Abstract

Core autophagy genes are generally not mutated in cancer

Early evidence suggested that BECN1 (ATG6) was a tumor suppressor gene. This was based on reports that one allele of BECN1 was lost with high frequency in breast, ovarian, and prostate cancer; that BECN1 overexpression suppressed human breast cancer cell line growth; and that mice with allelic loss of Becn1 are tumor-prone (Aita et al. 1999; Liang et al. 1999; Qu et al. 2003; Yue et al. 2003). However, BECN1 is adjacent to the known tumor suppressor gene breast cancer 1 (BRCA1) on chromosome 17.

Hereditary breast cancer commonly results from the presence of a pathogenic germline missense mutation in BRCA1 (or BRCA2 or PALB2) followed by somatic deletion of the remaining wild-type BRCA1 allele. These deletions are typically large, deleting BRCA1 along with hundreds of other genes, including BECN1. Genomic analysis of BECN1 in The Cancer Genome Atlas (TCGA) demonstrated that allelic loss of BECN1 does not occur independently of codeletion with BRCA1, suggesting that BRCA1 loss is the driver mutation in hereditary and sporadic breast cancer (Laddha et al. 2014; White et al. 2015). Moreover, recurrent somatic mutations that are a hallmark feature of tumor suppressor genes are not found in BECN1 in human cancers (Aita et al. 1999; Laddha et al. 2014). Less BECN1 expression resulting from concurrent BECN1 and BRCA1 deletion (Laddha et al. 2014; Tang et al. 2015), however, may reduce autophagy in these cancers, which may impair their response to stress and nutrient deprivation that may be therapeutically exploitable.

Insight into the role of autophagy in human cancer is also provided by a more global analysis of the mutational status of the genes encoding the entire core autophagy machinery in TCGA. The core autophagy machinery is generally not targeted by high-frequency somatic single-nucleotide mutations across cancers in TCGA (Lebovitz et al. 2015). Although a small number of recurrent mutations in RB1CC1/FIP200, ULK4, and ATG7 are found, these are rare and limited to few tumor types. These large-scale analyses indicate that the core autophagy machinery largely escapes genomic mutation in human cancers. It will be interesting to explore the significance of the minority of tumors in which these rare mutations are found.

Analysis of the human cancer transcriptomes indicates largely invariable ATG gene expression (Lebovitz et al. 2015). This suggests a pattern of sustained transcriptional expression similar to genes unable to tolerate reduced expression in cancer. As such, the core autophagy machinery is expressed and may be protected from alteration in the majority of human cancers (Lebovitz et al. 2015). It is also interesting to note that the microphthalmia/TFE (MiTF/TFE) subfamily of basic helix–loop–helix leucine zipper transcription factors that promote autophagy and lysosomal biogenesis is encoded by oncogenes (Ferguson 2015), which may contribute to expression of the autophagy pathway in human cancers. Thus, the majority of human cancers expresses a functionally intact autophagy pathway, and, in a subset of these, the pathway is transcriptionally up-regulated.

Regulation of core autophagy genes and cargo receptors in cancer

The first complex known to regulate autophagy is the ULK1–ATG13–FIP200 complex (Ganley et al. 2009). ATG13 and ULK1 are phosphorylated and inhibited by mTOR in a nutrient-dependent fashion to regulate autophagy initiation. Recent work revealed a direct interaction between focal adhesion (FA) kinase (FAK) family-interacting protein of 200 kDa (FIP200) and ATG16L1 (a component of the ATG12–ATG5–ATG16L1 complex), which is crucial for ULK1 complex-dependent autophagy independent of ULK1 kinase activity (Gammoh et al. 2013). Intriguingly, amino acid starvation can engage a protein phosphatase (PP2A-B55α) to efficiently dephosphorylate and activate the ULK1 complex. This finding explains a puzzling observation in the field: Nutrient starvation triggers autophagy more rapidly than rapamycin despite the fact that both nutrient starvation and rapamycin suppress mTORC1 activity equally well. It also reveals one possible mechanism accounting for the high basal autophagy activity in “autophagy-addicted” cancer cells. The seemingly contradictory coexistence of intact mTORC1 function and strong autophagy activity in cancer cells may be due to the expression of phosphatase activity targeting ULK1, which offsets the autophagy inhibitory function of mTORC1, allowing cancer cells to simultaneously reap the benefits of both mTOR signaling and autophagy activation (Wong et al. 2015).

Independent of its well-characterized role in autophagy, FIP200 interacts with other proteins to regulate diverse cellular functions (Gan and Guan 2008). To address potential contributions of its autophagy-dependent and autophagy-independent action in cancer, a Fip200 conditional knock-in mutant allele (cKI) was generated that specifically disrupts FIP200 interaction with its autophagy partner, ATG13, to block autophagy but not its binding to other cell signaling molecules. Changing residues 582–585 (LQFL) to AAAA (FIP200-4A mutant) in FIP200 demonstrates that this region is required for interaction with ATG13 and abolishes its canonical autophagy function in vitro. Specifically blocking FIP200 interaction with ATG13 using FIP200-4A mutant knock-in mice prevents autophagy in vivo, providing direct support for the essential role of the ULK1/ATG13/FIP200/ATG101 complex in the process beyond previous studies that knock out individual components. The nonautophagic functions of FIP200 (for instance, promoting resistance to TNF-α-induced apoptosis) are retained in this cKI model, as embryogenesis is unaffected, unlike FIP200 knockout mice. However, FIP200-mediated canonical autophagy is required to support neonatal survival and tumor cell growth (Chen et al. 2016).

Selective autophagy provides a means to control the turnover of specific proteins and organelles in a signal-dependent manner. However, our understanding of these systems and how cargo is selected is rudimentary. Recently, progress has been made in elucidating the identity and regulation of cargo receptors for degradation of ferritin (ferritinophagy) and damaged mitochondria (mitophagy). While mitophagy employs a ubiquitin signal on damaged mitochondria to link with the autophagy cargo receptors OPTN and NDP52, ferritin is recognized by a new adaptor (NCOA4), which localizes to autophagosomes and is required for delivery of ferritin to the lysosome, where iron is released for use by the cell (Dowdle et al. 2014; Mancias et al. 2014). The expression of NCOA4 itself is regulated by iron. In a high-iron milieu, NCOA4 associates with the HERC2 ubiquitin ligase and is degraded, while, in a low-iron milieu, NCOA4 accumulates and delivers ferritin to the autophagosome. NCOA4 is required for erythropoiesis in vertebrates (Mancias et al. 2015). Future work will uncover the role of other cargo receptors and the six ATG8 proteins in the control of selective autophagy. Additionally, with the realization that the regulated expression of specialized cargo receptors allows the cell to promote one type of selective autophagy over another (Green and Levine 2014), it is also timely to consider how one can exploit inactivation of such selective adaptor proteins to parse out the different roles of autophagy in cancer to determine whether specific types of autophagy play more significant roles than others in tumorigenesis (Mancias and Kimmelman 2016).

The autophagy cargo receptor p62/SQSTM1 binds ubiquitin on cargo to deliver cargo to autophagosomes by docking onto microtubule-associated protein 1A/1B light chain 3 (LC3) on autophagosomes. P62 itself is an autophagy substrate that accumulates when autophagy is inhibited (Katsuragi et al. 2015). p62 accumulation induced by autophagy inactivation contributes to the development of benign hepatomas in mouse models, but the underlying mechanism is not known (Komatsu et al. 2007; Takamura et al. 2011). In addition to its role as a cargo adapter, p62 also serves as a signaling adaptor for TRAF6, KEAP1, and components of the mTORC1 complex, promoting survival and proliferative functions, possibly explaining the tumor promotion observed upon its accumulation. In support of this concept, accumulation of p62 has been observed in premalignant liver disease, and high expression of p62 in hepatocellular carcinoma (HCC) predicts poor prognosis (Umemura et al. 2016). In multiple physiologically relevant HCC mouse models, p62 is necessary for the activation of mTORC1 in HCCs driven by TSC2 deficiency (Umemura et al. 2016). A role for p62 in promoting NFR2 activation and HCC has also been identified (Saito et al. 2016). Adenovirus-mediated overexpression of p62 in hepatocytes in vivo in the absence of any other oncogenic stimuli drives HCC through the activation of mTORC1 in a manner that is independent of p62's role as a cargo receptor in autophagy. These results establish for the first time that the ability of p62 to activate mTORC1 (along with NRF2) in hepatocytes in vivo is a new critical driver of liver cancer initiation (Umemura et al. 2016). In contrast, reduction of p62 in stroma promotes tumorigenesis of prostate cancer through interleukin-6 (IL6) production (Valencia et al. 2014). Roles of other cargo receptors in cancer have also been reported. Sequestration of the cargo receptor NDP52 promotes noncanonical nuclear factor κB (NF-κB) signaling in cancer cells (Newman et al. 2012), and ubiquitylation of the cargo receptor OPTN increases autophagy and p62 elimination, suppressing tumorigenesis (Liu et al. 2014). Thus, the cargo receptors whose levels are controlled by autophagy can play a role in cancer.

Given the recent appreciation for the importance of mitochondria in cancer (Zong et al. 2016), a selective form of autophagy of particular interest is that of mitophagy, which removes damaged mitochondria. Emerging findings indicate that there are multiple mechanisms by which mitochondria can be targeted for degradation at the autophagolysosome (Chourasia et al. 2015a). Elimination of depolarized mitochondria by mitophagy depends on the action of the PINK1 kinase and its ability to recruit the PARKIN E3 ubiquitin ligase that promotes ubiquitination of key mitochondrial proteins, which serve as “eat me” signals to cargo receptors targeted by nascent autophagosomes (Youle and Narendra 2011; Lazarou et al. 2015). Not all mammalian cells express high levels of PARKIN; therefore, there is an ongoing search for other mitophagy regulators, especially in cancer cells. Additional PARKIN-independent mitophagy adaptors include BCL2/adenovirus E1B 19-kDa-interacting protein 3 (BNIP3), BNIP3L/NIX, and FUN14 domain-containing protein 1 (FUNDC1) that are induced by specific developmental signals and physiological stresses (for example, hypoxia, nutrient deprivation, loss of the tumor suppressors p53 or pRB, and activation of NF-κB or FoxO transcription factors). These mitophagy adaptors interact directly with processed LC3 and promote an overall reduction in mitochondrial mass as an adaptive response to stresses (Novak et al. 2010; Hanna et al. 2012; Liu et al. 2012; Chourasia et al. 2015a). BNIP3, a BH3 domain-containing protein that, in some contexts, is considered a proapoptotic protein, likely plays a more important role in mitophagy than in the regulation of apoptosis during nutrient stress. BNIP3 is induced by nutrient deprivation in the liver, where it plays a role in promoting mitophagy and lipid oxidation (Glick et al. 2012). BNIP3 expression is induced during the early stages of tumorigenesis but is lost in advanced cancer (Sowter et al. 2001, 2003; Okami et al. 2004; Abe et al. 2005; Akada et al. 2005; Erkan et al. 2005). Therefore, BNIP3 acts as a tumor suppressor in mouse models (Manka et al. 2005; Chourasia et al. 2015b). Loss of BNIP3 is observed in triple-negative breast cancer (Koop et al. 2009; van Diest et al. 2010; Chourasia et al. 2015b), and mice carrying a targeted deletion of BNIP3 sustain increased rates of tumor growth, acquire invasiveness, and increase spontaneous metastasis (Chourasia et al. 2015b). Tumors in these mice have increased mitochondrial mass but reduced mitochondrial respiration and metabolite uptake. Increased production of reactive oxygen species (ROS) leads to increased hypoxia-inducible factor 1α (HIF-1α) levels and activity, glycolytic metabolism, and angiogenesis (Chourasia et al. 2015b). Thus, BNIP3, a HIF-1α target, may act as a tumor suppressor downstream from HIF-1α, feeding back to promote mitochondrial integrity, reduce ROS production, and limit the protumorigenic activity of HIF-1α.

Interestingly, tumor suppressor activity has also been attributed to PARKIN, which is deleted in human ovarian, breast, bladder, and lung cancers among others (Cesari et al. 2003), and this activity is associated with promoting oxidative metabolism (Zhang et al. 2011). Thus, in contrast to inhibition of general autophagy that blocks tumor progression (Wei et al. 2011; Yang et al. 2011; Guo et al. 2013a; Rosenfeldt et al. 2013; Strohecker et al. 2013), genetic deletion of two different mitophagy modulators (PARKIN and BNIP3) leads to increased tumorigenesis and accelerated progression to malignancy (Zhang et al. 2011; Chourasia et al. 2015b). There are a number of major differences between loss of canonical autophagy genes and inactivation of specific mitophagy pathways. First, inactivation of BNIP3 does not eliminate all forms of mitophagy and simply reduces stress-induced mitophagy. Perhaps, if all forms of mitophagy are inactivated, this would have a different phenotypic consequence more similar to inhibition of canonical autophagy genes. Note that autophagy supplies substrates for mitochondrial metabolism, the loss of which significantly impairs mitochondrial function beyond what would be expected for loss of mitophagy alone (Guo et al. 2016). In addition, genetic autophagy inhibition produces an accumulation of oncogenic p62, whereas genetic mitophagy inhibition does not, further distinguishing the function of autophagy from mitophagy.

Evidence from mouse models that autophagy suppresses cancer initiation

A large body of work in genetically engineered mice deficient for essential autophagy genes in specific tissues has demonstrated that autophagy prevents an array of degenerative and inflammatory diseases (Mizushima and Komatsu 2011). The mechanisms by which autophagy is thought to maintain tissue homeostasis include the removal of toxic, unfolded, or mutant proteins and the elimination of damaged mitochondria and other organelles. Connections between metabolic defects and degenerative and inflammatory disease caused by loss of autophagy remain to be explored. Arguably, the disease in which the functional status of autophagy has been investigated the most is cancer.

Despite the predominant absence of genetic inactivation of the core autophagy machinery in human cancers thus far, animal models in which autophagy loss promotes tumorigenesis have been identified. Becn1^+/− mice are prone to lung and liver tumors and mammary hyperplasia late in life (Qu et al. 2003; Yue et al. 2003). It is important to note that, in Becn1^+/− cells, the expression levels of the tumor suppressor p53 are also significantly reduced, again suggesting that there may be autophagy-independent reasons for the unique ability of the Becn1^+/− genotype to produce spontaneous tumors compared with other genetic autophagy defects (Liu et al. 2011). For instance, mosaic deletion of Atg5 in mice or Atg7 in mouse livers produces only benign hepatomas, suggesting that complete and specific autophagy deficiency promotes liver tumor initiation but restricts progression to malignant disease (Takamura et al. 2011). Similarly, loss of Atg5 or Atg7 in the mouse pancreas in the setting of a Kras mutation promotes benign pancreatic intraepithelial neoplasia (PanIN) formation but also prevents progression of this PanIN to malignant disease (Rosenfeld et al. 2014; Yang et al. 2014). Autophagy loss in the liver and pancreas causes oxidative stress, chronic tissue damage, and inflammation, all factors that are well known to promote tumor initiation (Mathew et al. 2009). This chronic state of tissue damage, cell death, and inflammation promotes compensatory proliferation, producing cells vulnerable to mutation and oncogenic transformation (White 2012). The situation may be similar in a mouse model of breast cancer in which allelic loss of becn1 promotes expansion of the stem cell pool susceptible to transformation by Wnt pathway activation (Cicchini et al. 2014). However as mentioned above, this may be due to autophagy-independent mechanisms as well (Cicchini et al. 2014).

Another example relating chronic inflammation to loss of autophagy is in Crohn's disease (CD) caused by the ATG16L1 Thr300Ala disease-associated polymorphism. CD is associated with chronic inflammation of the colon and intestine. Patients with CD are also predisposed to the development of cancer in these areas of inflammation. Atg16L1^T300A knock-in mice manifest the Paneth cell abnormalities associated with CD and also display defects in Goblet cells. Embryonic fibroblasts from these mice surprisingly have only a modest impairment of basal and induced autophagy but rather have a larger defect in autophagy of bacteria and increased production of IL-1β (Lassen et al. 2014). This suggests that the primary problem created by the ATG16L1 Thr300Ala disease-associated polymorphism is one of impaired anti-bacterial host defense, which in turn leads to chronic inflammation, tissue damage, and an elevated cancer risk.

As mentioned above, accumulation of the autophagy cargo receptor and substrate p62 in autophagy-deficient livers promotes tumorigenesis (Komatsu et al. 2007). Accumulation of p62 also promotes lung and mammary tumorigenesis induced by autophagy deficiency (Moscat and Diaz-Meco 2012; Wei et al. 2014). p62 activates NRF2 by binding to its inhibitor, KEAP1 (Komatsu et al. 2010; Lau et al. 2013), and also activates mTOR and NF-κB as discussed above, and its aberrant accumulation in autophagy-deficient tissues can activate these oncogenic signaling pathways. p62 deficiency blunts mammary and lung tumorigenesis and liver tumorigenesis induced by Atg7 loss (Moscat and Diaz-Meco 2012; Wei et al. 2014). In Braf^V600E-driven lung tumors, autophagy deficiency transiently stimulates tumor growth by enhancing activation of NRF2, which may be due to accumulation of p62 (Strohecker et al. 2013). Later, these tumors without Atg7 slow their growth and fail to transition from benign to malignant tumors.

The recurring theme from mouse models in which autophagy deficiency promotes tumorigenesis is that autophagy (1) prevents oxidative stress, activation of the DNA damage response, chronic tissue damage, and inflammation and (2) prevents accumulation of the autophagy substrate p62 and promotes activation of mTOR and NRF2. These events are a source of genomic instability and oncogenic signal transduction that promote cancer initiation (Karantza-Wadsworth et al. 2007; Mathew et al. 2007, 2009) commonly found in states of chronic inflammation induced by a broad array of stimuli. With the possible exception of the CD ATG16L1 disease-associated polymorphism, the question of whether this mode of tumor promotion upon suppression of autophagy occurs in humans has not been reported. Indeed, humans with loss of function of core ATG genes and autophagy, as is the situation in these mouse models, have not been identified. It is possible that these mechanisms come into play in humans only indirectly, through constitutive activation of mTOR, for example, which inhibits autophagy. Alternatively, autophagy loss may not be relevant to the vast majority of human cancers.

Evidence that autophagy promotes growth of advanced cancer

A first indication that autophagy may promote cancer came from the observations that autophagy is up-regulated in hypoxic tumor regions, suppresses tumor-induced inflammation, and promotes tumor cell survival (Degenhardt et al. 2006). As the functional status of autophagy is clearly influenced by the tumor microenvironment and the immune system, this revealed the importance of assessing autophagy in genetically engineered mouse models (GEMMs) of cancer. This approach has allowed investigation of whether or not tissue type or cancer genetics influence the role for autophagy in cancer. GEMMs have dissected the tumor cell-autonomous functions of autophagy and provided insight into a therapeutic window for autophagy inhibition in cancer therapy. A summary of the findings from these models is presented in Table 1.

Table 1.

Phenotypes of autophagy deficiency in GEMMs for cancer

Open in a separate window

Therapeutic window for autophagy inhibition

The GEMMs described above provide many examples where autophagy inactivation in tumor cells in vivo can suppress tumor initiation, growth, survival, maintenance, and malignancy. As autophagy is also important for the function of normal tissues, an important question is whether tumors are more autophagy-dependent than normal tissues to an extent sufficient to provide a therapeutic window. To address this question, autophagy was ablated in adult mice through systemic, conditional deletion of Atg7. In contrast to constitutive deletion of Atg7 or Atg5, which causes neonatal lethality following loss of placental nutrition from the wild-type mother (Kuma et al. 2004; Komatsu et al. 2005), adult mice survive Atg7 deficiency for 2–3 mo (Karsli-Uzunbas et al. 2014). Loss of adipose tissue occurs early after deletion of Atg7 in adult mice, possibly due to accumulation of mitochondria and increased FAO. Later, Atg7-deleted adult mice display liver, pancreas, muscle, and kidney damage and testicular degeneration, with the majority of the mice dying from neurodegeneration (Karsli-Uzunbas et al. 2014). Other tissues (e.g., lung) are surprisingly unaffected. Thus, adult mice are less autophagy-dependent than neonates, which may reflect differences in states of tissue differentiation and metabolic demand. Although adult mice are able to tolerate systemic Atg7 ablation in the short term, they are exquisitely sensitive to starvation, as fasting the Atg7-ablated mice causes rapid death due to hypoglycemia (Karsli-Uzunbas et al. 2014). Why some tissues (e.g., brain) are more autophagy-dependent than others (e.g., lung) is an interesting question.

With this understanding of the role of autophagy in adult mice, the next point to address is whether spontaneously arising tumors are more sensitive to conditional, systemic autophagy ablation than most normal tissues. Initiation of Kras-driven, p53-deficient lung tumorigenesis is surprisingly unaffected by the systemic absence of Atg7 in mice (Karsli-Uzunbas et al. 2014). Apparently, oncogenic signaling by Kras in the absence of p53 does not require autophagy to initiate tumor formation. In contrast, systemic Atg7 conditional deletion in mice with established Kras-driven, p53-deficient NSCLC (i.e., Atg7 deletion in both the host mouse and tumor) causes striking tumor regression (Karsli-Uzunbas et al. 2014). The anti-tumor activity of systemic, conditional Atg7 deletion is surprisingly greater than Atg7 deletion only in the tumor cells (Guo et al. 2013a), suggesting a role for host autophagy as well as tumor cell-autonomous autophagy in tumor maintenance (Karsli-Uzunbas et al. 2014).

In summary, these findings indicate that tumors can be more autophagy-dependent than most normal tissues and that there may be a therapeutic window for autophagy inhibition for cancer therapy and prevention. Note that complete, irreversible genetic ablation of autophagy provides the example of the most extreme phenotype unlikely to be replicated by small molecule autophagy inhibitors. Models for genetically switchable ATG deficiency will be informative to assess the reversibility of an autophagy-deficient phenotype in both normal and tumor tissues. To address potential toxicities, the lack of brain penetration of an autophagy inhibitor should limit toxicity, and fasting should be avoided.

New roles for autophagy identified in cancer progression

In addition to its established roles in tumor cell survival and metabolic adaptation, more recent work suggests that autophagy promotes cellular processes that favor invasion and metastasis. Recent work has begun to reveal that autophagy more directly supports tumor cell motility and invasion via cell-autonomous pathways. Autophagy promotes the dynamic assembly and disassembly of cell–matrix FAs, which is essential for efficient migration. Live-cell imaging revealed that genetic autophagy inhibition reduces FA assembly and disassembly rates, thereby increasing FA lifetime. Autophagosomes primarily associate with leading-edge FAs during disassembly, suggesting that autophagy locally and directly facilitates the degradation and destabilization of cell–matrix contact sites. Autophagy is required for tumor cell motility, as autophagy inhibition results in a block to cell migration and invasion in vitro and reduces metastasis in vivo (Sharifi et al. 2016). Autophagy promotes FA disassembly, mediated by the interaction of processed LC3 with paxillin, a key FA component, and this interaction was promoted by phosphorylation of paxillin by oncogenic SRC (Sharifi et al. 2016), suggesting that autophagy promotes tumor cell migration and metastasis (Macintosh et al. 2012; Lock et al. 2014; Qiang et al. 2014; Kenific et al. 2016). Furthermore, the selective autophagy cargo receptor neighbor of BRCA1 (NBR1) is required for efficient FA disassembly in motile cells, indicating that a selective autophagy process analogous to mitophagy enables the specific turnover of adhesions during migration (Kenific et al. 2016). Further work is needed to ascertain the role of NBR1 and other selective autophagy receptors during cancer invasion and metastasis in vivo.

In a three-dimensional organotypic culture model, autophagy regulators (ATGs) promote tumor cell motility and extracellular matrix invasion by Ras transformed epithelial cells. In further support, lung metastases are attenuated upon ATG knockdown in Ras transformed breast cancer xenografts. Notably, this invasion defect in autophagy-deficient cells is completely rescued upon treatment with conditioned medium from autophagy-competent counterparts, indicating that intact autophagy is required for the elaboration of secreted proinvasive factors. Autophagy-dependent secretion of the proinvasive cytokine IL6 is one of the critical factors for invasion (Lock et al. 2014). Measurement of circulating cytokines secreted by tumor cells in an autophagy-dependent manner could be useful to gauge autophagy levels within patient tumors (Kraya et al. 2015). These studies support the idea that autophagy promotes the coordinate secretion of cytokines favoring tumor cell invasion and broach the hypothesis that autophagy may direct non-cell-autonomous functions in tumor cells as well as associated stromal constituents by impacting secretory capacity.

Along these lines, autophagy has been shown to mediate a critical metabolic cross-talk in pancreatic cancer between the epithelial tumor cells and a specialized type of tumor-associated fibroblast known as a stellate cell. In this case, it is autophagy in the stellate cells that promotes the secretion of nonessential amino acids, likely through the autophagic degradation of protein (Sousa et al. 2016). Here, alanine has a critical role, as the tumor cells take up the secreted alanine and use its carbon to fuel the TCA cycle and biosynthesize fatty acids (Sousa et al. 2016). The utilization of alanine allows other carbon sources, such as glucose, to be used for biosynthetic functions such as the synthesis of serine, which is important in multiple cancer types to provide nucleotides for rapidly dividing cells. Importantly, inhibition of autophagy selectively in the stellate cells blocks this cross-compartmental metabolism and attenuates the promotion of tumor growth in both cotransplantation assays and in vitro studies using conditioned medium from stellate cells (Sousa et al. 2016). In a related fashion, several ATGs have been uncovered to promote the efficient biogenesis and egress of exosomes, further suggesting important functions for components of the core autophagy machinery in the tumor microenvironment (Murrow et al. 2015).

Autophagy's role in tumor immunity remains controversial. While early reports suggested that chemotherapy-induced immunity in highly immunogenic mouse cancers is autophagy-dependent (Michaud et al. 2011), concern was raised that this model did not mirror clinical experience (Amaravadi 2011). More recently, the interaction between T cells and natural killer cells and the tumor cells that they are engaged in killing suggests that tumor cell autophagy may serve to intercept perforin and granzyme effectors released by cytotoxic immune cells, blunting the efficacy of the anti-tumor immune response (Akalay et al. 2013; Baginska et al. 2013). Damage-associated molecular pattern molecules (DAMPS) such as high-mobility box group 1 (HMGB1), which are critical regulators of inflammation, also control autophagy levels within tumor cells (Zhu et al. 2015). Finally, autophagy was shown to be critical for the development of immunosuppressive T regulatory cells (Wei et al. 2016). Taken together, these findings indicate that the role of autophagy in anti-tumor immunity may be highly context-dependent and warrants further study.

Arguments against targeting the autophagy pathway in cancer

While the studies described above provide compelling evidence that inhibiting autophagy may have value for cancer therapy, arguments can be made opposing this concept. First, inhibiting autophagy may be too toxic. While this is always an issue with drug development, the GEMMs have identified the major toxicity concerns, which are not beyond what has been encountered with other anti-cancer drugs. Moreover, small molecule inhibitors targeting the autophagy pathway are not likely to be capable of irreversible, complete target inhibition, as is the case of genetic deletion of an ATG.

Second, human cancer cell lines can be unaffected by autophagy suppression (e.g., Maycotte et al. 2014; Mandelbaum et al. 2015). Human cancer cell lines often tolerate genome-editing approaches that delete Atg genes when grown in two-dimensional (2D) culture with replete culture medium (Eng et al. 2016), which is in contrast to the findings from spontaneously arising tumors. This is not likely an argument against targeting autophagy, as even cell lines derived from tumors deleted for ATGs that display an in vivo tumor growth defect can grow normally in the nutrient-rich conditions of 2D culture, although they are sensitive to starvation (Guo et al. 2013a; Strohecker et al. 2013). The high-nutrient conditions of in vitro cell culture conditions may be inappropriate for testing the function of a pathway responsive to nutrient stress and may miss key anti-cancer effects that would be evident in the in vivo microenvironment. There may also be clonal selection, adaptation, and development of resistance mechanisms that occur during generation of human cancer cell lines selected for deletion for Atgs in vitro.

Third, while there are numerous reports showing growth defects from autophagy inhibition in transplanted tumors in immunocompromised mice, these tend to be less impressive than what is seen in autochthonous models (Eng et al. 2016). Immunocompromised mice neglect the role of autophagy defects in activating the host immune system. Indeed, a common property of autophagy-deficient tumor cells is activation of innate and adaptive immune responses (Degenhardt et al. 2006; Guo et al. 2011; Mathew et al. 2014; Xie et al. 2015), which have been shown to contribute to anti-tumor activity (Wei et al. 2011; Levy et al. 2015). Moreover, the use of fast-growing tumors and a short time frame of analysis in nude mice is not a physiological assessment of tumorigenesis and impaired metabolic stress adaptation (Eng et al. 2016), although this is a reasonable approach to demonstrate dependency on oncogenic signaling (e.g., the effect of a BRAF inhibitor on BRAF mutant tumor cells).

Last, the anti-tumor activity produced by autophagy deficiency in the GEMMs has been proposed to be attributed to autophagy defects in both tumor and normal host tissues, and it has been put forth that it is host and not tumor cell-autonomous autophagy that promotes tumor growth (Nyfeler and Eng 2016). While the data do support non-tumor cell-autonomous roles for autophagy in tumor growth, there are extensive data that also support the importance of autophagy in the tumor cells themselves. Indeed, with a singular exception (Karsli-Uzunbas et al. 2014), the recombinase in the aforementioned autochthonous models delete the essential autophagy gene with activation of the oncogene and/or deletion of the tumor suppressor gene in the tumor cell compartment. There is no normal host tissue deficient for autophagy in these models. Tissue-specific recombinase expression also does not confer Atg deletion in fibroblasts, the major tumor stromal component. In the GEMMs for lung tumorigenesis, adenovirus delivery of the recombinase activates tumorigenesis with or without Atg deletion in isolated individual lung epithelial cells, resulting in tumors surrounded by normal stroma and lung tissue. While both host autophagy and tumor cell-autonomous autophagy contribute to tumor growth, survival, and malignancy, the interplay between them needs more in depth study and does not account for the vast majority of tumor growth defects due to Atg deletion in the GEMMs.

Potential therapeutic targets in the autophagy pathway

As autophagy has a demonstrated role in promoting tumor initiation, growth, survival, maintenance, malignancy, and metastasis in varying settings, therapeutic targeting of autophagy for cancer therapy may have value. Efforts in both the pharmaceutical industry and academia have focused on the development of small molecule inhibitors targeting the components of the autophagy pathway. Here we discuss several autophagy targets for small molecule inhibition.

Identifying potential responders to autophagy inhibition in patient populations

Unlike genetic mutations in oncogenes that help steer patients to specific ATP mimetic-based targeted therapy, specific mutated oncogenes identifying the tumors that would most benefit from autophagy inhibition are still evolving. Candidate biomarkers for potentially responsive tumors include those with abundant autophagosomes, hypoxic or nutrient stressed tumors, and tumors with activation of the MiT/TFE family. Based on preclinical work, Kras/p53 mutant PDAC, NSCLC, and melanoma are also a good starting point.

A defining feature of PDAC is its hypovascular and fibrotic microenvironment that promotes conditions of extreme hypoxia and nutrient deprivation (Ryan et al. 2014). PDAC cells adapt to life under limited nutrient conditions by relying on scavenging pathways for growth; namely, macropinocytosis and autophagy (Yang et al. 2011, 2014; Commisso et al. 2013; Perera and Bardeesy 2015). Importantly, both of these pathways converge on the lysosome for breakdown of scavenged material, and recent studies have shown that the lysosomal compartment is greatly expanded in PDAC cells (Perera et al. 2015). This finding suggests that increased lysosome biogenesis and function may be critical to the nutrient-scavenging program in PDAC to ensure efficient breakdown and recycling of cellular components and internalized material. The MiT/TFE family of transcription factors (TFE3, MITF, and TFEB) (Settembre et al. 2011, 2013) are upstream regulators of autophagy and lysosome function, which confers PDAC cells with increased catabolic capacity (Perera et al. 2015).

MiT/TFE factors are subject to negative regulation via direct phosphorylation by mTORC1 (Roczniak-Ferguson et al. 2012; Settembre et al. 2012; Martina et al. 2014); that is, under nutrient-replete conditions, mTOR is activated and localized to the lysosome, where it phosphorylates and inactivates the MiT/TFE proteins. Conversely, upon starvation, mTOR is switched off, enabling nuclear translocation of unphosphorylated MiT/TFE proteins. This mechanism of regulation ensures that catabolic processes such as autophagy are suppressed under conditions of high nutrient availability, while activation of biosynthetic reactions are favored. In PDAC, the MiT/TFE proteins bypass mTORC1-mediated negative surveillance and are constitutively localized in the nucleus to activate autophagy and lysosome gene expression (Perera et al. 2015). Thus, PDAC cells have found a way to break the dichotomy between states of biosynthesis and catabolism. By doing so, they maximize growth processes associated with high mTOR activity while simultaneously benefiting from the nutrient scavenging and metabolic adaptation associated with activation of catabolic programs. Similarly, dephosphorylation of other mTORC1 substrates, such as ULK1, also promotes elevated autophagy in the context of high mTORC1 signaling (Wong et al. 2015), as described above. Thus, some tumors activate the MiT/TFE family by either genetic alteration or oncogenic signaling, and this may be a useful biomarker for patient selection for responsiveness to autophagy inhibition.

Another example of a clinically feasible biomarker that could enrich for cancers that are autophagy-dependent is the BRAF^V600E mutation. BRAF mutant cancers may be especially sensitive to autophagy inhibition when combined with BRAF inhibition. BRAF mutant pediatric glioma cells are highly dependent on autophagy (Levy et al. 2014), while BRAF wild-type gliomas are autophagy-independent (Levy and Thorburn 2012). Similar findings were found in BRAF mutant melanoma (Ma et al. 2014). Autophagy markers were up-regulated in a large subset of samples of patients treated with BRAF or BRAF and MEK inhibitors. Mechanistically, BRAF inhibitors produce cytoprotective autophagy through an intermediate unfolded protein response. Genetic inactivation of autophagy also enhances the efficacy of BRAF inhibitors in GEMMs with tumors driven by oncogenic Braf, which are especially autophagy-dependent (Strohecker et al. 2013; Xie et al. 2013; Strohecker and White 2014). This opens up the possibility that pharmacological autophagy inhibition could provide a different strategy to deal with drug resistance: Rather than targeting the same pathway a different way, an entirely different process (i.e., autophagy) is inhibited to circumvent resistance to a kinase inhibitor.

Future directions

The gap in our knowledge currently is the limited understanding of what autophagy specifically does at the molecular level and how this is influenced by tissue type, tumor genetics, and disease state. For example, recycling by autophagy is important to supply substrates for metabolism and survival of NSCLC, but is this true in other cancers and is the purpose the same? Will this inform combination therapy, especially with jointly targeting metabolism? When is protein and organelle quality control important? Are there adaptive or resistance mechanisms to autophagy inactivation? Does host autophagy promote tumor growth and, if so, how? Pancreatic stellate cells use autophagy to provide alanine to PDAC metabolism, but what about tumors that lack this abundant stromal compartment? Analogously, is there a role of host autophagy in supplying tumors with circulating nutrients? Does autophagy inhibition generally promote an anti-tumor immune response, and will it enhance response to checkpoint inhibitors? What are the best targets in the autophagy pathway for inhibitor development? How can autophagic flux be measured in human tissue samples? Can autophagy inhibitors delay the onset of hereditary breast cancer and CRC as the GEMM cancer models suggest as a strategy for cancer prevention? Answering these and other questions will inform our understanding of the role of catabolic processes in health and disease.

Acknowledgments

Footnotes

References