Eicosapentaenoic Acid Demethylates a Single CpG That Mediates Expression of Tumor Suppressor CCAAT/Enhancer-binding Protein δ in U937 Leukemia Cells*

Polyunsaturated fatty acids (PUFAs) inhibit proliferation and induce differentiation in leukemia cells. To investigate the molecular mechanisms whereby fatty acids affect these processes, U937 leukemia cells were conditioned with stearic, oleic, linolenic, α-linolenic, arachidonic, eicosapentaenoic, and docosahexaenoic acids. PUFAs affected proliferation; eicosapentaenoic acid (EPA) was the most potent on cell cycle progression. EPA enhanced the expression of the myeloid lineage-specific transcription factors CCAAT/enhancer-binding proteins (C/EBPβ and C/EBPδ), PU.1, and c-Jun, resulting in increased expression of the monocyte lineage-specific target gene, the macrophage colony-stimulating factor receptor. Indeed, it is known that PU.1 and C/EBPs interact with their consensus sequences on a small DNA fragment of macrophage colony-stimulating factor receptor promoter, which is a determinant for expression. We demonstrated that C/EBPβ and C/EBPδ bind the same response element as a heterodimer. We focused on the enhanced expression of C/EBPδ, which has been reported to be a tumor suppressor gene silenced by promoter hypermethylation in U937 cells. After U937 conditioning with EPA and bisulfite sequencing of the −370/−20 CpG island on the C/EBPδ promoter region, we found a site-specific CpG demethylation that was a determinant for the binding activity of Sp1, an essential factor for C/EBPδ gene basal expression. Our results provide evidence for a new role of PUFAs in the regulation of gene expression. Moreover, we demonstrated for the first time that re-expression of the tumor suppressor C/EBPδ is controlled by the methylation state of a site-specific CpG dinucleotide.

Increasing evidence from animal and in vitro studies indicates that fatty acids, especially the long-chain polyunsaturated fatty acids (PUFAs), 2 affect carcinogenesis (1). n-3 PUFAs inhibit the growth of tumor cells both in vivo and in vitro (2,3), decrease metastasis and cachexia (4,5), and increase the cytotoxic effects of some chemotherapeutic agents (6), although the results are not always consistent (7,8). In addition, n-3 PUFAs reduce cell proliferation and induce differentiation and apoptosis in hepatocarcinoma and leukemia cells (9 -12). Although not completely known, several molecular mechanisms whereby n-3 PUFAs may modify the carcinogenic process have been proposed. These include alteration in cell membrane composition and function, modulation of mitochondrial calcium homeostasis, suppression of the biosynthesis of proinflammatory molecules, influence on signal transduction, alteration of hormone-stimulated cell growth, inhibition of angiogenic mediators, production of free radicals and reactive oxygen species, and influence on transcription factor activity and gene expression (2,(13)(14)(15)(16)(17).
In hepatocarcinoma cells, n-3 PUFAs modulate the expression of CCAAT/enhancer-binding protein (C/EBP) transcription factors (18,19). C/EBPs (␣, ␤, ␦, ␥, ⑀, and ) represent a family of master regulators that play an essential role in controlling cell proliferation and differentiation processes of several cell types, including myeloid cells (20). Members of the C/EBP family are structurally related, each consisting of an N-terminal transactivating region, a central basic DNA-binding domain, and a C-terminal leucine zipper motif for dimerization (20). During hematopoiesis, C/EBP␣, C/EBP␤, and C/EBP␦ are predominantly expressed in the granulocyte and monocyte lineages, whereas C/EBP⑀ is found in the middle to later stages of differentiation of granulocytes and T cells (21)(22)(23)(24)(25). The most compelling evidence for a crucial role of the C/EBPs in myeloid cell differentiation and maturation has come from studies on knock-out mice. C/EBP␣-deficient mice fail to undergo myeloid differentiation beyond the myeloblast stage and, therefore, lack mature neutrophils (26), whereas the phenotype of C/EBP␤-deficient mice indicates a potential role in the activation and/or differentiation of macrophages (27). Moreover, overexpression of C/EBP␣, C/EBP␤, C/EBP␦, and C/EBP⑀ induce granulocyte differentiation in myeloblastic cell lines (28), suggesting that in myeloid cells C/EBP family mem-bers can compensate in vivo for the lack of one of the other C/EBP proteins.
Besides C/EBPs, other transcription factors and co-activators contribute to myeloid cell fate (29). First of all, PU.1 drives the transcription of monocyte-specific genes, including the macrophage colony-stimulating factor (M-CSF) receptor (30,31). PU.1 and C/EBPs can bind to and activate the M-CSF receptor promoter, and their combinatorial activities are essential to mediate the M-CSF receptor expression level (32). In addition, the co-activator partner protein c-Jun cooperates with PU.1 (33) and C/EBPs (34) during monocyte differentiation, although it is able itself to induce partial monocyte differentiation in a variety of myeloid cell lines (35,36). c-Jun does not directly bind to the M-CSF receptor promoter but enhances the ability of PU.1 to transactivate it (37). Synergism among PU.1, C/EBPs, and c-Jun is essential to activate monocyte target genes (34). Among these, M-CSF receptor is critical for monocyte cell survival and proliferation and is activated early during the monocyte differentiation process (38 -40).
In the present study, we evaluated the effects of fatty acid conditioning of the U937 promonocytic cell line on proliferation, cell cycle progression, and the differentiation program in relation to chain length and the number of double bonds. We found that eicosapentaenoic acid (EPA) treatment reduced cell cycle progression and induced monocyte-specific M-CSF receptor expression by enhancing C/EBP␤, C/EBP␦, PU.1, and c-Jun expression. Considering that C/EBP␦ was reported to be a tumor suppressor gene (41,42) that is silenced by promoter hypermethylation in U937 cells and re-expressed by proximal promoter demethylation (43), we analyzed the same promoter region (Ϫ370 to Ϫ20) after EPA conditioning of U937 cells. We found a site-specific CpG demethylation that was a determinant for the binding activity of Sp1 transcription factor to induce C/EBP␦ gene expression.

Preparation of Albumin-bound Fatty Acids
A stock solution of each fatty acid (5 or 10 mM) was prepared by diluting the free fatty acid in ethanol and precipitating it with the addition of NaOH (final concentration, 0.25 M). The precipitate was dried under nitrogen, reconstituted with 0.9% (w/v) NaCl, and stirred at room temperature for 10 min with defatted BSA (final concentration, 10% (w/v) in 0.15 M NaCl). Each solution was adjusted to pH 7.4 with NaOH and stored in aliquots at Ϫ20°C protected from light under nitrogen. The fatty acid/ BSA molar ratio was 3:1 or 6:1.

Cell Culture
The human promonocytic cell line U937 (CRL-1593.2) obtained from the American Type Culture Collection (Manas-sas, VA) was cultured in RPMI 1640 medium supplemented with 10% fetal calf serum, 2 mM L-glutamine, and 1% penicillinstreptomycin at 37°C in a humidified incubator aerated with 5% CO 2 . U937 cells were seeded at a density of 0.3 ϫ 10 6 cells/ml for all experiments. Cells were incubated with fatty acid/BSA solutions at the indicated concentrations and times.

Flow Cytometry Analysis
Cell Cycle and Apoptosis-U937 cells were treated with fatty acids for 24 h (50 -200 M final concentration) and analyzed by flow cytometry as indicated below. After washing, the 200 ϫ g cell pellets were resuspended in 1 ml of hypotonic PI solution (50 g ml Ϫ1 in 0.1% sodium citrate plus 0.1% Triton X-100; Sigma). The samples were placed overnight in the dark at 4°C, and the PI fluorescence of individual nuclei was measured using an EPICS XL-MCL TM flow cytometer (Beckman Coulter, Inc., Miami, FL). Analysis of apoptosis was performed as described by Nicoletti et al. (44), and data were processed by an Intercomp computer and analyzed with SYSTEM II TM software (Beckman Coulter, Inc.). The cell cycle was analyzed by measuring DNA-bound PI fluorescence in the orange-red fluorescence channel (FL2) through a 585/42-nm bandpass filter with linear amplification. Analysis of distribution profiles was performed with ModFit LT software (Verity Software House, Topsham, ME) to determine fractions of the population in each phase of the cell cycle (G 0 /G 1 , S, G 2 /M). At least 15,000 events were collected for each sample. Cells were gated on FL2-area versus FL2-width plots to exclude aggregates and debris from analysis (45).
Forward and Side Scatter-Intact U937 cells were recovered after treatment with 100 M fatty acids, resuspended in 1 ml of PBS, and identified by forward and right angle (side) light scatter using the same flow cytometer equipped with a 15-milliwatt argon ion laser (488 nm). Cell viability was determined by counting triplicate samples for trypan blue dye-excluding cells.

[ 3 H]Thymidine Incorporation Assay
Triplicate samples of 1 ϫ 10 5 U937 cells suspended in 200 l of RPMI 1640 medium were cultured in the presence or absence of 100 M fatty acids for 24 h. [ 3 H]thymidine (specific activity, 6.7 Ci/mmol) (Amersham Biosciences) was added to the cultures at 2.5 Ci/well. After a 4-h incubation, cells were harvested with a multiple suction-filtration apparatus (Mash II) on a fiberglass filter (BioWhittaker) and counted in a ␤ counter apparatus (Packard Instrument Co.).

Chromatin Immunoprecipitation
ChIP assays were performed on U937 cells grown in standard conditions or with 100 M OA or 100 M EPA for 24 h using the EZ-ChIP kit in accordance with the protocol provided by the manufacturer (Millipore-Upstate). Briefly, after cross-linking with 1% formaldehyde and quenching with 125 mM glycine, cells were recovered, resuspended in SDS lysis buffer in the presence of protease inhibitors, and sonicated to obtained 200 -1000-bp chromatin fragments. Aliquots of the fragmented chromatin corresponding to about 10 6 cells were precleared with protein G-agarose. The precleared chromatin (1%) was kept as "input," and the residual was incubated overnight with the following antibodies: anti-C/EBP␣ (sc-61X), anti-CEBP␤ (sc-150X), anti-C/EBP␦ (sc-636X), and anti-PU.1 (sc-352X). Five micrograms of each antibody and 5 g of rabbit IgG were used. Antibody-DNA complexes were captured by incubation with protein G-agarose, eluted, and subjected to cross-linking reversal. The M-CSF receptor sequence containing the C/EBP and PU.1 binding sites was amplified using Mx3000P Real-Time PCR System with Brilliant SYBR Green qPCR Master Mix. The sequences of the primers used for ChIP were as follows: forward (Ϫ87/Ϫ66), 5Ј-GACTGCGACCCCTCCCTCTTG-3Ј; reverse (ϩ15/ ϩ36), 5Ј-CCTCCTCCTTGGGCTGATCCTC-3Ј. The rela-tive amount of immunoprecipitated DNA fragment (123 bp) was determined based on the threshold cycle (Ct) for each PCR product. Data were quantitatively analyzed according to the formula (47).

DNA Isolation and Quantitative DNA Methylation Analysis of C/EBP␦ CpG Island
Genomic DNA from unsupplemented U937 cells or U937 grown for 24 h with 100 M OA or 100 M EPA was extracted using a FlexiGene DNA kit in accordance with the protocol instructions (Qiagen). DNA methylation levels were quantified using the Methyl-Profiler qPCR Primer Assay for human C/EBP␦ (MePH28341-1A) (SABiosciences-Qiagen) in accordance with the protocol provided by the manufacturer. Primers are designed by an optimized computer algorithm to ensure that the amplicon contains cutting sites for both methyl-sensitive and methyl-dependent enzymes and are specifically designed for analyzing the DNA methylation status of CpG islands using restriction enzyme digestion (DNA Methylation Enzyme kit MeA-03, SABiosciences-Qiagen) followed by SYBR Green-based real time PCR detection. Briefly, each genomic DNA was subjected to four separate treatments according to the instructions provided by the manufacturer. (i) For the mock digest (Mo), no enzymes were added in the reaction. The product of the mock digest represented the total amount of input DNA for real time PCR detection. (ii) For the methylation-sensitive digest (Ms), cleavage was carried out with a methylationsensitive enzyme, which digested unmethylated DNA. The remaining hypermethylated DNA was detected by real time PCR. (iii) For the methylation-dependent digest (Md), cleavage was carried out with a methylation-dependent enzyme, which digested methylated DNA. The remaining unmethylated DNA was detected by real time PCR. (iv) For the double digest (Msd), both enzymes were added, and all DNA molecules (both methylated and unmethylated) were digested. This reaction measures the background and the fraction of input DNA refractory to enzyme digestion. The four mixtures were incubated at 37°C overnight. The enzymes were inactivated at 65°C for 20 min. The resulting DNA was stored at Ϫ20°C or utilized for real time PCR as described above. The specific RT-PCR program was as follows: 95°C for 10 min and 40 cycles of 97°C for 15 s and 72°C for 60 s as indicated in the instruction manual. The relative amount of each DNA fraction (methylated and unmethylated) was calculated using a standard ⌬Ct method, normalizing the amount of DNA in each digestion against the total amount of input DNA in the mock digest. The amount of hypermethylated target DNA copies (C HM ) is defined as ( where C R represents the amount of target DNA copies that are resistant to enzyme digestion and is defined as 2 Ϫ⌬Ct(Msd Ϫ Mo) . The amount of unmethylated DNA (C UM ) can be determined as

Bisulfite Modification of Genomic DNA and Sequencing
Genomic DNA from unsupplemented U937 cells or U937 grown for 24 h with 100 M OA or 100 M EPA was extracted as described above. Bisulfite modification of genomic DNA was done as described (48,49). Briefly, 1 g of genomic DNA (50-l volume) was denatured by using mild heat at alkaline pH by adding 3.5 l of 3 M NaOH and incubating for 10 min at 37°C. Immediately, 10 l of 10 mM hydroquinone (Sigma) and 520 l of 3 M sodium bisulfite (Sigma) were added and incubated for 12-16 h at 50°C. Bisulfite-treated DNA was purified and eluted in 50 l of H 2 O. The conversion of uracil was completed by alkaline desulfonation by adding 5 l of 3 M NaOH and incubating at 37°C. The treated DNA was purified using a QIAquick gel extraction kit and eluted in 30 l of H 2 O. Sequencing primers for bisulfite-modified DNA were designed by using the on-line program MethPrimer (available at the University of California San Diego web site). Bisulfite-modified DNA was amplified using sequencing primers for the C/EBP␦ promoter. The primers for the first PCR were as follows: forward primer, 5Ј-GATTTTATTTTTAATTTYGAGGAGY-3Ј (Y ϭ C/T); reverse primer, 5Ј-AAACTATCACCTCRCTA-AACCCAACCC-3Ј (R ϭ C/T). Following the initial amplification, an aliquot of the initial PCR products was used as template DNA in a nested PCR. The primers for the nested PCR were as follows: forward, 5Ј-GATTTTATTTTTAATTTYGAG-GAGY-3Ј (Y ϭ C/T); reverse primer, 5Ј-CCTTTTCTAAC-CCCRACTAARTA-3Ј (R ϭ C/T). Nested PCR conditions were as follows: 95°C for 5 min and 35 cycles of 94°C for 30 s, 50°C for 30 s, and 72°C for 60 s followed by 10 min at 72°C. The amplified products were confirmed by electrophoresis on an agarose gel. The nested 350-bp PCR products (Ϫ370/Ϫ20) were subcloned into pCR2.1 TA cloning vector (Invitrogen). Single clones were selected and cultured, and plasmid DNA was isolated using a GeneElute plasmid miniprep kit (Sigma) and sequenced using T7 primer at the Genechron-Ylichron S.r.l. Laboratory (Rome, Italy). Eight to 10 clones were sequenced for each sample.

Electrophoretic Mobility Shift Assays and Antibody Interference Assays
Nuclear extracts from U937 cells were prepared, and electrophoretic mobility shift/antibody interference assays were performed as described previously (50). Complementary pairs of site-specific methylated and unmethylated oligonucleotides were from Integrated DNA Technologies-TEMA Ricerca (Bologna, Italy). Oligonucleotides were annealed by boiling for 5 min at 95°C followed by cooling and end-labeled with [␥-32 P]dATP (Amersham Biosciences) using T4 polynucleotide kinase (New England Biolabs, Beverly, MA). Labeled fragments were separated from residual [␥-32 P]dATP using Quick Spin columns (Roche Applied Science) according to the manufacturer's instructions. The upper strand C/EBP␦ promoter oligonucleotide sequence used was 5Ј-CGGGCGGGGCGTGCAC-GTCAGCCGGGGCTAG-3Ј (underlined C corresponds to the C demethylated after EPA treatment). Binding reactions were carried out by mixing 20 g of nuclear extract with 4 l of 5ϫ binding buffer (40 mM HEPES, pH 8.0, 200 mM KCl, 0.4 mM EDTA, 20 mM MgCl 2 , 40% glycerol, 20 mM dithiothreitol, 0.05% bromphenol blue) in a total volume of 18 l and incubating at 4°C for 1 h. Approximately 90,000 cpm (2 l of ϳ0.5 ng/l) [␥-32 P]dATP-labeled probe were added per reaction and incubated for 15 min at room temperature. The reaction mixtures were loaded onto a 6% polyacrylamide gel (0.75 mm thick) and separated for 3 h at 200 V in 0.5ϫ Tris borate-EDTA. The gel was dried for 30 min under vacuum and exposed to autoradiograph film using an Instant Imager autoradiography system (PerkinElmer Life Sciences). For antibody interference assays, 3.0 l of anti-Sp1 antibody (Santa Cruz Biotechnology Inc. sc-59) was incubated with the nuclear extract and binding buffer for 1 h at 4°C in a total volume of 18 l. After addition of the probe, subsequent steps were identical to those for the band shift assays.

Statistical Analyses
All average results are presented as mean Ϯ S.D. One-way or two-way (where appropriate) analyses of variance with Bonferroni's post-test were used. Values were considered significant at p Ͻ 0.05.

Effects of PUFAs on Cell Viability, Proliferation, and Cellular
Morphology of U937 Cells-Variable effects of PUFAs on proliferation, differentiation, and apoptosis in leukemia cells have been reported in relation to cell line and fatty acid concentration (9,11,12). To study the effects of fatty acids on the percent distribution in the various phases of the cell cycle, U937 cells were analyzed by flow cytometry after incubation with fatty acids of increasing carbon chain length and double bond number. Twenty-four hour treatment with LNA, AA, EPA, and DHA increased significantly the percentage in G 0 -G 1 phase in a concentration-dependent manner compared with untreated cells. EPA exhibited the greatest effect when compared with LNA, AA, and DHA (Fig. 1A). Contrarily, SA, OA, and LA had no effect at any of the concentrations studied (Fig. 1A). For untreated U937 cells, 43.4% were in G 0 -G 1 phase, 44.5% were in S phase, and 11.9% were in G 2 -M phase (Fig. 1B). After 24-h supplementation with LNA, AA, EPA, and DHA (100 M final concentration), the percentage of cells in G 0 -G 1 phase increased (maximal effect was 64% with EPA) paralleled by a decrease in the percentage of cells in S and G 2 -M phases (33 and 2.5%, respectively, with EPA). Only minor changes were observed in SA-, OA-, and LA-treated cells in the same experimental conditions (Fig. 1B). No fatty acid treatment, except SA (51), induced U937 apoptosis up to a 200 M concentration for 24, 48, and 72 h (not shown). When cells were exposed to 100 M LNA, AA, EPA, and DHA for 24 h, a significant reduction in the cell number was observed as compared with untreated cells (p Ͻ 0.001), whereas SA, OA, and LA had no significant effect on cell proliferation (Fig. 1C). The increased number of cells in G 0 -G 1 phase and the decrease of total cell number induced by LNA, AA, EPA, and DHA treatments were in agreement with the reduced DNA synthesis. Indeed, LNA, AA, EPA, and DHA inhibited significantly [ 3 H]thymidine incorporation (p Ͻ 0.001). As expected, no effect on [ 3 H]thymidine incorporation was found in SA, OA, and LA treatments (Fig.  1D). EPA treatment produced the greatest effect on reduction of cell cycle progression.
Flow cytometry analyses, measuring forward and side scatter, were used to obtain information about the morphology of U937 leukemia cells after fatty acid treatments. Analysis of the cells showed a significant increase both in the forward and side AUGUST 5, 2011 • VOLUME 286 • NUMBER 31

JOURNAL OF BIOLOGICAL CHEMISTRY 27095
scatter of the U937 cells following 24-h treatment with 100 M AA, EPA, and DHA, indicating an increase in cell size and granularity. SA, OA, LA, and LNA addition resulted in a lower increase in forward and side scatter (Fig. 1E).
EPA Increases C/EBP, PU.1, and c-Jun Expression-We examined whether the effects on cell viability, proliferation, and cellular morphology induced by PUFA conditioning could be related to enhanced expression of lineage-specific transcription factors involved in the cell cycle progression and differentiation process of myeloid cell lines. Fatty acid treatments (100 M) had no effect on the content of any of the C/EBP␣ isoforms ( Fig.  2A). LNA, AA, EPA, and DHA increased the protein content of C/EBP␤, C/EBP␦, PU.1, and c-Jun, whereas SA, OA, and LA had a lesser or no effect and were unable to simultaneously induce all four transcription factors ( Fig. 2A). The enhanced protein expression levels of C/EBP␤, C/EBP␦, c-Jun, and PU.1 let us to suppose that U937 cells undergo the early phase of the myeloid differentiation process. Indeed, C/EBP␤, C/EBP␦, c-Jun, and PU.1 are involved in the transcriptional control of granulocyte and monocyte development (25).
To determine whether PUFA treatment affected C/EBP, PU.1, and c-Jun protein levels due to transcriptional events, RT-PCR was performed. We selected OA and EPA as potential inactive and active inducers, respectively, of C/EBP␣, C/EBP␤, C/EBP␦, PU.1, and c-jun mRNA levels. To assess the expression kinetic profile, we evaluated mRNA content in unsupplemented U937 cells and after 1-, 3-, and 24-h treatment with 100 M OA or EPA. A simultaneous and significant increase of C/EBP␤, C/EBP␦, PU.1, and c-jun mRNA levels was observed after 3 h of EPA treatment that further increased up to 24 h (Fig. 2B). On the contrary, OA did not induce any significant change. In agreement with the unchanged protein content, OA and EPA conditioning had no effect on C/EBP␣ mRNA levels (Fig. 2B).
EPA Enhances M-CSF Receptor Expression-The increased C/EBP␤, C/EBP␦, PU.1, and c-jun mRNA levels in response to EPA suggested a potential transcriptional effect on their target genes. M-CSF receptor is expressed very early during the monocyte differentiation process; PU.1, C/EBPs, and co-activator partner protein c-Jun are the transcriptional activators of M-CSF receptor (30,32). The M-CSF receptor mRNA levels were evaluated in U937 cells after 1-, 3-, and 24-h treatment with 100 M OA or EPA. Similarly to C/EBP␤, C/EBP␦, PU.1, and c-jun (Fig. 2B), M-CSF receptor mRNA levels increased significantly at 3-and 24-h EPA treatments (p Ͻ 0.001 versus control U937 untreated cells) (Fig. 3A), whereas OA had no effect. M-CSF receptor protein was not or barely detectable in the U937 untreated cell line. Twenty-four hour conditioning with 100 M SA, OA, and LA did not produce a significant increase. As expected, EPA, DHA, LNA, or AA addition resulted in evident protein increase (Fig. 3B).

C/EBP␤, C/E␦BP␦, and PU.1 Bind and Activate M-CSF Receptor Promoter after EPA Treatment-To investigate whether C/EBP and PU.1 transcription factors induced by EPA
were able to bind their consensus sequences on M-CSF receptor promoter (30,32), ChIP analysis were performed in U937 cell after 24 h of OA and EPA treatments using C/EBP and PU.1 antibodies. RT-PCR amplification of the Ϫ87/ϩ36 (123-bp) promoter region containing the elements under investigation (30,32) was performed in the immunoprecipitated DNA. No amplification product was found with normal rabbit IgG as a negative control for the antibodies utilized. The results of ChIP experiments demonstrated a significant ability of C/EBP␤, C/EBP␦, and PU.1 to bind the M-CSF receptor consensus sequences after EPA treatment, whereas C/EBP␣ was ineffective (Fig. 3C). No significant effect was exerted by OA. These results suggest that both C/EBP␤ and C/EBP␦ specifically bind to the same C/EBP consensus sequence most likely as a C/EBP␤/C/EBP␦ heterodimer. As a consequence, their enhanced expression following EPA treatment is particularly relevant.
EPA Induces C/EBP␦ Expression Acting as Site-specific Demethylating Agent-C/EBP␦ is silenced by hypermethylation in U937 cells, and the demethylating agent 5-aza-2Ј-deoxycytidine is able to induce its expression (43). We verified the effect of 5-aza-2Ј-deoxycytidine on C/EBP␦ protein levels in our experimental conditions and compared it with that of EPA. Western blot analysis of U937cells after 2-and 5-day treatment with 1 M 5-aza-2Ј-deoxycytidine as well as after 24-h 100 M EPA treatment showed increased protein levels (Fig. 4A). We hypothesized that EPA could exert its effects on protein expression by acting as a demethylating agent.
The analysis of C/EBP␦ gene (Ϫ3000/ϩ1269) using the EMBOSS (European Molecular Biology Open Software Suite) or MethPrimer on-line software programs retrieved six putative islands. To verify whether EPA treatment demethylates C/EBP␦ CpG islands, the percent content of CpG DNA methylation was quantified using the Methyl-Profiler qPCR Primer Assay. Quantitative RT-PCR indicated that the amount of C/EBP␦ hypermethylated DNA copies decreased significantly (p Ͻ 0.001) after EPA conditioning compared with OA-treated or untreated cells (Fig. 4B). The 5Ј-upstream promoter region containing a CpG island with high CG content and a determinant for C/EBP␦ gene re-expression (43) (Fig. 4C) was analyzed in detail. We performed bisulfite sequencing of this region and found a high degree of methylation in agreement with the low C/EBP␦ protein level. EPA treatment induced a site-specific CpG demethylation in all the sequenced clones (Fig. 4D). The same was not found for OA treatment, which induced random and low level CpG demethylation. Indeed, the specific CpG dinucleotide demethylated after EPA treatment is located in a critical region of C/EBP␦ proximal promoter between the TATA box and an Sp1 binding site (Fig. 4C) and may be functional for C/EBP␦ expression.
To evaluate the influence of this specific CpG methylation on the activity of Sp1 binding to the C/EBP␦ gene promoter, the Sp1 site was investigated by comparative electrophoretic mobility shift assay. C/EBP␦ gene promoter-specific oligos con-taining an unmethylated or methylated specific CpG (Fig. 5) were incubated with U937 nuclear extracts, and the binding activity was compared. Fig. 5 shows that the signal of the shifted band is stronger with the unmethylated compared with the methylated probe. Incubation of the electrophoretic mobility shift assay reactions with an anti-Sp1 antibody supershifted the unmethylated probe consistent with Sp1-specific binding to the probe. Overall, these experiments indicate that demethylation of a specific CpG can increase Sp1 binding to C/EBP␦ promoter.

DISCUSSION
In this study, we evaluated the effects of PUFAs, relative to saturated and monounsaturated fatty acids, on U937 promyelo-FIGURE 2. Effect of fatty acids on myeloid lineage-specific transcription factors. A, U937 cells were treated with 100 M fatty acids for 24 h. Total cell lysates (50 g of protein) were subject to Western blotting with the indicated antibodies as described under "Experimental Procedures." For each protein, one representative of four experiments is reported. Images of independent blots were acquired using the VersaDoc Imaging System, and signals were quantified using Quantity One software. The -fold change in each protein was compared with control U937 cells (C) and was calculated after correction for ␤-tubulin cytic leukemia cells. PUFAs inhibited DNA synthesis and cell cycle progression and promoted changes in cell morphology by increasing size and granularity of U937 cells. At the same time, M-CSF receptor expression increased. M-CSF receptor, specifically induced in the early phase of monocyte differentiation commitment, is expressed on the monocyte-macrophage cell lineage, and as such, it is a useful marker to discriminate between monocyte and granulocyte progenitor cells and their differentiated progeny (39). Expression of the M-CSF receptor gene is under stringent control of both extracellular and intracellular stimuli and appears to occur primarily at the transcriptional level (40). M-CSF receptor mRNA levels are under the control of monocyte lineage-specific transcription factors such as PU.1 and C/EBPs; the activity and specificity of the M-CSF receptor promoter are mediated by a small DNA fragment con-taining binding sites for PU.1 and C/EBPs (30,32). CEBP␣, C/EBP␤, and C/EBP␦ can all bind to and activate the M-CSF receptor promoter, a function that has been ascribed to their highly similar basic leucine zipper domains (52). Moreover, studies on the differences in the binding affinities of recombinant C/EBPs to the M-CSF receptor promoter demonstrated that C/EBP␤ binding is relatively weaker than CEBP␣ and C/EBP␦ binding (32).
We found that EPA treatment induced C/EBP␤ and C/EBP␦ expression and promoted C/EBP␤/C/EBP␦ heterodimer binding to the same response element on the M-CSF receptor promoter, resulting in enhanced M-CSF receptor expression. In addition, EPA treatment induced an increase on c-jun mRNA and protein levels. This last result is in agreement with previous data indicating that c-jun and M-CSF receptor mRNA increase during monocyte differentiation of U937 cells (53). Interestingly, the M-CSF receptor promoter has no consensus binding site for c-Jun transcription factor (37). As a consequence, c-Jun does not directly bind to M-CSF receptor promoter, but it associates with PU.1 and C/EBP␤ via its basic domain (34,37). Moreover, co-activator c-Jun recruitment by C/EBP␤ to DNA facilitates RNA polymerase II recruitment (34). c-Jun, PU.1, and C/EBP␤ have been shown to physically interact with each other and enhance the transcription of monocyte-specific genes via binding to their respective sites on DNA (33,54). c-Jun expression has been shown to be differentially up-regulated during monocyte, but not granulocyte, differentiation of myeloid cell lines, implying a specific role for c-Jun expression in monocyte development (55). Our results confirm that there is not a single master myeloid transcription factor that alone governs myeloid lineage commitment, but multiple transcription factors work cooperatively and coordinately to regulate both temporal and lineage-specific genes. In this light, the simultaneous effects induced by EPA treatment on myeloid lineage-specific PU.1, C/EBP␤, C/EBP␦, and c-jun gene expression suggest a complex regulatory network.
Among PU.1, C/EBP␤, C/EBP␦, and c-Jun transcription factors, we focused on the induction of the tumor suppressor C/EBP␦. Indeed, C/EBP␦ is specifically inactivated by promoter CpG island hypermethylation in U937 and other leukemia cell lines and in acute myeloid leukemia in response to treatment with a demethylating agent associated with gene re-expression (43). DNA hypermethylation is a common mechanism for Cell lysates (50 g of protein) were loaded. One representative of three experiments is shown. Images of independent blots were acquired using the VersaDoc Imaging System, and signals were quantified using Quantity One software. The -fold change of C/EBP␦ protein was compared with control U937 cells and was calculated after correction for ␤-tubulin loading differences. Data are the mean Ϯ S.D. of three separate experiments. B, cells were treated with 100 M fatty acids for 24 h, and the methylated DNA levels of C/EBP␦ CpG islands were quantified using the Methyl-Profiler qPCR Primer Assay as described under "Experimental Procedures." The means Ϯ S.D. (error bars) of three separate experiments are shown (*, p Ͻ 0.001 versus OA-treated or U937 untreated cells). C, C/EBP␦ gene proximal promoter DNA sequence contains a CpG island. The underlined sequences indicate the forward and reverse nested primers utilized for cloning and sequencing after bisulfite reaction. The cloned fragment (350 bp) contains 32 CpGs (in bold). The arrow indicates the CpG demethylated nucleotide after EPA treatment. D, sequencing of the individual clones generated by PCR after bisulfite reaction. Black and white circles represent methylated and unmethylated CpGs, respectively.
tumor suppressor gene inactivation in hematopoietic malignancies (56). Methylation of promoter CpG islands is associated with a closed chromatin structure and transcriptional silencing of the associated genes. Aberrant methylation of normally unmethylated promoter 5Ј-CpG-rich areas is the most commonly studied epigenetic mechanism associated with the transcriptional silencing of known and candidate tumor suppressor genes (57). The pattern of promoter methylation found in hematopoietic malignancies can be considered to be aberrant and a cancer-specific phenomenon; a disease-specific methylation pattern of key CpG islands has been found for particular genes (57).
In the present study, both 5-aza-2Ј-deoxycytidine and EPA conditioning induced C/EBP␦ protein expression (Ref. 43 and Fig. 4A). C/EBP␦-enhanced expression is ascribed to demethylation of a site-specific CpG downstream of the Sp1 site in the C/EBP␦ proximal promoter (Fig. 4C) required for C/EBP␦ basal transcriptional activation (58). Indeed, methylation of this specific CpG impaired Sp1 binding to its consensus sequence (Fig.  5). Our results are consistent with the site-specific CpG demethylation of C/EBP␦ proximal promoter in primary breast cancer (59). Moreover, evidence demonstrating gene promoter site-specific methylation as a mechanism of tumor suppressor gene silencing in various types of cancer cells has been reported (60, 61). A single CpG demethylation appears to be the molecular event associated with a putative antitumor gene expression in prostate carcinogenesis (62). On this basis, we can assume that the CpG demethylation events are not all equal within a CpG island. Although it is generally accepted that high concentrations of demethylating agents induce the maximal gene reactivation (63), paradoxically, global demethylation may result in increased tumorigenicity (64). Moreover, a strong demethylating action prevents ascertaining whether any of the CpG dinucleotides exhibits differential susceptibility toward the methylation-demethylation process and hinders locating the critical regulatory sequences within CpG islands. We identified a CpG dinucleotide in the C/EBP␦ proximal promoter as a key element that proved useful in identifying hypersensitive regions essential for gene regulation. This site may serve as a dynamic switch to activate the C/EBP␦ gene through a possible chromatin conformation change.
Surprisingly, even 3-h EPA treatment was able to significantly induce C/EBP␤, C/EBP␦, PU.1, c-jun, and M-CSF receptor mRNA level increases (Figs. 2B and 3A). These results led us to conclude that the site-specific C/EBP␦ promoter demethylation occurred within 3 h, suggesting the involvement of some active demethylation mechanism(s) occurring in the absence of DNA replication (the cell cycle is ϳ24 h in these cells). The existence of demethylating enzymes has been previously postulated when rapid demethylation of genes occurs (65,66).
The effects induced by EPA treatment on PU.1, C/EBP␤, C/EBP␦, c-Jun, and M-CSF receptor protein levels are similar to the effects of LNA, AA, and DHA treatments. On this basis, a potentially similar mechanism for all PUFA conditioning cannot be excluded. Although the molecular mechanisms underlying the CpG site-specific demethylation process and the potential common PUFA mechanisms need to be investigated, our findings provide for the first time evidence of a tight correlation among C/EBP␦ site-specific CpG demethylation, M-CSF receptor expression, and the beginning of the differentiation program induced by EPA in U937 leukemia cells.