Dissociation of histone H3 from the KSHV genome is concomitant with viral DNA replication

In this study we asked what histone modifications are associated with the KSHV genome during latency and how they change upon reactivation. For this, we used the well-characterized recombinant KSHV-positive primary effusion lymphoma cell line, TRExBCBL1-RTA, which expresses a Doxycycline (Dox) inducible myc/His-tagged RTA incorporated into the cellular genome [39]. We chose the RTA-mediated reactivation of KSHV instead of chemical inducers such as TPA, TSA or sodium butyrate because these chemicals can globally affect both cellular and viral gene expression, while RTA ensures robust and specific viral reactivation [39]. Figure S1 showed that Dox treatment (1ug/ml) of the TRExBCBL1-RTA cells for 6, 12 and 24 hours rapidly induced myc/His-RTA expression, resulting in a gradual induction of the KSHV gene expression cascade (Figure S1A, B) [39], whereas viral DNA replication was apparent only at 24 hpi, which was in correlation with the induction of late gene expressions (Figure S1B and C). To analyse the changes in the KSHV nucleosome structure upon reactivation, ChIP experiments were performed with a histone H3 specific antibody. We measured the abundance of specific DNA sequences in the histone H3 immunoprecipitates with qPCR using a set of primer pairs specific for both the promoter and the coding regions of various viral genes including RTA, LANA, K2, ORF8, ORF25, ORF56 and ORF64 (Figure S2A, B). This revealed comparable levels of H3 association throughout the viral genome during latency (0 hpi), which did not significantly change on most of the genomic regions at 12 hpi, but dropped sharply at 24 hpi (Figure S2B). This is in agreement with previous findings showing that disassembly of the viral chromatin is concomitant with viral DNA replication during lytic infection [40]. In contrast, H3 levels remained constant on cellular promoters during KSHV lytic reactivation, suggesting that H3 dissociation occurs specifically on the viral genome (Figure S2C).

Distinct patterns of activating and repressive histone modifications on the KSHV genome during latency and reactivation

In order to investigate the epigenome of KSHV during latency and lytic reactivation, we tested whether repressive histone modifications associated with lytic genes are responsible for maintaining the repression of lytic gene expression during latency and whether reactivation induces the deposition of activating histone modifications onto the viral genome for lytic gene expression. To uncover the global distribution of histone modifications on the KSHV chromatin during latency and reactivation, we mapped the genome-wide distribution of activating histone marks [acetyl-H3 (AcH3) and H3K4me3] and repressive marks (H3K9me3 and H3K27me3) on the KSHV genome (Figure 1 and Figure S3, S4, S5, S7). The ChIP-on-chip experiments were carried out with chromatins prepared from both non-induced (0 hpi/latency) and Dox-induced (12 hpi) TRExBCBL1-RTA cells. Figure 1 shows the average of two independent ChIP-on-chip biological replicates (Figure S3 and S4). To get a high resolution of the localization of histone modifications and proteins of interest on the KSHV genome, we used a 15-bp tiling microarray that contains 8942 overlapping 60 nucleotide-long oligos spanning the entire KSHV genome.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 1. Genome-wide mapping of histone modifications on the KSHV genome during latency and reactivation.

Each ChIP-on-chip experiment is an average of two biological replicates. The histone H3 and histone modification ChIPs were performed with non-induced and doxycycline-induced (12 hpi) TRExBCBL1-Rta cells followed by the hybridization of the labelled ChIP and input DNAs onto a custom designed KSHV-specific 15-bp tiling microarray. See the Material and methods for details. Orange colour indicates 0 hpi-ChIP/input ratio while the black line shows 12 hpi-ChIP/input. Numbers in the left upper corners show the maximum values of Cy5/Cy3. Missing probes in specific genomic regions are shown below the genome scale (**). The alternating dark and light blue squares atop display the viral ORFs where the white triangle indicates ORFs that are expressed from the reverse DNA strand. The “hpi” stands for hours post-induction.

https://doi.org/10.1371/journal.ppat.1001013.g001

Because changes in H3 occupancy may affect the enrichment of histone modifications on the viral genome, we first investigated the global distribution of H3 on the viral genome during latency and upon reactivation. The histone H3 ChIP-on-chips revealed that histone H3 enrichment levels were comparable throughout the KSHV genome at 0 hpi and did not significantly change on most parts of the KSHV genome at 12 hpi (Figure 1, S3, S4). The histone H3 ChIP analysis also showed similar results (Figure S2B). However, it should be noted that a small viral genomic region between 15 and 30 kb, which contains mostly KSHV unique genes, displayed a detectable decrease of H3 occupancy at 12 hpi (Figure 1). In contrast to the uniform distribution of H3, the different histone modifications displayed distinct patterns and were enriched in specific KSHV genomic regions during latency and reactivation (Figure 1). Immunoblot analysis revealed that the expression of H3 as well as the global levels of cellular histone modifications did not change upon KSHV reactivation (Figure S2 D), thus any change of the histone modifications on the KSHV genome is likely to be a consequence of the reactivation-induced change in the KSHV epigenome. The genome-wide mapping of histone modifications showed that both activating and repressive histone modifications were associated with the KSHV genome during latency (Figure 1). While both the activating AcH3 and H3K4me3 marks co-localized on the KSHV genome so did the repressive H3K9me3 and H3K27me3 marks. However, the activating and repressive histone modifications were mutually exclusive on the bulk of the KSHV genome (e.g. 30–60 kb and 90–120 kb) (Figure 1). As expected, the latency-associated locus (118–128 kb) where KSHV genes that are constitutively expressed during latency are located, was enriched with the activating H3K4me3 and AcH3 histone modifications but depleted for the repressive H3K9me3 and H3K27me3 (Figure 1). Unexpectedly, we also found that several regions of the KSHV genome (e.g. 1–30 kb and 60–90 kb) that are not associated with latency-gene expression were also highly enriched in both H3K4me3 and AcH3 histone marks (Figure 1). Strikingly, the PcG protein-mediated repressive histone modification, H3K27me3, was widely distributed throughout the KSHV genome, whereas the H3K9me3 repressive histone modification was restricted mainly to two genomic regions (30–60 kb and 95–115 kb) encoding a number of late genes (Figure 1). Interestingly, the activating H3K4me3 and AcH3 histone modifications were absent in these two genomic regions where the repressive H3K9me3 and H3K27me3 histone modifications coexisted, suggesting that KSHV genes in these regions have a strongly repressive heterochromatin structure during latency (Figure 1).

The RTA-induced initiation of KSHV lytic gene expression program results in the redistribution of histone modifications on viral genome. We calculated the 12hpi/input ratio to view the changes in histone modifications (Figure 1). Since H3 enrichment was comparable between 0 and 12 hpi on most parts of the viral genome and viral DNA replication had yet to occur by 12 hpi, we also hybridized the 12 hpi-ChIPs against the 0 hpi-ChIPs, which allowed us to specifically observe changes in histone modification levels on the KSHV genome upon lytic reactivation (Figure S7C). Others have also recently applied similar ChIP-on-chip analyses in studies investigating the epigenetic reprogramming of the host genome by the adenoviral protein E1a [41]. Our ChIP-on-chip analysis showed that the enrichment of the activating histone modifications was elevated the most when it was concomitant with the reduction of repressive histone marks in the genomic regions between 1 and 30 kb containing several early genes and between 68 and 77 kb, which encodes the IE proteins ORF45, ORF48, ORF50 (RTA) and K8 (KbZIP) (Figure 1 and S7). These changes in the viral chromatin are indicative of robust transcriptional activation and are consistent with the KSHV gene expression profile in that expression of the lytic genes in these genomic regions is induced in the early phase of reactivation [39], [42], [43]. In contrast, only minor changes in the levels of the AcH3 and H3K4me3 activating marks were detected in the genomic regions that encode large number of late genes (30–60 kb and 95–115 kb), whereas significant changes of the H3K27me3 repressive mark were observed at 12 hpi (Figure 1 and S7).

In summary, these results demonstrate that the activating histone modifications of the latent KSHV genome are preferentially enriched in the constitutively active latency-associated genomic locus and in the early-lytic gene-containing genomic regions. In contrast, the H3K9me3 and H3K27me3 repressive histone marks are primarily enriched in the genomic regions that encode many late genes during latency (0 hpi), and they remain associated with these regions during the early phase (12 hpi) of reactivation as well.

Profiling the histone modification patterns within the regulatory regions of KSHV genes

While only a few genes are expressed during latency, all viral genes are expressed upon lytic reactivation in a temporal and sequential order. This suggests that distinctive histone modifications may be associated with the different viral promoters to ultimately determine the timing and rate of viral gene expression. Thus, we attempted to delineate the characteristics of the chromatin structures associated with the regulatory regions of KSHV genes during latency and lytic reactivation (Figure 2, S6, S8). For this, we aligned the KSHV open reading frames (ORFs) relative to their translational start sites (TSS) and plotted the signal intensities of probes derived from the ChIP-on-chip analysis across a 2-kb region spanning 1 kb on either side the TSS (please see Materials and Methods for details). The rationale of this strategy was based on a number of considerations. (i) While the transcriptional start sites have only been identified for a few KSHV genes, they are generally within a few hundred base pairs of their TSS, showing that the TSS can be used as a reference point. (ii) Due to the compact structure of the KSHV genome, the promoters are usually closely localized upstream of the TSS. (iii) The 1 kb downstream region of the TSS was included in the analysis because several KSHV genes have introns close to the TSS at their 5′ ends, which may contain gene regulatory elements. Based on these factors, the 2-kb sequences around the TSS were considered to be the gene regulatory regions of KSHV genes. In fact, a similar strategy has also been used to analyse the recruitment of the KSHV transactivators, Rta and K-bZIP, to eighty-three putative KSHV promoters in TRExBCBL1-RTA cells [44]. Using the signal intensities of the probes derived from the ChIP-on-chip analysis, we performed an average linkage hierarchical clustering within each gene class (La or latent, IE, E, L), which gave us a comprehensive overview of the repressive and activating histone modifications across all the 2-kb TSS regions (Figure 2 and S8).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 2. Hierarchical clustering of histone modifications associated with the regulatory regions of viral genes.

Based on their expression patterns the viral genes were grouped as latent (La), IE, E and L genes and hierarchical clustering was performed within the groups. The rows display the histone modification patterns along the −1 kb to +1 kb genomic regions relative to the translational start site (TSS) of each viral gene, which we assigned for the gene regulatory regions. The 1 kb regions are divided into twenty 50 bp fragments that show the average of log2 ratio of probe signal intensities derived from the average of the biological replicates of ChIP-on-chip experiments. Blue and yellow colours represent lower-than-average and higher-than-average for enrichment, respectively, whereas gray shows missing values for enrichment due to lack of probes in those genomic regions. I-V represents the clusters of genes that have similar histone modification patterns. (A) Distinctive histone modification patterns are associated with the KSHV genes of different expression classes during latency. (B) Changes in the enrichment of histone modifications during reactivation (12hpi).

https://doi.org/10.1371/journal.ppat.1001013.g002

IE gene class.

By definition, IE genes are the first set of expressed viral genes whose induction does not require de novo expression of any viral proteins and can be rapidly induced upon reactivation of KSHV from latency. We found that although the IE genes (K4.2, K8, ORF45, ORF48, ORF50/RTA) are silenced during latency, their 2-kb regulatory regions are enriched in the activating histone modifications H3K4me3 and AcH3, and this is further increased upon lytic reactivation at ORFs 48, 50 and K8 (Figure 2, S8, S9). Since RTA is responsible for the switch between latency and lytic replication, not only is its promoter tightly repressed during latency, but its silencing should also be rapidly reversible upon reactivation. We found that the RTA promoter is enriched in both H3K4me3 and H3K27me3 during latency, suggesting that it possesses a bivalent chromatin that maintains the repression of the RTA promoter while keeping it poised for rapid activation (Figure 2, S8). An extensive ChIP analysis further confirmed the enrichment of AcH3, H3K4me3 and H3K27me3 and the depletion of H3K9me3 on the 2.5-kb promoter region of RTA during latency (Figure 3A). Upon lytic reactivation, however, the repressive H3K27me3 gradually decreased, while the activating histone modifications H3K4me3 and AcH3 increased on the RTA promoter (Figure 3A). These changes in the histone modification pattern are in concert with the induction of RTA expression.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 3. Dynamic association of histone modifications with viral genes during latency and reactivation.

(A) Time-course ChIP analysis of histone modifications on the RTA promoter during latency and reactivation. Cellular controls can be seen in panel D. (B) Colocalization of H3K4me3 and H3K27me3 on the RTA promoter is confirmed by sequential ChIP assays. The first ChIP was performed with either H3K4me3-specific or H3K27me3-specific antibody, followed by the elution of the immunoprecipitated DNAs and a second ChIP with either H3K27me3 or H3K4me3 antibody. LANA and ORF25 promoters were used as controls. (C) Time-course ChIP analysis of histone modifications on the selected latent (LANA), early (K2, ORF56) and late (ORFs 8, 25, 64) genes. Cellular controls can be found in panel D. Pr: promoter, in: within gene body. (D) ChIP assays of histone modifications on cellular promoters. The promoters of the repressed cellular MYT1 and HTF6 genes as well as the active promoter of the actin (ACT) gene were also tested using the same ChIP samples that had been used in panels A and C. ND: not detectable.

https://doi.org/10.1371/journal.ppat.1001013.g003

To confirm that the changes in the enrichment of AcH3, H3K4me3 and H3K27me3 were not due to differences in the efficiency of ChIPs between 0 and 12 hpi and that the low level of H3K9me3 was not due to a low efficiency of the H3K9me3-ChIP, three cellular promoters were also included as controls. Figure 3D shows that the levels of histone modifications on the transcriptionally repressed promoters of MYT1 and HTF6 genes as well as on the transcriptionally active promoter of the actin (ACT) gene did not significantly change between 0 hpi and 12 hpi, indicating that the efficiencies of the ChIPs were comparable at different time points (Figure 3D).

Dynamic association of the Polycomb group proteins and transcriptional activators with the KSHV genome during latency and lytic reactivation

Because of the genome-wide distribution of H3K27me3 on the KSHV genome, we performed a ChIP-on-chip assay to determine whether EZH2, the H3K27me3 histone methyltransferase of the PcG proteins, associates with the KSHV genome. (Figure 4A). ChIP-on-chip assays showed that EZH2 almost completely colocalized with H3K27me3 throughout the entire KSHV genome during both latency and reactivation (Figure 4A, S7, S10). Additional ChIP assays showed that EZH2 and SUZ12 (another subunit of the PcG complex PRC2) were found on the RTA and ORF25 promoters enriched with H3K27me3, but not on the LANA promoter (Figure 4B).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 4. Genome-wide binding of EZH2 to the KSHV genome correlates with the repression of lytic genes.

(A) ChIP-on-chip was performed for EZH2 and its genome-wide binding was compared with the distribution of H3K27me3 on the KSHV genome during both latency and reactivation. Labels are the same as in Figure 1. The H3K27me3 graph was taken from Figure 1. (B) EZH2 and SUZ12 binding to the H3K27me3-rich lytic promoters are shown by independent ChIP assays. EZH2-interacting PcG protein SUZ12 is enriched only where EZH2 is present. (C) Recruitment of transcription activators to the activated lytic promoters during KSHV reactivation. An anti-RNA polymerase II antibody (H-224) that recognizes the RNAPII independently from its phosphorylation state (total RNAPII) was used for ChIP of total RNAPII, while the anti-RNA polymerase II antibody CTD4H8 specifically immunoprecipitates RNAPII phosphorylated at the 5^th serine of its C-terminal domain (RNAPII Ser5).

https://doi.org/10.1371/journal.ppat.1001013.g004

Besides its regulatory role in cellular gene expression, the CBF1 transcription activator has been shown to play an active role in KSHV gene expression as well [45], [46]. Since the RTA promoter contains several CBF1 binding sites (Figure S11 A) [47], [48], we studied the recruitment of CBF1 onto the RTA promoter during reactivation. The ChIP assay showed that CBF1 was recruited to the RTA promoter, primarily at 1.4 kb upstream of the RTA translational start site where three putative CBF1 binding sites are closely located (Figure 4C and Figure S11 A). Furthermore, not only does this putative CBF-binding region of RTA promoter efficiently binds CBF1 in vitro, but its deletion also resulted in a dramatic decrease of the RTA-mediated autoactivation of its own promoter (Figure S11 B, C). Additional ChIP experiments revealed that RNAPII is also recruited to the promoter regions of RTA and LANA upon reactivation (Figure 4C). Recruitments of CBF and RNAPII to the RTA and LANA promoters were specific since they were not recruited to the promoter of the late gene, ORF25, whose expression was still blocked at 6 and 12 hpi (Figure 4C and S1B). The increase of RNAPII over the 3-kb upstream region of RTA may be not surprising given that this genomic region also includes the promoters of other lytic genes (ORFs 45, 46, 47, 48) as well as an alternative upstream promoter for RTA [49]. These results illustrate the dynamic associations of the PcG complex and transcriptional activators with KSHV genome during latency and lytic reactivation.

Polycomb group protein EZH2 and its H3K27me3 histone mark are required for the maintenance of KSHV latency

The dynamic association of EZH2 with KSHV genome suggests that PcG-mediated H3K27me3 histone modification is involved in the repression of lytic gene expression during latency. To address this issue, HA-tagged JMJD2A, JMJD3 and UTX histone demethylases were expressed in Vero-rKSHV.219 cells to test if the elimination of histone methylations can trigger KSHV reactivation (Figure 5A). JMJD2A is an H3K9me3-specific histone demethylase [50], JMJD3 and UTX are H3K27me3-specific histone demethylases [21], [51], and UTXmut is an enzymatically inactive form of UTX that contains a single point mutation (H1146A) in the Fe²⁺⁺ ion binding site [51]. JMJD3 and UTX have been shown to eradicate the H3K27me3 repressive mark, resulting in the upregulation of PcG-targeted gene expression [21], [52]. JMJD2A and JMJD3 or UTX expression detectably suppressed the steady-state levels of H3K9me3 or H3K27me3, respectively, in transfected Vero cells (Figure S12B, C). Since Vero-rKSHV.219 cells express red fluorescent protein (RFP) from the KSHV lytic PAN promoter and green fluorescent protein (GFP) from the EF-1α promoter, RFP expression has been extensively used as a marker of KSHV lytic reactivation [53]. Immunofluorescence analysis revealed that JMJD3 and UTX efficiently triggered KSHV reactivation, while JMJD2A and UTXmut did not (Figure 5A, B). Furthermore, coexpression of JMJD2A and JMJD3 showed no significant synergistic effect on KSHV reactivation (Figure 5A, B). Finally, JMJD2A, JMJD3, UTX and UTXmut were expressed at comparable levels (Figure S12A). These results bespeak the importance of the H3K27me3 histone modification in the maintenance of KSHV latency.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 5. Polycomb group proteins are involved in the maintenance of latency of KSHV.

(A) Overexpression of the wild-type HA-tagged H3K27me3 histone methyltransferases (HMTs) UTX and JMJD3 triggers the lytic reactivation of KSHV in Vero-rKSHV.219 as shown with the expression of RFP. In contrast, the H3K9me3 HMT JMJD2A and the enzymatically inactive UTXm showed little or no effect on KSHV lytic reactivation. (B) Quantification of RFP positive cells. (C and D) JSC-1 cells were treated with 5 uM DZNep for 1, 2 and 3 days and the cells were harvested for immunoblot analysis with the indicated specific antibodies against cellular proteins and histone modifications (C) or viral proteins (D). “Dpt” indicates days post-treatment. Whole cell lysate of NaB-treated JSC-1 cells was used as a control for Zta immunoblot. (E) JSC-1 cells were treated with DZNep as described in (C) and total RNAs were isolated for RT-qPCR analysis of some selected KSHV, EBV and cellular mRNAs. (F and G) BCBL-1 cells were infected by lentivirus expressing the indicated shRNAs and were then subject to immunoblotting analysis with the indicated antibodies (F) or RT-qPCR analysis was performed for the indicated viral transcripts (G).

https://doi.org/10.1371/journal.ppat.1001013.g005

A small molecule, 3-Deazaneplanocin A (DZNep), has been shown to inhibit the expression of the Polycomb repressive complex 2 (PRC2) components (EZH2, SUZ12, and EED), resulting in the suppression of H3K27me3 histone methylation and the upregulation of PcG target genes in vivo [54]. To further test the role of H3K27me3 histone methylation in KSHV latency, we treated KSHV and EBV co-infected primary effusion lymphoma JSC-1 cell line with DZNep, followed by immunoblotting assays (Figure 5 C and D). DZNep treatment dramatically decreased the EZH2 and SUZ12 and, thereby, H3K27me3 levels, ultimately resulting in the induction of the expression of polycomb-targeted cellular MYT1 gene (Figure 5E). However, H3K9me3, histone H3, and actin levels were not affected under the same conditions (Figure 5C). We found that DZNep treatment efficiently induced the reactivation of KSHV, but not EBV: KSHV Rta and K8 expressions were detected as early as 2 days after DZNep treatment, while the EBV IE protein Zta was not induced (Figure 5D). Real time quantitative RT-PCR also showed the induction of other early (ORF56) and late KSHV genes (ORFs 8, 25, 64), suggesting that PRC2 depletion activates the gene expression cascade of KSHV from the repressed latent state (Figure 5E).

Besides the downregulation of EZH2 and SUZ12, DZNep also induced apoptosis of JSC-1 cells, detected by monitoring the cleavage of the apoptosis marker PARP (Figure S13 A, B). To exclude the possibility that apoptosis may have influenced KSHV reactivation, we treated JSC-1 with stauorosporine (STS), which also induced apoptosis in JSC-1 as shown by the cleavage of PARP (Figure S13 C). In contrast to DZNep treatment, STS treatment affected neither the steady state levels of EZH2 and H3K27me3, nor did it reactivate KSHV from latency (Figure S13 C, D). This suggests that DZNep-mediated H3K27me3 reduction triggers KSHV reactivation. This was further supported by the fact that specific shRNA-mediated depletion of EZH2 induced the expression of number of lytic genes (Figure 5F, G). In summary, the depletion of the PcG proteins in latently infected cells induces the lytic reactivation of KSHV, suggesting that PcG proteins play an important role in the maintenance of KSHV latency.

Cell cultures and chemical treatments

TRExBCBL1-RTA is a KSHV-positive cell line, which expresses a doxycycline-inducible Rta gene [39]. It was maintained in RPMI 1640 medium (Cellgro) supplemented by 10% Tet system approved FBS (Clontech), 100 U/ml penicillin, 100ug/ml streptomycin and 20ug/ml hygromycin B. KSHV- and EBV-positive cell line JSC-1 was grown in RPMI 1640 medium (Cellgro) containing 10% FBS (Clontech), 100 U/ml penicillin and 100ug/ml streptomycin. For 293A, HeLa and Vero-rKSHV.219 cell lines DMEM medium (Invitrogen) containing 10% FBS (Clontech), 100 U/ml penicillin and 100ug/ml streptomycin was used. 1 ug/ml of Doxycycline (Dox) was used to induce the expression of myc/His-tagged Rta in TRExBCBL1-RTA. JSC-1 was stimulated with 5, 1, 0.2 or 0.04 uM of 3-deazaneplanocin (DZNEP) or 200 nM of staurosporine (STS).

Plasmids, transfection and luciferase assay

Expression plasmids HA-JMJD3 and HA-UTX were obtained from Kristian Helin (University of Copenhagen). HA-JMJD2A expression vector was gift from Yang Shi (Harvard University). HA-UTXm was generated by replacing the C terminus of the UTX in HA-UTX with PCR cloning to change histidine to alanine at amino acid 1146. pLuc3kb, pLuc1.9kb and pLuc0.9kb reporter plasmids were generated by inserting the Luciferase gene derived from pGL3-Basic and different sized fragments of the promoter region of KSHV Rta into pcDNA5/FRT. 293A and Vero.219 cells were transfected by Polyfect (Qiagen) according to the manufactrer's specifications. Luciferase assay was performed by using the Promega system. The luciferase activity values are the average of at least three independent experiments.

Immunofluorescent analysis

Vero cells were transfected with plasmids expressing HA-tagged JMJD2A, JMJD3, UTX or UTXm respectively. 3 days post transfection, cells were fixed by 4% paraformaldehyde and then permeabilized by 0.2% Triton X100. 10% goat serum was used for blocking aspecific binding of antibodies followed by incubation of cells with antibodies against HA-tag, H3K9me3 and H3K27me3. After extensive washing with PBS, FITC- and TRITC-conjugated secondary antibodies were applied followed by Hoechst staining.

DNA affinity purification assay

Biotinylated DNA was made with biotin-conjugated primers and PCR amplification of sequences that are derived from the KSHV RTA promoter. KSHV (U75698.1 at GenBank) coordinates for fragment A and B are 69861–70500 and 70511–71130, respectively. The assay was performed essentially as described by Atanasiu et. al. [75] except that streptavidine resin (Stratagene) was used to pull down the DNA/protein complexes and proteins were eluted from the streptavidine resin with Laemmli sample buffer (Sigma).

Antibodies for ChIPs and western blotting

The following antibodies were used in ChIPs and/or western blotting: rabbit anti-histone H3 (Abcam ab1791), rabbit anti-H3K27me3 (Millipore 07-448), rabbit anti-H3K9me3 (Millipore 07-442), rabbit anti-H3K4me3 (Millipore 04-745), rabbit anti-acetyl-histone H3 (AcH3) (Millipore 06-599), rabbit anti-RNA polymerase II (H-224) (total RNAPII) (Santa Cruz sc-9001), mouse anti-RNA polymerase II (CTD4H8) (RNAPII Ser5) (Millipore 05-623), rabbit anti-CBF1 (Abcam ab25949), mouse anti-EZH2 (BD Biosciences 612666), rabbit anti-SUZ12 (Abcam ab12073). For western blotting the following antibodies were used: mouse anti-PARP (BD Biosciences 556494), mouse anti-actin (Abcam), anti-JMJD2A (Bethyl A300-861A), anti-JMJD3 (Abgent AP1022a), KSHV specific antibodies such as mouse anti-Rta gift from Koichi Yamanishi (Osaka University, Japan), rabbit anti-K8.1, mouse anti-K8 (Abcam ab36617), rabbit anti-K2 (ABI 13-214-050), rat anti-LANA (ABI 13-210-100), Epstein Barr Virus specific antibody such as mouse anti-Zebra (Zta) (Argene 11-007).

Chromatin Immunoprecipitation assay (ChIP)

TRExBCBL1-RTA cells (4×10⁷) were treated for 6, 12 and 24 hr with 1 ug/ml doxycycline. After fixation of the cells with 1% (v/v) formaldehyde for 10 min at RT, the cross-linking was stopped by adding glycine (final concentration 125 mM) for 5 min at RT. Cells were washed (3×) with cold PBS and then resuspended in Cell Lysis Buffer (5 mM Tris-HCl, pH 8.0, 85 mM KCl, 0.5% NP40, 1× protease inhibitor cocktail (Roche)) and incubated on ice for 10 min. After centrifugation (5 min, 5000 rpm at 4°C), the pellet was resuspended in 1.8 ml of RIPA buffer (10 mM Tris-HCl, pH 8.0, 1 mM EDTA, pH 8.0, 140 mM NaCl, 0.1% SDS, 0.1% sodium deoxycholate, 1% Triton X-100, 1 mM PMSF, 1× protease inhibitor cocktail), sonicated and centrifuged at 13000 rpm for 10 min at 4°C to remove cell debris. Aliquots of the supernatant (cellular and viral chromatin) were stored at −80°C.

For preparation of input DNA 20ul of chromatin was incubated in 100ul of TE buffer containing 50ug/ul RNase A for 30 min at 37°C. The samples were then adjusted to contain 0.5% SDS and 0.5mg/ml Proteinase K (Invitrogen) and incubated for 1 hr at 37°C. Formaldehyde crosslinks were reversed by adding sodium chloride (final concentration 300 mM) to the samples and incubated overnight at 65°C. DNA was extracted first by one volume of phenol/chloroform/isoamyl alcohol (25∶24∶1) saturated with 10 mM Tris, pH8.0 and 1 mM EDTA and then purified once by one volume of chloroform. DNA was precipitated by cold absolute ethanol, 10% (v/v) of 3 M sodium acetate, pH 5.2 and 1ul of 15 mg/ml Glycogen Blue (Ambion) at −80°C at least for 1 hr following by wash with 70% ethanol. Finally the input DNA was dried at RT and resuspended in 20ul of water.

For ChIPs chromatin containing 10ug of DNA was first diluted in 500ul of RIPA buffer and precleared by Sepharose A beads. Immunoprecipitation was carried out with 1–2ug of antibodies overnight at 4°C. Next day to pull down the DNA/protein complexes, Protein-A/G agarose was added for 4 hr. Immunoprecipitation was washed sequentially with RIPA buffer once briefly and once for 10 min followed by washing with LiCl buffer (10 mM Tris-HCl, pH 8.0, 1mM EDTA, pH 8.0, 250 mM LiCl, 0.5% NP-40, 0.5% sodium deoxycholate) once for 10 min and with TE buffer two times for 10 min. The DNA/protein-Protein A/G agarose complex was resuspended in 100ul of TE buffer containing 50ug/ul RNase A and incubated for 30 min at 37°C. The Proteinase K treatment, crosslink reversal and DNA purification was done exactly as it was described at the preparation of input DNA. Both input and ChIP DNAs were measured by qPCR. Based on the standard curves for each primer pairs the enrichment of proteins and histone modifications on specific genomic regions were calculated as percentage of the immunoprecipitated DNA compared to input DNA. Each data points in ChIP figures were averages of at least three independent ChIPs using three independent chromatins.

Sequential ChIP (seqChIP) was performed as follows. ChIPs were carried out as described above except that after the first ChIP (ChIP I.) the immunoprecipitated DNA/protein complexes were eluted from the Protein A/G agarose by 100ul of Elution buffer (50 mM Tris-HCl, pH 7.5, 10 mM EDTA, pH 8.0, 1% SDS) by heating for 10–15 min at 65°C. 10% of the elution was saved for qPCR while the rest was diluted with RIPA buffer up to 1 ml and the second immunoprecipitation (ChIP II.) was performed exactly as described above for ChIP. The second ChIP was also measured by qPCR and the enrichment of the histone modifications on specific genomic regions was calculated as percentage of the immunoprecipitated DNA compared to the total amount of DNA eluted after the first ChIP. Each ChIP values were averages of three independent ChIPs using three independent chromatins.

Quantitative real-time PCR (qPCR)

qPCR was performed using iQ SYBR Green Supermix (Bio-Rad) and CFX96 real-time PCR machine (Bio-Rad). PCR program was as follows: after an initial preincubation step at 95°C for 3 min, there were 40 cycles, each consisting of 95°C for 10sec, 64.5 or 59°C depending on the primers (Table S1) for 20 sec and 72°C for 20sec. The last amplification cycle was followed by a melt curve analysis to make sure about the specificity of the qPCR amplification. As for ChIP 0.45ul of ChIP DNA and 4.5ng of input DNA were measured in qPCR. Quantification of the amplifications in qPCR was based on standard curves for each primer pairs. Sequences of the primers used in ChIP experiments can be found in Table S1.

RNA isolation and RT-PCR

Total cellular RNA was extracted with Tri reagent (Sigma) according to the manufacturer's instructions. 1ug of total RNA was treated by DNase I (Sigma), reverse transcribed by iScript cDNA Synthesis kit (Bio-Rad) and the cDNA was measured by either conventional PCR or qPCR. The relative quantification of gene expression was calculated with the ddCt method, where actin mRNA was used for normalization. RT-PCR graphs were made based on the average of at least two independent experiments.

Lentiviral shRNA knockdown

The shRNA constructs were prepared by using the pLKO.1 lentiviral vector. The target sequences are listed in Table S1. Supernatants from 293T cells transfected by the shRNA and packaging vectors were collected 60 hours post transfection followed by concentration of the virus (24000 rpm, 1.5 hr, 4°C) and used for spinning infections (1800 rpm, 45 min) of one million of BCBL-1 cells in the presence of 10 ug/ml polybrene. 5 days post infection cells were harvested for western blot and RT-qPCR analysis.

KSHV tiling microarray and ChIP-on-chip

The 15 bp-tiling KSHV microarray contains both KSHV and human probes and was manufactured by Agilent Technologies. The probes were spotted on 8×15K array format. The majority of the oligonucleotides are overlapping 60-mer probes covering the entire KSHV genome (U75698 and U75699 at GenBank). In addition, probes specific for the gene regulatory regions of 71 human genes were also included in the microarray. The human probes are derived from the Agilent Human CoC 2×244k microarray design. For ChIP-on-chip 30 ug of chromatin and 5–6 ug of antibodies were used per ChIP. The ChIP and DNA purification was performed exactly as described above. ChIP DNA was amplified by Complete Whole Genome Amplification (WGA2) kit (Sigma-Aldrich) and purified with QIAquick PCR Purification kit (Qiagen) according to the instructions of the manufacturers. Labeling, hybridisation and scanning of the microarrays were done at the Functional Genomics Core, Microarray Service at City of Hope (California). Probe signals were extracted using the Agilent Feature Extraction software (version 10.5.1.1). Each ChIP-on-chip experiment was performed two times and the average of the biological replicates is shown in Figures 1, 2, 4A, S5, S6, S8. The biological replicates can be found in Figures S3, S4, S5.

Normalization, visualization and cluster analysis of the ChIP-on-chip data

Log₂ (Cy5/Cy3) of the mean signals of the probes were calculated, and imported into R software (version 2.9.0) (http://www.r-project.org). Using the Bioconductor package “aroma.light,” lowess normalization as implemented by the function “normalizeLoess,” was used to correct for die bias. The log₂ ratios for samples at 0hpi and 12hpi were scaled to have the same median absolute deviation (MAD) using the function “normalizeBetweenArrays,” from the “limma” Bioconductor package [76], [77], [78], [79].

To visualize enrichment across the KSHV genome, the normalized log₂ ratios for each overlapping probe were averaged and a score was assigned to each base pair (bp). A moving average was calculated across the genome by sliding a 150-bp window stepwise by 1 bp unidirectionally. When calculating the moving average, the window is shifted past regions where there are missing probes so that regions with missing probes do not have any weight. Regions where no probes exist were assigned an enrichment value of “0.” Log₂ ratios were converted to linear scale, and tracks in BED format were created for each sample. Tracks were then imported into Genetrix software (Epicenter Software) for visualization.

For the KSHV promoter array analysis promoter regions were defined as 1000 bp upstream and downstream of the translational start sites (TSS) of each viral gene. The coordinates of ORFs are listed in Table S2 [80]. Normalized log₂ ratios were calculated as described above. Log₂ ratios were averaged for each 50 bp nonoverlapping window upstream and downstream of the TSS within the promoter for each gene. The TSS was assigned a single score. The resulting matrix was applied to perform hierarchical clustering with Cluster 3.0 (http://bonsai.ims.u-tokyo.ac.jp/~mdehoon/software/cluster/software.htm). The results of clustering were then imported into Java TreeView for visualization (version 1.1.4r2) [81].

While the 0 hpi-ChIPs were compared to the 0 hpi-DNA inputs in ChIP-on-chips the 12 hpi-ChIP-on-chips were analysed in two different ways. (1) 12 hpi-ChIPs were compared to 0 hpi-ChIPs (Figure S3, S4, S5, S6, S7, S8, S10) or (2) the 12 hpi-ChIP/input ratio was calculated by multiplying the 0 hpi-ChIP/input ratios by the 12 hpi-ChIP/0 hpi-ChIP ratios (Figure 1 and 2).