CMS-121

CMS121 — a synthetic derivative of the flavonol fisetin from the Salk Institute (Schubert and Maher labs); described in its own literature both as an acetyl-CoA carboxylase 1 (ACC1) inhibitor and as a fatty acid synthase (FASN) inhibitor

CMS-121 (CMS121 — a synthetic derivative of the flavonol fisetin from the Salk Institute (Schubert and Maher labs); described in its own literature both as an acetyl-CoA carboxylase 1 (ACC1) inhibitor and as a fatty acid synthase (FASN) inhibitor) is a cognitive & mood research compound. A fisetin derivative that acts on the first committed steps of de novo fat synthesis inside the cell. Its own literature names two adjacent enzymes: the Salk group reported that CMS121 raises acetyl-CoA by inhibiting acetyl-CoA carboxylase 1 (ACC1), the enzyme that carboxylates acetyl-CoA to malonyl-CoA, and separately identified fatty acid synthase (FASN) — the enzyme that condenses malonyl-CoA with acetyl-CoA into palmitate — as a molecular target. Either way the pathway consequence runs the same direction: less newly made fatty acid, and less polyunsaturated substrate available for the iron-dependent lipid peroxidation that drives oxytosis/ferroptosis. Downstream the group reports reduced lipid peroxidation, reduced inflammation and preserved mitochondrial homeostasis, plus increased histone H3K9 acetylation in rapidly aging mice, which is the acetyl-CoA pool showing up as an epigenetic mark. In diabetic and in normally aging mice the same mechanism reads out peripherally as lower triglycerides, lower free fatty acids and a lower liver inflammatory profile.

How CMS-121 works

A fisetin derivative that acts on the first committed steps of de novo fat synthesis inside the cell. Its own literature names two adjacent enzymes: the Salk group reported that CMS121 raises acetyl-CoA by inhibiting acetyl-CoA carboxylase 1 (ACC1), the enzyme that carboxylates acetyl-CoA to malonyl-CoA, and separately identified fatty acid synthase (FASN) — the enzyme that condenses malonyl-CoA with acetyl-CoA into palmitate — as a molecular target. Either way the pathway consequence runs the same direction: less newly made fatty acid, and less polyunsaturated substrate available for the iron-dependent lipid peroxidation that drives oxytosis/ferroptosis. Downstream the group reports reduced lipid peroxidation, reduced inflammation and preserved mitochondrial homeostasis, plus increased histone H3K9 acetylation in rapidly aging mice, which is the acetyl-CoA pool showing up as an epigenetic mark. In diabetic and in normally aging mice the same mechanism reads out peripherally as lower triglycerides, lower free fatty acids and a lower liver inflammatory profile.

Proposed benefits

Researched as a geroneuroprotector — a small molecule studied for age-related cognitive decline, brain lipid peroxidation and the metabolic side of aging.

Where to get CMS-121

The evidence for CMS-121

Graded by what exists behind each claim.

✅ Clinically validated

• A phase 1 exists and it is complete. 100 healthy volunteers, four parts: single oral doses of 50mg to 1800mg, repeated doses of 150 to 900mg once daily for 7 days, a dedicated elderly cohort at 600mg once daily for 7 days, and a food-effect cohort at 600mg. 149 treatment-emergent adverse events were reported, 138 of them mild and 11 moderate; no deaths and no serious adverse events. The events more common on drug than placebo were gastrointestinal symptoms and headache.
• Cardiodynamics were measured and were clean. Mean QTcF intervals stayed below 450 msec, changes from baseline stayed below 10 msec, and there were no individual outliers with an absolute QTcF above 450 msec or a change of 30 msec or more. There were no clinically significant laboratory findings.
• What the phase 1 did not do is test whether it works. It was a safety, tolerability, pharmacokinetics and cardiodynamics study in healthy people. There is no efficacy trial of CMS121 in Alzheimer's disease, in age-related cognitive decline, in metabolic disease or in anything else — not a failed one, a nonexistent one. Note also that the results are currently published as a medRxiv preprint rather than in a peer-reviewed journal, which is a real distinction and not a technicality.
• The pharmacokinetics carry one finding that matters more than the safety numbers: exposure was higher and the terminal half-life considerably longer in elderly subjects than in young ones, and a high-fat meal raised exposure by about 50%. Age and breakfast both change the dose.

📊 Correlative data

• There is no human observational or association data for CMS121 itself. What exists is a class signal from a different drug, and it points the opposite way to the mouse data: in people with steatohepatitis, pharmacological acetyl-CoA carboxylase inhibition with firsocostat produced hypertriglyceridemia significant enough that a 66-patient trial was run specifically to mitigate it, comparing icosapent ethyl against fenofibrate. Triglycerides rose on the ACC inhibitor and fenofibrate held them roughly flat while icosapent ethyl did not. CMS121 is not firsocostat and its own phase 1 reported no clinically significant laboratory findings over 7 days — but if ACC1 is genuinely the target, triglycerides are the marker to watch and the human direction of travel for the enzyme class is up, not down.
• The parent compound carries the adjacent literature. Fisetin is an orally active flavonol with published neuroprotective and cognition-enhancing activity in several animal models, and the medicinal chemistry program that produced CMS121 was motivated by fisetin's specific liabilities — a relatively high EC50 in cell assays, low lipophilicity, high topological polar surface area and poor bioavailability. CMS121 was selected from 160 synthetic fisetin derivatives.
• The ALS record belongs to the parent, not to this molecule. Fisetin itself has been reported to exert antioxidant and neuroprotective effects in multiple mutant hSOD1 models of amyotrophic lateral sclerosis — in Drosophila, in NSC34 cells and in transgenic mice — by activating ERK. A PubTator3 search on 29 September 2026 for CMS121 together with amyotrophic lateral sclerosis returned no study of CMS121 in any ALS model. If you came here for the ALS claim, it is fisetin's.

🧪 Theoretical / extrapolated

• The mechanism has two published names and they are one enzyme apart. The Salk group reported that CMS121 and J147 increased acetyl-CoA levels in cell culture and in mice via inhibition of acetyl-CoA carboxylase 1, producing neuroprotection and increased acetylation of histone H3K9 in SAMP8 mice at a site linked to memory enhancement. A year later the same group identified fatty acid synthase as a molecular target of CMS121 and showed that inhibiting FASN protects against Alzheimer-related toxicities. ACC1 makes malonyl-CoA out of acetyl-CoA; FASN consumes malonyl-CoA and acetyl-CoA to make palmitate. Both are de novo lipogenesis and both predict less newly made fatty acid, but they predict different things about which metabolite piles up, and no published experiment resolves which enzyme is engaged at the exposures a person actually reaches.
• The cell-death process being targeted is named and it is not apoptosis. Oxytosis/ferroptosis is driven by lipid peroxidation, and the argument for a lipogenesis inhibitor in the brain is substrate control: make less polyunsaturated fatty acid, have less of it to peroxidize. In transgenic AD mice CMS121 reduced cognitive loss while reducing brain lipid peroxidation and inflammation and modulating lipid metabolism.
• The mouse record is broad and the framing is aging, not one disease. In rapidly aging SAMP8 mice, CMS121 and J147 reduced cognitive decline and metabolic and transcriptional markers of aging while preserving mitochondrial homeostasis; fed only in the last quarter of the lifespan, both compounds improved physiological markers in brain AND kidney, with distinct tissue-specific protein changes in each. In SAMP8 mice treated to 13 weeks of age, CMS121 gave significantly better auditory brainstem response thresholds at 12 and 16 kHz and preserved more inner-hair-cell ribbon synapses.
• The peripheral metabolic data is the most quantified part of the file. In db/db leptin-receptor-deficient mice fed CMS121 for 6 months: a 5% reduction in body weight, improved glucose tolerance, lower HbA1c and insulin, lower blood and liver triglycerides and free fatty acids, lower liver NF-kappaB, IL-18, caspase 3 and C-reactive protein, and lower urinary NGAL, clusterin and albumin, with renal NDUFB8, UQCRC2 and VDAC restored. In normally aging wild-type mice fed it for 6 months: a 40% decrease in body weight GAIN — gain, not weight — improved glucose and lipid indexes, lower hepatic caspase 1, caspase 3 and NOX4, higher adipose Nrf1, TFAM, GLUT4 and electron-transport markers, a higher resting metabolic rate, and raised plasma short-chain acylcarnitines and butyrate metabolites. The authors describe the profile as mimicking the effects of ketogenic diets.
• One model result is explicitly partial and the title says so. In R6/2 and YAC128 Huntington's mice, CMS121 slowed motor dysfunction and increased median lifespan, with the YAC128 motor benefit becoming more pronounced as the mice aged, and reduced disease-related changes in proteasome and oxidative-phosphorylation gene expression. The paper is titled *partially* attenuates, which is the correct word and worth keeping.

How to read these tiers: they say how much human evidence exists, not how well something works — and ✗ flags harm, never a disappointing trial. How the evidence tiers work →

What CMS-121 actually does

This is a flavonol that was taken to a medicinal chemist, and the target it ended up hitting is a fat-synthesis enzyme, not an antioxidant pathway. The starting material was fisetin, an orally active flavonol with neuroprotective and cognition-enhancing activity in animal models. Chiruta 2012 states the problem with using it as a drug in plain terms — a relatively high EC₅₀ in cell-based assays, low lipophilicity, high topological polar surface area and poor bioavailability — and describes the multitiered screen built to fix it. CMS121 was selected from 160 synthetic fisetin derivatives Ates 2023. So the compound in the capsule is not a plant molecule and should not be argued for like one.

The mechanism has two published names, and they are one enzyme apart in the same pathway. Currais 2019 reports that CMS121 and its sibling J147 increased the levels of acetyl-CoA in cell culture and in mice via the inhibition of acetyl-CoA carboxylase 1 (ACC1), with neuroprotection and increased acetylation of histone H3K9 in SAMP8 mice, a site the authors link to memory enhancement. One year later Ates 2020 identified fatty acid synthase (FASN) as a molecular target of the same compound. These are consecutive steps: ACC1 carboxylates acetyl-CoA to malonyl-CoA; FASN then condenses malonyl-CoA with acetyl-CoA to build palmitate. Inhibit either and de novo lipogenesis falls.

The distinction is not academic, because the two predict different metabolites. Block ACC1 and acetyl-CoA accumulates while malonyl-CoA falls — which is exactly the pattern Dafre 2024 reports, with raised plasma short-chain acylcarnitines and butyrate metabolites and diminished de novo lipogenesis in mice. Block FASN and malonyl-CoA should accumulate instead, and malonyl-CoA is itself a signal: it inhibits carnitine palmitoyltransferase-1 and therefore restrains fatty-acid entry into mitochondria. No published experiment separates the two at the exposures a person reaches. A single enzyme-activity curve for CMS121 against ACC1 and against FASN, run side by side, would settle it, and the literature does not contain one.

Why a lipogenesis brake is a neuroprotection strategy at all. Ates 2020 frames it through oxytosis/ferroptosis — a cell-death process characterized by increased lipid peroxidation, distinct from apoptosis and iron-dependent. The logic is substrate control rather than radical scavenging: make less polyunsaturated fatty acid and there is less of it available to peroxidize. That is a genuinely different argument from fisetin's own, which was direct antioxidant activity plus raising intracellular glutathione, and it is the reason this page does not inherit fisetin's antioxidant framing.

The same mechanism has a peripheral face, and it is the more measurable one. A brake on de novo lipogenesis in the liver reads out as triglycerides, free fatty acids and hepatic inflammatory proteins, and that is what Zahid 2023 found in db/db mice: blood and liver triglycerides and free fatty acids decreased, with liver NF-κB, IL-18, caspase 3 and C-reactive protein all lower. A compound whose brain claim is invisible outside a laboratory has a metabolic claim anybody can draw in a tube, which is the single most useful fact on this page.

Cell, rodent, human — and where it stops

The chain here is unusually complete on one axis and completely empty on the other. Cells and mice, then a phase 1 in people, then nothing — because the phase 1 measured safety and drug levels, not whether anything got better.

Cells. Ates 2020 identified FASN as the molecular target and showed FASN inhibition protecting against Alzheimer-related toxicities in culture, with the mechanism framed as reduced lipid peroxidation in an oxytosis/ferroptosis model. Currais 2019 demonstrated the acetyl-CoA rise in cell culture before showing it in mice. Culture dishes: no animal, no person.

Mouse, model one — the disease model. In transgenic Alzheimer's mice, CMS121 reduced cognitive loss while reducing brain lipid peroxidation and inflammation and modulating lipid metabolism Ates 2020.

Mouse, model two — the aging model, which is the more interesting design. SAMP8 mice age rapidly. Currais 2019 reported CMS121 and J147 reducing cognitive decline and metabolic and transcriptional markers of aging while preserving mitochondrial homeostasis. Kepchia 2021 then fed the compounds only in the last quadrant of the lifespan — when the mice already had cognitive deficits and were beginning to develop chronic kidney disease — and both compounds improved physiological markers in brain and kidney at once, with distinct tissue-specific protein changes in each organ. Treating late rather than from youth is the design that matters for a human reader, and it is rarer than it should be.

Mouse, model three — the metabolic models, and the most quantified data on the page. db/db leptin-receptor-deficient mice on a CMS121 diet for 6 months: 5% lower body weight, improved glucose tolerance, lower HbA1c and insulin, lower blood and liver triglycerides and free fatty acids, lower liver NF-κB, IL-18, caspase 3 and CRP, lower urinary NGAL, clusterin and albumin, and renal NDUFB8, UQCRC2 and VDAC restored Zahid 2023. Normally aging wild-type mice, same 6-month diet: a 40% decrease in body weight gain, improved glucose and lipid indexes, lower hepatic caspase 1, caspase 3 and NOX4, higher adipose Nrf1, TFAM, GLUT4 and electron-transport-chain markers, and a higher resting metabolic rate Dafre 2024.

Mouse, models four and five — where the result is partial, and the page keeps the word. In R6/2 and YAC128 Huntington's mice, CMS121 slowed motor dysfunction and increased median lifespan, with the YAC128 benefit growing as the mice aged; the paper is titled partially attenuates Ates 2023. In SAMP8 mice treated to 13 weeks, ABR thresholds were significantly better at 12 kHz and 16 kHz and more inner-hair-cell ribbon synapses were preserved Pham 2023 — a specific frequency band, not hearing in general.

Humans, and this is where the chain stops. Maher 2025 dosed 100 healthy volunteers for up to 7 days and reported tolerability, kinetics and cardiodynamics. No efficacy endpoint has ever been measured in a person — not a cognitive scale, not a biomarker of Alzheimer's pathology, not HbA1c. The specific obstacle to transfer is duration and design: every mouse benefit came from months of continuous dietary exposure in an animal with an engineered disease or an accelerated-aging genotype, while the human record is seven days in healthy people. Nothing about the phase 1 makes the mouse results more likely to be true in a person; it makes them safe to go and test.

One correction to the story as it is usually told. The ALS literature attached to this family belongs to fisetin, not to CMS121: Wang 2018 reports antioxidant and neuroprotective effects in multiple mutant hSOD1 models — Drosophila, NSC34 cells and transgenic mice — via ERK activation, using the parent flavonol. A PubTator3 search on 29 September 2026 for CMS121 with amyotrophic lateral sclerosis returned no study of CMS121 in any ALS model.

CMS-121 pharmacokinetics — how much of it actually gets in

The pharmacokinetics are the strongest part of this file, which is unusual for anything in this catalog. Maher 2025 is a 100-subject phase 1 (NCT05318040) in four parts — 48 single-ascending-dose, 32 multiple-ascending-dose, 8 elderly, 12 food-effect. Single oral doses ran 50 mg to 1800 mg; repeated dosing ran 150, 300, 600 and 900 mg once daily for 7 days; the elderly cohort took 600 mg once daily for 7 days.

What was measured. T_max was about 1.5 hours after a single dose and about 2–3 hours on repeated dosing. Exposure was dose-proportional or slightly more than dose-proportional. The predominant circulating species was the metabolite CMS121-C2, and urinary excretion was minimal — the paper puts it below 0.411% — which says the clearance route is metabolic, not renal, and that a urine assay is the wrong place to look for it.

The half-life is not published as a number, and here is the arithmetic that bounds it anyway. The paper reports only that the terminal elimination half-life of CMS121 and its metabolites was considerably longer in elderly subjects. But it also reports that seven days of once-daily dosing produced no appreciable accumulation in total exposure, while peak concentrations did accumulate at the lower doses. An accumulation ratio near 1 on a 24-hour dosing interval means the drug is essentially gone before the next dose: four to five half-lives inside 24 hours puts the effective terminal half-life in young adults at roughly 5 hours or less. That is a bound derived here from the accumulation statement, not a measured value — and it is explicitly the bound that fails in the elderly, where the same paper says exposure is higher and elimination slower. The age group most likely to buy this is the age group the published kinetics say gets a bigger dose from the same capsule.

Food is a dose decision. At 600 mg, a high-fat high-calorie meal raised systemic exposure by approximately 50% and delayed T_max to about 4 hours. A 50% swing is larger than the gap between two adjacent dose levels in the trial's own design, so taking it with breakfast on some days and fasted on others is a self-inflicted dose-ranging study.

What the mouse work cannot tell you about exposure. Every efficacy study cited on this page delivered CMS121 in the diet — 6 months in db/db mice Zahid 2023, 6 months in aging wild-type mice Dafre 2024, the last quadrant of the lifespan in SAMP8 mice Kepchia 2021. Diet delivery means near-continuous low-level exposure with no peak, which is pharmacokinetically the opposite of one capsule a day. The abstracts read here do not state mg/kg, so no body-surface-area conversion is offered; and none is needed, because human dose levels exist. Running the comparison the other way for scale: 900 mg at 70 kg is 12.9 mg/kg/day, which ×12.3 corresponds to about 158 mg/kg/day in a mouse — a large dose by any rodent standard, quoted as a sanity check on the human numbers rather than as a claim about what the animals received.

What would have to be true, and how you would know it was not

Four predictions. The first two are cheap and would be the first honest evidence in a human being; the third cuts against the product and comes from the enzyme class's own human record; the fourth is about who is taking it.

1. If the lipogenesis brake is engaged at the dose taken, the lipid panel moves before anything cognitive does. Triglycerides and free fatty acids fell in both mouse models Zahid 2023 Dafre 2024. In a person that is a fasting lipid panel with triglycerides at baseline and 6 weeks, with free fatty acids if the lab offers them. This is the whole ballgame for target engagement: a compound whose published mechanism is de novo fatty-acid synthesis should be detectable in the cheapest test in the building. If nothing moves at all, the honest conclusion is that the capsule count is too low — one 25 mg capsule is a sixth of the lowest dose ever dosed repeatedly in a person.

2. The glycemic prediction, and it should be slow and only in the dysglycemic. db/db mice showed lower HbA1c and insulin with improved glucose tolerance Zahid 2023; aging wild-type mice showed improved glucose indexes and a higher resting metabolic rate Dafre 2024. The order of movement in a person would be fasting insulin, then HbA1c, and the retest windows are not optional: HbA1c integrates roughly 90 days, so a value at 6 weeks is noise. In someone already normoglycemic the prediction is nothing, because there is no defect for the mechanism to correct.

3. THE PREDICTION THAT CUTS AGAINST THE PRODUCT, and it comes from the enzyme rather than from a complaint. If ACC1 is genuinely the target Currais 2019, then the human record for that enzyme class predicts triglycerides going UP, not down. Lawitz 2023 exists because pharmacological ACC inhibition with firsocostat produced hypertriglyceridemia in steatohepatitis patients serious enough to require a dedicated 66-patient mitigation trial; in that study fenofibrate held triglycerides roughly flat (−2 mg/dL at six weeks) while icosapent ethyl did not (+41 mg/dL). CMS121 is not firsocostat, the potencies and tissue distributions differ, and CMS121's own 7-day phase 1 reported no clinically significant laboratory findings Maher 2025. But seven days is not six weeks, and this is the one marker where the mouse data and the human class data point in opposite directions. Draw triglycerides at baseline and 6–12 weeks. A rise is the mechanism working in the direction nobody advertises.

4. The exposure prediction, which is about the reader rather than the molecule. The phase 1 found higher exposures and considerably longer terminal half-lives in elderly subjects Maher 2025. The prediction follows arithmetically: an older person taking the dose that was tolerated by 20-somethings is running a higher exposure, and the tolerability read from the young cohorts does not transfer. Nothing in the published record tells you the size of that gap, because no number was given. The measurable consequence to watch is the adverse-event profile the trial actually reported — gastrointestinal symptoms and headache — appearing at doses that were uneventful in the trial's young arms.

What nobody has tested yet

Five experiments nobody has run, in the order that would change the page fastest.

1. Any efficacy endpoint, in any person. The phase 1 established that 100 healthy adults could take up to 1800 mg once or 900 mg daily for a week without incident Maher 2025. Nobody has followed it with a cognitive scale, an amyloid or tau measure, an HbA1c or a liver-fat measurement. The compound is therefore in the rarest position in this catalog — a human dose that is known, attached to a human benefit that has never been looked for.

2. Which enzyme it actually inhibits, at human concentrations. One paper says ACC1 Currais 2019; the next says FASN Ates 2020; the hearing paper's own title calls it a fatty acid synthase inhibitor Pham 2023. A single side-by-side enzyme assay across the concentration range the phase 1 shows is reachable would resolve it, and it matters because the two predict opposite movements in malonyl-CoA — which in turn sets whether fatty-acid oxidation is relieved or restrained.

3. A repeat-dose human study longer than a week. Seven days is the entire human exposure record. Every mouse result came from 6 months of dietary dosing Zahid 2023 Dafre 2024, and the one lipid-related liability suggested by the enzyme class showed up in humans over weeks Lawitz 2023. A 12-week study with a lipid panel and a CMP would cost very little and would be the first data that means anything for daily use.

4. Whether the aging benefit survives in an animal that is not genetically accelerated. SAMP8 is a senescence-accelerated strain and db/db mice lack a leptin receptor; the closest thing to a normal animal in this file is the wild-type aging study, which measured body weight gain, glucose and lipid indexes and tissue proteins rather than lifespan Dafre 2024. No lifespan study in normal mice has been published for CMS121. The Huntington's models gave an increase in median lifespan Ates 2023 — in mice engineered to die early of one disease, which is a different claim.

5. The experiment the dose format demands. Nobody has published anything about CMS121 below 50 mg, and the vendor's unit is 25 mg. A dose-response with a target-engagement readout — triglycerides, free fatty acids, or a plasma acylcarnitine panel of the kind Dafre 2024 used in mice — down through 25 and 50 mg would tell a reader whether one or two capsules does anything at all. As it stands, the bottom of the studied range is six capsules and everything below that is unexplored rather than ineffective.

CMS-121 — its own safety story, not its class's

Start with what is unusual here: this compound has a real human safety read, and it is reassuring as far as it goes. Across 100 volunteers, Maher 2025 reported 149 treatment-emergent adverse events, 138 of them mild and 11 moderate, with no deaths and no serious adverse events. The events more common on drug than on placebo were gastrointestinal symptoms and headache. Cardiodynamics were measured properly: mean QTcF stayed below 450 msec, changes from baseline below 10 msec, with no individual outlier above 450 msec or changing by 30 msec or more. There were no clinically significant laboratory findings and no C-SSRS abnormalities.

Now the boundaries of that read, because they are narrow. Seven days maximum. Healthy volunteers. Results in a medRxiv preprint, not a peer-reviewed paper. And one finding inside it is a caution rather than a reassurance: elderly subjects had higher exposures and considerably longer elimination half-lives, so the tolerability of 600 mg in a 25-year-old is not the tolerability of 600 mg in a 70-year-old, and the size of that difference was not published as a number.

The specific open risk is a lipid risk, and it comes from the target rather than from the compound. If ACC1 is the enzyme Currais 2019, the human precedent for inhibiting it is hypertriglyceridemia: firsocostat raised triglycerides in steatohepatitis patients enough that Lawitz 2023 was run to manage it, with fenofibrate succeeding and icosapent ethyl not. The mouse data for CMS121 runs the other way — triglycerides and free fatty acids fell in db/db and in aging wild-type mice Zahid 2023 Dafre 2024 — and CMS121's own week-long human exposure showed no lab signal. Mitigation is one blood draw: a fasting lipid panel at baseline and again at 6–12 weeks. If triglycerides climb, that is the class effect appearing, and it is manageable and well characterized in the literature above.

A second mechanism-derived caution, and it is about who should not take it. De novo lipogenesis is not a vestigial pathway. It is how the body builds fatty acids when dietary supply is insufficient, and it is most active in the liver, in lactating mammary tissue, and in the developing fetus and neonate. A drug whose entire purpose is to restrain that pathway has no business being taken in pregnancy or while breastfeeding, and no study has looked. Equally, the phase 1 enrolled healthy volunteers only, so anyone with existing liver disease is outside the studied population — which is exactly the population in whom the ACC-inhibitor lipid effect was documented.

The kidney is the organ this compound's own animal data says to watch, and the direction is favorable. Urinary NGAL, clusterin and albumin all fell in db/db mice Zahid 2023, and SAMP8 mice beginning to develop chronic kidney disease improved on kidney markers as well as brain markers Kepchia 2021. That makes a CMP with eGFR and a urine albumin-to-creatinine ratio the right baseline pair — not because harm is predicted, but because it is the one place the animal literature makes a falsifiable claim a reader can actually test.

Finally, the honest framing of the whole risk picture. The likeliest harm from this product is not toxicity. It is paying for an exposure nobody has studied: the lowest repeated human dose is six capsules, the bottle is 60 capsules and 1,500 mg total, and at the top studied daily dose the whole bottle is under two days. A reader taking one capsule a day is not running a low dose of a studied drug; they are running an unstudied dose of a drug with a well-characterized studied range printed one paragraph above.

Sources read for this page

• Chiruta C, Schubert D, Dargusch R, Maher P. Chemical modification of the multitarget neuroprotective compound fisetin. Journal of Medicinal Chemistry 2012;55(1):378-89 · PMID 22192055
• Currais A, Huang L, Goldberg J, Petrascheck M, Ates G, Pinto-Duarte A, Shokhirev MN, Schubert D, Maher P. Elevating acetyl-CoA levels reduces aspects of brain aging. eLife 2019;8:e47866 · PMID 31742554
• Ates G, Goldberg J, Currais A, Maher P. CMS121, a fatty acid synthase inhibitor, protects against excess lipid peroxidation and inflammation and alleviates cognitive loss in a transgenic mouse model of Alzheimer's disease. Redox Biology 2020;36:101648 · PMID 32863221
• Kepchia D, Currais A, Dargusch R, Finley K, Schubert D, Maher P. Geroprotective effects of Alzheimer's disease drug candidates. Aging (Albany NY) 2021;13(3):3269-3289 · PMID 33550278
• Zahid S, Dafre AL, Currais A, Yu J, Schubert D, Maher P. The Geroprotective Drug Candidate CMS121 Alleviates Diabetes, Liver Inflammation, and Renal Damage in db/db Leptin Receptor Deficient Mice. International Journal of Molecular Sciences 2023;24(7):6828 · PMID 37047807
• Pham TB, Boussaty EC, Currais A, Maher P, Schubert DR, Manor U, Friedman RA. Attenuation of Age-Related Hearing Impairment in Senescence-Accelerated Mouse Prone 8 (SAMP8) Mice Treated with Fatty Acid Synthase Inhibitor CMS121. Journal of Molecular Neuroscience 2023;73(4-5):307 · PMID 37097512
• Ates G, Taguchi T, Maher P. CMS121 Partially Attenuates Disease Progression in Mouse Models of Huntington's Disease. Molecular Neurobiology 2023 (online 21 Oct 2023); doi:10.1007/s12035-023-03711-2 · PMID 37864765
• Dafre AL, Zahid S, Probst JJ, Currais A, Yu J, Schubert D, Maher P. CMS121: a novel approach to mitigate aging-related obesity and metabolic dysfunction. Aging (Albany NY) 2024;16(6):4980 · PMID 38517358
• Maher P, Christopher R, Evans R, Raschke W. Safety, Pharmacokinetics, and Cardiodynamics of CMS121, a Novel Small Molecule Fisetin Derivative with Neuroprotective Properties, in Phase 1 Healthy Adult Volunteers. medRxiv 2025 (preprint, posted 5 Mar 2025); doi:10.1101/2025.02.28.25323123 · PMID 40093210
• Wang TH, Wang SY, Wang XD, Jiang HQ, Yang YQ, Wang Y, Cheng JL, Zhang CT, Liang WW, Feng HL. Fisetin Exerts Antioxidant and Neuroprotective Effects in Multiple Mutant hSOD1 Models of Amyotrophic Lateral Sclerosis by Activating ERK. Neuroscience 2018;379:152-166 · PMID 29559385
• Lawitz EJ, Bhandari BR, Ruane PJ, Kohli A, Harting E, Ding D, Chuang JC, Huss RS, Chung C, Myers RP, et al. Fenofibrate Mitigates Hypertriglyceridemia in Nonalcoholic Steatohepatitis Patients Treated With Cilofexor/Firsocostat. Clinical Gastroenterology and Hepatology 2023;21(1):143-152.e3 · PMID 34999207

CMS-121 — interference & stacking

Bloodwork to run alongside CMS-121

Baseline first, then again at 8–12 weeks.

The Brain Fog & Cognition panel covers these in one order — 12 markers, $233.06 with the discount applied.

Check results you already have → · All 103 markers A–Z

CMS-121 — frequently asked questions

What CMS-121 is used for

CMS-121 appears under 3 goals in the goal router.

Related Cognitive & Mood compounds

← Explore the full Protocol Vault

↑ Back to on this page