Cell Cycle Association of the Retinoblastoma Protein Rb and the Histone Demethylase LSD1 with the Epstein-Barr Virus Latency Promoter Cp

Abstract

The Epstein-Barr virus C promoter (Cp) regulates the major multicistronic transcript encoding the EBNA-LP, 1, 2, and 3 genes required for B-cell proliferation during latency. The growth-transforming potential of these viral genes suggests that they must be tightly regulated with the host cell cycle and differentiation process. To better understand Cp regulation, we used DNA affinity purification to identify cellular and viral proteins that bind to Cp in latently infected cells. Several previously unknown factors were identified, including the cell cycle regulatory proteins E2F1 and Rb. E2F1 bound to a specific site in Cp located in the core Cp region 3′ of the known EBNA2-responsive RBP-Jk (CSL, CBF1) binding site. The histone H3 K4 demethylase LSD1 (BCC110) was also identified by DNA affinity and was shown to form a stable complex with Rb. Coimmunoprecipitation assays demonstrated that E2F1, Rb, and LSD1 bind to Cp in a cell cycle-dependent manner. Rb and LSD1 binding to Cp increased after the S phase, corresponding to a decrease in histone H3 K4 methylation and Cp transcription. Coimmunoprecipitation and immunofluorescence assays reveal that LSD1 interacts with Rb. Surprisingly, LSD1 did not coimmunoprecipitate with E2F1, suggesting that it associates with Rb independently of E2F1. Depletion of LSD1 by small interfering RNAs inhibited Cp basal transcription levels, and overexpression of LSD1 altered the cell cycle profile in p53-positive (p53⁺), but not p53-negative (p53⁻), HCT cells. These findings indicate that Cp is a cell cycle-regulated promoter that is under the control of Rb and the histone demethylase LSD1 in multiple latency types.

Epstein-Barr virus (EBV) is a human gamma-1 herpesvirus that establishes a lifelong latency in over 90% of the world's population (26, 37). During latency, the virus exists predominantly as a chromatin-associated, multicopy episome in the nuclei of resting B lymphocytes (47, 52). Latent infection is associated with several malignancies, including Burkitt's lymphoma, Hodgkin's disease, nasopharyngeal carcinoma, and lymphoproliferative disorders in the immunosuppressed. Different viral-transcription patterns can be observed in each of these malignancies (57). Regulation of the EBV C promoter, or Cp, is important to the biology of EBV because it is the key control point distinguishing latency types and the expression of the viral oncogenes EBNA1, 2, 3A, 3B, 3C, and LP (4). Several cellular transcription factors, including CBF1 (also referred to as CSL or RBP-Jκ,), CBF2 (or Auf1), NF-Y, C/EBP, Sp1, and Egr-1, bind within Cp to regulate its activity (5, 17, 22, 23, 34). In addition, the viral protein products of Cp, EBNA-LP, EBNA2, and EBNA3A-C act in conjunction with various cellular factors to autoregulate Cp transcription (22, 30, 38, 59). Deletion mapping of the Cp region showed that sequences −433 to −245 upstream of the Cp initiation site (particularly around −370) are important for the EBNA2/CBF1/CBF2 response, −119 to −112 for C/EBP, −99 to −91 for Sp1/Egr-1, and −71 to −63 for NF-Y (16, 25, 34, 50).

Cell cycle regulation of transcription is required for the temporal coordination of DNA synthesis and cellular division. Cell cycle-regulated genes are commonly controlled by the E2F family of transcription factors and the Rb tumor suppressor protein (13, 19, 31, 49). Hyperphosphorylation of Rb by G₁/S cdk-cyclins disrupts its interaction with E2F and its ability to repress transcription (20, 27). Rb can inhibit transcription through its association with multiple corepressor complexes that contain histone deacetylases and other chromatin-modifying enzymes (7, 8, 31). Rb-mediated transcription repression can lead to epigenetically stable heterochromatin formation through its interactions with histone H3 K9 methylases (1, 33) as well as with DNA cytosine methylase complexes (36, 39). It remains unclear how E2F and Rb select a particular corepressor complex and whether the repression is dynamically reversed every cell cycle or epigenetically stable for multiple cell division cycles.

The recent discovery of numerous histone demethylases that function in corepressor and coactivator complexes indicates that histone methylation is a dynamic modification (28, 44, 51). Several histone demethylases have been identified, including the flavin adenine dinucleotide-binding monoamine oxidases (32, 46) and those of the JmjC domain-containing proteins (53). LSD1 (also known as BHC110 and KIAA0601) is a flavin adenine dinucleotide-binding monoamine oxidase with histone H3 lysine 4- and lysine 9-specific demethylase activities (15, 29). The lack of LSD1 in mice causes embryonic death, and conditional null alleles show defects in transcription repression as well as activation of some target genes (55). LSD1 has been isolated in several corepressor complexes, including the CO-REST complex required for repression of neural genes in nonneural tissue (29), the CtBP corepressor associated with Notch signaling (54), the ZEB1 corepressor complex involved in pituitary-cell specification (54), and the derepression of androgen receptor-dependent transcription (32, 55).

In this study, we used DNA affinity chromatography coupled with mass spectrometry to determine the factors that bind Cp. Based on the results of DNA affinity, we identified an unanticipated cell cycle dependence on transcription and histone modifications at the Cp control region and the potential regulation of Cp by the LSD1 histone demethylase.

MATERIALS AND METHODS

Cell lines and antibodies.

MutuI is a type I latency B-cell line derived from Burkitt's lymphoma. Raji (ATCC) is a type III latency B-cell line derived from Burkitt's lymphoma, and LCL3456-EBV is a type III latency B-cell line derived from primary lymphoblasts transformed with EBV strain B95-8. These cell lines were maintained in RPMI medium supplemented with 10% fetal bovine serum (FBS) and glutamine, penicillin, and streptomycin sulfate (G-P-S). HeLa, 293, and IMR90 cells were maintained in Dulbecco modified Eagle medium supplemented with 10% FBS and G-P-S. The HCT 116 sublines 40-16 and 379.2 are colon cancer-derived cells, where 40-16 is p53^+/+ and 379.2 is p53^−/−. These cells were maintained in McCoy's 5A medium supplemented with 10% FBS plus G-P-S.

The rabbit polyclonal antibodies used were anti-E2F1 (Santa Cruz), anti-Rb (Santa Cruz), anti-LSD1 (Abcam), anti-Orc2 (Pharmingen), anti-dimethyl K4 H3 (Upstate), anti-trimethyl K9 H3 (Upstate), anti-acetyl H3 (Upstate), anti-acetyl H4 (Upstate), anti-H3 (Upstate), and control rabbit immunoglobulin G (IgG) (Santa Cruz). Mouse monoclonal anti-EBNA2 was obtained from DakoCytomation.

Plasmids and primers.

N1258 (a Cp-luciferase [Cp-Luc] construct) was a kind gift from P. Ling, Baylor Medical School, Houston, TX. N1179 (pGEX-DP1) and N1180 (pGEX-E2F1) were kind gifts from S. Chellappan, University of South Florida, Tampa Bay, FL. All primer sequences used are available upon request.

DNA affinity chromatography.

DNA affinity chromatography was performed as previously described (2).

EMSAs.

Electromobility shift assays (EMSAs) were performed as described previously (10).

Coimmunoprecipitation assays.

Soluble nuclear extract fractions were obtained from Raji cells via the Dignam extraction method. The soluble nuclear fraction was then dialyzed overnight in NET buffer (150 mM NaCl, 1 mM EDTA, 10 mM Tris [pH 7.5]) containing 10% glycerol. After dialysis, 1-ml aliquots of the nuclear fraction were placed in Eppendorf tubes. Antibodies to rabbit IgG, Rb, LSD1, and E2F1 were added to each aliquot, and the tubes were rotated at 4°C overnight. Next, 100 μl of 50% protein A slurry was added to each tube and incubated at 4°C for 2 h. Protein A beads were then pelleted by centrifugation at 3,000 rpm on a tabletop for 2 min. The pellets were then washed six times with NET buffer. After the final wash, the pellets were resuspended in 50 μl of 2× Laemmli buffer and boiled at 95°C for 10 min, and the solubilized proteins were assayed by Western blotting.

Indirect immunofluorescence.

IMR90 cells grown on coverslips were washed twice with cold phosphate buffered-saline (PBS), preextracted in buffer {0.5% Triton X-100, 10 mM PIPES [piperazine-N,N′-bis(2-ethanesulfonic acid)] [pH 7], 100 mM NaCl, 300 mM sucrose, 3 mM MgCl₂} for 5 min on ice, and fixed with 3.7% formaldehyde at room temperature. After being washed twice with PBS, the cells were permeabilized in PBS containing 0.5% NP-40 for 10 min at room temperature and blocked by being incubated in PBG (0.2%, wt/vol, cold water fish gelatin [Sigma] and 0.5%, wt/vol, bovine serum albumin [BSA] [Sigma] in PBS) for at least 30 min. The coverslips were incubated with primary antibodies diluted in PBG for 1 to 2 h at room temperature. Monoclonal anti-Rb was used at a dilution of 1:600, and rabbit anti-LSD1 antibody was diluted at 1:300. The coverslips were then washed three times in PBG for 5 min each and incubated with fluorochrome-conjugated secondary antibodies (Alexa Fluor) diluted in PBG (1:600) for 1 h at room temperature. After being washed three times for 3 to 5 min each time in PBG, the coverslips were stained with 0.5 mg/ml of Hoechst 33258 (Sigma) diluted in PBG and mounted on Vectashield medium (Vector Laboratories) on a microscope slide. Images were taken with a 100× lens objective on a Nikon E600 upright microscope (Nikon Instruments, Inc., Melville, NY) using Image-Pro Plus software (Media Cybernetics, Silver Spring, MD) and Adobe Photoshop, version 7.0, for image processing.

ChIP assays.

For chromatin immunoprecipitation (ChIP) assays, we used the protocol provided by Upstate Biotechnology with minor modifications as previously described (14) and additional modifications as follows. DNA was sonicated to between 200- to 350-bp fragments on a Diagenode Bioruptor according to the manufacturer's protocol, and real-time PCR was performed with a Sybr green probe in an ABI Prism 7000, using 1/100th to 1/2,500th of the ChIP DNA according to the manufacturer's specified parameters.

Quantitative RT-PCR.

Reverse transcriptase PCR (RT-PCR) was done as previously described (9). Real-time PCR was performed with Sybr green probes in an ABI Prism 7000 using 1/100th to 1/2500th of the cDNA according to the manufacturer's specified parameters. EBNA2 transcripts were normalized to cellular β-actin or to the 36BD4 gene, which encodes the acidic ribosomal phosphoprotein PO (6).

BrdU incorporation assay.

To determine the effect of LSD1 on the cell cycle, HCT 379.2 and 40-16 cells were subjected to overexpression of N799 (a FLAG control vector) and N1222 (a FLAG-LSD1 vector). After 48 h, the cells were pulsed with 30 μM of BrdU for 15 min., washed twice with PBS, trypsinized, and harvested. The cell pellets were resuspended in 300 μl of ice-cold PBS and fixed by the slow, dropwise addition of 700 μl of ice-cold ethanol. The cells were then centrifuged, washed once with PBS, resuspended in 1 ml of PBS containing 0.5 mg/ml of RNase A, and incubated at 37°C for 30 min. The cells were then washed with 5 ml of PBS and pelleted. The cell pellets were then denatured by the addition (dropwise) of 1 ml of a 2N HCl/0.5% Triton X-100 solution. The acid was removed by washing the cells once with 1 ml of PBS and once with 1 ml of PBS plus 0.2% Tween 20. The cells were then stained with 20 μl of monoclonal anti-BrdU (BD Biosciences) in 200 μl PBS containing 1% BSA at 4°C overnight, washed twice with 1 ml of PBS-T (PBS plus 0.1% BSA plus 0.2% Tween 20), and stained with 1 μl of fluorescein isothiocyanate-conjugated anti-mouse IgG (Vector Laboratories) in 200 μl PBS containing 1% BSA at room temperature for 30 min in the dark. The cells were then washed 3 times with 1 ml of PBS-T and three times with 1 ml of PBS (a 5-min incubation with rotation at 4°C per wash). Finally, the cells were resuspended in 500 μl of PBS containing 5 μg/ml of propidium iodide and analyzed by flow cytometry.

RESULTS

The Cp promoter regulates transcription of the multicistronic RNA encoding the LP and EBNA proteins in proliferating lymphoblastoid cells during EBV latency (Fig. 1A). To further investigate the mechanism of Cp regulation, we isolated cellular factors that bind Cp by DNA affinity purification (Fig. 1B). Raji cell nuclear extracts were incubated serially with streptavidin magnetic beads containing biotinylated DNA from control plasmid pBKS (Stratagene) followed by an ∼500-bp region of Cp. Several enriched bands were detected by colloidal blue staining and subjected to identification by liquid chromatography tandem mass spectrometry. Among the identified proteins were the previously characterized CBF1, Sp1, and EBNA2 proteins. We also found E2F1, Rb, LSD1, MEF2D, and Mi-2α. The specific binding of these proteins to Cp was confirmed by Western blot analysis (Fig. 1C). Several other proteins that were not identified by liquid chromatography tandem mass spectrometry, like HDAC1, HDAC2, and ORC2, were used as negative controls for the specificity of DNA affinity purification. To determine if these proteins were binding to the same DNA element, we compared two regions of the Cp promoter known to have transcription regulatory function (Fig. 1D). The well-characterized EBNA2 response element bound CBF1 and EBNA2 as expected. Interestingly, the region immediately 3′ of CBF1 (probe B; EBV coordinates 10971 to 11070), important for the constitutive activity of Cp, bound E2F1, Rb, LSD1, MEF2D, and Sp1 (Fig. 1D).

FIG. 1.

E2F1, Rb, and LSD1 bind to the EBV latency promoter Cp. (A) Schematic of the EBV genome, the Cp-regulated latency transcript, and the promoter regulatory region. Probes used for DNA affinity and EMSA assays are indicated by letters. Probes A and B are biotinylated DNA fragments spanning EBV coordinates 10904 to 10990 and 10971 to 11077, respectively. (B) Colloidal staining of DNA affinity pull-down using Raji nuclear extract. The first lane contains the input, which represents 10% of the total nuclear proteins used for DNA affinity. The second lane is BKS, which is a DNA affinity purification using a random, biotinylated DNA fragment of pBluescript KS+. The third lane is Cp, which is a DNA affinity purification using biotinylated DNA spanning EBV coordinates 10904 to 11077. The lines next to the Cp lane indicate gel fragments that were cut and analyzed by mass spectrophotometry. (C) Western blot of protein pull-down by DNA affinity probed with different antibodies to proteins that were identified by mass spectrophotometry for Fig. 1A. (D) Western blot of protein pull-down by DNA affinity using probe A and probe B DNA fragments (see panel A).

To identify the precise location of the E2F binding site within Cp, we used EMSA with three DNA oligonucleotide probes spanning the sequence 10971 to 11070 (Fig. 2). Since E2F1 typically binds as a heterodimer with DP1, we expressed and purified both E2F1 and DP1 as glutathione S-transferase (GST) fusion proteins from Escherichia coli. The recombinant GST-E2F1 was tested alone or with its heterodimer partner, GST-DP1, for binding to probe A (sequence 10971 to 11010), probe B (sequence 11011 to 11040), or probe C (sequence 11041 to 11070). We found that GST-E2F1 bound to probe B but not to probe A or C and was only slightly stimulated by the addition of GST-DP1 (Fig. 2A). The ability of GST-E2F1 to bind in the absence of DP1 is probably a consequence of the homodimerization provided by the GST fusion. The specificity of binding for probe B was further demonstrated by competition assays, which demonstrates that the specific probe B competitor (C) but not a nonspecific competitor was capable of inhibiting DNA binding (Fig. 2B). Inspection of the sequence in probe B revealed that a noncanonical E2F binding site (GCGGGAGA) is located at position 11030 in the EBV genome (Fig. 2C).

FIG. 2.

Identification of E2F1 binding sites in Cp. (A) EMSA was used to analyze DNA binding of purified GST-E2F1 with or without GST-DSP1 to DNA probe A, B, or C, as indicated above each lane. (B) EMSA with purified GST-E2F1, GST-DSP1, and radiolabeled probe B with cold probe B competitor (C) or cold probe C nonspecific competitor (NC). (C) DNA sequences of probes A, B, and C used for EMSA for panels A and B and alignment of the consensus E2F binding site and the E2F binding site in Cp.

To determine if any of the proteins identified by DNA affinity chromatography bound to Cp in vivo, we used a ChIP assay (Fig. 3). Since Cp has been shown to have differential histone H3 K4 methylation patterns in different latency types, we were particularly interested in investigating the interaction properties of the histone H3 K4 demethylase LSD1. We compared the binding of E2F1, Rb, and LSD1 at Cp and control OriLyt regions of EBV in three different cell types. In general, we found that E2F1, Rb, and LSD1 bound to Cp but not to OriLyt in all three cell types. Control IgG did not precipitate Cp DNA, indicating that the E2F1, Rb, and LSD1 antibodies were enriched relative to a negative-control antibody. Some differences in the relative binding of these factors in the different cell types are worth noting. E2F1 and LSD1 were enriched relative to Rb in LCL3456 cells, whereas Rb was enriched relative to E2F1 and LSD1 in the Burkitt's lymphoma lines MutuI and Raji. This may reflect the fact that Cp activity is highest in LCL3456, moderate in Raji, and absent in MutuI cells. These findings indicate that E2F1, Rb, and LSD1 bind to Cp in vivo in different latency types, but that their relative levels may vary according to the transcription activity of Cp.

FIG. 3.

E2F, Rb, and LSD1 bind Cp in vivo. ChIP was used to detect protein binding to Cp in vivo, using antibodies to E2F1, Rb, LSD1, or the IgG control. ChIP DNA was analyzed by real-time PCR at the Cp (black bars) or OriLyt (open bars) site in MutuI (A), LCL3456 (B), and Raji (C) cells.

Since E2F and Rb proteins are known to control cell cycle transcription, we asked whether Cp was regulated in a cell cycle-dependent manner (Fig. 4). We found that EBNA2 transcription was highly elevated in Raji cells relative to that in MutuI cells (Fig. 4A). We therefore asked whether EBNA2 transcription was cell cycle regulated in Raji cells. Raji cells were synchronized by double thymidine blocking and release. EBNA2 mRNA levels were compared to those of cellular β-actin, using RT-PCR. We found that EBNA2 mRNA levels increased ∼2.5-fold in S phase relative to those in G₁ and then returned to G₁ levels by the G₂/M phase (Fig. 4A).

FIG. 4.

Cell cycle-regulated transcription, Rb-LSD1 binding, and histone modifications at Cp. (A) RT-PCR of EBNA2 across the cell cycle. Raji cells were collected at different time points following thymidine release, and the total mRNA was harvested for each time point. (B) ChIP assay of Rb, LSD1, and E2F1 binding at Cp across the cell cycle. The color legend is given in panel A for different points across the cell cycle. Raji cells were treated with thymidine and released. Cells were collected at different time points after thymidine release (0 h, 1 h, 2 h, 4 h, 8 h, and 12 h) for ChIP analysis. An aliquot was removed, stained with propidium iodine, and analyzed by FACS. (C) ChIP assay of acetylated histone H3, acetylated histone H4, dimethyl K4 histone H3, and trimethyl K9 histone H3 binding at Cp across the cell cycle. See the color legend in panel A for different points across the cell cycle. (D) The same process as for panel C was used, except that ChIP DNA was analyzed for the Wp sequence. (E) FACS analysis of cell cycle distribution for the experiments shown in panels A to C. Asyn, asynchronous cells.

We next examined whether protein interactions and histone modifications were cell cycle regulated at Cp. The interactions of E2F1, Rb, and LSD1 with Cp were examined in cell cycle-synchronized Raji cells, using a ChIP assay (Fig. 4B). We found that Rb protein was associated with Cp in G₁ and G₂/M but was significantly reduced in S phase (Fig. 4B). LSD1 binding had a very similar cell cycle profile to that of Rb. E2F1 bound Cp most detectably in G₁/S and in early and mid-S phase. The low levels of E2F1 binding suggest that other members of the E2F family may also be exchanged with E2F1 during the cell cycle, as has been reported at several cellular promoters. The cell cycle association of Rb and LSD1 prompted us to examine changes in histone acetylation and methylation statuses across the cell cycle (Fig. 4C). We found that histone H3 and H4 acetylation peaked in early S phase, consistent with the increased transcription activity of Cp. We also found that histone H3 methyl K4 was highly elevated in early and mid-S phase at Cp (Fig. 4C). In contrast, H3 methyl K9 was not significantly enriched at Cp, although a slight elevation was observed at late S phase (Fig. 4C). A similar cell cycle pattern of histone modifications was also observed at Wp, although the H3 methyl K4 signal was significantly less than that observed at Cp (Fig. 4D).

Since the Raji cell is a type III Burkitt cell with a poorly characterized Cp initiation site, we assayed the cell cycle transcription of EBNA2 and chromatin binding patterns at Cp in LCL3456 cells (Fig. 5). Lymphoblastoid cell lines (LCLs) cannot be cell cycle synchronized using a double thymidine block as was used for Raji cells (Fig. 4). Consequently, we used centrifugal elutriation to fractionate cells at different stages of the cell cycle (Fig. 5A) (60). Elutriated fractions were assayed for cell cycle distribution by propidium iodide staining and fluorescence-activated cell sorter (FACS) analysis (Fig. 5A). Elutriated fractions were then analyzed for EBNA2 mRNA expression using quantitative RT-PCR (Fig. 5B). We found that EBNA2 mRNA increased approximately fivefold in G₂/M (fraction 32) relative to G₁ (fractions 12 to 16). This indicates that EBNA2 mRNA is cell cycle regulated in LCL cells, similar to that observed in Raji. We used a ChIP assay to measure Rb, LSD1, and E2F1 binding to Cp at different cell cycle stages (Fig. 5C). We found that Rb and LSD1 binding to Cp was low in G₁/S but increased in G₂/M. In contrast, E2F1 binding peaked at early S phase (fraction 20) and then decreased in G₂/M (fraction 28 to 32). The extent of Rb and LSD1 binding to Cp was notably lower in LCL cells relative to that observed in Raji cells, while E2F1 was notably higher in LCL cells (compare Fig. 5C with 4B). The histone modifications at Cp were also examined in LCLs. Acetylated histone H3 and H3 methyl K4 were most enriched in mid-S phase (fraction 24), while H3 methyl K9 was most enriched in G₂/M (fraction 32) (Fig. 5D). The data suggest that Cp histone modifications and EBNA2 transcription are cell cycle regulated in LCLs similar to what was observed in Raji cells, although Rb and LSD1 bound to a lesser extent in LCLs, where Cp may be more active. However, in both cell types the enrichment of Rb and LSD1 in G₂/M corresponded to a decrease in H3 methyl K4, suggesting that Rb and LSD1 are in part responsible for the cell cycle changes in histone H3 methylation at Cp.

FIG. 5.

Cell cycle regulation of Cp transcription and chromatin in EBV-positive LCLs. (A) LCL3456 was subjected to centrifugal elutriation to fractionate cells enriched in specific stages of the cell cycle. Fractions 12, 16, 20, 24, 28, 32, and 40 were analyzed by FACS analysis after propidium iodide staining. (B) RT-PCR analysis of EBNA2 mRNA abundance was measured by real-time PCR and normalized to the cellular gene 36B4. (C) ChIP assay with antibodies to Rb, LSD1 (left), and E2F1 (right) quantified for different cell cycle fractions at Cp. (D) ChIP assay with antibodies to Ac-H3, MeH3-K4 (left), and MeH3-K9 (right), essentially as described for panel C. Asyn, asynchronous cells.

We next explored the possibility that LSD1 associated with Cp through an interaction with Rb and/or E2F proteins. We first tested the ability of LSD1 to interact with Rb by immunoprecipitation (IP) from Raji cell nuclear extracts (Fig. 6). We found that IP with Rb antibody pulled down significant amounts of LSD1 (Fig. 6A). Similarly, IP with LSD1 pulled down detectable amounts of Rb (Fig. 6B). These findings indicate that Rb and LSD1 form a stable complex in vitro. To determine if these proteins interact in vivo, we used indirect immunofluorescence assays in IMR90 cells (Fig. 6C). Immunofluorescence experiments showed a strong colocalization of 30% of Rb foci with LSD1 and 22% of LSD1 foci with Rb. These findings suggest that Rb and LSD1 can be detected in a common protein complex that colocalizes in punctate foci in the nucleus.

FIG. 6.

Coimmunoprecipitation and colocalization of LSD1 and Rb. (A and B) Coimmunoprecipitation of LSD1 (A) and Rb (B). Panel A is a Western blot probed with α-LSD1 and has an input lane containing 5% IP material and three IP lanes for rabbit IgG, rabbit α-Rb, and rabbit α-LSD1. The membrane was stripped and probed with α-Rb for panel B. (C) Immunofluorescent colocalization of LSD1 and Rb in IMR90 cells. Panel a shows IMR90 cells stained with mouse antibody to Rb. Panel b shows IMR90 cells stained with rabbit antibody to LSD1. Panel c shows a merged picture of the top two panels. The fourth panel shows an enlarged section of panel c. α, anti-; IB, immunoblot. Magnification, ×100.

Cell cycle regulation of transcription by Rb depends, in part, on changing protein interactions during the cell cycle. To determine if the Rb-LSD1 interaction was cell cycle regulated, we analyzed the Rb-LDS1 interaction at different stages of the cell cycle (Fig. 7A to C). The Rb-LSD1 interaction was monitored by IP from Raji cells synchronized in G₁/S, mid-S, or G₂/M. We found that IP with Rb antibody pulled down LSD1 in G₁/S and to a lesser extent in G₂/M (Fig. 7A). However, LSD1 was not detected in the IP from mid-S-phase-arrested cells (Fig. 7A). The reciprocal experiments show that Rb coprecipitates with LSD1 in G₁/S and G₂/M, while S phase IP pulled down a faster-migrating form of Rb (Fig. 7B). Presumably, this faster-migrating form was not immunoprecipitated by the Rb antibody and therefore did not pull down LSD1. These results suggest that the LSD1-Rb complex is disrupted, or significantly altered, in mid-S phase.

FIG. 7.

Cell cycle interaction between LSD1 and Rb. (A to C) LSD1, Rb, or IgG was immunoprecipitated at different stages across the cell cycle and probed with LSD1 (A)- or Rb (B)-specific antibodies. The FACS profile of stage-specific extracts is indicated in panel C. (D) Immunoprecipitations of Rb, LSD1, E2F1, or control IgG were probed with antibodies to Rb, LSD1, E2F1, or anti-methyl lysine (methyl-K), as indicated below each panel. α, anti-; Asyn, asynchronous cells; IB, immunoblot.

Rb is known to associate with E2F proteins, and it is likely that E2F is required for the specific recruitment of Rb and LSD1 to Cp. We therefore tested the interactions between E2F1, Rb, and LSD1 in asynchronous cells (Fig. 7C). We found that Rb was present in the IPs of Rb, LSD1, and E2F1, as expected (Fig. 7C, top panel). We found that LSD1 was present in the IPs of Rb and LSD1 but was not readily detected in the IP of E2F1 (Fig. 7C, second panel). E2F1 was found in the IP of Rb and E2F1 but was not detected in the IP of LSD1 (Fig. 7C, third panel). These findings suggest that Rb and LSD1 are in a complex that is distinct from the E2F1 complex.

Since LSD1 is a lysine demethylase, we tested whether any of the LSD1-associated proteins were reactive to a methyl lysine antibody (Fig. 7C, lower panel). Interestingly, we found that Rb, LSD1, and E2F1 IPs contained significant methyl lysine-reactive species other than histone proteins. The methyl lysine-reactive proteins comigrate with Rb and LSD1, suggesting that these proteins may be subject to regulation by lysine methylation similar to that for histone tails.

The function of LSD1 in Cp transcription was investigated using small interfering RNA (siRNA) depletion of LSD1 (Fig. 8A to C). We were unable to effectively deplete LSD1 from lymphoid cell lines but found that ∼80% of LSD1 could be depleted from 293 cells, as detected by Western blotting (Fig. 8A). Cp transcription was measured in 293 cells using a Cp-Luc reporter construct. We found that siRNA depletion of LSD1 inhibited Cp transcription by approximately sixfold, while control siRNA had no significant inhibitory effect on Cp. These results indicate that LSD1 contributes to the basal transcription activity of Cp in 293 cells.

FIG. 8.

LSD1 regulates Cp transcription and cell cycle progression. (A) Western blot of LSD1 and control ORC2 in 293 cells treated with siLSD1 or control siRNA (siCtrl). (B) Luciferase assays of Cp-Luc in 293 cells treated with siLSD1 or siCtrl. Control pGL3 is indicated above the first bar. (C) Western blot of HCT 40-16 cells with a cytomegalovirus (CMV)-FLAG vector or a CMV-FLAG-LSD1 expression vector, assayed with anti-FLAG antibody (top panel) or anti-ORC2 (lower panel) as a loading control. (D) HCT 379.2 (p53⁻) cells were transfected with CMV-FLAG or CMV-FLAG-LSD1 and assayed for cell cycle distribution using the BrdU-propidium iodide dual-staining method. (E) The same process as for panel D was used, except that HCT 40-16 (p53⁺) cells were transfected and analyzed. α, antibody.

The association of LSD1 with Rb suggests that LSD1 may have a global effect on cell cycle control. Others have found that LSD1 overexpression can affect cellular proliferation (32, 45). We therefore tested whether LSD1 overexpression had any effect on cell cycle distribution. In latently infected EBV-positive cells, LSD1 had little detectable effect on cell cycle progression. We reasoned that this failure to alter the cell cycle may be a result of inactivation of p53 by the EBV-encoded proteins EBNA3C and EBNA LP. To test this hypothesis, we compared the effect of LSD1 overexpression in HCT cell derivatives that were either p53 null (HCT 379.2) or contained wild-type p53 (HCT 40-16) (Fig. 8D to F). In cells lacking functional p53, LSD1 overexpression had no detectable effect on cell cycle distribution (Fig. 8E). In contrast, LSD1 overexpression in p53 wild-type HCT 40-16 cells caused an accumulation of cells in G₁ and a reduction of S and G₂/M cells. These findings suggest that LSD1 can alter cell cycle distribution in a p53-dependent manner.

DISCUSSION

Cell cycle regulation of transcription is thought to be essential for the coordinated expression of genes required for DNA synthesis and the completion of cell division (42). The EBV Cp controls transcription of a multicistronic RNA encoding six EBNA proteins that are essential for EBV immortalization of primary B lymphocytes (57). The EBNA proteins have multiple functions that include stimulation of Notch signaling, disruption of normal cell cycle restriction functions, and autoregulation of viral gene expression (57). Cp has been studied in great detail and has been shown to respond to numerous cellular and viral transcription regulatory proteins, including the EBNA2-responsive CBF1 and CBF2 and the B-cell-responsive Egr-1 and NF-Y. While Cp controls the stable expression of EBNA genes required for B-cell proliferation and development, Cp is also important for the programmed downregulation of EBNA genes required for the long-term maintenance of latent infection in hosts with healthy immune responses to EBV latent gene products. This developmental downregulation of Cp has been linked to the DNA methylation of key CpG residues in Cp (24). How Cp orchestrates the expression of EBNA genes during type III latency and the suppression of EBNA genes during type I latency is largely unknown.

In an effort to better understand Cp regulation, we initiated a biochemical analysis of cellular proteins that interact with Cp. Using DNA affinity purification, we identified several proteins that have not yet been recognized as Cp-interacting proteins. In particular, we found members of the E2F family (E2F1) and the tumor suppressor protein Rb. We also found DNA binding proteins, including C/EBP and MEF2D, that have known functions in regulating the BZLF1 promoter during lytic cycle reactivation (48, 56). We also identified several proteins involved in histone modifications, including the histone H3 K4 demethylase LSD1 (46). LSD1 was of particular interest, since our previous studies had implicated H3 K4 methylation in the chromatin regulation of latency types (9, 10, 12). In these previous studies, we found that histone H3 trimethyl K4 (H3mK4) was elevated at OriP and upstream of the EBNA1 binding sites at Qp in all cell types examined (12). The levels of H3mK4 appeared to spread from OriP toward Cp and the W repeats in type III latent infections (LCL and Raji cells) but was more restricted to the regions surrounding OriP in type I latent infections (Mutu I and Kem I cells) (9, 10). The mechanism regulating the spread of H3mK4 in different cell types has not been established but could be linked to differential activities or the recruitment of histone methylase and demethylases at Cp in these cell types.

LSD1 has been implicated in transcription repression and activation through an association with sequence-specific DNA binding proteins (29, 46, 54). Here, we showed by DNA affinity (Fig. 1) that LSD1 associates with Cp in vitro and by ChIP assay (Fig. 3) that it associates with Cp in vivo. We found that LSD1 bound to the same region of DNA that also bound E2F1 and Rb and that LSD1 forms a stable complex with Rb (Fig. 6). LSD1 and Rb were recruited to Cp in a similar cell cycle-dependent manner, suggesting that they coordinately regulate Cp (Fig. 4 and 5). Surprisingly, LSD1 did not coimmunoprecipitate with the E2F1-Rb complex, suggesting that the LSD1-Rb complex forms independently from E2F1 (Fig. 7D). This raises the question of how LSD1 is recruited to Cp if it does not associate with E2F1-Rb. Several possibilities may account for this observation. LSD1 has been shown to interact with several activators and repressor proteins, which may recruit LSD1 to Cp independently of E2F1 (45). Alternatively, LSD1 may bind to Rb when associated with another E2F family member, and several E2F proteins are known to exchange across the cell cycle at E2F binding sites (3). It is also possible that LSD1 binding to Rb may initiate the dissociation of the E2F-Rb complex. This last possibility is intriguing, since we also observed that Rb or Rb-related proteins were reactive to methyl lysine-specific antibodies, suggesting that LSD1 may regulate Rb interactions through demethylation of nonhistone substrates.

The functional significance of Rb and LSD1 at Cp is not completely clear. Depletion of LSD1 by siRNA in 293 cells caused a loss of Cp-Luc reporter plasmid activity, suggesting that LSD1 contributes to the basal activity of Cp in these cells (Fig. 8B). LSD1 has been reported to regulate cellular proliferation in some cell types, but the mechanism of this cell cycle alteration was not characterized in detail (45). The interaction of LSD1 with Rb would suggest that LSD1 depletion should affect cell cycle progression by altering the expression of cellular genes regulated by Rb. We found that forced expression of LSD1 did alter the cell cycle profile in p53⁺, but not p53⁻, HCT cells (Fig. 8E to F). LSD1 depletion may have indirect effects on the cell cycle, and this may have affected Cp expression in transient assays. Despite the indirect effects, our data demonstrate that LSD1 binds directly to Cp in vivo and in vitro. We therefore favor the model that LSD1 functions at Cp in a complex with Rb to confer cell cycle control of Cp transcription.

Cell cycle transcription regulation of EBNA1 has been previously demonstrated in studies of Qp (11, 40). Qp is the alternative start site for EBNA1-only transcription and is required for the continuous expression of EBNA1 in type I latency when Cp is silenced (43). Qp contains EBNA1 binding sites that are thought to repress transcription initiation (41). Activation is mediated by interferon response factor and E2F family binding sites, which contribute to the cell cycle regulation of Qp through the Rb pathway (11, 35, 40, 58). The cell cycle regulation of EBNA1 is thought to be consistent with the replication function of EBNA1, although no change in EBNA1 protein levels has been observed during the cell cycle. Our data suggest that Cp, like Qp, is regulated by the Rb pathways and is subject to cell cycle controls. It is likely that the S phase increase in RNA expression of EBNA genes through Cp and Qp helps to coordinate expression of viral genes that effect cellular proliferation, like EBNA2 and EBNA3C. Interestingly, Cp can be subject to epigenetic repression, and Rb is thought to function in heterochromatin formation (27, 31). Mice lacking Rb were unable to form telomeric heterochromatin, which led to genetic instability and premature ageing (18, 21). We suggest that Rb recruitment at Cp contributes to the epigenetic silencing that is observed in type I latency. Changes in Rb abundance or occupancy at Cp, or the Rb-interacting complexes, may determine whether Cp will be silenced through mitosis to form type I latency or will be transiently suppressed only at the end of S phase, as is observed in type III latency. In this way, cell cycle factors may be intimately linked with heritable epigenetic silencing.