Urolithin A

Urolithin A is a postbiotic: a molecule the body does not make and plants do not contain, produced instead by gut bacteria acting on food. The substrates are ellagitannins and ellagic acid, found in pomegranates, walnuts, strawberries, raspberries and oak-aged wine. When faecal samples from six volunteers were incubated anaerobically with ellagic acid, the pomegranate ellagitannin punicalagin, or a walnut extract, every culture produced urolithin A — but at very different rates and concentrations, which is why urinary excretion of urolithins after a pomegranate varies so much between people [1]. That variability is now treated as a trait: people are sorted into urolithin metabotypes according to which urolithins their microbiota can make, and the urolithins are far better absorbed than the ellagitannins they come from [7].

Chemically it is a small tricyclic lactone — a dibenzo[b,d]pyran-6-one with hydroxyls at the 3 and 8 positions — and it looks nothing like the large, sugar-laden tannins it comes from. That is the point: the bacteria strip the molecule down to something that can cross the gut wall.

The compound became interesting in 2016, when it was identified as the first natural compound in its class to induce mitophagy. It did so in cells and in living animals after oral dosing, prevented the age-related accumulation of dysfunctional mitochondria in C. elegans and extended the worms' lifespan, mobility and pharyngeal pumping, and improved exercise capacity in two mouse models of age-related muscle decline as well as in young rats [3]. Synthetic urolithin A has since been sold as Mitopure and taken through four placebo-controlled human trials.

What the human trials found. The pattern is consistent: pre-registered primary endpoints missed, secondary endpoints positive.

• First-in-human, 2019. Healthy sedentary elderly adults took a single dose or four weeks of daily urolithin A. The primary outcome was safety, and it was favourable. The compound was bioavailable in plasma at every dose, and at 500 mg and 1000 mg for four weeks it modulated plasma acylcarnitines and skeletal-muscle mitochondrial gene expression [8].
• Older adults, 4 months, 2022. Sixty-six adults aged 65 to 90 took 1000 mg a day or placebo. Neither primary endpoint — six-minute walk distance, and maximal ATP production in the hand muscle by magnetic resonance spectroscopy — improved significantly. Muscle endurance, a secondary endpoint, improved at two months in both the hand and the leg muscle tested, and plasma acylcarnitines, ceramides and C-reactive protein fell by four months [4].
• Middle-aged adults, 4 months, 2022. Eighty-eight overweight, untrained adults took 500 mg, 1000 mg or placebo. Peak power output, the primary endpoint, did not improve significantly (+4.3% and +3.9%). Hamstring average peak torque rose 12% at 500 mg and 9.8% at 1000 mg against placebo, which itself declined; peak oxygen uptake rose about 10% in the high-dose group. Lean and fat mass were unchanged [2].
• Immune ageing, 4 weeks, 2025. Fifty healthy middle-aged adults took 1000 mg a day. Urolithin A expanded peripheral naive-like, less terminally exhausted CD8+ cells by 0.50 percentage points (95% CI 0.16 to 0.83) and raised CD8+ fatty-acid oxidation capacity by 14.72 percentage points (95% CI 6.46 to 22.99). Mitochondrial biogenesis in CD8+ cells increased, as did CD56dim CD16bright NK cells and non-classical monocytes, activation-elicited TNF secretion and bacterial uptake by monocytes [5].
• Trained runners, 4 weeks, 2025. Forty-two competitive male distance runners took 1000 mg a day through an altitude camp. The 3000 m time trial did not improve, and there was no significant treatment effect on VO2max. Perceived exertion fell and post-exercise creatine kinase dropped sharply; muscle proteomics showed mitochondrial pathways up and inflammatory pathways down, with mitophagy markers trending up but no measured change in mitochondrial function [6].

Read together, these trials describe a compound that reliably shifts mitochondrial and inflammatory biology and unreliably shifts what people can do. The molecular signal is real; the functional signal is small, inconsistent across endpoints, and absent in people who are already fit.

Mitophagy, not antioxidant chemistry. Urolithin A sits in the polyphenol aisle, but its mechanism has little to do with scavenging radicals. Ellagitannins are potent antioxidants in vitro and largely lose that property on conversion to urolithins; what urolithin A does instead is switch on the quality-control pathway that removes damaged mitochondria [3][7].

Mitophagy is selective autophagy of mitochondria. In its best-characterised form, PINK1 accumulates on the outer membrane of a depolarised mitochondrion, recruits and activates the ubiquitin ligase Parkin, and Parkin tags the organelle for engulfment. The 2016 screen identified urolithin A as the first natural compound to induce this process both in vitro and in vivo following oral consumption [3].

The human muscle evidence is the strongest part of the case. Vastus lateralis biopsies taken before and after four months of supplementation showed, at 500 mg a day, enrichment of the "Parkin-mediated ubiquitin and proteasomal systems" pathway, with the ubiquitin-conjugating enzymes UBE2N and UBE2R2 raised. Targeted immunoblotting confirmed a rise in phospho-Parkin at Ser65 — the fraction of Parkin that PINK1 phosphorylates once it has translocated onto a mitochondrion. At 1000 mg a day the dominant change was metabolic rather than degradative: TCA-cycle, fatty-acid oxidation, electron-transport-chain and oxidative-phosphorylation proteins rose dose-dependently. BNIP3, which drives the PINK1/Parkin-independent mitophagy route, fell at the higher dose, which argues for an effect specific to the PINK1/Parkin axis rather than a general upregulation of autophagy. The fusion and fission machinery did not move — MFN1, MFN2, DRP1 and OPA1 were all unchanged — and VDAC, a marker of raw mitochondrial content, was unchanged too [2].

That last detail matters. Urolithin A appears to improve the quality of a muscle's mitochondria rather than the number of them. Falling plasma acylcarnitines in both four-month trials point the same way, since acylcarnitines accumulate when fatty acids enter the mitochondrion faster than they are oxidised [2][4].

No binding target has been identified for the mitophagy effect. This is the honest gap. The discovery paper describes urolithin A as inducing mitophagy without naming a receptor or enzyme it occupies to do so [3], and the later human work does not fill that in. "Mitophagy activator" names what happens downstream, not what the molecule binds.

Where a direct receptor target does exist, it is the aryl hydrocarbon receptor, and it is weak. In colon epithelial cells, urolithin A raised Cyp1A1 message, protein and enzyme activity, drove AhR into the nucleus, and gave 2–4-fold induction in an XRE-luciferase reporter. The calibration is the interesting part: at 50 µM it matched β-naphthoflavone, a low-affinity AhR ligand, while the high-affinity ligands FICZ and TCDD produced larger responses at nanomolar concentrations and about 15-fold reporter induction. In AhR-knockout mice, oral urolithin A produced no Cyp1A1 induction at all [9]. So AhR is a real direct target engaged at micromolar concentrations, and calling urolithin A an AhR agonist without that qualifier overstates it.

Downstream of AhR sits Nrf2. Urolithin A raised Nrf2 message and protein and moved it into the nucleus, inducing its target genes NQO1 and haem oxygenase-1, and raised the tight-junction proteins claudin-4, occludin, ZO-1 and TJP3. Colon explants and mice lacking either Nrf2 or AhR failed to respond, and knocking down AhR or Cyp1A1 in cells abolished the claudin-4 response, so the order runs AhR → Cyp1A1 → Nrf2 → tight junctions. Functionally, pretreatment stopped LPS-induced leakage of FITC-dextran across an epithelial monolayer, and oral dosing attenuated colitis in mouse models [9].

Mitophagy feeds a second, less obvious pathway in T cells. In mice, urolithin A expanded T memory stem cells and naive T cells and conferred strong anti-tumour CD8+ immunity. The route is mechanically specific: PINK1-mediated mitophagy released the mitochondrial phosphatase Pgam5 into the cytosol, where it dephosphorylated β-catenin, driving Wnt signalling and compensatory mitochondrial biogenesis [10]. The 2025 human trial was designed against that model and found a matching, if small, shift in circulating T-cell phenotype and fatty-acid oxidation capacity [5].

In the brain, timing appears to decide the outcome. Urolithin A attenuated memory impairment and neuroinflammation in APP/PS1 mice [11]. A 2026 study then deliberately compared prevention with rescue and found an asymmetry. Given to 16-month-old C57BL/6 mice for eight weeks, urolithin A raised hippocampal ATP, increased fusion, antioxidant and biogenesis proteins and lowered phosphorylated tau — and did not restore memory in the Morris water maze. Given to 5-month-old senescence-accelerated SAMP8 mice, before their cognitive decline begins at six months, it raised ATP roughly four-fold, cut reactive oxygen species and unfolded-protein load, lowered PHF-1 tau, raised the postsynaptic protein PSD95 and preserved spatial memory [12]. Both arms used 1.5 mg/kg intraperitoneally three times a week, a dose and route no one uses, but it is a rare study that asked whether the molecular changes translate and reported honestly that in old animals they did not.

Joints. Urolithin A improved mitophagy and mitochondrial respiration in primary chondrocytes from both healthy and osteoarthritic human donors, and reduced cartilage degeneration, synovial inflammation and pain in a mouse osteoarthritis model [13]. No human osteoarthritis trial is cited here.

Supplement urolithin A is synthetic. That is a deliberate solution to the metabotype problem: because production depends on carrying the right gut bacteria, eating the precursor food is not a reliable way to get the molecule [1][7]. Only about 15% of participants in the largest muscle trial had measurable plasma urolithin A before supplementation began [2].

In circulation, most of what is measured is conjugated. Plasma urolithin A-glucuronide and urolithin A-sulfate both rose dose-dependently alongside the parent compound over four months [2], matching rat ADME work in which glucuronidated and sulfonated forms were the predominant metabolites after both oral and intravenous dosing [14]. Whether the conjugates are inert carriers or contribute activity is not settled by any study cited here.

Tolerability in the human trials has been unremarkable. Across four months at 1000 mg a day in adults aged 65 to 90, 33 adverse events were recorded, none serious, with no statistical difference between urolithin A and placebo and no changes in vital signs, blood biochemistry, haematology or urinalysis [4]. In the 88-person middle-aged trial, adverse events were likewise diverse and not concentrated in the treated groups; a slight excess of musculoskeletal events was attributed to the muscle-biopsy procedure [2].

The animal toxicology was done properly and is reassuring. Synthetic urolithin A was not genotoxic across a battery of assays. In 28-day and 90-day oral rat studies at up to 5% of the diet, there were no alterations in clinical parameters, blood chemistry or haematology, no target organs and no specific toxic mechanism; the no-observed-adverse-effect level was the highest dose tested, 3451 mg/kg/day in males and 3826 mg/kg/day in females [14]. That is several thousand times the human trial doses on a weight basis.

Two caveats sit outside the toxicology. No trial cited here has run longer than four months, so the effect of switching on a mitochondrial quality-control pathway for years is untested. And urolithin A engages the aryl hydrocarbon receptor [9] — weakly, but AhR is a transcription factor with broad effects on xenobiotic metabolism and immunity, and no cited study followed that out over long-term use.

No clinical interaction study is cited here. Two mechanistic points are worth flagging rather than ignoring.

Urolithin A induces Cyp1A1 through AhR in colon and liver, confirmed in vivo by its absence in AhR-knockout mice [9]. Cyp1A1 induction is the classic route by which AhR ligands change the clearance of other substrates. The induction there was modest — about 5–6-fold in liver by the intraperitoneal route, against roughly 30-fold for a high-affinity ligand — and nothing cited here measured drug clearance in people, so this is a mechanism worth knowing about rather than a documented interaction.

Because the proposed benefit runs through mitochondrial quality control, it overlaps conceptually with NAD+ precursors, CoQ10 and PQQ. None of those combinations has been trialled.