Coordinate Regulation of TET2 and EBNA2 Controls the DNA Methylation State of Latent Epstein-Barr Virus

EBV genome DNA methylation correlates inversely with TET2 expression.

To investigate the relationship between EBV DNA methylation status and TET2 gene expression, we first validated that DNA methylation was elevated at key viral regulatory regions in a cell line with type I latency, but not in two related cells with type III latency (Fig. 1A). We compared the isogenic Burkitt lymphoma cell lines Mutu I and Mutu III with either type I or type III latency gene expression patterns, respectively. We also assayed an LCL cell line generated with the same EBV genome as Mutu I cells. All LCLs have type III latency. The DNA methylation state was analyzed by a methylcytosine-DNA immunoprecipitation (MeDIP) assay using real-time PCR amplimers at viral regulatory regions for EBER noncoding RNAs, the family of repeat (FR) region of OriP, and the promoter regions for LMP2A, LMP1, Cp, Qp, Zp, or Rp (Fig. 1A). As expected, we found that LMP2A, LMP1, and Cp were highly methylated in type I but not in type III latency cells. We also observed high levels of Rp methylation in type I but not type III latency. No methylation was observed at EBER, Qp, or Zp in either type I or type III. These findings confirm that key viral regulatory regions have elevated DNA methylation in type I compared to type III latency (37, 38).

We next assayed whether TET2 expression correlated with EBV latency type (Fig. 1B and C). Reverse transcription-quantitative PCR (RT-qPCR) assays revealed that in Mutu III and LCL cells, TET2 mRNA was readily detected, while this was undetectable in Mutu I cells (Fig. 1B). Western blotting revealed similar patterns for TET2 protein expression in Mutu III and LCL but not in Mutu I (Fig. 1C). We also observed that IDAX, a protein reported to bind and destabilize TET2, was expressed at similar levels in all three cell types (Fig. 1C). EBNA2 protein expression was also tested to confirm the latency type of each cell (Fig. 1C). These findings suggest that TET2 expression is coordinately regulated with EBV latency type.

TET2 mRNA and protein expression correlates with EBV type III latency patterns in BL and LCL cells.

To determine if the expression of TET2 correlated with EBV latency type more generally, we expanded our survey of Burkitt's lymphoma (BL) cells that have well-characterized type I or type III latency gene expression patterns (Fig. 2A and B). RT-qPCR revealed high levels of TET2 mRNA in Kem III and two different LCLs but not in Kem I or Akata cells, which have type I latency (Fig. 2A). Western blot analysis showed that the TET2 protein is expressed at measurable levels in type III latency cell lines but is essentially undetectable in type I latency cell lines, including EBV-negative Akata and DG75 BL cell lines (Fig. 2B). Raji, a BL-derived cell line with atypical type III latency expression features, had low TET2 mRNA but expressed an aberrantly high-molecular-weight form of TET2 (Fig. 2A and B), potentially consistent with its unusual gene expression pattern.

To determine if additional regulators of DNA methylation had a similar pattern of gene expression, we assayed other TET family members, TET1 and TET3, as well as DNA methyltransferase DNMT1, DNMT3a, and DNMT3b genes (Fig. 2C). We found that TET3, DNMT1, and DNMT3a genes were slightly elevated in LCLs compared to Mutu I, but none of these candidate genes were as enriched in LCLs relative to Mutu I as was TET2. The TET3 protein level was then examined in all the cell lines as illustrated in Fig. 2B. The Western blot analysis showed that most of the BL and LCL cells expressed similar levels of the TET3 protein, except DG75, which had a slightly higher level, and Raji, which had a slightly lower level of TET3 (Fig. 2D). These findings suggest that TET2 expression correlates better than several other DNA methylation regulators with EBV methylation status and latency type.

TET2 depletion deregulates the EBV latency gene expression program.

To determine if TET2 expression is important for EBV latency, we depleted TET2 using lentivirus short hairpin RNA (shRNA) transduction in LCLs (Fig. 3). We tested five different shRNAs and found only one (shTET2-1) that could deplete TET2 consistently. As shown in Fig. 3A, shTET2-1 had approximately 50% depletion efficiency, and a less efficient shTET2-2 had ∼38% depletion efficiency. We found that partial depletion of TET2 led to a partial reduction of latency proteins EBNA2, LMP1, and EBNA3C and a corresponding increase in lytic proteins Zta and EA-D (Fig. 3A). Similar changes in EBV gene expression were observed by RT-qPCR (Fig. 3B). TET2 depletion by shTET2-1 led to a reduction of EBNA2, EBNA3s, LMP1, and LMP2A mRNA, with a corresponding increase in Zta and Rta. Lytic cycle activation was further corroborated by qPCR of viral DNA (Fig. 3C). These findings indicate that TET2 depletion in type III latency leads to the loss of latency gene expression and an increase in lytic gene transcription and viral DNA replication.

TET2 depletion alters the EBV epigenome.

To investigate the mechanism of how TET2 depletion may alter the EBV epigenome, we next assayed cytosine hydroxymethylation (5hmc) and methylation (5mC) by DIP assay in LCLs (Fig. 4A and B). We found that EBV regulatory regions, including the EBERs, FR, Cp, LMP1p, and Zp, had elevated levels of 5hmc, which were extensively lost after shRNA depletion of TET2 (Fig. 4A). In contrast, 5mC increased after shRNA depletion of TET2 at all EBV regulatory sites, especially FR, LMP2Ap, Cp, and Zp. Qp remained mostly resistant to these DNA methylation-associated changes. Our previous study has shown that EBF1 bound to EBER and FR and that both EBF1 and RBP-jκ bound to LMP2Ap, LMP1p, and Cp in LCLs (25). Many of these EBF1 and RBP-jK binding sites were dependent on EBNA2 expression and showed epigenetic variance between type I and type III cell lines (25). We now show that TET2 depletion in an LCL led to a reduced association of EBF1 and RBP-jκ to their binding sites at LMP2Ap, LMP1p, and Cp (Fig. 4C and D). Neither EBF1 nor RBP-jκ bound significantly to the Qp or Zp region, indicating that their ChIP binding was specific. Interestingly, TET2 depletion also led to a loss of H3K4me3 at most of these viral regulatory elements (Fig. 4E). Histone H3 was not changed significantly, except for a loss at Qp and Zp (Fig. 4F). The effect of TET2 depletion on cellular genes is shown in Fig. 5.

TET2 depletion alters the cellular epigenome.

We have previously found that EBNA2-responsive promoters are regulated similarly on the viral and host genomes (25). Therefore, we next asked whether TET2 depletion had a similar effect on cellular regulatory elements that are known to be responsive to EBNA2. We found that cellular genes for interleukin-7 (IL-7), HES1, and FCER2, which were previously found to be responsive to EBNA2, were regulated in a fashion similar to that of EBV genes after shTET2 depletion in LCLs (Fig. 5). Like EBV regulatory elements (Fig. 4), IL-7, HES1, and FCER2 showed high levels of 5hmc that could be depleted by shRNA targeting of TET2 (Fig. 5A). Depletion of TET2 also led to an increase in DNA methylation at IL-7 and HES1 (Fig. 5B), partial loss of EBF1 at FCER2, and partial loss of RBP-jκ at IL-7 and FCER2 (Fig. 5C and D). A more significant loss of H3K4me3 was observed, while total H3 did not change significantly (Fig. 5E and F). The mRNA level of IL-7, HES1, or FCER2 was examined in LCL with or without shTET2 depletion (Fig. 5G). The IL-7 transcription level was slightly increased, while HES1 and FCER2 transcription levels were modestly decreased when TET2 was knocked down. Taken together, these findings indicate that TET2 is important for maintaining hydroxymethylation at viral and cellular regulatory elements and the consequent epigenetic programming important for maintaining type III latency.

Identification of EBNA2, EBF1, and RBP-jκ binding sites on the TET2 gene.

To understand how the TET2 gene may be coregulated with EBV type III latency, we analyzed existing ChIP-seq data sets from LCL (type III) and Mutu I (type I) for EBNA2 (LCL only), EBF1, RBP-jκ, H3K4me3, and H3K36me3 around the TET2 gene locus (Fig. 6). Significant H3K4me3 was observed at the TET2 transcription start site (TSS), and significant H3K36me3 was observed through the TET2 gene body in LCL relative to Mutu I, consistent with TET2 transcriptional activation in LCL relative to Mutu I. A significant EBNA2 peak (EBNA2_A) was identified at a position ∼350 kb upstream of the TET2 TSS (Fig. 6A and B). This position is also upstream of the transcription start site for IDAX, which is a divergently transcribed neighboring gene of the TET2 gene. This EBNA2 peak (EBNA2_A) overlapped a significant EBF1 peak common to both Mutu I and LCL and a smaller RBP-jκ peak that was enriched in LCL relative to Mutu I. Another significant EBNA2 peak (EBNA2_B), which is 80 kb away from the TET2 TSS, overlapped a smaller RBP-jκ peak. The two major RBP-jκ binding sites, upstream of the TSS (RBP_L) and within the TET gene body (RBP_R), also overlapped relatively small EBNA2 peaks. The two RBP-jκ binding sites were confirmed by conventional ChIP-qPCR and shown to be highly enriched in LCLs but were not detected in Mutu I cells (Fig. 6C). Taken together, these findings indicate that EBNA2 can colocalize with EBF1 or RBP-jκ at multiple sites throughout the TET2 gene locus and that RBP-jκ binding to TET2 is regulated in a latency type-dependent manner.

Epigenetic regulation of the TET2 gene locus is similar to that of EBV epigenotypes.

We next investigated the DNA methylation status of the TET2 gene locus in cell lines with either type I or type III EBV latency (Fig. 6D). We found that the regions overlapping the major transcription start site for TET2 (TSS_1 and TSS_2) were highly methylated in Mutu I (type I) cells but not detectably methylated in Mutu III or LCL (type III cells). We did not detect high levels of methylation at the RBP-jκ sites (RBP_L or RBP_R) in either type I or type III cells. We next tested whether DNA methylation at the TSS was associated with transcriptional silencing in type I latency (Fig. 6E). Mutu I cells treated with 5′ azacytidine showed a large (>10-fold) increase in TET2 mRNA relative to control-treated samples by RT-PCR (Fig. 6E). This increase in transcription correlated with an increase in histone H3K4me3 at the TET2 TSS (Fig. 6F). The levels of TET2 mRNA expression and H3K4me3 did not reach those of LCLs, suggesting that additional factors other than DNA demethylation drive TET2 transcription. Nevertheless, 5′ azacytidine had no effect on TET2 mRNA or H3K4me3 methylation in LCLs, further confirming the hypomethylation status of TET2 TSS in LCLs (Fig. 6E, right panel, and F).

EBNA2 is required for TET2 expression.

The contribution of EBNA2 in regulating TET2 expression was tested using the EREB2.5 LCLs with conditional expression of EBNA2 (Fig. 7A and B). In the absence of estradiol (E2), EBNA2 is sequestered in the cytoplasm and fails to bind and activate target genes in the nucleus (39). Western blot analysis revealed that the removal of E2 led to a rapid loss of the TET2 protein (Fig. 7A). EBNA2 protein levels do not decrease as rapidly, but previous studies indicate that its nuclear localization and transcription function is rapidly curtailed. Inactivation of EBNA2 led to a corresponding decrease in TET2 mRNA to virtually undetectable levels by 72 h (Fig. 7B). These changes corresponded to a complete loss of EBNA2 binding to the TET2 gene loci by ChIP assay (Fig. 7C) and an ∼2-fold decrease of RBP-jκ binding to its binding sites (Fig. 7D). These findings indicate that EBNA2 is required for TET2 mRNA expression, and this function likely depends on EBNA2 binding to the TET2 gene loci.

RBP-jκ is required for TET2 expression in type III latently infected B lymphocytes.

To better understand how EBNA2 may regulate TET2 in type III latency, we asked whether TET2 expression was dependent on the EBNA2-interacting partner RBP-jκ. RBP-jκ binding sites were identified in the TET2 gene locus by ChIP-seq (Fig. 6A and B) and validated by ChIP-qPCR (Fig. 6C). To determine whether RBP-jκ was important for TET2 expression in LCLs, we depleted RBP-jκ using three different shRNA-targeting sequences (Fig. 7E and F). In all three samples, TET2 protein levels were reduced, correlating with the depletion of RBP-jκ (Fig. 7E). In addition, the depletion of RBP-jκ correlated with the loss of TET2 mRNA expression (Fig. 7F). These findings indicate that RBP-jκ is functionally important for TET2 mRNA expression in LCLs.

TET2 colocalizes and coimmunoprecipitates with EBNA2.

To better understand how TET2 may coregulate EBNA2-dependent genes, we tested its ability to associate with viral and cellular chromatin sites that are also known to be bound and regulated by EBNA2 (Fig. 8A and B). ChIP-qPCR with TET2 antibody revealed that TET2 is enriched in LCLs at EBER, LMP2Ap, LMP1p, and Qp relative to Mutu I (Fig. 8A). Similarly, TET2 was enriched at cellular IL-7 and FCER2 genes, especially in LCLs (Fig. 8B). These findings indicate that TET2 co-occupies sites bound and regulated by EBNA2.

We next tested whether the EBNA2 protein interacts with the TET2 protein by coimmunoprecipitation (co-IP) (Fig. 8C to E). Immunoprecipitation of TET2 in LCL or Mutu III (Fig. 8C) or 293T cells transfected with either FLAG-tagged EBNA2 (Fig. 8D) or Myc-tagged EBNA2 (Fig. 8E) revealed a strong co-IP of EBNA2 with antibody to EBNA2 (Fig. 8C), FLAG epitope (Fig. 8D), or Myc epitope (Fig. 8E). We were unable to coprecipitate TET2 with EBNA2 antibody IP in LCL or Mutu III cells. However, we observed a weak co-IP of TET2 in 293T cells with FLAG-EBNA2 and Myc-EBNA2, suggesting that this interaction may be sensitive to EBNA2 antibodies. Taken together, these studies suggest that an interaction between EBNA2 and TET2 may direct TET2 to RBP-jκ target sites on the EBV and cellular genomes (Fig. 9).

Cells.

EBV-positive BL cell lines Mutu I, Mutu III, Kem I, and Kem III were obtained from Jeffrey Sample, Penn State University Hershey Medical School, Hershey, PA. DG75 and Raji were obtained from ATCC. Akata cell lines with or without EBV were obtained from Elena Mattia, Sapienza University of Rome, Rome, Italy. LCL cell lines were generated from human B cells (provided by the Wistar Institute Phlebotomy Core) transformed with Mutu I virus. LCLs, DG75, Raji, Akata, Mutu I, Mutu III, Kem I, and Kem III cells were maintained in RPMI containing 12% fetal bovine serum (FBS) and antibiotics (penicillin and streptomycin). EREB 2.5, a lymphoblastoid cell line expressing the estrogen-inducible EBNA2-estrogen receptor (ER) fusion protein (55), was maintained in RPMI containing 12% FBS, antibiotics (penicillin and streptomycin), and 1 μM estradiol. 293T cells were obtained from ATCC and grown in Dulbecco's modified Eagle medium (DMEM) with 10% FBS and antibiotics.

Lentiviral transduction.

pLKO.1 vector-based shRNA constructs for TET2 (shTET2-1, TRCN0000144344; shTET2-2, TRCN0000122172) or RBPjκ (sh RBPjκ-1, TRCN0000016203; shRBPjκ-2, TRCN0000016204; shRBPjκ-3, TRCN0000016205) were obtained from Open Biosystems. Control shRNAs (shControl or shCtrl) were generated in the pLKO.1 vector with the target sequence 5′-TTATCGCGCATATCACGCG-3′. Lentiviruses were produced by the use of envelope and packaging vectors pMD2.G and pSPAX2 as described previously (56). Mutu I or LCL cells were infected with lentiviruses carrying pLKO.1-puro vectors by spin infection at 450 × g for 90 min at room temperature. The cell pellets were resuspended and incubated in fresh RPMI medium and then treated with 2.5 μg/ml puromycin at 48 h after the infection. The RPMI medium with 2.5 μg/ml puromycin was replaced every 2 to 3 days. The cells were collected after 8 days of puromycin selection and then subjected to further analyses, as indicated.

MeDIP and hMeDIP assays.

Genomic DNA was purified using the Wizard Genomic DNA purification kit (A1120; Promega) and then subjected to methylcytosine-DNA immunoprecipitation (MeDIP) or hydroxymethylcytosine-DNA-IP (hMeDIP) assays. The MeDIP or hMeDIP assays were performed using the MagMeDIP kit (C02010021; Diagenode) or the hMeDIP kit (C02010031; Diagenode). qPCR assays were then performed as described previously (57). Quantification of precipitated DNA was determined using real-time PCR and the delta C_t (cycle threshold) method for relative quantitation (ABI 7900HT Fast real-time PCR system). Primers for DIP assays are listed in Table S2 in the supplemental material.

RNA extraction and quantitative RT-PCR.

RNA was isolated from 2 × 10⁶ cells using the RNeasy kit (Qiagen) and then further treated with DNase I by using the DNase treatment and removal kit (Ambion). RT-PCR was performed as previously described (58). Real-time PCR was performed with a SYBR green probe in an ABI Prism 7900, and the delta C_T or delta delta C_T method was used for relative quantitation. Primer sequences for RT-PCR are listed in Table S1 in the supplemental material.

Coimmunoprecipitation.

Co-IP was performed as described previously (59). Rabbit IgG (Santa Cruz Biotechnology sc-2027), rabbit anti-TET2 (EMD Millipore ABE364), Myc (Cell Signaling 2278), mouse IgG (Santa Cruz Biotechnology sc-2025), or mouse anti-FLAG M2 (Sigma F1804) was used in co-IP.

Antibodies used for Western blotting.

Antibodies used for Western blotting were rabbit polyclonal RBP-jκ (Abcam AB25949), TET2 (Proteintech 21207-1-AP), TET3 (GeneTex GTX121453), GAPDH (glyceraldehyde-3-phosphate dehydrogenase; Cell Signaling catalog number 2118), Myc (Cell Signaling catalog number 2278); rabbit polyclonal anti-Zta was generated from full-length Zta at Pocono Rabbit Farms; also used were rat monoclonal anti-EBNA2 (EMD Millipore MABE8), mouse monoclonal anti-TET2 (EMD Millipore MABE462), LMP1 (Dako M0897), EA-D (Abcam ab49668), sheep polyclonal anti-EBNA3C (Exalpha F125P), actin-peroxidase antibody (Sigma A3854), FLAG-peroxidase antibody (Sigma A8592).

ChIP-qPCR assays.

ChIP-qPCR assays were performed as described previously (57). Quantification of precipitated DNA was determined using real-time PCR and the delta C_T method for relative quantitation (ABI 7900HT Fast real-time PCR system). Rabbit IgG (Santa Cruz Biotechnology sc-2027), rabbit anti-EBF1 (EMD Millipore AB10523), RBP-jκ (Abcam AB25949), histone H3K4me3 (EMD Millipore 07-473), histone H3 (EMD Millipore 07-690), and TET2 (Proteintech 21207-1-AP) were used in ChIP assays. Mouse IgG (Santa Cruz Biotechnology sc-2025) and mouse anti-EBNA2 (gift from Paul Farrell, UK) antibodies were also used in ChIP assays. Primers for ChIP assays are listed in Table S2 in the supplemental material.

ChIP-seq and genomic data processing.

ChIP-seq and genome data processing of histone H3K36me3 were performed as described previously (20). Rabbit anti-histone H3K36me3 (Abcam ab9050) was used in ChIP-seq. The EBNA2, RBP-jκ, EBF1, and H3K4me3 genome browser tracks were generated previously (20) and are available at NCBI GEO data sets.