Metformin’s blood sugar control starts in the brain, not just the liver, study finds

By Vijay Kumar MalesuReviewed by Susha Cheriyedath, M.Sc.Aug 3 2025

Scientists uncover how low-dose metformin targets brain pathways to lower blood sugar, opening fresh avenues for safer and smarter diabetes therapies.

Study: Low-dose metformin requires brain Rap1 for its antidiabetic action. Image Credit: Kateryna Kon / Shutterstock

In a recent study published in the journal Science Advances, researchers tested whether low, clinically relevant doses of metformin lower blood glucose via inhibition of Ras-related protein 1 (Rap1) in the ventromedial hypothalamic nucleus (VMH) of the brain.

Classic models place metformin’s action in the liver through adenosine 5′-monophosphate-activated protein kinase (AMPK), but newer work adds adenosine 3′,5′-cyclic monophosphate (cAMP) signaling, mitochondrial targets, and even gut-mediated effects, including glucagon-like peptide-1 (GLP-1) and growth and differentiation factor 15 (GDF15). The central nervous system tightly regulates glucose through hypothalamic circuits, so even small drug signals in the brain can shift whole-body metabolism.

There remains uncertainty about the relative importance of these pathways at clinically relevant metformin doses. Could low doses of metformin work by a neural route? The current study addresses this question and highlights the need for further research to dissect brain-to-organ pathways.

The researchers used mice to test a brain-based pathway. They compared normal littermates with Rap1ΔCNS mice, a forebrain-specific Rap1 knockout generated by deleting Rap1a and Rap1b in calcium/calmodulin-dependent protein kinase II alpha (CaMKIIα)-expressing neurons. All mice were given a high-fat diet to raise blood sugar (hyperglycemia). They received single or repeated intraperitoneal doses of antidiabetic agents like metformin (a biguanide), rosiglitazone (a thiazolidinedione), exendin-4 (a GLP-1 receptor agonist), glibenclamide (a sulfonylurea), dapagliflozin (an SGLT2 inhibitor), and insulin, with blood glucose tracked over time. Dose-response testing used metformin at 50–150 mg/kg and glucose tolerance tests (GTTs) with area under the curve (AUC) analysis.

To probe central action, metformin was delivered by intracerebroventricular (ICV) injection (1–30 μg) to diet-induced obese mice, with food-intake controls and body-weight monitoring. Electrophysiology in hypothalamic slices assessed how metformin alters the firing of steroidogenic factor-1 (SF1) neurons in the VMH. Gain-of-function experiments expressed constitutively active Rap1 (Rap1V12) using adeno-associated virus (AAV) in VMH or a Rosa26-lox-stop-lox (LSL)-Rap1V12 × CaMKIIα-Cre cross to elevate CNS Rap1 activity. Outcomes included blood glucose, glucose tolerance, and c-Fos mapping of neuronal activation.

Deleting Rap1 in forebrain neurons produced a selective defect in metformin responsiveness. In littermate controls, metformin lowered glycemia, but Rap1ΔCNS mice did not show significant glucose reductions to metformin despite normal responses to other antidiabetic agents. Thus, global glucose-lowering capacity was intact, yet metformin’s effect was specifically lost when brain Rap1 was absent.

Dose-response studies sharpened this selectivity. At 50–150 mg/kg, metformin reduced blood glucose in controls in a dose-dependent fashion (quantified by AUC), but the same doses failed in Rap1ΔCNS mice. GTTs showed that low-dose metformin improved tolerance in controls, whereas Rap1ΔCNS mice gained this benefit only at suprapharmacologic exposures (≥200 mg/kg), implying that high concentrations can bypass the brain pathway. This highlights that the requirement for brain Rap1 is specific to low, clinically relevant doses of metformin, while higher, less clinically relevant doses likely act through peripheral mechanisms.

Directly targeting the brain confirmed sufficiency. ICV metformin (as low as 1–10 μg) acutely lowered blood glucose in diet-induced obese mice and in leptin-deficient (ob/ob) and streptozotocin-treated models, independent of food intake and without weight loss, indicating a centrally mediated glycemic effect at tiny doses compared with systemic delivery.

c-Fos mapping localized metformin-responsive neurons to the VMH. Electrophysiology showed that metformin depolarized VMH SF1 neurons and increased firing; this response was largely abolished when Rap1 was removed from SF1 neurons, implicating a Rap1-dependent VMH node as the metformin target.

Gain- and loss-of-function genetics further cemented causality. In Rap1CNSV12 mice (constitutively active Rap1 in forebrain), fasting glycemia and intolerance were higher, and metformin no longer improved glucose excursions during GTTs. Similarly, forcing Rap1V12 expression bilaterally in VMH using AAV blunted both acute and chronic glucose-lowering by metformin and markedly impaired metformin-induced improvements in glucose tolerance. Conversely, deleting Rap1 specifically in SF1 neurons lowered glycemia to the same degree as metformin and eliminated any additional acute or chronic effect of the drug. Together, these manipulations show that metformin’s therapeutic effect requires Rap1 inhibition within VMH SF1 neurons.

Pharmacological context matters, as brain and cerebrospinal fluid metformin concentrations at therapeutic dosing are ~0.5–10 micromolar, far below hepatic or intestinal levels. In this range, metformin activated SF1 neurons and reduced Rap1 activity, consistent with a highly sensitive central mechanism that dominates at low doses, while higher, less clinical doses likely recruit peripheral pathways and can bypass the CNS Rap1 requirement. The study does not exclude the possibility of direct effects of metformin on peripheral tissues such as the liver and intestine at higher doses.

In mice lacking neural Rap1, baseline blood glucose is often reduced, which may limit the observable effect of further metformin administration ("floor effect"). However, even in glycaemia-matched groups, metformin failed to lower glucose in Rap1ΔCNS mice but remained effective in controls.

The study also points to the potential involvement of other regulators, such as exchange protein directly activated by cAMP 2 (EPAC2), in activating Rap1 in the brain, as well as possible connections to the lysosomal AMPK pathway. While this was not directly tested in this study, it represents a promising avenue for future research.

To summarize, this study identifies a brain-first mechanism for metformin at therapeutic exposure: low doses inhibit Rap1 in VMH SF1 neurons to lower blood glucose. The effect is selective for metformin among United States Food and Drug Administration-approved agents and is lost when CNS Rap1 is deleted or constitutively activated, but restored at very high, less clinically relevant doses that likely act peripherally. While the findings highlight the importance of the VMH Rap1 pathway at low, clinically relevant doses, they do not rule out peripheral mechanisms at higher doses or in other contexts. For patients and clinicians, this brain pathway helps explain why modest doses work safely and consistently, and it points to central Rap1 signaling as a target to refine diabetes therapies that coordinate liver, muscle, and gut.