Two Thirds Fermentation, and the Enzyme That Runs Backwards

Part five takes the resemblance everyone reaches for first: that cancer cells and parasitic worms burn sugar the same wrong way. We open on Warburg's own 1927 table, where the tumor pulled an average of 70 mg of glucose out of every 100 cc of blood and sent back 46 mg of lactic acid — two thirds fermented and the rest respired, in his own arithmetic — and we correct a number this show's own Atlas page got backwards. Then the dates, properly: 1923 for the slice work, 1925 for the Coris in living animals, and a three-way exchange in Science in 1956 in which nobody conceded on the record. Then what carbon-13 infused into patients in the operating room actually shows, which cuts both ways, and the four numbers in circulation that no measurement supports. Then the instrument, because the rest of the season depends on it: what a PET scan is, how FDG gets trapped, and why the result is a map of where glucose is being trapped rather than a map of dividing cells — in a mouse tumor its brightest compartment was granulation tissue and macrophages. Then the worm as real chemistry rather than metaphor: a schistosome with no Pasteur effect at 5 percent oxygen, an adult Ascaris suum running mitochondrial complex II backwards as a quinol-fumarate reductase with the atomic coordinates published, malate dismutation and its branched end products, rhodoquinone built from tryptophan by the kynurenine route and not by any shikimate pathway — and our own mitochondria running the same reaction in liver, kidney and brain. Then the drugs, the diets and the alkaline trade, answered with trial numbers, and Warburg's own 1927 experiment against the diet his measurement is used to sell. Two fabrications are retired on air by name: an author list invented for a real paper, and a rhodoquinone pathway that does not exist.

Show notes

Start with the table, not the slogan. In 1927 Warburg, Wind and Negelein published an experiment in the Journal of General Physiology, free to read in full. They anesthetized tumor-bearing rats, sampled the vein draining the tumor against the abdominal aorta, and measured what went in against what came out.

Out of every 100 cubic centimeters of blood, a tumor removed an average of 70 milligrams of glucose, 48 to 95 across six Jensen sarcomas. The jugular, renal, iliac and portal beds removed 2 to 16 milligrams, the range his text gives, though one iliac value in the table reads 18. The tumor sent back an average of 46 milligrams of lactic acid, the mean of eight of ten tumors with two outliers set aside. Normal tissue added none. Fermentation means breaking sugar down part way and not handing the leftover electrons to oxygen; respiration means burning the carbon all the way down, oxygen taking the electrons at the end. Doing the first with oxygen present is aerobic glycolysis, and the observation that tumors do it is the Warburg effect (Warburg, Wind and Negelein 1927).

Then his own arithmetic, which is in the paper. Forty-six over seventy. About 66 percent of the glucose the tumor consumed went to fermentation, and the rest was respired. But it is his number, and it says two thirds. The slogan built on this paper is that cancer cells ferment sugar instead of breathing. His table says two thirds fermented and one third burned with oxygen, in a living animal, in 1927.

Which brings me to a number this show got wrong on its own page. The familiar line is that glucose falls to 57 percent of arterial concentration passing through a tumor. That is inverted. In Table II the 57 percent sits in a column of percentage decrease: mean arterial glucose 124 milligrams per 100 cubic centimeters, mean tumor-vein glucose 54. Venous glucose was 43 percent of arterial; the amount lost was 57 percent. A percentage decrease read as a percentage remaining. The paper's own prose says the glucose falls to 57 per cent, contradicting its table, and the page copied the prose. Corrected: a tumor is worse supplied with glucose than the tissue around it.

And the same paper reports what almost nobody quotes. Pieces of tumor held in oxygenated serum with no glucose still showed normal respiration and fermentation once glucose was given back, and killing most of the cells took removing oxygen and glucose together, over about four hours.

The dates get told wrong too, and that is not pedantry. Fermentation in tumors was not first found in a vein. It was found in 1923, in a dish, by Warburg and Minami, where a rat hepatoma formed lactate at 70 times the rate of normal liver, kidney or heart. Their paper is in Biochemische Zeitschrift and not in PubMed, so that figure reaches us secondhand, through Angela Otto's review, and I am saying so (Otto 2016).

The first demonstration in a living animal was not his either. In 1925 Carl and Gerty Cori compared the veins draining the two wings of hens carrying a Rous sarcoma in one wing: on the tumor side, 23 milligrams less glucose and 16 milligrams more lactic acid, plus one human forearm tumor at 12 and 9. The 1927 paper reports both as its starting point, so the claim that Warburg measured this first is refuted by the paper cited for it. And when the field marked the centenary in Nature Metabolism in 2023, its title dates the founding observation to 1923; the record carries no abstract, so that is as far as I can take it (Thompson 2023).

The name came last: Racker applied the phrase Warburg effect to tumors in the early 1970s, half a century after the measurement (Otto 2016). And the most repeated number in the field belongs to neither experiment. 'Approximately tenfold more glucose to lactate' is Koppenol, Bounds and Dang's 2011 restatement of Warburg's 1920s work in general, most of it slice work. The artery-and-vein data cannot give a tenfold lactate ratio: the denominator would be zero, because normal tissues released no lactic acid at all.

February 24, 1956. Warburg publishes 'On the origin of cancer cells' in Science, volume 123, issue 3191, turning the measurement into a theory of origin: respiration irreversibly injured, fermentation the prime cause. Five months and seventeen days later, Sidney Weinhouse answers him in the same journal, volume 124, issue 3215. Mind that volume: this show's own page says Weinhouse answered in the same volume, and he did not — a journal volume mistaken for a journal. Nor was it a two-man exchange: Warburg replied in that same issue, and Dean Burk with A. L. Schade followed. Three voices in one issue, and nobody conceded anything on the printed record.

Now the part a show with rules says out loud. None of those four pieces carries an abstract in PubMed, two have no DOI, none is in PubMed Central, and all four are indexed as Journal Article, not Review. That settles what I may do with them: not what Weinhouse argued or what Warburg gave ground on, only that the exchange happened, and where. Say it, because of the story it kills — that Warburg's explanation stood unchallenged for most of a century and was overturned recently. It was contested in print within six months, by name, in Science.

The same rule disposes of a document that travels with this story: a 1951 American intelligence translation of a 1950 Russian paper by V. V. Alpatov, said to link anaerobic metabolism and hoarded glycogen in worms and tumors. This show could not open it, so nothing from it goes on air. What is checkable is the man: Alpatov published in Nature on 7 December 1946 on the optical isomers of mepacrine acting on a bacterium (Alpatov 1946). Not on tumors.

So the measurement stands and the explanation does not — said in a review rather than with new data (Koppenol 2011). Here is the data. Nine patients with non-small cell lung cancer were infused with glucose labeled with carbon-13, a heavier traceable form of ordinary carbon, in the operating room, so each tumor could be compared with that patient's own benign lung. Enhanced glycolysis and enhanced glucose oxidation were both common, with evidence of multiple nutrients being oxidized in every tumor (Hensley 2016). That is not a cell that has lost the ability to breathe.

But refuse to overcorrect, because the answer cuts both ways. Clear cell renal cell carcinoma, traced the same way, showed suppressed flow through pyruvate dehydrogenase, the gate admitting sugar carbon into the mitochondrion, and reduced labeling of the TCA cycle, the loop that burns fuel carbon and hands electrons to the chain ending at oxygen. Among human tumors studied this way, the authors write, it is 'the first to demonstrate a convincing shift toward glycolytic metabolism' (Courtney 2018). In 2024 the same program traced more than 80 kidney cancer patients in the operating room: clear cell tumors had suppressed TCA labeling and lower electron transport chain activity than the adjacent kidney — and then the result nobody predicted, their metastases had more TCA labeling than the primaries (Bezwada 2024).

Inside one organ, the answer flips outright. Translocation renal cell carcinoma, driven by fusions of the TFE3 gene, is rewired toward oxidative phosphorylation, the oxygen-using mitochondrial route that makes most of a normal cell's ATP, against what its authors call the highly glycolytic nature of other renal cancers. Cell line, mouse and CRISPR work — gene editing, used here to delete genes one at a time and see which the cells cannot live without — not a clinical study (Li 2025). Nature Metabolism ran a commentary beside it under a headline that is this whole beat in ten words: 'Translocation renal cell carcinoma says no to the Warburg effect' (Pan and Cracan 2025). It depends on the tumor.

The numbers nobody checks, and who is actually burning them

Three numbers carry this story in popular telling. All three are wrong. The first is speed: that fermentation makes ATP, the cell's energy currency, a hundred times faster than the mitochondrion does. It is wrong inside the review it is hung on. Liberti and Locasale do carry a ten- to hundredfold figure, but it is the rate of converting glucose to lactate, not of making ATP, and the same review states that ATP synthesized over any given period is comparable between the routes (Liberti and Locasale 2016). The error swaps the speed of a conversion for the speed of energy production.

The second is thirty-six ATP per glucose, straight from the textbook. It was never measured. What is measurable is the P/O ratio, ATP made per atom of oxygen consumed: about 2.5 with substrates delivering electrons through NADH, the main carrier feeding the respiratory chain, and about 1.5 with succinate — fractional, not whole (Hinkle 2005) — with a ceiling of 33.45 ATP per glucose (Mookerjee 2017). Say about thirty. The memorized number came from a proton stoichiometry the structure of ATP synthase does not support.

The third is the premise under the first two: that tumors are furnaces. TCA flux was suppressed in all five primary solid tumor models, and the raised glycolysis did not make up the difference. Their sentence: instead of being hypermetabolic, as commonly assumed, solid tumors generally produce ATP at a slower than normal rate (Bartman 2023). The exception is in the title, lung metastases of breast cancer; the limit is five mouse models.

Now the percentages. Hao Wu, Minfeng Ying and Xun Hu measured the split in nine cancer cell lines. In ordinary medium, glycolysis supplied 23.7 to 52.2 percent of the ATP and oxidative phosphorylation 47.8 to 76.3 percent. Moved into tumor-like lactic acidosis, 20 millimolar lactate at pH 6.7, the split went to 5.7 to 13.4 percent glycolytic and 86.6 to 94.3 percent oxidative, with all but one of the nine consuming lactate rather than making it (Wu, Ying and Hu 2016). Same cell; what changed was around it.

Separately, eight of those nine converted 79 to 92 percent of the glucose they consumed into lactate — and that is where a widely repeated number went wrong, because those percentages are about the fate of glucose and the 23.7-to-52.2 figure is about the share of ATP. The circulating 79 and 91 percent were blended from two reviews. One is Yaojie Fu's 2017 review, giving 79 percent for HeLa and 91 percent as the oxidative ATP share in — its own word — 'MCFs,' plural, naming no cell line at all (Fu 2017). The other is Michelle Potter's, restating Zu and Guppy, where glycolytic ATP averaged 17 percent across forty years of data (Potter 2016). Those numbers appear only in Potter's restatement, and Zu and Guppy's conclusion is the whole sentence, not the half usually quoted: there is no evidence cancer cells are inherently glycolytic, 'but that some tumors might indeed be glycolytic in vivo as a result of their hypoxic environment' (Zu and Guppy 2004). Then the number nobody has: what fraction of a tumor cell's ATP comes from glycolysis in an actual patient. There is no reliable published figure.

And while Fu's review is open, correct the thing its title names. The reverse Warburg effect — the proposal that a tumor's fibroblasts do the fermenting and feed lactate to the cancer cells — is not a recent discovery. It was proposed and named in 2009 (Pavlides 2009), and lactate-fueled respiration was shown in mice the year before (Sonveaux 2008). Say 'proposed in 2009,' never 'shown': no human study has measured lactate flux from stroma to cancer cell.

So where does the fermentation come from, if not broken mitochondria? It is installed by the genes that drive growth, and the Atlas page carries that section: MYC switching on lactate dehydrogenase A directly (Shim 1997) and running a separate glutamine program without PI3K or AKT, the two enzymes a growth-factor or insulin signal runs through on its way into a cell (Wise 2008); AKT raising glucose consumption without changing oxidative phosphorylation (Elstrom 2004); HIF-1 actively shutting the pyruvate dehydrogenase gate rather than surrendering to low oxygen (Kim 2006); mutant KRAS maintaining a pancreatic tumor, not just starting it (Ying 2012).

That is a clean circuit diagram, and the next paper takes a hammer to it. Shawn Davidson and colleagues traced labeled glucose or glutamine in mice with Kras-driven lung tumors: glutamine use was minimal, and the tumors put more glucose carbon into the TCA cycle than normal lung did, the opposite of the same tumor type in culture (Davidson 2016). Genotype does not set the phenotype, as Pavlova, Zhu and Thompson put it (2022).

Which brings us to lactate, and to one sentence this show has gotten wrong in print. In mice, lactate has the highest turnover flux of any metabolite measured, exceeding glucose by 1.1-fold fed and 2.5-fold fasted, and in fasted animals glucose reaches the TCA cycle mainly via circulating lactate (Hui 2017). Mouse numbers; say so. In humans, Brandon Faubert and colleagues infused carbon-13 lactate into five patients with non-small cell lung cancer and found extensive labeling of their tumors' TCA metabolites, clearest in the tumors that lit up brightly on the sugar scan this episode comes to in a moment. The head-to-head comparison, the one showing lactate predominating over glucose, was done in mouse xenografts, not in those five patients (Faubert 2017). The show's own page states that predominance as a human finding. It is not one — a species transfer, the commonest way a true result becomes a false sentence.

Lactate also signals, and the enzymology is unsettled in public. Di Zhang and colleagues found 28 lactylation sites on core histones, the proteins DNA is spooled around, the lactate-derived tag switching genes on directly (Zhang 2019). Then the arithmetic objection: lactyl-CoA, which donates the tag, sits a thousandfold below acetyl-CoA, making p300 a doubtful lactyl-tagger. Two answers followed within months — AARS1, borrowed from the protein-building machinery (Ju 2024), and ACSS2 as a true lactyl-CoA synthetase coupled to KAT2A, with a co-crystal structure, driving PD-L1 expression and immune evasion in brain tumors (Zhu 2025) — and in 2025 a review from Yingming Zhao's laboratory, which found lactylation, names writers, erasers, readers, synthetases and an emerging route that skips lactyl-CoA entirely (Sheng 2026). Three candidate enzymes, and the founding laboratory saying the chemistry is unsettled. All cells and mice; none of it human. The immunology is the same class and lives on the lactate page: deleting MCT1 — monocarboxylate transporter 1, one of the pumps that move lactate across a cell membrane — from regulatory T cells mattered inside the tumor and not in the periphery (Watson 2021), low-lactate melanomas grew slowly only in immunocompetent mice (Brand 2016), and tumor lactic acid pushed macrophages toward the M2 state (Colegio 2014). Lactate is waste is dead; lactate is the problem is not established.

Last, the question that sounds settled and is not. Who eats the glucose inside a tumor? Bradley Reinfeld and colleagues measured it by cell population in several mouse cancer models: myeloid cells had the greatest capacity to take up tumor glucose, then T cells, and the cancer cells were last, with the highest glutamine uptake instead. That contradicts the most repeated line in popular cancer metabolism — and a famous paper from the other direction, where tumor glucose consumption in a mouse sarcoma restricted T cells and lowered their interferon-gamma output, with checkpoint antibodies against CTLA-4, PD-1 and PD-L1 restoring glucose in the microenvironment (Chang 2015). Both mouse. Disputed, not resolved.

Thirty seconds on that 2015 paper for another reason. It is the paper a chatbot cited with a correct PubMed Central link, correct journal, date and institutions, attached to an author list with six people invented or misnamed and six of the real fourteen gone. The real list runs Chang, Qiu, O'Sullivan, Buck, Noguchi, Curtis, Chen, Gindin, Gubin, van der Windt, Tonc, Schreiber, Pearce and Pearce. Name the error, not the machine: a fabricated author list bolted onto a genuine record. A correct link is not a correct citation — the first of two fabrications this part retires by name.

And since the question is who eats the glucose, settle a list that circulates about parasites: four organisms named as hijacking glucose while living in oxygen-rich human blood. Only one qualifies. Red cells infected with Plasmodium falciparum consume glucose 50 to 100 times faster than uninfected cells, most of it leaving as lactic acid (Roth 1990). Giardia lives in the gut lumen and Trichomonas vaginalis in the urogenital tract, so neither is a bloodstream organism, and Toxoplasma's lactate output this show could not verify either way, so it stays off. Then correct malaria against itself: carbon-13 tracing shows a conventional oxidative TCA cycle in both blood stages, flux low in the asexual parasites and markedly higher in the gametocytes, which die when it is blocked (MacRae 2013).

The lethal version gets its cause told backwards. In 306 Kenyan children with severe malaria the acidosis was predominantly high-anion-gap in at least 43 percent, and the same abstract implicates hypovolemia, renal impairment and anemia (English 1997) — while microvascular imaging put severity down to obstruction of the small vessels by parasitized red cells more than to low blood volume, which is why the field moved away from fluid loading (Hanson 2014). The lactate is largely a perfusion failure, not parasite output. And Leishmania does not eat our lactate: a fructose-1,6-bisphosphatase knockout places the amastigote in a glucose-poor phagolysosome living on amino acids (Naderer 2006). The parasite eats our protein, which is both true and worse.

A map of where glucose is being trapped

Say what the machine does before trusting it. PET is positron emission tomography. A tracer atom decays by throwing out a positron, the electron's antimatter twin; it meets an electron, and the two annihilate into a pair of gamma rays flying off in opposite directions. A ring of detectors catches both ends and works backwards to where they came from, and millions of pairs make a map of where one molecule has collected. Here the molecule is sugar: FDG, fluorodeoxyglucose, glucose with radioactive fluorine-18 where the second carbon should carry a hydroxyl group. A cell cannot tell the difference. FDG enters through the same GLUT transporters that admit glucose, and hexokinase, the first enzyme of sugar breakdown, clamps a phosphate on, which normally stops glucose leaking back out. The next enzyme cannot act on it, so it piles up. Gallagher and colleagues named that in 1978: metabolic trapping (Gallagher 1978). That record has no abstract, no digital object identifier, and two typographical errors in its title.

The number on the report is the standardized uptake value, SUV: radioactivity in a region divided by the dose spread over the patient's weight, with SUVmax the brightest point in a lesion. The European procedure guideline defines it, with the fasting and timing rules that make centers comparable (Boellaard 2015). Oncologists long worked to a rough line of about 2.5 — the decision rule in what follows.

What does a bright spot track? Not division. In rats carrying three tumors, FDG uptake correlated with messenger RNA for the transporter GLUT1 at 0.83 and hexokinase at 0.77, and did not track how fast the cells were dividing (Haberkorn 1994). Rat data.

Then the finding that stops this being tidy. In nineteen melanomas proven under the microscope, GLUT-1 was present in seventeen and tracked brightness at P below 0.0001, hexokinase II was expressed in none, and Ki-67, the marker of dividing cells, showed no relationship to uptake (Park 2012). Be careful what that licenses. Hexokinase II is one of four isoforms, and the one that antibody test looks for; hexokinase I, commoner in many tissues, was not measured. So this tumor type traps the tracer without the isoform the textbook names, not with no enzyme. Nineteen patients, called preliminary by their authors — but 'the scan shows how fast the cancer is dividing' is wrong, and wrong in the data the scan was built on.

Who else traps glucose? In 1992 Ryuichi Kubota and colleagues injected mice carrying transplanted tumors and laid film against thin tissue slices. The granulation tissue around the tumor and the macrophages crowding the dead margins held more FDG than the living cancer cells, and up to 29 percent of that tumor's glucose use came from cells that were not tumor (Kubota 1992). Mouse data, and the engine of every false positive here. An activated immune cell is a glucose-hungry cell. That is not a flaw in the tracer; it is the tracer working.

Across 40 studies and 1,474 focal lung lesions, FDG-PET's working point was about 96.8 percent sensitivity against 77.8 percent specificity — sensitivity being the share of real cancers a test catches, specificity the share of non-cancers it clears, so one benign lesion in five lit up anyway (Gould 2001). Thirteen years later, the same journal, the reversal. Across 70 studies and 8,511 nodules, pooled sensitivity was 89 percent and specificity 75 percent. But adjusted specificity — that figure after other differences between studies are statistically accounted for — was 61 percent where infectious lung disease is endemic against 77 percent where it is not. A 16-point gap, with heterogeneity at an I-squared of 87 and 82 percent. Then the conclusion in full, because the conditional clause is what gets cut: 'These data do not support the use of FDG-PET to diagnose lung cancer in endemic regions unless an institution achieves test performance accuracy similar to that found in nonendemic regions' (Deppen 2014). Be exact about scope: endemic infectious lung disease, and the abstract names no parasite. This show's own page has credited that meta-analysis with a parasite finding it does not contain, and I am correcting that here.

The single-center series name the benign diagnoses. In 191 Indian patients with an indeterminate lung nodule, median SUVmax was 11.2 for the cancers against 10.3 for the tuberculous ones, and at the 2.5 cutoff specificity was 34.7 percent (Purandare 2017). Scanning twice after one injection, sold as the fix, separated nothing where tuberculosis is endemic (Sahlmann 2004; Sathekge 2010).

Then the bright spots needing no pathogen. Brown adipose tissue appeared in 7.5 percent of 1,013 women and 3.1 percent of 959 men across 3,640 scans (Cypess 2009). Reactive armpit nodes after COVID-19 vaccination appeared in 38.6 percent of 5,010 subjects pooled across 25 studies — and give the second half: most were faint, the higher-activity ones that cause real errors were less frequent, and reading them against blood pool rather than by eye brought the figure below 20 percent (Adin 2025). In 2025 the European and American nuclear medicine societies called hybrid FDG-PET 'the method of choice for a wide variety of infectious and inflammatory disorders' (Abikhzer 2025). Every parasitic false positive here is the tracer doing by accident what another subspecialty uses it for on purpose.

Now the parasites. Geometry first: where the uptake sits is the only useful clue. Of seventeen proven lung hydatid cysts — the fluid-filled cysts of the dog tapeworm — scanned before anyone knew what they were, ten showed the doughnut sign, a bright ring around a cold center, five were FDG-negative outright, and the brightest reached an SUVmax of 15.8, which in a lung nodule reads as cancer (Yoldaş 2022).

In alveolar echinococcosis the uptake is perilesional — in the host tissue around the lesion, not inside it. Across 120 scans of 70 patients on antiparasitic drugs, rescanning at three hours changed the interpretation in 32.5 percent of cases, and the authors say the parasite may stay viable even after perilesional uptake has gone (Caoduro 2013). How often is it hot? Ninety-three percent of hepatic lesions in 61 patients (Brumpt 2019); and in a 203-patient cohort, among the 89 with both PET and magnetic resonance imaging, 96.4 percent of Kodama type 2 lesions and 90.6 percent of type 3, the commonest types, while type 5 showed none (Eberhardt and Haggenmüller 2025).

Reuter's follow-up is sharper. Of fifteen patients rescanned a median of 6.5 years after a 1999 baseline, eleven had lesions inactive at baseline; only two stayed inactive across 81 months, and in two, activity appeared anew at 80 and 82 months. Their words: 'Lack of metabolic activity indicates suppressed parasite activity and is not equivalent to parasite death.' Give them the other half this show has left out: suppression can last years, so treatment can be paused, and relapses were reliably caught (Reuter 2008).

In 121 patients from the German echinococcosis database, mean SUVmax was 6.0, reaching 18.0 — malignant territory for a benign lesion. The authors call it an indirect reflection of lesion activity, and the scan the method of choice for inflammatory activity (Kratzer 2025). In twenty patients scanned with Patlak analysis, which separates tracer trapped in cells from tracer still in blood, the uptake ratios tracked the patient's own antibody level against hydatid fluid — Patlak ratio 0.85, SUVratio 0.73, the only correlations the paper reports, a co-author affiliated to the maker of the hardware it advocates (Husmann 2025) —. And the oncology line has no authority here: a cutoff of 2.09 separated alveolar from cystic disease at an area under the curve of 0.99 (Shen 2024).

The prospective test in cystic echinococcosis came out negative: of sixteen patients only one had uptake in the cyst, and the one cyst viable at surgery was FDG-cold (Salvador 2021). One study has measured how often the sugar scan misses a parasitic lesion, by running a second tracer against it: in 8 patients with cerebral alveolar echinococcosis carrying 35 lesions, carbon-11 choline found all 35 and FDG 24, an accuracy of 68.57 percent (Ayituhongman 2025). One center, prospective and weak — and the only measured false-negative rate for FDG in a parasitic lesion that exists.

Then the patients who were cut. A smoker's lung nodule was resected; inside was a granuloma around a clotted artery holding a Dirofilaria — in this form the larvae die before maturing and lodge in the pulmonary arteries (Stone 2015). A 17-year-old girl lit up from neck to abdomen and was worked up for malignancy, the picture mimicking lymphoma; the pathology was cysticercosis (Jiang 2014), and two parasitologists contested that reading the next year in a letter with no abstract, so I report that the dispute exists, not what it says (Galán-Puchades 2015). And the most treacherous: a woman with a raised CA19-9, a blood marker of bile duct cancer, and uptake in liver and a celiac node, read as cholangiocarcinoma, when the cause was the liver fluke Clonorchis sinensis (Son 2025) — Part 002 and the liver-fluke page, because that fluke both mimics the cancer and causes it. The pattern is not historical either: two more lesions were read as metastatic malignancy and as chondrosarcoma in 2026 and 2025 (Joseph 2026; Avcı 2025).

Here is the mirror that proves the mechanism: change the background, and the sign flips. A review of 29 HIV-positive patients with brain lesions found eleven with cerebral toxoplasmosis, cold at mean SUVmax 3.5, and five with primary brain lymphoma at 18.8 — so the contrast rests on sixteen patients (Lewitschnig 2013). Cortex is the hungriest tissue in the body, so against it an infected lesion is a hole. The signal is about host tissue, not the organism.

Now the gap, stated as a gap. No study has separated parasite FDG from host FDG in a single human lesion. The host side rests on mouse autoradiography, rat inflammation and the perilesional localization in patients. So the honest formulation is not 'entirely host.' It is overwhelmingly host, with a documented but unquantified parasite contribution. The settling study can be named: take resected FDG-avid parasitic lesions, count tracer in sorted cells or on film, and report counts per gram in worm against host rim. Until someone does, anyone who tells you how much of the glow is the worm is guessing.

Documented how? Nicolas Salem and colleagues showed that live adult Schistosoma mansoni in a dish take up FDG in proportion to worm number. Then athymic nude mice, bred without a thymus, stripping out most of the T-cell-driven granulomatous response: FDG uptake in the portal vein and liver tracked worms recovered at an R-squared of 0.58 across 40 animals, and 0.85 in the 17 carrying more than 50 worms. The dish removes the host; the nude mouse most of its inflammation.

The tapeworm has the machinery too. Echinococcus multilocularis lacks a gut and absorbs glucose across its syncytial tegument, the living skin covering it; of two cloned transporters, EmGLUT1 but not EmGLUT2 carried 2-deoxy-D-glucose in frog oocytes (Kashiide 2018). The limit goes in the same sentence: that is the parent molecule, the fluorine-18 version was never tested, and nobody has measured the clinical signal it accounts for. The parasite drives the host's half too: Echinococcus protoscoleces pushed macrophages toward the M2 state through PI3K, AKT and mTOR (Zhang 2022) — against a separate finding that M1 and M2 polarization by themselves did not raise deoxyglucose uptake, where only bacterial lipopolysaccharide did (Tavakoli 2013). Two culture papers disagreeing; say so rather than choose.

The worm's own engine, and ours

Everything so far has been ours — the tumor's chemistry, and the scan we built out of it. The next stretch is the worm's, and then we come back to ourselves, which is where this gets uncomfortable. In 1975 Schiller, Bueding, Turner and Fisher cultured adult Schistosoma mansoni for twelve days and found the same movement, pairing, glucose use and lactic acid production at 5 percent oxygen as at none. No Pasteur effect: oxygen did not slow the fermentation, and the worms were undamaged. What oxygen changed was eggs — 118 viable eggs per worm pair with oxygen against virtually none without (Schiller 1975). A blood-dwelling parasite fermenting with oxygen present is, in form, the tumor phenotype — and it is 1975 biochemistry. Then refuse to generalize: Fasciola hepatica miracidia cannot function anaerobically at all, in a paper confessing that the belief that free-living stages are aerobic 'originated mainly intuitively' (Boyunaga 2001).

Now the gut worm. An adult Ascaris suum, the roundworm of pigs, lives where there is very little oxygen, and runs mitochondrial complex II backwards. Complex II is succinate dehydrogenase, the only enzyme in both the TCA cycle and the electron transport chain; in us it takes electrons off succinate and puts them on coenzyme Q. In the adult worm it runs the reverse, taking electrons off a quinol and putting them onto fumarate — which is why its name changes: quinol-fumarate reductase. And somebody has the atomic coordinates: four subunits, five cofactors, in its authors' description essentially the aerobic enzyme but for an extra peptide on the smallest anchor subunit, with the reversal attributed to the low redox potential of rhodoquinol (Shimizu 2012).

It is not even a different machine from the larva's: the adult's fumarate reductase and the larva's succinate dehydrogenase share the same iron-sulfur subunit (Amino 2000). What the adult lost is the destination. Cytochrome oxidase — complex IV, which hands electrons to oxygen and makes water — is undetectable in its muscle mitochondria (Komuniecki 1993), while the worm still makes a textbook cytochrome c, the courier that normally carries electrons to complex IV, at a mammalian redox potential (Takamiya 1996). It kept the courier and lost the address.

What the reversed enzyme is for has a name: malate dismutation. One pool of malate is split two ways at once. Part is oxidized to pyruvate and carbon dioxide, by a malic enzyme that in this worm uses NAD where ours uses NADP — the real host-parasite difference, and the end of the folk claim that worms do not breathe out carbon dioxide (Clancy 1992) — and part is reduced through fumarate to succinate. The halves balance each other's electron books, and nothing at the end needs oxygen. Notice what comes out: succinate, acetate and propionate, and for Ascaris muscle the branched acids 2-methylbutyrate and 2-methylvalerate (Suarez de Mata 1991). Not lactate. The popular comparison reaches for lactic acid because it is the Warburg molecule, and for this worm it is the wrong end product. Succinate, fumarate and malate are the four-carbon series; lactate is the three-carbon outlier. Those are pages, not pictures.

Then correct 'adult helminths are strictly anaerobic,' and with a date, because it is sold as a recent overturning. Adult Paragonimus westermani, the lung fluke, carries cyanide-sensitive succinate oxidase and, in the same preparation, NADH-fumarate reductase eighteenfold above bovine heart, with both quinones present at once. The authors' own word is 'facultatively anaerobic,' and the year is 1994 (Takamiya 1994). Haemonchus contortus larvae had a large anaerobic capacity before they were parasitic, while the adults reverse complex II (Roos and Tielens 1994). And 'entirely debunked' fails the other way: a 2025 re-examination argues the model was over-generalized from Ascaris and cannot be extrapolated to other helminths (Vairoletti 2025). Two companion claims are wrong: no record places alternative oxidase, a plant and fungal enzyme, in a helminth genome, and 'cytochrome o' is bacterial.

Which brings us to the molecule it all turns on, and the second fabrication retired here by name. Rhodoquinone is coenzyme Q's red cousin: the same fat-soluble electron shuttle in the mitochondrial membrane, with an amine where Q carries a methoxy group, and a redox potential — how willingly a molecule gives up electrons — low enough that electrons run off it onto fumarate. That lower potential is the whole reason complex II can be made to run backwards on purpose. The claim to retire is that helminths build it by the shikimate pathway, using an enzyme called Rqka, 'rhodoquinone biosynthesis protein A,' swapping a methoxy group on ubiquinone for an amino group. Animals have no shikimate pathway, and there is no such gene and no such protein. Nor is ubiquinone the precursor: deleting kynureninase in Caenorhabditis elegans abolished rhodoquinone and left ubiquinone untouched (Roberts Buceta 2019).

The real route is better than the invention. In animals rhodoquinone comes from tryptophan, broken down along the kynurenine pathway to anthranilic and 3-hydroxyanthranilic acid and prenylated by a COQ-2 isoform unique to these species — so the amine is there at the start, not swapped in at the end (Del Borrello 2019; Salinas 2020). Across more than a hundred helminth genomes, only kynureninase and the tryptophan dioxygenases proved essential, and the worms kept that stripped-down pathway not to make NAD, which is what it is for in us, but to build the red quinone — paying with total dependence on NAD salvage, which FK866 exploits by killing cultured worms (Comas-Ghierra 2023). Nor is there an evolutionary mystery: bacteria and some protists reach the same molecule by an independently evolved route using rquA, spread by lateral gene transfer (Salinas 2020). The rhodoquinone page carries the rest: Fasciola builds its own from mevalonate rather than modifying host ubiquinone, a documented negative (Van Hellemond 1996), and the molecule turns up in every examined eukaryote that reduces fumarate during anoxia (Van Hellemond 1995).

And then the reversal, dated February 2025. Rhodoquinone was detected in mitochondria purified from certain mouse and human tissues, delivering electrons to fumarate by reversing succinate dehydrogenase, independent of ambient oxygen — and, in the abstract's own words, that chain 'is strictly present in vivo and is undetectable in cultured mammalian cells' (Valeros 2025). That erases the clean worm-and-human dividing line. Then hold the evidence class steady: absolute quantification in mammalian tissue became possible only in 2026, when pure rhodoquinone standards were synthesized — in the same laboratory that reported the detection (Do 2026). The same small community supplies the standards used across these papers, so citing three of them is not three independent confirmations. And the dispute is live: a 2026 review of anthelmintic pharmacology still rests its selectivity argument on that system being 'absent in humans' (Vinaud and Bezerra 2026).

Now turn the instrument around, because our own mitochondria run the worm's reaction. In 2021 Jessica Spinelli and colleagues asked what happens in a mammalian mitochondrion when oxygen cannot be reduced — no oxygen, or a drug blocking the enzyme that uses it. The chain does not stall: complex I, the gateway where electrons carried by NADH enter, went on working, and so did dihydroorotate dehydrogenase, the enzyme that builds pyrimidines. What piled up was ubiquinol, the reduced form of coenzyme Q, and that pushed the succinate dehydrogenase complex backwards, so the electrons landed on fumarate — the terminal electron acceptor, the molecule at the end of the line that in every textbook is oxygen (Spinelli 2021).

The numbers are the whole point. Fumarate reduction was undetectable at 20 percent oxygen, which is room air; stimulated by a drop to 15 percent; maximal at 3. Under hypoxia or antimycin A, which blocks complex III, the reversal exceeded forward succinate oxidation about fourfold. Then the tissue map, in mice. Liver, kidney and brain net-reverse the complex even at atmospheric oxygen — measured in tissue cultured outside the animal, because the in-vivo protocol was a bolus injection whose labeling never reached steady state, so the paper says forward and reverse activities could not be compared there. Under hypoxia, pancreas, white fat, thymus and lung join them; heart and calf muscle do not net-reverse (Spinelli 2021).

Two corrections belong here. First, a number: the claim in circulation is that mammalian mitochondria reverse complex II below 1 percent oxygen. That is wrong by more than an order of magnitude, in the direction that matters, since the reaction is stimulated at 15 percent and peaks at 3. Call it the threshold error — below 1 percent it sounds like a last resort; stated correctly it is something a mouse liver does in air. Second, sourcing: the claim was hung on Banerjee and Kumar's 2022 piece as in vivo isotope tracing. It is a review of two primary papers, and PubMed types it as one. Cite it anyway for one sentence: the fumarate-succinate and ubiquinol-ubiquinone pairs are nearly iso-potential, about 10 millivolts apart, so complex II is 'poised for facile reverse electron transfer' (Banerjee and Kumar 2022). Ours is balanced on a knife edge; the worm installed a lower-potential carrier and fixed the direction. And a caution: 'cancer runs the TCA cycle backwards under hypoxia' usually means reductive carboxylation, glutamine carbon pushed into citrate by IDH1 and IDH2 to build fat (Filipp 2012).

Hypoxia is not the only way to push our complex II backwards, and the other way lives in the gut. Hydrogen sulfide is both something our own cells make and a respiratory poison: it hands its reducing power to coenzyme Q through sulfide quinone oxidoreductase while inhibiting complex IV, and at the concentrations that poison complex IV a redox cycle runs between the two, complex II in reverse. Knocking complex II out of mouse intestinal epithelium cut thiosulfate, the biomarker of sulfide oxidation, to about a third of control (Kumar 2022). And the place is no anaerobic box: a 2025 review describes a steep radial oxygen gradient running, in its authors' words, from a virtually anoxic lumen to a highly vascular lamina propria (Brake and Banerjee 2025). Our own gut chemistry runs our own complex II backwards, in the compartment the adult worm occupies — though that review has no primary data, and its authors say the methods for quantifying it in living animals are still being built.

Now the tumor claim, at its exact strength. Tomitsuka, Kita and Esumi started from an honest premise, worth hearing as published, in their words and spelling: 'Increased glycolysis is the principal explanation for how cancer cells generate energy in the absence of oxygen. However, in actual human tumor microenvironments, hypoxia is often associated with hypoglycemia because of the poor blood supply. Therefore, glycolysis cannot be the sole mechanism for the maintenance of the energy status in cancers.' Their abstract spells it tumor; this show says tumor. So they put DLD-1, Panc-1 and HepG2 into exactly that condition, and NADH-fumarate reductase activity rose while NADH-oxidase activity fell. And pyrvinium pamoate, an anthelmintic approved for pinworm, inhibited it in both parasite and mammalian mitochondria (Tomitsuka 2012). One drug, one reaction, two kingdoms. Then the conflict of interest: three authors of the review that first proposed carrying this reaction from parasites to cancer cells (Sakai 2012) are the same three who published the experiment. Not fraud, and no reason to discard the measurement — a reason not to count proposal and test as two independent voices.

Now the ceiling, which is low: three cell lines, no animal tumor model, no patients. The cytotoxicity is conditional: a paper by Isao Ishii and colleagues — a review in substance, though PubMed types it only as a journal article — reports potent killing only during glucose starvation, and names five competing routes for the drug, from the fumarate reductase system to Wnt signaling and the androgen receptor (Ishii 2012). Five mechanisms for one molecule is what an unsettled mechanism looks like. The human record is thinner: an open-label single-arm protocol dosing the drug before pancreatic surgery (Ponzini 2023). A protocol is a plan, and searched in October 2026 PubMed indexes no results from it. One more limit comes from the 2025 rhodoquinone paper: because that chain is undetectable in cultured cells, every one of those experiments was run without it (Valeros 2025). A real echo, not the same engine.

Which leaves the biggest hole in the season's parallel, said rather than filled in: nobody has measured fumarate reduction in a human tumor in a patient. The settling study is possible with instruments that exist — infuse a labeled tracer in the operating room and report net fumarate reduction against forward succinate oxidation, by histology — then ask whether rhodoquinone is in human tumor mitochondria at all.

What is sold from the idea, and what the comparison licenses

Drugs first: tried properly, failed properly. IACS-010759 is a potent, selective inhibitor of complex I. Two phase I trials ran it: seventeen patients with relapsed acute myeloid leukemia under NCT02882321, twenty-three with solid tumors under NCT03291938. Its therapeutic index was too narrow, and the dose-limiting toxicities were neurotoxicity and raised blood lactate — Warburg's own readout as a side effect. No recommended phase 2 dose, limited activity, both trials discontinued (Yap 2023), and the compound was discovered and trialed by units of one cancer center.

Metformin inhibits the same complex weakly, and two large randomized trials found nothing. A hazard ratio compares how fast events happen in one arm of a trial against the other, moment by moment: one means the arms are the same, and below one favors the treatment. MA.32 randomized 3,649 patients with high-risk non-metastatic breast cancer, and after futility was declared in the receptor-negative group the primary analysis ran in the 2,533 who were hormone-receptor-positive: invasive disease-free survival gave a hazard ratio of 1.01, 95 percent confidence interval 0.84 to 1.21, p equals .93. In the receptor-negative patients it was the same, 1.01, interval 0.79 to 1.30, and grade 3 non-hematological toxicity was commoner on the drug, 21.5 against 17.5 percent (Goodwin 2022). In STAMPEDE, 1,874 men with metastatic prostate cancer gave an overall survival hazard ratio of 0.91, the interval 0.80 to 1.03 (Gillessen 2025).

The rest runs the same way. Telaglenastat blocks glutaminase, the enzyme that begins glutamine breakdown; in CANTATA, 444 patients with advanced kidney cancer gave 0.94, the interval 0.74 to 1.21, with two authors employed by the sponsor (Tannir 2022). Devimistat aims at fuel entry into the citric acid cycle; in phase 3, 528 patients with metastatic pancreatic cancer gave 0.95, the interval 0.77 to 1.18 (Philip 2024). AZD3965 was the first MCT1 inhibitor given to people, the same lactate pump the regulatory T cell work turned on: in 40 patients it reached target concentrations, produced seven dose-limiting toxicities, five of them reversible retinal changes, and reported no efficacy — and patients with retinal or cardiac disease were excluded, because that pump sits in eye and heart (Halford 2023).

Then the ones sold hardest. Dichloroacetate: the randomized test in 45 patients with head and neck cancer lowered serum lactate as predicted, raised three-month complete response from 37.5 to 71.4 percent, and made no difference to five-year survival (Powell 2022). And 3-bromopyruvate, evidenced least: 'it eradicated tumors in all 19 of 19 rodents' comes from an editorial by the head of the originating laboratory, with no new data and no controls (Pedersen 2012), and the indexed human record is two case reports — a fibrolamellar liver cancer indexed with a fatal outcome (Ko 2012), and a man with stage IV melanoma in whom the drug had, in the authors' words, minimal anticancer effect (El Sayed 2014). This show's page says one case report. It is two.

Diets next. Ketogenic diets in glioma are safe, feasible and not shown to extend survival. A 2025 review of 18 studies found its two randomized trials showed no significant survival benefit (Gritsch 2025). A 2026 systematic review took in 43 reports, only 20 clinical — eighteen on ketogenic diet, two on methionine restriction — and found the evidence supports feasibility and metabolic activity, not efficacy (Ghoche 2026). ERGO2 gave 50 patients with recurrent brain tumors re-irradiation with or without a calorie-restricted ketogenic diet, and missed its endpoint: progression-free survival at six months 20 percent against 16 (Voss 2020). A 2026 meta-analysis reporting 29.4 months against 14.6 cannot carry that weight, because historical controls are not a comparison group (Firdous 2026). The much-quoted finding that diet patients with below-median glucose lived longer is an after-the-fact subgroup (Voss 2022).

Outside the brain, a 2026 trial randomized 32 evaluable patients with metastatic pancreatic cancer to chemotherapy with or without a ketogenic diet: progression-free survival 8.5 months against 6.2, hazard ratio 0.53, overall survival 13.7 against 10.2, both intervals including no effect (Jameson 2026).

The legitimate version of the dietary argument is narrower, and it is about drugs. PI3K is the enzyme insulin signals through, so inhibiting it raises blood glucose, and the compensating insulin switches the pathway back on inside the tumor. Blocking that feedback in mice, by diet or by drugs, greatly improved the efficacy-to-toxicity ratio of PI3K inhibitors (Hopkins 2018). And the harm, also in mice: in animals carrying cancers that make interleukin-6, a ketogenic diet slowed tumor growth but brought on cachexia, the wasting of muscle and fat, sooner, and shortened survival (Ferrer 2023). One sugar correction belongs here: fructose enhances tumor growth in mice, but not as tumor fuel, because the cancer cells lack ketohexokinase-C and the liver does the fructolysis (Fowle-Grider 2024). And one definition straightened: the obesity paradox is not that excess fat worsens prognosis but that overweight and obese patients sometimes survive better (McQuade 2018), while the paper cited as proving fat mass causes tumors is a narrative review saying 'we have limited insight' (Park 2011).

Alkalinity borrows the vocabulary and has none of the evidence. A systematic review screened 8,278 citations and included one study: no randomized trials of acid load or alkaline water as treatment (Fenton and Huang 2016). Cesium chloride, sold as high pH therapy, has produced no confirmed regression, and near 6 grams a day causes low potassium and lethal arrhythmia (Melnikov and Zanoni 2009; O'Brien 2008). The legitimate cousin is oral bicarbonate buffer therapy, whose own authors call 'alkalizing agents' a misnomer, because a buffer raises tumor pH toward 7.4 without alkalinizing normal tissue; their three trials foundered on taste and accrual (Gillies 2022). The best pro-alkaline human evidence is a retrospective series of 129 colorectal cancer patients at the clinic providing the therapy, where urinary pH at or above 7.0 was associated with longer survival — a design that shows association, not effect, and whose authors go no further than calling for prospective evaluation (Suzuki 2026).

The mechanism underneath is not clean either. Ras-transfected hamster cells engineered to make under 4 percent as much lactic acid, grown as tumors in mice, measured pH 6.65 against 6.78 in the parental tumors (Newell 1993). A tumor making almost no lactic acid was just as acidic, carbon dioxide a significant source. Then the document that refutes the trade. In the same 1927 paper, Warburg, Wind and Negelein accepted that insulin slowed tumor growth and glucose sped it up, and calculated that a modest fall in blood sugar would cut tumor fermentation to a quarter — but they separated slowing a tumor from killing it. His words, in the English text: 'Even if it were possible to remove the blood-sugar entirely in living animals, the life of the tumor would not be threatened.' He then held tumor-bearing rats at very low blood sugar with insulin, to convulsions, with tumor respiration and fermentation nearly normal. The man whose measurement sells the diet published the experiment against it, ninety-nine years ago.

So what does the comparison license? Three things. Aerobic fermentation is a strategy, not a symptom: adult Schistosoma mansoni used glucose and made lactic acid at the same rate at 5 percent oxygen as at none (Schiller 1975). Our own mitochondria do run complex II backwards onto fumarate, liver, kidney and brain net-reversing it in tissue cultured at atmospheric oxygen (Spinelli 2021), and since February 2025 some human tissues are known to carry the worm's quinone (Valeros 2025). And the convergence is on an environment both lineages inherited the chemistry for: no enzyme of core anaerobic energy metabolism is specific to one eukaryotic supergroup, which reads as presence in the common ancestor followed by differential loss (Müller 2012). A hypoxic tumor cell reaching for fumarate is not imitating a worm; both are reaching for something old. And the cleanest parasite-side fuel result the season has is not in this part: Theileria, which installs a reversible Warburg-like state in the cells it infects and loses it when killed. Part 003 carries it.

It does not license that cancer is a parasite, that parasites cause cancer through metabolism, or that shared chemistry is shared mechanism. Three parasites are Group 1 human carcinogens, the human evidence sufficient, as Part 001 established from IARC's Volume 100B — and the route Part 001 takes from that volume's full text is chronic inflammation and injury, not fuel; the PubMed record carries no abstract, so that attribution belongs to Part 001, not the record. The glycogen parallel fails the same way. Worms and tumors do both hoard glycogen, but in clear cell renal cell carcinoma, the cancer most famous for it, knocking out glycogen synthase, deleting both phosphorylases and running xenografts changed nothing — 'a secondary, and apparently dispensable, consequence' of constitutive HIF-1-alpha signaling (Xie 2021). Then refuse to over-generalize, because in pancreatic cancer glycogen mobilized by PYGL does fuel glycolysis and liver metastasis (Ji 2023). Same feature, opposite answer, two tissues.

What the fuel chapter changes is this season's counting question, at the instrument. The old version of this episode said a scan tells you something is there and a biopsy tells you what it is. Replace it: this scan tells you where glucose is being trapped, which is not the same thing as where cancer is. Its brightest compartment in a mouse tumor was granulation tissue and macrophages; its adjusted specificity for lung nodules falls from 77 to 61 percent where infectious lung disease is endemic; it is active in over 90 percent of the commonest lesion types of alveolar echinococcosis. The old episode was right that the diagnosis of cancer is morphological, made by looking at shape — and that is no dismissal of the scan, because the scan decides which lesions become biopsies, and where these parasites live it decides least reliably. And where the old version said nothing in the workflow ever asks a question a non-human organism could answer, amend it: one step does. FDG asks where glucose is being trapped, and a worm's inflamed host capsule answers loudly, in the wrong direction.

So the obstacle list has four items, not three. A core needle samples a thousandth of a tumor. The antibodies and primers are aimed at human targets. Nobody orders the test. And the instrument that finds and stages these lesions is least specific precisely where these parasites are endemic: nobody has measured what share of that 16-point gap is parasitic rather than tuberculous or fungal, and no study has separated parasite tracer uptake from host uptake in one human lesion.

The closing discipline: four numbers this part will not say. 'Thirty-six ATP per glucose' — the measured ratio is fractional, and gives about thirty (Hinkle 2005). 'About 5 ATP per glucose for malate dismutation' — searched for specifically, and nothing supports it. 'Seventy-nine and 91 percent of ATP from oxidative phosphorylation' — two reviews blended, and the likeliest origin is a measurement of a different quantity, the fate of the glucose a cell consumes rather than its ATP share (Wu, Ying and Hu 2016), though nobody has traced them back, which is the point. And 'fermentation makes ATP a hundred times faster' — that figure is the rate of making lactate (Liberti and Locasale 2016).

Why cells do it is open. Two cell-culture explanations stand five years apart, unreconciled: demand for NAD+ exceeding demand for ATP (Luengo 2021), against demand for compartmentalized cytosolic ATP with hexokinase detached from the mitochondrial membrane (Huggler 2026), whose abstract still calls aerobic glycolysis an enigmatic phenotype. A 2025 review from the laboratory behind the growth-program idea says no explanation fits all the conditions where it is seen (Li 2025). And a PubMed search on 9 October 2026 returned 5,355 records for 'Warburg effect,' 21 typed as retracted publications, including a 1981 Science paper titled 'Warburg effect revisited' (Racker and Spector) and a 2022 propofol paper (Qu). Two halves of this comparison go elsewhere: membrane potential and charge to the bioelectricity part, collagen and movement to its own.

Sources

Every paper referenced on air, in the order it comes up. Links go to the publisher via DOI.