Epigenetic processes control the organization and output of the genome in somatically heritable ways and do not involve changes in DNA sequence. They are fundamental to the physiological and pathological behaviour of cells and involve interactions between chemical marks on the nucleic acid and protein components of chromatin. The marks are placed on the macromolecules by enzymes (“writers”) and can be removed by another group of enzymes referred to as “erasers”. For these marks to be functional, they are interpreted by “readers” which can either activate or suppress gene expression to ensure accurate read-out of genetic information in different cell types. Because epigenetic states are heritable during meiotic and somatic cell division, understanding the mechanisms for their inheritance is an area of intense interest.

The genetic basis of cancer has been accepted for a long time and the output of “The Cancer Genome Atlas” (TCGA) project has unequivocally shown the presence of gene mutations in almost all the thousands of cancers that have been sequenced.¹  However, a surprising outcome of this collaborative effort was that about a third of newly detected mutations were in genes that control the structure and function of the epigenome. Perhaps this should have been expected, given that pathologists have long relied on changes in chromatin architecture, visible under the light microscope, to make a diagnosis of malignancy. These mutations may be at least partially responsible for this altered nuclear morphology. Another outcome of TCGA was the finding that all human tumours, without exception, contain alterations in the patterns of DNA methylation.²  Given that this process is fundamental to the development and function of vertebrates, alterations in methylation patterns could contribute to somatically heritable changes in gene expression which underlie malignancy.

Early studies to define how the functional output of the genome is controlled by chromatin were inclined to be compartmentalized in the sense that protein and nucleic acid biologists and biochemists tended to focus on their own fields. The mechanisms by which DNA methylation blocks transcription initiation through protein interactions are still not entirely clear. The issue of causality linking these protein and DNA modifications to functional output in normal development and cancer was controversial for many years. Despite these basic uncertainties, the field of epigenetic therapies for cancer developed with the hope of restoring growth control to malignant cells.

In this chapter, we will focus on DNA methyltransferase inhibitors (DNMTi) and histone deacetylase inhibitors (HDACi), which were the first chromatin-targeting drugs to be developed and validated in the clinic, resulting in their approval for treatment of certain malignancies by the American Food and Drug Administration (FDA) and other international regulatory organizations.

The fact that DNA contained a “fifth base”, subsequently identified as 5-methylcytosine was firmly established by paper chromatographic methods in 1948 by Hotchkiss.³  The function of this modification, which constituted about 1% of the total base composition of DNA, remained unclear until the seminal papers by Riggs⁴  and Holliday and Pugh⁵  independently hypothesized that it might play a role in controlling gene expression and epigenetic inheritance. Central to these hypotheses were the concepts that DNA methylation was predominantly found at CpG palindromes, which were either symmetrically methylated or unmethylated on both strands of DNA. Patterns of DNA methylation, in which some sites were modified but others not, were proposed to be important for gene control. Initial methylation during differentiation was postulated to be catalysed by so-called “de novo DNA methyltransferases”. A key additional hypothesis was that the patterns could subsequently be copied after cell division by a “maintenance methylase” which would recognize hemimethylated sites and convert them to full methylation of both cytosines in the CpG palindrome. These ideas were later verified to be essentially true and the postulated de novo methylases (DNMT3A and DNMT3B) and the maintenance enzyme (DNMT1) were identified and cloned and extensively studied.^6 ,7  DNA methylation is now accepted as being a heritable epigenetic mark essential for embryonic and post-natal development of vertebrates.

For a long time, the methylation mark was considered to be permanent and could only be removed by a passive process involving cell division in the absence of maintenance methylation. While this is certainly one possibility, the discovery of 5-hydroxymethylcytosine in mammalian DNA⁸  and the identification of the three Ten Eleven Translocase (TET) enzymes dramatically changed that concept and provided a mechanism for active demethylation for the first time.⁹  However, the paradoxical association of TET loss of function with genome wide DNA hypomethylation suggests that the exact relationship between methylation and demethylation remains an open question.¹⁰ 

The potential role of DNA methylation in cancer was proposed in the 1970s, given the hypotheses that it was important in gene control and might therefore contribute to aberrant cell behaviour in neoplasia.¹¹  Since that time, many studies have firmly established a key role for aberrations in this epigenetic process in all human cancers.¹²  Also, the identification of mutations in the DNMT3A and TET2 genes in leukaemia cells has demonstrated the potential for DNA methylation abnormalities to initiate abnormal growth control.¹³  Given the extensive literature on this subject, it will not be re-examined here.

The prototypical DNA methyltransferase inhibitors were developed in Prague in Czechoslovakia in the Institute for Organic Chemistry and Biochemistry in the mid-1960s^14 ,15  (Figure 1.1). 5-Azacytidine (5-AzaCR) and 5-aza-2′-deoxycytidine (5-Aza-CdR) were initially synthesized as nucleoside analogues predicted to be potential cytotoxic anticancer agents following incorporation into DNA. Clinical trials began initially with 5-Aza-CR in childhood leukemia¹⁶  and suggested that the drug might be effective in the treatment of haematological but not solid tumors.¹⁷  5-Aza-CR was then tested in the USA, initially in the treatment of acute leukaemia in children where it showed some activity, which was presumed to be due to cytotoxic activity.¹⁸  A review published soon after summarized these findings and concluded that the drugs might be of value in the treatment of leukaemia, however there was little response in patients treated with solid tumors.¹⁹  Perhaps the fact that solid tumours appeared to be refractory was not surprising given that most chemotherapeutic regimens rely on multiple drugs and few single-agent treatments are currently in use.

These early trials focused on determining the maximum tolerated dose as opposed to the maximum biologically effective dose. Because the drugs had strong myelosuppressive effects, the decision to use high doses led to the delayed development of the drugs for the treatment of leukaemia.

The fact that these nucleosides were potent inhibitors of DNA methylation was discovered by accident, when we observed the formation of differentiated cells in cultures of immortalized mouse embryo cells several days or weeks after exposure to the nucleoside.²⁰  Importantly, the authenticity of these new phenotypes was demonstrated,²¹  and the results were highly surprising in that they showed the reprogramming of cellular states by brief exposure to either 5-Aza-CR or 5-Aza-CdR. Subsequently, we demonstrated that there was a link between inhibition of DNA methylation and the formation of these differentiated cells.²²  An important consideration from this work was that the dose-response curves were not linear but showed a bell-shaped response. This suggested that if the drugs functioned as cellular re-programmers, it might be important to use doses below the maximum tolerated dose. However, this was not tested in the clinic for many years and, as mentioned earlier, delayed the development of the inhibitors for the treatment of myeloid tumours.

Considerable effort was put into the clinical testing of both 5-Aza-CR and 5-Aza-CdR over the next decades. In this regard, the contributions of Silverman et al.²³  Lübbert and co-worker²⁴  and Fenaux and Ades²⁵  were instrumental in the demonstration in 2009 that 5-aza-CR was effective in high-risk myelodysplastic syndromes.²⁶  This pivotal study, which represented a large randomized open-label phase III trial, clearly demonstrated the efficacy of DNA methyltransferase inhibitors (DNMTis) in the treatment of liquid tumours.

While the early clinical trials were being conducted, efforts were made to develop more stable versions of DNA methyltransferase inhibitors, which would resist both hydrolysis and deamination within the patient. Examples of the many compounds tested are shown in Figure 1.1. Included in these was zebularine, a stable inhibitor of DNA methyltransferases, neither subject to hydrolysis nor deamination, which was an effective inhibitor of methylation both in cells and in animal studies.²⁷ 

However, concerns were raised that the drug might be too toxic for treatment of humans so clinical trials were not pursued. Promising results were also obtained with Guadecitabine, originally called SGI-110, which is a dinucleotide prodrug of 5-aza-CdR and 2-deoxyguanosine. This drug was resistant to deamination and was effective in inducing demethylation both in cell culture²⁸  and in patients.²⁹  However, it did not show sufficient improvement over existing drugs³⁰  and clinical development has therefore been discontinued. The results of the above work resulted in the FDA approval for the use of DNMTis in the treatment of myeloid malignancies and they are now the standard-of-care for these diseases in much of the world.

The nucleoside analogues all require incorporation into the DNA of dividing cells to be effective, given that they are mechanism-based inhibitors of DNA methylation. They inhibit all three known DNA methyltransferases due to the formation of a covalent bond between the enzyme and the 6 position of the 5-azacytosine ring.³¹  This leads to the degradation of the enzymes with the result that the DNA synthesized subsequently is unmethylated. Pharmaceutical companies have attempted to develop inhibitors that do not require incorporation into DNA for a long time. Recently, a new tool compound called GSK 3685032 has shown the ability to specifically inhibit DNMT1 specifically both in culture and in animal models.³²  The development of such specific inhibitors targeting individual DNA methyltransferases is potentially a breakthrough which augurs well for the future development of the field.

Early protocols for treatments with DNMTis required either intravenous or subcutaneous injection, which limited the more general utilization of the drugs, given the need for patients to visit the clinic regularly. Drug companies, therefore focused on the development of oral formulations which would facilitate drug administration (Table 1.1). Two main approaches have been used, either the encapsulation of 5-aza-CR, for slow release, or 5-aza-CdR combined with a cytidine deaminase inhibitor (Cedazuridine). Both approaches have been FDA- approved (Table 1.1) for the treatment of myeloid malignancies. As mentioned earlier, Guadecitabine, which is a dinucleotide prodrug resistant to deamination, was effective in the treatment of MDS but was not further developed since it did not show a survival advantage when compared to existing treatments.

As has been the case with many drugs initially developed as single agents, the focus has switched to combining DNMTis with other therapies. There are now multiple treatment regimens showing survival advantages in which DNMTis are combined with Venetoclax, which is a BCL2 inhibitor.³³ 

The mechanisms by which DNMTis elicit responses in treated cells have been difficult to decipher. Much of the focus has been on their potential to reactivate tumour suppressor genes silenced by promoter CpG island methylation.¹²  However, our findings, taken together with those of the Baylin lab, that they could induce the expression of repetitive elements and a state of “viral mimicry” suggested the possible involvement of the immune system in responses.^34 ,35  The approach is based on the fact that more than 50% of the human genome consists of transposable elements including endogenous retroviruses and short and long interspersed nuclear elements. The expression of these potentially harmful sequences is controlled by epigenetic processes including DNA and histone methylation. Therefore, drug treatment can result in their expression and the induction of an innate and acquired immune response. The potential for epigenetic therapy to synergize with immune-oncology is now being actively tested.³⁶ 

Pilot clinical trials have demonstrated activation of transposable elements in leukaemia patients treated with DNMTis and this was associated with increased survival.³⁷  Recently, another Phase 2 trial showed that DNMTis, combined with immunotherapy, increased the survival of patients by three-fold relative to those patients who have become resistant to DNMTi therapy.³⁸  These developments suggest that there is a promising future for the treatment of leukaemia with DNMTis in combination strategies and future larger trials are indicated.

It has been apparent for a long time that DNMTis are not effective as single agents in the treatment of solid tumors.¹⁹  More recent trials have used the concept of “epigenetic priming” of tumours that have become resistant to standard therapies to re-sensitize them to drug vulnerabilities.^39 ,40  These are just two examples, among many others that have demonstrated the safety of the approaches but have not yet shown sufficient efficacy to become a standard of care. Recent combination trials with other cytotoxic agents and immunotherapies have demonstrated safety and small potential for increased survival,⁴¹  but further development is needed. Perhaps solid tumours are relatively refractory to DNMTi therapy because these inhibitors are unstable in plasma. All the current nucleoside DNMTis require incorporation into DNA to be effective, so perhaps the development of more specific inhibitors, which do not require DNA incorporation such as GSK3685032, will result in better patient outcomes.

The nucleosome, which is the fundamental organizing structure in chromatin, consists of about 147 base pairs of DNA wrapped around a histone octamer of two molecules each of histones H2A, H2B, H3 and H4. Sixty years ago, the post-translational acetylation of these histones was first discovered by Vincent Allfrey, who immediately suggested that the modification might have a role in controlling gene expression.⁴²  Unlike the situation with DNA methylation, which at first appeared to be quite stable, it was soon realized that histone acetylation was more dynamic, and the existence of an acetylation–deacetylation cycle was established.^43 ,44  The level of acetylation was therefore determined by an interplay between acetylation and deacetylation. Links between DNA methylation and histone acetylation were suggested by Tazi and Bird with the demonstration that unmethylated CpG rich regions of DNA (CpG islands) were hyperacetylated.⁴⁵ 

Interestingly, the development of histone deacetylase inhibitors (HDACi) also began in the mid 1970s due to unexpected observations on the effects of butyric acid on cellular differentiation. Cultured erythroleukemia cells can be induced to undergo a form of erythroid differentiation by organic solvents such as dimethyl sulfoxide and begin producing haemaglobin.⁴⁶  Leder and Leder discovered that butyric acid was a far more potent inducer of differentiation than dimethyl sulfoxide.⁴⁷  This result was followed up by Riggs et al., who showed that n-butyrate suppressed histone deacetylation in HeLa and Friend erythroleukemia cells.⁴⁸  They found that rapid, dramatic and reversible changes in histone acetylation occurred in the presence of n-butyrate. Soon after this, the lab of Vincent Allfrey showed that the effect of butyrate was not due to an increase in the rate of histone acetylation but rather to a decrease in the rate of histone deacetylation.⁴⁹  They also showed that an alteration in nucleosome structure in highly acetylated chromatin was accompanied by an increase in DNA accessibility.⁵⁰  An important additional discovery was that the suppression of histone deacetylation led to the accumulation of multi-acetylated forms of histones H3 and H4 in treated cells. These findings strongly suggested a link between the acetylation status of histones H3 and H4 and gene expression. However, it was not until the seminal work of Allis that a causal relationship between gene activity and histone acetylation was firmly demonstrated.⁵¹ 

These pioneering discoveries led to the development of HDACis that have been an area of intense activity for the treatment of cancer.^27 ,52  Vorinostat was the first HDACi approved by the FDA for the treatment of cutaneous T-cell lymphoma⁵²  and a total of four drugs have now been approved by the FDA.

Interestingly, although the relationship between the DNA and histone components of chromatin was poorly understood in the early 1980s, two early papers addressed the issue of whether inhibition of DNA methylation by 5-aza-CR concomitantly with inhibition of histone deacetylation by sodium butyrate would act synergistically in causing gene activation. Jahangeer et al. showed this synergism in the induction of the β-adrenergic receptor in HeLa cells exposed to both chemicals at the same time.⁵³  They interpreted their results to suggest that the induction of β-adrenergic receptors in HeLa cells by butyrate and 5-aza-CR involved separate mechanisms, which turned out to be true. Importantly, Ginder et al. showed this synergism in animals by exposing adult white leghorn chickens to both chemicals.⁵⁴  The importance of the establishment of open chromatin was also explored by measuring the expression of the embryonic globin gene and showing an increase in DNAse1-hypersensitivity at the promoter of the gene. These ground-breaking experiments, which were the first to link DNA methylation, histone deacetylation and chromatin opening at relevant promoters received little attention and have been cited only 40 and 168 times, respectively, over the last four decades.

The possibility that combination treatments of DNMTis with HDACi might have value in the reactivation of tumour suppressor genes aberrantly silenced by promoter CpG island methylation was investigated in a landmark paper from the Baylin laboratory.⁵⁵  This study conclusively showed that although trichostatin (an HDACi) could not reactivate extensively methylated promoters by itself, it increased the level of gene expression induced by 5-aza-CdR alone. This led to the concept that combinations of epigenetic drugs might have value in the clinic⁵⁶  and to clinical trials to test this approach.⁵⁷  Many other combinations are undergoing testing, including combinations of DNMTs with inhibitors of the polychrome repressive complex 2 (PRC2)⁵⁸  such as Tazemetostat which is now FDA approved. These studies point to the value of combining drugs based on their potential synergies on chromatin modifications.

This book will discuss many new potential drugs targeting the epigenome for therapy. Although the first two classes of inhibitors discussed above are now approved for specific indications, it is important to remember that their clinical development took a long time and many false starts and that is not unusual in oncology. For example, it took many years before a breakthrough occurred in the application of immunotherapy to therapy. The availability of epigenome maps and an ever-increasing understanding of how chromatin functions in a holistic way will be of great value in the development and testing of new drugs. Most therapeutic regimens use combinations of drugs and the same will almost certainly apply to epigenetic therapy.

This work was supported by the US National Cancer Institute (NCI; grants R35CA209859 to P.A.J. and R50CA243878 to M.L.).

The authors thank Dr Kirsten Grønbaek for the helpful comments, Dr Gangning Liang for his help with preparing the references, and Ryan Burgos for help with preparing Table 1.1.