Ammonia in Disguise

A tumor has a nitrogen problem. One transamination solves it — and the molecule that does the solving also carries carbon out and a standing order for glucose back. What happens to the nitrogen that misses the exit is stranger still.

Measure ammonia in a tumor and you will find some. Now ask how much nitrogen is actually leaving that tumor. Nobody has run that comparison directly — but almost everything we know about how tumors handle nitrogen says the second number should be much larger than the first.

The difference would be travelling under another name.

Much of a tumor's surplus nitrogen leaves disguised as alanine — a neutral, soluble, perfectly ordinary amino acid that nobody looks at twice in the bloodstream. The same molecule carries carbon out and a request for glucose back. One enzyme, one step, three services. And the fraction that does not get out quietly enforces the metabolic state that lets the rest escape. That is a claim about a flux nobody has measured directly. What follows is the case for it.

Part I / The Nitrogen Problem

The Reaction Nobody Looks at Twice

Alanine aminotransferase does something almost embarrassingly simple. It takes pyruvate — the end product of glycolysis, the chain of reactions that breaks glucose apart, and a molecule every cell makes by the bucketful — and swaps an oxygen atom for an amino group, a nitrogen with its hydrogens, borrowed from the amino acid glutamate.

pyruvate + glutamate ⇄ alanine + α-ketoglutarate

That is the whole reaction. Two versions of the enzyme run it — one in the body of the cell (ALT1/GPT) and one inside the mitochondria (ALT2/GPT2) — and it is fully reversible, meaning it can run in either direction. Which way it goes depends only on what is abundant on each side.

If you have ever had liver blood tests, you have already met this enzyme without being introduced. ALT is one of the numbers on the report, and it is there for a simple reason: when liver cells are damaged, their contents spill into the blood, and this is one of the things that spills.

It is not usually presented as a piece of cancer biology, and that is a shame — because in a tumor this enzyme is not spilling out of anything. It is running a logistics operation.

Ammonia: A Weapon the Tumor Can Also Die From

Start with the problem the operation exists to solve.

A fast-growing tumor burns through amino acids. It consumes glutamine at rates second only to glucose, and many tumors also break down the branched-chain amino acids — leucine, isoleucine and valine, the ones most abundant in muscle protein. Every amino acid broken down liberates a nitrogen atom that the cell now has to do something with. The naive solution is to release it as ammonia.

And here is where it gets interesting, because ammonia in a tumor is not simply waste. It is a weapon. A broad metabolic survey of a mouse model of metastatic colorectal cancer found a robust accumulation of ammonia inside the tumor, and those elevated levels pushed the local T cells into exhaustion — a worn-out state in which they stop multiplying and stop killing. Clearing the ammonia reactivated them, shrank the tumors, extended survival, and improved the response to anti-PD-L1, one of the checkpoint immunotherapies. A separate line of work describes glutamine-derived ammonia killing effector T cells outright — the ones that do the actual destroying of tumor cells — by a mechanism provisionally called "ammonia death," alongside evidence that ammonia favors regulatory T cells, whose job is to damp immune responses down.

One caveat on that study, because it bears on how much weight the word "weapon" can carry. The ammonia built up there not because the tumor was producing more of it, but because the tumor had lost the capacity to clear it, through a breakdown in urea-cycle gene expression downstream of the transcription factor HNF4α. That is a failure of disposal rather than an act of production. The same group's later experiments point the same way: a drug that lowers ammonia by acting on the gut microbiota did nothing to tumor size, while agents that add clearance capacity to the animal worked — which locates the source in the cancer epithelium rather than in a dysbiotic microbiome.

For the present argument that is almost better. It is a tumor whose nitrogen valve is broken, and the consequence is precisely the immunosuppression a working valve would have to give up.

So ammonia suppresses the immune system. A tumor benefits from that.

But ammonia is indiscriminate. It disrupts pH, mitochondria, and nitrogen metabolism in whatever cell it reaches, and the cell producing it is the one standing closest. A tumor that simply dumped all its surplus nitrogen as free ammonia would poison its own stroma, its own mitochondria, and eventually itself.

This is the same shape as the ROS problem. Reactive oxygen species — the damaging by-products of oxygen metabolism — are useful to a tumor at one concentration and lethal at another, and the whole redox economy exists to hold them in that window. Ammonia is the nitrogen-side analogue. The tumor's task is not to eliminate it and not to maximize it. It is to regulate it — to keep enough free ammonia around to do the immunosuppressive work while routing the rest somewhere harmless. (See also: ammonia in cancer development.)

Where the Two Economies Share a Substrate

That analogy is better than an analogy. The two economies draw on the same molecule at the same step: glutaminase splits glutamine into glutamate and ammonia, and in most proliferating cells that single reaction is both the largest source of free ammonia and the source of the glutamate that is simultaneously glutathione's committed precursor and the nitrogen donor the transaminase loads onto pyruvate. The products land in both ledgers in fixed proportion. Pull harder on glutamine and you get more ammonia to dispose of, more glutamate to dispose of it with, and more raw material for antioxidant defence — in one step.

The clearest demonstration is old and comes from an organ nobody studies for cancer. Working in the functioning rat kidney, Welbourne followed glutamine into both fates at once, and driving the γ-glutamyl cycle pulled glutamine uptake and ammoniagenesis up together while supporting glutathione turnover. The 1979 title names both currencies in a line. And the cycle is a loop rather than a drain: γ-glutamyl transpeptidase cleaves glutathione at the cell surface, the cysteine is salvaged for resynthesis, and the γ-glutamyl group returns through 5-oxoproline to glutamate — back into the pool the transaminase draws from. Glutathione turnover does not consume the amino donor. It returns it.

Tumors lean on this. GGT1 is a marker of cisplatin resistance: it keeps intracellular glutathione topped up, generates a cysteine-containing dipeptide that binds the drug outside the cell before it can enter, and protects against ferroptosis. What nobody appears to have done is put the two halves on one axis — the GGT literature is written by people studying drug resistance and does not mention ammonia; the ammonia literature does not mention glutathione turnover. They are describing one branch point from opposite sides.

            If the coupling is real, it is a prediction. Inhibit glutaminase and both ledgers move at once: less ammonia, less glutamate, less glutathione, less capacity to export nitrogen as alanine. Which is a reason to be careful about reading glutaminase inhibition as purely anti-anaplerotic — the same block is also a redox intervention and a nitrogen-disposal intervention.
        

The Valve Itself

Transamination is the valve. Instead of cutting the nitrogen loose, the cell hands it to a carrier. Glutamate accepts nitrogen from other amino acids, alanine aminotransferase moves it onto pyruvate, and the nitrogen leaves as alanine — soluble, neutral, unremarkable, and invisible to the assay we use to look for nitrogen stress.

Alanine is not the only way out, and it would weaken the case to pretend otherwise. Glutamine is itself a nitrogen carrier, two nitrogens to a molecule, and tumors both consume and release it. Glutamate leaves through the cystine antiporter. And there is an exit nobody would have guessed: working in cancer cells under low oxygen, one group found glutamine nitrogen routed into secreted dihydroorotate and orotate — pyrimidine intermediates dumped before they ever became nucleotides — while ammonia release actually fell.

That result deserves a pause, because it is this whole argument in miniature. Glutamine uptake went up. Ammonia output went down. The nitrogen went out wearing something else. Anyone watching the ammonia would have concluded the cell's nitrogen economy had gone quiet at precisely the moment it was running hardest.

So the claim is not that alanine is the only disguise. It is that alanine is the one with the most machinery behind it: a reaction nearly every cell can run, a substrate the glycolytic tumor already has in surplus, and a return leg that pays the tumor back in glucose.

Pyruvate is what makes this possible. It is the acceptor. A cell with nitrogen to shed and no pyruvate available has a problem. A cell with pyruvate in surplus has a disposal route that costs it nothing and poisons nobody.

And a tumor that runs on glycolysis has pyruvate in surplus by definition. The Warburg effect — the tumor habit of burning glucose down to lactate even when there is plenty of oxygen available to do better — is usually told as a story about energy: inefficient fuel use, lactate as exhaust. Seen from the nitrogen side it reads differently. A cell running glycolysis hard is continuously manufacturing the exact molecule it needs in order to export nitrogen safely.

            The waste product and the disposal vehicle are the same thing. Which means the tumor's glycolytic phenotype is not only a fuel strategy. It is also, quietly, a nitrogen-handling strategy — and the two cannot be separated, because one produces the substrate the other requires.
        

Part II / What Alanine Carries

What Alanine Actually Carries

Once the nitrogen is loaded onto pyruvate, the resulting molecule is doing three jobs at once. They are usually described in three separate literatures, by three separate communities, and almost never in the same paragraph. And in one tumor type the whole arrangement runs backwards.

Carbon

Carbon, moved

Three carbons the tumor did not finish oxidizing, exported intact rather than wasted.

Nitrogen

Nitrogen, shed silently

A nitrogen atom disposed of without ever appearing as free ammonia.

The liver

A bill sent to the liver

Carbon the liver will rebuild into glucose and return — at its own expense.

The reverse case

Carbon, delivered

In pancreatic tumors the stroma runs the reaction the other way and feeds the cancer.

The Reverse Case: The Stroma Delivers

Start with the exception, because it is the clearest demonstration that this carrier runs in both directions. In pancreatic cancer the alanine moves into the tumor rather than out of it.

Those tumors sit inside a dense wall of fibrous tissue, and the cells that build that wall — pancreatic stellate cells — turn out to be feeding the cancer. They secrete alanine, which outcompetes carbon from both glucose and glutamine as a fuel for the tumor's mitochondria, and for the fats and amino acids it needs to grow.

The mechanism is the part that should make you sit up. The stellate cells are not passively leaking. They are digesting their own internal components in order to produce that alanine — a normal survival process called autophagy, by which a cell recycles its own parts under stress — and the cancer cells are the ones sending the signal that switches it on. Block autophagy in the stellate cells alone, and tumor growth in mice largely stops.

Note what this costs the tumor: nothing. In an environment where glucose and blood-borne nutrients are genuinely scarce, it has a parallel supply line that drugs aimed at glucose metabolism do not touch, paid for by somebody else's self-digestion.

A Standing Order to the Liver

Now the forward direction. Alanine released into the blood goes to the liver, which does two things with it. It strips the nitrogen off and routes it into the urea cycle, and it takes the three carbons left over — now pyruvate again — and builds glucose. That glucose re-enters circulation, where a glucose-hungry tumor is waiting.

tumor: pyruvate + N → alanine → blood → liver
liver: alanine → NH₃ → urea | alanine → pyruvate → glucose → back to tumor

This is the glucose–alanine cycle, described in healthy physiology decades ago as the way exercising muscle offloads nitrogen and recycles carbon. A tumor does not invent it. A tumor subscribes to it.

Three separate services, one carrier. An ammonia assay sees none of it — the nitrogen in transit is bound into an ordinary amino acid the whole way.

The clearest measurement comes from an unlikely model: adult zebrafish carrying BRAF-driven melanoma. Those tumors consumed roughly fifteen times more glucose than any other tissue measured, and yet the animals' circulating glucose stayed normal. The reason was a tumor–liver alanine cycle. The melanoma exported glucose-derived alanine, which gave the liver carbon for gluconeogenesis and simultaneously let the tumor dump excess nitrogen from branched-chain amino acid breakdown — a pathway the authors found switched on in human melanoma too. Block the cycle pharmacologically and tumor burden fell.

Notice who pays. Rebuilding glucose is expensive; the liver spends its own energy reconstructing a fuel the tumor declined to finish burning. The equivalent loop that recycles lactate, the Cori cycle, already costs a net four ATP per turn, and the liver carries that cost. The glucose–alanine cycle is less efficient still, because the nitrogen it delivers has to be converted to urea on top. Wasteful loops of exactly this kind, running between organs, are one of the standing explanations for cancer cachexia — the progressive wasting that is recorded as the cause of death in up to a fifth of cancer deaths, and that nobody has been able to treat.

Part III / The Architecture

Where It Stops Looking Opportunistic

Take the three services together.

A single transamination lets a tumor move carbon it cannot finish oxidizing, shed nitrogen it cannot safely hold, and place a standing order with the liver for the fuel it most depends on. Three problems, three solutions, one molecule, one enzyme, one step. There is no elaborate machinery being assembled. Just pyruvate, which the tumor already has in surplus, and a transferase almost every cell carries.

That is the moment the picture changes. On its own, each service reads as a cell making do. Together they read as an architecture.

And the enzyme pays out in both directions. Remember that the reaction produces α-ketoglutarate alongside the alanine — a small molecule that an entire family of enzymes requires in order to function at all. Run the reaction one way and α-ketoglutarate accumulates, switching those enzymes on; among them are the ones that stiffen the collagen scaffolding around a growing metastasis. That is the mechanism by which breast cancer cells use pyruvate from their surroundings to build their foothold in the lung. Run it the other way and α-ketoglutarate falls, disabling the cell's oxygen sensor — which would otherwise destroy HIF-1α, the master switch for the low-oxygen survival program. A separate study found exactly this, with GPT2 overexpression lowering α-ketoglutarate, inhibiting PHD2 and stabilizing HIF-1α, which expanded the pool of stem-like cancer cells.

Run it forward

pyruvate → alanine

α-KG rises. The collagen-stiffening enzymes switch on, and the scaffolding hardens around a forming metastasis.

Run it backward

alanine → pyruvate

α-KG falls. The oxygen sensor stalls, the low-oxygen switch stays on, and the pool of stem-like tumor cells expands.

There is a further turn of the screw. If falling α-ketoglutarate stabilizes HIF, and HIF in turn induces the transaminase, the arrangement feeds itself. That is what a glioblastoma study found: hypoxia raised GPT2 at both message and protein level, and did so through HIF-2, which binds the hypoxia response element of the human GPT2 gene directly. Note the division of labour — HIF-2 does the inducing, and HIF-1 is not required. The low-oxygen program switches on the enzyme; the enzyme lowers α-ketoglutarate; lower α-ketoglutarate keeps the low-oxygen program on.

Two papers, the same enzyme, opposite effects on the same molecule, both working in the tumor's favor. That is not a contradiction in the literature. It is mass action — the ordinary chemical rule that a reversible reaction runs away from whatever is piled up. Whichever ingredient is abundant sets the direction. The tumor wins either way.

The Third Route to α-Ketoglutarate

Both of those routes move α-ketoglutarate by running the transaminase one way or the other. There is a third route, and it does not involve the enzyme at all. It involves the ammonia the valve failed to catch.

A cell facing more free ammonia than it wants does not simply tolerate it. It fixes it — and the acceptor it reaches for is α-ketoglutarate. Glutamate dehydrogenase runs the reaction backwards from its textbook direction, adding ammonia to α-ketoglutarate to make glutamate. Working in brain cells under the high-ammonia conditions of liver failure, one group labelled the nitrogen and watched it arrive in glutamate within minutes. The detoxification works. But the α-ketoglutarate it consumes is a TCA cycle intermediate, and draining it stalls the cycle — rapidly, measurably, and not because of any pH effect. Mitochondrial respiration fell within hours.

NH₃ + α-ketoglutarate + NAD(P)H → glutamate (ammonia handled, TCA cycle starved)

Two objections to that, and both are serious.

The first is concentration. The astrocyte experiments ran under liver-failure ammonia, well above anything a tumor is likely to see, and they concern a specific isoform — GDH2, whose regulatory profile differs from the GLUD1 form most cells rely on. Glutamate dehydrogenase does not run the fixing direction easily. One group measured its kinetics in cancer cells under near-physiological conditions and put the enzyme's half-saturation point for ammonium above 20 mM under conventional fitting — against intracellular tumor ammonium of roughly 1 to 5 mM. On that reading the reverse reaction barely happens. But GDH is cooperative, and fitting the same data to a model that accounts for the cooperativity dropped the figure to 6–12 mM; simulating the enzyme at physiological substrate and regulator concentrations, those authors predicted that cancer cells do fix nitrogen this way. The honest statement is that the step is plausible and contested, not established. It needs the cooperative model, and it needs local ammonia at the high end of what tumors reach. It is also not the only sink available — glutamine synthetase fixes ammonia onto glutamate at the cost of ATP, and in astrocytes that is normally the dominant route. The study above was specifically about the glutamate dehydrogenase arm.

The second objection is larger, and I have not seen it answered anywhere. Tumors consume glutamine partly in order to make α-ketoglutarate — that is what glutaminolysis is for, and it was shown decades ago to run far beyond what protein and nucleotide synthesis require. So the same glutamine catabolism that liberates the nitrogen load also floods the cell with the molecule I am claiming gets drained. Nitrogen pressure and α-ketoglutarate supply rise together. For the loop in this section to close, fixation has to outrun anaplerosis — in some compartment, at some moment, or in cells where glutaminolysis is already saturated.

I do not know that it does. What I will say is that the two processes are in direct competition, that the competition is measurable, and that nobody appears to have measured it.

Now consider what α-ketoglutarate is doing elsewhere in the cell. It is the required partner of an entire enzyme family — the same one mentioned above — and the most consequential members are the prolyl hydroxylases, the proteins that function as the cell's oxygen sensor. They cannot work without it. And we know the inhibition is competitive at that exact site, because raising α-ketoglutarate pharmacologically reverses it — that is how the oncometabolite story in succinate- and fumarate-driven tumors was worked out.

So ammonia does not need to inhibit anything. It removes the ingredient.

            The result is functional hypoxia. Oxygen is present in the cell and cannot be used — a lesion of the same functional kind as cyanide poisoning rather than of the same kind as a poor blood supply, though the site differs: cyanide blocks the terminal oxidase, while α-ketoglutarate depletion starves the cycle upstream of it. Whatever the cell does next, it does on glycolysis — because glycolysis is the only ATP source it has left.
        

You could carry this one step further and say HIF-1α is therefore stabilized, giving a full pseudohypoxic program. But a separate study found that starving cells of amino acids depleted α-ketoglutarate and inactivated the prolyl hydroxylases without stabilizing HIF at all — no change in its target genes. So the HIF layer is not automatic. But the functional-hypoxia argument does not depend on it.

Where the Loop Closes

Put that beside the pyruvate argument from Part I and the thing becomes self-reinforcing.

Glycolysis manufactures the pyruvate that lets nitrogen leave safely as alanine. Nitrogen that fails to leave becomes ammonia. Ammonia is fixed onto α-ketoglutarate, which stalls the TCA cycle and pushes the cell back onto glycolysis — which makes more pyruvate, which restores the export capacity.

The tumor's glycolytic commitment is not only permissive for nitrogen disposal. It is partly caused by it.

A tumor falling behind on nitrogen disposal is therefore pushed harder into the Warburg phenotype — and the push is what rescues it. Which suggests the arrow in Part I may run both ways. Glycolysis supplies the acceptor; nitrogen pressure helps keep glycolysis running.

            This puts ammonia in a category it is not usually placed in. Succinate, fumarate and 2-hydroxyglutarate are called oncometabolites because they disable the α-ketoglutarate-dependent enzymes and produce a hypoxia-like state in an oxygenated cell. Ammonia reaches a comparable endpoint by removing the co-substrate rather than competing for the site — a different mechanism, a similar result. Unlike the others it requires no mutation in a TCA enzyme. Any cell with a heavy nitrogen load and an inadequate exit can generate it.
        

The Asymmetry That Decides Who Benefits

All of which raises an obvious objection. Alanine is a common amino acid. The tumor microenvironment is not a private room. If alanine is floating around, why do the immune cells not simply use it too?

They would like to. In 2019 a group at Harvard and Princeton asked what a resting T cell needs in order to wake up and become a killer, and found something unexpected: T cells require alanine from outside themselves in order to activate. Deprive them of it and the whole process falters. Memory T cells — the ones held in reserve from a previous encounter — need it again to wake up a second time.

This is strange on its face. Alanine is one transamination away from pyruvate, and T cells have plenty of pyruvate. Why would they need to import it?

The answer is the whole point. T cells barely make alanine aminotransferase at all. They cannot run the reaction. What they do instead, on waking up, is switch on alanine transporters — pumps that pull it in from outside. And when you feed them labelled alanine and follow the atoms, they do not even burn it. They use it to build protein.

	Tumor cells	T cells
Alanine aminotransferase	High; GPT2, the mitochondrial form, rises with tumor grade	Barely present
Can convert pyruvate → alanine	Yes, in either direction as conditions favor	No
Can import alanine	Yes	Yes, via transporters induced on activation
Role in the exchange	Sets the direction and the supply	Takes whatever is available

The tumor is the price-setter. The T cell is a price-taker.

That asymmetry is not a detail. The same molecule arrives as a strategic asset for one cell type and as a dependency for the other — and only one of them controls the tap. We already know which way the local default points: in pancreatic tumors the stroma is running an alanine delivery service, and it is aimed at the cancer.

So the immune cell in that space is competing for a nutrient it cannot make, against a cell that can make it, in an environment that cell has already organized. And it is doing so while the free ammonia the tumor did release is independently pushing it toward exhaustion.

            Both halves of the nitrogen flux work against the immune system, in different ways. The fraction released as ammonia attacks T cells directly. The fraction exported as alanine denies them a nutrient they cannot synthesize and hands the tumor carbon, glucose and α-ketoglutarate. There is no setting of this dial at which the T cell comes out ahead.
        

Part IV / The Hand on the Throttle

Who Actually Sets the Rate

Everything so far has described the loop as if the tumor were running it. For the tumor's half, that is fair. For the liver's half, it is not — and this is the part of the picture that almost never gets drawn.

The liver does not perform gluconeogenesis because alanine has arrived. It performs gluconeogenesis because glucagon and cortisol have told it to. Alanine is the substrate; the hormones are the instruction. Glucocorticoids induce the two rate-limiting enzymes of the pathway, which is the step the zebrafish work measured as elevated in tumor-bearing animals. Deliver all the alanine you like to a liver that has not been told to make glucose, and comparatively little happens.

Cortisol does a second thing at the other end of the loop. It mobilizes muscle protein, releasing amino acids into circulation — including the branched-chain amino acids that feed the transamination in the first place. This is one of the central mechanisms by which cancer wasting proceeds.

So a rise in cortisol feeds the arrangement at both ends simultaneously: more substrate arriving at the tumor, and a liver under firmer orders to process what comes back. The tumor does not have to drive this. The host's endocrine state does it.

And there is a third effect, which is where this section connects to everything above. Glucocorticoids suppress T cells directly — they block activation, recruit regulatory T cells, and push macrophages toward the unhelpful phenotype. In a review of 640 lung cancer patients across two centres, those taking the equivalent of at least 10 mg of prednisone daily when checkpoint immunotherapy began had worse response rates and shorter survival, with a hazard ratio of 1.7 for overall survival after adjustment for performance status, smoking history and brain metastases. That is observational, and worth reading with care: steroids were prescribed most often for breathlessness, fatigue and brain metastases, all of which carry their own prognosis, so some residual confounding is likely. The study also looked only at steroids already in use when treatment started; whether steroids begun later, to manage immune side effects, carry the same penalty is a separate question. On this mechanism, the vulnerable moment is the one where T cells are first being activated.

            Three mechanisms, one hormone, all pointing the same way. Cortisol supplies the tumor's nitrogen load from muscle, instructs the liver to complete the carbon half of the loop, and disables the immune response to the result. Nothing else in this article does all three.
        

The Fasting Problem

            A note before this section. What follows is an argument about one pathway, not a recommendation. A single mechanism, however tidy, is a poor basis for decisions about eating during cancer treatment — fasting and protein intake both have large literatures pointing in several directions at once, and the people who should weigh those against an individual situation are that person's oncology team. Nothing here is a reason to change what anyone is eating.
        

This has an uncomfortable implication for the intervention people reach for first.

Prolonged fasting raises glucagon and cortisol. That is not a side effect; it is the mechanism by which fasting maintains blood glucose. But those are precisely the two signals that activate hepatic gluconeogenesis and mobilize muscle protein — which is to say, on this particular pathway, fasting pushes in the same direction the tumor does.

Whether that is outweighed by fasting's other effects — on autophagy, on IGF-1 signalling, on sensitizing tumors to chemotherapy — is a separate question with its own substantial literature, and I am not arguing it is not. But the specific claim that fasting starves a tumor of glucose does not survive this loop intact. The liver has a workaround, and the tumor is helping supply the raw material for it.

The protein question splits along the same seam. This loop does not consume protein. It consumes leucine, isoleucine and valine, because those are what the branched-chain transaminase acts on. That makes composition the variable worth attending to rather than quantity — and an isolated BCAA supplement is the limiting case, a concentrated dose of exactly the three amino acids the tumor is equipped to strip. Whey sits at the high end among ordinary foods and legume proteins toward the low end; the per-food figures are in USDA FoodData Central for anyone wanting to compare specific ones. And since net muscle breakdown probably matters more than what arrives on the plate, cutting protein hard enough to lose muscle may release more substrate than it withholds.

Part V / Consequences

What This Means for Measurement

If most of a tumor's surplus nitrogen leaves as alanine rather than ammonia, then ammonia concentration is a poor proxy for nitrogen flux. The two can move independently — and a tumor that is good at transamination can run an enormous nitrogen turnover while keeping free ammonia at a level that looks almost unremarkable.

That is worth stating plainly, because it cuts against an intuition the field has been building on. A low or moderate tumor ammonia reading does not mean the nitrogen economy is quiet. It may mean the opposite: that the valve is working.

            The measurement that would settle it is not difficult. Label the nitrogen in glutamine or the branched-chain amino acids, feed it to a tumor, and count how it leaves — as ammonia, as alanine, as glutamine, as anything else. Then compare that split across tumor types, and against how well each responds to immunotherapy. Nobody appears to have published it.
        

A Warning About Lactate Oxidase

There is a class of therapy where this matters immediately.

Tumor lactate suppresses immune function, so a reasonable idea is to destroy it — and several groups are now delivering a bacterial enzyme called lactate oxidase into tumors, wrapped in nanocapsules, carried by engineered gut bacteria, or packaged into the tiny vesicles those bacteria shed. The published results are encouraging: lactate falls, T cells move in, checkpoint immunotherapy works better.

But look at what the enzyme actually does. Lactate oxidase converts lactate into pyruvate. Which is to say: a therapy designed to disarm the tumor's immunosuppression manufactures, inside the tumor, the exact substrate the transaminase needs — the acceptor for nitrogen disposal, the carbon for the liver loop, the input to the α-ketoglutarate switch.

lactate + O₂ → pyruvate + H₂O₂ → then what?
→ alanine + α-KG | → back to lactate | → acetyl-CoA | → acetate

Nobody has tested this. The pyruvate left behind is almost never measured in these studies; alanine, as far as I can find, never. And the pro-tumor effects of surrounding pyruvate that have been demonstrated are mostly about preparing the ground where metastases take root, rather than about the growth of the original tumor — so experiments designed as these are, reporting the size of the primary tumor, would not detect the problem even if it were happening.

The Orchestrator That Never Shows Its Face

I am not concluding ammonia is secretly in charge, though it's looking pretty suspicious. Every estimate of how much ammonia matters in cancer has been made with an instrument that cannot see most of the traffic — one that reads a well-run nitrogen economy as a quiet one. We do not know how large ammonia's role is. We know that the way we have been looking could only ever have made it look small.

A tumor that handles nitrogen well keeps the free ammonia in a band: high enough to exhaust T cells, low enough not to damage itself. Everything above that band goes out bound to pyruvate, as an amino acid so ordinary it does not register as a finding. The nitrogen does its work and leaves without a signature.

And a tumor that handles it badly is not simply a tumor in trouble. The ammonia it fails to export drains the molecule its oxygen sensor depends on, strands it on glycolysis, and hands it back the pyruvate it needed all along. The failure mode and the correction are the same event.

Which is my main takeaway. We have been treating ammonia as the thing that happens when the system breaks down. It may be better understood as one of the things holding the system in place.

Oncometabolites: A New Candidate

Ammonia in Disguise

The Drug That Should Have Worked

The Amino Acid You Can Actually Move

The Enzyme the Tumour Threw Away

The Oncometabolite That Needs No Mutation

Research Blog - Terms of Use

Synergies for Cancer Treatment

Pages

Pages

Sunday, October 4, 2026