Molecular Pathways: Trafficking of Metabolic Resources in the Tumor Microenvironment

No potential conflicts of interest were disclosed.

The following editor(s) reported relevant financial relationships: P.S. Steeg reports receiving commercial research grants from Geron, GlaxoSmithKline, and Sanofi.

The members of the planning committee have no real or apparent conflicts of interest to disclose.

Upon completion of this activity, the participant should have a better understanding of how stromal cells in the tumor microenvironment support adjacent tumor cells by supplying metabolic resources, including lactate, amino acids, and free fatty acids. The participant should also gain knowledge on therapeutic approaches aimed at reversing the metabolic reprogramming of the tumor microenvironment.

This activity does not receive commercial support.

The identification of cancer as a genetic disease, compellingly established by the detection of genomic derangements within malignant cells, led researchers to focus on alterations in tumor suppressor genes and oncogenes. However, in the last decade, our genetic view has been expanded by the observation that tumors are thriving organs with multiple cell types within a distinctive extracellular matrix (ECM), and that all these components can affect tumor progression and response to therapies. This view has added significant complexity to the study of human tumors, as it takes into consideration the effects of fibroblasts, mesothelial cells, immune cells, adipocytes, and endothelial cells on tumor growth. During transformation and metastasis, cancer cells recruit these cell types to surround themselves with a supportive tumor microenvironment (TME). Over time, the tumor and the adjacent cells coevolve and even metastasize together (1). Stromal cells are recruited by paracrine growth factors (e.g., PDGF and VEGF) secreted by cancer cells and then, in turn, secrete cytokines (e.g., HGF, TGFβ, and CCL5; refs. 2–4), which accelerate the aggressiveness of cancer cells. The progression of cancer is further supported by the TME, in which low levels of inflammation mediated by immune cells create microenvironmental conditions promoting the invasion of epithelial tumor cells (5). Three-dimensional organotypic cultures using primary cells have allowed modeling of these complex interactions in cell culture (6, 7).

These reciprocal interactions between cancer and stromal cells have been worked out in detail, but minimal emphasis has been placed on the metabolic alterations in the TME. While it is now accepted that cancer cells undergo unique metabolic alterations that facilitate growth (8), most studies of cancer cell metabolism have narrowly focused on changes in the cancer cells, generally ignoring the possible contributions of the TME. However, it has long been known that, in normal tissue, different cell types cooperate to adapt to metabolic demands. For example, adipocytes provide energy to exercising muscle fibers by supplying fatty acids (9), and lactate is shuttled between muscle fibers for use as energy (10).

Given that normal cells have codependent metabolic relationships, that tumors of different origins use different metabolic pathways, and that, even within the same tumor, epithelial cancer cells have different metabolic states (11), it is probable that individual cell types within and surrounding a tumor also have different, interdependent, metabolic states. Our goal is to review the substrates (lactate, amino acids, and fatty acids) that cancer cells use to generate energy, with a focus on how adjacent stromal cells serve as a unique and targetable source of these metabolic building blocks (Fig. 1).

Lactate is an end product of glycolysis during anaerobic metabolism, a gluconeogenic precursor, and a regulator of the cellular redox state. Under steady-state conditions, lactate is either used within cells or secreted into the extracellular space, where it is available for neighboring cells (“cell–cell lactate shuttle”; ref. 12). In exercising muscles, lactate is produced by rapidly contracting fast-twitch muscle fibers through glycolysis and is exported outside of the cell through monocarboxylate transporter (MCT4). Lactate is then either transported to the liver to be utilized as a substrate for gluconeogenesis (“Cori cycle”) or taken up by adjacent slow-twitch oxidative fibers through MCT1 and, as a precursor for glucose, replenishes energy stores.

A similar mechanism has been described in tumor cells (11, 13). Regulated by the O₂ gradient, hypoxic tumor cells distant from blood vessels use anaerobic glycolysis, producing lactate, which is released by the cell through MCT4. Lactate then diffuses along a concentration gradient to the well-oxygenated tumor cells lining the tumor vasculature, enters the cells through MCT1, and is oxidized to produce energy through oxidative phosphorylation. Sonveaux and colleagues in 2008 showed that, when MCT1 is inhibited, the oxidative tumor cells are forced to utilize glucose for energy production and, as glucose is not available, the tumors necrose (11). These findings demonstrated, for the first time, that metabolic cooperation exists between aerobic and hypoxic tumor cells, leading to metabolic heterogeneity within the tumor organ. These results also proved that cancer cells are capable of the opportunistic optimization of available metabolic substrates within their microenvironment to increase tumor survival and growth.

Evidence that trafficking of energy resources also occurs between tumor and stromal cells began to emerge as early as 2006, when colorectal cancers were analyzed immunohistochemically and shown to overexpress lactate dehydrogenase (LDH) 5 and MCT1, suggesting that they metabolize glucose through anaerobic glycolysis and export lactate. Interestingly, adjacent fibroblasts expressed proteins involved in lactate uptake and metabolism (MCT1/2, LDH1) and a low level of a glucose transporter (GLUT1). On the basis of the immunohistochemical expression patterns of LDH5, MCT1/2, and GLUT1 in the cancer cells and fibroblasts, the authors' hypothesis was that the fibroblasts use tumor-derived lactate as an energy source (13). This hypothesis was subsequently supported by findings from Martinez-Outschoorn and colleagues, which demonstrated that cancer cells utilize lactate produced by cancer-associated fibroblasts (CAF). In coculture experiments, MCF7 breast cancer cells induced oxidative stress and activated hypoxia-inducible factor alpha (HIF1α) in adjacent fibroblasts, which resulted in autophagic destruction of the mitochondria and increased anaerobic glycolysis and lactate production by the stromal cells (14, 15). The lactate produced by the CAFs was then utilized by the cancer cells, a phenomenon termed the “reverse Warburg effect” (16). In this bidirectional exchange, the cancer cells actively induce metabolic changes in normal fibroblasts, converting them into CAFs; during this process, TGFβ ligands activate HIF1α and NF-κB and caveolin-1 is downregulated, resulting in impaired mitochondrial function. Because oxidative phosphorylation is reduced, CAFs, in turn, perform aerobic glycolysis and secrete lactate, which is used as an energy source by the cancer cells (17). Caveolin-1 plays a central role in stromal metabolic changes; however, the clinical relevance of caveolin-1 in cancer is unclear. Caveolin-1 is upregulated in breast, ovarian, and hepatic cancer while it is downregulated in pancreatic and renal cancer (18). Similarly, high levels of lactate efflux by stromal cells, as measured by MCT4, have been associated with decreased survival in triple-negative breast cancers (19). On the basis of these combined data, normalization of stromal metabolism and reprogramming of TME metabolic pathways could be a rational way to target tumor growth.

It has long been recognized that glutamine is the amino acid most highly utilized by cancer cells and that many cancer cells are reliant on the presence of exogenous glutamine, a phenomenon termed “glutamine addiction.” Cancer cells use glutamine to replenish tricarboxylic acid cycle (TCA) intermediates (anaplerosis), as a nitrogen donor for nucleotide and amino acid synthesis, and for protein translation (20). In fact, glutamine is at the intersection of genetic, epigenetic, and metabolic aberrations in cancer, as exemplified in an elegant study by Terunuma and colleagues (21). Here, metabolomic profiling of breast tumors and tumor-adjacent normal tissue showed that 2-hydroxyglutarate (2HG) was 4.6-fold higher in tumors than normal tissue, was functionally linked to both glutamine metabolism and MYC activation, and resulted in a global increase in DNA methylation and a poor clinical outcome (21). This study adds to numerous studies indicating a relationship between MYC activation and glutamine utilization (22).

In a process that mirrors the transferring of lactate, cancer cells are also able to obtain glutamine from the TME. Metabolomic profiling of CAFs revealed an overall catabolic phenotype in which CAFs produce several potent metabolic substrates for cancer cell use, including glutamine (23). The hypothesis that cancer cells take advantage of glutamine produced by CAFs is supported by the behavior of MCF7 breast cancer cells cultured with CAFs. In coculture, the cancer cells show (i) reduced glutamine synthesis (measured by expression of glutamate-ammonia ligase; GLUL), (ii) increased glutamine catabolism [measured by expression of glutaminase (GLS) and glutamate dehydroxylase 1 (GLUD1)], and (iii) increased expression of a glutamine uptake transporter (SLC6A14; ref. 24).

In another example of altered amino acid metabolism, both cancer and stromal cells (tumor-associated macrophages and CAFs) increase consumption of tryptophan and arginine through modified expression of enzymes, including indoleamine 2,3-dioxygenase, tryptophan 2,3-dioxygenase, arginase, and nitric oxide synthase (25–27). The depletion of tryptophan and arginine not only results in the production of tumor-promoting metabolites (e.g., kynurenine, ornithine, and nitric oxide) that have known roles in migration, invasion, and cell survival, but also suppresses activation of T cells and immunosurveillance (25, 27, 28). In a second example, p62, an adaptor protein for the atypical protein kinase C serine/threonine kinases, has been shown to function as a tumor suppressor in the stroma of prostate cancer, through a mechanism requiring cystine and glutamine uptake as a result of mTORC1/c-myc activation. Loss of p62 in the tumor microenvironment reduced mTORC1/c-myc, impaired expression of key cystine and glutamine transporters, and reduced NADPH production through the pentose phosphate pathway, changes that resulted in impaired glutathione (GSH) production. This metabolic reprogramming promoted tumor growth through production of IL6 and TGFβ and maintenance of the CAF phenotype (29).

Stromal cystine may also promote drug resistance in chronic lymphocytic leukemia (CLL; ref. 30). Mesenchymal stromal cells take up cystine and convert it to cysteine, which is released into the microenvironment and taken up by CLL cancer cells for GSH synthesis. High GSH expression in the CLL tumor cells augments survival and reduces drug-induced cytotoxicity. This effect of GSH in cancer cells is not consistent with the effects of p62 loss in stromal cells described above, which indicates an association between impaired GSH production and the promotion of tumor growth. Such inconsistencies demonstrate the complex and at times contradictory metabolic effects in different components of the TME, which might also be very tumor type specific. Overall, what has emerged from the studies described above is yet another example of the metabolic coupling of cancer cells and the tumor stroma, in which the TME serves as an essential and renewable source of several types of amino acids that are used as metabolic building blocks for cancer cells.

Adipocytes are a major constituent of the TME in renal cell, breast, and ovarian cancer, but, until recently, they have been considered passive bystanders to cancer progression. Reports in breast (31) and ovarian (32) cancer demonstrated that normal tumor-adjacent adipocytes undergo at least three functional changes that support tumor growth. First, in the presence of cancer cells, adipocytes increase secretion of inflammatory cytokines (i.e., IL6 and IL1β), proteases (i.e., MMP-11/stromelysin-3), and plasminogen activator inhibitor-1 (PAI-1). Second, dedifferentiation of the adjacent adipocytes occurs, as evidenced by a loss of terminal differentiation markers. In a positive feedback loop, MMP-11/stromelysin-3 further promotes adipocyte dedifferentiation (33). Third, there are fewer lipid droplets in adipocytes, indicating decreased lipid accumulation and/or delipidation. These functional changes characterize the transformation of normal adipocytes into “cancer-associated adipocytes” (CAA), which are primed to supply energy, in the form of fatty acids, to adjacent cancer cells.

One of the first clues that cancer cells obtain energy substrates from adjacent adipocytes came from the observation that, in both patient samples and in vitro coculture experiments, ovarian cancer cells adjacent to adipocytes not only accumulate lipids but also have an increased rate of fatty acid β-oxidation (32). Upon further investigation, activation of AMP-activated protein kinase (AMPK) was noted in cancer cells cultured with adipocytes. AMPK is a central metabolic sensor, that, when activated (phosphorylated), induces energy producing processes, including fatty acid oxidation. It is evident that a similar phenomenon occurs in prostate cancer as the translocation of lipids from adipocytes to prostate cancer cells has been visualized by Fourier transform infrared spectroscopy (34). Interestingly, prostate cancer cells use almost no glucose and rely almost entirely on fatty acid oxidation for energy production (35), resulting in the limited utility of F18-2DG PET imaging in patients with well-differentiated prostate cancer (36).

While adipose tissue is predominantly comprised of adipocytes, it also contains endothelial cells, macrophages, and fibroblasts. This amalgam of stromal cells acts as an endocrine organ, regulating energy homeostasis in the TME through the secretion of adipokines and inflammatory cytokines, a process that seems to be particularly relevant in breast and ovarian cancer (37). Cancer-associated cachexia is the clinical manifestation of the exploitive relationship between cancer cells and adipose tissue. In patients with advanced cancer, cachexia is a result of adipose atrophy induced by increased lipolysis in adipocytes. Accordingly, mice lacking the ability to increase lipolysis [deleted hormone sensitive lipase (Hsl) or adipose triglyceride (Atgl) genes] do not develop cachexia (38). These findings regarding HSL and ATGL in cachexia suggest that the metabolic exchange between stromal and cancer cells can be uncoupled and raises the possibility of novel metabolically targeted therapeutic approaches for cancer.

Broadly, therapeutic approaches aimed at reversing metabolic reprograming in the TME include drugs that inhibit the cancer cell's ability to seize the metabolism of stromal cells or drugs that interrupt the cancer cell's ability to use the energy resources produced by the stromal cells (Table 1).

Treatments aimed at inhibiting the cancer cell's ability to co-opt the metabolism of stromal cells are primarily focused on stopping the induction of CAF and CAA phenotypes. An antimalarial drug, chloroquine, blocks oxidative stress and autophagy in fibroblasts, preventing the production of high energy mitochondrial fuels and disrupting the ability of cancer cells to induce a CAF metabolic phenotype (39). On the basis of promising preclinical results, several clinical trials are underway to determine whether chloroquine is an effective adjuvant treatment for cancer (40). The challenge of chloroquine use as a cancer therapeutic is the paradoxical role of autophagy in cancer cells. Specifically, autophagy can promote tumor growth in some contexts and suppress tumor growth in other contexts; therefore, it is possible that the antiautophagic effect of chloroquine could stimulate cancer growth.

Therapies aimed at interrupting the transformation of adipocytes into CAAs by inhibiting dedifferentiation include a class of antidiabetic drugs, thiazolidinediones. Thiazolidinediones are ligands for the transcription factor, peroxisome proliferation-activated receptor (PPARγ), which regulates the terminal differentiation of adipocytes (41). In liposarcoma, a tumor characterized by dysfunctional adipocyte differentiation, clinical testing demonstrated that a thiazolidinedione, troglitazone, inhibited dedifferentiation of adipocytes (42). Unfortunately, thiazolidinedione use has been associated with cardiovascular side effects and bladder cancer, resulting in the withdrawal of two thiazolidinediones, rosiglitazone and pioglitazone, from the market. These adverse events may hamper further development of thiazolidinediones as cancer therapeutics.

Effort has also been made to develop therapies aimed at interrupting the ability of cancer cells to use the energy resources produced by the stroma. The strategy of blocking the energy substrates of cancer cells has a precedent in pediatric acute lymphoblastic leukemia (ALL), as l-asparaginase treatment kills ALL cells by depriving the cells of asparagine and glutamine. Unfortunately, significant toxicity prohibits use of the medication in adults (43). Another agent, sodium phenylbutyrate, which decreases plasma glutamine, has undergone phase I testing for the treatment of solid tumors (44, 45) but awaits advanced trials to test its clinical efficacy. An inhibitor of MCT 1/2 (AZD3965), takes aim at lactate transfer between tumor cells and tumor/stromal cells and is in phase I trials (46). In addition, the pyruvate dehydrogenase kinase inhibitor, dichloroacetate, reduces lactate production and may be utilized to interfere with the reverse Warburg effect (47). Interruption of the use of free fatty acids by cancer cells is being attempted using fatty acid binding protein-4 (FABP4) inhibitors and metformin. The inhibition of FABP4, which reversibly binds long chain fatty acids, was shown to have a protective effect in ovarian cancer in vitro and in mouse models (32).

At least one therapy, metformin, may work by both interrupting the cancer cell–induced metabolic changes in the stroma and blocking cancer cell use of stromal supplied energy. Metformin is a commonly used treatment for diabetes that has been associated with improved cancer outcomes in epidemiologic studies (48) and is currently undergoing phase II/III testing as adjuvant treatment in several cancer types. In the TME, metformin has been shown to block adipocyte-mediated lipid accumulation in ovarian cancer cells (49) and to restore caveolin-1 expression in fibroblasts, reversing the CAF phenotype induced by cancer cells (15). If metformin proves to attenuate protumorgenic metabolic changes in the TME, it might also be repurposed in the primary prevention of cancer, as it is likely that these changes are important in the early steps of carcinogenesis. Moreover, as metformin has an excellent safety profile, its use can be justified in those at high risk, but without evidence of disease.

Recent reports indicating that targeted therapies lead to a metabolic response of the entire tumor organ extend the concept of metabolic interdependence in the TME even further. In a preclinical study, treatment of various tumors with receptor tyrosine kinase (RTK) inhibitors (sunitinib, sorafenib) induced anaerobic glycolysis and lactate production in cancer, stromal, and endothelial cells (50). When the RTK inhibitor treatment was stopped, the metabolism of the tumor organ adapted again by reducing glycolysis and shifting to an increased TCA, leading to lipogenesis and rapid tumor regrowth. The mechanistic insight gained from this reversal was exploited by combining RTK inhibitor treatment and treatment with a fatty acid synthase inhibitor (orlistat); the combination abrogated tumor regrowth and metastasis (50). As outlined above, significant efforts are under way to develop therapeutics that target the recruitment of metabolic substrates from tumor stroma, representing a new frontier in cancer drug discovery.

The study of stroma–tumor metabolic interactions, while still in its infancy, has added yet another level of complexity to our efforts to understand the tumor organ. While untangling the metabolic pathways regulated in cancer and stromal cells appears to be a formidable task, several common themes have emerged:

1. Cancer cells impose a self-serving metabolic program on stromal cells, recruiting them to supply energy substrates that support tumor survival and aggression.

2. The metabolic networks in both cancer and stromal cells have high plasticity and are capable of fast temporal and spatial metabolic adaptation based on changing environmental cues.

3. Stromal cells detoxify the tumor microenvironment to reduce cancer cell apoptosis/autophagy and provide nutrients.

A feasible next step in delineating the complex metabolism of the TME is to use advanced untargeted mass-spectrometry profiling approaches as well as flux analysis, coupled with sophisticated bioinformatics analysis and standardization (51, 52). These approaches will allow us to obtain an integrated picture of the metabolic changes within the different tumor compartments. It is evident that we must study the tumor organ as a functional metabolic domain if we are to understand the metabolic dependence of cancer cells on the TME and identify novel therapeutics that are able to slow/limit tumor growth.

Conception and design: I.L. Romero, A. Mukherjee, L.M. Litchfield, E. Lengyel

Development of methodology: I.L. Romero, A. Mukherjee, L.M. Litchfield, E. Lengyel

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): I.L. Romero, A. Mukherjee, L.M. Litchfield, E. Lengyel

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): I.L. Romero, A. Mukherjee, L.M. Litchfield, E. Lengyel

Writing, review, and/or revision of the manuscript: I.L. Romero, A. Mukherjee, H.A. Kenny, L.M. Litchfield, E. Lengyel

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): E. Lengyel

Study supervision: I.L. Romero, E. Lengyel

The authors thank Gail Isenberg for editing the article.

I.L. Romero was supported by the Eunice Kennedy Shriver National Institute of Child Health and Human Development of the NIH under award number K12HD000849 and the American Board of Obstetrics and Gynecology. E. Lengyel was supported by the NCI of the NIH under award numbers R01CA111882 and R01CA169604, and a grant from Bears Care, the charitable beneficiary of the Chicago Bears Football Club.