Introduction

A major goal in cancer research is to increase the susceptibility of cancer cells to apoptosis-based strategies [1, 2]. However, the resistance of some cancer cells to chemotherapy could be ascribed not only to their inability to activate the apoptotic death program, but also to the protection exerted by autophagy [3, 4]. In fact, although the notion of autophagy as “programmed cell death (PCD) type II” has been addressed [5], autophagy is generally considered to have a protective function in many cellular stress conditions, being able to counteract nutrient deprivation by recycling energy through macromolecule degradation [6, 7].

In fact, autophagy is a very complex process with death or survival role, working in both normal and cancer cells, with different and sometimes opposite effects. By consequence, it is hard to define exactly its impact in disease and cancer prevention/treatment. Moreover, the cross talk between autophagy regulators and pro-survival/death factors results in plethora of signals that interact each other to decide the final fate of the cell, leading to the so-called apoptosis/autophagy paradox [7, 8].

Apoptosis: a short overview

Programmed Cell Death was first described as a controlled process occurring during insect metamorphosis by Lockshin and Williams [9–13] and during embryogenesis and development both of vertebrates and invertebrates [14, 15]. The term apoptosis comes from the Greek and denotes the falling off of leaves; Kerr and colleagues named cell shrinkage and chromatin margination as apoptosis, and observed that it occurs also under pathological conditions [16]. The hallmarks of apoptosis (often referred to Programmed Cell Death type I), which is an energy-dependent process, are nuclear shrinkage, chromatin condensation and apoptotic body formation. From a biochemical point of view, caspase activation, poly(ADP-ribose) polymerase-1 (PARP-1) proteolysis and DNA fragmentation into oligonucleosome fragments represent the typical markers of classical apoptosis (which is today described as caspase-dependent) [17].

Generally, the apoptotic signaling cascade is divided into two pathways, extrinsic and intrinsic, which are triggered either by soluble molecules that bind to plasma membrane receptors or by different stimuli that converge on mitochondria, respectively (schematized in Fig. 1). The extrinsic apoptotic pathway is promoted by soluble molecules belonging to the Tumor Necrosis Factor (TNF) family, normally secreted as homotrimers, which can bind to plasma membrane receptors of the TNF-Receptor (TNF-R) family, causing their trimerization and subsequent activation. TNF-Rs possess a Death Domain (DD) [18] that is responsible for the recruitment of other DD-containing proteins, such as TNF-R type 1-Associated Death Domain protein (TRADD) and Fas-Associated protein with Death Domain (FADD), which can bind to initiator caspases-8 and -10, allowing homodimerization and in turn, activation. The complex formed by TNF-R, FADD (and eventually TRADD) and caspases-8 and -10 is referred as Death-Inducing Signaling Complex (DISC) [19, 20]. Upon activation, caspases-8 and -10 cleave the effector caspases-3, -6, and -7, which are in charge of cellular dismantling during the final step of apoptosis [21].

Fig. 1
figure 1

Schematic diagram of caspase-dependent apoptosis. Extrinsic pathway (left) is triggered by TNF-like molecules and consequent activation of their receptors (TNF-R) causing DISC formation and promoting the activity of initiator caspases-8 and -10 and, in turn, of effector caspases (-3, -6 and -7). Intrinsic pathway (right) is triggered by intracellular stimuli sensed by Bid and converging on mitochondria, from where cytochrome c and other factors (e.g. Endonuclease G and AIF) are released. Anti-apoptotic proteins from Bcl-2 family inhibit mitochondria permeabilization and subsequent cell death. Upon mitochondrial membrane permeabilization, the apoptosome is formed, consisting of Apaf-1, cytochrome c and caspase-9. Each pathway converges on effector caspases that can promote DNA degradation and cellular protein dismantling. AIF Apoptosis-Inducing Factor, Apaf-1 Apoptotic Protease Activating Factor-1, DISC Death-Inducing Signaling Complex, FADD Fas-Associated protein with Death Domain, TNF Tumor Necrosis Factor, TRADD TNF-R type 1-Associated Death Domain protein

Full size image

The intrinsic apoptotic pathway is driven by various stimuli, e.g. DNA damaging agents, oxidative stress, hypoxia and growth factor deprivation, converging on mitochondria, the sensing organelles of this apoptotic pathway. The key event of the cascade, which ultimately leads to the activation of caspase-3, is represented by mitochondrial outer membrane permeabilization [22]. Mitochondrial integrity is controlled by several members of the Bcl-2 super-family, which are characterized by the presence of the BH3 (Bcl-2 Homology) domain, which can be present in multiple copies to ensure the interaction of one member to the other(s). Two subcategories of Bcl-2 proteins can be distinguished: pro-apoptotic and anti-apoptotic family members. Bax, Bak, Bid, Bad, Noxa and PUMA belong to the first group, while the second group includes Bcl-2, Bcl-XL, Mcl-1 and A1 [23]. During apoptosis, pro-apoptotic Bax and Bak dimerize to insert themselves into the outer mitochondrial membrane and trigger the intrinsic apoptotic pathway thanks to their ability to form holes into biological membranes [24]. Anti-apoptotic Bcl-2 proteins can disrupt this interaction, avoiding mitochondria permeabilization. The latter event is crucial for the execution of intrinsic apoptotic pathway, given that mitochondria contain several pro-apoptotic factors such as cytochrome c, AIF, Smac/DIABLO, HtrA2/Omi and Endonuclease G [22].

Following mitochondria permeabilization, cytochrome c translocates to the cytosol where it can bind to Apoptotic Protein Activating Factor-1 (Apaf-1), allowing the formation of a multi-protein platform called the apoptosome, a seven-spoked, wheel shaped complex essential for the recruitment and subsequent activation of caspase-9 [25]. Upon activation, caspase-9 in turn activates caspase-3, thus promoting the execution of apoptosis.

Caspases are controlled by specific cellular inhibitors called Inhibitor of Apoptotic Proteases (IAP), which can bind to them, thus blocking their function [26, 27]. To counteract IAP function, Smac/DIABLO and HtrA2/Omi are released from mitochondria in order to bind to or cleave IAPs, respectively [28, 29]. The last step of the apoptotic program is DNA degradation which is carried out by various nucleases [30], including Endonuclease G, which is released from mitochondria to the nucleus, where it cleaves DNA. DNA fragmentation occurs also during caspase-independent apoptosis, where it is mainly operated by L-DNase-II [31–34], and in parthanatos, which is characterized by the activation PARP-1, poly(ADP-ribose) over-production and AIF translocation from mitochondria to the nucleus [35]. Once the cell has been committed to death and the apoptotic program is activated, cellular fragments are engulfed by professional or amateur phagocytes near the apoptotic cell, thus avoiding inflammatory response (reviewed in [36]).

Autophagy: general properties

The term autophagy comes from the Greek words αύτος (autos), which means self, and φαγέω (fageo), which means eat; it is a self-degradative process which ensures the regular turnover of cellular components by sequestering damaged organelles and misfolded proteins, targeting them for lysosomal degradation [37]. Since its discovery, autophagy was considered a kind of disposal mechanism aimed at recycling of cellular components [38], but now it is clear that it is much more than a recycle bin, being involved in many cellular and physiological pathways, including development and differentiation [37].

The first insights into the molecular control of autophagy originated from yeast genetic analysis [39], leading to the identification of the first autophagy-related gene atg1 [40]. The further search for atg orthologues in different species demonstrated that they are conserved from slime mould to mammals [41]. To date, four types of autophagy have been described in mammalian cells, i.e. macroautophagy (hereafter referred as autophagy), microautophagy, chaperone-mediated autophagy and alternative macroautophagy [38, 42]. Moreover, specifically targeted autophagy, involving mitochondria turnover, has been described and termed mitophagy [43, 44].

The simplified pathway of autophagy signaling is shown in Fig. 2. The active search for the molecular determinants of autophagy allowed the identification of mammalian Target Of Rapamycin (mTOR) as the key negative regulator of autophagy. mTOR is a conserved Ser/Thr kinase that controls many cellular processes, including transcription and protein synthesis; under stress conditions, mTOR is inactivated, thus allowing autophagy to start through the formation of the ULK complex, composed of FAK-family Interacting Protein of 200 kDa (FIP200), Unc-51-Like Kinase (ULK) and ATG13 (reviewed in [45–47]). Then, in the proximity of the phagophore a multimeric PI3K complex is formed, including Beclin-1, Vacuolar Protein Sorting (VPS) 15, VPS34 and Activating Molecule in Beclin-1-Regulated Autophagy (AMBRA1) [48]. Beclin-1 is positively regulated by AMBRA1, which is phosphorylated and released from the dynein motor complex during autophagy initiation [49]. The PI3K complex is controlled positively by UV radiation Resistance-Associated Gene (UVRAG) and negatively by Rubicon [48].

Fig. 2
figure 2

Autophagy signaling pathways. a mTOR is the negative regulator of the formation of ULK complex (composed of FIP200, ATG13 and ULK), which is essential for autophagy execution. This complex phosphorylates AMBRA1, leading to the activation of PI3K complex (formed by P-AMBRA1, Beclin-1, VPS15 and VPS34). b In the left part, the subsequent cellular structures leading to autolysosome formation are illustrated, while the right part represents the molecular markers of each step. The phagophore colocalizes with the ATG5-ATG12/ATG16 complex during autophagosome formation. Finally, mature autophagosome, which contains LC3-II protein on its membrane, fuses with the lysosome, forming the autolysosome, where autophagosomal content is degraded. Phosphorylation of AMBRA1 is marked with a star. AMBRA1 Activating Molecule in Beclin 1-Regulated Autophagy, ATG Autophagy-related Gene, FIP200 FAK-family Interacting Protein of 200 kDa, LC3 microtubule-associated protein Light Chain 3, mTOR mammalian Target Of Rapamycin, PI3K Phosphatidyl Inositol 3 Kinase, ULK1 Unc-51-Like Kinase 1, UVRAG UV radiation Resistance-Associated Gene, VPS Vacuolar Protein Sorting

Full size image

As illustrated in Fig. 2, the key event of autophagy is the formation of the autophagosome, which proceeds through different steps (reviewed in [50]).

1. Formation of the so-called phagophore, i.e. the first nucleating site for autophagosome assembling [51]. The membranes mainly originate from endoplasmic reticulum [52–54], trans-Golgi and late endosomes [52, 55–57], even if de novo membrane formation may occur [58].

During phagophore assembling, an ubiquitin-like system regulates the formation of the ATG5-ATG12 heterodimer. Then, a protein complex associates to phagophore, consisting of ATG5-ATG12 and a homodimer of ATG16.

2. Formation of the autophagosome. Another ubiquitin-like system modulates the correct elongation of autophagosomal membranes and allows the processing of LC3-I (microtubule-associated protein Light Chain 3) into the isoform II. LC3-II is the most specific marker for autophagosome formation, being incorporated into the autophagosomal membrane during its elongation (for an exhaustive review, see [58]).

3. Formation of the autolysosome, which originates from the fusion of the autophagosome with the lysosome. Delivery and fusion of the autophagosome to lysosome are controlled by cytoskeleton and lysosomal membrane proteins, respectively [55]. In fact, drugs that specifically target microtubules, such as nocodazole, block autophagosome conversion into autolysosomes [59]. Autophagosome maturation can be regulated by lysosomal LAMP1/2 (Lysosomal Associated Membrane Protein); in this respect, a mutation in the LAMP2 gene has been found to be associated to the X-linked Danon disease, which is characterized by cardiomyocyte hypertrophy and accumulation of autophagosomes in the heart tissue [60].

4. Degradation of autolysosomal content. The final step, carried out by lysosomal hydrolases, consists of lipid, protein and sugar breakdown and consequent release of the catabolic products into the cytoplasm, thus rendering them available to be re-used, possibly acting as a novel source for energy production (see below).

Although the dynamics of the assembly of both ULK and PI3K complexes is clear, knowledge of the final part of the story is still limited [61]. Indeed, the actual body of evidence obtained by different research groups points out that autophagy is a very complex pathway, as it emerges also from a recent proteomic survey defining various functional modules within the process [62]. To enhance the availability of the many data so far collected, Homma et al. [63] provided the scientific community with the Autophagy Database (http://tp-apg.genes.nig.ac.jp/autophagy) that is an invaluable tool for classifying autophagy-related proteins and updating the specific literature, which increased dramatically in the last 10 years [63].

Autophagy as a source for energy production

The original concept that autophagy, under physiological conditions, is a recycling mechanism that provides cells with novel amino acids, monosaccharides and lipids originating from the breaking down of damaged or aged protein/organelles, has been modified on the basis of the recent evidence that autophagy controls also lipid metabolism (in particular lipid droplets) [64]. Lipid droplets (LD) are cytosolic structures constituted mainly by cholesterol and triglycerides, used by the cell as lipid storage [65, 66]. Singh et al. [64] showed that in the rat hepatocyte RALA255-10G cell line defective autophagy modulates lipid storage inside LD causing an increase in their size and number. In fact, after prolonged starvation, they observed a co-localization of LC3 with LD surface and an increase in the percentage of LD-containing autophagosomes [64]. Autophagic control of LD formation is intriguing because it provides the evidence that autophagy is essential to regulate the availability of intracellular free fatty acids and lipid load. Although free fatty acids are not the main resource for energy production [67], under starvation or stress conditions autophagy may contribute to energy production through free fatty acid release; these lipids may be delivered to mitochondria to produce energy by β-oxidation reactions [68].

Under stress conditions, autophagy provides cells with energy to ensure correct homeostasis [69]. As an example of this notion, it has been reported that the autophagic catabolism of intracellular substrates allows IL3-deprived lymphocytes to face adverse conditions by maintaining high ATP levels [70]. That autophagy is essential to ensure self-sustaining under critical conditions, was supported also by the evidence that, when autophagic functions are inhibited (by RNA interference or chemicals), starved lymphocytes undergo cell death [70]. In vivo experiments aiming at evaluating the impact of crucial autophagy genes on stress response were performed. A sudden deprivation of nutrients occurs after birth in mammals, causing a metabolic stress normally counteracted by autophagy; in fact, it was reported that atg5 −/− mice die at birth because of autophagy defect and consequent impaired production of recyclable amino acids [71]; moreover, it was found that inefficient autophagy could inhibit oocyte-embryo transition in mice [72]. Low intracellular ATP levels due to impaired autophagy in atg5 or Beclin-1 null mice affected embryonic development; in particular, the cavitation process appeared to be compromised due to an inefficient clearance of apoptotic bodies, which requires phosphatidylserine exposure and LPC secretion, both energy-consuming processes [73], as it is the case in general for apoptosis, which shares with autophagy some molecular determinants, as described in the following section.

Apoptosis and autophagy cross talk: crucial factors

The main interconnection routes linking apoptosis and autophagy are depicted in Fig. 3. The active search for common regulators of apoptosis and autophagy led to the discovery that Bcl-2 family members play a double (and opposite) role (reviewed in [74]). Indeed, Bcl-2 and Bcl-XL bind to and inhibit Beclin-1 through the Beclin-1 BH3 domain [75]. Under nutrient excess conditions, when autophagy is not necessary, the association between Bcl-2 and Beclin-1 is maximal, while decreases after cell starvation, i.e. when autophagy is essential to guarantee cell survival [76]. This latter condition is initially modulated by JNK1-mediated Bcl-2 phosphorylation, leading to its dissociation from Beclin-1, thus promoting autophagy. To face the deleterious effects of long term starvation, which cannot be rescued anymore by autophagy, phosphorylated Bcl-2 binds the pro-apoptotic protein Bax, thus inhibiting apoptosis [77, 78]. However, under extreme conditions, JNK1 mediates hyper-phosphorylation of Bcl-2, which detaches from Bax, thus facilitating apoptosis and consequently a safe cell death [78].

Fig. 3
figure 3

Apoptosis and autophagy cross talk. Upon autophagic stimuli (right), JNK and DAPK phosphorylate Bcl-2 and Beclin-1, respectively, thus abolishing their association. As a consequence, free P-Bcl-2 associates to Bax and inhibits apoptosis. Free P-Beclin-1 associates with VPS15 and VPS34 forming the PI3K complex, leading to ATG5-ATG12/ATG16 complex formation and, ultimately, autophagy. p53 (left) induces DRAM, mTOR and Bak synthesis, leading to apoptosis and/or autophagy. Bak induces mitochondrial outer membrane permeabilization, triggering cytochrome c translocation to the cytosol and, finally, apoptosis. Calpains may induce apoptosis by ATG5 cleavage and its consequent translocation to mitochondria; at the same time, truncated ATG5 (tATG5) inhibits autophagy. Upon apoptotic stimuli, active caspase-3 cleaves Beclin-1, thus blocking autophagy; then, Beclin-1 C-term fragment moves to the mitochondria, acting as a pro-apoptotic protein, thus amplifying the apoptotic cascade. Protein phosphorylation is marked with a star and cleavage with scissors. AMBRA1 Activating Molecule in Beclin 1-Regulated Autophagy, ATG Autophagy-related Gene, DAPK Death-Associated Protein Kinase, DRAM Damage-Regulated Autophagy Modulator, JNK c-Jun N-terminal Kinase, mTOR mammalian Target Of Rapamycin, VPS Vacuolar Protein Sorting

Full size image

To explain Bcl-2 binding properties, a speculative model was proposed, taking into account that the affinity of phosphorylated Bcl-2 toward pro-apoptotic proteins is higher than that toward Beclin-1 [78]. The hypothesis was based on the fact that, although Beclin-1 and pro-apoptotic Bcl-2 family members share the same BH3 domain [79], the absence of a hydrophobic amino acid at position 119 (which contains the polar Thr) lowers Beclin-1 affinity for Bcl-2 compared to other BH3-containing proteins [80]. The relevance of a finely tuned regulation of Beclin-1/Bcl-2 binding was further demonstrated by the evidence that Death-Associated Protein Kinase (DAPK) phosphorylates Beclin-1 on Thr119, thus promoting its dissociation from Bcl-2, and autophagy activation [81]. Of note, DAPK also participates in apoptotic bleb formation thanks to its interplay with cytoskeletal factors [82].

The existence of cross talk between autophagy and apoptosis is also supported by the double role of ATG5, which, in addition to the promotion of autophagy, enhances susceptibility to apoptotic stimuli. Overexpression of Atg5 sensitizes tumor cells to chemotherapy [83]; by contrast, silencing of this gene results in partial resistance to anticancer drugs. It was shown that during apoptosis ATG5 is cleaved by calpains. This event allows ATG5 translocation to mitochondria, where it interacts with Bcl-XL and controls cytochrome c release, caspase activation and apoptosis; on the contrary, in the absence of ATG5 within mitochondria, autophagy takes place [83, 84]. This sequence was observed in several apoptotic paradigms, suggesting that both calpain activation and ATG5 cleavage might be general phenomena in apoptotic cells [83, 84].

Another autophagic factor, i.e. ATG3, is controlled by FLICE-Inhibitory Protein (FLIP), which is a negative regulator of the extrinsic apoptotic pathway (being able to recognize and bind FADD through specific DED domains). Remarkably, within these domains, non-overlapping sequences are in charge of apoptosis inhibition and recognition and binding of ATG3; the latter event blocks the conjugation to LC3 and, in turn, autophagy [85]. This points out a double role of FLIP, which can control both apoptosis and autophagy at the same time.

Moreover, it has been shown that main apoptotic proteases, i.e. caspases, can degrade autophagic proteins. Caspase-3 [86] and other caspases [87, 88] are able to cleave and inactivate Beclin-1, thus inhibiting autophagy and consequently enhancing apoptosis progression [89]. Furthermore, Wirawan et al. [88] showed that caspase-mediated cleavage of Beclin-1 during apoptosis induced by IL-3 withdrawal from culture medium, not only blocks autophagy but enhances the power of the pro-apoptotic stimulus triggered by growth factor deprivation. Of note, N- and C-terminal fragments of Beclin-1 generated after the cleavage relocate to the nucleus and mitochondria, respectively. Remarkably, Beclin-1 C-terminal, which lacks the BH3 domain, induces cytochrome c and HtrA2/Omi release from mitochondria and subsequent amplification of the apoptotic stimulus. The results of this study are in line with other observations on cleaved ATG5 fate [83, 84], thus supporting a possible double role of key autophagic proteins in controlling autophagy and enhancing apoptosis.

A physical interaction between Beclin-1 and the apoptosis inhibitor Survivin was documented, which does not alter protein function but increases Survivin half-life, thus promoting at the same time autophagy and cell survival [90]. Given that many signals converge on Beclin-1 to determine the final fate of the cells, this protein appears to be essential in maintaining cellular homeostasis, acting as a molecular switch from autophagy to apoptosis.

A further link between these processes is represented by p62, also called sequestosome, a multifunctional protein that targets proteins to proteasome degradation and autophagy; p62 is implicated in autophagy progression as well as in apoptosis induction and cancer development [91]. LC3-II binds p62 to regulate protein packaging and delivering to the autophagosome, thus facilitating the clearance of misfolded/damaged proteins and deformed organelles through the autophagic machinery [92–94].

p62 accumulation was described in autophagy-defective cells, which suffer from proteasome inactivation and altered NF-κB regulation, and undergo tumorigenesis [95]. The recent observation of a p62 role in modulating LC3-II degradation by the proteasome indicates that p62 is a crucial factor in the interplay between autophagy and proteasome activity [96]. In the apoptotic scenario, the evidence for an effect of p62 on caspase-8 was provided; in fact, the critical initiator of the extrinsic apoptotic pathway requires p62 for its efficient polyubiquitination, aggregation and full activation [97].

A simplified summary of the double role of some death effectors in apoptosis and autophagy is presented in Table 1.

Table 1 Functions of autophagy and apoptosis effectors
Full size table

Also p53 is involved in the cross talk between apoptosis and autophagy. On the one hand, p53 regulates positively apoptosis by activating Bax and inducing mitochondrial depolarization; on the other hand, it increases the synthesis of DRAM and mTOR, both involved in the control of autophagy with opposite effects [98]. It is worth noting that especially the cytoplasmic form of p53 seems to be responsible for inhibition of autophagy, which can be rescued by blocking p53 itself [99, 100]. On the contrary, p53 post-transcriptionally down-regulates LC3 levels in starved cells, thus allowing a fine control of the autophagic flux, avoiding the “autophagic burst” that could be dangerous for cells [101].

An innovative approach to the analysis of programmed cell death, based on systems biology, revealed that although individual death processes, i.e. apoptosis, autophagy and programmed necrosis, are regulated by different signaling pathways, some regulators are common to more than one process [102, 103]. Taking advantage of the use of a platform based on single and double sets of RNAi-mediated perturbations targeting combinations of apoptotic and autophagic genes, Kimchi’s group identified several levels of connection between apoptosis and autophagy [104, 105].

Altogether, these observations support the existence of a cross talk between apoptosis and autophagy; however, how the intricate interconnection functionally affects the final outcome, remains a mystery.

Cancer: a ping-pong match between autophagy and apoptosis

The first correlation between autophagy and tumorigenesis relies on the observation that ectopic expression of Beclin-1 decreases cancer cell proliferation in vitro and tumorigenesis in vivo [106]. Thereafter, the evidence that mice lacking one copy of Beclin-1 gene show a high incidence of spontaneous cancers demonstrated that Beclin-1 is a haploinsufficient tumor suppressor [107, 108], as confirmed in various human cancers showing monoallelic deletion of Beclin-1 locus (17q21) (reviewed in [109]).

Of note, mutations of others autophagy-regulator genes, such as atg5, uvrag and atg4C were correlated to cancer conditions, e.g. natural killer-cell malignancy [110], colon and gastric carcinoma [111, 112] and fibrosarcoma [113].

The different functions af autophagy in normal and cancer cells are summarized in Fig. 4. As shown for normal cells, autophagy could ensure homeostasis under stress conditions (a, left) and modulate cancer cell dynamics by maintaining genome integrity, as supported by the evidence that in autophagy-deficient mice, the allelic loss of Beclin-1 causes an increase of DSBs (monitored by γH2AX level), chromosome aberrations, centrosome abnormalities, aneuploidy, gene amplifications and cancer proneness (a, right) [114]. Further evidence supporting this effect was based on the association between inefficient autophagy and accumulation of the scaffold protein p62, which may induce mitochondrial damage and hyper-production of reactive oxygen species (ROS), leading to a permanent status of oxidative stress resulting in DNA damage and cancer development [95].

Fig. 4
figure 4

Multiple roles of autophagy in normal and cancer cells. (a) Normal cells face stress conditions (represented by lightning bolt) by activating autophagy, thus recovering from the insult and maintaining tissue homeostasis (left); conversely, inefficient autophagy causes chromosomal instability, leading to neoplastic transformation and cancer development (right). (b) In the presence of hyper-proliferation signals from oncogenes, normal cells can take advantage from autophagy to trigger Oncogene-Induced Senescence (OIS), thus preventing cancer development (left), which occurs when autophagy-deficient cells are unable to block oncogene-driven transformation (right). (c) Cancer cells suffering from intrinsic metabolic, physical or pharmacological stress (lightning bolt) activate autophagy that protects from stress and allows their proliferation (left). When autophagy is inhibited, stress conditions trigger cell death through the activation of apoptosis or necrosis, causing tumor regression or inflammation, respectively (middle). Enforced autophagy (right) turns the physiological protective effect into cell death (PCD type II). DSB Double Strand Breaks, OIS Oncogene-Induced Senescente, PCD type II Programmed Cell Death type II

Full size image

Furthermore, as shown in Fig. 4b, autophagy could be involved in the so-called Oncogene-Induced Senescence (OIS), by which cells counteract dangerous signals from oncogenes (e.g. Ras) leading to hyper-proliferation (and, potentially, transformation) [115, 116]. Thus, OIS represents a safe manner to avoid tumorigenesis. To do this, senescent cells improve autophagy efficiency by enhancing the expression level of crucial factors, as demonstrated by LC3-II accumulation [117]. To better elucidate the role of autophagy in senescent cells, atg5 or atg7 depleted cells were monitored for proliferation and cancer markers and found to be able to bypass OIS and re-enter cell cycle [117]. Conversely, the enforced activation of autophagy through the over-expression of ULK3 was sufficient to drive cells to senescence [117].

The role of autophagy in overcoming tumor development by counteracting genomic instability and/or aberrant cell cycle allowed its label as a “guardian-of-the-genome” [118], thus supporting the cytoprotective function of this process toward cancer development (reviewed in [119, 120]).

The situation appears to be more complex when a tumor is already established and suffers from micro-environmental hypoxia and nutrient deprivation, and from DNA and cellular damage after anti-cancer therapy. These are the typical “SOS” signals triggering autophagy, elimination of intracellular damaged components and replenishment of energy, thus promoting cancer cell survival. In this respect, autophagy is the key player of a paradox for cancer biology, so that it has been defined “a cell death impostor” [121]. Of interest, a dual role of autophagy in controlling the metastatic potential of cancer cells was postulated, based on the effect on anoikis, which is the essential feature of migrating cells [122].

From the above considerations, it is obvious to wonder if the modulation of autophagy could have beneficial effects on apoptosis-refractory tumors. In order to minimize the protection of cancer cells by autophagy, a series of inhibitors was developed. The abrogation of autophagy proved to re-sensitize cancer cells to apoptogenic stimuli (Fig. 4c), thus allowing the setting of an innovative strategy based on autophagy inhibitors currently used in clinical trials [123]. To better define the molecular basis of the effect of inhibition of autophagy, it is crucial to investigate if apoptosis becomes the primary route to death of cancer cells in which autophagy is inhibited. However, we must keep in mind that the dangerous induction of excessive necrosis and consequent inflammation, which recruits in loco many inflammatory cells, such as cytotoxic T cells and natural killer cells, could promote tumor progression (Fig. 4c) [124, 125].

From the seminal paper of Boya et al. [126], showing that autophagy may be cytoprotective in nutrient-depleted cells, and that autophagy inhibition triggers apoptosis, accumulating evidence supported the existence of cross talk between type I and II cell death pathways, the latter being possibly a “brake” of apoptosis. Thus, inhibition of autophagy could help in overcoming apoptosis evasion of cancer cells.

The pioneering observation that the most important autophagy regulator, i.e. mTOR, may influence apoptosis activation strengthens the link between the two processes. In fact, mTOR inhibitor Rapamycin can induce apoptosis in dendritic cells, thus sensitizing cancer cells to cisplatin [127]. Moreover, it was recently reported that inhibition of mTOR pathway results in enhanced apoptosis in melanoma cells [128]. Accordingly, autophagy inhibition in human neuroblastoma cells was found to accelerate the apoptotic response to the herbicide paraquat [129]; in autophagy-inhibited colon cancer cells treated with bortezomib [130] and sulforaphane-treated human breast cancer cells, restoration of apoptosis was reported [131]. Cancer HepG2 cells treated with etoposide were resistant to apoptosis; however, when autophagy was inhibited, they responded to the drug by activating caspase-dependent apoptosis [132]. These results are the proof of the constant conversation between the two main players in the cell survival game.

Finally, it has to be mentioned that an alternative anti-cancer therapy based on autophagy-inducing agents revealed that apoptosis-resistant cancer cells can be killed through enforced autophagy, which acts in this case as Programmed Cell Death type II (reviewed in [109, 133, 134]) either by cooperating with other cell death mechanisms or murdering cells by itself (Fig. 4c) [135–137].

In conclusion, when a new drug is tested on cancer cells, it is advisable to investigate not only the existence of apoptotic hallmarks but also the presence of markers of autophagy, to avoid precipitous conclusions about the drug mechanism of action. As an example of this approach, we have recently described that the apoptogenic agent 2-methoxyestradiol could promote autophagy in human cancer cell lines characterized by apoptosis-resistance [138]. This observation could be exploited to manipulate autophagy (e.g. by inactivating crucial functions) and check for the impact of protective effect avoidance on drug response.

Concluding remarks

The growing interest toward the physiological role of autophagy is supported by studies on infectious diseases [139, 140], neurological disorders [141], cardiac remodelling [142] and longevity [143, 144]. Defective autophagy is associated with some disorders, e.g. collagen VI muscular dystrophies [145], Crohn’s disease, stroke, diabetes, neurodegenerative and inflammatory paradigms (reviewed in [146]). In order to facilitate the monitoring of autophagy, specific morphological and biochemical protocols have been developed [147–153]. A panel of molecules modulating autophagy (reviewed by [154]) is now widely used and considerable effort was directed toward the use of knockout animals (or cells) to study if and how different processes are perturbed by the absence of functional autophagic factors [150, 155]. As underlined by Rubinzstein, the final outcome of the research in the autophagy domain is to extrapolate to complex organisms the experience of yeast, that is, to manipulate autophagy in whole organisms [156].

Special attention is actually paid to the multifaceted scenario of tumor formation and growth, where autophagy could counteract the genomic instability that is a causative event of cancer, as well as facilitate tumor resistance to therapy, which generally relies on the inability to undergo apoptosis. In this respect, the intricate interconnection between autophagy and apoptosis requires a more accurate deciphering of the crucial signals governing their cross talk.

Despite the important progress made in elucidating the basic features of autophagy and its interconnection with apoptosis, the function of this process in cancer still represents a matter of debate. From the huge amount of published data, it is hard to associate a unidirectional function to the autophagic process given that the final outcome depends on the cell type, the stimulus a cell has to face and the ability to evade or not apoptosis in response to drug treatment.