine 27 (2). Levels of EZH2 decline during passage of primary MEFs, whereas knockdown of EZH2 results in premature senescence, in part due to a loss of H3K27 trimethylation of the Ink4a/Arf locus, leading to reduced binding of PRC1 and consequent activation of the locus (7). The regulation of EZH2 in primary cells is incompletely understood, although evidence in cancer cell lines suggest potential roles for transcriptional regulation by the pRB-E2F pathway (12) and post-transcriptional regulation by tumor suppressor miRNAs (13)(14)(15)(16)(17)(18)(19). Notably, KDM2B can also positive regulate EZH2 levels through an undefined pathway (2). Overall, although KDM2B appears to modulate PRC1 and PRC2 function, the molecular mechanisms and specific contribution of these processes to growth control downstream of KDM2B in primary cells has not been determined.
Here, we sought to elucidate the functional relationship between KDM2B and EZH2 in primary cells. We show that KDM2B and EZH2 are coordinately down-regulated in a series of primary cell types undergoing senescence. Moreover, we find that up-regulation of EZH2 is a critical component of KDM2Bdependent control of replicative mortality. This is mediated in part through the direct repression of miRNAs let-7b and miR-101, which are up-regulated during cellular aging and present a barrier to cellular immortalization by down-regulating EZH2 and its transcriptional activator E2F2. Importantly, this pathway is also required for sustained proliferation of immortal Ink4a/Arf null fibroblasts suggesting a broad role of this histone modifying enzyme in cell cycle progression beyond its antisenescence function in primary cells.

EXPERIMENTAL PROCEDURES
Cell Culture-MEFs were isolated from E13.5 C57BL/6 mouse embryos as described previously (2). For the isolation of murine mouse mesenchymal stem cells, bone marrow cells were collected from 6 -8-week-old C57BL/6 mice by crushing femurs and tibias. Nucleated cells were counted using a hemocytometer and seeded in 75 cm 2 flasks at a density of 1 ϫ 10 6 cells/cm 2 with complete medium consisting of high glucose DMEM, 10% (v/v) fetal bovine serum (Hyclone), and 1% penicillin/streptomycin. The non-adherent cell population was removed after 72 h, and fresh medium was added in the cultures. Thereafter, the medium was changed every three to 5 days for ϳ4 weeks; when 70 -80% confluent, adherent cells were harvested with trypsin-EDTA (Sigma) at 37°C for 5 min and replated at 1000 cells/cm 2 . The human cell lines IMR90 (CCL-186), HEK293T (CRL-11268), BJ human skin fibroblasts (CRL-2522) were bought from the American Type Culture Collection and cultured in DMEM supplemented with 10% (v/v) fetal bovine serum, penicillin, and streptomycin.
Quantitative Real-time PCR-RNA was isolated with the RNeasy mini kit (no. 74104; Qiagen). cDNAs were synthesized with QuantiTect Reverse Transcription kit (no. 205310; Qiagen). PCR reactions were carried out in duplicate with FastStart Universal SYBR Green (no. 04913850001; Roche Applied Science) in a Stratagene MX3005P continuous fluorescence detector. The primer sets are listed in the supplemental table.
ChIP-MEFs from two 20-cm dishes were fixed with 1% formaldehyde for 20 min, followed by two washes with PBS. Cells were lysed in 500 l of SDS lysis buffer (50 mM Tris-HCl (pH 8.0), 1% SDS, 150 mM NaCl, and 5 mM EDTA) plus protease inhibitor mixture I from Roche Applied Science (no. 11836170001) and were incubated on ice for 10 min. Lysates were sonicated in a Bioruptor sonicator three times, 8-min
Cell Proliferation-Cells were plated in duplicates in 12-well plates at concentrations as indicated in the corresponding figure legends and passaged in 1:2 or 1:3 ratio. Cells were counted with an automatic cell counter (Countess, Automated Cell Counter; Invitrogen) according to the instructions of the manufacturer.
Histone Methylation Maps-Chromatin coordinates for the H3K4me3, H3K27me3, and H3K36me3 methylation patterns at the genomic loci of mmu-let-7b and mmu-mir-101a in MEFs were computed through the Epigenomics Initiative supported by the Broad Institute and visualizd through the UCSC Genome Browser.
Cell Cycle Analysis-Logarithmically growing cells were washed twice with PBS, fixed overnight with methanol, and stained with propidium iodine (50 g/ml) in PBS containing 0.05% Nonidet P-40, and RNase A (0.5 mg/ml). After incubation for 30 min at 37°C, the DNA content of each cell cycle phase was determined by FACS with a FACSCalibur (BD Biosciences) flow cytometer. Raw data were analyzed with the FlowJo software by applying the Dean-Jett-Fox algorithm that fits G 0 /G 1 and G 2 /M phases with Gaussian curves and S phase with a second-degree polynomial curve.
Bioinformatics Analysis-Raw data for GSE9520 (24), GSE11954 (25), and GSE15161 (26) were downloaded from the Gene Expression Omnibus and analyzed with the dChIP analyzer. The Invariant Set Normalization method was used to normalize arrays, and the model-based method was used for probe selection and computing expression values (27). Background subtraction was performed with the mismatch probe (PM/MM difference) algorithm. Clustering of miRNAs (GSE9664, (28)) took place with the Cluster 3.0 and visualized with the JavaTreeView software. The miRNA TargetScan database (version 4.2) was used to identify EZH2 targeting miRNA. Statistical analysis took place with the GraphPad PRISM (version 5.01).

KDM2B Is Conserved Regulator of EZH2 Activity and of Onset of Senescence in Primary Cells-We have shown previously that
KDM2B is the only Jumonji domain containing histone demethylase that is down-regulated in MEFs undergoing senescence (2). Fig. 1, A-C, extends these findings, showing that the down-regulation of KDM2B is a hallmark of senescence in multiple types of primary mouse and human cells and that senescence is accompanied by a concomitant down-regulation of EZH2. To delineate the potential role of epigenetic changes in the silencing of the KDM2B upon cellular aging, we performed ChIP experiments in early and late passage MEFs. We found that cell passage induces a repressive epigenetic signature at the KDM2B promoter characterized by increased tri-methylation of H3K27, a repressive histone mark, and reduced trimethylation of H3K4, a permissive histone mark (Fig. 1D). Thus, cellular aging epigenetically silences KDM2B, and its down-regulation may be a prerequisite for multiple cell types to undergo senescence. Indeed, as in MEFs, knockdown of KDM2B accelerated the onset of senescence across the panel of primary cell lines ( Fig. 1E and supplemental Fig. S1A), whereas ectopic expression of KDM2B accelerated cell cycle progression and led to lifespan extension in a Jumonji domain-dependent manner (supplemental Fig. S1, B and C) (1,2) .
Because KDM2B and EZH2 are coordinately down-regulated during senescence, and because KDM2B knockdown and overexpression studies show that KDM2B regulates EZH2 levels (Fig. 1, F and G, and supplemental Fig. S1, D and E), we performed genetic complementation studies to evaluate the functional interactions between KDM2B and PcG proteins in the senescence program. Knockdown of EZH2 and BMI1 counteracted the capacity of KDM2B to promote immortalization of MEFs and to suppress p16 Ink4a expression and consequent upregulation of retinoblastoma-Ser 807/811 phosphorylation (Fig.  1H). Concurrent EZH2 overexpression did not enhance the effects of KDM2B on proliferation and p16 Ink4a , suggesting that KDM2B overexpression maximally activates EZH2, whereas BMI1 and KDM2B acted synergistically (Fig. 1I). Reciprocally, premature senescence caused by KDM2B knockdown was partially reverted by BMI1 overexpression, whereas EZH2 overexpression had only modest effects upon KDM2B depletion ( Fig.  1J and data not shown). We conclude that cross talk between KDM2B and PcG proteins is central to lifespan regulation in primary cells. It is important to note that unlike KDM2B or BMI1, the overexpression of EZH2 did not result in immortalization of MEFs (Fig. 1I) (12)). Therefore, although EZH2 function is absolutely required for KDM2B-mediated immortal growth, it cannot independently override additional control mechanisms (12). Collectively, these data indicate that KDM2B and BMI1 function cooperatively as master regulators of immortalization and that KDM2B acts upstream of BMI1 through the induction of EZH2 as well as through additional pathways.
KDM2B Counteracts Senescence-associated Up-regulation of miRNAs That Target EZH2-Next, we sought to explore what triggers the down-regulation of EZH2 in aging cells and how this process is counteracted by KDM2B. Because the epigenetic alterations induced by KDM2B relate to transcriptional repression (2,4,9,10,29), the role of this factor in EZH2 induction may be indirect. A number of miRNAs that regulate EZH2 abundance by targeting its 3Ј-UTR have been identified and are thus potential candidates for control of EZH2 expression during senescence; these include the let-7/98 family, miR-101, and miR-26 tumor suppressor miRNAs (supplemental Fig. S1F) (15,18,19). Notably, we found that let-7b, miR-101a, and to a lesser extent miR-26, are up-regulated in senescent MEFs as compared with early passage MEFs ( Fig. 2A). No significant increases in other members of the let-7/98 family were observed ( Fig. 2A and data not shown). Up-regulation of let-7b is also observed in serially passaged human mesenchymal stem cells (supplemental Fig. S1G), aging human skeletal muscle (30), and aging neural stem cells (31). Thus, the up-regulation of miRNAs, particularly let-7b, may contribute to the senescence program in primary cells.
KDM2B overexpression resulted in decreased levels of let-7b, whereas KDM2B knockdown caused up-regulation of this miRNA; miR-101 showed a similar profile but with less pronounced increases upon KDM2B depletion, and miR-26 was largely unaffected by KDM2B modulation (Fig. 2, B and C). KDM2B depletion minimally affected the expression of other members of the let-7/98 family, such as let-7a and let-7c (supplemental Fig. S1H). Importantly, let-7b overexpression induced premature senescence (Fig. 2D) and also counteracted the immortalization phenotype driven by KDM2B (Fig. 2E). Consistent with previous reports (32) overexpression of let-7b in MEFs reduced the content of cells in S phase and caused mitotic arrest (supplemental Fig. S1I), as also seen upon knockdown of either KDM2B or EZH2 (supplemental Fig. S1J). miR-101 and miR-26 had a smaller effect on the proliferation of MEFs compared with let-7b (supplemental Fig. S2A). Consistent with these findings, EZH2 was down-regulated by let-7b and less so by the other miRNAs (Fig. 2F). Significantly, the use of a let-7b antagonist reversed the effect of KDM2B depletion on EZH2 (supplemental Fig. 2G) suggesting that a KDM2Blet-7b axis regulates EZH2 levels in MEFs.
let-7b is an established tumor suppressor that inhibits a number of oncogenes, such as c-Myc, lin-28b, K-Ras, Hmga2 (33,34), as well as the E2F2 transcription factor (35). E2F2 levels decline during senescence (Fig. 2H), and in line with previous studies of let-7/98 targets (35), we found that E2F2 expression was also suppressed by let-7b (Fig. 2I). Importantly, overexpression of KDM2B potently increased the expression of E2F2 (Fig.  2J), as well as that of c-Myc, an additional let-7b target (supplemental Fig. S2B). Notably, E2F2 transcriptionally activates EZH2 (12); thus, induction of let-7b in aging MEFs appears to contribute to down-regulation of EZH2, by targeting of both E2F2 (35) and the 3Ј-UTR EZH2 (19). Importantly, ectopic expression of BMI1 largely rescued the proliferation of MEFs stably expressing let-7b (Fig. 2K), suggesting that let-7b restrains the proliferation of MEFs by opposing the function of PcG proteins. Collectively, these data show that cellular aging up-regulates miRNAs that target EZH2 and deregulate the cell cycle machinery and that these miRNAs are components of the senescence program in MEFs.
KDM2B Epigenetically Represses Expression of let-7b and miR-101-Next, we sought to determine whether KDM2B directly regulates the expression of the EZH2-targeting miRNAs through binding and histone demethylation of their genomic loci. We first optimized ChIP for endogenous KDM2B using the Ink4b/Ink4a/Arf locus as a control. This locus had previously been shown to be bound by ectopically overexpressed KDM2B (2). Here, we show that endogenous KDM2B strongly bound the Ink4a promoter but not to Arf or Ink4b ( Fig.  3A; supplemental Fig. S2C shows specificity of the antibody) and ectopic KDM2B potently repressed the expression of p16 Ink4a , and to a lesser extent p19 Arf , in a Jumonji domain-dependent manner ( Fig. 1G and supplemental Fig. S2D). KDM2B overexpression did not affect the levels of p15 Ink4b (supplemental Fig. S2D). Consistently, KDM2B strongly repressed the expression of a luciferase reporter driven by the Ink4a promoter, whereas deletion of the CxxC (ZF-CxxC) domain, which binds nonmethylated CpG islands (36), completely reversed the effect of KDM2B on p16 Ink4a (Fig. 1G and supplemental Fig.  S2E).
We subsequently found that KDM2B binds the let-7b and miR-101a loci, but not to miR26a-1, consistent with the observed effects of KDM2B on the expression of these miRNAs (Fig. 3B, compare with Fig. 2, B and C). Computational analysis of histone methylation maps in MEFs revealed a common epigenetic signature for let-7b and miR-101a characterized by the  SEPTEMBER 23, 2011 • VOLUME 286 • NUMBER 38 presence of the H3K36me3 mark, suggesting active transcription by RNA polymerase II but lack of either H3K4me3 or H3K27me3 (supplemental Fig. S3A). We found that overexpression of KDM2B decreased the H3K36 methylation at both the let-7b and miR-101a loci in a JmjC domain-dependent manner (Fig. 3C). Reciprocally, knockdown of endogenous KDM2B in MEFs caused increased methylation of H3K36 at these regions (Fig. 3, D and E). Because H3K36 methylation positively correlates with the recruitment of RNA polymerase II, we conclude that KDM2B-mediated demethylation of H3K36 contributes to the silencing of let-7b and miR-101a.

KDM2B-let-7b Pathway Regulates Proliferation of Ink4a/Arf
Null Fibroblasts-Both let-7b and EZH2 have been implicated in growth control independent of their functions in senescence (12,32,37). To study whether KDM2B has a comparable role, we proceeded to overexpress and knock down KDM2B in immortal Ink4a/Arf null fibroblasts. Notably, unlike passaged wild-type MEFs, MEFs immortalized by inactivation of either Ink4a/Arf or p53 tumor suppressors or by overexpression of the c-myc oncogene, showed sustained high levels of KDM2B expression (Fig. 4A, supplemental Fig. S3B, and data not shown). Knockdown of KDM2B in Ink4a/Arf null and c-myc-overexpressing fibroblasts markedly reduced cell proliferation ( Fig. 4B and supplemental Fig. S3C) in association with let-7b up-regulation (Fig. 4C) and EZH2 down-regulation (Fig. 4D). Levels of p15 Ink4b were unaffected (supplemental Fig. S3D). Knockdown of BMI1 or EZH2 (Fig. 4B) and overexpression of let-7b or a dominant-negative KDM2B Jumonji deletion mutant (1, 2) strongly inhibited cell proliferation (Fig. 4E). Therefore, KDM2B function in growth control and EZH2 regulation extends beyond the bypass of senescence in primary cells through a mechanism associated with regulation of let-7b miRNA.

DISCUSSION
Here, we show that epigenetic silencing of KDM2B contributes to the senescence of multiple types of primary embryonic and adult cells, and we define a let-7b-EZH2 pathway as an important component of KDM2B-mediated proliferative control and bypass of the senescence checkpoint. Thus, our data show that KDM2B has a conserved function in replicative lifespan and define a novel mechanism for EZH2 regulation. As both EZH2 and KDM2B contribute to the silencing of the Ink4/ Arf locus by direct chromatin modifications, the KDM2B-let-7b-EZH2 circuit appears to act in a feed-forward manner to reinforce the senescence phenotype (Fig. 4F).
Genetic studies in Drosophila (11) and biochemical purification in mammalian cells (2,9,10) have demonstrated interplay between KDM2B and Polycomb-mediated gene repression. Our genetic complementation studies highlighted the role of such functional interactions in the growth control of primary cells. KDM2B is required to maintain EZH2 expression levels in the immortalization process, and knockdown of either EZH2 or BMI1 counteracts cellular immortalization and repression of p16 Ink4a driven by KDM2B. BMI1 synergizes with KDM2B in facilitating cell proliferation and p16 Ink4a repression, effects that are consistent with the predicted function of EZH2-catalyzed H3K27 trimethylation in facilitating PRC1 recruitment. Whereas KDM2B and BMI1 are master regulators of immortalization, overexpression of EZH2 alone does not immortalize MEFs. Moreover EZH2 is much weaker at rescuing cell growth upon KMD2B knockdown in comparison to BMI1. Thus, KDM2B appears to act upstream of both EZH2 and BMI1 in cellular immortalization and functions both by maintaining EZH2 expression as well as through additional pathways.
KDM2B regulates the Ink4/Arf locus both indirectly through EZH2 modulation and by direct epigenetic modifications. Our chromatin immunoprecipitation of the endogenous protein and gene expression studies in MEFs showed that KDM2B binds to and potently represses the Ink4a promoter, whereas Arf binding and repression are much weaker (Fig. 3A and supplemental Fig. S2, D and E). We did not detect binding of KDM2B at the Ink4b promoter, and we did not observe reduced FIGURE 4. KDM2B represses let-7b and regulates the proliferation of Ink4a/Arf null fibroblasts. A, control MEFs, c-Myc-infected MEFs, and Ink4a/Arf null fibroblasts were analyzed at passage 5 by Western blotting for the relative expression of KDM2B. B, cumulative cell numbers of serially passaged Ink4a/Arf null fibroblasts (seeded at a concentration of 1 ϫ 10 5 cells/well) infected with lentiviruses to knock down the indicated proteins. C and D, Ink4a/Arf null fibroblasts were infected with shControl and shKDM2B lentiviruses and analyzed by qRT-PCR for the relative levels of let-7b (C) and by Western blotting (D) for the indicated proteins. E, Ink4a/Arf null fibroblasts seeded at a seeded at a concentration of 1 ϫ 10 5 cells/well were engineered to overexpress wild-type or demethylase deficient KDM2B, or let-7b. The graph shows the cumulative cell number for the indicated period of time. F, KDM2B represses let-7b and miR-101 (␣) and the Rb pathway (b) and up-regulates EZH2. EV, empty vector.
Ink4b levels upon KDM2B overexpression. Although these results differ from a prior report describing Ink4b as a critical target for KDM2B in MEFs (4), they are consistent with recently published data (38) showing that knockdown of the long form of KDM2B in MEFs caused the up-regulation of p16 Ink4a , and to a lesser extent p19 Arf , whereas no changes were seen in the p15 Ink4b . It should be noted that both Ink4a and Ink4b are dispensable for MEF senescence (6); thus, there are additional critical targets of KDM2B (and EZH2) in cellular lifespan control.
Another important finding of this study is that KDM2B function in growth control and EZH2 regulation extends beyond the bypass of senescence in primary cells through a mechanism that involves the regulation of let-7b. let-7b is an established tumor suppressor that inhibits the function of a number of oncogenes in addition to EZH2, including c-Myc, lin-28b, K-Ras, Hmg2a, and E2F2 (31,(33)(34)(35). Thus, the capacity of overexpressed KDM2B to promote immortalization of wild-type MEFs supports a role in tumor initiation, whereas the critical role for KDM2B in proliferation and let-7b regulation in MEFs immortalized by Ink4a/Arf deletion or c-Myc expression suggests sustained function in the maintenance of established cancers.