c-Rel Regulates Ezh2 Expression in Activated Lymphocytes and Malignant Lymphoid Cells*

Background: The mechanisms by which the expression of Ezh2 is regulated in normal and malignant cells are poorly understood. Results: c-Rel recruited to the Ezh2 locus up-regulates Ezh2 expression in activated lymphocytes and malignant lymphoid cells. Conclusion: c-Rel is a critical regulator of Ezh2 expression in lymphocytes and malignant lymphoid cells. Significance: We provide a mechanistic basis for rational combinatorial therapy for Ezh2-expressing cancers. The polycomb group protein Ezh2 is a histone methyltransferase that modifies chromatin structure to alter gene expression during embryonic development, lymphocyte activation, and tumorigenesis. The mechanism by which Ezh2 expression is regulated is not well defined. In the current study, we report that c-Rel is a critical activator of Ezh2 transcription in lymphoid cells. In activated primary murine B and T cells, plus human leukemia and multiple myeloma cell lines, recruitment of c-Rel to the first intron of the Ezh2 locus promoted Ezh2 mRNA expression. This up-regulation was abolished in activated c-Rel-deficient lymphocytes and by c-Rel knockdown in Jurkat T cells. Treatment of malignant cells with the c-Rel inhibitor pentoxifylline not only reduced c-Rel nuclear translocation and Ezh2 expression, but also enhanced their sensitivity to the Ezh2-specific drug, GSK126 through increased growth inhibition and cell death. In summary, our demonstration that c-Rel regulates Ezh2 expression in lymphocytes and malignant lymphoid cells reveals a novel transcriptional network in transformed lymphoid cells expressing high levels of Ezh2 that provides a molecular justification for combinatorial drug therapy.

Polycomb group proteins are highly conserved regulatory factors that were first identified as repressors of hox genes in Drosophila melanogaster (1). Three polycomb group complexes, including polycomb repressive complexes 1 and 2 and pho-repressive complex, have been identified and well characterized (2)(3)(4)(5). The histone methyltransferase enhancer of zeste homolog 2 (Ezh2) is a core component of polycomb repressive complex 2 that represses gene activity via structural modification of chromatin and contributes to the epigenetic regulation of gene expression during development and tumorigenesis (2,(5)(6)(7)(8)(9)(10)(11)(12). Ezh2 overexpression and somatic mutations are known to associate with several types of aggressive human cancer, including prostate cancer, breast cancer, lymphoma, and leukemia (13)(14)(15). However, the functional importance of Ezh2 in immune responses is poorly understood.
Our previous studies showed that Ezh2 has critical roles in lymphopoiesis (16,17). When Ezh2 was inactivated in bone marrow stem cells, lymphocyte development was blocked at the pro-to pre-B cell and pro-to pre-T cell transitions. Defective distal V H to D H J H recombination was also identified in Ezh2deficient pro-B cells (16). Inactivation of Ezh2 in germinal center B cells also revealed an essential role of Ezh2 in germinal center formation, whereas overexpression of hyperactive Ezh2 in the presence of a Bcl2 transgene promoted lymphomagenesis (18,19). In the T-lineage cells, Ezh2-deficient double negative thymocytes failed to progress to the double positive stage in response to anti-CD3 stimulation in vivo, implying that Ezh2 is involved in TCR-mediated T cell signaling (17). Taken together, these data suggest an indispensable role for Ezh2 in normal lymphocyte development, maturation, and activation, as well as lymphoid transformation.
Because Ezh2 is crucial for various biological and pathogenic processes, its expression requires precise regulation (16,17,20). Although a large number of Ezh2 expression profiling studies have focused on identifying Ezh2-regulated genes (21)(22)(23), only a handful of molecules are reported to regulate Ezh2 expression at the transcriptional level (21, 24 -27). For instance, E2F1/2 and HIF-1␣ up-regulate Ezh2 mRNA expression by binding directly to the Ezh2 promoter via their respective response elements (21,24). c-Myc can induce Ezh2 expression indirectly by either down-regulating miR-26a expression, which targets the Ezh2 mRNA 3Ј-UTR and suppresses Ezh2 protein levels, or by activating the retinoblastoma protein-E2F pathway (26,28,29). Elk-1, a downstream effector of the MEK-ERK signaling cascade, directly contributes to the up-regulation of Ezh2 in breast cancer cells (27). In contrast, p53 represses Ezh2 expression either through direct binding to the Ezh2 promoter or by downregulating E2F via p21 WAF1 to reinforce p53-mediated G 2 /M arrest (25). Although the deregulation of these transcription factors in epithelial tumors provides a clear link between Ezh2 expression and epithelial cancer, it remains unclear whether these or other transcription factors are responsible for the high levels of Ezh2 in lymphomas and leukemias or the up-regulation of Ezh2 in mature mitogen-stimulated lymphocytes.
Here, we identified c-Rel as a positive transcriptional regulator of Ezh2 expression in activated primary murine lymphocytes and human malignant lymphoid cells, where c-Rel recruitment to the first intron of the murine and human Ezh2 loci promoted Ezh2 expression. Treatment with the c-Rel inhibitor pentoxifylline (PTX) 2 not only reduced Ezh2 expression but also reduced the survival of human leukemia/lymphoma cell lines, including enhancing their sensitivity to the Ezh2-specific inhibitor, GSK126. Our results demonstrating that c-Rel is critical for regulating Ezh2 expression in normal and malignant lymphoid cells also provide a mechanistic basis for rational combinatorial therapy to treat cancers that express high levels of Ezh2.
Chemicals-Pentoxifylline (Sigma-Aldrich) was dissolved in PBS following the manufacturer's instruction. GSK126 (Active Biochem) was dissolved in DMSO (Sigma-Aldrich) following the manufacturer's instructions.
Plasmids-To generate Ezh2 minimal promoter (MP) reporter construct, MP of Ezh2 (Ϫ1915/ϩ55) was PCR-amplified from the genomic DNA of A20 cell line and subcloned into the HindIII site of pGL3-basic vector (Promega). To create Ezh2 reporter constructs with its enhancer region, a 220-bp fragment (including the first 5 bp of Ezh2 exon 2) upstream of Ezh2 exon 2 was PCR-amplified and subcloned into HindIII-NcoI site of pGL3 enhancer vector (Promega). Subsequently, fragments containing Ezh2 MP with different lengths of Ezh2 intron one region were subcloned into MluI-HindIII site of pGL3-basic vector. Different lengths of identified conserved fragment upstream of the Ezh2 MP were fused with Ezh2 MP and MP1694 construct at MluI site. pGL3-control (Promega) and pGL3 enhancer were used as controls for the luciferase assay. Selected transcription factors were PCR-amplified from their cDNA clone and subcloned into pEGFP-N1 (2A) expression vector. pEGFP-N1 (2A) was modified from pEGFP-N1 vector (Clontech) by adding a 2A peptide sequence into the N terminus of EGFP. 2A peptide sequence enables the bicistronic expression of the cDNA and EGFP reporter gene (31). All plasmids were verified by restriction enzyme digestions and sequencing (Axil Scientific Pte Ltd.).
Semiquantitative RT-PCR and RT-qPCR Analysis-Total RNA was purified using TRIzol (Invitrogen) as recommended. Reverse transcription was performed using SuperScript III (Invitrogen) with random hexamer. cDNA was amplified with primers specifically for Ezh2 and HPRT: Ezh2 forward, 5Ј-AACA-CCAAACAGTGTCCATGCTAC-3Ј; Ezh2 reverse, 5Ј-CTAAG-GCAGCTGTTTCAGAGAG AA-3Ј; HPRT forward, 5Ј-GCTGG-TGAAAAGGACCTCT-3Ј; and HPRT reverse, 5Ј-CACAGGAC-TAGAACACCTGC-3Ј. The abundance of transcripts of the housekeeping gene HPRT was used as a loading control. Quantification of PCR product was done using image processing software, ImageJ (National Institute of Health). Ezh2 expression was also assessed by real time RT-PCR with TaqMan gene expression assay, Mm00468464_m1, which amplifies exon 19 and 20 boundary of Ezh2. Ezh2 expression was detected with the ABI StepOnePlus real time PCR system (all from Applied Biosystems). Fold changes in expression were determined by the 2 Ϫ⌬⌬Ct method. Alternatively, RT-qPCR for ChIP was analyzed using KAPA SYBR Fast ABI Prism 2ϫ qPCR master mix (KAPA Biosystems, Inc.) with 0.15 M of each forward and reverse primer. The data were normalized to the expression levels of the HPRT.
DNA Affinity Precipitation Assay-Jurkat cells were harvested, washed once with PBS and resuspended in buffer A (0.25 M sucrose, 10 mM HEPES, pH 7.9, 5 mM MgCl 2 , and 0.5% Nonidet P-40). Cells were lysed on ice, and the nuclei were pelleted. Pelleted nuclei were washed in buffer A and pelleted again. Washed nuclei were resuspended in buffer B150 (25 mM HEPES, pH 7.9, 20% glycerol, 1.5 mM MgCl 2 , 0.1 mM EDTA, 150 mM KCl, 0.5 mM PMSF, and 0.5 mM DTT) followed by buffer B450 (20 mM HEPES, pH 7.9, 20% glycerol, 1.5 mM MgCl 2 , 0.1 mM EDTA, 450 mM KCl, 0.5 mM PMSF, and 0.5 mM DTT) dropwise and incubated on a rotator for 40 min at 4°C. Nuclear extracts were purified at 14,000 ϫ g for 15 min at 4°C to pellet the debris, and the supernatant as the nuclear extract was collected. Buffer B0 (25 mM HEPES, pH 7.9, 20% glycerol, 1.5 mM MgCl 2 , 0.1 mM EDTA, 0.5 mM PMSF, and 0.5 mM DTT) was added into the purified nuclear extract to lower salt concentration down to 150 mM. DNA probes were prepared via the hybridization of synthesized single-stranded oligonucleotide DNAs. Equal amounts of oligonucleotides, in annealing buffer (10 mM Tris, pH 8.0, 50 mM NaCl, 1 mM EDTA) was prepared to a total volume of 20 l. DNA mixture was heated at 95°C for 4 min and cooled to 30°C slowly. Dynabeads were washed in 1ϫ B&W buffer (5 mM Tris-HCl, pH 7.5, 0.5 mM EDTA, and 1 M NaCl), and subsequently, 200 pmol of prepared DNA probe mixture was added to prewashed Dynabeads and incubated at room temperature for 30 min. The probe bound beads were washed once in 1ϫ B&W buffer and followed by buffer B150, twice. 60 g of Jurkat nuclear extract was added with probebound beads to a total volume of 80 l in buffer C (25 mM HEPES, pH 7.9, 20% glycerol, 1.5 mM MgCl 2 , 0.1 mM EDTA, 150 mM KCl, 1 mg/ml BSA, 0.5 mM PMSF, and 0.5 mM DTT), and incubated at room temperature for 1 h to facilitate protein-DNA binding interaction. Dynabeads were washed three times with buffer B150 and DNA-bound protein was detected by immunoblotting.
Electroporation-Jurkat T cells suspended in unsupplemented, prewarmed RPMI, with 300 nM of human si-Rel or nontargeting siRNA (Dharmacon) in 4 -mm cuvettes (Bio-Rad) were electroporated (Bio-Rad GenePusler). After electroporation, cells were transferred into culture flask with prewarmed complete Jurkat media without antibiotics. Lympholyte M treatment was performed at 0, 6, and 24 h post-transfection to FIGURE 1. Ezh2 expression is up-regulated in activated lymphocytes. Purified mature naïve B and T cells were stimulated with anti-IgM antibody (5 g/ml)/IL-4 (10 ng/ml) and plate-bound anti-CD3 (10 g/ml)/anti-CD28 (5 g/ml) antibodies, respectively. A, total RNAs isolated from indicated postactivation time points were reverse transcribed, and Ezh2 mRNA levels were determined by semiquantitative PCR. Three-fold serial dilution of the cDNA templates was used for semiquantitative PCR amplification. The housekeeping gene HPRT was chosen as a loading control. PCR products were confirmed by sequencing. The data shown are the means Ϯ S.D. of more than three independent experiments. B, real time RT-qPCR was conducted using TaqMan gene expression assays (ABI) and normalized against HPRT. C and D, B and T cell activations were controlled by assessing down-regulation of the resting cell marker, CD62L, and up-regulation of the activation markers, CD69 and CD86 for B cells and CD69 and CD44 for T cells, 24 h postactivation. The data shown in this figure are representative of three independent experiments.

TABLE 1 Predicted transcription factor binding sites in Ezh2 locus (؊1915/؉1694)
High scoring transcription factor (TF) binding sites identified by TFSEARCH and rVista2.0 (37, 38) were summarized. Sequence logo was generated with WebLogo3 (59) based on TFSEARCH (37) (AP1, c-Rel, Ikaros-2, Pbx1, and SP1) or Children's Hospital Informatics Program MAPPER position weight matrix database (Meis1a and DP1). HEK293T cells were transfected with the indicated constructs and analyzed for luciferase activity 48 h post-transfection. Nontransfected cells (NT), as well as pGL3 enhancer (En) and pGL3-Control (Control) transfected cells, were used as internal controls for the luciferase assay. The relative luciferase activities (RLAs) of the respective constructs were normalized to Renilla luciferase activity. Relative fold changes were calculated (Calc.) using the RLA of control. The data shown are the means Ϯ S.D. of three independent experiments. The difference between the indicated pairs was determined by two-tailed Student's t test with equal variance. **, p Ͻ 0.005; ***, p Ͻ 0.001). C, the correct splicing of each reporter construct was examined by RT-PCR with the indicated primer pairs (arrows). Properly spliced constructs generated a 168-bp band, except in the case of the MP construct, which gave rise to a 99-bp band. NOVEMBER 14, 2014 • VOLUME 289 • NUMBER 46 enrich live cells. RNA from transfected cells was isolated at 6 and 24 h and reverse transcribed. RT-qPCR was performed to analyze levels of the indicated transcript.

c-Rel Regulates Ezh2 Expression
Luciferase Assay-HEK293T cells were transiently transfected with luciferase reporter construct, transcription factor expression vectors and Renilla luciferase expression vector (pRL-TK) (Promega). Cells were harvested for assay 48 h posttransfection. Firefly luciferase activity was determined by a Dual-Glo luciferase assay system (Promega) and normalized against Renilla luciferase activity. All assays were done according to manufacturer's standard protocol and detected using Fluoroskan Ascent FL (Thermo Scientific). For luciferase assays performed using Jurkat T cells, Jurkat T cells were transiently transfected with luciferase reporter construct, transcription factor expression vectors, and Renilla luciferase expression vector using X-tremeGene HP DNA transfection reagent (Roche) according to the manufacturer's instructions.
Annexin V and Propidium Iodide Staining-Cells were harvest and incubated with FITC annexin V (Biolegend) and Pro-pidium Iodide (Sigma) for 15 min at room temperature. Cells were acquired using FACSCalibur instrument (BD Biosciences).
Cell Viability Assay-1 ϫ 10 3 (Jurkat) or 1 ϫ 10 4 (MM1S) cells were seeded onto each well of 96-well plates and treated with GSK126 and PTX at various concentrations for 6 days. Medium and drugs were replaced at day 3. Cell viability was assessed using MTT assay (Molecular Probes, Invitrogen). Briefly, MTT was added to culture medium and incubated at 37°C for 4 h. In living cells, MTT was reduced to purple insoluble formazan, which was then solubilized by acidified isopropanol. The absorbance at 570 nm was read on a spectrophotometer (Tecan System Inc.) with reference wavelength at 620 nm. IC 50 values of the selected compounds were defined as the drug concentration that reduced the absorbance by 50% and were determined graphically.

Identification of Regulatory Regions within Intron 1 of the Murine Ezh2
Gene-Ezh2 protein is reported to be up-regulated in various proliferative lymphomas and germinal center lymphoblasts (18,32). To determine whether Ezh2 up-regulation is controlled by a transcriptional mechanism, we first performed semiquantitative and quantitative RT-PCR analysis of cDNAs isolated from purified peripheral B and T cells 24 and 48 h after antigenic stimulation (Fig. 1, A and B). Our results showed that Ezh2 mRNA expression levels were up-regulated ϳ1.5-2-fold in both activated B and T cells. To confirm the   NOVEMBER 14, 2014 • VOLUME 289 • NUMBER 46 activation lymphocytes in our experiments, we determined the expression levels of several surface markers by FACS. As expected, activated lymphocytes down-regulated CD62L and up-regulated CD69, CD86 (B cells), or CD44 (T cells) after 24 h of stimulation (Fig. 1, C and D).

c-Rel Regulates Ezh2 Expression
Because Ezh2 mRNA expression was induced in activated lymphocytes, we performed a comparative analysis of conserved noncoding sequences (CNSs) using VISTA Browser (33-35) to identify potential regulatory regions in the Ezh2 gene. The MP of the murine Ezh2 locus (Ϫ1915 to ϩ55, relative to the transcription start site) was defined based on an alignment with the known human EZH2 minimal promoter (25). Several stretches of highly conserved CNSs along the Ezh2 locus were identified, but in the current study, we focused on the CNS that covers a 3-kb intron 1 fragment adjacent to exon 1 ( Fig. 2A). To identify the cis-regulatory elements located between positions Ϫ1915 and ϩ3051, we generated a series of firefly luciferase reporter constructs containing the Ezh2 minimal promoter, untranslated exon 1, and various lengths of intron 1 CNS. Because splicing signals are shown to enhance transcription D and E, c-Rel was recruited to the Ezh2 locus in primary B and T cells. Cells were stimulated, and activation was confirmed as described in Fig. 1. B (D) and T (E) cells were fixed and sonicated. Chromatin was immunoprecipitated with anti-c-Rel antibody, and purified DNAs were amplified by qPCR (left) or semiquantitative PCR (right). IgG antibody was used as a negative control. For qPCR, samples were normalized to input and expressed as fold enrichment compared with IgG controls. F, c-Rel associates with mouse and human site 4. DNA affinity precipitation assays using biotinylated either mouse wild-type site 4 (Mmu), deleted mouse site 4 (Mmu*), or human site 4 probes (Hsa) were performed. Binding proteins were resolved on SDS-PAGE, and c-Rel was detected by specific antibody. Probe sequences are indicated in the low panel. Boxes show the position of site 4 or remaining nucleotide residue after site 4 deletion in Mmu* probe. G and H, recruitment of c-Rel to the Ezh2 locus is correlated with acetylated histone H3. ChIP was performed using activated B (G) and T (H) cells as described above with antibody specific to acetylated histone H3 (H3Ac) and DNAs were amplified by qPCR. A-D on the x axis indicate the positions of PCR products that are described in C. Bar graphs shown in the figure are the means Ϯ S.D. of more than three independent experiments. The differences between the indicated pairs were significant as determined by two-tailed Student's t test with equal variance. **, p Ͻ 0.005; ***, p Ͻ 0.001. (36), we also introduced into our reporter constructs a 220-bp intron 1 fragment upstream of Ezh2 exon 2 that contains a splice acceptor and a branch site (Fig. 2B). It is unlikely that transcripts initiated within intron 1, because no significant luciferase activity was detected with the reporter constructs lacking both the MP and exon 1 (Fig. 2B: constructs ϩ220, ϩ838, and ϩ1694). In addition, the proper splicing of each reporter construct was confirmed by RT-PCR with a specific primer pair (Fig. 2C). Two active regions (A1 and A2) and one repressive (R1) region were identified in these reporter assays (Fig. 2B). Because all three regulatory regions fall between ϩ715 and ϩ1694, the MP/ϩ1694 reporter was used for subsequent screening to identify transcription factors that regulate Ezh2 expression.

c-Rel Regulates Ezh2 Expression
c-Rel Is a Transcriptional Activator Controlling Ezh2 Expression-To identify transcription factor binding sites in the Ezh2 minimal promoter and intron 1 fragment, the sequence was analyzed using TFSEARCH and rVista2.0 programs (37,38). High scoring binding sites of seven transcription factors, including AP1, c-Rel, DP1, Ikaros, Pbx1, Meis1a, and SP1 were identified by both prediction methods. Sequence logos, positions, and their respective scores are summarized in Table 1. To test the regulatory effects of the predicted transcription factors on Ezh2 expression, we co-expressed each factor with a luciferase reporter construct containing only the Ezh2 MP or both the Ezh2 MP and partial intron 1 (MP/ϩ1694). E2F and p53 were used as positive and negative controls, respectively (21,25). As shown in Fig. 3, we identified Ikaros and SP1 as potential negative regulators and c-Rel as the lone activator of Ezh2 expression in our assay.
c-Rel Binds to Intron 1 of the Ezh2 Locus-Because c-Rel induced the most significant changes in the expression of the Ezh2 reporter, we performed additional assays to validate its regulatory effect. Using in silico analysis, we revealed four c-Rel binding sites within the Ezh2 MP/ϩ1694 construct. Sites 1 and 2 are located in the minimal promoter region, and sites 3 and 4 are within intron 1 of the MP/ϩ838 construct (Fig. 4A). Because all predicted c-Rel binding sites are located within the MP/ϩ838 construct, we utilized this reporter for subsequent experiments. c-Rel not only controlled the expression of the Ezh2 reporter in a dose-dependent manner (Fig. 4B), but also promoted endogenous Ezh2 expression in HEK293T cells (Fig.  4C). Deletion of site 1 and/or 2 did not have any effect on the expression of the Ezh2 MP reporter (Fig. 4D), which is in agreement with the earlier screening showing that c-Rel did not enhance the luciferase activity of the MP reporter (Fig. 3). Surprisingly, the luciferase activity in cells expressing the MP/ϩ838 reporter lacking site 3 was increased. This was due to the generation of an artificial c-Rel binding site (CGGGGCT-(G/T)GT) upon site 3 deletion (Fig. 4E). The elevated luciferase activity was abolished when the artificial c-Rel binding site was removed (3*). In contrast, deletion of site 4 significantly reduced the luciferase activity of the MP/ϩ838 construct (Fig.   FIGURE 6. c-Rel is critical for Ezh2 up-regulation in activated lymphocytes. Purified murine T and B cells were stimulated as described in Fig. 1. Total RNAs isolated from resting or activated control (wt) and c-Rel-deficient (c-Rel Ϫ/Ϫ ) lymphocytes (24 h upon stimulation) were reverse transcribed, and c-Rel (A), Ezh2 (B), and RelA (C) mRNA levels were determined by qPCR and normalized against HPRT. The data shown in the figure are the means Ϯ S.D. of three independent experiments. The differences between the indicated pairs were significant as determined by two-tailed Student's t test with equal variance. *, p Ͻ 0.01; **, p Ͻ 0.005; ***, p Ͻ 0.001; NS, not significant (p Ͼ 0.01). 4E). Comparable results were obtained using Jurkat T cells, a lymphoid lineage cell line (Fig. 4F). Although site 4 is not well conserved between human and mouse loci, they are both pre-dicted to be c-Rel binding sites. To determine whether c-Rel could also regulate the expression of Ezh2 through the human site 4 sequence, we generated a humanized reporter construct (site 4 Hsa) by replacing the mouse site 4 and adjacent sequence with the corresponding human sequence in the mouse reporter MP/ϩ838 construct (site 4 Mmu). Similar c-Relinduced luciferase activities were observed with both constructs (Fig. 4G). Taken together, our data suggest that position ϩ779 to ϩ788 (site 4) of the Ezh2 locus is the only authentic c-Rel binding site.
c-Rel Regulates Ezh2 Expression in Activated Lymphocytes-Next, we sought to determine whether c-Rel regulates Ezh2 expression in activated B and T cells. Because the transcriptional activity of c-Rel, a member of the NF-B family, is regulated by nuclear translocation following stimulus-dependent IB degradation (39 -41), we examined the translocation of c-Rel in activated lymphocytes. Our results showed that the nuclear translocation of c-Rel upon antigenic stimulation preceded the up-regulation of Ezh2 expression (Fig. 5, A and B), suggesting an involvement of c-Rel in the up-regulation of Ezh2 during lymphocyte activation.
To determine whether c-Rel is recruited to the newly identified c-Rel binding site in the endogenous Ezh2 locus upon lymphocyte activation, we performed ChIP assays. The positions of the PCR products are indicated in Fig. 5C. Although c-Rel did not associate with the Ezh2 locus in either resting lymphocytes or after 4 h of stimulation, a significant increase in c-Rel binding to this site (position C) in B and T cells was detected 24 h after antigenic stimulation (Fig. 5, D and E). The binding of c-Rel to mouse and human site 4 was further validated by DNA affinity precipitation assays using either mouse (Mmu) or human probes (Hsa) but not a mutated mouse probe (Mmu*), which did not interact with c-Rel (Fig. 5F). Because c-Rel frequently associates with the co-activator, p300 (42), we also determined the histone H3 acetylation levels across the Ezh2 promoter and intron 1 region. As expected, the increase in c-Rel binding correlated well with elevated chromatin accessibility as determined by the enrichment of acetylated histone H3 (Fig. 5, G  and H).
To further validate the c-Rel regulation of Ezh2 expression, we examined Ezh2 expression in c-Rel-deficient B and T lymphocytes. We found that c-Rel was not required for basal Ezh2 expression in resting lymphocytes. However, in contrast to activated control B and T cells, antigen-stimulated c-Rel-deficient lymphocytes, which still undergo most aspects of activation including the up-regulation of cell surface markers (30,43), failed to up-regulate Ezh2 expression (Fig. 6). Collectively, these data demonstrate that c-Rel regulates the induction of Ezh2 expression in activated lymphocytes.
To determine whether Ezh2 expression can be regulated by a c-Rel inhibitor, we pretreated lymphocytes with phosphodiesterase inhibitor PTX, a phosphodiesterase inhibitor reported to block c-Rel activation (44 -46). As expected, the nuclear translocation of c-Rel in activated B and T cells was compromised by PTX treatment (Fig. 7, A-C), with c-Rel recruitment to the Ezh2 locus and Ezh2 up-regulation during lymphocyte activation abolished (Fig. 7, A-E). Importantly, the up-regulation of several activation markers was not affected by PTX treatment in B cells and was only slightly reduced in T cells (Fig. 7, F and  G). We also noted that PTX treatment of T cells also altered the kinetics and magnitudes of Akt and Erk activation (Fig. 7H). Collectively, these results suggest that PTX suppresses Ezh2 expression by inhibiting c-Rel nuclear translocation.
c-Rel Controls Ezh2 Expression in Leukemia and Multiple Myeloma Cell Lines-To determine whether c-Rel is responsible for high levels of Ezh2 expression in transformed lymphoid lineage cells, we analyzed the human leukemic Jurkat T cell line and the multiple myeloma cell line MM1S, both of which expressed high levels of wild-type Ezh2 (data not shown). When these cell lines were treated with PTX for 24 h, nuclear levels of c-Rel plus its recruitment to the Ezh2 locus were reduced and coincided with a drop in Ezh2 expression levels (Fig. 8, A and B). The PTX effect was diminished at 48 h, most likely because of the relatively short half-life of this compound (47). c-Rel knockdown in Jurkat T cells also resulted in decreased Ezh2 expression (Fig. 8C). High levels of Ezh2 expression in Jurkat and MM1S cells coincide with their relative resistance to the Ezh2specific inhibitor GSK126 (Fig. 8, D and E) when compared with other cancer cell lines (48). Given that PTX treatment significantly decreased Ezh2 expression levels, reduced cell viability, and induced G 1 phase cell cycle arrest (Fig. 8, F and G, and data not shown), we decided to determine whether a combination of GSK126 and PTX would result in enhanced growth inhibition and death of these cells (Fig. 8, F and H). As predicted, the dosage of GSK126 required to achieve significant growth inhibition and cell death was reduced dramatically with the addition of PTX. For example, in the presence of 300 g/ml PTX, the growth IC 50 of GSK126 in Jurkat cells was reduced more than 10-fold (from IC 50 ϭ 8982 nM with GSK126 alone to 643 nM in combination), which is comparable with GSK126 sensitive cancer cell lines (48) (Fig. 8D). This result suggests that c-Rel is a novel therapeutic target for cancers expressing high levels of Ezh2 and that c-Rel inhibition may complement treatment with Ezh2-specific inhibitors.

DISCUSSION
In the current study, we showed that c-Rel is a critical positive regulator of Ezh2 expression in activated primary murine lymphocytes and human leukemia/lymphoma cell lines. Our experiments demonstrated that c-Rel recruited to intron 1 of the Ezh2 locus promotes Ezh2 expression. c-Rel regulation of Ezh2 expression was further validated by the lack of Ezh2 upregulation in activated murine c-Rel-deficient lymphocytes and c-Rel knockdown Jurkat T cells. Collectively, our results indicate that c-Rel is the most important positive regulator of Ezh2 identified to date in activated lymphocytes and lymphoid malignancies.
Treatment with PTX blocked c-Rel nuclear translocation and suppressed Ezh2 expression in primary murine lymphocytes and leukemia and multiple myeloma cell lines. Moreover, combinatorial treatment with the Ezh2-specific inhibitor GSK126 and PTX lowered the IC 50 of GSK126 significantly (more than 10-fold) and reduced the survival of these malignant cell lines. PTX (Trental) has already been used widely in the clinical treatment of a variety of inflammatory disorders (45,46). Here, our data provide a molecular justification for using combinatorial therapy in cancers expressing c-Rel and high levels of wild-type Ezh2 by showing that PTX targets the Ezh2 transcriptional activator, c-Rel, lowering Ezh2 expression to a level where the Ezh2-specific inhibitor GSK126 disables the function of any remaining Ezh2 protein.
Because other members of the NF-B family of proteins share DNA binding motifs that are similar to that of c-Rel, we also determined whether RelA, RelB, p50, and p52 are able to regulate Ezh2 expression. In our experiments, only RelA enhanced luciferase activity of an Ezh2 reporter, but it did not promote endogenous Ezh2 expression. Furthermore, RelA was neither up-regulated in primary lymphocytes upon antigenic stimulation (Fig. 6C) nor recruited to the c-Rel binding site in the Ezh2 locus (data not shown). The critical role of c-Rel and not RelA in the regulation of Ezh2 expression was supported by the capacity of PTX to block the nuclear translocation of c-Rel, but RelA in T cells, the former coinciding with a reduction in Ezh2 protein levels (data not shown and Fig. 7B). It is not surprising that RelA recognizes the c-Rel binding site on the Ezh2 reporter construct, because the DNA-contacting residues are conserved in both RelA and c-Rel (49,50). However, the endogenous Ezh2 locus in lymphocytes is almost certainly regulated by additional mechanisms to ensure preferential binding of c-Rel. One drawback of our primary reporter screen is that we did not use lymphoid lineage cells because of their low transfection efficiencies. This may have precluded the identification of other cell or tissue-specific transcriptional regulators. Notwithstanding the crucial regulatory effect of c-Rel on Ezh2 expression in lymphoid cells, our study demonstrating that even in nonlymphoid lineage cells, c-Rel can function as a potent Ezh2 regulator raises the possibility that c-Rel has a hitherto unappreciated role in the control of Ezh2 expression in a wider range of cell types.
Recently, Ezh2 has been shown to be essential for germinal center B cell survival and to be involved in the regulation of T cell polarization (18,51). Given the importance of c-Rel in lym-phocyte activation, c-Rel-regulated Ezh2 expression could be critical in these physiological processes. Activation of NF-B signaling pathways are also associated with malignancies in various organisms (52). Notably, c-Rel is the only member of the NF-B family of proteins able to transform B lineage cells in culture (53). Precisely how c-Rel promotes cellular transformation remains unclear, although c-Rel can regulate BCL-2 expression directly and E2F1/2 indirectly through hyperphosphorylation of retinoblastoma protein (pRB) (54,55). Importantly, c-Rel amplification and mutations in BCL-2 and EZH2 are commonly found in germinal center B cell-like diffuse large B cell lymphoma, whereas co-expression of BCL-2 and v-Rel or hyperactive Ezh2 promotes lymphoid transformation (19,56). Taken together, this information indicates that c-Rel is likely to play a central role in the intricate regulatory circuitry of B cell transformation. With targeting of the NF-B pathway emerging as a logical therapeutic strategy to promote apoptosis of malignant lymphoid cells and to potentially improve the efficacy of chemotherapeutic agents in childhood acute lymphoblastic leukemia patients (57,58), our current findings provide a mechanistic basis for exploring new combinatorial therapies to treat cancers expressing high levels of Ezh2.