Clinical & Experimental Metastasis

, Volume 26, Issue 1, pp 5–18

Epigenetic contributions to cancer metastasis

Authors

    • Departments of Biochemistry, Oncology and PaediatricsUniversity of Western Ontario
    • The London Regional Cancer Program, London Health Sciences Centre and the EpiGenWestern Research Group at the Children’s Health Research Institute
Review

DOI: 10.1007/s10585-008-9166-2

Cite this article as:
Rodenhiser, D.I. Clin Exp Metastasis (2009) 26: 5. doi:10.1007/s10585-008-9166-2

Abstract

The molecular basis of cancer encompasses both genetic and epigenetic alterations. These epigenetic changes primarily involve global DNA methylation changes in the form of widespread loss of methylation along with concurrent hypermethylation events in gene regulatory regions that can repress tissue-specific gene expression. Increasingly, the importance of these epigenetic changes to the metastatic process is being realized. Cells may acquire an epi-genotype that permits their dissemination from the primary tumour mass or the ability to survive and proliferate at a secondary tissue site. These epigenetic changes may be cancer-type specific, or in some cases may involve a common target gene providing a selective advantage to multiple metastatic cell types. In this review, I examine the growing volume of literature related to the epigenetic contributions to cancer metastasis. I discuss the functional importance of these epigenetic phenomena and how new epigenetic biomarkers may permit the identification of diagnostic signatures of metastasis and the development of new cancer therapies.

Keywords

EpigeneticMetastasisHypermethylationHypomethylation

Introduction

Cancer is a complex collection of diseases that differ on the basis of the tissue of origin, the genes within which mutations have occurred, as well as the clinical consequences of these causative molecular events [1]. These diverse clinical cancer phenotypes include age of onset, stage, progression and their propensity to metastasize to specific distant organs within the body. In fact, up to 90% of cancer deaths are due to the spread, or metastasis, of cancer cells from the primary site. Metastatic cells may differ genetically from cells in the primary tumour and differ clinically with respect to their susceptibility to treatment modalities [2]. It is well established that both genetic and epigenetic alterations contribute to the molecular basis of cancer. What is now becoming increasingly evident is the important contribution of epigenetic changes to the metastatic process. These epigenetic changes include functionally relevant modifications to chromatin that are essential for development and differentiation and which allow the genome to respond to developmental and environmental cues [3]. Here I review the present state of knowledge regarding epigenetic contributions to metastasis, which may provide a selective advantage for cells to disseminate from the primary tumour mass or to eventually grow and prosper at a distant site. DNA hyper- and hypomethylation changes have been identified at individual gene loci, and molecular technologies are now available to identify epigenetic contributions to metastasis at the genome level. These approaches can permit the identification of diagnostic and prognostic signatures of metastasis. Furthermore, this new information will allow us to unravel the functional importance of these epigenetic phenomena, with the potential to translate this knowledge into cancer therapies targeting epigenetic changes in metastatic cells.

The basics of cancer metastasis

Metastasis is a complex, multi-step biological process characterized by a series of distinct yet interrelated steps that vary in their timing and efficiency [2, 4]. The interruption of any of these steps, either due to non-permissive genetic mutations or through treatment interventions, can delay or halt the metastatic cascade. The initial step of metastasis occurs at the site of the primary tumour, where growth occurs through the initiation of blood vessels that provide nutrients and a potential conduit through which cells could spread. Cross talk between tumour and blood vessel endothelial cells leads to the release of angiogenic factors that induce the synthesis of new vessels to feed that tumour. Next, invasion of the stromal tissue surrounding the primary tissue occurs, as cells breach the basement membrane and invade into the surrounding tissues. Again, as with the previous step, genetic alterations in certain cells evolving within the tumour mass can provide these cells with the ability to disassemble the extracellular matrix. This allows some of these cells to eventually find their way to the lymphatic or blood microvessels that form their conduit for dissemination throughout the body. Once in the circulatory system, many metastasizing cells are destroyed, but some end up arrested in the capillary beds within distant organs, where they have the opportunity to extravasate into that new host tissue. These single cells have several potential fates, including being eliminated from their new site or remaining dormant until extracellular conditions or subsequent genetic changes within the cell permit further growth and proliferation [5]. In time, some of these distant micrometastases can commence a new round of colonization in the new host tissue involving the initiation of blood vessel growth at this secondary site.

The complexity of cancer metastasis not only relates to the cascade described above, but also with respect to the defining molecular characteristics and origin of the metastatic cell, the efficiency of the metastatic process, the tumour cell/stromal cell interactions at the secondary site and the molecular characteristics responsible for the entrance and release of metastasized cells from dormancy at the new host tissue site [5, 6]. Recent evidence from whole genome microarray studies indicates that complex gene signatures superimposed over the primary tumour genotype may exist in metastasized cells. These gene signatures are likely to define the efficiency of the metastatic process and the ability of individual metastatic cells to survive at specific secondary tissue sites [7]. It is likely that a proportion of these alterations in gene expression are the consequence of altered epigenetic patterns of DNA methylation and histone modifications [8].

The basics of cancer epigenetics

In general, the multistep process of cancer progression involves both genetic and epigenetic alterations that drive normal human cells into malignant derivatives [1, 8]. Evidence has been collected that metastasis, as well as the re-emergence of quiescent tumour cells at their secondary tissue site may be influenced, permitted (or in some cases prevented) by losses or gains in gene expression affecting the steps within the metastatic cascade. These alterations may be genetic in origin, resulting from sporadic mutational events or influenced by predisposing inherited mutations that lead to early onset familial cancers. In addition, evidence is beginning to emerge that gene-specific expression changes result from epigenetic modifications that either alter transcription directly (i.e. by inactivating metastasis suppressors or other genes) or through chromatin effects by increased genetic instability and the resulting chromosome imbalances that lead to gene dosage changes.

Several recent reviews regarding mammalian epigenomics [9], cancer epigenomics [8, 10] and the function of chromatin modifications [11], have provided excellent overviews concerning the epigenetic contributions to tumourigenesis, as well as highlighting potential diagnostic, prognostic and therapeutic applications from this rapidly progressing field [8]. Epigenetic regulation involves heritable modifications that do not change the DNA sequence but rather provide ‘extra’ layers of control that regulate how chromatin is organized and genes are expressed [12]. These adaptations include dynamic chromatin marks that are primarily controlled by reversible patterns of (a) DNA methylation and (b) histone protein modifications. These marks confer a structural adaptation of chromosomal regions that can register, signal or perpetuate altered activity states [13]. Such instructions ensure proper patterns of gene expression and chromosomal integrity by ‘tagging’ targeted DNA sequence with a special status relative to the surrounding DNA. DNA methylation is established and maintained by DNA methyltransferases (DNMTs) at cytosines within CpG dinucleotides clustered within the 5′ regulatory regions of most housekeeping genes. These cytosines are usually unmethylated in actively transcribed genes while, in contrast, methylated cytosines are generally associated with silenced DNA.

Epigenetic regulation also involves post-translational modifications to histone proteins that can orchestrate DNA organization and gene expression. Histone modifying enzymes are recruited to ensure that a receptive DNA region is either accessible for transcription, or that DNA is targeted for silencing (by histone deacetylases; HDACs). These histone modifications are themselves complex in their type (e.g. including acetylation, methylation and phosphorylation) and with respect to the specific histones (H3, H4) and lysine residues (K4, K9, K20 and K27) on which these modifications occur. In general, transcriptionally active chromatin has predominantly unmethylated DNA along with high levels of acetylated histones, while silenced chromatin has been remodelled with methylated DNA and deacetylated histones [8]. Specific combinations of histone modifications are responsible for regulating gene expression pattern [9, 14] via an epigenetic ‘tag’ that marks DNA with a special status that activates or silences genes and permits (or impedes) the recruitment of transcription complexes to CpG-rich gene regulatory regions. These reversible modifications ensure that specific genes can be expressed (or repressed) in response to specific developmental or biochemical cues such as changes in hormone levels, dietary components, drug exposures or from extracellular cues from the surrounding cellular environment.

The proper timing, establishment and maintenance of epigenetic patterns of DNA methylation and/or histone acetylation are essential for normal development. In contrast, temporal or spatial epigenetic errors are causative factors for virtually all human disease states, including cancer, genetic disorders and paediatric syndromes, as well as being contributing factors in autoimmune diseases and aging [12]. DNA methylation changes are the better characterized of these changes, in part due to molecular technologies that have allowed gene-specific and genome-wide changes to be identified. Dynamic, aberrant patterns of DNA methylation and histone modifications that occur in cancer cells can modify chromatin accessibility by transcription factors, alter gene expression patterns and increase the risk of genomic instability [8]. The primary effects on DNA methylation patterns in cancer cells include transcriptional repression due to promoter hypermethylation [15] and an overall loss of methylation across the cancer cell genome [16]. Key genes down-regulated by promoter hypermethylation include tumour suppressor genes that have anti-proliferative functions [12, 15], and may be hypermethylated in a tissue-specific manner [1719]. Global genomic hypomethylation in cancer cells increases the risk of genomic instability and can drive the tumourigenic process [16, 20]. In addition, proteins such as the DNMTs that are directly responsible for establishing and maintaining DNA methylation may have altered expression patterns in human cancers [21]. Characterizing these epigenetic patterns has yielded substantial insight into the epigenetic contributions to cancer progression, and the responsiveness of epigenetic patterns to environmental cues [22, 23].

Although much recent work has focused on defining epigenetic contributions to all aspects of tumour progression, similar contributions to metastasis are not well understood. Conceptually, the dynamic and plastic nature of epigenetic patterning in cells makes this process a logical one by which metastasis could occur, given the responsiveness of these epigenetic patterns to environmental cues and their roles in perpetuating altered chromatin states [13]. The secondary tissue site may provide a welcoming home for metastasized cells because the new environment presents an appropriate ‘soil’ within which the extravasating cells can proliferate, leading to a secondary tumour. On the other hand, dormancy may occur, reflecting a restrictive genetic and epigenetic response by a cell to the novel environmental conditions it experiences at its new host environment. These negative environmental signals may include interactions with surrounding cells that are non-permissive to growth, or deficiencies in essential nutrients such as folate that are required for DNA synthesis and methyl metabolism [24].

Genetic contributors to metastasis

Over the past 5 years, our understanding of the molecular characteristics of metastasis has been best understood in the context of genetic changes identified using gene-specific, tissue-specific as well as whole genome approaches. Individual metastasis suppressor genes have been identified that, when lost or mutated, are permissive to the metastatic or invasive phenotype [25, 26]. For the most part, these studies have involved candidate gene approaches in which the expression status of individual genes has been evaluated in cell lines and/or tumour material in a tissue-specific manner. For example, significantly reduced levels of mRNA expression of the metastasis suppressor genes BRMS1, KISS1 (kisspeptin), KAI1 (CD82) and Mkk4 (MAP2K4; mitogen-activated protein kinase kinase 4) have been shown in breast cancer brain metastasis [27], with specific suppression of BRMS1 modifying several metastasis-related phenotypes [28]. In other cases, certain of these metastasis suppressor genes, such as KAI1, are associated with tumour progression and/or metastasis in a variety of human tumours [29]. Similarly, KISS1 inhibits metastasis in both in vivo melanoma and breast carcinoma models [30], RhoGLI is a marker for aggressive human cancers [31] and down-regulation of NDP kinase A (nm23/NME1) significantly correlates with distant metastases and poor survival [32].

Whole genome approaches using microarray platforms have identified prognostic gene sets that predict a short interval to distant metastases (i.e. a poor prognosis signature [3336]) or have identified gene sets that mediate metastasis from a specific primary tissue to a tissue-specific host site [7, 3745]. Minn and co-workers [39] identified a 54 gene breast cancer set that marks and mediates breast cancer metastasis to the lungs. The complexity of this gene set is evident on several levels, since at least two separate classes of genes may be present in this set; one class that confers both breast tumourigenicity and lung metastagenicity and a second class that confers functions advantageous to cells in that lung environment. Additionally, Kang and co-workers identified a functionally diverse gene set that when overexpressed, cooperatively promotes the metastasis of breast cancer cells to bone [40]. In head and neck small cell carcinoma (HNSCC), a 685 gene signature associated with lymph node metastasis can be dissected into multiple components, including down-regulation of tumour cell specific genes and up-regulation of stromal cell specific genes [46]. Finally, a gene expression signature has been identified that distinguishes primary from metastatic adenocarcinoma, with tumours possessing the signature being more likely associated with metastasis and poor clinical outcome [47]. These reports underscore the complexity of metastasis as a multigenic process and support the concept that some sub-populations of cells in the primary tumour possess a gene set that is permissive for metastasis and/or for the colonization and growth of those cells at specific secondary sites.

Epigenetic contributors to metastasis

The epigenetic contributions to metastasis are less well understood than the expression changes in the metastasis-associated genes and gene sets described above. For the most part, attempts to identify epigenetic contributions have focused on the mapping of increased DNA methylation within the promoter regions of individual candidate genes. These hypermethylation events usually result in the repression of gene expression that may permit tumour cells to metastasize or have a selective advantage at the secondary tissue site. Mapping such DNA methylation changes requires the use of specialized technologies to identify these alterations on either a gene-specific or genome-wide basis [8]. Detecting aberrant methylation within the promoter regions usually involves methylation-sensitive restriction enzymes to define differences in CpG methylation status within CpG islands, as well as PCR-based methods that combine sodium bisulphite treatment with genomic sequencing to map methylation at CpGs across a gene’s promoter region [48]. The presence of aberrant methylation patterns requires further verification to determine their functional effects in gene expression, by analyzing mRNA levels via Realtime RT-PCR or Northern analyses that can confirm gene expression changes.

To a large extent, our understanding of the relationship between epigenetic alterations and metastasis is at a similar stage as our genetic knowledge from 5 years ago. As a result, the literature regarding these epigenetic contributions to metastasis is complex, somewhat haphazard and largely descriptive. Table 1 provides a compilation of the genes affected in different cancer types. Most reports relate DNA methylation changes in individual genes or in the context of individual cancers. In some reports, cell line models have been employed while others have combined these in vitro approaches with xenograft models and to clinical material [54, 76]. Furthermore, only a few studies have taken a functional approach to manipulate gene expression patterns by exposing cells to demethylating agents such as 5-aza-2′-deoxycytidine (5-aza-C) that can ‘reset’ DNA methylation patterns at hypermethylated promoters [64, 107, 119]. While DNA methylation changes have been specifically identified in many genes, these studies have provided only a limited understanding of the functional mechanisms affected by such epigenetic alterations and their relationship to metastasis. As described below, disease-specific studies have been undertaken to understand the role of epigenetic regulation in metastasis in both cell line model systems and in clinical materials.
Table 1

Hypermethylated gene promoters associated with metastasis

Downregulated Gene(s)

Gene name

Reference

Bladder cancer

    MIM (MTSS1)

Metastasis suppressor 1

[49]

Breast cancer

    TFPI2

Tissue factor pathway inhibitor 2

[50]

    CXCL12

Chemokine (C-X-C motif) ligand 12

[51]

    CIITA

Class II, major histocompatibility complex, transactivator

[52]

    uPA

PLAU; urokinase plasminogen activator

[53]

    CDH1

E-cadherin

[54]

    ESR2

Estrogen receptor 2

[55]

    CST6

Cystatin M/6

[56]

    ESR1, BRCA1, CDH1

Estrogen receptor 1, breast cancer susceptibility gene 1; E-cadherin

[57]

    PITX2

Paired-like homeodomain 2

[58]

    ITIH5

Inter-alpha (globulin) inhibitor H

[59]

    ID4

Inhibitor of DNA binding 4

[60]

    hMLH1

mutL Homolog 1, colon cancer, nonpolyposis type 2

[61]

    LATS1 LATS2

Large tumor suppressor, homolog 1/2

[62]

    MGMT

O-6-Methylguanine-DNA methyltransferase

[63]

    CLCA2

Calcium activated chloride channel 2

[64]

    IRF8

Interferon regulatory factor 8

[65]

    p16INK4a, CEA

Cyclin-dependent kinase inhibitor 2A, carcinoembryonic antigen-related CAM 5

[66]

    TIMP3

Tissue inhibitor of metalloproteinase 3

[67]

    Multiple genes

Cyclin D2, RAR-β(, TWIST, RASSF1A, HIN-1

[68]

    Multiple genes

RASSF1A, APC, TWIST, CDH1, GSTP1, RAR-β2

[69]

    Multiple genes

MRP-1/CD9, PMP-22, CASP-3

[70]

    Multiple genes

p16, THBS2, CDH1, RAR-β2, MINT1, MINT2, MINT31

[71]

Infiltrating breast cancer

    CDH1

E-cadherin

[72]

Endometrial carcinomas

    14-3-3 sigma

SFN; stratifin

[73]

extrahepatic bile duct carcinoma

    p16INK4a

Cyclin-dependent kinase inhibitor 2A

[74]

Cholangiocarcinoma

    Multiple genes

CDH1, MGMT and 22 other genes

[75]

Colorectal cancer

    CXCL12

Chemokine (C-X-C motif) ligand 12

[76]

    PIK3CG

Phosphoinositide-3-kinase, catalytic, gamma polypeptide

[77]

    p16INK4a, p14ARF

Cyclin-dependent kinase inhibitor 2A

[78]

    APAF1

Apoptotic peptidase activating factor 1

[79]

    TPEF (TMEFF2)

Transmembrane protein with EGF-like and two follistatin-like domains

[80]

    Multiple genes

APC, CDH1, and 10 other genes

[81]

Esophageal small cell carcinoma

    PLC delta1

Phospholipase C, delta 1

[82]

    Lin-7C

Lin-7 homolog C

[83]

    p14ARF

Cyclin-dependent kinase inhibitor 2A

[84]

    P300

E1A binding protein p300

[85]

Gastric cancer

    CDH1

E-cadherin

[86]

    CDH1

E-cadherin

[87]

    EphA7

EPH receptor A7

[88]

    LIMS2

LIM and senescent cell antigen-like domains 2

[89]

    DAPK

Death-associated protein kinase 1

[90]

    KLF4

Kruppel-like factor 4

[91]

    CTNNB1

Catenin (cadherin-associated protein), beta 1

[92]

    SOCS1

JAK binding protein

[93]

    SYK

Spleen tyrosine kinase

[94]

    Multiple genes

By microarray analyses

[95]

Gastrointestinal

    HHIP

Hedgehog interacting protein

[96]

Glioblastoma

    TIMP3

Tissue inhibitor of metalloproteinase 3

[97]

Head and neck small cell carcinoma

    Multiple genes

By restriction landmark genomic scanning; RLGS

[98]

Non-small cell lung carcinoma

    BRCA1/2, XRCC5

Breast cancer susceptibility genes 1 and 2; Ku80

[99]

Lung cancer

    Multiple genes

APC, CDKN2A, GSTP1, DAPK, MGMT

[100]

Melanoma

    GnTV (MGAT5)

N-Acetylglucosaminyltransferase V

[101]

Nasopharyngeal carcinoma

    Thy1

Thy-1 cell surface antigen

[102]

Oral squamous cell carcinoma

    CDH1

E-cadherin

[103]

    p16INK4a/p14ARF

Cyclin-dependent kinase inhibitor 2A

[104]

Ovarian cancer

    Multiple genes

p16, RARβ, CDH1, H-cadherin, APC, GSTP1, MGMT, RASSF1A

[105]

Pancreatic endocrine neoplasms

    Multiple genes

p16/INK4a, APC and 9 other genes

[106]

Prostate cancer

    TIMP2

Tissue inhibitor of metalloproteinase 2

[107]

    ESR2

Estrogen receptor 2

[108]

    ALOX15

Arachidonate 15-lipoxygenase

[109]

    Multiple genes

GSTP1, APC, PTGS2

[110]

    Multiple genes

CDH1, APC and 14 other genes

[111]

Renal cell carcinoma

    p14ARF

Cyclin-dependent kinase inhibitor 2A

[112]

    DAL1

Erythrocyte membrane protein band 4.1-like 3

[113]

Thyroid cancer

    SERPINB5

Maspin

[114]

Uveal melanoma

    TIMP3

Tissue inhibitor of metalloproteinase 2

[115]

Multiple cancers; brain mets

    DAPK

Death-associated protein kinase 1

[116]

Multiple cancer types

    RUNX3

Runt-related transcription factor 3

[117]

LN multiple cancers

    MGMT, p16INK4a

O-6-Methylguanine-DNA methyltransferase, cyclin-dependent kinase inhibitor 2A

[118]

Breast cancer

Multi-gene approaches have an important role in the diagnosis, prognosis and treatment of metastatic breast cancer [3941, 120] and this strategy has also been directed at genes that are epigenetically inactivated. One approach has been to concurrently determine the DNA methylation and expression status of candidate genes in matched normal tissue, primary cancer and metastatic tissues from individual patients. For example, Shinozaki and co-workers assessed six known tumour suppressor genes and related genes. They showed that hypermethylation in E-cadherin (CDH1) was significantly associated with primary breast tumours demonstrating sentinel lymph node (SLN) metastases, with measurable methylation also evident in the other genes: CDH1 (90% of patients), RASSF1A (59%), RAR-β2 (48%), APC (34%), TWIST (28%) and GSTP1 (24%) [69]. Furthermore, in some cases promoter hypermethylation was correlated with specific tumour characteristics. For example, this group also showed that GSTP1 methylation was associated with increased tumour size, CDH1 methylation was highly associated with estrogen receptor (ESR1) negative tumours and RAR-β2 hypermethylation was more frequent in HER2-positive tumours. In another smaller study, similar multi-gene correlations linked breast SLN metastases with epigenetic alterations in CDH1 and RAR-β2 as well as with p16 and thromospondin 2 (THBS2) [71], while another group correlated hypermethylation of ESR1 and BRCA1 with hypermethylated CDH1 in breast LN metastases [57]. CDH1 was also shown to be frequently methylated in infiltrating breast cancer [72] and in a mouse xenograft model, 5-aza-C could restore CDH1 expression and suppress growth of both primary tumour and lung metastases [54]. On the basis of these studies and those in other cancers described later in this review [81, 86, 87], it is evident that CDH1 is a common epigenetic target in cancer metastasis and a potential candidate for targeted epi-therapy.

Promoter methylation has also been observed in a wide variety of other essential molecular pathways in the context of metastatic breast cancer. These include genes involved in apoptosis (caspase 3 [70]; interferon regulatory factor 8 IRF8 [65]), DNA repair (MGMT [63]; hMSH2/hMLH1 [61]), the regulation and composition of the extracellular matrix (TIMP3 [67]; TFPI2; tissue factor pathway inhibitor 2 [50]; ITIH5 [59] inter-alpha (globulin) inhibitor H5), transcription (ID4 [60]; PITX2 [58]) and the cell cycle (p16 [66]; LATS1/LATS2 [62]). In addition, epigenetic silencing of the lysosomal cysteine protease inhibitor cystatin 6 (CST6) is more frequently observed in metastatic lesions than in primary cancers [56] and the epigenetic silencing of the chemokine CXCL12 (rather than its receptor CXCR4) contributes to the metastatic potential of mammary carcinoma cells [51], as well colorectal cells (described below; [76]). Finally, there is some evidence that the loss of expression due to epigenetic inactivation observed in breast cancer metastasis may in fact be reversible. The loss of ESR2 expression is more pronounced in invasive ductal than in lobular carcinoma and if present in the primary tumour will persist in the axillary LN metastasis [55]. This group also showed that treatment of ESR2 cell lines with DNA methyltransferase inhibitors restored ESR2 expression. Similarly, the metastasis-associated calcium-activated chloride channel gene (CLCA2) can be re-activated with 5-aza-C [64], while promoter hypomethylation of the urokinase plasminogen activator gene (uPA/PLAU), a gene that is only expressed in highly invasive cells, can be reversed with S-adenosylmethionine. This treatment results in the significant inhibition of uPA expression, tumour cell invasion in vitro as well as tumour growth and metastasis in vivo [53].

Colorectal cancer

The complexity of colorectal cancer (CRC) is compounded by the heterogeneity of genetic and epigenetic contributions responsible for the progression of this disease. Alterations in genes involved in proteolysis, cell-cell adhesion, angiogenesis and cell survival contribute to CRC [121]. The Wnt signalling pathway is frequently targeted in CRC and other cancers [122], and it appears that inactivation of certain genes may be responsible for Wnt activation and the conferring of a selective advantage for metastatic cells with such targeted hypermethylation events. For example, Tang and co-workers showed that CDH1, WIF1 (WNT inhibitory factor 1) and SFRP1 (secreted frizzled-related protein 1) are more frequently methylated in lung adenocarcinomas derived from CRC metastasis, while APC is more commonly methylated in lung primary adenocarcinomas [81]. Immunosurveillance pathways may also be targeted by hypermethylation events. Wendt reported the over-expression of DNA methyltransferase enzymes and the loss of expression of the CXCL12 chemokine in primary colon carcinomas and colon cell lines due to promoter methylation [76]. Furthermore, xenografts of stably transfected, CXCL12-expressing HT29 colon cells showed that in vivo metastatic tumour formation was prevented when expression of CXCL12 was re-established. Concurrent with CXCL12 re-activation was an increase in caspase 3/7 activity and apoptosis in these transfected cells, suggesting that the loss of CXCL12-CXCR4 autocrine signalling through CXCL12 confers resistance to apoptosis in colorectal tumours [76]. Similarly APAF1, a key factor in the mitochondrial apoptotic pathway, also appears to be involved in metastatic progression. A proportion of CRC liver metastases exhibited allelic imbalance (AI) of 12q23 and also showed methylation within the APAF1 promoter, although the link between this hypermethylation and AI was not correlated [79]. A high degree of promoter methylation in the TPEF gene (TMEFF2; tomoregulin) has also been observed in liver metastasis [80]. As well, p16INK4 and p14ARF, also display simultaneous hypermethylation associated with lymph node metastasis and higher tumour grade and along with TPEF, these epigenetic changes may be useful prognostic biomarkers for CRC progression [78]. The methylation-dependent silencing of these and other genes [77, 96] shows that like other cancer types, epigenetic alterations can contribute to colorectal cancer cell invasion and metastasis and can affect clinical outcome through the silencing of key genes in a multitude of regulatory and signalling pathways.

Gastric cancer

Gastric cancer is a common cancer (particularly in Asia) that involves a multi-step cascade initiated by a variety of genetic, epigenetic and environmental factors (reviewed in [123]). Of the two main histological types of gastric cancer, diffuse-type gastric cancer (DGC) exhibits a higher frequency of invasion and lymph node metastasis than the more differentiated, intestinal-type of gastric cancer [119]. A wide variety of genetic events in gastric cancer are well characterized [123]). In contrast, de novo epigenetic events have been described in a limited number of genes, particularly in the context of promoter hypermethylation of genes in the APC/CDH1 pathway [15]. CDH1 methylation has been shown to be significantly higher in gastric tissues with lymph node (LN) metastases than in those without LN metastasis, and CDH1 methylation has been associated with serosal invasion [86]. A recent study has also shown that methylation of CDH1 is a frequent early event in gastric cancer progression and is significantly correlated with LN metastasis [87]. Hypermethylation and loss of β-catenin (CTNNB1) expression, an integral component of the WNT signalling pathway, has been reported in a sub-group of primary gastric cancer, cell lines and in metastases [92]. Other genes reported to be epigenetically down-regulated in gastric cancer metastasis are diverse in their roles and functions. Hypermethylation of Death-associated protein kinase (DAPK) was observed in intestinal, diffuse and mixed type of gastric cancer and correlated with the presence of LN metastasis, advanced stage and poor survival [90]. Similar reports have correlated LN metastasis and poor survival with hypermethylation of the spleen tyrosine kinase SYK1 [94] and with promoter hypermethylation and hemizygous deletion of Kruppel-like factor 4 (KLF4) [91]. Furthermore, SOCS1 (JAK binding protein) hypermethylation is associated with LN metastasis and advanced tumour stage [93]. Finally, several reports have shown similar gene-specific hypermethylation events correlated with metastasis and the loss of gene expression that could be restored when cells were treated with 5-aza-C and/or the histone deacetylase inhibitor trichostatin A (TSA; [89, 119, 124]). These reports include a genome-wide microarray approach that has implicated hypermethylation of sixteen genes of interest in the frequently metastatic DGC type of gastric cancer. These genes include PGP9.5 (UCHL1; ubiquitin carboxyl-terminal esterase L1), which may serve as a useful biomarker for aggressive disease [119].

Prostate cancer

Epigenetic contributors to prostate cancer are under intensive investigation and these reports have been compiled in several recent reviews [18, 125]. One approach has been to identify epigenetically targeted genes that can act as markers for early-stage prostate cancer and/or represent signatures of disease progression or metastasis. For example, Yegnasubramanian and co-workers evaluated the promoter methylation status of sixteen candidate genes in a panel of 83 metastatic prostate cancers and eight lymph nodes from 28 men [111]. Five genes (GSTP1, APC, PTGS2, MDR1 and RASSF1a) were highly methylated in prostate cancer tissues compared to normal tissues. Interestingly, the hypermethylation profile of the metastases was similar on the whole to their respective primary cancers, with the methylation profiles also being quite homogeneous across all metastases for any given patient. These observations support a clonal model of prostate metastasis in which a hypermethylation fingerprint is established early and then maintained throughout tumour progression, perhaps through the overexpression of the DNA methyltransferase enzyme DNMT1. Additional studies also support this clonal model and evoke the diagnostic and prognostic value of hypermethylation at certain genes (GSTP1, APC, PTGS2; [110]) with respect to prostate cancer metastasis. An inverse relationship between ESR2 promoter methylation in LN and bone metastases has been described whereby metastatic prostate cells regain ESR2 expression concurrent with the loss of promoter methylation [108]. The dynamic nature of gene methylation patterns has also been seen in the context of TIMP2, a member of the family of natural inhibitors of matrix metalloproteinases (MMP). TIMP2 is frequently methylated and its expression is suppressed in human prostate tumours and cell lines, yet this gene can be re-expressed in metastatic prostate cell lines after combined treatment with 5-aza-C and TSA [107]. Similar re-activation experiments have been performed in which 5-aza-C and TSA can re-activate the inflammatory response-related ALOX15 gene, which also exhibits promoter hypermethylation in prostate cell lines and tumour metastases [109].

Finally, in addition to these studies focussed on hypermethylation in prostate metastasis, disregulated gene expression resulting from altered patterns of histone protein modifications has also been implicated in prostate metastasis. For example, EZH2 is a member of the Polycomb group (PcG) complex of genes that methylates lysine 27 of histone 3 (H3K27), subsequently leading to transcriptional repression of its gene targets [126]. Individual target genes of EZH2 have been identified [127, 128], as has a polycomb repression signature of fourteen direct gene targets of EZH2 that correlates with poor clinical outcomes in prostate and other cancers [126]. In addition, EZH2 may serve as a recruitment platform for DNA methyltransferases [129]. The overexpression of EZH2 in metastatic prostate cancer [130], including in peripheral circulating tumour cells [131] make it a promising candidate for the detection of metastatic prostate cancer cells and as a therapeutic target for treatment of the disease.

Other cancers

Similar reports have implicated a role for hypermethylation events and the inactivation of gene expression as contributors to metastasis in other cancers. In some cases, these events are occurring in genes that appear to be common targets for promoter hypermethylation, including CDH1 [103], DAPK [116], p16INK4a/p14ARF [74, 84, 104, 112, 118] and TIMP3 [97]. In other cases, cancer-specific hypermethylation profiling has been undertaken using candidate gene approaches (as in esophageal squamous cell carcinoma [8285]). Finally, multi-gene studies have been performed in the context of lung cancer [99, 100], cholangiocarcinoma [75] and pancreatic endocrine neoplasms [106].

Promoter hypomethylation

DNA hypomethylation events in cancer are less well defined in comparison to gene-specific hypermethylation and mostly involve the demethylation of intron/exon gene regions and multi-copy repetitive DNA sequences that may contribute to genomic instability (Table 2; [8, 16]). It is somewhat paradoxical that hyper- and hypomethylation occur concurrently in cancer cells, although these are likely independent events [18, 143]. Choi and co-workers showed higher levels of hypomethylation in LINE-1 and Alu elements in neuroendocrine tumours with lymph node metastasis [132] with the tendency for methylation at these repeats to decrease in prostate adenocarcinoma progression. As well, a strong correlation has been found between hypomethylation on chromosome 8 and the presence of metastasis in prostate carcinoma [133].

Hypomethylation events in single-copy genes have been identified in association with re-activation of expression and both tumour progression and metastasis (as reviewed in [16]). For example, overexpression and hypomethylation of S100A4, a calcium-binding protein previously implicated in metastasis, has been correlated with gene activation in colon adenocarcinoma cell lines [135], associated with poor differentiation and/or higher grade in pancreatic ductal adenocarcinomas [136] and endometrial carcinoma [137] and also correlated with medulloblastoma development [138]. Promoter hypomethylation in the protease-activated receptor 2 gene (PAR2) has been associated with gastric cancer [139], as has hypomethylation of MAGE (melanoma antigen) gene family members [140, 141]. Finally, certain genes such as uPA/PLAU [53] and synuclein gamma (SNCG; [142]), in which hypomethylation has been reported in a wide number of different cancers with high invasive or metastatic potential, offer the opportunity for genetic and epigenetic approaches to knock down expression of these metastasis-associated genes.
Table 2

Hypomethylated gene promoters associated with metastasis

Upregulated Gene(s)

Gene name

Reference

Breast cancer

    uPA (PLAU)

Urokinase plasminogen activator

[53]

Neuroendocrine tumours

    LINE-1, Alu repeats

Repetitive elements

[132]

Prostate cancer

    Chromosome 8

 

[133]

    uPA, MMP2, VEGF

PLAU, matrix metallopeptidase 2, vascular endothelial growth factor A

[134]

Colon cancer

    S100A4

S100 calcium binding protein A4

[135]

Pancreatic adenocarcinoma

    S100A4

S100 calcium binding protein A4

[136]

Endometrial cancer

    S100A4

S100 calcium binding protein A4

[137]

Medulloblastoma

    S100A4

S100 calcium binding protein A4

[138]

Gastric cancer

    PAR2 (F2RL1)

Coagulation factor II (thrombin) receptor-like 1

[139]

    MAGE genes

Melanoma-associated antigen A1, A3 genes

[140]

    MAGE genes

Melanoma-associated antigen A1, A2, A3 genes

[141]

Breast and ovarian cancer

    SNCG

Synuclein gamma

[142]

Translational applications of epigenetic targets in metastasis

Ultimately, the primary goal of biomedical research is to improve the human condition. In the context of epigenetic contributors to cancer metastasis, several avenues of progress related to diagnosis, prognosis and treatment have been undertaken to meet this objective. First, as described in previous sections of this review, many studies have identified specific genes associated with metastasis that have altered expression due to epigenetic changes. Whole epigenome signatures of metastasis have yet to be reported, in part due to the fact that the necessary promoter microarray platforms have only recently become available [144]. As a result, diagnostic and prognostic signatures of metastasis are incomplete or lacking. An underlying premise in identifying epigenetic signatures is that once identified as contributing to specific steps related to metastasis, these epigenetic alterations may be correctable by ‘epigenetic drugs’ that can restore the normal epigenetic patterns (and expression patterns) of these genes [145]. For example, several approaches have been undertaken to re-establish normal DNA methylation patterns in metastatic cells at gene targets such as CDH1 [54, 145]. DNMT over-expression has been linked to poorer overall survival and shorter metastasis-free survival [146]. Nucleoside analogues of cytidine (such as 5-aza-C) are inhibitors of DNMT through irreversible binding that leads to DNMT depletion and hypomethylation [147] and various studies have used these inhibitors to demethylate hypermethylated genes, restore their expression and in some cases, inhibit growth of the primary tumor and/or metastases in comparison with untreated controls [54, 64, 89, 107, 108, 119, 124, 148]. As well, antisense oligonucleotides targeting DNMTs have been developed [149], although no objective responses have been reported in Phase II clinical trials [150]. In addition, non-nucleoside analogues have been used to re-establish expression of tumour suppressor genes with concurrent anti-metastasis effects [151] and several histone deacetylase inhibitors [152] have been shown to modulate expression of metastasis genes [153, 154]. Finally, in contrast to strategies to re-activate gene expression of silenced genes, one report has shown that the use of S-adenosylmethionine and antisense oligonucleotides (directed at the methyl DNA-binding domain protein; MBD2), can inhibit tumour-promoting genes uPA/PLAU, VEGF and MMP2, with a concurrent inhibition of tumour cell invasion in vitro and growth in vivo [134]. Again, the concern with these approaches in the clinical context will be establishing the stability and specificity of the epigenetic reprogramming initiated by these therapeutic compounds.

Summary

The molecular approaches and platforms developed to identify gene-specific alterations in the DNA methylation and histone modifications have allowed significant strides to be made in our understanding of epigenetic contributions to metastasis. Progress will continue on several fronts. There is the need to develop epigenetic signatures of metastasis that will be used concurrently with genomic signatures to better map the evolving molecular landscape on which metastasis occurs so that diagnostic and prognostic markers can be identified. In particular this involves identifying critical steps in metastasis for which epigenetic changes are important contributors. Already it is evident that genetic and epigenetic alterations have considerable overlap, with the inappropriate loss or gain of gene expression in the cancer cell frequently due to the hyper- or hypomethylation events described in this review. As the functional relevance of these epigenetic events are confirmed, more efforts to translate this knowledge into the clinic will be undertaken. Already epigenetic drugs are being developed to reset gene expression either through the de-methylation (or remethylation) of under (or over) expressed genes respectively. One significant challenge to be met will be to re-set the epigenetic status of specific genes, without destabilizing the cancer epigenome as a whole and risking the emergence of more aggressive metastatic disease. In addition, the development of such targeted epi-therapies will provide an additional weapon for use in combination with more traditional therapeutic strategies. Ultimately a thorough understanding of the dynamic nature of epigenetic chromatin modifications and the epigenetic plasticity of metastatic cells in multiple tissue environments will lead to novel targeted strategies to control and cure metastatic disease.

Acknowledgements

I gratefully acknowledge financial support by grant #016506 from the Canadian Breast Cancer Research Alliance, with special funding support from the Canadian Breast Cancer Foundation and the Cancer Research Society. I also acknowledge funding support from the London Regional Cancer Program Small Grants Competition and the Lawson Health Research Institute Internal Research Fund. Finally, I thank my colleagues and the trainees and staff in my lab for stimulating discussions and critique during preparation of this review.

Copyright information

© Springer Science+Business Media B.V. 2008