Trends in Immunology Review The Contribution of Epigenetics to Cancer Immunotherapy Lorea Villanueva,1,2,5 Damiana Álvarez-Errico,1,5 and Manel Esteller 1,2,3,4, * Effective anticancer immunotherapy treatments constitute a qualitative leap in Highlights cancer management. Nonetheless, not all patients benefit from such therapies Recent immunotherapy treatments are because they fail to achieve complete responses, suffer frequent relapses, or showing promising results in cancer management, but a proportion of pa- develop potentially life-threatening toxicities. Epigenomic signatures in immune tients do not fully benefit from therapy and cancer cells appear to be accurate and promising predictors of patient out- and fail to achieve durable responses or comes with immunotherapy. In addition, combined treatments with epigenetic experience relapse. drugs can exploit the dynamic nature of epigenetic changes to potentially modu- Chimeric antigen receptor (CAR) vector late responses to immunotherapy. Candidate epigenetic biomarkers may provide insertion at epigenetic chromosomal a rationale for patient stratification and precision medicine, thus maximizing the sites may determine the potency and chances of treatment success while minimizing unwanted effects. We present durability of CAR-T cell cytotoxic re- a comprehensive up-to-date view of potential epigenetic biomarkers in immuno- sponses, as shown for the complete remission of a chronic lymphocytic leuke- therapy and discuss their advantages over other indicators. mia (CLL) patient with TET2 disruption caused by CAR19 vector integration. Predicting Immunotherapy Success or Failure: The Importance of Developing Reliable ‘Biomarkers’ Novel biomarkers of response are Immunotherapy (see Glossary) has become a game changer in cancer treatment. It relies on needed, and epigenetic biomarkers may represent solid candidates for filling enhancing the patient’s immune defenses to combat tumor cells. However, the significant this niche, thereby contributing to proportion of nonresponding cases and treatment-associated toxicities remain an obstacle to theragnosis and precision medicine. therapeutic success. Therefore, it is important to discover and develop good predictors of Combinatorial approaches of epige- therapy efficacy and/or toxicity such that treatments can be personalized and the chances of a netic drugs with immunotherapy may cure increased. We discuss in depth the latest research on epigenetic marks as biomarkers for synergize to reshape the tumor micro- immunotherapy, their potential and limitations, and the preclinical and clinical studies currently environment and restore an effective being conducted; we provide comprehensive and novel insights into their present and future antitumor response, thus overcoming immunotherapy limitations for some usage in the clinic. cancer types. Cutting-Edge Immunotherapy-Based Strategies and Their Limitations Immunomodulatory agents targeting the interferon (IFN)-α2b pathway or the interleukin (IL)-2/IL-2R axis (aldesleukin) were approved by the US FDA in 1986 and 1992 for the treatment of leukemia 1 Josep Carreras Leukemia Research and melanoma, respectively. Subsequently, the monoclonal antibody (Ab) rituximab, which targets Institute (IJC), Badalona, Barcelona, Catalonia, Spain B cell CD20, was approved by the FDA for non-Hodgkin lymphoma. Since then, Ab-based thera- 2 Centro de Investigación Biomédica en pies have substantially improved with the development of bispecific and trispecific antibodies Red Cancer (CIBERONC), Madrid, Spain 3 [1] that may allow more precise and effective treatment strategies. Therapeutic vaccination against Institucio Catalana de Recerca i Estudis Avançats (ICREA), Barcelona, Catalonia, cancer also emerged as a promising immunotherapeutic strategy at the beginning of the 21st Spain century, leading to FDA approval of the dendritic cell (DC)-based vaccine sipuleucel_T for the treat- 4 Physiological Sciences Department, ment of prostate cancer in 2010. In this context, the arrival of immune checkpoint blockade School of Medicine and Health Sciences, University of Barcelona (UB), (ICB) therapy was a tipping point that created new possibilities for clinical applications. Barcelona, Catalonia, Spain 5 These authors contributed equally Since FDA approval in 2011 of the cytotoxic T-lymphocyte antigen 4 (CTLA-4) targeting ICB ipilimumab to treat metastatic melanoma, many immunotherapeutic approaches targeting immune *Correspondence: checkpoint molecules have been included in clinical practice guidelines for cancer treatments. mesteller@carrerasresearch.org Pembrolizumab and nivolumab [anti-PD-1 (programmed cell death) monoclonal antibodies (Abs)] (M. Esteller). 676 Trends in Immunology, August 2020, Vol. 41, No. 8 https://doi.org/10.1016/j.it.2020.06.002 © 2020 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). Trends in Immunology were approved by the FDA in 2014 and 2015, respectively, for the treatment of various solid and Glossary hematological malignancies. PD-L1 blockers atezolizumab and avelumab were FDA-approved in Apoptosis: a highly regulated 2017 and durvalumab in 2018 for the treatment of bladder, lung, and Merkell cell carcinoma. programmed cell death mechanism that One of the latest FDA-approved ICB molecules is the anti-PD-1 monoclonal Ab cemiplimab for occurs in the absence of cell membrane breakage, ensuring the correct disposal metastatic cutaneous squamous cell carcinoma. Other promising immune receptors such as of dysfunctional cells. LAG-3, TIM-3, TIGIT, VISTA, OX40, CD40, 4-1BB, and GITR are currently under investigation in Bispecific and trispecific antibodies: preclinical studies [2]. genetically engineered monoclonal immunoglobulins that respectively target two or three antigens/epitopes The success of immunotherapy success has extended to adoptive cell therapies with the approval of simultaneously. chimeric antigen receptor (CAR) T cell therapy by the FDA, followed by other promising CAR-T cell-related encephalopathy strategies such as tumor-infiltrating lymphocytes (TILs) and T cell receptor (TCR)- syndrome (CRES): neurotoxic engineered cells that are currently being tested in various advanced clinical trials for different side-effect developed in some patients following CAR-T cell infusion. malignancies (clinicaltrials.gov). Novel CAR-based strategies using natural killer cells (CAR-NK) in Chimeric antigen receptor (CAR) conjunction with lymphodepleting chemotherapy are also showing encouraging results in a single- T cell therapy: an immunotherapy arm dose-escalation Phase I/IIa trial (NCT03056339i, recruiting) in which 73% of patients with approach in which T cells are genetically engineered in the laboratory to recognize non-Hodgkin's lymphoma or chronic lymphocytic leukemia (CLL) responded to treatment [3]. targeted antigens expressed in tumor cells. Nonetheless, beyond tumor intrinsic resistance to ICB, there are innate and adaptive mecha- CpG islands: genomic regions enriched nisms that affect antitumor-specific T cell activation, expansion, and exhaustion [2,4], leading to in cytosine and guanine dinucleotides; CpG islands are located in DNA response failure. In addition, treatment can be associated with potentially life-threatening regulatory regions such as promoters. autoimmune-like toxicities leading to specific organ damage [5]. In general, the incidence Cytokine-release syndrome (CRS): of immune-related adverse effects (irAEs) is higher with combined therapies than with anti- a systemic inflammatory response that is PD-1/anti-PD-L1 Ab (66%) or anti-CTLA-4 Ab monotherapies (72%) [5]. In addition, combined triggered, for example, by CAR-T cell activation following targeted antigen blockade showed a higher fatality rate (1.23%) than anti-PD-1/anti-PD-L1 Abs (0.37%) and recognition. CTLA-4 blockers (1.08%) administered separately, according to a meta-analysis conducted in Deconvolution analysis: a 19 217 ICB-treated patients with 25 types of solid cancers, mainly melanoma and lung cancer computational method to infer the proportion of components and their [6]. Severe-grade toxicities are mostly associated with anti-CTLA-4 therapies, where colitis is the relative contributions to a particular primary cause of irAES-related death, affecting 8–22% of patients with melanoma treated with signal from a complex sample. anti-CTLA-4 Ab [7]. Pulmonary toxicity (5% with anti-PD-1/anti-PD-L1 Abs, or combined treatment Epigenetic biomarker: a modification with anti-CTLA-4 Ab) and hepatitis incidence (3–9% with anti-CTLA-4 Abs) are low, but with poten- or signature used as a qualitative and/or quantitative putative indicator of tially fatal consequences, mainly in melanoma and lung cancer [5]. A meta-analysis of 38 clinical physiological/pathological conditions, or trials carried out in different solid cancers showed that endocrine affectations such as pituitary in response to treatment. or thyroid toxicity can appear after CTLA-4 or combined therapy, leading to gonad dysfunction Epigenetic drug: targets epigenome- [5]. In the case of CAR-T cell therapy, severe adverse toxicities, including cytokine-release related pathological changes to allow the treatment of several disorders, including syndrome [8], and CAR-T cell-related encephalopathy syndrome (CRES) [9], are more cancer. frequent. Thus, beyond most clinical indications and guidelines, it is essential to develop new Epigenetic fingerprints: molecular biomarkers that will enable us not only to predict the intensity and duration of a clinical signatures comprising sets of epigenetic modifications relative to a particular response (if any) but also the risk of developing important adverse effects, and thereby rationalize condition. the use of these expensive customized therapeutic strategies. Exhaustion: a state of cell dysfunction accompanied by progressive loss of Epigenetic Biomarkers: Pros and Cons effector function that occurs after Molecules that act as predictors of the cancer response to immunotherapy (termed biomarkers) persistent antigen stimulation. Immune checkpoint blockade (ICB): are commonly used in personalized cancer immunotherapy: these include PD-L1 expression an immunotherapy that abolishes [10], tumor-associated antigens (TAAs) [11], HLA expression [12], TCR repertoire assessments inhibitory signals in tumor-associated [13], tumor mutational burden and neoantigen identification [14], mismatch repair deficiency immune cells, thus supporting an [15], the presence of TILs [16], and the presence of cells within the tumor microenvironment effective anticancer T cell response. Immunological synapse: contact (TME) that might potentially inhibit antitumor immune responses [e.g., regulatory T cells between an antigen-presenting or target (Tregs), tumor-associated macrophages (TAMs), and/or polymorphonuclear myeloid- cell and a lymphoid lineage cell that derived suppressor cells (PMN-MDSCs)] [4,17], as well as the incidence of genetic alterations orchestrates lymphocyte activation and a subsequent immune response. [18] that are associated with treatment response. The epigenetic control of these events has been Trends in Immunology, August 2020, Vol. 41, No. 8 677 Trends in Immunology extensively demonstrated, supporting the relevance of using specific epigenetic changes (Box 1) Immunotherapy: a therapeutic intervention aiming to enhance the as potential biomarkers for immunotherapy. Cancer cell-intrinsic epigenetic alterations have been patient's own immune system. associated with carcinogenesis [19,20], tumor progression [21,22], and treatment resistance as Lineage switch: a conversion a result of lineage switch in B cell lymphoma and leukemia [23]. Finally, treatment itself may phenomenon affecting hematopoietic cause epigenetic alterations, as evidenced from DNA methylation of relevant genes including lineages: one cell lineage transforms into a different lineage that exhibits distinct CD96, HHLA2, CCR5, CXCR5, and CCL5 in CD8+ T cells purified from healthy donors following lineage-defining markers. immunotherapy [24] (see Outstanding Questions). Mismatch repair deficiency: a cellular defect in mismatch repair (i.e., genetic Epigenetic biomarkers may offer additional advantages, such as low patient invasiveness, mutations in the machinery that corrects mistakes due to DNA replication), given that changes in DNA methylation can be measured in liquid biopsies and body fluids [25]. resulting in the accumulation of They can also convey a layer of information about the life habits and conditions of patients, and mutations by the cell. reveal details about the origin and evolution of a given disease [26]. In some instances, as Polymorphonuclear myeloid- derived suppressor cells discussed later for lung cancer [27], they can serve as diagnostic, prognostic, predictive, and (PMN-MDSCs): immune cells involved therapy-monitoring indicators, facilitating the fostering of theragnosis and precision medicine in an immunosuppressive response. [26]. Despite the great potential of epigenetic modifications as putative biomarkers, some caveats Proto-oncogenes: genes that regulate cell proliferation and differentiation; can potentially drive cancer when mutated or Box 1. Epigenetic Modifications: DNA Methylation, Histone Modifications, and Noncoding RNAs aberrantly expressed. The term epigenetics comprises mitotically heritable changes that affect gene expression without modifying the primary Regulatory T cells (Tregs): a genome DNA sequence. Epigenetic modifications include DNA methylation, modification of histone tails, and noncoding subpopulation of CD4+ T cells that (nc) RNAs. Epigenetic alterations are key in directing the aberrant expression of tumor-associated genes that drive cellular dampen the immune response, malignant transformation and cancer progression. preventing autoimmune diseases or promoting tumor progression by DNA Methylation suppressing antitumor responses. The most abundant mark on DNA occurs via addition of a methyl (CH3) group by DNA methyltransferases (DNMTs) at the 5′ T cell receptor (TCR)-engineered position on the pyrimidine ring of cytosine residues in the context of a CpG dinucleotide. DNA methylation affects transcrip- cells: cells encoding an engineered tion factor binding and controls accessibility to regulatory regions in the DNA, modulating gene expression [85]. Aberrant TCR that has been inserted as part of an modifications of 5-methylcytosine (5mC) profiles can promote uncontrolled cell proliferation and survival, as well as immunotherapeutic strategy. supporting tumor growth. Silencing of tumor-suppressor genes through promoter hypermethylation is a hallmark of many TCR repertoire: TCR antigen-specific cancer types [86]. Removal of 5mCs occurs by sequential oxidation by TET enzymes (in dividing mammalian cells), diversity produced by V(D)J gene generating 5-hydroxymethylcytosine (5hmC), 5-carboxycytosine (5caC), 5-formylcytosine (5fmC), and unmethylated C. rearrangement during T cell development. Accumulating evidence suggests that oxidized intermediate forms could represent bona fide epigenetic marks [87,88]. Th1 cytokines: produced by CD4+ helper T cells associated with Post-translational Histone Modifications proinflammatory responses. Transcriptional regulation: control These control gene expression by modulating chromatin conformation and accessibility. Families of histone-modifying mechanisms of RNA transcription in enzymes target specific residues at histone tails in response to external signals. The most frequent histone modifications response to various signals that include acetylation, methylation, and phosphorylation, but others including citrullination, ubiquitination, ADP-ribosylation, coordinate gene expression. deamination, formylation, O-GlcNAcylation, propionylation, butyrylation, crotonylation, proline isomerization, and the Tumor-associated macrophages recently discovered lactylation have also been described [89]. (TAMs): myeloid cells associated with the TME that modulate the antitumor Acetylation. This mark consists of the reversible addition of an acetyl group (-CH3CO) to lysine residues and is associated response. with gene transcription activation. Histone acetyltransferases (HATs) are chromatin writers that add acetyl groups, Tumor escape: evasion of the immune whereas histone deacetylases (HDACs) are erasers that remove this mark. response by tumor cells owing to loss of immunogenicity or immunosuppressive Methylation. The addition of methyl groups (-CH3) to lysine and/or arginine residues in histone tails is catalyzed by several mechanisms. histone methyl transferases (HMTs), whereas histone demethylases (HDMs) catalyze removal. Depending on the specific Tumor-infiltrating lymphocytes modified residue, methylation marks can be categorized as activating (e.g., H3K4me3) or repressive (H3K9me3, (TILs): immune cells that migrate into H3K27me3). the tumor and can potentially recognize and eliminate cancer cells. Phosphorylation. Histone phosphorylation occurs mainly at serine (S), threonine (T), and tyrosine (Y) residues of histone tails and Tumor intrinsic resistance: is positively associated with transcription and an accessible chromatin conformation. Histone H3 phosphorylation at tyrosine nonresponsiveness to therapy owing to 41 (H3Y41) is enriched at active promoters close to transcription start-sites (TSS) together with the H3K4me3 mark. failure of the immune system to elicit an effective antitumor immune response. ncRNAs Tumor microenvironment (TME): the ncRNAs represent ~90% of human genome-derived RNAs, and include several different types of ncRNA that are classified surroundings of a tumor composed of according to size. ncRNAs N200 bp in length include long ncRNAs (lncRNAs), pseudogene transcripts, and circular RNAs resident and recruited immune cells, (circRNAs). miRNAs, tRNA-derived small RNAs (tsRNAs), and PIWI-interacting RNAs (piRNAs) are ncRNAs b200 bp in fibroblasts, secreted soluble molecules, length. Dysregulation of ncRNA expression contributes to the pathogenesis of many diseases including cancer [90]. blood vessels, and extracellular matrix. 678 Trends in Immunology, August 2020, Vol. 41, No. 8 Trends in Immunology and specific requirements need to be addressed. Candidate epigenetic biomarkers must Tumor mutational burden: the total number of nonsynonymous somatic be tested in homogeneous and reliable patient cohorts (clinical trials). In addition, sensitive mutations accumulated by the tumor cell and precise detection methods need to be developed to facilitate the discovery of new genome. epigenetic fingerprints (see Outstanding Questions). Finally, clinical laboratories must be Whole-genome bisulfate adapted for marker analysis, which should accompany other prognostic and diagnostic sequencing (WGBS): a next- generation sequencing technique; approaches (Box 2). consists of deep sequencing an entire genome after sodium bisulfite Epigenetic Modifications That Modulate Antitumor Immunity: Can They Help To modification to analyze DNA Predict Immunotherapy Success? methylation. The sum of accumulated genetic mutations and aberrant changes in the chromatin landscape disrupts cellular homeostasis, and this can lead to cancer initiation and promote tumor progres- sion. Genetic and epigenetic mechanisms cooperate intricately and can act as cancer drivers. Frequent mutations in cancer include those leading to aberrant expression and/or activity of chromatin-modifying enzymes [26]. Box 2. Laboratory Methods for Analyzing Epigenetic Modifications Methods for studying epigenetic changes can interrogate a single locus or cover the whole genome. DNA Methylation Methylation can be analyzed by methods that include: (i) Bisulfite conversion of DNA. Genomic DNA is treated with sodium bisulfite, which changes unmethylated cytosines to uracil, whereas 5mC is unchanged. The identification and quantification of modified compared with unmodified cyto- sines at single loci can be assessed by pyrosequencing. By contrast, DNA methylation arrays and whole-genome bisulfite sequencing (WGBS) – based on high-throughput next-generation sequencing (NGS) – allow genome-wide (with single-base resolution) or whole-genome interrogation, respectively. (ii) Enzymatic digestion of DNA. This is based on the recognition and fragmentation of genomic DNA by methylation- sensitive restriction enzymes such as MspI, and subsequent analysis of the fragments by NGS. (iii) Reduced representation bisulfite sequencing (RRBS). Combines enzymatic digestion and bisulfite conversion to generate fragments enriched in CpG-rich regulatory regions, such as promoters, that are then analyzed by parallel sequencing, reducing costs compared with sequencing of the entire genome. (iv) Affinity-based methods (such as methylated DNA immunoprecipitation sequencing, MeDIP). These are based on enrichment of selected DNA fragments by pulling down genomic DNA with either antibodies or DNA-binding proteins. The enriched fraction is then analyzed by sequencing. One advantage of affinity-based methods is that they allow interrogation of relevant DNA methylation variants such as 5hmC, as in the 5hMeDIP protocol. Post-translational Histone Modifications These are mainly assessed by immunoprecipitation with specific antibodies (e.g., against histone modifications or histone variants) to pull down protein–DNA complexes. The DNA is then analyzed by PCR, arrays, or massively parallel sequencing (ChIP-PCR, ChIP-on-ChIP, or ChIP-seq, respectively). Chromatin Structure and Accessibility Analysis of these key epigenetic features is informative regarding the molecular mechanisms of pathogenicity. Methods to study accessibility and nucleosome positioning are based on enzymatic digestion and isolation of either nucleosome- bound or accessible regions of DNA coupled to NGS. Micrococcal Nuclease (MNase)-Seq Allows interrogation of nucleosome-protected DNA by digesting nucleosome-free genomic regions with MNase from Staphylococcus aureus. Undigested DNA represents DNA bound to nucleosomes, which is recovered and sequenced. Assay for Transposase-Accessible Chromatin (ATAC)-Seq This method assesses the accessibility of chromatin; it is based on the cutting and addition of adapters to the genomic DNA by Tn5 transposase. In this process, called tagmentation, DNA is amplified by PCR and analyzed by NGS. An open chromatin conformation correlates with the frequency of sequences that map to a given genomic region. Recently, heterogeneity in samples can also be assessed by single-cell ATAC-Seq (scATAC-Seq) protocols that rely on a first step of single-cell separation before ATAC library preparation and sequencing. Trends in Immunology, August 2020, Vol. 41, No. 8 679 Trends in Immunology Epigenetic Changes in Tumor Cells and Tumor-Associated Immune Cells The genetic component of tumorigenesis includes mutations that cause activation of oncogenic genes or inactivation of tumor-suppressor genes. However, at the epigenetic level, transcriptional regulation of tumor-associated genes can drive oncogenesis. Cancerous cells often feature increased transcription of proto-oncogenes as a result of promoter hyperacetylation [28,29], concomitant with repression of tumor-suppressor genes by promoter hypoacetylation and DNA hypermethylation [30]. Many cancer types show a global loss of DNA methylation through- out the genome, with gains of DNA methylation specifically located at CpG islands, thus repressing genes that control cancer growth, such as tumor-suppressor genes [31,32]. Recently, changes in oxidized intermediate 5hmC (5-hydroxymethylcytosine) profiles have been associated with tumor aggressiveness and a less-differentiated phenotype, both in KRAS mutant mouse models of pancreatic ductal adenocarcinoma (including Pdx1-cre;LSL-KrasG12D;Col1a1-TRE- shp53-shRenilla;Rosa26-CAGs-LSL-rtTA-IRES-mKate2) [33] and in some human hematological malignancies such as acute myeloid leukemia (AML) [34]. Accompanying extensive modification of the tumor cell epigenome, reconfiguration of the TME and tumor-driven rewiring of immune cell chromatin landscapes can also modulate the extent and quality of the antitumor immune response, the potential response to immunotherapy, and overall disease outcome. A protumorigenic immunosuppressive TME epigenetically affects immune cells, and can lead to tumor escape [28,35]. From the extent of cellular accumulation of TILs in the TME, tumors may be arbitrarily classified as 'hot', when highly or modestly infiltrated, or 'cold', if they are noninfiltrated tumors [36]. Inflamed 'hot' TMEs are compatible with effective type 1 T helper (Th1) cell IFN-γ-mediated antitumor immune responses, as shown by the correla- tion between the presence of an IFN-responsive genes signature in some tumors (e.g., melanoma, and head and neck squamous cell carcinoma, HNSCC), and a better response to immunotherapy, compared with tumors lacking the IFN-γ signature [37]. For the above reasons, it is crucial to understand how tumors shape an immunosuppressive TME. TILs, particularly cytotoxic CD8+ T cells (CTLs), often display an ‘exhausted’ phenotype resulting from persistent antigen stimulation [38,39]; an immunosuppressive TME can contribute to tumor-specific CTL exhaustion, as shown in a mouse model of tamoxifen-inducible liver cancer (ASTxCre-ERT2) [38]. CD8+ T cell exhaustion features specific chromatin-accessible regions associated with an altered transcriptional profile, including enrichment for genes such as Pdcd1 (encoding PD-1), and the Ifng and Il10 signaling pathways [40]. Finally, concomitant extensive epigenetic reprogramming [41] leading to silencing of genes involved in effector differentiation, such as Ifng, Tcf7, Myc, and Ccr7, have resulted in loss of CD8+ T cell effector function in mice [42,43] (Figure 1). Of relevance, anti-PD-1 Ab therapy can partially reverse CD8+ T cell exhaustion, but achieving a durable antitumor CD8+ T cell effector response has proved challenging [44]. Chromatin states can differentiate dysfunctional tumor- infiltrated PD-1-positive CD8+ T cells, whose exhaustion might be reversed, from those that are potentially refractory to anti-PD-1 Ab therapeutic intervention [45]. Tumor antigen-specific CD8+ T cells are rendered dysfunctional during tumorigenesis in a tamoxifen-inducible hepatocellular carcinoma murine model (ASTxCre-ERT2), and they fail to produce IFN-γ and tumor necrosis factor (TNF)-α as a result of chromatin remodeling, as evidenced from ATAC (assay for transposase- accessible chromatin)-seq data [45]. Such epigenetic reconfiguration can be reverted by ex vivo treatment with IL-15 [45]. Similar chromatin-accessibility profiles have been described in human TILs from melanoma and non-small cell lung tumors (NSCLC) [45]. From another angle, double- stranded RNAs (dsRNAs) can accumulate in cancer cells upon derepression of endogenous retro- viruses (ERVs) by epigenetic drugs such as DNA demethylating agents and lysine demethylase 1 (LSD1) inhibitors in ovarian cancer cell lines [46]. This is relevant because cytoplasmic accumulation of dsRNAs mimics viral infection, triggering an IFN response within the TME and promoting robust antitumoral immunity [46]. In cell lines derived from hematological cancers and ovarian cancer, dual 680 Trends in Immunology, August 2020, Vol. 41, No. 8 Trends in Immunology Trends in Immunology Figure 1. Intrinsic Mechanisms Driving CD8+ T Cell Exhaustion in Cancer. In addition to immunosuppressive factors in the tumor microenvironment (TME), transcriptional regulation and epigenetic restraining through DNA methylation and repressive histone marks can induce a dysfunctional CD8+ T cell state. (A) An increase in transcription factors such as TOX, NRA4, EOMES, and NFAT, that upregulate the expression of inhibitory receptor genes, including Pdcd1, Havcr2, and Lag3 (which encode PD-1, TIM3, and LAG3) among others, is associated with a terminally ‘exhausted’ phenotype [40,42,100]. (B) EZH2, part of the PRC2 H3K27me3 chromatin writer complex, is a key regulator of transcriptional silencing of memory- associated genes that leads to terminally differentiated effector CD8+ T cells [41]. This repressive mark has been associated with tumor-specific CTL dysfunction in pre- and early malignant lesions [38]. DNA methylation enzymes, such as DNMT1 and DNMT3B, are upregulated in exhausted CD8+ T cells [38] and DNMT3A genome-wide de novo methylation can promote terminal exhaustion [43]. Black arrows represent an upregulation of epigenetic enzymes recruitment and red crosses represent transcriptional repression. This figure was created using BioRender (https://biorender.com/). Abbreviations: CTL, cytotoxic T lymphocyte; DNMT, DNA methyltransferase; H3K27me3, histone H3 trimethylated on lysine 27. inhibition of methyltransferase G9a and DNA methyltransferases (DNMTs) caused upregulation of nonoverlapping groups of ERV transcripts, with subsequent induction of viral defense genes such as those encoding IRF7 and STAT1 [46,47]. In colorectal cancer initiating cell lines (CICs), treatment with the demethylating agent 5-AZA (5-aza-2-deoxycytidine) induces IFN response factors such as IRF7 and OASL by prompting dsRNA upregulation and activating the MDA5/MAVS/IRF7 signaling pathway [48]. Thus, targeting of cancer cells with epigenetic drugs can induce viral mimicry leading to activation of IFN responses that promote effective immune responses; these, if combined with checkpoint inhibition therapy, might potentially enhance antitumoral responses and increase therapeutic success. Epigenetic Modifications as Putative Biomarkers for Immunotherapy Several preclinical studies in mice, cell lines, and patient samples propose potential epigenetic biomarkers whose fingerprints are associated with immune evasion or response signatures that can be used as indicators of immunotherapy efficacy (Table 1 and Outstanding Questions). Of note, most studies to date describe biomarkers in bulk tissue preparations, and further information regarding biomarkers at the cellular level is still missing. Deconvolution anal- ysis of transcriptomic and epigenomic data from tumor tissues is a potent tool for identifying the cell types that express dysregulated genes [49]. Indeed, DNA methylation-based deconvolution Trends in Immunology, August 2020, Vol. 41, No. 8 681 Trends in Immunology Table 1. Potential Epigenetic Biomarkers Corresponding to an Immunocompetent Phenotype Putative biomarker Mark Source Malignancy Refs CD8+ T cell infiltration H3K27me3 at Th-1 genes Mouse Ovarian cancer [35] Lysine demethylase 1 (LSD1) Mouse Melanoma [51] inhibition H3K27me3 at CXCL9 Human Glioblastoma [52] Mouse Ovarian cancer CCL5 methylation Human, mouse Melanoma, breast, ovarian, [53] colon, and lung Nuclear receptor-binding SET TCGA HNSCC [54] domain protein 1 (NSD1) Human cell lines inactivation Mouse PDCD1LG2 methylation TCGA Gastric adenocarcinoma [55] Singapore cohort CD8+ T cell exhaustion PRF1 methylation Human (ex vivo) Urothelial bladder cancer [63] (UBC) CD8+ T cell genome-wide Mouse Pan-cancer [43] methylation Immunological synapse H3K27me3 H3K27ac Human Gastric adenocarcinoma [64] and antigen presentation H3K27me3 Mouse Melanoma [57] H3K27me3 and H3K27ac Human Small cell lung cancer [58] at MHCI Human cell lines (SCLC) Mouse HLA, CD40, CD80, CD86 TCGA Pan-cancer [56] methylation Novel ICB treatments ADORA2A methylation TCGA HNSCC [61] TNFRSF9 methylation Human Melanoma [62] analysis revealed that aberrant activation of the transforming growth factor (TGF)-β pathway in cancer-associated fibroblasts could drive immunosuppression in tumors that were originally 'hot', thus preventing a successful response to ICB [50]. Epigenetic silencing by EZH2- mediated histone methylation and DNMT1 DNA methylation can constitute tumorigenic mecha- nisms. In whole-tissue studies, inhibition of the epigenetic modifiers EZH2 and DNMT1 in tumor cells from an ovarian cancer ID8 model in mice increased the expression of Th1 chemokine genes Cxcl9 and Cxcl10 by the tumor [35]. Combined epigenetic treatment and immunotherapy promoted increased CD8+ T cell infiltration and improved the anti-PD-L1 Ab therapeutic effect compared with immunotherapy in the absence of EZH2 and DNMT1 inhibition [35]. Thus, epige- netic modulation may revert the immune evasion caused by epigenetic silencing of Th1 chemokines [35]. Moreover, histone demethylase LSD1 genetic ablation in mouse syngeneic melanoma tumor cells promoted CD4+ and CD8+ T cell tumor infiltration and increased tumor cell immunogenicity in mice bearing LSD1-deficient B16 tumors, relative to their wild-type (WT) counterparts [51]. Furthermore, leukemia inhibitory factor (LIF)-mediated epigenetic silencing of Cxcl9 in TAMs isolated from an ID8 ovarian cancer mouse model correlated with reduced CD8+ T cell tumor infiltration and limited anti-PD-1 Ab efficacy upon LIF blockade [52]. In human ovarian cancer, CCL5 and CXCL9 overexpression leads to CD8+ T cell tumor infiltration and longer patient survival and response to anti-PD-1 Ab ICB compared with nonoverexpressing cells [53]. In addition, epigenetic silencing of Ccl5 by DNA methylation in ID8 mouse ovarian cancer tissue led to impaired TIL recruitment and CXCL9 secretion by TAMs, as evidenced from immunostaining and multispectral microscopy in Cxcl9-expressing tumors compared with Cxcl9low tumors, yielding 'cold' tumors that decrease animal survival and the response to anti- 682 Trends in Immunology, August 2020, Vol. 41, No. 8 Trends in Immunology PD-1 Ab [53]. In addition, The Cancer Genome Atlas (TCGA) identified a subtype of HNSCC, characterized by inactivation of histone methyltransferase nuclear receptor-binding SET domain protein 1 (NSD1), as well as by DNA hypomethylation, showing a 'cold' tumor phenotype with low immune cell infiltration as well as low PD-1 expression in CD8+ T cells, relative to active NSD1- expressing tumors [54]. The authors also demonstrated in a HNSCC xenograft mouse model that Nsd1 knockdown in tumor cells resulted in downregulation of chemoattractant cytokines and tumor immune desertification relative to control transduced WT tumors [54]. Hence, epige- netic mechanisms that promote tumor recruitment of lymphocytes might synergize with subse- quent ICB immunotherapy. The epigenetic status of PDCD1LG2 (encoding PD-L2) in cancer cells also appears to be relevant, at least in some tumor types; specifically, methylation of several CpGs within this locus have been associated not only with CD8+ T cell tumor infiltration but also with a high tumor mutational burden in patients with gastric adenocarcinomas [55]. From another perspective, analysis of TGCA DNA methylation data from 20 human solid tumors revealed that hypomethylation of immunological synapse genes such as HLA, CD40, and CD80 in immuno- genic tumors such as melanoma correlated with effector CD4+ and CD8+ T cell infiltrates modu- lating the tolerogenicity of the TME [56]. Three different murine models of melanoma (B16-F10, RIM3, and NrasQ61KInk4a−/−) have demonstrated upregulation of Ezh2 expression on tumors upon anti-CTLA-4 Ab treatment [57]. Such upregulation results in the epigenetic repression of tumor immunogenic genes, such as MHC-I genes and Cxcl9, reducing antigen processing and presentation to immune cells relative to controls [57]. Accordingly, whole-genome CRISPR/ Cas9 screening in the erythroleukemia cell line K562 showed a repressive role of PRC2 on MHC-I genes in tumor cells, leading to tumor growth and resistance to CTL-mediated tumor cell killing in a murine model of small-cell lung cancer (SCLC). Moreover, EZH2 inhibition allowed restoration of HLA gene expression [58]. In another study, ATAC-seq on sorted IFN-γ-treated melanoma cells revealed an increase of open chromatin regions relative to untreated cells, leading to higher STAT-1 occupancy on loci such as MHC-II and CD274, and driving transcriptomic fea- tures of resistance to anti-CTLA-4 Ab treatment [59]. Indeed, the search for novel, robust, and valid ICB targets is an active field of research. Molecules with immunomodulatory effects, such as Adora 2A and galectin 3, can be epigenetically controlled in some cancers including prostate and HNSCC, respectively [60], and are being proposed as potential biomarkers for patient strat- ification. Specifically, low promoter ADORA2A DNA methylation in tumors from HNSCC patients correlated with immune cell infiltration and overall survival (OS) [61]. Within the immune compart- ment, CD8+ T cells show epigenetic signatures that might help to determine their antitumoral potential upon ICB therapy. For instance, the T cell co-stimulatory receptor TNFRSF9 (also known as 4-1BB/CD137), whose expression is regulated by DNA methylation, may be a promising therapeutic target for melanoma treatment [62]. A recent study in a TCGA cohort of cutaneous melanoma showed that DNA methylation status of the human TNFRSF9 promoter correlated with the extent of lymphocyte (including NK cell) and macrophage infiltration into the tumor [62]. Owing to the regulatory role of DNA methylation on TNFRSF9 expression in tumor- infiltrating immune cells, and given that TNFRSF9 promotes immune activation, these findings suggest a layer of epigenetic regulation of TNFRSF9 expression that contributes to enhanced lymphocyte recruitment and activation by tumors, thereby promoting antitumoral responses [62]. Finally, increased expression and promoter hypomethylation of TNFRSF9 correlated signifi- cantly with higher progression-free survival (PFS) and robust response to anti-PD-1 Ab therapy in melanoma patients, supporting its possible suitability as a predictive biomarker of the response to immunotherapy [62]. Analysis by whole-genome bisulfate sequencing (WGBS) of isolated virus-specific CD8+ T cells from a mouse model of chronic infection with lymphocytic choriomeningitis virus (LMCV) strain clone 13 revealed the acquisition of exhaustion-associated DNA-methylation programs through de novo Trends in Immunology, August 2020, Vol. 41, No. 8 683 Trends in Immunology DNMT3a-mediated methylation of pivotal genes such as Ifng, Myc, Tcf7, Ccr7, and Tbx21 [43]. The same differentially methylated regions were observed in tumor-infiltrating PD-1highTim3+CD8+ T cells isolated from a TRAMP-C2 prostate tumor mouse model. Of relevance, combined PD-L1 blockade and treatment with DNA demethylating agents reverted the exhaustion program and in- duced proliferation of tumor-infiltrating PD1+CD8+ T cells, thus reducing tumor growth in these mice compared with PD-L1 monotherapy [43]. Recently developed single-cell technologies allowing the interrogation of the heterogeneity of complex samples are shedding light on the contri- butions of different cell types to cancer responses to immunotherapy. In one study, single-cell transcriptomic analysis of CD8+ TILs revealed two cellular subsets, progenitor and terminally exhausted, that differed in their chromatin accessibility and transcription of Slamf6 and Havcr2 (encoding Tim-3), and which largely determined the cytotoxic capacity of these cells in response to PD-1 blockade in mice bearing B16-OVA tumors [40]. In another study, DNA methylation of sorted exhausted CD8+ tumor tissue-resident memory T cells (Trms) from urinary bladder cancer patients was interrogated. Tumor-infiltrating Trm cells showed that low DNA methylation in an enhancer region of the PRF1 locus correlated with high expression of the perforin protein, and allowed restoration of Trm antitumoral cytotoxic activity upon TCR ex vivo stimulation in the pres- ence of IL-15 [63]. Furthermore, single-cell transcriptional profiling identified a population of dys- functional PD-1highCD8+ T cells with reduced chromatin accessibility in TCF7 regions; these cells were associated with exhaustion and a failed response to checkpoint blockade in nonresponder melanoma patients relative to responders [42]. Such evidence may support the rationale for using epigenetic ‘biomarkers’ and transcriptional programs to delineate therapeutic strategies that might drive or complement anticancer treatments [42]. Epigenetic modifications relevant to the immunotherapy response have also been identified in patients treated with immunotherapeutic agents (Table 2). Comparison of histone modification pro- files [histone H3 trimethylated or monomethylated on lysine 4 (H3K4me3, H3K4me1) or acetylated on lysine 27 (H3K27ac)] in promoters of human primary gastric tumors, gastric cancer cell lines, and normal gastric tissue revealed altered chromatin profiles in somatic promoters that led to dysfunctional antigen presentation and tumor depletion of immunogenic peptides, hence re- ducing tumor antigenicity, strongly suggesting an epigenetically controlled mechanism of immune evasion of immunotherapy [64]. Early studies showed that tumor PDCD1LG1 (coding for PD-L1) promoter methylation was associated with anti-PD-1 Ab resistance in patients with NSCLC refractory to treatment with EGFR tyrosine kinase inhibitors, and who were treated with nivolumab (anti-PD-1 Ab) [65]. The value of unambiguously assessing epigenetic biomarkers to facilitate the pre- diction of response to immunotherapy was first demonstrated in matched pretreatment and post-ICB samples in a cohort of melanoma patients. Specifically, lower CTLA4 methylation in malignant mela- noma tissue samples from a TGCA cohort was associated with a stronger response to anti-PD-1 and Table 2. Epigenetic Hallmarks Identified in a Cohort of Cancer Patients Treated with Immunotherapy Epigenetic mark Malignancy Treatment Outcome Refs PD-L1 methylation NSCLC Anti-PD-1 Ab Anti-PD-1 Ab resistance [65] Low CTLA-4 methylation Melanoma Anti-PD-1 Ab Response to therapy and overall survival [66] Anti-CTLA-4 Ab H3K27ac/H3K4me3 Gastric cancer Anti-PD-1 Ab Resistance to anti-PD-1 Ab [70] TET2 disruption CLL CART19 cell Enhanced CAR-T cell activity and [68] complete remission FOXP1 demethylation NSCLC Anti-PD-1 Ab Clinical benefit with PD-1 blockade [27] and EPIMMUNE epigenetic signature Global low methylation Lung cancer Anti-PD-1 Ab Poor prognosis and lower tumor immunity [67] 684 Trends in Immunology, August 2020, Vol. 41, No. 8 Trends in Immunology anti-CTLA-4 Ab therapy and increased OS compared with the non-ICB cohort [66]. Subsequent studies have highlighted the role of epigenetics in predicting or modulating the immune response to immunotherapy. A multicenter study conducted in tumor tissue from NSCLC patients treated with anti-PD-1 Ab compared the presence and absence of an epigenetic signature (named EPIMMUNE), and showed that the presence of EPIMMUNE was associated with improved progression-free sur- vival (PFS) and OS, thus proposing EPIMMUNE as a good predictor of immunotherapy response [27]. A second independent study described a global DNA-methylation signature in TGCA tissue samples of lung cancer and melanoma that correlated loss of DNA methylation with tumoral CD8+ T cell infiltration, revealing poor prognosis associated with weaker tumor immunity that was indepen- dent of mutation burden [67]. Moreover, disruption of the ten-eleven translocation methylcytosine dioxygenase 2 (TET2) gene in a single clone of CAR-T cells was associated with enhanced cytotoxic activity of a CAR-T against CD19 antigen (CART19) upon ex vivo restimulation. Infusion of CART19 in a CLL patient showed clonal expansion of CART19 cells and a potent antitumoral response that led to complete remission in that patient [68] (see Outstanding Questions). These observations emerged from a completed nonrandomized pilot/Phase I study (single-arm) in patients with CD19+ leukemia and lymphoma (NCT01029366ii). Patient outcome correlated with vector integration site, including proximity to epigenetic chromosomal features such as the repressive mark H4R3me2 and the chro- matin hyperacetylation-associated gene BRD3 [69]. In another example, transcriptomic data revealed that the extent of alternative promoter usage at 2732 sites in gastric cancer biopsies correlated with immune evasion and resistance to anti-PD-1 Ab immune checkpoint inhibition in gastric cancer pa- tients [70], suggesting that alternative promoter usage might have potential as a predictive biomarker for immunotherapy. In conclusion, the leverage of epigenetic changes controlling tumor-associated immune responses to reinforce immunotherapy may be a promising route for therapeutic intervention in cancer patients and certainly deserves further robust investigation because many unknowns remain, particularly when attempting to predict responses across different tumor types. In particular, given the need to associate reliable biomarkers with immunotherapy success, epigenetic changes that might accompany and/or be predictive of treatment outcomes are desirable, and may be relevant to the development of precise diagnostic and stratification strategies (Figure 2). Epigenetic Drugs To Boost Immunotherapy-Based Strategies The highly dynamic nature of epigenetic regulation has been harnessed from a therapeutic point of view through the development of drugs (epidrugs) that target the chromatin-modifying machinery and affect epigenetic changes (Box 3). The reversibility of epigenetic modifications has enabled pharmacological interventions with DNA hypomethylating agents (HMAs) and histone deacetylase inhibitors (HDACis). Clinical and Preclinical Trials Combining Epidrugs and Immunotherapy A challenging aspect of cancer management is the development of resistance to treatment, and this is a major limitation of many anticancer therapeutic approaches. The immunomodulatory effect of epigenetic drugs can be leveraged to enhance the antitumor effect of immunotherapy, as shown in transplantable murine models of mammary carcinoma and mesothelioma where combining the DNA HMA 5-aza-2′-deoxycytidine (5-AZA-CdR) with an anti-CTLA-4 monoclonal Ab led to increased antitumor activity relative to controls [71]. This study set the rationale for a single- arm Phase Ib trial using the DNA HMA guadecitabine combined with ipilimumab (anti-CTL4 Ab) in patients with unresectable stage III/IV melanoma (NCT02608437iii). Treatment safety and tolerance were assessed in the trial, together with treatment-related immunomodulatory effects including activation of immune-related pathways, upregulation of HLA-I by melanoma cells, and in- creased numbers of CD8+PD-1+ T cells and CD20+ B cells at the tumor site post-treatment Trends in Immunology, August 2020, Vol. 41, No. 8 685 Trends in Immunology Trends in Immunology Figure 2. Epigenetic Mechanisms in the Tumor Microenvironment (TME) That Are Relevant to Cancer Immunotherapy. DNA and histone methylation can shape the TME by regulating immune-related gene expression. DNA and histone methylation by DNMT1 and EZH2, respectively, result in downregulation of the chemoattractant cytokine genes Cxcl9 and Cxcl10, thereby dampening CD8+ T cell recruitment to mouse ovarian tumors [35]. Interferon (IFN)-γ- induced Cxcl9 secretion by antigen-presenting cells (APCs) is impaired because of decreased CD8+ T cell infiltration upon Ccl5 methylation in tumor cells [53]. Leukemia inhibitory factor (LIF)-mediated EZH2 recruitment to the Cxcl9 promoter triggers its epigenetic silencing in tumor-associated macrophages (TAMs) [52]. Immunological synapse formation and antigen presentation, which are necessary for an effective antitumor cytotoxic response, may be epigenetically silenced in tumor cells by methylation [58]. Finally, methylation of genes in programmed cell death 1 (PD-1)+CD8+ T cells can induce an ‘exhausted’ state that is refractory to effector response rescue with anti-PD-1 antibody therapies [43]. Red crosses represent transcriptional repression. This figure was created using BioRender (https://biorender.com/). Abbreviations: CK, cytokine; DC, dendritic cell; DNMT, DNA methyltransferase; Th1, type 1 T helper cell. relative to baseline [72]. Treatment with hypomethylating drugs has also been shown to cause upregulation of CD274, PDCD1LG2, and CTLA4 in myelodysplastic syndrome (MDS) and AML CD34+ cells; patients resistant to treatment with the demethylating agent decitabine showed higher expression of such genes relative to sensitive patients [73]. In addition, PDCD1 promoter demethylation in CD4+ and CD8+ T cells from MDS or AML patients after 5-AZA treatment tended to show a poorer response to treatment and shorter overall response (OR) than patients with a methylated PDCD1 promoter [74]. This has established a rationale for 686 Trends in Immunology, August 2020, Vol. 41, No. 8 Trends in Immunology Box 3. Epidrugs Currently Accepted for Clinical Use Outstanding Questions DNMT Inhibitors Are we putting the ‘CART’ before the Hypomethylating agents (HMAs) currently approved for clinical use are the nucleotide analogs 5-aza-2-deoxycytidine (5- horse? Is it necessary to further AZA) and decitabine (5-AZA-2′-deoxycytidine, 5-AZA-CdR). These molecules are incorporated into the DNA of dividing characterize the human epigenome cells, causing irreversible binding and degradation of DNMTs, and subsequent activation of DNA damage response before immunotherapy administration (DDR) pathways [91]. The activation of the DDR triggers apoptosis of highly proliferative cells, promoting cytotoxicity in because treatment itself may cause addition to the epigenetic effect of HMAs. epigenetic alterations in tumor cells, and because most immunotherapy HDAC Inhibitors treatments are administered as Vorinostat (suberoylanilidehydroxamic acid, SAHA) inhibits the enzymatic activity of class I and II HDACs [92]. It was first second-line therapy? Could this new approved by the FDA in 2006 for the treatment of progressive, persistent, or recurrent (relapsed/refractory) cutaneous knowledge contribute to new combi- T cell lymphoma (CTCL) [93]. Romidepsin (cyclic peptide and depsipeptide) blocks HDAC activity, causing apoptosis. It natorial strategies to overcome immu- was approved by the FDA for CTCL in 2007 and for peripheral T cell lymphoma (PTCL) in 2011. The hydroxamate-type notherapy resistance? class I, II, and IV HDAC inhibitor belinostat was approved for relapsed or refractory PTCL in 2014. Finally, in 2015, the FDA approved panobinostat, a nonselective pan-HDAC inhibitor (in combination with bortezomib and dexamethasone) Is it worth studying DNA methylation for the treatment of multiple myeloma (MM). changes in patient-derived CD8+ T lymphocytes induced by immunother- Emerging Epigenetic-Based Therapies apy as a means of monitoring bio- markers of immunotherapy response? Drug discovery devoted to the search for novel compounds that specifically target or reverse cancer-promoting epigenetic modifications is a fast-expanding field. Upon screening with a library of several epigenetic targeting compounds, an aurora Should efforts be devoted to identifying kinase inhibitor has been shown to kill thyroid cancer cell lines efficiently, making it a potentially strong candidate for novel unique universal biomarkers for each epigenetic treatment of this cancer type [94]. A newly developed mutant IDH1 inhibitor (DS-1001b) has shown promising type of cancer, or should they instead antitumor properties in chondrosarcoma cell lines [95]. Ricolinostat, the first selective HDAC6 inhibitor [96], is currently be devoted to defining 'response being tested in a Phase Ib/Phase II clinical trial to treat MM in combination with pomalidomide and dexamethasone. This signatures' for patient stratification? ongoing open-label trial recruited a total of 101 patients, with one study arm to assess safety and best dose (Phase Ib) as well as overall response to treatment (Phase II) in relapsed and refractory MM (NCT01997840xiii) Other small molecules that How much do epigenetic tests reflect selectively inhibit HDAC6 are showing greater efficacy and promising results in preclinical settings [97,98]. Novel experi- tumor heterogeneity? Does it make mental approaches to promote targeted epigenome editing, such as directed integration of CpG-free DNA to induce de more sense to aim for single-cell reso- novo DNA methylation and to allow the correction of abnormal epigenetic modifications on disease-causing loci, may have lution? The TME is important in treat- therapeutic potential and warrant further investigation [99]. ment resistance, but cell clonality could be the key to understanding specific epigenetic changes that drive treatment failure or success. Could evaluating the success of combinatorial therapy with HMAs and ICB in AML and MDS patients this provide a rationale for target inser- [75,76]. Moreover, treatment with CDK4/6 inhibitors reduced DNMT1 mRNA and protein tion strategies in the design of CAR-T cell therapies? expression, causing hypomethylation of immune-related genes (such as those encoding STATs and IRFs), in mouse tumor models (MMTV-rtTA/tetO-HER2, MMTV-PyMT) and pa- Will technology allow the use of tients with breast and colon carcinoma, enhancing antitumor immunity by both promoting epigenetic biomarkers in routine antigen presentation and reducing Treg cell expansion relative to controls [77]. The combina- clinical settings to predict or monitor immunotherapy response? It is still tion of CDK4/6 inhibitors with anti-PD-L1 Ab therapy in these mouse models led to tumor re- necessary to simplify the methods gression and resistance to rechallenge with tumoral cells compared with untreated mice [77]. and decrease the costs of measuring epigenetic biomarkers. In line with this, a novel dual inhibitor that targets both G9a and DNMT has been used in combi- Is the medical community prepared nation with anti-PD-L1 Ab therapy in a mouse transgenic model of bladder cancer (PtenloxP/loxP; to adopt the use of epigenetic Trp53loxP/loxP; Rb1loxP/loxP; Rbl1−/−), and enhanced the antitumor response by increasing CD8+ T biomarkers as a means to develop cell and NK cell tumor infiltration and inducing tumor regression relative to both untreated mice precision medicine? How can the use of epigenetic biomarkers be exploited and anti-PD-L1-Ab-only treated mice [78]. In the case of HDAC inhibitors, entinostat is a synthetic so that these are translated from the benzamide that selectively inhibits class I and IV HDACs and is being investigated in clinical trials in bench to the bedside? combination with anti-PD-1 Ab therapy (pembrolizumab) for colorectal cancer and ICB-resistant melanoma [79–81] and NSCLC [82–84]. Figure 3 summarizes selected ongoing clinical trials. Concluding Remarks Immunotherapy treatments have led to a major breakthrough on the road to monitoring, control- ling, and eliminating particular malignancies. However, because every cancer is unique, efforts must be concentrated on disentangling the way in which each tumor type manifests at the molecular level and, consequently, how treatments can affect the development and potential Trends in Immunology, August 2020, Vol. 41, No. 8 687 Trends in Immunology Trends in Immunology Figure 3. Timeline of the Combinatorial Use of Epigenetic Drugs with Immunotherapy in Clinical Trials. Examples of ongoing clinical trials using different combinations tested for solid tumor and hematological malignancies are shown. Drug names of DNA methyltransferase inhibitors (DNMTis) and histone deacetylase inhibitors (HDACis), as well as immune checkpoint inhibitor molecules, are shown. Cancer types treated in the trial are illustrated with respect to the affected organ. The state of the trial (recruiting, active, or completed) is indicated. Overall response (OR), progression-free survival (PFS), partial response (PR), and stable disease (SD) are shown for each clinical trial according to published data (NCT02890329iv, NCT02638090v, NCT02437136vi, NCT02508870vii, NCT02397720viii, NCT02845297ix, NCT03765229x, NCT02775903xi, NCT02260440xii) [75,76,79–81,83]. Abbreviations: AML, acute myeloid leukemia; 5-AZA, 5-aza-2-deoxycytidine; CRC, colorectal cancer; CTLA-4, cytotoxic T-lymphocyte antigen 4; MDS, myelodysplastic syndrome; NSCLC, non-small cell lung cancer, PD-1, programmed cell death 1. eradication of each individual tumor. This knowledge must be translated into updated medical guidelines if specific and effective therapies for different cancer types are to be deployed. One way to implement personalized medicine for cancer relies, in part, on identifying robust bio- markers that might not only predict response to immunotherapy but also factor in individual differences and dynamic changes caused by the disease course itself and/or by the treatment. In this respect, epigenetic biomarkers, including epigenetic signatures, might represent sensitive predictors of the predisposition of each patient for response to treatment. As such, they might offer valuable mechanistic information to potentially prevent or interfere with a desired therapeutic response. Nonetheless, epigenetic and novel immunotherapy treatments are both nascent tools in clinical practice, and further research will be necessary to discover and develop reliable epige- netic biomarkers. By leveraging this knowledge, new-generation epigenetic drugs suitable for combining with immunotherapy might be designed (see Outstanding Questions). We emphasize that the availability of effective epigenetic biomarkers and new combinatorial approaches are likely to increase the chances of success in cancer therapeutics. Acknowledgments We thank the CERCA (Centres de Recerca de Catalunya) Programme/Generalitat de Catalunya for institutional support. This work was supported by the Health Department PERIS (Plan Estratègic de Recerca i Innovació en Salut) project SLT/002/16/ 00374 and AGAUR (Agència de Gestió d'Ajuts Universitaris i de Recerca) projects 2017SGR1080 of the Catalan Govern- ment (Generalitat de Catalunya); the Ministerio de Ciencia e Innovación (MCI), Agencia Estatal de Investigación (AEI), and 688 Trends in Immunology, August 2020, Vol. 41, No. 8 Trends in Immunology European Regional Development Fund (ERDF) project RTI2018-094049-B-I00; the Cellex Foundation; and La Caixa Bank- ing Foundation (LCF/PR/GN18/51140001). Disclaimer Statement M.E. is a consultant for Ferrer International and Quimatryx. Resources i https://clinicaltrials.gov/ct2/show/NCT03056339 ii https://clinicaltrials.gov/ct2/show/NCT01029366 iii https://clinicaltrials.gov/ct2/show/NCT02608437 iv https://clinicaltrials.gov/ct2/show/NCT02890329 v https://clinicaltrials.gov/ct2/show/NCT02638090 vi https://clinicaltrials.gov/ct2/show/NCT02437136 vii https://clinicaltrials.gov/ct2/show/NCT02508870 viii https://clinicaltrials.gov/ct2/show/NCT02397720 ix https://clinicaltrials.gov/ct2/show/NCT02845297 x https://clinicaltrials.gov/ct2/show/NCT03765229 xi https://clinicaltrials.gov/ct2/show/NCT02775903 xii https://clinicaltrials.gov/ct2/show/NCT02260440 xiii https://clinicaltrials.gov/ct2/show/NCT01997840 References 1. Kantarjian, H. et al. (2017) Blinatumomab versus chemother- 15. Le, D.T. et al. (2017) Mismatch repair deficiency predicts apy for advanced acute lymphoblastic leukemia. N. Engl. response of solid tumors to PD-1 blockade. Science 357, J. Med. 376, 836–847 409–413 2. Marin-Acevedo, J.A. et al. (2018) Next generation of immune 16. Tumeh, P.C. et al. (2014) PD-1 blockade induces responses checkpoint therapy in cancer: new developments and chal- by inhibiting adaptive immune resistance. Nature 515, lenges. J. Hematol. Oncol. 11, 39 568–571 3. Liu, E. et al. (2020) Use of CAR-transduced natural killer cells in 17. Neubert, N.J. et al. (2018) T cell-induced CSF1 promotes CD19-positive lymphoid tumors. N. Engl. J. Med. 382, melanoma resistance to PD1 blockade. Sci. Transl. Med. 10, 545–553 eaan3311 4. Sade-Feldman, M. et al. (2016) Clinical significance of circulating 18. Wang, M. et al. (2019) Role of tumor gene mutations in treat- CD33+CD11b+HLA−DR− myeloid cells in patients with stage IV ment response to immune checkpoint blockades. Precis. melanoma treated with ipilimumab. Clin. Cancer Res. 22, Clin. Med. 2, 100–109 5661–5672 19. De Carvalho, D.D. et al. (2012) DNA methylation screening 5. Kennedy, L.B. and Salama, A.K.S. (2020) A review of cancer identifies driver epigenetic events of cancer cell survival. immunotherapy toxicity. CA Cancer J. Clin. 70, 86–104 Cancer Cell 21, 655–667 6. Wang, D.Y. et al. (2018) Fatal toxic effects associated with 20. Yoshimi, A. et al. (2019) Coordinated alterations in RNA splicing immune checkpoint inhibitors: a systematic review and meta- and epigenetic regulation drive leukaemogenesis. Nature 574, analysis. JAMA Oncol. 4, 1721–1728 273–277 7. Gupta, A. et al. (2015) Systematic review: colitis associated 21. Orouji, E. et al. (2019) Histone methyltransferase SETDB1 with anti-CTLA-4 therapy. Aliment. Pharmacol. Ther. 42, contributes to melanoma tumorigenesis and serves as a new 406–417 potential therapeutic target. Int. J. Cancer 145, 3462–3477 8. Fitzgerald, J.C. et al. (2017) Cytokine release syndrome after 22. Giacopelli, B. et al. (2019) Developmental subtypes assessed chimeric antigen receptor T cell therapy for acute lymphoblastic by DNA methylation-iPLEX forecast the natural history of leukemia. Crit. Care Med. 45, e124–e131 chronic lymphocytic leukemia. Blood 134, 688–698 9. Schuster, S.J. et al. (2017) Chimeric antigen receptor T Cells in 23. Bueno-Costa, A. et al. (2020) B-cell leukemia transdifferentiation refractory B-cell lymphomas. N. Engl. J. Med. 377, 2545–2554 to macrophage involves reconfiguration of DNA methylation for 10. Topalian, S.L. et al. (2012) Safety, activity, and immune corre- long-range regulation. Leukemia 34, 1158–1162 lates of anti-PD-1 antibody in cancer. N. Engl. J. Med. 366, 24. Aznar, M.A. et al. (2018) CD137 (4-1BB) costimulation modifies 2443–2454 DNA methylation in CD8+ T cell-relevant genes. Cancer 11. Shukla, S.A. et al. (2018) Cancer-germline antigen expression Immunol. Res. 6, 69–78 discriminates clinical outcome to CTLA-4 blockade. Cell 173, 25. Liu, L. et al. (2018) Targeted methylation sequencing of plasma 624–633 cell-free DNA for cancer detection and classification. Ann. 12. Rodig, S.J. et al. (2018) MHC proteins confer differential sensi- Oncol. 29, 1445–1453 tivity to CTLA-4 and PD-1 blockade in untreated metastatic 26. Berdasco, M. and Esteller, M. (2019) Clinical epigenetics: melanoma. Sci. Transl. Med. 10, eaar3342 seizing opportunities for translation. Nat. Rev. Genet. 20, 13. Han, J. et al. (2020) TCR repertoire diversity of peripheral 109–127 PD-1+CD8+ T cells predicts clinical outcomes after immuno- 27. Duruisseaux, M. et al. (2018) Epigenetic prediction of response to therapy in patients with non-small cell lung cancer. Cancer anti-PD-1 treatment in non-small-cell lung cancer: a multicentre, Immunol. Res. 8, 146–154 retrospective analysis. Lancet Respir. Med. 6, 771–781 14. Rizvi, N.A. et al. (2015) Mutational landscape determines sensi- 28. Kugel, S. et al. (2015) Identification of and molecular basis for tivity to PD-1 blockade in non-small cell lung cancer. Science SIRT6 loss-of-function point mutations in cancer. Cell Rep. 348, 124–128 13, 479–488 Trends in Immunology, August 2020, Vol. 41, No. 8 689 Trends in Immunology 29. Kugel, S. et al. (2016) SIRT6 suppresses pancreatic cancer 56. Berglund, A. et al. (2020) Methylation of immune synapse through control of Lin28b. Cell 165, 1401–1415 genes modulates tumor immunogenicity. J. Clin. Invest. 130, 30. Ma, J. et al. (2010) Promoter hypermethylation and histone 974–980 hypoacetylation contribute to pancreatic-duodenal homeobox 57. Zingg, D. et al. (2017) The histone methyltransferase Ezh2 controls 1 silencing in gastric cancer. Carcinogenesis 31, 1552–1560 mechanisms of adaptive resistance to tumor immunotherapy. Cell 31. Feinberg, A.P. and Vogelstein, B. (1983) Hypomethylation Rep. 20, 854–867 distinguishes genes of some human cancers from their normal 58. Burr, M.L. et al. (2019) An evolutionarily conserved function of counterparts. Nature 301, 89–92 polycomb silences the MHC class I antigen presentation pathway 32. Lister, R. et al. (2009) Human DNA methylomes at base resolution and enables immune evasion in cancer. Cancer Cell 36, 385–401 show widespread epigenomic differences. Nature 462, 315–322 59. Benci, J.L. et al. (2016) Tumor interferon signaling regulates a multi- 33. Morris, J.P. et al. (2019) α-Ketoglutarate links p53 to cell fate genic resistance program to immune checkpoint blockade. Cell during tumour suppression. Nature 573, 595–599 167, 1540–1554 34. Cimmino, L. et al. (2017) Restoration of TET2 function blocks 60. Ahmed, H. et al. (2009) Evidence of heavy methylation in the aberrant self-renewal and leukemia progression. Cell 170, galectin 3 promoter in early stages of prostate adenocarci- 1079–1095 noma: development and validation of a methylated marker for 35. Peng, D. et al. (2015) Epigenetic silencing of TH1-type early diagnosis of prostate cancer. Transl. Oncol. 2, 146–156 chemokines shapes tumour immunity and immunotherapy. 61. Vogt, T.J. et al. (2018) Detailed analysis of adenosine A2a Nature 527, 249–253 receptor (ADORA2A) and CD73 (5′-nucleotidase, ecto, NT5E) 36. Derks, S. et al. (2020) Characterizing diversity in the tumor- methylation and gene expression in head and neck squamous immune microenvironment of distinct subclasses of gastro- cell carcinoma patients. Oncoimmunology 7, e1452579 esophageal adenocarcinomas. Ann. Oncol. Published online 62. Fröhlich, A. et al. (2020) Comprehensive analysis of tumor May 6, 2020. https://doi.org/10.1016/j.annonc.2020.04.011 necrosis factor receptor TNFRSF9 (4-1BB) DNA methylation 37. Ayers, M. et al. (2017) IFN-γ-related mRNA profile predicts clinical with regard to molecular and clinicopathological features, response to PD-1 blockade. J. Clin. Invest. 127, 2930–2940 immune infiltrates, and response prediction to immunotherapy 38. Schietinger, A. et al. (2016) Tumor-specific T cell dysfunction is in melanoma. EBioMedicine 52, 102647 a dynamic antigen-driven differentiation program initiated early 63. Hartana, C.A. et al. (2018) Tissue-resident memory T cells are during tumorigenesis. Immunity 45, 389–401 epigenetically cytotoxic with signs of exhaustion in human 39. Wherry, E.J. et al. (2007) Molecular signature of CD8+ T cell urinary bladder cancer. Clin. Exp. Immunol. 194, 39–53 exhaustion during chronic viral infection. Immunity 27, 670–684 64. Qamra, A. et al. (2017) Epigenomic promoter alterations 40. Miller, B.C. et al. (2019) Subsets of exhausted CD8+ T cells dif- amplify gene isoform and immunogenic diversity in gastric adeno- ferentially mediate tumor control and respond to checkpoint carcinoma. Cancer Discov. 7, 630–651 blockade. Nat. Immunol. 20, 326–336 65. Zhang, Y. et al. (2017) PD-L1 promoter methylation mediates the 41. Gray, S.M. et al. (2017) Polycomb repressive complex 2- resistance response to anti-PD-1 therapy in NSCLC patients with mediated chromatin repression guides effector CD8+ T cell EGFR-TKI resistance. Oncotarget 8, 101535–101544 terminal differentiation and loss of multipotency. Immunity 46, 66. Goltz, D. et al. (2018) CTLA4 methylation predicts response to 596–608 anti-PD-1 and anti-CTLA-4 immunotherapy in melanoma pa- 42. Sade-Feldman, M. et al. (2018) Defining T cell states associated tients. JCI Insight 3, e96793 with response to checkpoint immunotherapy in melanoma. Cell 67. Jung, H. et al. (2019) DNA methylation loss promotes immune 175, 998–1013 evasion of tumours with high mutation and copy number load. 43. Ghoneim, H.E. et al. (2017) De novo epigenetic programs inhibit Nat. Commun. 10, 4278 PD-1 blockade-mediated T cell rejuvenation. Cell 170, 142–157 68. Fraietta, J.A. et al. (2018) Disruption of TET2 promotes the 44. Pauken, K.E. et al. (2016) Epigenetic stability of exhausted therapeutic efficacy of CD19-targeted T cells. Nature 558, T cells limits durability of reinvigoration by PD-1 blockade. 307–312 Science 354, 1160–1165 69. Nobles, C.L. et al. (2019) CD19-targeting CAR T cell immuno- 45. Philip, M. et al. (2017) Chromatin states define tumour-specific therapy outcomes correlate with genomic modification by T cell dysfunction and reprogramming. Nature 545, 452–456 vector integration. J. Clin. Invest. 130, 673–685 46. Liu, M. et al. (2018) Dual inhibition of DNA and histone methyl- 70. Sundar, R. et al. (2019) Epigenomic promoter alterations pre- transferases increases viral mimicry in ovarian cancer cells. dict for benefit from immune checkpoint inhibition in metastatic Cancer Res. 78, 5754–5766 gastric cancer. Ann. Oncol. 30, 424–430 47. Daskalakis, M. et al. (2018) Reactivation of endogenous retro- 71. Covre, A. et al. (2015) Antitumor activity of epigenetic viral elements via treatment with DNMT- and HDAC-inhibitors. immunomodulation combined with CTLA-4 blockade in synge- Cell Cycle 17, 811–822 neic mouse models. Oncoimmunology 4, e1019978 48. Roulois, D. et al. (2015) DNA-demethylating agents target colo- 72. Di Giacomo, A.M. et al. (2019) Guadecitabine plus ipilimumab rectal cancer cells by inducing viral mimicry by endogenous in unresectable melanoma: the NIBIT-M4 clinical trial. Clin. transcripts. Cell 162, 961–973 Cancer Res. Cancer Res. 25, 7351–7362 49. Chakravarthy, A. et al. (2018) Pan-cancer deconvolution of 73. Yang, H. et al. (2014) Expression of PD-L1, PD-L2, PD-1 and tumour composition using DNA methylation. Nat. Commun. CTLA4 in myelodysplastic syndromes is enhanced by treat- 9, 3220 ment with hypomethylating agents. Leukemia 28, 1280–1288 50. Chakravarthy, A. et al. (2018) TGF-β-associated extracellular 74. Ørskov, A.D. et al. (2015) Hypomethylation and up-regulation matrix genes link cancer-associated fibroblasts to immune of PD-1 in T cells by azacytidine in MDS/AML patients: a ratio- evasion and immunotherapy failure. Nat. Commun. 9, 4692 nale for combined targeting of PD-1 and DNA methylation. 51. Sheng, W. et al. (2018) LSD1 Ablation stimulates anti-tumor Oncotarget 6, 9612–9626 immunity and enables checkpoint blockade. Cell 174, 549–563 75. Daver, N. et al. (2019) Efficacy, safety, and biomarkers of 52. Pascual-García, M. et al. (2019) LIF regulates CXCL9 in tumor- response to azacitidine and nivolumab in relapsed/refractory associated macrophages and prevents CD8+ T cell tumor- acute myeloid leukemia: a nonrandomized, open-label, phase infiltration impairing anti-PD1 therapy. Nat. Commun. 10, 2416 II study. Cancer Discov. 9, 370–383 53. Dangaj, D. et al. (2019) Cooperation between constitutive and 76. Zeidan, A.M. et al. (2019) Clinical outcomes of older patients (pts) inducible chemokines enables T cell engraftment and immune with acute myeloid leukemia (AML) receiving hypomethylating attack in solid tumors. Cancer Cell 35, 885–900 agents (HMAs): a large population-based study in the United 54. Brennan, K. et al. (2017) NSD1 inactivation defines an immune States. Blood 134, 646 cold, DNA hypomethylated subtype in squamous cell carcinoma. 77. Goel, S. et al. (2017) CDK4/6 inhibition triggers anti-tumour Sci. Rep. 7, 17064 immunity. Nature 548, 471–475 55. Lingohr, P. et al. (2019) Clinicopathological, immune and molecular 78. Segovia, C. et al. (2019) Inhibition of a G9a/DNMT network correlates of PD-L2 methylation in gastric adenocarcinomas. triggers immune-mediated bladder cancer regression. Nat. Epigenomics 11, 639–653 Med. 25, 1073–1081 690 Trends in Immunology, August 2020, Vol. 41, No. 8 Trends in Immunology 79. Azad, N.S. et al. (2018) ENCORE 601: a phase 2 study of 90. Ferreira, H.J. and Esteller, M. (2018) Non-coding RNAs, epige- entinostat in combination with pembrolizumab in patients netics, and cancer: tying it all together. Cancer Metastasis Rev. with microsatellite stable metastatic colorectal cancer. J. Clin. 37, 55–73 Oncol. 36, 3557 91. Stresemann, C. et al. (2008) Azacytidine causes complex DNA 80. Johnson, M.L. et al. (2017) ENCORE 601: a phase II study of methylation responses in myeloid leukemia. Mol. Cancer Ther. entinostat (ENT) in combination with pembrolizumab (PEMBRO) in 7, 2998–3005 patients with melanoma. J. Clin. Oncol. 35, 9529 92. Nishioka, C. et al. (2008) Histone deacetylase inhibitors induce 81. Johnson, M.L. et al. (2016) Preliminary results of ENCORE growth arrest and apoptosis of HTLV-1-infected T-cells via 601, a phase 1b/2, open-label study of entinostat (ENT) blockade of signaling by nuclear factor kappaB. Leuk. Res. in combination with pembrolizumab (PEMBRO) in patients 32, 287–296 with non-small cell lung cancer (NSCLC). J. Clin. Oncol. 34, 93. Mann, B.S. et al. (2007) FDA approval summary: vorinostat for e20659 treatment of advanced primary cutaneous T-cell lymphoma. 82. Saltos, A.N. et al. (2018) Phase I/Ib study of pembrolizumab Oncologist 12, 1247–1252 and vorinostat in patients with metastatic NSCLC (mNSCLC). 94. Dalva-Aydemir, S. et al. (2019) Toward in vitro epigenetic drug J. Clin. Oncol. 36, 9046 design for thyroid cancer: the promise of PF-03814735, an 83. Gray, J.E. et al. (2019) Phase I/Ib study of pembrolizumab plus aurora kinase inhibitor. Omics J. Integr. Biol. 23, 486–495 vorinostat in advanced/metastatic non–small cell lung cancer. 95. Nakagawa, M. et al. (2019) Selective inhibition of mutant IDH1 Clin. Cancer Res. 25, 6623–6632 by DS-1001b ameliorates aberrant histone modifications and 84. Levy, B.P. et al. (2019) Randomised phase 2 study of impairs tumor activity in chondrosarcoma. Oncogene 38, pembrolizumab plus CC-486 versus pembrolizumab plus 6835–6849 placebo in patients with previously treated advanced non- 96. Cao, J. et al. (2018) Ricolinostat (ACY-1215) suppresses prolifera- small cell lung cancer. Eur. J. Cancer 108, 120–128 tion and promotes apoptosis in esophageal squamous cell carci- 85. Lea, A.J. et al. (2018) Genome-wide quantification of the noma via miR-30d/PI3K/AKT/mTOR and ERK pathways. Cell effects of DNA methylation on human gene regulation. eLife Death Dis. 9, 817 7, e37513 97. Pérez-Salvia, M. et al. (2018) In vitro and in vivo activity of a new 86. Lokk, K. et al. (2014) DNA methylome profiling of human tis- small-molecule inhibitor of HDAC6 in mantle cell lymphoma. sues identifies global and tissue-specific methylation patterns. Haematologica 103, e537–e540 Genome Biol. 15, 3248 98. Shuttleworth, S.J. (2016) KA2237 and KA2507: novel, oral cancer 87. Tahiliani, M. et al. (2009) Conversion of 5-methylcytosine to immunotherapeutics targeting PI3K-p110β/p110δ and HDAC6 5-hydroxymethylcytosine in mammalian DNA by MLL partner with single-agent and combination activity. Cancer Res. 76, 3996 TET1. Science 324, 930–935 99. Takahashi, Y. et al. (2017) Integration of CpG-free DNA 88. Ito, S. et al. (2011) Tet proteins can convert 5-methylcytosine induces de novo methylation of CpG islands in pluripotent to 5-formylcytosine and 5-carboxylcytosine. Science 333, stem cells. Science 356, 503–508 1300–1303 100. Seo, H. et al. (2019) TOX and TOX2 transcription factors cooperate 89. Zhao, Z. and Shilatifard, A. (2019) Epigenetic modifications of with NR4A transcription factors to impose CD8+ T cell exhaustion. histones in cancer. Genome Biol. 20, 245 Proc. Natl. Acad. Sci. U. S. A. 116, 12410–12415 Trends in Immunology, August 2020, Vol. 41, No. 8 691