Brain-penetrant mTOR inhibition clears aggregates in models, but human dosing has not been shown to reach the brain — and mTORC2 is needed for memory in mice

Mechanism-to-outcome gap · editorial confidence 70% · evidence stands at: Clinical outcome in humans — negative on target engagement · Atlas ID H8

Editorial confidence is the curator's judgement of how likely the hypothesis is to hold, given the studies in the Atlas. It is set by hand, not computed from data, and it is not a calibrated probability.

The gap

In animal and cell models, blocking mTOR clears out the toxic protein clumps involved in diseases like Alzheimer's and improves memory-related behaviour - a fairly strong case at the lab level. In fact, one long mouse study found rapamycin actually improved learning and memory and reduced anxiety. In people the picture is different: two trials reported in 2025 and neither showed a benefit - and in both, the drug could not be shown to have reached the brain at all, so they do not really test the idea. The worry that these drugs also block mTORC2 is backed by a study in this Atlas: mice whose forebrain neurons lacked mTORC2 formed weaker long-term memories, and raising mTORC2 activity improved memory (HUA2013). That was shown in mice and flies, not in people.

Technical framing: The preclinical CNS case for mTOR inhibition is unusually deep; the human evidence is thin rather than absent. Aggregate-clearance evidence: SPI2010 rapamycin lowered amyloid-beta and abolished cognitive deficits in an AD mouse model, with increased neuronal autophagy; ZHO2009 mTORC1 inhibition suppressed anatomical/behavioral abnormalities; CAC2010 mapped the mTOR-amyloid-tau triangle onto cognition; TAN2024 (preprint) restored tau-induced metabolic, mitochondrial and cognitive deficits; EHN2008 reversed learning deficits in Tsc2+/- mice. Human evidence: SAX2026, the TRON multicentre randomised controlled trial, tested everolimus against placebo on memory and executive function in tuberous sclerosis complex (38 randomised, ages 16-60, 24 weeks). Effect sizes were small (Cohen's d 0-0.465) and responder rates were high in BOTH arms (20/23, 87% everolimus vs 9/12, 75% placebo), which the authors attribute to practice and placebo effects - an uninformative trial rather than a positive one. It is disease-specific, short and not an ageing trial, so it does NOT close the cognitive-AGEING gap; but it does mean that not every outcome in this corpus comes from animals or models. HAL2012 is on the benefit side: chronic rapamycin ENHANCED learning and memory in young mice, BLOCKED age-associated cognitive decline in older mice, and DECREASED anxiety- and depression-like behaviour, with elevated midbrain monoamines. The mTORC2 liability is sourced by HUA2013. Conditional deletion of Rictor in the postnatal mouse forebrain reduced mTORC2 activity and selectively impaired long-term memory and the late phase of hippocampal LTP, with the same deficit in dTORC2-deficient flies; hippocampal actin polymerisation fell, restoring it rescued both L-LTP and memory, and a compound that raised mTORC2 activity converted early LTP into late LTP and enhanced memory. Loss of function and rescue in the same study, which is why it can carry the claim. mTORC2 is therefore required for memory consolidation - and mTORC2 is the arm chronic rapamycin also disrupts (LAM2012, SAR2006). Net: strong mechanism on both sides of the trade-off (A and M – animal and molecular), two human trials that reported in 2025 without demonstrating brain exposure or target engagement, so they do not cleanly test efficacy, one uninformative disease-specific human trial, and an unquantified balance between clearing aggregates (mTORC1/autophagy) and preserving plasticity (mTORC2).

What changed since this question was written

Two human trials have now reported, and in neither was brain exposure or brain target engagement demonstrated, so neither is a clean test of efficacy. An eight-week open-label phase 1 in mild cognitive impairment and early dementia made CNS penetrance its primary endpoint and found rapamycin undetectable in cerebrospinal fluid before or after treatment, with no cognitive change, while CSF p-tau181, GFAP and neurofilament light rose (Gonzales, Commun Med 2025, PMID 40394335, NCT04200911, n=10). Read that result for what it measures: a CSF concentration below the limit of detection is not the same measurement as drug absent from brain tissue, and the assay's sensitivity and sampling times have not been independently adjudicated here. A six-month single-arm phase IIa at 7 mg/week in early Alzheimer's found no change in cerebral glucose metabolism on FDG-PET and no cognitive change; CSF total tau rose 22% and neurofilament light 24%, which at n=13 without a control arm is at least as compatible with neuronal injury as with aggregate clearance (Svensson, medRxiv 2025, NCT06022068, preprint). A null imaging endpoint is not itself a measure of brain exposure either. A randomised 12-month phase 2 with cognitive secondary endpoints reads out in 2026 (REACH, NCT04629495, n=40).

What is still open

Not whether mTOR inhibition helps human cognitive ageing, because that has not been tested yet. Whether any CNS-penetrant mTOR inhibitor can achieve brain mTORC1 inhibition at tolerable doses. Every human ageing and Alzheimer's trial found uses sirolimus or everolimus, both P-glycoprotein substrates and poor CNS penetrators, and no CNS-optimised mTOR inhibitor appears in the ageing or dementia trial registry. The mTORC2 clause makes the design harder rather than easier: an inhibitor that did reach the brain chronically would be expected to erode long-term memory (Huang, Nat Neurosci 2013, PMID 23455608), so the target is a window, not a direction.

The hypothesis

Educated guess: a drug that actually gets into the brain, given in pulses rather than continuously, might help clear the protein clumps of neurodegeneration without blunting the part of the pathway that memory needs. The catch found in 2025: at the doses used in ageing trials, neither of the two trials managed to measure the drug getting into the brain — which is not the same as proving that none of it arrives.

Technical framing: A CNS-penetrant, intermittent or mTORC1-selective regimen can reach brain mTORC1 at a level that enhances autophagic clearance without chronic mTORC2 suppression. The first half — reaching the brain at all — is the part that two trials at geroscience doses were unable to demonstrate, and it has to be demonstrated before the second half can be tested.

How it could be tested

Two-step: (i) mouse tauopathy/amyloid models comparing chronic pan-rapamycin vs intermittent vs mTORC1-selective (bi-steric) dosing - endpoints aggregate load, autophagic flux (LC3-II/p62), hippocampal LTP, and Morris-water-maze memory, to isolate the mTORC2/plasticity cost. (ii) A biomarker-anchored early-AD human pilot (CSF/plasma p-tau and amyloid PET as surrogate) on the winning schedule. Prediction: intermittent/selective arms clear aggregates with preserved LTP and memory; chronic pan-inhibition clears aggregates but degrades plasticity.

Related studies

Year	Evidence	Study
2010	A	Inhibition of mTOR by rapamycin abolishes cognitive deficits and reduces amyloid-beta levels in a mouse model of Alzheimer's disease SPI2010 Connected the longevity drug to a specific age-related disease. Long-term rapamycin prevented memory deficits and lowered toxic amyloid-beta in an Alzheimer's mouse model - and the benefit tracked with INCREASED autophagy in neurons. Suggested that the same autophagy boost that may slow aging could also help clear disease-causing proteins.
2009	A	Pharmacological inhibition of mTORC1 suppresses anatomical, cellular, and behavioral abnormalities in neural-specific Pten knock-out mice ZHO2009 In neural Pten-knockout mice, rapamycin prevented and reversed neuronal hypertrophy and improved a subset of the PTEN-associated abnormal behaviours.
2010	A	Molecular interplay between mammalian target of rapamycin (mTOR), amyloid-beta, and Tau: effects on cognitive impairments. CAC2010 Revealed a vicious cycle: amyloid-beta RAISES mTOR activity, and high mTOR in turn blocks the autophagy needed to clear amyloid and tau - so the disease feeds itself. Rapamycin broke the loop in 3xTg-AD mice, rescuing memory and lowering BOTH amyloid and tau, with autophagy induction shown to be necessary for the reduction in amyloid levels.
2008	A	Reversal of learning deficits in a Tsc2+/- mouse model of tuberous sclerosis EHN2008 Rapamycin reverses learning and memory deficits in a Tsc2+/- tuberous sclerosis model.
2024	PP	Targeting mTOR restores tau-induced metabolic, mitochondrial, and cognitive deficits in a tauopathy mouse model TAN2024 In mice engineered to overexpress a phosphomimetic tau variant, one week of rapamycin partially mitigated tau-driven mitochondrial dysfunction and cognitive impairment in the Morris water maze - extends the mTOR-autophagy-neurodegeneration link (already seen with Huntington's) to a direct tau-phosphorylation mechanism relevant to Alzheimer's.
2012	A	Chronic inhibition of mammalian target of rapamycin by rapamycin modulates cognitive and non-cognitive components of behavior throughout lifespan in mice. HAL2012 Asked whether the lifespan-extending dose of rapamycin harms or helps the aging BRAIN. Reassuringly, it enhanced learning and memory in young mice, prevented age-related cognitive decline in old ones, and even reduced anxiety and depression-like behavior - linked to boosted brain monoamines. Cognitive benefit, not cost.
2026	H	Everolimus for the treatment of neuropsychological deficits in tuberous sclerosis complex: findings from the TRON multicentre randomised controlled trial SAX2026 In a multicentre RCT of TSC patients (n=38 randomised, 2:1 everolimus:placebo), 24 weeks of everolimus produced similarly large 'responder' rates for neuropsychological improvement as placebo (87% vs 75%), suggesting large practice/placebo effects rather than a clear drug benefit on cognition -- despite everolimus's established efficacy for TSC tumours and epilepsy. Adverse events were more frequent with everolimus (88% vs 61.5%).
2013	A	mTORC2 controls actin polymerization required for consolidation of long-term memory HUA2013 Conditional deletion of Rictor in the postnatal mouse forebrain reduced mTORC2 activity and selectively impaired long-term memory and the late phase of hippocampal LTP, with a comparable long-term-memory deficit in dTORC2-deficient flies. Hippocampal actin polymerisation was reduced, and restoring it rescued both L-LTP and long-term memory; a compound that raised mTORC2 activity converted early LTP into late LTP and enhanced memory. This is the loss-of-function AND rescue evidence that mTORC2 is required for memory consolidation - the arm chronic rapamycin also disrupts, and the study Open Question H8 needs for its mTORC2 liability leg, which until 2026-09-05 rested on no supporting study at all.
2012	A	Rapamycin-induced insulin resistance is mediated by mTORC2 loss and uncoupled from longevity LAM2012 Chronic rapamycin (about 2 mg/kg a day, close to the lifespan-extending dose) made mice glucose intolerant and insulin resistant. A clamp test placed the problem in the liver, which kept making glucose, while muscle and fat took up glucose normally. Removing mTORC1 from the liver did not cause this. Rapamycin did break up mTORC2 (it separated mTOR from Rictor) in liver, fat and muscle, and deleting Rictor in the liver alone reproduced the glucose problem. In the genetic part, female mice with one working copy each of mTOR and mLST8 had about 50% less mTORC1 signalling in the liver, normal mTORC2 signalling and normal glucose tolerance, and lived 14.4% longer on average (p = 0.046). Males did not. Boundary: mouse drug and genetics work, not people, and the lifespan gain is female-only. It supports separating the two effects, but the paper does not test a drug that blocks mTORC1 without touching mTORC2.
2006	M	Prolonged rapamycin treatment inhibits mTORC2 assembly and Akt/PKB SAR2006 The molecular explanation for rapamycin's dark side. Short-term rapamycin only hits mTORC1, but prolonged treatment can also inhibit the assembly of NEW mTORC2 in susceptible cell types, reducing mTORC2 abundance and cutting Akt signaling over time. The wording matters: this is prevention of new assembly, not the dismantling of complexes already built, and the response depends on cell context and length of exposure. This is the mechanistic root of the insulin-resistance side effect later shown in mice (see Lamming 2012) - crucial for anyone dosing rapamycin for longevity.
2024	S	Targeting ageing with rapamycin and its derivatives in humans: a systematic review LEE2024 Described by its authors as the first systematic review of rapamycin/rapalogs in humans for aging. Screened 18,400 articles, included 19 studies. Found improvements in immune, cardiovascular, and skin (integumentary) parameters; NO significant effect on endocrine, muscular, or neurological systems. No serious adverse events in healthy people, but more infections and raised cholesterol/triglycerides in people with age-related disease. This is the highest-tier human-evidence summary in the whole Atlas - it aggregates many individual human studies rather than reporting one.
2026	H	Rapamycin increases cerebral blood flow and modulates metabolic, inflammatory, and microbiome profiles in healthy middle-aged APOE4 carriers: a pilot single-arm trial. AWARE2026 Single-arm pilot trial: 1 mg/day rapamycin for 4 weeks in 23 cognitively normal adults aged 45 to 65. Cerebral blood flow rose by more than 15% across several brain regions in the nine APOE4 carriers, while non-carriers showed no significant change; metabolic and inflammatory profiles improved and the gut microbiome was modulated. Uncontrolled, small and genotype-specific, so this is early evidence only.
2008	A	Response of a neuronal model of tuberous sclerosis to mammalian target of rapamycin (mTOR) inhibitors: effects on mTORC1 and Akt signaling lead to improved survival and function MEI2008 mTOR inhibition rescues a neuronal tuberous-sclerosis model, supporting rapamycin for TSC neurology.

Revision history

ShowHide

2026-08-30: SAX2026 (TRON trial) added as human evidence. HAL2012 had been cited as a 'cognitive downside signal'; it reports the opposite and moved to the benefit side.
2026-09-05: The mTORC2-memory leg previously cited LAM2012 and LEE2024, neither of which establishes it, and was flagged as an editorial extrapolation. HUA2013 now sources it.
2026-09-23: The card previously claimed no human cognitive-ageing or neurodegeneration outcome. Two 2025 trials reported null cognitive results, and in both the drug was not shown to reach the brain. The title was changed accordingly (URL kept).
2026-09-29: Title changed to 'has not been shown to reach the brain' and 'in mice'; SPI2010 no longer described as 'autophagy-dependent'; the beginner text no longer says the mTORC2 study is missing; CSF biomarker rises in Gonzales 2025 added.
2026-10-03: AWARE2026 added - human single-arm pilot (n=23) showing 1 mg/day sirolimus for 4 weeks raised cerebral blood flow in APOE4 carriers only; a CNS-level physiological effect in humans, but brain mTORC1 inhibition was not demonstrated.
2026-10-03: MEI2008 added - in neuronal Tsc1-knockout mice both rapamycin and RAD001 reached brain tissue and reduced phospho-S6 (mouse, high-dose developmental disease model; says nothing about human geroscience dosing).
2026-10-10: Audit N01 (residual A03). The card said both 2025 trials 'failed upstream of efficacy' and that reaching the brain 'has now failed twice at geroscience doses'. Not demonstrating exposure is not the same as demonstrating that exposure failed. The wording now keeps exposure, target engagement, downstream pharmacodynamics and clinical efficacy apart, and records that an undetectable CSF concentration is not a measurement of drug in brain tissue and that a null FDG-PET endpoint is not a measure of exposure.

← All open questions · Frontier questions

Was this page useful?