The metabolic cross-talk between epithelial cancer cells and stromal fibroblasts in ovarian cancer progression: Autophagy plays a role.

The impressive and intriguing similarity in the cellular and molecular composition of the stroma formed during the healing of a wound and of the stroma surrounding epithelial cancer has been known for a long time. The wound repair by secondary intention is accompanied by a sequence of events in the stroma that include clotting, neovascularization, recruitment of macrophages and lymphocytes, activation of myofibroblasts, and release of pro‐inflammatory cytokines (e.g., TNF‐α, TGF‐β, IL‐6, and IL‐1β). This sequence of events causes an intense remodeling of the ECM, with dynamic alternation of degradation and synthesis of fibrous proteins, eventually leading to the formation of a scar. Strikingly, a similar scenario occurs during the evolution of infiltrative epithelial tumors, with the exception that cancer cell proliferation and necrosis maintain a pro‐inflammatory environment that eventually results in a desmoplastic stroma. 40 This analogy has suggested to Dvorak the paradigm that “cancer is a wound that never heals.” 40 , 41 This paradigm perfectly fits with the inflammatory process that associates with ovarian cancer development and progression. 42 , 43 , 44 At each ovulation, the follicle wall ruptures to release the ovum and is thereafter repaired through controlled inflammatory events much alike the wound healing. 45 During this process, macrophages and fibroblasts are recruited to the wounded epithelial surface, and an enormous amount of cytokines/chemokines and matrix‐remodeling enzymes (including prostaglandins, bioactive eicosanoids, plasminogen activators, collagenases, interleukins, 33 tumor necrosis factor α [TNF‐α], and various growth factors) are released on site. 46 , 47 , 48 Therefore, the ovarian surface epithelial cells adjacent to the site of ovulation are exposed to an inflammatory and oxidative environment that enhances the risk of malignant transformation. 49 , 50 In addition, the chemokines and cytokines released in this context attract and promote the adhesion of extra‐ovarian malignant cells to the ovary. 51 Thus, the enhanced risk of ovarian carcinogenesis associated with repetitive ovulation would arise from the continuous creation of an inflammatory microenvironment that favors either the local malignant transformation or the homing of extra‐ovarian malignant cells and, thereafter, cancer progression. Conceivably, factors other than inflammation, for instance the balance between estrogens and progesterons, are also involved in ovarian tumorigenesis. 52 , 53 Yet, it is a fact that inflammation in the peritoneal cavity determines a stromal environment permissive for ovarian cancer progression. 54 , 55 , 56 , 57

Active CAFs are marked by the characteristic expression of alpha‐smooth muscle actin (ASMA) and fibroblast‐activated protein (FAP). 58 Studies conducted in a multitude of carcinomas have shown that CAFs may originate from different sources. 59 CAFs have been reported to derive from: (1) the conversion of fibroblasts locally present in the ECM, 60 (2) the differentiation of bone marrow derived precursor cells, 61 (3) the trans‐differentiation of malignant epithelial cells (through EMT, epithelial‐mesenchymal transition) 62 or of endothelial cells (through EndMT, endothelial‐mesenchymal transition), 63 and (4) the differentiation of mesenchymal stem cells (MSCs). 64 , 65 Ovarian cancer cells release chemotactic cytokines and growth factors that recruit and activate the cells in the stroma (Fig. 1 ). The conditioned medium of ovarian (SKOV3) cancer cells, as well as TGF‐β1, could induce the trans‐differentiation of stromal fibroblasts into ASMA‐expressing myofibroblasts. 66 The conversion of adipose‐derived and bone marrow derived MSCs and of peritoneal fibroblasts into CAFs was observed in a xenograft transplant of human ovarian cancer SKOV3 cells transgenically expressing the homeobox gene HOXA9. 67 This effect was mediated by the release into the peritoneum of TGF‐β2 driven by HOXA9 from the ovarian cancer cells. 67 Interestingly, the concomitant transplantation of ovarian cancer cells expressing transgenic green fluorescent protein could exclude the conversion of these cells into CAFs, which suggests that generation of CAFs through EMT of epithelial ovarian cancer cells is a rare event. 67 The conversion of adipose‐derived MSC into CAFs within ovarian cancer stroma can be induced also by lysophosphatidic acid (LPA) released by the ovarian cancer cells. 64 An important issue is whether the characters acquired by CAFs are stably maintained also in the absence of malignant cells, or whether these phenotypic changes are reversible. Both genetic and epigenetic factors could be involved in stabilizing the phenotype of CAFs. However, loss of heterozygosity and alterations of chromosomal copy number were found extremely rare in CAFs from ovarian cancer cells. 68 It is therefore likely that the activated phenotype of CAFs is maintained through the epigenetic regulation of gene expression. In support of this hypothesis, a distinct pattern of DNA methylation has been reported in CAFs isolated from breast and prostate cancers. 69 , 70 Such a study has yet to be done in CAFs isolated from ovarian cancer. For sure, micro‐RNAs play a big role in modulating the phenotypic characters of fibroblasts in the ovarian cancer microenvironment. 71 Differences in the expression of a subset of 11 micro‐RNAs have been reported between normal omental fibroblasts and CAFs from omental tumors, with miR‐31 and miR‐214 being the most downregulated and miR‐155 being the most upregulated in the latter. 72 Interestingly, playing with the transfection of specific miRNAs and anti‐miRNAs to mimic this deregulation, it was possible to induce a functional conversion of normal fibroblasts into CAFs and vice versa. 72 Origin and tumorigenic roles of ovarian cancer associated fibroblasts Note : Ovarian cancer cells secrete cytokines and soluble factors that induce the recruitment, differentiation, and activation of cancer‐associated fibroblasts (CAFs). The latter secrete a set of soluble factors, including extracellular matrix components and cytokines, that affect the behavior and the fate of ovarian cancer cells (for abbreviations, see the text). The cytokines released in the cancer stroma elicit both paracrine and autocrine stimulations on cancer cells, fibroblasts, immune cells, and endothelial cells, thus amplifying and perpetuating the effects.

A few studies link the activity of cytokines with the autophagy process in the CAFs. Starvation‐induced differentiation of fibroblasts in myofibroblasts, as marked by the de novo synthesis of ASMA and increased expression of stress fibers, strictly depended on induction of autophagy. 139 Similarly, one can hypothesize that the differentiation of stromal fibroblasts into myofibroblasts induced by TGF‐β1 and other cytokines secreted by ovarian cancer cells 66 occurs through the upregulation of autophagy. In this same line, it is interesting observation that IL‐1β secreted by ovarian cancer cells attenuates the expression of p53 in neighboring CAFs, 31 given that cytoplasmic p53 is a known downregulator of autophagy. 140 As said before, sustained autophagy (as induced by IL‐2) is necessary for myofibroblasts survival in the stroma. 138

Autophagy is greatly influenced by an array of factors in the tumor microenvironment, such as hypoxia, pH, oxidative stress, ammonia, glucose and amino acid availability, cytokines, hormones, and growth factors. 141 , 142 , 143 , 144 , 145 , 146 , 147 The physical interaction of tumor cell with surrounding cells (inflammatory cells, fibroblasts) in the matrix also influences the autophagy compliance, and consequently also the survival or death, of the tumor cell. In addition, the cytokines released by both CAFs and ovarian cancer cells have an impact on the composition of the stroma by recruiting other cells, and thus contribute to create a microenvironment that ultimately affects the regulation of autophagy in ovarian cancer cells. It has been proposed that the functional cross‐talk between tumor cells and stromal cells through the exchange of soluble factors finally results in the reprogramming of the latter toward a metabolic state that is permissive for the growth and metastasization of the cancer cells. 148 In this respect, autophagy may represent the target and the pivotal driver at the same time of such metabolic reprogramming. Lisanti and colleagues have recently proposed a new paradigm of the mutual interaction between CAFs and cancer cells, in which the former supply energetic metabolites to the latter. 26 , 149 In such situation, autophagy in cancer cells would be maintained at low basal level, while autophagy in CAFs would be upregulated. A number of inflammatory cytokines present in the stroma and the hydrogen peroxide released by ovarian cancer cells could induce autophagy in CAFs. 150 , 151 As a result, CAFs would fuel the metabolism of neighboring cancer cells with autophagy metabolites such as amino acids, fatty acids, ketones, and lactate, 26 , 149 , 152 , 153 thus supporting the growth and propagation of cancer cells while keeping autophagy at a minimal level. In this model, the so‐called “Warburg effect,” that is, the aerobic glycolysis, occurs in CAFs rather than in cancer cells, which suggested the term “reverse Warburg effect.” 154 , 155 In a similar fashion, in pancreatic cancer the tumor‐induced autophagy in CAFs leads to the secretion of alanine, which outcompetes glutamine, in turn, to fuel the cancer cells in low‐glucose microenvironment. 25 This same scenario is likely to occur also in the case of ovarian cancer. 156 Thus, an integrated model of the interplay between CAFs and ovarian cancer cells cannot disregard the modulation of autophagy in both the cell types (see Fig. 3 ). Our unpublished data support the view that high infiltration of CAFs associates with a low level of Beclin‐1‐dependent autophagy in ovarian cancer cells. We have previously reported that the patients bearing an ovarian cancer highly expressing autophagy‐active Beclin‐1 could experience a better prognosis as compared to those bearing an autophagy‐defective cancer. 120 Also, we have shown that ovarian cancer cell migration induced by IL‐6 occurs via downregulation of autophagy. 23 Taken together, it is tempting to speculate that poor prognosis in ovarian cancer patients is due to the downregulation of autophagy in cancer cells operated by CAFs, as predicted in the model proposed by Lisanti and colleagues. If this interpretation is correct, to be effective the therapeutic strategies targeting autophagy should consider the different modulation of autophagy in CAFs and in cancer cells as influenced by the microenvironmental conditions. 24 The reciprocal regulation of autophagy between cancer cells and cancer‐associated fibroblasts Note : The cytokine‐mediated cross‐talk and the exchange of metabolites between cancer cells and CAFs reciprocally regulate the actual level of autophagy in the cells. Cytokines, reactive oxygen species (ROS), and ammonia released by cancer cells as well as hypoxia (which triggers the HIF1α pathway) and lack of nutrients in the tumor microenvironment induce autophagy in CAFs, which leads to loss of mitochondria (mitophagy) and of Caveolin 1. The latter events favor glycolysis. Because of increased autophagy and glycolysis in CAFs, metabolites (including glutamine), lactic acid, and ketones are supplied to cancer cells. As a result, in these cells autophagy is downregulated, while anabolism is stimulated, with consequent growth of cancer.

The tumor stroma is characterized by intense infiltration of reactive immune and inflammatory cells, neoformation of a nonfunctional network of blood vessels, recruitment of fibroblasts, and neodeposition of collagen and fibrin, besides other ECM proteins. In such a reactive stroma (also referred to as desmoplastic), fibroblasts, either locally resident or recruited from other anatomical sites, undergo a pronounced alterations in the phenotype and expression profile of cellular and secreted proteins. A complex stromal–epithelial interaction reciprocally influences the dynamic changes in the structure and composition of ECM and of tumor stroma, and alterations in this interaction play an important role in the development of cancer. 157 In vivo, the actual level of autophagy in cancer cells is different in the different areas of the tumor considered, depending on the local level of nutrients, oxygen, and on the mixture of soluble factors, as provided by the vasculature and stromal composition. Given the pathophysiological role of autophagy in the regulation of cell survival and cell death in response to metabolic and genotoxic stresses, it appears clear that, by influencing the level of autophagy in the cancer cells, the stroma composition indirectly contributes to the onset of chemo resistance and dormancy in ovarian cancer cells, two conditions that negatively impact on prognosis. CAFs could release exosome in the stromal compartment to deliver metabolic substrates and micro‐RNAs to cancer cells, thus reprogramming the energetic metabolism of the latter. 158 The reciprocal stimulation through physical contact and soluble factors between CAFs and cancer cells not only reprograms the expression of secretory factors, but also modifies in a dynamic way the metabolism in both cell types. In brief, we may assume that the stromal microenvironment acts as an epigenetic modifier of autophagy, with a fallout on the behavior of malignant cells and, consequently, the prognosis of cancer's patients. Presently, autophagy is a target process in the therapy of cancer 159 and of ovarian cancer in particular. 160 With increasing recognition of the role of CAFs in carcinogenesis and progression, it is believed that targeting the metabolic cross‐talk between epithelial cancer cells and CAFs could open novel therapeutic strategies to fight ovarian cancer. 161 Potential therapeutic strategies presently under investigation include, among others, inhibitors of the signaling pathways triggered by TGF‐β and HIF‐1α, two known inducers of autophagy in CAFs (Fig. 3 ). 162 , 163 Thus, learning how the actual level of autophagy in cancer cells and in stromal cells modulated by the tumor microenvironment may help to predict whether inhibitors or inducers of autophagy will achieve a long‐term therapeutic benefit.

Ovarian cancer is the seventh most common cancer in women, with the incidence of around 9.4 and 5.0 age‐standardized rate per 100,000 in developed and developing countries, respectively. 1 , 2 Despite the progress in drug discovery and improvement in the management, ovarian cancer remains the leading cause of death from gynecological cancer. 2 , 3 Worldwide, as many as 240,000 women are diagnosed with ovarian cancer, and approximately half of them die each year. 2 Early diagnosis could give a high probability of recovery, yet it is difficult because of the unspecificity of the symptoms in the early phases of the disease. On the other hand, late stage ovarian cancers are aggressive, featuring rapid growth, dissemination, chemo/radio resistance, and recurrence. 4 Ovarian cancer progression denotes not only the emergence of more aggressive clones, but also reflects the dynamic changes in the microenvironment, in its cellular and molecular composition, as well as in the heterotypic interactions between tumor and stromal cells. 5 Alike other solid tumors, the ovarian cancer tissue is composed of epithelial‐derived cancer cells embedded in a tumor stroma, which consists of heterogeneous cell types and a mixture of amorphous components. 6 The latter forms the extracellular matrix (ECM) and includes structural and specialized proteins and proteoglycans. Various cell types are found in the stroma of ovarian cancer, including immune and inflammatory cells (such as lymphocytes, macrophages, and mastocytes), endothelial cells, adipocytes, and the “cancer‐associated fibroblasts” (CAFs). 7 , 8 The latter definitely represent the major cell component in the activated stroma. For long time the potential role in carcinogenesis of stromal cells has been neglected, as they were regarded just as part of an inflammatory reaction induced by necrotic cancer cells. It is now recognized that the stroma composition and architecture, in terms of vascularization, type of cells, and of their secretion, play a role in the establishment and progression of cancer cells. This fact was appreciated since the “seed and soil” paradigm of metastasis formation formulated by Stephen Paget, in which the stroma (the soil) is a determinant factor in allowing the cancer cell (the seed) to take root. 9 , 10 It is now well established that the stroma contributes to ovarian tumorigenesis and progression. 11 , 12 An important notion is that the stroma itself, in terms of cell and molecular composition and architecture, is subject to dynamic changes that parallel the growth and progression of the tumor, which suggests that the two compartments coevolve and influence each other during cancer progression. 13 The bidirectional communication between stromal and ovarian cancer cells has an impact on the metabolism, and consequently on the behavior, of the various actors. 14 Autophagy is one such metabolic process that occurs in both cancer and stromal cells and that is influenced by and in turn influences the microenvironment. 15 Autophagy is a lysosome‐driven process of macromolecules and organelles degradation that plays a fundamental role in cell and tissue homeostasis. 16 In cancer cells, autophagy is clearly deregulated and contributes to abnormal growth and development of metastasis and of immune, radio, and chemo resistance. 17 , 18 Autophagy plays a role in cancer stemness 19 , 20 and in cell reprogramming, 21 as well as in tumor cell dormancy 22 and in tumor invasion. 23 In the last decade, it has become evident that autophagy in stromal cells, particularly in CAFs, adipocytes, and immune cells also has a role in the carcinogenesis process, since it is involved in the secretion of cytokines and of other soluble factors that impinge on the metabolism of epithelial cancer cells. Recent studies support the view that a “metabolic symbiosis” exists between CAFs and cancer cells. In this model, cancer cells induce a rise of autophagy in CAFs, which in turn provide the cancer cells with energetic metabolic substrates and so reducing their autophagy needs. 24 , 25 Thus, the metabolic cross‐talk between stromal and cancer cells reciprocally affects autophagy regulation, 26 which might reflect in behavioral changes driving cancer relapse and metastasis as we will see later. In this review, we present the “state of art” of the current knowledge on the role of stromal cells (particularly CAFs) and of their cytokines in the development and progression of ovarian cancer, with an emphasis on the role played by autophagy in the cross‐talk between epithelial cancer cells and CAFs. This latter aspect has not been studied extensively in ovarian cancer, yet the data available support the view that, as demonstrated in other cancer models, autophagy is differentially regulated in ovarian cancer and in CAFs. Understanding how CAFs and ovarian cancer cells reprogram each other's metabolism and, in particular, how autophagy is modulated in both these cells might shed lights on novel signaling pathways that could be targeted for the treatment of ovarian cancer.

The contribution of CAFs in cancer progression has been widely discussed in many different cancer models, 73 , 74 , 75 , 76 including ovarian cancer. 5 , 8 , 31 , 77 Here, we will briefly review the principal ways through which CAFs promote ovarian cancer progression (see also Fig. 1 ). Ovarian CAFs express and secrete, among others, high levels of IL‐6, cyclo‐oxygenase 2 (COX‐2), and of chemokine (CXC motif) ligand CXCL‐1. 31 , 78 The latter promotes cell proliferation by binding to its receptor CXCR2, which is highly expressed on ovarian cancer cells. 79 Ovarian CAFs could facilitate cancer cell spreading and invasion by releasing the hepatocyte growth factor (HGF), 80 and collagenolytic proteases such as the Matrix Metalloprotease MMP‐2 81 , 82 and the urokinase‐like activator of the plasminogen (uPA). 83 The numerous cytokines and soluble factors released by CAFs reported to promote ovarian cancer progression are listed in Table 1 . Cytokines and soluble factors secreted by CAFs and proposed activity in ovarian cancer Promotes cell proliferation and angiogenesis Induces EMT, invasion and metastasis Induces MMPs production Induces cell growth, cell motility, and angiogenesis Induces cancer invasion, dissemination in the peritoneal cavity, and metastasis CXCR4‐dependent pathway Cross‐talk with VEGF to induce strong angiogenesis MMP‐2, MMP‐9 MT1‐MMP Direct effect in angiogenesis Cross‐talk with MMPs and CXCL12 Induces expression of EZX2 leading to cancer cell migration VEGFR‐2‐dependent pathway: RAS/Raf/MAPK, PLC‐γ, PI3K/AKT Synergistic angiogenic effects Note : CCL5, cisplatin‐induced chemokine (C‐C motif) ligand 5; TGF‐β, transforming growth factor β; VCAN, versican; MFAPA5, microfibrillar‐associated protein; NPPB, natriuretic peptide B; NPR1, NPPB receptor; FAP, fibroblast activation protein 1α, SDF‐1, stromal‐derived factor‐1 or CXCL‐12, CXCL‐11, CXCL‐1; IL‐6, interleukin 6; EZH2, enhancer of zeste homologue 2; LPA, lysophosphatidic acid; ECM, extracellular matrix. The secretion of versican in the ECM is another means through which CAFs may promote ovarian cancer motility, spreading, and invasion. Under TGF‐β stimulation, CAFs release high amount of versican, along with MMP‐9, resulting in increased aggressiveness of ovarian cancer. 84 CAFs are found abundantly in the proximity of the neo‐formed blood vessels. It has been demonstrated that under LPA stimulation, the active fibroblasts express and release the CXC chemokine ligand 12/stromal cell derived factor 1α (SDF‐1α), the Vascular Endothelial Growth Factor (VEGF‐A), and IL‐6, which promote the recruitment of endothelial cells and the angiogenic sprouting in the tumor context. 85 These events clearly favor the growth and metastasization of ovarian cancers. 86 Finally, CAFs contribute to cancer progression also by suppressing the immune response. It has been reported that a high ratio of tumor infiltrating T‐cytotoxic versus T‐regulatory (Treg) lymphocytes associates with a better prognosis in ovarian cancer patients. 87 , 88 Conversely, the presence of Treg lymphocytes in the tumor microenvironment of ovarian cancer tissues creates an immunosuppressive condition that negatively affects the patient's survival. 89 , 90 Besides ovarian cancer cells, also the CAFs release immune‐modulatory cytokines that eventually suppress or limit the immune response 8 , 57 (Fig. 2 ). The cross‐talk between ovarian cancer cells and cancer‐associated fibroblasts in the regulation of autophagy and its role in cancer progression Note : Autophagy in cancer cells opposes malignant progression and promotes dormancy. Cytokines and other factors secreted by ovarian cancer cells activate autophagy in CAFs, which then secrete a vast array of soluble factors including proteases, cytokines, growth factors, and metabolites that ultimately create a microenvironment favorable to cancer growth, metastasization, and onset of chemoresistance. The inflammatory stroma could also create the conditions for awakening the dormant cancer cells, thus favoring cancer relapse (for abbreviations and a detailed description, refer to the text).