Tandemly Integrated HPV16 Can Form a Brd4-Dependent Super-Enhancer-Like Element That Drives Transcription of Viral Oncogenes

ABSTRACT In cancer cells associated with human papillomavirus (HPV) infections, the viral genome is very often found integrated into the cellular genome. The viral oncogenes E6 and E7 are transcribed from the viral promoter, and integration events that alter transcriptional regulation of this promoter contribute to carcinogenic progression. In this study, we detected highly enriched binding of the super-enhancer markers Brd4, MED1, and H3K27ac, visible as a prominent nuclear focus by immunofluorescence, at the tandemly integrated copies of HPV16 in cells of the cervical neoplasia cell line W12 subclone 20861. Tumor cells are often addicted to super-enhancer-driven oncogenes and are particularly sensitive to disruption of transcription factor binding to the enhancers. Treatment of 20861 cells with bromodomain inhibitors displaced Brd4 from the HPV integration site, greatly decreased E6/E7 transcription, and inhibited cellular proliferation. Thus, Brd4 activates viral transcription at this integration site, and strong selection for E6/E7 expression can drive the formation of a super-enhancer-like element to promote oncogenesis.

O ncogenic human papillomaviruses (HPVs) are the cause of cervical cancer, and HPV genomes, which normally replicate extrachromosomally, are often found integrated into the host genome of these cancer cells (1). Commonly, either a single viral genome or multiple tandemly repeated viral genomes are integrated into the host DNA (2). The E6 and E7 oncogenes are expressed from the integrated genomes, most often as a fusion transcript expressed from the 3= junctional copy of HPV and the adjacent cellular DNA (3). Viral genome integration promotes carcinogenesis in a number of ways, but in almost all cases, the cancer-derived cells are dependent on expression of the E6 and E7 oncogenes for continued proliferation. Integration often occurs in the HPV E2 open reading frame, which disrupts the ability of E2 to repress E6 and E7 gene expression (1). The resulting dysregulation of E6 and E7 causes disruption of cell cycle control, leading to genetic instability and carcinogenesis (4). Even when the E2 gene remains intact, methylation of E2 binding sites inhibits binding and renders the viral promoter resistant to E2 regulation (5). Integration events also occur in the E1 gene: this not only removes the downstream E2 gene, but also eliminates the growthsuppressive properties of the E1 protein (6,7). In many cases, the E6-E7 virus-cell fusion transcript expressed from integrated DNA is more stable than the viral message, again increasing E6 and E7 levels (8). Only rarely does insertional mutagenesis result in modified expression of cellular oncogenes or tumor suppressors (9). In this study, we identify an additional mechanism of E6 and E7 oncogene upregulation. We show that multiple tandem copies of integrated HPV16 can act as a Brd4-dependent super-enhancerlike element that drives transcription of the E6 and E7 oncogenes.
Brd4, a double bromodomain and extraterminal domain (BET) protein, plays an essential role in cellular transcription by binding acetylated histones and recruiting positive transcriptional complexes to promoters. Brd4 also plays an important role in transcriptional regulation and replication of papillomaviruses (reviewed in reference 10). The viral E2 protein binds to Brd4 and stabilizes its association with chromatin (11,12). However, in the context of the HPV early promoter, Brd4 and E2 primarily repress viral transcription (13)(14)(15), in part because E2 interacts with the C-terminal domain (CTD) of Brd4 and blocks the formation of the Brd4-pTEFb complex (16). In this way, Brd4 commonly acts as an E2-dependent repressor of E6 and E7 oncogene transcription.
Super-enhancers have been defined as the spatial clustering of large groups of traditional enhancers that control genes responsi-ble for cell identity (17)(18)(19). Super-enhancers are also associated with the expression of oncogenes; for example, a super-enhancer often drives the MYC gene in multiple myeloma (20). Binding of transcription factors, cofactors, and chromatin regulators is enriched at super-enhancers; specifically, super-enhancers can be identified by a high density of mediator, acetylated H3K27, and Brd4 (18,19). Super-enhancers are particularly sensitive to disruption of transcription factor binding. Super-enhancers often drive gene products to which the cells are "addicted," which has great therapeutic potential as it allows targeted disruption of gene expression from super-enhancer-regulated disease-causing genes (20). Here we show that there is highly enriched binding of the super-enhancer marker mediator (MED1), acetylated H3K27, and Brd4 at the integrated locus of HPV16 in 20861 cells. Furthermore, we show that disruption of Brd4 binding to this locus reduces HPV16 oncogene expression, resulting in greatly decreased cellular proliferation and induction of senescence.

RESULTS
Brd4 is highly enriched at the locus of HPV integration in W12 20861 cells. W12 cells were originally derived from an HPV16positive CIN1 (cervical intraepithelial neoplasia grade 1) lesion (21). The original cells consisted of a mixed population of cells containing either extrachromosomal or integrated HPV16 genomes, but subsequently a series of clonal cell lines were derived that stably maintained the viral genome extrachromosomally or as integrants in the host genome (2). Analysis of Brd4 by immunofluorescence (IF) in the W12 subclone 20861 revealed a large and prominent focus of Brd4 signal in the nucleus of virtually every cell (Fig. 1A). This subclone has approximately 30 copies of the HPV16 genome integrated in tandem at a single locus (2), raising the possibility that the observed Brd4 focus corresponded to the HPV16 integration site. In comparison, as shown in Fig. 1A and B, no similar conspicuous focus of Brd4 could be detected in HPVnegative 1A keratinocytes, the 20863 subclone of W12 (with extrachromosomal HPV16 genomes), the 6E subclone of CIN612 cells (with similarly tandemly integrated copies of HPV31), and the 9E subclone of CIN612 cells (with extrachromosomal HPV31 genomes). To prove that the prominent Brd4 focus corresponded to the HPV16 integration site, immunofluorescent in situ hybridization (immuno-FISH) was performed to simultaneously detect viral DNA and the Brd4 protein. As shown in Fig. 1C, the Brd4 focus overlaps the HPV integration locus in about 94% cells.
Brd4 binding at the HPV integration locus is disrupted by BET inhibitors. The bromodomains of Brd4 interact with chromatin through acetylated histone tails (22). Several inhibitors, which mimic acetylated lysine, have been developed to disrupt the binding of BET bromodomains to chromatin (23,24). To show that Brd4 binding to the HPV integration locus was mediated through acetylated histone tails, 20861 cells were treated with 1 M iBET762 (or controls) for 22 h before detection of Brd4 by immunofluorescence. As shown in Fig. 2, detectable Brd4 binding at the HPV integration locus was reduced to less than 5% of cells.
The HPV integration locus is not a site of genetic instability. Ustav et al. have shown that integrated HPV can undergo multiple rounds of replication in the presence of the E1 and E2 replication proteins, and this induces the recruitment of markers of the DNA damage response (25). We also find that Brd4 can be recruited to HPV integration sites, but only in the presence of E1 and E2 (unpublished data). Integrated HPV genomes generally do not express the E1 and E2 proteins, but nevertheless we examined the 20861 Brd4 focus/HPV integration site for the presence of ␥-H2AX and Rad51, markers of DNA damage and homologous recombination. Only rare (3 to 4%) cells showed any colocalization of Brd4 with these markers (see Fig. S1 in the supplemental material). Therefore, Brd4 binding is not due to genetic instability of the HPV integration locus.
Brd4 does not bind to the HPV integration locus throughout mitosis. We have also shown that Brd4 and HPV1 E2 bind to host chromatin at regions associated with common fragile sites of the host genome that have been designated PEB-BLOCs (for persistent E2 and Brd4 broad localizations of chromatin) (26). Notably, Brd4 binds PEB-BLOCs throughout the cell cycle, in contrast to its interaction with cellular promoter regions from which it is displaced in mitosis (26). Therefore, we analyzed the presence of the prominent speckle of Brd4 in 20861 cells and found that it does not persist strongly throughout mitosis (see Fig. S2 in the supplemental material). We conclude that Brd4 binding to the HPV integration site in 20861 cells is not related to PEB-BLOCs.
Brd4 activates HPV16 transcription in 20861 cells. Based on the findings described above, it seemed most likely that Brd4 was regulating transcription at the integrated HPV locus. Brd4 has been shown to both activate and repress transcription of the HPV early promoter at integrated loci (13,16,27). To analyze whether Brd4 regulated viral transcription from the integrated locus of HPV16, 20861 cells were treated with the bromodomain inhibitor iBET762 (or controls) and the early E6*I spliced transcript levels ( Fig. 3A) were determined by real-time quantitative PCR (qPCR). As shown in Fig. 3B, inhibition of Brd4 binding to chromatin by iBET762 treatment dramatically suppressed viral transcription in 20861 cells. In comparison, E6*I transcription was also measured in the 20863 subclone, which contains extrachromosomal copies of HPV16. Viral transcription also decreased in 20863 cells after iBET762 treatment, indicating that Brd4 also plays an activation role in these cells. However, the fold change in transcription after BET inhibitor treatment was much greater in 20861 cells than in 20863 cells. These findings were also confirmed with a second bromodomain inhibitor, JQ1 (Fig. 3C). These data indicate that Brd4 activates viral transcription from both integrated and extrachromosomal HPV16 in these cells. However, transcription of the integrated HPV locus in 20861 cells is higher and more sensitive to disruption of Brd4 binding than in 20863 cells.
The BET inhibitors, iBET762 and JQ1, are acetyl-lysine mimics that inhibit binding of a specific class of bromodomains to acetylated histones (23,28). Although these compounds bind with highest affinity to Brd4, they can also inhibit binding to close BET family members (29). To further prove that downregulation of HPV16 transcription was due to displacement of Brd4 from the integration locus, we downregulated Brd4 expression using small interfering RNA (siRNA) technology. As shown in Fig. S3 in the supplemental material, downregulation of Brd4 expression also decreased HPV16 E6*I expression 2.8-fold. This confirms that Brd4 activates transcription of integrated HPV16 in 20861 cells.
pTEFb, the transcriptional pause release factor, colocalizes with Brd4 at the HPV integration locus. Brd4 is classically an activator of transcription and promotes transcriptional elongation by interaction with pTEFb (30) and through a number of factors that bind the ET region (31). CDK9 is the catalytic subunit of pTEFb, and so we examined the HPV integration locus for CDK9. As shown in Fig. 4A, cyclin-dependent kinase 9 (CDK9) is highly enriched at the prominent Brd4 focus, implying that it functions to activate transcription at this locus. Colocalization of Brd4 and CDK9 suggests that Brd4 activates viral transcription from integrated HPV16 in 20861 cells by interacting with pTEFb to enhance transcriptional elongation by RNA polymerase II (Pol II).
The HPV integration locus in 20861 cells is bound by factors related to super-enhancers. Super-enhancers are greatly enriched in H3K27ac and have a high density of binding of MED1 and Brd4 (17). They are also often composed of enhancers that have become tandemly repeated (18), and they often drive expression of oncogenes to which the cell is addicted (20). The latter two properties are also characteristics of many HPV integration loci. Therefore, to test whether the 20861 HPV integration locus had these additional characteristics of super-enhancers, we analyzed this region for the presence of H3K27ac and MED1. As shown in Fig. 4B, the HPV16 integration locus is greatly enriched in H3K27ac and MED1, as detected by immunofluorescence. This was further confirmed by chromatin immunoprecipitation (ChIP), which showed great enrichment for all three markers of super-enhancers over the viral genome in 20861 cells (Fig. 5). Not unexpectedly, the strongest signal in the viral genome was observed over the viral upstream regulatory region (URR). As shown in Fig. 5A, the enrichment of the super-enhancer markers exceeded that of known cancer-associated super-enhancers within the CCND2 and FOSL1 gene loci (18,20). Enrichment was also observed over the viral URR in 20863 cells (Fig. 5A). However, displaying the data as the percentage of bound immunoprecipitated chromatin DNA relative to the total amount of input chromatin (percentage of input) does not take into consideration the wide differences in copy number among the integrated and extrachromosomal HPV genomes. Therefore, the data are shown in Fig. 5B adjusted for copy number. This shows the very high enrichment of superenhancer markers bound to each genome copy in the HPV16 integration locus in 20861 cells. When we consider that there are 30 copies of the HPV16 genome in tandem array (2), we can appreciate the extremely strong enrichment, which corresponds to that observed by immunofluorescence (Fig. 4). Therefore, we propose that the tandem array of integrated HPV genomes can form a Brd4-dependent super-enhancer-like element in 20861 cells.
Proliferation of 20861 cells is sensitive to disruption of Brd4 binding. Brd4 binding to the integrated HPV16 genome in 20861 cells is almost completely disrupted by bromodomain inhibitors (Fig. 2), and this disruption greatly reduces viral transcription (Fig. 3). Past studies have shown that HPV cancer cells are addicted to E6 and E7 oncogene production (32)(33)(34). Therefore, we hypothesized that the proliferation of 20861 cells would be particularly sensitive to inhibition of Brd4 binding to viral DNA. To test this, cells were plated at low density, and growth curves were determined in the presence of a titration of the bromodomain inhibitors JQ1ϩ and iBET72ϩ or their respective negative-control stereoisomers (JQ1Ϫ and iBET72Ϫ, respectively). Proliferation of 20861 cells was compared to the growth of 20863 cells, and an HPV-negative cervical cancer cell line, C-33A. As shown in Fig. 6 for iBET72ϩ (see Fig. S4 in the supplemental material for JQ1ϩ), C-33A cells were relatively resistant to disruption of Brd4 binding as measured by colony growth. Both inhibitors dramatically reduced proliferation of 20861 and 20863 cells. This finding mirrors the reduction in HPV16 mRNA levels shown in Fig. 4, where both subclones showed a substantial decrease in E6 and E7 viral transcription. This confirms that growth of HPV-containing 20863 and 20861 cells is dependent on continued E6 and E7 transcription, which is regulated by Brd4 expression and can be modulated by BET inhibitors.
Gene expression analysis confirms that W12 cells undergo senescence in response to decreased E6 and E7 gene expression. Using amplification of papillomavirus oncogene transcripts (APOT), we have determined that in 20861 cells, HPV16 is integrated in chromosome 2 p23.2 (unpublished). (A complete analysis of the integration site sequence will be published at a later date.) The viral splice donor at nucleotide 880 is fused to an intergenic cryptic acceptor. There are no obvious genes encoded in the vicinity of this region that could account for the high occupancy of super-enhancer markers in the integrated viral DNA. To ensure that the sensitivity of W12 20861 cells to the bromodomain inhibitors was due to downregulation of E6 and E7, and not due to downregulation of cellular oncogenes driven by either the HPV16  Table S1 and Fig. S5 in the supplemental material. In general, the vast majority of genes showed a similar response in both cell lines. The few genes that had notable difference in magnitude of gene expression changes are listed in Table S2 and have no obvious connection to cellular proliferation. Therefore, the responses of 20861 and 20863 to BET inhibitors are similar, and growth suppression is primarily due to downregulation of E6 and E7. It is well characterized that downregulation of E6 and E7 expression in cells dependent on their growth results in p21 Cip1 -dependent growth arrest and senescence, and our findings mirror these observations (35)(36)(37). A comparison of expression changes in senescence-related genes identified by Wells et al. (35) is shown in Table S3 in the supplemental material. Although these data cannot show that the BET inhibitors directly and solely inhibit proliferation through downregulation of the E6 and E7 viral proteins, taken together with studies of others (35)(36)(37), it seems likely. Also, in conclusion, while we had hypothesized that the growth of 20861 cells would be more sensitive to inhibition of Brd4 binding than 20863 cells, it seems that any cell line that has become addicted to E6 and E7 oncogene function is exquisitely sensitive to their downregulation.

DISCUSSION
We have shown that the tandemly repeated HPV16 integration locus in W12 20861 cells has properties similar to those described for super-enhancers. The chromatin at this locus is so highly enriched in the super-enhancer markers Brd4, MED1, and H3K27ac, that they are observed as a very prominent nuclear domain by indirect immunofluorescence. HPV cancer cells are "addicted" to HPV E6/E7 oncogene expression, and downregulation of these proteins results in reduction of cellular proliferation and induction of senescence (33,36). Treatment of 20861 cells with acetyllysine histone mimics disrupts Brd4 binding, resulting in decreased HPV16 E6 and E7 transcription and induction of senescence.
The role of Brd4 in regulating E6/E7 transcription has been well studied, but most investigations have focused on the ability of E2 to repress the major early promoter in a Brd4-dependent man- ner (13,14,16,38). E2 recruits Brd4 via the E2 binding sites adjacent to the viral promoter and inhibits transcription, in part because E2 binding to the C terminus of E2 competes with Brd4 recruitment of pTEFb (16). In W12 20861 cells, we show clearly that Brd4 activates transcription from the integrated HPV16 P 97 promoter. Similarly, it has been shown that Brd4 activates the viral early promoter in integrated HPV (in the absence of E2) in both HeLa and Caski cells (16). Notably, in our study Brd4 also activates transcription from the P 97 promoter in the extrachromosomal HPV genomes in 20863 cells, even though E2 is assumed to be present (39). These findings are consistent with those of Helfer et al. (40), who also showed that iBET inhibitors reduced HPV transcription in 20863 cells. Bechtold et al. have demonstrated that the P 97 promoter in the extrachromosomal HPV16 genomes in 20863 cells is impervious to repression by E2, and the authors hypothesize that chromatin renders the promoter inaccessible to E2 binding (39). We also show that disruption of Brd4 binding downregulates E6-E7 transcription and results in growth repression/cell death. As shown by others, and confirmed here by gene expression analysis, this is most likely due to reactivation of the p53 and Rb tumor suppressor pathways leading to induction of irreversible cellular senescence mediated by p21 Cip1 (35)(36)(37). Both 20861 and 20863 cells were sensitive to treatment with BET inhibitors, showing that Brd4 can activate transcription of both integrated and extrachromosomal HPV16 genomes. It has been shown that E2 expression does not inhibit E6/E7 expression in 20863 extrachromosomal genomes (39) and is usually not expressed from integrated genomes.
Both 20861 and 20863 cells are sensitive to treatment with BET inhibitors, and viral transcription is dramatically downregulated in both cases. Furthermore, both integrated and extrachromosomal HPV genomes are enriched in Brd4, MED1, and H3K27ac, so what is the significance of the HPV16 super-enhancer-like element? Super-enhancers are defined by the high intensity of these factors, not just their presence at enhancers. 20861 cells express much higher levels of E7 and have increased plating efficiency and a growth advantage compared to 20863 cells (2). However, although the super-enhancer-like element drives very high levels of E6 and E7, the BET inhibitors efficiently disrupt E6/E7 transcription in both cases, leading to cellular senescence. Almost all HPVcontaining keratinocyte cells are completely dependent on the expression of E6 and E7, and therefore it would be expected that BET  inhibitors could reduce E6/E7 expression (if transcription was dependent on Brd4) and result in growth suppression. Further studies are essential to determine how often Brd4 activates viral transcription in HPV infections and cancers, and if this is widespread, iBET inhibitors could be therapeutic for a wide range of HPVassociated diseases. A very recent study by Groves et al. (using a different series of W12 subclones and with HPV integrated at different sites than 20861 cells) also showed that E6 and E7 expression levels correlate with the balance of active chromatin modifications at the HPV integration site (41). However, as shown here, the prominence of the Brd4 nuclear focus in the 20861 subclone is somewhat unusual. Kalantari and colleagues demonstrated that, compared to other subclones, the integrated HPV16 genomes in 20861 cells were notable for particularly high levels of E6 and E7 expression and absence of DNA methylation (42). In contrast, many other cells such as the well-studied cell line Caski, are highly methylated and only the 3=-junctional copy of the viral genome is transcriptionally active (3). Cellular super-enhancers have also been shown to form in DNA methylation valleys (DMVs) (18). We hypothesize that the tandem copies of unmethylated URR in 20861 cells create a high density of binding sites for transcription activation complexes that include Brd4 and MED1; these act as a super-enhancer-like element for viral transcription from the promoter at the virus-cell junction copy. Thus, the ability of integrated HPV to act as a super-enhancer for viral transcription may depend on the copy number and/or methylation status of the integrated copies. It is becoming apparent that each HPV integration site has unique characteristics that depend on viral and cellular breakpoints, rearrangement, and amplification of viral and cellular junctional sequences and epigenetic regulation (41,43). Furthermore, the organization of integration sites can be dynamic and driven by selection for increased viral oncogene expression (43).
RNA preparation. Total cellular RNA was extracted using an RNeasy minikit (Qiagen). RNA integrity was verified using the Bioanalyzer 2100 (Agilent Technologies). One microgram of total cellular RNA was reverse transcribed using a Transcriptor first-strand cDNA kit (Roche).
IF. For immunofluorescence (IF) analysis, cells were cultured on coverslips and fixed at room temperature in 4% paraformaldehyde (PFA)-PBS for 20 min. Cells were permeabilized in 0.1% Triton X-100 and stained with primary and fluorescent secondary antibodies using standard procedures. Coverslips were mounted in ProLong Gold containing 4=,6diamidino-2-phenylindole (DAPI) for analysis by confocal microscopy.
IF-FISH. Cells grown on coverslips were fixed in cold methanol-acetic acid (3:1) for 3 min and 4% paraformaldehyde-PBS for 20 min. Immunofluorescence was performed as described above. Antibodies were fixed in situ with methanol-acetic acid (3:1) at room temperature for 10 min and 2% paraformaldehyde-PBS at room temperature for 2 min. Cells were treated with RNase A and dehydrated in a 70%, 90%, and 100% ethanol series for 3 min each. Full-length HPV16 DNA was fluorescently labeled using the Ulysis nucleic acid labeling kit (Life Technologies). Seventyfive to 360 ng labeled fluorescence in situ hybridization (FISH) probe in hybridization buffer (Empire Genomics) was added to the coverslip, and DNA was denatured at 75°C for 5 min, followed by hybridization at 37°C overnight. Cells were washed with 1ϫ phosphate-buffered detergent (Qbiogene) for 5 min at room temperature, 1ϫ wash buffer (0.5ϫ SSC, 0.1% SDS) for 5 min at 65°C, and 1ϫ phosphate-buffered detergent (Qbiogene) for 5 min at room temperature. Coverslips were mounted in ProLong Gold containing DAPI for analysis by confocal microscopy.
Confocal microscopy and image analysis. Images were collected using a Leica TCS-SP5 laser scanning confocal imaging system and processed using Leica, Las AF Lite software. Data and images shown were obtained from z-stacks of 10 slices encompassing the entire cell and combined using maximum projection, or from single optical slices (as indicated). Colocalization was determined using Leica Microsystems Las AF Lite software. Single z-stack images were taken at the foci. A region of interest line (shown in white) was drawn simultaneously in all channels (blue, red, and green) so that it spanned the nucleus and intersected the foci (Fig. 4). Data of relative staining intensity versus distance was exported and processed using Microsoft Excel.
Microarray expression analysis. Gene expression analysis was carried out in the Research Technologies Branch, NIAID. RNA integrity was verified using a Bioanalyzer. Five hundred nanograms of RNA was amplified and labeled using the Illumina TotalPrep RNA amplification kit (Applied Biosystems), hybridized to Illumina HumanHT-12 v4 Expression Bead-Chip, and scanned using the Illumina HiScan-SQ. Signal data were extracted from image files with the Gene Expression module (v.1.9.0) of the GenomeStudio software (v.2011.1) from Illumina, Inc. Signal intensities were converted to log 2 scale. Analysis of variance (ANOVA) was performed on the normalized signals to test mRNA expression differences between 20861 and 20863 cells treated with iBET72ϩ and iBET72Ϫ. Significance was determined by a false discovery rate at 0.05 to account for multiple comparisons. Statistical analysis was performed primarily in JMP/Genomics 7.0 (SAS Institute Inc., Cary, NC).
Accession number(s). All data have been uploaded to GEO under accession no. GSE75987.