Autophagy, a key mechanism of oncogenesis and resistance in leukemia

Besides the improvement in knowledge of the mutational markers involved in the emergence of leukemias, numerous small- and large-scale functional approaches have been developed to identify important cellular functions that become deregulated en route to, or as a result of, the development of leukemias. Macroautophagy, hereafter referred to as autophagy, is a degradative pathway that is currently under investigation for its role in leukemia progression or resistance to chemotherapy.

Autophagy is an adaptive survival mechanism that recycles amino acids through degradation of damaged organelles and macromolecules under conditions of cellular stress, such as nutrient deprivation or caloric restriction.¹  In mammal cells, autophagy is a multistep process starting with an initiation and a nucleation step in which there is a de novo isolation of a portion of the intracellular membrane termed a phagophore.^2,3  During a subsequent step of elongation, the phagophore invaginates, and its ends can fuse to generate a double-membraned vesicle referred to as the autophagosome, which is delimited by several lipid layers that sequester the cytoplasmic content and/or organelles. In a following maturation step, the autophagosomes fuse with endosomes to become amphisomes, which will themselves fuse with lysosomes to become autolysosomes capable of enzymatic degradation of the original intravesicular contents⁴  (Figure 1A).

From yeast to mammals, autophagosome biogenesis is controlled by several protein complexes composed of ATG proteins (autophagy-related genes). Of a total of 40 ATG proteins identified to date, ∼50%, play an essential role in the formation and elongation of the phagophore. ATG proteins bind transiently once recruited to the membranes of the phagophore and the autophagosome.^1,3 

The ATG proteins can be classified into 3 functional groups according to the autophagy step that they control: (i) the complex composed of the serine/threonine kinase ULK1 (or its yeast homolog Atg1), mATG13, FIP200, and ATG101, is involved in the initiation step; (ii) the Atg6/BECLIN-1-Atg14/ATG14L-VPS34-VPS15 and BECLIN-1-UVRAG-BIF-1-VPS34-VPS15 complexes are required for phagophore nucleation; and (iii) the 2 conjugation systems composed of ATG5-ATG12 and Atg8/MAP1LC3-II are essential for the elongation and closure of the autophagosome. ATG7 is absolutely required for the establishment of these 2 conjugation systems and therefore for the autophagosome biogenesis (Figure 1A). In contrast to initial impressions, autophagy not only is a recycling mechanism that helps cells to circumvent their nutrient stress but also regulates cell differentiation, cell death, and cell cycle.^5-7  This plethora of newly characterized functions underlies a key role of autophagy in leukemogenesis and chemoresistance in leukemias.

Several recent publications demonstrate that autophagy is required for the life-long maintenance of the murine hematopoietic stem cell (HSC) compartment and that deletion of ULK1-interacting partners, FIP200, ATG7, or ATG5, impairs survival and transplantation capability of normal HSCs.^8-11  In addition to these effects, Lin⁻Sca1⁺Kit⁺ cells from Vav-ATG7^−/− mice accumulate mitochondria, mitochondrial superoxide, and DNA damage and display increased proliferation, leading to a lethal preleukemic phenotype in mice.⁹  Bone marrow cells from these mice exhibit more elevated IC-NOTCH1 than their normal counterpart, and inhibition of Notch signaling in the context of autophagy deficiency restores normal HSCs differentiation capacity¹²  (Figure 1B). Basal autophagy activation can then modulate downstream signaling pathways (ie, Notch) that are involved in the control of proper HSC proliferation and differentiation. Impaired autophagy switches HSCs from a normal to a preleukemic state that is likely more favorable to overt leukemic transformation as is exemplified by the fact that heterozygous loss of ATG5 enhances disease progression and aggressiveness of an MLL-ENL-driven acute myeloid leukemia (AML) mouse model.¹¹ 

In line with these results, an analysis of whole exome sequencing results from 223 cases with myeloid neoplasms, including myelodysplastic syndromes, AML, and myeloproliferative neoplasms, established that 14% of these patients carry missense mutations or copy number alterations in at least 1 relevant gene involved in autophagy.¹³  For instance, expression of WIPI-1, a gene involved in phagophore nucleation, is significantly reduced in blasts from a cohort of 98 patients with AML in comparison with granulocytes harvested from 13 healthy donors.¹⁴  In silico analyses establish that multiple key autophagy genes are comprised within regions which are commonly heterozygously deleted in AML, especially in a subset of patients with complex karyotype.¹¹  Among these genes, expression level of ATG12, which is involved in autophagosome maturation, is frequently decreased in patients with de novo AML.^11,15  Human Atg8 homologs, including GAPARAPL1, GABARAPL2 (GATE-16), and MAP1LC3B, are also found among those depleted genes, and their expression is markedly decreased in AML blasts.¹¹  Another study performed on a different cohort of patients with AML demonstrated that GABARAPL1 and GABARAPL2 are significantly downregulated in patients exhibiting AML1-ETO and PML-RARA translocations compared with healthy donors.¹⁶ 

Although autophagy inhibition accelerates AML development, other groups reported that ATG7 or ATG4B knockdown alters the viability of primary chronic myeloid leukemia (CML) CD34⁺ progenitor cells,^17,18  likely by increasing mitochondrial oxidative phosphorylation and ROS accumulation.¹⁷  Consistent with these observations, a core of 5 autophagy genes, ATG4A, ATG4B, ATG4C, ATG5, and BECLIN-1, is found overexpressed in CML CD34⁺ cells compared with normal CD34⁺ cells, confirming that autophagy may have an opposite role in CML vs AML. Therefore, autophagy could play a versatile role in disease development that may depend on (i) the type of progenitor and oncogenic driver engaged in leukemic transformation and (ii) the state of leukemia expansion (tumor initiation vs progression).

Impairment of autophagy in preleukemic cells could also appear as an adaptive mechanism intended to stabilize oncogenic proteins that would otherwise be cleared by the autophagic machinery. Indeed, several recent publications demonstrate that fusion proteins such as PML-RARA or BCR-ABL can be degraded in autophagic vesicles; therefore, autophagy inhibition would help stabilize these oncogenic fusion proteins^19-22  (Figure 1B).

All-trans retinoic acid (ATRA), alone or combination with arsenic trioxide (ATO), has been shown to restore differentiation in acute promyelocytic leukemia cells exhibiting the PML-RARA translocation.²³  Several proteolytic mechanisms that contribute to the degradation of the RARA fusion protein are required for sustained remission and long-term cure in response to ATRA and ATO. These combined compounds have been shown to promote autophagy-mediated degradation of the PML-RARA oncoprotein in an mammalian target of rapamycin (mTOR)-dependent manner,²¹  and through redistribution of the cation-dependent mannose-6-phosphate receptor that leads to autophagosome maturation and vesicle acidification.²⁴  Recently, upregulation of high mobility group box 1HMGB1, a chromatin-associated nuclear protein and a critical regulator of autophagy, has also been reported in response to ATRA.^25,26  High mobility group box 1 translocates into the cytosol and regulates the interaction between ubiquitin-binding adaptor protein p62/SQSTM1 and PML-RARA, and in so doing, promotes PML-RARA degradation.

Imatinib, the front-line treatment of CML, inhibits the oncogenic BCR-ABL tyrosine kinase activity and also induces its autophagic proteolysis.¹⁹  BCR-ABL degradation can also be induced by ATO and requires activation of the lysosomal protease cathepsin B, which multiple groups have described as a direct mediator of BCR-ABL cleavage.^27-29  Interestingly, although autophagy stimulation is initially required for BCR-ABL degradation upon Imatinib treatment, it also promotes leukemic cell recovery following drug withdrawal,³⁰  suggesting that there is a threshold of autophagy induction that confines either its pro-death or its pro-survival function. Therefore, inhibiting autophagy after disease remission may have the potential to prevent emergence of a minimal residual disease able to fuel BCR-ABL–driven leukemia relapse. Hedgehog signaling is an essential pathway that was shown to stimulate the stemness potential of CML cells, thereby maintaining a reservoir of leukemic stem cells potentially involved in disease resistance and relapse.³¹  A recent study determined that combined inhibition of autophagy and Hedgehog pathways synergizes to enhance CML cell death resulting from Hedhehog pathway suppression in imatinib-resistant CML cells.³²  In addition to BCR-ABL, another tyrosine kinase, FLT3, which is found in a FLT3-ITD mutant form in ∼20% of AML cases, is also degraded by autophagy in response to treatment with proteasomal inhibitor bortezomib. Autophagy-mediated FLT3-ITD degradation overcomes resistance to an FLT3-specific inhibitor, quizartinib, which emerges from mutations in the kinase domain of FLT3.³³ 

Interestingly, the oncogenic fusion AML1-ETO is not degraded by autophagy despite the fact that autophagy is induced by AML1-ETO–targeting drugs, such as histone deacetylase inhibitors valproic acid and vorinostat.³⁴  In AML1-ETO–positive leukemias, autophagy has a prosurvival role, as its inhibition reduces viability of histone deacetylase inhibitor–treated AML cells.

As exemplified in AML1-ETO–positive leukemias, autophagy may act as a prosurvival mechanism in hematopoietic disease, and numerous studies demonstrate that autophagy activation may promote resistance to standard chemotherapies. Therefore, targeting autophagy may account for a promising therapeutic alternative for combination therapies (Table 1).

The chemotherapy regimen currently used to treat patients with AML is based on a combination of anthracyclines with cytarabine. Despite its antileukemic properties, cytarabine induces mTOR-dependent cytoprotective autophagy in AML cell lines and primary leukemic blasts but not in normal leukocytes.³⁵  Blockade of autophagy by either ATG7 depletion or treatment with bafilomycin A1 or the lysosomotropic agent, chloroquine, markedly increases cytarabine cytotoxic effects in AML cells^35-37  (Figure 1B; Table 1). Accordingly, hydroxychloroquine, a structural analog of chloroquine, promotes cell death in cytarabine-resistant cells.³⁸  Cytoprotective autophagy against the cytarabine/anthracycline combination is also stimulated by the bone marrow–derived mesenchymal stromal cells. Concomitant silencing of ATG7 in both AML and mesenchymal stromal cells results in a greater susceptibility of leukemic blasts to these genotoxic agents than the depletion of ATG7 only in AML cells.³⁹  S100A8, a member of the S100 calcium-binding protein family, has been shown to play a major role in autophagy-mediated resistance to these front-line therapies. S100A8 is upregulated in chemoresistant leukemic cells and physically interacts with the autophagy regulator BECLIN-1 to stimulate autophagy. Hence, S100A8 knockdown blocks autophagy and increases chemosensitivity of leukemic cells to ATO and anthracyclines.^40,41 

Statins are another class of antileukemic agents that are currently being tested in several clinical trials in combination with conventional chemotherapies used in AML. They act to lower cholesterol levels by inhibiting 3-hydroxy-3-methylglutaryl-coenzyme A reductase, the enzyme that catalyzes the rate-limiting step of cholesterol biosynthesis.⁴²  The decrease in the intracellular cholesterol level induces a cytostatic effect on leukemic cells⁴²  and stimulates cytoprotective autophagy. In this context, autophagy suppression potentiates the antileukemic effect of statins.⁴³  Another example of a therapy currently being tested in a clinical trial in patients with relapsed AML is the recombinant human arginase, an arginine-degrading enzyme. AML blasts are arginine auxotrophic, meaning that they rely on extracellular arginine for survival and proliferation.^44,45  Despite the fact that recombinant arginase blocks leukemic cell expansion by degrading extracellular arginine, it also increases autophagosome formation in AML.⁴⁶  Inhibiting autophagy using 3-methyladenine or chloroquine significantly enhances arginase-induced cell growth inhibition and apoptosis (Table 1).

Activating mutations in NOTCH1 are common in T-cell acute lymphoblastic leukemia (T-ALL), and glutaminolysis is a critical pathway for leukemia cell growth downstream of NOTCH1 and a key determinant of the response to anti-NOTCH1 therapies. Inhibition of NOTCH1 signaling impairs glutaminolysis and triggers autophagy as a salvage pathway supporting leukemia cell metabolism. Consequently, inhibition of glutaminolysis and autophagy synergistically enhances the antileukemic effects of anti-NOTCH1 therapy in T-ALL.⁴⁷  In T-ALL, glucocorticoids such as dexamethasone also exhibit marked antiproliferative effects; however, glucocorticoid resistance represents a major obstacle to complete successful treatment of this disease. Autophagy activation is a recently described mechanism of resistance to dexamethasone resulting from an inhibition of mTOR pathway.⁴⁸  Autophagy inhibition overcomes dexamethasone resistance in lymphoid-malignant cells (Table 1). Conversely, direct targeting of mTOR or AKT pathway with RAD001 or MK-0226, respectively, in B-cell progenitor ALL and T-ALL has been shown to trigger autophagy-associated cell death.^49-51 

Current treatments for B-cell disorders combine high-dose chemotherapy regimens and immunotherapy via the use of chemotherapy-coupled monoclonal antibodies targeting receptors expressed on malignant cells (ie, CD20). In line with this, a recent publication demonstrated that high doses of hydroxychloroquine and chemotherapeutic agent, chlorambucil, both loaded into biodegradable nanoparticles coated with an anti-CD20 antibody, strongly enhance malignant B-cell death compared with noncoupled cytotoxic agents.⁵² 

The proteosomal inhibitor bortezomib also exerts a significant cytotoxicity against B-cell acute lymphoblastic leukemia and multiple myeloma cells in which it provokes unwanted cytotoxic protein accumulation by direct inhibition of the β type-5 proteosomal subunit⁵³ ; however, subsequent endoplasmic reticulum stress induced by bortezomib activates the unfolded protein response (UPR) and autophagy as rescue mechanisms intended to compensate the loss of proteasomal activity and restore protein homeostasis. Several mechanisms have been established for autophagy induction in B-cell acute lymphoblastic leukemia and multiple myeloma treated with proteasome inhibitors. These mechanisms include an increased formation of the BECLIN-1-VPS34 complex, an enhanced de novo expression of the autophagic cargo receptor and adapter protein, SQSTM1/p62, or the upregulation of an essential regulator of the UPR-induced autophagy, GRP78/BiP.^54-56  Thus, impairment of autophagy activation potentiates the cytotoxic accumulation of intracellular protein aggregates induced by first- and second-generation proteasome inhibitors^54,55,57-59  (Table 1). Recently, a high-throughput compound screen identified metformin, a widely used antidiabetic agent, as a modulator of bortezomib-induced autophagy. Despite the fact that metformin was extensively described as an autophagy activator especially through the activation of the adenosine 5′-monophosphate–activated protein kinase, it can also impair bortezomib-induced GRP78 upregulation and subsequent autophagy and enhance the antimyeloma effect of bortezomib, confirming the therapeutic benefit of simultaneously modulating both proteasome- and autophagy-dependent pathways.^60,61 

Level of autophagy activation plays a critical role in cancer progression. Although reduced autophagic activity through mutations or downregulation of important autophagy regulators supports preleukemia development, and eventually cooperates with oncogenic fusions (such as BCR-ABL or PML-RARA) to promote full leukemia transformation, reactivation of this mechanism by front-line therapies can afterward participate in disease relapse. Thus, numerous recent studies suggest that inhibiting autophagy may be an efficient approach to improve the chemotherapeutic antileukemic regimens. In line with this, choloroquine and hydroxychloroquine are currently being tested in 39 clinical trials in solid and liquid tumors (Table 1). More particularly, in multiple myeloma, hydroxychloroquine is tested in combination with bortezomib (NCT00568880⁵⁹ ). In CML, hydroxychloroquine is combined with imatinib (NCT01227135), whereas in AML, this compound is combined with mitoxantrone and etoposide (NCT02631252). Considering that the antineoplastic effects of these agents may only partly rely on the inhibition of autophagy,^62,63  future clinical studies using more specific inhibitors should bring further demonstration of the therapeutic benefit of targeting autophagy in hematopoietic malignancies. These compounds include the specific inhibitor of the autophagy-initiating serine/threonine kinase ULK1, SBI-0206965, which exhibits in combination with mTOR inhibitors noticeable cytotoxic effects in lung cancer,⁶⁴  the USP10/13 deubiquitinase selective inhibitor, Spautin-1, which inhibits autophagy by promoting degradation of the BECLIN-1-VPS34 complex,⁶⁵  or the selective VPS34 inhibitors, PIK-III and SAR405, which block autophagosome maturation.^66,67  Of note, a novel orally available autophagy inhibitor, ROC-325, that is well tolerated in vivo and significantly more potent than hydroxychloroquine, seems already to hold great promise for AML treatment.⁶⁸ 

A.P. is supported by the ATIP/AVENIR, the St Louis Association research grants, the European Hematology Association nonclinical advanced research grant, and received the Prix Laurette Fugain from the Laurette Fugain Association. P.A. is supported by grants from the Association pour la Recherche sur le Cancer Foundation and the Institut National du Cancer (2012-045).

Contribution: P.A. and A.P. wrote the manuscript.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Alexandre Puissant, Hôpital St Louis, Jean Bernard Center, 16 rue de la Grange aux Belles, 75010 Paris, France; e-mail: alexandre.puissant@inserm.fr.