DNA Cytosine Methylation in the Bovine Leukemia Virus Promoter Is Associated with Latency in a Lymphoma-derived B-cell Line

Bovine leukemia virus (BLV) proviral latency represents a viral strategy to escape the host immune system and allow tumor development. Besides the previously demonstrated role of histone deacetylation in the epigenetic repression of BLV expression, we showed here that BLV promoter activity was induced by several DNA methylation inhibitors (such as 5-aza-2′-deoxycytidine) and that overexpressed DNMT1 and DNMT3A, but not DNMT3B, down-regulated BLV promoter activity. Importantly, cytosine hypermethylation in the 5′-long terminal repeat (LTR) U3 and R regions was associated with true latency in the lymphoma-derived B-cell line L267 but not with defective latency in YR2 cells. Moreover, the virus-encoded transactivator TaxBLV decreased DNA methyltransferase expression levels, which could explain the lower level of cytosine methylation observed in the L267LTaxSN 5′-LTR compared with the L267 5′-LTR. Interestingly, DNA methylation inhibitors and TaxBLV synergistically activated BLV promoter transcriptional activity in a cAMP-responsive element (CRE)-dependent manner. Mechanistically, methylation at the −154 or −129 CpG position (relative to the transcription start site) impaired in vitro binding of CRE-binding protein (CREB) transcription factors to their respective CRE sites. Methylation at −129 CpG alone was sufficient to decrease BLV promoter-driven reporter gene expression by 2-fold. We demonstrated in vivo the recruitment of CREB/CRE modulator (CREM) and to a lesser extent activating transcription factor-1 (ATF-1) to the hypomethylated CRE region of the YR2 5′-LTR, whereas we detected no CREB/CREM/ATF recruitment to the hypermethylated corresponding region in the L267 cells. Altogether, these findings suggest that site-specific DNA methylation of the BLV promoter represses viral transcription by directly inhibiting transcription factor binding, thereby contributing to true proviral latency.

otic bovine leucosis, a neoplastic proliferation of B cells (1)(2)(3)(4)(5). BLV is structurally and biologically closely related to human T-lymphotropic viruses HTLV-I and HTLV-II (4,6). The majority of BLV-infected cattle are asymptomatic carriers of the virus. Only about 30% of BLV-infected animals develop a preneoplastic condition termed persistent lymphocytosis, with 2-5% developing B-cell leukemia and/or lymphoma after a long latency period (7). The virus can be inoculated experimentally into sheep, in which the disease appears earlier and with higher frequencies, providing a helpful model for understanding BLV and HTLV-induced leukemogenesis. BLV infection is characterized by viral latency in a large majority of infected cells and by the absence of viremia. These features are thought to be due to the transcriptional repression of viral expression in vivo (8 -11). Latency is likely to be a viral strategy to escape the host immune response and allow tumor development.
The BLV transcriptional promoter is located in the 5Ј-long terminal repeat (5Ј-LTR) and is composed of the U3, R, and U5 regions. Transcription initiates at the U3-R junction of the 5Ј-LTR. BLV gene expression is induced at the transcriptional level by the virus-encoded transactivator Tax BLV (12,13). Transactivation by Tax BLV requires the presence of three 21-bp enhancer elements (called Tax-responsive elements (TxREs)) located in the U3 region of the 5Ј-LTR (12,14). Each TxRE contains a core octanucleotide sequence corresponding to an imperfectly conserved cyclic AMP-responsive element (CRE), which binds cellular transcription factors: the CRE-binding protein (CREB), the CRE-modulator isoform (CREM), and activating transcription factor-1 and -2 (ATF-1 and ATF-2) (14 -17). Besides the imperfect CRE consensus, each TxRE also contains an E-box sequence, which overlaps each of the three CRE motifs (4,18). A PU.1/Spi-B site (19) and a glucocorticoid responsive-element (20 -23) are also present in the U3 region. In addition, BLV expression is regulated by LTR sequences located downstream of the transcription initiation site: an upstream stimulatory factor-binding site in the R region (24) and an interferon regulatory factor-binding site in the U5 region (25).
During retroviral infection, the RNA viral genome is retrotranscribed into double-stranded DNA, which becomes integrated into the host genome and is organized into chromatin as all cellular genes. This chromatin environment is likely to be a key parameter for the control of viral gene expression because transcriptional activation by cellular or viral trans-acting factors is dependent on chromatin accessibility. BLV expression is controlled by chromatin structure. Indeed, our laboratory reported previously that histone acetylation is an important epigenetic modification for Tax BLV -dependent and -independent viral transcriptional regulation (26 -29). BLV promoter activity is induced by deacetylase (HDAC) inhibitors, and this is correlated with an increased histone-4 acetylation at the viral promoter (27). Our laboratory has also demonstrated that HDAC inhibitors and Tax BLV synergistically activate BLV promoter transcriptional activity in a CRE-and CREB-dependent manner. This synergism is mediated by an HDAC inhibitor indirect action that requires de novo protein synthesis and that increases the level of CREB/ATF bound to the BLV promoter (28).
To further investigate the role of chromatin structure in the control of BLV gene expression, we studied in this report the involvement of another epigenetic modification, DNA methylation, in BLV gene expression regulation. The enzymes that establish and maintain specific DNA methylation patterns include the de novo DNA methyltransferases DNMT3A and DNMT3B and the maintenance DNA methyltransferase DNMT1 (30,31). In mammalian cells, DNA methylation occurs predominantly at cytosines in CpG dinucleotides of transcriptional regulatory regions and is generally associated with gene silencing, either directly by inhibiting the binding of transcription factors to their recognition sequences or indirectly by preventing transcription factors from accessing their target sites through attachment of methyl-CpG-binding proteins, which "read" DNA methylation patterns. These proteins (MeCPs) recruit histone deacetylases and histone methyltransferases, thereby resulting in formation of a closed repressive chromatin structure (31,32). Here, we have demonstrated that DNA cytosine hypermethylation is associated with BLV postintegration latency in a lymphoma-derived B-cell line, L267. Mechanistically, site-specific methylation in BLV promoter CRE elements inhibits CREB/CREM/ATF-1 in vitro binding and decreases by 2-fold BLV LTR-driven transcription. Importantly, by chromatin immunoprecipitation (ChIP) assays, we have demonstrated in vivo the recruitment of the transcription factors CREB and CREM, and to a lesser extent ATF-1, to the hypomethylated CRE region of the YR2 5Ј-LTR, whereas we detected no CREB/CREM/ATF recruitment to the hypermethylated corresponding region in the L267 cells. Interestingly, we show that Tax BLV protein decreases the expression levels of DNMT1, DNMT3A, and DNMT3B, which could, at least in part, explain the lower level of DNA methylation observed in the L267 LTaxSN 5Ј-LTR compared with the L267 5Ј-LTR. These results indicate that DNA methylation contributes to proviral latency, a viral strategy to escape host immune response and allow tumor development.

EXPERIMENTAL PROCEDURES
Cell Lines and Cell Culture-The Raji cell line, a human B-lymphoid Epstein-Barr virus-positive cell line derived from a Burkitt's lymphoma, and the Epstein-Barr virus-negative B-lymphoid cell line DG75 were maintained in RPMI 1640-Glutamax I medium (Invitrogen) supplemented with 10% fetal bovine serum, 50 g of streptomycin/ml and 50 units of penicillin/ml. The human epithelial HeLa cell line is derived from a cervical carcinoma and is transformed by human papillomavirus type 18. HeLa cells were cultured in Dulbecco's modified Eagle's-Glutamax I medium (Invitrogen) supplemented with 5% fetal bovine serum, 50 g of streptomycin/ml and 50 units of penicillin/ml. L267 is a clonal lymphoma-derived B-cell line established from a BLV-infected sheep (S267) injected with naked proviral DNA of an infectious BLV variant described previously (33)(34)(35). The tumor-derived L267 provirus displays a wild-type sequence that is identical to the infectious provirus used for sheep S267 inoculation (34,35). The L267 LTaxSN cell line results from the transduction of native L267 with the pLTaxSN retroviral vector expressing the tax cDNA following cocultivation with the PG13 LTaxSN producer cell line and G418 selection of transduced cells (34,36). L267 and L267 LTaxSN were maintained in Opti-MEM medium (Invitrogen) supplemented with 10% fetal bovine serum, 1 mM sodium pyruvate, 2 mM glutamine, non-essential amino acids, and 100 g of kanamycin. YR2 is a cloned B-lymphoid cell line established from peripheral blood lymphocytes isolated from a BLV-infected sheep (10,36,37). YR2 cells were maintained in Opti-MEM-Glutamax I medium (Invitrogen) supplemented with 10% fetal bovine serum, 50 g of streptomycin/ml, and 50 units of penicillin/ml. All cells were grown at 37°C in an atmosphere of 5% CO 2 .
Transient Transfection and Luciferase Assays-Raji and DG75 cells were transfected using the DEAE-dextran procedure as described previously (38). At 22 h after transfection, the cells were mock-treated or treated with increasing concentrations of procaine (Sigma), procainamide (Sigma), 5-aza-2Ј-deoxycytidine (Sigma), or zebularine (Calbiochem) as indicated. At 72 h after induction, the cells were lysed and assayed for luciferase activity (Promega). Luciferase activities were normalized with respect to protein concentrations using the detergent-compatible protein assay (Bio-Rad). HeLa cells were transfected using FuGENE TM 6 (Roche Applied Science) as described previously (24). All transfection mixtures contained pRL-TK (50 ng), in which a cDNA encoding Renilla luciferase is under the control of the herpes simplex virus thymidine kinase promoter region and which is used as an internal control for transfection efficiency. At 48 h posttransfection, luciferase activities (firefly and Renilla) were measured in cell lysates by using the dual luciferase reporter assay system (Promega).
Bisulfite Cytosine Methylation Analysis-Samples of total genomic DNA (700 ng) isolated with the DNeasy tissue kit (Qiagen) and digested overnight with EcoRI in a volume of 20 l were prepared for methylation analysis as described previously by Olek et al. (39). Briefly, the digested DNA, which had been boiled for 5 min, chilled quickly on ice, and incubated in 0.3 M NaOH at 50°C for 15 min, was mixed with 2.5 volumes of 2% low melting point agarose (SeaPlaque agarose, FMC Bioproducts). Ten-microliter volumes of agarose-DNA mixture were pipetted into 750 l of chilled mineral oil to form agarose beads. Aliquots of 1 ml of 2.5 M sodium metabisulfite (Sigma) and 125 mM hydroquinone (Sigma) (pH 5.0) were then added to mineral oil containing up to four agarose beads. The reaction mixtures were kept on ice for 30 min and then incubated in the dark at 50°C for 3.5 h. The agarose beads were equilibrated four times (15 min each time) with 1 ml of Tris-EDTA (pH 8.0) and DNA desulfonated in 500 l of 0.2 M NaOH (twice for 15 min each time) at room temperature. Finally, the beads were washed with 1 ml of Tris-EDTA (pH 8.0) (3 times for 10 min each time) and then with 1 ml of sterile water (twice for 15 min each time) and kept at 4°C.
ChIP Assays-ChIP assays were performed using the ChIP assay kit (EZ ChIP technology, Upstate). YR2 and L267 cells were cross-linked for 10 min at room temperature with 1% formaldehyde. To detect chromosomal flanking regions, pellets were sonicated (Bioruptor sonicator) to obtain DNA fragments at an average size of 400 bp. Chromatin immunoprecipitations were performed with an antibody directed against CREB (Santa Cruz Biotechnology, catalog number sc-186), CREM (sc-440), ATF-1 (sc-243), and ATF-2 (sc-242). To test aspecific binding to the beads, a purified IgG was used as a control for immunoprecipitation (Vector Laboratories, catalog number I-1000). Quantitative PCR reactions were performed with the MESA GREEN qPCR MasterMix (Eurogentec). Relative quantification using standard curve method was performed for each primer pair, and 96-well optical reaction plates were read in an Applied Biosystems ABI Prism 7300 real-time PCR instrument (absolute quantification method). Fold enrichments were calculated as percentages of input values following this formula: Immunoprecipitated DNA/total DNA. Primer sequences used for quantification in a region overlapping the three CRE sites of the BLV LTR (Fw, 5Ј-GCCAGAAAAGCTGGTGACG-3Ј and Rv, 5Ј-GGGTGTGGATTTTTCGGG-3Ј) and in the U5 region of the BLV LTR, where no CRE binding site has been reported (Fw, 5Ј-CCTCTGACCGTCTCCACGT-3Ј and Rv, 5Ј-AAGT-AAGACAGGAAACAAGCGC-3Ј), were designed using the software Primer Express 2.0 (Applied Biosystems).
Plasmid Constructs and Generation of Constructs with Sitespecific Methylation-The eukaryotic expression vector pSG-Tax BLV was a kind gift from Drs. Richard Kettmann and Luc Willems (Faculty of Agronomy, Gembloux, Belgium). The eukaryotic pcDNA3.1/Myc-His-derived expression vectors for DNMT1, DNMT3A, and DNMT3B (pCMV-Myc-Dnmt1, pDNMT3aMyc, and pDNMT3bMyc, respectively) were kind gifts from Drs. François Fuks and Tony Kouzarides (42,43). The non-episomal plasmid pLTRwt-luc and the episomally replicating plasmid pLTR*wt(direct)-luc were described previously (in Refs. 24 and 27, respectively) and contain the luciferase gene under the control of the complete 5Ј-LTR of the 344 BLV provirus. Strain 344 is an infectious and pathogenic molecular clone (33). The vector pLTRwt-luc was used as a substrate for mutagenesis with the QuikChange site-directed mutagenesis method (Stratagene) using 50 ng of pLTRwt-luc DNA/reaction. Following PCR, the samples were treated with DpnI for 2 h to remove the DNA template. The products were purified using the QIAquick method (Qiagen), reannealed, and precipitated. DNA concentration was determined by UV absorbance before transfection. Mutations in the CRE1 and CRE2 sites were generated with the following pairs of mutagenic oligonucleotides: CRE1 (Fw, 5Ј-CAGACAGAGTGGTCAGCTGCC-3Ј; Rv, 5Ј-GGCAGCTGACCACTCTGTCTG-3Ј) and CRE2 (Fw, 5Ј-AAGCTGGTGTGGGCAG CTGGT-3Ј; Rv, 5Ј-ACCAGC-TGCCCACACC AGCTT-3Ј) (mutations are highlighted in boldface type, and the CRE motif is underlined). These mutated plasmids were designated pLTRmutCRE1 and pLTRmutCRE2, respectively. All mutated constructs were fully sequenced after identification (Genome Express).
The pLTRwt-luc vector with methylation either at the Ϫ154 CpG, or the Ϫ129 CpG or at both CpGs in combination (Ϫ154 CpG and Ϫ129 CpG) was generated using site-specific methylated oligonucleotides (purchased from Eurogentec) and the QuikChange site-directed mutagenesis kit (Stratagene) following a protocol adapted from Martinowich et al. (44). In brief, two oligonucleotide primers, each complementary to opposite strands of the parental vector and methylated at the Ϫ154 CpG (Fw, 5Ј-CAGACAGAGAC Me GTCAGCTGCC-3Ј; Rv, 5Ј-GGC-AGCTGAC Me GTCTCTGTCTG-3Ј) or the Ϫ129 CpG (Fw, 5Ј-AAGCTGGTGAC Me GGCAGCTGGT-3Ј; Rv, 5Ј-ACCAGCT-GCC Me GTCACCAGCTT-3Ј), were extended by the PfuTurbo DNA polymerase. The product of this reaction was then treated with the endonuclease DpnI, which is specific for methylated and hemimethylated DNA (target sequence 5Ј-Gm 6 ATC-3Ј), and digested the parental DNA template to select for mutationcontaining synthesized DNA. This experimental strategy led to the production of circular closed plasmids with site-directed methylation on both strands. These CpG methylated plasmids were designated pLTR-154me, pLTR-129me, and pLTR-154me/-129me, respectively, and transfected directly after the mutagenesis reaction without amplification in bacteria. As an additional control, we generated pLTRwt-luc and pLTR*wt(direct)-luc derivatives, which were hypermethylated only in the LTR (i.e. in which every CpG of the BLV LTR, but no CpG elsewhere in the pLTRwt-luc or pLTR*wt(direct)-luc plasmid, were methylated in vitro by the SssI methyltransferase). To this end, we methylated in vitro the whole pLTRwt-luc (or pLTR*wt(direct)-luc) plasmid with SssI and then cloned back the purified fragment corresponding to the LTR in the parental reporter vector pGL3basic (or pREP10), thereby generating pLTRmSssIme (or pLTR*mSssIme).
ELISA-After culture, the cells were harvested by centrifugation, and the supernatants were removed. The BLV p24 protein in the supernatants was titrated by an immunoenzymatic capture technique as described by Portetelle et al. (45).

Activation of BLV Promoter Transcriptional Activity by
Inhibitors of DNA Methylation-Our laboratory had previously demonstrated that histone acetylation is an important chromatin modification for Tax BLV -dependent and -independent transcriptional regulation of BLV (26 -28). To further study the molecular mechanisms underlying the epigenetic control of BLV expression, we examined the potential role of DNA methylation in BLV transcriptional activity. To this end, we tested the effects of DNA methylation inhibitors on transcriptional activity of the BLV promoter (the 5Ј-LTR) in transient transfection experiments (Fig. 1). A pLTRmSssIme reporter plasmid containing the firefly luciferase (luc) gene under the control of the complete BLV 5Ј-LTR (nucleotides Ϫ211 to ϩ320) (24), which had been methylated previously in vitro by the Sss1 methyltransferase (see "Experimental Procedures"), was transiently transfected into the human B-lymphoid Raji cell line. Transfected cells were mock-treated or treated with increasing concentrations of procaine (0 -10 mM), procainamide (0 -10 mM), 5-aza-2Ј-deoxycytidine (5-AZAdc) (0 -200 nM), or zebularine (0 -1 mM) for 72 h and assayed for luciferase activity. As shown in Fig. 1A, the BLV LTR was activated in a dose-dependent manner by each of the four DNA methylation inhibitors used. pLTRmSssIme presented procaine inductions varying from 1.3-to 4.9-fold, procainamide inductions varying from 1.3-to 3.2-fold, 5-AZAdc inductions varying from 1.6-to 2.9fold, and zebularine inductions varying from 1.0-to 9.3-fold. Similar results were observed using the human B-lymphoid DG75 cell line (data not shown). Bisulfite sequencing analyses revealed that the increased transcriptional activity of the BLV LTR following treatment of the cells with each of the four DNA methylation inhibitors used was correlated with a partial CpG demethylation in the U3 and R regions of the BLV LTR (data not shown).
Because transiently transfected DNA does not always form proper chromatin structure, we next wanted to confirm the above described transient transfection results in a nucleosomal context. To this end, we used an episomally replicating BLV LTR-luc reporter plasmid (pLTR*wt(direct)-luc), described previously by our laboratory from a modified pREP10 episomal vector that contained the Epstein-Barr virus replication origin and encoded nuclear antigen EBNA-1 (27). Several studies have reported that pREP-based episomal constructs display the hallmarks of proper chromatin structure when transiently transfected into cells (46 -49). We transiently transfected into Raji cells the episomally replicating pLTR*mSssIme plasmid, where the BLV promoter had been methylated previously in vitro by the Sss1 methyltransferase. Transfected cells were mocktreated or treated with increasing concentrations of the four precited DNA methylation inhibitors and assayed for luciferase activity. As shown in Fig. 1B, the BLV promoter was activated in a dose-dependent manner by procaine (up to 4.0-fold), procainamide (up to 3.4-fold), 5-AZAdc (up to 4.5-fold), and zebularine (up to 4.7-fold). Moreover, a cellular model in which a BLV 5Ј-LTR-luciferase cassette was stably integrated into the B-lymphoid DG75 cell line (construction named clone 27; its characterization is part of a separate, unpublished study from our laboratory) was also used to address the effect of DNA

. Activation of BLV LTR activity by DNA methylation inhibitors.
A, Raji cells were transiently transfected using the DEAE-dextran procedure with 500 ng of pLTRmSssIme, where the BLV LTR had previously been methylated in vitro by the SssI methyltransferase. At 22 h after transfection, cells were mock-treated or treated with increasing concentrations of procaine (0, 2.5, 5, 7.5, and 10 mM), procainamide (0, 2.5, 5, 7.5, and 10 mM), 5-AZAdc (0, 50, 100, 150, and 200 nM), or zebularine (0, 0.05, 0.1, 0.25, and 0.5 mM). Luciferase activities were measured in cell lysates 96 h after transfection and normalized to protein concentrations. Results are presented as histograms indicating the procaine, procainamide, 5-AZAdc, and zebularine inductions of pLTRmSssIme with respect to its basal activity, which was assigned a value of 1. Means Ϯ S.E. are shown. Results from a representative experiment of three independent transfections are shown. B, Raji cells were transiently transfected using the DEAE-dextran procedure with 500 ng of the episomal vector pLTR*mSssIme, where the BLV LRT had previously been methylated in vitro by the SssI methyltransferase. At 22 h posttransfection, cells were mock-treated or treated with increasing concentrations of procaine (0, 0.1, 0.5, 1, and 10 mM), procainamide (0, 0.1, 0.5, 1, and 10 mM), 5-AZAdc (0, 250, 500, 1000, and 2500 nM), or zebularine (0, 0.05, 0.075, 0.1, and 1 mM). Luciferase activities were measured in cell lysates 96 h after transfection and normalized to protein concentrations. Results are presented as histograms indicating the procaine, procainamide, 5-AZAdc and zebularine inductions of pLTR*mSssIme with respect to its basal activity, which was assigned a value of 1. Means Ϯ S.E. are shown. The results from a representative experiment of three independent transfections are shown. methylation inhibition. Using clone 27, we observed that increasing concentrations of DNA methylation inhibitors enhanced BLV 5Ј-LTR-directed luciferase expression in a dosedependent manner (data not shown). These results demonstrate that the BLV LTR-directed gene reporter expression is activated in a dose-dependent manner in response to selective DNA methylation inhibitors in the context of non-episomal and episomal LTR methylated constructs, thereby suggesting a role of DNA methylation in BLV transcriptional regulation.
Hypermethylation of the U3 and R Regions in the BLV 5Ј-LTR in L267 Lymphoma-derived B-cell Line-To further examine the potential role of DNA methylation in transformation-associated BLV silencing, we investigated the DNA methylation status of the 5Ј-LTR in the context of integrated BLV proviruses. To this end, we used three different BLV-infected cell lines representing two distinct viral transcriptional states: the YR2 and L267 cell lines, which represent the silent/nonproductive state, and the L267 LTaxSN cell line, which represents the productive state. On the one hand, the YR2 cell line is an ovine B-cell line, which was cloned from leukemic B cells of a BLVinfected sheep (10,36,37,50). This cell line contains a single monoclonally integrated silent provirus in which two Glu-to-Lys amino acid substitutions in the Tax BLV protein impair the infectious potential of the integrated provirus (33). The L267 cell line is a B-cell line isolated from sheep S267, a BLV-infected sheep injected with naked proviral DNA of an infectious BLV variant (33)(34)(35). Provirus expression is completely suppressed in L267 lymphoma B cells, as demonstrated conclusively by their incapacity to generate infection in the in vivo sheep model, and the tumor-derived L267 provirus displays a wild-type sequence that is identical to the infectious provirus used for sheep S267 inoculation (34,35). In contrast to the YR2 cell line, which represents a defective latency state, the L267 cell line represents a true latency state, i.e. a reversible transcriptionally silent but replication-competent provirus. On the other hand, the L267 LTaxSN cell line is a virus-expressing cell line resulting from transduction of the native L267 cell line with a retroviral vector pLTaxSN expressing a wild-type Tax BLV protein. Transduction of this Tax BLV coding vector restores viral expression from the silent integrated provirus (34).
To verify the productive versus nonproductive state of our cell lines, we first measured BLV transcripts by reverse transcription qPCR in each of the three cell lines. We observed very low levels of viral transcripts in the YR2 and L267 cell lines, as compared with the L267 LTaxSN cell line, which abundantly produces BLV mRNAs ( Fig. 2A). Similarly, measurement of p24 major capsid antigen by ELISA in the supernatants of all cultured cell lines showed a dramatic increase in the production of this structural antigen in the L267 LTaxSN cell line compared with the silent L267 and YR2 cell lines (Fig. 2B). These results indicate that exogenous Tax BLV is capable of restoring viral expression from silent integrated proviruses, in agreement with previous studies (34).
To determine the methylation status of each CpG dinucleotide in the 5Ј-LTR of integrated BLV proviruses, we used the sodium bisulfite sequencing technique to analyze genomic DNA from the YR2, L267, and L267 LTaxSN cell lines. To distinguish between the 5Ј-and 3Ј-LTRs, which have the same nucle-otide sequence, we used reverse primers located in the gag gene (Fig. 2C, primers NB and ND) or forward primers located in the X region (primers NE and NG) for the nested PCR, respectively. As shown in Fig. 2C, the U3 region and the 5Ј-half of the R region of the 5Ј-LTR in L267 cells were densely methylated. Indeed, six consecutive CpG dinucleotides in the U3 region and nine consecutive CpG dinucleotides in the R region, overlapping the transcription initiation site, were strongly methylated. Among these methylated CpGs, the Ϫ154 CpG and the Ϫ129 CpG (located in the distal CRE1 and middle CRE2 sites, respectively) were methylated in 100% of the cloned sequences tested for the L267 cell line. These two methylated Ϫ154 and Ϫ129 CpGs are indicated by asterisks in Fig. 2C. In the L267 LTaxSN cells, only the 5Ј-half of the R region was methylated although to a lesser extent than in the L267 cells. Remarkably, the 5Ј-LTR in the YR2 cell line was completely unmethylated.
In all retroviruses, LTRs localized at the 5Ј and 3Ј extremities of the proviral genome have identical nucleotide sequences but display different functions; the 5Ј-LTR functions as a transcriptional promoter, whereas the 3Ј-LTR has a polyadenylation function. Therefore, we addressed the question of differential CpG methylation patterns in the sequentially identical but functionally different BLV LTRs. Our bisulfite sequencing analysis of the 3Ј-LTR revealed a lack of CpG methylation in the three tested cell lines: YR2, L267, and L267 LTaxSN (Fig. 2C).
In conclusion, our results indicate that the BLV true latency state correlates with CpG hypermethylation in the 5Ј-LTR U3 and R regions of the integrated proviral DNA, directly demonstrating for the first time the association of this epigenetic modification with a specific viral expression state. This could be a viral strategy to escape host immune response and allow tumor onset and propagation. Importantly, these methylation marks were not observed in the YR2 proviral system, representative of a defective latency state.
Ectopically Expressed DNMT1 and DNMT3A, but Not DNMT3B, Down-regulate BLV Promoter Activity-To address the role of DNMTs in transcriptional regulation of the BLV promoter, transient cotransfections of Raji cells were performed with the pLTRwt-luc reporter construct and increasing amounts of expression vectors coding for DNMT1, DNMT3A, or DNMT3B. As shown in Fig. 3A and B, DNMT1 and DNMT3A strongly repressed basal LTR-directed gene expression, suggesting that these DNMTs could negatively regulate BLV promoter activity. In contrast, DNMT3B had no effect on reporter gene expression (Fig. 3C), suggesting a certain functional specificity of DNMTs in the transcriptional regulation of the BLV promoter. Western blot experiments allowed us to verify the efficiency of each DNMT overexpression in correlation with the level of BLV 5Ј-LTR promoter activity (Fig. 3). These results confirmed that the inability of DNMT3B to repress BLV 5Ј-LTR activity was not due to a lack of DNMT3B protein expression (Fig. 3C).
In addition, we performed ChIP experiments to evaluate in vivo the recruitment of DNMTs to the BLV promoter using chromatin samples prepared from the transiently transfected cells. We immunoprecipitated the chromatin either with an antibody directed against the c-Myc epitope fused to the different DNMTs present in the recombinant expression vectors we had used in the overexpression experiments or with antibodies directly targeting DNMT1, DNMT3A, or DNMT3B. However, we did not succeed in detecting any statistically relevant ChIP signal compared with the negative control (consisting of a purified IgG antibody). In additional ChIP experiments, we used chromatin samples from YR2 and L267 cells to study DNMT recruitment in vivo to the hypomethylated and hypermethylated BLV promoters, respectively. Despite many attempts, we could not detect in vivo the recruitment of these DNMTs to the BLV 5Ј-LTR region in our ChIP assays neither with the chromatin extracted from the transiently transfected cells nor with the chromatin prepared from the BLV-infected cell lines YR2 and L267 (data not shown).
Moreover, we evaluated the DNA methylation profile of the BLV LTR in the LTR-reporter constructs following overexpression of each DNMT. Methylation-sensitive restriction analyses revealed that overexpression of DNMT1 and DNMT3A (and to a lesser extent DNMT3B) increased CpG methylation in the BLV LTR. However, the results we obtained by the bisulfite sequencing technique revealed in each case a very weakly methylated profile (data not shown), which did not correspond to the methylation pattern we observed in the infected cell lines (see Fig. 2C), thereby suggesting that the establishment of the DNA methylation profile during or after natural integration into the host cell genome can not be recapitulated using LTR-reporter constructs.
Taken together, our results suggest that the activity of the BLV promoter is negatively regulated by the DNA methyltransferases DNMT1 and DNMT3A but not by DNMT3B. However, we could not demonstrate the recruitment of these enzymes in vivo in the BLV 5Ј-LTR region.
Synergistic Activation of BLV Promoter by Tax BLV and DNA Methylation Inhibitors-To examine the potential link between DNA methylation and Tax BLV transactivation, we tested the effect of two DNA methylation inhibitors (5-AZAdc and procaine) on basal and Tax BLVinduced BLV LTR activity. To this end, human B-lymphoid Raji cells were transiently cotransfected with the pLTR-luc reporter construct (methylated in vitro by Sss1 before transfection) and with increasing amounts of a Tax BLV expression vector, pSG-Tax BLV . Twenty-two hours posttransfection, transfected cells were mock-treated or treated with either 5-AZAdc (250 M) or procaine (7.5 mM) for 72 h and assayed for luciferase activity ( Table 1). As expected, in the absence of 5-AZAdc or procaine, Tax BLV transactivated the BLV promoter in a dose-dependent manner up to 985-fold (Table 1, Tax BLV fold activation, lines 2-6). Treatment of transfected cells with 5-AZAdc or procaine alone resulted in a 2.77-or 5.31-fold activation of luciferase expression, respectively ( Table 1, line 1). Remarkably, when cells were both cotransfected with increasing amounts of the Tax BLV expression vector and treated with 5-AZAdc (or procaine), a strong synergism was observed between Tax BLV and these DNA methylation inhibitors, resulting in transactivations ranging from 134-to 2094-fold in the presence of 5-AZAdc (Table 1, lines 2-6, Tax BLV ϩ 5-AZAdc fold activation) and from 350-to 3824fold in the presence of procaine (Table 1, lines 2-6, Tax BLV ϩ procaine fold activation). Transcriptional activators synergize when their combination produces a transcriptional rate that is greater than the sum of the effects produced by individual activators (51). Transfection of 1 ng of pSG-Tax BLV led to a 52.8-fold stimulation of transcription in the absence of DNA methylation inhibitors, whereas in the presence of 5-AZAdc or procaine, it led to a 134-or 350-fold stimulation, respectively (  Table 1 are presented in Fig. 4 as histograms indicating the synergistic activation (-fold) by Tax BLV and 5-AZAdc and by Tax BLV and procaine (Fig. 4, A and B, respectively). This synergism between Tax BLV and 5-AZAdc (or procaine) persisted even at saturating amounts of the Tax BLV transactivator (data not shown), indicating that the observed effect was not the consequence of increased Tax BLV expression because of activation of simian virus 40 (SV40) promoter by 5-AZAdc or procaine. Moreover, we confirmed by Western blot experiments that 5-AZAdc or procaine did not increase the amount of Tax BLV protein by activating the SV40 promoter (data not shown). Similar results were observed using another B-lymphoid cell line (the DG75 cell line; data not shown).
Synergistic activation by ectopically expressed Tax BLV and inhibitors of DNA methylation required intact CRE binding sites in the three BLV TxREs, because point mutations in these three CRE motifs abrogated the synergistic effect (pLTR-mut3CRE-luc; Table 1, lines 7-12, and Fig. 4). This implied that the synergistic effect was mediated by interactions at the CRE motifs and not at the otherwise intact LTR BLV DNA sequences. In conclusion, these results demonstrate that the DNA methylation inhibitors 5-AZAdc and procaine synergistically enhance the transcriptional activity of the BLV promoter mediated by Tax BLV in a CRE-dependent manner.
Methylation at Ϫ154 CpG and Ϫ129 CpG in BLV 5Ј-LTR CRE1 and CRE2 Sites, Respectively, Impairs CREB, CREM, and ATF-1 Binding-Our bisulfite sequencing results indicate that the central CpGs located in the 5Ј-LTR CRE1 and CRE2 sites (i.e. Ϫ154 CpG and Ϫ129 CpG) are hypermethylated in the integrated BLV provirus of L267 cells (Figs. 2C and 5). This hypermethylation correlates with the transcriptionally silent viral state in these cells. We investigated the in vitro effect of methylation at the Ϫ154 CpG (distal CRE1 site) and at the Ϫ129 CpG (middle CRE2 site) on transcription factor binding by EMSAs. For these studies, we used oligonucleotide probes corresponding to the TxRE1 (nucleotides Ϫ164 to Ϫ144) or to the TxRE2 (nucleotides Ϫ139 to Ϫ119) sites of the BLV 5Ј-LTR, which were either unmethylated (referred to as TxRE1 or TxRE2) or methylated at the Ϫ154 or Ϫ129 CpG (referred to as TxRE1-154me or TxRE2-129me, respectively). These probes were incubated with nuclear extracts from the latently BLVinfected B-cell line L267 treated (Fig. 6, A and B, lanes 2 and 7) or not (lanes 1 and 6) with a combination of PMA ϩ ionomycin (PϩI), two known activators of BLV expression (26,52,53); from the defective cell line YR2 treated (lanes 4 and 9) or not (lanes 3 and 8) with PMA ϩ ionomycin; or from PBMCs derived from a BLV-infected sheep (BLV-infected sheep M298 presenting a persistent lymphocytosis) (lanes 5 and 10). When the EMSAs were carried out with the unmethylated oligonucleotide TxRE1 and TxRE2 probes, we observed, as reported previously by our laboratory (17), a minor and two major retarded bands (called C1, C2, and C3, respectively) (Fig. 6, A  and B, lanes 1-5). Our laboratory had previously reported that the C1 complex contains the ATF-2 transcription factor and the C2 complex contains ATF-1, CREB, and CREM proteins (14,15,17,27). Supershift experiments did not identify the nature of the proteins present in complex C3 (17). Treatment of YR2 or L267 cells with PMA ϩ ionomycin increased the intensity of the retarded DNA-protein complex C2 as compared with mock-treated cells (Fig. 6, A and B, compare lanes 1 and 2 and  lanes 3 and 4). Moreover, we observed a strong binding of CREB/CREM/ATF-1 proteins to the TxRE1 and TxRE2 probes when these probes were incubated with nuclear extracts prepared from BLV-infected PBMCs (Fig. 6, A and B, lanes 5).
Remarkably, when the methylated TxRE1-154me probe or the methylated TxRE2-129me probe was used, we observed the complete disappearance of complex C2, even with nuclear extracts from infected PBMCs or from PMA ϩ ionomycintreated YR2 or L267 cells (Fig. 6, A and B, compare lanes 1-5  with lanes 6 -10). Following PMA ϩ ionomycin treatment of YR2 cells, complex C1 seemed to migrate slower than in mock-treated cells (Fig. 6A, compare lanes 3 and 4), but we could not identify by supershift experiments any potential additional (co)factor(s) associated with complex C1 following PMA ϩ ionomycin treatment (data not shown). However, complex C1 was not observed when PBMC nuclear

Synergistic activation of BLV promoter activity by Tax BLV and inhibitors of DNA methylation
Raji cells were transiently cotransfected using the DEAE-dextran procedure with 500 ng of either the pLTRwt-luc (lines 1-6) or pLTR-mut3CRE-luc (lines 7-12) (both methylated in vitro by the SssI methyltransferase before transfection) and with increasing amounts of pSG-Tax BLV (from 0 to 100 ng of plasmid DNA). To maintain the same amount of transfected DNA and to avoid squelching artifacts, the different amounts of pSG-Tax BLV cotransfected were complemented to 100 ng of DNA by using the pSG5 empty vector. Twenty-two hours posttransfection, cells were mock-treated or treated with either 5-AZAdc (250 M) or procaine (7.5 mM) for 72 h. Luciferase activities were measured in cell lysates 96 h after transfection and normalized with respect to protein concentrations. Results are presented as relative light units (RLU), Tax BLV fold activation, Tax BLV ϩ 5-AZAdc fold activation, and Tax BLV ϩ procaine fold activation of the reporter constructs (pLTRwt-luc and pLTR-mut3CRE-luc) with regard to their respective basal activity level, which was arbitrarily set at a value of 1. The Tax BLV ϩ 5-AZAdc (or Tax BLV ϩ procaine) fold synergism was determined as described previously (51) using the following formula: [Fold activation by Tax BLV ϩ 5-AZAdc (or Tax BLV ϩ procaine)]/{[fold activation by Tax BLV alone] ϩ [fold activation by 5-AZAdc (or procaine) alone]}. Values represent the means of triplicate samples with S.E. indicated in parentheses. An experiment representative of four independent transfections is shown. extracts were used (Fig. 6, A and B, lanes 5), suggesting a minor function of this complex.
To confirm these results, unlabeled double-stranded oligonucleotides were prepared and used as competitors in the EMSAs with the unmethylated TxRE1 or TxRE2 probe and nuclear extracts from BLV-infected PBMCs (supplemental Fig.  S1, A or B, respectively). As shown previously by our laboratory (17), formation of complex C2 was competed out by molar excesses of the TxRE1 or TxRE2 unlabeled homologous oligonucleotides but not by the same molar excesses of a heterologous oligonucleotide of unrelated sequence. Moreover, the C2 complex was not or was weakly competed out by the same molar excesses of methylated TxRE1-154me (supplemental Fig. S1B) or TxRE2-129me (supplemental Fig. S1A) unlabeled oligonucleotides, respectively, thereby confirming that formation of complex C2 was dependent on the presence of an unmethylated TxRE1 or TxRE2 sequence (supplemental Fig. S1). Altogether, these results indicate that methylation at the Ϫ154 CpG or at the Ϫ129 CpG abolish in vitro binding of the CREB/CREM/ATF-1 transcription factors to the CRE1 or CRE2 site, respectively.
To confirm these results in vivo, we conducted ChIP experiments using the YR2 and L267 cell lines, representative of defective and true latency, respectively. These two cell lines allowed us to compare CREB/CREM/ATF recruitment to the hypomethylated YR2 BLV 5Ј-LTR with their recruitment to the hypermethylated L267 BLV 5Ј-LTR. We used specific antibodies directed against CREB, CREM, ATF-1, and ATF-2 proteins and a purified IgG antibody as a control. Two different probes were designed in the BLV 5Ј-LTR to compare the recruitment of these proteins to the TxRE site region with their recruitment to the U5 region in which no CRE sites had been reported previously. As shown in Fig. 6C, we observed the recruitment of transcription factors CREB and CREM, and to a lesser extent ATF-1, to the hypomethylated CRE region of YR2 5Ј-LTR. In contrast, no CREB/CREM/ATF recruitment was observed to the hypermethylated corresponding region in the L267 cell line (Fig. 6C). Moreover, in both cell lines, the ATF-2 protein, which is part of complex C1 (see the EMSAs in Fig. 6, A and B, and as reported previously by our laboratory (17)), did not seem to be recruited in vivo to the 5Ј-LTR region (Fig. 6C). This latter result is in agreement with our in vitro binding studies showing that complex C1, which contains the ATF-2 transcription factor, is not the major retarded band observed in gel shift assays using the TxRE probes (Fig. 6, A and B), suggesting a less important function of ATF-2 compared with that of CREB/CREM/ ATF-1. Altogether, these results indicate that CpG methylation impairs in vivo CREB/CREM/ATF-1 binding to the CRE binding sites of the BLV 5Ј-LTR.
Methylation at the Ϫ129 CpG in the CRE2 Site Down-regulates BLV 5Ј-LTR Transcriptional Activity-To assess the transcriptional regulatory function of methylation at the Ϫ154 and Ϫ129 CpGs, we modified the pLTRwt-luc vector so that the Ϫ154 CpG and/or the Ϫ129 CpG were methylated by a PCRmediated mutagenesis protocol (see "Experimental Procedures" and Ref. 44). As controls, we also generated, by the same technique, an unmethylated pLTRwt-luc as well as unmethylated mutant reporter constructs named pLTRmutCRE1 or pLTRmutCRE2, which contained point mutations abolishing CREB/CREM/ATF-1 transcription factor binding to the CRE1 or CRE2 site, respectively. Directly after purification, the PCR products from these site-directed mutagenesis experiments were transiently transfected into the human epithelial HeLa cell line (which presents high transfection efficiency). At 44 h posttransfection, cells were lysed and assayed for luciferase activity. As shown in Fig. 7A, methylation at the Ϫ129 CpG caused a 46% decrease in luciferase activity. Methylation at the Ϫ154 CpG did not affect basal BLV LTR-directed gene expression. Moreover, when combined with methylation at the Ϫ129 CpG, methylation at the Ϫ154 CpG did not further decrease BLV promoter activity compared with what we observed with the plasmid methylated only at the Ϫ129 CpG.
Point mutation in the CRE1 site (pLTRmutCRE1) or in the CRE2 site (pLTRmutCRE2) caused a 35 or 71% decrease, respectively, of basal LTR-directed gene expression. As an additional control, we generated a pLTRwt-luc derivative that was hypermethylated only in the LTR (i.e. in which every CpG of the 5Ј-LTR, but no CpG elsewhere in the pLTRwt-luc plasmid, was methylated in vitro by SssI methyltransferase). This construction was named pLTRmSssIme (see "Experimental Procedures"). As shown in Fig. 7A, when transiently transfected the  Table 1. Values represent the means of triplicate samples Ϯ S.E. are indicated. An experiment representative of four independent transfections is shown.
pLTRmSssIme exhibited a 72% decrease in luciferase activity compared with that observed with the unmethylated pLTRwtluc construct.
To validate our method, we performed additional independent transfection experiments using the pLTRwt-luc, pLTRmutCRE1, and pLTRmutCRE2 constructs resulting from the same PCR mutagenesis technique but followed by a classical amplification step in bacteria and not directly transfected after purification of the mutagenized PCR products. These constructs were transiently transfected into HeLa cells. At 44 h posttransfection, cells were lysed and assayed for luciferase activity. As shown in Fig. 7B, mutation in the CRE1 or the CRE2 site reduced BLV promoter activity by 39 or 75%, respectively. Remarkably, these reductions were almost identical to those observed by directly transfecting the same plasmids after mutagenesis (compare pLTRmutCRE1 and pLTRmutCRE2 in Fig. 7), thereby validating the mutagenesis technique we used in Fig. 7A.
Altogether, these results demonstrate that methylation at the Ϫ129 CpG significantly reduces BLV promoter-driven reporter gene expression, whereas methylation at the Ϫ154 CpG has no effect on the BLV promoter activity. These data suggest that methylation at the Ϫ154 CpG position in the BLV promoter, even if it impairs CREB/CREM/ATF-1 transcription factor binding, does not interfere with basal transcriptional activity of this promoter ex vivo.
Transactivator Protein Tax BLV Down-regulates DNMT Expression-In our bisulfite sequencing experiments, we observed a partial loss of DNA methylation of the BLV 5Ј-LTR in L267 LTaxSN cells compared with L267 cells (Fig. 2C). A possible explanation is that the transactivating Tax BLV protein could decrease the expression level of DNA methyltransferases. In order to address this issue, we quantified DNMT1, DNMT3A, and DNMT3B protein expression levels in the L267 LTaxSN cell line compared with the L267 cell line by Western blot experiments. Remarkably, DNMT1 and DNMT3B protein expression was strongly down-regulated in L267 LTaxSN cells compared with L267 cells (Fig. 8A). However, we could not assess the protein level of DNMT3A because no ovine antibody against DNMT3A was available commercially and the murine anti-DNMT3A antibody did not reveal any specific band in either L267 or L267 LTaxSN cellular extracts. We next measured DNMT mRNA expression levels in these two cell lines by reverse transcription qPCR. As shown in Fig. 8B, we observed significant decreases in DNMT1, DNMT3A and DNMT3B mRNA levels in L267 LtaxSN cells compared with L267 cells, whereas BLV transcripts were more abundant in the productive L267 LtaxSN cells than in the silent L267 cells. These results suggest that Tax BLV down-regulates DNMT expression at the transcriptional level, which could, at least in part, explain the lower level of DNA methylation we observed in the L267 LTaxSN 5Ј-LTR compared with that observed in the L267 5Ј-LTR.
On the basis of our results, we concluded that Tax BLV -dependent down-regulation of DNMT expression could lead to a lower level of DNA methylation in the BLV promoter region, notably in the CRE site region. The transcription factors CREB/ CREM/ATF-1 could subsequently bind to their cognate sites and allow Tax BLV -mediated transactivation of the BLV 5Ј-LTR, thereby facilitating virus escape from latency.

DISCUSSION
BLV proviral latency appears to represent a viral strategy to escape host immune system and allow tumor development. Our laboratory had previously demonstrated the role of histone deacetylation in BLV transcriptional repression (26 -29). In this report, we further studied the epigenetic control of BLV expression and showed that BLV promoter activity was induced by several DNA methylation inhibitors (such as 5-AZAdc). Moreover, ectopically expressed DNMT1 and DNMT3A, but not DNMT3B, down-regulated BLV promoter activity. However, we failed to detect any recruitment in vivo of DNMTs to the  5 and 10). Sequence-specific retarded bands of interest are shown. A minor and two major protein-DNA complexes were observed and named C1, C2, and C3, respectively (indicated by arrows). The presence of ATF-2 in the minor protein-DNA complex C1 and the presence of ATF-1, CREB, and CREM in the major protein-DNA complex C2 were reported previously by our laboratory (17). B, EMSA analysis of nuclear factors interacting with the methylated or unmethylated TxRE2 probes. The nucleotide sequence of the methylated TxRE2-129me double-stranded oligonucleotide is shown with the CRE2 site indicated by an arrow on the coding strand and with the methylated cytosine residues indicated by the abbreviation me. The unmethylated TxRE2 double-stranded oligonucleotide (lanes 1-5) or its methylated version, TxRE2-129me (lanes 6 -10), was used as probe and incubated with nuclear extracts from the BLV latently infected B-cell line L267 (30 g) treated (lanes 2 and 7) or not (lanes 1 and 6) with PMA ϩ ionomycin for 30 min; from the defective cell line YR2 (15 g) treated (lanes 4 and 9) or not (lanes 3 and 8) with PMA ϩ ionomycin; or from BLV-infected ovine PBMCs (10 g) (lanes 5 and 10). Sequence-specific retarded bands of interest are shown. A minor and two major protein-DNA complexes were observed (C1, C2, and C3, respectively; indicated by arrows). The presence of ATF-2 in the minor protein-DNA complex C1 and the presence of ATF-1, CREB, and CREM in the major protein-DNA complex C2 were reported previously by our laboratory (17). C, CREB/CREM/ATF-1 are recruited in vivo to the BLV 5Ј-LTR CRE region. YR2 and L267 cells were crosslinked for 10 min at room temperature with 1% formaldehyde. To detect chromosomal flanking regions, pellets were sonicated to obtain DNA fragments of an average size of 400 bp. Chromatin immunoprecipitations were performed with an antibody directed against CREB, CREM, ATF-1, or ATF-2. To test aspecific binding to the beads, a purified IgG was used as a control for immunoprecipitation. Quantitative PCR reactions were performed with oligonucleotide primers hybridizing in a region overlapping the three CRE sites of the BLV LTR or in the U5 region of the BLV LTR, where no CRE binding site has been reported previously. BLV promoter neither in the context of transiently transfected LTR-luc reporter constructs nor in the latently infected cell lines YR2 and L267. It should be stressed that several viral promoters have been shown to be regulated by CpG methylation (54 -57), but in vivo DNMT recruitment to these viral regions has never been reported so far. This could be explained by a rapid kinetics of DNMT action or by a transient and/or weak recruitment of these enzymes to the viral promoters.
Importantly, using the sodium bisulfite sequencing technique, we showed that DNA cytosine hypermethylation in the 5Ј-LTR U3 and R regions of the integrated proviral DNA was associated with a BLV true latency state in the lymphoma-derived B-cell line L267 but not with a defective latency state in the YR2 cell line. Indeed, the YR2 cell line is a B-cell line characterized by a silent defective Tax BLV -mutated provirus (33,36), whereas the L267 cell line is a transformed B-cell clone isolated from a lymphomatous BLV-infected sheep, which carries a single full-length, nondefective, transcriptionally silent provirus (36,37). This latter cell line represents a true latency state, i.e. a reversible, transcriptionally silent, but replicationcompetent provirus. These results suggest that BLV gene expression is inactivated in tumoral cells by epigenetic mechanisms such as DNA methylation or by genetic alterations such as inactivating mutations in the tax gene. . Down-regulation of BLV promoter activity by site-specific methylation. A, using site-directed mutagenesis we generated BLV promoter luciferase constructs with site-specific methylation at the Ϫ154 CpG or/and Ϫ129 CpG (pLTR-154me, pLTR-129me, and pLTR-154/-129me, respectively) as well as its unmethylated counterpart, pLTRwt-luc. By the same technique, we generated mutated BLV promoter luciferase constructs containing point mutation abolishing transcription factor binding to the CRE1 or CRE2 motif (pLTRmutCRE1 or pLTRmutCRE2, respectively). The products from these mutagenesis reactions (500 ng of DNA) were transfected directly into HeLa cells using the FuGENE procedure. Each transfection included 50 ng of the internal control plasmid, pRL-TK, in which the herpes simplex virus thymidine kinase promoter drives Renilla luciferase gene expression. Luciferase activities (firefly and Renilla) were measured in cell lysates 44 h after transfection. The results are expressed as luciferase firefly /luciferase Renilla . As an additional control, we generated a pLTRwt-luc derivative that was hypermethylated only in the LTR (i.e. in which every CpG of the BLV LTR, but no CpG elsewhere in the pLTRwt-luc plasmid, was methylated in vitro by SssI methyltransferase) and named pLTRmSssIme. The results are presented as histograms indicating the luciferase activity of each reporter construct relative to that measured with the pLTRwt-luc vector, which was arbitrarily assigned an activity value of 100%. The down-regulation of BLV promoter activity by methylation(s) or by mutation(s) is also indicated (-fold) at the top of the histogram representation. Means Ϯ S.E. from five independent transfections performed with different DNA products are shown. B, HeLa cells were transiently cotransfected using the FuGENE procedure with 500 ng of pLTRwt-luc, pLTRmutCRE1, or pLTRmutCRE2 and 50 ng of the internal control plasmid, pRL-TK. Here, in contrast to the plasmids used in A, the reporter plasmids were amplified in bacteria prior to transfection. Luciferase activities (firefly and Renilla) were measured in cell lysates 44 h after transfection. The results are expressed as Luciferase firefly /luciferase Renilla . The results are presented as histograms indicating the luciferase activity of each reporter construct relative to that measured with the pLTRwt-luc vector, which was arbitrarily assigned a value of 100% of activity. The down-regulation of BLV promoter activity by mutations is indicated (in -fold) at the top of the histogram representation. Means Ϯ S.E. from triplicate samples are represented. An experiment representative of two independent transfections is shown.

FIGURE 8. Down-regulation of DNMT expression levels by Tax BLV .
A, nuclear extracts from the two BLV-infected cell lines, L267 and L267 LTaxSN , were analyzed by Western blotting with an antibody directed against DNMT1, DNMT3B, or DNMT3A. We could not assess the protein level of DNMT3A because no ovine antibody against DNMT3A was available commercially, and the murine anti-DNMT3A antibody did not reveal any specific band in either the L267 or the L267 LTaxSN cellular extracts. B, total RNA samples were extracted from L267 or L267 LTaxSN cells and digested by DNase I. First strand cDNA was synthesized by reverse transcription, and qPCR reactions were performed with oligonucleotide primers amplifying DNMT1, DNMT3A, or DNMT3B transcripts and ␤-actin transcripts. The results are presented as the ratio of DNMT to ␤-actin. Means Ϯ S.E. from triplicate samples are represented. An experiment representative of three independent experiments is shown. As a control, we performed quantitative PCR with oligonucleotide primers amplifying the 5Ј-UTR region of the BLV as presented in Fig. 2A.
Efficient BLV transcription requires the virus-encoded Tax BLV protein, which acts through three TxREs located in the 5Ј-LTR. The binding of CREB, CREM, and ATF transcription factors to the three imperfect CRE sites located in each TxRE has been proposed to allow Tax BLV -mediated transcriptional activation. In this report, we showed that DNA methylation inhibitors synergistically enhanced the transcriptional activation of the BLV promoter by Tax BLV in a CRE-dependent manner. Mechanistically, methylation of the CpG located at the Ϫ154 position or at the Ϫ129 position relative to the transcription start site impairs in vitro CREB/CREM/ATF-1 transcription factor binding to the distal CRE1 or middle CRE2 site, respectively. Moreover, methylation at the Ϫ129 CpG alone was sufficient to decrease by 2-fold BLV promoter-driven reporter gene expression. By ChIP experiments, we demonstrated in vivo the recruitment of the transcription factors CREB and CREM, and to a lesser extent ATF-1, to the hypomethylated CRE region of the YR2 5Ј-LTR, whereas we detected no CREB/CREM/ATF recruitment to the hypermethylated corresponding region in the L267 cell line, thereby indicating an inverse correlation between hypermethylation of, and in vivo CREM/CREM/ATF-1 binding to, the BLV promoter. As Tax BLV recruitment to the BLV LTR requires CREB/ CREM/ATF binding to the TxREs, CpG methylation also impairs Tax BLV -mediated transactivation (see above). These findings suggest that site-specific DNA methylation of the BLV promoter represses viral transcription by directly inhibiting transcription factor binding and that CpG methylation contributes to true proviral latency, together with histone deacetylation.
The methylation status of the 5Ј-LTR of BLV has been controversial so far. Indeed, an earlier study using methylationsensitive restriction enzymes has shown that the proviral DNA is hypermethylated in BLV-induced tumors (58). Later on, another study reported by Tajima et al. (59) examined the extent of DNA methylation of the U3 region and part of the R region of the 5Ј-LTR in BLV-infected cattle and in experimentally BLV-infected sheep at various stages of the disease; and they detected no or minimal CpG methylation of the U3/R region by the bisulfite genomic sequencing method even in tumor cells. The first study (58) is in agreement with our results in the true latent L267 cell line, whereas the second study (59) seems more compatible with what we observed in the defective latent YR2 cell line. Therefore, a possible explanation for these discrepancies could be that the tumor cells examined in Tajima's study carried proviruses silenced by genetic alteration(s), such as deletions in the BLV provirus or mutations into functionally important domains of the Tax BLV protein, that are well established mechanisms of transformation-associated virus silencing in ovine B-cell tumors (10,36,60).
We also studied the BLV 5Ј-LTR methylation status in the L267 LTaxSN cell line in which expression of Tax BLV via the pLTaxSN vector relieves the repressive state of the L267 provirus and allows viral expression, as shown by viral RNA expression and by virus release in the naive host (34). In this productive L267 LTaxSN cell line, the U3 region is almost completely devoid of methylated CpGs, and the 5Ј-half of the R region is methylated but to a lesser extent compared with the L267 cell line. These results reinforce the inverse correlation between the degree of methylation of cytosine residues at CpG dinucleotides in proviral DNA and the transcriptional activity of the BLV promoter.
Importantly, our results indicate that Tax BLV expression decreases DNMT1, DNMT3A, and DNMT3B expression levels, which could explain, at least in part, the lower level of DNA methylation observed in the L267 LTaxSN 5Ј-LTR compared with the L267 5Ј-LTR. In addition to playing a direct tumorigenic function, Tax BLV and its more extensively studied HTLV-1 counterpart Tax HTLV-1 modulate the expression of several genes, either up-or down-regulating their expression through mechanisms including Tax interaction with cellular transcription factors, chromatin-modifying enzymes, and proteins involved in the posttranscriptional control of mRNA expression (61). Selective down-regulation of gene expression by Tax BLV may result from its interaction with methyl-CpG-binding domain 2 demethylase (MBD2), with histone methyltransferase Suv39h1, and/or with HDACs, interactions already reported for Tax HTLV-1 (Ref. 62-64 and our laboratory). 13 Proteins from other oncogenic viruses have been shown to disturb the host-DNA methylation system by up-regulating DNMT expression, thereby maintaining hypermethylation of tumor suppressor genes (64 -66). Here, we show that Tax BLV downregulates DNMT expression, which leads to a lower level of methylation in the BLV promoter region and contributes to virus escape from latency.
Despite their similar sequences, the 5Ј-LTR contains the transcriptional promoter, whereas the 3Ј-LTR has a polyadenylation function. In this report, we also have shown the hypomethylated status of the 3Ј-LTR by bisulfite genomic sequencing analysis of the proviruses integrated in the three BLV-infected cell lines (YR2, L267, and L267 LTaxSN ), revealing a different methylation pattern of the 3Ј-LTR compared with the 5Ј-LTR. Such a differential level of methylation between the two LTRs has already been reported for the HTLV-I virus, where the 5Ј-LTR and the 3Ј-LTR are hypermethylated and not methylated at all, respectively, in the same population of adult T-cell leukemia cells (67)(68)(69)(70). The unmethylated state of the HTLV-I 3Ј-LTR is consistent with its transcriptionally active role in promoting the expression of the HTLV-I bZIP (HBZ) gene, which is important for the proliferation of adult T-cell leukemia-and HTLV-I-infected cells and is encoded by the minus strand of the provirus (71).
Mechanistically, we have shown that CpG methylation in the BLV 5Ј-LTR impairs in vitro and in vivo CREB/CREM/ATF-1 binding to their cognate sites in the U3 region. Furthermore, CpG methylation at the Ϫ129 position alone was sufficient to reduce by 2-fold the BLV LTR-driven luciferase expression in our transient transfection experiments. The literature has reported a few examples of such transcriptional repression by direct inhibition of factor binding by site-specific methylation, such as for the mouse Bdnf (brain-derived neurotrophic factor) gene (44) or the interferon-␥ gene (72). In addition to directly prevent the binding of transcription factors to their cognate 13 C. Van Lint, unpublished results. binding sites, CpG methylation could also act through attachment of methyl-CpG-binding proteins, which in turn recruit repressive chromatin modification enzymatic complexes.
Moreover, inhibition of CREB/CREM/ATF-1 binding impairs Tax BLV -mediated transcriptional activation of the BLV promoter (as strongly suggested by the fact that DNA methylation inhibitors and Tax BLV synergistically activated BLV gene expression in a CRE-dependent manner). This binding inhibition could also impair the recruitment of the histone acetyltransferase coactivators CREB-binding protein (CBP)/p300; their role in BLV transcriptional regulation, recently demonstrated by our laboratory (17), leads to a reduction in the level of histone acetylation, reinforcing the transcriptionally silenced state of integrated BLV proviruses.
In conclusion, our results constitute direct evidence for an interplay between the DNA methylation status of, and CREB/ CREM/ATF protein binding to, the BLV promoter, a mechanism that could contribute to true proviral latency. Together with previous data on the role of chromatin modifications in BLV expression, the present work contributes to a deeper understanding of the complex epigenetic control of BLV transcriptional latency.