We need to talk about the Warburg effect

We are approaching the 100th anniversary of Otto Warburg’s first description of the metabolic phenotype bearing his name—a propensity for tumours to metabolize glucose anaerobically rather than aerobically, even when oxygen is available. Generations of scientists have studied the Warburg effect, yet misconceptions persist about its causes and relationship to oxidative metabolism in the mitochondria. Here, we review the definition of the Warburg effect and discuss its place within a modern understanding of cancer biology.

In the 1920s, Otto Warburg showed that cultured tumour tissues have high rates of glucose uptake and lactate secretion, even in the presence of oxygen (aerobic glycolysis). Those three metabolic properties—glucose uptake, lactate secretion and oxygen availability—constitute the Warburg effect as he defined it. The importance of oxygen in this definition is sometimes overlooked. Long before Warburg, Pasteur showed that oxygen suppresses the fermentation of sugars, thus identifying conversion of glucose to lactate as an expected response to hypoxia. If this logic is applied to cancer, tumours may be hypoxic, and hypoxia may induce lactate formation in tumours as it does elsewhere. But that’s not the Warburg effect. What set tumours apart in Warburg’s analysis was the lack of proportion between glycolysis and respiration—that is, the tumour samples took up glucose and converted it to lactate even when there was sufficient oxygen to convert glucose to CO2, which other tissues prefer to do and which we now know is much more productive in terms of ATP synthesis (an excellent review can be found in ref. 1).

Let us consider two aspects of the Warburg effect: the high glycolytic rate and its relationship to fuel oxidation in the mitochondria. A special relationship exists between glucose metabolism and rapid cell proliferation, at least in cultures, as observed in cancer cells as well as non-malignant cells2. When stimulated by growth factors to proliferate, most cells take up glucose and secrete a fraction of the carbon back into the culture medium as lactate, just as Warburg described in his tumours. Withdrawing glucose or inhibiting glycolysis is often deleterious to cancer cell proliferation and tumorigenesis in experimental models3,4. These observations led to the perception that high rates of lactate secretion are required to support malignant cell proliferation.

How is glycolysis related to cell proliferation? Rapid glucose uptake and metabolism allow cells to feed several non-mitochondrial pathways that contribute to macromolecular synthesis (Fig. 1). These include the pentose-phosphate pathway, which produces ribose for nucleotides and NADPH for reductive biosynthesis; the hexosamine pathway, which is required for protein glycosylation; serine–glycine–one-carbon metabolism, which feeds glutathione, nucleotides and methylation reactions; and glycerol synthesis for the production of complex lipids. In cancer cells, these pathways are often activated in response to oncogenic signalling.

Fig. 1: Glycolysis and TCA-cycle metabolism supply anabolic pathways.

Activities related to glycolysis are in black, and activities related to pyruvate oxidation and the TCA cycle are in red. Both processes support cell growth by feeding branch pathways required for anabolism. In the Warburg effect, glycolysis terminates with lactate production and secretion despite the presence of oxygen. These latter steps (blue) provide a means of recycling NADH to NAD+ but result in loss of carbon from the cell upon lactate release. GLUT, glucose transporter; MCT, monocarboxylate transporter; MPC, mitochondrial pyruvate carrier; glucose-6P, glucose-6-phosphate; fructose-6P, fructose-6-phosphate; fructose-1,6-biP, fructose-1,6-bisphosphate; DHAP, dihydroxyacetone-phosphate; GA3P, glyceraldehyde-3-phosphate; 3PG, 3-phosphoglycerate; acetyl-CoA, acetyl coenzyme-A; α-KG, α-ketoglutarate; Succ-CoA, succinyl-CoA; OAA, oxaloacetate; NAD+, oxidized nicotinamide adenine dinucleotide; NADH, H+, reduced nicotinamide adenine dinucleotide.

But converting glucose to lactate and then secreting the lactate does not provide carbon for biosynthetic pathways—it eliminates carbon from the cell. Therefore, if the Warburg effect supports anabolic metabolism, it does so indirectly. The truth is that the reason why many proliferating cells display the Warburg effect is still not fully understood. One possibility is that the Warburg effect may allow cells to maintain large pools of glycolytic intermediates, and having full pools of these metabolites favours engagement of the pentose-phosphate pathway and other biosynthetic pathways that branch from glycolysis5. Presumably, this advantage would apply even if much of the glycolytic flux were to terminate in lactate secretion. That said, the Warburg effect is not a universal feature of proliferating cancer cells, even in culture. An analysis of more than 80 lung cancer cell lines found a nearly tenfold range in the rates of glucose uptake and lactate secretion. In some cell lines, glucose uptake exceeded lactate secretion, thus resulting in a conspicuously un-Warburgian form of metabolism6.

Does having a high glycolytic rate predict anything helpful about the state of the mitochondria? This question is a point of confusion, in no small part because of Warburg’s writing on the matter. Warburg thought that aerobic glycolysis was a symptom of impaired oxidative metabolism. His seminal hypothesis paper, ‘On the origin of cancer cells’, took this idea a step further by arguing that “the irreversible injuring of respiration” not only characterizes cancer but causes it7. But Warburg’s own experiments revealed persistent oxygen consumption in tumour tissue. The rate of respiration was low relative to what might have been predicted by the high rate of glucose uptake, but respiration itself did not appear to be impaired1. Recent work strongly supports an important role of respiration—and mitochondrial metabolism more broadly—in cancer cell and tumour growth. Respiration enables cell proliferation by allowing cells to produce intermediates from the tricarboxylic acid (TCA) cycle that are required for anabolism (Fig. 1). Respiration is linked to the TCA cycle, because the reduced electron carriers generated as the cycle turns are re-oxidized by the electron-transport chain. Ablating components of the electron-transport chain in Kras-mutant mouse lung adenocarcinoma cells suppresses tumour growth, and the acquisition of functional mitochondrial DNA from adjacent stromal cells in vivo restores tumour growth8,9.

Human tumours also respire. Sequences from 1,675 tumour biopsies from more than 30 different types of cancer have revealed a selection against mitochondrial-DNA mutations that have deleterious effects on respiration10. Other evidence of mitochondrial activity in human tumours comes from intra-operative 13C tracing experiments, which have revealed prominent fuel oxidation in tumours from the human brain and lung11,12. In lung tumours, the extent of glucose oxidation correlates with the content of proliferating cancer cells13. Mitochondrial metabolism in cancer has practical importance, because several inhibitors of oxidative phosphorylation and related pathways have now advanced into clinical trials.

Given the importance of mitochondria in cancer, the oft-cited ‘glycolytic switch’, taken to be synonymous with the Warburg effect, is misleading and probably not very useful as a general concept. ‘Glycolytic’ tumours—that is, tumours that take up glucose and process it to pyruvate through the glycolytic pathway—also oxidize pyruvate in the mitochondria. This phenomenon is very clear in human lung cancer, in which intraoperative infusions with [13C]glucose result in abundant labelling in both glycolytic and TCA-cycle metabolites from the tumour. The widespread clinical use of [18F]fluorodeoxyglucose (FDG) positron emission tomography contributes to the perception of a glycolytic switch, but this tracer reports glucose uptake without providing any information about glucose’s metabolic fates. Sequential FDG positron emission tomography and [13C]glucose infusions in the same patients with lung cancer have revealed that the FDG signal is not particularly predictive of glycolytic-intermediate labelling and certainly does not indicate suppressed glucose oxidation13,14. In summary, there is little evidence supporting a glycolytic switch in these tumours. Even when glucose uptake is activated, oxidative metabolism persists in the tumour and may exceed that in adjacent non-malignant tissue. New metabolic-imaging agents have added further evidence of concomitant FDG uptake and oxidative metabolism in mouse models of cancer, and some of these agents may soon be used to assess human cancer15.

The Warburg effect has also been studied in settings of physiological cell proliferation, including in T cells and haematopoietic stem and progenitor cells. Here, growth-factor signalling causes a disproportionate activation of glycolysis relative to pathways involved in oxidative metabolism. Therefore, the glycolytic rate may far exceed the maximal rate of mitochondrial pyruvate oxidation, thus making lactate secretion unavoidable when glucose is abundant, as it usually is in culture. But genetic inhibition of mitochondrial complex III in T cells and haematopoietic stem cells renders them functionally incompetent in vivo, thus demonstrating the importance of mitochondrial function in these cells16,17. Moreover, recent isotope tracing studies have reported that activating T cells in vivo rather than in culture stimulates glucose oxidation in the TCA cycle; these findings emphasize the value in studying metabolic responses in a physiological environment18.

In several forms of cancer, particularly in the kidney, mutations in the TCA-cycle enzymes succinate dehydrogenase and fumarate hydratase impair oxidative metabolism19. These mutations cause fumarate and/or succinate to accumulate and inhibit α-ketoglutarate-dependent dioxygenases involved in demethylation of histones, RNA and DNA; the resulting epigenetic changes are thought to contribute to transformation. In clear cell renal cell carcinoma, loss of the von Hippel–Lindau tumour suppressor results in pseudohypoxic stabilization of hypoxia-inducible factors and the activation of a transcriptional program that suppresses pyruvate oxidation19. These tumours also accumulate l-2-hydroxygluatrate, which, like succinate and fumarate, inhibits α-ketoglutarate-dependent dioxygenases and enhances DNA methylation20. Cell lines derived from these tumours are highly glycolytic under aerobic conditions. But even these cells use residual or reprogrammed aspects of mitochondrial metabolism to maintain pools of TCA-cycle intermediates for anabolism and generate oncometabolites such as l-2-hydroxygluatrate.

Where does this leave us with the Warburg effect? In short, genetically defined impairments in oxidative metabolism may stimulate aerobic glycolysis in cancer, but, in general, aerobic glycolysis does not predict loss of oxidative metabolism. Even tumours that are hardwired to suppress pyruvate oxidation and produce lactate can rewire their mitochondrial metabolism to synthesize oncometabolites and essential TCA-cycle intermediates for anabolism. This is a major conclusion from the past decade of research on the Warburg effect, driven by advances in studying cancer metabolism in vivo. Perhaps the route to this understanding is fitting, given that, among Warburg’s many contributions, his discovery of the ‘oxygen-transferring ferment of respiration’ (that is, the function of cytochrome c oxidase) was recognized with the Nobel Prize in Physiology or Medicine in 1931. Ironically, Warburg’s name has become synonymous with glycolysis, even though he is one of the most important figures in the history of mitochondrial research.


  1. 1.

    Koppenol, W. H., Bounds, P. L. & Dang, C. V. Nat. Rev. Cancer 11, 325–337 (2011).

  2. 2.

    Lunt, S. Y. & Vander Heiden, M. G. Annu. Rev. Cell Dev. Biol. 27, 441–464 (2011).

  3. 3.

    Fantin, V. R., St-Pierre, J. & Leder, P. Cancer Cell 9, 425–434 (2006).

  4. 4.

    Patra, K. C. et al. Cancer Cell 24, 213–228 (2013).

  5. 5.

    Newsholme, E. A., Crabtree, B. & Ardawi, M. S. Biosci. Rep. 5, 393–400 (1985).

  6. 6.

    Chen, P. H. et al. Mol Cell 76, 838–851.e835 (2019).

  7. 7.

    Warburg, O. Science 123, 309–314 (1956).

  8. 8.

    Weinberg, F. et al. Proc Natl Acad. Sci. USA 107, 8788–8793 (2010).

  9. 9.

    Tan, A. S. et al. Cell Metab. 21, 81–94 (2015).

  10. 10.

    Ju, Y. S. et al. eLife 3, e02935 (2014).

  11. 11.

    Fan, T. W. et al. Mol. Cancer 8, 41 (2009).

  12. 12.

    Maher, E. A. et al. NMR Biomed. 25, 1234–1244 (2012).

  13. 13.

    Hensley, C. T. et al. Cell 164, 681–694 (2016).

  14. 14.

    Kernstine, K.H. et al. Ann. Thorac. Surg. https://doi.org/10.1016/j.athoracsur.2019.10.061 (2019).

  15. 15.

    Momcilovic, M. et al. Nature 575, 380–384 (2019).

  16. 16.

    Sena, L. A. et al. Immunity 38, 225–236 (2013).

  17. 17.

    Ansó, E. et al. Nat. Cell Biol. 19, 614–625 (2017).

  18. 18.

    Ma, E. H. et al. Immunity 51, 856–870.e855 (2019).

  19. 19.

    Linehan, W. M. et al. Cancer Discov. 9, 1006–1021 (2019).

  20. 20.

    Shim, E. H. et al. Cancer Discov. 4, 1290–1298 (2014).

Download references

Author information



Corresponding authors

Correspondence to Ralph J. DeBerardinis or Navdeep S. Chandel.

Ethics declarations

Competing interests

R.J.D. is an advisor for Agios Pharmaceuticals. N.S.C. is an advisor for Rafael Pharmaceuticals.

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

DeBerardinis, R.J., Chandel, N.S. We need to talk about the Warburg effect. Nat Metab 2, 127–129 (2020). https://doi.org/10.1038/s42255-020-0172-2

Download citation

Further reading