Histone deacetylases as therapeutic targets in hematologic malignancies
Among the epigenetic modifications associated with the development of human cancer, alterations of the acetylation status of histones play a prominent role (see [
1] for review). The extent of acetylation and deacetylation on different positions of core histones is determined by the antagonistic activity of histone acetylases (HAT) and histone deacetylases (HDAC) and alters the nucleosomal conformation of both transformed and non-transformed cells. Deacetylation of histones by HDACs hinders the accessibility of DNA to transcription factors that are involved in determining malignant cell behaviour, thereby changing their activity, subcellular localisation and interaction partners. In addition, acetylation is an important post-translational modulation of a wide range of nuclear and cytoplasmic proteins involved in the regulation of a multitude of cellular functions (e.g., p53, tubulin, heat-shock protein 90). A disrupted equilibrium between HDACs and HATs, with preponderance of deacetylase activity, leads to transcriptional repression of a diverse set of genes involved in the regulation of cell proliferation, differentiation and apoptosis. Aberrant gene transcription caused by abnormal activity of HATs and HDACs is commonly observed in leukemia and lymphoma [
2‐
6]. Accordingly, modulation of protein lysine acetylation through inhibition of histone deacetylases (HDACs) is currently being considered as an attractive new therapeutic strategy for acute myeloid and lymphoid leukemias and myelodysplastic syndromes.
Epigenetic alterations in acute lymphoblastic leukemia
Transcriptional silencing of genes due to epigenetic mechanisms is an important alteration in acute lymphoblastic leukemia (ALL), with research so far having focused on the role of DNA methylation. An exception is the t(12;21)(p13;q22) (TEL-AML1), which is restricted to precursor B-cell lineage leukemia and is the most common (~25%) translocation in childhood acute lymphoblastic leukemia (ALL). Like AML1 in the CBF acute myeloid leukemias, the abnormal TEL-AML1 fusion protein can bind to core enhancer sequences. Instead of activating transcription through recruitment of co-activators and HATs, it recruits co-repressors and HDACs. Since TEL-AML1-induced transcriptional repression was shown to be reversed by HDAC inhibitors, TEL-AML1-positive ALL may be considered likely to benefit from treatment with HDAC inhibitors [
21,
22] Preclinical studies testing this hypothesis are reviewed elsewhere in this paper.
Down-regulation of microRNAs (miRNA) by epigenetic mechanisms, most notably by DNA methylation, may contribute to tumorigenesis. A recent study explores the epigenetic alterations of miRNAs in ALL by analyzing the methylation and chromatin status of the miR-124a loci in ALL [
23]. Expression of miR-124a was down-regulated in ALL by hypermethylation of the promoter and histone modifications including decreased levels of acetylated histone H3. Epigenetic down-regulation of miR-124a induced an up-regulation of its target, cyclin-dependent kinase 6 (CDK6) as well as phosphorylation of retinoblastoma (Rb), and contributed to the abnormal proliferation of ALL cells both in vitro and in vivo. CDK6 inhibition by inhibition of HDACs decreased ALL cell growth in vitro. Although an analysis of 353 patients diagnosed with ALL revealed an association between a higher relapse and mortality rate and hypermethylation of the tumor suppressor microRNA Hsa-miR-124a, these results nevertheless provide the rationale for therapeutic strategies in ALL that either target the epigenetic regulation of microRNAs and/or directly target the CDK6-Rb pathway, e.g. by deacetylase inhibitors.
DAC inhibitors in hematologic malignancies: molecular targets and biochemical / pharmacologic properties
At present, 18 HDAC isoforms are known which are grouped into four classes. From a functional as well as from a translational point of view, class I HDACs (isoforms 1,2,3,8) are the best characterized proteins of this family. Less extensive information is available on function and expression of class II isoforms [
4‐
7,
9,
10] while little is known about class III HDACs (the sirtuins) and HDAC11, which on the basis of distinct structural properties has been suggested to constitute an HDAC class on its own.
A large variety of well established as well as novel HDAC inhibitors possess antineoplastic activity in vitro and in animal models in vivo. Some of these inhibitors are unselective, i.e. they target class I and II but not class III, while others target only specific HDAC classes or isoforms. As these compounds also inhibit deacetylation of numerous non-histone proteins, they should preferably be referred to more broadly as deacetylase inhibitors (DACi). DACi may be subdivided into two categories, which inhibit nuclear and cytoplasmic deacetylases, respectively.
The chemical structure of DACi encompasses three subunits: (1) a zinc-chelating group, (2) a usually hydrophobic “spacer” group, and (3) a catalytic domain which determines the specificity of the compound. DACi inhibitors are classified by structure and include the short-chain fatty acids valproic acid and sodium butyrate, the cyclic tetrapeptides romidepsin (depsipeptide, FK228, FR901228), Trapoxin A, and Apicidin, the hydroxamic acids vorinostat (suberoylanilide hydroxamic acid, SAHA), trichostatin A (TSA), LAQ824, panobinostat (LBH529) and PXD101, the benzamides MS-275, CI-994, und MGCD-0103, cyclic tetrapeptides, electrophilic ketones (trifluoromethylketone), and others (depudecin, SNDX-275, and isothiocyanates) [
4]
Although these DACi differ in structure, potency and possibly HDAC enzyme selectivity, they target primarily class I and II HDACs and do not affect the activity of the class III sirtuins (reviewed by Marchion and Münster [
24]) It is becoming increasingly clear that the downstream effects of HDAC inhibition, which ultimately leads to growth inhibition and apoptosis in different tumor types, depend upon both the HDAC inhibitor and the cell type [
25‐
27]. Only very recently, research began to focus on the histone acetylation status in human tumours in general and the specific expression of HDAC isoforms in solid as well as hematological malignancies.
Non-histone targets of DACi in acute myeloid leukemia
HATs and DACs affect the acetylation status of lysine residues not only of histones but also of transcription factors (TF) (eg, p53, E2F1, GATA1, RelA, Y1, MAD/MAX, TFIIE and TFIIF, and hormone receptors). Altered acetylation of TFs may affect their DNA binding and transcriptional activity [
4‐
6,
28]. In addition, DACs have been shown to deacetylate a multitude of proteins other than histones or transcription factors, e.g. the cytoskeleton protein-tubulin, β-catenin, DNA repair enzymes and the heat shock protein 90 (hsp90) [
29‐
35]. Accordingly, modulation of DAC function by DACi may not only affect gene transcription but also modify the stability of proteins, as well as the ability of proteins to interact with DNA and other proteins involved in important biologic functions in the leukemic cells [
36]. The sheer number of transcription factors known to be acetylated suggests that the acetylation of these nonhistone proteins may have as much regulatory effect on transcription as the acetylation of histone proteins. A more comprehensive description of nonhistone targets of acetylation has been reviewed recently [
36‐
38].
The molecular basis for the relatively selective antitumor activity of DACi is unknown. Insigna et al. investigated the effects of DACis on leukemias expressing PML-RAR or AML1-ETO [
39]. Even though these oncoproteins are known to initiate leukemogenesis through deregulation of HDACs, it was shown that oncogene expression is not sufficient to confer DACi sensitivity to normal cells. DACi-induced induction of apoptosis in leukemic cells was found to be p53 independent and depend upon activation of the death receptor pathway (TRAIL and Fas signaling pathways). Interestingly, TRAIL, DR5, FasL and Fas were upregulated by DACis in the leukemic cells, but not in normal hematopoietic progenitors, indicating that sensitivity in leukemias is a property of the fully transformed phenotype and depends on DACis-induced activation of a specific death pathway.
DACi-mediated induction of TRAIL was also identified as a mediator of the selective anticancer action by Nebbioso et al. [
40]. Expression of TRAIL, by directly activating the TNFSF10 promoter, triggered tumor-selective death signaling in acute myeloid leukemia (AML) cells, without inducing apoptosis in normal CD34(+) progenitor cells. DACi induced proliferation arrest, TRAIL-mediated apoptosis and suppression of AML blast clonogenicity occurred irrespective of karyotype.
Heat shock proteins (HSPs) are molecular chaperones that stabilize folding and conformation of both normal and oncogenic proteins. These chaperones thereby prevent the formation of protein aggregates. HSPs are often overexpressed in human malignancies, including AML (see [
41] for review), and are the main chaperones required for the stabilization of multiple oncogenic kinases involved in the development of AML. HSP90 client proteins are involved in the regulation of apoptosis, proliferation, autophagy and cell cycle progression; several of these proteins are considered possible therapeutic targets for treatment of AML. The results from initial phase I/II clinical trials testing HSP90 inhibitors have documented that HSP90 inhibition can mediate antileukemic effects in vivo. HSP90 activity is also regulated by posttranscriptional modulation, and HSP90 inhibition can thereby be indirectly achieved through increased acetylation caused by DACi (see section
Combination with other agents)
Clinical experience with deacetylase inhibitors in MDS and AML
To date, clinical examination of DACi in patients with AML has met with limited success. Current experience is mostly restricted to phase I or phase II studies involving small numbers of patients.
Valproic acid is the only DACi that has been clinically investigated in larger numbers of patients with MDS. An early clinical trial involving 18 MDS patients showed an overall response rate of 37% (1 partial response (PR) and 6 hematologic improvements (HI)). All-
trans retinoic acid (ATRA) did not enhance the clinical activity in patients who had not already demonstrated a response to VPA [
76]. In a subsequent report of 43 patients with MDS, the overall response rate was 35% (1 PR, and 15 HI) [
77]. Efficacy of the VPA +/- ATRA treatment seems to be inversely correlated with the stage of the disease. According to Kuendgen et al., low or intermediate-I IPSS scores or a normal bone marrow blast count proved to be good predictors for response [
78].
The combination of VPA with ATRA was also tested in AML patients who were considered ineligible for intensive chemotherapy. A disappointingly low response rate of only 5% with no CR was observed among 40 patients receiving this combination [
79] Similarly, none of 26 patients with high-risk AML achieved a CR in another study examining combined treatment with VPA and ATRA [
80]. Results did not improve when ATRA was added later on in another pilot study with 8 AML patients [
81]. Nevertheless, 5 out of 11 patients with de novo AML responded to a therapy with VPA, ATRA and theophylline (1 CR, 2 complete remissions with incomplete recovery of peripheral blood counts (CRi), 2 HI) [
82].
In view of the development of the novel, considerably more potent DAC inhibitors vorinostat, panobinostat, romidepsin and the isotype-specific MGCD0103, the use of VPA in patients with AML, as a single agent or in combination with ATRA, does not appear very promising. The role of VPA in combination with hypomethylating agents will be addressed in a later section of this review.
Vorinostat was investigated as a single agent in a phase I study with 31 AML and 10 MDS patients. The maximal tolerated dose was 200 mg BID. Seven patients experienced hematologic improvement, including 4 CR in patients with AML. Increased histone acetylation was observed at all dose levels [
83]. These favorable results were not confirmed in a randomized phase II trial, in which only one out of 37 AML patients achieved CR [
84].
Panobinostat (LBH589) was administered in a phase I study as a 30-min infusion on days 1 to 7 of a 21-day cycle. Fifteen patients with AML (
n = 13), ALL (
n = 1), or MDS (
n = 1) were treated. At doses <11.5 mg/m
2, i.v. panobinostat was well tolerated with consistent, albeit transient antileukemic and biologic effects: in 8 of 11 patients peripheral blasts declined, but rebounded following the 7-day treatment period. The median acetylation of histones H2B and H3 in CD34(+) and CD19(+) cells increased significantly during therapy, as did apoptosis in CD14(+) cells [
85]. Oral panobinostat was evaluated in patients with advanced hematologic malignancies. Doses of ≥ 40 mg weekly lead to 2 CRs out of 26 evaluable AML patients [
86].
Romidepsin is a potent, bicyclic tetrapeptide with DAC-inhibitory activity. Twenty patients with AML were treated at 13 mg/m(2)/d on days 1, 8, and 15 of a 28-day cycle. Antileukemic activity was observed in 5 of 7 patients with chromosomal abnormalities known to recruit HDACs, including those involving core binding factor (CBF). Two patients had clearance of bone marrow blasts and 3 patients had a greater than 50% decrease in bone marrow blasts. These responses were associated with a significant increase in MDR1, p15 and p14 expression [
87]. In a phase II study in patients with high-risk MDS and AML, one patient achieved a CR, 6 patients experienced stable disease [
88].
The hydroxamate DACi belinostat was administered as a 30-min i.v. infusion on days 1-5 of a 21-d cycle to 16 patients with advanced hematologic malignancies. No complete or partial remissions were noted in these heavily pre-treated patients [
89].
In contrast to the potent pan-DACi panobinstat, belinostat and vorinostat, the orally administered benzamide MGCD0103 is a selective inhibitor of HDAC1, 2, 3 (class 1) and 11 (class 4) and does not inhibit class 2 HDACs. In a phase I study, 3 of 29 patients with leukemia or MDS achieved a CR [
90].
A phase I dose escalation study of the synthetic benzamide derivative MS-275 was completed in 75 patients with advanced acute leukemias. In spite of biologic alterations such as increase in protein and histone H3/H4 acetylation, p21 expression, and caspase-3 activation in bone marrow mononuclear cells, no clinical responses were seen [
91].
Toxicity of deacetylase inhibitors
The most frequently adverse effects observed during DACi treatment for hematologic malignancies include gastrointestinal toxicity with nausea, vomiting and diarrhea, as well as fatigue, thrombocytopenia, neutropenia and non-specific ECG changes such as flattened T-waves, ST-segment depression, and QT-prolongation [
92].
These changes are dose-dependent and represent a class effect of DACi. The pathogenesis of thrombocytopenia is unclear, but does not appear to involve cytotoxic mechanisms, as they are reversible within a few days of drug discontinuation. The other adverse events typically also rapidly revert to normal, or are easily controlled by prompt institution of supportive therapy, e.g. in the case of diarrhea. Nevertheless, the different DACi display some notable differences regarding their toxicity profiles.
Epigenetic therapy combining deacetylase- and DNA-methyltransferase-inhibitors and/or ATRA in AML and MDS
The first clinical studies to combine DACi and hypomethylating agents in patients with MDS and AML were conducted with VPA (see Table
1). Promising results were obtained in each of these studies [
93‐
96] indicating both improved and accelerated responses with combined epigenetic therapy: the time to response was significantly shorter with combined therapy (1–3 cycles) than with single-agent DNMT (4–6 cycles). VPA was targeted to therapeutic plasma levels of at least 50 μg/mL to increase the efficacy of the hypomethylating agent [
96], but no association between histone acetylation and response was observed [
97]. Further correlative studies demonstrated reversal of p15 or CDH-1 promoter methylation during the first cycle of therapy in all six responding patients, whereas no demethylation was observed in any of the six non-responders [
97].
Table 1
Clinical trials evaluating DACi in combination with DNMT inhibitors
Patient numbers | N = 54 | N = 25 | N = 53 | N = 62 | N = 32 |
Diagnosis | AML (89%), MDS (11%) | AML | AML (92%) MDS (8%) | MDS | AML (56%) MDS (44%) |
Response (CR, CRi) | 22% | 32% | 28% | 12% | 9% |
No of previously untreated patients | 11 | 12 | 33 | na | 8 |
Response of previously untreated patients | 50% | 58% | 42% | na | 25% |
Survival of responding patients | median 15 months | 3–10 months | > 5 months | na | 8–19+ months |
Even though these trials were designed primarily as dose-finding studies, the 20–30% remission rate (CR and CRi) was higher than generally observed with 5-azacitidine or decitabine when used as a single agent. This also holds true for a combination trial of SAHA and decitabine: of 61 patients evaluable for response, a CR or CRi was achieved by 18% patients with MDS, 8% with relapsed/refractory AML, and 36% with untreated AML. [
98] Even higher CR/CRi rates of up to 40 to 60% were achieved in untreated patients with AML by combining a hypomethylating agent with VPA. A recent follow up of these untreated patients revealed that responders received more cycles of therapy and had significantly longer survival. Non-responding patients had a higher WBC and higher bone marrow blast count at the start of therapy. In general, patients who relapse after combination epigenetic therapy appeared to have a poor prognosis [
99].
Numerous further combination studies using potent HDACi are currently being conducted, the results of which are being awaited with interest.
Combination of deacetylase inhibitors with cytotoxic agents
The rationale for phase I trials of the sequential combination of vorinostat followed by cytotoxic agents in patients with acute leukemias stems from
in vitro studies showing that the sequence-dependent interaction of vorinostat and cytarabine or etoposide arrested the cells in G1 or G2 phase during vorinostat treatment and allowed recovery into S phase after removal of vorinostat [
68]. Concurrent administration of vorinostat and idarubicin for 3 days was demonstrated in a phase I trial in 41 patients with refractory leukaemia (90% AML), with 2 CRs, 1 CRi and 4 marrow responses. Correlative studies demonstrated histone acetylation in patients on therapy and modulation of CDKN1A and TOP2A (topoisomerase II) gene expression and a dose-related elevation in plasma vorinostat concentrations [
100].
Preliminary results of a phase II study of vorinostat followed by idarubicin and cytarabine with 45 mostly high-risk AML patients aged > 65 years were reported. Induction therapy consisted of oral vorinostat 500 mg TID (days 1 to 3), idarubicin 12 mg/m
2 iv (days 4 to 6) and cytarabine (1.5 g/m
2 as a continuous infusion on days 4 to 7) followed by a maximum of 5 consolidation cycles with dose-reduced chemotherapy. Complete remission (CR) after one course of therapy was achieved in 35 patients and 1 patient achieved a CRi for an overall response rate of 80%. No excess toxicity with the addition of vorinostat has been observed compared to standard induction therapy [
101].
DACi-associated drug resistance
The rationale for developing deacetylase inhibitors as anti-leukemic agents is based on evidence showing induction of cellular differentiation, growth arrest, and apoptosis of malignant cells, but there is no evidence that only genes involved in leukemogenesis are targeted by DACi. Recent reports indicate that expression of the multidrug resistance-1 (MDR1) gene is also regulated by epigenetic mechanisms, raising the possibility that this and other drug transporters able to counteract the cytotoxicity of various anti-leukemic drugs may be upregulated by DACi. This was in fact shown to be the case in a recent
in vitro study demonstrating that in AML cells, expression of MDR1, breast cancer resistance protein (BCRP), and multidrug resistance-associated proteins (MRP) 7 and 8 were induced by phenylbutyrate, valproate, vorinostat or trichostatin A in a dose- and time-dependent manner. The pattern of gene induction by different DACi was shown to be cell line specific and associated with hyperacetylation of histone proteins in the promoter regions of MDR1, BCRP, and MRP8. Drug-induced apoptosis was impaired in KG-1α cells treated with phenylbutyrate, resulting in resistance to daunorubicin, mitoxantrone, etoposide, vinblastine, paclitaxel, topotecan, gemcitabine, and 5-fluorouracil [
102].
Similarly, the combination of ATRA with depsipeptide (FK228) induced MDR1 expression in NB4 promyelocytic leukemia (APL) cells, which normally do not express MDR1 and are highly sensitive to anthracyclines. Upregulation of MDR1 expression occurred via increased H4 and H3-Lys9 acetylation of the MDR1 promoter, and prevented doxorubicin-induced growth inhibition and apoptosis in APL cells. G1 cell-cycle arrest and upregulation of p21 mRNA may have further impaired the induction of apoptosis of cells in G2 phase. Conversely, initial exposure to doxorubicin followed by ATRA/FK228 treatment enhanced apoptosis [
103]. These results indicate that epigenetic mechanisms leading to a drug resistance phenotype broader than the “classic multidrug resistance” may be activated by DACi exposure of AML cells, an effect which might impair therapeutic efficacy. This highlights the importance of investigating mechanism-based sequential therapies in clinical trials that combine DAC inhibitors with other agents commonly used for treatment of acute leukemias.
DACi as treatment for acute leukemias and MDS: a critical appraisal and future perspectives
Despite the profound preclinical activity of DACi against acute myeloid leukemia cells, their clinical development as treatment for myeloid malignancies has been less straightforward than for several lymphoid malignancies, most notably cutaneous T cell lymphoma. The to date modest success in AML and MDS highlights the limitations of our understanding of the complex interactions between epigenetic and genetic changes in these malignancies, and the paucity of our knowledge concerning the biological function of individual HDAC enzymes and the pleiotropic cellular effects of DACi. While it has been established that treatment of leukemic cells with DACi induces cell death, differentiation and/or cell-cycle arrest, they may also effect neoplastic growth by influencing the tumor microenvironment, regulating host immune responses or modifying the properties of normal hematopoietic cells. Thus, the beneficial effect of valproic acid in patients with low risk MDS may actually be a reflection of the unique ability of VPA to enhance normal hematopoietic function, whereas patients with high risk MDS or AML generally do not benefit because of the insufficient anti-leukemic activity of VPA in conjunction with the depletion of normal, potentially VPA-responsive progenitor cells.
In addition, the initial concept that DACi mediate their biological effects only through the regulation of gene expression via direct hyperacetylation of histones is no longer valid, following realisation that DACi acetylate diverse non-histone proteins, thereby regulating a broad range of cellular functions independent of transcriptional mechanisms. It is therefore not surprising that the actions of DACi are cell-context dependent, and will differ according to the differential expression and function of the individual HDACs in a given leukemia. HDAC enzyme expression is therefore likely to determine differential sensitivity to various types and dosages of DAC inhibitors, as well as their toxicity profile.
Limited clinical experience has so far demonstrated the relative safety of DACi currently in clinical testing, but adverse effects that become particularly relevant with long-term use have been noted, some of which, such as fatigue and diverse gastrointestinal complaints, appear to be class effects. Moreover, safety aspects will need to be reevaluated when DACi are combined with other anticancer agents, an obvious next step in their clinical development given available evidence that the clinical benefit derived from epigenetic and chromatin modifiers will be accrued when they are combined with chemotherapy or other targeted anti-leukemic agents. The potential of DACi to upregulate drug efflux pumps is a notable example for potentially detrimental drug-drug interactions, although there is as yet no evidence that this is clinically relevant. A rational selection of appropriate combination partners for DAC inhibitors will require an improved understanding of the specific epigenetic and genetic aberrations of each hematologic malignancy, which will have to be complemented by the identification of predictive biomarkers. If these challenges can be met, DAC inhibitors should develop into an important element of novel, targeted treatment strategies for leukemia.