Cohesins Repress Kaposi's Sarcoma-Associated Herpesvirus Immediate Early Gene Transcription during Latency

Cells and plasmids.

KSHV-positive PEL cells (BCBL1, BC-1, and BC-3) and both Epstein-Barr virus (EBV)- and KSHV-positive cells (JSC-1) were grown in RPMI medium (Gibco BRL) containing 10% heat-inactivated fetal bovine serum and the antibiotics penicillin and streptomycin (50 U/ml). 293T cells were cultured in Dulbecco's modified Eagle's medium with 10% fetal bovine serum and antibiotics. All cells were cultured at 37°C in a 5% CO₂ environment. KSHV lytic activation was induced by the addition of 1 mM sodium butyrate (and/or 20 ng/ml tetradecanoyl phorbol acetate [TPA]) for 8 h or 24 h before analysis as indicated in the figure legends. Where indicated, cells were treated with 13 μM ganciclovir (GCV; Sigma) at the time of short hairpin RNA (shRNA)-expressing lentivirus infection to inhibit KSHV lytic replication.

Generation of shRNA lentiviruses and lentivirus infection.

Lentiviruses were generated by transient transfection of 293T cells by use of a three-plasmid system (a pLKO plasmid expressing shRNA against cellular CTCF, Rad21, SMC1, or SMC3; pMD2.G [envelope plasmid expressing vesicular stomatitis virus glycoprotein]; and psPAX2 [packaging plasmid]). Viral supernatant was harvested at 2 days posttransfection, concentrated, and stored at −20°C until use. For shRNA knockdown experiments, PEL cells were infected with concentrated lentiviruses by spin infection for 90 min at 450 × g and 25°C in the presence of 8 μg/ml of Polybrene (Sigma). After spin infection, cells were seeded at 0.4 × 10⁶/ml for 48 h before puromycin selection. Cells were then harvested at 1 day or 3 days post-puromycin selection, as indicated in the figure legends.

ChIP.

Chromatin immunoprecipitation (ChIP) assays were performed as described previously (12). Briefly, cells were fixed in 1% formaldehyde for 15 min. DNAs were sonicated to 200- to 400-bp DNA fragments on a Diagenode Bioruptor according to the manufacturer's protocol. Quantification of precipitated DNA was determined using SYBR green probe real-time PCR and the Absolute Quantification program (ABI 7900HT Fast real-time PCR system; Applied Biosystems). Antibodies used in the ChIP assay are listed below. Primers for ChIP assays (sequences are available upon request) were designed using Primer Express, version 2.0 (Applied Biosystems). PCR data were normalized to input values that were quantified in parallel for each experiment. IPs were performed in triplicate for each antibody, and the experiments were repeated at least two times.

ChIP-Seq.

ChIP with high-throughput sequencing (ChIP-Seq) was performed as described previously (43). Briefly, 1 × 10⁷ BCBL1 cells were used per assay, with either rabbit anti-cellular SMC1 (Bethyl) or anti-CTCF (Millipore) antibody or control rabbit IgG (Santa Cruz Biotechnology). A Diagenode Bioruptor was used to sonicate genomic DNA into 150- to 300-bp DNA fragments according to the manufacturer's protocol. DNA fragments of ∼150 to 300 bp were isolated by agarose gel purification, ligated to primers, and subjected to Solexa-based sequencing using the manufacturer's recommendations (Illumina, Inc.). Peak-calling analysis of the KSHV genome was performed as described previously (43).

Antibodies.

The following antibodies were used for ChIP assays: anti-IgG (Santa Cruz Biotechnology), anti-CTCF (Millipore), anti-SMC1 (Bethyl), anti-Rad21 (Abcam) anti-trimethyl H3K4 (Millipore), anti-H3K27 (Millipore), anti-acetylated H3K9 (Millipore), RNA polymerase II (Millipore), and RNA polymerase II pS2 (Abcam) and pS5 (Abcam) antibodies. The following mouse monoclonal antibodies were used for ChIP assays: anti-IgG (Santa Cruz Biotechnology) and RNA polymerase II (Millipore) antibodies. Rabbit polyclonal anti-SMC1 (Bethyl) and anti-SMC3 (Bethyl), rat anti-KSHV LANA antibody (Advanced Biotechnologies Inc.), and mouse monoclonal anti-actin (Sigma) and Rad21 (Santa Cruz Biotechnology) antibodies were used for Western blotting. Anti-KSHV ORF45 antibody was provided by Yan Yuan, University of Pennsylvania, and anti-KSHV ORF50 antibody was provided by Erle Robertson, University of Pennsylvania.

Isolation of KSHV genomic DNA and quantification of its copy number.

The intracellular KSHV DNA copy number was determined by quantitative PCR (qPCR) analysis of purified total genomic DNA. Briefly, 10⁶ BCBL1 cells were resuspended in SDS lysis buffer (1% SDS, 20 mM NaCl, 4 mM EDTA, 20 mM Tris, pH 8.0) with proteinase K for at least 6 h at 50°C. The cell lysate was then subjected to phenol-chloroform extraction and ethanol precipitation. Precipitated DNA was then assayed by real-time PCR, using primers for the KSHV ORF50 promoter region (5′-CCC GCC CAG AAA CCA GTA G-3′ and 5′-TGC GGA GTA AGG TTG ACT TTT TAA-3′) and for normalization to the cellular actin DNA signal (5′-GCC ATG GTT GTG CCA TTA CA-3′ and 5′-GGC CAG GTT CTC TTT TTA TTT CTG-3′).

RT-PCR.

Total RNA was isolated using an RNeasy kit (Qiagen) and then further treated with DNase I (New England Biolabs) at 37°C for 30 min. For cDNA synthesis, 2 μg of total RNA was incubated with a 5 μM concentration of random decamers (Ambion), 150 U of Superscript II RNase H⁻ reverse transcriptase (RT; Invitrogen), 1.6 U of RNase inhibitor, a 1 mM concentration of each deoxynucleoside triphosphate (dNTP), and 3.3 mM dithiothreitol (DTT) in a 20-μl reaction mixture according to the manufacturer's protocol (Invitrogen). After heat inactivation at 65°C for 10 min, the sample was diluted with 180 μl distilled water. Diluted RT products were analyzed by real-time PCR with a SYBR green probe, using the ΔC_T method (ABI Prism 7900 sequence detection system; Applied Biosystems). The sequences of primers used for RT-PCR assay are available upon request. The level of beta-actin in each sample was used as an internal control for each quantitative PCR. A quantitative PCR assay with RNA but without reverse transcription was conducted to serve as a negative control for each reaction.

PFGE.

BCBL1 cells infected with lentivirus expressing shControl, shRad21, or shCTCF shRNA were resuspended in 1.0% agarose plugs and incubated in lysis buffer (0.2 M EDTA [pH 8.0], 1% SDS, 1 mg/ml proteinase K) at 50°C for 48 h. The agarose plugs were washed twice in Tris-EDTA (TE) buffer (pH 7.5). Pulsed-field gel electrophoresis (PFGE) was performed for 22 h at 14°C, with a linear ramping pulse of 60 to 120 s through 120°C as described previously (Bio-Rad CHEF Mapper). DNA was transferred to nylon membranes by established methods for Southern blotting. The DNA was detected by hybridization with a ³²P-labeled probe specific for the KSHV TR region and visualized with a Molecular Dynamics phosphorimager.

shRNA depletion of cohesin subunits results in KSHV reactivation in BCBL1 cells.

To investigate the function of cohesins in regulation of KSHV latency and reactivation, we generated a puromycin-resistant lentivirus expressing shRNAs targeting SMC1, SMC3, and Rad21. BCBL1 cells were infected with lentivirus and selected with puromycin for 5 days postinfection. Western blotting was used to measure the extent of shRNA depletion (Fig. 1A). SMC1 levels were partly depleted by shSMC1 as well as by shSMC3. SMC3 was partially depleted by shSMC3. Rad21 was depleted by shRad21 as well as by shSMC1 and shSMC3. The codependent depletion of cohesin subunits may not be unexpected, since these proteins are known to form a complex that may be unstable in the absence of any one major component. More relevant to KSHV, we observed that KSHV immediate early proteins ORF50 and ORF45 were elevated in BCBL1 cells infected with any of the cohesin-targeting shRNAs but not with the shControl vector (Fig. 1A). Actin levels were not significantly affected by lentivirus infection or shRNA depletion of cohesins (Fig. 1A, bottom panel).

To determine if cohesin depletion altered the DNA structure of the KSHV genome, we assayed KSHV DNA by PFGE and Southern blot analysis (Fig. 1B). We found that KSHV genomes were present as an equal mixture of circular and linear forms in BCBL1 cells infected with shControl lentivirus (Fig. 1B, lane 1). In contrast, BCBL1 cells infected with shSMC1, shSMC3, or shRad21 lentivirus showed a large (>10-fold) increase in linear and sublinear forms of the KSHV genome (Fig. 1B, lanes 2 to 4).

We also assayed the effect of CTCF depletion on BCBL1 cells (Fig. 1C and D). CTCF was depleted efficiently (>80%) (Fig. 1C, top panel). However, unlike shRad21, shCTCF did not induce ORF45 expression (Fig. 1C). We also noted that shRNA depletion of CTCF did not induce linear genome formation as analyzed by PFGE (Fig. 1D, lane 5). To determine if the increase in linear genomes induced by shRad21 was due to activation of the viral lytic cycle, we used GCV to inhibit viral DNA polymerase (Fig. 1D, lanes 2 and 4). We found that addition of GCV to shRad21-infected cells prevented the accumulation of linear genomes (Fig. 1D, lane 4), indicating that the amplification of linear genomes reflects viral lytic replication. The ability of cohesin depletion to induce the KSHV lytic cycle was further confirmed by RT-PCR analysis of KSHV lytic genes for ORF50, ORF73, ORF25, and PAN (Fig. 1E). Depletion of SMC1, SMC3, or Rad21 resulted in an ∼10- to 50-fold increase in KSHV lytic mRNA in BCBL1 cells. Additionally, we showed by qPCR that the KSHV DNA copy number increased ∼10- to 100-fold after cohesin depletion (Fig. 1F).

Rad21 depletion causes KSHV reactivation in multiple PEL cell lines.

To investigate whether cohesins regulate KSHV latency in PEL cells other than BCBL1 cells, we assayed the effect of Rad21 depletion in the BC1, BC3, and JSC-1 cell lines (Fig. 2). PEL cells were infected with shControl or shRad21 lentivirus and then assayed by Western blotting to assess Rad21 depletion (Fig. 2A). shRad21 lentivirus infection depleted Rad21 levels to various extents, with BC1 and BC3 cells showing only ∼30% reduction and with more robust reductions in BCBL1 and JSC-1 cells. JSC-1 and BCBL1 cells showed a strong induction of the lytic antigen ORF45, while BC3 showed only a modest activation and BC1 showed little or no induction. To further assess the effects of Rad21 depletion on KSHV reactivation, we analyzed KSHV DNA by PFGE (Fig. 2B). We found that shRad21 lentivirus infection increased linear and sublinear forms of the viral genome in BC3, JSC-1, and BCBL1 cells but had little effect in BC1 cells. qPCR analysis of the KSHV DNA copy number indicated that shRad21 increased the genome copy number in BC3, JSC-1, and BCBL1 cells but not in BC1 cells (Fig. 2C). These findings suggest that depletion of Rad21 induces KSHV lytic replication in multiple different PEL cells. The failure of BC1 cells to reactivate in response to shRad21 lentivirus infection may be explained partly by the poor depletion of Rad21 in these cells (∼50% depleted). However, BC1 cells also failed to respond efficiently to NaB-induced lytic activation (data not shown), suggesting that they have a more restrictive latency than that of other PEL cells tested.

Cohesins bind at CTCF sites within the immediate early control region.

To generate a high-resolution map of CTCF and cohesin binding sites on the KSHV genome in BCBL1 cells, we performed ChIP-Seq experiments with CTCF and SMC1 antibodies (Fig. 3). ChIP-Seq peaks for CTCF and SMC1 were identified using a Poisson distribution model (modified MACS program) and then normalized to an IgG control (Fig. 3A) (43). As reported previously, most cohesin (e.g., SMC1) peaks overlapped extensively with CTCF peaks (56, 62, 68, 78). However, several CTCF peaks were not co-occupied with cohesins (nucleotides [nt] ∼22,000, 27,000, and 120,000). A major CTCF-cohesin peak was detected within the first intron of the LANA transcript (nt ∼127,400), consistent with our previous findings (32, 34, 68). We also found that CTCF and cohesins bound efficiently to the TR region, which had not been reported previously. Since we previously found that the major CTCF-cohesin peak was capable of forming a 3C-measured DNA loop with the region upstream of the ORF50 promoter, we focused on CTCF and SMC1 binding to this region (Fig. 3B and C). We found that CTCF bound to sites in the ORF45 gene (nt ∼67,755), the ORF46 gene (nt ∼68,744 to 68,763), the ORF49 start site (nt ∼72,294 to 72,312), and the ORF50 termination site (nt ∼74,575 to 74,594). A consensus CTCF binding motif was identified for each of these peak regions (data not shown). SMC1 bound to all of these sites, but with slightly different occupancies. While CTCF bound more extensively at the ORF45-ORF46 peak, SMC1 bound more extensively at the ORF49 and ORF50 regions.

Rad21 depletion alters histone and RNA polymerase II posttranslational modifications at the KSHV immediate early control region.

To analyze epigenetic changes induced by Rad21 depletion, we performed a conventional ChIP assay with a series of primers spanning the regions of the lytic immediate early gene cluster (Fig. 4). Conventional ChIP assays confirmed that CTCF bound at the ORF45 (primer B), ORF46 (primer D), and ORF49 (primer K) regions (Fig. 4B, blue bars). The histone modifications H3K4me3, H3K9Ac, and H3K27me3 were examined throughout this region. We found that H3K4me3 was elevated throughout the region between the CTCF sites at positions ∼69,000 (primer D) and ∼72,600 (primer K). H3K9Ac was elevated at the ORF46 region (primer E), while H3K27me3 was notably depleted at this position relative to the surrounding regions. The effect of shRad21 or shCTCF depletion was also measured using ChIP assay across this region (Fig. 4B and C, pink and green bars). We found that CTCF binding was reduced by shCTCF, as expected, but was also reduced by shRad21 (Fig. 4B, top graph). shRad21 depletion also caused a loss of H3K4me3, H3K9Ac, and K3K27me3 at all positions in the lytic control region that were tested. In contrast, shCTCF depletion had little effect on histone modifications in this region.

To further assess the effect of shRad21 on KSHV lytic transcription control, we assayed RNA Pol II and its phosphorylated CTD isoforms (pS2 and pS5) by ChIP assay (Fig. 4C). In shControl lentivirus-infected cells, we found that RNA Pol II was distributed weakly throughout the immediate early control region, with a minor peak of RNA Pol II pS5 occurring within the ORF46 gene (primer E). In contrast, shRad21 depletion produced a significant increase in RNA Pol II occupancy at the ORF47 locus (primer F) and a general increase in RNA Pol II pS2 and a decrease in pS5 throughout the immediate early gene locus. Depletion of CTCF produced a slight increase in total RNA Pol II occupancy at several primer positions but had no effect on either phospho-isoform of RNA Pol II. Taken together, these findings suggest that Rad21 depletion causes a general loss of histone occupancy and an increase in RNA Pol II pS2 at the lytic control region. Since pS2 corresponds to the elongation-competent form of RNA polymerase, this suggests that Rad21 depletion promotes RNA polymerase loading and elongation throughout the lytic control region.

Rad21 depletion activates immediate early gene transcription similarly to chemical inducing agents.

The effects of shRad21 and shCTCF on transcription of the ORF45 to ORF50 genes within the lytic control region were assayed by RT-qPCR (Fig. 5). Depletion of Rad21 resulted in large increases in ORF45 (primers B, C, and D), ORF47 (primer F), ORF49 (primer K), and ORF50 (primer L) (Fig. 5A, dark gray bars). Depletion of CTCF showed no similar increase in transcription throughout this region (Fig. 5A, light gray bars). Early gene transcription was also measured by chemical induction with NaB for 8 h (Fig. 5B) or with NaB and TPA for 24 h (Fig. 5C). We found that chemical induction for short (8 h) and long (24 h) times produced a pattern of transcription nearly identical to that observed for Rad21 depletion. The effect of Rad21 depletion on the KSHV lytic control region was re-evaluated using a new set of primer pairs optimized for RT-PCR of KSHV mRNAs (17) rather than detection by ChIP primers optimized for viral DNA fragments. The new mRNA-specific primers revealed that KSHV gene transcripts for all viral genes in this locus were activated by Rad21 depletion, with transcripts for ORF45, ORF46, and ORF50 being the most abundant (Fig. 6A). The same Rad21-depleted cells were evaluated by ChIP assay of Rad21, SMC1, and SMC3 binding (Fig. 6B to D). The ChIP assay revealed that Rad21 depletion led to a complete loss of all cohesin subunits from the KSHV lytic control region, suggesting that Rad21 is essential for cohesin binding to this locus. The early induction of ORF45, ORF46, ORF47, and ORF49 is consistent with previous studies reporting these genes to be immediate early genes similar to ORF50 (77, 84). Our findings suggest that these genes are coordinately activated by lytic-inducing agents and that Rad21 plays a key role in regulating transcription of this lytic control region.

NaB treatment disrupts CTCF-cohesin binding at the lytic control region.

To explore the possibility that NaB acts on cohesin binding to regulatory chromatin, we assayed the effect of NaB treatment on CTCF and cohesin binding in ChIP assays (Fig. 7). BCBL1 cells were treated with or without NaB for 8 h prior to ChIP analysis. We found that both CTCF and cohesin binding was partially disrupted after 8 h of NaB treatment. The most significant losses were observed for cohesin subunits SMC1 and Rad21 at the ORF49 locus (primer K). These findings suggest that loss of cohesin and CTCF binding at the immediate early locus is an early event during KSHV lytic induction by NaB.

Chromatin organization of the lytic control region.

Others have shown that a stem cell-like bivalent pattern of histone modifications is observed at key regions of the KSHV genome, including the ORF50 gene promoter region (73). Our data support the finding that the region upstream of the ORF50 gene promoter is enriched in H3K4me3 and H3K27me3. However, at the ORF46 region, histones were enriched in H3K4me3 and H3K9ac and depleted of H3K27me3. Interestingly, CTCF binding was more robust at the ORF45 and ORF47 sites, while cohesin binding was more robust at the ORF50 transcription initiation site. Histone H3K4me3 formed a broad peak bounded by CTCF sites near the transcription start sites for ORF45 (nt ∼68,763) and ORF50 (nt ∼72,294). RNA Pol II was enriched at the ORF47 region, but no transcription was detected, and RNA Pol II remained in the phospho-S5 form associated with polymerase pausing. We propose that the region between CTCF sites at the ORF45 and ORF50 transcription start sites constitutes a functional chromatin element, which we refer to as the lytic control region (Fig. 8). Depletion of the cohesin subunit Rad21 led to a complete disruption of this chromatin organization, suggesting that cohesins are important for maintaining the histone modification pattern and paused RNA polymerase between the divergently transcribed immediate early genes.

Coordinated control of lytic cycle immediate early genes.

The genomic organization of KSHV suggests that ORF45 to ORF49 may be regulated coordinately with ORF50, since their genes share an upstream promoter regulatory region. Previous studies have shown that the ORF50 and K8 genes can be expressed as a multicistronic immediate early transcript (84) and that the ORF50 gene encodes the primary activator of lytic gene transcription (22, 46, 70, 84). The ORF45 gene was also shown to be an immediate early gene (84) as well as a component of the KSHV tegument (86), and the murine herpesvirus 68 (MHV68) orthologue of ORF45 is required for the immediate early phase of viral replication (29). The ORF46 gene encodes viral uracil deglycosylase, the ORF47 and ORF48 genes encode components of the entry machinery for gL, and the ORF49 gene encodes an orthologue of BRRF1, which has been implicated in EBV lytic gene activation (21, 24, 27). Furthermore, the ORF47-ORF46-ORF45 gene region can be expressed as a multicistronic transcript (77), and the gene cluster is potently activated by Rta, along with autoactivation of the ORF50 gene (6, 8, 63, 76). These observations suggest that the entire locus is coordinately regulated with lytic reactivation signals that activate ORF50 transcription. Our data indicate that several of these genes are coordinately regulated during the immediate early stages of lytic reactivation. We found that sodium butyrate treatment leads to a rapid loss of cohesin binding that correlates temporally with transcription activation of ORF47-ORF46-ORF45 and ORF50. Taken together, these findings support the model that the lytic control region consists of a cohesin-regulated divergent promoter that can initiate transcription of ORF50 as well as ORF47-ORF46-ORF45 in response to sodium butyrate treatment.

Coordinated control of lytic and latent gene expression by cohesins.

Cohesins have been shown to both positively and negatively regulate transcription (11, 47, 61, 67), and more recently, they have been implicated in the control of RNA Pol II pausing (18, 33). The transcriptional effects of cohesins may be related to other biochemical properties, including functions associated with chromatin boundaries, insulators, and DNA looping. The chromatin boundary and insulator functions have been linked closely to CTCF, which co-occupies many of the cohesin sites. Interestingly, we found that depletion of CTCF disrupted cohesin binding but did not result in viral lytic gene activation (Fig. 4). This suggests that cohesins have a more specialized function in transcription repression at the KSHV lytic control region. Cohesin depletion led to an increase in RNA Pol II S2 at the ORF50 promoter region, with a corresponding loss of several histone modifications. CTCF depletion has a much less significant effect on lytic transcription, RNA Pol II recruitment, and histone modifications. CTCF depletion by shRNA may not be as complete as that of Rad21 and other cohesins, and this may partly explain the failure of CTCF depletion to disrupt latency. However, CTCF depletion does lead to reductions of the genome copy number and of LANA expression, consistent with previous findings that CTCF is required for maintaining latent gene expression and for genome maintenance (32, 33, 68). Thus, while CTCF and cohesins may co-occupy many DNA binding sites, they can have functionally separable activities in gene regulation.

Regulation of KSHV lytic control region by cohesins and other factors.

The lytic control region is regulated by numerous cellular factors and signaling pathways, including RBP-jK, C/EBP, octamer binding protein, YY1, SP1, and HIF1α-mediated hypoxia (6–8, 25, 42, 45, 57, 63, 75, 77, 81). Autoactivation of lytic transcripts by Rta has been shown to function through RBP-jK, which binds at several sites in the ORF50 upstream control region. RBP-jK also binds within the latency control region and can be regulated by LANA (30, 38, 44). We note that many RBP-jK sites in the lytic and latency control regions overlap or are in close proximity to the CTCF-cohesin binding sites (45). Although it was not examined in this study, we speculate that RBP-jK may contribute to cohesin function in the structural organization of the KSHV genome. RBP-jK is known to associate with corepressor complexes in the absence of coactivators such as Notch or Rta (16, 83). Thus, CTCF-cohesin may interact with other regulatory factors, including RBP-jK, to maintain bivalent chromatin architecture and histone modifications that can readily respond to environmental stimuli. It is also possible that cohesin depletion has additional indirect and global effects that lead to robust reactivation of KSHV genomes. However, since CTCF-cohesin binds to the ORF50 region (Fig. 3 and 4), and since this binding is lost during early stages of lytic reactivation (Fig. 7), we favor the model that cohesins function as direct repressors of transcription at the KSHV lytic control region.