Introduction

Oncogenesis has long been conceived as a merely cell-intrinsic process resulting from the accumulation of (epi)genetic defects that underlie the acquisition of a malignant phenotype.1 From this perspective, the causal nexus between the cell-intrinsic functions of macroautophagy (herein referred to as autophagy) and the multistep process through which a normal cell first becomes neoplastic (malignant transformation) and then forms an ever more aggressive and clinically manifest cancer (tumor progression; exhaustively reviewed in ref. 2) readily comes into sight. As a catabolic process consisting of the sequestration and the lysosomal breakdown of cytoplasmic material, autophagy contributes to the maintenance of metabolic homeostasis (by supplying substrates for bioenergetic metabolism from disposable cellular components and/or energy stores in conditions of nutrient deprivation)3 and genomic stability (through the turnover of potentially genotoxic (dam)aged organelles and protein aggregates)4 in every cell lineage. Thus, the transient loss of autophagic functions (favoring malignant transformation) followed by the restoration of autophagic proficiency (supporting the viability of established neoplasms in harsh environmental conditions) perfectly fits in the multistep model classically adopted to describe oncogenesis.2,5 The demonstration that immunodeficient mice are more susceptible to spontaneous and chemical carcinogenesis than their immunocompetent counterparts has unveiled the notion that cancer is not (only) an (epi)genetic disease of individual cells, but (also) an immunological disorder.

According to the immunosurveillance model, the clinical manifestation of cancer invariably results from the escape of neoplastic cells from recognition by the innate or adaptive arm of the immune system.7 Tumors elude immunosurveillance in a dual manner: (1) as cancer cell variants with limited immunogenicity are selected by the immune system (immunoediting), and/or (2) as neoplastic cells actively suppress tumor-targeting immune responses (immunosubversion).8,9 Therefore, the (re)activation of adaptive immune responses, usually mediated by CD8C cytotoxic T lymphocytes (CTLs) that recognize tumor-specific (neo)-antigens, is of the utmost importance for controlling tumor progression. Conventional anticancer agents often fail to mediate therapeutic effects in immunocompromised mice, suggesting that their mode of action involves immunological mechanisms.10 Growing evidence indicates indeed that multiple chemotherapeutics as well as radiation therapy can stimulate anticancer immune responses, either as they directly act on the immune system (“off-target” immunostimulation), and/or as they exacerbate the immunogenicity of malignant cells (“on-target” immunostimulation).10 Immunogenicity is a function of antigenicity, i.e., the ability of cancer cells to be targeted by CTLs upon recognition of tumor-associated epitopes (exposed on the cell surface in association with MHC class I molecules) and adjuvanticity, i.e.

the provision of immunostimulatory signals in response to sublethal or lethal stress.11 Different classes of chemotherapeutics including (but not limited to) anthracyclines (doxorubicin, epirubicin, mitoxantrone), platinum-based compounds (oxaliplatin) and alkylating agents (cyclophosphamide), as well as some forms of radiation therapy, photodynamic therapy, and oncolytic virotherapy, succeed in restraining tumor growth as they trigger immunogenic cell death (ICD), a specific form of cellular demise that culminates in the activation of an adaptive tumor-targeting immune response.12 The ability of some anticancer agents to kill malignant cells in ‘the right way’ converts them into prototypical antigen-donor cells (ADCs) while enhancing the proficiency of antigen-presenting cells (APCs), hence building the foundations for a robust CTL-driven adaptive immune response.12 The cancer cell-extrinsic effects of autophagy cooperate with their cancer cell-intrinsic counterparts for the inhibition of malignant transformation. As an example, autophagic responses in myeloid cells have profound anti-inflammatory functions (related to the inhibition of inflammasome activation), and hence prevent the establishment of chronic inflammatory foci that support transformation.13 Conversely, the cancer cell-extrinsic functions of autophagy rival with their cancer cell-intrinsic counterparts in the context of anticancer immunosurveillance. 

15 In many instances, cancer cells progressively modify the tumor microenvironment upon the recruitment (and/or local differentiation) of myeloidderived suppressor cells (MDSCs) and CD4CIL2RA/CD25C FOXP3C regulatory T (TREG) cells, the latter playing a major role in the inhibition of CTL activity.16 In a murine model of KRASG12D-driven pulmonary carcinogenesis, the ablation of the essential autophagy gene Atg5 (autophagy related 5) ensuing the intranasal delivery of an adenovirus encoding the Cre recombinase accelerates the manifestation of malignant lesions as it increases the frequency of tumor-infiltrating TREG cells.17 Manipulations designed to deplete (i.e., the injection of an antibody specific for IL2RA/ CD25 [interleukin 2 receptor, a chain]) or functionally inhibit TREG cells (i.e., the administration of an antibody targeting IZUMO1R/FOLR4/R4 [IZUMO1 receptor, JUNO])18 retard the development of autophagy-deficient tumors, underscoring the concept that a proficient autophagic program in malignant cells facilitates their immunological control.17 The subversion of immunosurveillance upon autophagy inhibition affects the capacity of cancer cells to release immunostimulatory signals commonly referred to as damage-associated molecular patterns (DAMPs), which are sensed by specific pattern recognition receptors (PRRs) expressed by immune cells.19 Most likely, DAMP release also occurs during early oncogenesis as cancer cells suffer (and sometimes succumb to) oncogenic stress.

 In the context of non-small cell lung cancer (NSCLC), the autophagy-dependent secretion of immunostimulatory ATP (see below)20 is counteracted by the KRAS (KRAS proto-oncogene, GTPase)-driven overexpression of ENTPD1/CD39 (ectonucleoside triphosphate diphosphohydrolase 1), an ecto-enzyme that initiates the conversion of ATP into immunosuppressive adenosine, which also involves NT5E/CD73 (50 -nucleotidase, ecto).21 While ATP binds to purinergic receptors such as P2RY2 (purinergic receptor P2Y, G-protein coupled 2) on immature DCs to favor their recruitment, adenosine works as a chemoattractant for TREG cells through its action on ADORA2A (adenosine A2a receptor) and ADORA2B (adenosine A2b receptor).21 Hence, when autophagy is inhibited, malignant cells preferentially recruit TREG cells over DCs, thus generating an immunosuppressive tumor microenvironment. Immunohistochemical analyses of human breast carcinoma lesions revealed that the presence of cytoplasmic MAP1LC3B/ LC3 (microtubule associated protein 1 light chain 3 b) puncta linked to reduced levels of SQSTM1/p62 (sequestosome 1), which together are indicative of a functional autophagic response, correlated with an improved ratio of CTLs over TREG cells.22 Similarly, in a rodent model of non-alcoholic fatty liver disease (NAFLD)-driven hepatocellular carcinoma, in which autophagy is disabled by the accumulation of toxic lipid droplets in hepatocytes, tumor progression was associated with the depletion of tumor-infiltrating CD4C T lymphocytes.

23 Altogether, preclinical and clinical evidence suggests that, at least in some cancer types, the autophagic proficiency of malignant cells modulates tumor infiltration by myeloid and lymphoid cells to support the establishment of an immunostimulatory tumor microenvironment. Of note, autophagy may not only boost the adjuvanticity of cancer cells but also exacerbate their antigenicity.24 As a matter of fact, the immunopeptidome of autophagy-competent cells substantially diverges from that of their autophagyincompetent cells. Owing to the block in protein translation that characterizes the initiation of autophagic responses, as well as to the role of autophagy in miRNA homeostasis, autophagic cells are indeed characterized by a unique repertoire of novel MHC class I epitopes.14 Such peptides can be either presented on the surface of cancer cells or processed and crosspresented by DCs (upon loading on MHC class I molecules) through a cascade of events that is facilitated by autophagy.24 Autophagy also mediates immunostimulatory functions as it supports the survival and function of APCs and CTLs. Besides its effects on the processing of exogenous MHC class II epitopes 2164 F. PIETROCOLA ET AL. (reviewed in ref. 25), autophagy is involved in antigen crosspresentation by DCs, although the underlying mechanisms remain to be precisely elucidated.

First, most pharmacological blockers of autophagy are rather nonspecific as they target PIK3C3/Vps34 (phosphatidylinositol 3-kinase catalytic subunit type 3), such as wortmannin and 3-methyladenine, or affect lysosomal acidification, such as chloroquine and its derivatives.28 While the efficacy of these drugs, alone or in combination with a plethora of chemotherapeutics or with radiation therapy, has been extensively corroborated in vitro or in patient-derived cancer cell lines implanted into immunodeficient mice,27 the link between their anticancer activity and autophagy remains questionable. Thus, chloroquine sensitizes mouse cancer cells to the antineoplastic effects of the DNA-damaging agent cisplatin irrespective of the expression of ATG12 (autophagy-related 12).29 The knockout of ATG7 (autophagy-related 7) fails to affect the antiproliferative activity of chloroquine in multiple cancer cell lines expressing mutant KRAS.30 Finally, the ability of chloroquine to reduce the growth of melanoma cell xenografts is tied to antiangiogenic effects rather than to bona fide autophagy inhibition.31 Second, most studies suggesting that autophagy inhibition improves the efficacy of cancer therapy, including those in which autophagy was genetically blocked with short-hairpin RNAs (shRNA) targeting essential components of the autophagic machinery, were performed in immunodeficient mice, which obviously excludes any crosstalk between autophagy and the immune system from the working hypothesis.32 Third, several groups used models of genetically driven cancers in which the overexpression of an oncogene was coupled to tissue-specific ablation of floxed Atg5 or Atg7. 

In the context of KRASG12D- and BRAFV600E-driven pulmonary and pancreatic oncogenesis,17,33,34 autophagy-incompetent adenomas fail to evolve into adenocarcinomas, confirming the involvement of proficient autophagic responses in tumor progression. Although these findings were obtained in immunocompetent hosts, the evidence that autophagy-mediated repression of tumor growth relies on the presence of a functional TRP53/p53 (tumor protein p53) system17,35 complicates data interpretation. Moreover, it is uncertain whether a total and irreversible inhibition of autophagy (as obtained by the deletion of Atg5 or Atg7) truly reflects the effects of pharmacological autophagy inhibitors (which would be expected to be partial and reversible). Indeed, combined treatment with chloroquine or hydroxychloroquine fails to yield any significant improvement in the efficacy of multiple anticancer agents in clinical studies.32 Instead, it has been proposed that such treatments might initiate a vicious cycle facilitating malignant transformation in other tissues.2,13 In summary, although much emphasis has been laid on autophagy inhibition as a potential strategy to reduce the fitness of malignant cells, whether this kind of therapeutic intervention provides actual benefits to cancer patients remains to be formally demonstrated.

12 as well as (5) release of ANXA1 (annexin A1), which interacts with FPR1 (formyl peptide receptor 1) on maturing DCs to ensure their recruitment into close proximity of dying cancer cells.38 All these functional aspects represent an indispensable condition for chemotherapy-induced ICD to be AUTOPHAGY 2165 perceived as immunogenic, and constitute a unique tool for the identification of novel ICD inducers.39 As a modifier of cellular adjuvanticity, autophagy is actively involved in the release of ICD-associated DAMPs. Thus, autophagy-deficient cells fail to expose phosphatidylserine (a prominent phagocytic signal) on the plasma membrane as they die, as a consequence of pre-mortem intracellular ATP depletion.40 Along similar lines, an intact autophagic machinery is required for the routing of ATP from lysosomes to the extracellular milieu in the course of ICD.20 Through this mechanism, autophagy favors the generation of an extracellular ATP gradient that engages purinergic P2RY2 and P2RX7 receptors on myeloid cells, including DCs and their precursors.21 In an immunocompetent syngeneic background, the efficacy of mitoxantrone- or oxaliplatin-based chemotherapy against mouse MCA205 fibrosarcomas and mouse CT26 colorectal carcinomas was reduced by the inhibition of autophagy in malignant cells with Atg5- or Atg7-targeting shRNAs.41 The limited response of autophagy-deficient tumors to chemotherapy with ICD inducers can be reproduced: (1) in tumor-bearing athymic nu/nu mice (which are constitutively immunodeficient, because they lack T lymphocytes); (2) in wild-type mice that are depleted from CTLs by means of a CD8-targeting antibody;

 (3) in wild-type mice bearing ENTPD1-overexpressing tumors. Conversely, the co-administration of the ENTPD1 inhibitor ARL67156 rescues the response of autophagy-deficient tumors to mitoxantrone and oxaliplatin by increasing pericellular ATP levels.41 Consistently, in a murine model of BRAFV600E-driven melanoma in which delivery of the Cre recombinase into melanocytes engenders oncogene activation and loss of the oncosuppressor Pten, the systemic administration of mitoxantrone mediates optimal therapeutic effects only in autophagy-proficient tumors. Indeed, mitoxantrone loses its efficacy when Atg7 is ablated in melanoma cells as well as when CD4C T lymphocytes are depleted by means of specific antibodies.42 Similarly, the irradiation of CT26 colorectal carcinomas delays tumor progression only if malignant cells are implanted in immunocompetent settings and if tumors can mount an effective autophagic response. Thus, while the stable downregulation of Atg5 by a specific shRNA sensitizes CT26 cells to radiation therapy when tumors are established in athymic nu/nu mice, this maneuver causes relative radioresistance in immunocompetent syngeneic C57BL/6 mice. Also in this model, the administration of ARL67156 restores the response of autophagy-deficient neoplasms to treatment as it promotes tumor infiltration by CTLs.43 Altogether, these findings delineate a landscape in which autophagy-dependent ATP release in response to immunogenic cancer treatment is required for the induction of a therapeutically relevant tumor-targeting immune response. 

We surmise that these events could favor ATP release and activate TMEM173/STING (transmembrane protein 173), a cAMP-dependent protein that stimulates the expression of type I IFN-related genes, hence mimicking anthracyclines in their capacity to elicit viral mimicry in cancer cells.37 Culturing malignant cells in nutrient deprived-medium or fasting mice for 2 d (with ad libitum access to water) are strong triggers of autophagy. Nutrient starvation (by itself or combined with anticancer agents) restrains tumor progression unless malignant cells have activating mutations in PIK3CA (phosphatidylinositol 3-kinase, catalytic, a polypeptide), which operates downstream of IGF1R (insulin-like growth factor 1 receptor) to suppress autophagy.46 Similarly, the reduction in circulating IGF1 (insulin-like growth factor 1) levels normally associated with fasting cycles of 48 h synergizes with immunogenic chemotherapies (including doxorubicin, cyclophosphamide, mitoxantrone and oxaliplatin) in limiting the growth of mouse 4T1 breast carcinomas, B16 melanomas and/or MCA205 fibrosarcomas evolving in immunocompetent syngeneic mice.47 The ability of fasting to synergize with chemotherapy could be replicated by IGF1R inhibitors, yet is obliterated: (1) if tumors are inoculated into athymic nu/nu mice;48 (2) upon systemic administration of excess IGF1;47 (3) when nonimmunogenic chemotherapeutics such as cisplatin are used;49 (4) if MCA205 cells are genetically manipulated to deplete ATG5 and ATG7.