Epigenetic Cancer Therapy: Rationales, Targets and Drugs

Home , Cytidine deaminase, Decitabine, Gemcitabine

M Rius and F Lyko

Division of Epigenetics, DKFZ-ZMBH Alliance, German Cancer Research Center, Heidelberg, Germany

The fundamental role of altered epigenetic modification many tumors show an aberrant epigenetic modification patterns in tumorigenesis establishes epigenetic regulatory pattern. These tumor-specific epigenetic alterations or enzymes as important targets for cancer therapy. Over the epigenetic mutations, affect a variety of cancer-related past few years, several drugs with an epigenetic activity genes. The most prominent example for epimutations is have received approval for the treatment of cancer provided by the widely observed epigenetic silencing of patients, which has led to a detailed characterization of tumor suppressor genes that is closely associated with their modes of action. The results showed that both promoter hypermethylation (Esteller, 2007; Jones and established drug classes, the histone deacetylase (HDAC) Baylin, 2007). It has long been known that DNA inhibitors and the DNA methyltransferase inhibitors, show methyltransferase inhibitors can effectively induce DNA substantial limitations in their epigenetic specificity. HDAC demethylation and phenotypic changes related to the inhibitors are highly specific drugs, but the enzymes have a reactivation of epigenetically silenced genes (Jones and broad substrate specificity and deacetylate numerous Taylor, 1980). These findings were later adapted to the proteins that are not associated with epigenetic regulation. targeting of epigenetic mutations in cancer and thus Similarly, the induction of global DNA demethylation by established the fundamental concept of epigenetic non-specific inhibition of DNA methyltransferases shows cancer therapy (Egger et al., 2004). pleiotropic effects on epigenetic regulation with no apparent Research over the past decade has shown that tumor-specificity. Second-generation azanucleoside drugs epigenetic regulation occurs on several mechanistic have integrated the knowledge about the cellular uptake and levels, including DNA methylation, covalent histone metabolization pathways, but do not show any increased modifications and small regulatory RNAs (Esteller, specificity for cancer epigenotypes. As such, the traditional 2007; Jones and Baylin, 2007). Thus, in addition to rationale of epigenetic cancer therapy appears to be in need DNA methyltransferase (DNMT) inhibitors, histone of refinement, as we move from the global inhibition of deacetylase (HDAC) inhibitors have also contributed to epigenetic modifications toward the identification and shape epigenetic cancer therapy. Altogether, there are targeting of tumor-specific epigenetic programs. Recent currently four FDA-approved drugs with an epigenetic studies have identified epigenetic mechanisms that promote mode of action: the DNA methyltransferase inhibitors self-renewal and developmental plasticity in cancer cells. 5-azacytidine (Vidaza) and decitabine (20-deoxy-5-aza- Druggable somatic mutations in the corresponding epige- cytidine, Dacogen) and the HDAC inhibitors suberoy- netic regulators are beginning to be identified and should lanilide hydroxamic acid (SAHA, Zolinza) and facilitate the development of epigenetic therapy approaches romidepsin (Istodax). Various additional DNMT and with improved tumor specificity. HDAC inhibitors are currently being evaluated in Oncogene (2012) 31, 4257–4265; doi:10.1038/onc.2011.601; preclinical studies and in clinical trials. However, there published online 19 December 2011 are currently no clinical trials with drug candidates directed against other epigenetic targets. This situation Keywords: DNA methylation; chromatin; DNA methyl- warrants a detailed discussion of DNMT and HDAC transferase inhibitors; histone deacetylase inhibitors; inhibitors before novel approaches towards an epigenetic metabolic activation; stem cells therapy of cancer are considered. 5-Azacytidine and decitabine have shown significant clinical benefit in the treatment of myelodysplastic syndrome (Silverman et al., 2002; Kantarjian et al., 2006) and are also used for the treatment of myeloid Introduction leukemias. Both drugs are particularly effective when administered at low doses, which are sometimes below The traditional rationale for the development of 10% of their maximum tolerated doses (Issa and epigenetic cancer drugs is based on the observation that Kantarjian, 2009). These conditions are generally assumed to maximize DNA demethylation responses, and several studies have confirmed that drug-induced Correspondence: Dr F Lyko, Deutsches Krebsforschungszentrum, Im DNA demethylation occurs in patients (Issa et al., 2005; Neuenheimer Feld 580, 69120 Heidelberg, Germany. Mund et al., 2005; Soriano et al., 2007; Stresemann and E-mail: [email protected] Received 14 September 2011; revised and accepted 21 November 2011; Lyko, 2008). However, a direct link between DNA published online 19 December 2011 demethylation and clinical responses has not been Epigenetic cancer therapy M Rius and F Lyko 4258 demonstrated yet and there are currently no established required the presence of HR23B. These data suggest DNA methylation biomarkers that accurately predict that HDAC inhibitors target aberrant proteasome patient responses (Fandy et al., 2009). This remains as activity in cancer cells, which is possibly mediated by an important challenge that is related to the complex hyperacetylation of proteasome-associated non-histone mode of action of azanucleosides (see below) and that proteins. will have to be addressed in future studies to firmly Moreover, when yeast cells were incubated in the establish clinical proof-of-concept for epigenetic therapy presence of the HDAC inhibitor valproic acid, the drug with these drugs. specifically counteracted the activation of the ATR- The HDAC inhibitors SAHA and romidepsin are mediated DNA damage checkpoint response (Robert approved for the treatment of cutaneous T-cell lympho- et al., 2011). Surprisingly, the effect on DNA damage ma (Olsen et al., 2007; Piekarz et al., 2009). Numerous repair involved the degradation of key recombination additional HDAC inhibitors are currently undergoing proteins by autophagy (Robert et al., 2011). These clinical testing. The high number of developmental observations were explained by a model where protein drugs is related to the excellent druggability of the acetylation marks DNA recombination proteins for HDAC enzyme family, which permits the development transport to the vacuole and subsequent degradation, of potent isoenzyme-specific inhibitors (Bolden et al., and where deacetylation of these factors is required for 2006). It is important to notice that histone deacetyla- their stabilization in subnuclear compartments with tion has been shown to synergistically interact with severely damaged DNA. In this model, HDAC inhibi- DNA methylation in the epigenetic silencing of cancer tors would affect major oncological pathways (DNA genes (Cameron et al., 1999). This observation has also damage response and autophagy) without invoking any provided the scientific rationale for various clinical trials effects on epigenetic gene regulation. Altogether, these with combinations of DNMT and HDAC inhibitors. findings further illustrate why HDAC inhibitors should Somewhat surprisingly, these studies have yet to reveal not be considered as specific epigenetic drugs. It is synergistic clinical effects, which might again be related possible (although not proven) that targeting of other to complexities in the drugs’ modes of action that are histone modifying enzymes, like histone methyltrans- only beginning to become elucidated (see below). ferases and histone demethylases, might permit a higher degree of epigenetic specificity (see below).

Are HDAC inhibitors epigenetic drugs? Epigenetic cancer therapy with DNMT inhibitors The important role of HDAC inhibitors in current concepts of epigenetic cancer therapy is often inter- The most frequently used epigenetic drugs are the preted to reflect a high specificity of HDACs for histone cytosine derivatives 5-azacytidine and decitabine. Both substrates. However, similar to the situation with DNA drugs are undergoing a series of metabolic activation methyltransferase inhibitors, a direct link between drug- steps before they can become incorporated into nucleic induced histone hyperacetylation and clinical responses acids (Figure 1). Incorporation of azacytosine bases into remains to be established. It is now becoming increas- DNA and RNA triggers a multitude of stress and ingly clear that the enzymatic activity of HDACs is not damage signaling pathways (Hollenbach et al., 2010), restricted to histone proteins, and treatment of human but also induces the formation of covalent adducts cancer cell lines with highly specific HDAC inhibitors between DNA and DNA methyltransferases. This can induce hyperacetylation of 1750 proteins (Choudh- covalent trapping results in the depletion of the enzymes ary et al., 2009). This finding strongly suggests that non- and thus causes a global reduction of DNA methylation histone proteins constitute the large majority of HDAC (Stresemann and Lyko, 2008). A global reduction of substrates. In addition, these studies also suggest that DNA methylation has been shown to have antitumoral the complexity and functional significance of the human effects in a mouse model for intestinal tumorigenesis protein acetylome are similar to the more established (Laird et al., 1995). In this study, a combination of protein phosphorylome. It is conceivable that these hypomorphic alleles for the DNA methyltransferase 1 modification signatures have combinatorial roles in gene caused a significant reduction in the size and regulating the biological activity of proteins (Sims and number of intestinal polyps. To date, these observations Reinberg, 2008). Drug-induced alterations of non- still represent the most important proof-of-concept data histone protein modification signatures could therefore for the antitumoral effects of global DNA demethylation. have a major role in the pharmacological responses to It is generally assumed that the therapeutic benefit of HDAC inhibitors (Xu et al., 2007). global DNA demethylation is related to the demethyla- A prominent example for this notion is provided by tion and reactivation of aberrantly silenced genes. the identification of HR23B, a protein that is involved in However, global demethylation is not selective for DNA repair and in proteasomal targeting of ubiquiti- cancer-specific methylation marks and epigenetic side nylated substrates, and that has a major role in effects could severely curtail the specificity of demethy- regulating the sensitivity of cancer cells to HDAC lation therapy (see below). It is likely that individual inhibitors (Fotheringham et al., 2009). Efficient induc- DNA methyltransferases, like DNMT3B, have specific tion of apoptosis by HDAC inhibitors was closely roles in the establishment and/or maintenance of cancer- associated with reduced proteasomal activity and specific DNA methylation changes (Linhart et al., 2007).

Oncogene Epigenetic cancer therapy M Rius and F Lyko 4259 clinical candidates are likely to emerge. Thus, additional screening efforts will be required for the identification of small-molecule scaffolds with a high potency in the inhibition of specific DNMT enzymes. Because of the inherently low specificity of azanucleoside drugs, enzyme-specific DNMT inhibitors will also be important for unambiguously establishing clinical proof-of-concept for demethylation therapies. It is important to realize that a causal link between drug- induced demethylation and therapeutic responses to 5-azacytidine or decitabine has not been established yet. A clinical relevance of the epigenetic effects is currently mostly supported by characteristic details in the progression of clinical responses in cancer patients: for example, both drugs trigger delayed hematologic responses. This contrasts the responses observed with cytarabine, a chemically related drug that has no known epigenetic activity and induces immediate cytotoxic responses (Issa and Kantarjian, 2009). Altogether, these findings suggest that nucleoside DNMT inhibitors have genuine epigenetic activity in patients, but important details remain to be established.

Metabolic activation and cellular transport of DNMT inhibitors

The continued clinical development of azanucleosides requires detailed knowledge of the drugs’ cellular metabolism. Both drugs utilize different pathways for their intracellular metabolization, as the azacytosine ring is bound to ribose in 5-azacytidine and to deoxyribose in decitabine. In this context, three enzymes appear to be particularly relevant: (1) deoxycytidine kinase (dCyd kinase), which catalyzes the initial rate- limiting step in the synthesis of the monophosphorylated Figure 1 Cellular pathways for the metabolic activation of form of decitabine; (2) uridine-cytidine kinase (Urd-Cyd DNMT inhibitors. Members of the hENT, hCNT, the proton oligopeptide cotransporter familiy (SLC15) and the SLC22 family kinase), which is assumed to be responsible for the of organic cation and anion transporters are known to mediate the monophosphorylation of 5-azacytidine; and (3) ribonu- uptake of nucleoside analogs. Inside the cells, 5-azacytidine (5-aza- cleotide reductase, which reduces 5-azacytidine dipho- CR) and decitabine (5-aza-20-deoxycytidine; 5-aza-dCR) are sphate to 5-azadeoxycytidine diphosphate (Figure 1). metabolized to nucleotides by different enzymes before incorpora- Deficiencies in dCyd kinase activity have long been tion into DNA and RNA. Metabolic intermediates can also be effluxed from the cells by members of the ABC family. Several linked to decitabine resistance. This was confirmed by enzymes are a potential source of drug resistance and are marked the selection and characterization of decitabine-resistant in red. rat leukemic cell lines that revealed mutations in the dCyd kinase gene (Stegmann et al., 1995). Similar results were obtained in more recent studies with human cancer Thus, the availability of enzyme-specific DNMT in- cells, where decitabine sensitivity could be restored by hibitors would facilitate the targeting of cancer-specific transfection of the wild-type dCyd kinase (Qin et al., epigenetic alterations. Small molecules that selectively 2009). dCyd kinase has also been used to predict in vivo inhibit individual DNMT enzymes in biochemical assays sensitivity and resistance to other cytidine analogs, such have been described recently (Kuck et al., 2010) and the as gemcitabine and cytarabine, in patients (Flasshove development of enzyme-specific inhibitors thus appears et al., 1994; Kroep et al., 2002). Taken together, these feasible. However, even though we have seen a rapid results clearly imply dCyd kinase as a key enzyme in the increase in the number of candidate DNMT inhibitors metabolic activation of decitabine and a potentially over the past few years, the epigenetic potency of 5- important factor in clinical decitabine resistance. azacytidine and decitabine presently remains unmatched As outlined above, 5-azacytdine is metabolized by a (Chuang et al., 2005; Stresemann et al., 2006; Medina- distinct biochemical pathway, and it is assumed that the Franco et al., 2011). Moreover, the known small- monophosphorylation step is catalyzed by Urd-Cyd molecule DNMT inhibitors would require a substantial kinase. This assumption is based on the observation that amount of preclinical development before competitive both uridine and cytidine inhibit the phosphorylation

Oncogene Epigenetic cancer therapy M Rius and F Lyko 4260 and nucleic acid incorporation of 5-azacytidine with a epithelial tissues. Gene expression analyses that demon- consequent reduction of its toxicity in leukemia cell lines strated a positive correlation between hENT1 expression (Li et al., 1970). However, Urd-Cyd kinase failed to and 5-azacytidine potency in human cancer cell lines phosphorylate 5-azacytidine in biochemical assays (Van provided a first indication for an important role of Rompay et al., 2001), and the enzyme has not been hENTs in 5-azacytidine uptake (Huang et al., 2004). demonstrated to be rate-limiting in the metabolism of 5- Furthermore, low expression levels of hENT1 and azacytidine. As such, additional work will be required to hENT2 were detected in cells resistant to decitabine experimentally confirm the role of Urd-Cyd kinase in (Qin et al., 2009). As hENT transporters show ubiqui- the phosphorylation of 5-azacytidine and to further tous expression, they can have a major role in the characterize the biochemical pathway(s) for the meta- sensitivity and resistance of cancer cells to azanucleoside bolic activation of 5-azacytidine. DNMT inhibitors. The ability of hCNT transporters to Ribonucleotide reductase is another critical enzyme mediate cellular uptake of 5-azacytidine and decitabine and represents an important link between the metabo- (Rius et al., 2009) probably becomes more relevant for lism of 5-azacytidine and decitabine. This enzyme the treatment of epithelial cell-derived tumors, where the enables the metabolites of 5-azacytidine to become expression levels of hCNT proteins can be higher. incorporated into DNA and is thus rate limiting for the In addition to the hCNT and hENT proteins, several synthesis of deoxyribonucleotides from ribonucleotides members of the ABC superfamily and particularly the (Figure 1). Importantly, the enzyme only reduces a members of the ABCC subfamily have been described to fraction of 5-azacytidine metabolites to deoxycytidine mediate the cellular efflux of nucleoside drug metabo- metabolites, whereas a major part is phosphorylated and lites (Cole and Deeley, 2006). However, little is known incorporated into RNA (Li et al., 1970). This fractional about specific transporters that mediate the efflux of conversion of 5-azacytidine to deoxycytidine metabo- 5-azacytidine and decitabine metabolites from cells. At lites has also been used to explain the differential effects present, only ABCC4 has been identified as an efflux of 5-azacytidine and decitabine in human cancer cell pump for 5-azacytidine (Rius et al., 2010). Further lines (Hollenbach et al., 2010). For example, decitabine studies are required because the ABC family represents a often causes DNA hypomethylation at lower concentra- well-established factor in cancer drug resistance (Leo- tions, whereas 5-azacytidine has greater effects on nard et al., 2003). reducing cell viability and protein synthesis (Hollenbach Lastly, the efficacy of 5-azacytidine and decitabine is et al., 2010; Hagemann et al., 2011). Further insight also influenced by their plasma stability. In this context, could come from the combined use of azanucleosides cytidine deaminase (CDA), an enzyme involved in with ribonucleotide reductase inhibitors. Initial studies pyrimidine salvaging, has a major role. The enzyme have suggested that DNA demethylation by both catalyzes the hydrolytic deamination of (deoxy-)cytidine 5-azacytidine and decitabine could be inhibited by the to (deoxy-)uridine and is also able to deaminate 5- ribonucleotide reductase inhibitor hydroxycarbamide azacytidine and decitabine to 5-azauridine and 5-aza-20- (Choi et al., 2007), and it will be important to investigate deoxyuridine. Cytidine deaminase is ubiquitous in the the effect of drug treatment on the nucleotide pool and human body and represents an important determinant on the resulting effects on cell cycle regulation. for the biotransformation rates of cytarabine and Apart from their metabolic activation, the cellular gemcitabine, two cytidine analogs that are closely transport across the plasma membrane represents an related to azanucleosides (Bhatla et al., 2009; Maring additional, important factor in the response of cancer et al., 2010; Parmar et al., 2011). cells to azanucleoside drugs. In particular, reduced The increasing knowledge of the mechanisms mediat- uptake into the cells or increased efflux from the cells ing cellular drug uptake and metabolization has also results in reduced intracellular drug levels and, conse- fueled the development of improved azanucleoside quently, in the development of drug resistances. Several derivatives. This is exemplified by S110 and CP-4200, families of membrane transporters have been recognized two second-generation azanucleoside drugs (Figure 2). to have a role in the transport of nucleosides and S110 is a dinucleotide of decitabine and deoxyguanidine, nucleoside analogs across the plasma membrane in and is highly resistant to cytidine deaminase (Yoo et al., human cells (Pastor-Anglada et al., 2005; Damaraju 2007). CP-4200 is an elaidic acid ester of 5-azacytidine, et al., 2009; Huber-Ruano and Pastor-Anglada, 2009). which utilizes a modified and potentially less selective These include the SLC28 and SLC29 families encoding transport uptake mechanism (Brueckner et al., 2010). the concentrative and equilibrative nucleoside transpor- Importantly, both drugs have retained robust demethy- ters, respectively (hCNT/SLC28 and hENT/SLC29) and lating activity, despite their extensive chemical modifica- the ATP-binding cassette (ABC) family (Figure 1). tions (Yoo et al., 2007; Brueckner et al., 2010). In The hCNT and hENT proteins represent the best- addition, CP-4200 also causes widespread DNA known nucleoside transporters. While hCNTs mediate demethylation and reactivation of tumor suppressor the uptake of nucleosides against a concentration genes under conditions where inhibition of hENT- gradient and in a Na þ -dependent manner, hENTs dependent drug transport completely blocked the facilitate the transport of nucleosides in a bidirectional demethylating effects of 5-azacytidine (MR and FL, manner (Huber-Ruano and Pastor-Anglada, 2009). unpublished data). Both CP-4200 and S110 have shown Compared with hENTS, hCNTs have a higher affinity significant in vivo activity in mouse tumor models for the substrates, but are mostly expressed in specialized (Brueckner et al., 2010; Chuang et al., 2010), which

Oncogene Epigenetic cancer therapy M Rius and F Lyko 4261 shown that the demethylation of a highly methylated LINE-1 antisense promoter in an intron of the cMet proto-oncogene results in the induction of an illegitimate cMet transcript, which is associated with reduced cMet protein expression and decreased cMet receptor signaling. These findings provide an important example for abnormal transcriptional regulation induced by the demethylation of repetitive elements and suggest that many genomic loci might be subject to the epigenetic side effects of global DNA demethylation (Lizardi, 2010). Recent epigenomic studies have also shown that epigenetic gene silencing by DNA methylation and collaborating epigenetic mechanisms can be observed at numerous developmental genes and in normal cells (Mohn and Schubeler, 2009). For example, differentia- Figure 2 Second-generation DNMT inhibitors. A second-generation-associated genes are methylated in mouse and tion of nucleoside DNMT inhibitors has been developed recently to increase drug stability and to overcome transport-related drug human hematopoietic progenitor cells and become resistance. The dinucleotide derivative of decitabine, S110, is demethylated during differentiation (Ji et al., 2010; resistant to the hydrolytic cleavage by the plasma cytidine Bocker et al., 2011). A functional role of DNA deaminase, in contrast to the parental drug decitabine, which is methylation in the differentiation of adult progenitor rapidly deaminated in blood. The elaidic acid ester of cells is further supported by the observation that 5-azacytidine, CP-4200, enters into cells through a hENT- independent mechanism, thus overcoming transporter-related drug hematopoietic stem cells from mice with reduced DNA resistance. Chemical structures used for the modification of the methyltransferase 1 activity failed to suppress key azacytosine scaffold are marked in blue. differentiation genes and lost their ability to differentiate into lymphoid progeny (Broske et al., 2009). While the further illustrates the potential of the azacytosine same study also revealed a therapeutic potential of scaffold for continued preclinical development. Impor- global DNA hypomethylation for restricting the self- tantly, the development of chemically modified azacy- renewal capacity of leukemia stem cells, the data clearly tosine derivatives might also allow the targeting of demonstrates that DNA methylation is required to tumor entities that are insensitive to 5-azacytidine and protect normal hematopoietic stem cells from lineage decitabine. restriction. In a more general sense, these studies illustrate that tightly controlled DNA demethylation is an important Epigenetic side effects of global DNA demethylation factor in cellular differentiation (Figure 3). The findings imply that epigenetic drugs that induce global DNA Apart from the hypermethylation of cancer genes, demethylation and/or the global loss of collaborating human cancer cells are also characterized by an overall histone modifications could adversely affect the func- DNA hypomethylation (Ehrlich, 2002). Genomic hypo- tionality of adult progenitor cell populations in patients. methylation has long been associated with genomic Moreover, numerous large blocks with significant instability and could thus represent a tumor-promoting hypomethylation were described in colon cancer methy- epigenetic alteration. As the established DNA methyl- lomes (Hansen et al., 2011). These blocks showed transferase inhibitors show no specificity for tumor increased gene expression variation, which suggested suppressor genes but rather affect epigenetic modifica- that a further decrease in DNA methylation beyond the tions globally, their ability to induce genomic instability normal levels of differentiated cells might induce remained an important safety issue. Indeed, mouse phenotypic variability (Figure 3). Drugs that induce models with reduced levels of Dnmt1 and correspond- global DNA hypomethylation could thus promote ingly reduced levels of genome-wide DNA methylation oncogenic transformation in cancer patients (Hansen showed a high incidence of lymphomas, which was et al., 2011), and it appears critical that more specific accompanied by genetic instability (Eden et al., 2003; epigenetic drugs become identified and developed. This Gaudet et al., 2003). Further analyses in this model might not be achievable through the inhibition of DNA suggested that DNA hypomethylation causes the methyltransferases and would therefore require the activation and transposition of endogenous retroviral targeting of other epigenetic factors and pathways. elements (Howard et al., 2008). This finding also implied that the activation of (normally silent) repetitive elements might be a particularly important factor in A new rationale for epigenetic cancer therapy: targeting the cellular response to global DNA demethylation. the epigenetic plasticity of cancer cells More recently, the consequences of global DNA demethylation on the epigenetic regulation of repetitive When the traditional rationale for epigenetic cancer elements were further characterized in human cancer therapy was first developed, it was assumed that cells (Weber et al., 2010; Wolff et al., 2010). It was hypermethylation-associated gene silencing is only

Oncogene Epigenetic cancer therapy M Rius and F Lyko 4262 nic stem cells and the relationship between H3K27 trimethylation, and de novo DNA methylation is assumed to reflect the presence of a stem cell-like epigenetic program in cancer cells (Ohm and Baylin, 2007). In agreement with this notion, a recent analysis also identified bivalent chromatin structures in human tumor samples (Aiden et al., 2010). An attractive strategy towards the development of more specific epigenetic cancer therapies thus resides in the targeting of the enzymes that instruct cancer-specific de novo methylation and/or cancer stem cell phenotypes. An intriguing candidate is the histone H3 lysine 27 methyltransferase EZH2. This enzyme, together with its closely related paralog EZH1, is critical for stem cell identity and for the establishment of H3K27 methyla- Figure 3 DNA demethylation in cellular differentiation and tion marks within bivalent chromatin structures of oncogenic transformation. In adult progenitor cells, DNA methy- embryonic stem cells (Shen et al., 2008). Several studies lation is found in differentiation-associated genes, preventing cells have described overexpression of EZH2 in human from differentiation. Demethylation causes the activation and expression of differentiation genes. Further demethylation has been cancers, including prostate and breast cancer, as well associated with increasing gene expression variability and increased as in lymphoma (Varambally et al., 2002; Simon and phenotypic variability of cancer cells (Hansen et al., 2011). The use Lange, 2008). Importantly, while cancer genome sequen- of hypomethylating drugs in cancer patients may thus interfere cing approaches detected inactivating EZH2 mutations with normal epigenetic regulation in adult progenitor cells and/or in myeloid leukemias (Ernst et al., 2010; Nikoloski et al., induce a tumor-promoting epigenetic program. 2010), they also uncovered specific heterozygous point mutations of EZH2 in lymphomas (Morin et al., 2010). observed in the context of epigenetic mutations. How- These point mutations cause an enhanced capacity of ever, it is now clear that hypermethylation-associated EZH2 to di and trimethylate H3K27 (Sneeringer et al., gene silencing also has an important role in the 2010) and could thus drive the establishment of cancer- regulation of developmental genes and therefore does specific epigenetic programs. Intriguingly, the mutant not represent a cancer-specific epigenetic alteration. EZH2 enzyme appears to be druggable, which should Consequently, the traditional rationale for epigenetic allow the development of potent inhibitors with a high cancer therapy requires further development to provide degree of cancer specificity for the treatment of patients sufficient tumor selectivity for therapeutic success. that carry the corresponding mutation. A higher degree of specificity could be achieved if Additional cancer-specific epigenetic drug targets epigenetic cancer therapy could be directed against should be identified through ongoing studies that cancer-specific epigenetic modification patterns. While investigate the mechanisms underlying the phenotypic many details remain to be resolved, several studies have plasticity of cancer cells. These approaches will identify provided evidence for cancer-specific epigenetic path- the factors that establish and maintain the embryonic ways. For example, chromosomal translocations affect- chromatin environment conductive to self-renewal and ing the MLL gene cause mixed lineage leukemia that is developmental plasticity in cancer cells. Successful characterized by highly defined gene expression changes targeting of the corresponding enzymes would result in associated with aberrant H3K79 methylation (Bernt reduced proliferation and reduced developmental plas- et al., 2011). In agreement with this notion, a potent ticity, thus limiting both tumor growth and the self- small-molecule inhibitor of the H3K79 methyltransfer- renewal capacity of tumor stem cells (Figure 4). Apart ase DOT1L selectively killed cancer cells that harbored from EZH2/EZH1, candidates should include the as yet the MLL gene translocation (Daigle et al., 2011); thus unknown enzyme(s) required for H3K4 trimethylation providing a compelling paradigm for epigenetic therapy in stem cells. The TET enzymes represent additional with a very high degree of cancer specificity. factors that are associated with epigenetic regulation It has also been shown that genes affected by de novo and cell fate determining mechanisms. These enzymes DNA methylation during tumorigenesis are pre-marked catalyze the hydroxylation and further modification of by histone H3 lysine 27 trimethylation, suggesting methylated DNA and probably have an important role that this chromatin modification functions as an in the demethylation of DNA (Ito et al., 2011). TET instructive mechanism in the establishment of cancer- proteins are associated with bivalent chromatin domains specific DNA methylation patterns (Ohm et al., 2007; and have an important role in targeting EZH2 to Schlesinger et al., 2007; Widschwendter et al., 2007). Of chromatin (Williams et al., 2011; Wu et al., 2011). The note, H3K27 trimethylation is also an important wealth of data generated by cancer genome and component of ‘bivalent’ chromatin structures that are epigenome-sequencing projects is very likely to identify defined by the presence of both activating H3K4 additional somatic mutations in epigenetic regulators trimethylation and repressive H3K27 trimethylation that have important roles in the establishment and (Bernstein et al., 2006). Bivalent chromatin structures maintenance of cancer-specific epigenetic modification represent a defining molecular characteristic of embryo- patterns. Some of these mutations will likely be

Oncogene Epigenetic cancer therapy M Rius and F Lyko 4263 the targeted enzymes and cancer specificity of the targeted epigenetic pathways. At the same time, the azacytosine scaffold has shown unmatched potency in the inhibition of DNA methylation. This has renewed the scientific interest in the cellular uptake and metabolization pathways of azanucleosides and has allowed the development of second-generation DNMT inhibitors. More importantly, however, the genomic and epigenomic profiling of tumors is currently revolutionizing our molecular understanding of cancer and allows fundamentally novel insights into the epigenetic programs driving tumorigenesis. Several mutations have been identified in factors associated with DNA methylation (TET2, IDH1/2, DNMT3A) and could conceivably increase or decrease global methylation levels. It is possible that these mutations could serve as patient stratification biomarkers for the treatment with demethylating drugs. Additional mutations include druggable somatic mutations in histone modification factors that are connected to the developmental plasticity of cancer cells. Many additional epigenetic target candidates are likely to be discovered by combining data from cancer genome sequencing projects with data from corresponding epigenome Figure 4 A revised rationale for the development of epigenetic sequencing projects. Careful characterization of these drugs. Cancer stem cells are characterized by a specific epigenetic factors will allow the distinction between epigenetic program that underpins their phenotypic plasticity and thus represents a therapy-resistant source for phenotypically diverse drivers and passengers and will lead to the development tumor cells. Treatment of cancer stem cells with epigenetic drugs of a diverse array of highly specific epigenetic drugs. restricts their phenotypic plasticity, thus limiting tumor growth and decreasing the cellular complexity of tumor cells. druggable and thus give rise to novel epigenetic drug Conflict of interest candidates. Maria Rius declares no conflict of interest. Frank Lyko received a commercial research grant from Clavis Pharma. Conclusions and perspective

The field of epigenetics has spawned four approved cancer drugs, thus underscoring its great promise for Acknowledgements cancer therapy. However, the increasing use of these drugs, both in the laboratory and in the clinical practice, We thank Bodo Bru¨ckner for critically reading the manuscript. also identified important limitations. These limitations This work is supported by grants from the Deutsche are largely defined by a lack of specificity on several Forschungsgemeinschaft (RI 2121/1-1) and Deutsche Kreb- levels, including drug specificity, substrate specificity of shilfe (grant number 109773).

References

Aiden AP, Rivera MN, Rheinbay E, Ku M, Coffman EJ, Truong TT Bocker MT, Hellwig M, Breiling A, Eckstein V, Ho AD, Lyko F. et al. (2010). Wilms tumor chromatin profiles highlight stem cell (2011). Genome-wide promoter DNA methylation dynamics of properties and a renal developmental network. Cell Stem Cell 6: human hematopoietic progenitor cells during differentiation and 591–602. aging. Blood 117: e182–e189. Bernstein BE, Mikkelsen TS, Xie X, Kamal M, Huebert DJ, Cuff J Bolden JE, Peart MJ, Johnstone RW. (2006). Anticancer activities of et al. (2006). A bivalent chromatin structure marks key develop- histone deacetylase inhibitors. Nat Rev Drug Discov 5: 769–784. mental genes in embryonic stem cells. Cell 125: 315–326. Broske AM, Vockentanz L, Kharazi S, Huska MR, Mancini E, Bernt KM, Zhu N, Sinha AU, Vempati S, Faber J, Krivtsov AV et al. Scheller M et al. (2009). DNA methylation protects hematopoietic (2011). MLL-rearranged leukemia is dependent on aberrant H3K79 stem cell multipotency from myeloerythroid restriction. Nat Genet methylation by DOT1L. Cancer Cell 20: 66–78. 41: 1207–1215. Bhatla D, Gerbing RB, Alonzo TA, Conner H, Ross JA, Meshinchi S Brueckner B, Rius M, Rivera Markelova M, Fichtner I, Hals PA, et al. (2009). Cytidine deaminase genotype and toxicity of cytosine Sandvold ML et al. (2010). Delivery of azacytidine to human cancer arabinoside therapy in children with acute myeloid leukemia. Br cells by elaidic acid esterification increases therapeutic drug efficacy. J Haematol 144: 388–394. Mol Cancer Ther 9: 1256–1264.

Oncogene Epigenetic cancer therapy M Rius and F Lyko 4264 Cameron EE, Bachman KE, Myohanen S, Herman JG, Baylin SB. transportome in cancer chemosensitivity and chemoresistance. Cancer (1999). Synergy of demethylation and histone deacetylase inhibition Res 64: 4294–4301. in the re-expression of genes silenced in cancer. Nat Genet 21: Huber-Ruano I, Pastor-Anglada M. (2009). Transport of nucleoside 103–107. analogs across the plasma membrane: a clue to understanding drug- Choi SH, Byun HM, Kwan JM, Issa JP, Yang AS. (2007). induced cytotoxicity. Curr Drug Metab 10: 347–358. Hydroxycarbamide in combination with azacitidine or decitabine Issa JP, Gharibyan V, Cortes J, Jelinek J, Morris G, Verstovsek S et al. is antagonistic on DNA methylation inhibition. Br J Haematol 138: (2005). Phase II study of low-dose decitabine in patients with 616–623. chronic myelogenous leukemia resistant to imatinib mesylate. J Clin Choudhary C, Kumar C, Gnad F, Nielsen ML, Rehman M, Walther TC Oncol 23: 3948–3956. et al. (2009). Lysine acetylation targets protein complexes and co- Issa JP, Kantarjian HM. (2009). Targeting DNA methylation. Clin regulates major cellular functions. Science 325: 834–840. Cancer Res 15: 3938–3946. Chuang JC, Warner SL, Vollmer D, Vankayalapati H, Redkar S, Ito S, Shen L, Dai Q, Wu SC, Collins LB, Swenberg JA et al. (2011). Bearss DJ et al. (2010). S110, a 5-aza-20-deoxycytidine-containing Tet proteins can convert 5-methylcytosine to 5-formylcytosine and dinucleotide, is an effective DNA methylation inhibitor in vivo and 5-carboxylcytosine. Science 333: 1300. can reduce tumor growth. Mol Cancer Ther 9: 1443–1450. Ji H, Ehrlich LI, Seita J, Murakami P, Doi A, Lindau P et al. (2010). Chuang JC, Yoo CB, Kwan JM, Li TW, Liang G, Yang AS et al. Comprehensive methylome map of lineage commitment from (2005). Comparison of biological effects of non-nucleoside DNA haematopoietic progenitors. Nature 467: 338–342. methylation inhibitors versus 5-aza-20-deoxycytidine. Mol Cancer Jones PA, Baylin SB. (2007). The epigenomics of cancer. Cell 128: 683– Ther 4: 1515–1520. 692. Cole SP, Deeley RG. (2006). Transport of glutathione and glutathione Jones PA, Taylor SM. (1980). Cellular differentiation, cytidine analogs conjugates by MRP1. Trends Pharmacol Sci 27: 438–446. and DNA methylation. Cell 20: 85–93. Daigle SR, Olhava EJ, Therkelsen CA, Majer CR, Sneeringer CJ, Song J Kantarjian H, Issa JP, Rosenfeld CS, Bennett JM, Albitar M, DiPersio J et al. (2011). Selective killing of mixed lineage leukemia cells by a et al. (2006). Decitabine improves patient outcomes in myelodysplastic potent small-molecule DOT1L inhibitor. Cancer Cell 20: 53–65. syndromes: results of a phase III randomized study. Cancer 106: 1794– Damaraju VL, Sawyer MB, Mackey JR, Young JD, Cass CE. (2009). 1803. Human nucleoside transporters: biomarkers for response to nucleo- Kroep JR, Loves WJ, van der Wilt CL, Alvarez E, Talianidis I, Boven E side drugs. Nucleosides Nucleotides Nucleic Acids 28: 450–463. et al. (2002). Pretreatment deoxycytidine kinase levels predict in vivo Eden A, Gaudet F, Waghmare A, Jaenisch R. (2003). Chromosomal gemcitabine sensitivity. Mol Cancer Ther 1: 371–376. instability and tumors promoted by DNA hypomethylation. Science Kuck D, Singh N, Lyko F, Medina-Franco JL. (2010). Novel and 300: 455. selective DNA methyltransferase inhibitors: docking-based Egger G, Liang G, Aparicio A, Jones PA. (2004). Epigenetics in virtual screening and experimental evaluation. Bioorg Med Chem human disease and prospects for epigenetic therapy. Nature 429: 18: 822–829. 457–463. Laird PW, Jackson-Grusby L, Fazeli A, Dickinson SL, Jung WE, Li E Ehrlich M. (2002). DNA methylation in cancer: too much, but also too et al. (1995). Suppression of intestinal neoplasia by DNA little. Oncogene 21: 5400–5413. hypomethylation. Cell 81: 197–205. Ernst T, Chase AJ, Score J, Hidalgo-Curtis CE, Bryant C, Jones AV Leonard GD, Fojo T, Bates SE. (2003). The role of ABC transporters et al. (2010). Inactivating mutations of the histone methyltransferase in clinical practice. Oncologist 8: 411–424. gene EZH2 in myeloid disorders. Nat Genet 42: 722–726. Li LH, Olin EJ, Buskirk HH, Reineke LM. (1970). Cytotoxicity and Esteller M. (2007). Cancer epigenomics: DNA methylomes and mode of action of 5-azacytidine on L1210 leukemia. Cancer Res 30: histone-modification maps. Nat Rev Genet 8: 286–298. 2760–2769. Fandy TE, Herman JG, Kerns P, Jiemjit A, Sugar EA, Choi SH et al. Linhart HG, Lin H, Yamada Y, Moran E, Steine EJ, Gokhale S et al. (2009). Early epigenetic changes and DNA damage do not (2007). Dnmt3b promotes tumorigenesis in vivo by gene-specific predict clinical response in an overlapping schedule of 5-azacytidine de novo methylation and transcriptional silencing. Genes Dev 21: and entinostat in patients with myeloid malignancies. Blood 114: 3110–3122. 2764–2773. Lizardi PM. (2010). As we bring demethylating drugs to the clinic, we Flasshove M, Strumberg D, Ayscue L, Mitchell BS, Tirier C, Heit W better know the DICE being cast. Oncogene 29: 5772–5774. et al. (1994). Structural analysis of the deoxycytidine kinase gene in Maring JG, Wachters FM, Slijfer M, Maurer JM, Boezen HM, Uges DR patients with acute myeloid leukemia and resistance to cytosine et al. (2010). Pharmacokinetics of gemcitabine in non-small-cell lung arabinoside. Leukemia 8: 780–785. cancer patients: impact of the 79A4C cytidine deaminase polymorph- Fotheringham S, Epping MT, Stimson L, Khan O, Wood V, Pezzella ism. Eur J Clin Pharmacol 66: 611–617. F et al. (2009). Genome-wide loss-of-function screen reveals an Medina-Franco JL, Lopez-Vallejo F, Kuck D, Lyko F. (2011). important role for the proteasome in HDAC inhibitor-induced Natural products as DNA methyltransferase inhibitors: a compu- apoptosis. Cancer Cell 15: 57–66. ter-aided discovery approach. Mol Divers 15: 293–304. Gaudet F, Hodgson JG, Eden A, Jackson-Grusby L, Dausman J, Mohn F, Schubeler D. (2009). Genetics and epigenetics: stability and Gray JW et al. (2003). Induction of tumors in mice by genomic plasticity during cellular differentiation. Trends Genet 25: 129–136. hypomethylation. Science 300: 489–492. Morin RD, Johnson NA, Severson TM, Mungall AJ, An J, Goya R Hagemann S, Heil O, Lyko F, Brueckner B. (2011). Azacytidine and et al. (2010). Somatic mutations altering EZH2 (Tyr641) in follicular decitabine induce gene-specific and non-random DNA demethyla- and diffuse large B-cell lymphomas of germinal-center origin. Nat tion in human cancer cell lines. PLoS One 6: e17388. Genet 42: 181–185. Hansen KD, Timp W, Bravo HC, Sabunciyan S, Langmead B, Mund C, Hackanson B, Stresemann C, Lubbert M, Lyko F. (2005). McDonald OG et al. (2011). Increased methylation variation in Characterization of DNA demethylation effects induced by 5-aza- epigenetic domains across cancer types. Nat Genet 43: 768–775. 20-deoxycytidine in patients with myelodysplastic syndrome. Cancer Hollenbach PW, Nguyen AN, Brady H, Williams M, Ning Y, Richard N Res 65: 7086–7090. et al. (2010). A comparison of azacitidine and decitabine activities in Nikoloski G, Langemeijer SM, Kuiper RP, Knops R, Massop M, acute myeloid leukemia cell lines. PLoS One 5: e9001. Tonnissen ER et al. (2010). Somatic mutations of the histone Howard G, Eiges R, Gaudet F, Jaenisch R, Eden A. (2008). Activation methyltransferase gene EZH2 in myelodysplastic syndromes. Nat and transposition of endogenous retroviral elements in hypomethy- Genet 42: 665–667. lation induced tumors in mice. Oncogene 27: 404–408. Ohm JE, Baylin SB. (2007). Stem cell chromatin patterns: an Huang Y, Anderle P, Bussey KJ, Barbacioru C, Shankavaram U, Dai Z instructive mechanism for DNA hypermethylation? Cell Cycle 6: et al. (2004). Membrane transporters and channels: role of the 1040–1043.

Oncogene Epigenetic cancer therapy M Rius and F Lyko 4265 Ohm JE, McGarvey KM, Yu X, Cheng L, Schuebel KE, Cope L et al. Sneeringer CJ, Scott MP, Kuntz KW, Knutson SK, Pollock RM, (2007). A stem cell-like chromatin pattern may predispose tumor Richon VM et al. (2010). Coordinated activities of wild-type plus suppressor genes to DNA hypermethylation and heritable silencing. mutant EZH2 drive tumor-associated hypertrimethylation of lysine Nat Genet 39: 237–242. 27 on histone H3 (H3K27) in human B-cell lymphomas. Proc Natl Olsen EA, Kim YH, Kuzel TM, Pacheco TR, Foss FM, Parker S et al. Acad Sci USA 107: 20980–20985. (2007). Phase IIb multicenter trial of vorinostat in patients with Soriano AO, Yang H, Faderl S, Estrov Z, Giles F, Ravandi F et al. persistent, progressive, or treatment refractory cutaneous T-cell (2007). Safety and clinical activity of the combination of lymphoma. J Clin Oncol 25: 3109–3115. 5-azacytidine, valproic acid, and all-trans retinoic acid in Parmar S, Seeringer A, Denich D, Gartner F, Pitterle K, Syrovets T acute myeloid leukemia and myelodysplastic syndrome. Blood 110: et al. (2011). Variability in transport and biotransformation of 2302–2308. cytarabine is associated with its toxicity in peripheral blood Stegmann AP, Honders MW, Willemze R, Landegent JE. (1995). mononuclear cells. Pharmacogenomics 12: 503–514. De novo induced mutations in the deoxycytidine kinase (dck) gene in Pastor-Anglada M, Cano-Soldado P, Molina-Arcas M, Lostao MP, rat leukemic clonal cell lines confer resistance to cytarabine (AraC) Larrayoz I, Martinez-Picado J et al. (2005). Cell entry and export of and 5-aza-20-deoxycytidine (DAC). Leukemia 9: 1032–1038. nucleoside analogues. Virus Res 107: 151–164. Stresemann C, Brueckner B, Musch T, Stopper H, Lyko F. (2006). Piekarz RL, Frye R, Turner M, Wright JJ, Allen SL, Kirschbaum MH Functional diversity of DNA methyltransferase inhibitors in human et al. (2009). Phase II multi-institutional trial of the histone cancer cell lines. Cancer Res 66: 2794–2800. deacetylase inhibitor romidepsin as monotherapy for patients with Stresemann C, Lyko F. (2008). Modes of action of the DNA cutaneous T-cell lymphoma. J Clin Oncol 27: 5410–5417. methyltransferase inhibitors azacytidine and decitabine. Int Qin T, Jelinek J, Si J, Shu J, Issa JP. (2009). Mechanisms of resistance J Cancer 123: 8–13. to 5-aza-20-deoxycytidine in human cancer cell lines. Blood 113: Van Rompay AR, Norda A, Linden K, Johansson M, Karlsson A. 659–667. (2001). Phosphorylation of uridine and cytidine nucleoside Rius M, Keller D, Brom M, Hummel-Eisenbeiss J, Lyko F, Keppler analogs by two human uridine-cytidine kinases. Mol Pharmacol D. (2010). Vectorial transport of nucleoside analogs from the apical 59: 1181–1186. to the basolateral membrane in double-transfected cells expressing Varambally S, Dhanasekaran SM, Zhou M, Barrette TR, Kumar- the human concentrative nucleoside transporter hCNT3 and the Sinha C, Sanda MG et al. (2002). The polycomb group protein export pump ABCC4. Drug Metab Dispos 38: 1054–1063. EZH2 is involved in progression of prostate cancer. Nature 419: Rius M, Stresemann C, Keller D, Brom M, Schirrmacher E, Keppler D 624–629. et al. (2009). Human concentrative nucleoside transporter Weber B, Kimhi S, Howard G, Eden A, Lyko F. (2010). Demethyla- 1-mediated uptake of 5-azacytidine enhances DNA demethylation. tion of a LINE-1 antisense promoter in the cMet locus impairs Met Mol Cancer Ther 8: 225–231. signalling through induction of illegitimate transcription. Oncogene Robert T, Vanoli F, Chiolo I, Shubassi G, Bernstein KA, Rothstein R 29: 5775–5784. et al. (2011). HDACs link the DNA damage response, processing of Widschwendter M, Fiegl H, Egle D, Mueller-Holzner E, Spizzo G, double-strand breaks and autophagy. Nature 471: 74–79. Marth C et al. (2007). Epigenetic stem cell signature in cancer. Nat Schlesinger Y, Straussman R, Keshet I, Farkash S, Hecht M, Genet 39: 157–158. Zimmerman J et al. (2007). Polycomb-mediated methylation on Williams K, Christensen J, Pedersen MT, Johansen JV, Cloos PA, Lys27 of histone H3 pre-marks genes for de novo methylation in Rappsilber J et al. (2011). TET1 and hydroxymethylcytosine in cancer. Nat Genet 39: 232–236. transcription and DNA methylation ﬁdelity. Nature 473: 343–348. Shen X, Liu Y, Hsu YJ, Fujiwara Y, Kim J, Mao X et al. (2008). Wolff EM, Byun HM, Han HF, Sharma S, Nichols PW, Siegmund KD EZH1 mediates methylation on histone H3 lysine 27 and comple- et al. (2010). Hypomethylation of a LINE-1 promoter activates an ments EZH2 in maintaining stem cell identity and executing alternate transcript of the MET oncogene in bladders with cancer. pluripotency. Mol Cell 32: 491–502. PLoS Genet 6: e1000917. Silverman LR, Demakos EP, Peterson BL, Kornblith AB, Holland JC, Wu H, D0Alessio AC, Ito S, Xia K, Wang Z, Cui K et al. (2011). Dual Odchimar-Reissig R et al. (2002). Randomized controlled trial of functions of Tet1 in transcriptional regulation in mouse embryonic azacitidine in patients with the myelodysplastic syndrome: a study stem cells. Nature 473: 389–393. of the cancer and leukemia group B. J Clin Oncol 20: 2429–2440. Xu WS, Parmigiani RB, Marks PA. (2007). Histone deacetylase Simon JA, Lange CA. (2008). Roles of the EZH2 histone methyl- inhibitors: molecular mechanisms of action. Oncogene 26: transferase in cancer epigenetics. Mutat Res 647: 21–29. 5541–5552. Sims III RJ, Reinberg D. (2008). Is there a code embedded in proteins Yoo CB, Jeong S, Egger G, Liang G, Phiasivongsa P, Tang C et al. that is based on post-translational modiﬁcations? Nat Rev Mol Cell (2007). Delivery of 5-aza-20-deoxycytidine to cells using oligodeox- Biol 9: 815–820. ynucleotides. Cancer Res 67: 6400–6408.

Oncogene