Chromatin-Related Drivers Inducing CSC Formation

Leukemias represent a paradigm of hierarchical cancers maintained by LSCs. Numerous studies have identified mutated epigenetic regulators that favor the acquisition of uncontrolled self-renewal ability and initiate the disease. A prominent example is offered by mixed lineage leukemia (MLL)-associated leukemia, which is characterized by chromosomal rearrangements involving the KMT2A/MLL gene. KMT2A/MLL encodes a histone methyltransferase that orchestrates several essential cellular processes through modification of chromatin, mainly regulating accessibility to enhancer regions [4]. Oncogenic MLL fusion proteins created by translocations induce LSC formation in acute myeloid leukemia (AML) and acute lymphoblastic leukemia (ALL) 4, 5. Importantly, MLL fusion proteins can initiate the oncogenic process both in hematopoietic stem cells and in short-lived progenitors, suggesting that they are able to reprogram committed cells and actively confer de novo self-renewal capacity 6, 7, 8. Notably, efficient leukemic reprogramming of myeloid progenitors by MLL chimeras requires the action of a MLL antagonist, the Polycomb group (PcG) protein BMI1, to ensure repression of tumor-suppressor genes that would otherwise counteract the effect of the oncogenic fusions 9, 10. Furthermore, the observations that different cell types respond differently to the presence of MLL fusion proteins [11], and that diverse types of leukemia can be induced by the same oncoprotein, suggest that the epigenetic landscape of the cancer cell-of-origin may influence the effect of MLL chimeric proteins 12, 13. Recurrent mutations in MLL proteins have also been identified in a variety of solid tumors 14, 15, 16, 17, pointing to MLL proteins as general cancer drivers. In addition to histone modifiers, mutated structural proteins regulating the higher-order structure of chromatin, such as cohesins, have been shown to enforce stem cell transcriptional programs and have been implicated in the emergence of LSCs [18]. Moreover, inactivating mutations disrupting the function of chromatin-remodeling complexes, which are found at high frequency in various types of cancers, have been linked to aberrant activation of stem cell-related pathways 19, 20, 21.

Probably the most compelling evidence supporting a key role of chromatin in the acquisition of uncontrolled self-renewal comes from studies in glioblastoma (GBM), a highly aggressive form of brain cancer characterized by an undifferentiated phenotype and a high frequency of CSCs. Recent sequencing efforts have identified gain-of-function mutations in genes encoding histone H3 in about one third of pediatric GBMs. The gene mainly affected is H3F3A, and a K27M substitution is the most common alteration 22, 23. The primary mechanism leading to oncogenesis induced by the K27M mutation is inhibition of the Polycomb repressive complex 2 (PRC2), which results in genome-wide reduction in the repressive histone H3 trimethylated lysine 27 (H3K27me3) mark [24] and re-establishment of an earlier developmental program in neural precursor cells and consequent acquisition of oncogenic self-renewal ability [25]. The identification of a histone protein as a key driver in GBM has particularly important implications because it demonstrates a direct and major role of chromatin in the emergence of CSCs. Interestingly, H3.3 mutations are only found in pediatric GBM, suggesting a different mechanism of CSC formation in adult patients. Nevertheless, chromatin dysregulation is likely to play a crucial role in adult GBM as well, considering that approximately half of adult GBMs harbor mutations in at least one chromatin modifier [26].

Transcriptional Intratumoral Heterogeneity

The presence of at least two functionally distinct subsets of cells (tumorigenic self-renewing CSCs, and non-tumorigenic differentiated cells) in many cancers has been the basic observation supporting the hierarchical model of cancer development over the past two decades. Recent single-cell transcriptomic studies have extended our understanding of the gene expression programs that shape tumor architecture and have confirmed that aberrant differentiation programs establish cellular hierarchies within individual cancers. In a pioneering study, colon cancers were shown to contain multiple cell types with transcriptional profiles resembling those of the cellular lineages making up the normal epithelium. Importantly, single cancer cells could recapitulate the lineage diversity of the primary tumors in transplantation assays, demonstrating that multilineage differentiation represents a key source of intratumoral transcriptional heterogeneity [37]. Similar studies in glioblastoma and oligodendroglioma have confirmed this finding, showing that brain tumors contain subpopulations of undifferentiated cells characterized by stem cell and proliferation gene signatures, and subsets of cells that have lost self-renewal potential through neural differentiation 38, 39. The approach used in these studies is powerful because it allows reconstruction of cellular hierarchies from genome-wide expression signatures in primary tumors, avoiding the caveats associated with xenograft assays. However, one limitation of these studies is their descriptive nature, which cannot assign causality to the identified gene signatures in the absence of functional validation.

The observation that IDH-induced oligodendrogliomas contain differentiated, non-self-renewing cells underscores the diverse, and at times antithetic, role of epigenetic mechanisms in cancer development. During tumorigenesis, mutations in IDH proteins drive CSC emergence partly through alteration of DNA methylation profiles and epigenetic reprogramming of committed oligodendrocytes. However, during tumor growth, additional epigenetic changes occur and establish developmental hierarchies that restrict the proliferative potential of some cells, counteracting the effect of the initiating mutations (Figure 2).

Open in a separate window

Figure 2

Epigenetic Mechanisms Integrate Cell-Intrinsic and Cell-Extrinsic Changes Affecting Cancer Cells and Generate Functional Intratumoral Heterogeneity. Schematic depiction of the distinct layers of alterations that affect cells during cancer development via epigenetic mechanisms. Initiating mutations (pink dash inside the nuclei) affect the cell epigenome (cylinders inside the cell nuclei) either directly, when mutations hit epigenetic regulators, or indirectly, when mutations in other drivers trigger gene expression changes mediated by chromatin remodeling and DNA methylation [100]. In either case, epigenetic reprogramming translates mutations into malignant phenotypes and promotes the acquisition of uncontrolled self-renewal. Secondary mutations occurring during tumor growth (red and black dashes) and signals from tumor microenvironment (pink, red, and black dots outside cells) induce further changes in the epigenome of the cells, either enhancing (red) or inhibiting (black) cancer cell self-renewal in a subclone- and context-dependent manner. The phenotype of each cell within a tumor is the result of all these alterations, which collectively shape the epigenome of the cell and determine which cells drive cancer growth. In the combined layers, the cell with a pink nucleus and a thin arrow represents a cell that has maintained the self-renewal ability conferred by the initiating events. Cells with red nuclei and thick arrows represent cells with enhanced self-renewal ability due to either favorable secondary mutations (cell on the right) or signaling (cell on the left). Cells with grey nuclei represent cells that have lost self-renewal ability due to either deleterious secondary mutations or signaling.

Epigenetic Regulators That Inhibit Cancer Cell Self-Renewal and Establish Cellular Hierarchies

What are the mechanisms that generate distinct epigenetic states within tumors and confer distinct functional properties to CSCs and differentiated cells? Epigenetic regulators belonging to two distinct groups have so far been identified: those that inhibit cancer cell self-renewal and establish differentiation hierarchies, and those that are hijacked by CSCs to avoid differentiation and sustain their phenotype (Table 1). Within the first group, the linker histone H1.0 plays an important role in restricting cancer cells long-term proliferative potential and determines functionally distinct subsets of cells within individual tumors [40]. In numerous cancer types, H1.0 levels are highly heterogeneous, with low levels in cells expressing CSC markers and high levels in differentiated cells. Functional characterization of the impact of H1.0 expression on cancer cells revealed that only cells able to stably repress H1.0 preserve a chromatin configuration compatible with self-renewal capacity. Re-expression of H1.0 in subsets of cells during tumor growth induces genome-wide silencing of oncogenic and self-renewal genes, and promotes differentiation into non-tumorigenic cells [40]. Thus, only cells expressing low H1.0 levels can maintain CSC properties. As discussed above for mutations in H3.3, the observation that an integral component of chromatin acts as an important regulator of cancer cell differentiation states highlights the important role of epigenetics in specifying tumor organization and affecting tumor maintenance. Another example supporting this concept comes from MLL-driven AML, in which high levels of the histone demethylase KDM5B revert the histone modification patterns and gene expression programs established by MLL during leukemogenesis, thereby extinguishing LCS potential [41]. Similarly, in glioma the histone methyltransferase G9a and the related global increase in H3K9me2 inhibit self-renewal of CSCs in vitro [42]. Although the presence of cellular hierarchies is recognized in many cancers, our knowledge of how they are established and how cancer cells lose self-renewal capacity during tumor growth is still rudimental. A comprehensive identification of the molecular players that drive cancer cell differentiation is central to understanding, and possibly exploiting, the mechanisms that naturally inhibit tumor maintenance and deprive cancer cells of their malignant properties.

Epigenetic Regulators That Sustain Cancer Cell Self-Renewal

Although many cancer cells succumb to differentiation during tumor growth, CSCs evade this process and preserve their self-renewal capacity acquired during transformation. Not surprisingly, considering the importance of epigenetic regulators in normal stem cell maintenance, many chromatin-related proteins and DNA-methylating enzymes are essential to maintaining the CSC state. Remarkably, most of the proteins implicated in CSC maintenance are typically not mutated but are co-opted in their wild-type state by CSCs to avoid differentiation and sustain their malignant properties. Prominent examples of hijacked proteins are Polycomb complex proteins. The PRC2 catalytic subunit EZH2 has a tumor-promoting role in many malignancies, and inhibition of its activity or genetically induced loss of the protein strongly impairs tumor growth 43, 44, 45, 46, 47, 48, 49, 50. Several molecular mechanisms, not always dependent on EZH2 methyltransferase activity, have been shown to underlie the role of EZH2 in cancer, including repression of tumor suppressors 43, 44, 45, 46, 47, 48, 49, 50, activation of oncogenic NOTCH signaling [47], stabilization of β-catenin [48], and inhibition of DNA damage repair and consequent induction of secondary mutations [50]. Regardless of the specific mechanism, the commonality in all these cases is that cells become dependent on EZH2 to preserve their self-renewal potential, as a consequence of the cellular changes induced by transformation and other alterations occurring during tumor growth. Other PcG proteins such as a non-canonical PRC1.1 complex in AML [51] and BMI1 in glioma [52] exert similar functions. Histone modifiers in general have often been reported as positive regulators of CSC self-renewal (Table 1) 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64. Furthermore, the chromatin-remodeling complex SWI/SNF sustains high levels of c-MYC by regulating enhancer function and is required for maintenance of self-renewing LSCs in MLL-driven leukemia [65]. Interestingly, in many cases, functional dependency on chromatin regulators is observed in neoplasms initiated by mutations targeting other epigenetic regulators, highlighting the complex role of epigenetics in cancer and the existence of synthetic lethality interactions that could be exploited for therapeutic purposes 44, 45, 46, 53, 57, 64.

Interplay between Extracellular Signaling and Epigenetics in Generating Intratumoral Functional Heterogeneity

Variations in oxygen and nutrient concentrations are a likely source of phenotypic variation within tumors, and histone modifiers have been reported to act as sensors of hypoxia [83] or are linked to metabolism [84]. Similarly, DNA methylation patterns are affected by changes in the environment [85]. Thus, cells exposed to distinct conditions in their local environment may respond by modifying their epigenome. Furthermore, classic cell-to-cell signaling pathways, such as WNT, TGF-β, SHH, and NOTCH, which tightly regulate self-renewal and differentiation during embryonic development and adult tissue homeostasis, affect CSC self-renewal either positively or negatively [86]. In physiological conditions, the heritable switch from division to differentiation induced or prevented by these pathways is determined by changes in chromatin and DNA methylation [87]. Although the downstream effectors of aberrant developmental pathways in cancer are largely unknown, it stands to reason that epigenetic factors are likely to be crucial mediators that translate extracellular signaling into differential phenotypes within cancers. In line with this view, functional interactions between TGF-β and two histone modifiers, PRMT5 [88] and G9a [89], modulate cancer cell phenotypes in various types of carcinoma through changes in histone methylation. Moreover, the transcriptional regulator CDK9 has been found to act as a downstream effector of the NOTCH pathway to sustain GBM stem cells [90]. In addition, cAMP, a second messenger involved in transduction of various extracellular signals, promotes breast CSC differentiation via the histone demethylase PHF2 [91]. Because chromatin-based mechanisms represent the last level of signal transduction cascades, and several ‘epigenetic’ drugs have already been developed, finding novel mechanisms where chromatin-related factors mediate CSC-sustaining signaling pathway activity could have important therapeutic implications (Figure 3, Key Figure). As an additional layer of complexity, distinct epigenetic states can in turn determine the response to microenvironment signaling, as exemplified by breast and glioma CSCs in which the tumor-suppressive or tumor-promoting effect of paracrine TGF-β signaling is determined by the DNA methylation status of SMAD target genes 92, 93. Thus, the interplay between signaling and the epigenome has emerged as a crucial force that shapes tumor architecture, and future characterization of these complex interactions will certainly provide invaluable insights into CSC biology (Figure 2).

Open in a separate window

Figure 3

Key Figure: The Complex Role of Epigenetic Mechanisms in Cancer

The diagram summarizes mechanisms that affect cellular plasticity in cancer through changes in chromatin and DNA methylation. Mechanisms operating during carcinogenesis unleash cellular plasticity. During tumor growth, cellular plasticity is exploited by cancer cells (hijacking) and is further modulated either positively, to favor adaptation of cells and cancer growth (enhancing), or negatively (restricting). Blue, red/purple, and grey cells represent normal cells, self-renewing, and differentiated cancer cells, respectively. Double arrows indicate reversible transitions between cell states. Mechanisms involving indirect reprogramming of the epigenome by oncogenic hits or environmental cues are not discussed in this review because they have been described elsewhere 100, 101. Abbreviation: CSC, cancer stem cell.

Reversible Epigenetic States and Cancer Cell Plasticity

A major difference between normal cellular hierarchies and cancer is that, although cell fate decisions triggered by environmental cues are generally stable and heritable in normal cells, cancer cells maintain an intrinsic plasticity that allows them to easily change their phenotype in response to new signals and possibly switch between cellular states. Evidence exists in some cancers that differentiated cells can reacquire self-renewal ability and revert to a CSC state (Figure 3). This occurs, for example, in basal carcinomas of the breast, in which TGF-β stimulation induces conversion of non-tumorigenic CD44^lo cells to CD44^hi CSCs. Importantly, this cellular plasticity appears to be dependent on the chromatin status of the ZEB1 promoter because a poised, bivalent configuration allows reversion to CD44^hi CSCs upon stimulation, while the presence of repressive marks renders CD44^lo cells insensitive to TGF-β [94]. Another example highlighting how chromatin states can influence cancer cell plasticity comes from melanoma, a type of cancer that does not appear to develop in a hierarchical manner [95]. Even in the absence of ‘classical' CSCs, Roesch et al. identified a subpopulation of cells that is required for continuous tumor growth and is marked by the expression of the histone demethylase JARID1B. However, instead of being a stable subpopulation, JARID1B⁺ cells are a dynamic subset whose composition changes over time as cells gain and lose JARID1B expression, and transiently acquire stemness properties depending on the tumor context [96]. These observations support a model of transient stemness, sustained by dynamic epigenetic states, in which at any given time a self-renewing but changing subpopulation of cells exists among the bulk of tumor cells. An important implication of these findings, which can be extended to classical hierarchically organized tumors, is that the local tumor context appears to be key in determining, through epigenetics, which cells within a tumor can act as tumor-maintaining CSCs, arguing against a deterministic model of CSC identity.

Work in the laboratory of P.S. is supported by the Francis Crick Institute, which receives its core funding from Cancer Research UK (FC001152), the UK Medical Research Council (FC001152), and the Wellcome Trust (FC001152).