Epigenetics of cervical cancer. An overview and therapeutic perspectives

Molecular Cancer 2005 and BioMed Central
An Open Access Research article

Published 25 October 2005 

Overview of cervical cancer
Cervical cancer remains one of the greatest killers of women worldwide. It is difficult to foresee a dramatic increase in cure rate even with the most optimal combination of cytotoxic drugs, surgery, and radiation; therefore, testing of molecular targeted therapies against this malignancy is highly desirable. A number of epigenetic alterations occur during all stages of cervical carcinogenesis in both human papillomavirus and host cellular genomes, which include global DNA hypomethylation, hypermetylation of key tumor suppressor genes, and histone modifications. The reversible nature of epigenetic changes constitutes a target for transcriptional therapies, namely DNA methylation and histone deacetylase inhibitors. To date, studies in patients with cervical cancer have demonstrated the feasibility of reactivating the expression of hypermethylated and silenced tumor suppressor genes as well as the hyperacetylating and inhibitory effect upon histone deacetylase activity in tumor tissues after treatment with demethylating and histone deacetylase inhibitors. In addition, detection of epigenetic changes in cytological smears, serum DNA, and peripheral blood are of potential interest for development of novel biomolecular markers for early detection, prediction of response, and prognosis.

Epidemiology and treatment
Cervical cancer remains one of the greatest killers of women worldwide. According to Globocan 2000, it is estimated that in 2000 the numbers of patients diagnosed with and those who died from this disease were 470,606 and 233,372, respectively [1]. It is remarkable that these rates occur despite the fact that cervical cancer is a model for early detection due to its long and relatively well-known natural history, which offers an excellent opportunity for its detection before lesions become invasive [2].

Cervical cancer is currently staged clinically according the International Federation of Gynecology and Obststrics (FIGO) guidelines. In terms of treatment, invasive disease can be divided into three main groups: 1) early stage going from microinvasive disease IA1, IA2 to macroscopic disease confined to cervix and measuring <4 cm, IB1; 2) locally advanced FIGO stages IB2-IVA, and 3) IVB and recurrent disease [3].

Treatment of early stages
The recommended treatment for IA1 patients is either a local procedure such as conization or total hysterectomy depending on the patient's desire to remain fertile, whereas for IA2 patients the recommended procedure is a radical one including pelvic lymphadenectomy. On average, 8% of cases shows positive pelvic lymph nodes. As many women at this disease stage deserve to preserve fertility, radical trachelectomy is becoming an option for these patients. The same can apply for IB1 patients. In early cases that are surgically treated, the presence in the surgical specimen of a combination of intermediate-risk factors (vascular and lymphatic permeation, tumor size >2 cm, and deep cervical stroma invasion) or high-risk factors (positive pelvic lymph nodes, parametrial infiltration, and positive surgical margins) dictates use of adjuvant radiation or chemoradiation respectively. As a group, the prognosis of early-stage cases is fairly good with 5-year survival exceeding 90% [4,5]

Treatment of locally advanced stages
Results of treatment for these patients are far from optimal. In this regard, treatment of locally advanced cervical cancer has experienced no major changes for nearly 80 years during which exclusive radiation was considered the standard of care; thus, 5-year survival for stages IB2, IIB, IIIB, and IVA are 72.2, 63.7, 41.7, and 16.4%, respectively, according the 1998 Annual Report on the Results of Treatment in Gynaecological Cancer [6]. The lengthy permanence of this unimodal treatment was due, on the one hand, to the classical concept that cervical cancer is a disease that progresses in an orderly fashion (local, then regional, and at the very last, systemic); therefore, it could be effectively treated with a local modality such as radiation instead of a systemic modality such as chemotherapy. On the other hand, the role of surgery for locally advanced cases failed to treat the disease successfully by radical surgical procedures [7]. Over the last 20 years, however, an increasingly number of trials that incorporate either chemotherapy and/or surgery with radiation (neoadjuvant chemotherapy followed by radiation, neoadjuvant chemotherapy followed by surgery plus minus adjuvant radiation, and concurrent chemoradiation) have been performed in an attempt to improve treatment results. Radiation concomitant with cisplatin-based chemotherapy is considered the current standard of treatment. This combined modality produces an absolute increase in 5-year survival of 12% as compared with radiation alone. On the other hand, neoadjuvant chemotherapy when followed by surgery – but not when followed by radiation – yields a 15% increase in absolute 5-year survival. These data emerged from three meta-analyses of the literature based on individual patient analysis [8,9].

Treatment of IVB and recurrent disease
Patients with cervical cancer may present at diagnosis with distant metastases (stage IVB) or have, after primary treatment, pelvic recurrence, distant metastases, or a combination of both. Recurrence rates vary from up to 20% to 70% in early stages and locally advanced disease, respectively [10,11] and the majority of recurrences occur within 2 years of diagnosis. At this stage, cervical cancer is even more difficult to treat because this clinical situation is the result of a more malignant phenotype resulting from accumulation of genetic defects during tumor progression and previous therapies; thus, the tumor at this stage is commonly resistant to chemotherapy. Moreover, these patients frequently have a poor performance status that limits use of agressive chemotherapy and the majority of patients die as a result of uncontrolled disease. In this setting, the few patients who recur after initial surgical treatment can be salvaged with concurrent chemoradiation if the disease is local or locoregional [12]. Those who receive primary radiation or chemoradiation and have pelvic disease can be offered an ultraradical procedure such as pelvic exenteration; nonetheless, this procedure is currently limited to patients with small and central tumors that in these situations, pelvic exenteration may offer 5-year survival for up to 50% of patients [13]. Although some efforts have been devoted to extending the exenterative procedures to patients with higher disease burdens by use of intraoperative radiation [14], laterally extended pelvic exenteration [15], or pre-exenterative chemotherapy [16] none of these options are widely used. Unfortunately, patients with IVB and those with distant metastases – with or without pelvic relapse – have no option other than systemic chemotherapy that in this setting has limited value; cisplatin is the most active single agent [17] and more recently in combination with topotecan has shown a modest increase in time to progression and median survival as compared with single agent cisplatin. In any case, median survival remains between 6 and 12 months [18].

 

Molecular pathogenesis of cervical cancer 

Human papillomavirus
Current experimental and epidemiologic information undoubtedly points to the human papillomavirus (HPV) as the primary causal agent in development of cervical carcinoma. Therefore, the study of its carcinogenic role continues to represent the mainstream research on the molecular biology of cervical cancer, with the idea that prophylactic and therapeutic applications of knowledge from this field could benefit millions of women afflicted, or at risk to be afflicted, with HPV-induced cervical cancer.

HPV classification is predicted on DNA sequence homology. At least 200 types have been identified and these have been classified into 16 groups [19]. Genital HPVs are classified according to their potential to induce malignant transformation as follows: high-risk types (16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, 68, 73, and 82); probable high-risk types (26, 53, and 66), and low-risk types (6, 11, 40, 42, 43, 44, 54, 61, 70, 72, 81, and CP6108) [20]. Among high-risk strains, HPV 16 and 18 are those most closely associated with cervical carcinoma and are found in >50% and 20% of squamous cell carcinomas, respectively [21].

A large body of knowledge supports the view that high-risk HPV types (HR-HPV) have the ability to transform cells into a malignant phenotype. Nevertheless, only a minority of cervical lesions infected with HR-HPV inevitably progress to cervical carcinoma, as indicated by frequent spontaneous clearance of HPV infection and the long delay between onset of persistent infection and emergence of the malignancy. For that reason, studies have been focused on analyzing the participation of possible viral and cellular factors governing HPV-induced malignancy.

HPV structure
Human papillomaviruses (HPV) belong to the Papovaviridae family. They consist of a 72-capsomere capside containing the viral genome. Capsomers are composed of two structural proteins: the 57 kD late protein L1, which accounts for 80% of the viral particle, and the 43–53 kD minor capside protein L2. The HPV genome consists of eight kilobasepairs (Kbp) and is a double-stranded DNA molecule. Arrangement of the 8–10 open reading frames (ORFs) within the genome is similar in all papillomavirus types, and partly overlapping ORFs are arranged on a sole DNA strand. The genome can be divided into three regions: the long control region (LCR) without coding potential; the region of early proteins (E1–E8), and the region of late proteins (L1 and L2) [22].

Early gene products
E1 and E2 encode proteins that are vital for extrachromosomal DNA replication and completion of the viral life cycle [23]. A hallmark of HPV-associated cervical carcinoma is loss of the expression of viral E2 protein [24]. A fusion product consisting of the small E8 ORF with part of the E2 protein has been described. This fusion protein is able to repress viral DNA replication as well as transcription, and is therefore believed to play a major role in the maintenance of viral latency observed in the basal cells of infected epithelium [25,26]. The E4 protein is expressed in the later stages of infection when complete virions are being assembled, and is not known to have transforming properties; however, it is considered to play an important role in the maturation and replication of the virus [27]. The E5 in open reading frame is often deleted in cervical carcinoma cells, indicating that it might not be essential in maintaining the malignant transformation of the host cell. Nevertheless, it has been reported that E5 protein possesses a weak transforming activity [28].

 

E6 and E7 proteins
E6 and E7 are the most important oncogenic proteins. Transcription of E6 and E7 genes has been observed to occur always in cervical carcinomas, being the first indication of a main role of these viral genes in HPV-associated tumorigenesis. The immortalization and transforming potential of E6 and E7 proteins have been demonstrated in tissue culture and in experimental animal models [29]. From the studies of E6-p53 and E7-pRb models, numerous actions have been identified of viral gene products on cellular proteins. Therefore, several findings hint at possible ways by which HPV-infected cells may escape controls governing cell growth and proliferation.

The E6 protein of high-risk HPV anogenital types shows weak oncogenic potential in the majority of established cell lines, and cooperation with E7 protein is required for full transforming capacity. Discovery of the inactivation of the tumor suppressor genes p53 and pRB by E6 and E7 oncoproteins provided a basic explanation of how high-risk HPV types induce their oncogenic effects on cervical cells [30]. E6 has many interactions with cellular proteins; nevertheless, its key action is inhibition of the function of tumor suppressor protein p53 by enhancing its degradation through the ubiquitin pathway [31,32]. To inhibit p53 function, E6 requires a cellular protein called E6-associated protein (E6AP). In non-infected cells, ubiquitin-mediated degradation of p53 is triggered by the mdm-2 protein, while in HR HPV-infected cells the E6-E6AP complex replaces mdm-2 in control of cellular p53 levels. This shift shortens the p53 half-life and reduces its levels in cervical carcinoma cells to less than one half of the level found in normal ephithelial cells [33]. It is known that increase in p53 levels plays a critical role in the induction of genes that results in cell cycle arrest [34], allowing repair of damaged DNA or activation of apoptotic pathways [35]. Therefore, cells expressing E6 maintain low levels of functional p53, altering normal response to DNA damage and favoring accumulation of genomic mutations. Binding of the E7 oncoprotein on pRB provides a complementary function. Binding releases transcription factor E2F that activates expression of genes that stimulate DNA synthesis in the cell. If earlier E6 action had freed the same cell from p53 control, that cell survives into the S phase with damaged DNA and, through E7 action, is able to replicate the HPV DNA [36]. Oncogenic properties of E6 and E7, as well as their effects on p53 and pRB, have provided the general basis for further investigations of the role of HPV in carcinogenesis in the HPV-infected cervix. Research in the action of the two oncoproteins have shown how they subvert key cell cycle and regulatory processes such as cyclins, cyclin-dependant kinases (CDKs), and cyclin-dependant kinase inhibitors (CDIs), among other interactions, to transform and immortalize the host cell [37].

DNA integration
HVP DNA is usually extrachromosomal or episomal in benign cervical precursor lesions. Cancer tissues may contain both episomal and integrated HPV DNAs at the same time, although integration appears to occur more frequently in HPV 18- than in HPV 16-associated cervical cancer [28]. During HPV DNA integration, the viral genome usually breaks in the E1/E2 region. The break generally leads to loss of the E1 and E2 regions. Loss of E2 results in uncontrolled and increased expression of E6 and E7 oncogenic proteins. Increased expression of E6 and E7, meanwhile, has been observed to lead to malignant transformation of host cells and to tumor formation [38]. HPV viral integration into host genome DNA is associated with progression from polyclonal to monoclonal status in cervical intraepithelial neoplasia (CIN), and these events play a fundamental role in the progression from low- to high-grade cervical neoplasia [39].

Epigenetics
Epigenetics can be defined as the study of genoma function that is contained outside of DNA itself and by means of which stable alterations in gene expression are set. Epigenetics is a well-established phenomenon that plays a major role in a diversity of biological processes such as embryonic development, cancer biology, and immune system response, among many others. The two most widely studied epigenetic changes are DNA methylation and histone acetylation; however, the picture is much more complicated than this, with new players coming onto the scene such as the RNA interference phenomenon, which has proven to be implicated in transcriptional silencing through small duplex RNA molecules that recruit silencing complexes to the chromatin [40,41,22].

DNA methylation
DNA methylation is a covalent chemical modification that occurs at the cytosine ring, resulting in the addition of a methyl (CH3) group at the carbon 5 position. According to the fact that DNA is made up of four bases and that therefore 16 possible dinucleotide combinations can occur, the CpG dinucleotide should have a frequency of 6%. However, the actual presence is only 5–10% of its predicted frequency. The human genome is not methylated uniformly and contains regions of unmethylated segments interspersed with methylated region. In contrast to the remainder of the genome, smaller regions of DNA, called CpG islands – and ranging from 0.5–5 kb and occurring on average every 100 kb – have distinctive properties. These regions are unmethylated, GC-rich (60–70%), have a ratio of CpG to GpC of at least 0.6, and thus do show no suppression of dinucleotide CpG frequency. Approximately one half of all genes in humans have CpG islands, and these are present in both housekeeping genes and genes with tissue-specific patterns of expression [43-46]. At least three functional DNA methyltransferases (DNMTs) have been identified; the most abundant is DNMT1, which preferentially methylates hemi-methylated DNA. DNTM1 localizes to replication foci and is responsible for maintaining proper methylation levels during replication and possibly DNA repair [47,48]. Other known functional methyltransferases are DNMT3a and DNMT3b, which are responsible for de novo methylation during embryogenesis [49]. DNMT3a and DNMT3b have equal preferences for hemi-methylated and non-methylated DNA, and thus have been classified as de novo methyltransferases[50]. In addition to DNMTs, the machinery of methylation includes demethylases, methylation centers triggering DNA methylation, and methylation protection centers [51]. The effect of DNA methylation on gene transcription can only be seen in the context of chromatin remodeling players. DNA methylation can directly interfere with transcriptional factor binding and thus inhibit replication [52], in addition to the ability of DNA methyltransferases DNMT1, DNMT3a, and DNMT3b to repress transcription in a methylation-independent manner [53]. Methyl-CpG binding proteins, which can recognize methylated DNA, have been shown to associate with large protein complexes containing HDACs and chromatin-remodelling activities, and it has also been suggested that DNA methylation could produce gene silencing by methyl binding domain proteins that recruit histone methyltransferases, which methylate lysine 9 in histone H3 and subsequently repress gene transcription [54]; as a result, histones are deacetylated and gene transcription is most often repressed.

 

Histones and post-translational modifications
How double-strand DNA is packaged into the dynamic structure of chromatin is crucial for the process of transcriptional control by regulating transcription factor accessibility to DNA regulatory sequences. Chromatin is constituted of nucleosomes, which are comprised of 146 base pairs of DNA wrapped around a core of two copies each of H2A, H2B, H3, and H4 histone proteins. These proteins suffer post-translational modifications that play a prominent role in gene expression regulation and signal transduction pathways such as methylation, acetylation, ubiquitination, phosphorylation, and sumoylation, which determine chromatin architecture and ultimately gene transcription [57].

The most widely studied modification is acetylation. Addition of charge-neutralizing acetyl groups to lysine residues on histones disrupts interactions with DNA, resulting in chromatin decompactation, greater access of DNA to transcription factors, and the presence of a transcriptionally active genomic locus. This post-translational modification depends on the net local balance between activities of histone acetyltransferase (HAT) and histone deacetylase (HDAC). In general, increased levels of histone acetylation (hyperacetylation) are found in more decondensed euchromatin, whereas decreased levels of acetylation (hypoacetylation) are characteristic of more condensed heterochromatin [58]. However, this mechanical model is an oversimplification of how gene transcription is regulated as additional histone modifications influence transcription [59].

Histone methylation can occur on lysine and arginine residues, giving the cell another layer of regulatory options, for example, lysine 9 in histone H3. It is currently known that histone arginine methylation is more dynamic, correlating well with gene activation and its loss from target arginines in H3 and H4 with gene inactivation. In contrast, lysine methylation appears to be a more stable mark. In this sense, methylation of lysine 4 in histone H3 correlates with gene activation, whereas methylation of lysines 9 and 27 in histone H3 correlates with repression [61-63]. Phosphorylation is another important and long-appreciated histone modification often associated with chromosome/chromatin condensation that includes mitosis, meiosis, apoptosis, and DNA damage, events regulated by different histone kinases (for example, members of the Aurora/AIK family [64-66]. Along with this post-translational modification of histone proteins, sequence-specific DNA binding by transcription factors and other protein remodeling factors determine a histone code for gene-specific transcriptional control that may dictate which modification or specific combinations of histone modifications can affect distinct downstream events by altering chromatin structure and/or generating a binding platform for protein effectors that can specifically recognize the modification and initiate gene transcription or repression [67].

Epigenetic alterations in cancer
Because of the close interplay between DNA methylation and histone modifications, it is expected that both mechanisms are operating in disease processes such as cancer; nonetheless, for the majority of tumor types the epigenetic defects could be just one of the many molecular cell alterations that lead to the malignant phenotype.

DNA methylation and cancer
Abnormalities in DNA methylation have long been associated with cancer. Both hypo- and hypermethylation play a prominent role in carcinogenesis, and their contribution shows scarcely defined boundaries. It has long been known that in cancer cells both alterations coexist: malignant tumors show global hypomethylation and regional hypermethylation. Whether one must precede the other or whether both should start at the same time remains to be elucidated. In terms of carcinogenesis, the first observations in fact were done on hypomethylation [69]; later, the discovery of regional hypermethylation as a means to silence the tumor suppressor genes expression gained the most attention [70].

 

Hypermethylation and gene silencing
Observations that tumor suppressor genes can be inactivated not only through structural changes (mutation, deletion) but also by lack of expression due to promoter hypermethylation positioned tumor suppressor gene epigenetic silencing as a well-established oncogenic process [71]. The first suppressor gene known to be hypermethylated and silenced was RB [72], which was followed by multiple publications describing similar findings for a variety of tumor suppressor genes, among them p16, MLH1, VHL, and E-cadherin [73].

Whether gene promoter hypermetylation is the cause or consequence for the tumor suppressor gene silencing is still a matter of controversy; nevertheless, these views are not mutually exclusive. That DNA methylation is causal has been shown by the ability of diverse pharmacologic compounds and molecular techniques to reactivate gene expression upon inhibition of DNA methylation in cancer cells [74].

On the other hand, other findings suggest that hypermethylation-induced gene silencing could be secondary to changes that determine gene expression, such as chromatin modification, so that methylation helps to maintain the silenced status of the gene. Strong support for the second view came from experiments showing that methylation of histone H3 lysine 9 – that is, chromatin modification – occurred, along with re-silencing of p16 in absence of DNA methylation in cells in which p16 had previously been activated by knocked out of DNA methyltransferase [75] and by data demonstrating p16 silencing in mammary epithelial cells that had escaped senescense and had demethylated the promoter [76].

 

Hypomethylation and gene activation
It is known that tumor cells have global DNA hypomethylation that can be as high as 60% less than their normal counterparts [77]. This hypomethylation occurs mainly in the body of genes (coding regions and introns), as well as in pericentromeric regions of chromosomes rich in repetitive DNA sequences [79]. Interestingly, hypomethylation is progressive from premalignant conditions to fully developed malignancies [80]. The main mechanisms put forward in attempting to explain cancer causation by hypomethylation include chromosome instability and reactivation of transposable elements and/or inappropriate gene activation [81]

There are two pieces of convincing evidence linking hypomethylation with chromosomal instability. The congenital disorder ICFs syndrome (immunodeficiency, chromosomal instability, and facial anomalies) caused by mutations at DNMT3b demonstrates loss of methylation in classical satellite DNA and mitogen-inducible formation of bizarre multiradial chromosomes that contain arms from chromosomes 1 and 16 [82]. This disorder, however, is not associated with cancer, but common somatic tumors such as breast, ovarian, and other epithelial tumors commonly have unbalanced chromosomal translocations with breakpoints in the pericentromeric DNA of chromosomes 1 and 16 [83]. In mouse models with an inactivated allele of NF1 and p53 genes, introduction of a hypomorphic DNMT1 allele caused a 2.2-fold increase in LOH frequency [84].

Finally, some reports have stressed the fact that many CpG islands are normally methylated in somatic tissues [85], and that demethylation could lead to activation of nearby genes such as HRAS. Indeed, experimental demonstration exists that hypomethylation leads to activation of genes important for cancer development, including promoter CpG demethylation and overexpression of 14-3-3sigma, maspin, heparanase, and S100A4 in several tumor types [86-88]. The question here is whether over-expression was indeed caused by hypomethylation or whether promoters are hypomethylated secondary to its high transcriptional activity. There are data showing that the sole hypomethylation as achieved by pharmacologic means is not sufficient to activate gene expression. In this context, some genes are not permisive for expression; this means that despite the fact that methylation is relieved the necessary ancillary factors to activate transcription are not present. Others are permissive and therefore reactivated by demetylation, whereas for others hypomethylation does not affect their levels of expression but can be over-expressed due to activation of signalling pathways known to activate them [89].

 

Chromatin and cancer
All classical genetic alterations – for instance, mutations in oncogenes or tumor suppressor genes of malignant cells – eventually affect gene transcription (mutant Ras, HER2 amplifications), or are transcription factors in themselves (c-myc, p53). It is therefore not surprising that the machinery of transcription control could be directly involved in the carcinogenesis process. Although the complex nature of the regulation of transcription is clear, certainly a disruption in the balance of activities of enzymes in charge of maintaining histone acetylation status is expected to occur in cancer. Among histone acetylases, the coding genes of p300/CBP have been found altered in some neoplasms. Mutations have been observed in epithelial tumors such as lung, esophageal, ovarian, and gastric tumors [90-93]. Chromosomal translocations involving CBP or p300 that in turn disrupt transcription by its fusion with partern genes are well-described molecular defects leading to hematologic malignancies such as some forms of acute myeloid leukemias [94,95]. Histone deacetylase activity leading to aberrantly repressed transcription was one of the first described leukemogenesis events. In acute promyelocytic leukemia, PML-RARα translocation generates a fusion gene product that represses transcription by associating with a co-repressor complex that contains HDAC activity [96]. Similar mechanisms account for other types of leukemia and lymphomas such as AML1-ETO and LAZ3/BCL6, respectively [97,98].

Despite the fact that participation of DNA methylation and chromatin in the carcinogenic process is unquestionable, it must be borne in mind that the split of epigenetics and genetics as separate types of defects in cancer is very artificial. In fact, according to the definition of epigenetics as genetic information not contained in the DNA sequence itself, current evidence demonstrates that primary genetic defects (mutations in genes with no known primarily methylating or chromatin-modifying activity such as growth factor receptors, adhesion molecules, etc,, or mutations in genes that in themselves affect DNA methylation or chromatin such as DNMTs or HAT/HDAC) are those leading to altered DNA methylation and chromatin changes. Demonstration that exogenous or endogenous carcinogens without causing primarily gene mutations lead to epigenetic abnormalities should prove that epigenetics is by itself one of the carcinogenic steps.

Internal server error