Chronic myeloid leukemia (CML) is a multi-lineage myeloproliferative neoplasm that originates from HSCs and is characterized by uncontrolled proliferation of hematopoietic cells, particularly an excessive number of granulocytes in the peripheral blood. More than 95% of patients harbor a characteristic reciprocal chromosomal translocation product, called BCR-ABL1, that originates in a HSC and leads to overexpression of its protein (p210^BCR-ABL), with constitutively activated tyrosine kinase activity, giving rise to and sustaining the CML clone [55,56,57]. The identification of BCR-ABL1 as the major driver in the pathogenesis of CML led to the development of targeted therapies with ABL1 tyrosine kinase inhibitors (TKIs) such as Imatinib, Dasatinib, and Ponatinib [58,59]. Treatments with TKIs in the initial chronic phase of CML have been shown to induce remarkable haematological and cytogenetic responses, however, treatments of advanced stage patients or individuals with ABL1 kinase domain mutations are much more challenging and most importantly, none of these monotherapies are curative [60,61,62]. Strong evidence indicates that LSCs, including quiescent LSCs, are insensitive to TKIs and constitute a critical source of leukemia recurrence and a significant reservoir for the emergence of TKI-resistant sub-clones, necessitating lifelong treatment for most patients [57,63,64,65,66,67,68,69,70]. Interestingly, it has been reported that CML stem cells are not strictly dependent on the tyrosine kinase activity of BCR-ABL1 for survival and that they may be effectively protected by the bone marrow microenvironment and/or exploit signaling pathways such as autophagy [51,71,72,73].

The connection between autophagy and CML was first documented by an increased autophagic response following treatment with imatinib mesylate (IM) or other non-targeting drugs in CML cell lines and primary patient samples [51,74]. This response was attributed to ROS generation, ER stress, and a dose-dependent increase in gene and protein expression of key autophagy players [51,75,76]. The increase in autophagy induces CML cells to become senescent rather than apoptotic, causing them to be more elusive to treatment [77]. Thus, the regulation of autophagy in CML has been largely explored in the context of BCR-ABL1, but may be influenced by different signaling pathways. For example, the mammalian target of rapamycin (mTOR) in the TORC1 signaling pathway is a well-known negative regulator of autophagy, and BCR-ABL1 has been shown to suppress autophagy via the PI3K/AKT/FOXO4/ATF-5 pathway by stimulating transcription of mTOR [78,79,80]. On the other hand, it has been shown that autophagy is induced by BCR-ABL1 via the rapamycin-insensitive mTORC2 signaling complex and helps CML cells to recover from TKI treatment [74,81]. The kinase mTOR can be part of two distinct complexes, TORC1 and TORC2 [80]. While the TORC1 complex and its negative regulation on autophagy has been extensively studied, much less is known about the TORC2 complex that stimulates and mediates autophagy during amino acid starvation [80]. TORC2 promotes autophagy via its downstream target Ypk1, which in turn inhibits calcineurin, allowing activation of the eIF2α kinase Gcn2 and phosphorylation of 4E-BP1 T37/46 and S65 to activate autophagy [80].

Key enzymes in the LC3 conjugation cascade, such as ATG7 and ATG4B, have both been implicated as critical for CD34⁺ CML cell survival, while genetic inhibition of ATG4B or ATG7 was able to increase apoptosis of CML stem and progenitor cells and sensitize them to TKIs [6,28]. Interestingly, knockdown of ATG7 also led to metabolic reprogramming that resulted in increased oxidative phosphorylation and mitochondrial ROS accumulation and caused primary CML cells to differentiate into erythroid cells [28]. Mechanistically, a recent study demonstrated that ULK1 could inhibit ATG4B activity and LC3 processing by phosphorylating ATG4B on serine 316, resulting in inhibition of its catalytic activity both in vitro and in vivo [82]. Notably, this activity seems to be regulated by PP2A-mediated dephosphorylation. MST4, a member of the mammalian sterile20-like (STE) serine/threonine kinase (STK) family, was found to stimulate autophagy by activating ATG4B through phosphorylation of ATG4B S383, which stimulates ATG4B to act on LC3, thereby increasing autophagic flux [83]. Of even more interest, inhibiting ATG4B can enhance the anti-tumor effects of radiotherapy in a solid tumor model. Our group investigated ROS-mediated autophagy in the context of protein–protein interactions between BCR-ABL1, the scaffold protein AHI-1, and the cytoskeletal modulator DNM2 [75]. Here, the identified AHI-1-BCR-ABL-DNM2 complex was shown to regulate endocytosis and ROS generation, both of which lead to autophagy induction in CML, but knockdown of DNM2 disrupted these activities and reduced autophagy induction in primitive leukemic cells [75]. Ianniciello et al. also demonstrated that proficient autophagy is required for CD34⁺ CML cells to proliferate when they transition from hypoxic to normal oxygen conditions, which mimics the leukemic commitment as the cells migrate away from the hypoxic bone marrow niche [84]. Together, these findings reinforce the concept that autophagy is a critical pro-survival pathway in CML and contributes to drug resistance and disease progression. The recognition of the innate and acquired mechanisms that make CML stem cells resistant to TKIs has thus prompted considerable interest in developing new treatment strategies, including targeting autophagy.

Unlike CML, AML is a very heterogeneous disease and represents much more context-dependent complexity regarding the role of autophagy in disease initiation and progression. Acute leukemias are characterized by the rapid accumulation of clonal, immature myeloid blasts, with incomplete differentiation, in the patients’ bone marrow and peripheral blood, and are often accompanied by multi-lineage cytopenias [85,86]. Although major progress has been made in identifying different molecular and genetic subgroups, AML therapies and long-term patient outcomes have not improved significantly over the past 40 years [87,88]. One of the foremost hurdles to a cure is thought to be the presence of LSCs or leukemia-initiating cells that, in many patients, constitute a small subpopulation of self-renewing cells at the top of a hierarchical organization with resistance to chemotherapy, causing treatment failure and relapse [89,90,91,92]. Immunodeficient mouse models show that these leukemic cells can produce AML upon xenotransplantation, including in secondary and tertiary recipients, and generate non-LSC committed progenitors unable to be serially transplanted, all features of functionally true LSCs [90,91,93,94,95]. Initial immunophenotypic analyses demonstrated that the majority of AML patient cells express the surface antigen CD34⁺, with LSCs mostly residing in the CD34⁺CD38⁻ subfraction [91,96,97]. However, more recent studies revealed that many patients harbor, at minimum, two distinct LSC populations that coexist in the CD34⁺CD38⁻ (Lymphoid-primed multipotent progenitor (LMPP)- or multipotent progenitor (MPP)-like LSCs) and in the CD34⁺CD38⁺ (granulocyte-macrophage progenitor (GMP)-like LSCs) subfractions or less often in the CD34⁻ subfraction, with the former, more immature CD34⁺CD38⁻ subpopulation containing a higher LSC frequency in comparative xenotransplantation experiments [98,99,100,101]. Moreover, some AML patients lack expression of CD34, the CD34⁻ AML cells, and here, LSCs can mostly be detected in the CD34⁻ compartment, with very few LSCs being present in the much smaller CD34⁺ fraction [99,100,102,103], highlighting the heterogeneity of AML and the complexity of studying and understanding the disease.

AML patients usually harbor genetic abnormalities that often originate from chromosomal translocations and rearrangements, such as t(8;21), t(15;17), inv(16), t(6;9), t(9;11), or t(11;19), and characterize, in some cases, a particular prognostic leukemic subtype [104,105,106,107]. Recent molecular studies demonstrated that additional mutations in receptor kinases, key signaling mediators, proto-oncogenes, or epigenetic enzymes, for example FLT3-ITD, TP53, c-KIT, or IDH1/2 mutations, often determine the course and severity of the disease [108,109,110,111,112,113]. Interestingly, sequencing and in silico studies have shown that a high frequency of AML patients carry often heterozygous deletions, missense mutations, or copy number changes of key autophagy genes, particularly AML patients with complex karyotypes [42,114,115]. For example, heterozygous chromosomal loss of 5q, 16q, or 17p correlate with the encoded regions for the autophagy genes ATG10 and ATG12, GABARAPL2 and MAP1LC3B, or GABARAP, respectively [42]. In line with these observations, Watson et al. also demonstrated that human AML blasts exhibit low expression of several autophagy genes, including ATG10, ATG5, ATG7, BECN1, GABARAP, GABARAPL1/2, and MAP1LC3B, decreased autophagic flux, and high ROS levels [42]. Furthermore, this study showed that heterozygous loss of Atg5 or Atg7 in a MLL-ENL AML mouse model led to more aggressive leukemia progression, suggesting a tumor-suppressive role for autophagy [42]. Similarly, Jin et al. confirmed that Ficoll-enriched leukemic blasts from AML patients express significantly lower transcript levels of ULK1, FIP200, ATG14, ATG5, ATG7, ATG3, ATG4B, and ATG4D compared to granulocytes from healthy donors [52]. In addition, Rudat et al. determined, in a large RNAi screen for “rearranged during transfection” receptor tyrosine kinase (RET) effectors, that mTORC1-mediated suppression of autophagy can stabilize mutant FLT3 in AML, while an increase in autophagy was achieved through RET inhibition and led to FLT3 depletion [53]. In contrast, Heydt et al. showed that FLT3-ITD increases autophagy in AML cell lines and patient cells via ATF4 and that inhibition of autophagy or ATF4 abolishes FLT3 inhibitor resistance [116]. Moreover, a recent investigation by Folkerts et al. revealed that various leukemic cell lines, and purified CD34⁺ cells from AML patients, exhibit inconsistent levels of basal autophagic flux, with particularly high levels in immature ROS^low LSC blasts and adverse AML risk groups, such as those with TP53 mutations [117]. In this study, knockdown of ATG5 in primary AML cells resulted in impaired engraftment of human cells in immunodeficient NSG mice, an observation that is in contrast to previous work and would rather suggest a tumor-promoting role for autophagy in this context [117]. Interestingly, other reports showed that ATG5 or ATG7 are required for the efficient initiation of AML in the context of MLL-AF9, the most common alteration found in infant AML, with poor prognosis, but that autophagy is no longer needed for the maintenance of established AML or LSC functions in secondary xenotransplantation experiments [54,118,119,120]. However, in a different murine MLL-ENL AML model, knockout of Atg5 or Atg7 decreased the number of functional LSCs, increased activation of mitochondria and ROS levels in these cells, and extended survival of leukemic mice [121]. Together, these representative, variable data suggest a highly complex, context-dependent role for autophagy in leukemic transformation vs. maintenance and LSC properties in AML.

Notably, and similar to CML, many studies in AML have shown another consistently critical role for autophagy with treatment implications: pro-survival protection of leukemic cells upon chemotherapy. For instance, it has been demonstrated that treatment of AML cells with cytarabine (AraC), anthracyclines, or sorafenib activates and increases autophagic flux in these cells, including LSCs, allowing them to resist chemotherapy [54,121,122]. However, in these cases, targeting of autophagy by genetic or pharmacological means, combined with chemotherapy, seems to hold great promise in developing more effective treatment strategies, as discussed later.

Lymphocytic leukemias arise in the bone marrow from a lymphoid progenitor cell, or in the case of chronic lymphocytic leukemia (CLL) from marginal zone B-cells, and take on either a blast-like T-cell or B-cell phenotype [123,124]. Acute lymphocytic leukemia (ALL) is the most common pediatric cancer and, much like AML, ALL is widely diverse in its presentation and driver mutations and progresses quickly if left untreated. CLL—the most common adult leukemia—is more indolent and differentiated than ALL but is also genetically diverse [123]. Within ALL cases, the most common phenotype is the B-cell-like leukemia (B-ALL), and CLL only presents as B-cell-like. Targeted treatments for lymphocytic leukemia are still limited due to the molecular heterogeneity, but clinical advances have greatly improved the prognosis of these diseases with non-targeted chemotherapies [123,124,125,126]. However, subsets of patients acquire therapeutic resistance through unknown mechanisms, and both ALL and CLL have identified autophagy as a potential resistance pathway.

An autophagic treatment response has been identified in lymphoblastic leukemia, specifically after metabolic chemotherapeutics used to target ALL. Standard treatment of B-ALL is the administration of synthetic glucocorticoids (GC) to slow down cellular glucose intake and induce cell death [127]. Following GC treatment in B-ALL, there is a reduction in glucose and glutamine cellular intake but unexpectedly an increase of glutamine synthesis [128]. Dyczynski et al. showed that in the absence of normal nutrient intake, there is increased autophagic flux to catabolically make up for the lost nutrients; a byproduct of this catabolism, ammonia, is then utilized by glutamate–ammonia ligase to synthesize glutamine intracellularly, allowing the cells to provide their own glucose supply [128]. B-ALL patients that were GC-resistant also had 36 differentially expressed autophagy genes (26 upregulated and 10 downregulated) compared to GC-sensitive patients, suggesting the autophagic GC response is important for treatment-mediated apoptosis [129]. ALL treatment with l-asparaginase (l-asp) also targets metabolism by hydrolyzing glutamine and asparagine to glutamic acid and aspartic acid, respectively, thus reducing aerobic metabolism [126]. Upon l-asp treatment, ALL cells induce autophagy to compensate for the metabolic stress, and inhibition of autophagy using CQ induces treatment sensitivity in vitro and in vivo [130].

There are genetic connections between lymphoblastic leukemia driver mutations and autophagy. B-ALL patients with the ETV6-RUX1 translocation have significantly increased expression of the autophagy initiating lipidase VPS34 compared to normal hematopoietic stem/progenitor cells, and the presence of the ETV6-RUNX1 fusion protein was found to have a causal relationship to VPS34 overexpression through epigenetic regulation [131]. In CLL, a positive prognostic factor is the deletion of 13q14, containing microRNA-15 and -16, both of which target anti-apoptotic B-cell CLL/lymphoma 2 (BCL-2) leading to BCL-2 upregulation in CLL [132]. BECN1 is negatively regulated by the binding of BCL-2, and therefore BCL-2 upregulation suppresses BECN1 activity and inhibits autophagy [133]. Another positive prognostic factor of CLL is the expression of SLAMF1, which is involved in autophagosome recruitment and autophagic flux activation [134]. These positive CLL prognostic markers (13q14 deletion and stable SLAMF1 expression) have opposing downstream mechanistic effects on autophagy, yet both positively affect CLL prognosis, showcasing once more how molecular heterogeneity can change the role of autophagy.