Autophagy and senescence in cancer-associated fibroblasts metabolically supports tumor growth and metastasis via glycolysis and ketone production

Abstract

Senescent fibroblasts are known to promote tumor growth. However, the exact mechanism remains largely unknown. An important clue comes from recent studies linking autophagy with the onset of senescence. Thus, autophagy and senescence may be part of the same physiological process, known as the autophagy-senescence transition (AST). To test this hypothesis, human fibroblasts immortalized with telomerase (hTERT-BJ1) were stably transfected with autophagy genes (BNIP3, CTSB or ATG16L1). Their overexpression was sufficient to induce a constitutive autophagic phenotype, with features of mitophagy, mitochondrial dysfunction and a shift toward aerobic glycolysis, resulting in L-lactate and ketone body production. Autophagic fibroblasts also showed features of senescence, with increased p21(WAF1/CIP1), a CDK inhibitor, cellular hypertrophy and increased β-galactosidase activity. Thus, we genetically validated the existence of the autophagy-senescence transition. Importantly, autophagic-senescent fibroblasts promoted tumor growth and metastasis, when co-injected with human breast cancer cells, independently of angiogenesis. Autophagic-senescent fibroblasts stimulated mitochondrial metabolism in adjacent cancer cells, when the two cell types were co-cultured, as visualized by MitoTracker staining. In particular, autophagic ATG16L1 fibroblasts, which produced large amounts of ketone bodies (3-hydroxy-butyrate), had the strongest effects and promoted metastasis by up to 11-fold. Conversely, expression of ATG16L1 in epithelial cancer cells inhibited tumor growth, indicating that the effects of autophagy are compartment-specific. Thus, autophagic-senescent fibroblasts metabolically promote tumor growth and metastasis, by paracrine production of high-energy mitochondrial fuels. Our current studies provide genetic support for the importance of "two-compartment tumor metabolism" in driving tumor growth and metastasis via a simple energy transfer mechanism. Finally, β-galactosidase, a known lysosomal enzyme and biomarker of senescence, was localized to the tumor stroma in human breast cancer tissues, providing in vivo support for our hypothesis. Bioinformatic analysis of genome-wide transcriptional profiles from tumor stroma, isolated from human breast cancers, also validated the onset of an autophagy-senescence transition. Taken together, these studies establish a new functional link between host aging, autophagy, the tumor microenvironment and cancer metabolism.

Figure 1. Fibroblasts overexpressing BNIP3 show a loss of Cav-1 expression, with constitutive activation of the autophagic program. BNIP3 was stably overexpressed in hTERT fibroblasts via transduction with lenti-viral vectors. Lv- represents fibroblasts transduced with the vector alone control, namely Lv-105 (puro). (A) Note that BNIP3 overexpression in fibroblasts is sufficient to cause a dramatic reduction in caveolin-1 (Cav-1) protein expression, seen by immunoblot analysis. (B) Similarly, BNIP3 overexpression drives the upregulation of LC3-I and -II and cathepsin B, under basal cell culture conditions.(C) Overnight starvation drives even stronger expression of BNIP3, as well as Beclin1 and ATG16L1, which are other key markers of autophagy. Thus, BNIP3 overexpression in fibroblasts is indeed sufficient to drive a loss of Cav-1 expression via the activation on an autophagic program. In all three panels, immunoblotting with β-actin is shown as a control for equal protein loading.

Figure 2. Fibroblasts overexpressing BNIP3 promote tumor growth without a measurable increase in angiogenesis. To assess the functional in vivo effects of BNIP3 overexpression in fibroblasts, we employed a mouse xenograft model. Briefly, MDA-MB-231 breast cancer cells were co-injected with BNIP3 fibroblasts or vector alone control fibroblasts, into the flanks of nude mice. n = 10 per experimental group. Lv-105 represents fibroblasts transduced with the vector alone control, namely Lv-105 (puro). (A) Note that at 4 weeks post-injection, BNIP3 fibroblasts promoted an ~2.1-fold increase in tumor growth. (B) However, no significant differences in tumor neo-vascularization were observed. Thus, autophagic BNIP3 fibroblasts are sufficient to drive tumor growth, without increased angiogenesis.

Figure 3. BNIP3L and Beclin1 overexpression in fibroblasts drives a loss of Cav-1 expression, and promotes tumor growth. To investigate the involvement of stromal BNIP3L and Beclin1 in tumor formation, we stably overexpressed both genes in hTERT fibroblasts. Lv- represents fibroblasts transduced with the vector alone control, namely Lv-105 (puro). (A) Note that BNIP3L and Beclin1 overexpression in fibroblasts is sufficient to induce a loss of Cav-1 expression, as observed by immunoblot analysis. Blotting with β-actin is shown as a control for equal protein loading. (B) BNIP3-, Beclin1- or control- fibroblasts were co-injected with MDA-MB-231 epithelial breast cancer cells, into the flanks of nude mice. At 4 weeks post-injection, the mice were sacrificed and the tumors were collected. n = 10 per experimental group. For both BNIP3L- and Beclin1 fibroblasts, a significant increase in tumor growth was observed. Thus, overexpression of both autophagic genes (BNIP3L and Beclin1) is sufficient to drive tumor growth.

Figure 4. Fibroblasts overexpressing Cathepsin B (CTSB) show a loss of Cav-1 expression, with constitutive activation of the autophagic program. (A) To investigate if stromal cathepsin B expression plays a significant functional role in breast cancer pathogenesis, we overexpressed the CTSB gene in fibroblasts, via lenti-viral transduction. Note that both the CTSB-precursor and activated-cleaved form were observed by immunoblot analysis, in stably transfected hTERT fibroblasts. (B) Cathepsin B expression was also evaluated after 12h and 16h of starvation (in Hepes-buffered HBSS). Note that CTSB continued to accumulate during starvation, consistent with its stabilization within lysosomes, during autophagy. (C) We also evaluated Cav-1 expression after overnight starvation. As expected, we detected a strong reduction in Cav-1 protein levels, validating the connection between autophagy and loss of Cav-1 protein expression. (D) Immunoblot analysis of CTSB fibroblasts demonstrated that cathepsin B overexpression is sufficient to drive the induction of autophagy. Note that cathepsin B overexpression strongly induces several autophagy/mitophagy markers (BNIP3, Lamp1, Beclin1 and ATG16L1). In all panels, β-actin expression was assessed as a control for equal protein loading. Lv- represents fibroblasts transduced with the vector alone control, namely Lv-105 (puro).

Figure 5. Fibroblasts overexpressing Cathepsin B (CTSB) drive enhanced tumor growth, without a measurable increase in angiogenesis.(A) The in vivo functional effects of stromal CTSB were evaluated using a xenograft model. MDA-MB-231 cells were co-injected with CTSB fibroblasts or vector alone control fibroblasts, into the flanks of nude mice. Note that CTSB fibroblasts promoted a significant ~2.1-fold increase in tumor growth, as compared with control fibroblasts. n = 10 per experimental group. Lv-105 represents fibroblasts transduced with the vector alone control, namely Lv-105 (puro). (B) Quantitation of neo-vascularization, via immunostaining with CD31, did not show any significant increases in tumor angiogenesis. Thus, autophagy in the tumor stroma can promote tumor growth, independently of angiogenesis.

Figure 6.

Fibroblasts overexpressing ATG16L1 show constitutive activation of the autophagic program. (A) ATG16L1 functions as a clathrin adaptor protein during autophagy, facilitating the recruitment of plasma membrane, by driving autophagosome formation via endocytosis. To assess the functional role of stromal ATG16L1 in breast cancer development, we stably overexpressed ATG16L1 in hTERT fibroblasts. (B) Note that ATG16L1 overexpression induces the upregulation of Beclin1 and Lamp1, under basal cell culture conditions. (C) Similarly, increases in cathepsin B and LC3 were also observed after overnight starvation (in Hepes-buffered HBSS). In all three panels, β-actin expression was assessed as a control for equal protein loading. Lv- represents fibroblasts transduced with the vector alone control, namely Lv-105 (puro).

Figure 7. Fibroblasts overexpressing ATG16L1 drive enhanced tumor growth, without a measurable increase in angiogenesis.(A) In vivo, ATG16L1-fibrolasts also promoted tumor growth, resulting in an ~1.6-fold increase in tumor volume, relative to control fibroblasts. n = 10 per experimental group. (B) Increases observed in tumor growth were independent of tumor angiogenesis, as we previously observed for BNIP3, and CTSB. Lv-105 represents fibroblasts transduced with the vector alone control, namely Lv-105 (puro).

Figure 8. BNIP3-, CTSB- and ATG16L1- fibroblasts all show mitochondrial dysfunction: OXPHOS and MitoTracker. (A) We evaluated the functional consequences of autophagic genes expression on mitochondrial activity in fibroblasts, by determining the levels of mitochondrial enzymes associated with the respiratory chain. Note that a strong reduction in key components of complex I, III, and IV was observed, for all the genes examined. Furthermore, in CTSB fibroblasts there is a downregulation of the expression of complex II. Lv- represents fibroblasts transduced with the vector alone control, namely Lv-105 (puro). (B) Note that all three types of autophagic fibroblasts (BNIP3, CTSB, and ATG16L1) show dramatic reductions in MitoTracker staining, indicative of a loss of mitochondrial membrane potential.

Figure 9. BNIP3-, CTSB- and ATG16L1- fibroblasts all show mitochondrial dysfunction, with increased production of L-lactate or ketone bodies. (A) Loss of functional mitochondrial causes changes in cell metabolism, leading to the accumulation of L-lactate. Note that BNIP3- and CTSB fibroblasts both show significant increases in L-lactate production, between 25-to-37%. However, ATG16L1 fibroblasts did not show any increases in L-lactate accumulation. (B, C) Mitochondrial dysfunction can also activate ketone body production, resulting in the accumulation of 3-hydroxy-butyrate. Note that only ATG16L1 fibroblasts showed increases in ketone production, resulting in an up to 2.3-fold accumulation of 3-hydroxy-butyrate. The data were normalized either for cell number or for protein content per well. Thus, BNIP3- and CTSB fibroblasts produce L-lactate, while ATG16L1 fibroblasts produce ketone bodies, as a consequence of autophagy and the resulting mitochondrial dysfunction. (D) Autophagic fibroblasts were co-cultured with MDA-MB-231-GFP cells, and mitochondrial activity was visualized by MitoTracker staining. Note that all three autophagic fibroblast cell lines (BNIP3, CTSB, and ATG16L1) increased the MitoTracker staining (RED; Upper panels) in adjacent MDA-MB-231 cells (GREEN; Lower panels), during co-culture. This is consistent with the notion that autophagic fibroblasts provide mitochondrial fuels, such as L-lactate and ketone bodies, for the anabolic growth of cancer cells.

Figure 10. Autophagic fibroblasts promote experimental metastasis and lung colonization. (A) MDA-MB-231 cells were co-injected with fibroblasts into the tail vein of nude mice. Note that all three autophagic fibroblast cell lines (BNIP3, CTSB, and ATG16L1) enhanced the metastatic capacity of MDA-MB-231 cells, as compared with vector-alone control fibroblasts. BNIP3 fibroblasts and CTSB fibroblasts increased metastasis by ~2.5-fold and ~7-fold, respectively. Interestingly, ATG16L1 fibroblasts showed the largest capacity for increasing metastasis, driving an ~11-fold increase. (B) ATG16L1 fibroblasts also showed increases in myofibroblast markers, such as calponin and vimentin, indicating that autophagy may be also sufficient to promote myofibroblast differentiation. β-actin expression was assessed as a control for equal protein loading.

Figure 11. Autophagic fibroblasts show the induction of p21(WAF1/CIP1), a CDK inhibitor, and morphologic cell hypertrophy. (A) To assess the induction of a senescence phenotype, we examined the expression of CDKI’s (cyclin-dependent kinase inhibitors) and cyclin D1, in all three autophagic fibroblast cell lines. BNIP3 fibroblasts showed an induction of p21(WAF/CIP1) and p16(Ink4a), without any changes in p19(ARF). Conversely, CTSB- and ATG16L1 fibroblasts showed an induction of p21 and p19, with reductions in p16. Thus, the most consistent change observed across all three autophagic fibroblast cell lines was the induction of p21. (B) Senescent cells also appear flatter and/or hyper-trophic, as protein synthesis may continue without cell division. Note that BNIP3-, CTSB-, and ATG16L1 fibroblasts all appear more hyper-trophic (“wider”), as seen by light microscopy 72 h after cell plating.

Figure 12. Autophagic fibroblasts show increased protein mass per cell, and the induction of β-galactosidase activity. (A) A quantitative view of cell hypertrophy was achieved by measuring the total amount of protein per cell. Previously, it was demonstrated that an increase in cell protein content is related to the induction of senescence. Note that BNIP3- and ATG16L1 fibroblasts both showed significant cell hypertrophy (between 30-to-32%), using this approach, while CTSB fibroblasts also showed a trend toward cell hypertrophy. (B) β-galactosidase activity is the gold standard for measuring the onset of senescence. Thus, we quantitatively measured β-galactosidase activity by FACS analysis. Note that all three autophagic fibroblast cell lines showed an increase in β-galactosidase activity, as reflected by an increase in i) the % of β-Gal-positive cells, and ii) β-Gal-Mean-Intensity. (C) Similar results were obtained with more conventional β-Gal-staining methods, such as in BNIP3 overexpressing fibroblasts.

Figure 13. β-galactosidase is predominantly expressed in the tumor stroma of human breast cancer patients. We investigated the compartmentalization of β-galactosidase in a panel of human breast cancer samples, using specific-antibodies directed against β-galactosidase. Note that β-galactosidase is largely confined to the tumor stroma. Thus, expression of β-galactosidase in vivo may be a stromal phenomenon, reflecting the onset of senescence in the tumor microenvironment.

Figure 14.

ATG16L1 expression in human breast cancer cells drives autophagy and reduces tumor growth. (A, B) To understand the compartment-specific role of autophagy in tumor growth, we also stably overexpressed ATG16L1 in human breast cancer cells. ATG16L1 overexpression in MDA-MB 231 cells, by itself, is indeed sufficient to drive autophagy and mitophagy under baseline cell culture conditions, as evidenced by the loss of Cav-1 expression and the upregulation of BNIP3 (B). Similarly, MDA-MB 231 cells harboring ATG16L1 display the upregulation of certain autophagy markers (Beclin1 and CTSB), after overnight starvation (A). (C) Note that ATG16L1 overexpression in MDA-MB-213 cells induces p21 expression, but no significant changes were detected in p19 and/or Cyclin D1 protein levels. (D) We also assessed the capacity of autophagic MDA-MD-231 cells to undergo tumor growth in vivo, after implantation in the flanks of nude mice. Note that ATG16L1 overexpression caused a near 2-fold reduction in tumor volume. (E) However, we did not observe any differences in tumor neo-vascularization, as assessed by quantitation of CD31-positive vessels.

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 in stromal fibroblasts is then sufficient to confer the cancer-associated fibroblast phenotype, resulting in autophagy, mitophagy, and a shift toward aerobic glycolysis. Autophagy also drives the onset of senescence, via the autophagy-senescence transition. Autophagic-senescent fibroblasts then produce high-energy nutrients (L-lactate, ketone bodies, glutamine, and free fatty acids), which “fuel” mitochondrial metabolism (OXPHOS and β-OX; oxidative phosphorylation and β-oxidation) in adjacent cancer cells, resulting in the onset of anabolic tumor growth. This simple model could explain why chronological aging is one of the most significant risk factors for the development of cancer. AST, autophagy-senescence transition.