Urolithin A vs. Urolithin B

Urolithins are small molecules that gut microbes can produce from ellagic acid and ellagitannin-rich foods. Pomegranates and walnuts provide precursors; they do not deliver a standardized dose of every downstream metabolite. The microbial conversion varies between people, which makes “I ate the precursor” different from “my tissues received this molecule.”^[2][3][22]

A 2026 experiment illustrates why the chemistry matters. A affected platelet activation in an ex vivo human system; removing or repositioning its second hydroxyl group, or attaching a glucuronide at the studied site, did not preserve that effect. This is structure–activity evidence. It does not show that an oral supplement prevents blood clots.^[1]

Microbial production patterns are sometimes called metabotypes. UM-A predominantly identifies A as the final urolithin; UM-B includes B and/or iso-A alongside A. UM-0 indicates no detectable production under the tested conditions. “B producer” therefore does not mean someone makes only B, and these categories are not a simple ranking of healthy versus unhealthy guts.^[22]

Direct supplementation bypasses part of this production bottleneck. In a randomized crossover study of 100 adults, a product containing 500 mg of A produced more consistent circulating exposure than about 240 mL of pomegranate juice. It was an exposure comparison, not proof that the supplement was better for every health outcome or that eating fruit was ineffective.^[2]

Pharmacokinetics describes what the body does to a compound: absorption, distribution, metabolism and elimination. Pharmacodynamics describes what the compound does to a biological system. A pathway can look excellent while delivery remains inadequate. Equally, detecting a metabolite in blood establishes exposure without proving a benefit.

The unconjugated parent molecule is often called the aglycone. After absorption, phase II enzymes attach groups such as glucuronide or sulfate. Those conjugates can have different distribution and activity. “Free” in this discussion means unconjugated; it should not be confused with the pharmacological concept of the fraction unbound to plasma proteins.^[4][7]

The early A study included 60 older volunteers across its single- and repeated-dose portions. Conjugates predominated in circulation; short-term tolerability was favorable. These observations make human delivery more credible, but they do not establish absolute oral bioavailability, a hypertrophy threshold or the best personal dosing schedule.^[4]

B does appear after precursor consumption. The pomegranate study detected B-glucuronide in urine from some volunteers. That answers whether humans can form and eliminate a B metabolite. It cannot supply the half-life, clearance or muscle concentration of a capsule containing isolated B.^[3]

Laboratory methods can make this distinction easy to miss. An oak-extract study reported a median B serum signal of 22.3 ng/mL after eight weeks. Its samples were treated with β-glucuronidase before analysis. The reported number therefore includes B released from conjugates; it is not a measurement of that concentration of circulating unconjugated B. The intervention was also oak extract, not purified B.^[5]

Conjugation is not necessarily a permanent dead end. Experimental work shows that β-glucuronidases can release urolithin aglycones from conjugates. Local deconjugation is a plausible contributor to tissue activity. Its extent and location still need measurement; it cannot be assumed to reproduce a muscle-cell bath containing a fixed micromolar concentration.^[6]

Nor are all conjugates simply inactive. Direct immune-cell comparisons found that glucuronides failed to reproduce some cytokine effects of the parent compounds, while other signaling effects persisted. The appropriate question is which molecular form, in which tissue, at which exposure?^[7]

A’s best-known mechanism is mitophagy: selective disposal of damaged mitochondria through cellular recycling machinery. The appeal is improved quality control in a tissue with demanding energy needs. Foundational experiments linked A to mitophagy, longer worm lifespan and better muscle function in rodents. Those species and endpoints matter; a worm-lifespan result does not quantify human longevity.^[8]

Think of maintenance as preserving the machinery that supports work. Clearing a malfunctioning component and building more contractile tissue are different tasks. They may support each other, but neither makes the other inevitable. A trial must measure performance and muscle size separately if it wants to claim both.

B enlarged C2C12 mouse myotubes at 15 μM over 24 hours. AR knockdown and pharmacological blockade supported androgen-receptor involvement; rapamycin supported mTORC1 dependence. Akt phosphorylation did not increase. Direct AR binding was not established, so “natural SARM” would overstate the experiment.^[9]

In mice, pumps delivering 10 μg/day for 28 days increased tibialis fiber area by 11.9%. There were six animals per arm, with three per arm in that histology comparison. Continuous pump delivery is a different exposure from swallowing a supplement. This is a promising lead for translation, not a human muscle-gain estimate.^[9]

The mechanism should also remain tissue-specific. In activated microglial models, B influenced inflammatory and oxidative-stress pathways, including Nrf2/HO-1 and NF-κB, with signaling that differed from the muscle experiment. A single molecule need not turn the same intracellular switch in the same direction everywhere.^[17]

A 2026 study compared A and B in cultured myotubes derived from nine post-mortem donors, aged 55–96. It gives us something more informative than comparing unrelated experiments: both molecules tested within one human-cell platform. It still does not reproduce digestion, blood metabolism, training or the physiology of a living participant.^[10]

A preferentially affected mitochondrial and inflammatory programs; B affected lipid-metabolism and muscle-related programs. Both showed supportive fusion-like observations. Donor histories were limited, and pathway predictions from gene expression need functional validation. This makes “A only cleans up, B only builds” too rigid. The field is discovering overlapping effects, not two neatly separated products.^[10]

The result sharpens the next question: can an orally administered product reproduce a relevant response at concentrations actually reached in human muscle? A concentration chosen for a cell experiment is an experimental condition, not an established target for a consumer.

The most useful comparison currently has an uneven design: several randomized A trials, and no controlled trial of directly administered B establishing human hypertrophy identified in our search through October 10, 2026. That is an evidence gap, not a finding that B cannot work.

The middle-aged trial offers a concrete strength result. Average hamstring torque changed by +12.0% with 500 mg, +9.8% with 1,000 mg and −9.8% with placebo. Those are each arm’s changes from baseline; they are not three placebo-adjusted effect sizes. The corresponding between-group P values were .027 and .029. Walking and aerobic-capacity improvements were not statistically significant between groups, and lean and fat mass were unchanged.^[12]

In the older-adult trial, walking distance improved in both groups, so a headline quoting only A’s improvement would overstate the treatment evidence. Adverse-event rates did not significantly differ, and no serious adverse events were reported. The modest sample, predominantly female and entirely White, limits generalizability.^[11]

A smaller trial in 20 resistance-trained men reported advantages in selected contraction and repetition tests, but not bench or squat one-repetition maximum. Its duration reporting is inconsistent across the title, methods and figure captions. It is worth tracking, without making it the foundation of a hypertrophy claim.^[23]

These studies support a more precise ambition than “builds muscle”: improving particular aspects of performance in particular settings. A can remain interesting even when a trial does not grow lean mass. The distinction helps readers choose the endpoint they care about rather than treating endurance, strength and muscle size as interchangeable.

Muscle is only one tissue with expensive maintenance demands. Immune cells, neurons and vascular cells also adapt their metabolism. That makes broader applications biologically plausible—but each needs its own outcome evidence. An effect on an isolated immune cell cannot automatically become a claim about infection resistance.

A 2025 randomized trial assigned 50 adults aged 45–70 to A at 1,000 mg/day or placebo for 28 days. It found changes in CD8-cell phenotype and metabolism, including a 0.50-percentage-point treatment difference in the reported naive-like phenotype measure (95% CI, 0.16–0.83). This is human immune biology, not demonstrated prevention of infections or cancer. A January 2026 correction replaced a duplicated figure panel; we assessed the corrected article.^[14][15]

A fresh trial published October 9, 2026 studied 50 middle-aged adults with obesity for four weeks. A at 1,000 mg/day did not significantly improve the primary brachial flow-mediated-dilation outcome. Some task-evoked cerebrovascular responses improved, while task performance did not. Interesting brain hemodynamics deserve follow-up; calling this proven cognitive enhancement would move beyond the data.^[16]

B’s neurological work includes experimental aging models with behavioral and neuronal outcomes. These can help identify a mechanism and a candidate intervention, but an induced-aging mouse model is not a diagnosis of Alzheimer’s disease in a person. Similarly, a high-fat/high-sugar mouse experiment cannot establish a human fat-loss effect or prove that microbiome changes caused every observed benefit.^[18][19]

The shared opportunity is considerable: small metabolites connecting diet, microbes and tissue biology. The strongest next studies will ask whether those connections produce clinically useful changes, rather than assembling a long list of pathways and calling the list a benefit.

A has short-term human tolerability data from several trials. That provides a stronger starting point than natural occurrence alone. It does not settle rare adverse effects, multi-year exposure, every medication combination, pregnancy or all disease populations. B’s safety cannot be inferred from A’s trials because the molecules and biological responses differ.^[4][11][14]

The foundational B mouse experiment also reported lower testis weight despite unchanged measured testosterone. It does not prove a human reproductive effect, but it prevents describing B as demonstrably “non-suppressive” or consequence-free.^[9]

Regulatory language needs equal precision. FDA’s response to GRAS Notice 791 stated that it had no questions regarding the notifier’s conclusion for A under specified intended food uses. That is not FDA approval of A as a muscle-building drug, and it says nothing equivalent about B.^[21]

Funding is part of interpretation. Several major A trials were funded by Amazentis and involved investigators with company ties. The 2026 vascular trial reported foundation funding and no declared conflicts. Commercial involvement does not erase an experiment, but independent replication, prespecified endpoints and complete null-result reporting strengthen confidence.^{[4][11][12][13][14][16]}

A-plus-B combinations are easy to imagine: maintenance alongside protein-turnover support. That is a research hypothesis. Complementary-looking pathways do not establish additive benefits, the right timing, an exposure ratio or safety. No convincing human combination trial demonstrating superior hypertrophy was identified in this review.

Readers should also preserve product identity. A study of purified A is not automatically a study of a pomegranate extract, a different formulation or a multi-ingredient “mitochondrial” blend. A study of B in cells is not an evaluation of the purity or performance of a retail product bearing its name. Ingredient identity, route and the measured outcome belong together.

For B, the next useful step is human pharmacology: verified compound identity, oral exposure, parent-versus-conjugate measurements and tolerability. Muscle sampling could test whether an exposure actually engages the proposed biology. A null result here would still be valuable because it could distinguish a delivery problem from an ineffective target.

A comparative trial should not rely on lean mass alone. Muscle imaging, force production and practical function answer different questions; nutrition and training need enough control to interpret them. The primary outcome, sample size and analysis should be chosen before investigators see which result looks best.

For A, larger independent trials can clarify who benefits most and whether selected laboratory-performance signals become durable improvements in everyday life. For B, carefully designed human studies can test whether its muscle-cell effects survive absorption and metabolism. A direct comparison would finally let the molecules compete on the same meaningful endpoint.