Skip to main content

08-12-2015 | Treatment | Article

Improving cancer immunotherapy with DNA methyltransferase inhibitors

Cancer Immunology, Immunotherapy

Authors: Mohammad H. Saleh, Lei Wang, Michael S. Goldberg

Publisher: Springer Berlin Heidelberg


Immunotherapy confers durable clinical benefit to melanoma, lung, and kidney cancer patients. Challengingly, most other solid tumors, including ovarian carcinoma, are not particularly responsive to immunotherapy, so combination with a complementary therapy may be beneficial. Recent findings suggest that epigenetic modifying drugs can prime antitumor immunity by increasing expression of tumor-associated antigens, chemokines, and activating ligands by cancer cells as well as cytokines by immune cells. This review, drawing from both preclinical and clinical data, describes some of the mechanisms of action that enable DNA methyltransferase inhibitors to facilitate the establishment of antitumor immunity.


The accumulation of genomic mutations in oncogenes and tumor suppressor genes has long been regarded as the core driver of tumor progression [ 1], but research over the past decade has highlighted the growing importance of epigenetic mechanisms in neoplastic disease [ 2]. The aberrant orchestration of epigenetic machinery can give rise to cancer, rendering the understanding and, as required, inhibition of these mechanisms very important.
Global hypomethylation of genomic DNA increases during carcinogenesis, contributing to chromosomal instability, reactivation of transposable elements, and loss of imprinting [ 3, 4]. In contrast, the focal hypermethylation of promoter CpG islands and accumulation of histone modifications silences many tumor suppressor genes [ 3].
This emerging role for epigenetics in cancer has inspired investigation of the therapeutic potential of drugs that selectively target and reverse abnormalities that arise from epigenetic modifications [ 4]. To date, there are at least four known types of DNA modifications and 16 kinds of histone modifications, including methylation, acetylation, phosphorylation, and ubiquitylation [ 5]. Of the five epigenetic modifiers currently approved by the US Food and Drug Administration, two inhibit DNA methyltransferases (DNMTs: azacitidine and decitabine) and three inhibit histone deacetylases (HDACs: vorinostat, romidepsin, and belinostat). Although there is considerable evidence for the anticancer effects of HDAC inhibitors [ 6, 7], recent studies suggest that the majority of HDAC substrates are non-histone, non-epigenetic proteins [ 8, 9]. This review will thus focus on the epigenetic effects of DNMT inhibition by azacitidine (AZA) or decitabine (DAC).
AZA and DAC are prodrug cytosine analogs—a ribonucleoside and a deoxyribonucleoside, respectively—that are incorporated into nucleic acids as azacytosine-guanine pairs following their uptake and phosphorylation [ 10]. DNMTs, which naturally bind CG pairs, become trapped in a covalent bond to DNA when they attempt to methylate the azacytosine. Such adducts result either in cell death at high analog doses or in a DNA damage response at low analog doses that degrades DNMTs and subsequently reactivates genes that were previously silenced by hypermethylation [ 10, 11].
While much interest has focused on the reactivation of tumor suppressor genes by DNMT inhibitors (DNMTi), recent evidence suggests that the inhibition of DNMTs produces cancer cell-extrinsic immunomodulatory effects. This has prompted several groups to study the effects of DNMTi in combination with cancer immunotherapy. Herein, we summarize the mechanisms of DNMTi on antitumor immunity and its synergy with immunotherapy in preclinical and clinical settings.

DNMTi boost tumor immunogenicity

Immune evasion is an emerging hallmark of tumor formation and progression [ 1]. One of the most effective evasion tactics taken by cancer cells is the impairment of antigen presentation. Specifically, levels of the class I major histocompatibility complex (MHC I) are downregulated by irreversible gene-disabling mutations or by reversible inactivating methylation patterns [ 12]. Addressing the latter using DNMTi has yielded promising results in a variety of cancer types. In addition to upregulating MHC I, DNMTi can increase tumor immunogenicity by increasing levels of antigens displayed in MHC I, particularly cancer-testis antigens (CTAs) [ 13, 14]. CTAs are a family of tumor-associated antigens expressed on tumors, but not normal tissues, aside from the testes and placenta. CTA expression is regulated primarily by DNA methylation, so DNMTi treatment increases CTA expression on tumor cells, enabling host cytotoxic T lymphocytes (CTLs) to discriminate them from healthy cells [ 15].
Both DAC and AZA are currently approved for the treatment of patients with myelodysplastic syndromes (MDS). The impact of these DNMTi may extend beyond their cytotoxicity, as they influence antigen presentation in hematologic cells as well. DAC treatment was shown to increase expression of MHC I and II molecules on a chronic lymphocytic leukemia (CLL) cell line and, to a lesser extent, on a primary CLL culture [ 16]. It also induced the de novo gene expression of multiple CTAs, including one of the most immunogenic CTAs, NY-ESO-1, and upregulated baseline levels of others, such as multiple members of the melanoma-associated antigen (MAGE) family [ 16]. Similar results on CTA expression were observed following in vitro treatment of acute myeloid leukemia (AML) cells with DAC, most notably inducing de novo expression of the CTA SSX2 in a majority of cell lines tested [ 17]. These data are supported by translational studies, as clinical administration of AZA induced SSX2 in all eight AML patients examined [ 17]. Similarly, the addition of AZA to an HDAC inhibitor for treatment of Hodgkin lymphoma patients resulted in a broader antitumor T cell repertoire, suggesting an increase in antigen presentation [ 18].
Importantly, there is ample evidence for increased immunogenicity by DMNTi therapy beyond MDS, most notably in melanoma and ovarian cancer, but also in several other cancer types (Table  1). In vitro treatment of a melanoma cell line with DAC upregulated the surface expression of MHC I, increased IFN-γ release by tumor-specific CTLs, and enhanced target cell lysis [ 19]. Similar results were observed with ovarian cancer cell lines and xenograft melanoma models [ 20, 21]. Immunizing mice with DAC-treated melanoma cells induced a significant anti-NY-ESO-1 humoral response over multiple challenges, indicating an antitumor memory response [ 22]. Increased CTA expression was also increased on melanoma patient samples following ex vivo treatment with DAC, notably showing a reduction in intratumor CTA heterogeneity post-treatment [ 23].
Table 1
Summary of cancer/testis antigens induced or upregulated by DNMT inhibition in different cancer types
Cancer/testis antigen
In vitro
In vivo
Ex vivo
Sar [ 33]
CT- 7
Sar [ 33]
CT- 10
Sar [ 33]
CLL [ 16]
CLL [ 16]
GAGE1- 6
MM [ 35]; RCC [ 30]; Sar [ 33]
Mel [ 22]
Mel [ 23]; MM [ 35]
CLL [ 16]
CLL [ 16]
CLL [ 16]
CLL [ 16]
CLL [ 16]
CLL [ 16]
CLL [ 16]; Gli [ 29]; Sar [ 33]
AML [ 17]; CLL [ 16]; Sar [ 28, 30, 49]; MM [ 35]; NB [ 34]; Ov [ 21]; Thy [ 32]; CC, MC, LC, RCC [ 30]
Mel [ 22]
Gli [ 29]; Mel [ 23]; Sar [ 28]
MM [ 35]; RCC [ 30]
Mel [ 22]
Mel [ 23]; MM [ 35];
CLL [ 16]; Gli [ 29]; MM [ 35]; NB [ 34]; Ov [ 21]; Sar [ 28, 30, 33]; RCC, CC, MC, LC [ 30]
Mel [ 22]
Gli [ 29]; Mel [ 23] MM [ 35]; Sar [ 28]
LC, MM [ 36]
CLL [ 16]; MM [ 35]; Ov [ 21]; RCC [ 30]; Sar [ 33]; Thy [ 32]
Mel [ 22]
CLL [ 16]; Gli [ 29]; Mel [ 23]; MM [ 35]
Ov [ 21]
Gli [ 29]; Mel [ 23]
CLL [ 16]
Ov [ 21]; Sar [ 33]
Mel [ 22]
Ov [ 21]
CLL [ 16]
AML [ 17]
AML [ 17]
NY- ESO- 1
AML [ 17]; CLL [ 16]; Gli [ 29]; MM [ 35]; NB [ 34]; Ov [ 21]; Sar [ 28, 30, 33]; Thy [ 25]; CC, MC, LC, Mel, RCC [ 30]
Gli [ 29]; Mel [ 22]; Thy [ 25]
Gli [ 29]; Mel [ 23]; MM [ 35]; Sar [ 28]
LC, MM [ 36]
NY- SAR- 35
CLL [ 16]
AML [ 17]; CLL [ 16]
CLL [ 16]
CC, Hep, LC, Lym, MC, Mel, NB, Sar [ 31]
CC, LLC, MC [ 26]
CLL [ 16]
Mel [ 23]
AML [ 17]
RCC, Sar [ 30]
AML [ 17]
SSX1- 5
Mel [ 23]
CLL [ 16]; Gli [ 29]
CLL [ 16]; Gli [ 29]
AML [ 17]; CLL [ 16]; MM [ 35]
CLL [ 16]; Gli [ 29]; MM [ 35]
AML [ 17]
AML [ 17]; CLL [ 16]; Gli [ 29]
Gli [ 29]
TAG- 1
Ov [ 21]
TAG- 2a
Ov [ 21]
TAG- 2b
Ov [ 21]
Ov [ 21]
TPX- 1
CLL [ 16]
CLL [ 16]
CLL [ 16]
Columns indicate treatment conditions, with ex vivo referring to the treatment of primary culture samples
AML acute myeloid leukemia, CC colon carcinoma, CLL chronic lymphoid leukemia, Gli glioma, Hep hepatoma, LC lung carcinoma, LLC lewis lung carcinoma, Lym lymphoma, MC mammary carcinoma, Mel melanoma, MM mesothelioma, NB neuroblastoma, Ov ovarian carcinoma, RCC renal cell carcinoma, Sar sarcoma, Thy thyroid cancer
P1A is the murine analog of MAGE/GAGE/BAGE genes
Given the preclinical data illustrating the effects of DAC on NY-ESO-1 levels, a phase I dose-escalation trial was initiated to evaluate the ability of DAC to augment vaccination against NY-ESO-1. Sixty percent of evaluable ovarian cancer patients who received the combination of DAC, NY-ESO-1 vaccine, and doxorubicin achieved disease stabilization or partial clinical response [ 24]. The therapy was associated with manageable toxicities, increased NY-ESO-1 serum antibodies, and enhanced antigen-specific T cell responses.
MHC I upregulation by DNMTi has been demonstrated in mammary, lung, colon, and thyroid histotypes [ 25, 26] as well as human papilloma virus (HPV)-associated carcinomas [ 27], sarcomas [ 28], and gliomas [ 29]. DNMTi-mediated increases of levels of various CTAs have also been observed in a majority of these cancers [ 25, 26, 2833] as well as in neuroblastomas [ 31, 34], mesotheliomas [ 35], and renal, esophageal, pleural [ 36], and liver [ 31] cancers (summarized in Table  1). Clearly, downregulation of MHC I and tumor-associated antigens is a fundamental mechanism of tumor immune evasion, and DNMTi can uncloak such tumors.

DNMTi stimulate NK cell- and CD8 T cell-mediated cytotoxicity

In addition to rendering cancer cells more recognizable to T cells in antigen-specific manner, epigenetic modifiers can also enhance cytotoxic natural killer (NK) and T cell function. Specifically, epigenetic modifiers can induce expression of chemokines as well as activating ligands on the surface of tumors cells [ 4, 37]. DAC and valproic acid, an HDAC inhibitor, upregulate MICA, a ligand for natural-killer group 2, member D (NKG2D) expressed by NK cells as well as activated CD8 T cells [ 4, 38]. DAC similarly upregulates the related ligand MICB through promoter DNA demethylation and DNA damage [ 39]. MICA and MICB are stress response ligands and are commonly expressed within tumors.
DAC also intrinsically enhances NK cell antitumor cytotoxicity through various mechanisms. DNA methylation regulates the expression of killer immunoglobulin-like receptors (KIRs), receptors that bind MHC I to allow NK cells to recognize abnormal cells that have downregulated MHC I [ 40]. The epigenetic modulation of KIR is a known mechanism of immune escape in cancer. Restoring proper expression of KIR with immunosensitizing drugs is an attractive approach to improving cancer immunotherapy. Surprisingly, while DAC enhances NK cell responsiveness to activating stimuli by permitting transcription of genes involved in NK cell reactivity, AZA impairs NK cell-mediated cytotoxicity and IFN-γ production [ 41].
DNMTi-mediated demethylation promotes transcription of genes involved in CTL reactivity, most notably antitumor cytokines. For example, the transcription of IL-2 in non-proliferating T lymphocytes is correlated with the demethylation of a promoter-enhancer region of the IL-2 gene upon activation [ 42]. This demethylation pattern of the IL-2 locus (as well as that of IFN-γ) was observed early after CD8 T cell antigen exposure and is maintained throughout CD8 memory development in the presence of CD4 help [ 43]. This suggests that IL-2 levels could be enhanced and maintained via DNMTi-mediated hypomethylation. Indeed, comparing IL-2 production by recent thymic emigrant peripheral murine T cells to their more mature but still naïve counterparts revealed a reduced difference in IL-2 levels upon culturing in the presence of AZA [ 44].
Similarly, culturing primary mouse CD8 T cells in the presence of AZA increased their expression of IFN-γ up to 25-fold. Their expression of IL-3, a T cell growth and differentiation cytokine as well as a myeloid proliferation enhancer, also increased by up to 14-fold in a dose-dependent manner [ 45]. The IFN-γ locus is highly methylated at three specific CpG sites in naïve CD8 T cells, while it is unmethylated in effector T cells, permitting the production of large amounts of IFN-γ. In memory T cells, these sites are partially methylated but rapidly demethylate upon stimulation [ 46], suggesting an activation-dependent demethylation spectrum to control production levels. Indeed, AZA treatment of naïve CD8 T cells increases the number of IFN-γ-producing cells to a quantity that is comparable to the number of IFN-γ-producing memory cells [ 46].
Consistent with these results, following DAC treatment, we observed a significant increase in the recruitment of and antitumor cytokine production by NK cells and CD8 T cells in the ascites of mice bearing orthotopic synergenic ovarian tumors [ 20]. Specifically, the percentage of these cells that produced IFN-γ and TNF-α increased markedly [ 20]. Interestingly, NK cells produced both IFN-γ and TNF-α while, T cells produced either IFN-γ or TNF-α. The enhanced infiltration of NK cells and CD8 T cells into subcutaneous tumors following DAC treatment was also confirmed. The mechanism underlying the propensity of these cytolytic lymphocytes to migrate to the tumor site was evaluated. The ovarian cancer cells were treated with DAC ex vivo, and microarray analysis was performed. Interestingly, the most highly enriched Gene Ontology was that of “immune system process,” and top gene candidates included chemokines and mediators of innate immunity (Fig.  1) [ 20]. All candidates that were analyzed by qPCR were successfully validated, and Irf7 (innate immune signaling), Ccl5 and Cxcl10 (chemokines that attract NK and T cells), and GitrL (a ligand that activates these leukocytes) were among the top hits.
DNMTi can influence the behavior of helper cells as well. Culturing human CD4 T cells in the presence of AZA revealed that transcription of IL17A, a proinflammatory cytokine, is directly induced by the demethylation of its promoter region [ 47]. In vitro AZA treatment of CD4 T cells isolated from 68 MDS patients led to increased IL-17 production by these cells [ 48]. Moreover, the peripheral blood of AML and MDS patients treated with AZA contained significantly increased levels of IL-17A-secreting CD4 T cells [ 49]. Similarly, DAC treatment of helper T cells demethylated Th1-specific promoters, induced naïve cells toward Th1 polarization, and depolarized Th2 cells, thereby promoting production of IFN-γ [ 50].
Taken together, these studies point to the potential use of DNMTi for the enhancement of cytotoxic NK and CD8 T cells as well as helper CD4 cells, particularly through the induction of critical immunostimulatory cytokines.

DNMTi decrease suppression by regulatory adaptive and innate immune cells

Clinical data suggest that the reduction of immunosuppressive regulatory immune cells may be as important as stimulating effector cell-mediated antitumor immunity [ 51, 52]. Notably, in addition to promoting tumor immunogenicity and effector cell function, epigenetic modifiers can decrease natural and tumor-induced immunosuppression. A number of regulatory T cells (Tregs) present in peripheral blood samples from a pool of 68 MDS patients were reduced following treatment with AZA. This result is consistent with the observed reduction in proliferative capacity and suppressive function of Tregs upon AZA treatment of patient CD4 T cells in vitro [ 48].
Levels of myeloid-derived suppressor cells (MDSCs), another major immunosuppressive population often found in the tumor microenvironment, can also be modulated with DNMTi. In the 4T1 mouse model, a combination regimen of DAC, entinostat (an HDAC inhibitor), anti-CTLA-4, and anti-PD-1 dramatically decreased circulating granulocytic MDSCs and brought their level down to those observed in non-tumor-bearing mice. This treatment also significantly decreased intratumoral MDSCs and Tregs [ 53]. However, combination with HDACi is not essential for achieving similar results in other tumor models. In a syngeneic murine ovarian cancer model, DAC treatment reduced the number of tumor-associated MDSCs in the peritoneal cavity [ 20]. Consistently, the frequency of tumor-infiltrating MDSCs was also reduced in the B16 murine melanoma model in response to DAC treatment [ 54].
Similarly, mice bearing subcutaneous TRAMP-C2 (derived from prostatic adenocarcinoma) or TC-1/A9 (derived from primary lung epithelia) tumors showed a decreased percentage of MDSCs in their tumor microenvironments and spleens upon AZA treatment. A higher dose of AZA was also shown to counter the accumulation of MDSCs induced by cyclophosphamide treatment. To elucidate the mechanism underlying these effects, splenic MDSCs cultured in vitro in the presence of AZA showed a reduction in the percentage of CD11b+/Gr-1+ MDSCs and an increase in the percentages of CD11c+ and CD86+/MHCII+ cells. This suggests that DNMTi target MDSCs, at least partially, by inducing their differentiation toward DCs [ 55]. This explanation is supported by earlier research showing that the addition of DAC to a GM-CSF-supplemented culture of tumor-infiltrated CD11b+ myeloid cells promoted the generation of pure CD11b+ CD11c+ mature antigen presenting cells that produced lower levels of immunosuppressive and proinflammatory cytokines [ 56]. Taken together, these studies suggest that the success of DNMTi-based treatments can be partially attributed to their ability to counter lymphocyte- and myeloid-driven immunosuppression within the tumor milieu.

Rationale for combination of DNMTi with immunotherapy

There is also emerging evidence supporting the use of DNMTi to specifically activate a broad variety of other immunomodulatory genes in multiple cancers. Both DAC and AZA induce sustained upregulation of genes related to antigen presentation, inflammation, and immune responses in leukemic and breast cancer cells in vitro [ 57]. Similarly, the treatment of non-small cell lung cancer (NSCLC) lines with AZA enriches gene expression related to antigen presentation, type I and type II interferon signaling, apoptosis, viral defense, and numerous immune-related transcription factors [ 58]. Moreover, a gene expression and methylation analysis of 63 cancer cell lines of breast, colorectal, and ovarian origins revealed similar enrichment for immunostimulatory pathways, such as cytokine signaling and inflammation [ 59]. This ability to widely prime the immune system makes DNMTi a great candidate for combinations with other immunotherapies.
Immune checkpoint blockade is of particularly relevance as a combination to DNMTi. Administering DNMTi alone or in combination with an HDACi to patients previously untreated for MDS or AML resulted in upregulated expression of CTLA-4, PD-1, PD-L1, and/or PD-L2 in a majority of patients, suggesting that antibodies against these targets may compensate for any negative impact that DNMTi may have on antitumor immunity. Of note, responding patients exhibited far lower levels of PD-L1 and PD-L2 expression than those patients who were resistant to therapy, and lower baseline PD-L1 expression correlated significantly with improved overall response rate [ 60]. The CTLA-4 and PD-1 axes are central in orchestrating multiple delicate inhibitory pathways, and their upregulation can dampen T cell activation and enhance tumor immune evasion if not countered by immune checkpoint blockade therapy. PD-1 and CTLA-4 are also implicated in the induction and expansion of Tregs, with inhibition of CTLA-4 resulting in a reduced MDSC frequency [ 61]. Additionally, CTLA-4 [ 62] and PD-1 [ 63] blockade elevate IFN-γ levels at murine tumor sites, further supporting the value of combining immune checkpoint blockade with DNMTi.
As mentioned above, survival of mice bearing 4T1 tumors was significantly enhanced when they were treated with the combination of epigenetic modification and checkpoint blockade relative to epigenetic therapy alone [ 53]. The addition of anti-PD-1 and anti-CTLA-4 also leads to a reduction in levels of circulating granulocytic MDSC [ 53]. These results are supported by our recent study on the synergistic antitumor effect of DAC with anti-CTLA-4 [ 20]. While neither therapy produced durable responses as a monotherapy in the orthotopic syngeneic murine model of ovarian carcinoma that we evaluated, the efficacy of anti-CTLA-4 was potentiated by combination with DAC, drastically extending the survival of mice (Fig.  2). Mechanistically, this synergy was explained by the ability of DAC to increase the expression of chemokines that recruit NK cells and CD8 T cells and to promote the production of IFN-γ and TNF-α by these cells. Furthermore, it was observed that this combination promotes differentiation of naive T cells into effector T cells and prolongs cytotoxic T lymphocyte responses. These data suggest that epigenetic modifiers can induce not only cancer cell-intrinsic effects but also cancer cell-extrinsic effects that alter the localization and function of leukocytes in the tumor microenvironment.

Recent and current clinical trials

The aforementioned evidence has led to multiple clinical trials employing DNMTi in combination with a variety of immune-based therapies. One study (NCT01799083) used low-dose DAC followed by an infusion of cytokine-induced killer cells to treat solid tumors and observed an 80 % objective response rate with minimal adverse effects, compared to a 50 and 72 % rate for DAC alone or DAC combined with chemotherapy, respectively [ 64]. Another recently completed phase I study (NCT01241162) combined DAC with a vaccine cocktail of CTA peptides (NY-ESO-1, MAGE-A1, and MAGE-A3) to treat pediatric neuroblastomas and sarcomas. This combination was well tolerated, with 60 % (6/10) of evaluable patients developing a response to the CTA vaccines and 20 % (2/10) remaining disease-free at least 2 years after their last vaccination [ 65]. This regimen will be repeated in an ongoing trial (NCT02332889, currently recruiting) to treat gliomas, medulloblastomas, and central nervous system primitive neuroectodermal tumors. Another currently recruiting study (NCT01690507) aims to treat elderly MDS and AML patients using DAC combined with a modified CAG regimen (a cocktail of cytarabine, aclacinomycin, and the G-CSF cytokine) followed by an infusion of HLA haploidentical T cells.
Of particular interest is a study (NCT01928576) seeking to combine DNMTi and checkpoint blockade to treat NSCLC patients. Given the low PD-L1 expression in primary NSCLC samples and the fact that AZA upregulates the transcription and protein expression of PD-L1 in vitro [ 58], patients are to be treated by AZA (with or without entinostat) followed by the anti-PD-1 drug nivolumab. Given the encouraging preclinical results reported by us and others, the outlook for more clinical trials combining immune checkpoint blockade and epigenetic modification is promising.


DNMTi have been in clinical development for years, but they have conferred limited utility as cytotoxic agents, particularly against solid tumors. Recent studies suggest that these molecules may be utilized optimally at lower doses as priming agents to enhance the efficacy of immunotherapy. Preclinical data indicate that DNMTi are useful across diverse tumor types, and we hope that translational data will confirm that these compounds can be applied broadly in the clinical setting to break immune tolerance and promote antitumor immunity, particularly in the context of combination with cancer immunotherapy.


We thank the Ovarian Cancer Research Fund (Liz Tilberis Scholar award) and the Susan F. Smith Center for Women’s Cancer for supporting this work.

Compliance with ethical standards

Conflict of interest

The authors declare no conflict of interest.
Acute myeloid leukemia
Chronic lymphocytic leukemia
Cancer testis antigen
Cytotoxic T lymphocyte
DNA methyltransferases
DNA methyltransferase inhibitor
Histone deacetylase
Melanoma-associated antigen
Myelodysplastic syndrome
Myeloid-derived suppressor cell
Major histocompatibility complex class I
NK cells
Natural killer cells
Natural-killer group 2, member D
Non-small cell lung cancer
Killer immunoglobulin-like receptor
Regulatory T cells

Related topics