The Amino Acid You Can Actually Move

Methionine restriction is the most promoted metabolic intervention in cancer. A randomized trial in metastatic breast cancer has now tested whether a diet people can really follow delivers it. It did not — and the amino acids it did lower are the ones nobody was aiming at.

Nearly every cancer cell is methionine-addicted. The dependency is reproducible, the mechanism is understood, and a diet low in methionine is something a motivated person can attempt without a prescription. That combination has made methionine restriction the most widely promoted metabolic intervention in oncology outside the ketogenic diet.

So it is worth knowing what happened when somebody finally ran the experiment properly: an eight-week randomized controlled trial of an intensive, fully-provisioned plant-based diet in women with metastatic breast cancer, on active systemic therapy, with ninety-five percent compliance. The intervention worked. It lowered a great many things.

Methionine was not one of them.

A whole-food plant-based diet significantly reduced leucine, isoleucine, lysine, phenylalanine, threonine, alanine, glutamate, proline and tyrosine in these patients. The reduction in methionine did not reach significance. The diet marketed as a way to lower methionine did not measurably lower methionine — but it did lower the amino acids that carry nitrogen out of a tumor, which may be the more interesting result.

This follows on from Ammonia in Disguise and The Drug That Should Have Worked. The first argued that a tumor's surplus nitrogen leaves largely as alanine; the second, that interventions aimed at one metabolic currency tend to be absorbed by another. This is the third case, and the first where the intervention is something people are already doing to themselves.

Part I / Why Methionine Looked Like the Answer

A Dependency That Is Real and Strange

Methionine addiction is not a claim about cancer cells being unable to make methionine. They can. They synthesize normal or excess amounts and still require the exogenous supply — which is the part that makes it interesting rather than trivial.

The explanation that has held up is a flux problem. Cancer cells run transmethylation reactions so hard that regeneration from homocysteine cannot keep pace, and histone H3 lysine hypermethylation appears to be the major sink. When cells revert to methionine independence, those hypermethylated marks are reduced or lost, and so is their malignancy. The methyl economy, not the amino acid supply, is what's addicted.

The strongest therapeutic data came in 2019. Dietary methionine restriction combined with 5-fluorouracil shrank chemoresistant RAS-mutant colorectal cancer in patient-derived xenografts, and combined with radiation it extended sarcoma tumor tripling time by 52% — from 17.5 to 26.6 days. The mechanism was tumor-cell-autonomous flux through one-carbon metabolism, affecting redox and nucleotide handling.

            One detail from that paper decides how the rest of this article reads. In the sarcoma model, methionine restriction on its own did not alter tumor growth at all. It worked only alongside radiation. Restriction is a sensitizer, not a treatment — which means the version taken without concurrent therapy is not a weaker form of the tested intervention. It is a different thing, and in that model it did nothing.
        

Part II / The Axis Nobody States

Two Experiments, Opposite Answers

The immunological literature on methionine restriction contradicts itself, and the contradiction is usually reported as uncertainty. It is better than that: the studies disagree in a patterned way.

On one side, a 2023 study found that methionine restriction impairs antitumour immunity through the gut microbiota. Restriction cut faecal hydrogen sulfide production; transplanting faeces from restricted mice was by itself enough to reduce CD3⁺ and CD8⁺ T cells in recipients. Tumors progressed faster. Most damningly, anti-PD-1 suppressed tumor growth in control-diet animals but not in restricted ones — and restriction had no effect at all in germ-free mice, confirming the microbiota was carrying it.

On the other side, restriction begun five days after tumor implantation and run alongside checkpoint inhibitors was five times more effective at reducing tumor size than immunotherapy alone, in male mice. Surface PD-L1 rose, CD8⁺ T cells moved from the tumor periphery into the tumor, and the effect ran through cGAS-STING and interferon signalling — inhibiting STING blunted the rise in MHC-I. A 2025 preprint reports the same direction across seven syngeneic models.

Study	When restriction started	Result
Ji et al., Nature Metabolism 2023	3–4 weeks before tumor implantation	Immunity impaired; anti-PD-1 failed
Gao et al., Nature 2019	Concurrent with chemotherapy or radiation	Sensitized to both
Morehead et al., 2023	Day 5 after implantation, alongside checkpoint inhibitors	5× tumor reduction over immunotherapy alone
Preprint, September 2025	Concurrent with implantation	T cells primed; checkpoint blockade enhanced

Every benefit is short and concurrent with treatment. The single harm is weeks of restriction before anything else happens.

The 2025 authors name the difference themselves: their work began restriction at tumor injection, the contrary study began three to four weeks earlier, and they propose that duration and depth determine the therapeutic effect through distinct mechanisms. Chronic restriction has time to remodel a microbiome and draw down a T cell pool. Acute restriction alongside treatment does not.

            Which puts the popular version on the wrong side of the line. Nobody adopting a low-methionine diet does it for ten days around a chemotherapy cycle. They do it indefinitely, starting whenever they decide to. That is the protocol with harm data behind it, and it is the one most easily done without medical supervision.
        

Part III / What a Real Diet Did

The Trial That Should Have Settled It

All of the above is mouse work. The question that decides whether any of it reaches people is narrower and more boring: can a diet someone will actually eat lower their plasma methionine?

The trial that bears on this is unusually good. Women with stage 4 breast cancer on stable systemic therapy were randomized 2:1 to an eight-week whole-food plant-based diet or usual diet. The intervention arm received three prepared meals and a side dish every day, weekly assessment visits, and weekly physician calls. Thirty of thirty-two completed. Ninety-five percent met the dietary prescription.

It worked as a dietary intervention, comprehensively. Weight fell 6.6%. LDL cholesterol went from 104.6 to 82.2 mg/dL, fasting insulin from 16.8 to 11.2 uIU/mL, insulin resistance from 4.4 to 2.7. Sex hormone binding globulin rose. IGF-1 fell 10% within the intervention group. A companion analysis found reductions in IL-6, IL-8, TNF-α and leptin, along with CA15-3 and VEGF-C.

And it was safe in the population most people would worry about. Adverse events were infrequent and mild. The weight loss happened without portion or calorie restriction — subjects ate 16% more food by mass while taking in 26% fewer calories — and the authors record it as occurring without signs or symptoms of progressing disease or cachexia.

Then They Measured the Amino Acids

The amino acid data was published separately, and it is the part that matters here.

Fell significantly	Did not
Isoleucine, leucine, lysine, phenylalanine, threonine	Methionine, tryptophan
Alanine, glutamate, proline, tyrosine	Cysteine, serine, glycine · valine trended up

Nine of twenty amino acids fell significantly. The one the diet is reputed to target was not among them.

The authors' own summary of the problem is the useful line: reductions in dietary amino acids did not predict reductions in serum. Eating less of something is not the same as having less of it in your blood, and for methionine the gap was wide enough to swallow the entire effect.

One caveat, stated plainly because it cuts against the strength of this result: the amino acid analysis was within-group only, eighteen patients, with the control arm excluded. It is weaker evidence than the parent trial. But it is also the best evidence that exists for this question in this population, and it points one way.

            This retrospectively explains something odd about the preclinical literature. Every study showing benefit used an 80 to 86 percent reduction — 0.86% to 0.12% methionine by weight in mouse chow, or 17.2 down to 2.9 mg/kg/day in the one human feeding study, which used six healthy volunteers over three weeks. That always looked like researchers being heavy-handed. It may instead be what it takes to move the analyte at all. Methionine is defended by protein turnover and remethylation, and a change in which foods you eat is apparently not enough to overcome that.
        

So the argument for dietary methionine restriction has three steps, and they are not equally supported. That an achievable diet can be delivered to metastatic patients on therapy: now well demonstrated. That lowering plasma methionine sensitizes tumors to treatment: reasonable preclinical support. That the achievable diet lowers plasma methionine: the step this trial tested, and did not confirm.

Part IV / What It Moved Instead

Read That List Again

Leucine. Isoleucine. Alanine. Glutamate.

Those are not a random nine amino acids. They are, with one exception, the working parts of the nitrogen export pathway described in the first article of this series — the branched-chain amino acids that donate nitrogen to the transaminase, the glutamate that carries it, and the alanine that takes it out of the tumor and bills the liver for glucose on the way.

The intervention was aimed at the lower pathway and landed on the upper one.

Be careful with this, because it is the kind of pattern that is easy to over-read. Valine is a branched-chain amino acid and it trended upward, which the clean version of this story does not accommodate. Lysine, phenylalanine, threonine, proline and tyrosine fell too, and they are not part of the nitrogen export route. A plant-based diet lowers many amino acids for the ordinary reason that it contains less protein of lower density, and some of what fell will be that and nothing more.

What survives the caution is narrower but still worth stating: the two branched-chain amino acids that feed the transaminase, the glutamate that accepts their nitrogen, and the alanine that carries it out all fell significantly, while methionine did not. If you were designing a dietary intervention against the nitrogen economy rather than the methyl economy, this is roughly the result you would have hoped for — and nobody was designing for it.

            The trial is reported as a methionine study that didn't work. It may be better understood as a nitrogen-handling study that nobody has analyzed yet. The dataset exists; the amino acids were measured; the question of whether those reductions track with anything — progression markers, inflammatory markers, the IGF-1 fall — has not been asked.
        

Part V / What to Keep

Where This Leaves Methionine

Not nowhere. The dependency is real, the sensitization data are reproducible across several groups, and the immunological picture resolves sensibly once you sort the studies by duration. What has changed is which delivery route looks viable.

Short, deep restriction concurrent with treatment, in a selected tumor, remains a live research strategy. That is what the positive studies tested, and it is what the clinical protocols have used: formula-based or heavily engineered diets, run for weeks rather than months, around a specific therapy.

Food-based restriction as a way of reaching that state is the part this trial puts in question. Not because patients can't comply — they complied at ninety-five percent — but because compliance didn't produce the biochemical change the strategy depends on.

And the pharmacological route has already had its trial. The MAT2A inhibitor AG-270 lowered plasma SAM by 54 to 70 percent in genotype-selected patients with confirmed target engagement, and returned two partial responses in forty. Development was discontinued; a successor compound is in trials.

The Pattern, Third Time

Three articles, three currencies, one shape. Alanine: the tumor sets the supply and the T cell takes the price. Glutamine: blocking the enzyme starves the immune cell with no bypass before it starves the tumor with five. Methionine: tumors outcompete T cells for it through the SLC43A2 transporter, and the T cells lose SAM, lose H3K79me2, lose STAT5, and stop working.

Note what the SLC43A2 authors did about it. They did not restrict methionine further. They supplemented it, and separately blocked the tumor's transporter — and both restored the T cells. Take the tumor's advantage away rather than the nutrient away from everyone. That is the same conclusion the ammonia work reached by enhancing clearance instead of blocking production, and it keeps arriving from different directions.

            If there is a general lesson across the three, it is about selectivity. A metabolic intervention is not selective because it targets a pathway the tumor depends on. Every cell in the microenvironment uses these pathways. It is selective only if the tumor has fewer ways around the block than the cells you are trying to protect — and on current evidence that is almost never the case. The tumor is the one with the stroma, the scavenging, the isoform redundancy and the transporters. The T cell is the one with nothing.
        

What I Would Measure Next

The specific experiment this trial makes available: take the amino acid panel already collected, and test whether the fall in alanine, glutamate and the branched-chain amino acids tracks with the fall in CA15-3, VEGF-C, IL-6 or IGF-1. The data are in hand. Nobody has looked, because nobody was asking about nitrogen.

And the measurement that would settle the larger question is still the one the first article ended on, still unpublished: label the nitrogen, feed it to a tumor, and count how it leaves.

            What this is and is not. This is an argument about how to design the next study. It is not dietary advice, and nothing here is a reason for anyone to change what they are eating during cancer treatment. The trial discussed above was conducted with provided meals, weekly physician contact and continuous monitoring, in patients on stable systemic therapy — conditions that have no equivalent in self-administration. Decisions about nutrition during treatment belong with an oncology team that knows the individual case.
        

Research Blog - Terms of Use

Synergies for Cancer Treatment

Pages

Pages

Monday, October 5, 2026