Human papillomavirus E7 oncoprotein induces KDM6A and KDM6B histone demethylase expression and causes epigenetic reprogramming

Abstract

Human papillomaviruses (HPVs) are small double-stranded DNA viruses that are associated with epithelial hyperplasias. A subgroup of high-risk HPVs is etiologic agents of cervical carcinomas as well as other anogenital cancers and oropharyngeal tumors (reviewed in ref. 1). Because of frequent integration of the viral genome into a host cell chromosome, E6 and E7 are the only viral proteins that are consistently expressed in HPV-associated cancers. E6 and E7 have oncogenic activities, and their expression is necessary for the induction and maintenance of the transformed phenotype (reviewed in ref. 2).

The HPV E6 and E7 proteins lack intrinsic enzymatic activities and do not act as DNA binding transcription factors; rather, they reprogram their host cells by associating with cellular signaling molecules. High-risk HPV E6 proteins target the p53 tumor suppressor for degradation, thereby thwarting p53-mediated transcriptional cytostatic and cytotoxic responses to cellular stress signals. High-risk HPV E7 oncoproteins associate with and degrade the retinoblastoma tumor suppressor (pRB), which acts as a cell cycle-specific repressive subunit of several E2F transcriptional complexes, and thereby, subvert cell cycle-dependent E2F transcriptional activities (reviewed in ref. 2). The HPV E6 and E7 oncoproteins also associate with enzymes that modulate histone acetylation and thus, broadly regulate the transcriptional competence of host cell chromatin (3–8).

We have previously reported that the HPV16 E7 oncoprotein associates with E2F6-containing polycomb transcriptional repressor complexes (PRCs) and that the detection of these complexes is reduced in HPV16 E7-expressing cells (9). PRCs require the histone H3 lysine 27 trimethyl (H3K27me3) mark to associate with and transcriptionally silence chromatin (reviewed in ref. 10). PRCs have been most extensively studied in Drosophila melanogaster (reviewed in ref. 11), where they establish and sustain lineage-specific epigenetic silencing of Homeobox (HOX) genes during development (12, 13). HOX proteins are master regulators of transcriptional programs that create and maintain cellular identities. Other PRC-regulated genes include the INK4A-ARF tumor suppressor locus, which encodes p16^INK4A, an inhibitor of cyclin-dependent kinases (CDK) 4 and 6, and p14^ARF, an inhibitor of mdm2-mediated p53 degradation (14). Silencing through H3K27me3 involves PRC2 and PRC1. The catalytic subunit of PRC2, EZH2, is a methyl transferase that catalyzes di- and trimethylation of H3K27, which is then recognized by PRC1 (reviewed in ref. 15). The repressive H3K27me3 mark can be removed by one of two known histone demethylases, KDM6A (UTX) and KDM6B (JMJD3). KDM6B has been shown to remove H3K27me3 marks from the p16^INK4A promoter during ras/raf oncogene-induced cellular senescence (16, 17), whereas KDM6A has been implicated in the removal of H3K27me3 marks from the promoters of several genes encoding RB-binding proteins (18).

Here, we report that HPV16 E7 expression results in a dramatic reduction of the H3K27me3 mark necessary for the binding of PRC1 through transcriptional induction of the histone demethylases KDM6A and KDM6B. We discovered that increased expression of the cervical carcinoma biomarker p16^INK4A is specifically linked to KDM6B induction. Induction of KDM6B and its transcriptional target p16^INK4A by HPV16 E7 is not dependent on pRB inactivation and hence, is not linked to E2F activation. Because HPV16 E7 simultaneously inactivates the critical mediator of p16^INK4A-induced senescence, pRB, HPV16 E7-expressing cells escape senescence and continue to proliferate, and several known KDM6A- or KDM6B-regulated HOX genes are expressed at higher levels in such cells. Hence, HPV16 E7 expression causes epigenetic reprogramming of host cells at the level of histone methylation. Because these HPV16 E7-induced alterations in H3K27me3 levels and associated transcriptional changes are rapidly reversible on silencing of E7 expression and because depletion of KDM6A and KDM6B inhibit proliferation of CaSki cervical cancer cells, our results suggest that KDM6A and KDM6B may be targets for therapy of HPV-associated lesions and cancers.

Results

H3K27me3 Levels Are Reduced in HPV16 E7-Expressing Primary Human Epithelial Cells.

We have previously shown that HPV16 E7 binds to E2F6-containing PRCs and that detection of these complexes is reduced in HPV16 E7-expressing cells (9). E2F6-containing PRCs bind to H3K27me3 transcriptional repressive marks. To determine if the reduced detection of E2F6-containing PRCs in HPV16 E7-expressing cells was accompanied by alterations in the H3K27me3 transcriptional repressive mark, we compared the levels of mono-, di-, and trimethylated H3K27 by immunofluorescence in donor and passage-matched primary human foreskin keratinocyte (HFK) populations that were engineered to express HPV16 E7 (HFK/E7) or were infected with an empty retroviral control vector. These experiments revealed a striking reduction in the intensity of H3K27me3 staining in HFK/E7 cells. In contrast, there was no detectable reduction of H3K27me1 and H3K27me2 or total H3 staining in HFK/E7 compared with control HFKs (Fig. 1A). A reduction of H3K27me3 levels in HFK/E7 was also detected by Western blotting (Fig. 1B).

Open in a separate window

Fig. 1.

The repressive H3K27me3 mark is specifically reduced in HPV16 E7-expressing primary human epithelial cells. (A) Immunofluorescence analysis of histone H3 methylation in primary human epithelial cells (HFKs) and donor- and passage-matched HPV16 E7-expressing HFKs (HFK/E7). (B) Western blot analysis of H3K27me3 levels in HFKs and HFK/E7s. Lysates were separated by SDS/PAGE, transferred, and probed for H3K27me3 (me3) and total H3 as a loading control.

Expression of the H3K27-Specific Demethylases KDM6A and KDM6B Is Increased in HPV16 E7-Expressing Primary Human Epithelial Cells.

H3K27me3 repressive marks are placed by the histone methyltransferase containing PRC2 (19–22) and are removed by the histone demethylases KDM6A and KDM6B (23–26). To determine if levels of the histone demethylases KDM6A and KDM6B were increased in HPV16 E7-expressing cells, we performed immunofluorescence analyses in HFK/E7 cells and donor- and passage-matched control HFKs. These experiments revealed a striking increase of KDM6A and KDM6B levels in HFK/E7s compared with control HFKs (Fig. 2A). Increased expression of KDM6A and KDM6B in HFK/E7 cells was also detected by Western blot analysis (Fig. 2B). Quantitative real-time RT-PCR (qRT-PCR) experiments revealed that KDM6A and KDM6B mRNA levels are increased in HFK/E7 cells (Fig. 2C), and thus, the mechanism of induction is at least in part transcriptional. Consistent with a previous report (27), EZH2 mRNA expression was also increased in HFK/E7 cells (Fig. S1).

Open in a separate window

Fig. 2.

Increased expression of KDM6A and KDM6B in HPV16 E7-expressing primary human epithelial cells. (A) Immunofluorescence analysis of KDM6B (Upper) and KDM6A (Lower) expression in donor- and passage-matched populations of primary human foreskin keratinocytes (HFKs; Left) and HPV16 E7-expressing HFKs (HFK/E7; Right). (B) Western blot analysis of KDM6A and KDM6B levels in HFKs and HFK/E7 cells. Lysates were separated by SDS/PAGE, transferred, and probed for KDM6A, KDM6B, and HPV16 E7. A pRB blot is shown to document functional E7 expression, and an actin blot is included as a loading control. (C) Quantitative real-time RT-PCR analysis of KDM6A and KDM6B mRNA expression in HFK and HFK/E7 cells. The bar graph shows averages and SDs from three independent experiments, each performed in triplicate. Increases in HFK E7 cells are statistically significant (*), with P values < 0.005.

HPV16 E7-Mediated Induction of the Cervical Cancer Biomarker p16^INK4A Is Mediated by KDM6B.

The CDK4/6 inhibitor and tumor suppressor p16^INK4A is highly expressed in high-risk HPV-associated lesions and cancers and is an excellent biomarker for such malignancies (28, 29). KDM6B controls induction of p16^INK4A in response to oncogenic stress by ras/raf (16, 17). Given that HPV16 E7 induces expression of both p16^INK4A (30) and KDM6B (Fig. 2), we investigated if p16^INK4A positivity correlated with the loss of the H3K27me3 mark and analyzed p16^INK4A expression and H3K27 trimethylation by immunofluorescence in four HPV16-positive cervical intraepithelial neoplasia (CIN) specimens. In areas with strong p16^INK4A staining, we detected little, if any, H3K27me3 staining (Fig. 3A Upper), whereas the presumably normal adjacent tissue on the same slide that exhibited weak p16^INK4A staining showed a robust H3K27me3 signal (Fig. 3A Lower). Similar results were obtained with organotypic raft cultures prepared from primary HFKs and HPV16-immortalized HFKs (Fig. S2). To determine if increased expression of p16^INK4A is mechanistically linked to HPV16 E7-induced KDM6B expression, we depleted KDM6B in monolayer cultures of HPV16-immortalized HFKs by transfecting a pool of specific siRNA duplexes or control siRNA and analyzed KDM6B, p16^INK4A, and p14^ARF levels by Western blotting. KDM6B depletion caused a decrease in p16^INK4A levels compared with control siRNA transfected cells, whereas p14^ARF levels were unchanged (Fig. 3B).

Open in a separate window

Fig. 3.

HPV16 E7-mediated induction of KDM6B is critical for the expression of the cervical cancer biomarker p16^INK4A. (A) Coimmunofluorescence staining of p16^INK4A and H3K27me3 in an HPV16-positive cervical intraepithelial neoplasia (CIN) specimen. Upper and Lower are from different areas of the same specimen. Similar staining patterns were detected in three additional CIN specimens. Hoechst staining is shown to visualize nuclei, and a phase picture shows cellular morphology. (B) Monolayer cultures of HPV16-immortalized HFKs were transfected with KDM6B-specific siRNA duplexes or control siRNA (CTL), and p16^INK4A and p14^ARF expression was analyzed by Western blotting at 72 h posttransfection. An actin blot is shown as a loading control (Left). Averages and SDs of KDM6B, p16^INK4A, and p14^ARF levels from three independent experiments are shown on the bar graph in Right. Statistically significant decreases (P < 0.05) are indicated by an asterisk.

HPV16 E7-Mediated Induction of KDM6B and Its Target p16^INK4A Is Not Dependent on pRB Inactivation.

To determine whether HPV16 E7-mediated pRB inactivation provides a necessary oncogenic stress signal that causes induction of KDM6B expression, we compared expression of KDM6B and its target p16^INK4A by Western blotting in donor- and passage-matched primary human fibroblasts with expression of WT HPV16 E7 or the pRB-binding/degradation-deficient HPV16 E7 delD21-C24 mutant as well as control vector-transduced cells. Expression of the HPV16 E7 delD21-C24 mutant robustly increased KDM6B and p16^INK4A levels (Fig. 4). Hence, up-regulation of KDM6B and its transcriptional target p16^INK4A by HPV16 E7 is not dependent on pRB inactivation.

Open in a separate window

Fig. 4.

HPV16 E7-mediated induction of KDM6B and its transcriptional target p16^INK4A are not dependent on the ability of HPV16 E7 to degrade pRB. Western blot analysis of KDM6B and p16^INK4A expression in donor- and passage-matched populations of primary human fibroblasts (HFFs) expressing WT HPV16 E7, the pRB binding/degradation-deficient HPV16 delD21-24 mutant, and control vector-infected HFFs. An E7 blot shows similar expression of WT and mutant E7, and an actin blot is show as a loading control.

Deregulated HOX Gene Expression in HPV16 E7-Expressing Primary Epithelial Cells.

HOX A–D loci are well-established transcriptional targets of PRCs (31). To determine whether the HPV16 E7-mediated increases in KDM6A and KDM6B expression cause changes in HOX gene expression, we compared expression of the 39 HOX A–D genes in HFK/E7 cells with donor- and passage-matched control HFKs by qRT-PCR. The mRNA levels of a number of HOX genes were significantly increased in HFK/E7 cells (Fig. 5). Hence, HPV16 E7 expression causes epigenetic reprogramming of primary human epithelial cells.

Open in a separate window

Fig. 5.

Dysregulation of HOX gene expression in HPV16 E7-expressing primary human epithelial cells. Quantitative real-time RT-PCR of HOX mRNA expression in primary human foreskin keratinocytes (HFKs) and donor- and passage-matched HPV16 E7-expressing HFKs (HFK/E7). Bar graphs represent averages and SDs of three experiments, each performed in triplicate. HOX genes that exhibit significant (P < 0.05) up-regulation are marked (*). Although all 39 genes in the HOX A–D clusters were analyzed, those HOX genes that were expressed in HFKs at levels too low to evaluate are not shown.

HPV16 E7-Mediated Induction of KDM6B and p16^INK4A and Decreases in H3K27me3 Levels Are Associated with Decreased H3K37 Trimethylation and Increased KDM6B Binding to the INK4A Promoter and Are Reversible.

To determine if induction of KDM6B by HPV16 E7 is reversible, we used U2OS cells with doxycycline-inducible expression of HPV16 E7. On induction of HPV16 E7 expression, we observed increases in KDM6B and p16^INK4A expression (Fig. 6A), with concomitant decreases in the H3K27me3 mark (Fig. 6B). These changes were abolished when HPV16 E7 expression was reduced by removal of doxycycline (Fig. 6). Hence, the observed induction of KDM6B expression, associated decreases in H3K27me3 levels, and induction of p16^INK4A expression are a direct consequence of HPV16 E7 expression, and these alterations are reversible on silencing of HPV16 E7 expression.

Open in a separate window

Fig. 6.

HPV16 E7-mediated induction of KDM6B and p16^INK4A and decreases in H3K27me3 levels are associated with decreased H3K37 trimethylation and increased KDM6B binding to the INK4A promoter and are reversible. (A) Western blot analysis of KDM6B, p16^INK4A, and HPV16 E7 expression in U2OS-tet on cells with doxycycline-inducible expression of HPV16 E7 treated with doxycycline as indicated. Lysates were separated by SDS/PAGE, transferred, and probed with the indicated antibodies. An actin blot is shown as a loading control. (B) Immunofluorescence analysis of histone H3 lysine 27 trimethylation in U2OS-tet on cells with doxycycline-inducible expression of HPV16 E7 treated with doxycycline as indicated. Hoechst stain is shown to visualize nuclei. (C and D) ChIP assays using lysates from U2OS-tet on cells with or without doxycycline-inducible expression of HPV16 E7. In C, an antibody specific for H3K27me3 was used, and in D, an antibody specific to KDM6B was used. qPCR was used to measure the degree of enrichment. The H3K27me3 signal was normalized to histone H3 bound, and the KDM6B results are presented as a percentage of bound/input. Statistically significant changes (P < 0.05) are indicated by an asterisk.

To compare the ratio of H3K27me3 to total H3 throughout the INK4A-ARF locus before and after induction of HPV16 E7, we performed ChIP experiments. Consistent with our previous results (Fig. 1), the H3K27me mark was virtually absent from the INK4A-ARF locus when HPV16 E7 was expressed (Fig. 6C). There was an increase in the H3K27me3 mark around the transcriptional start site of INK4A on depletion of KDM6B but not when KDM6B was depleted (Fig. 6C). Consistent with this result, we observed increased KDM6B binding around the INK4A transcriptional start site on HPV16 E7 expression (Fig. 6D). Because of the lack of appropriate antibodies, KDM6A binding at the INK4A-ARF locus could not be evaluated by ChIP. Consistent with the observed increase in EZH2 expression, we observed increased EZH2 binding along the INK4A-ARF locus (Fig. S3).

KDM6A and KDM6B Depletion Causes Growth Suppression in a Cervical Carcinoma Line.

To investigate whether high-level expression of KDM6A and KDM6B cells may modulate the transformed phenotype of HPV16-transformed cells, we depleted KDM6A or KDM6B in the HPV16-positive CaSki cervical carcinoma cells followed by colony formation assays. The number of colonies was reduced by ∼50% when either KDM6A or KMD6B was depleted (Fig. 7).

Open in a separate window

Fig. 7.

KDM6A and KDM6B depletion causes growth suppression in a cervical carcinoma line. KDM6A or KDM6B were depleted in the HPV16-positive CaSki cervical carcinoma cells followed by colony formation assays. Averages and SDs for three independent experiments are shown. Statistically significant changes (P < 0.05) are indicated by an asterisk.

Discussion

We discovered that HPV16 E7 expression causes a marked reduction in the H3K27me3 repressive epigenetic mark (Fig. 1), and a similar decrease was also evident in HPV16-positive cervical lesions (Fig. 3A). A previous report showed that HPV16 E7 induces expression of the histone methyl transferase component of PRC2, EZH2 (27), and we confirmed this result in our studies (Fig. S1). EZH2 overexpression and increased binding to the p16^INK4A promoter (Fig. S3) may seem at odds with our finding that the H3K27me3 mark is decreased in HPV16 E7-expressing cells (Fig. 1). It has been shown, however, that EZH2 overexpression does not result in increased PRC2 activity but instead, enhances PRC4 formation (32). PRC4 causes histone H1K26 deacetylation and methylation (32), which then serves as a binding site for the lethal(3)malignant brain tumor-like 1 (L3MBTL1) polycomb group protein. Hence, increased EZH2 expression in E7-expressing cells may be predicted to cause enhanced H1K26 methylation. Intriguingly, HPV16 E7 is associated with the L3MBTL2 protein (9), which is closely related to L3MBTL1 (33), and it will be interesting to determine whether the predicted increase in EZH2-containing PRC4 complexes and association with L3MBTL2 may alter H1K26 methylation (Fig. 7).

The observed decrease in the repressive H3K27me3 mark is mediated by transcriptional induction of the two H3K27-specific histone demethylases, KDM6A and KDM6B (Fig. 2). The concept that a viral oncoprotein affects expression of histone demethylases to alter the epigenetic memory of host cells has not been previously appreciated. KDM6A and KDM6B have each been implicated in tumor development. KDM6A is a bona fide tumor suppressor, and somatic mutations have been detected in multiple human tumor types. KDM6B is also a potential tumor suppressor, because the KDM6B gene is located at 5q31, an area that is frequently lost in multiple cancer types (reviewed in ref. 34). Conversely, KDM6B overexpression may also have oncogenic activities, because its expression is increased in prostate cancer, most dramatically in metastatic tumors (35). Our results that KDM6A or KDM6B depletion inhibits proliferation of CaSki cervical cancer cells (Fig. 7) are consistent with this model.

The experiments presented here also establish the elusive mechanism for induction of the CDK4/6 inhibitor and cervical cancer biomarker p16^INK4A by HPV16 E7. A previous study suggested that HPV16 E7 may induce p16^INK4A expression through an E2F-dependent pathway and that increased p16^INK4A expression may represent a consequence of HPV E7-mediated pRB tumor suppressor inactivation (30). High-level p16^INK4A expression, however, is a unique hallmark of high-risk HPV-associated lesions and cancers and is not generally observed in tumors that have lost pRB tumor suppressor function. Moreover, expression of the simian vacuolating virus (SV) 40 large-tumor antigen, which also potently activates E2F, does not enhance p16^INK4A expression. Our results show that HPV16 E7-mediated p16^INK4A induction is independent of pRB inactivation (Fig. 4) and is mediated by KDM6B. Hence, it is tempting to speculate that increased p16^INK4A expression in high-risk HPV-associated lesions and tumors may represent a cellular response to HPV16 E7-induced oncogenic stress similar to what has been reported for ras/raf (16, 17). Induction of p16^INK4A as a consequence of ras/raf-mediated oncogenic stress has been recently documented in vivo by an analysis of mice expressing K-ras^G12D in enterocytes (36). Interestingly, p16^INK4A expression was no longer observed in malignant tumors, supporting the notion that p16^INK4A induction in response to an oncogenic stimulus may represent a critical cell intrinsic tumor suppressive mechanism that needs to be overcome during malignant progression, presumably by an additional cellular mutation (36). Although HPV16 E7 triggers increased p16^INK4A expression, it simultaneously neutralizes the key mediator of p16^INK4A-induced senescence, pRB, and HPV16 E7-expressing cells continue to proliferate, despite high levels of p16^INK4A expression. Based on these findings, one might suggest the somewhat heretical hypothesis that mutations that target the pRB pathway, which occur in the majority of human solid tumors, are not to induce uncontrolled proliferation, as is commonly assumed, but are to abrogate the p16^INK4A-mediated senescence response induced by oncogenic stress (Fig. 8).

Open in a separate window

Fig. 8.

Model for epigenetic reprogramming by the HPV16 E7 oncoprotein. Details are in the text.

Unlike ras/raf, which only causes increased expression of KDM6B (16, 17), HPV16 E7 expression also triggers increased expression of the related KDM6A (Fig. 2). It is possible that KDM6A induction by HPV16 E7 expression also represents a cellular defense response, perhaps mediated by different molecular triggers than KDM6B. A systematic investigation into the molecular triggers and mechanisms that drive KDM6A and KDM6B induction by HPV16 E7 is currently in progress.

It is clear, however, that HPV16 E7 induces expression of these histone demethylases through pathways that are distinct from the normal mechanism of KDM6B induction during epithelial differentiation, where recruitment to promoters is enhanced, whereas overall KDM6B levels are not altered (37). The transcriptional induction of KDM6A and KDM6B by HPV16 E7, however, is reminiscent of the induction of these enzymes during wound healing (38). Because HPV infects the proliferating basal layer of the epithelium through microabrasions, the initial stages of the HPV life cycle likely occur in a wound-healing environment. Moreover, cellular activities related to wound healing may increase the efficiency with which HPV genomes become established in basal cells (39), and HPV infection has been reported to maintain cells in a wound healing-like state (40). HPV16 E7-mediated transcriptional induction of KDM6A and KDM6B may, therefore, contribute to maintaining a wound-healing state that may be beneficial for the viral life cycle.

Because HPV16 E7-expressing cells continue to proliferate in the presence of high-level KDM6A and KDM6B expression, one predicts that such cells are epigenetically reprogrammed. Consistent with this model, our studies document dysregulated HOX gene expression in HPV16 E7-expressing cells (Fig. 5). Of particular interest are HOXC5 and HOXC8, which have previously been reported to be up-regulated in cervical carcinomas (41). HOX genes have been implicated in the regulation of adult skin (42–44). Thus, one possible consequence of abnormal HOX gene expression in HPV-infected cells may be an altered differentiation state of the epithelium that may be conducive to the viral life cycle.

In addition, alterations in HOX gene expression patterns may modify the stemness of the infected host cell. One might, therefore, speculate that HPV16-infected basal epithelial cells may be reprogrammed to a more stem-like state. This would represent an attractive mechanism to guarantee long-term maintenance of HPV genomes in infected epithelial cells. Such a mechanism may also have important implications with respect to the oncogenic activities of high-risk HPVs. One possibility is that HPV16 E7-induced epigenetic alterations, potentially in concert with other activities of the HPV16 E7 and/or HPV16 E6 oncoproteins, may contribute to the development of a cancer stem cell (reviewed in ref. 45).

Dysregulation of HOX gene expression is frequently observed in human tumors, and some HOX genes are bona fide oncogenes or tumor suppressors (reviewed in ref. 11). In addition to altering the stemness of the cell, dysregulated HOX gene expression has also been linked to other pathways that promote tumorigenesis, including differentiation, invasion and epithelial-to-mesenchymal transition (EMT), apoptosis, proliferation, and receptor signaling (reviewed in ref. 11).

Alterations of patterns of DNA methylation and histone modifications are often found early in tumorigenesis and may be key initiating events in certain cancers (reviewed in ref. 46). Because, unlike genetic mutations, epigenetic aberrations are potentially reversible, they constitute a ripe target for therapeutic intervention. Our findings that HPV16 E7-induced epigenetic and related transcriptional alterations are reversible on silencing of E7 expression (Fig. 6) and that KDM6A or KDM6B depletion inhibits proliferation of CaSki cervical cancer cells suggest that inhibition of KDM6A and/or KDM6B may be therapeutic modalities for HPV-associated lesions and cancers.

Materials and Methods

Supplementary Material

Acknowledgments

Footnotes

References