Cancer Metabolism Historical Perspectives: A Chronicle of Controversies and Consensus

Abstract

A century ago, Otto Warburg's work sparked the field of cancer metabolism, which has since taken a tortuous path. As evidence accumulated over the decades, consensus views of causes of cancer emerged, whereby genetic and epigenetic oncogenic drivers promoted immune evasion and induced new blood vessels and neoplastic metabolism to support tumor growth. Neoplastic cells abandon social cues of intercellular cooperation, escape tissue confinement, metastasize, and ultimately kill the host. Herein, key milestones in the study of cancer metabolism are chronicled with an emphasis on carbohydrate metabolism. The field began with a cancer cell–autonomous view that has been refined by a richer understanding of solid cancers as growing, immune-suppressive, complex organs comprising different cell types that are nourished by a variety of nutrients and variable amounts of oxygen through abnormal neovasculatures. Based on foundational historical studies, our current understanding of cancer metabolism offers a hopeful outlook for targeting metabolism to enhance cancer therapy.

The origin of life is thought to arise in part from a nascent nonenzymatic glycolytic pathway that is common to all self-sustaining life forms (Ralser 2018). In fact, the last hypothetical universal common ancestor (LUCA) of all cells is surmised to use nonoxidative sugar metabolism (Weiss et al. 2016). LUCA evolved long before the Great Oxidation Event resulting from the proliferation of photosynthetic organisms some 2.4 billion years ago (Holland 2006). The availability of oxygen enabled the emergence of mitochondria as synthetic organelles to power eukaryotic metabolism and promote the evolution of metazoans. Oxygen availability also leads to reactive, toxic metabolic wastes. These byproducts, in addition to exogenous genotoxins, can corrupt the life-propagating encoded information, leading to cell growth arrest, death, or neoplastic cell transformation. In this way, the evolved metabolic processes of metazoans not only enable such complex life forms but also imperil them. In turn, deregulated neoplastic cell growth and proliferation also depend on these evolved metabolic pathways, but in ways that may distinguish neoplastic from physiologic metabolism.

EARLY CONCEPTS OF CANCER BIOLOGY

In 1885, Ernst Freund reported that blood glucose levels were elevated in individuals with cancer (Freund 1885) and proposed that glucose must sustain cancer. On this subject, the New York Times reported on December 23, 1887 that blood from the German Prince Frederick III, who married Queen Victoria's daughter, was to be analyzed for excess sugar to determine whether his laryngeal nodule was a cancer (Baron 1999). During this time, a small laryngeal biopsy by Morell Mackenzie, a leading head and neck surgeon sent by Queen Victoria, was rendered nonmalignant by Rudolph Virchow (Baron 1999). Based on the biopsy results, Mackenzie suggested that extensive surgical resection with high morbidity should not be undertaken. Unfortunately, for Prince Frederick, the nodule was not benign and progressive cancer was later diagnosed (Mackenzie 1888). The result of the Freund “diagnostic” blood test could not be found in the detailed Mackenzie report (Mackenzie 1888). In any event, the validity of Freund's test was questioned by a subsequent December 24, 1887 New York Times precis pointing to the possibility that the association of high blood sugar and cancer could be coincidental in patients with preexisting diabetes mellitus, and that one condition did not cause the other. Intriguingly, diabetes with hyperinsulinemia is now known to be a major risk factor for developing cancer (Gallagher and LeRoith 2020).

In parallel around the same time, some of the first information about cancer's interactions with the immune system and genomic alterations were uncovered. The surgeon William Coley noted that a patient with sarcoma underwent complete remission after a severe postsurgical wound infection with Streptococcus pyogenes and surmised that the infection was critical for the cure (Hoption Cann et al. 2003). In the 1890s, he developed a vaccine of killed bacteria (known as “Coley's toxins”) to inoculate his patients and found complete remission of a sarcoma in his first patient. Arguably, this was the first cancer immunotherapy approach without the knowledge of the immediate cause of cancer or the role of tumor immunity. In 1902, Theodor Boveri speculated that malignant tumors could result from “certain abnormal chromosome constitution” based on his studies of chromosomes of fertilized urchin eggs and the appearance of tumor-like growths when chromosomes were present in imbalanced numbers (Boveri 2008), laying the groundwork for modern cancer genomics.

In the early 1900s, the cause of aneuploidy and the substance of heredity in chromosomes were unknown, but a clue for the cause of cancer was reported in a seminal 1911 paper (Rous 1911) by Peyton Rous, documenting that chicken tumor cell-free extracts can induce avian cancers. In this paper, Rous also noted spontaneous regression of these chicken tumors, which were found to have an “accumulation of lymphocytes,” now known as tumor-infiltrating lymphocytes. Rous’ observations were initially obscured but later led to the discovery of the Rous sarcoma virus, which proved to be foundational for the discovery of retroviral oncogenes.

The biochemistry of cancer was boosted by Otto Warburg who won the Nobel prize in 1931 for his discovery of cytochrome c oxidase (Warburg 1928). It was not his Nobel discovery, but a series of papers (Warburg 1930) in the 1920s underscoring the connection between altered glucose metabolism and cancer that brought Warburg into the limelight of cancer research. He provided evidence for what he believed to be the key cause of cancer—damaged cellular respiration.

In 1933, Hans Krebs found that among amino acids, glutamate-exposed guinea pig kidney consumed the most oxygen and noted an accompanying diminished ammonia level (Krebs 1935). In the same year, Dickens and Greville also noted that spleen, Jensen rat sarcoma, and rat or chick embryos produced large amounts of ammonia in the absence of sugar (Dickens and Greville 1933). Krebs reported in 1935 the findings of an enzyme system for the synthesis of glutamine from glutamate and ammonia as well as the enzyme hydrolysis of glutamine, reversing the reaction (Krebs 1935). These observations documented the existence of glutamine synthetase and glutaminase, which are now known to play critical roles in tumor glutamine metabolism (Altman et al. 2016).

In 1945, Leuchtenberger and colleagues reported the striking finding of complete remissions of spontaneous murine mammary tumors treated with folic acid (Leuchtenberger et al. 1945). Based on these results, Sidney Farber treated 11 children with lethal acute lymphocytic leukemia (ALL) with folate and observed an “acceleration phenomenon” in the bone marrow of these patients (Farber 1949). This unexpected acceleration of the leukemia led Farber to the idea of using antifolates to treat leukemia. Farber reported clinical responses of childhood ALL to aminopterin in the landmark 1948 paper (Farber and Diamond 1948), underscoring the importance of inhibiting one-carbon metabolism in cancer and laying the foundation for modern chemotherapy (Stine et al. 2022). It should be noted that acute leukemias, which can proliferate in circulation, are distinctly different from solid tumors that have complex tumor microenvironments and grow slower. The rapid proliferation of leukemias requires heightened metabolism that renders them more responsive to cytotoxic therapies. A recent study using in vivo isotopic labeling and mass spectrometry (Fig. 1) underscores the difference between liquid and solid tumors, showing that tricarboxylic acid (TCA) cycling is higher in leukemia compared to solid tumors (Bartman et al. 2023).

Figure 1.

From Warburg manometer to mass spectrometry. Cancer metabolism research over the last century has been advanced by emerging technologies from the (A) Warburg manometer, to (B) mass spectrometry, and (C) illustration showing spatial mass spectrometry imaging of a tissue section that generates relative densities of metabolite signal intensity across the tissue section. The graph (right) depicts cell counts as a function the distribution of signal intensities (Ci) of a metabolite. Each technological advance offers additional views of the complexity of cancer metabolism particularly in the context of the tumor immune microenvironment. (GC) Gas chromatography, (LC) liquid chromatography, (EC) electrochemical. (Figure created with BioRender.com.)

Cancer neovascularization emerged in 1945 (Algire et al. 1945) as another key concept in cancer biology and one intimately connected to energy delivery and waste disposal. The work by Algire et al. recorded the appearance of new blood vessels in grafted normal tissues or tumor grafts. They observed that vascularization of normal transplanted tissues increased over a week and the emergence of arterioles and venules became visible. In contrast, tumors recruited new capillaries rapidly over 3 days evolving into large vessels that did not develop into arterioles or venules. The tumor neovasculature is disordered but the tumors continued to be able to recruit new vasculature as they grew (Folkman et al. 1971). Without neo-angiogenesis, solid tumors would be limited to sizes less than ∼200 µm in diameter, the limit of tissue oxygen diffusion (Carmeliet and Jain 2000).

In retrospect, synthesis of observations from the late 1800s into the 1950s provides a picture of solid cancer as a neoplastic mass, often with genomic changes, which can arise from a cell-free viral tumor extract, requires neovascularization, consumes glucose to produce lactate, consumes amino acids, converts glutamine to glutamate and ammonia, and is sensitive to one-carbon metabolism inhibition. The apparent conflicting observations that folate reduced the growth of mammary tumors but accelerated childhood leukemia suggest that fast-growing liquid tumors require folate for neoplastic growth and perhaps in the case of mammary tumors, folate may be required for the function of the antitumor arm of the immune system, although that was not appreciated at the time (Ron-Harel et al. 2016). More clearly, the observations of complete remissions induced by Coley's toxin underscored the importance of tumor immunity even before key concepts of innate and adaptive immunity were known.

OTTO WARBURG, CARL AND GERTY CORI, AND AEROBIC GLYCOLYSIS

Otto Warburg was a meticulous quantitative biochemist who innovated the “Warburg” manometric apparatus (Fig. 1A) that permitted precise measurements of glucose and oxygen consumption as well as carbon dioxide and lactate production by thin slices of normal or cancer tissues. His early studies of sea urchin eggs led to the finding that, upon fertilization, there was a rapid rise in oxygen consumption. Hence, he postulated that cancer tissue, being proliferative, would consume high amounts of oxygen relative to normal tissues. Instead of higher oxygen consumption, he reported in 1924 (Warburg 1930) that the Flexner–Jobling rat liver carcinoma tissue slices did not take up more oxygen than normal liver, but rather the carcinoma produced more lactate than normal liver under oxygenated conditions (Warburg 1925). Known as the Pasteur effect first described in 1861 (Racker 1974), oxygen was documented to suppress glycolysis in yeast. The converse whereby glucose suppresses respiration is known as the Crabtree effect (Crabtree 1929). As such, the heightened glycolytic feature in cancer tissues bypasses the Pasteur effect resulting in aerobic glycolysis, the ability to undergo glycolysis in the presence of oxygen that was coined the “Warburg effect” by Efraim Racker (Fig. 2; Racker 1972).

Figure 2.

Pasteur, Crabtree, and Warburg effects. Generalized mammalian cells are depicted with the consumption of glucose through mitochondrial oxidation or glycolysis. Pasteur described the ability of oxygen to suppress yeast glycolysis that produces ethanol (not lactate as illustrated for mammalian cells), a phenomenon known as the Pasteur effect. Conversely, Crabtree found that some yeast strains demonstrated the ability of glucose to suppress respiration, known as the Crabtree effect. Warburg hypothesized that damaged mitochondria in cancer cells result in enhanced aerobic glycolysis, which bypasses the Pasteur effect, termed the Warburg effect. (Figure created with BioRender.com.)

The propensity of cancers to take up glucose avidly and convert the vast majority to lactate, or the Warburg effect, became a paradigm for cancer research in the early and mid-1900s. This concept, generated largely from in vitro experiments, was studied in tumors by Carl and Gerty Cori and reported in 1925 (Cori and Cori 1925a,b). They found that glucose levels tend to be diminished in mouse and rat tumors compared to normal muscle. Likewise, tumor lactate levels were diminished compared to muscle in fasting animals. Hence, they reasoned that tumor lactate might be washed away by blood circulation and surmised that an increase in glucose by intraperitoneal injection could reveal the propensity of tumors to produce high lactate levels. Indeed, when glucose was administered, tumor glucose levels rose significantly and were accompanied by an elevation of tumor lactate, a phenomenon that was not seen with normal liver. Further, they found that blood lactate levels were more elevated in tumor-bearing animals than in non-tumor-bearing animals after glucose administration. Thus, they concluded that the in vivo experiments were not contradictory to the in vitro findings of Warburg, but rather the production of tumor lactate depends on the availability of circulating glucose. Their continued studies of glucose and lactate metabolism led to the 1947 Nobel prize discovery of the Cori cycle (Cori and Cori 1929), the conversion of glucose by muscle to lactate that in turn is converted to liver glycogen, which can be mobilized to produce circulating glucose (Fig. 3A).

Figure 3.

In vivo Warburg effect, the Cori cycle, and in vivo cancer positron emission tomography (PET) imaging. (A) By sampling arterial and venous blood across rodent normal organs, such as the liver, and tumors (green), Crabtree observed that increased glucose resulted in a higher lactate venous output from tumors than normal tissues, which tend to take up lactate from arterial blood. Warburg also documented that tumors have a propensity to convert high levels of glucose to lactate in vivo. In the case of liver, glycogen is produced through gluconeogenesis from muscle-generated lactate and in turn glucose released from glycogen can then be used by muscle in an interorgan circuit termed the Cori cycle. (B) The Warburg effect is exploited clinically to diagnose and monitor human cancers using an ¹⁸F-fluorodeoxyglucose PET scan. Normal heart and liver also accumulate ¹⁸F-labeled deoxyglucose, but tumors tend to have abnormally high uptake of the tracer (green). (Figure created with BioRender.com.)

Corroborating earlier studies of Rous sarcomas (Cori and Cori 1925b), Warburg and colleagues published in 1927 (Warburg et al. 1927) the study of tumor metabolism in vivo. The experimental approach was meticulous, requiring the dissection of normal or tumor arterial and venous vessels from the anesthetized animal for the collection of efferent and afferent blood (Fig. 3A). The major blood vessels were sampled, and a drop in glucose level was found in each case from the arterial to venous side. Compared to these normal differences in glucose levels, the drop across Jensen sarcomas were pronounced, suggesting that the consumption of glucose was higher in the tumor. When measuring arteriovenous differences in lactic acid level, they found that most organs consumed lactate, except for the brain (i.e., comparing levels in arterial vs. venous “Jugularis”). In contrast to evidence of lactate consumption by normal tissues, in all 10 Jensen tumors, venous lactate was much higher than arterial levels, indicating that Jensen tumors consumed glucose and produced lactate. These studies corroborated the findings by the Coris a few years earlier (Cori and Cori 1925b) and supported the notion of the Warburg effect in tumors (Fig. 3A). Intriguingly, these historical studies are now largely substantiated by more sophisticated mass spectrometry (Fig. 1B) with the use of isotopically labeled substrates such as glucose, lactate, or 2-deoxyglucose and modeling of metabolite distributions in vivo (Faubert et al. 2017; Liu et al. 2020; Bartman et al. 2023). Although lactate produced from glucose can be oxidized by tumors (Faubert et al. 2017), the Warburg effect has been documented in solid tumor models (Bartman et al. 2023), underscored by the utility of ¹⁸F-2-deoxyglucose clinical imaging of human cancers (Fig. 3B; Som et al. 1980; Nolop et al. 1987).

CANCER METABOLISM CONTROVERSIES

The dogma of the Warburg effect providing an oversimplified view of cancer metabolism began to be challenged with controversies that crescendoed into the 1960s. Crabtree sought to determine whether Warburg metabolism is an “exclusive feature of malignant tissues” and whether anaerobic versus aerobic glycolysis have any relationships to the magnitude of respiration (Crabtree 1928). Crabtree cited several publications documenting that nonmalignant tissues, such as the retina, placenta, and leukocytes, have high aerobic glycolysis, thereby questioning the validity of the Warburg hypothesis. Further, Crabtree used the Warburg manometer to study infectious nonmalignant lesions, such as pigeon pox, chicken vaccinia, or human warts or papillomas. He found that the excess glycolysis in pigeon pox slices was in the same order as those found by Warburg for tumor slices. Crabtree documented the elevation of glycolysis in Rous sarcoma tumors but surmised from his findings that changes accompanying the Warburg effect “… are not specific for malignant tissues but are a common feature of pathological overgrowths.” Crabtree's historical findings presaged later studies that document virus-induced cellular glycolytic metabolism (Bissell et al. 1972; Thai et al. 2014).

The debate on the role of the Warburg effect in malignancies continued with camps on both sides digging into their positions. In studies of normal and tumor tissues, Elliott and Baker (1935) did not find differences in the Warburg effect between normal and tumor tissues (Elliott and Baker 1935) as compared to studies by Dickens (Dodds and Dickens 1940). Further, Boyland reported for the British Empire Cancer Campaign in 1940 and mentioned glycolysis, which can be high in normal tissues, and hence is “… therefore impossible to consider this characteristic to be peculiar to tumours” (Boyland 1940a). This report resulted in a debate in Nature (March 30, 1940) between Dickens and Boyland about the merits of the Warburg effect in cancer (Boyland 1940b).

The ongoing debate on the Warburg effect was punctuated by Warburg's 1956 Science article that provided an overview titled “On the Origin of Cancer Cells” (Warburg 1956). He wrote with dogmatic authority, unshaken by contradictory data, that cancer cells have injured respiration (Fig. 2), and the resulting aerobic glycolysis causes cancer. He dismissed the roles of carcinogens and viruses in cancer, stating that “From this point of view, mutation and carcinogenic agent are not alternatives, but empty words, unless metabolically specified. Even more harmful in the struggle against cancer can be the continual discovery of miscellaneous cancer agents and cancer viruses, which, by obscuring the underlying phenomena, may hinder necessary preventive measures and thereby become responsible for cancer cases.” Warburg's views on damaged respiration that drives glycolysis as a cause of cancer were challenged by Sidney Weinhouse (Weinhouse 1956) citing that isotope tracer studies revealed no difference between tumor and normal tissue in their conversion of glucose to carbon dioxide. Dean Burk (Burk and Schade 1956), another key figure, tipped the scale toward Warburg's aerobic glycolysis as a feature of cancer, but Burk acknowledged the validity of Weinhouse's objection to the concept of damaged mitochondria as a driver for malignancies. Warburg was wrong to dismiss an active role of mitochondria in tumorigenesis, in particular, since evidence shows the importance of mitochondrial function in cancer (Vasan et al. 2020).

Efraim Racker was a prolific biochemist who contributed fundamental insights into carbohydrate metabolism. His entry into cancer metabolism began with fundamental studies of glycolysis in the Ehrlich ascites tumor cells, demonstrating that the conversion of glucose to lactate in cell extracts could be enhanced by the additional of purified phosphofructokinase and glyceraldehyde-3-phosphate dehydrogenase together with hexokinase, thereby defining the limiting glycolytic steps in ascites tumor extracts (Wu and Racker 1959). Skeptical of Warburg's damaged mitochondria hypothesis, Racker proposed that there are multiple causes of cancer, which share in common inefficient sodium-potassium ATPase pumps associated with aerobic glycolysis (Racker 1972). In 1981, Racker and Spector (1981) reported that the Src oncogenic kinase phosphorylates and suppresses the ATPase pump and thereby promotes aerobic glycolysis. This putative first link between an oncogene and the Warburg effect further overshadows observations of aerobic glycolysis in normal cells, such as mitogen-activated lymphocytes (Hedeskov 1968) that in retrospect were perhaps the first reported glimpse of immunometabolism.

Racker's striking report of a link between an oncogene and tumor metabolism was, unfortunately, the result of scientific misconduct by his graduate student Spector (Racker 1989). The harbinger of misconduct was uncovered by the finding that ¹²⁵Iodine was spiked in his student's experiments to mimic the results of ³²P in the phosphorylation studies. The notion that Src drives the Warburg effect evaporated with this scandal. However, in 1983, Cooper in Hunter's laboratory and colleagues (Cooper et al. 1983) reported that enolase, phosphoglycerate kinase, and lactate dehydrogenase (LDH) were tyrosine phosphorylated in cells transformed by the Rous sarcoma virus bearing the v-Scr oncogene, but the functional significance was unclear. During this time, an early study of positron emission tomography (PET) using ¹⁸F-fluoro-2-deoxyglucose (FDG) showed enhanced glucose tumor uptake, assumed to be the Warburg effect, correlated with the degree of malignancy of cerebral gliomas (Di Chiro et al. 1982). The use of FDG PET (Fig. 3B) to detect altered cancer metabolism expanded (Hillner et al. 2008) and is now a standard of practice in clinical oncology.

The Warburg effect controversies distracted the literature from the key findings of Krebs (1935) and Dickens (Dodds and Dickens 1940) on the conversion of glutamine to glutamate and ammonia by normal tissues and the Jensen rat sarcoma. Glutamine was further shown by Eagle and coworkers in 1956 to be essential for mammalian cell growth in vitro (Eagle et al. 1956), providing the basis for Basal Medium Eagle. In 1983, consumption of glutamine was found to be increased in stimulated rat lymphocytes resulting in the production of glutamate, aspartate, and ammonia (Ardawi and Newsholme 1983). Brand reported (Brand et al. 1984) that concanavalin A–activated lymphocytes increased expression of glycolytic enzymes, enhancing glucose metabolism by 54-fold, whereby glucose was converted 90% to lactate and 1% was consumed for respiration. This contrasts with resting lymphocytes that oxidize 27% of the glucose to CO₂. Glutamine use increased by eightfold in stimulated lymphocytes, producing glutamate, ammonia, aspartate, and CO₂. These foundational observations were largely forgotten in the current literature, but undoubtedly paved the way for recent studies on the use of glucose and glutamine for cancer metabolism (Cairns et al. 2011; DeBerardinis and Chandel 2016; Pavlova et al. 2022). In this respect, a recent tumor nutrient-partitioning study documents highest use of glutamine by tumor cells versus highest use of glucose by tumor myeloid cells in a mouse syngeneic MC38 colon tumor cell model (Reinfeld et al. 2021).

Warburg's controversial views on carbohydrate metabolism as the primary cause of cancer dominated the dialog on the biochemistry of cancer and ushered in an era of research on cancer metabolic pathways until the late 1970s when proto-oncogenes were discovered as precursors of viral oncogenes that drive neoplastic transformation. At the turn of the decade, in the 1980s, many oncogenes were discovered and documented to be altered in human cancers, opening a new chapter in cancer research focusing on the genetics of cancer (Varmus 1984). At this point, the interest in metabolism began to wane partly due to controversies over Warburg's dogmatic views and whether cancer metabolism is any different than normal metabolism. The field of cancer metabolism was further displaced by the view that oncogenes and tumor suppressors are the primary drivers of cancer with metabolism playing a subservient role to genetics.

ONCOGENES, TUMOR SUPPRESSORS, AND ALTERED TUMOR METABOLISM

The Src oncogene, fraudulently linked to the Warburg effect by Spector, appeared again with Ras in 1987, when Flier and coworkers (Flier et al. 1987) reported that rodent fibroblasts transfected with these oncogenes increased the mRNA expression of a glucose transporter and had increased uptake of 2-deoxyglucose. This connection between oncogenes and glucose uptake was further supported by the finding in 1989 that Ras and c-Mos-transformed NIH3T3 fibroblasts expressed more GADPH than control cells (Persons et al. 1989). Intriguingly, Myc expression did not result in increased glucose transporter expression or glucose uptake in the Flier study (Flier et al. 1987), but the levels of GAPDH in NIH3T3 appeared to correlate with Myc expression in the Persons study (Persons et al. 1989). However, the detailed mechanistic links between these oncogenes and elevation of the glucose transporter mRNA were missing. Within a decade of these findings, Myc-dependent genes in Myc-transformed Rat1a fibroblasts were identified based on the notion that the product of the MYC oncogene behaves as a transcription factor (Kato et al. 1990; Lewis et al. 1997).

To identify Myc-responsive genes, control or anchorage-independent Myc-transformed Rat1a fibroblasts were grown in suspension cultures. Through representational difference analysis, a form of PCR-assisted subtraction cloning, over 20 putative Myc-responsive genes were identified (Lewis et al. 1997). Among these, lactate dehydrogenase A (LDHA) was transcriptionally induced in Rat1a-Myc cells as evidenced by nuclear run-on assays and Myc-binding sites that are required for Myc transactivation of an LDHA promoter-luciferase reporter (Shim et al. 1997). Importantly, Myc transformation was dependent on LDHA. The finding of LDHA among putative Myc target genes functionally linked Myc to the Warburg effect, providing a firm mechanistic link between an oncogene and aerobic glycolysis. Semenza and coworkers (Wang et al. 1995) cloned the hypoxia-inducible factor (HIF) gene, which was shown to induce the expression of many glycolytic genes under hypoxic conditions (Firth et al. 1994; Semenza et al. 1994). The induction of these genes by HIF to mediate anaerobic glycolysis contrasts with the ability of Myc to induce glycolysis under aerobic conditions (Dang and Semenza 1999).

In addition to the hypoxic stabilization of HIF-1α and HIF-2α proteins, HIF-1 is also thought to be stabilized by upstream oncogenic signaling. In this regard, HIF-driven metabolic rewiring downstream of oncogenic drivers contributes to neoplastic glycolytic metabolism and angiogenesis. Activation of mTORC1 by amino acids and growth signaling through RHEB induces glucose metabolism through increasing Myc and HIF-1α activity and expression (Düvel et al. 2010). It is intriguing to note that MYC is central to PI3K inhibitor resistance (Muellner et al. 2011) and oncogenic alterations of metabolism downstream of PI3K-Akt (Hoxhaj and Manning 2020). Moreover, RAS induces pancreatic cancer glycolytic metabolism (Reinfeld et al. 2021) in a MYC-dependent manner (Ying et al. 2012). In this context, it should be noted that the RAS-ERK pathway has been shown to increase Myc expression and protein levels (Farrell and Sears 2014). As such, the potential collaboration between MYC and HIF signaling downstream of oncogenic pathways could be central to the Warburg effect seen in different cancers.

Subsequent to the observation on MYC-associated, glucose-deprivation-induced cell death (Shim et al. 1998), MYC overexpressing human cells were found to be addicted to glutamine (Yuneva et al. 2007), suggesting a role for MYC in regulating glutamine metabolism. In this respect, the Thompson (Wise et al. 2008) and Dang (Gao et al. 2009) laboratories independently reported the regulation of glutaminolysis by MYC, which activates glutaminase for the conversion of glutamine to glutamate and subsequent catabolism through the TCA cycle. Further, MYC is broadly involved in regulating many metabolic pathways including nucleotide and lipid metabolism (Dang 2012).

Based on their studies of the metabolism of activated T cells, Thompson and coworkers in 2002 reported that costimulation via CD28 triggered a PI3K-Akt-dependent activation of glycolysis (Frauwirth et al. 2002). While Akt was a known oncogene, first identified as the cellular homolog of v-Akt found in the rodent AKT8 retrovirus, activating mutations of PIK3Ca (PI3K) in human cancers were not reported until 2004 (Samuels et al. 2004). In this respect, activated Akt was documented to drive aerobic glycolysis (Elstrom et al. 2004) and subsequent studies underscore the ability of Akt to directly phosphorylate and activate HK2 and PFKBP2 (Hoxhaj and Manning 2020).

Loss-of-function of tumor suppressors also contributes to altered oncogenic metabolism (Levine and Puzio-Kuter 2010; Humpton and Vousden 2016). For example, increased expression of the tumor suppressor PTEN, which opposes PI3K, resulted in heightened oxidative metabolism in vivo (Garcia-Cao et al. 2012), which is the phenotypic converse of the activation of glycolysis by PI3K (Hu et al. 2016). The tumor suppressor p53 tends to diminish glycolysis in favor of a more heightened oxidative metabolism (Humpton and Vousden 2016). This is in part driven by p53 activation of TIGAR as reported (Bensaad et al. 2006). Further, p53 induces synthesis of cytochrome c oxidase (SCO2) to drive mitochondrial respiration, such that loss of wild-type p53 decreased SCO2 expression, resulting in increased glycolysis (Matoba et al. 2006). Conversely, mitochondrial function affects p53 response. Inhibition of mitochondrial complex III or dihydroorotate dehydrogenase (DHODH) activity depletes pyrimidines and activates p53 (Ladds et al. 2018; Mick et al. 2020). In this respect, p53 is both downstream and upstream of metabolic perturbations. The tumor-suppressive effects of tuberous sclerosis complex TSC1 and TSC2 and alteration of metabolism are largely through their ability to inhibit mTOR activity (Manning and Cantley 2003). The tumor suppressor retinoblastoma (RB) has been implicated in glutamine metabolism, such that loss of Rb enhanced E2F-mediated expression of ASCT2- and E2F-independent increase in GLS (Reynolds et al. 2014).

Intriguingly, at the same time that canonical oncogenes were shown to impact metabolism, several core metabolic enzymes were shown to behave as tumor suppressors. Inherited mutations of several nuclear-encoded mitochondrial components, including succinate dehydrogenase subunits SDHB, SDHC, and SDHD and fumarate hydratase (FH), predispose to family syndromes of cancers such as pheochromocytoma, paraganglioma, leiomyosarcoma, and chromophobe renal cell carcinoma (Gottlieb and Tomlinson 2005). These findings suggest that these enzymes are tumor suppressive and the mechanism underlying their tumor-suppression function in part involves HIF stabilization (Selak et al. 2005) and epigenetic modification. For example, SDH mutation causes an accumulation of succinate, which inhibits α-ketoglutarate-dependent prolyl hydroxylases and stabilizes HIF-1α (Selak et al. 2005), whereas FH mutations cause an accumulate of fumarate, which inhibits α-ketoglutarate-dependent demethylases and leads to epigenomic alterations that drive epithelial–mesenchymal transition (EMT) (Sciacovelli et al. 2016). These direct links between metabolic enzyme mutations and familial cancer underscore the importance of metabolic perturbation as a cancer driver.

EMERGING CANCER METABOLISM CONSENSUS

Considering general principles, it is apt to distinguish between maintenance and proliferative metabolism (Vander Heiden et al. 2009). Maintenance metabolism is required to sustain and renew cellular structures and functions by providing ATP to support membrane potentials and protein synthesis. These processes are diurnally dynamic, driven by the circadian clock core transcription factor Clock:Bmal1, whose oncogenic perturbation is documented (Sancar and Van Gelder 2021). As such, normal metabolic studies in vivo can be affected by this diurnal fluctuation that enables daily oscillation of cellular metabolism to synchronize with organismal feeding and fasting cycles.

Proliferative metabolism, on the other hand, can result from normal growth signaling such as activation of T cells, proliferation of bone marrow cells required to replace cellular blood components, or proliferation of the gut epithelium. Upon growth stimulation, signaling through Ras-MEK-ERK signaling cascade activates and stabilizes MYC to induce metabolic and growth-related mRNAs such as those for glucose or amino acid transporter to import nutrients for cell growth (Dang 2012). The influx of amino acids and growth signal transduction through PI3K-Akt-TSC2-RHEB activates mTOR to induce translation and protein synthesis (Cantor and Sabatini 2012). Together, MYC and mTOR can be envisioned to amplify transcription and translation (Hoxhaj and Manning 2020), respectively, of growth signaling and drive proliferative metabolism (Fig. 4). The hypoxia-independent stabilization of HIF-1 is not necessary but can contribute to proliferative metabolism and induction of tumor neovascularization (Fig. 4). Tumor suppressors such as PTEN and TSC2 attenuate the growth signaling pathways driven by PI3K and mTORC1, respectively (Cantor and Sabatini 2012). Hence, loss of these tumor suppressors increased signaling through these oncogenic pathways and their effects on metabolism. P53 can attenuate Myc function by sensing an overactive Myc-Arf axis (Zindy et al. 1998) or suppress proliferation by sensing DNA replication or ribosomal stress (Lindström et al. 2022). p53 can sense ribosomal stress when MDM2 is bound to specific ribosomal subunits and release p53 from its grip. Increased p53 function, in turn, inhibits glycolysis and increases respiration (Fig. 4).

Figure 4.

Oncogenic alterations of metabolism. The diagram shows a cell with an activated growth factor receptor (left) triggering signal transduction down the Ras-MEK-Erk pathway to activate Myc. Growth signal is also transmitted down the PI3K-Akt pathway to activate mTORC1, which senses amino acids for full activation. Myc, in turn, activates genes involved in anabolic metabolism, driving glycolysis, glutaminolysis, nucleotide, lipid, and protein synthesis. mTORC1 amplifies growth signaling by stimulating translation and protein synthesis for mass accumulation that includes its direct activation of nucleotide and lipid synthesis. The hypoxia-inducible factor (HIF) can be induced by mTORC1 and stabilized under hypoxia to induce anaerobic glycolysis. On the other hand, the tumor suppressor p53 suppresses glycolysis and induces mitochondrial respiration. Note that the prevalent human oncogenes Ras and PI3K are upstream of Myc and mTORC1, enabling transcriptional and translational amplification of oncogenic cell growth and proliferation. Fatty acids as an energy source through oxidation is depicted. (1C) One-carbon, (PPP) pentose phosphate pathway. (Figure created with BioRender.com.)

A key question is whether there are key differences between normal proliferative versus oncogenic metabolism (Vander Heiden et al. 2009). As discussed previously, the Warburg effect can be observed in cancers and normal tissues. For example, whereas resting T cells use less glycolysis and rely on oxidative metabolism, T-cell receptor (TCR) stimulation of murine T cells induces a proliferative metabolic program resembling that of malignant lymphocytes (Madden and Rathmell 2021). Specifically, stimulation of T cells with anti-CD3 and anti-CD28 drives glycolysis and glutaminolysis in a Myc-dependent fashion that enables proliferation, which does not depend on HIF-1α (Fig. 5A; Wang et al. 2011). Upon withdrawal of stimulation, T cells undergo apoptosis and some attain a resting memory oxidative metabolic state. In contrast to normal T cells, oncogenic NOTCH-driven T-cell lymphomas are dependent on constitutive MYC expression, which drives a constitutive proliferation metabolic profile that cannot return to a resting state (Zhou et al. 2022). In this regard, a difference in normal versus neoplastic proliferative metabolism is that the former can be turned off. In contrast, the latter is constitutively turned on, rendering the malignant state addicted to a constant supply of nutrients. Normal cells have mechanisms that sense nutrient deprivation such as AMPK, which can induce cell growth arrest. However, MYC-addicted cells are vulnerable to glucose or glutamine deprivation–induced cell death as are AKT-addicted cells (Shim et al. 1998; Elstrom et al. 2004; Yuneva et al. 2007). Given these observations, are there sufficient therapeutic indices to exploit metabolism for cancer therapy?

Figure 5.

T-cell activation and the effect of diets on tumor and immune cell metabolism and fate. (A) Anti-CD3 plus anti-CD28 activation of resting primary murine T cells, which respires via β-oxidation, requires Myc, which activates glycolysis and glutaminolysis to drive biomass accumulation for cell proliferation. (B) Dietary high fiber or choline is shown to produce microbial short chain fatty acids (SCFAs) or trimethylamine (TMA) and liver-derived trimethylamine oxide (TMAO). SCFAs activate effector T cells, whereas TMAO polarizes macrophages toward inflammatory M1 states that increase coronary artery inflammation or enhance immune checkpoint blockade cancer therapy. Ketogenic diet affects the gut microbiota and tumor immunity; results from ongoing clinical studies are pending. Whether a serine/glycine deprivation diet proves to increase cancer therapy response in humans remains to be established. (C) The illustration depicts a scale balancing cells that have antitumor activity, such as cytotoxic CD8⁺ T cells (CD8), natural killer (NK) cells, or inflammatory M1 macrophages, versus cells, such as myeloid-derived suppressor cells (MDSCs), regulatory T (Treg) cells, or alternatively activated M2 macrophages that assist cancer cell growth. (CTL) Cytotoxic T-lymphocyte. (Figure created with BioRender.com.)

METABOLIC THERAPY AND LESSONS LEARNED

When considering metabolic vulnerabilities of cancers, recent studies have highlighted the importance of tissue-specific metabolic effects of oncogenic drivers, metabolic plasticity, diet, as well as the impact of these features on the microbiome and antitumor immunity. Different oncogenes induce different metabolic profiles in the same organ. For example, in contrast to MYC, which drives glutamine and glucose metabolism in MYC-inducible liver cancer, MET oncogene-driven liver cancer expresses glutamine synthetase and hence appears less dependent on exogenous glutamine (Yuneva et al. 2012). On the other hand, the same oncogene can induce different metabolic effects in different tissues. Kras effects on branched chain amino metabolism are different in Kras, p53-loss-driven murine pancreatic adenocarcinoma versus non-small-cell lung cancer (NSCLC) (Mayers et al. 2016). In the former, branched chain amino acid (BCAA) uptake is diminished, whereas in NSCLC, the tumors incorporate BCAA into proteins. Hence, tissue-specific effects of oncogenes add to the complexity of tumor metabolism in vivo when considering the metabolic vulnerabilities of cancers.

Metabolic plasticity (Fendt et al. 2020) and metabolic stress such as activation of AMPK or the integrated stress-response pathways induce resistance to inhibition of cancer metabolism (Costa-Mattioli and Walter 2020). Metabolic plasticity was elegantly illustrated by Yuneva and coworkers using a MYC-inducible model of mouse HCC (Méndez-Lucas et al. 2020). They demonstrated that lost glutaminase (Gls) extended survival as seen with pharmacological Gls inhibition (Xiang et al. 2015). However, loss of hexokinase 2 (Hk2) did not extend survival. Intriguingly, loss of both Gls and Hk2 further extended survival, but these double-knockout (KO) tumors eventually caused the demise of their hosts, indicating yet other ways that allow for neoplastic cells to circumvent metabolic blocks. This genetic evidence for metabolic plasticity is underscored by the cooperation between metabolic inhibitors, such as a combination of inhibitors of LDH and mitochondrial complex I, to slow tumor growth (Oshima et al. 2020).

Whereas loss of Hk2 did not extend survival of MYC-induced HCC, Hk2 is documented to be required for initiation and maintenance of murine KRas-driven lung cancer and ErbB2-driven breast cancer (Patra et al. 2013). Further, systemic deletion of Hk2 also reduced tumorigenesis in a diethylnitrosamine-induced murine model of HCC (DeWaal et al. 2018), and, importantly, loss of Hk2 did not affect T-cell proliferation or T-cell-mediated viral immunity (Mehta et al. 2018). Notably, some human multiple myelomas do not express hexokinase 1 and are highly sensitive to decreased HK2 (Xu et al. 2019). These observations suggest that HK2 is an example of an enzyme that appears to be cancer specific.

Under nutrient-depleted conditions, decreased mTOR activity and activation of AMPK induce ULK activity to drive autophagy, whereby autophagosomes are formed and destined for lysosomal degradation to recycle metabolites for survival (Onodera and Ohsumi 2005; Rabinowitz and White 2010). Further, mitophagy—a form of autophagy—is necessary to cull dysfunctional mitochondria. The maintenance of an NAD⁺/NADH ratio >>1 to drive oxidative anabolism is essential for cell function. As such, under nutrient deprivation, autophagy maintains NAD⁺ levels (Kataura et al. 2022). When the conversion of NADH to NAD⁺ is saturated via mitochondrial NADH malate-aspartate and glycerol-3-phosphate dehydrogenase shuttles, aerobic glycolysis is induced to regenerate cytosolic NAD⁺ to drive GADPH-mediated catalysis (Wang et al. 2022). Hence, pathways that can regenerate cytosolic NAD⁺ when NADH is in excess could increase in activity when other pathways are limited. Excessive NADH levels induce reductive stress (Mick et al. 2020) and activate as the transcriptional corepressor CtBP to generate an adaptive transcriptome (Di et al. 2013).

In the study of MYC-driven murine liver cancer (Méndez-Lucas et al. 2020), the loss of Psat1, which drives one-carbon metabolism through serine and glycine, did not affect survival. However, withdrawal of serine and glycine from the diet as done previously by Vousden et al. (Maddocks et al. 2017) prolonged the survival of Psat KO but not wild-type tumor-bearing animals (Méndez-Lucas et al. 2020). These findings underscore that the effect of diet depends on the metabolic wiring of the tumor cells (Kalaany and Sabatini 2009; Lien and Vander Heiden 2019). What was not accounted for in these studies is the effect of diet on the host microbiome or immunity. Since the availability of dietary L-serine can affect the gut microbiota during inflammation (Kitamoto et al. 2020), whether a serine/glycine deprivation diet influences tumor immunity beyond a cancer cell–autonomous effect remains to be established. In this regard, the ketogenic diet can alter the host microbiome (Fig. 5B; Ang et al. 2020) and curb several models of mouse tumorigenesis. Ketogenic diet curbs tumor growth in a model of mouse pancreatic adenocarcinoma in combination with chemotherapy (Yang et al. 2022). Further, this combination also worked in immunocompromised mice, but sustained response was observed only in mice with an intact immune system. Intriguingly, a ketogenic diet alters gut and serum metabolome in dogs with implications on tumor immunity (Allenspach et al. 2022).

Dietary choline can induce inflammation through its conversion to trimethylamine (TMA) by the gut microbiota, and in turn oxidized by the liver to trimethylamine oxide (TMAO), which is well-implicated in provoking coronary artery disease (Wang et al. 2015). Dietary choline or administration of TMAO induces inflammatory M1 macrophages (Fig. 5B) that increase graft-versus-host response as well as response of tumors to immune checkpoint blockade (Wu et al. 2020; Mirji et al. 2022). As such, there is much more to learn about the effects of components of diet, such as high fiber content that generates microbial short chain fatty acids with immune modulatory activities (Fig. 5B), on the microbiota that in turn increase cancer therapy responses (He et al. 2021; Spencer et al. 2021).

The idea of targeting metabolism for cancer treatment was championed by Sidney Farber who targeted nucleotide synthesis with the “anti-metabolite” aminopterin and subsequently methotrexate, which is still used clinically (Farber and Diamond 1948; Farber 1949). Together with asparaginase, an active therapy in lymphoblastic leukemias, this demonstrates that metabolic therapies can be active anticancer agents. However, other metabolic therapies have proven less efficacious. The fact that 2-deoxyglucose can inhibit glycolysis made it a candidate for studies in cancer patients, but studies from decades ago showed that it did not produce clear benefit, and patients had side effects such as diaphoresis (Landau et al. 1958). Likewise, the glutamine analog 6-diazo-5-oxo-1-norleucine (DON), which targets a multitude of glutamine using enzymes and hence is imprecise, was also tested in humans, but it appeared too toxic for clinical use (Magill et al. 1957). It is notable that a DON prodrug has significant preclinical efficacy against several tumor models in immunocompetent mice (Leone et al. 2019). Whether clinical trials on DON prodrug prove to be effective remains to be seen. Recent failures of metabolic inhibitors in the clinic result from either lack of activity or intolerable side effects. For example, CB-839, which is a highly specific glutaminase (GLS) inhibitor with little associated side effects, failed in a study of patients with renal cell carcinoma due to a lack of efficacy (Tannir et al. 2022). The use of the mitochondrial complex I inhibitor IACS-07549 failed in clinical studies because of neurotoxicity (Yap et al. 2023). However, an exception is the successful implementation of specific inhibitors for mutant isocitrate dehydrogenases, IDH1 and IDH2, for the treatment of cancers such as acute myelogenous leukemia (DiNardo et al. 2018). Here, the therapeutic index is widened by the specificity of the drugs for mutant versus wild-type enzymes.

CONCLUDING REMARKS

Over the past century, Warburg's studies on cancer metabolism and those reporting the use of Coley's toxin for cancer therapy lay the foundation for current studies that offer a hopeful outlook for new cancer therapeutic opportunities. Given the profound success of cancer immunotherapy, it should be noted that the use of metabolic inhibitors can interfere with or potentiate the antitumor arm of the immune system (Leone et al. 2019; Hermans et al. 2020). As such, the development of metabolic inhibitors to target cancer cells should also account for its potential adverse effect on antitumor immunity (Fig. 5C). Notably, a major challenge to effective immunotherapy is tumor acidity (Boedtkjer and Pedersen 2020; Tu et al. 2021; Gillies et al. 2022), whose mitigation in the clinical setting has not been sufficiently addressed by current research. The rapid improvement of mass spectrometry imaging (Ma and Fernández 2022) of tissues down to the single-cell level should provide the spatial resolution necessary to gain a richer understanding of the tumor immune microenvironment metabolic states (Fig. 1C) and potentially expose novel cancer metabolic vulnerabilities. The emerging field of immunometabolism (Buck et al. 2017; Leone and Powell 2020; Madden and Rathmell 2021; Stine et al. 2022) offers a richer understanding of metabolic vulnerabilities of immune versus cancer cells that is anticipated to provide novel druggable opportunities to enhance immunity while diminishing cancer cell viability.