Human Papillomavirus Carcinogenesis: an Identity Crisis in the Retinoblastoma Tumor Suppressor Pathway

Abstract

Viruses are obligate intracellular parasites and need to reprogram host cells to establish long-term persistent infection and/or to produce viral progeny. Cellular changes initiated by the virus trigger cellular defense responses to cripple viral replication, and viruses have evolved countermeasures to neutralize them. Established models have suggested that human papillomaviruses target the retinoblastoma (RB1) and TP53 tumor suppressor networks to usurp cellular replication, which drives carcinogenesis. More recent studies, however, suggest that modulating the activity of the Polycomb family of transcriptional repressors and the resulting changes in epigenetic regulation are proximal steps in the rewiring of cellular signaling circuits. Consequently, RB1 inactivation evolved to tolerate the resulting cellular alterations. Therefore, epigenetic reprograming results in cellular “addictions” to pathways for survival. Inhibition of such a pathway could cause “synthetic lethality” in adapted cells while not markedly affecting normal cells and could prove to be an effective therapeutic approach.

HUMAN PAPILLOMAVIRUS-ASSOCIATED CANCERS

Papillomaviruses are ubiquitous viruses with 8-kb double-stranded circular genomes. They specifically infect squamous epithelial tissues and cause hyperplastic lesions, referred to as warts. Based on their tissue tropism, the ∼200 human papillomavirus (HPV) types can be divided into mucosal and cutaneous types. Mucosal HPVs are classified as “low risk” and “high risk” for malignant progression of the lesions that they cause. Approximately 5% of all human cancers, including >99% of cervical carcinomas, many penile, vulvar, vaginal, and anal carcinomas, and a growing fraction of head and neck squamous cell carcinomas, are caused by high-risk HPVs (1). With an estimated 4,100 deaths in 2015 in the United States, one woman succumbs to cervical cancer approximately every 2 h in this country (and every 2 min worldwide) (2). Given the low vaccination rates in the United States and the fact that the available prophylactic vaccines do not inhibit malignant progression in individuals who are already infected, and because HPV-associated cancers often develop years to decades after the initial infection, these dire numbers are not likely to change in the foreseeable future.

HPV-ASSOCIATED CANCERS ARE FUELED BY E6 AND E7 EXPRESSION

High-risk HPV-associated cancers are generally nonproductive infections, i.e., viral proteins are produced but no viral progeny is generated. Importantly, the HPV E6 and E7 genes are the only viral genes that are consistently expressed in these cancers. Although the recently identified, recurrent cellular mutations in cervical carcinoma likely also contribute to oncogenesis (3), cervical cancer lines remain “addicted” to E6/E7 expression, and extinguishing HPV E6/E7 expression in cervical cancer lines causes senescence (4).

High-risk HPV E6 and E7 are potently oncogenic in cell-based transformation assays and in transgenic mouse models. In contrast, the low-risk HPV E6 and E7 proteins are only weakly transforming or do not transform at all. High-risk HPV E7 proteins target the retinoblastoma tumor suppressor RB1 for degradation, and high-risk HPV E6 proteins target the TP53 tumor suppressor, as well as cellular proteins that contain postsynaptic density protein, Drosophila discs large tumor suppressor, and zonula occludens 1 (PDZ) domains for degradation. In addition, E6 proteins stimulate telomerase (TERT). In contrast, low-risk HPV E7 proteins bind, but do not degrade, RB1, and low-risk HPV E6 proteins do not bind or target TP53 and PDZ proteins for degradation, nor do they stimulate TERT (1).

TEXTBOOK MODEL OF HPV ONCOGENESIS

Textbooks and review articles have popularized a simple scheme illustrating how the RB1 and TP53 tumor suppressor pathways are connected and how their functional inactivation is essential for carcinogenesis (Fig. 1A). According to this model, RB1 inactivation by E7 is the initiating oncogenic hit. RB1 inactivation is driven by the need of HPVs to ensure proliferative competence of infected cells, which is achieved through aberrant expression of genes regulated by E2F transcription factors. Many of these E2F-regulated genes are required for S-phase entry and progression. Promiscuous S-phase entry and proliferation, however, cause E2F-dependent upregulation of the p14^ARF tumor suppressor, which triggers TP53 stabilization through inhibition of the MDM2 ubiquitin ligase. Since TP53 activation leads to cell cycle arrest in G₁ and/or apoptosis, high-risk HPV E6 proteins evolved to target TP53 for ubiquitylation by the E6-associated UBE3A (E6AP) ubiquitin ligase, which results in TP53 proteasomal degradation (1). High-risk HPV E7 proteins also abrogate TP53 transcriptional activities (5). In addition, E2F activation by E7 is thought to stimulate expression of the CDK4/CDK6 (CDK4/6) inhibitor and tumor suppressor p16^INK4A, which, through CDK4/6 inhibition, would result in RB1 tumor suppressor activation (6). This is functionally inconsequential, though, since the RB1 tumor suppressor is inactivated.

FIG 1.

HPV carcinogenesis. (A) According to the currently engrained model, the E7 protein initiates carcinogenesis by inactivating the retinoblastoma tumor suppressor protein (RB1). The resulting activation of E2F transcription factors causes TP53 activation, a cellular defense response that results in G₁ cell cycle arrest and/or apoptosis. To thwart this response, the E6/UBE3A (E6AP) ubiquitin ligase complex targets TP53 for ubiquitylation and proteasomal degradation. To counteract telomere erosion, E6 also activates telomerase (TERT), which results in unlimited, aberrant cellular proliferation. (B) The revised model suggests that HPV E7 expression triggers a p16^INK4A-mediated cellular defense response that is reminiscent of RAS oncogene-induced senescence (OIS). De novo p16^INK4A expression is triggered by removal of repressive trimethyl marks on lysine 27 of histone H3 (H3K27me3) by the KDM6B histone demethylase. High-risk HPV E7 proteins have evolved to degrade RB1 to short-circuit the p16^INK4A-mediated activation of RB1 tumor suppressor activity that is designed to trigger G₁ cell cycle arrest and senescence (see also panel C). E7 expression also causes epigenetic derepression of the p14^ARF promoter through a KDM6B-independent mechanism. As a consequence of removal of H3K27me3 repressive marks, HPV E7-expressing cells show dysregulated homeobox (HOX) gene expression, suggestive of loss of Polycomb-mediated repression and epigenetic reprogramming. HPV E7-expressing cells rapidly acquire an “addiction” to the KDM6A and KDM6B H3K27 demethylases and to p16^INK4A tumor suppressor expression. See the text for details. (C) Oncogenic stress, such as that induced by the RAS oncoprotein, triggers a cell intrinsic tumor suppressor pathway, and this phenomenon is known as oncogene-induced senescence (OIS). OIS is initiated by transcriptional induction of the KDM6B H3K27 demethylase, which results in derepression of genes typically silenced by Polycomb group (PcG) proteins, such as the p16^INK4A tumor suppressor, which inhibits CDK4/6 activity by abrogating complex formation with D-type cyclins (CCND). CDK4/6 inhibition causes accumulation of hypophosphorylated, active, RB1 tumor suppressor, which triggers G₁ cell cycle arrest by forming a repressor complex with E2F/DP transcription factors. RB1 tumor suppressor activation also causes cellular senescence through a pathway that is largely E2F independent.

This elegant and well-published model, however, has a number of significant limitations. Although low-risk HPV E7s less efficiently bind to and do not degrade RB1, they activate E2F-dependent promoters and cause cellular hyperproliferation; however, they are nononcogenic. The cutaneous HPV1 E7 protein does not transform despite binding RB1 and activating E2F as efficiently as HPV16 E7. Furthermore, an RB1 binding-competent, but RB1 degradation-deficient, HPV16 E7 mutant efficiently activates E2F but does not transform. Since expression of low-risk HPV E7 proteins or an RB1 binding-competent/RB1 degradation-deficient HPV16 E7 mutant does not cause TP53 stabilization, E2F activation per se cannot be sufficient to cause TP53 activation (5). This is consistent with the finding that low-risk HPV E6 proteins did not evolve to target TP53 for degradation (7). Hence, despite its elegance and simplicity (Fig. 1A), the current model of HPV transformation requires significant refinement.

REVISED MODEL OF HPV ONCOGENESIS

Recent studies suggest that p16^INK4A expression in response to high-risk E7 expression is the main carcinogenic determinant, and degradation of RB1 by high-risk HPV E7 proteins has evolved from the need to negate the p16^INK4A cytostatic response (Fig. 1B). Low-risk HPV E7 proteins do not trigger p16^INK4A, consistent with the finding that p16^INK4A expression is a biomarker specifically for high-risk HPV-infected lesions and cancers. High-risk HPV E7 induces p16^INK4A independent of RB1 inactivation, through an epigenetic mechanism involving the KDM6B histone demethylase (8). KDM6B catalyzes removal of trimethyl marks on lysine 27 of histone H3 (H3K27me3), which are critical for gene silencing by Polycomb repressive complexes (PRCs) (9). This is reminiscent of RAS oncogene-induced senescence (OIS), a cellular defense response involving p16^INK4A-mediated inhibition of CDK4/6 and resulting in RB1-mediated cell cycle arrest and senescence (10, 11) (Fig. 1C). RB1 binding and E2F activation by E7 is not sufficient to thwart RB1 senescence signaling; therefore, high-risk HPV E7s evolved to degrade RB1 (12). High-risk (but not low-risk) HPV E7 proteins also trigger expression of p14^ARF, which is normally silenced by PRCs. HPV16 E7 expression causes a decreased incidence of H3K27me3 marks on the p14^ARF promoter, but the mechanism is distinct from p16^INK4A derepression and does not involve KDM6B (8).

This revised model is entirely consistent with the well-established “thrust and parry” concept of viral oncogenesis: HPVs generate oncogenic stimuli by associating with and functionally reprogramming key cellular control circuits. The infected cells attempt to mount innate defense responses to neutralize the invading virus, which the virus, in turn, has evolved to blunt. However, and in contrast to the “old” model, the ability of high-risk HPV E7 proteins to subvert RB1 and activate E2F is not the primary initiating thrust. Instead, RB1 degradation is a strategy that high-risk HPVs evolved to muffle the p16^INK4A-mediated cellular OIS response.

HOW DO HIGH-RISK HPV E7 PROTEINS TRIGGER ONCOGENIC STRESS?

Similar to the RAS oncogene, HPV16 E7 triggers p16^INK4A expression through transcriptional activation of the KDM6B H3K27 demethylase (8, 10, 11). It does not appear, however, that E7 functions through a RAS-dependent mechanism, and E7 lacks RAS-like activities in standard oncogene cooperation experiments (5). Moreover, E7 also induces expression of the KDM6A H3K27 demethylase (8, 13). One interesting possibility is that E7 may induce OIS as a consequence of causing replicative stress (14), which may interfere with the ability of Polycomb repressive complexes (PRCs) to silence the ARF-INK4A locus, leading to p16^INK4A and p14^ARF expression. In conjunction with the induction of TP53, this would lead to a robust apoptotic or senescence response, which HPVs would have to subvert via subsequent targeting of RB1 and TP53 by E7 and E6, respectively.

Replication stress, which arises from unrepaired DNA damage, has been correlated with tumorigenesis in many contexts (15). However, the degree to which replication stress is a driver or consequence of tumorigenesis is not clear. High-risk, but not low-risk, HPV E7 proteins are well known for their ability to induce genomic instability, but this has been regarded traditionally as a driver of malignant progression rather than an activity which contributes to initiation of carcinogenesis. E7 induces aberrant centriole synthesis, inhibits the anaphase-promoting complex to potentially cause a prometaphase delay, and causes accumulation of double-stranded DNA breaks, which has been linked to induction of replicative stress (14, 16).

Although PRCs play well-known roles in the maintenance of cell identity and proliferation, recent evidence has implicated Polycomb group (PcG) proteins in the cellular response to DNA damage (17). However, the precise role for PcG proteins in DNA repair has not been elucidated. One possibility is that PRC-mediated modification of histones is required for the change in chromatin conformation necessary to allow access of DNA repair machinery to the DNA. Alternatively, PcG activity could serve to repress transcription while repair is ongoing. Histone H2A is mono- and polyubiquitylated during a DNA damage response (DDR), including monoubiquitylation at Lys119 by PRC1 (17). Therefore, an incessant induction of a DDR by E7 could result in persistent, high levels of H2A K119 ubiquitylation at sites of DNA repair, which may interfere with the normal dynamic of monoubiquitylation/deubiquitylation of H2A necessary for PRC repression. On the other hand, a persistent DNA damage response could alter PRC activity by recruiting PcG proteins to sites of DNA damage and away from normal target genes that regulate cellular identity and cell fate decisions (17).

A relationship between Polycomb and replicative stress was demonstrated recently for Drosophila melanogaster. Specifically, persistent replicative stress and an ongoing DNA damage response led to alterations in cell fate decisions among stem cells and their progeny. Furthermore, Polycomb mutant phenotypes could be enhanced by abrogation of DNA repair signaling, resulting in persistence of unrepaired double-stranded DNA breaks (18). The Drosophila genome encodes two RB homologs (Rbf1 and Rbf2), two E2F subunits (dE2F1 and dE2F2), and one DP subunit (dDP) (19). However, Drosophila does not appear to encode INK4-related CDK4/6 inhibitors. Consequently, a cellular senescence response is not triggered by replicative stress in flies. Indeed, replication stress is a common feature of the rapid endocycles that occur during Drosophila embryogenesis.

Given the role of PRCs in signaling the cellular DNA damage response, it is tempting to speculate that replicative stress induced by high-risk HPV E7 proteins, which is at least in part independent of RB1 inactivation (14), may be the initial oncogenic stimulus that causes OIS signaling, leading to induction of p16^INK4A and other PRC-repressed genes. High-risk HPV E7s have evolved to overcome OIS by targeting RB1 for degradation, leading to outgrowth of cells that aberrantly express genes normally repressed by PRCs, including homeobox (HOX) genes (8, 13). It is, therefore, conceivable that HPV E7-expressing cells are epigenetically reprogramed, which may be the basis for E7's ability to contribute to the epithelial-to-mesenchymal transition (EMT). Degradation of cellular identity due to increased epigenetic plasticity may also contribute to tumor heterogeneity.

(WHY) DOES IT MATTER?

Cynics may argue that this model merely suggests a more winding and convoluted path by which high-risk HPV infection causes unlimited proliferation of cells with defective RB1 and TP53 tumor suppressor circuits (Fig. 1). However, “epigenetic reprogramming,” presumably as a consequence of increased KDM6A and KDM6B expression and evidenced by aberrant HOX gene expression, has important biological consequences. Since high-risk HPV E7-expressing cells become rapidly addicted to KDM6A and KDM6B (8, 20), this effect of E7 could be targeted therapeutically; the GSK-J4 small-molecule KDM6A/B selective inhibitor induced apoptosis in high-risk HPV E7-expressing cells (20). This suggests the exciting possibility of epigenetic therapies for high-risk HPV-associated lesions and cancers.

KDM6B addiction is linked to p16^INK4A expression; consequently, high-risk HPV E7-expressing cells are dependent upon expression of the p16^INK4A tumor suppressor (20). This implies that the p16^INK4A tumor suppressor is pro-oncogenic in cells with inactive RB1, and it will be interesting to determine whether the RB1 tumor suppressor may be similarly pro-oncogenic in cells that have lost p16^INK4A expression. Addiction to p16^INK4A is related to CDK4/6 inhibition, suggesting that there is a yet-to-be-discovered CDK4/6 substrate(s) that causes cell death when phosphorylated by CDK4/6 in cells with an inactive RB1 tumor suppressor. Since CDK4/6 inhibition is required for proliferation/survival of high-risk HPV E7-expressing cells, CDK4/6 inhibitors may not be useful therapeutic agents for HPV-associated lesions and cancers. Some HPV-negative tumors, including triple-negative breast cancers, treatment-resistant prostate carcinomas, serous ovarian adenocarcinomas, and lung cancers, frequently also show high-level p16^INK4A expression. It will be interesting to determine whether some of these tumors are similarly addicted to KDM6B and/or p16^INK4A expression.

Given that removal of repressive H3K27me3 marks by KDM6 enzymes is only one of the necessary steps for derepression of Polycomb-regulated genes, such as the ARF-INK4A and HOX genes, it will be important to determine whether E7-expressing cells develop addictions to any of the other factors that contribute to E7-mediated p16^INK4A expression or are components of the E7/KDM6A “synthetic lethality” axis. Such factors could represent previously unappreciated therapeutic targets for high-risk HPV-associated lesions and cancers.

In summary, we propose that modulation of Polycomb-mediated repression represents a cellular response to high-risk HPV E7 expression and that RB1 degradation by E7 has evolved to overcome this response. Moreover, dysregulating Polycomb repression may degrade cell identity, thereby contributing to tumor heterogeneity. Induction of persistent replicative stress and double-stranded DNA breaks by E7 may be the key oncogenic hit that triggers this effect on PRCs. The resulting alterations in the cellular epigenome generate addictions to specific epigenetic enzymes, which may be harnessed therapeutically.