Functions of autophagy in the tumor microenvironment and cancer metastasis

Abstract

Macro-autophagy is an ancient and highly-conserved self-degradative process that plays a homeostatic role in normal cells by eliminating organelles, pathogens and protein aggregates. Autophagy, as it is routinely referred to, also allows cells to maintain metabolic sufficiency and survive under conditions of nutrient stress by recycling the by-products of autophagic degradation, such as fatty acids, amino acids and nucleotides. Tumor cells are more reliant than normal cells on autophagy for survival in part due to their rapid growth rate, altered metabolism and nutrient deprived growth environment. How this dependence of tumor cells on autophagy affects their progression to malignancy and metastatic disease is an area of increasing research focus. Here, we review recent work identifying critical functions for autophagy in tumor cell migration and invasion, tumor stem cell maintenance and therapy resistance and cross-talk between tumor cells and their microenvironment.

Graphical abstract

Autophagy is upregulated as tumors progress to become malignant and this review addresses how autophagy modulates cancer metastasis. Specifically, we examine the critical role played by autophagy in tumor cell migration and invasion, in maintaining the cancer stem cell phenotype and tumor cell dormancy as well as how autophagy affects interactions of tumor cells with components of the tumor microenvironment. Finally, we discuss how these functions make autophagy inhibition an attractive therapeutic option for metastatic cancers.

Autophagy & Cancer

The process of macro-autophagy involves the formation of double-membraned vesicles called autophagosomes, around the cellular content that is to be degraded, known as autophagic cargo [1] [2]. Such cargo typically includes organelles of all types, intra-cellular pathogens and protein aggregates. Induction of autophagy is controlled at the transcriptional and post-translational level (figure 1). Autophagy (Atg) genes are transcriptionally regulated by the ATF4 and MIT/TFE transcription factors [3, 4], and as such are induced by cellular stresses including amino acid deprivation and ER stress (ATF4) known to promote autophagy, as well as by other signals that inhibit mTOR, a negative regulator of MIT/TFE factors (figure 1)[4]. However, autophagy is primarily regulated in a post-translational manner to permit a rapid response to nutrient stress. The formation of autophagosomes is regulated by a pre-initiation complex involving the Ulk1/Ulk2 serine/threonine kinase that is sensitive to amino acid supply and cellular energy status, as a result of being regulated negatively by mTOR and positively by AMPK (figure 1) [5, 6]. As part of the pre-initiation complex with ATG13 and FIP200, Ulk1/2 phosphorylates Beclin1 to activate the lipid kinase activity of Vps34 (a class III PI3K), the catalytic component of the initiation complex, increasing phospho-inositol-3-phosphate (PIP3) production and recruiting further components of the autophagy machinery to drive the process forward (figure 1) [1]. A key consequence of increased initiation complex activity is the recruitment and activation of the conjugation complexes, including ATG5-ATG12/ATG16L that in turn recruits processed LC3 and related molecules to the expanding phagophore (figure 1) [1, 6]. Processed LC3 at nascent phagophores is key to selecting cargo for degradation and proteins, or their cargo adaptors, at cellular structures destined for degradation contain specific motifs known as LC3-interacting region (LIR) motifs that promote their interaction with LC3 (figure 1) [7, 8]. Closure of the phagophore to form a mature autophagosome requires LC3-related proteins leading to fusion with the lysosome resulting in degradation of autophagosomal cargo and recycling of constituent amino acids, nucleotides and fatty acids (figure 1) [1, 2, 9]. Novel non-canonical functions for autophagy proteins are emerging, including roles in anti-viral immunity, secretory processes and endocytic trafficking that are distinct from autophagy but rely on a select cadre of Atg gene products [10, 11]. Together with the increasing variety of cargo that autophagy seems capable of turning over, this has complicated our ability to dissect at a mechanistic level how autophagy influences disease processes, such as cancer. In particular, the pleiotropic nature of autophagy has made it challenging to determine whether autophagy is tumor-promoting, and therefore a valid therapeutic target, or indeed tumor-suppressive which might limit its value as a target in cancer treatment [12].

Figure 1. Autophagy overview.

Autophagy is exquisitely sensitive to amino acid deprivation in addition to other nutrient stresses. Amino acid deprivation induces AMPK and suppresses mTORC1 to promote Ulk1 kinase activity as part of the pre-initiation complex (Ulk1/FIP200/Atg13) that in turn activates the lipid kinase activity of the initiation complex (Beclin1/VPS34 and associated proteins). This is turn recruits the Atg12/Atg5/Atg16L conjugation complex that promotes processing and insertion of lipid conjugated LC3 into nascent phagophore membranes. LC3 and related proteins interact with cargo to be turned over via autophagy and also promotes closure of the autophagosome around the cargo. Fusion of these double-membraned autophagosomes with the lysosome results in proteolytic degradation of the cargo and release from the autophagolysosome of amino acids, nucleotides and lipids derived from degraded cargo. In addition to induction of autophagy by these post-translational mechanisms, autophagy is also stimulated by transcriptional control of Atg genes, notably by ATF4 and MiT/TFE transcription factors that are activated by amino acid deprivation amongst other stresses.

Much of the focus on the role of autophagy in cancer has been aimed at its role in tumor metabolism [9] where its ability to modulate mitochondrial turnover has been proposed as the main explanation for the altered metabolic phenotype in tumors forming in autophagy-deficient GEM models. Such models, including Ras-driven lung and pancreatic cancers [13–16], showed reduced fatty acid oxidation, a switch to Warburg metabolism and increased ROS, consistent with failure to degrade mitochondria in an appropriate manner [12, 17]. More recently, the role of autophagy in modulating how the tumor microenvironment sustains growing tumors has also come to the fore [18, 19]. Beyond its role in modulating tumor metabolism, autophagy performs other functions that likely contribute to its role in cancer and response to cancer therapy [12].

Metastasis is a multi-step process in which tumor cells at the primary tumor site undergo epithelial-to-mesenchymal transition (EMT) [20], acquire invasive and other properties that allow them to escape through the basement membrane into the vasculature, in a part of the metastatic cascade known as intravasation [21]. This is followed by selective pressure on escaped tumor cells to survive in the circulation where they sometimes become lodged, or if they are able to survive at all, use their invasive properties to extravasate at a distant site [21]. Frequently, such tumor cells alight in an environment that does not favor their growth and where they are again subject to evolutionary pressures to survive [22]. However, even if tumor cells survive such seemingly unfavorable conditions, they may not be able to expand and can remain dormant at such sites for years [22, 23]. Recently, it has emerged that primary tumors can release signals that mobilize other cell types to establish a pre-metastatic niche ahead of disseminating tumor cells landing there to promote the survival and outgrowth of metastatic lesions [24–26]. Throughout all these steps in the metastatic cascade, tumor cells also need to evolve the means to evade immune surveillance and T-cell mediated killing [27]. Intriguingly, many features of the metastatic tumor cell, whether that be mesenchymal properties or ability to escape immune surveillance, are associated with a stem-like phenotype and linked to drug resistance [20, 28]. Autophagy is up-regulated in primary human breast, glioblastoma, melanoma, esophageal cancer and hepatocellular carcinoma upon progression to advanced metastatic disease and expression of autophagy markers in these cancers is associated with poor prognosis [29–31] [32], highlighting novel and important roles for autophagy at different points in the metastatic cascade. Here, we review recent work defining new functions of autophagy in cancer metastasis including how autophagy promotes acquisition of pro-migratory and invasive properties, maintains tumor cell stemness and drug resistant phenotypes, and molds the co-evolution of tumors with their microenvironment.

Autophagic control of tumor cell migration and invasion

Increased motility is required for tumor cells to escape the primary tumor site and to successfully colonize secondary sites during metastasis [23, 33, 34]. This involves protrusion of the plasma membrane in the forward direction, adhesion of cellular integrins to the extra-cellular matrix (ECM) and induced signaling through focal adhesions [35, 36], polymerization of the actin cytoskeleton and contraction of the cell body [37], and finally detachment of the cell from the ECM behind the direction of movement [36]. Cells vary in how they migrate (amoeboid versus mesenchymal, for example) depending on cell type and microenvironment [36, 38], but all cell migration involves exquisite coordination of the processes described above, in some fashion. Over the past several years, increasing evidence has emerged to demonstrate a direct role for autophagy in tumor cell motility and invasion during metastasis [39–42], including through turnover of components of the cell migration machinery [43, 44] and ECM proteins [45], as well as other roles such as modulation of the tumor cell secretome [46].

Various studies have demonstrated a key role for autophagy in focal adhesion dynamics during cell migration and invasion [42–44] [47–49]. Inhibition of focal adhesion kinase (FAK), such as occurs during cell detachment from the sub-stratum, resulted in autophagic degradation of active SRC kinase. This required c-CBL, a SRC binding protein and E3 ubiquitin ligase that contains a LIR motif and interacts directly with processed LC3B to promote autophagic turnover of SRC [42]. SRC turnover by autophagy also required Ambra1, a multi-functional protein known as an inhibitor of the Beclin1-VPS34 initiation complex but that alternatively can interact with FAK to limit active SRC at focal adhesions and promote SRC degradation by autophagy [49]. Inhibition of autophagy restored SRC expression at focal adhesions in adhesion-stressed cells but was associated with cell death [42] indicating that autophagic turnover of SRC was required in response to cell detachment or FAK inactivation (figure 2). Interestingly, FAK inhibition increased autophagic flux suggesting that FAK activity inhibits autophagy, possibly by sequestering or otherwise interfering with FAK-interacting protein of 200 kD (FIP200) that although first identified as a protein that interacts with and inhibits FAK kinase activity [50], has subsequently been identified as the mammalian homologue of yeast autophagy gene Atg17 (Autophagy related gene 17) and a key component of the pre-initiation complex in mammals that interacts with ULK1 and ATG13 (figure 1) [51]. FAK may also inhibit autophagy through its inhibitory interaction with Tuberous Schlerosis Complex 2 (TSC2) resulting in activation of mTOR (figure 2)[52, 53]. It will be important in future studies to determine whether FAK requires FIP200, TSC2 or other signaling mechanisms to modulate autophagy and to assess the extent to which FIP200 mediates cross-talk between autophagy and cell migration.

Figure 2. Coordinate control of autophagy and cell migration.

Autophagy and cellular motility are coordinated at the molecular level through reciprocal control of both processes by common modulators including Focal adhesion kinase (FAK), FAK-Interacting Protein 200 kD (FIP200) and Paxillin (PXN). Autophagy promotes cell motility through turnover of focal adhesion complexes and specifically via degradation of focal adhesion component, Paxillin (PXN). Autophagy also promotes turnover of SRC in response to FAK inhibition to prevent cell death in response to cell detachment. Both FAK (through TSC2) and PXN (through ill-defined mechanisms) feed back to modulate autophagic flux. Similarly, there is a reciprocal interaction between autophagy and members of the Rho family of small G proteins involved in cell migration. Autophagy promotes Rho turnover via p62/Sqstm1 while ROCK1 that acts downstream of Rho stimulates autophagy through phosphorylation of Beclin1.

Recent work from our laboratory demonstrated a critical role for autophagy in the motility and invasion of metastatic tumor cells in vitro and for tumor metastasis in vivo [44]. Inhibition of autophagy reduced tumor cell motility due to decreased focal adhesion disassembly. This was attributed to accumulation of Paxillin (PXN), a core component of focal adhesions [44, 48] and PXN was identified as a LC3-interacting protein that contains a conserved LIR motif (figure 2) [44]. The interaction between PXN and LC3B was promoted by oncogenic SRC and required the Y40 residue at position +1 of the LIR motif in PXN [44], a site previously identified as a target of SRC phosphorylation [54]. Consistently, the ability of oncogenic SRC to promote cell motility and invasion was dependent on phosphorylation of Y40, interaction of PXN with LC3 and functional autophagy (figure 2) [44]. The targeting of PXN for autophagic degradation in the highly metastatic tumor cells studied did not require either of the cargo adaptors p62/Sqstm1 or Near BRCA1 (NBR1) [44] but a different mechanism may be at play in other cell types since in Ras-transformed MCF10A breast epithelial cells, focal adhesion turnover by autophagy was specifically dependent on NBR1 (figure 2) [43]. In addition, c-CBL has also been reported to be required for targeting PXN to autophagosomes for degradation [48], in addition to its role in promoting SRC turnover [42]. Similar to FAK that is both a regulator of autophagy and regulated by autophagy, PXN is required for efficient autophagosome formation in MEFs [55], is phosphorylated by Ulk1 and along with vinculin relocates from focal adhesions to autophagosomes in response to nutrient deprivation [55]. These studies highlight a critical role for autophagy in focal adhesion dynamics in tumor cells and a reciprocal role for focal adhesion components in modulating autophagy.

An intriguing reciprocal relationship also exists between control of the Rho family of small GTPases and autophagy during cell migration. RhoA, Rac1 and CDC42 GTPases modulate cell motility by promoting formation of membrane protrusions, lamellopodia and filopodia respectively [36, 56, 57]. The ability of rho1 to induce hemocyte migration during wound healing in Drosophila was dependent on atg1 (autophagy related gene-1) and ref(2)P the fly homologue of cargo adaptor p62/sqstm1 [40]. Chemical inhibition of autophagy prevented blood cell migration to larval wound sites in flies while knockdown of beclin1 or ulk1 prevented mouse macrophages spreading in response to inflammatory signals [40]. p62/Sqstm1 has since been shown to target mammalian RhoA to the autophagosome for degradation [58] with the failure to turn over RhoA in cells knocked down for ATG5 resulting in RhoA build-up at the midbody during mitosis, cytokinesis defects and aneuploidy [58]. Conversely, Rho signaling has been implicated in the regulation of autophagy [59, 60] with Rho-associated kinase 1 (ROCK1) identified as a regulator of starvation-induced but not basal autophagy [59]. Inhibition of ROCK1 resulted in the formation of enlarged, immature autophagosomes leading the authors to suggest that ROCK1 promotes autophagy by limiting time spent in early phagophore elongation phases of autophagy [60]. ROCK1 is also activated by amino acid deprivation leading to direct phosphorylation of Beclin1 by ROCK1 on Thr119 causing disruption of the Beclin1/Bcl-2 complex resulting in derepression of autophagy (figure 2) [61, 62]. Meanwhile, Rac1 plays a role in modulating Rab7, a different small GTPase that promotes maturation of late stage autophagosomes and lysosomal fusion [63]. The efficient cycling of Rab7 GTPase activity as required for autophagolysosome maturation in response to amino acid starvation is controlled by armus, a RabGAP that localizes to autophagosomes through direct interaction with LC3 [64, 65]. Rac1 also interacts with armus and promotes autophagic recycling of E-Cadherin during EGF-stimulated cell scattering [64]. Failure to inactivate Rac1 during amino acid starvation blocked autophagy due to Rac1 binding to LC3 interfering with the interaction of armus with LC3, thereby preventing Rab7 localization to maturing autophagolysosomes [65]. Finally, RhoA, Rac1 and CDC42 were also identified within an autophagy-centered human gene interaction network built on known autophagic responses to microenvironmental cues (figure 2) [66]. These and other studies highlight yet again how cell migration is molecularly coordinated with control of autophagy and vice versa

Autophagy in Tumor Stem cells and Metastatic Dormancy

Autophagy is required for normal tissue stem cell maintenance and differentiation [67–69] with hematopoietic stem cells dependent on autophagy for survival [70, 71] and muscle stem cells dependent on autophagy to prevent senescence [72]. More specifically, mitophagy has been directly implicated in preventing stem cell aging and promoting stem cell self-renewal [73–75] such that defective mitophagy in mammary stem cells inhibited preferential segregation of younger mitochondria to daughter stem cells and older mitochondria to daughter non-stem cells [73]. The requirement for mitophagy in the self-renewal of hematopoietic stem cells has been linked to elimination of metabolically active mitochondria thereby maintaining HSCs in a glycolytic state with low levels of oxidative metabolism [76–78]. The balance between glycolysis and oxidative metabolism is key to determining whether stem cells remain quiescent or undergo differentiation [74, 79–81]. By limiting mitochondrial mass through mitophagy and stimulating glycolysis while limiting oxidative metabolism, stem cells can maintain their slow cycling, self-renewing state (figure 3) [80]. By contrast, inhibition of glycolysis and enhancement of mitochondrial respiration promotes differentiation that involves mitochondrial remodeling, reduced mitophagy, dispersed cytoplasmic localization, and increased expression of enzymes involved in the TCA cycle and the electron transport chain [79, 81–84].

Figure 3. Requirement for autophagy in stem cell maintenance.

Autophagy is required for maintenance of a stem-like phenotype both in normal tissue stem cells but also in certain types of cancer stem cells. Stresses (hypoxia, TGF-β) that induce stem-like properties, including self-renewal capability, EMT, up-regulation of CD44, also induce autophagy that is required for many of these stem-like features, including up-regulation of CD44. Recent work has highlighted a specific role for mitophagy in promoting a stem-like state in cancers, possibly by limiting ROS production and promoting glycolytic metabolism that may limit differentiation.

Cancer stem cells (CSCs) are defined functionally as those relatively rare cells in a tumor that have the capacity to self-renew and to differentiate to regenerate all aspects of tumor heterogeneity [23, 85, 86]. Various reports have correlated the expression of CSC markers, such as CD44, with increased invasiveness and metastasis [87–91]. Indeed, stresses such as hypoxia and TGF-β that promote epithelial to mesenchymal transition (EMT) and a more motile tumor cell phenotype, also induce features of CSCs, including increased self-renewal and upregulation of CD44 (figure 3) [34, 85, 86, 92–94]. Significantly, stresses such as hypoxia and TGF-β also induce autophagy, alongside EMT and stemness [32, 95] and numerous studies have now linked increased autophagy to maintenance of CSCs (figure 3) [29–32, 96]. In human breast ductal carcinoma in situ (DCIS), increased autophagy was detected in those rare tumor cells that showed tumor-initiating capacity and migratory ability and CD44+ CSCs were dependent on autophagic flux for their ability to form mammospheres in vitro and tumors in vivo [29, 96–98]. In esophageal squamous cell carcinoma (ESCC), autophagy was required for the upregulation of CD44 that accompanied induction of EMT (figure 3) [30]. Here again, it was mitophagy that was specifically up-regulated in ESCC cells undergoing EMT [30], possibly in response to increased ROS production and membrane depolarization that is known to activate Parkin-dependent mitophagy [99]. Inhibition of Parkin-mediated mitophagy attenuated up-regulation of CD44 and led to cell death [30]. Similar to what has been described for pluripotent stem cells, reduced mitochondrial mass has also been reported to distinguish CSCs from non-CSCs (figure 3) [100, 101]. However, quiescent PDAC cells with CSC properties were recently described as relying on oxidative respiration to maintain their stem cell state and were less dependent on glycolysis than proliferating cells in the bulk of the tumor [102]. Clearly, the role of mitophagy in CSCs needs to be addressed more directly to determine whether reduced mitochondrial mass arising from increased mitophagy in response to EMT, cancer therapy or other challenges, is indeed required to establish and/or maintain tumor stem cell properties.

The “re-awakening” of tumor cells at distant sites leading to outgrowth of macrometastatic disease many years after primary tumors were “successfully” treated has led to the concept of metastatic dormancy [22, 23, 103, 104]. Various studies have shown a role for autophagy in promoting tumor cell survival during tumor dormancy [105]. For example, inhibition of autophagy blocked re-emergence of ovarian tumors after dormancy induced by the ARHI tumor suppressor [105]. Pre-clinical studies also showed that autophagy inhibition in the Eμ-Myc mouse model of B-cell lymphoma following treatment with alkylating agents blocked tumor recurrence validating a role for autophagy in promoting both tumor dormancy and drug resistance [106]. Autophagy may promote the dormancy of disseminated tumor cells simply by supplying key amino acids and other nutrients or alternatively, as discussed above, autophagy may play a more instructive role by eliminating mitochondria, modulating redox balance and actively promoting the stem cell state [22, 107, 108]. Indeed, CSC markers are up-regulated on disseminated tumor cells in the bone marrow of breast cancer patients [109] and such observations have led to the suggestion that dormant tumor cells are in fact CSCs that depend on autophagy to survive at secondary sites over extended periods of time and expand later as metastatic lesions made up of both CSCs and non-stem tumor cells representing the full heterogeneity of rapidly growing tumors [22, 105]. If as suggested, dormant tumor cells are preferentially reliant on autophagy for survival, then this provides a rationale for combining autophagy inhibition with conventional therapeutic responses to eliminate dormant tumor cells and prevent subsequent metastatic outgrowth.

Autophagy contributes to the therapy resistance of numerous cancers, including resistance to conventional genotoxic therapies [110, 111], prostate cancer to androgen ablation therapy [112], breast cancer to SERMs [113–115], GIST to Imatinib [116, 117], lung cancer to tyrosine kinase inhibitors [118–120], glioblastoma to temozolomide [121], myeloma to bortezomib [122, 123], melanoma and brain cancers to B-Raf inhibitors [124–126], as well as resistance of various cancers to PI3K inhibitors [127]. Thus, there is a compelling rationale to combine these therapeutic approaches with agents that inhibit autophagy, such as chloroquine, an anti-malarial drug that inhibits autophagy by increasing the pH of the lysosome and blocking lysosomal proteases [128, 129]. Early clinical trials combining chloroquine with more conventional therapies, such as temozolomide for glioblastoma treatment showed promise with autophagy inhibition more than doubling survival times [128, 130, 131]. One particularly successful trial applying chloroquine to conventional doxorubicin treatment of non-Hodgkins lymphoma in dogs demonstrated both that effective chloroquine doses were well-tolerated but also improved overall drug response and progression-free survival [132]. More recent studies have reported efficacy of adjuvant chloroquine treatment in clinical trials for an extended range of human cancers, as reviewed in more detail elsewhere [128, 129].

However, the ability to target autophagy in cancer has also been limited by the high doses of agents such as chloroquine required to elicit an effect in humans [128, 129, 133], the unpredictable effects of lysosomal inhibitors used to block autophagy on the activity of mTORC1 that is regulated at the lysosome also [134, 135], the unknown off-target effects of compounds such as chloroquine [128] and the possible adverse consequences of systemic autophagy inhibition [136]. An additional hurdle to assessing the efficacy of autophagy inhibition in cancer treatment is the lack of reliable markers of autophagy in human tumors [129]. Nevertheless, the development of improved derivatives of chloroquine that are efficacious at lower doses [133] and simultaneously target both autophagy and mTORC1 activity at the lysosome [137], in addition to development of targeted inhibitors of ULK1 [138, 139], VPS34 [140] and other enzymes required for autophagy [141] holds significant promise for our ability to overcome the tumor-promoting effects of autophagy in cancer therapy resistance and metastatic dormancy [12].

Autophagy in the tumor microenvironment

Cancer cells co-evolve with their tumor microenvironment and the role of autophagy in modulating how the cancer cell interacts with other cell types in the surrounding milieu is emerging as a key topic in determining whether autophagy inhibition is likely to be effective in cancer treatment [129]. How autophagy modulates the interaction of tumor cells with components of both the innate and adaptive immune systems is particularly complex, as has been reviewed extensively elsewhere [142, 143]. Autophagy promotes the release of damage-associated molecular patterns (DAMPs) and ATP from dying tumor cells thereby recruiting CD8+ cytotoxic T lymphocytes (CTLs) that synergize with conventional therapeutics to eliminate cancers [144, 145]. Autophagy also promotes tumor immune surveillance through trafficking of tumor-antigens through the lysosome for cross-presentation on dendritic cells and by suppressing infiltration of tumor-promoting FoxP3+ T-regulatory cells [145, 146]. These anti-tumor activities of autophagy in eliciting an immune response would argue that inhibiting autophagy is not justified in in combination with conventional cancer therapy. However, the role of autophagy in modulating tumor immunity may be context dependent with other studies using less immunogenic tumor models indicating that anti-tumor immunity is not adversely affected by autophagy inhibition [147] and that in fact autophagy promotes escape from immune surveillance in some models [148, 149] possibly through indirect mechanisms relying on the role of autophagy in promoting EMT and secretion of immune-modulatory cytokines [150]. A recent study has identified a novel role for autophagy in suppressing the anti-tumor activity of certain types macrophages in a mouse model of pancreatic ductal carcinoma (PDAC) [151]. Inhibiting autophagy through inducible expression of a dominant negative Atg4B allele (Atg4B^CA) decreased PDAC tumor growth and induced tumor regression of already formed tumors in an autochtonous model of PDAC [151]. However, when autophagy was similarly inhibited in an orthotopic transplant model using primary tumor cells from the autochtonous model, tumor growth was inhibited but tumor regression of existing tumors was not observed [151]. Inhibiting autophagy by expression of Atg4B^CA caused macrophage accumulation in autochtonous PDAC tumors while depletion of macrophages prevented tumor regression induced by autophagy inhibition without affecting rates of tumor cell growth. These findings indicate strongly that autophagy plays both a cell intrinsic role promoting tumor cell growth but also has a non-cell autonomous effect suppressing the anti-tumor activities of macrophages [151]. It was not determined how autophagy inhibition resulted in macrophage recruitment to the tumor although production of DAMPs by dying tumor cells as a result of autophagy inhibition could provide an explanation. As reviewed elsewhere [142] [143], further investigation is required to clarify when and how autophagy modulates the interaction of tumor cells with immune cells in the tumor microenvironment.

Several recent reports have illuminated how autophagy can be induced in certain other non-tumor cells that make up the tumor microenvironment to promote tumor cell growth and invasion [18, 152]. Examination of the interaction between pancreatic ductal adenocarcinoma (PDAC) cells and pancreatic stellate cells (PSCs), a specialized type of fibroblast that makes up a large part of the desmoplastic microenvironment in PDAC, revealed that autophagy was up-regulated specifically in PSCs when grown with pancreatic tumor cells but not when co-cultured with normal ductal epithelia. This in turn promoted protein catabolism in PSCs generating alanine to fuel the TCA cycle, oxygen consumption and lipid synthesis in the adjacent tumor cells (figure 4) [18]. The effects of increased alanine uptake by tumor cells and its conversion to pyruvate to fuel TCA cycle and lipid synthesis allowed glucose-derived carbon to be more efficiently used to promote serine and glycine biosynthesis with knock-on effects for nucleotide production. These effects of co-culture of tumor cells with PSCs or PSC-conditioned media was blocked by inhibiting autophagy in PSCs or by knocking down the GPT1 alanine transaminase (that converts alanine to pyruvate) in tumor cells. The growth stimulatory effects of PSCs on PDAC growth were observed both in vitro and in vivo in both subcutaneous and orthotopic mouse models using human PDAC cells and both human and mouse PSCs [18]. Consistent with a role in promoting tumor growth, there was selection against expression of the autophagy inhibitory activity of Atg4BCA in PSCs in an autochtonous mouse model of PDAC [151].

Figure 4. Autophagy in the tumor microenvironment supports tumor cell growth through supply of nutrients and other factors.

The growth of pancreatic ductal adenocarcinoma (PDAC) is supported by a highly complex and dense desmoplastic stroma that includes pancreatic stellate cells (PSCs). Autophagy plays a critical role in the tumor-promoting activity of PSCs that supply associated PDAC tumor cells with the breakdown products of autophagy, including amino acids such as alanine. Inhibition of alanine metabolism in tumor cells neutralized the pro-tumor activity of PSCs. Autophagy is induced in PSCs by their association with PDAC cells although how tumor cells signal to PSCs to induce autophagy is not yet clear. Autophagy in PSCs was also required for production of pro-inflammatory cytokines such as IL-6, that may be feeding back to both promote tumor cell motility and mobilization of metabolites systemically in the animal, but also may also contribute to the signal that induces autophagy in PSCs in the first place. Additional metabolites and factors may also contribute to additional aspects of how autophagy in the tumor microenvironment promotes tumor growth and metastasis.

Complementary work highlighted the role played by autophagy in activating PSCs to produce pro-migratory and invasive cytokines such as IL-6, in addition to extracellular matrix proteins [153]. Conversely, inhibition of autophagy in PSCs resulted in adoption of a quiescent state, reduced expression of IL-6 and ECM proteins, attenuated migration/invasion properties and reduced metastasis to the liver of co-injected PDAC cells in an orthotopic mouse model (figure 4) [153]. Autophagy is known to promote secretion of IL-6 and other factors [46, 154, 155] and thus inhibiting autophagy as a means to limit tumor cell migration and invasion, in addition to tumor cell growth is well-justified.

In pancreatic cancer, the desmoplastic reaction that is dependent on PSC activation is a barrier to therapy both by inducing hypoxia and stem-like phenotypes in the tumor cells, and by physically preventing access of the vasculature and drugs to the tumor [156]. Efforts to improve delivery of drugs to PDAC tumors, by degrading the desmoplastic “fortress” around these tumors [157], have failed and this has been attributed to selection for more aggressive tumor cells and increased vasculature leading to easier egress of Tregs, reduced tumor immune surveillance and easier escape of more aggressive mesenchymal tumor cells into the periphery [158, 159]. Thus efforts to limit autophagy in PSCs may provide a more promising alternative in PDAC treatment since it would both limit the ability of PSCs to supply tumor cells with nutrients such as alanine [18], as well as pro-invasive factors, such as IL-6 [153] but importantly autophagy inhibition would also be predicted to limit acquisition of stem-like, drug-resistant features and other malignant properties by the tumor itself, as described above (figure 3).

Genetic analysis of the role of autophagy in tumor growth in Drosophila has further emphasized the significance of autophagy in the tumor microenvironment [152]. Here, it was shown that autophagy is up-regulated systemically in flies bearing malignant Ras^V12;scrib−/− tumors in their eye-antennal imaginal discs but not with benign Ras^V12 tumors [152]. Ablation of autophagy in the tumor had only a modest effect on tumor growth but inhibition of autophagy in either the non-tumor epithelia surrounding the tumor or in the entire animal had a marked effect on both tumor growth and invasiveness [152]. Indeed, transplant of quiescent Ras^V12;scrib−/−;Atg13−/− autophagy deficient tumors into autophagy competent host flies could rescue tumor growth while transplant into autophagy deficient host flies could not [152]. In this model, induction of autophagy in the tumor microenvironment, which was referred to as non-cell-autonomous autophagy (NAA) was independent of tumor growth since dominant negative PI3K blocked tumor cell growth but the tumor was still able to induce autophagy in neighboring non-tumor cells. Intriguingly, the ability of the tumor to induce NAA was dependent on tumor-specific activation of TNF, JAK/STAT, JNK and Hippo pathway signaling via Yorkie [152] with Yorkie transcriptional targets upd1 and upd3 (both IL-6 like cytokines in flies) able to induce NAA [152]. The Yorkie target ImpL2, an insulin/IGF antagonist secreted by over-proliferating tissues was previously shown to cause systemic tissue wasting in flies [160, 161] but was not required to induce NAA in flies with Ras^V12;scrib−/− tumors [152]. In line with the work from Kimmelman and colleagues [18], silencing of the amino acid transporter slimfast in tumors, reduced tumor growth although uptake of a specific amino acid or other possible mediators of the non-cell-autonomous/autophagy-dependent signal feeding tumor cells was not identified [152].

These various studies collectively demonstrate that malignant tumors induce autophagy in non-tumor cells making up their microenvironment in a manner that further promotes tumor growth and progression (figure 4). Intriguingly, the fly studies suggest that tumor growth can induce autophagy systemically in tissues throughout the animal and it remains to be determined which factors drive this phenomenon but likely signals that circulate, again perhaps IL-6. Critically, these reports demonstrate a key role for autophagy in non-tumor cells in producing metabolites that fuel tumor cell growth. In the case of mammalian pancreas cancer studies, this appears to be driven by autophagy-dependent production of alanine from associated non-tumor cells that is taken up by the tumor and fed into the TCA cycle and lipid biosynthetic pathways amongst other [18] with a similar mechanism identified in flies where slimfast was required for the positive effects of NAA on tumor growth [152]. Production of alanine by cancer-associated fibroblasts in PDAC predominantly affected tumor growth rate [18] but autophagy in non-tumor cells also drove tumor cell invasion and metastasis [152, 153], perhaps through production of pro-migratory cytokines, such as IL-6.

These studies open up new perspectives on the role of autophagy in cancer progression and metastasis. Particularly intriguing is the possibility that diet and metabolic stresses inducing autophagy in distant organs could influence tumor growth and metastasis at another site or indeed that tumor growth may influence systemic metabolism by promoting autophagy in key organs, such as the liver. Also, there is the possibility that other metabolites in addition to alanine, produced by protein catabolism during autophagy in non-tumor tissue may play a signaling or metabolic role in promoting tumor progression and metastasis. These are all avenues of future investigation.

Abbreviations

AMPK

adenosine monophosphate activated kinase

ATG/Atg

human/mouse autophagy gene

ATF4

activating transcription factor 4

ATP

adenosine triphosphate

CBL

Casitas B-lineage lymphoma gene

CD44

cell differentiation antigen 44

CSC

cancer stem cell

CTL

cytotoxic T lymphocyte

DAMP

damage associated molecular pattern

DCIS

ductal carcinoma in situ

ECM

extracellular matrix

EMT

epithelial-mesenchymal transition

ER

endoplasmic reticulum

FAK

focal adhesion kinase

FIP200

FAK interacting protein 200 kD

GPT1

glutamate-pyruvate transaminase-1

HSC

hematopoietic stem cell

IL-6

interleukin 6

JAK

janus kinase

JNK

c-Jun N-terminal kinase

LC3

microtubule associated light chain 3

MIT/TFE

microphthalmia-associated transcription factor/transcription factor E3

PDAC

pancreatic ductal carcinoma

PSC

pancreatic stellate cell

PXN

paxillin

ROS

reactive oxygen species

SERM

selective estrogen receptor modulator

SRC

sarcoma oncoprotein encoded by c-SRC

STAT

signal transducers and activators of transcription

TCA

tricarboxylic acid

TGF-β

transforming growth factor beta

TNF

tumor necrosis factor

mTOR

mammalian target of rapamycin

mTORC1

mammalian target of rapamycin complex 1

ULK1

Unc51-like kinase-1

VPS34

Vacuolar protein sorting-associated protein 34