The role of autophagy in tumorigenesis is complex; its action may be compartment- and cell type-specific.23
Autophagy was originally linked to reduced tumor growth when Beclin1, a crucial autophagy regulator, was first reported to be deleted in many epithelial cancers, indicating that autophagy can function as a tumor suppressor.64
In dramatic support of this simple concept, recombinant expression of Beclin1 in epithelial cancer cells drives autophagy and effectively suppresses tumorigenesis.65
Similarly, deficiencies in other autophagy regulators have been linked to a pro-tumorigenic phenotype.66
More specifically, autophagy-deficient cancer cells show increased DNA damage and genomic instability as well as increased ROS levels and oxidative stress.67
Several oncogenes have been shown to repress autophagy in cancer cells, supporting the overall idea that catabolic processes might be suppressed in tumor cells in order to favor the accumulation of cell biomass.68
However, just the opposite effects may be occurring in the tumor stromal microenvironment (). Thus, a catabolic tumor stroma, driven by autophagy in cancer-associated fibroblasts, could “energize” the anabolic growth of cancer cells by providing essential nutrients and mitochondrial fuels (L-lactate, ketones, glutamine and free fatty acids) in a paracrine fashion.2,69
We have previously termed this hypothesis the “autophagic tumor stroma model of cancer metabolism”2,7
Figure 15. Two-compartment tumor metabolism is fueled by the autophagy-senescence transition (AST). In this model, cancer cells secrete hydrogen peroxide (H2O2), which induces oxidative stress in neighboring normal fibroblasts. Oxidative stress (more ...)
More specifically, we have demonstrated that oncogene-induced activation of H2
production in cancer cells drives oxidative stress in cancer-associated fibroblasts (CAFs), which then undergo autophagy and mitophagy.3,4,70
Loss of mitochondrial function increases aerobic glycolysis, driving the production and release of micro-nutrients, such as L-lactate and ketone bodies into the tumor microenvironment.11,49,71-73
Ketones and L-lactate, in turn, can act by a paracrine mechanism on cancer cells, driving mitochondrial biogenesis and OXPHOS (oxidative phosphorylation) in tumor cells, generating a positive feedforward loop to support anabolic tumor growth.9,10,74-76
Here, we developed a genetically tractable model system to directly study the compartment-specific role of autophagy in tumor growth and metastasis. First, we expressed autophagy-associated genes (such as BNIP3, CTSB, and ATG16L1) in hTERT-immortalized fibroblasts to genetically generate constitutively autophagic fibroblasts. Then, we validated that these genetically modified fibroblasts undergo autophagy, mitophagy, develop significant mitochondrial dysfunction and metabolically shift toward glycolysis, with increased L-lactate and ketone body production. Importantly, these autophagic fibroblasts effectively promoted tumor growth and metastasis, mimicking the behavior of cancer-associated fibroblasts.
Interestingly, we demonstrated that overexpression of a single autophagy gene is sufficient to drive autophagy induction and promote tumor growth. Our study establishes that autophagy, by itself, is sufficient to enhance tumor growth, independently of neo-vascularization. Autophagy action is clearly compartment-specific, since we also show that autophagy induction in human breast cancer cells greatly diminishes tumor growth. Thus, the induction of autophagy in stroma cells releases nutrients into the tumor microenvironment, which may support cancer cell growth. On the other hand, activation of autophagy in cancer cells drives the consumption of cellular components and effectively reduces tumor growth in MDA-MB-231 cells harboring ATG16L1.
A loss of stromal Cav-1 expression, specifically in cancer-associated fibroblasts, is a new important prognostic marker that predicts early tumor recurrence, lymph node metastasis and tamoxifen resistance, as well as poor clinical outcome in breast cancer patients.15-20
The proposed mechanism underlying a loss of stromal Cav-1 in cancer-associated fibroblasts is that Cav-1 undergoes lysosomal degradation, due to the induction of autophagy, secondary to oxidative stress.5,6,23,51
Our current data mechanistically validated this hypothesis, as genetic induction of autophagy/mitophagy with numerous autophagy-associated genes was indeed sufficient to drive a loss of Cav-1 protein expression in fibroblasts. Thus, the autophagic destruction of Cav-1 is biomarker for poor prognosis, as the autophagic tumor stroma provides recycled nutrients to fuel anabolic tumor growth.
We also show that autophagy-induced mitochondrial dysfunction drives a shift toward glycolysis in stromal fibroblasts, with associated increases L-lactate or ketone body production. Remarkably, autophagic fibroblasts also promoted tumor cell metastasis, likely via the production of high-energy mitochondrial fuels, such as L-lactate and ketone bodies, which could be “burned” by OXPHOS in adjacent cancer cells. In this regard, ketogenic ATG16L1 fibroblasts showed the highest efficiency in promoting distant metastasis, resulting in an ~11-fold increase in lung colonization. Consistent with these findings, ketones are a more powerful mitochondrial fuel as compared with L-lactate. Ketones can produce more energy than L-lactate, as they require substantially less oxygen consumption and can be “burned” by mitochondria under ischemic and/or hypoxic conditions.77,78
Thus, there may be a direct connection between ketone body production/utilization and cancer cell metastasis.
Ketone bodies are a group of organic molecules are produced both in the cytosol and in mitochondria, which are always present at a low level in healthy individuals. However, dietary manipulations and certain pathological conditions can increase the levels of these compounds in vivo. Pathologies leading to abnormal glucose metabolism, such as poorly controlled diabetes mellitus, can drive increased ketone body production.79
Epidemiologic studies suggest that people suffering from diabetes are at a higher risk of many forms of cancer (including breast, colon, pancreas and etc.) and diabetes is also associated with reduced survival after cancer.80,81
Breast cancer mortality rates remain high, primarily due to the metastasis of primary tumors to distant organs, such as the lungs.82
In diabetic patients, it is possible that breast cancer metastasis may be augmented by metabolic dysfunction, such as a tendency toward ketone body production. Although these data need to be confirmed with additional studies, they suggest that ketone bodies could play a crucial role in breast cancer metastasis.
Recently, several studies have linked autophagy with senescence. For example, ULK-3, a homolog of ATG1, induces autophagy in fibroblasts, conferring the acquisition of a senescence phenotype.27,61
Another protein proposed as a link between autophagy and senescence is cathepsin B. More specifically, CTSB is able to proteolytically cleave SIRT1, allowing for SIRT1 inactivation and driving the establishment of a senescence phenotype.83
Thus, autophagy and senescence may be viewed as a continuum or two related phases of the same overall biological process.
Our current results also directly support a molecular link between autophagy and senescence in the tumor stroma. For example, all the autophagic fibroblast cell lines we examined (overexpressing either BNIP3, CTSB, or ATG16L1) showed key features of a senescence phenotype, such as (1) induction of p21(WAF1/CIP1), a well-characterized CDK inhibitor, (2) a change in cell shape, consistent with cellular hypertrophy and (3) the upregulation of β-galactosidase activity. To confirm the clinical relevance of our findings, we also examined the distribution of β-galactosidase in human breast cancers by immunohistochemistry, using specifc antibody probes. Interestingly, the distribution of β-galactosidase (and thus the senescence phenotype) was largely confined to the tumor stroma and absent from epithelial cancer cells. Thus, senescence may be primarily confined to the tumor stromal compartment. Interestingly, β-galactosidase is also a known lysosomal enzyme and senescent cells show an increase in lysosomal mass,32,37
also supporting the general idea that senescence and autophagy are two sides of the same coin. Senescent cells also show a shift toward aerobic glycolysis,35,36
probably secondary to the onset of mitophagy.
Importantly, Judy Campisi’s group has previously shown that senescent fibroblasts promote tumor growth.24-26,84
Her laboratory attributed the growth-promoting activity of senescent fibroblasts to the senescence-associated secretory phenotype (SASP), which results in the secretion of specific growth factors and cytokines, as well as extracellular matrix remodeling.24,25
However, she estimated that the SASP accounts for only ~50% of the tumor-promoting activity of senescent fibroblasts.24,25
The other 50% still remains unaccounted for.24,25
Importantly, many of the same stimuli that she used to generate senescent fibroblasts, such as acute exposure to hydrogen peroxide and oxidative stress, are now known to be inducers of autophagy, and are also inducers of myofibroblast differentiation.29-31
Thus, we propose here that the other 50% of the tumor-promoting activity of senescent fibroblasts may actually reflect their autophagic/catabolic phenotype, mitochondrial dysfunction and a shift toward aerobic glycolysis, which results in production of mitochondrial fuels (such as L-lactate, ketones, glutamine and free-fatty acids). These mitochondrial fuels can then “feed” the TCA cycle and OXPHOS in adjacent cancer cells, allowing for the anabolic growth of tumor cells. In accordance with our new “metabolic” hypothesis, certain sub-types of senescent fibroblasts promote tumor growth but do not show the senescence-associated secretory phenotype (SASP).26
Thus, inhibition of autophagy and senescence in cancer-associated fibroblasts is now a new attractive metabolic therapeutic target for “cutting off the fuel supply” to rapidly proliferating cancer cells that are energetically dependent on the tumor stroma.