Genome-wide siRNA screen identifies SMCX, EP400, and Brd4 as E2-dependent regulators of human papillomavirus oncogene expression

Abstract

An essential step in the pathogenesis of human papillomavirus (HPV)–associated cancers is the dysregulated expression of the viral oncogenes. The papillomavirus E2 protein can silence the long control region (LCR) promoter that controls viral E6 and E7 oncogene expression. The mechanisms by which E2 represses oncogene expression and the cellular factors through which E2 mediates this silencing are largely unknown. We conducted an unbiased, genome-wide siRNA screen and series of secondary screens that identified 96 cellular genes that contribute to the repression of the HPV LCR. In addition to confirming a role for the E2-binding bromodomain protein Brd4 in E2-mediated silencing, we identified a number of genes that have not previously been implicated in E2 repression, including the demethylase JARID1C/SMCX as well as EP400, a component of the NuA4/TIP60 histone acetyltransferase complex. Each of these genes contributes independently and additively to E2-mediated silencing, indicating that E2 functions through several distinct cellular complexes to repress E6 and E7 expression.

Papillomaviruses infect squamous epithelial cells and cause a variety of epithelial lesions (1). A subset of the more than 140 human papillomavirus (HPV) types infect mucosal squamous epithelia, including that of the anogenital tract; these are classified as either low-risk or high-risk, depending on the lesions associated with infection. Infections with high-risk HPVs (i.e., HPV16 or HPV18) cause squamous intraepithelial lesions that can progress to cancer, most notably cervical cancer. HPV is considered the major cause of human cervical cancer, the second most common cancer in women worldwide (2).

The viral proteins E6 and E7 account for the oncogenic potential of high-risk HPVs in part through their ability to target and degrade p53 and Rb, respectively (1). The long control region (LCR) is the upstream enhancer and promoter region that drives transcription of E6 and E7. An early step in cervical carcinogenesis frequently involves the integration of HPV DNA into cellular chromosomes in a manner that disrupts the E1 and/or E2 ORF (3 –6). Because E2 is capable of repressing expression from the LCR, its loss leads to the increased, dysregulated expression of the HPV oncogenes E6 and E7 (7 –9). Expression of E2 in HPV-positive cervical cancer cells causes a growth arrest and senescence (10, 11) due to E2-mediated repression of E6 and E7 expression (12, 13). The E2 protein consists of an N-terminal transactivation domain and a C-terminal DNA binding and dimerization domain (14, 15). The E2 proteins from different papillomavirus types are well conserved both functionally and at the amino acid level. The mucosal HPV LCRs contain four E2 binding sites (E2BS) and it has been hypothesized that E2 binding to the two promoter proximal binding sites within the LCRs competes with the binding of cellular transcription factors Sp1 and TBP at the promoter (16, 17). Yet, this steric hindrance model cannot fully explain E2-mediated transcriptional silencing; forms of E2 that can still bind the promoter but contain deletions or specific amino acid substitutions in the E2 transactivation domain are unable to repress E6/E7 transcription (11, 18, 19). This suggests that E2 silencing involves the recruitment of specific factors to the LCR. Although the bromodomain protein Brd4 has been previously implicated in E2-mediated transcriptional repression of the LCR (20), other work suggests that additional factors are also involved in E2-mediated repression (21).

To develop a more comprehensive understanding of E2-mediated repression of the E6/E7 promoter, we performed an unbiased, genome-wide siRNA screen to identify cellular genes that repress expression from the HPV18 LCR. From more than 21,000 human genes, the screen identified 96 that are involved in repression of the HPV18 LCR. Although the majority of these cellular genes function in concert with E2, some act independent of E2 to repress the HPV18 LCR. Several Gene Ontology (GO) processes, including transcription, chromatin modification and DNA replication, were over-represented in the list of 96 repressors. Among these genes, we validated the cellular demethylase JARID1C/SMCX and EP400, a component of the NuA4/TIP60 histone acetyltransferase complex that was first identified through its ability to bind the adenovirus E1A protein, as mediators of E2 repression (22 –25). In addition, we confirmed a role for the E2-binding protein Brd4 in transcriptional repression (20). Knock down of each of these proteins individually resulted in partial de-repression of the LCR. Our screen revealed that a single cellular protein or pathway was not sufficient for E2’s full transcriptional repression of the HPV18 LCR. Instead, our findings support the hypothesis that E2 recruits and coordinates the activities of several distinct cellular pathways to down-regulate the expression of the papillomavirus oncogenes.

Results

Genome-Wide siRNA Screen.

The initial screen used C33A cells, an HPV-negative human cervical carcinoma cell line engineered to stably express FLAG-HA–tagged bovine PV1 (BPV1) E2 from a bicistronic mRNA that also encodes the IL2 receptor α subunit (IL2Rα) (26, 27). An E2-repressible reporter in which luciferase expression is controlled by the HPV18 LCR was introduced into these cells, and single cell clones were isolated and characterized (28). Transfection with siRNAs targeting BPV1 E2 resulted in a reduction of E2 protein below detection levels and a 7- to 11-fold increase in luciferase activity (Fig. 1A). Luciferase activity was not affected by control siRNA (siC#1) or siRNA against USP15, a cellular protein not involved in E2 repression. The screen protocol was optimized in C33A/BE2/18LCR c4 cells using E2#3 siRNA as the positive control and USP15 siRNA as the negative control. The screen, represented in Fig. S1A, was performed with the Dharmacon human siGENOME SMARTpool library of 21,121 pools. Each pool comprises four independent siRNA duplexes against the target gene. Luciferase activity was measured 72 h post-transfection, and z-scores from three replicate experiments for each SMARTpool were averaged and standard deviations (SD) determined. A SMARTpool was considered a hit if the z-score was ≥2. The hit rate in the primary screen was 2.4%. Of the 511 hits (Dataset S1), indicated by the red box in Fig. 1B, 400 were classified as weak (z-scores ≥2 and <3); 102 as moderate (z-scores ≥3 and <5); and nine as strong hits (z-scores ≥5). Because the screen measured an increase in luciferase activity and positive z-scores, we did not account for SMARTpools that decreased cell viability or negatively affected cellular proliferation. Thus this screen would have missed cellular genes that could be involved in repression of the HPV18 LCR, but, when knocked down, reduced cell viability. To validate hits from the genome-wide siRNA screen, a deconvolution secondary screen was performed in which the four duplexes from each of the 511 SMARTpool hits were individually screened at a final concentration of 20 nM (Fig. S1B). From the 511 primary hits, 391 (77%) validated with at least one siRNA duplex, whereas 120 initial hits failed to validate with any of the individual siRNAs (Dataset S1). The majority of these nonvalidated targets (109 of 120) had been classified as weak in the primary screen. This level of nonconcordance is similar to that observed in other high-throughput screens using the Dharmacon library (29, 30). A total of 231 cellular genes were selected for further analysis (130 that confirmed with two or more duplexes, and 101 that confirmed with one duplex and a z-score ≥5).

Fig. 1.

Genome-wide siRNA screen to identify cellular proteins involved in E2-mediated repression of LCR. (A) Characterization of C33A/BE2/18LCR c4 cell line. Cells were mock-transfected or transfected with the indicated siRNAs at a final concentration of 20 nM. Cell extracts were harvested 72 h post-transfection to determine relative luciferase units (RLU) (Top) or indicated protein levels (Bottom). The experiment was performed in triplicate. Bars indicate mean RLU normalized to protein concentration ± SD; representative Western blots are shown. (B) Primary screen z-score distribution of all SMARTpools analyzed. Each Dharmacon SMARTpool library plate was screened in triplicate with data point displaying the mean ± SD. Area boxed in red indicates the 511 SMARTpools with an average z-score ≥2. (C) C33A/16E2/18LCR c1 cell line was characterized as described in A. Bars illustrate the mean of three experiments ± SD. Asterisk indicates a nonspecific band in the HA blot for HPV16 E2.

Additional Secondary Screens Identified 96 Cellular Genes that Contribute to Repression of HPV18 LCR.

Our primary screen used C33A cells that expressed BPV1 E2 because repression of the HPV18 LCR by BPV1 E2 has been well studied (7, 9, 17). To identify cellular proteins that contribute to the repression mediated by the E2 protein from a high-risk HPV type, we determined which of these 231 hits were also involved in HPV16 E2-mediated repression of the HPV18 LCR. We generated the C33A/16E2/18LCR c1 clonal cell line to conduct this screen. No significant change in luciferase activity was observed with siC#1 or siRNA against USP15 (Fig. 1C). However, transfection with siRNAs against HPV16 E2 resulted in a dramatic decrease in HPV16 E2 protein levels and a 13- to 17-fold increase in luciferase activity. To perform the secondary screen, we transfected C33A/16E2/18LCR c1 cells with each of the four individual duplexes against the 231 hits and measured luciferase activity 72 h post-transfection. Based on these results (Dataset S2), the list of candidate genes was reduced to 141 that scored in both BPV1 E2 and HPV16 E2 repression screens.

To identify cellular genes that scored as hits because they either directly or indirectly decreased E2 protein levels, we performed a quantitative in-cell western analysis. C33A/BE2/18LCR c4 cells transfected with individual siRNAs were fixed, permeabilized, and stained to measure E2 expression at 72 h post-transfection. An experimental plate from this analysis is displayed in Fig. S2. Cells were stained with Alexa680-succinimidyl ester to label lysines within cells for normalization (red channel, Fig. S2A) and immunostained for HA to detect E2 protein (green channel, Fig. S2B). Figure S2C is a composite image of S2A and S2B. Twenty genes negatively affected E2 expression and/or stability because their siRNA-mediated knock down led to decreased E2 protein levels (Dataset S2). One of these genes was IL2Rα. As the E2 and IL2Rα proteins were translated from the same mRNA via an IRES in the bicistronic reporter, this was an expected result that validated the assay. Because decreased E2 levels could result in a de-repression of transcription from the HPV18 LCR, we eliminated these candidates from further analysis. Based on manual examination of the candidate list, we excluded an additional 25 cellular genes because only one of the four duplexes consistently scored positive (15/25) or the intended target of the duplexes (10/25) was unclear (i.e., Entrez Gene ID was discontinued, predicted, hypothetical, or inferred).

Next, we determined which of the remaining cellular genes repressed the HPV18 LCR in an E2-dependent manner and which were capable of repressing transcription from the HPV18 LCR in the absence of E2. We generated a clonal C33A cell line (C33A/18LCR c4) that stably expressed the HPV18 LCR-luciferase reporter in the absence of any papillomavirus E2 protein. Transient transfection of a BPV1 E2 expression construct confirmed that the HPV18 LCR in these cells could be repressed 3-fold by E2 (Fig. S3A). Introduction of siRNA against luciferase mRNA suppressed luciferase activity 6-fold (Fig. S3B), and luciferase levels were unaffected by siC#1 or USP15 siRNA. We introduced individual siRNA duplexes against the 96 remaining candidate genes into these cells, and determined that 58 of the 96 were dependent on E2 for HVP18 LCR repression and 34 functioned in an E2-independent manner. The E2-dependence of four cellular proteins remained unclear from this analysis (Dataset S3).

Of 22 broad GO processes, the 96 cellular genes identified were significantly enriched in five GO processes at a P value < 0.01 (Table S1). Because our screen was designed to identify cellular proteins that play a role in transcriptional repression of the HPV18 LCR, it was not surprising to find enrichment in biosynthesis, cell cycle, chromatin modification, DNA replication, and transcription.

Validation of E2-Dependent Hits in Repression of HPV18 LCR Using a Non–High-Throughput Format.

From the 58 E2-dependent hits, we focused our validation studies on three cellular proteins, namely, Brd4, EP400, and SMCX. The SMARTpools against each of these three proteins were strong hits in the primary screen and all four individual duplexes scored positive in the secondary screens with either BPV1 E2 or HPV16 E2. EP400, a known component of multiple histone acetylase complexes, was of interest because it has been described to interact with viral proteins from adenovirus and SV40 and was found to interact with five other E2-dependent hits identified in our screen (Fig. S4) (24, 25, 31, 32). SMCX was selected based on its role as a histone demethylase associated with transcriptional repression (22, 23, 33). The E2-interacting bromodomain protein Brd4 was selected for further analysis, as it has previously been implicated in E2-dependent transcriptional repression (20). To validate cellular genes identified in the high-throughput screens, individual siRNA duplexes were tested for their effect on luciferase activity in C33A/16E2/18LCR c1 cells in a non–high-throughput format. As shown in Fig. 2, each of the four siRNA duplexes against Brd4, EP400, or SMCX efficiently reduced target protein levels. The siRNA-mediated knock down of Brd4 increased HPV16 E2 protein levels (Fig. 2A), which is consistent with data from the in-cell western analysis in which Brd4 knock down increased BPV1 E2 protein expression (Dataset S2). Z-scores from the screens correlated with the luciferase activity measured in the individual assays for all three genes. Those duplexes with the greatest z-scores in the screens caused the largest increase in luciferase activity in the non–high-throughput experiments. Several other siRNA duplexes directed against E2-dependent hits (BRD8, CAPN1, CD2BP2, DEFA3, ENPP7, ENY2, EPC, LRP1, PEX14, PPIB, PRAP1, SKP1A) were further analyzed in a non–high-throughput format (Fig. S5). All but one of these siRNAs caused at least a 2-fold increase in luciferase activity when compared with C33A/16E2/18LCR c1 cells transfected with siC#1. Taken together, these experiments validated the high-throughput screens and indicate that a number of distinct cellular proteins and pathways contribute to E2-mediated transcriptional repression of the HPV18 LCR.

Fig. 2.

Brd4, EP400, and SMCX contribute to E2-mediated transcriptional repression of the HPV18 LCR. C33A/16E2/18LCR c1 cells were transfected with the indicated siRNAs at 20 nM. Cell extracts were harvested 72 h posttransfection to determine RLU, as well as HPV16 E2 (HA), actin, Brd4 (A), EP400 (B), and SMCX (C) protein levels. Experiments were performed in triplicate, with each bar representing the average ± SD and representative immunoblots displayed. The long and short isoforms of Brd4 are indicated by the lines in A. Asterisk indicates a nonspecific band in the HA blots for HPV16 E2.

Brd4, SMCX, and EP400 Each Contribute to E2-Mediated Transcriptional Repression of Endogenous HPV18 LCR in HeLa Cells.

HeLa cells are an HPV18-positive epithelial cell line originally derived from a cervical cancer in which E6 and E7 are expressed from the LCR and the E2 ORF has been disrupted (4). The continued expression of E6 and E7 is required for cell growth. Exogenous expression of E2 in HeLa cells arrests growth, inducing apoptosis and/or senescence through repression of E6 and E7 transcription (7, 10, 11, 13, 34). We asked whether the cellular genes that we had identified (i.e., Brd4, EP400, and SMCX) contributed to the ability of E2 to repress transcription from the endogenous HPV18 LCR in HeLa cells. To avoid confounding effects because of E2-induced senescence, we used HeLa/16E6/16E7 cells for these experiments. These HeLa cells had been engineered to express HPV16 E6 and HPV16 E7 from retrovirus promoters that are not subject to E2-mediated repression (35).

HeLa/16E6/16E7 cells stably transduced with an expression plasmid for HPV16 E2 or the control plasmid pOZN were transfected with the indicated siRNAs and harvested 72 h posttransfection. The level of transcription from the HPV18 LCR was determined by quantitative PCR using primers specific for HPV18 E7 mRNA and normalized to the housekeeping transcript G6PD. There was a 4-fold decrease in HPV18 E7 mRNA in the HPV16 E2-expressing cells compared with those with the control plasmid pOZN, indicating that, as expected, HPV16 E2 repressed the endogenous HPV18 LCR. E2-mediated repression was alleviated by siRNA targeting HPV16 E2 (Fig. 3A). HPV18 E7 mRNA levels also increased in the cells that received siRNAs targeting Brd4, SMCX, or EP400, demonstrating that these cellular proteins contribute to repression of the endogenous HPV18 LCR in HeLa cells. HPV18 E7 mRNA levels remained largely unchanged in the similarly treated control HeLa/16E6/16E7/pOZN cells (Fig. 3B). We did observe a modest 2-fold effect with the siRNA duplex to SMCX in the control HeLa/16E6/16E7/pOZN cells, suggesting that SMCX may have E2-dependent as well as E2-independent roles in repression of the HPV18 LCR in HeLa cells. This differs from the results in C33A cells where the siRNA duplex to SMCX did not affect transcription from the HPV18 LCR luciferase reporter in C33A/18LCR c4 cells (Dataset S3), no doubt reflecting some level of cell line variability. Knock-down efficiency of the intended targets was confirmed by Western blot analysis (Fig. 3 C and D). Taken together, these results further validate the genes identified in the high-throughput screens and indicate that Brd4, SMCX, and EP400 each contribute to the ability of HPV16 E2 to inhibit transcription from the endogenous HPV18 LCR in HeLa cells.

Fig. 3.

Brd4, EP400, and SMCX are E2-dependent corepressors of the HPV18 oncogenes in HeLa cells. HeLa/16E6/16E7/16E2 (A and C) or HeLa/16E6/16E7/pOZN cells (B and D) were transfected with the indicated siRNAs at a final concentration of 50 nM and cells were harvested at 72 h post-transfection. (A and B) HPV18 E7 transcript levels, normalized to G6PD transcript levels, were determined by qRT-PCR. Each siRNA transfection was performed at least three times; bars indicate the mean RNA level ± SD relative to values obtained in siC#1-transfected cells. (C and D) Cell extracts were separated by SDS/PAGE and E2, Brd4, SMCX, EP400, and actin protein levels were visualized by immunoblot analysis. Representative immunoblots are displayed.

Brd4, SMCX, and EP400 Have Additive Effects on E2-Dependent Transcriptional Repression of HPV18 LCR.

The genome-wide siRNA screen and subsequent secondary screens illustrated that E2-mediated repression of the HPV18 LCR involves several cellular proteins. Because knock down of any individual cellular protein did not decrease E2-mediated repression to the same extent as the knock down of either BPV1 E2 or HPV16 E2 in C33A cells, we hypothesized that E2-mediated repression was not mediated by a single cellular protein. To determine whether Brd4, SMCX and EP400 have additive effects on E2-mediated repression, siRNAs were transfected into C33A/16E2/18LCR c1 cells singly or in combination and luciferase levels determined 72 h post-transfection (Fig. 4). Neither siC#1 or GAPDH siRNA changed luciferase activity. In contrast, there was an 18-fold increase in luciferase activity when cells were transfected with 16E2#2. Consistent with previous experiments, transfection with siRNAs against Brd4, EP400, or SMCX significantly increased luciferase activity but not to the same extent as 16E2#2 siRNA. Varying the final siRNA concentration did not alter the level of knock down detectable by Western blot analysis, nor did it significantly affect the change in luciferase activity observed. When Brd4, EP400 and SMCX siRNAs were transfected into C33A/16E2/18LCR c1 cells in pair-wise combinations, the effect on luciferase activity was greater than that achieved with any of the siRNAs individually. The increase in luciferase levels in Brd4/EP400 combined or Brd4/SMCX combined siRNA-transfected cells was comparable to that observed with HPV16 E2 siRNA alone. When any of these three duplexes was individually paired with siC#1 or GAPDH siRNA, luciferase levels were similar to those observed in cells transfected with Brd4, EP400 or SMCX siRNA alone. Luciferase activity, and hence the decrease in E2-mediated transcriptional repression, was greatest in C33A/16E2/18LCR c1 cells simultaneously transfected with Brd4, EP400, and SMCX siRNAs. Adding siC#1 or GAPDH siRNA did not further enhance luciferase levels, confirming that the changes in luciferase levels are caused by the knock down of these specific proteins. We speculate that the alleviation of E2-mediated repression in cells simultaneously transfected with siRNAs against Brd4, EP400, and SMCX is greater than in those transfected with siRNA against HPV16 E2 because E2 levels are reduced, but not eliminated, by siRNA knock down and a certain level of E2-mediated repression likely remains in siRNA-treated cells. It is possible that decreasing the expression of three of the major effectors to E2-mediated repression results in a more dramatic phenotype than reduction of E2 levels alone. Western blot analysis confirmed that each siRNA efficiently knocked down its target protein, even when added to cells in combination (Fig. 4). The additive effect of knocking down Brd4, EP400, and SMCX suggests that E2 recruits multiple proteins to the HPV18 LCR to mediate repression of its oncogenes.

Fig. 4.

Brd4, EP400, and SMCX contribute additively to E2-mediated transcriptional repression. C33A/16E2/18LCR c1 cells were transfected with the specified siRNAs at the concentrations indicated or a concentration of 10 nM/siRNA duplex in the combined transfection assays. After 72 h, cell extracts were harvested, RLU determined and protein expression examined by Western blot. The experiment shown was performed in triplicate and corresponding immunoblots are displayed. Each bar indicates the mean ± SD.

BPV1 E2 and HPV16 E2 Bind SMCX and EP400.

Because Brd4, EP400, and SMCX contribute to E2-dependent repression of the HPV18 LCR and an interaction between E2 and Brd4 is well established (27, 36), we asked whether SMCX and/or EP400 complex with E2. C33A cells were transfected with HA-tagged BPV1 E2 alone or in combination with FLAG-tagged SMCX. Cell lysates were harvested 48 h post-transfection, and proteins were immunoprecipitated with anti-HA or anti-SMCX antibodies. Input and precipitated proteins were separated by SDS/PAGE and visualized by immunoblot analysis. Immunoprecipitation with an anti-HA antibody efficiently pulled down the HA-tagged BPV1 E2 protein (Fig. 5A) as well as Brd4, as expected. Some SMCX also immunoprecipitated with BPV1 E2, indicating that these proteins are present in a complex, perhaps together with other cellular proteins. SMCX was not detected in the immunoprecipitates from cells transfected with the control plasmid pOZN, indicating that the coimmunoprecipitation of SMCX and BPV1 E2 is specific. These results were confirmed in a reciprocal immunoprecipitation of SMCX where the presence of BPV1 E2 was detected in cells cotransfected with both BPV1 E2 and SMCX (Fig. 5B). No E2 was detected in lysates immunoprecipitated with control IgG. An interaction between HPV16 E2 and SMCX was detected in immunoprecipitates from C33A cells cotransfected with FLAG-tagged HPV16 E2 and HA-tagged SMCX (Fig. S6A). To determine whether E2 is able to form a complex with EP400, C33A cells were transfected with HA-tagged BPV1 E2 and FLAG-tagged EP400 (Fig. 5C). Cell lysates were harvested 48 h posttransfection. Immunoprecipitation with an anti-HA antibody efficiently pulled down the HA-tagged BPV1 E2, as well as EP400, indicating E2 and EP400 are present in the same complex. EP400 also coimmunoprecipitated with HPV16 E2 in cotransfected cells (Fig. S6B). We conclude that E2 forms a complex or complexes with SMCX and EP400, and note that these interactions may be direct or require additional cellular proteins.

Fig. 5.

SMCX and EP400 form a complex with E2. (A and B) C33A cells were transfected with expression plasmids for HA-BPV1 E2, FLAG-SMCX, or the corresponding parental plasmids. At 48 h post-transfection, cells were harvested and proteins immunoprecipitated with the indicated antibodies (HA, SMCX, IgG). Total cellular proteins and bound proteins were separated by SDS/PAGE and visualized by immunoblot analysis using antibodies against HA (detection of BPV1 E2), SMCX, Brd4, and actin. (C) C33A cells were transfected with expression plasmids for HA-BPV1 E2, FLAG-EP400, or the corresponding control plasmids and cell extracts were harvested 48 h posttransfection. Proteins were immunoprecipitated and visualized as described in A, with an anti-HA antibody used for detection of BPV1 E2 and an anti-EP400 antibody used for detection of EP400.

Discussion

To investigate the mechanisms by which the papillomavirus E2 protein controls expression of E6 and E7, we performed an unbiased, genome-wide siRNA screen to identify cellular genes that contribute to repression of the LCR promoter. We identified 511 SMARTpools that, when introduced into cells, resulted in de-repression of the HPV18 LCR (Fig. 1B and Dataset S1). We then performed a series of secondary screens to focus on genes whose involvement in transcriptional repression was conserved among different E2 proteins and did not decrease E2 protein levels (Dataset S2). This left 96 cellular genes that contribute, either directly or indirectly, to repression of the HPV18 LCR. Using C33A/18LCR cells not expressing any papillomavirus E2 protein, we determined that 58 genes contributed to repression of the HPV18 LCR in an E2-dependent manner and that 34 functioned in the absence of E2; the E2-dependence of 4 of the candidate genes remained unclear (Dataset S3).

We focused our subsequent studies on three E2-dependent cellular genes, Brd4, EP400, and SMCX. A protein interactome consisting of the E2-dependent candidate corepressors suggested a central role for EP400 in E2-mediated transcriptional repression (Fig. S4). A member of the SWI2/SNF2 family of chromatin-remodeling proteins, EP400 was first identified through its interaction with adenovirus E1A and was subsequently shown to bind SV40 large T antigen (25, 31, 32). EP400 is a component of the NuA4/TIP60 histone acetylase complex (24, 25, 37). Three other members of NuA4 (i.e., BRD8, EPC1, and YEATS/GAS41) were hits in our primary screen, indicating that this complex is involved in E2-mediated repression of the HPV18 LCR. TIP60, the histone acetyl transferase component of the NuA4/TIP60 complex, specifically acetylates histone H2A on K5, K14 of histone H3, as well as K5, K8, K12, and K16 of histone H4 (38). These are the same acetylated lysine residues that Brd4 binds preferentially during interphase and mitosis (39). Evidence suggests that histone acetylation may function as a recruitment signal for proteins involved in E2-mediated repression of the LCR (20); it is reasonable to speculate that one role of the NuA4/TIP60 complex in E2-mediated transcriptional repression may be to facilitate Brd4 recruitment to the HPV18 LCR.

Although the NuA4/TIP60 complex is frequently associated with transcriptional activation, EP400 has been shown to function as a transcriptional repressor, at least partly by antagonizing TIP60 function (40 –42). EP400 represses specific subsets of cellular genes in an adenovirus E1A-dependent manner (43, 44). Here we show that E2 and EP400 coimmunoprecipitate with one another (Fig. 5C and Fig. S6B). We propose that the papillomavirus E2 protein may alter EP400 complex formation and direct EP400 as well as other components of the NuA4/TIP60 complex to novel targets within an infected cell. Relocation of these proteins to the HPV18 LCR could result in transcriptional inhibition of the E6 and E7 oncogenes.

The second protein selected for more detailed study was SMCX, also known as JARID1C. SMCX possesses demethylase activity against tri- and di-methylated H3K4 (22, 23). The association of SMCX with DNA-binding proteins, including REST, is thought to contribute to the transcriptional repression of other genes (23). In addition, SMCX inhibits the transcriptional activity of Smad3 in a manner independent of its enzymatic activity (33). Here, we show that this demethylase and transcriptional repressor is involved in E2-mediated repression of the papillomavirus oncogenes and establish that E2 and SMCX are found in the same complex (Fig. 5 A and B and Fig. S6A). We hypothesize that E2 recruits SMCX to the HPV18 LCR to maintain a transcriptionally inactive promoter. In support of this idea, specific nucleosome structures along the HPV16 and HPV18 LCRs have been shown to play a role in E6 and E7 repression (45).

Finally, our third target for further investigation was the E2-interacting protein Brd4, which has previously been implicated in several aspects of the papillomavirus replication cycle —initially in E2-mediated genome maintenance and subsequently in E2’s transcriptional activation function (27, 46). Its role in E2-mediated transcriptional repression was first documented in biochemical studies by Wu et al., which indicated that Brd4 was necessary and sufficient for E2-mediated repression of the HPV11 LCR (20). Previous work from our laboratory suggested that there are Brd4-independent mechanisms for E2-mediated transcriptional repression (21). In these studies, neither shRNA-mediated knock down of Brd4 nor expression of the C-terminal domain (CTD) of Brd4 restored LCR activity in the presence of E2. In the present study, we confirm a role for Brd4 in repression of the HPV oncogenes and show that it is one of several cellular proteins necessary for E2-mediated transcriptional repression. Although the siRNA-mediated knock down of Brd4 resulted in a partial alleviation of repression of the HPV18 LCR, it did not abolish E2-mediated repression. Taken together, these studies suggest Brd4-dependent as well as Brd4-independent mechanisms of E2-mediated transcriptional repression.

There are two reasons that may explain why the previous study from our laboratory did not identify a Brd4-dependent contribution to E2-mediated transcriptional repression (21). First, the shRNA-mediated knock down used in the former experiments was not as effective at knocking down Brd4 when compared with the Brd4 siRNAs used in this study. Second, the Brd4-CTD has been shown to function in a dominant negative manner in other E2-mediated functions, and the former study assumed that it would also function as a dominant negative if Brd4 contributed to E2-mediated transcriptional repression. We now believe that the Brd4-CTD does not function as a dominant negative for Brd4’s role in E2-mediated repression of the HPV18 LCR, and this is supported by the recent work of Yan et al. (47).

Our study establishes that the papillomavirus E2 protein uses multiple cellular proteins to inhibit expression of its oncogenes. Several of these control transcription through histone modification. Network analysis provides evidence for interactions among the top 96 cellular candidates that we identified, perhaps best reflected by the finding that four members of the NuA4/TIP60 histone acetylase complex are involved in E2-mediated repression. Although this study has focused on the role of E2 in regulating HPV18 LCR transcriptional activity in C33A and HeLa human cervical cancer cell lines, we plan to extend these studies to normal human squamous epithelial cells and to examine the role of these proteins in E2’s functions in the context of virus/host cell interactions during the course of infection and epithelial cell differentiation. It will also be informative to determine whether these proteins affect other parts of the papillomavirus replication cycle. A deeper understanding of the cellular proteins and pathways involved in silencing the expression of the papillomavirus E6 and E7 oncogenes will provide new insights into the development of therapeutic approaches for HPV-associated malignant and premalignant lesions.

Experimental Methods

Please refer to SI Text for detailed explanations of experimental methods, including the following: cell culture, cell lines and DNA plasmids; siRNAs and transfections; quantitation of luciferase activity; analysis of protein expression; siRNA screens; quantitative in-cell Western analysis; protein–protein interaction network analysis; gene ontology analysis; quantitative real time-PCR; and transient DNA transfections and immunoprecipitations.