Genome-wide hypomethylation at CpG sites has been reported in many tumor types, such as brain tumor, gastric cancer, liver cancer, and breast cancer [24, 25]. Though global DNA hypomethylation usually occurs in intergenic regions and contributes less to genetic mutations, it can result in genome instability via gene mutations, deletions, inversions, translocations, and amplifications [26]. For example, Long Interspersed Nucleotide Element 1 (LINE-1) is a key component of interspersed DNA repeats. LINE-1 hypomethylation has been reported in several cancer types and portends to a worse prognosis [27, 28, 29, 30]. A recent meta-analysis pointed out that LINE-1 hypomethylation in tissue samples may serve as an epigenetic marker for cancer risk [30]. In addition, several studies highlighted that hypomethylation of cancer-specific CGI may drive the overexpression of oncogenic driver genes [31]. A well-illustrated example is the relationship between BCL-2 and B-cell chronic lymphocytic leukemia [32].

In healthy tissues, the CpG dinucleotides at promoter sites of tumor suppressor genes (TSG) are generally unmethylated and transcriptionally active. However, approximately 5–10% CGI are hypermethylated to prompt TSG silencing in cancer cells [33], such as VHL, RB1, CDKN2A, GATA4, and MHL1 [34, 35]. DNA hypermethylation was first observed in colon cancer. Studies also indicated that CGI methylated phenotype was related to BRAF mutations in colon cancer [36, 37, 38], whereas glioblastoma displaying promoter hypermethylation was significantly associated with IDH1 mutations [39, 40]. Hypermethylation of mismatch repair gene MHL1 has been found in human colorectal cancers, which is related to microsatellite instability [41]. Another example is the hypermethylation of CDKN2A (p16), leading to the inactivation of this gene in esophageal adenocarcinoma [42].

There is a bidirectional interplay between genetic and epigenetic alterations. In addition to genetic mutations driving the dysregulated epigenetic landscape, DNA methylation-induced mutagenesis are also one of the major sources of genetic alterations, resulting in the formation of tumor-associated neoantigens [43]. DNA methylation enables the cytosine base to be more susceptive to deamination, either spontaneous or mutagen motivated, leading to C to T transitions. DNA hypermethylation of transposable elements is critical to epigenetic silencing, as DNA 5-methycytosine is mostly found to locate in the transposable elements [44]. For example, cancer-testis antigens (CTA) could be suppressed by DNA methylation in most somatic cells, but they can be specifically expressed in normal male germ cells [43]. Importantly, reactivation of CTA-coding genes by demethylation can contribute to the presence of neoantigens with immunogenicity and thus enhance immune surveillance [45].

DNA methylation changes can regulate the differentiation of naïve CD8${}^{+}$ T cells into effector T cells, consisting of cytotoxic and memory subtypes [48, 49, 50, 51] (Fig. 3). Naïve CD8${}^{+}$ T cells will be activated into an ‘effector’ state after antigen exposure, which is characterized by the rapid secretion of cytokines and effector proteins, as well as the capacity for infiltration and migration into the TME. However, continuous antigen stimulation would inevitably result in T cell exhaustion that is subdivided into a ‘plastic dysfunctional state’ and a ‘fixed dysfunctional state’ [52]. This process is accompanied by an epigenome-wide remodeling wherein hundreds of thousands of genes are variously methylated [53], which are collectively maintained by DNMT1, DNMT3A, and TET2. In the developmental trajectory, genes that prompt the activation, proliferation, and differentiation of T cells are demethylated and highly expressed, such as IFNG and GZMB [53]. Conversely, unnecessary genes are frequently methylated and regressed, such as TCF7. Naïve CD8${}^{+}$ T cells more commonly undergo demethylation compared to cytotoxic and exhausted CD8${}^{+}$ T cells, whereas effector genes of cytotoxic CD8${}^{+}$ T cells generally experienced hypermethylation to hypomethylation from naïve to cytotoxic T cell fate, such as GZMB, IFNG, CCL4, and CCL3 [54]. Furthermore, memory CD8${}^{+}$ T cells maintain demethylated effector genes, inducing the rapid and robust immune response of memory CD8${}^{+}$ T cells to re-challenge with antigens [52]. DNMT3A is a methyltransferase in charge of de novo methylation. Studies have shown that genetic ablation of DNMT3A may induce fewer effector CD8${}^{+}$ T cells [55]. In addition, the loss of DNMT3A can abolish de novo methylation which serves to establish immunological memory [52].

Fig. 3.

DNA methylation plays a key role in the differentiation of CD8${}^{\mathrm{+}}$ T cells from naïve to effector status. DNA methylation changes can lead to the formation of different subtypes of CD8${}^{+}$ T cells, including effector T cells and exhausted T cells. Effector genes of cytotoxic CD8${}^{+}$ T cells generally experienced hypermethylation to hypomethylation from naïve to cytotoxic T cell fate. In memory CD8${}^{+}$ T cells, naïve T cell-associated genes are methylated but effector T cell-associated genes are demethylated. These methylation levels are maintained by DNMT1, DNMT3A, and TET2.

Consistently, DNA methylation is responsible for the fate-decision of naïve CD4${}^{+}$ T cells, modulating their differentiation into various effector subtypes, such as Th1, Th2, Th17, Treg, and memory T cells (Fig. 4). Th1 cells secrete type I cytokines (IL-12 and IFN-$\gamma$) that prompt tumor suppression, whereas Th2 cells secrete type II cytokines (IL-4, IL-5, and IL-13) that polarize immunity towards tumor progression [48, 50, 51]. The methylation status of immune genes is linked to the immune response in the TME. The differentiation of naïve CD4${}^{+}$ T cells into Th1 and Th2 relies on different epigenetic remodeling of certain genes. It has been demonstrated that some genes present opposite patterns of expression. For example, IFN-$\gamma$ remains demethylated at promoter regions and upregulates IFN-$\gamma$ production upon Th1 cell differentiation [56]. In contrast, the IFN-$\gamma$ promoter is methylated through de novo methylation catalyzed by DNMT3A during the development of Th2 cells and Th17 cells [57, 58]. Previous studies reported DNMT3A deletion may result in the failure of the formation of Th2 and Th17 cells, mainly due to the absence of de novo methylation at the IFN-$\gamma$$$locus [59]. Furthermore, IL-4 and IL-17 genes are demethylated and Th2 and Th17 subtypes are activated [59, 60]. It has been found that a global loss of DNA methylation occurs in the formation of CD4${}^{+}$ memory T cells [61]. There is, strong evidence to indicate that differentiation of CD4${}^{+}$ naïve T cells into Th1, Th2, Th17, Treg, and memory T cells depends on the dynamic changes of DNA methylation at gene promoter and enhancers, which is maintained by DNMT1, DNMT3A, and TET2.

Fig. 4.

DNA methylation plays a key role in the activation of naïve CD4${}^{\mathrm{+}}$ T cells into effector T cells, such as Th1, Th2, Th17, and Treg cells. Naïve CD4${}^{+}$ T cells are stimulated by IL-12 and IFN-$\gamma$ and form Th1 cells, of which process the methylation status is maintained by DNMT1 and TET2. Th1 cells secrete type I cytokines (IL-12 and IFN-$\gamma$) that prompt tumor suppression. The IFN-$\gamma$promoter is methylated through de novo methylation catalyzed by DNMT3A during the development of Th2 cells and Th17 cells. Also, the IL-4 gene is demethylated and upregulated in Th2 cell, while IL-17 is demethylated and highly expressed in Th17 cell. For the Treg cells, FOXP3 gene is demethylated at promoters and enhancers, followed by the increased expression of FOXP3.

Activation of naïve T cells initially requires the interactions between T cell receptor (TCR) present on T cells and MHC complex (signal 1) expressed on antigen-presenting cells (APC), followed by the activation of co-stimulatory molecules (signal 2) provided by mature dendritic cells. When both signals are present, T cell activation initiates an autonomous intracellular signaling cascade, leading to T cell expansion and differentiation. Global DNA methylation remodeling plays a critical role in priming and activation of cytotoxic effector T cells. For example, demethylation at a promoter-enhancer region of the IL-2 loci, significantly drives the increased levels of IL-2 cytokines that are required for T cell activation and proliferation [62]. Promoter demethylation is observed in several effector genes, such as GZMK and GZMB [53].

T cells become dysfunctional or exhausted when they are exposed to continuous antigen stimulation, resulting in the failure to produce effective immune response. Generally, exhausted T cells display high levels of inhibitory receptors, such as PD-L1, leading to reduced effector function and suppressed T cell proliferation [63, 64, 65]. It is recognized that DNA methylation contributes to regulating PD-L1 expression. Compelling evidence implicates that PD-L1 promoter is demethylated within exhausted CD8${}^{+}$ T cells in multiple tumors [66, 67, 68], whereas PD-L1 in effector CD8${}^{+}$ T cells remained in the methylated status. PD-1 blockade has the ability to reverse T cell exhaustion in both chronic infection and tumor settings. But this reinvigoration is transient, as T cells become re-exhausted after the withdrawal of PD-1 blockade [69]. Based on ATAC-seq analysis, this is mainly because immune checkpoint inhibitor (ICI) is unable to completely revert exhausted T cells into effector T cells [70]. Furthermore, de novo DNA methylation by DNMT3A at post-effector stage is required for the exhaustion phenotype [70]. Genetic ablation of DNMT3A significantly induced the generation of effector cytokines in the mice model infected with lymphocytic choriomeningitis virus (LCMV) [70]. PD-1 blockade does not erase exhaustion-related DNA methylation. The ablation of DNMT3A combined with a DNA demethylating agent before PD-1 blockade, will result in pronounced T cell expansion and effective immune responses. Overall, aberrant DNA methylation at specific gene loci confers T cell exhaustion [54, 63, 71]. ICI can successfully rejuvenate exhausted T cells that have been epigenetically remodeled.

TSG silencing caused by promoter hypermethylation is a common mechanism of tumorigenesis, which greatly facilitated the discovery of DNA methyltransferase inhibitors (DNMTi) in the past few decades [100, 101, 102, 103]. DNMTi, also termed as hypomethylating agents, are the most widely utilized epigenetic drugs, especially for treating hematologic malignancies. Analogues of the nucleoside cytidine involve 5-azacytidine (5-AZA), 5-aza-2’deoxycytidine (decitabine), and SGI-110 (guadecitabine), which make DNMT inaccessible to DNA and consequently cause DNA hypomethylation [104, 105, 106]. Both 5-AZA and decitabine have been approved by the Food and Drug Administration (FDA) [107]. These two drugs are currently used as first-line therapy for myelodysplastic syndrome (MDS), bringing improved response rates and prolonged survival [108, 109]. Impressive success has also been achieved in the treatment of acute myeloid leukemia (AML) and chronic myelomonocytic leukemia (CMML) (Table 1). In addition, DNMTi has shown improved outcomes in solid tumors, such as ovarian cancer and non-small cell lung cancer [110, 111, 112].

There is a frequently overlooked phenomenon that the genome-wide DNA hypomethylation occurs in several solid tumors [113], which is related to tumor initiation and progression. TET2 mutations are detected in approximately 20–25% of MDS, 7–23% AML, and over 53% CMML [114, 115, 116]. Further, previous studies indicated that TET1 acts as a suppressor of hematological malignancies [117, 118]. However, there are no approved TET inhibitors to reverse DNA hypomethylation, though several agents are being tested at preclinical stages. For example, the novel compound Bobcat339 has been shown to directly block the enzymatic activity of TET1/2 [119]. Since the JAK/STAT pathways are involved in TET1 transcription, STAT inhibitor UC-51423 may avoid aberrant TET1 function [120]. An alternative approach is to target SAM that donates the methyl group and manipulates the level of methylation. More importantly, emerging studies suggest that SAM could modulate cancer immunity by affecting the functionality of CD8${}^{+}$ T cells [121].

Since DNA methylation and histone modification usually work in parallel, considerable efforts are being made to explore various combinations that may significantly increase the efficacy of a single agent and decrease drug resistance. For instance, in a murine ovarian cancer model, the combination of DNMTi and histone deacetylase (HDACi) is capable of increasing the cytotoxic activity of CD8${}^{+}$ cells and NK cells, and promoting effector T cell infiltration in the tumor microenvironment (TME) [122]. Similarly, as the location of DNA methylation and enhancer of zeste homolog 2 (EZH2)-mediated histone acetylation in chromatin are usually mutually exclusive, EZH2 is able to maintain quiescence of particular genes when CGI become activated due to DNA hypermethylation [123, 124]. Therefore, the combination of DNMTi with EZH2 inhibitors is a promising method to enhance anti-tumor responses.

Numerous clinical trials are evaluating the combination of epigenetic modulators with immune checkpoint inhibitors (ICIs) [125]. The rationale for this proposal is that epigenetic drugs may assist immunotherapy to increase the ability of cytotoxic T cells to attack malignant cells [126, 127, 128, 129]. ICI, an exciting cancer therapy which has emerged in the past several years, has been approved by FDA for non-small cell lung cancer (NSCLC) , melanoma, and renal cancer [130, 131, 132]. There was an unexpected finding that a group of advanced NSCLC patients who had relapsed after a combination of azacytidine (DNMTi) and entinostat (HDACi), received a robust and durable response when they were subsequently enrolled in a clinical trial of ICI [129]. However, it is unknown whether the favorable outcomes reflect the effectiveness of combined therapy or are merely responses to the exposure to ICI. Numerous efforts have been made to figure out the underlying mechanisms as to how epigenetic therapy mediates the augmentation of ICI. It is thought that DNMT inhibitors could reverse immune evasion, as they are able to upregulate the expression of tumor-associated antigens (TAAs) and MHC molecules on the tumor cell surface [133, 134]. Recent studies highlighted that ‘viral mimicry’ produced by DNMTi was tightly associated with increased interferon that may trigger immune attraction [126, 127]. Interests have been switched into the activation of endogenous retroviruses (ERV) which are generally methylated and suppressed in most somatic cells. Mounting evidence suggests that ERV expressed in clear cell renal cell carcinoma could encode peptides that induce T cell and B cell immunoreactivity [135]. As cytosine methylation is critical to the regulation of ERV, reactivation of ERV could be achieved by exposure to DNMTi, and might be effective in reversing immune tolerance for ICI therapy [126, 127, 136, 137].

Aberrant DNA methylation is an epigenetic memory signature that is central to the pathobiological of IDH-mutant tumors, such as gliomas and AML [138]. DNMTi could be used in combination with isocitrate dehydrogenase (IDH) inhibitors, reducing the production of oncometabolites. More specifically, inhibition of mutant forms of IDH, an enzyme in the TCA cycle, could impact the demethylation status of DNA and histone. Findings also demonstrated that DNMTi might be more effective than IDH inhibitors [139], but the efficacy of combination therapy is still unknown. Another example is the combination use of DNMTi with serine metabolism that has been shown to more aggressively kill liver kinase B1-deficiency tumors carrying KRAS [140]. Overall, despite great attentions in targeting both metabolism reprogramming and epigenetic remodeling, it is unknown whether these two hallmarks function in a synergistical manner.

Preclinical studies suggest that DNMTi may improve outcomes when combined with standard cytotoxic drugs. Epigenetic regulation might be a potential mechanism leading to drug resistance to cytotoxic drugs, and thus DNMTi could reverse these pathways and improve the durability of clinical responses [141]. One ongoing clinical trial is being tested to combine DNMTi with cytotoxic agents, making ovarian cancers re-sensitive to these standard drugs [142, 143, 144, 145, 146]. Furthermore, a series of clinical trials are underway to explore combinations of DNMTi with chemotherapy, in an attempt to restore chemosensitivity among patients whose disease have relapsed.

A triple combination of DNMTi, HDACi, and ICI, has also been shown to trigger a stronger anti-tumor effect. The application of ICI following combined inhibition of DNMT1 and EZH2 would make ovarian cancer cells to express CXCL9 and CXCL10, which could activate T helper 1 cells and attract tumor-infiltrating lymphocytes (TILs) [147].