Metformin has been prescribed for more than sixty years, and nobody can agree on how it works. The liver was the standard answer. Then the gut made a strong case. Last year, a team at Baylor College of Medicine added a third candidate that got the most attention of the three, because it was the least expected: the brain.
The finding is real, and the experiments are careful. But it carries a complication that most coverage skipped past, and it is not a small one. Metformin is a positively charged, water-loving molecule that does not readily cross into the brain, and the experiments that most directly implicate the brain delivered the drug straight into the ventricles, bypassing the barrier entirely.
What the Baylor Team Actually Showed
The work centered on Rap1, a small GTPase, in the ventromedial hypothalamus, a region long implicated in whole-body glucose control.
Published in Science Advances by Hsiao-Yun Lin, Weisheng Lu, and colleagues, with corresponding author Makoto Fukuda, the study reported that mice with forebrain-specific Rap1 knockout resisted the glucose-lowering effect of low-dose metformin while remaining sensitive to other antidiabetic agents. Centrally administered metformin inhibited brain Rap1 and reduced high blood sugar. Forcing Rap1 into an active state increased glucose and completely abolished metformin's effect.
Metformin also activated a specific subset of neurons in the ventromedial hypothalamic nucleus, and that activation required Rap1. Fukuda, an associate professor of pediatrics and nutrition at Baylor, said the team "investigated whether and how the brain contributes" to metformin's antidiabetic effect, in contrast to the liver and gut explanations that dominated the field. The relevant cells were SF1 neurons, which showed a measurable increase in activity upon metformin administration to the brain.
Both loss-of-function and gain-of-function results pointed in the same direction. On its own terms, that is a strong package.
The Barrier Problem
Here is the part that complicates the headline. Metformin exists as a monoprotonated cation across the physiological pH range. A positive charge impedes passive diffusion across cell membranes, so its distribution depends almost entirely on transporters, principally the organic cation transporters of the SLC22A family.
Those transporters are abundantly expressed in the small intestine, liver, and kidney, which is why the drug concentrates in these organs. At the blood-brain barrier, the picture is murkier. A frequently cited measurement puts cerebrospinal fluid metformin at roughly 4 percent of plasma levels. A review of metformin's brain actions notes that a 2020 analysis using improved separation of cerebral microvessels found that OCT1 and OCT2 were not expressed in mouse, rat, or human cerebral microvessels, contradicting earlier assumptions that these transporters ferry metformin across the barrier.
Other work disagrees on the details. A 2023 in vitro blood-brain barrier model concluded that metformin uses organic cation transporters to cross an endothelial co-culture, and a 2025 systematic review of metformin and the central nervous system described rapid brain penetration after intravenous dosing with quantifiable cerebrospinal fluid levels and dose-dependent saturation.
The honest summary is that metformin reaches the brain in small and disputed amounts, and that how much arrives at any given nucleus at ordinary oral doses is not settled.
Why That Does Not Sink the Finding
The Baylor design partly anticipates this, which is worth noting.
The strongest evidence is genetic rather than pharmacological. Mice lacking Rap1 in the forebrain lost their response to low-dose metformin given systemically, while still responding to other drugs. That result does not depend on knowing the brain concentration. Something about brain Rap1 is required for a systemically administered drug to work, whether the drug arrives directly or acts through an upstream signal that does.
The team also emphasized that the brain appears to respond to far lower concentrations than the liver or intestine require, with neuronal activation detectable at doses as low as 1 microgram delivered centrally. If the relevant threshold in hypothalamic neurons is low enough, a 4 percent cerebrospinal fluid fraction could be sufficient. That is a plausible reconciliation. It is also, at present, an argument rather than a measurement.
The limitations extend beyond the barrier question. Every experiment was conducted in mice, most of which were genetically modified. Intracerebroventricular injection is a research tool, not a route any patient uses. No human data support the pathway. And metformin's mechanism has produced confident answers before, which were later revised, which is precisely the history the Baylor group is adding to rather than closing.
Fukuda has said the team plans to examine whether the same Rap1 signaling explains metformin's other documented brain effects, including reports that it slows brain aging, and that the findings open a door to drugs targeting the pathway directly. Both are reasonable next steps, and neither has happened yet.
None of this is a reason for anyone to change how they take metformin. Anyone with questions about their diabetes regimen should raise them with a prescribing clinician.
Key Questions Answered
What did the study find?
Low-dose metformin required Rap1 in the ventromedial hypothalamus to lower blood glucose in mice. Deleting brain Rap1 blocked the effect; forcing Rap1 activation abolished it.
Does metformin get into the brain?
Only to a limited degree. It is a charged molecule that depends on transporters, and a commonly cited figure places cerebrospinal fluid concentration at roughly 4 percent of plasma concentration. Evidence on brain transporter expression is inconsistent.
Why does that matter?
Some of the most direct experiments delivered metformin into the brain's ventricles, bypassing the barrier. That demonstrates what the brain can do with the drug, not how much of the oral dose reaches it.
What is the strongest evidence for the brain pathway?
The genetic result. Mice lacking brain Rap1 stopped responding to systemically administered low-dose metformin but remained responsive to other antidiabetic agents.
Was this tested in people?
No. All experiments used mice, most of them genetically modified.
Does this change how metformin should be taken?
No. The findings concern the mechanism, not dosing or indication. Nothing about prescribing has changed.
Why does the mechanism matter at all?
Knowing which tissue drives the effect would let drug designers build compounds that concentrate there, potentially delivering the benefit with fewer side effects elsewhere.
