You eat what you are: autophagy inhibition as a therapeutic strategy in leukemia

Abstract

A deeper understanding of the role of autophagy, literally ‘self-eating’, in normal and cancer cell biology has emerged over the last few years. Autophagy serves as a vehicle for cells to respond to various stressors including genomic, hypoxic and nutrient stress, and to oppose mechanisms of ‘programmed’ cell death. Here, we review not only mechanisms of cell death and cell survival but also the early successes in applying autophagy inhibition strategies in solid tumors using the only currently available clinical inhibitor, oral hydroxychloroquine. In acute leukemia, currently available chemotherapy drugs promote cell death and demonstrate clinical benefit, but relapse and subsequent chemotherapy resistance is common. Increasing preclinical data suggest that autophagy is active in leukemia as a means of promoting cell survival in response to chemotherapy. We propose coupling autophagy inhibition strategies with current cytotoxic chemotherapy and discuss synergistic combinations of available anti-leukemic therapies with autophagy inhibition. Furthermore, novel autophagy inhibitors are in development and promise to provide new therapeutic opportunities for patients with leukemia.

INTRODUCTION

Acute leukemias are clonal malignancies of the hematopoietic system characterized by accumulation of immature cell populations in the bone marrow or peripheral blood. Acute myeloid leukemia (AML) is the most common type of acute leukemia in adults but has the lowest survival rate of all leukemias.¹ In spite of intensive treatment, more than 10% of AML patients fail to respond to initial therapy and over 60% relapse with resistant disease.² Achieving complete response depends on multiple factors including age, the presence of cytogenetic and molecular abnormalities, and performance status and organ function at diagnosis. However, these factors alone do not fully explain the observed patterns of resistance.³ Various mechanisms have been proposed to explain chemotherapy resistance in patients with AML, including multidrug resistance and inability of current therapies to eradicate leukemia stem cells (LSCs). Recent studies have also suggested that autophagy represents an important mechanism of resistance. An understanding of the multiple mechanisms contributing to treatment failure is necessary to develop more effective therapeutic approaches for patients with acute leukemia.

Several groups of proteins that promote the efflux of cytostatic drugs have been investigated in drug resistant acute leukemia.⁴ One group is the adenosine triphosphate (ATP)-binding cassette transporters (ABC transporters). P-gp, encoded by the multidrug resistance gene 1 (MDR1 or ABCB1), is involved in resistance to several drugs that are commonly used in AML, including anthracyclines. Expression of P-gp is an adverse prognostic factor for attaining a complete response and survival in adult leukemia, particularly in elderly patients.⁵ The breast cancer resistance protein, encoded by the ABCG2 gene, and the multidrug-resistant protein MRP3 also correlate with diminished survival in AML.⁴ However, studies of involving pharmacologic inhibition of P-gp have not demonstrated a substantial impact on chemotherapy outcomes.⁴

Recent studies have determined that LSCs have a central role in leukemia pathogenesis, and that failure to eradicate these cells is an important factor in patient outcomes.⁶ A number of genetic and cellular adaptations have been found to confer chemotherapy resistance in LSCs. Standard chemotherapy kills leukemia cells that are progressing through the cell cycle, but spares LCSs that are in a quiescent state in the bone marrow. This dormancy is also associated with an extremely low rate of oxidative phosphorylation that further renders LSCs insensitive to toxins affecting bioenergetics or inhibitors targeting key signaling pathways. Inducing LSCs to enter the cell cycle before chemotherapy treatment enhances their susceptibility to chemotherapeutic drugs, facilitating their eradication and sustained disease remission.⁶ Interestingly, autophagy is enhanced when both normal and malignant cells are in a resting G0/G1 state,^7,8 suggesting that LSCs may be resistant to chemotherapy in part through enhanced autophagy.

Autophagy is a lysosomal-degradation process that serves as a vehicle for clearing dysfunctional organelles and for maintaining genomic stability by clearing excised genomic fragments. As such, it regulates the responses of eukaryotic cells to cellular stress including those arising from hypoxia, genomic instability, endoplasmic reticulum stress, nutrient stress and facultative intracellular infections. In the setting of chemotherapy, radiation therapy and immunotherapy, autophagy serves as a major means for resistance from cytotoxic stress.⁹ The precise mechanism for how autophagy inhibits cell death is unclear but may involve limiting DNA-damage response-mediated apoptosis^10–12 or, alternatively, activation of the high-mobility group box 1 (HMGB1)/RAGE (receptor for advanced glycation end products) signaling axis.¹³ In addition, autophagy helps promote cell survival through nutrient recycling and the clearance of damaged organelles or oxidized and aggregated proteins.¹⁴ Recently, increasing evidence suggests that the role of autophagy changes rather dramatically during the period over which tumors develop. This complexity should be kept in mind when analyzing data from preclinical and clinical studies. Specifically, during the early phase of tumor initiation, autophagy appears to be suppressed, allowing the emergence of genomic instability.^15–20 Later in tumor development, autophagy is more active, leading to enhanced cancer cell survival, as described in this review. Therefore, efforts to inhibit autophagy are particularly warranted in established tumors. Notably, emerging data indicate that autophagy is a major contributor to chemotherapy resistance in AML, and its inhibition may represent an important therapeutic strategy. In this review, we will first describe the autophagic process and its regulation and then review evidence of the role of autophagy in chemotherapy resistance, summarizing the results of the first clinical trials of autophagy inhibition in advanced cancers. Finally, we will discuss the potential role of autophagy induction in AML and how this process may be manipulated to therapeutic advantage.

THE REGULATION AND MEASUREMENT OF AUTOPHAGIC FLUX

The term ‘autophagy’ was introduced by Christian René, viscount de Duve, the Nobel Prize-winning discoverer of lysosomes and peroxisomes in 1963. He first used this term to describe the morphologic features observed in glucagon-induced autophagy.²¹ We now recognize macroautophagy as the primary autophagic mechanism. Other mechanisms include microautophagy (direct delivery of cargo into the lysosome through membrane invaginations) and chaperone-mediated autophagy (translocation of soluble proteins into the lysosome through chaperone proteins).²² When individuals simply refer to autophagy, as in this review, they primarily are referring to macroautophagy. The process of autophagy begins with the formation of an isolation membrane, termed the phagophore (Figure 1).²³ Subsequent vesicle elongation allows engulfment and sequestration of organelles as well as bulk cytosolic material. This double-membrane structure is called the autophagosome.²¹ Loaded autophagosomes are transported along tubulin tracks to a perinuclear location where they fuse with lysosomes to form autolysosomes, enabling destruction of the engulfed material by lysosomal acid hydrolases. Finally, release and recycling of amino acids, nucleotides and fatty acids occurs to meet the anabolic and bioenergetic needs of the cell.²⁴ The complete process of autophagy from formation of isolation membranes to degradation in the autolysosomes is referred to as autophagic flux.

Figure 1.

Overview of the stages of autophagy. Autophagy begins with membrane formation to create the phagophore, which gradually elongates to engulf defective organelles and proteins, creating the mature autophagosome. Ultimately, the autophagosome fuses with a lysosome, where lysosomal hydrolases degrade and recycle the autophagosome contents.

First identified and studied in yeast, autophagy is mediated by multiple members of the autophagy gene (Atg) family. The ATG proteins form various complexes that function as effectors of the different phases of autophagy described above. In mammals, induction of autophagy involves the unc-51-like kinase 1 (ULK1) (or its homolog, ULK2) protein. The preautophagosomal structure has been identified in yeast. It is less clear what ‘seeds’ autophagy in mammalian cells, although recent studies have implicated the endoplasmic reticulum-Golgi intermediate compartment as the primary membrane source able to recruit ATG14, an early autophagosome marker.²⁵ Once initiated, additional proteins are recruited to ULK1/2, including the 200-kDa FAK-family interacting protein (FIP200), ATG13 and ATG101, to create the phagophore.²⁶ This complex interacts with two major regulatory complexes, the mammalian target of rapamycin 1 (mTORC1) and AMP-kinase. Interestingly, mTORC1 resides on the lysosomal membrane, sensing amino-acid flux and availability.²⁶ Under nutrient-rich conditions, mTORC1 directly interacts with ULK1/2, leading to phosphorylation of ULK1/2 and subsequent phosphorylation of ATG13, thereby inhibiting autophagy.²⁷ The class I phosphoinositol-3 kinase/AKT pathway is activated by insulin and growth factors, and interacts with mTORC1 to promote inhibition of autophagy.²⁸ However, in the setting of starvation, activating phosphorylation of ULK1/ULK2 occurs, leading to autophagosome formation.²⁹ AMP kinase, also active in the setting of starvation, promotes autophagy through negative regulation of mTORC1 and direct phosphorylation of activating sites on ULK1. Autophagosome formation is also dependent on the generation of phosphatidylinositol-3-phosphate, which is generated by class III phosphoinositol-3 kinases (PI3KC3/VPS34).³⁰ PI3KC3/VPS34, Beclin-1 and PIK3R4/p150 are three core components of the complex, PtdIns3K, which serves as a regulatory platform for phosphorylation/dephosphorylation of phosphatidylinositol-3-phosphate.³⁰ PtdIns3K can act to positively or negatively regulate autophagy, dependent on its interaction with other host factors. Interactions with host cofactors such as ATG14L and UVRAG serve to facilitate autophagy, whereas interaction with the Rubicon complex (Rubicon-UVRAG) results in negative regulation of autophagy.^31–33 HMGB1, a nuclear protein, can also induce autophagy when it translocates to the cytoplasm and binds with Beclin-1.³⁴ Of note, HMGB1 is also released extracellularly during necrotic cell death, and in addition to its other known roles as a damage-associated molecular pattern molecule, extracellular HMGB1 also triggers autophagy in cancer cells.³⁵

The most frequently employed ‘biomarker’ assays for autophagy involve immunoblotting for LC3-II protein or visualization of cytoplasmic ‘puncta’ representing LC3-II-containing autophagosomes. LC3 is a mammalian homolog of the yeast ATG8 protein that is cleaved by ATG4 during autophagy induction, converting proLC3 into LC3-I. The ATG5-ATG7 system then conjugates phosphotidyethanolamine to LC3-I, forming LC3-II. Lipidated LC3 (LC3-II) protein is easily detected as a faster migrating species on immunoblots. LC3-II is also critical for formation of the autophagosome. In cells expressing an exogenous GFP (green fluorescent protein)-LC3 fusion protein, fluorescence microscopy can be used to visualize and quantify autophagasome formation, observed as fluorescent ‘puncta’ in the cytoplasm.³⁶ Inhibition of autophagy at an early stage, for example, by suppression of PI3KC3/VPS34, results in decreased production of LC3-II. By contrast, inhibition of autophagic flux at a late stage using bafilomycin or chloroquine leads to increased levels of LC3-II and cytoplasmic puncta (autophagasomes) due to decreased degradation of the LC3-II protein by lysosomal proteases.³⁷ Autophagic vesicle (AV) accumulation, visualized through electron microscopy, has also been used as a surrogate for autophagic flux.³⁸ Intracellular levels of the scaffolding protein p62/SQSTM1 are tightly controlled by autophagy.³⁹ As such, levels of p62/SQSTM1 are inversely correlated with the degree of autophagy, and decreased p62/SQSTM1 may used to demonstrate autophagy induction.³⁹ Autophagy can be largely considered as a non-selective process in response to cellular oxidants, endoplasmic reticulum stress or starvation, but also includes highly selective organelle or pathogen targeting¹⁴ and clearance using molecules that have been better defined in yeast.⁴⁰

AUTOPHAGY MEDIATES CHEMOTHERAPY RESISTANCE IN PRECLINICAL MODELS

Autophagy promotes normal cellular functioning in the setting of stress by eliminating defective organelles, invasive microorganisms or protein aggregates, thereby preserving the integrity of the cell. Not surprisingly, autophagy inhibition early in the development of a cancer promotes tumor development.⁴¹ Indeed, deletion of genes important for autophagy induction, including ATG6/Beclin-1, UVRAG,³² ATG5,⁴² or Bax interacting factor 1⁴³ promote tumor development by accelerating inflammatory responses,⁴⁴ enhancing survival bioenergetics⁴⁵ and propagating oxidative stress.⁴⁶ Thus, defects in autophagy have been linked to early tumorigenesis in a number of mouse models.^47–49

Similar to the development of solid tumors in mice with impaired autophagy, defective autophagy in hematopoietic stem cells promotes the development of myelodysplastic syndrome and leukemia. Specifically, deletion of Atg7 in murine hematopoietic stem cells results in myeloid dysplasia and a phenotype resembling human AML.⁵⁰ Furthermore, many of the commonly deleted chromosomal regions in human myelodysplastic syndrome and AML, including 5q, 7q and 16p, encompass genes encoding critical autophagic proteins.⁵¹ These findings suggest that impaired autophagy may contribute to the development of myelodysplastic syndrome in humans.

Following tumor initiation, during the course of tumor progression, cells accumulate genetic damage that interferes with normal apoptotic cell death. This leads to an ‘autophagic switch’, whereby the cancer cells become more dependent on autophagy for survival (Figure 2). Mounting evidence indicates that established tumor cells utilize autophagy to resist radiation- or chemotherapy-induced cell death.^38,52,53 Pharmacologic or genetic inhibition of autophagy promotes the response of tumor cells to chemotherapy, radiation therapy and immunotherapy in a number of solid tumor and leukemia models.^52,54–57

Figure 2.

Representation of the regulation and timing of the autophagy pathway in the progression of cancer. Autophagy is often inhibited in early-stage cancer, allowing tumor development, but in late-state cancers, autophagy and accompanying chemotherapy resistance is high. Autophagy can be induced by mTORC1 inhibitors such as rapamycin or inhibited by chloroquine and its derivatives as well as bafilomycin. Damage-associated molecular pattern molecules (DAMPs) are released during the tissue damage that occurs during tumor progression and facilitates immune responses. HMGB1 is a prototypical DAMP that also regulates autophagy.

Autophagy also regulates the immune response to cancer.⁴⁴ Autophagy⁵⁸ is necessary for the anti-tumor immune response evoked by apoptotic tumor cells in response to chemotherapy by facilitating the release of ATP, an important immunogenic signal, promoting Interleukin 1 release. Autophagy-competent colorectal carcinoma cells treated with chemotherapy release more ATP than cells that are autophagy incompetent. Furthermore, autophagy-competent tumor cells injected into mice recruit dendritic and T cells to sites of tumor cell death and elicit an anti-tumor immune response, which does not occur in mice injected with tumor cells that are autophagy deficient. However, the immune response to autophagy-deficient tumor cells can be restored by artificially restoring extracellular ATP concentrations.⁵⁸ Furthermore, in a Kras-driven lung cancer model, mice with autophagic defects upregulate genes responsible for inflammation and increased levels of regulatory T cells (Tregs). Depletion of the Tregs in these autophagy-deficient mice slows the growth of the Kras-driven tumors to rates similar to those observed in autophagy-competent mice.⁴⁸ Together, these observations demonstrate that functional autophagy, particularly in the early stage of tumor development, is important in modulating the immunogenicity of cell death. Inhibiting autophagy through administration of chloroquine promotes NK and T cell mediated cytotoxicity during IL-2 therapy.^59,60 The role of autophagy in regulating anti-cancer immune responses merits further investigation, but in general autophagy can be noted to be important early in an immune response to promote antigen cross-presentation by dendritic cells, but once immune effectors are generated, that they are facilitated in their action by limiting autophagy.

LESSONS FROM CLINICAL STUDIES OF AUTOPHAGY INHIBITION IN SOLID TUMORS

In light of the sensitizing effect of autophagy inhibition on chemotherapy-induced cell death, a number of clinical trials have been launched combining the first-generation autophagy inhibitor hydroxychloroquine (HCQ) with individual chemotherapeutic agents in patients with solid tumors. Chloroquine (CQ) derivatives, such as HCQ, impair lysosomal function and autophagic flux at a late step, resulting in accumulation of AVs.⁶¹ HCQ is an inexpensive oral drug that has been prescribed for patients with malaria, rheumatoid arthritis (RA), systemic lupus erythematosus and HIV for many years. It is considered as a safe and effective medication for patients with RA. The typical dose of HCQ for RA is 400 mg/day. HCQ has a long half-life and requires weeks to achieve peak concentration when given at the typical 400 mg/day dose. However, when much higher loading doses are given over a short period of time, high blood concentrations of HCQ can be quickly attained, which correlate with a more rapid improvement in RA outcomes.⁶² Administration of higher doses of HCQ, up to 1200 mg/day, do not substantially increase toxicities in patients with RA. The main side effects related to higher dosing of HCQ are gastrointestinal, including nausea, vomiting, diarrhea, and abdominal pain, and ophthalmic, including temporary visual changes and increased retinal pigment.⁶² The latter typically only occurs after years of use and has not been a problem in cancer patients treated to date.

A randomized, placebo-controlled trial of CQ in patients with glioblastoma multiforme established the safety of adding CQ derivatives to cytotoxic chemotherapy and radiation.⁶³ In vitro studies have determined that CQ and HCQ exhibit similar potencies for inhibiting autophagy, but the favorable safety profile of HCQ has led to its preferred use in the majority of early-phase clinical trials that combine autophagy inhibition with conventional chemotherapy. These studies, reported in Table 1, have established that HCQ is generally well tolerated with most combinations of chemotherapy. Five of the seven published studies did not reach a maximally tolerated dose, and collectively established 1200 mg/day of HCQ as the recommended Phase 2 dose.^64–68 However, one clinical trial that combined HCQ with daily temozolimide and concurrent radiation for glioblastoma multiforme surprisingly noted grade 4 neutropenia and thrombocytopenia in all patients at a dose of 800 mg/day.⁶⁹ Another trial described Grade 3 fatigue as the dose-limiting toxicity at 800 mg/day dose when HCQ was combined with vorinostat.⁷⁰ Common adverse effects reported on these studies included nausea, diarrhea, constipation, anorexia, rash, fatigue, anemia and weight loss.

Table 1.

Summary of clinical trials of autophagy inhibition in solid tumors

Reference	Phase	Tumor	Therapy	MTD	DLTs	Response
Rosenfeld et al.69	3+3 Phase I/Phase II	Newly dx GBM	HCQ+RT+ TMZ (Continuous)	600 mg/day HCQ	Gr 4 neutropenia in 2/3 pts and Gr 4 thrombocytopenia in 4/3 pts at 800 mg/day	No significant improvement in expected overall survival, no significant autophagy inhibition at MTD
Rangwala et al.67	3+3 Phase I	Melanoma and advanced solid tumors	HCQ + TMZ (intermittent)	No MTD. RP2D 1200 mg/day HCQ	1/6 pts with Gr 3 rash at 1000 mg/day and 1/6 pts with Gr 3 n/v at 1200 mg/day	PR in 3/29 pts, and SD in 8/29 patients. Trend toward higher AUC in those with a PR/SD than those with PD (P = 0.0635)
Rangwala et al.66	3+3 Phase I/Phase II expansion cohort in melanoma	Melanoma and advanced solid tumors	HCQ + TEM	No MTD. RP2D 1200 mg/day HCQ	1/6 pts with Gr 4 thrombocytopenia at 200 mg/day	No PR. SD in 14/21 in dose-escalation cohort, and 14/19 melanoma patients in the both cohorts
Vogl et al.68	3+3 Phase I	Relapsed/Refractory multiple myeloma	HCQ + Bortezomib	No MTD. RP2D 1200 mg/day HCQ	1/6 pts with Gr 3 thrombocytopenia and Gr 4 neutropenia at 1200 mg/day	3/22 very good partial response, 3/22 with minor response, and 10/22 with stable disease
Mahalingam et al.70	3+3 Phase I	Advanced solid tumors	HCQ + VOR	600 mg/day HCQ	3/6 pts with Gr 3 fatigue and 1/6 pts with gr 2 seizure at 800 mg/day	PR 1/24, SD in 2/24
Goldberg et al.64	3+3 Phase I	Advanced NSCLC	HCG + Erlotinib	No MTD. RP2D 1000 mg/day	No DLTs	PR 1/19, SD in 2/19
Lotze et al.65	Phase I/II Storer design	Neoadjuvant pancreatic cancer	HCG + gemcitabine	No MTD	No DLTs	R0 resection rate 81%, 14/31 with >50% dec CA19-9

Abbreviations: GBM, glioblastoma multiforme; Gr, Grade; HCQ, hydroxychloroquine; MTD, maximum tolerated dose; NSCLC, non-small cell lung cancer; PD, progressive disease; PR, partial response; pts, patients; RT, radiation therapy; SD, stable disease; TEM, temsirolimus; TMZ, temozolomide; VOR, vorinostat.

Early signs of efficacy have been observed in the majority of the published clinical trials incorporating HCQ, with responses more commonly seen in patients receiving high doses of HCQ (800–1200 mg/day). Effective autophagy inhibition was also demonstrated in individual studied patients.^66–69 The dose of HCQ appeared to correlate with the degree of autophagy inhibition,⁶⁷ as measured by AV formation, as well as LC3-II and p62/SQSTM1 levels. Although autophagy inhibition could be demonstrated in peripheral blood mononuclear cells, particularly at higher doses, tumor cell autophagy was more significantly inhibited.^68,70

THE ROLE OF AUTOPHAGY IN LEUKEMIA CELL DEATH

Although much of our understanding regarding the role of autophagy in chemoresistance has been pioneered in solid tumors, emerging evidence indicates that autophagy has an important role in leukemia. Preclinical investigations have documented autophagy induction in leukemia cells following treatment with conventional chemotherapy. For example, treatment of K562, a myeloid leukemia cell line, with daunorubicin enables conversion of LC3-I into LC3-II, a punctate distribution of endogenous LC3-II, and accumulation of AVs assessed by electron microscopy.⁷¹ Similar autophagy induction, as revealed by rapid conversion of LC3-I into LC3-II, a decrease in p62/SQSTM1, and an increase in AVs, has been demonstrated after cytarabine treatment of the human leukemia cell lines, REH and HL-60, as well as peripheral blood mononuclear cells from patients with chronic myelogenous leukemia (CML).⁷²

The mTOR pathway has a critical role in the survival and growth of cancer cells, and aberrant activation of this pathway appears important in the pathophysiology of AML.^73–75 In response to growth factor signaling, the mTOR signaling complex I (mTORC1) directly blocks the induction of autophagy by promoting inhibitory phosphorylation of ULK1.²⁷ Rapamycin, an mTORC1 inhibitor, has yielded mainly cytostatic effects in animal models and limited responses in clinical trials.⁷⁶ This is likely due to potent induction of autophagy by this compound. A distinct mTOR complex, mTORC2, is insensitive to rapamycin and is responsible for phosphorylation of an activating site on AKT.⁷³ The antileukemic properties of dual mTORC1/mTORC2 inhibitors are currently being explored.^77,78 When dual mTORC1/C2 inhibitors are used to treat AML cell lines, increased expression of LC3-II, appearance of punctate GFP-LC3-positive autophagosomes and decreased expression of p62/SQSTM1 all occur, consistent with the induction of autophagy.^77,78 Furthermore, combined treatment of primary leukemic blasts with an autophagy inhibitor and a dual mTORC1/2 inhibitor results in markedly enhanced suppression of colony formation.⁷⁷ This compelling preclinical data using the combination of mTORC and autophagy inhibitors has also been confirmed in BCR/ABL-expressing leukemic cells,⁷⁹ and holds promise as a novel approach for anti-leukemia therapy. Clinical trials evaluating this combination strategy are clearly warranted.

Further evidence for induction of autophagic flux in leukemic cells has been revealed through the study of HMGB1. In AML cell lines treated with the anti-leukemia agents cytarabine and etoposide, the addition of HMGB1 neutralizing antibodies results in a significant increase in leukemia cell death.⁸⁰ Likewise, treatment with HMGB1 protein enhances drug resistance, an effect that is dependent on increased autophagic flux, as the resistance can be reversed through pharmacologic inhibition of autophagy. Treatment of leukemia cell lines with HMGB1 increases LC3-II abundance with a corresponding decrease in p62/SQSTM1.⁸¹ In sum, autophagy is induced in leukemia cells, at least in part by the release of HMGB1 from dying and stressed cells, thus enabling subsequent chemotherapy resistance.

Acetylation of autophagic proteins also appears to be an important means of autophagy regulation in leukemia.^82–84 The role of acetylation was demonstrated in leukemia cells while exploring the apoptotic effects of histone deacetylase inhibitors (HDACi). HDACi induce caspase-dependent apoptosis as well as autophagy.^85–89 In DS-AMKL (Down syndrome-acute megakaryoblastic leukemia) cell lines, treatment with an HDACi leads to apoptosis and cell-cycle arrest, and also results in acetylation of ATG7 as well as multiple other proteins involved in autophagy.⁹⁰ Interestingly, in primary AML cells a lower degree of basal autophagy is predictive of response to the HDACi valproic acid.⁹⁰ Moreover, valproic acid-resistant cell lines exposed to CQ are sensitized to the pro-apoptotic effects of HDACi.

The synergistic activity of HDACi and autophagy inhibitors has also been demonstrated in t(8;21) AML.⁹¹ In both t(8;21) cell lines and primary patient specimens, combined treatment with valproic acid and CQ reduced cell viability to levels significantly lower than either treatment alone. However, Torgerson et al.⁹¹ and others have demonstrated that autophagy induction is not responsible for the degradation of AML-ETO that is commonly observed following HDACi treatment. Instead, HDACi-induced autophagy may act to promote cell survival by limiting drug-induced accumulation of ubiquitinated proteins, as indicated by the enhanced caspase activity and apoptosis that is seen when autophagy inhibition is combined with HDACi treatment.

It has been proposed that the effects of HDACi on autophagy in leukemia cells may be dependent on the duration of exposure, as 12 h of HDACi exposure to DS-AMKL cells led to autophagy induction but further exposure, up to 24 h, consistently inhibited autophagy.⁹⁰ The inhibition of autophagy by prolonged exposure to HDACi is presumably due to the acetylation of critical autophagic proteins.⁹⁰ A similar time-dependent accumulation of LC3-II, consistent with induction of autophagy, was seen when t(8;21)-positive AML cells were exposed to HDACi for up to 8 h, but the LC3-II levels appear to decrease following 16 and 24 h of exposure.⁹¹ Regardless of the effect of HDACi treatment on autophagy induction or inhibition, it is clear that HDACi-induced leukemia cell death is enhanced when autophagy is also inhibited.

Pharmacologic inhibition of autophagy also enhances the efficacy of the HDACi suberoylanilide hydroxamic acid against imatinib-resistant CML.⁸⁷ Suppression of p53 expression demonstrated that these effects occur independently of p53, but are associated with enhanced apoptosis, as demonstrated by a marked increase in caspase-3-positive cells.

A heightened level of autophagy is well documented in CML following treatment with imatinib, possibly the result of increased endoplasmic reticulum stress that was otherwise inhibited by active BCR/ABL.⁹² In addition, reduced expression of microRNA-30a, an inhibitor of autophagy, appears to be involved in the induction of autophagy during imatinib treatment.⁹³ In both CML cell lines and primary cells, treatment with a microRNA-30a mimic leads to decreased autophagy, and enhanced toxicity of imatinib.⁹³ Furthermore, autophagy inhibition also enhances the impact of interferon-α on CML cells.⁹⁴ Interestingly, the level of baseline autophagy activity before treatment may be a biomarker for response to imatinib in patients with newly diagnosed CML.⁹⁵ The difference in pre-treatment Atg4B transcript levels in CD34⁺ cells from imatinib-responding versus non-responding patients is significantly different (P = 0.014), with higher levels detected in the imatinib non-responders.⁹⁵ Suppression of ATG4B protein expression in the imatinib non-responders in vitro led to decreased cell viability and impaired cell proliferation. Although these findings require validation in a larger cohort of samples, they suggest that ATG4B may represent a biomarker predicting poor response to imatinib alone, while also identifying patients who may derive particular benefit from autophagy inhibition concurrent with imatinib therapy. Taken together, these findings underscore the significance of autophagy induction in primitive CML cells as a means of cell survival during periods of stress, and the therapeutic potential of autophagy inhibition as a means of enhancing cell death in the typically chemotherapy-resistant LSCs.^92,95,96

While enhanced autophagy allows advanced cancers to survive stressful conditions, including chemotherapy treatment, the role of autophagy during arsenic trioxide (As₂O₃) therapy may be different. Induction of autophagy independent of the mTOR/AKT pathway appears to occur in As₂O₃-treated AML cell lines.^97,98 Autophagy induction in this setting is necessary for the anti-leukemic properties of As₂O₃, although the mechanism whereby autophagy mediates the anti-leukemic effects of As₂O₃ remains unclear.⁹⁷ Interestingly, additional studies have demonstrated that autophagy is responsible for As₂O₃-induced degradation of BCR/ABL protein. Similarly, autophagy induction by all-trans retinoic acid appears to regulate degradation of the PML-RAR oncoprotein, possibly through interaction with p62/SQSTM1.^99,100 In contrast to the pro-survival function of autophagy in advanced cancers, these findings suggest that autophagy induction is important for promoting the effectiveness of As₂O₃ and all-trans retinoic acid, underscoring the complexity of autophagy in anti-leukemia therapies.

Another caveat regarding the application of autophagy inhibition in the clinical setting concerns tumors with mutant p53. In a Kras model of pancreatic cancer, mice with impaired autophagy due to Atg5 or Atg7 deletion, but harboring wild-type p53, developed precancerous lesions that never progressed to pancreatic ductal adenocarcinoma.⁴⁹ By contrast, Atg7^–/– mice with mutant p53 not only developed precancerous lesions, but also rapidly progressed to pancreatic ductal adenocarcinomas that resulted in early death.⁴⁹ These findings have led some to issue warnings over the use of autophagy inhibition in the setting of p53 mutant tumors.¹⁰¹ However, subsequent studies using genetically engineered mouse models of pancreatic cancer with loss of heterozygosity of p53, more accurately representing human tumor biology, showed that while genetic autophagy inhibition led to increased tumor development, these lesions did not progress to invasive cancer or cause early death.¹⁰² Our interpretation of these findings is that early, artificial, p53 ablation enhances the susceptibility to autophagy inhibition and accelerates tumor growth, but in established tumors with heterozygous p53 mutation, autophagy inhibition does not enhance tumor progression. It is notable that p53 pathway alterations occur in less than 20% of de novo AML cases.¹⁰³

NEWLY DESCRIBED MECHANISMS OF CELL DEATH

Numerous investigations have demonstrated crosstalk between apoptosis and autophagy signaling pathways. At the same time, our understanding of cell death mechanisms continues to evolve, with the discovery of novel pathways that operate independently of apoptosis and necrosis. Newly recognized pathways of programmed cell death include necroptosis, ferroptosis, pyroptosis and NETosis. Necroptosis is a form of regulated necrosis dependent on receptor-interacting protein kinases 1 and 3. It can be triggered by tumor necrosis factor-α and results in cell death through a caspase-independent pathway.¹⁰⁴ Necroptosis has been implicated in a variety of diseases,¹⁰⁴ including leukemia,^105,106 and targeting this pathway may offer new therapeutic advantages. Ferroptosis is a recently described nonapoptotic form of cell death, first reported in Ras-mutated cell lines, in which the small molecule erastin triggers an iron-dependent form of cell death.^107,108 Although the exact role of iron in the process of ferroptosis has not been fully elucidated, there is strong evidence that cell death by ferroptosis is primarily mediated by reactive oxygen species.¹⁰⁸ Pyroptosis is dependent on caspase-1 and is triggered by pathogens. Cell death arising by pyroptosis was originally described more than 20 years ago, although classical apoptosis was thought to be the underlying mechanism.¹⁰⁹ However, caspase-1 does not have a role in apoptotic cell death,^110,111 while the apoptotic proteases caspase-3, -6 and -8 are not involved in pyroptosis.¹¹² Neutrophils have recently been found to activate a newly defined form of pathogen-induced cell death called NETosis. Neutrophils, upon stimulation with bacteria, interleukin-8, phorbol myristate acetate or lipopolysaccharide, can form net-like, extracellular strands composed of deoxyribonucleic acid, histones, HMGB1 and granular proteins, termed neutrophil extracellular traps.¹¹³ These neutrophil extracellular traps are instrumental in breaking down virulence factors and eliminating bacteria.¹¹³ NETosis was recently recognized as a mode of cell death that is distinct from apoptosis and regulated necrosis based upon its insensitivity to caspase inhibition and necrostatin-1 (which inhibits necroptosis).¹¹⁴ While the relevance of these distinct pathways to leukemia pathophysiology has yet to be determined, the awareness of alternative pathways of cell death may offer both critical insights into chemotherapy resistance and new avenues for development of anti-cancer therapeutics.

TARGETING AUTOPHAGY AS AN OPPORTUNITY FOR THERAPEUTIC ADVANCES IN LEUKEMIA

The preclinical findings discussed herein clearly demonstrate that HCQ can effectively inhibit autophagy in leukemia cells in vitro and frequently enhances the response to treatment with individual chemotherapy regimens. Based on findings made in the treatment of patients with solid tumors, HCQ can be safely paired with a variety of chemotherapeutic agents. Autophagy inhibition with HCQ during AML therapy may limit chemotherapy resistance and enhance response, and the development of clinical trials incorporating HCQ with multi-agent drug therapy seems to be a next rational step in the field. The dose of HCQ necessary to achieve profound autophagy inhibition may limit the practicality of this approach. More potent autophagy inhibitors are critically needed, and candidate compounds such as Lys05 are currently being evaluated.¹¹⁵

Relapse of disease in AML is a common phenomenon, and chemotherapy resistance in this setting frequently occurs with complete response rates ranging from 20 to 60% using modern treatment regimens.^116–118 A number of promising chemotherapeutics have been combined with autophagy inhibition in leukemia cells with compelling results in vitro, particularly HDACi and dual mTORC1/C2 inhibitors.^77,78,90,91 Both of these classes of inhibitors have been combined with HCQ in early-phase studies of advanced solid studies. For example, when HCQ was combined with temsirolimus, an mTORC1 inhibitor, in patients with advanced solid tumors and melanoma, 67% (23/34) of patients experienced prolonged stable disease, including those with tumor types not typically considered responsive to rapamycin analogs.⁶⁶ The combination of HCQ with the HDACi vorinostat in advanced solid tumors resulted in dose limiting toxicities (grade III fatigue) at 800 mg/day of HCQ and but failed to consistently achieve autophagy inhibition in peripheral blood mononuclear cell at lower doses, although tumor specimens did reveal evidence of autophagy inhibition.⁷⁰ However, with the short-term duration of therapy used in AML treatment, similar toxicities of fatigue may be less problematic. Both HDACi and mTOR inhibitors, particularly dual mTORC1/2 inhibitors, are promising for successful pairing with HCQ or other autophagy inhibitors in AML clinical trials, possibly as part of a multi-agent chemotherapy regimen.

Autophagy inhibition holds potential for patients with CML as well, particularly due to the possibility of achieving cell death in LSCs that are typically resistant to therapy. Indeed, a multicenter phase 2 clinical trial randomizing patients to imatinib or imatinib plus hydroxychloroquine in patients with CML is currently underway.¹¹⁹ Although newer generation tyrosine kinase inhibitors often effectively treat patients with imatinib resistance, the T315I mutation remains problematic and combination of tyrosine kinase inhibitors with autophagy inhibitors may be beneficial.^92,93 Degradation of BCR/ABL through autophagy induction with As₂O₃ suggests that autophagy inhibition in CML may be more complex and suggests caution in its early application. Finally, the role of biomarkers to predict benefit from autophagy inhibition is particularly appealing, and further studies validating the expression of ATG4B, or other proteins involved in autophagy such as HMGB1, sRAGE and LC3, as predictive for imatinib response are critical.⁹⁵ Similar studies in AML may be even more helpful given the multifactorial and often treatment-resistant nature of the disease.