Combining epigenetic drugs with other therapies for solid tumours - past lessons and future promise

Page created by Don Swanson
 
CONTINUE READING
Combining epigenetic drugs with other therapies for
        solid tumours - past lessons and future promise
      Daphné Morel, Daniel Jeffery, Sandrine Aspeslagh, Geneviève Almouzni,
                                         Sophie Postel-Vinay

     To cite this version:
    Daphné Morel, Daniel Jeffery, Sandrine Aspeslagh, Geneviève Almouzni, Sophie Postel-Vinay. Com-
    bining epigenetic drugs with other therapies for solid tumours - past lessons and future promise. Nature
    Reviews Clinical Oncology, Nature Publishing Group, 2019, 17 (2), pp.91-107. �10.1038/s41571-019-
    0267-4�. �hal-02371032�

                                  HAL Id: hal-02371032
                https://hal.archives-ouvertes.fr/hal-02371032
                                      Submitted on 9 Dec 2020

    HAL is a multi-disciplinary open access            L’archive ouverte pluridisciplinaire HAL, est
archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents
entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non,
lished or not. The documents may come from émanant des établissements d’enseignement et de
teaching and research institutions in France or recherche français ou étrangers, des laboratoires
abroad, or from public or private research centers. publics ou privés.
                                              Copyright
REVIEWS
                                  Combining epigenetic drugs with
                                  other therapies for solid tumours —
                                  past lessons and future promise
                                  Daphné Morel 1,5, Daniel Jeffery                 2,5
                                                                                      , Sandrine Aspeslagh3, Geneviève Almouzni                  2,5
                                                                                                                                                    *
                                  and Sophie Postel-Vinay 1,4,5*
                                  Abstract | Epigenetic dysregulation has long been recognized as a key factor contributing to
                                  tumorigenesis and tumour maintenance that can influence all of the recognized hallmarks of
                                  cancer. Despite regulatory approvals for the treatment of certain haematological malignancies, the
                                  efficacy of the first generation of epigenetic drugs (epi-drugs) in patients with solid tumours has
                                  been disappointing; however, successes have now been achieved in selected solid tumour subtypes,
                                  thanks to the development of novel compounds and a better understanding of cancer biology that
                                  have enabled precision medicine approaches. Several lines of evidence support that, beyond their
                                  potential as monotherapies, epigenetic drugs could have important roles in synergy with other
                                  anticancer therapies or in reversing acquired therapy resistance. Herein, we review the mechanisms
                                  by which epi-drugs can modulate the sensitivity of cancer cells to other forms of anticancer therapy,
                                  including chemotherapy, radiation therapy, hormone therapy, molecularly targeted therapy and
                                  immunotherapy. We provide a critical appraisal of the preclinical rationale, completed clinical
                                  studies and ongoing clinical trials relating to combination therapies incorporating epi-drugs. Finally,
                                  we propose and discuss rational clinical trial designs and drug development strategies, considering
                                  key factors including patient selection, tumour biomarker evaluation, drug scheduling and response
                                  assessment and study end points, with the aim of optimizing the development of such combinations.
1
 ATIP–Avenir Group,              Over the past decade, remarkable advances in antican-          epi-drugs, which are almost exclusively DNA methyl-
UMR981, INSERM (French           cer therapy have been made, owing notably to the devel-        transferase (DNMT) inhibitors or histone deacetylase
National Institute of Health     opment of precision medicine and immunotherapy                 (HDAC) inhibitors (Box 1). However, success with
and Medical Research),
                                 approaches as well as continuous improvements in the           certain third-generation epi-drugs used according to
Gustave Roussy Cancer
Campus, Villejuif, France.       available therapeutic options and strategies. Nevertheless,    precision medicine paradigms has provided new hope
2
 Nuclear Dynamics Unit —
                                 the persistent ability of cancers to resist treatment          in the field of epigenetic therapy. For example, the iso­
UMR3664, National Centre         remains a major challenge that limits the effectiveness        citrate dehydrogenase 1 (IDH1) inhibitor ivosidenib
for Scientific Research,         all anticancer therapies.                                      received FDA approval on the basis of the results of
Institut Curie, Paris, France.       Epigenetic dysregulation has major roles in tumour         a phase I trial involving patients with IDH1-mutant
3
 University Hospital Brussels,   development, progression and acquisition of therapeutic        acute myeloid leukaemia (AML), only 4 years after the
Brussels, Belgium.
                                 resistance, in both solid and haematological malignan-         first patient was enrolled3. In addition, tazemetostat,
4
 Drug Development                cies1. To date, however, the efficacy of ‘epi-drugs’ — drugs   an inhibitor of the histone-lysine N-methyltransferase
Department (DITEP), Gustave
Roussy Cancer Campus,
                                 that target enzymes involved in epigenetic regulation of       enhancer of zeste homologue 2 (EZH2), has demon-
Paris-Saclay University,         genome function (Fig. 1) — has mostly been confined to         strated efficacy in patients with EZH2-mutant haemato-
Villejuif, France.               haematological cancers2, perhaps in part because solid         logical malignancies or highly aggressive solid tumours
5
 These authors contributed       tumours tend to arise from more-differentiated or even         harbouring genetic aberrations affecting the SWI/SNF
equally: Daphné Morel, Daniel    terminally differentiated cells with a reduced capability      chromatin remodelling complex, such as SMARCB1
Jeffery, Geneviève Almouzni      for epigenetic reprogramming. Furthermore, biomark-            (also known as INI1) or SMARCA4 mutations or dele-
and Sophie Postel-Vinay
                                 ers for patient selection have crucially been lacking in       tions4. Furthermore, responses to bromodomain and
*e-mail: genevieve.almouzni@
                                 early phase trials of epi-drugs, which have largely been       extra-terminal domain (BET) inhibitors, which tar-
curie.fr; sophie.postel-vinay@
gustaveroussy.fr                 conducted according to the historical ‘one size fits all’      get certain bromodomain-containing protein (BRD)
https://doi.org/10.1038/         approach. This strategy probably hampered the devel-           family members, have been observed in patients with
s41571-019-0267-4                opment of first-generation and second-generation               BRD4-rearranged NUT midline carcinoma5.

Nature Reviews | Clinical Oncology
Reviews

Key points                                                                                  comprise ten biological capabilities acquired during
                                                                                            oncogenesis that drive and/or enable the development
• The first generation of drugs targeting the epigenome (epi-drugs) were developed          of cancer, as defined by Hanahan and Weinberg10.
  using a ‘one size fits all’ approach and proved to have disappointing efficacy in         Various epi-drugs are able to enhance the effective-
  patients with solid tumours.                                                              ness of four major pillars of the anticancer treatment
• The potential of epi-drugs to modulate the sensitivity of tumours to other anticancer     armamentarium, namely genotoxic and/or cytotoxic
  drugs and to overcome therapy resistance presents major new avenues of clinical           treatments (encompassing chemotherapy and radio-
  investigation.
                                                                                            therapy), hormone therapy, molecularly targeted therapy
• A new generation of epi-drugs, which were developed to be more specific for their         and immunotherapy11 (Fig. 2). In the following sections of
  targets and have promising activity in certain biomarker-selected populations, is
                                                                                            this Review, we describe the synergistic mechanisms
  now entering early phase clinical trials and are showing promising efficacy.
                                                                                            of epi-drug combinations stratified according to each of
• Epi-drug development should follow a precision-medicine approach, with further
                                                                                            these four pillars of therapy, first discussing the preclini­
  identification of robust predictive biomarkers for patient selection and subsequent
  implementation of this strategy in clinical trials.
                                                                                            cal evidence, followed by examples of corresponding
                                                                                            clinical trials. The examples provided were selected to
• Sequential treatment schedules and/or dosing at less than the maximum tolerated
  dose might improve the efficacy and tolerability of epi-drugs when used in
                                                                                            illustrate the specific challenges to the development
  combination with various other anticancer therapies.                                      of epi-drug combinations (including those relating to
                                                                                            patient selection, treatment schedules, toxicities and
                                                                                            efficacy assessments) and to highlight the most recent
                                 Cancers continuously evolve, in part through               and most prominent clinical successes and failures.
                             dynamic and reversible epigenetic changes that confer          We also provide comprehensive lists summarizing past
                             a fitness advantage. This epigenetic plasticity — a term       and ongoing trials of epi-drugs in combination with
                             used hereinafter in reference to reversible changes in         other anticancer therapies (Supplementary Tables 1
                             epigenetic marks on DNA, histone and non-histone               and 2, respectively).
                             proteins, as well as the functional consequences of these
                             alterations — is also involved in primary and acquired         Combinations with cytotoxic agents
                             resistance to various anticancer therapies, through mod-       Chemotherapies. In preclinical studies assessing the
                             ulation of tumour cells or their microenvironment6,7           effects of epi-drugs in combination with chemotherapy,
                             (Supplementary Figure 1). Epi-drugs might therefore            synergy has mostly been achieved with DNA-damaging
                             be most effective when used in combination with other          agents. HDAC inhibitors and DNMT inhibitors pro-
                             treatments and present particularly attractive strate-         mote global chromatin relaxation, which facilitates DNA
                             gies for sensitizing cancer cells to a given therapy and       damage by genotoxic agents and interferes with
                             for overcoming acquired resistance mechanisms in a             DNA-damage repair (DDR)12 (Fig. 3a). HDAC inhib-
                             dynamic fashion. Of note, cancer stem cells, which are         itors further impede DDR by disrupting ATM signal-
                             characterized by their abilities to self-renew and initiate    ling, modulating the post-translational modification
                             cancer8, probably also have important roles in thera-          of non-histone substrates of HDACs, including p53
                             peutic resistance. The inherent phenotypic plasticity          and X-ray repair cross-complementing protein 6 (also
                             of these stem-like cells is key to this resistance and is      known as Ku70), and by interfering with the regula-
                             regulated by both transcription factors and epigenetic         tion of histone marks that regulate DDR13. In addition,
                             changes. Indeed, some third-generation epi-drugs,              combined use of HDAC inhibitors and DNMT inhibi­
                             such as lysine-specific histone demethylase 1A (LSD1)          tors at low doses might revert resistance to cytotoxic
                             inhibitors, have been shown to disrupt the molecular           agents, in part, by inducing removal of acquired epi­
                             changes underlying this plasticity. However, this topic        genetic alterations that drive the resistance phenotype14,15.
                             has been reviewed elsewhere9 and will not be discussed         Furthermore, epigenetic silencing of genes involved in
                             further herein.                                                shaping the tumour microenvironment (such as TGFBI)
                                 Despite robust preclinical evidence supporting             has been associated with resistance to taxanes and can be
                             epi-drug-containing combination therapies, clinical            reversed with DNMT inhibitors, such as 5-azacytidine16,
                             success with first-generation and second-generation            whereas defects in genes encoding components of chro-
                             epi-drugs has been limited. With the new generation of         matin remodelling complexes (such as SMARCA4) are
                             novel, selective epi-drugs currently entering clinical test-   correlated with sensitivity to taxanes and cisplatin17,18.
                             ing, learning from past successes and failures is crucial to   Interestingly, Fillmore et al.19 showed that inhibition of
                             maximizing future patient benefits. Herein, we provide         EZH2 increases the sensitivity of SMARCA4-mutant or
                             a critical review of preclinical and clinical data support-    EGFR-mutant non-small-cell lung cancers (NSCLCs) to
                             ing the use of epi-drugs in combination with other anti-       DNA topoisomerase 2 (TOP2) inhibitors, but conversely
                             cancer therapies in the treatment of patients with solid       promotes resistance in tumours that are wild type for
                             tumours. We also highlight the high stakes and chal-           both genes. These findings highlight the importance
                             lenges of successful epi-drug development and propose          of considering genomic characteristics of the tumour
                             clinically relevant approaches to optimize this process.       when selecting patients for therapy with particular epi-
                                                                                            drugs. Other third-generation epi-drugs, such as BET
                             An appraisal of epi-drug combinations                          inhibitors and protein arginine N-methyltransferase 5
                             Epigenetic regulation influences all of the recognized         (PRMT5) inhibitors, also interfere with DDR processes,
                             hallmarks of cancer, which can each be targeted using          raising the possibility of synergy with cytotoxic therapy
                             various therapeutic approaches. These hallmarks                (reviewed in detail elsewhere20,21).

                                                                                                                            www.nature.com/nrclinonc
Reviews

     Nuclear
                                                                                                phase II studies24,25 conducted to evaluate this strategy
     position                                                                                   by combining decitabine with carboplatin in the treat-
                                                                                                ment of women with recurrent and/or progressing ovar-
                              Nucleus                   Chromatin
  Higher-order                                          remodellers                             ian cancer illustrate certain challenges associated with
  chromatin                                                                                     epi-drug–chemotherapy combinations. In one of these
                                                                                                trials25, low-dose decitabine (10 mg/m2) was admin­
                                                                                                istered on days 1–5 of a 28-day cycle, which seemed to
                                                                                                result in better tolerability than high-dose decitabine
                                                                                                (45 or 90 mg/m2) administered on day 1 of each cycle
                                                                                 Histone        in the other trial24 (Supplementary Table 1). Differences in
                                                                                 PTM
                                                                                                the trial designs and patient populations precludes any
                                                                      Ac     BRD readers
                 Structural RNA                                                                 formal comparison of these trials, although the level of
                                          Nucleosome                                            efficacy — in terms of objective response rate (ORR)
                                                                                   BETi         and progression-free survival (PFS) — in the trial of
                                                                                                low-dose decitabine (which included patients with
                                                                        Me
                                                  Ac                            Histone PTM     platinum-refractory disease) also seemed to be superior
                              DNA                                               writers/        to that observed in the trial of the high-dose regimen
                                                                                erasers         (in which patients with partially platinum-sensitive dis-
                                                                  HMT        HDM
                                            HAT        HDAC                                     ease were enrolled), thus supporting the selection of the
                 DNA
                 modifiers                                                                       low-dose regimen for further testing in a subsequent
  DNMTi           DNMT                     HATi         HDACi     HMTi        HDMi              phase III trial (Supplementary Table 2). Interestingly, both
                                                        (SIRTi)   (PRMTi)                       studies showed a high incidence of carboplatin-related
                              Me                                                                hypersensitivity reactions with the combination (35%
                                                  Ac      Me                                    and 67%), potentially resulting from decitabine-induced
    TETi              TET
                                                                                                demethylation of genes encoding cytokines involved in
                                                                        Gene                    commitment of naive T cells to the T helper 1 (TH1) and
                IDH         α-KG                  Trans-acting                                  T helper 2 (TH2) lineages and IgE-mediated hypersensi-
                                                  factors                                       tivity reactions24,26. Global decitabine-induced 5-methyl-
 IDHi       Mutant            D-2-HG                                                            cytosine demethylation of DNA from peripheral blood
             IDH
                                                                                                mononuclear cells (PBMCs) was also noted in both
Fig. 1 | Targeting the epigenome. Genome function is regulated at multiple levels by            trials24,25. This effect was dose-dependent27, which is in
>800 epigenetic enzymes. A predominant aspect of this regulation involves variations in         contrast to the findings of other clinical trials indicating
the organization of genomic DNA within chromatin structures, the basic unit of which            that the pharmacodynamic modulation induced by cer-
is the nucleosome core particle consisting of 146 bp of DNA wrapped around a histone            tain epi-drugs reaches a plateau at doses well below the
octamer. Epigenetic modifiers acting at this level comprise four nonexclusive categories:       maximum tolerated dose4,28. Notably, Matei et al.25 identi-
the ‘writers’, which add specific marks to DNA or the core histones H2A, H2B, H3, and H4;       fied that patients with a PFS duration of >6 months have
‘readers’, which recognize these marks; ‘erasers’, which remove them; and the ‘shapers’,        higher numbers of demethyl­ated genes in tumour biopsy
which remodel chromatin, mobilize nucleosomes or mediate histone exchange. Current
                                                                                                samples and in PBMCs, suggesting that PBMC DNA
epigenetic drugs (epi-drugs) target proteins belonging to the first three categories. These
epi-drugs include inhibitors of DNA-modifying enzymes, such as DNA methyltransferases           methylation status could provide information on tumour
(DNMTs) that can methylate certain cytosine or adenosine nucleotides of DNA and                 DNA methyl­ation and serve as a pharmacodynamic
ten-eleven translocation (TET) enzymes that catalyse the oxidation of 5-methylcytosine in       or predictive biomarker.
DNA to initiate the process of active DNA demethylation. Another category of epi-drugs              Overall, despite the preclinical data showing that
with effects on DNA methylation target gain-of-function mutant forms of isocitrate              HDAC inhibitors and DNMT inhibitors potentiate
dehydrogenase (IDH) enzymes, which produce the oncometabolite d-2-hydroxyglutarate              the activity of cytotoxic chemotherapy, clinical results
(D-2-HG) that competitively inhibits TET hydroxylases and thereby results in aberrant           have been disappointing: three out of the five key ran-
DNA methylation. By contrast, wild-type IDH enzymes support DNA demethylation by                domized trials of such combinations were halted owing
producing α-ketoglutarate (α-KG), which is a cofactor for TET proteins. Other epi-drugs         to a lack of efficacy and/or unfavourable toxicity pro-
inhibit writers or erasers of histone arginine and/or lysine post-translational modifications
                                                                                                files when compared with chemotherapy alone. Such
(PTMs), including histone acetyltransferases (HATs), histone deacetylases (HDACs),
histone methyltransferases (HMTs) and other protein arginine methyltransferases                 undesired systemic toxicities and limited efficacy might
(PRMTs) and histone demethylases (HDMs), which can also act on non-chromatin factors.           warrant investigation of targeted delivery of epi-drugs,
An additional class of epi-drugs inhibits readers of histone PTMs, such as bromodomain          for example, as payloads of antibody–drug conjugates
and extra-terminal domain (BET) family proteins (BRDs), which bind acetylated lysine            (ADCs) or liposomal particles. Moreover, most — if
residues of histones (as well as other non-chromatin factors). Epi-drugs targeting              not all — trials of epi-drug–chemotherapy combina-
chromatin remodelling factors and histone chaperones are also under development,                tions have been performed in unselected patient popu­
but are beyond the scope of this manuscript. SIRTi, sirtuin inhibitor. Figure adapted from      lations (that is, without the use of predictive molecular
ref.161, Springer Nature Limited.                                                               biomarkers). Thus, appropriate patient selection using
                                                                                                rational epigenetic biomarkers might produce clinically
                                      Findings of preclinical studies have demonstrated the     meaningful improvements in the therapeutic window
                                   potential of DNMT inhibitors to induce the re-expression     and new therapeutic opportunities. For example, evalu­
                                   of epigenetically silenced genes and thereby restore         ating SLFN11 or PCFT promoter hypermethylation
                                   sensitivity of ovarian cancers to platinum-based chemo-      using clinical grade assays could potentially enable the
                                   therapy22,23. However, the discordant results of two         identification of patients who would benefit most from

Nature Reviews | Clinical Oncology
Reviews

Box 1 | generations of epi-drugs                                                               and radiotherapy was well tolerated in 21 patients with
                                                                                               localized pancreatic ductal adenocarcinoma and led to
Multiple generations of epigenetic drugs (epi-drugs) have been developed.                      four R0 resections among 11 patients who underwent
First-generation and second-generation epi-drugs include DNA methyltransferase                 surgical exploration and a median overall survival (OS)
inhibitors (DNMTi) and histone deacetylase inhibitors (HDACi), some of which have
                                                                                               duration of 1.1 years for the entire cohort, which accord-
already been approved for the treatment of certain haematological malignancies2.
                                                                                               ing to the investigators compared favourably with the
DNMTi are pyrimidine analogues that are incorporated into DNA during replication and
form covalent DNMT–drug adducts that trigger the activation of DNA damage response             results of other phase I studies of chemoradiotherapy
and can, ultimately, lead to apoptosis, thus explaining the dual epigenetic and cytotoxic      in such a patient population38. Second, the combina-
effects of these drugs. With the exception of sirtuin inhibitors, which inhibit nicotinamide   tion of vorinostat and vectorized internal radiotherapy
adenine dinucleotide (NAD+)-dependent class III histone deacetylases, HDACi inhibit the        with 131I-metaiodobenzylguanidine in children with
Zn2+-dependent catalytic activity of either all (pan-HDACi) or single HDAC enzymes             relapsed and/or refractory high-risk neuroblastoma led
(for example, HDAC6i). First-generation DNMTi (azacytidine and decitabine) and HDACi           to an ORR of 12% at all dose levels and 17% at the recom­
(vorinostat and romidepsin) have unfavourable pharmacokinetic characteristics and              mended phase II dose;39 this regimen is now being
target selectivity, which led to the development of second-generation DNMTi (such as           evaluated in a randomized phase II trial (Supplementary
zebularine and guadecitabine) and HDACi, (including the hydroxamic acid HDACi                  Table 2). Why these two studies had promising results
belinostat and panobinostat, the benzamide HDACi tucidinostat (also known as
                                                                                               whereas other clinical trials of epi-drug–radiotherapy
chidamide) and carboxylic acid derivative HDACi (such as valproic acid)), with improved
chemical properties under physiological conditions.                                            combinations did not is unclear, although the appar-
  Third-generation epi-drugs include, among others, bromodomain and extra-terminal             ently greater permissivity of tumours to epigenetic
domain inhibitors (BETi), histone methyltransferase inhibitors (HMTi) and histone              reprogramming in the latter study might be explained
demethylase inhibitors (HDMi) and protein arginine methyltransferase inhibitors (PRMTi).       by the paediatric nature of the disease (which is likely to
BETi disrupt the recognition of acetylated lysine residues on histones, a mark associated      have arisen during early development, when epigenetic
with active transcription, by BET-containing reader proteins, including certain members        processes have key roles in determining cell fate). These
of the bromodomain-containing protein (BRD) family. HMTi (including enhancer of zeste          results call for the implementation of biomarkers for
homologue 2 inhibitors and DOT1-like histone-lysine methyltransferase inhibitors),             patient selection. For example, a vorinostat sensitivity
HDMi (such as lysine-specific histone demethylase 1A inhibitors) or PRMTi (for example,        transcriptional signature (sig-139), which is enriched
targeting PRMT1 or PRMT5) reduce the catalytic activity of the specific target enzyme
                                                                                               for the expression of genes involved in DNA synthe-
or, in the specific case of embryonic ectoderm development (EED) protein inhibitors,
destabilize the enzymatic complex containing the target protein.                               sis, repair and chromatin remodelling as well as RNA
                                                                                               synthesis, splicing and processing has been identified
                                                                                               using a panel of cell lines, with positive associations
                              the use of DNMT inhibitors to reverse resistance to              with PFS and OS observed in patients with newly diag-
                              platinum-based agents29 or pemetrexed30, respectively.           nosed glioblastoma treated with vorinostat plus standard
                                                                                               temozolomide and radiation therapy40.
                              Radiotherapy. In preclinical studies, HDAC, DNMT,
                              EZH2 or BET inhibition potentiates the antitumour                Combinations with hormone therapy
                              activity of radiotherapy by disrupting DDR and the               Oestrogen receptor-positive breast cancer. Oestrogen
                              cell cycle and increasing oxidative stress31–35 (Fig. 3a).       receptor-α (ERα), the pivotal oncogenic driver of 70%
                              Transient HDAC inhibition with valproic acid (before             of breast cancers, is released from heat shock protein 90
                              and up to 24 hours after irradiation) results in radio-          (HSP90) upon binding with its endogenous ligands,
                              sensitization of human glioma cell lines — a schedule            predominantly oestradiol. The ERα–oestrogen complex
                              that might limit the systemic toxicities associated with         subsequently translocates into the nucleus and binds to
                              chronic HDAC inhibition36. Interestingly, BET inhi-              oestrogen-responsive elements — a process tightly regu-
                              bition can decrease the risk of radiation-induced lung           lated by chromatin-remodelling factors41,42. Preclinically,
                              fibrosis and has a radioprotective effect on irradiated          HDAC inhibitors have been shown to increase the anti-
                              non-malignant lung tissue in mouse models, but con-              tumour efficacy of, or reverse resistance to, hormone
                              versely radiosensitizes thoracic cancer cells (that is,          therapies in models of ERα-positive breast cancer43–45.
                              NSCLC, breast or oesophageal cancer cells) in vitro37.           HDAC inhibition interferes with ERα signalling pathway
                              This potential ability to differentially modulate radio-         at multiple levels, including transcriptional suppression
                              sensitivity according to cell lineage, if recapitulated in       of ESR1 (which encodes ERα), reductions in ERα mRNA
                              humans, could provide new therapeutic opportunities.             and protein stability and modulation of the transcrip-
                                  Certain DNMT inhibitors, such as the cytidine                tion of ERα targets as well as other genes involved in cell
                              analogues decitabine and 5-azacytidine, can be incor-            proliferation and metastasis46,47 (Fig. 3b). HDAC6 inhibi-
                              porated into DNA and/or RNA; thus, these agents are              tion, specifically, leads to HSP90 hyperacetylation, which
                              particularly strong radiosensitizers in all tissues and          disrupts the HSP90–ERα association and favours ERα
                              the combination of these epi-drugs with radiotherapy             ubiquitination and proteasomal degradation48. Selective
                              is not recommended owing to concerns regarding toxi­             HDAC6 inhibitors are, therefore, interesting candidates
                              city. In clinical trials, combining HDAC inhibitors with         for future clinical evaluation48,49.
                              radiotherapy, usually concurrently and sometimes                     BET inhibition also suppresses the growth of
                              in addition to chemotherapy, has mostly resulted in              tamoxifen-resistant tumour cells and xenografts in
                              increased levels of toxicity and minimal patient benefit         combination with the anti-oestrogen fulvestrant50,51; the
                              (Supplementary Table 1). Two phase I studies, however,           underlying mechanisms include inhibition of the BRD3
                              have provided encouraging results. First, the combina-           and/or BRD4-dependent recruitment of the histone-
                              tion of the HDAC inhibitor vorinostat with capecitabine          lysine N-methyltransferase NSD2 (also known as WHSC1)

                                                                                                                              www.nature.com/nrclinonc
Reviews

                        Genotoxic and/or cytotoxic
                        chemotherapy and radiotherapy                                                                            Targeted therapies
                        DNMTi and/or HDACi                                                                                       DNMTi and/or HDACi
                        • Facilitate access to DNA                                                                               • Revert EMT
                        • Re-express tumour suppressors
                        • Increase ROS                                                                                           HDACi
                                                                                                                                 • Promote HIF1α
                        DNMTi, EZH2i and/or LSD1i                                                                                  degradation
                        • Differentiate cancer stem cells                                                                        • Reduce oncoprotein
                        EZH2i                                                                                                      stability
                        • Sensitive to TOP2i                                           Genome                                    BETi
                        BETi and/or PRMTi                                              instability                               • Prevent oncogene
                        • Decrease DDR capabilities                                    and mutation     Activating                 transcription
                                                                       Resisting                        invasion and
                                                                       cell death                       metastasis

                                                             Enabling                                             Inducing
                                                             replicative                                          angiogenesis
                                                             immortality

                                                                                     Hallmarks of cancer

                                                                                                                      Tumour-
                                                            Evading growth                                            promoting
                                                            suppressors                                               inflammation

                                                                     Sustaining                            Avoiding
                                                                     proliferative                         immune
                                                                     signalling                            response
                                                                                       Deregulating
                                                                                       cellular
                                                                                       energetics

                            Hormone therapy                                                                                   Immunotherapy

                        HDACi, HATi and/or BETi                                                                         DNMTi, HDACi and/or EZH2i
                        • Reverse endocrine                                                                             • Increase antigen presentation
                          resistance                                                                                      and chemokine and/or IFN
                                                                                                                          expression
                                                                                                                        DNMTi, HDACi and/or LSDi
                                                                                                                        • Enable viral mimicry
                                                                                                                        • M2→M1 macrophage
                                                                                                                          polarization
                                                                                                                        BETi
                                                                                                                        • Anti-inflammatory

                       Fig. 2 | Epi-drugs can enhance the activity of other anticancer therapies to counteract the hallmarks of cancer.
                       Current prominent types of anticancer therapeutics have counteracting effects against various hallmarks of cancer
                       (defined by Hanahan and Weinberg10 in 2011), as indicated by the coloured arcs surrounding the hallmarks of cancer wheel.
                       Cytotoxic and/or genotoxic chemotherapies and radiotherapy (green arc), hormone therapy (orange arc), immunotherapy
                       (blue arc) and targeted therapies (purple arc) each counteract specific hallmarks, with some overlap. Epigenetic drugs
                       (epi-drugs) can enhance the activity of these anticancer therapies through various mechanisms; examples of the relevant
                       categories of epi-drugs and their augmenting effects observed in preclinical studies are indicated in the figure for each type
                       of therapy. Accordingly, the combination of epi-drugs with other anticancer therapies can lead to sensitization to treatment
                       and/or reversal of resistance. BETi, bromodomain and extra-terminal domain inhibitor; DDR, DNA damage repair; DNMTi,
                       DNA methyltransferase inhibitor; EZH2i, enhancer of zeste homologue 2 inhibitor; EMT, epithelial-to-mesenchymal
                       transition; HATi, histone acetyltransferase inhibitor; HDACi, histone deacetylase inhibitor; HIF1α, hypoxia-inducible
                       factor 1α; IFN, interferon; LSD1i, lysine-specific demethylase 1 inhibitor; PRMTi, protein arginine methyltransferase inhibitor;
                       ROS, reactive oxygen species; TOP2i, DNA topoisomerase 2 inhibitor. Adapted with permission from ref.10, Elsevier.

                       which monomethylates and/or dimethylates histone H3                a selective inhibitor of the CREB-binding protein (CBP)
                       lysine 36 to generate H3K36me1 and H3K36me2 marks                  and histone acetyltransferase p300 (EP300) transcrip-
                       that are commonly associated with active transcription,            tional co-activator proteins, via a mechanism involving
                       and disruption of elongation-associated phosphory­                 disruption of the core ERα signalling complex54.
                       lation events on RNA polymerase II that results in stalled             The results of two phase II studies demonstrate
                       transcription, thus ultimately reducing the expression of          that combinations of HDAC inhibitors and an anti-
                       ESR1 and ERα target genes50–53. Similarly, single-agent            oestrogen or aromatase inhibitor have acceptable
                       activity and synergism with anti-oestrogen therapies               tolerability, with promising efficacy (Supplementary
                       has been demonstrated in preclinical studies of CPI-1,             Table 1). The first trial, a single-arm study of vorinostat

Nature Reviews | Clinical Oncology
Reviews

a            Irradiation                       Damage to DNA                                 b
                                               and/or proteins

                                                                                                                                              Hormone precursor
              H2O                                                  Premetrexed                                                                               Aromatase
                                                                                                                                                             inhibitor
    Oxidative                                  Cyt c            Microtubule
                          HDACi                                                                                                                            Oestrogen
    stress                                                                                                                                                 or androgen
                                                                                                                         HDACi
               HO•                                                                                                                Anti-androgen              Anti-oestrogen
                                                  DNMTi          Taxanes
                                                                                             Proteasome                                 HSP90

                                 Platinum salts            PRMTi
                                                                                                                                        HR
                                 Topoisomerase-i           EZH2i
                                 PARPi
                                      Chromatin
                                      relaxation         DNMTi                      PCFT                  HDACi                                                  HDACi
                                 PRMTi                                                                       BETi
                                 DNMTi
                                                                                             Transcription                                                         RNAPII
    γH2AX                                                   Methylated DNA                   factor
                            DNA damage repair
                                                                                                                                      BETi
                                                                                                                                                                   HR target
                                                            BETi               BRCA1                          HR promoter             CBPi                         genes
                                                                               RAD51
                            ATM Ku70                                           RBBP8
      Acetylated                                                               TOPBP1                                                                HRE
      protein                                                                                                                                                   Tumour
                                HDACi                                                                                                                           proliferation

c                                                                                            d

                                                                                                                     Regulatory
                                                                        E-cadherin                                   T cell
                                                       VEGFRi

                         HERi                                                                                    EZH2i                 M2 macrophage

                                                                                                                                                     HDACi
                                                                                                                                                                LSD1i
                                                                                                                                                     DNMTi
                                                                                                                                  Cytotoxic
                                                                                                                                  T cell
                                                       HDACi
            RAS
                            PI3Ki       PI3K                 HDACi

                                                                       HIF1α                 CCL5, CXCL9,                TCR
            RAF          BRAFi          AKT             HIF1α                                CXCL10, IFNα,
                                                                                             IFNβ                                  M1 macrophage
                          mTORi      mTOR
                                                                                                                         MHC
            MEK          MEKi
                                                                  Proteasome

            ERK                                                                                     Type I                                dsRNA
                                                                                                    interferon
                                                                                                    response

                                                       Oncoproteins

                  BETi
                                                            HDACi                                            EZH2i
                                 AXL, FGFR2,                                                                                                            LSD1i
                                                                                                                         Methylated
                                 HER3 (ERBB3),                                                                           histone
                                 MET, YAP                                     BIM                         ISG                                                            ERV
                                                                              TRAIL                                                     DNMTi
                                                                              CDH1
                                                                              (E-cadherin)
        Kinome adaptation

                                                                                                                                                  www.nature.com/nrclinonc
Reviews

◀   Fig. 3 | Rationale for epi-drug combination, with examples of synergy or reversal               the 8.3-month OS benefit observed in ENCORE 301
    of resistance mechanisms. Overview of epigenetic drug (epi-drug)-induced modulation             resulted from beneficial effects on hormone resistance,
    of sensitivity to four pillars of anticancer therapy: radiotherapy, cytotoxic chemotherapy      cancer stem cells, cell differentiation, anticancer immu-
    and DNA damage repair-targeting therapies (part a); hormonotherapy (part b); targeted           nity and/or sensitization to subsequent therapies is
    therapies (part c); and immunotherapies (part d). Prominent processes are detailed here
                                                                                                    unknown. Regardless, this finding led to the designa-
    but others exist. Histone deacetylase inhibitors (HDACi) and DNA methyltransferase
    inhibitors (DNMTi) synergise with genotoxic and/or cytotoxic therapies by interfering
                                                                                                    tion of entinostat plus exemestane as a breakthrough
    with the DNA damage response (DDR), notably through increased DNA damage and                    therapy by the FDA in postmenopausal women with
    disruption of the function of DDR proteins. Bromodomain and extra-terminal domain               advanced-stage hormone receptor-positive, endocrine-
    inhibitors (BETi) globally decrease DDR protein levels, resulting in functional DDR defects.    resistant breast cancer, and supported the develop-
    Protein arginine methyltransferase 5 inhibitors (PRMTi), among other mechanisms, more           ment of the E2112 study58. Results of this randomized,
    specifically induce aberrant splicing of DNA repair factors, thereby impairing their activity   double-blind, placebo-controlled phase III trial, in which
    (part a). HDACi interfere with hormone receptors (HR) at multiple levels, including             OS and PFS are co-primary end-points, are expected in
    transcriptional regulation, mRNA and protein stability (notably through modulation of           2021 and should shed light on the mechanisms under-
    acetylation levels of their chaperone protein HSP90) and transcription of nuclear target        lying the benefits of adding entinostat to exemestane
    genes. BETi suppress the transcription of the HR themselves or their target genes (part b).
                                                                                                    (if confirmed) and on relevant biomarkers for such
    HDACi modulate the stability of several tyrosine kinase receptors and downstream
    proteins, notably through modulation of acetylation levels of HSP90. HDAC4/6 inhibitors         combinations. Importantly and further supporting
    more specifically reduce hypoxia-inducible factor 1α (HIF1α) stability and activity, thereby    the above strategy, the randomized, double-blind,
    synergizing with anti-angiogenic agents. BET inhibitors prevent the transcription of            placebo-controlled, phase III ACE study, in which the
    alternative tyrosine kinase receptors and proteins that drive resistance through the            addition of the HDAC inhibitor tucidinostat to exemes-
    ‘kinome adaptation’ mechanism (part c). Treatment of cancer cells with enhancer of zeste        tane was evaluated in postmenopausal patients with
    homologue 2 inhibitors (EZH2i), lysine-specific histone demethylase 1A inhibitors (LSD1i),      advanced-stage hormone receptor-positive breast can-
    or DNMTi favours chemokine-dependent T cell attraction following enhanced expression            cer that had progressed after previous endocrine therapy,
    of CC-chemokine ligand 5 (CCL5), CXC-chemokine ligand 9 (CXCL9), CXCL10 or CXCL12.              revealed a significant prolongation of PFS in the experi­
    LSD1i and DNMTi also stimulate type I interferon response through increased expression          mental arm (7.5 months versus 3.8 months; HR 0.74,
    of endogenous retroviral elements (ERV), which eventually triggers T cell-mediated
                                                                                                    95% CI 0.58–0.98; P = 0.03)59.
    antitumour immunity following double-strand RNA (dsRNA) stress. Epi-drugs can also
    modulate immune cell transcriptional programmes. EZH2i elicit functional alterations in
    regulatory T cells and enhance the cytotoxicity of effector T cells, while other epi-drugs,     Hormone-sensitive prostate cancer. Observations similar
    such as LSD1i, DNMTi and HDACi, decrease myeloid-derived suppressor cell activity               to those described in preclinical models of ERα-positive
    and favour a M1-like macrophage-mediated antitumour immune response (part d).                   breast cancer have been reported in models of both
    ATM, ataxia telangiectasia mutated; BRAFi, BRAF inhibitor; CBPi, CREB-binding protein           hormone-sensitive and castration-resistant prostate can-
    inhibitor; Cyt c, cytochrome c; HERi, HER (ErbB) family receptor tyrosine kinase inhibitor;     cer models (CRPC)60–63 (Fig. 3b). Additionally, BET inhi-
    HRE, hormone response element; ISG, interferon-stimulated genes; MEKi, MEK                      bition prevents binding of the androgen receptor (AR)
    inhibitor; mTORi, mTOR inhibitor; PARPi, poly(ADP-ribose) polymerase inhibitor; PI3Ki,          to its target binding sites in the genome and the subse-
    PI3K inhibitor; RNAPII, RNA polymerase II; TCR, T cell receptor; VEGFRi, VEGFR inhibitor.       quent recruitment of RNA polymerase II; accordingly,
                                                                                                    addition of the BET inhibitor JQ1 to the anti-androgen
                                                                                                    enzalutamide or of the BET inhibitor OTX-015 to the
                                   combined with tamoxifen in 43 patients with ER-positive,         AR-agonist ARN-509 markedly reduced the growth of
                                   endocrine-resistant metastatic breast cancer55, revealed         enzalutamide-resistant prostate cancer xenografts in
                                   durable (lasting >24 weeks) objective responses in 19%           mice64,65. Early phase clinical trials of BET inhibitors com-
                                   and disease stabilization in 21% of patients. In the second      bined with next-generation anti-androgens in patients
                                   trial, ENCORE 301, a randomized placebo-controlled               with CRPC are ongoing (Supplementary Table 2).
                                   study of entinostat added to exemestane in 130 women                 In a phase II trial in patients with CRPC resistant to
                                   with ER-positive, endocrine-resistant advanced-stage             second-line antiandrogen therapy66, the addition of the
                                   breast cancer56, the entinostat combination resulted in          HDAC inhibitor panobinostat to the second-generation
                                   numerical but not statistically significant improvements         anti-androgen bicalutamide increased the median PFS
                                   in PFS (median 4.3 months compared with 2.3 months in            in comparison with that of historical control groups
                                   the placebo group; P = 0.11) but a striking significant          (34 weeks versus 16 weeks). However, this potential
                                   improvement in the exploratory OS end-point (median              benefit was only observed with high-dose panobino­
                                   28.1 months versus 19.8 months; P = 0.036). In particu-          stat (40 mg triweekly), which was associated with an
                                   lar, patients with a percentage change in the level of           increase in the incidence of severe toxicities: treatment
                                   PBMC pan-protein lysine acetylation above the study              interruption and dose reductions were needed in 27.5%
                                   median had a better median PFS than those with a per-            and 41% of patients, respectively, versus 11.5% and 4%
                                   centage change below the median (8.6 months versus               of patients receiving 20 mg triweekly. This finding sug-
                                   2.8 months; HR 0.32, 95% CI 0.13–0.79). Immunological            gests that more selective compounds with an improved
                                   analyses revealed significant depletion of blood granu-          therapeutic window might be required for the long-term
                                   locytic and monocytic myeloid-derived suppressor cells           clinical development of such combinations.
                                   (MDSCs; P = 0.029 and P = 0.002, respectively), as well
                                   as decreased CD40 expression in immature and mono-               Combinations with targeted therapy
                                   cytic MDSCs (P = 0.007 and P = 0.011, respectively) and          ErbB-targeted therapies. Preclinical evidence shows that
                                   increased HLA-DR expression in CD14+ monocytes                   resistance to agents targeting ErbB (also known as HER)
                                   and enrichment of this CD14+HLA-DR hi monocyte                   family receptor tyrosine kinases (RTKs) can be epige-
                                   subset (P = 0.015 and P = 0.0004, respectively)57. Whether       netically driven and thus reverted by epi-drugs (Fig. 3c).

    Nature Reviews | Clinical Oncology
Reviews

          Examples include the reversion of epithelial-to-              the basis of EGFR mutational status or using epigenetic
          mesenchymal transition (EMT) that can occur with              biomarkers73. Likewise, in the phase I/II IVORI study
          ErbB-targeted therapies through HDAC inhibition8              involving unselected patients with NSCLC74, gefinitib
          or the hyperacetylation of HSP90 chaperone proteins           plus vorinostat resulted in a median PFS of 9.1 months
          resulting from HDAC6 inhibition, which decreases              that did not seem to be better than historical data75. In
          interactions of HSP90 with stabilization of multiple          this regard, an ongoing randomized, placebo-controlled,
          RTKs, including EGFR (ErbB-1) 67. Target-specific             double-blind phase II/III trial (NCT02416739) of
          mechanisms also exist. A deletion polymorphism that           gefitinib or erlotinib with or without the HDAC3 inhib-
          results in alternative splicing of BCL2L11 (encoding          itor nicotinamide in patients with EGFR-mutant NSCLC
          the pro-apoptotic BCL-2-like protein 11, also known           should provide more relevant data and could potentially
          as BIM) is associated with resistance of EGFR-mutant          enable the identification of patient selection biomarkers
          NSCLC cell lines to EGFR tyrosine kinase inhibi-              for this combination (Supplementary Table 2).
          tors (TKIs) and shorter PFS durations in series of
          patients with EGFR-mutant NSCLC who received such             Anti-angiogenic agents. Tumour angiogenesis is mainly
          TKIs (6.6 months compared with 11.9 months in those           controlled by hypoxia-inducible factor 1α (HIF1α) and
          without the polymorphism) 68; notably, alternative            its transcriptional target VEGF. Inhibition of HDAC4
          splicing of BCL2L11 has been shown to be HDAC3-               and/or HDAC6 reduces HIF1α stability and transcrip-
          dependent and can be reverted by vorinostat in preclini­      tional activity by promoting its acetylation, polyubiq-
          cal models69. The clinical relevance of this approach is      uitination and subsequent proteasome degradation76.
          currently being evaluated in a phase I trial of vorino­       Accordingly, HDAC inhibitors and VEGF inhibitors
          stat and gefitinib in patients with EGFR-mutant NSCLC         have synergistic anti-angiogenic and antitumour effects
          harbouring the BCL2L11 deletion polymorphism                  in mouse models of various cancers77,78.
          (Supplementary Table 2).                                          Multiple early stage clinical trials have been condu­
              Targeting BRD4 using BET inhibitors also over-            cted to evaluate the safety of combinations of HDAC
          comes resistance to ErbB-targeted agents in preclinical       inhibitors and anti-angiogenic agents, including
          models of various tumour types, including anti-EGFR           vori­nostat with bevacizumab for clear cell renal cell
          anti­bodies in head and neck squamous cell carcinoma          carcinoma (RCC)79, resminostat with sorafenib for
          (HNSCC) and the dual EGFR and HER2 (ErbB-2)                   hepatocellular carcinoma 80 and panobinostat with
          TKI lapatinib in HER2-positive breast cancer, notably         bevacizumab for high-grade glioma81,82 (Supplementary
          through the abrogation of the transcription of alterna-       Table 1). All of these trials provide evidence that the
          tive kinases that drive acquired resistance, such as MET,     combinations are tolerable and provided evidence of
          AXL, FGFR2 or HER3 (ErbB-3) — a phenomenon                    clinical benefit79–82. The most recently reported phase I
          referred as ‘kinome adaptation’70–72 (Fig. 3c). In in vitro   study was conducted to assess the combination of
          models of lapatinib-resistant breast cancer, the combi-       abexinostat and pazopanib, with durable responses in
          nation of JQ1 and lapatinib has even been reported to be      seven of ten patients with pazopanib-refractory solid
          superior to lapatinib plus other TKIs targeting the bypass    tumours, including one response lasting >3.5 years83.
          pathways, and removal of lapatinib from the media             Among the 22 patients with RCC (including ten and
          allowed the cells to regrow despite continued JQ1 mono­       eight patients with prior progression on pazopanib
          therapy, suggesting that BET inhibition re-sensitized         and anti-angiogenic agents, respectively), the ORR was
          resistant cells rather than simply synergizing with lapa­     27% and the median duration of response was 10.5 months.
          tinib70,71. RNA sequencing (RNA-seq) revealed that            Notably, histone acetylation and HDAC2 expression
          the effect of BET inhibition was remarkably specific,         in PBMCs were positively correlated with durable
          as JQ1 monotherapy resulted in the downregulation             responses to treatment83. These promising results led to
          of only 8% of all expressed genes, but suppressed 27% of      the initiation of the ongoing randomized, double-blind,
          lapatinib-induced genes when used in combination              placebo-controlled phase III RENAVIV trial of the same
          with lapatinib70,71. Importantly, in HNSCC cell lines with    combination, with a crossover design, which will ena-
          acquired resistance to the anti-EGFR antibody cetuxi-         ble a thorough evaluation of clinical benefit in patients
          mab, this heterogeneous kinome adaptation is associated       with VEGFR TKI-naive advanced-stage RCC and could
          with increased levels of BRD4 expression72, which might       provide opportunities for biomarker identification
          therefore serve as patient selection biomarker for this       (Supplementary Table 2).
          combination.
              Multiple phase I and/or II studies have shown the         PI3K, AKT and/or mTOR inhibitors. Findings of several
          feasibility of combining HDAC inhibitors and ErbB             preclinical studies indicate the potential of combinations
          inhibitors (predominantly EGFR TKIs in patients with          of HDAC inhibitors and PI3K–AKT–mTOR path-
          NSCLC, but also the anti-HER2 antibody trastuzumab            way inhibitors, including the observations that HDAC
          in patients with breast cancer) although no clear signs       inhibition can destabilize AKT84,85, suppress survivin
          of delays or reversal of resistance have been observed        expression86, inhibit EMT and invasiveness, or increase
          (Supplementary Table 1), perhaps because patients with        oxidative stress87,88, and induce autophagy89. Inhibi­
          epigenetically driven resistance to trastuzumab were not      tion of HDACs and the PI3K–AKT–mTOR pathway
          selected. For example, patients with NSCLC enrolled in        also converges on the upregulation of tumour suppres-
          the ENCORE-401 randomized phase II study of erlotinib         sor genes, such as FOXO1 in MYC-driven medullo­
          plus either entinostat or placebo were not selected on        blastoma90 (Fig. 3c). Synergy also occurs between mTOR

                                                                                                       www.nature.com/nrclinonc
Reviews

                       inhibitors and BET inhibitors in preclinical models of             BET inhibitors also synergize with BRAF inhibi-
                       triple-negative breast cancer91. Interestingly, BET inhi-      tors in preclinical models of BRAF-mutant melanoma,
                       bition can both prevent and revert resistance to mTOR          which leads to increased cytotoxicity and epigenetic
                       inhibition in such models91. Similar to the aforemen-          modulation of dedifferentiated precursor cells102,103.
                       tioned observations relating to countering the kinome          Furthermore, BET inhibition can overcome resistance
                       adaptation mechanism of resistance, transcriptional            to dual MEK and TBK1 inhibition in KRAS-driven
                       upregulation of RTKs upon PI3K inhibition can be               NSCLC models by blocking transcriptional activation
                       overcome by preventing BRD4 binding at the pro-                of YAP1, which is a core component of the Hippo sig-
                       moter regions of RTK genes using BET inhibitors92.             nalling pathway that often contributes to tumorigene-
                       Furthermore, when single-agent treatments are inef-            sis104. Interestingly, the investigators who conducted this
                       fective, dual PI3K–BET inhibition can induce apoptosis         study anticipated that continuous BET inhibition would
                       across multiple cancer cell lines and mammary tumour           not be tolerable, and used in vitro pharmacodynamic
                       models92. Similar results have been reported for pre-          modelling to optimize an intermittent sequential admin-
                       clinical models of epithelial ovarian cancer and neuro­        istration schedule that was tolerable and proved to be
                       blastoma, in which PI3K inhibition overcomes resistance        equally effective in mice. Clinical data from studies eval-
                       to BET inhibitors93,94. These data pose a considerable         uating such combinations are not currently available, but
                       challenge to the drug development dogma that drugs             considering that the toxicities have been a major chal-
                       that do not display single-agent activity should not           lenge to the application of epi-drug combinations, with
                       progress beyond phase I testing95 and suggests that cer-       long-term toxicities often being a limiting factor96,105,106,
                       tain drugs might only be effective when rationally used        such intermittent scheduling approaches should be
                       in combinations. This approach might be especially             encouraged and applied more systematically.
                       pertinent with PI3K inhibitors and first-generation epi-
                       drugs, the single-agent activity of which has thus far been    PARP inhibitors. Preclinical synergy between HDAC
                       limited in patients with solid tumours.                        inhibitors and poly(ADP-ribose) polymerase (PARP)
                           Phase I trials in which mTOR inhibitors were               inhibitors has been reported in models of triple-negative
                       combined with vorinostat in patients with various              breast107, prostate108 and ovarian cancer109. Mechanisti­
                       advanced-stage solid tumours have provided some evi-           cally, HDAC inhibitors induce a ‘BRCAness’ pheno-
                       dence of therapeutic activity, but thrombocytopenia            type following downregulation of proteins involved in
                       limited doses to levels below the full dose of each agent      homologous recombination, including RAD51, BRCA1
                       individually, suggesting that synergistic effects also occur   and UHRF1. Notably, the combination of olaparib and
                       in non-malignant cells96,97 (Supplementary Table 1).           vorinostat is to be evaluated in patients with metastatic
                       Again, these findings warrant the evaluation of discon-        breast cancer (NCT03742245). Furthermore, in the past
                       tinuous and/or sequential drug administration sched-           2 years, the results of four independent studies have
                       ules, targeted drug delivery and/or biomarker-driven           shown a strong synergy between PARP inhibitors and
                       patient selection in order to improve the therapeutic          BET inhibitors, notably, in preclinical models of homo­
                       index. Early phase trials of HDAC inhibitors combined          logous recombination-proficient ovarian cancer110–113
                       with mTOR inhibitors are ongoing in patients with              (Fig. 3a). Mechanistically, BET inhibitors downregulate
                       various types of solid tumour, including the combina-          various proteins involved in homologous recombina-
                       tion of everolimus and panobinostat in patients with           tion, including BRCA1, BRCA2, RAD51 and RBBP8
                       H3.3 or H3.1 K27M-mutant glioma (NCT03632317;                  (also known as CtIP), suggesting broad effects on the
                       Supplementary Table 2).                                        expression DDR-related genes rather than a targeted
                                                                                      synthetic lethality (as epitomized by PARP inhibition
                       BRAF and/or MEK inhibitors. In preclinical studies,            in cancers associated with germline BRCA mutations).
                       HDAC inhibitors revert acquired resistance to BRAF             Interestingly, a reciprocal observation was made in mod-
                       and/or MEK inhibitors, notably, by upregulating trans­         els of BET inhibitor-resistant CRPC, whereby acquired
                       cription of the caspase 8 inhibitor c-FLIPL98, modulating      polycomb repressive complex 2 (PRC2)-mediated silenc-
                       PI3K signalling99 or downregulating melanocyte line-           ing of DDR-related genes led to increased sensitivity to
                       age pathways (restoration of which drives adaptive drug        PARP inhibitors114. Combinations of PARP and BET
                       resistance in melanoma cells)100 (Fig. 3c). In 2019, HDAC      inhibitors have not yet been evaluated in clinical studies.
                       inhibition was further shown to drama­tically enhance          We anticipate that rational patient selection and optimal
                       the effectiveness of BRAF and/or MEK inhibition in             dosing and scheduling will be key to maximizing the
                       preclinical models of both sensitive (BRAF-mutant)             therapeutic window and, thus, potential clinical utility,
                       and resistant (NRAS-mutant or NF1-mutant) RAS                  considering that thrombocytopenia is a dose-limiting
                       pathway-driven melanoma101. Specifically, HDAC inhibi­         toxicity (DLT) of HDAC, PARP and BET inhibitors.
                       tors induced chemical synthetic lethality via suppression
                       of the non-homologous end-joining DDR pathway on               Epi-drug–epi-drug combinations. Owing to the interplay
                       a background of BRAF and MEK inhibitor-mediated                among various epigenetic processes, combining several
                       disruption of the alternative homologous recombina-            epi-drugs might be a promising approach to anticancer
                       tion pathways101. Importantly, MGMT expression was             therapy. For example, overexpression of KMT2A (which
                       found to be characteristic of tumours with inherent            encodes lysine N-methyltransferase 2A, also known as
                       defects in DNA repair and to predict responsiveness to         MLL1) drives resistance of EZH2-aberrant cancers to
                       such combination therapy101.                                   EZH2 inhibition via the interaction of MLL1 with the

Nature Reviews | Clinical Oncology
Reviews

          CBP and EP300 co-activators, which results in losses           by divergent results obtained with EZH2 inhibitors.
          of H3K27 methylation and reciprocal gains in H3K27             Inhibition of EZH2 in T cells leads to enhanced anti-
          acetylation, thereby triggering oncogenic transcriptional      tumour responses with ICIs in preclinical models by
          reprogramming; however, BET inhibition disrupts sig-           eliciting functional alterations in regulatory T (Treg)
          nalling mediated by H3K27 acetylation and re-sensitizes        cells and by increasing the cytotoxicity of effector
          cancer cells to EZH2 inhibitors115. Similarly, BET inhi-       T cells 134. In line with these results, a patient with
          bition can reverse resistance to DNMT inhibitors in            SMARCB1-negative chordoma and disease progression
          preclinical models of colorectal cancer (CRC)116.              on EZH2 inhibition with tazemetostat had a durable
                                                                         and exceptional abscopal response to radiotherapy last-
          Combinations with immunotherapies                              ing >2 years; comparisons of tumour biopsy samples
          Antitumour immunity involves the complex interplay             obtained prior to treatment and during treatment with
          among immune, cancer and stromal cells. Specific               tazemetostat revealed a substantial increase in intra­
          DNA-modifying or histone-modifying enzymes and                 tumoural and stromal infiltrates of CD8+ cytotoxic T cells
          histone chaperones, including DNMTs, KMT1A (also               and FOXP3+ Treg cells, as well as enhanced expression
          known as SUV39H1) and chromatin assembly factor 1              of the immune-checkpoint proteins programmed
          (CAF-1), respectively, contribute to both the immuno­          cell death 1 (PD-1) and lymphocyte activation gene 3
          genicity of cancer cells and the lineage commitment            (LAG3) on T cells135. These findings suggest that EZH2
          and/or maturation of immune cells117,118; thus, epi-drugs      inhibition can promote sustained antitumour immune
          can potentially be used to modulate antitumour                 responses and lead to immune-checkpoint activa-
          immunity119,120. Several mechanisms of resistance to           tion135, although the consequences of these effects,
          immune-checkpoint inhibitors (ICIs)121 can be reverted,        particularly with regard to ICI sensitivity, remain
          in preclinical models, by epigenetic manipulations             unclear at present. Contradictory pro-tumour effects
          (Fig. 3d) — a topic that we have reviewed elsewhere120.        have been reported preclinically with the EZH2 inhib-
          Here, we review important new data that have been              itor GSK126, which resulted in increased numbers of
          reported in the past few months.                               MDSCs but decreased numbers of CD4+ T cells and
              Preclinical modulation of adaptive immunity has            IFNγ+CD8+ T cells in the tumour microenvironment136.
          been described with the ablation of LSD1 or inhibition         Together, these observations suggest that the cellular
          of HDACs or DNMTs, which can enhance antigen pres-             composition of the tumour and drug administration
          entation, in part through the induction of dsRNA pro-          schedule can influence the immunomodulatory effects
          duction from endogenous retrovirus (ERV) genes and the         of EZH2 inhibition.
          resultant type I interferon response, leading to increased         Some emerging clinical findings underscore the
          T cell infiltration of tumours and synergistic effects with    promise of combinations of epi-drugs and ICIs, notably
          ICIs122–124. Similarly, DNMT inhibitors can further syner-     in tumours that are resistant and/or refractory to ICIs
          gize with HDAC inhibitors to induce ERV transcription,         (Supplementary Table 1). For example, the combination
          promote CC-chemokine ligand 5 (CCL5)-dependent                 of the anti-PD-1 antibody pembrolizumab and vorino­
          T cell infiltration and favour T cell memory and effector      stat in a phase I/Ib study resulted in one confirmed partial
          phenotypes, and to downregulate immunosuppressive              response (PR) and eight stable disease responses among
          MYC signalling, thereby reverting immune evasion in            24 patients with ICI-resistant metastatic NSCLC137.
          NSCLC models125. Interestingly, such an effect has also        Moreover, preliminary results from the ENCORE-601
          been described with CDK4/6 inhibitors, which indirectly        phase Ib/II trial evaluating the combination of pembroli-
          reduce the activity of DNMT1 and thereby increase ERV          zumab and entinostat demonstrate a PR in one patient
          expression126; this observation highlights how some            with microsatellite-stable CRC (MSS-CRC), PRs in nine
          non-epigenetic drugs can have clinically relevant indi-        patients and a complete response in one patient with
          rect epigenetic effects. Moreover, chemokine-dependent         ICI-resistant melanoma and PRs in seven patients
          T cell attraction has also been described following            with ICI-resistant NSCLC (among 16, 53 and 67 patients,
          enhanced expression of CXCL12 resulting from DNMT              respectively)138–140; in the NSCLC expansion cohort, a
          inhibition127 or of CCL5, CXCL9 and CXCL10 from                high level of peripheral CD14+CD16− HLA-DRhi clas-
          LSD1128 or EZH2 inhibition129. Finally, the development        sic monocytes at baseline (defined as ≥13.1% of live
          of selective clinical-grade inhibitors of the histone-lysine   PBMCs per millilitre of blood) was correlated with
          N-methyltransferase KMT1A might bring future clini-            a better ORR (21.1% versus 6.5%) and median PFS
          cal benefits via enhancement of the long-term memory           (5.3 months versus 2.7 months)138. Contrary to these
          reprogramming capacity of CD8+ T cells118.                     single-arm phase I–II trials of HDAC inhibitors, how-
              With regard to innate immunity, LSD1, DNMT and             ever, the only randomized study with results reported
          HDAC inhibitors decrease the activity of MDSCs                 to date revealed an increased risk of toxicities and no
          and promote an M1-like macrophage-mediated anti-               therapeutic benefit from combining pembrolizumab
          tumour immune response in mouse models of breast               with the second-generation DNMT inhibitor CC-486
          or pancreatic cancer130,131. HDAC inhibition can also          (oral azacytidine)141. The choice of full-dose CC-486 and
          enhance the cytolytic activity of NK cells and TLR9            the absence of patient selection, either for likely sensi-
          agonist-induced abscopal responses (that is, in lesions        tivity to pembrolizumab or CC-486, might at least par-
          beyond the one treated with radiotherapy)132,133.              tially explain these results. Results from other studies
              Of note, modulation of immunity using epi-drugs            of epi-drug plus ICI combinations are difficult to inter-
          will probably require some fine-tuning, as illustrated         pret at present because most data are preliminary and

                                                                                                         www.nature.com/nrclinonc
You can also read