Nutrient metabolism and cancer in the in vivo context: a metabolic game of give and take

Abstract

Nutrients are indispensable resources that provide the macromolecular building blocks and energy requirements for sustaining cell growth and survival. Cancer cells require several key nutrients to fulfill their changing metabolic needs as they progress through stages of development. Moreover, both cell‐intrinsic and microenvironment‐influenced factors determine nutrient dependencies throughout cancer progression—for which a comprehensive characterization remains incomplete. In addition to the widely studied role of genetic alterations driving cancer metabolism, nutrient use in cancer tissue may be affected by several factors including the following: (i) diet, the primary source of bodily nutrients which influences circulating metabolite levels; (ii) tissue of origin, which can influence the tumor's reliance on specific nutrients to support cell metabolism and growth; (iii) local microenvironment, which dictates the accessibility of nutrients to tumor cells; (iv) tumor heterogeneity, which promotes metabolic plasticity and adaptation to nutrient demands; and (v) functional demand, which intensifies metabolic reprogramming to fuel the phenotypic changes required for invasion, growth, or survival. Here, we discuss the influence of these factors on nutrient metabolism and dependence during various steps of tumor development and progression.

This review discusses the changes in the nutrient metabolism and dependence during various steps of tumor development and progression.

Glossary

AIM

Apoptosis inhibitor of macrophage

ALL

Acute lymphoblastic leukemia

APC

Adenomatous polyposis coli

ARG1

Arginase 1

BCAA

Branched‐chain amino acids

CAFs

Cancer‐associated fibroblasts

ccRCC

Clear cell renal cell carcinoma

CD36

Cluster of differentiation 36

CRC

Colorectal cancer

CTC

Circulating tumor cell

CXCL16

Chemokine (C‐X-C motif) ligand 16

ECM

Extracellular matrix

EMT

Epithelial‐to-mesenchymal transition

FABP4

Fatty acid‐binding protein 4

FADS2

Fatty acid desaturase 2

GLS1

Glutaminase1

H3K27me3

Trimethylation of histone 3 lysine 27

HCC

Hepatocellular carcinoma

HFD

High‐fat diet

HIF1a

Hypoxia‐inducible factor 1‐alpha

IDO1

Indoleamine 2,3‐dioxygenase

KRAS

Kirsten rat sarcoma oncogene

lncRNA

Long non‐coding RNAs

LUAD

Lung ductal adenocarcinoma

MCT

Monocarboxylate transporter

MDSCs

Myeloid‐derived suppressor cells

MET

Mesenchymal‐to-epithelial transition

MET

MET proto‐oncogene (Hepatocyte growth factor receptor)

MITF

Microphthalmia transcription factor

mTOR

Mammalian target of rapamycin

MYC

MYC proto‐oncogene (Myelocytoma)

NK

Natural killer cell

NSCLC

Non‐small‐cell lung carcinoma

OXPHOS

Oxidative phosphorylation

PC

Pyruvate carboxylase

PD‐1

Programmed cell death protein 1

PDAC

Pancreatic ductal adenocarcinoma

PDK1

Pyruvate dehydrogenase kinase‐1

PDX

Patient‐derived xenograft

PGC‐1α

PPAR‐gamma coactivator 1 alpha

PHGDH

Phosphoglycerate dehydrogenase

PML

Promyelocytic leukemia

PSCs

Stroma‐associated pancreatic stellate cells

PTEN

Phosphatase and tensin homolog

ROS

Reactive oxygen species

SREBP

Sterol regulatory element‐binding protein

TCA

Tricarboxylic acid cycle

TGF‐β

Transforming growth factor‐beta

TIGAR

TP53‐induced glycolysis regulatory phosphatase

VHL

von Hippel–Lindau tumor suppressor

Introduction

Over time, tumor cells need to adapt to the changing circumstances imposed by the different stages of cancer development (initiation, proliferation, invasion, and metastasis formation) and their surrounding environment to successfully sustain growth demands. One of the most fundamental aspects of this adaptation is the extensive metabolic rewiring that tumor cells undergo in order to support cell survival and proliferation (Vander Heiden & DeBerardinis, 2017). Growing evidence shows that cellular metabolism is thoroughly reprogrammed in cancer cells, enabling tumor initiation, promotion, and progression (Vazquez et al, 2016; Lunt & Fendt, 2018; Kreuzaler et al, 2019). The process of metabolic reprogramming can be highly influenced by nutrient availability to tumor cells at the local environment. Cancer cells residing in different organs have access to different nutrients and their metabolites. Both the presence and the abundance of those will determine the metabolic phenotype that cancer cells adapt to overcome different stages during the tumorigenic process (Lorendeau et al, 2015; Elia et al, 2018). In this way, identical cells in different microenvironments (including tissues, organs, or different parts of the same tumor) exhibit distinct metabolic programs that lead to a variation in phenotypes (Elia & Fendt, 2016; Rinaldi et al, 2018).

As a tumor mass grows and expands, the tumor cells within undergo fluctuations in nutrient availability. Nutrients and metabolic byproducts, such as lactate, are present in a gradient in solid tumors, from tumor edges that have greater access to oxygen and nutrients, to the tumor cores with elevated hypoxia and metabolic byproducts. Additionally, diet‐driven changes may alter the circulating nutrients and impact the function of circulating immune cells—both of which play roles in altering nutrient availability to solid tumors (Follain et al, 2020). As such, studying the nutrient and metabolite composition at the intratumoral level and in the tumor interstitial fluid can provide clues for what nutrients are available to tumor cells, what are being taken up for cellular processes, and what metabolites are secreted as a byproduct for the tumor microenvironment. Here, we review the current research on various nutrient networks that the tumor cell is capable of using during the different stages of cancer development, focusing exclusively in an in vivo setting where the complexity of interactions makes the difference in tumor nutrient dependency.

Evolution of nutrient dependence during cancer development

Nutrient availability to tumor cells is a decisive factor driving metabolic rewiring in cancer. Recent studies provide evidence that cancer cells use different metabolic processes depending on their microenvironment (da Cunha et al, 2019; Doglioni et al, 2019). For example, some tumor cells use the nutrient glutamine as a primary carbon source to feed into metabolic pathways required for in vivo growth, whereas in other tissues or malignant situations glutamine acts as a signaling molecule for tumor progression (Pan et al, 2016; Cacace et al, 2017). Currently, several in vitro models are being developed to more closely resemble in vivo cell environments in order to study metabolic dependencies (Coloff et al, 2016; Cantor et al, 2017; Elia et al, 2017; Vande Voorde et al, 2019). Although these models are useful for determining the molecular mechanisms driving tumorigenesis, the use of in vivo systems to complement these models is crucial for the translational applications of in vitro findings. These in vivo models have the advantage of presenting the most physiologically relevant contexts to study cancer, where the diversity of nutrient environments, oxygen tensions, and stromal cell populations—that are difficult to recapitulate in vitro—can be incorporated into experiments. Based on such studies, it was found that diet, oncogenic mutations, tissue of origin, and tumor microenvironment can all impact metabolic phenotypes in an in vivo situation. The role that these factors play in the metabolic adaptations that tumor cells acquire over the stages of cancer will be discussed in this section.

The first hurdle—nutrient requirements of cancer initiation and growth

Does diet play a role in tumor initiation and progression?

Clinical epidemiologic research complemented by preclinical dietary studies

Changes in whole‐body metabolism caused by diet can affect local nutrient availability in the interstitial fluid of organs. Thus, dietary nutrient intake affects the metabolites that are available to malignant cells during all stages of disease (Sullivan et al, 2019a). In humans, diet has already been implicated in affecting the risk of cancer development, primarily with dietary carbohydrates and fat being of interest. Several epidemiological studies link nutritional excess and the subsequent increase in body adiposity, weight, and obesity with increased rates and risk of various cancers in humans (Mayne et al, 2016; Goncalves et al, 2019a). The Continuous Update Project run by the World Cancer Research Fund and American Institute for Cancer Research has recently reported that overweight or obese individuals have an increased risk for a number of cancers (WCRF/AICR, 2018). Several meta‐analyses show a strong correlation between obesity and weight gain with digestive tract malignancies (liver, colon, rectal, pancreatic, and biliary cancers) and hormone‐linked cancer in women (breast and endometrial cancers) and men (prostate) (Dong et al, 2017; Kyrgiou et al, 2017; WCRF/AICR, 2018). While there is clear evidence that diet and excess nutrient intake may affect the risk of certain types of cancers, this does not necessarily mean that this translates to disease outcomes. For example, a recent meta‐analysis of 9 clinical studies suggests that triple‐negative breast cancer disease‐free progression and overall survival is unaffected by the presence of obesity (Mei et al, 2018). In brief, there appears to be a link between diet, nutrient excess, and adiposity with site‐specific cancers; however, clarifying the causal mechanisms remains an area of intense study.

Despite the importance of nutrient availability for tumor development, studying the role of diet on nutrient exposure to cancer tissues in patients is extremely challenging. There are numerous barriers to interpret nutritional interventions such as feasibility, insufficient magnitude of dietary change, nonadherence to protocols, or the design of appropriate controls (Mayne et al, 2016). In this respect, clinical observations are complemented with dietary studies in murine models to build on the mechanistic interplay between nutrient availability and cancer development, since nutritional interventions and conditions are easier to control in research animal diets (Lien & Vander Heiden, 2019). To date, a number of rodent dietary interventions have been used to study the effect of diet on cancer initiation and progression, including high‐fat diets (HFD), Western diets (high fat with carbohydrates and cholesterol), ketogenic diets (low carbohydrate), and nutrient‐restricted diets. These dietary interventions result in altered tissue and plasma nutrient profiles in animal models of disease. For example, HFD and Western diets have been shown to have the most profound effects on circulating glucose and free fatty acid concentrations throughout the exposure to modified animal feed (Lai et al, 2015; Wang et al, 2016). HFD can also influence the fate of glucose in tumors by altering the whole‐organism energy balance. Lipid‐driven perturbations in hepatic metabolism indirectly feed demanding colorectal tumor growth by stimulating hyperinsulinemia and hyperglycemia, essentially increasing growth signaling and glucose for use by the tumors in extra‐hepatic sites (Wang et al, 2018). The effect of ketogenic diets on whole‐organism physiology may benefit cancer therapy, with animals experiencing improvements in hyperglycemia and hyperinsulinemia resulting from PI3K inhibitors (Hopkins et al, 2018). Restriction diets that limit specific nutrients (such as serine‐ and glycine‐free diets) can increase survival in lymphoma and intestinal cancer and influence proliferation in melanoma and breast cancer models (Maddocks et al, 2017; Sullivan et al, 2019a). Additionally, the gut microbiome has re‐emerged as an important physiological factor in cancer development. As such, the effects of diet on the microbiome and cancer have become an area of increasing interest. The effects of these diets in tumor initiation, progression, as an intervention, and how the gut microbiome may be involved are discussed below.

Influence of diet during tumor initiation

Dietary modulations combined with well‐defined genetic or chemical induction models of murine cancers are widely used to study the role of nutrient‐induced changes in tumor initiation. High‐fat and sugar diets have been shown to accelerate tumor initiation in murine models of hepatocellular carcinoma (HCC) (Asgharpour et al, 2016; Tsuchida et al, 2018). These studies found that prolonged exposure to a Western diet (HFD with glucose, fructose, and cholesterol) provokes localized inflammation in the liver, lipotoxicity, and systemic changes in insulin levels and sensitivity, which, at a molecular level, activates cellular stress and oncogenic pathways relevant for HCC (Asgharpour et al, 2016; Tsuchida et al, 2018). Interestingly, this suggests that a Western diet may prime healthy tissue to develop tumorigenesis. In colorectal cancer (CRC), this diet was found to drive CRC initiation but not progression in adenomatous polyposis coli (APC) mutant mice—which are predisposed to develop intestinal tumors—where fat and sugar exposure increased the number but not the size of tumors. The mechanism behind this outcome may be an altered metabolism, oxidative stress, inflammation, or immune cell infiltration that accelerates the loss of heterozygosity of the Apc gene (Niku et al, 2017).

On the other hand, tumor initiation driven by sugar availability (fructose and glucose) has been reported in the presence of certain oncogenic mutations. In this respect, loss of the macrophage‐secreted apoptosis inhibitor of macrophage (AIM) in the context of a high fructose diet significantly increases HCC development, affecting insulin signaling and reactive oxygen species scavenging (Ozawa et al, 2016). In a CRC model with APC loss in intestinal stem cells, fructose supplementation by water increased tumor initiation alongside increased body weight and impaired whole‐body metabolism (Goncalves et al, 2019b). However, when the authors used oral gavage of fructose to limit weight gain and whole‐body metabolic disturbances, no impact on CRC initiation (as determined by total tumor numbers), but an impact on progression (observed by increased tumor size and grade of disease), was found (Goncalves et al, 2019b). Interestingly, the combination of dietary glucose and fructose, but not each sugar alone, significantly affected cell metabolism of CRC tumors. The presence of fructose in these tumors—even at a moderate dose—enhances glucose metabolism by depleting ATP levels, thereby activating glycolysis (via activation of the glycolytic enzyme phosphofructokinase). This metabolic rewiring led to an accelerated de novo lipogenesis which enhances intestinal tumor growth and grade, an enhancement that can be abolished by preventing fructose metabolism (Goncalves et al, 2019b). These studies suggest synergistic effects of the genetic predisposition and diet resulting in a significantly accelerated tumorigenesis, though further investigations in cancer patients are required.

Influence of diet during tumor progression

Although the role of dietary sugar and fat in cancer progression in patients remains unclear, an increased number of mouse studies provide evidence for the influence of diet on metastatic disease. In breast cancer models, dietary sugar was found to facilitate the development of secondary lung tumors (Jiang et al, 2016a; Fan et al, 2017). Moreover, it was found that human colon cancer tumor xenografts injected subcutaneously and orthotopically displayed accelerated growth and progression in diet‐induced obese and insulin‐resistant mice (O'Neill et al, 2016). Additionally, HFD has been reported to favor metastatic progression in xenograft models from cell lines of various tissue origin (Kim et al, 2011; Park et al, 2012; Pascual et al, 2017; Chen et al, 2018b). For instance, oral squamous carcinoma cells expressing the fatty acid transporter CD36 are sensitive to circulating blood fat levels resulting in increased metastatic potential triggered through HFD or palmitic acid exposure (Pascual et al, 2017). Furthermore, HFD‐induced lipid accumulation in prostate tumors enhanced the metastatic progression in mouse models with co‐deletion of phosphatase and tensin homolog (Pten) and promyelocytic leukemia (Pml) genes—two common oncomutations that cooperatively promote metastasis in human prostate cancer. The metastatic phenotype induced by dietary lipids might be driven by hyperactivation of an aberrant sterol regulatory element‐binding protein (SREBP) prometastatic lipogenic program, which is observed in metastatic human prostate cancer (Chen et al, 2018b). HDF consumption has also been shown to stimulate colon cancer progression as a consequence of small increases in fat mass in the tumor and adipose tissue which contribute to the infiltration of leukocytes in the tumor. The cross‐talk between cancer cells, leukocytes, and adipocytes within the tumor mass may promote the generation of growth factors, cytokines, and chemokines in situ, thereby stimulating tumor growth, angiogenesis, and metastasis (Park et al, 2012). Interestingly, some of these HFD‐related phenotypes can be observed in the absence of obesity and metabolic syndrome, suggesting that environmental dietary factors, rather than the development of metabolic diseases, can boost the progression of tumors.

Therapeutic window of dietary interventions

Cancer‐preventing diets are being explored for their potential to impact tumor development. Preventative effects on cancer progression and development have been suggested through consuming a Mediterranean diet, ketogenic diet (high‐fat/low‐carbohydrates), and applying caloric restriction (Lashinger et al, 2016; Hopkins et al, 2018; Piazzi et al, 2019; Castejón et al, 2020). Restriction of specific nutrients, such as serine and glycine, was found to impair colorectal cancer and lymphoma at various stages of initiation and progression, but showed a less clear effect in pancreatic cancers (Maddocks et al, 2017). In breast cancer, serine restriction inhibited tumor growth, albeit only with the loss of phosphoglycerate dehydrogenase (PHGDH), the rate‐limiting enzyme in serine biosynthesis, highlighting both the dependency of breast cancer tumor growth on serine and the metabolic adaptations that cancer cells use to sustain their growth (Sullivan et al, 2019a). Conversely, in a recent study on squamous cell carcinoma, progenitor stem cells were found to rely on serine availability, but not serine biosynthesis and PHGDH expression, to support stem cell populations and subsequent tumor initiation and burden (Baksh et al, 2020). This effect was associated with the presence of the repressive chromatin mark histone 3 lysine 27 trimethylation (H3K27me3), which controls terminal differentiation of stem cells, thus inhibiting differentiation and escape from becoming a malignant cell. H3K27me3 demethylation was discovered in tumors grown in mice fed serine‐ and glycine‐free diets, and PHGDH loss also prevented H3K27 methylation on this diet. Ultimately, PHGDH loss stimulated tumor growth in the presence of serine, by supporting tumor stem cell proliferation, but the loss of serine from the diet prevented this effect (Baksh et al, 2020). Dietary restriction of the essential amino acid methionine disrupted one‐carbon metabolism in humans, which has been shown to improve therapeutic responses of chemo‐ and radiotherapy‐resistant colorectal cancers and sarcomas in patient‐derived xenograft (PDX) models of tumor growth and progression (Gao et al, 2019). Furthermore, ketogenic diets or diets supplemented with ketones, a metabolic byproduct of fat catabolism, have been studied for potential anti‐cancer effects, with relevant translational results in glioblastoma patients (Poff et al, 2015; Martuscello et al, 2016; Feng et al, 2019). Currently, the benefit of these low‐carbohydrate (ketogenic) diets is tested in overweight women with endometrial cancers (http://www.clinicaltrials.gov/ identifier NCT03285152). Interestingly, dietary interventions administered in conjunction with anti‐cancer treatments can increase treatment efficacy. For instance, taking advantage of the insulin‐sensitizing effects of the ketogenic diets, response to PI3K inhibitors (which is dampened in the presence of high insulin levels) can be improved in multiple tumor types with varied mutational profiles in mice (Hopkins et al, 2018). Furthermore, the improvement of endocrine therapy by reducing insulin signaling via ketogenic diet is currently tested in women with estrogen receptor–positive breast cancer (http://www.clinicaltrials.gov/ identifier NCT03962647). On the other hand, growing preclinical and clinical evidence shows that short‐term fasting can enhance the efficacy of a variety of chemotherapeutic agents against multiple tumor types and can reduce side effects (Peng et al, 2012; de Groot et al, 2015; Dorff et al, 2016; Bauersfeld et al, 2018). Currently, large randomized clinical trials are validating the effect of calorie restriction and intermittent fasting on treatment side effects in cancer patients (http://www.clinicaltrials.gov/ identifiers NCT01802346, NCT00936364, NCT03162289, among others). These works highlight the importance of considering diet effects on therapy response in both, the preclinical and clinical settings.

Implications of gut microbiome in the diet‐cancer interaction

An emerging field looking at the intersection of diet and cancer involves the study of the microbiome, the commensal bacteria, and other microbial populations that live in symbiosis within the body. CRC has been especially highlighted to have interactions with the gut microbiome, the single largest home of symbiotic microbes in the body (Sender et al, 2016). Since the gut microbiome metabolizes nutrients that mammals are unable to break down for absorption and use (such as fiber), it is especially sensitive to dietary composition. Obesity in humans and HFD feeding in mice both drive population lost of diversity in the gut microbiome (Dalby et al, 2017; Foley et al, 2018). Further, the presence of colon cancer alters the microbiome populations as compared to healthy controls (Dejea et al, 2014; Feng et al, 2015). Beyond the gastrointestinal tract, interactions with the immune system have driven further research linking cancer development, immunotherapies, and the gut microbiome. Several studies have shown the complex interplay between the microbiome and cancer that brings microbiota modulation as a viable strategy for improving the clinical efficacy of anti‐cancer treatment. Ma and colleagues found that the metabolism of bile acids by microbes regulated the chemokine (C‐X‐C motif) ligand 16 (CXCL16) which attracted active natural killer (NK) T cells to control liver tumor growth (Ma et al, 2018). In humans with non‐small‐cell lung cancer (NSCLC), differential microbiome composition or antibiotic exposure negatively impacted clinical response to programmed cell death protein 1 (PD‐1) blockade (Routy et al, 2018). Similarly, microbiome composition may affect responses to anti‐PD‐1 immunotherapy in melanoma patients by modulating antigen presentation and effector T‐cell function (Gopalakrishnan et al, 2018). Recent work has begun to explore the diet‐microbiome‐cancer nexus, with HFD‐driven changes to the gut microbiome driving colon cancer development in Ras‐mutated mice (Schulz et al, 2014). In a study on multiple myeloma, major changes in human microbiomes were measured in patients versus healthy controls, and subsequent fecal microbiome transfers, specific microbe supplementation, and glutamine‐deficient diets in a mouse model of myeloma showed that microbe‐synthesized glutamine was found to be a critical driver of tumor growth (Jian et al, 2020). Further, it stands to say that many of the dietary interventions discussed above will impact the microbial flora and may play a role in cancer initiation, growth, and response to therapy, and thus, the microbiome is a critical consideration when discussing causality of diet effects on cancer (Zitvogel et al, 2017).

Taken together, it emerges that diet and environment affect circulating nutrient availability. This directly or indirectly alters nutrient availability to tumors. In the consequence, diet interventions can affect tumor growth and development at least in mouse models (Fig 1). While causality is limited in prospective and retrospective clinical studies on diet, cancer risk, and outcomes, the functional and mechanistic preclinical studies combined with ongoing large‐scale genetic and transcriptomic analysis of human tumor samples will be important to further increase our understanding of nutrient availability and its impact on cancer development.

Figure 1. Diet can determine tumor cell metabolism which may have effects on initiation and progression of different type of cancers.

Diet intake influences nutrient availability in the body, affecting what nutrients are available during early initiation and tumor progression stages of cancer development. Various mechanisms are involved in diet‐induced cancer growth, primarily due to altered cellular metabolism, ROS production and diminished ROS scavenging, and altered immune cell infiltration or function. (A) Effects of high‐fat diet (HFD) and Western diets on initiation and progression of various cancers. Excess lipids consumed through HFD feeding can drive tumorigenesis and promote tumor growth by changing cellular metabolism to be more glycolytic, increasing levels of inflammation, increasing the production of ROS species that damage DNA and promote cellular stress mechanisms and DNA damage, and by affecting the amounts and types of immune cells that infiltrate the growing tumors. The diet also affects the composition of the gut microbiome, which impacts the growth of colorectal cancer, and also affects immune cell infiltration into tumors. Nutrients produced by microbes in the gut are also associated with promotion of cancer growth. (B) Diets high in fructose can affect initiation of hepatocellular carcinoma, but promote progression of tumors in other sites including colorectal and breast. Mechanisms involved in fructose‐driven cancer development and progression includes metabolic rewiring of cancer cells to be more glycolytic, allowing for increased de novo lipogenesis to sustain uncontrolled cell growth and altered ROS scavenging. Abbreviations and symbols: ROS (Reactive oxygen species), Δ (alterated), ↑ (high influence).

Do oncogenic mutations and tissue origin drive specific nutrient dependencies during tumor initiation?

One of the first steps in oncogenic transformation is the gain or loss of function of a gene regulating proliferative pathways (Hanahan & Weinberg, 2011). Beyond modulating proliferative signaling, a number of these genes are also involved in regulating cancer cell metabolism to support the rapid growth and division of cancer cells (Iurlaro et al, 2014). With the development and characterization of genetically modified models and in vivo metabolic tracer studies, a number of mutations have been identified that dictate nutrient requirements of malignant tissues among different organ sites and cancer types (Elia et al, 2016; Fernández‐García et al, 2020). One of the first studies showing this effect in vivo is from Yuneva and colleagues, where liver tumors—with distinct activating oncogenic mutations—exhibit unequal utilization of some nutrients such as glucose and glutamine. Here, myelocytoma (MYC)‐induced liver tumors increase lactate production from glucose and catabolism of both glucose and glutamine through the tricarboxylic (TCA) cycle, whereas hepatocyte growth factor receptor (MET)‐induced liver tumors use glucose to produce glutamine (Yuneva et al, 2012). Intriguingly, the tissue context can also affect the metabolic reprogramming outcome driven by the oncogenic mutation. In contrast to liver tumors, lung adenocarcinomas induced by MYC maintain glucose catabolism alongside elevated levels of glutamine (Yuneva et al, 2012). In comparison, Kirsten rat sarcoma oncogene (KRAS)‐driven non‐small‐cell lung carcinomas show the same increase in oxidative glucose usage to fuel the TCA cycle but minimal glutamine utilization (Davidson et al, 2016), suggesting that both, the metabolic reprogramming by MYC and KRAS, are also tissue of origin dependent (Fig 2). Another interesting example in which an oncogenic mutation and the tissue of origin can drive drastically different metabolic changes is observed in KRAS‐driven lung and pancreatic tumors. Kras mutation in lung tissue upregulates branched‐chain amino acid (BCAA) uptake and their catabolism and incorporation into proteins, whereas pancreatic tumors bearing the same mutation decrease BCAA uptake compared to adjacent normal tissues (Mayers et al, 2016). Furthermore, despite the presence of the same oncogenic mutations (KRAS activation and p53 loss), different levels of BCAA and their catabolites were found in the tumor interstitial fluid (the extracellular fluid that perfuses tumors) in lung (LUAD) and pancreatic ductal adenocarcinoma (PDAC) (Sullivan et al, 2019a). Thus, the tissue of origin is one of the major determinants of the tumor microenvironment composition and the nutrients available to tumor cells.

Figure 2. Organ‐specific tumor metabolism is influenced by the tissue of origin and oncogenic driver mutations.

MYC‐induced lung tumor (upper left panel) shows lactate production from glucose and catabolism of glucose through the tricarboxylic cycle (TCA). While it is likely these tumors still consume glutamine, the intracellular levels are high, suggesting that glutamine synthesis exceeds consumption. KRAS‐driven lung tumor (upper right panel) resembles the increased oxidative glucose usage feeding the TCA cycle with minimal glutamine utilization. MET‐induced liver tumor (bottom right panel) mainly oxidizes glucose, with apparent net glutamine synthesis. MYC‐induced liver tumors also (bottom left panel) exhibit increased glucose uptake. However, in these tumors is observed an increased lactate production. In addition, these tumors use glutamine to fuel the citric acid cycle.

Apart from the tissue‐specific differences, there is an additional impact of variable oncogene expression on the cancer cell metabolic phenotypes. Elevated mutational burden observed in late‐stage lung tumors (specifically KRAS copy gain) has been associated with increased glucose metabolism and oxidative TCA cycle (Kerr et al, 2016). Conversely, changing nutrient availability by diet interventions has been shown to affect genetic programs. HFD fed mice bearing KRAS mutant pancreatic acinar cells enhanced KRAS activation, resulting in a metabolic shift in pancreatic tumor cells toward an aggressive glycolytic phenotype (Wang et al, 2019a). Further, the growth of KRAS‐driven tumors was not affected by the removal of serine and glycine from the diet, while APC‐ or MYC‐driven pancreatic tumors were sensitive to serine/glycine depletion resulting in reduced growth (Maddocks et al, 2017).

Nonetheless, these are only a few examples of driver mutations which—depending on the nutrient composition in the environment and the tissue of origin—activate specific metabolic pathways to support tumor cell growth. Other oncogenic mutations may have a key role in shaping metabolic needs during tumor development (Sullivan & Vander Heiden, 2019). However, the scope of how and when these changes occur is far from clear, the recent works discussed above highlight the diverse metabolic changes elicited by similar intrinsic factors (tissue of origin or genetic program). Studies using metabolite tracing approaches may be useful to clarify what global metabolic changes are happening in the clinical context of disease and whether connections can be made to disentangle oncogenic programs and their influences on metabolism in human cancer.

Fuel for established tumors—nutrients that support cancer development

Can the nutrient availability imposed by the local environment define the metabolic features of cancer cells?

Once the intrinsic features of transformed cells have activated proliferative programs, tumor growth is sustained by the interaction of cancer cells with their microenvironment. The organ environment can then redefine oncogene‐imposed metabolic dependencies of cancer cells, meaning that not only the oncogenic drivers and the tissue of origin, but also the local environment influences the development of the tumor (Lunt & Fendt, 2018; Rinaldi et al, 2018). The local environment is characterized by fluctuating metabolic properties. One of the extrinsic features that contribute to this deregulation, along with oxygen and pH, is the nutrient composition of the local environment. Diet‐driven changes in circulating nutrients (Lai et al, 2015), the function of draining lymph nodes (García Nores et al, 2016), and the tissue vasculature (Carmona‐Fontaine et al, 2017) all play roles in altering nutrient availability to solid tumors in this organ environment. Therefore, each organ has a unique metabolic profile making its local environment a particular soil for cancer cells to grow (Table 1).

Table 1.

Presence and availability of some of the most common nutrients implicated in tumor metabolism in blood and different tissues

Nutrient	Human (concentration in μM)		Mouse (concentration in μM)		Reported location in humang		Concentration in tumor interstitial fluid (μM)
	Intracellulara	Blooda	Intracellularb	Bloodd	Organa	Tissuea	Mousec
Alpha‐ketoglutarate	290‐350	0‐23	800	8‐27	NA	NA	8
Citrate	NA	30‐400	584	250‐600	Prostate	All tissues	340
Dihydroxyacetone phosphate (DHAP)	60‐220	16	1630	20‐300	NA	NA	750
Fumarate	NA	NA	485	NA	Prostate	NA	2.5
Glucose	NA	3000‐6000	NA	5000‐11000	Lung (5600e), brain (1000), abdomen (8000‐10000f), bladder, Liver, prostate, pancreas, kidney	Epithelial, muscle, adipose	2700‐4500f
Glucose 6‐phosphate	26‐50	5‐30	675	2‐15	Liver, kidney	Muscle (smooth and skeletal), adipose	107
Glyceraldehyde 3‐phosphate	5.70‐7.70	5	141	30‐100	Kidney	Smooth muscle	160
Lactate	600‐3,500	500‐2,000	NA	3,000‐8,000	Lung, abdomen (5,600f)	All tissues	8900‐16700f
Malate	2,800‐3,600	3‐21	1,390	20‐100	Prostate	NA	155
Phosphoenolpyruvate	15‐19	14.5	12	0.1‐2.5	Prostate	NA	15
3‐Phosphoglycerate	45	47	NA	1‐2.3	Prostate	NA	30
Pyruvate	27‐127	20‐250	5880	40‐200	Testicle, spleen, brain, liver, kidney, pancreas	Muscle (smooth, skeletal and cardiac), adipose	40
Succinate	NA	6‐16	352	100‐500	Spleen, brain, liver, prostate, kidney, pancreas	Muscle (smooth and skeletal), adipose	110
Alanine	NA	200‐300	6980	500‐1000	Prostate	All tissues	1100
Arginine	NA	60‐130	255	110‐200	Intestine, spleen, bladder, liver, testicle, prostate, kidney, pancreas	Muscle (smooth and skeletal), epithelial, adipose	2
Asparagine	NA	30‐80	215	50‐120	Prostate	All tissues	115
Aspartate	NA	6‐20	14900	15‐70	Prostate	All tissues	400
Cysteine		30‐300	84	NA	Intestine, spleen, liver, testicle, prostate, kidney	Muscle (smooth and skeletal), epithelial	NA
Cystine	60‐120	60‐120	NA	30‐135	NA	All tissues	50
Glutamate	1200‐1600	20‐145	63800	50‐170	Intestine, spleen, prostate, kidney, pancreas	Muscle (smooth and skeletal), epithelial, adipose	1050
Glutamine	NA	390‐900	17200	500‐900	Intestine, testicle, spleen, prostate, kidney, pancreas	NA	800e
Glycine	NA	120‐325	3710	80‐350	Intestine, spleen, bladder, brain, prostate, kidney, pancreas	Epithelial	490
Histidine	NA	80‐240	410	45‐165	Prostate	All tissues	95
Hydroxyproline	NA	3‐30	NA	NA	Spleen, liver, skin, prostate, kidney, pancreas	Muscle (smooth and skeletal)	NA
Methionine	NA	10‐45	639	60‐350	Spleen, liver, prostate, kidney, pancreas	Smooth muscle	75
Proline	NA	100‐350	1230	45‐135	Prostate	All tissues	120
Serine	NA	50‐200	4860	10‐50	Prostate	All tissues	90

NA, Not available.

^aWishart et al (2018). Estimated concentrations in μM from studies in healthy adult compiled in the Human Metabolome Database.

^bPark et al (2016). Measurements of absolute concentrations (μM) in immortalized baby mouse kidney cells (iBMK cells).

^cSullivan et al (2019a). Microenvironmental metabolites in murine pancreatic adenocarcinoma interstitial fluid.

^dSullivan et al (2019a). Plasma concentration (μM) in control (healthy) mice.

^eFisher (1984). Measurements in isolated perfused rat lung.

^fAnn Burgess & Sylan (1962). Measurements in mouse tumor tissues.

^gHealthy organs and tissues in which nutrients have been verified and their concentration (when known) in normal conditions.

Adaptations to lung environment

For most healthy tissues, glucose serves as a primary source of energy meaning that this metabolite is highly available in most organs. For example, glucose oxidation in the lung has been estimated to be comparable to most other metabolically active organs (O'Neil & Tierney, 1974; Fisher, 1984). Although the lung nutrient environment has been primarily studied in murine models, there are strong indications of remarkable dependence on oxidative glucose metabolism in different human lung tumors (Fan et al, 2009; Davidson et al, 2016; Hensley et al, 2016). Interestingly, lung cancer cells grown in vitro rely on glutamine as the preferred anaplerotic carbon source to support the TCA cycle, but upon returning to the in vivo lung environment, glucose metabolism supports TCA anaplerosis (Davidson et al, 2016). In addition, even under well‐oxygenated conditions, lactate production in the lung is elevated when compared to many other tissues (Fisher, 1984) presumably to minimize local oxygen consumption, thereby enhancing overall oxygen delivery to other tissues (Liu & Summer, 2019). This makes lactate highly available in the local environment, allowing it to be used as a carbon source to support the growth of human non‐small‐cell lung cancers (Hensley et al, 2016; Faubert et al, 2017). Similarly, elevated pyruvate availability (compared to blood) in the lung environment favors pyruvate carboxylase (PC)‐dependent anaplerosis to fuel the TCA cycle, which is required for tumor survival and proliferation (Christen et al, 2016) (Fig 3A). NSCLC patients show PC upregulation and enhanced PC activity (compared to Glutaminase1 (GLS1)) in the primary tumors compared to normal adjacent tissues (Sellers et al, 2015). Interestingly, breast cancer cells growing in the lung as secondary tumors also increase PC‐dependent anaplerosis compared to corresponding primary tumors (Christen et al, 2016). While primary triple‐negative breast cancers often use glutamine anaplerosis and are consequently susceptible to GLS1 inhibitors (Elia et al, 2016), different nutrient availability in the lung microenvironment may rewire anaplerotic pathways in cells with the same origin, losing this drug susceptibility. This switch between GLS1‐dependent anaplerosis and PC‐dependent anaplerosis can be recapitulated when growing breast cancer cells in vitro in the presence or absence of pyruvate (Christen et al, 2016). This suggests that, although tumor cells maintain some metabolic features of their tissue of origin (Gaude & Frezza, 2016), different in vivo organ microenvironments can drive different metabolic adaptations, even among cancer cells of the same origin.

Figure 3. Tumors have the ability to consume multiple nutrients to fuel the central carbon metabolism.

The plasticity of tumor cells allows them to adapt to the unique local environment of each organ, as noted in lung, brain, and pancreas. This flexibility allows them to take advantage of the available nutrients to fulfill energetic and biosynthetic demands promoting tumor growth.

Adaptations to brain environment

The human brain uses about 20% of the body's daily supply of glucose—making it the largest consumer of glucose of all body tissues (Erbslöh et al, 1958; Mergenthaler et al, 2013). Due to the lack of fuel storage, brain tissue requires a continuous supply of glucose and, under normal conditions, the supply of this metabolite from the blood is high in the brain microenvironment. Hence, genetically diverse human glioblastoma tumors (orthotopically transplanted into mice) utilize mitochondrial glucose oxidation to replenish biosynthetic intermediates of the TCA cycle and to contribute to a large glutamine pool during tumor growth (Marin‐Valencia et al, 2012). Apart from its mitochondrial fate, glucose has been shown to support protein and nucleic acid synthesis by preferentially supplying pools of macromolecular precursors such as glutamine, glutamate, and glycine in in situ studies of human brain tumor metabolism (Fig 3B) (Maher et al, 2012; Tardito et al, 2015). In fact, the limitation of serine and glycine in the brain environment restricts the growth of metastatic cells colonizing this organ. To overcome this metabolic constrain, disseminated cancer cells enhance de novo serine synthesis necessary for nucleotide production and cell proliferation, leading to a strong dependency on this biosynthetic pathway in metastatic brain tumors (Ngo et al, 2020). Additionally, brain cells can metabolically respond to physiological changes such as starvation or altered neuronal activity. Brain tissue can produce substantial amounts of acetate (Jang et al, 2019), which can be used (among others, such as ketone bodies or short and medium‐chain fatty acids) as an alternative fuel when blood glucose is low (Ebert et al, 2003; Mason et al, 2006; Deelchand et al, 2009). This may benefit brain‐resident tumors, as patient‐derived glioblastoma (but not normal brain tissue) has been shown to avidly consume available acetate, as opposed to glucose, to fuel the TCA cycle (Fig 3B) (Mashimo et al, 2014). Interestingly, brain metastases from different origins, even with established glucose oxidation phenotypes, can adapt to use acetate as an alternative energy source in the new environment. These observations suggest that the unique brain microenvironment may favor tumors of diverse origins that can utilize acetate as a main biosynthetic substrate (Mashimo et al, 2014) or efficiently synthesize serine for cell growth (Ngo et al, 2020).

Adaptations to other environments

Further metabolic plasticity in tumors has been reported in other organs such as the pancreas and liver. Poorly vascularized pancreatic tumors experience deficient nutrient availability due to perfusion limitations, causing extracellular protein catabolism to become a source for metabolic intermediates. In this organ, KRAS‐driven PDAC tumors, which are low in upper glycolytic intermediates, glutamine, and serine, can obtain sufficient amino acids via extracellular protein scavenging to fuel their metabolic requirements (Kamphorst et al, 2015; Davidson et al, 2017) (Fig 3C). The fact that PDAC tumors may present permeable and leaky blood vessels due to the high interstitial pressure and lymphatic deficiency may result in plasma protein accumulation in the microenvironment of these tumors which boosts protein scavenging (Kamphorst et al, 2015).

Under physiological conditions, the liver produces, stores, and releases glucose to maintain body homeostasis. This may make glucose highly available in this environment, and, contrary to lung tumors, the liver microenvironment is naturally more conducive to cells that display a high glycolytic profile. Along this line, mouse models of liver cancer show elevations of glycolytic flux profiles, lactate levels, and high anaplerotic glutamine utilization (Hu et al, 2011; Yuneva et al, 2012; Muir et al, 2017). Changes in glucose metabolism among distinct organs have been also found in cancer patients. Contrary to human lung (Hensley et al, 2016) and brain tumors (Maher et al, 2012), which show significant levels of glucose oxidation, tumors growing in the kidney display enhanced glycolytic flux with minimal glucose oxidation and turnover of the TCA cycle (Courtney et al, 2018). Cancer cells growing in the liver can also rewire palmitate metabolism by desaturating palmitate using fatty acid desaturase 2 (FADS2) to the unusual fatty acid sapienate which can potentially influence oncogenic lipid‐signaling networks (Vriens et al, 2019; Triki et al, 2020), suggesting an alternative mechanism of metabolic plasticity triggered by fatty acid metabolism.

Interestingly, tissues differ substantially in their susceptibility toward specific oncogenic events, meaning that some tumors display a predominance of mutations that affect metabolic networks based on their site of origin (Schneider et al, 2017). For example, von Hippel–Lindau tumor suppressor (VHL) is inactivated in clear cell renal cell carcinoma (ccRCC), whereas it is only rarely mutated in other tissues. VHL loss drives a pseudo‐hypoxic state by preventing the degradation of hypoxia‐inducible factor‐α (HIF‐α). This state is associated with increased glycolysis and suppressed glucose oxidation, which had been observed in ccRCC tumors compared to the adjacent kidney tissue (Courtney et al, 2018). Similarly, tissue‐specific co‐occurring mutations in the serine‐threonine kinase 11 (STK11) are a common feature of KRAS‐driven NSCLC. The co‐mutation of both genes influence glycolysis and enhanced oxidative phosphorylation, a reprogrammed glucose metabolism that has been reported in lung cancer patients (Faubert et al, 2017; Caiola et al, 2018). In contrast, drivers that affect metabolic networks in the liver (such as TP53) are less specific and the consequences of these mutations depend on the developmental context (Schneider et al, 2017).

Although the main evidence in humans has been shown in glucose enriched‐environments, emerging studies are pointing that nutrients may have an influential role in the tissue environment in modulating in vivo cancer metabolism. The flexibility that cancer cells exhibit may create opportunities for tumor cells to adapt to growth in new environments that initially seem hostile due to the nutrient milieu. Therefore, the environment dictates the nutrient availability in the tumor niche and only cancer cells with sufficient metabolic flexibility are able not only to adapt to these environmental conditions but can also rewire their metabolism to promote tumor growth.

Can tumor heterogeneity provide an advantageous situation even during nutrient fluctuations?

As discussed, the metabolic flexibility exhibited by cancer cells is advantageous for the adaptation to nutrient variability in the organ environment. However, the question remains: Is metabolic flexibility also important for cells to respond to changes within the tumor? This seems plausible, considering an increasing number of works that describe pronounced heterogeneity and metabolic flexibility within distinct regions of solid tumors (Hensley et al, 2016; Xiao et al, 2019; Tasdogan et al, 2020; Vivas‐García et al, 2020). This intratumoral metabolic heterogeneity is imposed by (i) intrinsic factors of the tumor cells, such as genetic alterations, or (ii) by local pathophysiological conditions, such as nutrient deprivation.

Intrinsic factors triggering tumor heterogeneity

Although tumorigenesis may begin with a homogeneous genetic background, it concludes with billions of malignant cells that are mutationally diverse and amplified by clonal selection (Marusyk & Polyak, 2010). Clonal heterogeneity within a tumor has been reported for a variety of malignancies (Marusyk & Polyak, 2010) and recent advances in single‐cell sequencing technology have further uncovered individual tumor heterogeneity in clinical samples, revealing the existence of multiple tumor cell states (Liu et al, 2017; Puram et al, 2017; Wang et al, 2019b). Interestingly, distinct tumor subpopulations exhibit striking differences in the metabolic features of these distinct cell states. For instance, high metabolic heterogeneity in the TCA cycle and oxidative phosphorylation (OXPHOS) has been observed at the single‐cell level in human melanomas and head and neck squamous cell carcinoma, suggesting that variation in mitochondrial activity may be the major contributor to intratumoral metabolic heterogeneity in these tumors (Xiao et al, 2019). In fact, an increased mitochondrial activity in epidermal stem cell populations can drive inhibition of de novo serine synthesis pathway, which in turn allows α‐ketoglutarate accumulation and the subsequently demethylation repression—beneficial for squamous cell carcinoma initiation (Baksh et al, 2020). In the context of therapy resistance, clusters of drug‐tolerant cells were found to be defined by the oncogene MITF (microphthalmia transcription factor) and a decreased proliferative gene expression signature in PDX melanoma models (Rambow et al, 2018). In squamous cell carcinoma, a subpopulation of slow‐cycling stem cells in the vicinity of tumor‐vasculature displayed activated transforming growth factor‐beta (TGF‐β) signaling, enhancing glutathione metabolism, which was associated with the development of therapy resistance (Oshimori et al, 2015). In both situations, the resistant cell subpopulations (with different metabolic features) were associated with disease relapse. In fact, subpopulations of stem cell‐like tumor cells are reported to support metabolic adaptations, protecting tumor cells from fluctuations in essential nutrients for tumor development (Ahmed et al, 2018; De Francesco et al, 2018). This would suggest that individual cells could become more or less stem‐like, impacting the metabolic profile and hence allow the tumor cells the flexibility to respond to changing microenvironments.

Extrinsic factors triggering tumor heterogeneity

The microenvironment within a tumor is not completely homogeneous, having regions that present different vasculature and different infiltrating normal cell populations. This spatial heterogeneity affects the nutrient composition that tumor cells are exposed to. Most tumors have partially structured habitats in which individual cells have varying proximity to the vasculature and hence different levels of oxygen and metabolite accessibility (Hoogsteen et al, 2007; Da Ponte et al, 2017). Interestingly, cancer cells engage into a metabolic symbiosis across different regions of the same tumor. That is, vessel‐distant (or hypoxic) tumor areas are characterized by increased expression of the lactate monocarboxylate transporter (MCT4) and release lactate. In contrast, tumor areas relatively distant from hypoxic areas preferentially express MCT1, which facilitates lactate uptake from the microenvironment resulting in oxidative metabolism (Allen et al, 2016; Jiménez‐Valerio et al, 2016; Pisarsky et al, 2016) (Fig 4). This metabolic symbiosis between lactate‐generating and lactate‐consuming cells is associated with adaptive resistance to anti‐angiogenic therapy in mouse models of breast cancer (Pisarsky et al, 2016), renal cell carcinoma (Jiménez‐Valerio et al, 2016), and pancreatic neuroendocrine tumors (Allen et al, 2016). Similarly, the level of perfusion within different tumors and even different tumor areas can determine glucose utilization as observed in lung cancer patients. Metabolic heterogeneous regions observed in the tumors of these patients suggest a metabolic symbiosis, in which tumor tissue with poorly perfused regions preferentially oxidize glucose whereas highly perfused tumor areas can rely on nutrients other than glucose (e.g., lactate) to support their metabolism (Hensley et al, 2016). These studies suggest that tumor zmetabolism heterogeneity observed in vivo is strongly influenced by the microenvironment and might confer an advantage for cells undergoing different situations.

Figure 4. Model of metabolic symbiosis within a tumor mass sustains metabolic limitations due to local variations in nutrient levels.

Tumor heterogeneity showed in three different regions within the tumor mass: in regions of deep hypoxia (blood vessels‐distant, in purple), oxidative phosphorylation (OXPHOS) is prevented and substrate availability is reduced. Cancer cells are highly dependent on glycolysis and thus release lactate through monocarboxylate transporter 4 (MCT4). In a moderate hypoxic environment (in pink), various substrates including lactate (exported from highly glycolytic cancer cells) may contribute to fuel the tricarboxylic acid (TCA) cycle and OXPHOS. These cancer cells are characterized by a high expression of MCT1. Under normoxia (surrounding tumor blood vessels, in red), cancer cells can easily exchange nutrients and oxygen into the bloodstream. In this situation, aerobic glycolysis and glucose oxidation in the mitochondria are fully active, and different nutrients (glucose, lactate) can fuel TCA cycle.

Tumor‐associated stromal cells reside in the tumor microenvironment and contribute to the nutrient composition in the milieu around them. Cancer cells can influence the stroma behavior to benefit from nutrients released by these tumor‐associated stromal components to overcome metabolic constraints within the tumor. For example, in ovarian cancer, tumor cell‐secreted metabolites such as lactate are posited to induce the upregulation of glutamine anabolic pathways in cancer‐associated fibroblasts (CAFs). This metabolic cross‐talk allows CAFs to harness carbon and nitrogen from noncanonical sources to synthesize glutamine that sustains tumor growth in nutrient‐deprived conditions (Yang et al, 2016). Conversely, in prostate cancer, CAFs, influenced by tumor HIF1α signaling, release lactate via MCT4 and pancreatic tumor cells take it up via MCT1, being this metabolic interaction clinically associated with poor prognosis (Fiaschi et al, 2012; Pértega‐Gomes et al, 2014). In the pancreatic tumor environment where glucose and serum‐derived nutrients are limited, cancer cells stimulate alanine secretion by induction of autophagy in stroma‐associated pancreatic stellate cells (PSCs) and selectively consume the PSC‐released alanine, using this carbon source in the TCA cycle (Sousa et al, 2016). These observations suggest that the tissue of origin may influence the behavior of tumor‐associated stromal cells residing in the milieu.

While robust data reveal how heterogeneity becomes particularly beneficial when tumors face strong selective pressures, such as chemotherapy (Ding et al, 2012; Kreso et al, 2013) or metastatic barriers (Wu et al, 2012; Casasent et al, 2018), further studies are needed to decipher the advantage of tumor heterogeneity during nutrient limitation. These initial studies suggest that metabolic heterogeneity within tumors allows cells to sustain cell survival in fluctuating nutrient environments, which might influence therapeutic vulnerabilities.

Traversing a perilous environment—the road to metastasis

What metabolic inputs support tumor cell survival in circulation?

Primary tumors may acquire functional mutations during tumor development that initiate the invasion‐metastatic cascade whereby cancer cells adopt an epithelial‐to‐mesenchymal (EMT) phenotype to disseminate from primary lesions into circulation (Lu & Kang, 2019; Vivas‐García et al, 2020). Once cancer cells leave the primary tumor and embark on this journey, they must undoubtedly adapt their metabolism to overcome the stresses and dangers they face in circulation.

To leave the primary tumor, cancer cells must develop a more motile and invasive cellular phenotype, which requires drastic transcriptional and metabolic changes. The process of primary tumor detachment causes a spike in reactive oxygen species (ROS) that circulating tumor cells (CTCs) need to counteract to survive (Friesen et al, 2004; Schafer et al, 2009). However, a non‐excessive amount of ROS can actually be beneficial since it boosts the activation of antioxidant programs, providing protection from cellular stress. In fact, detachment‐induced ROS may serve a protective role by priming cancer cells with enhanced oxidative defenses that may also aid in drug resistance (Fig 5A) (Gong et al, 2015; Elia et al, 2018; Cheung et al, 2020). In mouse models of breast and colon cancer metastasis, mitochondrial ROS (induced by chemotherapy agents such as doxorubicin) contribute to high baseline DNA damage in CTCs. Evidence suggests that this primes DNA repair pathways and antioxidant protein defenses and, hence, enhances chemotherapy resistance (Gong et al, 2015; Alix‐Panabières et al, 2020). However, a recent study in PDAC highlights the context‐dependent importance of ROS, where suppressing ROS supports tumor development while increased ROS after dissemination promotes metastasis in particular cancer settings (Cheung et al, 2020; Fendt & Lunt, 2020). This low‐to‐high shift in ROS regulation was mediated by expression of the pentose phosphate pathway‐promoting enzyme TIGAR (TP53‐induced glycolysis regulatory phosphatase) and is contrary to what has been shown to drive skin, melanoma, and lung cancer metastasis (Gal et al, 2015; Lignitto et al, 2019). The dynamic regulation of ROS has also been implicated in myeloid state cycling of hematopoietic stem cells, where low ROS maintains quiescence and self‐renewal while high ROS initiates their differentiation and migration (Ludin et al, 2014). Since stem‐like CTCs have been characterized as having more aggressive propensities for metastasis, it is likely this balance in redox state plays a key role in regulating metastasis at secondary sites. However, as exemplified by tumor dependent on TIGAR activity, the balance between excessive ROS and “just enough” to drive the phenotypic requirements of cancer progression is also dependent on genetic background and cancer type.

Figure 5. Cancer cell metabolism influences the tumor microenvironment and can support metastasis formation.

Cell types are indicated with respect to their location in the tumor microenvironment (labeled in the bottom edge), plus nutrients that have been shown to influence their activity in vivo. (A) The mechanistic consequences of tumor detachment and response to doxorubicin therapy can drive gene expression changes regulating nutrient utilization in circulating tumor cells. Abbreviations: Doxorubicin (DOX), GSH (glutathione), FAO (fatty acid oxidation), ROS (reactive oxygen species). (B) Immunoregulatory interactions can be elicited by metabolic byproducts of the intratumoral metabolism, which can act as immune regulators in the metastatic niche. (C) Nutrients can influence metastatic seeding and outgrowth by modulating immune responses and effector functions. Abbreviations: TCA (Tricarboxylic acid cycle), mTOR (mammalian target of rapamycin).

These studies suggest that metabolic pathways involved in protection from ROS support metastasis formation. Thus, antioxidant promoting metabolites such as cysteine and glutamine may be important to maintain ROS in CTCs and combat oxidative damage (Fig 5A) (Friesen et al, 2004; Knott et al, 2018; Combs & DeNicola, 2019). Cysteine and glutamine are metabolic precursors for the synthesis of glutathione, which maintains redox balance and keeps intracellular ROS at a relatively low level. Studies in mouse models of metastasis and tumorigenesis have elucidated the roles of glutamine and cysteine in regulating oxidative stress to promote survival. Glutamine can promote a stem cell phenotype by maintaining levels of ROS that regulate the β‐catenin pathway through dephosphorylation (Liao et al, 2017). In addition, oncogenic RAS transformation can redirect glutamine to maintain cellular redox balance by stimulating the transcription of cystine/glutamate antiporters to enhance glutathione levels, ultimately contributing to tumor progression (Lim et al, 2019). Reductive glutamine metabolism can also regulate oxidative stress by supporting NADPH production within the mitochondria to quench ROS via glutathione generation (Jiang et al, 2016b). Hence, dependence of cancer cells on fatty acid oxidation in vivo has been linked to the promotion of antioxidant defense via NADPH production as well as enhanced drug resistance in glioblastoma and gastric cancer (Pike et al, 2011; He et al, 2019).

These nutrients can provide defense not only against chemotherapy and detachment‐induced stress but also against the mechanical damage induced by circulatory shear stress. Using a model of metastatic breast cancer, Fu and colleagues show that ROS buffering via high manganese superoxide dismutase activity enhances in vivo breast cancer CTC resiliency to circulatory shear stress, increasing the number of cells that successfully metastasize to the lung (Fu et al, 2016). Interestingly, a microfluidics assay of breast, lung, and ovarian CTCs revealed that the high shear stress experienced in arteries during exercise is enough to cause ROS‐driven necrosis of nearly 90% of CTCs (Regmi et al, 2017). Taken together, these studies suggest the enhancement of antioxidant pathways after dissemination in CTCs may initially serve to combat the physical strains of circulatory travel across different physiological states, but then provide an additional benefit toward resisting chemotherapies with oxidative damage‐driven mechanisms. Hence, targeting the regulation of nutrients influencing ROS maintenance can improve current treatments by influencing the survival and drug‐resistant properties of CTCs.

How does the local environment of secondary organs influence the metabolic phenotypes of metastases?

Despite our understanding of metabolic demands and plasticity in primary tumors, the role of nutritional environments once CTCs reach secondary sites of metastasis is less clear. In the metastatic sites, disseminated cancer cells may adapt their metabolism to promote metastatic seeding and colonization. Cancer cells that seed in distant tissues can transition between quiescent and proliferative cell states, depending on the presence or absence of optimal growth signals and conditions (Wells et al, 2013; Lambert et al, 2017). For proliferative cancer cells, glutamine and asparagine have been shown to be instrumental in fueling the mesenchymal‐to‐epithelial (MET) switch in metastasis (Krall et al, 2016; Brabletz et al, 2018; Luo et al, 2018). In the 1960s, asparagine was first identified as a metabolic vulnerability for acute lymphoblastic leukemia (ALL)—whereby treatment of patients with the enzyme L‐asparaginase depletes asparagine from the plasma by converting it to aspartic acid and ammonia (Broome, 1961; Egler et al, 2016). More recently, roles for asparagine in metastasis have been uncovered. L‐asparagine content is selectively elevated in proteins that drive EMT during the metastatic invasion, and limiting de novo asparagine synthesis via asparaginase treatment significantly reduced the capacity for in vivo colonization in the 4T1‐T mouse model (Fig 5B) (Knott et al, 2018). This may be due to asparagine's role in stimulating the mammalian target of rapamycin (mTOR)‐mediated synthesis of proteins necessary for EMT during invasion, even in glutamine‐deprived secondary sites (Pavlova et al, 2018). In contrast to the nutrients that are involved in CTC seeding, much less attention has been given to the environmental cues that trigger dormant states in cancer (Aguirre‐Ghiso & Sosa, 2018). Investigating these influences may lead to further understanding of why some metastatic cancers can remain dormant for long periods before relapse.

Growing evidence indicates that primary tumors and their metastasis are metabolically different, suggesting that the local environment affects fitness for metastatic outgrowth. One of these adaptations has been observed during breast‐derived lung metastasis where, once early metastatic lesions are established, pyruvate further drives PC activity, which supports the TCA cycle anaplerosis needed for proliferation (Christen et al, 2016). Accordingly, genetic deletion of PC impaired pulmonary but not extra‐pulmonary metastasis—suggesting that PC activity could be important for organ tropism (Shinde et al, 2018). The metabolic changes induced by the organ microenvironment were also found to be reversible with the observation that transplanting liver metastasis‐derived cancer cells back into the primary subcutaneous site induced a parental metabolic state (Piskounova et al, 2015). This suggests that metastases of different tumors growing in the same organ might be metabolically more similar than metastases of the same primary tumor growing in different organs. Overall, oncogenic mutations drive nutritional requirements during (primary) tumor initiation and formation, whereas the nutrient status of the tumor microenvironment in distant organs is a key contributing factor supporting metastatic growth of the (secondary) tumor. Further work is needed in this area, particularly to establish specific nutrient content of different human organs and to understand how diet affects nutrient fluctuation and availability in the tumor microenvironment.

Does the phenotype of disseminated cancer cells drive the metabolic dependencies during metastasis?

Metabolic phenotypes of cancer cells can delimit the fate of those cells that are spread from the primary tumor to other tissues. For example, it has been shown that CTCs rely on lactate uptake by MCT1 to manage oxidative stress in PDX models of melanoma. MCT1 phenotype increases the chances of successful seeding by managing more surviving disseminated cancer cells and thereby increases metastasis formation. Interestingly, MCT1 inhibition showed little effect on primary subcutaneous tumors but depleted circulating melanoma cells and reduced metastatic burden (Tasdogan et al, 2020). This highlights an important consideration: The circulatory environment (as opposed to the primary tumor) may select for those CTC populations with a metabolism that sustains the unique demands of circulatory travel, and phenotypic heterogeneity in CTC fitness confers differences in metastatic potential. Once CTCs successfully extravasate and arrive at a distant organ, their phenotype can also dictate the nutrient requirements of these cells to successfully colonize the new tissue. For example, breast cancer cells colonizing lung tissue rely on pyruvate to remodel the extracellular matrix (ECM) of the metastatic niche (Elia et al, 2019), which benefits metastatic outgrowth. As a result, inhibiting the monocarboxylate transporter for pyruvate (MCT2) or mitochondrial alanine aminotransferase enzyme (involved in pyruvate metabolism) is sufficient to impair lung metastasis formation in an in vivo model of breast cancer metastasis (Elia et al, 2019).

The metabolic phenotypes of primary tumor subclones may also predict which secondary sites they will be able to colonize. For instance, depending on the site of metastasis, metastatic breast cancer cells display a different PPAR‐gamma coactivator 1 alpha (PGC‐1α) expression likely acquired from a heterogeneous primary tumor. Unlike liver or bone, in lung tissues, a PGC‐1α‐dependent breast metastatic phenotype drives bioenergetic flexibility in colonizing cells by promoting both the glycolytic and mitochondrial functions of cancer cells, features likely required to colonize this organ (Andrzejewski et al, 2017). In contrast, liver metastasis from breast cancer tumors engages distinct metabolic programs, characterized by an increase in aerobic glycolysis and a concomitant reduction in mitochondrial metabolism. This phenotype is driven by expression of HIF‐1α‐target Pyruvate dehydrogenase kinase‐1 (PDK1), whose activity is required for efficient liver metastasis (Dupuy et al, 2015). Although the metabolic plasticity of tumor cells may enhance their propensity for metastatic seeding (McGuirk et al, 2020), colonizing specific secondary sites may demand phenotypes that impose strict metabolic and nutritional requirements on CTCs. Consequently, different populations of cells leaving a heterogeneous primary tumor may be selected for upon reaching secondary sites based on the ability of their distinct metabolic programs to facilitate the requirements for tissue‐specific colonization.

Are there additional external factors that affect nutrient availability in the secondary organ?

It becomes apparent that the metabolic traits of secondary tumors are, in part, nutrient environment‐dependent. Additionally, there are complementary levels of manipulating the nutrient environment that influence the adaptation of metastatic cells to a new organ. One of these additional factors is the diet, which can affect nutrient availability to both the primary tumor as well as the tissues for potential secondary tumor establishment. Thus, increased dietary asparagine intake promotes metastatic progression and, more interestingly, asparagine restriction reduces metastasis without affecting the growth of the primary tumor (Knott et al, 2018). Another important regulator of the tumorigenic process is the tumor secretome. The secretome consists of key regulatory molecules, derived from the primary tumor or tumor‐associated cells, which influence the microenvironment of secondary organs—priming them for the cultivation of secondary tumors (Shinde et al, 2018). In this regard, recent metabolomics analysis of the in vivo secretome of lung cancer models revealed that increased succinate secretion into plasma by cancer cells induces macrophage‐dependent cytokine signaling in the metastatic niche (Wu et al, 2019). This signaling stimulates cancer cell migration and EMT, enhancing cancer metastasis. Interestingly, serum succinate has been found to be significantly higher in 97 lung cancer patients compared to 21 healthy patients and warrants further study to understand the effect on disease progression (Wu et al, 2019). Furthermore, the secretome can influence distal stromal cells to cultivate a more permissive tumor microenvironment for metastasis. In metastatic breast cancer, primary tumor‐secreted miRNA‐122 suppresses glucose utilization in stromal cells to accommodate the elevated energy demand of cancer cells during metastatic growth (Fong et al, 2015). In other words, these factors can support metastatic cells in the competition for limited nutrient supplies.

Another key component of both primary and metastatic tumor growth is the manipulation of the host immune system (Fig 5C). Immune avoidance is a well‐documented hallmark of cancer progression, and recent work has helped clarify the interplay between nutrient availability and tumor immunogenicity (Doglioni et al, 2019). For example, monocytes have been implicated in fueling proliferative cancer metabolism (Mondanelli et al, 2019; Vitale et al, 2019). Tumor cell metabolism can also deplete local nutrient levels, driving immune suppression, and immune cell‐tumor interactions that allow for more favorable tumor growth. An inverse relationship between T‐cell infiltration and glucose metabolism in squamous cell carcinoma has been described (Ottensmeier et al, 2016). This was attributed to the consumption of local glucose by tumors as a mechanism to exhaust glycolysis‐dependent T‐cell infiltrates into a quiescent‐like state. In addition to depleting resources, cancer cells can also stimulate immunoregulatory pathways in local lymphocytes to avoid cytotoxicity. Highly glycolytic tumors cultivate a tumor microenvironment with low pH and by secreting high concentrations of lactate, which suppresses T‐cell activation and NK cell function in the metastatic niche (Huber et al, 2017; Payen et al, 2019). In contrast to the direct depletion of the glycolytic substrate (glucose) from activating T cells, acidic tumor microenvironments halt T‐cell expansion by inhibiting the export of lactate by MCT1. This transporter requires a favorable concentration gradient for co‐transport of H+/lactate into the extracellular space; thus, acidic microenvironments with high lactate do not favor lactate export from T cells and result in suppression. In addition, lactate release from glycolytic tumors can stabilize HIF1α in tumor‐associated macrophages. This HIF1α stabilization in macrophages promotes the expression of a HIF1α‐stabilizing lncRNA that is delivered back to tumor cells through extracellular vesicles—fueling tumor growth, lactate production, and T‐cell/NK cell suppression in a positive feedback loop (Chen et al, 2018a; Vitale et al, 2019). Similar immunoregulatory mechanisms have been described for macrophages, myeloid‐derived suppressor cells (MDSCs), and dendritic cells in the in vivo tumor microenvironment. Tumors can stimulate these cells to upregulate key catabolic enzymes, such as arginase 1 (ARG1) or indoleamine 2,3‐dioxygenase 1 (IDO1), depleting local arginine and tryptophan which are essential for T‐cell proliferation and differentiation in vivo (Mondanelli et al, 2019). Thus, there is an emerging role for tumor‐associated immune cells in regulating the local nutrient microenvironment to promote anti‐inflammatory responses in cancer. Cancer cells can also interact with other stroma cells to shape the metabolic requirements to promote metastasis. For instance, adipocytes can transfer fatty acids to disseminated cancer cells arriving at organs exhibiting adipocyte‐rich environments for fueling tumor growth. It has been shown that primary human omental adipocytes promote homing, migration, and invasion of ovarian cancer cells by upregulation of fatty acid‐binding protein 4 (FABP4) on the adipocyte‐tumor cell interface (Nieman et al, 2011).

The commonly accepted “seed and soil” hypothesis of metastasis (Mathot & Stenninger, 2012) acknowledges the nutritional suitability of the soil (tissue) to promote seeding (metastasis). However, recent data suggest this concept needs to incorporate the “composition” of the soil, that is, the other physical components of secondary sites (such as vascular/mesothelial cell density, EMT signals, ECM composition) that affect the ability of seeds to penetrate the soil and surface. Cancer cells in secondary sites must receive extracellular invasion signals to trigger EMT, and cells must adapt their metabolism to support this invasion phenotype. ECM modifications are also necessary during metastatic seeding, where metabolic reliance on pyruvate in the lung is shaped by a phenotypic need for breast cancer cells to remodel ECM—rather than to fuel growth (Elia et al, 2019). Endothelial transmigration also depends on “soil density” and the ability of CTCs to process ECM proteins to navigate through dense tissues (Jiang et al, 2015). Some studies in murine models have also investigated the influence of tissue vasculature and endothelial surface molecules in determining metastatic potential (Ruoslahti & Rajotte, 2000; Bugyik et al, 2016). This “soil composition” seems to be therapeutically exploitable, where recent targeting of angiogenic tissue vasculature by a synthetic peptide disrupted the permeability of pre‐metastatic lung vasculature, preventing metastatic seeding in mouse models of lung cancer and melanoma (He et al, 2020). Thus, not only nutrient availability but phenotypic requirements of cancer cells in secondary sites dictate metabolic needs, as these necessitate changes in metabolism to support successful metastasis.

Concluding remarks

Due to the high relevance of the nutrient dependence in cancer cells to support continued growth, there is a significant therapeutic window for metabolism‐based cancer treatment. Although currently under development, targeting the nutrient requirements of cancer is emerging as an effective method of improving existing therapy approaches (Garcia‐Bermudez et al, 2020). Perhaps one of the earliest methods of targeting nutrient dependencies in blood cancers was the reduction of serum L‐asparagine via guinea‐pig‐isolated L‐asparaginase treatment to treat acute lymphoblastic leukemia (Dolowy et al, 1966). Since this landmark discovery, other means of “nutrient‐based metabolic therapy” have been explored. Similar to L‐asparaginase, an engineered cystathionine gamma‐lyase enzyme with a higher affinity for L‐cysteine than L‐cystathionine was designed to deplete serum cystine (Cramer et al, 2017). By systemically depleting cystine in prostate cancer xenograft mouse models, treated animals showed a marked impairment in tumor growth. In addition, since the influence of the diet on nutrient availability is a self‐evident fact, dietary interventions are emerging as synergistic with traditional treatments to improve the efficacy of anti‐cancer therapies (Lévesque et al, 2019). Alternatively, combining metabolic drugs with dietary interventions (Hopkins et al, 2018) and anti‐angiogenic therapies (Pisarsky et al, 2016) or targeting those nutrient transporters that are enriched in tumor cells hold great promise for diagnosis (Zhang & Wang, 2020) and therapeutics (Arensman et al, 2019; Elia et al, 2019). Thus, understanding the adaptations of tumor cells to nutrient fluctuations will open a new therapeutic window to improve the clinical outcome of cancer patients.

However, it still remains a significant challenge to successfully translate the nutritional dependencies that are found in cancer into patient‐relevant interventions. This is in part due to the stark contrast between broad nutrient composition available to tumors in various tissues in vivo compared to the selective nutrient availability in standard in vitro culture medium. Additionally, the inability to fully recapitulate the physiological complexity of the tumor microenvironment in vitro (i.e., nutrient gradients, stromal populations, tissue‐resident lymphocytes, varying diet in patients) to accurately recapitulate disease states merits consideration. For many nutrients, physiological concentrations in organs have been estimated in animal models. It remains to be seen whether the nutrients levels found in mouse tumors are the same as those in humans. This may drive discrepancies in drug response between tumor cells in vivo and cells in culture (Muir & Vander Heiden, 2018) or variable drug sensitivity within tumor due to the metabolic heterogeneity that exists between and within human tumors (Faubert & DeBerardinis, 2017). In fact, combination therapies that address intratumor heterogeneity need to be further considered to improve the outcome of some of the current metabolic‐based cancer treatments.

Unfortunately, as cancer recurrence is common for many patients, addressing the role of nutrient utilization and metabolism in persisting cells in different tissues is of substantial interest. Since recurrence often occurs in sites distinct from primary cancers, comprehensive investigations of the metabolic dependencies in various tissues or environments are desperately needed. As outlined, various populations of cells may arise during cancer progression—each “giving” and “taking” certain nutrients or soluble factors in an exchange with the local environment. Thus, cancer cells play an integral part in shaping the local microenvironment where subsequent daughter cells will need to adapt. To better understand the dynamic changes that occur during this cancer development, single‐cell RNA sequencing technology provides the power to analyze even rare populations of circulating and solid tumor cells for distinct metabolic signatures that develop under various contexts (Lambrechts et al, 2018; Rambow et al, 2018; Xiao et al, 2019). Leveraging the ability of single‐cell technology to unmask heterogenous cell populations with high throughput methods of cellular metabolomics represents the next major step for identification of metabolic signatures among cancer cell populations. This information may contribute to nutrient‐targeting therapies that can supplement common chemotherapeutics or elucidate novel targets for personalized cancer therapy.

In need of answers.

1. Could cancer type‐specific nutrient dependencies be translated into precision medicine targets? Could circulating nutrient levels be used to predict response to specific cancer treatments?

2. Can dietary interventions modify the local environment to be able to control the nutrient availability in tumor niches? Can these dietary interventions be combined with metabolic therapies to improve patient response? Is this a therapeutic window to translate dietary interventions to prevent metastatic diseases?

3. Since the role of diet has been shown to influence cancer progression, can dietary interventions modify circulatory nutrient levels that are supporting the requirements of CTCs? Can an altered nutrient intake prevent seeding of these disseminated cancer cells or the awakening of dormant cells?

4. Is there a specific nutrient that can stimulate metastatic cells to shift from the quiescent state to proliferative? Could this present an option to target cellular metabolism to prevent the growth of secondary tumors?

5. What factors determine which cells in a heterogeneous tumor are more likely to disseminate and enter the circulation? Is there a specific nutrient component allowing this population of cells to be more fit in a specific secondary organ?

Conflict of interest

S.‐M.F. has received funding from Bayer, Merck, and Black Belt Therapeutics and has consulted for Fund+. The other authors have no conflict of interest to declare.

EMBO Reports (2020) 21: e50635