How the mTOR field is put together

A pathway map shows how the molecules fit together. This page is the other half: the order in which people found out, who is meeting next, and what nobody has answered.

46 years from the soil sample to the clinic · 5 of 30 studies on that path done in humans · 0 with an ageing endpoint.

Timeline — how the field got here

Click a node to read what it unlocked and what it left open. All 30 milestones are also written out below, so nothing here depends on the graphic.

Oliver's mTOR Atlas · Timeline

How mTOR biology actually grew

Every node is a landmark study. Every edge is a claim about what one study made possible for the next. The vertical axis is real time, so the empty stretches are part of the argument. Nodes carry the Atlas's own evidence tiers: blue is direct human evidence. Count the blue.

This is an editorial reconstruction, not a citation network. Nothing on this canvas is computed from citation data, co-citation, or any bibliometric measure. Each edge is a judgement made by the curator — “study A is what made study B askable” — written down so it can be argued with. Two researchers reading the same papers could reasonably draw a different tree, and the absence of an edge is not evidence that no influence existed. Treat the shape as an interpretation of how the field moved, and the nodes and their metadata (PubMed-verified) as the checkable part.

Drag to pan · scroll/pinch or +/− to zoom

Six moments

A curator's selection, not a consensus list — from a soil sample on a Pacific island to a large trial that missed its primary endpoint.

How we got here

Every entry in this Atlas traces back to a real paper. Here are six moments that built the field — from a soil sample on a Pacific island to a large clinical trial that missed its primary endpoint. These six are a curator's selection, not a consensus list.

1975

A molecule from Rapa Nui

Vézina and colleagues isolate a compound from Streptomyces bacteria in a soil sample from Rapa Nui (Easter Island), with antifungal activity nobody can yet explain. They name it after the island: rapamycin. DOI: 10.7164/antibiotics.28.721

1994

The target gets a name

Sabatini and colleagues purify a rapamycin-binding protein from rat brain and propose it as the mammalian TOR homolog — they name it RAFT1. Brown and colleagues independently report the same protein as FRAP in Nature that same year. The name "mTOR" itself comes a year later, from Sabers and colleagues in 1995. Two decades of "what does this drug even do?" finally has a molecular answer. DOI: 10.1016/0092-8674(94)90570-3

2002–2004

One protein, two complexes

Kim et al. and, independently and simultaneously, Hara et al. discover Raptor, the partner protein that defines mTOR Complex 1 — this Atlas's own RAPTOR-MTORC1 edge cites both. Loewith and colleagues define the TORC1/TORC2 split in yeast the same year. Two years later, Sarbassov et al. discover Rictor in mammalian cells — defining a second, rapamycin-resistant complex, mTORC2. mTOR turns out to have two faces at the molecular level too. DOI: 10.1016/s0092-8674(02)00808-5 · DOI: 10.1016/j.cub.2004.06.054

2009

The first drug shown to extend mammalian lifespan

The NIA Interventions Testing Program, led by Harrison and colleagues, shows rapamycin extends lifespan in genetically heterogeneous mice — even when treatment starts late in life. First pharmacological intervention ever shown to do this in a mammal. DOI: 10.1038/nature08221

2018

From mice to people

Mannick and colleagues show low-dose mTOR inhibition reduces respiratory tract infections and upregulates antiviral interferon-response genes in elderly volunteers — building on their 2014 finding that a similar regimen improved influenza-vaccine response. Together, the first hints the mouse result might translate to humans. DOI: 10.1126/scitranslmed.aaq1564

2021

The reality check

The follow-up phase 3 trial — 1,024 participants — misses its primary endpoint. The drug still switched on the right antiviral genes; it just didn't reduce clinically symptomatic illness. This Atlas keeps that result next to the success that came before it, because a database that only remembers the wins isn't science. DOI: 10.1016/S2666-7568(21)00062-3

All 30 milestones, in order

Oldest first. The year rail is to scale in one respect only: where the field stood still, the gap says so.

1975

Rapamycin isolated

A · Chemistry & target tool

Vézina, Kudelski & Sehgal · J Antibiot · 1975 · Microbial

A soil bacterium from Rapa Nui yields a molecule with an unusually clean effect on cell growth. Nobody yet knows what it binds.

A soil microbe from Easter Island makes a molecule that clearly slows cell growth — but nobody yet knows what it latches onto inside the cell.
16 years
1991

TOR1/TOR2 in yeast

A · Chemistry & target foundational

Heitman, Movva & Hall · Science · 1991 · Yeast

Resistance mutations name the target: TOR1 and TOR2. The field's true origin — a drug converted into a genetic handle.

Yeast that had become resistant to the drug pointed straight at its target genes, TOR1 and TOR2 — turning a drug into a genetic tool, and starting the whole field.

What stayed open The paper frames rapamycin as an immunosuppressant that blocks T-cell activation. That framing survives eighteen years, until it is inverted (Araki, 2009).

Builds on Rapamycin isolated made it possible
1994

RAFT1 — mammalian TOR

A · Chemistry & target foundational

Sabatini et al. · Cell · 1994 · Mammalian cells

The yeast gene has a mammalian counterpart. mTOR exists, and the entire yeast literature becomes relevant to human biology.

The same gene turns out to exist in humans too — meaning everything learned in yeast could now apply to human biology.

Builds on TOR1/TOR2 in yeast is extended here
1994

FRAP — parallel find

A · Chemistry & target foundational

Brown, Albers … Schreiber · Nature · 1994 · Mammalian cells

Two labs isolate the same protein within weeks by different routes. Parallel discovery signals a ripe question — worth showing, not collapsing into one node.

A second lab finds the very same protein within weeks, using a different method — a sign this was a question the whole field was closing in on at once.

Builds on TOR1/TOR2 in yeast is extended here
8 years
2002

Raptor

B · Complex architecture mechanism

Kim, Sarbassov … Sabatini · Cell · 2002 · Mammalian cells

mTOR is not a lone kinase but the core of a complex. Raptor is the scaffold that lets mTORC1 reach its substrates.

mTOR isn't a lone enzyme — it works as part of a larger complex, and Raptor is the scaffold piece that lets it reach the right targets.

Builds on RAFT1 — mammalian TOR made it possible
2002

Raptor — parallel find

B · Complex architecture mechanism

Hara, Maruki … Yonezawa · Cell · 2002 · Mammalian cells

The same complex, back to back in the same issue of Cell, plus C. elegans genetics tying raptor to TOR function in a whole animal.

Another lab publishes the same discovery side by side, plus evidence in a living worm that this scaffold protein matters for a whole animal, not just cells in a dish.

Builds on RAFT1 — mammalian TOR made it possible
2002

TORC1 vs TORC2 in yeast

B · Complex architecture foundational

Loewith … Hall · Mol Cell · 2002 · Yeast

Yeast reveals the architecture before mammals do: two complexes, and rapamycin only touches one of them. This predicts mTORC2 two years early.

Yeast reveals the bigger picture two years before mammals do: there are actually two separate TOR complexes, and the drug only blocks one of them.

Builds on TOR1/TOR2 in yeast is extended here
2003

TSC2 is a GAP for Rheb

C · Upstream sensing mechanism

Inoki, Li, Xu & Guan · Genes Dev · 2003 · Mammalian cells

The growth-factor input is wired: Akt relieves TSC1/TSC2, TSC2 stops shutting off Rheb, Rheb turns on mTORC1. A disease gene becomes a pathway component.

The final piece connecting growth-factor signals to mTOR clicks into place: Akt turns off a brake (TSC), which then releases the gas pedal (Rheb) that switches mTOR on.

Builds on RAFT1 — mammalian TOR made it possible
2003

AMPK phosphorylates TSC2

C · Upstream sensing mechanism

Inoki, Zhu & Guan · Cell · 2003 · Mammalian cells

Energy status enters the pathway. Low ATP → AMPK → TSC2 → mTORC1 off. The mechanistic bridge to caloric restriction and, later, to metformin.

The cell's low-energy alarm (AMPK) is now wired into the same brake: when energy runs low, AMPK reinforces the brake on mTOR — the mechanistic link to calorie restriction, and later to the diabetes drug metformin.

Builds on TSC2 is a GAP for Rheb is extended here
2003

Lifespan+ in C. elegans

E · Ageing & lifespan lifespan

Vellai … Müller · Nature · 2003 · C. elegans

The first time TOR is tied to how long an animal lives. The pathway stops being about cell size and starts being about ageing.

For the first time, this pathway is linked to how long an animal lives — turning it from a story about cell size into a story about ageing.

Builds on TOR1/TOR2 in yeast is extended here
2004

Rictor / mTORC2

B · Complex architecture mechanism

Sarbassov … Sabatini · Curr Biol · 2004 · Mammalian cells

Two lineages meet: yeast genetics said a second complex should exist, mammalian biochemistry finds it. A convergence node — the shape a citation graph cannot show you.

Two separate lines of research — yeast genetics and human biochemistry — arrive at the same answer at the same time: the second complex, mTORC2, really exists.

Builds on TORC1 vs TORC2 in yeast converges here Raptor converges here
2004

Lifespan+ in Drosophila

E · Ageing & lifespan lifespan

Kapahi, Zid … Benzer · Curr Biol · 2004 · Drosophila

Replication in a second organism — and the effect depends on nutritional state, the first strong hint that TOR is the mechanism behind dietary restriction.

The same lifespan effect shows up in a second species, and it depends on how much food the animal gets — the first strong clue that this pathway is the actual mechanism behind the benefits of calorie restriction.

Builds on Lifespan+ in C. elegans is extended here
2005

mTORC2 phosphorylates Akt

B · Complex architecture mechanism

Sarbassov, Guertin, Ali & Sabatini · Science · 2005 · Mammalian cells

mTORC2 is the missing S473 kinase for Akt. The pathway closes on itself, and mTORC2 becomes a metabolic node you cannot casually switch off.

mTORC2 turns out to be the missing piece that fully switches on Akt, a major growth signal — tying the two mTOR complexes into one system that can't easily be split apart.

What stayed open The origin of the selectivity problem: chronic rapalog dosing eventually reaches mTORC2, which is where the insulin resistance and hyperlipidaemia seen in the clinic come from.

Builds on Rictor / mTORC2 made it possible
2005

Restriction acts via TOR

E · Ageing & lifespan lifespan

Kaeberlein … Kennedy · Science · 2005 · Yeast

Deleting TOR1 adds nothing on top of caloric restriction — they are the same road. Restriction stops being a mystery and becomes a pathway.

Removing the TOR gene and restricting calories turn out to do the exact same thing — they don't add up, because they're really the same intervention.

Builds on TOR1/TOR2 in yeast is extended here
2007

Temsirolimus, phase 3

F · Clinic translation human

Hudes … Motzer · N Engl J Med · 2007 · Human

The first hard human outcome for an mTOR inhibitor: median overall survival 10.9 months versus 7.3 on interferon. Also the first clean human safety picture — rash, oedema, hyperglycaemia, hyperlipidaemia.

The first solid evidence in actual patients: a related drug helps people with advanced kidney cancer live longer, and doctors get their first clear picture of its side effects (rash, swelling, high blood sugar, high cholesterol).

What stayed open An oncology endpoint in sick patients says almost nothing about dosing a healthy person for decades.

Builds on RAFT1 — mammalian TOR made it possible
2008

Rags sense amino acids

C · Upstream sensing mechanism

Sancak … Bar-Peled, Sabatini · Science · 2008 · Mammalian cells

Amino acids act through a separate input that works by moving mTORC1, not by activating it directly. Location becomes a mechanism.

Amino acids turn out to switch mTOR on in an unusual way — not by directly activating it, but by moving it to a different spot inside the cell where it can then be turned on.

Builds on Raptor made it possible
2009

mTORC1 → ULK1: autophagy

D · Downstream output mechanism

Hosokawa … Mizushima · Mol Biol Cell · 2009 · Mammalian cells

The direct molecular link from nutrient status to self-digestion — the wire connecting mTOR to nearly every longevity mechanism people care about.

The direct wiring between “nutrients are available” and “stop the cell's self-cleanup process” is found — the same wire connects mTOR to nearly every process linked to ageing that people care about.

What stayed open Nutrient input is dynamic, but this is measured as an on/off state. Whether the timing of mTORC1 release, rather than its average level, sets autophagic flux is unsettled.

Builds on Raptor made it possible
2009

Mouse lifespan +14%

E · Ageing & lifespan lifespan

Harrison, Strong … Miller · Nature · 2009 · Mouse

A drug started at 600 days — roughly a 60-year-old human — extends life in a mammal: about 14% in females and 9% in males, replicated at three independent sites in genetically heterogeneous mice. The heaviest node in the tree.

A drug given to mice as old as a 60-year-old human still extends their lives — about 14% in females, 9% in males — and three independent labs get the same result. The single biggest result in this whole field.

What stayed open It is a mouse. Seventeen years on, no human trial has tested a lifespan or healthspan endpoint. Everything above this node on the human side is inference.

Builds on Lifespan+ in C. elegans is extended here Lifespan+ in Drosophila is extended here Restriction acts via TOR is extended here
2009

Rapamycin boosts memory

F · Clinic reversal

Araki … Ahmed · Nature · 2009 · Mouse & primate

The drug filed for eighteen years as an immunosuppressant makes memory T cells better — in quantity and in quality, in mice and in macaques. The field's assumption inverts.

A drug that had been used for 18 years purely to suppress the immune system turns out to also make immune memory cells work better — flipping the field's assumption about what it does.

What stayed open Immunosuppression and immune enhancement depend on dose and timing. Nobody has mapped where that boundary sits in humans.

Builds on TOR1/TOR2 in yeast is overturned here Raptor made it possible
2010

mTORC1 on the lysosome

C · Upstream sensing mechanism

Sancak, Bar-Peled, Zoncu … Sabatini · Cell · 2010 · Mammalian cells

The lysosome — long treated as the cell's rubbish bin — turns out to be the platform where growth is decided.

The lysosome — long thought of as just the cell's trash bin — turns out to be the platform where the decision to grow gets made.

Builds on Rags sense amino acids is extended here
2011

v-ATPase: inside-out

C · Upstream sensing mechanism

Zoncu, Bar-Peled, Efeyan … Sabatini · Science · 2011 · Mammalian cells

The signal starts inside the lysosomal lumen and is read outwards. Counter-intuitive, and it relocates where 'nutrient status' physically lives.

The nutrient signal starts from inside the lysosome and gets read from the outside — a surprising, inside-out way for a cell to sense its surroundings.

Builds on mTORC1 on the lysosome is extended here
2011

Everolimus, BOLERO-2

F · Clinic translation human

Baselga … Hortobagyi · N Engl J Med · 2011 · Human

Progression-free survival roughly doubles when everolimus is added to endocrine therapy. mTOR inhibition becomes standard oncology care, and its side-effect profile gets characterised in thousands of people.

Adding this drug to standard hormone therapy roughly doubles how long advanced breast cancer stays under control — the drug becomes a normal part of cancer treatment, and doctors learn its side effects in thousands of patients.

Builds on Temsirolimus, phase 3 is extended here
2012

4E-BPs and TOP mRNAs

D · Downstream output reversal

Thoreen, Chantranupong … Sabatini · Nature · 2012 · Mouse cells

Ribosome profiling under a full mTOR inhibitor shows the translational program is almost entirely TOP-motif mRNAs, controlled by the 4E-BPs. Earlier models built on 5'UTR complexity find no support.

Blocking mTOR completely and checking exactly which genes stop being read shows the effect runs almost entirely through one protein family (4E-BPs) — overturning an earlier, more complicated model.

What stayed open Rapamycin only partially inhibits 4E-BP phosphorylation — so twenty years of rapamycin data were reading a partial inhibitor as if it were complete.

Builds on Raptor made it possible
2012

mTORC1 → TFEB

D · Downstream output mechanism

Settembre, Zoncu … Sabatini, Ballabio · EMBO J · 2012 · Mammalian cells

The sensing branch and the autophagy branch fuse: mTORC1 sits on the lysosome and phosphorylates the transcription factor that builds more lysosomes.

Two separate parts of the story merge into one: mTOR, sitting on the lysosome, directly controls the master switch (TFEB) that builds more lysosomes and turns on cleanup.

What stayed open TFEB shuttles in and out of the nucleus — an inherently time-dependent readout, and a natural place to test whether pulse frequency carries information.

Builds on mTORC1 on the lysosome converges here mTORC1 → ULK1: autophagy is extended here
2013

GATOR1 / GATOR2

C · Upstream sensing mechanism

Bar-Peled, Chantranupong … Sabatini · Science · 2013 · Mammalian cells

The negative arm of amino-acid sensing — and it is mutated in human cancers. Also the handle that makes the leucine sensor findable.

A whole new set of brake-and-release proteins for amino-acid sensing is found — some of which are mutated in human cancers, and one that later leads researchers to the actual leucine sensor.

Builds on Rags sense amino acids is extended here
2014

RAD001 in older adults

F · Clinic translation human

Mannick … Klickstein · Sci Transl Med · 2014 · Human

The first time the ageing branch touches a human. Low-dose everolimus raises influenza-vaccine response by about 20% in older adults and lowers PD-1 on T cells.

For the first time, the “ageing” side of this research reaches an actual human: a low dose of the drug improves older adults' response to a flu vaccine by about 20% and reduces a marker of immune exhaustion.

What stayed open The endpoint is a surrogate — antibody titre, not health, not lifespan.

Builds on Rapamycin boosts memory is extended here Mouse lifespan +14% is extended here
2015

Sestrin2: leucine sensor

C · Upstream sensing mechanism

Wolfson, Chantranupong … Sabatini · Science · 2015 · Mammalian cells

A twenty-year question closes: a protein binds leucine directly, with a Kd matching the concentration at which leucine half-maximally activates mTORC1. The cell has a leucine receptor.

A twenty-year mystery is solved: a protein is found that grabs onto leucine directly, at just the concentration where leucine is known to switch mTOR on. The cell really does have a dedicated leucine detector.

Builds on GATOR1 / GATOR2 made it possible
2016

RapaLink-1

F · Clinic tool

Rodrik-Outmezguine … Rosen, Shokat · Nature · 2016 · Cells & mouse

Linking the rapamycin pocket to the kinase pocket in a single molecule defeats resistance mutations that beat both earlier generations — and it is the chemistry that makes true mTORC1 selectivity thinkable. Direct ancestor of the bi-steric inhibitors now in human trials.

Chemists link two separate drug parts into a single molecule that gets around the resistance mutations that beat both earlier generations of the drug — the chemistry that first makes a much more selective mTOR-blocker seem possible, and the direct ancestor of the newer drugs now in human trials.

What stayed open Selectivity in cells is not selectivity in a person dosed for years.

Builds on mTORC2 phosphorylates Akt made it possible Temsirolimus, phase 3 is overturned here
2018

Phase 2a: fewer infections

F · Clinic translation human

Mannick … Klickstein · Sci Transl Med · 2018 · Human

264 older adults, six weeks of low-dose TORC1 inhibition, and a significant drop in infections reported over the following year. The strongest human signal the field has produced.

264 older adults took a low dose for six weeks, and over the following year they reported significantly fewer infections — the strongest positive human signal this field has produced so far.

What stayed open Phase 2a, self-reported infections. Closer to health than a titre — but still not ageing.

Builds on RAD001 in older adults is extended here
2021

Phase 3 misses endpoint

F · Clinic null human

Mannick … Shergill · Lancet Healthy Longev · 2021 · Human

1,024 participants. RTB101 did not reduce clinically symptomatic respiratory illness — 26% versus 25%, p=0.65. The antiviral genes still switched on; the clinical outcome did not move.

In a trial of over 1,000 people, the drug switched on the right antiviral genes just like before — but it did not actually reduce how many people got sick with respiratory illness (26% vs. 25%, essentially no difference).

What stayed open The biomarker moved and the patient did not. This is the node most databases would quietly omit, and the reason the Atlas records failures at the same weight as successes.

Builds on Phase 2a: fewer infections is overturned here

Where the field is now

The next three meetings in the Atlas calendar.

6–11 Sep 2026 · Grenå, Denmark

EMBO Workshop: Autophagy — Foundations, Functions and Physiology 2026

EMBO

22–26 Sep 2026 · Cold Spring Harbor, NY, USA

CSHL: Mechanisms of Aging 2026

CSHL

27–29 Sep 2026 · Lunteren, Netherlands

7th European Meeting on AMPK & Nutrient-Sensing Pathways 2026

Other

All 41 meetings, with country and relevance filters →

What the field doesn't know

Two kinds of not-knowing, and they are not the same thing. The first is an experiment nobody has run. The second is a link this corpus does not hold — which may mean it does not exist, or may mean the Atlas is incomplete.

Where the timeline stops

4 nodes in the figure have no children: the branch ends there because the experiment has not been done.

B · Complex architecture

mTORC1-only, in humans?

Bi-steric inhibitors spare mTORC2 acutely. Whether that holds under chronic human dosing — and whether sparing mTORC2 removes the metabolic toxicity — is still open; the Phase 1 signal (SCH2025) is consistent with it but was uncontrolled.

D · Downstream output

Frequency, not level?

Every mechanism on this branch was measured at steady state. If mTORC1 oscillates with feeding, growth-factor bursts and the cell cycle, then two cells with identical average activity could have opposite fates. Almost nothing in the literature is designed to detect that.

F · Clinic

Does timing matter?

Every human trial so far dosed on a fixed schedule chosen for convenience. If mTORC1 is phase-dependent, dose timing is an untested variable sitting inside every null result — including the one directly below this bud.

E · Ageing & lifespan

Human healthspan?

Untested — not disputed, not emerging. No completed trial has used a human ageing endpoint. This is the question the Atlas exists to keep visible.

Gaps computed from the corpus

Surfaced by joining the Atlas's entity graph to the kind of study behind each finding.

Contradiction / tension