The phosphatidylinositol 3-kinases are a family of enzymes that can phosphorylate phosphoinositides of cell membranes, and have thus been deemed as the initiators of cellular signal transduction. PI3Ks are divided into three classes. However, only Class I and III PI3Ks have been shown to be associated with autophagy (25). The Class I PI3K promotes the generation of phosphatidylinositol-3,4,5-trisphosphate (PIP3) at the plasma membrane, thereby increasing Akt/PKB recruitment. The activated Akt/PKB hinders its downstream TSC1-TSC2 complex, influences GTP binding, and motivates the mammalian target of rapamycin (mTOR), which acts as a negative regulator of autophagy by sensing variable ATPs, amino acids, and energy metabolism (2,26). In addition, mTOR can be directly inhibited by rapamycin treatment, leading to autophagy induction. Conversely, Class III PI3K activates autophagy. It phosphorylates PI to generate PI3P, which contributes to the movement of lysosomal enzymes for component degradation (27). Previous findings have shown that the Class III PI3K integral protein Beclin-1, which is also known as the mammalian homolog of an essential yeast autophagy gene (Atg6), can facilitate phagosome nucleation, maturation and clearance of apoptotic cells (28). Therefore, knockdown or silence of Beclin-1 inhibits autophagy and causes cell death (29).

The endoplasmic reticulum (ER) is in charge of protein folding. Under stressed circumstances, the accumulation of immature proteins provokes ER stress response and boosts autophagy, resulting in autophagic cell death (30). This process is mainly regulated via the IRE1α-JNK and PERK-eIF2α-ATF4 pathways (31,32).

AMP-activated protein kinase is the energy sensor. In energy-defective situations (decreased ATP/AMP ratio and starvation), the AMPK is excited by LKB1 and p53, leading to autophagy activation via the positive regulation of the TSC1-TSC2 complex or negative regulation of mTOR. AMPK can also directly suppress the raptor, which is a subunit of mTORC1 that inhibits autophagy activity (16,33).

The relationship between RAS and autophagy is multifaceted. In fact, RAS plays both positive and negative roles in autophagy with different pathways. On the one hand, RAS inhibits autophagy by directly activating Class I PI3K-Akt-mTOR1. By contrast, RAS upregulates Raf-1/MEK/ERK- or Rac1/MKK7/JNK-dependent autophagy activity with the help of ATG5/7 and Beclin-1 (34–36).

The Bcl-2 family is divided into three types, based on their BH domains and functions. Bcl-2, Bcl-xL and myeloid cell leukemia sequence-1 (MCL-1) act as anti-apoptosis pathways, while BAX-like proteins (BAX, BAK) and BH3-only proteins (BNIP3) promote apoptosis (2). Similarly, these proteins play a bidirectional role in autophagy. On the one hand, the BNIP3 can integrate with Bcl-2, which correspondingly decreases Bcl-2 binding to Beclin-1, thereby liberating Beclin-1 for autophagy induction (37,38). By contrast, Bcl-2, Bcl-xL and MCL-1 inhibit autophagy.

Death-associated protein kinase (DAPK) is associated with autophagy in cancer. Hyperactive DAPK influences its related protein kinase DRP-1, which increases plasma membrane flexing and blebbing, enabling phagocytosis (39). In addition to the aforementioned p53-AMPK network, DNA damage can trigger autophagy through p53-dependent damage-regulated autophagy modulator-1 (DRAM-1) activation (40).

In summary, autophagy networks are sophisticated. Dysfunctions of these signal pathways are involved in tumorigenesis, and targeting signal transduction proteins may provide potential therapeutic schemes for cancer.

The pro-death role of autophagy is vital for suppressing tumor progression and enhancing antitumor response. Fludarabine (Fd), a nucleoside analog that inhibits DNA/RNA synthesis and DNA repair, is widely used for chronic lymphocytic leukemia (CLL). Mahoney et al observed fludarabine-triggered autophagy activity in CLL cells. However, the inhibition of autophagy did not enhance or diminish cell death induced by fludarabine (47). Recently, a study by Sharma et al reported that fludarabine induced BECN-dependent autophagy by decreasing Mcl-1 expression, resulting in autophagic cell death in Fd-sensitive leukemic B cells (48). However, in Fd-resistant leukemic cells, Mcl-1 expression was increased, which inhibited BECN-dependent autophagic cell death. The Bcl-2 family also contributed to promoting cell survival via sequestering BECN. Moreover, activated AMPK-dependent autophagy assisted Fd-resistant cells escape cell death. Of note, adding the BCL-2 inhibitor obatoclax can induce cell death in both Fd-sensitive and Fd-resistant leukemic cells. The findings suggest that combining autophagy repressors such as Mcl-1, Bcl-2 and AMPK inhibitors, with fludarabine may improve the curative effect for CLL patients, especially those who acquire Fd-resistance after several cycles of fludarabine chemotherapy.

Metformin, an anti-diabetic agent, has already been proven to play a positive role in cancer control, including acute leukemia (47–50). Wang et al reported that metformin enhanced the anti-leukemic effect of sorafenib by triggering mTOR-dependent autophagy in FLT3-ITD-positive AML, which suggested the combination strategy of the two drugs in AML treatment (51). Metformin also induced a dose-dependent inhibition of lymphoma cell proliferation in vitro and in vivo via the negative control of the mTOR signal pathway (52). Metformin-induced autophagy enhanced the anti-lymphoma effect of doxorubicin and temsirolimus (mTOR inhibitor), offering a novel therapeutic strategy.

Sertraline, a widely used antidepressant drug, was found to induce leukemia cell death mediated by autophagy and apoptosis. It also enhanced the sensitivity of chemotherapy drugs, thereby providing potential alternative strategies for leukemia treatment (53).

Trocoli et al observed that all-trans retinoic acid (ATRA) can promote autophagy by reducing the activity of Bcl-2 and mTOR in acute promyelocytic leukemia (APL). In addition, the upregulation of Beclin-1 contributed to the stability of mature APL cells in a non-autophagic manner, although autophagy-related cell differentiation was not explored (54). Notably, by combining ATRA with RAD001 (mTOR1C inhibitor), Nishioka et al observed step-down cell growth and enhanced cell differentiation of AML cells (55). Similarly, Isakson et al reported that ATRA-dependent autophagy induced the degradation of PML-RARα, which is known to create abnormal granulocytic differentiation, via inhibition of the mTOR pathway as well (56). Thus, the simultaneous use of ATRA and mTOR suppressant may be a valid protocol for APL patients.

Arsenic trioxide (As₂O₃), another metalloid poison that has conventionally been used in APL patients, also induced cell death. We previously reported that As₂O₃ led to cell growth arrest by means of both apoptosis and Beclin-1-dependent autophagy (58). Goussetis et al also documented a MEK-ERK-mediated autophagic cell death in leukemia cells (59). The studies demonstrated As₂O₃-induced autophagy as the mechanism of cell death.

Autophagy in response to idarubicin served as a pro-death mechanism in leukemic cells as well. Idarubicin restrained mTOR either by upregulation of its upstream inhibitor AMPK, or downregulation of its activator Akt, thus leading to autophagic cell death. Pharmacological impairment (bafilomycin A1 or chloroquine) of autophagy partially reduced the cytotoxicity of idarubicin towards REH cells, which highlighted its cell-killing function (60).

Dexamethasone (Dex) also contributed to cell death in lymphoid leukemia and multiple myeloma. Several experiments observed Dex-induced autophagic cell death through the marked upregulation of promyelocytic leukemia protein (PML) and PML-dependent Akt dephosphorylation (61,62). In depth studies on PML and Akt-dependent intracellular pathways may provide optimized use of Dex in the treatment of lymphoid malignancies.

Besides its potential to induce cell death, autophagy is appropriated to serve as a cell survival mechanism as well.

For example, the tyrosine kinase inhibitor AG490 was found to induce both apoptosis and autophagy by inhibiting STAT3, reducing the expression of HSF1/HSP70, and downregulating Mcl-1 in primary effusion lymphoma (PEL) cells. A combination of AG490 with autophagy inhibitor bafilomycin is believed to enhance the cytotoxic effect (63,64). The HSP70 and its main transcription factor HSF1 have been reported as essential for PEL survival by affecting protein folding and cell stability, and HSP70 inhibition via 2-phenylethynesulfonamide could block the autophagic process and induce immunogenic cell death (65). Therefore, a combination of AG490 and HSP inhibitors may provide a potential treatment strategy. Another TKI, imatinib (IM), was also reported to activate a cytoprotective autophagy in chronic myeloid leukemia (CML) (66). Rothe et al identified that CML stem/progenitor cells obtained imatinib resistance due to its intensive pro-survival activities associated with elevated autophagy gene transcripts and proteins, especially ATG4B (67). Mancini et al describe a survival role of autophagy in imatinib-treated Bcr-Abl-positive cells, possibly connected with ER stress and following responses (68). The group of Howard Hughes Medical Institute suggested that imatinib-induced autophagy was associated with the Bcr-Abl/PI3K/Akt/FOXO4/ATF5/mTOR pathways (69). Recently, the combination of NVP-BEZ235 (dual PI3K/mTOR inhibitor) and imatinib was reported to significantly inhibit CML cell growth and proliferation, as well as enhance sensitivity to imatinib in imatinib-resistant CML cells (70). Inversely, imatinib-induced autophagy also isolated the mosaic gene Bcr-Abl in autophagosomes through the regulation of Beclin-1, which prevented disease progression (71). Therefore, autophagy inhibition of the aforementioned specific autophagic pathways would enhance the treatment effect of imatinib in CML.

Zhu et al also reported a pro-survival role of autophagy in CML treated with interferon. Interferon-1 activated both JAK1-STAT1 and NF-κB pathways, leading to increasing BECN1-ATG5-ATG7, one of the major regulators of the classical autophagy pathway. Inhibiting autophagy by using chloroquine enhanced the cytotoxic effect of interferon-1 (72).

Cytarabine was reported to increase the phosphorylation of AMPK/ERK, inhibit Akt and attenuate mTOR activity, thereby inducing cytoprotective autophagy in leukemic cells. Pharmacological (bafilomycin A1 and chloroquine treatment) or genetic (siRNA downregulation of either LC3b or autophagic cargo receptor p62) impairment of autophagy markedly increased cell death by accumulating DNA fragmentation and mitochondrial damage as well as activating oxidative stress, thus providing the combination strategy of cytarabine and autophagy inhibitor in leukemia treatment (73).

Notably, doxorubicin and melphalan were shown to motivate autophagy in multiple myeloma (MM) cells, which provided an adaptive condition for evading DNA-damaging stimulus. The increasing Beclin-1-PI3KC3 complex and Bcl-2 disintegrated from Beclin-1, functioned as a cell survival mechanism in this process. Thus, the inhibition of autophagy with DNA-damaging agents likely augmented anti-myeloma activities without additional side effects (74). In cases of acute lymphocytic leukemia (ALL) treated with bortezomib, cytoprotective autophagy was also associated with the balance between the Bcl-2-Beclin-1 complex and the Beclin-1-PI3KC3 complex, as its downregulation by chloroquine potentiated anti-ALL activity (75). The aforementioned results indicated a pivotal role of Bcl-2-Beclin-1-PI3KC3 complexes in autophagy and provided a novel target to improve clinical efficacy. Another experiment provided evidence of autophagic cell proliferation in aggressive B-cell lymphoma managed with bortezomib. In that study, ER stress-associated IRE1-JNK was stimulated, followed by phosphorylated Bcl-2 and elevated LC3-II, which was widely used for autophagic flex surveillance. Certainly, the addition of JNK inhibitor SP600125 enhanced the anti-lymphoma effects of bortezomib (76).

Anthracycline daunorubicin (DNR) revealed potent leukemia killing activity by DNA damage. However, DNR-induced autophagy was inversely cytoprotective, and authors of that study identified the activation of MAPK (MEK-ERK) in DNR-treated myeloid leukemia cells (77). Obviously, adding MEK1/2 inhibitor U0126 to daunorubicin would enhance chemotherapeutic response and reduce drug toxicities.

The example of the autophagic pro-survival role in CLL originated from the observation that dasatinib was capable of sustaining stable cellular metabolism from AMPK activation and consequent autophagy motivation, which was also the mechanism of drug resistance (78). However, Morita et al reported that dasatinib induced autophagy-dependent cell death in Bcr-Abl-positive leukemia cells with central nervous system (CNS) infiltration in vitro and in vivo, confirming the important mechanism of autophagy for CNS leukemia treatment (79).

Of note, sorafenib regulated the mTOR or Mcl-1 pathway in multiple myeloma with different outcomes (80): on the one hand, mTOR inhibition by sorafenib induced cytoprotective autophagy; on the other hand, sorafenib-triggered downregulation of Mcl-1 was essential to cell death, and the addition of ABT737 (Bcl-2 antagonist) improved the efficacy of sorafenib. As a result, a combination of sorafenib with autophagy inhibitor and ABT737 is a potentially new treatment strategy in multiple myeloma.

Recently, Takahashi et al demonstrated that L-asparaginase (L-asp)-triggered autophagy contributed to ALL cell survival by eliminating accumulated DNA damage and injured mitochondria with the help of a ROS-p53 loop, indicating that combination treatment with an autophagy inhibitor enhanced anti-leukemic effects of L-asp and overcame drug resistance for ALL patients (81).

Thus, the aforementioned studies fully demonstrated the dual role of autophagy in hematological malignancies, of which induction or inhibition may improve the therapeutic efficacy depending on the types of antitumor agents and pathways of stress responses.