{
 "meta": {
  "api_version": "1.0.0",
  "dataset_version": "2.0.0",
  "corpus_snapshot": "2026-10-01T14:49:35+0200",
  "source_commit": "03bba2194c4832a1bcdfb704d54dd458ab0df021",
  "license": "CC-BY-4.0",
  "cite": "Barton O. Oliver's mTOR Atlas. doi:10.5281/zenodo.22059963",
  "docs": "https://mtor-atlas.org/api/"
 },
 "count": 15,
 "data": [
  {
   "id": "H1",
   "kind": "open-question",
   "title": "Sensor-selective geroprotection without the metabolic penalty",
   "category": "Mechanism-to-outcome gap",
   "confidence": 0.75,
   "evidence_stands_at": "Mechanism in cells",
   "gap": "The amino-acid sensors are among the most pathway-specific upstream control points and among the least phenotype-tested; how readily each can be drugged has not been compared systematically. One exception exists and is held in this corpus: LEE2010 (Science 2010) showed that loss of Drosophila Sestrin produces fat accumulation, mitochondrial dysfunction, muscle degeneration and cardiac malfunction, all prevented by inhibiting TOR or activating AMPK. That is an organismal age-related-pathology phenotype (A – animal), though not a lifespan endpoint, and it is carried in the pathway model as the SESN2-AGING edge. Beyond that single fly result, the sensors (Sestrin2, CASTOR1, SAMTOR, v-ATPase) rest on M – molecular – work only, and no mammalian lifespan and no human ageing endpoint shows that manipulating ONE sensor arm reproduces the benefit of broader mTORC1 inhibition. Read the gap as a mechanism-attribution gap, not as an absence of mammalian data on the nutrients themselves. Dietary methionine restriction does extend mouse lifespan and does improve glucose and insulin measures (Miller 2005 and later work; outside this corpus, which is why no SID is cited here). What has not been shown is the SENSOR claim - that such a benefit is transmitted through SAMTOR to GATOR1, or through Sestrin2, rather than through any other consequence of eating less methionine.",
   "gap_beginner": "The amino-acid sensors in this pathway (Sestrin2, CASTOR1, SAMTOR, the lysosome's acid pump) are among the most pathway-specific control points in it - and among the least tested for whether changing them actually changes health or lifespan. There is one exception: in fruit flies, losing Sestrin caused fat build-up, mitochondrial problems, muscle degeneration and heart trouble, all preventable by turning TOR down. In mammals, the diet side of this does work - feeding mice less methionine makes them live longer and handle blood sugar better. What nobody has shown is that the benefit actually travels THROUGH one of these sensors, and no human study has tested it at all. So the question is not 'is there any evidence' - it is whether targeting one single sensor can reproduce what the broader drugs do.",
   "changed_since_written": "The premise has been disfavoured since 2018, outside this corpus. Methionine restriction delivers its metabolic benefit even in mice with constitutively active hepatic mTORC1, so suppressing mTORC1 is not required for it (Yu and Lamming, FASEB J 2018, PMID 29401631). Its known mediators are hepatic FGF21, the transsulfuration/H2S route (Hine, Cell 2014, PMID 25542313) and separable downstream arms (Wanders, FASEB J 2015, PMID 25742717) — and an H2S donor on its own did not extend lifespan in the ITP (Harrison, GeroScience 2023, PMID 38041783). Separately, the destination this question aims at was already reached genetically without any sensor: reduced mTORC1 alone extended life with normal glucose tolerance in female mTOR/mLST8 double-heterozygous mice (Lamming, Science 2012, PMID 22461615).",
   "still_open": "What fraction, if any, of methionine restriction's benefit actually runs through SAMTOR to GATOR1. This is genuinely untested: no SAMTOR knockout ageing study and no mammalian Sestrin lifespan study exists in the literature searched. But it is now a question about the size of a contribution, not about an unexplored route to geroprotection.",
   "hypothesis": "Stated as a null, which is the testable form: if methionine restriction's healthspan effects persist unchanged in a SAMTOR-null or SAM-binding-dead knock-in mouse, the sensor arm is dispensable and sensor-selective drugs are a dead end.",
   "hypothesis_beginner": "Turned around into a form that can fail: if mice engineered without a working SAM sensor still get the full benefit of a low-methionine diet, then that sensor is not how the benefit travels, and building drugs against it would be pointless.",
   "how_to_test": "SAMTOR-null or SAM-binding-dead knock-in mice on methionine-restricted against control diet; healthspan, glucose tolerance and mTORC2 activity as endpoints. One confound to design around: Sestrin loss is deleterious rather than geroprotective — it abolishes the benefits of exercise, acting through AKT and PGC1-alpha (Kim, Nat Commun 2020, PMID 31929512) — so a Sestrin knockout does not test this hypothesis. Secondary and cheaper: does hepatic PRMT1 inhibition, which improved insulin sensitivity in aged mice (Jiang, Cell Metab 2023, PMID 38006878), reproduce the metabolic half without mTORC2 loss?",
   "studies": [
    "GU2017",
    "SAX2015",
    "LEE2010",
    "LAM2012"
   ],
   "revision_log": [
    "2026-08-30: The evidence basis previously read 'supported ONLY by tier-D mechanistic studies and links to ZERO aging/longevity outcomes'. That was already false when displayed: LEE2010 had been added after the July 2026 external review, but this card had not been updated.",
    "2026-09-21: Scope clarified as a mechanism-attribution gap. An earlier beginner-register version said 'nothing equivalent has been shown in a mammal', which overstated the gap.",
    "2026-09-29: 'the most pharmacologically specific nodes' and 'most precise, most drug-friendly' softened, because no comparative analysis supports them. The Lamming 2012 result is now qualified as female mice."
   ],
   "url": "https://mtor-atlas.org/question/sensor-selective-geroprotection-without-the-metabolic-penalty/",
   "api_url": "https://mtor-atlas.org/api/v1/questions/H1.json"
  },
  {
   "id": "H10",
   "kind": "open-question",
   "title": "Suppression of senescence/SASP (geroconversion), not just autophagy, is the mediator of rapamycin's healthspan benefit",
   "category": "Mechanism-to-outcome gap",
   "confidence": 0.7,
   "evidence_stands_at": "Mechanism in cells",
   "gap": "H4 asks whether autophagy is REQUIRED for the lifespan benefit; this gap proposes a parallel, partly autophagy-INDEPENDENT mediator that is equally untested at the outcome level. Evidence that mTOR gates senescence: DEM2009 rapamycin decelerates cellular senescence (geroconversion from arrest to senescence); BLA2006 framed aging as quasi-programmed mTOR-driven cellular hyperfunction pharmacologically inhibitable by rapamycin; LAB2015 mTOR promotes the pro-tumorigenic SASP by driving translation of IL1A, the cytokine at the top of the NF-kB cascade (a second route, MK2/ZFP36L1 control of SASP mRNAs, is Herranz, Nat Cell Biol 2015, PMID 26280535, outside this corpus); CAS2009 mTOR drives Wnt-induced epidermal stem-cell exhaustion; LV2026 links Lamtor5-mTOR to cGAS-driven immunosenescence. Whether SASP suppression can occur independently of autophagy is an OPEN question in this Atlas, not a premise - and closing it requires a primary paper entered in the corpus. Human anchor: CHU2019 topical rapamycin REDUCED senescence markers (p16INK4a) in human skin (p=0.008) - the rare direct human PD readout, but on skin, not a systemic aging endpoint. Missing: no study in this corpus tests whether BLOCKING senescence suppression (or clearing senescent cells independently) reproduces or abolishes rapamycin's mammalian healthspan gain, so autophagy vs geroconversion as the dominant mediator is unresolved. Gaps here are computed against this Atlas, not against PubMed.",
   "gap_beginner": "Rapamycin's anti-ageing benefit is usually credited to boosting autophagy (cellular cleanup), but there is a competing explanation: rapamycin also calms \"senescent\" cells - old, damaged cells that pump out inflammatory signals. Whether it can do that WITHOUT autophagy is an open question: no study in this Atlas shows it. In humans, a topical version of the drug did reduce a marker of skin-cell senescence, but that is skin, not whole-body ageing. Nobody has directly tested, in one experiment, which of the two explanations actually drives the lifespan benefit.",
   "changed_since_written": "The dichotomy does not hold. Autophagy is required to build the SASP — the TASCC compartment physically couples mTOR to autolysosomes in senescent cells, and disrupting that localisation suppresses IL6 and IL8 synthesis (Narita, Science 2011, PMID 21512002) — and it also restrains the SASP, by degrading GATA4 through p62 (Kang, Science 2015, PMID 26404840). Both are true in different contexts, now formalised as a threshold model: below a damage threshold autophagy suppresses senescence, above it autophagy sustains senescent cells and their secretions (Bahar, Redox Biol 2026, PMID 41690118). So autophagy-deficient tissue with SASP suppression intact is not a state that can reliably be constructed. The occlusion design fails too: a senolytic must have its own robust lifespan effect to be occluded and fisetin did not extend lifespan in the ITP (Harrison, GeroScience 2023, PMID 38041783); senolytics are not clean, since D+Q and fisetin themselves inhibit mTOR; and non-additivity is weak evidence anyway, because two mechanistically distinct drugs can simply be additive (Gkioni, Nature Aging 2025, PMID 40437307). The senomorphic claim itself is essentially settled — rapamycin suppresses the SASP through IL1A translation (Laberge, Nat Cell Biol 2015, PMID 26147250).",
   "still_open": "Attribution of share, not a choice between two mechanisms. What fraction of rapamycin's lifespan benefit flows through senescent-cell burden, as against translation, autophagy, immune surveillance and tumour suppression? Nobody has partitioned it, and the partitioning is what a card like this should be asking for.",
   "hypothesis": "Rapamycin's benefit is partly mediated by suppressing senescence and the SASP, partly by other arms, and the open quantity is the share — not which of two candidates wins.",
   "hypothesis_beginner": "Rather than asking whether it is cleanup or calming inflamed cells, the honest question is how much of the benefit each one carries. They also turn out to depend on each other, so they were never really alternatives.",
   "how_to_test": "Genetic clearance rather than drugs. Does rapamycin still extend lifespan and healthspan in INK-ATTAC mice already cleared of p16-positive cells (Baker, Nature 2016, PMID 26840489)? Full additivity means senescence suppression contributes little; partial occlusion measures the share it does contribute. Pair it with a tissue readout: is rapamycin's SASP suppression lost in Atg7-deficient tissue, as the TASCC result predicts, or enhanced, as the GATA4/p62 arm predicts? That second test is what tells the two couplings apart.",
   "studies": [
    "DEM2009",
    "BLA2006",
    "LAB2015",
    "CAS2009",
    "LV2026",
    "CHU2019",
    "PYO2013"
   ],
   "revision_log": [
    "2026-08-30: The basis asserted that 'external work shows rapamycin still suppresses SASP in Nrf2-KO fibroblasts WITHOUT activating autophagy'. No study in the corpus supported it and no PMID was recorded, so the claim was withdrawn.",
    "2026-09-04: 'no study tests whether...' scoped to this corpus.",
    "2026-09-29: The MK2/ZFP36L1 mechanism had been attributed to LAB2015; LAB2015 shows IL1A translation, and MK2/ZFP36L1 is Herranz 2015."
   ],
   "url": "https://mtor-atlas.org/question/suppression-of-senescence-sasp-geroconversion-not-just-autophagy-is-the-mediator-of-rapamycins-healthspan-benefit/",
   "api_url": "https://mtor-atlas.org/api/v1/questions/H10.json"
  },
  {
   "id": "H2",
   "kind": "open-question",
   "title": "mTORC1-selective (mTORC2-sparing) dosing captures longevity without insulin resistance",
   "category": "Contradiction / tension",
   "confidence": 0.85,
   "evidence_stands_at": "Clinical outcome in humans (oncology); no ageing endpoint",
   "gap": "The lifespan benefit of rapamycin is strongly linked to mTORC1 inhibition, whereas chronic disruption of mTORC2 is a major mechanism behind the insulin resistance (LAM2012). The separation is well supported but NOT exhaustive: this Atlas's own S6K1-IRS1 edge records that the contribution of the S6K1->IRS-1 feedback arm to rapalog dysglycaemia in humans has never been apportioned against mTORC2 loss, so 'all benefit = mTORC1, all harm = mTORC2' overstates what has been shown. QUANTIFIED (full text): ARR2015 - daily 2 mg/kg rapamycin impaired glucose tolerance (+71% AUC) and inhibited BOTH mTORC1 and mTORC2 (AKT-S473), while weekly / every-5-days dosing caused NO glucose impairment and preserved mTORC1 inhibition with no detectable mTORC2 effect -> intermittent dosing decouples benefit from harm in vivo. Wording matters here: 'no detectable mTORC2 inhibition under this assay' is not the same claim as 'mTORC2 was not inhibited'. Dose-response MIL2014 (4.7/14/42 ppm; top dose +23% M / +26% F). PEARL tested 5 vs 10 mg/week but its primary endpoint was visceral fat (missed) - no mTORC1-selective human aging readout exists. Early clinical evidence now exists outside the rapamycin/rapalog class. Bi-steric mTORC1-selective inhibitors (>25-fold selectivity over mTORC2) were validated preclinically (MEN2023: RMC-6272 overcomes drug resistance via complete mTORC1 suppression, sparing mTORC2) and clinically (SCH2025: RMC-5552 Phase 1, n=57 solid tumors) -- treatment-related hyperglycemia was only 4%, which is consistent with the hypothesis that sparing mTORC2 reduces the glucose-intolerance signature seen with chronic rapamycin/rapalogs. Because SCH2025 was open-label, single-arm and had no rapamycin or ATP-site comparator, the low hyperglycaemia rate is a supportive signal, not a causal attribution: it is compared against historical rapalog experience, not against a randomised control. This is the strongest human evidence yet for the mechanistic premise of H2, but it is preliminary, and it comes from an oncology trial, not an aging/longevity endpoint -- the missing link is still a healthspan/lifespan readout for a selective inhibitor.",
   "gap_beginner": "The lifespan benefit of blocking mTOR is closely tied to mTORC1, while the main side effect (insulin resistance) comes largely from also disrupting mTORC2. That split is well supported, but it is not the whole story - a feedback loop through S6K1 also contributes to the blood-sugar problem, and nobody has worked out how much each one matters in people. In mice, giving rapamycin less often (weekly instead of daily) avoided the glucose problems while keeping the mTORC1 effect, with no mTORC2 effect detectable - which is not quite the same as saying mTORC2 was untouched. Newer drugs aimed only at mTORC1 showed a low rate of blood-sugar side effects in an early human cancer trial: encouraging, but that trial had no comparison group and was not about ageing.",
   "changed_since_written": "2025: a bi-steric mTORC1-selective inhibitor reached humans. In a phase 1 trial of RMC-5552 (Schram, Clin Cancer Res 2025, PMID 41056387, n=57) treatment-related hyperglycaemia occurred in 4% and was not dose-limiting — but mucositis occurred in 49%, so the dose-limiting toxicity moved rather than disappeared. The mechanistic premise also firmed up: bi-sterics reactivate 4E-BP1, which rapalogs cannot, while sparing mTORC2 (Mahauad-Fernandez, Cell Chem Biol 2025, PMID 40803322). No lifespan data exist for any bi-steric in any species.",
   "still_open": "Two things this question had merged. First, does a bi-steric extend lifespan in genetically heterogeneous mice? Untested, and not in any ITP cohort. Second, is rapamycin's glucose intolerance costing anything at all? Lamming 2012 already reported lifespan extension as separable from glucose homeostasis (in female mTOR/mLST8 double-heterozygous mice), and the highest ITP rapamycin dose produces both the largest lifespan gain and glucose intolerance — so the penalty this question proposes to remove may be a biomarker rather than a cost. A further complication: in male mice, lifespan extension by acarbose and 17-alpha-estradiol travels with raised hepatic mTORC2 signalling (Garratt, Aging Cell 2017, PMID 28834262), so mTORC2 tone is not purely a side-effect channel.",
   "hypothesis": "An intervention that suppresses the mTORC1/4E-BP1 arm while sparing mTORC2 -- now achievable with bi-steric mTORC1-selective inhibitors (RMC-6272/RMC-5552) in addition to intermittent/low-dose rapamycin -- yields the lifespan benefit without glucose intolerance.",
   "hypothesis_beginner": "Educated guess: a drug (or dosing schedule) that blocks mTORC1 while leaving mTORC2 alone should give the longevity benefit of rapamycin without the blood-sugar downside.",
   "how_to_test": "Two experiments, not one. (a) Bi-steric (RMC-6272 class) in an ITP-style lifespan protocol at a dose achieving measurable 4E-BP1 dephosphorylation — the actually untested claim. (b) To test whether the metabolic penalty is a cost: normalise glucose tolerance on top of 42 ppm rapamycin, with acarbose or an SGLT2 inhibitor, and ask whether lifespan moves. Note that bi-sterics (deeper, continuous suppression) and intermittent rapamycin (shallower, pulsed) are opposite manipulations and predict opposite things for muscle and immunity; they should not be pooled as one strategy.",
   "studies": [
    "LAM2012",
    "ARR2015",
    "MIL2014",
    "KEN2016",
    "THO2009",
    "MOE2025",
    "MEN2023",
    "SCH2025"
   ],
   "revision_log": [
    "2026-07-09: Bi-steric inhibitor evidence added (revised 2026-07-29).",
    "2026-08-30: The opening sentence previously read 'Longevity comes from mTORC1 but chronic rapamycin's harm comes from mTORC2 disruption', and the ARR2015 clause read 'inhibited mTORC1 only'. Both were more certain than the experiments support, which is a preferential separation.",
    "2026-09-29: The Lamming 2012 result is now qualified as female mice."
   ],
   "url": "https://mtor-atlas.org/question/mtorc1-selective-mtorc2-sparing-dosing-captures-longevity-without-insulin-resistance/",
   "api_url": "https://mtor-atlas.org/api/v1/questions/H2.json"
  },
  {
   "id": "H3",
   "kind": "open-question",
   "title": "Muscle-sparing, pulsed mTORC1 inhibition",
   "category": "Contradiction / tension",
   "confidence": 0.7,
   "evidence_stands_at": "Lifespan in mice; PD in humans",
   "gap": "mTORC1 is required for muscle (raptor-KO dystrophy; DRU2009 blocks human MPS; ROM2001 hypertrophy) yet mTOR inhibition is among the most reproducible pharmacological ways to extend mouse lifespan. QUANTIFIED: BIT2016, HIGH-DOSE ARM (8 mg/kg/day i.p. x 90 days from 20-21 months) -> males +60% post-treatment life expectancy (p=0.02); females NO survival benefit (p=0.261) with a shift toward aggressive haematopoietic cancers (round-cell tumours 16/16 vs 6/12 controls, p=0.002) and away from non-haematopoietic neoplasia. From the full text: the SAME paper's second arm - 126 ppm encapsulated rapamycin in diet, same 90 days, same starting age - significantly increased post-treatment survival in BOTH sexes independently, and a Cox model with robust standard errors found no evidence that sex modified the treatment effect (p=0.904). So 'transient dosing captures the benefit' is supported, and the female harm signal is specific to the high-dose injection regimen rather than a property of pulsed dosing as such. A pulsed / muscle-sparing design must therefore control for dose and route, not only for sex. This is still one experiment in one strain: whether the principle generalises across strain, species, treatment window and endpoint is untested.",
   "gap_beginner": "mTORC1 is needed to build muscle, yet turning it down is also one of the best-studied ways to extend lifespan in animals - a real tension. In one mouse study, a short course of rapamycin extended male life expectancy by 60%. At the high injected dose it did nothing for females and caused more aggressive blood cancers in them - but the same study also tested a lower dose given in food, and that one helped both sexes equally. So pulsed dosing is not automatically a problem for females; the dose and the route matter. It is still just one experiment.",
   "changed_since_written": "The premise is age-inverted. The worry about muscle comes from a single acute dose in young men, measuring protein synthesis over one to two hours (Drummond, J Physiol 2009, PMID 19188252). In aged muscle mTORC1 is already hyperactive and partial inhibition increased mass and fibre cross-sectional area in old rats (Joseph, Mol Cell Biol 2019, PMID 31308131); constitutive myofibre mTORC1 activation is itself a sarcopenia model, rescued by activating 4E-BP1 (Crombie, JCSM 2022, PMID 36398408); and exercise is now framed as normalising mTORC1 downward (Choy, PNAS 2025, PMID 41284871). The only head-to-head schedule comparison found weekly and thrice-weekly rapamycin both preserved hypertrophy, grip strength and exercise capacity — the schedule mattered for glucose, not for muscle (Elliehausen, Aging Cell 2025, PMID 40704394). In humans, 48 weeks of intermittent low-dose rapamycin raised lean mass in women at 10 mg/week (Moel, Aging 2025, PMID 40188830).",
   "still_open": "Not whether mTOR inhibition spares muscle. Two narrower things: whether in aged mammals there is a dose and schedule that improves muscle function rather than only preserving mass, and whether bi-steric inhibitors, which suppress mTORC1 output more deeply than rapamycin, cross from the protective into the atrophic regime. There are no bi-steric skeletal-muscle data at all.",
   "hypothesis": "In aged muscle, where mTORC1 is already elevated, partial inhibition improves contractile function rather than costing it — and any anabolic cost is confined to young or growing muscle.",
   "hypothesis_beginner": "Educated guess: in old muscle the pathway is stuck too far on, so turning it down helps rather than hurts. The worry about losing muscle may only apply to young, growing muscle.",
   "how_to_test": "Partial mTORC1 inhibition started at 18 months or later in genetically heterogeneous mice. Endpoints must be functional — grip strength, contractile force, neuromuscular junction integrity — not mass, because the two dissociate: mice with hyperactive mTORC1 gain mass and mitochondrial activity with no functional gain (Kalenta, J Appl Physiol 2026, PMID 41979886). Counterexample to carry honestly: four weeks of daily rapamycin in young rats did reduce fast-fibre mass independently of energy intake (Ato, PLoS One 2024, PMID 39637031).",
   "studies": [
    "DRU2009",
    "ROM2001",
    "MOE2025",
    "BIT2016"
   ],
   "revision_log": [
    "2026-08-30: Dose qualifier added from the BIT2016 full text (the 126 ppm dietary arm benefited both sexes).",
    "2026-09-29: 'the leading longevity lever' softened."
   ],
   "url": "https://mtor-atlas.org/question/muscle-sparing-pulsed-mtorc1-inhibition/",
   "api_url": "https://mtor-atlas.org/api/v1/questions/H3.json"
  },
  {
   "id": "H4",
   "kind": "open-question",
   "title": "Is autophagy actually REQUIRED for the mammalian lifespan benefit?",
   "category": "Mechanism-to-outcome gap",
   "confidence": 0.7,
   "evidence_stands_at": "Lifespan in mice (sufficiency only)",
   "gap": "Autophagy is assumed to mediate longevity but shown directly only in fly and worm - BJE2010 (rapamycin extends fly lifespan through autophagy and reduced translation) and HAN2008 (blocking autophagy genes abolishes the lifespan extension produced by dietary restriction in C. elegans) - plus Atg5 overexpression in mouse (PYO2013), which shows sufficiency, not necessity. No study IN THIS CORPUS tests whether BLOCKING autophagy abolishes rapamycin's mammalian lifespan extension - the epistasis experiment that would settle it has not been done in a mammal as far as this Atlas holds. Note the shape of the gap: it is specifically MAMMALIAN. The causal experiment does exist in invertebrates, which is why this card cites two of them. Gaps in this Atlas are computed against this corpus, not against PubMed. Note that this particular claim also survives an external check - a targeted PubMed query for an autophagy-deficient x rapamycin x mammalian lifespan experiment returns no records - but the wording should still say what it can support.",
   "gap_beginner": "Autophagy (the cell's self-cleanup process) is widely assumed to be why blocking mTOR extends lifespan, but that's only been directly shown in flies and worms, plus one mouse study that boosted autophagy a different way. Nobody has tested what happens to rapamycin's lifespan benefit in mice that can't do autophagy at all.",
   "changed_since_written": "Nothing has answered the question, but the experiment as proposed turns out not to be runnable. Inducible whole-body Atg7 deletion in an adult mouse limits survival to two or three months through neurodegeneration and infection, plus fatal fasting hypoglycaemia (Karsli-Uzunbas, Cancer Discov 2014, PMID 24875857); a lifespan study needs about three years. A null result would be unreadable — no benefit, or an animal that died of something rapamycin cannot fix — and chronic rapamycin also inhibits mTORC2, so failure in an Atg7 knockout would not implicate autophagy cleanly either. In flies, rapamycin already acts through autophagy and translation in parallel (Bjedov, Cell Metab 2010, PMID 20074526), so the expected mammalian answer is partly, which a survival curve cannot resolve.",
   "still_open": "Necessity, still untested in any mammal. The mammalian evidence runs one way only: raising autophagy extends life (Becn1-F121A knock-in, Fernandez, Nature 2018, PMID 29849149; Atg5 overexpression, +17.2% median, Pyo, Nat Commun 2013, PMID 23939249). Sufficiency is shown; necessity is not.",
   "hypothesis": "Autophagy is necessary: rapamycin will FAIL to extend lifespan in autophagy-deficient (e.g. inducible Atg7-KO) mice.",
   "hypothesis_beginner": "Educated guess: autophagy is actually required – mice that can't perform autophagy should get no lifespan benefit from rapamycin, even though the drug still works everywhere else.",
   "how_to_test": "Occlusion rather than ablation. Does rapamycin still extend lifespan in Becn1-F121A mice, which already carry maximal basal autophagy and are healthy? Full additivity means autophagy carries little of rapamycin's benefit; occlusion means it carries that much. No sick animals are needed and the line already exists. Supporting designs: tissue-restricted Atg7 deletion, which survives into old age (Yuan, Aging Cell 2020, PMID 32212304), and autophagy-hypomorphic rather than null backgrounds. One caveat on the Becn1 line: its longevity phenotype is abolished by a chronic high-phosphate diet (Shi, FASEB J 2020, PMID 31908069), so diet must be controlled.",
   "studies": [
    "BJE2010",
    "HAN2008",
    "PYO2013"
   ],
   "revision_log": [
    "2026-09-04: The basis previously read 'No study tests whether...', an absence claim about the whole literature; now scoped to this corpus.",
    "2026-09-21: MEL2003 was cited as the worm evidence. It tested bec-1 in daf-2 mutants (the insulin/IGF-1 pathway, not TOR and not dietary restriction), so it did not support the claim; HAN2008 replaced it."
   ],
   "url": "https://mtor-atlas.org/question/is-autophagy-actually-required-for-the-mammalian-lifespan-benefit/",
   "api_url": "https://mtor-atlas.org/api/v1/questions/H4.json"
  },
  {
   "id": "H5",
   "kind": "open-question",
   "title": "A sensor/autophagy biomarker panel as a surrogate endpoint for human trials",
   "category": "Human-endpoint gap",
   "confidence": 0.7,
   "evidence_stands_at": "Pharmacodynamics in humans",
   "gap": "QUANTIFIED human-endpoint desert: every completed human trial in this corpus with a clinical or functional aging/healthspan PRIMARY endpoint missed it or was safety-only. The exception is immune-response surrogates, which were met: MAN2014 (influenza-vaccine response, about 20%) and the MAN2018 phase 2a (self-reported infections) - which is why H7 reports positive human data while this card reports none on hard outcomes. PEARL (MOE2025): visceral fat NOT met (secondary bone density OR 0.24, 95% CI 0.06-0.93, p=0.04). RTB101 phase 3 (MAN2021): symptomatic-RTI primary NOT met. Resveratrol (POU2013): fully negative. CALERIE (ROM2016): the paper held here is the SAFETY analysis of the two-year randomised caloric-restriction trial in 218 non-obese adults - CR was safe and well tolerated overall, but bone density fell significantly more than in controls, so 'no significant differences' would be wrong. Scope note: CALERIE is a caloric-restriction trial, not an mTOR-inhibitor trial; it belongs here as a geroscience comparator, not as direct evidence on an mTOR intervention. The human 'wins' are disease indications (BOLERO-2 / BAS2012: PFS roughly doubled with exemestane, overall survival not significantly improved; RADIANT-3 / YAO2011: median PFS 11.0 vs 4.6 months, HR 0.35, p<0.001) or surrogate biomarkers (CHU2019: topical rapamycin epidermal p16 reduced, p=0.008). Not 'few trials' but none with a hard aging endpoint. A candidate real-time readout tool for the mTORC1-vs-mTORC2 arm of the panel now exists -- AIMTOR (BOU2020), a genetically encoded BRET biosensor validated in human and mouse cells that reads mTOR activity live, including subcellular resolution (lysosome/cytosol/nucleus/mitochondria). This does not fill the human-endpoint desert itself, but it derisks the panel's feasibility: the biosensor arm is no longer a proposed capability, it is a published, working assay.",
   "gap_beginner": "Among the human trials collected in this Atlas, none has shown a clinical or functional healthspan benefit: trials aimed at such outcomes either missed their main goal or only checked safety. The goals that were met are immune-response measures, such as a better flu-vaccine response in older adults, not proof of healthier or longer lives. The wins that do exist are for specific diseases (like certain cancers), not for ageing itself. Testing ageing directly takes decades, so researchers want a faster stand-in measurement, and a tool that can track mTOR activity live inside human cells now exists – making that stand-in more realistic to build, even though it hasn't been done yet.",
   "changed_since_written": "The field has explicitly separated three things this question merges: a target-engagement marker, a biomarker of ageing, and a validated surrogate endpoint (Moqri, Cell 2023, PMID 37657418). The 2025 Longevity Biotechnology Association recommendations for trial biomarkers list clocks, omics and functional measures, and do not include a pathway-activity panel (Herzog, npj Aging 2025, PMID 41436473). The current answer to shortening geroscience trials is hierarchical clinical composites with biomarkers placed at the lowest tier (Abdellatif, Nature Aging 2026, PMID 42342910), alongside organ-specific proteomic clocks (Goeminne, Cell Metab 2024, PMID 39488213). On this panel's own components: there is no human blood assay for SAMTOR or GATOR state, and the first measurement of physiological autophagic flux in a large human trial reached significance only post hoc (Bensalem, J Physiol 2025, PMID 40345145, n=121, six months).",
   "still_open": "Split in two, because the halves have different answers. As a pharmacodynamic readout for choosing dose and schedule, an mTORC1 against mTORC2 substrate panel is tractable and trials are running. As a surrogate endpoint it is not: surrogacy requires showing the marker mediates clinical benefit across intervention classes, and no ageing biomarker of any kind — pathway panel, epigenetic clock or organ clock — has met that bar. The consensus papers say so themselves.",
   "hypothesis": "Two hypotheses. Tractable: a blood mTORC1/mTORC2 substrate-phosphorylation panel plus an autophagic-flux assay can select dose and schedule for mTOR-directed geroprotectors in humans. Still negative: that any ageing biomarker satisfies formal surrogate-endpoint criteria sufficient for regulatory acceptance.",
   "hypothesis_beginner": "Two different claims that were being treated as one. The easier one: blood tests can show the drug is hitting the pathway, which helps pick a dose. The hard one, still unmet: a blood test that stands in for living longer or healthier, well enough that a regulator would accept it instead of waiting.",
   "how_to_test": "Back-test the panel across the animal longevity studies in the corpus (does it separate lifespan-extending from null interventions?), then validate prospectively against EVERLAST's daily-vs-weekly arms. The mTORC1-vs-mTORC2 substrate-phosphorylation arm can now build directly on AIMTOR (BOU2020) rather than needing de novo assay development.",
   "studies": [
    "MOE2025",
    "MAN2021",
    "POU2013",
    "ROM2016",
    "CHU2019",
    "BAS2012",
    "YAO2011",
    "LEE2024",
    "BOU2020"
   ],
   "revision_log": [
    "2026-07-09: AIMTOR (BOU2020) biosensor added as a candidate readout.",
    "2026-08-30: The basis previously cited RADIANT-3 (YAO2011) as 'PFS 5% vs 2%'. Those were the grade 3/4 hyperglycaemia rates from the adverse-event table, not a progression-free-survival result.",
    "2026-09-04: Trial acronyms paired with Atlas study codes inline.",
    "2026-09-23: The hypothesis was split in two (it was previously one).",
    "2026-09-29: 'Every trial missed' limited to clinical and functional healthspan endpoints; the immune-response surrogates that were met (MAN2014, MAN2018) are now named."
   ],
   "url": "https://mtor-atlas.org/question/a-sensor-autophagy-biomarker-panel-as-a-surrogate-endpoint-for-human-trials/",
   "api_url": "https://mtor-atlas.org/api/v1/questions/H5.json"
  },
  {
   "id": "H6",
   "kind": "open-question",
   "title": "Sex dimorphism in mTOR-longevity responses is pervasive and sometimes direction-flipping",
   "category": "Contradiction / tension",
   "confidence": 0.8,
   "evidence_stands_at": "Lifespan in mice",
   "gap": "Emerged directly from the full-text effect sizes: S6K1-KO +19% median lifespan in FEMALES / no effect in males (SEL2009); rapamycin favors females (MIL2014 all doses in F vs top-2 in M; HAR2009 +38% F vs +28% M; FOK2014 +16% F vs +11% M), and so, more narrowly, does a genetic mTOR hypomorph (WUX2013 +22% F/+19% M); yet acarbose favors MALES (+22% vs +5%, HAR2014). Sex changes not just the magnitude but sometimes the DIRECTION of the response. BIT2016 needs a qualifier: its male-benefit / female-harm result holds only for the high-dose 8 mg/kg/day injection arm. In the same paper's 126 ppm dietary arm, post-treatment survival rose significantly in both sexes and a Cox model found no evidence that sex modified the treatment effect (p=0.904). BIT2016 therefore supports DOSE-dependent dimorphism, not intrinsic dimorphism, and is as much a boundary case for this hypothesis as evidence for it. The thesis stands on SEL2009, MIL2014, HAR2009, FOK2014, WUX2013 and HAR2014.",
   "gap_beginner": "Male and female animals often respond differently to these drugs – sometimes just by how much, but sometimes the effect flips direction entirely. Rapamycin tends to help females somewhat more; a different drug, acarbose, helps males more. Females also build up higher drug levels in their blood after the same dose, which may explain part of it.",
   "changed_since_written": "The pharmacokinetic explanation was never a tested claim. It was one speculative line in a discussion — lifespan extension was larger in females at each dose, perhaps reflecting sexual dimorphism in blood levels (Miller, Aging Cell 2014, PMID 24341993). What is demonstrated instead is endocrine: gonadectomy flips drug responses in both directions, castrated males losing most responses and ovariectomised females gaining male-like ones (Garratt, Aging Cell 2017, PMID 28834262); 17-alpha-estradiol improves sarcopenia in intact but not castrated males at the same dose and exposure (Garratt, Aging Cell 2019, PMID 30740872); and it engages the pathway in males only, whereas rapamycin and acarbose do so in both sexes (Shen, Aging Cell 2021, PMID 33742521). Newer work adds something the card did not anticipate: the direction flips by endpoint, not only by drug. Rapamycin favours females for lifespan and motor decline (Singh, J Gerontol A 2026, PMID 41863332), acarbose and 17-alpha-estradiol favour males for lifespan — yet all three favour females for trabecular bone (Dall'Ara, bioRxiv 2026, preprint, not peer reviewed).",
   "still_open": "Two narrow questions, neither of which is pharmacokinetics against pharmacodynamics. Does the female advantage in rapamycin lifespan survive titrating both sexes to equal whole-blood trough concentrations? Nobody has run this, and it is the only falsifiable form of the PK claim. And is the governing variable gonadal hormone milieu rather than chromosomal sex? The claim that exposure-matching would narrow the gap for rapamycin but not for acarbose is untested in either direction and should not be read as established.",
   "hypothesis": "The governing variable is gonadal hormone milieu rather than chromosomal sex, and which sex benefits depends on the endpoint measured as well as on the drug.",
   "hypothesis_beginner": "Educated guess: what decides the difference is not being male or female as such, but which sex hormones are circulating — and the answer can flip depending on whether you measure lifespan, bone or movement.",
   "how_to_test": "Two experiments. First the pharmacokinetic test nobody has run: titrate males and females to equal whole-blood rapamycin trough concentrations and re-measure lifespan. Second, four-core-genotype mice plus gonadectomy with and without hormone replacement, across rapamycin, acarbose and 17-alpha-estradiol, separating chromosomal sex from hormonal environment. Measure more than one endpoint in the same animals — survival, bone and motor function — because they do not point the same way.",
   "studies": [
    "SEL2009",
    "MIL2014",
    "HAR2009",
    "FOK2014",
    "WUX2013",
    "BIT2016",
    "HAR2014"
   ],
   "revision_log": [
    "2026-08-30: BIT2016 qualified by dose arm: the female harm signal belongs to the high-dose injection arm only.",
    "2026-09-29: WUX2013 is a genetic mTOR hypomorph and is no longer grouped with the rapamycin studies."
   ],
   "url": "https://mtor-atlas.org/question/sex-dimorphism-in-mtor-longevity-responses-is-pervasive-and-sometimes-direction-flipping/",
   "api_url": "https://mtor-atlas.org/api/v1/questions/H6.json"
  },
  {
   "id": "H7",
   "kind": "open-question",
   "title": "Low-dose mTOR inhibition rejuvenates the aging immune system without net immunosuppression",
   "category": "Contradiction / tension",
   "confidence": 0.75,
   "evidence_stands_at": "Clinical outcome in humans — negative",
   "gap": "mTOR inhibition is the textbook immunosuppressant - rapalogs are licensed for transplant-rejection prophylaxis and transplantation is the largest human exposure dataset for the class. Sourcing note: HEI1991 is the yeast paper that identified TOR1/TOR2 through rapamycin-induced G1 arrest in S. cerevisiae. It establishes the cell-cycle mechanism, NOT the clinical immunosuppression claim, which currently has no dedicated citation in this corpus. Yet the SAME pathway at LOW dose is one of the few mTOR interventions with POSITIVE human data. QUANTIFIED: MAN2014 - 6 weeks low-dose RAD001 (everolimus) in elderly IMPROVED influenza-vaccine antibody titers (~20% rise vs placebo) and reduced exhausted PD-1+ CD4/CD8 T cells. MAN2018 - the phase 2a trial: six weeks of a low-dose combination of BEZ235 (RTB101) plus everolimus was associated with fewer self-reported infections over the following year and upregulated interferon-induced antiviral genes. The phase 2b result - fewer laboratory-confirmed respiratory tract infections, with only the 10 mg/day RTB101 dose effective - is reported in MAN2021, together with the phase 3. But MAN2021 (phase 3 PROTECTOR-1) MISSED its primary endpoint (clinically symptomatic respiratory illness NOT reduced) - so the human immune 'win' collapsed at scale. Mechanistic anchors. ARA2009 (Nature 2009) is direct SUPPORT for this hypothesis, not a counterweight: rapamycin had immunostimulatory effects on memory CD8 T-cell generation in mice and in non-human primates, increasing both the quantity and the quality of the memory response, and acting cell-intrinsically through mTORC1. The genuine counterweights are ZEN2013 (mTORC1 is REQUIRED for Treg function, so over-inhibition would blunt regulatory immunity) and the transplant dose range itself; LV2026 shows age-related Lamtor5 loss itself drives immunosenescence. The unresolved tension remains: the dose that rejuvenates adaptive immunity and the dose that suppresses it are not separated.",
   "gap_beginner": "High-dose, long-term mTOR blocking is a classic immune-suppressing treatment, used to stop organ-transplant rejection. But oddly, a low dose of a similar drug improved older adults' response to a flu vaccine and reduced infections in earlier trials. A larger follow-up trial, though, didn't confirm a real-world benefit against symptomatic respiratory illness. So there may be a \"sweet spot\" dose that helps rather than suppresses immunity. One drug was taken all the way to a large trial on that idea and it did not work, and newer trials suggest the trade-off may not be a single dial at all.",
   "changed_since_written": "The hypothesis was tested at phase 3 scale and failed. RTB101 monotherapy for 16 weeks in adults aged 65 and over, double-blind and placebo-controlled across 54 sites, did not reduce clinically symptomatic respiratory tract infections (Mannick, Lancet Healthy Longev 2021, PMID 33977284); the programme ended there. One nuance matters: the 2018 positive result was a combination of BEZ235 and RAD001 in a different population, so the negative is decisive for RTB101 monotherapy in the general elderly, not for the mechanism. Newer data also cut against the no-immunosuppression half. Switching transplant recipients to everolimus left neutralising antibody titres unchanged, lowered spike-specific T-cell responses and raised bacterial infections from 11.1% to 27.3% in the same trial (Messchendorp, Clin Infect Dis 2026, PMID 40231961). Weekly sirolimus at 6 mg plus home exercise in adults 65 to 85 came out favouring placebo on the primary functional endpoint, with 99 against 63 adverse events and one possibly drug-related pneumonia (Stanfield, JCSM 2026, PMID 41985884) — and that dose sits inside the supposedly safe window. PEARL established tolerability over 48 weeks but its primary endpoint was null and it carried no immune endpoint (Moel, Aging 2025, PMID 40188830).",
   "still_open": "Net immunosuppression is the wrong axis. The effect is dissociative rather than scalar: the same trial can show raised antibody titres alongside lowered antigen-specific T-cell responses and more bacterial infections. What remains open is whether any dose or schedule gives a net favourable balance read per compartment — humoral against cellular, viral against bacterial against fungal — and whether the 2018 combination approach, untested since 2021, behaves differently from monotherapy. Reversal of T-cell exhaustion, which this card asserts, has never been the primary endpoint of a human trial of low-dose mTOR inhibition; MAN2014 reported fewer PD-1-positive T cells only as a supporting readout.",
   "hypothesis": "There is a dose and schedule of intermittent mTORC1 inhibition that improves humoral and antiviral responses in older adults without a net infection cost. For RTB101 monotherapy at phase 3 scale the answer is already no; for combination TORC1 inhibition and other schedules it is untested since 2021.",
   "hypothesis_beginner": "Educated guess, now partly tested and partly not: there may be a low dose that wakes up an old immune system without leaving it exposed. One drug was tried in a large trial and did not work. Whether a different dose, schedule or drug combination could is still unknown — and the newer data suggest the trade-off may not be a single dial at all.",
   "how_to_test": "Dose-ranging elderly RCT with a validated immune-function primary endpoint (T-cell exhaustion markers + neoantigen/vaccine seroconversion + lab-confirmed infection incidence), testing several low/intermittent rapalog schedules against placebo, with paired PBMC mTORC1 vs mTORC2 phosphosignatures to map the rejuvenation-vs-suppression crossover. Prediction: an inverted-U where mid-low doses maximize vaccine response and minimize infections, with immunosuppression only above a definable trough.",
   "studies": [
    "MAN2014",
    "MAN2018",
    "MAN2021",
    "HEI1991",
    "ARA2009",
    "ZEN2013",
    "LV2026",
    "LEE2024"
   ],
   "revision_log": [
    "2026-08-30: The basis previously read 'ARA2009 mTOR drives memory-CD8 differentiation ... so over-inhibition would blunt immunity', which reversed the paper's finding; ARA2009 supports this hypothesis.",
    "2026-09-23: The basis previously ended 'and no human trial has hit a durable clinical immune endpoint'. A phase 3 trial did test one and missed it, which is a negative result rather than an absence of one.",
    "2026-09-29: The phase 2b result (laboratory-confirmed infections, 10 mg RTB101) was attributed to MAN2018; it is reported in MAN2021. MAN2018 is the phase 2a combination trial with self-reported infections."
   ],
   "url": "https://mtor-atlas.org/question/low-dose-mtor-inhibition-rejuvenates-the-aging-immune-system-without-net-immunosuppression/",
   "api_url": "https://mtor-atlas.org/api/v1/questions/H7.json"
  },
  {
   "id": "H8",
   "kind": "open-question",
   "title": "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",
   "category": "Mechanism-to-outcome gap",
   "confidence": 0.7,
   "evidence_stands_at": "Clinical outcome in humans — negative on target engagement",
   "gap": "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 and failed upstream of efficacy rather than at it, one uninformative disease-specific human trial, and an unquantified balance between clearing aggregates (mTORC1/autophagy) and preserving plasticity (mTORC2).",
   "gap_beginner": "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.",
   "changed_since_written": "Two human trials have now reported, and both failed upstream 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). 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 randomised 12-month phase 2 with cognitive secondary endpoints reads out in 2026 (REACH, NCT04629495, n=40).",
   "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.",
   "hypothesis": "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 has now failed twice at geroscience doses and must be demonstrated before the second half can be tested.",
   "hypothesis_beginner": "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, the drug does not measurably get into the brain at all.",
   "how_to_test": "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.",
   "studies": [
    "SPI2010",
    "ZHO2009",
    "CAC2010",
    "EHN2008",
    "TAN2024",
    "HAL2012",
    "SAX2026",
    "HUA2013",
    "LAM2012",
    "SAR2006",
    "LEE2024"
   ],
   "revision_log": [
    "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."
   ],
   "url": "https://mtor-atlas.org/question/brain-penetrant-mtor-inhibition-clears-neurodegenerative-aggregates-in-models-but-has-zero-human-cognitive-aging-endpoint-and-mtorc2-is-needed-for-memory/",
   "api_url": "https://mtor-atlas.org/api/v1/questions/H8.json"
  },
  {
   "id": "H9",
   "kind": "open-question",
   "title": "Rapamycin + AMPK-axis drugs (acarbose / metformin): at least additive in male mice, untested in humans",
   "category": "Evidence desert",
   "confidence": 0.7,
   "evidence_stands_at": "Lifespan in mice (answered); no human data",
   "gap": "Animal combination data are suggestive but the human evidence for combining mTOR inhibition with AMPK-axis geroprotectors is a desert. QUANTIFIED (STR2022, NIA Interventions Testing Program C2017 cohort): rapamycin + acarbose started at 9 months extended MALE lifespan beyond what either of the two prior ITP rapamycin-only cohorts achieved, while in FEMALES the combination was neither better nor worse than rapamycin alone - the authors relate this to the limited survival benefit acarbose alone had shown in earlier female cohorts. CAVEAT: STR2022's rapamycin-only comparison is HISTORICAL, against prior cohorts rather than a concurrent rapamycin-only arm, so 'more potent than either component' is the authors' suggestion, not a within-experiment randomised result. That is precisely why this gap remains open. HAR2014 established acarbose's male-skewed lifespan effect. The AMPK arm converges back onto mTOR mechanistically: ZHO2001 metformin acts via AMPK, GWI2008 AMPK phosphorylates raptor to gate mTORC1, HOW2017 metformin inhibits hepatic mTORC1 dose-dependently via AMPK - which raises the opposite possibility that metformin + rapamycin are partly REDUNDANT on the mTORC1 node. Human signal exists only for observational data (BAN2014: type-2 diabetics on metformin outlived non-diabetic controls - a result subject to the usual immortal-time and prescription-bias caveats of the metformin-survivorship literature, and not treated here as causal). No study IN THIS CORPUS - animal or human - cleanly separates ADDITIVE (distinct pathways) from REDUNDANT (same node) for these pairs, and this corpus holds no human combination trial. Gaps here are computed against this Atlas, not against PubMed. LOGIC NOTE: distinct proximal mechanisms do not entail additivity. Two drugs can hit different targets and still converge downstream, interact non-linearly, or show sex- and strain-dependent interaction - which is exactly what the ITP combination data themselves show.",
   "gap_beginner": "In mice, combining rapamycin with a diabetes drug called acarbose extended life more than either drug alone - but only in males; in females it was no better than rapamycin by itself. A different diabetes drug, metformin, reaches mTORC1 partly through AMPK, the cell's fuel sensor. It therefore presses on the same checkpoint that rapamycin blocks, only from another direction, so combining them might not add much. But be careful with that reasoning: two drugs hitting different targets do not automatically add up - they can still overlap further downstream, or interact in unexpected ways. Nobody in this Atlas's collection has run a human trial combining any of these.",
   "changed_since_written": "The mouse half has not been open since 2016, outside this corpus. The NIA Interventions Testing Program tested metformin with rapamycin (Strong, Aging Cell 2016, PMID 27312235) and rapamycin with acarbose (Strong, Aging Cell 2022, PMID 36179270); the acarbose combination gave male mice longer lifespan than rapamycin alone, with no comparable benefit in females. Both conclusions rest on historical comparison against earlier rapamycin-only cohorts rather than a contemporaneous factorial, so at least additive in males is defensible and synergistic is not. A three-drug combination also beat every single agent on healthspan measures in 20-month-old mice (Jiang, Sci Rep 2022, PMID 35508491). The category itself is also questionable: acarbose is an intestinal alpha-glucosidase inhibitor with a strong male bias, and low-dose metformin reaches AMPK through the lysosomal PEN2 route rather than complex I (Ma, Nature 2022, PMID 35197629), so grouping them as the AMPK axis imports a shared mechanism neither clearly uses at geroprotective doses.",
   "still_open": "The human half, entirely: no trial has tested rapamycin with an AMPK-axis or glucose-lowering drug against either agent alone on any ageing endpoint, and TAME remains unfunded with no rapamycin arm. A more consequential question has appeared alongside it. Metformin abolished the VO2max gain from 16 weeks of training and blunted training-induced vascular insulin sensitivity in 91 adults (Malin, JCEM 2026, PMID 41160096). Since exercise is the best-evidenced human geroprotective intervention, an AMPK-axis drug may subtract from a stack rather than add to it. For contrast, trametinib and rapamycin combine additively for mouse lifespan in both sexes (Gkioni, Nature Aging 2025, PMID 40437307) — combination with dose-lowering may be a more tractable route to benefit without toxicity than engineering complex selectivity.",
   "hypothesis": "In mice the answer is in: at least additive in males for rapamycin with acarbose, with no comparable female benefit, on cross-cohort rather than factorial evidence. What is hypothesised and untested is the human case — and the live alternative, raised by the exercise data, is that AMPK-axis drugs antagonise other geroprotective interventions in people rather than complementing them.",
   "hypothesis_beginner": "In mice this has been answered: pairing rapamycin with acarbose did better than rapamycin alone in males, though not in females. In people nothing has been tried. And there is a warning sign — in a recent human trial metformin cancelled out the fitness gains from exercise training, so adding it to a stack may take away rather than add.",
   "how_to_test": "Factorial mouse lifespan/healthspan study (vehicle / rapamycin / acarbose / metformin / rapamycin+acarbose / rapamycin+metformin), sex-stratified, with mTORC1 (S6K/4E-BP1) and AMPK (raptor-Ser792) readouts plus glucose tolerance. Test for statistical super-additivity. Prediction: rapamycin+acarbose is additive on lifespan; rapamycin+metformin is sub-additive on lifespan but restores glucose tolerance. Follow with a human factorial healthspan-biomarker trial on the additive pair.",
   "studies": [
    "HAR2014",
    "ZHO2001",
    "GWI2008",
    "HOW2017",
    "BAN2014",
    "ARR2015",
    "MOE2025",
    "STR2022"
   ],
   "revision_log": [
    "2026-08-30: The basis previously read 'No study - animal or human -', an absence claim about the whole literature; now scoped to this corpus.",
    "2026-09-05: The ITP combination publication, previously flagged as missing, was entered as STR2022 (PMID 36179270).",
    "2026-09-23: Title changed after the literature review (URL kept).",
    "2026-09-29: Title changed from 'answered in mice' to 'at least additive in male mice'; the beginner text no longer says metformin and rapamycin share AMPK."
   ],
   "url": "https://mtor-atlas.org/question/rapamycin-ampk-axis-drugs-acarbose-metformin-potentially-additive-redundant-or-synergistic-and-untested-as-a-combination-in-humans/",
   "api_url": "https://mtor-atlas.org/api/v1/questions/H9.json"
  },
  {
   "id": "F1",
   "kind": "frontier",
   "title": "Does mTORC1 carry information in the pattern of its activity over time, not its average level?",
   "category": "Framing gap",
   "evidence_stands_at": "Mechanism in cells",
   "gap": "Almost the entire literature reports mTORC1 as a level — phospho-S6K or phospho-4E-BP1 at one time point, in a lysate, after a step change in nutrients. But a real cell never sees a step: it sees amino-acid pulses after meals, insulin pulses, growth-factor pulses, mechanical load during exercise, and a circadian rhythm underneath all of it. If the pathway responds to the shape of that input rather than to its mean, then a single-time-point measurement is not an underpowered measurement of the right thing. It is a measurement of the wrong thing.",
   "gap_beginner": "Nearly every study measures mTOR activity the way you would read a thermometer once: a single number, from a single moment, usually right after the cell was given a sudden jolt of nutrients. But real cells never get a sudden jolt. They get meals, exercise, sleep and waking — a rhythm. If what matters is the rhythm rather than the average, then the standard measurement is not just imprecise. It is answering a different question from the one being asked.",
   "changed_since_written": "Two things made this askable rather than merely sayable. Live single-cell reporters of mTORC1 activity now resolve minutes rather than hours, so the shape of a response can be observed rather than inferred from lysates. And the 2025 structural work showed that activation on the lysosomal membrane is a multi-step spatial process — Rag and Ragulator, then Rheb, then direct membrane engagement of mTOR and Raptor — rather than a single binary switch (Nature 2025, structural basis for mTORC1 activation on the lysosomal membrane). A multi-step process has intermediate states, and intermediate states are what let a pathway distinguish a brief pulse from a sustained one.",
   "still_open": "Whether any downstream output actually reads pulse duration or frequency rather than integrated activity. The cleanest version: do autophagy and cap-dependent translation have different temporal filters, so that the same total mTORC1 activity delivered as one long pulse or several short ones produces different amounts of each? We have not found that comparison published in a mammalian cell.",
   "why_it_matters": "If true, it reframes intermittent dosing from a way of lowering exposure into a way of choosing which outputs you engage, and it puts feeding, exercise and circadian timing inside the pharmacology rather than beside it.",
   "how_to_test": "Optogenetic or chemogenetic control of mTORC1 activity in single cells, delivering matched total activity in different temporal patterns — one sustained pulse against several brief ones — with simultaneous live readouts of autophagic flux and of cap-dependent translation. The discriminating result is a divergence between the two outputs under equal integrated activity. A negative result is equally informative and would justify the field's steady-state convention.",
   "related_questions": [
    "H3",
    "H2"
   ],
   "url": "https://mtor-atlas.org/questions/frontier/does-mtorc1-encode-information-in-dynamics-not-average-activity/",
   "api_url": "https://mtor-atlas.org/api/v1/questions/F1.json"
  },
  {
   "id": "F2",
   "kind": "frontier",
   "title": "Is mTORC1 one pathway or several separable outputs — and which of them carries ageing?",
   "category": "Framing gap",
   "evidence_stands_at": "Lifespan in mice",
   "gap": "The field asks how much mTORC1 to inhibit, and lately which complex to spare. It rarely asks which of mTORC1's outputs needs inhibiting at all. Yet the outputs behave differently: the 4E-BP1 arm and the S6K1 arm separate in muscle, where activating 4E-BP1 rescues a sarcopenia model and deleting S6K1 does not; bi-steric inhibitors reactivate 4E-BP1 where rapalogs cannot; and lysosomal localisation and mTORC1 activity can be uncoupled, so different substrates need not move together.",
   "gap_beginner": "Most of the argument is about how hard to press the brake, and recently about which of the two brakes to press. Hardly anyone asks which of the things the brake controls actually needs slowing. And those things do come apart: in muscle, fixing one downstream branch rescues the problem while removing another does nothing.",
   "changed_since_written": "Substrate selectivity has moved from a curiosity to a stated paradigm shift in the 2026 review literature, on the back of results showing that where mTORC1 sits and which substrates it reaches are not the same variable as how active it is. On the therapeutic side, bi-steric inhibitors are the first tool that changes the output profile rather than only the output amount (Mahauad-Fernandez, Cell Chem Biol 2025, PMID 40803322), and they have now been through a human phase 1 (Schram, Clin Cancer Res 2025, PMID 41056387).",
   "still_open": "Which outputs are required for geroprotection. No lifespan experiment has compared mTORC1 outputs at matched mTORC1 activity. Single-output genetics exists - S6K1 deletion extends female mouse lifespan (SEL2009) and 4E-BP mediates dietary-restriction lifespan extension in flies (ZID2009) - but there is no 4E-BP1-arm-only ITP cohort, no S6K1-arm-only cohort at matched mTORC1 activity, and no bi-steric lifespan data in any species.",
   "why_it_matters": "It would change what a geroprotective drug is supposed to do. Sparing mTORC2 is a question about avoiding harm; choosing outputs is a question about where the benefit lives, and it may explain why deeper inhibition is not reliably better.",
   "how_to_test": "An ITP-style lifespan protocol comparing rapamycin against a bi-steric at doses matched for total mTORC1 inhibition but differing in 4E-BP1 engagement. If lifespan tracks 4E-BP1 dephosphorylation rather than overall mTORC1 suppression, the target is an output, not the complex. Genetic arm: 4E-BP1 and S6K1 manipulations on a common background with the same endpoints.",
   "related_questions": [
    "H2",
    "H3",
    "H4"
   ],
   "url": "https://mtor-atlas.org/questions/frontier/which-mtorc1-outputs-actually-carry-ageing/",
   "api_url": "https://mtor-atlas.org/api/v1/questions/F2.json"
  },
  {
   "id": "F3",
   "kind": "frontier",
   "title": "Is mTOR signalling decided by where the complex is active, rather than by how much of it is active?",
   "category": "Framing gap",
   "evidence_stands_at": "Mechanism in cells",
   "gap": "The lysosome is usually described as the place mTORC1 sits. The structural work of 2025 describes something else: a signalling platform where Rag and Ragulator, Rheb, and direct membrane engagement have to be assembled in order, and where substrate access depends on that assembly. If access is positional, then two cells with identical total mTORC1 activity can be running different programmes, and no lysate will tell them apart.",
   "gap_beginner": "The lysosome is normally described as mTOR's parking space. Newer structural work describes it more like a workbench: what the enzyme can reach depends on how it is docked. If that is right, two cells with the same amount of activity can be doing completely different jobs — and grinding them up to measure the total would hide exactly that.",
   "changed_since_written": "Structures published in 2025 resolved both the assembly on the lysosomal membrane and the dynamic regulation by amino acids through the GATOR2, Sestrin2 and CASTOR1 interfaces (Nature 2025). The authors are explicit that how nutrient binding is converted into GATOR2 activation remains unresolved — so the mechanism is now detailed enough to generate positional predictions, and those predictions are untested.",
   "still_open": "Whether substrate identity actually depends on location in a living cell, as opposed to in a reconstituted system. The specific unanswered question: does the same amount of active mTORC1 produce a different substrate profile when it is tethered to the lysosome than when it is tethered elsewhere?",
   "why_it_matters": "It decides whether pathway-activity biomarkers can mean anything. A blood panel reports how much, and if the biology runs on where, the panel is measuring a quantity the cell does not use.",
   "how_to_test": "Forced relocalisation. Tether active mTORC1 to the lysosomal membrane, to the trans-Golgi and to the plasma membrane at matched activity, and read the substrate profile at each — S6K1, 4E-BP1, ULK1, TFEB. A differing profile at equal activity establishes that position is an independent variable; an identical profile retires the idea.",
   "related_questions": [
    "H5",
    "H1"
   ],
   "url": "https://mtor-atlas.org/questions/frontier/is-mtor-signalling-decided-by-where-not-how-much/",
   "api_url": "https://mtor-atlas.org/api/v1/questions/F3.json"
  },
  {
   "id": "F4",
   "kind": "frontier",
   "title": "Does the timing of mTOR inhibition have to match the body's own pulses — feeding, exercise, sleep?",
   "category": "Framing gap",
   "evidence_stands_at": "Lifespan in mice",
   "gap": "Intermittent dosing is currently justified as a way to lower exposure and dodge side effects. But if the pathway reads pattern rather than level, the interesting variable is not how often the drug is given but whether the drug's troughs line up with the body's own peaks — the post-meal window, the post-exercise anabolic window, the circadian trough. No lifespan study has ever varied the phase of dosing while holding the dose and the interval constant.",
   "gap_beginner": "Giving the drug once a week instead of daily is usually explained as a way to lower the dose and avoid side effects. But if the pathway cares about rhythm, the question is not how often you give it — it is whether the gaps in the drug line up with the moments the body actually needs the pathway on, like after a meal or after exercise. Nobody has tested that alignment.",
   "changed_since_written": "The one head-to-head schedule comparison in exercising mice found that weekly and thrice-weekly rapamycin both preserved hypertrophy and grip strength, and that the schedule mattered for glucose rather than for muscle (Elliehausen, Aging Cell 2025, PMID 40704394). That is evidence that schedule does something — and evidence that the field's stated reason for choosing a schedule was the wrong one. Intermittent dosing has also now shown lifespan extension in its own right, in female C57BL/6J mice (Arriola Apelo, J Gerontol A 2016, PMID 27091134).",
   "still_open": "Phase, as distinct from frequency. Does the same weekly dose given before against after the feeding window, or before against after exercise, produce different lifespan, muscle and glucose outcomes? This is a cheap experiment by ageing-research standards and it has not been run.",
   "why_it_matters": "It is the practical form of F1 and it is testable now with existing drugs. If phase matters, every current human trial has been dosing blind to a variable that decides the answer.",
   "how_to_test": "A lifespan cohort at one dose and one interval, with dosing phase as the only variable — aligned with the light-phase trough, aligned with the feeding window, aligned post-exercise. Endpoints: lifespan, grip strength and contractile force, glucose tolerance, and mTORC1 and mTORC2 substrate phosphorylation sampled at several times of day rather than once.",
   "related_questions": [
    "H3",
    "H2",
    "H7"
   ],
   "url": "https://mtor-atlas.org/questions/frontier/must-inhibition-match-the-bodys-own-rhythm/",
   "api_url": "https://mtor-atlas.org/api/v1/questions/F4.json"
  },
  {
   "id": "F5",
   "kind": "frontier",
   "title": "Which human ageing phenotypes are causally reversible by mTOR modulation — and is it the same mechanism in each tissue?",
   "category": "Human-endpoint gap",
   "evidence_stands_at": "Clinical outcome in humans — mixed and partly negative",
   "gap": "Does rapamycin extend human lifespan is not a testable question on any useful timescale, and treating it as the goal has let a set of much sharper questions go unasked. Immune ageing, sarcopenia, metabolic dysfunction, neurodegeneration, fibrosis and stem-cell exhaustion are separate phenotypes with separate readouts, and there is no reason to assume one drug reaches them all or reaches them by the same route.",
   "gap_beginner": "Asking whether a drug makes people live longer cannot be answered in any reasonable amount of time, and aiming at it has crowded out better questions. Ageing is not one thing: the immune system, muscle, metabolism and the brain each age in their own way, and a drug might reach some of them and miss others entirely.",
   "changed_since_written": "The human results that have arrived are tissue-specific and they do not agree with each other. A phase 3 trial of low-dose TORC1 inhibition failed to reduce respiratory infections in older adults (Mannick, Lancet Healthy Longev 2021, PMID 33977284). In the brain, one 2025 trial found rapamycin undetectable in cerebrospinal fluid, while CSF p-tau181, GFAP and neurofilament light rose (Gonzales, Commun Med 2025, PMID 40394335), and a second found no change in cerebral glucose metabolism on FDG-PET (Svensson, medRxiv 2025, preprint). Brain exposure was never demonstrated. In muscle, 48 weeks of intermittent dosing raised lean mass in women rather than lowering it (Moel, Aging 2025, PMID 40188830). Same drug, three tissues, three different kinds of answer.",
   "still_open": "Which phenotypes are reachable at tolerable exposure, and whether the mechanism is shared. The pattern in the failures is worth naming on its own: both the immune and the brain trials failed upstream of efficacy, at target engagement, not at mechanism. So the prior question is what counts as demonstrated mTOR target engagement in a given human tissue — and no trial that has failed so far was designed to answer it.",
   "why_it_matters": "It replaces one unanswerable question with several answerable ones, and it explains the run of negative human trials as an exposure problem rather than a refutation of the biology.",
   "how_to_test": "Phenotype-first rather than drug-first. Pick one tissue where target engagement can be demonstrated directly — skin, blood or muscle, all biopsiable — fix the regimen against a measured pharmacodynamic endpoint in that tissue, and only then test a functional outcome. Run the same design across tissues and compare the pharmacodynamic signature that accompanies benefit in each. Whether the signature is shared answers the mechanism half.",
   "related_questions": [
    "H5",
    "H7",
    "H8"
   ],
   "url": "https://mtor-atlas.org/questions/frontier/which-human-ageing-phenotypes-are-causally-reversible/",
   "api_url": "https://mtor-atlas.org/api/v1/questions/F5.json"
  }
 ]
}