Epigenetics in cancer stem cells

Cancer stem cells (CSCs) define a small, unique subset of cells with self-renewal ability and the capacity to generate the different cell types that constitute the whole tumor [1]. These cells are termed CSCs because of their “stem-like” properties commonly shared with normal tissue stem cells. Such properties include extensive self-renewal ability (symmetrical and asymmetrical) and differentiation capacity. It should be noted that a general capacity to differentiate is not a mandatory feature of CSCs and that the ability of CSCs to differentiate and repopulate the cell types found in the original tumor is of greater significance. More importantly, CSCs should demonstrate potent tumor-initiation capacity. This property is usually demonstrated by injecting limited number of CSCs into an orthotopic in vivo environment to generate the bulk tumor. Nevertheless, the concept of CSC is of significant importance as it highlights the need to eradicate the CSC populations to achieve an effective cure.

The first clear evidence of CSCs being a key tumor-initiating subset of cancer cells was demonstrated in acute myeloid leukemia (AML) where prospective CSCs were isolated using cell surface markers that identify normal haematopoietic stem cells and evaluated for their tumor-initiating properties [2, 3]. Since then, similar identifications of tumor-initiating populations have been identified in multiple solid tumors that includes brain, breast, liver, ovary, prostate, lung, melanoma and colon cancers, by using different cell surface markers or through side population (SP) analysis [4‐11]. For instance, in malignant glioma and medulloblastoma tumors, a putative neural stem cell marker, CD133 has been demonstrated to be adequate and essential to initiate and recapitulate the tumor upon xenotransplantation into immune-compromised mice [5]. However, this finding has been disputed as tumors can also arise from CD133-negative cells in a subset of glioma tumors [12]. In addition, CD133 surface marker expression has been demonstrated to change according to disease state and progression, further complicating its role as a bona fide CSC marker in brain tumors [13, 14]. In liver cancers such as hepatocellular carcinoma (HCC) and intrahepatic cholangiocarcinoma (ICC), similar use of cell surface markers such as epithelial cell adhesion molecule (EpCAM), cytokeratin 19 (CK19), CD133, CD90, CD44, CD24, and CD13 has been applied to define a subpopulation of liver cancer cells as CSCs [15]. Importantly, it has recently been shown that these CSC markers are not specific to liver CSCs, and that distinct populations of liver CSCs express different surface markers possibly due to the strong intra- and inter-heterogeneity and varied etiology of liver cancer [16]. As a result, CSC studies have begun to move away from the reliance of cell surface markers to identify tumor-initiating cells and have begun to identify other complementary methods of measuring the functional activities of CSCs that may serve to identify CSCs as well as the molecular mechanisms that regulate CSCs [17].

Currently, the central theme of the CSC model is the ability of a subset of cells at the apex of the hierarchy to propagate tumors and promote tumor progression as compared to the non-tumorigenic cells within the bulk tumor. One of the gold standards to functionally identify CSCs is the capacity of these cells to regenerate a phenotypic copy of the original tumor in an orthotopic transplantation model. Non-CSCs, by definition, lack this ability and fail to generate tumors in the transplantation model. It is important to note that the CSC hierarchy model may not be ubiquitous for all cancers and that some tumorigenic cells are common in certain cancers. It is also important to note that such transplantation assays measure the tumorigenic potential of the cells to form tumors and not their actual fate. For example, alterations in tumorigenic assays carried out by Quintana and colleagues showed that CSC frequency could be increased by changing several experimental parameters such as the use of extracellular matrix (ECM) in the form of matrigel, prolonging the duration for tumor formation, and varying the severity of immune-compromised mice used [18]. This study highlighted that the tumor-initiating capacity may be an artificial consequence of the conditions employed in xenograft mouse models.

While analyzing CSC surface marker expression in primary tumors has been often performed to study the clinical impact of CSCs on tumor progression, more often than not, this has resulted in ambiguous data possibly due to the fact that CSC properties that sustain the primary tumor phenotype are defined by more than just specific marker expression [19, 20]. Analysis of key signalling pathway activity that resembles those functioning in stem-like cells, is more likely to accurately interrogate the clinical contribution of CSCs. An example of such studies was carried out by Lim et al. in BRCA1 mutation-associated breast tumors, where the authors prospectively isolated distinct subpopulations of normal and tumorigenic epithelial cells from BRCA1 mutation heterozygous individuals and found that luminal progenitors were highly represented in BRCA1 mutation-associated breast tumors, more than the stem cell population [21]. This suggests that luminal progenitors are more likely the cells-of-origin for BRCA1 mutation-associated breast tumors, which was later confirmed in a transgenic mouse model study carried out by Molyneux and colleagues [22]. These studies highlight the predictive capability of gene expression mapping of pathway activation rather than specific marker identity. In a separate study, John Dick and colleagues demonstrated that tumor-initiating AML stem cells contribute to disease progression and patient survival outcome, underscoring the importance of functionally defining the CSCs [23]. More importantly, the contribution of CSCs, with preferential activation of core stem cell programs, to patient survival outcome has been demonstrated. The study by Shats et al. showed that a stemness gene signature derived from embryonic stem cells (ESCs) could predict a breast cancer patient cohort sensitive to drugs linked to this signature using a Connectivity Map [24], demonstrating the clinical contribution of CSCs to patient outcome [25]. Collectively, these studies highlight that CSCs that perpetuate tumors are not merely defined by surface marker expression, but more importantly and more accurately by their gene expression profiles and consequent pathway activations.

Epigenetic regulation of the genome is one of the primary means by which genetic code is altered to control cellular developmental hierarchies. Epigenetic mechanisms such as histone modifications, DNA methylation, chromatin remodelling and even changes in noncoding RNAs including miRNAs together govern the epigenome landscape that dictate the outcome of cell fate specification without changes to the DNA sequences. Such changes in the genome is important during normal mammalian development and ESCs differentiation [26]. Importantly, gene expression profiles change during cellular differentiation according to not only a network of transcription factors but also the “epigenomic landscape” of the cell. For the purpose of this review, we will focus our discussions on two primary mechanisms of epigenetic regulation: histone modifications and DNA methylation.

Histone methylation occurs predominantly on lysine (K) and arginine (R) residues and these methylation marks serve as docking sites for histone readers [27]. Both lysine and arginine methylation can occur on both histones and non-histone proteins. The highly conserved histone lysine methylation occurs at three different levels: mono-, di-, and tri-methylation. Such modifications are commonly associated with gene activation or repression, depending on the target histone modification. For instance, histone H3 lysine 4 (H3K4), histone H3 lysine 36 (H3K36), and histone H3 lysine 79 (H3K79) are associated with gene activation whereas histone H3 lysine 9 (H3K9), histone H3 lysine 27 (H3K27) and histone H4 lysine 20 (H4K20) are associated with gene repression. The N-terminal tails of histones frequently undergo other post-translational modifications, which play significant roles in various DNA-templated processes, including transcription [28]. Hence, aberrations in histone modifications can lead to deregulated gene expression as seen in various human disease and malignancies.

DNA methyltransferases (DNMTs) are a class of enzymes involved in transferring a methyl group from S-adenosyl methionine (SAM) to cytosine bases of CpG dinucleotides at gene promoters and regulatory regions [29]. CpG dinucleotides are concentrated in short CpG-rich regions known commonly as “CpG islands”. In humans, CpG islands occupy about 60% of the gene promoters. CpG promoter islands can be methylated during development that results in long-term gene silencing. One classic example of such naturally occurring CpG methylation is the X-chromosome inactivation and the imprinted genes. DNA hypermethylation has also been associated with the silencing of tumor suppressor genes as well as differentiation genes in various cancers [30]. The reduced expression of these genes may then contribute to the formation of CSCs within tumor cell populations [31, 32]. Indeed, the importance of DNA methylation in maintaining CSC properties have been reported in leukemic, lung and colon stem cells [33‐35]. The accumulation of epigenetic abnormalities has been suggested to be an early event that predisposes these tumor cells to acquire further mutations and genomic instability. This is supported by the fact that epigenetic machinery is crucial for the maintenance of normal stem and progenitor cells and that any epigenetic deregulation can lead to accumulation of cells with increased stemness properties and self-renewal ability, thus giving rise to CSCs.

As epigenetic mechanisms have important functions in modulating stem cell properties in cancer cells, targeting components of these epigenetic pathways would help in eradicating both CSCs and the bulk tumor population. Inhibitors of epigenetic modulatory enzymes such as HDACs and DNMTs have been widely studied and many are currently in clinical trials for treatment of a variety of cancers. In addition, deregulation of chromatin remodelling has been associated with tumorigenesis and tumor progression, thus making chromatin remodelling proteins viable targets for small molecule inhibitors as well. Indeed, many of these therapeutic strategies aim to induce differentiation of CSCs and to sensitise these cells to chemotherapy, with the ultimate goal of reducing tumor relapse and improving patient survival. Here, we review the development of various epigenetic therapies designed to target different components of the epigenetic machinery. A summary of these epigenetic drugs and their clinical status can be found in Table 1.