Deregulated expression of Polycomb-group oncogenes in human malignant lymphomas and epithelial tumors

Abstract

Genes belonging to the Polycomb-group (PcG) are epigenetic gene silencers with a vital role in the maintenance of cell identity. They contribute to regulation of various processes in both embryos and adults, including the cell cycle and lymphopoiesis. A growing body of work has linked human PcG genes to various hematological and epithelial cancers, identifying novel mechanisms of malignant transformation and paving the way to development of new cancer treatments and identification of novel diagnostic markers. This review addresses the current insights in the role of PcG genes in development of human malignancies.

POLYCOMB GENES, PROTEINS AND COMPLEXES

Polycomb-group (PcG) genes play a vital role in stable inheritance of tissue- and cell type-specific gene silencing patterns and are responsible for maintenance of cellular identity through many successive cell generations. PcG genes were first discovered as epigenetic gene silencers during embryogenesis and have been found to play a role in development of the heart (1,2), the skeleton (3–5) and the nervous system (6,7). However, PcG genes play a central role in regulation of various adult processes, including the cell cycle, lymphopoiesis and X-inactivation (8), and several PcG genes have been identified as oncogenes.

PcG proteins use conserved protein domains to form multimeric complexes (Table 1) (reviewed in 9–13). In humans, the Polycomb repressive complex 1 (PRC1) contains the BMI1/MEL18, RING, HPC and HPH PcG proteins. Another complex, PRC2, is composed of EED, EZH, SUZ12 and YY1. These proteins represent the evolutionary conserved complex ‘core’, but the fine composition of PcG complexes is determined by cell type and tissue type and by proliferation status (10,14–17). Several PcG proteins interact or colocalize with various non-PcG proteins, including the transcription modulators CtBP (18), E2F6 (14,19), KyoT2 (20), RYBP (21), AF9 (22), SSX (23) and the MAP/KAP kinase 3pK (24). All of these proteins may contribute to the silencing activity of PcG complexes and their ability to bind chromatin, and many are involved in oncogenesis.

MECHANISTIC ASPECTS OF GENE SUPPRESSION BY PcG COMPLEXES

PcG complexes bind chromatin and are directly involved in remodeling of chromatin structure and, as a consequence, control of gene activity. Two of the best-known forms of chromatin modification are histone tail acetylation and methylation, which, as a rule-of-thumb, result in gene activation and silencing, respectively (25). PcG complexes associate with, or contain, various enzymes that modify histone tails, including histone deacetylases (HDACs), histone methyl transferases (HMTs) and an ubiquitin ligase (Table 1). In PRC2, the EED protein recruits the HDAC1/2 HDACs (26), whereas EZH2 is a known HMT that methylates lysine 27 of histone H3 (27–29). The other PcG complex, PRC1, is associated with HMT activity and contains CBX4/HPC2, which has SUMO E3 ligase activity, and RING1/RNF2, which are E3 ubiquitin ligases that ubiquitinate lysine 119 of histone H2A (30–34). Most of these enzymatic activities are evolutionary conserved (35–38), and transcriptional silencing of PcG target genes is closely linked to introduction of ‘epigenetic marks’ such as methylated lysine residues on histone tails (39–41). In addition, interference with key components of PcG complexes removes these epigenetic marks and relieves repression of PcG target genes (39,42). It is now commonly accepted that alteration of chromatin structure is the main mode of action of PcG complexes.

A hypothetical mechanism of PcG-mediated gene silencing (reviewed in 9,10) proposes that maintenance of gene silencing requires both PcG complexes. This process is initiated by HDAC activity of the PRC2 ‘initiation’ complex, which deacetylates core histone tails. This is followed by methylation of these residues, most likely through HMT activity of EZH2 in PRC2. Experiments in Drosophila demonstrated transient interaction between components of the two complexes at the time when silencing is established (43). Also, methylated histone 3 tails are thought to form a docking site for the PRC1 ‘maintenance’ complex. Recruitment of the PRC1 complex may lead to additional methylation by the SUV39H1 HMT in PRC1 and establishment of a stable pattern of gene suppression.

PcG GENES AND THE CELL CYCLE

PcG genes are the focus of intense scientific scrutiny because of their direct role in cell cycle control and malignant transformation (Table 2). The best-known example is regulation by the Bmi1 PcG protein of a major cell cycle checkpoint in which the inhibitors p16^INK4A and p19^ARF (p14^ARF in humans) play a central role. Lymphocytes in mice with transgenic overexpression of Bmi1 downregulate p16^INK4A and p19^ARF and have a high predisposition for B- and T-cell lymphomas (44–48). This observation directly links the PRC1 complex, containing Bmi-1, to regulation of apoptosis and senescence in mice. During normal cell cycle control, the retinoblastoma (Rb) gene product is hyperphosphorylated (pRb) by the complex of cyclin D and cyclin-dependent kinases 4 and 6 (Cdk4/6). Rb normally binds to the E2F transcription factor, but pRb is unable to do so and allows transcription of E2F-dependent promoters. This results in cell cycle progression because genes that are necessary for the G1/S transition are transcribed. The role of p16^INK4A in this process is such that it prevents binding of Cdk4/6 to cyclin D and thereby inhibits phosphorylation of Rb, resulting in inhibition of E2F-mediated transcription, cycle arrest and senescence. In addition, p19^ARF inhibits p53 degradation by sequestering MDM2, which results in p53-mediated cell cycle arrest and apoptosis.

These observations have produced a model of cell cycle regulation by the PRC1 protein Bmi1, where Bmi1 downregulates p16^INK4A and p19^ARF. This results in cell cycle progression because Rb is hyperphosphorylated and p53 is degraded. Notably, in the absence of Bmi1, p16^INK4A and p19^ARF are upregulated and cells become subject to apoptosis/senescence pathways (7, 48). Although these observations suggest that the PRC1 complex is the main player in regulation of the cell cycle, evidence is rapidly increasing that the PRC2 complex is intricately involved in this process as well. For instance, overexpression of the PRC2 EZH2 gene in human breast epithelial cell lines promoted transformation and conferred invasive capacity (49), demonstrating that transformation capacity is not limited to PRC1. The situation becomes even more complex by the discovery that the PRC2 genes EZH2 and EED are themselves downstream target genes of the E2F–pRb pathway (50,51), suggesting that the PRC1 complex may contribute to regulation of components of PRC2.

The development of malignant lymphomas in Bmi1-transgenic animals and the transforming capability of Bmi1 in primary MEFs immediately raise the question of whether human BMI-1 possibly contributes to development of cancer by a similar mechanism. One indication that it may indeed contribute to oncogenesis came from studies of primary human fibroblasts, where overexpression of BMI1 resulted in an extended replicative lifespan by inhibition of p16^INK4A (52). Although BMI1 overexpression did not result in immortalization of the fibroblasts, similar studies in mammary epithelial cells demonstrated that experimental overexpression of BMI1 produced elevated telomerase activity and immortalization (53), suggesting that BMI1 can contribute to malignant transformation by multiple mechanisms which are cell type-specific.

These studies, and the work in mutant mice, collectively identified BMI1 and EZH2 as bona fide oncogenes with cell type-specific transforming capabilities and have naturally placed these genes in the spotlight of cancer research in humans. However, the extent to which artificial overexpression of these genes in cell models reflects the situation in patients is unclear. However, a rapidly growing body of work, has demonstrated that neoplastic cells in various human cancers all display abnormal patterns of PcG gene expression.

PcG ONCOGENE EXPRESSION IN HUMAN MALIGNANT LYMPHOMAS

Altered PcG gene expression is widespread in human malignant lymphomas, and neoplastic cells in primary nodal large B-cell lymphomas (LBCL) (54,55), mantle cell lymphomas (MCL) (56,57) and Hodgkin's lymphoma (HL) (58–60) abnormally express BMI1 and several of its PRC1 binding partners. In the normal counterparts of these malignancies, the mature B-cells in germinal centers (GC), expression of the majority of PcG genes encoding PRC1 and PRC2 is separated in resting and activated cells (16,61,62). For instance, BMI1 and several of its binding partners in PRC1 are primarily detected in resting B-cells in the GC mantle zone and in non-dividing centrocytes of the GC follicle. In contrast, these genes appear silent in proliferating follicular centroblasts which express the PRC2 genes EZH2 and EED instead. However, cycling neoplastic cells in human malignant lymphomas have lost the mutually exclusive expression of core components of PRC1 and PRC2. These cells coexpress all PcG proteins constituting the PRC1 and PRC2 complexes, suggesting an inability to downregulate the PRC1 complex during cell division (55,56,58,59).

Aberrant BMI1 expression in cycling neoplastic cells of malignant lymphomas correlates well with development of lymphomas in Bmi1 transgenic mice and strongly suggests that human BMI1 contributes to lymphomagenesis. This is supported by the distinct expression patterns of individual PRC1 PcG genes in clinically defined primary LBCL subgroups (55) and their close correlation with biological behavior (63–66). For instance, dividing neoplastic cells in primary nodal LBCL overexpress BMI1 and several of its binding partners and have an unfavorable to intermediate prognosis (55). In contrast, primary cutaneous LBCL have a better prognosis and express these binding partners as well, but in the absence of BMI1. Also, re-analysis of a previously published database of gene expression in diffuse LBCL, using a new clustering algorithm, indicated that the BMI1 oncogene is predominantly expressed in the subset of non-GC phenotype DLBCL (67). These results collectively show that BMI1 expression in malignant B-cell lymphomas correlates with their biological behavior, suggesting that BMI1 plays a role in development of these malignancies. In addition, BMI1 expression patterns may be diagnostically relevant (55).

An immediate question is whether there is any evidence that BMI1 expression in human malignant lymphomas results in suppression of p16^INK4A/p14^ARF, as it does in mice. In HL and HL-derived cell lines, the presence of BMI1 in tumor cells was not related to absence of p16^INK4a expression: ∼50% of the HL patients showed strong nuclear co-expression of BMI1 and p16^INK4a (58). Importantly, neoplastic cells that were positive for both BMI1 and p16^INK4a also expressed the proliferation marker MIB1, suggesting that they were in cycle despite the presence of p16^INK4a. More recent unpublished studies of diffuse LBCL also demonstrated that the presence of BMI1 in MIB1^POS neoplastic cells does not necessarily coincide with the absence of p16^INK4a or p14^ARF (van Galen et al., manuscript in preparation). These findings agree with studies of p16^INK4a in aggressive LBCL and HL published by others, which showed that these tumors frequently express p16^INK4a and p14^ARF and that loss of p16^INK4a/P14^ARF is mainly associated with promoter hypermethylation, gene loss or mutation (68). Although the currently available evidence does not allow us to rule out that BMI1, being an epigenetic regulator, contributes to hypermethylation of the p16^INK4a/p14^ARF promoter, the PcG expression studies collectively show that the presence of BMI1 in human malignant LBCL does not clearly correlate with the absence of p16^INK4a. What explains this discrepancy with the mouse studies? A recent screening of human cells using almost 8000 siRNA constructs for different human genes showed that BMI1 was not among the genes that allow circumvention of the p53 pathway (69). Therefore, the target genes of BMI1 in human lymphocytes may be different from those in mouse lymphocytes. An alternative explanation is that the PRC1 complex in human lymphocytes behaves differently than in mouse lymphocytes, for instance, because of distinct variation in mouse and human PcG complex composition. This could lead to altered activity of the PcG complex, because several binding partners of BMI1 have opposing roles (70,71). Also, major players in the Rb–p16 pathway are capable of directly interacting with members of the PRC1 PcG complex, and the amount of BMI1 and HPC2 may determine whether the PcG silencing complex is targeted toward the p16^INK4a/p14^ARF locus or toward cyclin A and cdc2 genes (72).

PcG ONCOGENE EXPRESSION IN HUMAN EPITHELIAL TUMORS

Is abnormal PcG gene expression particular for hematological malignancies or is PcG gene expression also disturbed in other human cancers? Our recent work and that of others demonstrated that neoplastic cells in solid tumors, including medulloblastomas (6), and tumors originating from liver (73,74), colon (74), breast (49,75,76), lung (74,77), penis (78) and prostate (79), all display disproportionate expression of PcG genes. Also, some of these patterns are associated with loss of differentiation by tumor cells, metastatic behavior and a worse prognosis.

Similar to GC B-cells, healthy epithelial cells in lung and breast tissue primarily employ PRC1 genes, including BMI1, in the absence of EZH2 (75,80). However, in premalignant precursor lesion of lung carcinomas and bronchial squamous cell carcinomas, increased expression of BMI1 is observed in dividing cells (77,80). At the same time, expression of EZH2 and the proliferation marker MIB1 is elevated in BMI1^POS cells, according to the severity of the histopathological stage (80). These patterns suggest that altered expressions of BMI1 and EZH2 are early events in development of lung carcinoma and that they precede high proliferation rates. Contrary to the situation in HL (58), upregulation of BMI1 in lung carcinomas correlates with downregulation of p16^INK4a in lung carcinomas (77), penile carcinomas (78) and colorectal cancer (81). This could indicate that the mechanism of cell cycle regulation by BMI1 in epithelial cells is different from the mechanism in lymphocytes, which could be related to cell type-specific variation in PcG complex composition and target gene recognition (10,15).

Interestingly, development of breast cancer appears to follow a similar pattern (75). Major differences in PcG gene expression between normal breast tissue and ductal hyperplasia, well-differentiated ductal carcinoma in situ or well-differentiated invasive carcinomas, could not be demonstrated using immunohistochemistry (75). However, poorly differentiated DCIS and invasive carcinomas were found to frequently express EZH2 in combination with proteins constituting the PRC1 complex (75), and BMI1 overexpression was observed in invasive ductal breast carcinoma and axillary lymph node metastases (82). In addition, whereas the majority of BMI-1/EZH2 double-positive cells in poorly differentiated DCIS were resting, poorly differentiated invasive carcinoma displayed an enhanced rate of cell division within BMI-1/EZH2 double-positive neoplastic cells (75). This suggests that deregulated expression of EZH2 appears to be associated with loss of differentiation and development of poorly differentiated breast cancer in humans and that increased EZH2 expression precedes high frequencies of proliferation during breast cancer development. These patterns are in agreement with an earlier study showing that EZH2 protein levels were strongly associated with breast cancer aggressiveness and that overexpression of EZH2 promotes anchorage-independent growth, cell invasion and transformation (49). Interestingly, increased expression of EZH2 has also been observed in two studies of prostate cancers using a combination of gene expression profiling and tissue arrays (79,83) and has been proposed to be a marker for metastatic prostate cancer (79). However, conflicting results have been reported in studies of prostate (84) and ovarian cancers (85), and it has been suggested that enhanced EZH2 expression may be associated with a subset of tumors defined by high proliferation indices (84,86) and not necessarily reflect a direct contribution of EZH2 to carcinogenesis or metastasis.

CONCLUDING REMARKS

Evidence is rapidly increasing that PcG genes are a novel class of oncogenes and anti-oncogenes, which may in future years become central to the development of novel cancer therapies based on epigenetic gene silencing (87,88). Not only are PcG genes such as BMI-1 and EZH2 capable of cellular transformation, but also they are vital for cell survival. To date, abnormal PcG gene expression has been described in most human cancers. Also, the correlation between PcG expression and biological behavior of clinically defined cancer subtypes suggests that these genes play a central role in oncogenesis, and holds a promise for discovery of novel diagnostic markers (55).

Our current knowledge does not allow us to provide a clear explanation for altered PcG expression in human malignancies yet. Some PcG genes are located on chromosomal regions known to be associated with chromosomal abnormalities (89–91) (Table 2), and gene amplifications of BMI1 and EZH2 have been described in MCL (57) and solid tumors (50) and L3MBTL deletions in myeloid leukemia (91). However, it remains to be determined whether PcG overexpression observed in human tumors is directly related to oncogenesis. In fact, one could even argue that inappropriate PcG expression in cancer cells may not be abnormal at all. For instance, in human LBCL, not only the BMI1 PcG gene appears to be expressed abnormally, but also its binding partners RING1 and HPH1 (55). Is it possible that these genes, present on different chromosomes, are coordinately deregulated, in addition to all the other genetic abnormalities that have been described in malignant lymphomas? Or is there an alternative explanation? Recently, a minor follicular B-cell subpopulation was discovered that may represent a transitional stage between proliferating centroblasts and resting centrocytes and that expresses all major components of the two PcG complexes (16). In this regard, this population of cells closely resembles dividing neoplastic cells in nodal LBCL and may represent its normal counterpart (54,55). What could be the relevance of this observation? In Drosophila, coexpression and actual interaction between the two PcG complexes occur only transiently during early larval development (43). At later stages, the two complexes are separated, like they are in the majority of healthy GC B-cells (16,62). Are healthy follicular B-cells, in which PRC1 and PRC2 core components are coexpressed, more susceptible for malignant transformation, because chromatin-modifying enzymatic activities that are normally separated in PRC1 and PRC2 are active at the same time? And does the ‘abnormal PcG expression pattern’ in nodal LBCL then merely reflect a gene expression profile that is frozen at a time of transformation?

Another aspect of PcG gene expression in tumors that we need to understand in more detail is how PcG genes affect the biological behavior of tumor cells. Obviously, inappropriate expression epigenetic regulators may result in loss of normal gene silencing pathways and thus contribute to loss of cell identity. One possibility is that this underlies the observed epigenetic alterations seen in many tumor cells (87,92). A very recent development (reviewed in 93,94) is the finding that BMI1 is vital for survival of healthy and leukemic stem cells. It has been proposed that BMI1 may not only confer a growth advantage to cells, but also confer ‘stemness’ to cells which might place PcG genes central in the process of dedifferentiation (92). In future years, fundamental epigenesis research shall answer many of our questions and direct us to novel biological problems. Polycomb research is clearly leading us to a better understanding of new theories in cancer research, but it will also shine a new and brighter light on classic tumor cell biology.

References