mTOR Core · Lesson 06 · Intermediate · 30 min
Nutrient Sensing
The question
How does a cell know which nutrients it has?
What you should be able to do
After this lesson you should be able to
- Trace a chain of three inhibitory steps and state the sign of the whole chain.
- Predict the direction of an mTORC1 readout after deleting a sensor, a GATOR2 subunit or a GATOR1 subunit.
- Distinguish a protein that binds a nutrient from a protein that is required for the response to it.
- Evaluate what an amino-acid withdrawal-and-re-addition experiment can say about a fed animal.
Research skill Competing hypotheses
The core idea
A cell does not just notice 'there is food'. It has separate detector proteins for separate molecules: one for leucine, one for arginine, one for a molecule made from methionine.
The wiring is the strange part. A detector that has caught its amino acid stops blocking the next protein in the chain — and that protein blocks the next one, which blocks the next.
Count the blocks: food arriving releases a chain of brakes rather than pressing an accelerator. The resting state of this system is off, and actively held off.
The mechanism, in outline
Follow leucine. With no leucine, its detector Sestrin2 holds down a complex called GATOR2. Leucine binding to Sestrin2 lets GATOR2 go. Free GATOR2 blocks GATOR1, and GATOR1 is the thing that switches the Rag proteins off.
Arginine works the same way with a different detector, and a molecule derived from methionine with a third one.
One detector sits inside the lysosome membrane and reports what is in the lumen — so a single amino acid can be reported from two different places.
Not everything goes this route: glutamine has been shown to reach mTORC1 without the Rag proteins at all.
Follow the double negatives
The leucine and arginine branches through their sensors to the Rag GTPases. Every link drawn here is an inhibition taken from the Atlas pathway model — which is why an amino acid arriving makes the pathway more active.
GATOR1 restrained · Rag GTPases in their active state
Both amino acids are bound to their sensors, so neither sensor is inhibiting GATOR2, GATOR2 is free to inhibit GATOR1, and GATOR1 stops switching the Rags off. Count the blunt ends on the way through: the pathway is on because three inhibitions are released, not because anything was pushed.
- Leucine
- Essential BCAA; its cytosolic concentration is read by Sestrin2 and (contested) by LARS.The amino acid the cell watches most closely.Binds Sestrin2 with ~20 µM Kd — within the range over which intracellular leucine actually fluctuates, which is the main argument that Sestrin2 is a physiological sensor rather than a binder.
- Arginine
- Sensed twice: by cytosolic CASTOR1 and by the lysosomal transporter SLC38A9.A second amino acid the cell counts.Two-sensor architecture lets the cell distinguish cytosolic from lysosomal arginine pools; the functional division of labour is still argued.
- Sestrin2
- Leucine-binding negative regulator: without leucine it holds GATOR2 inactive.A leucine detector that works as a brake.Sestrin2 is a *negative* regulator. Leucine binding releases GATOR2 — the pathway is switched on by removing a brake. Also stress-inducible via p53, so it sits at a stress/nutrient junction.
- CASTOR1
- Arginine-binding inhibitor of GATOR2.An arginine detector that works as a brake.Homodimer (or heterodimer with CASTOR2); arginine binding dissociates it from GATOR2. Same double-negative logic as Sestrin2.
- GATOR2
- Pentameric complex (WDR24, WDR59, MIOS, SEH1L, SEC13) inhibiting GATOR1.A brake on the brake.Structures (Valenstein 2022) resolved the cage-like architecture and the Sestrin2/CASTOR1 binding surfaces; how it inhibits GATOR1 catalytically is still argued.
- GATOR1
- DEPDC5–NPRL2–NPRL3; GAP for RagA/B.The switch-off machine for the Rags.DEPDC5 loss causes focal epilepsy — a human phenotype that establishes GATOR1's physiological relevance beyond cell lines.
- Rag GTPases
- RagA/B–RagC/D heterodimers; nucleotide state determines mTORC1 recruitment.The taxi that brings mTORC1 to the lysosome.Note the inversion: RagA/B is active when GTP-loaded, RagC/D when GDP-loaded. The Rags control mTORC1 *location*, not its catalytic activity.
This is a simplified teaching model with a fixed set of states, not a simulation. It shows the direction each control pushes the pathway, not how much, how fast, or what any particular cell would do.
Every state of this model
| Leucine | Arginine | Readout | What it means |
|---|---|---|---|
| present | present | GATOR1 restrained · Rag GTPases in their active state | Both amino acids are bound to their sensors, so neither sensor is inhibiting GATOR2, GATOR2 is free to inhibit GATOR1, and GATOR1 stops switching the Rags off. Count the blunt ends on the way through: the pathway is on because three inhibitions are released, not because anything was pushed. |
| present | absent | GATOR2 partly inhibited · Rag signalling reduced | One sensor is free and the other is still holding GATOR2. The teaching model draws this as a reduced rather than an absent response; how much a single missing amino acid actually costs varies with cell type and is not a fixed fraction. |
| absent | present | GATOR2 partly inhibited · Rag signalling reduced | The mirror image, and worth comparing with the state above: the pathway can report which amino acid is missing at the sensor, and by the time the signal reaches the Rags that information is gone. |
| absent | absent | GATOR1 active · Rag GTPases switched off | Both sensors unoccupied, both inhibiting GATOR2, GATOR1 free to act as a GAP on the Rags. This is the resting state of the system: held off, and released by the arrival of nutrients rather than switched on by them. |
Why double negatives, and why it matters for reading data
Wiring things this way has an advantage: if a part breaks, the system tends to fail towards 'off' rather than towards uncontrolled growth.
For a reader it has a cost. Signs are easy to get backwards, and a surprising number of confusions in this field are sign errors rather than disagreements about data.
Genes for one of these blocking complexes are mutated in some cancers — the same logic as the missing brake in Lesson 04, one step further up.
What 'sensor' does and does not mean
These proteins were identified largely through binding assays, structures and cell work using amino-acid withdrawal followed by re-addition — a stimulus far sharper than anything a fed animal experiences. Calling a protein a sensor says that its behaviour changes when the molecule binds it, in the systems tested. It does not establish how much of a tissue's response to a meal runs through that protein, and it does not tell you the concentration range over which the response is graded in a living animal. The Atlas has a standing open question about exactly this gap.
What does the evidence say?
These are Atlas studies, with the Atlas's own evidence tier. Each card links to the full record — nothing here restates it.
Amino acid sufficiency signals through mTOR to p70 S6K and 4E-BP1 via a common effector, first linking nutrients to mTOR.
Study page →Found the OFF switch for amino acid signaling: the GATOR1 complex is a GAP that shuts the Rag GTPases (and thus mTORC1) down when amino acids run low, while GATOR2 opposes it.
Study page →Sestrin2 is a direct leucine sensor whose leucine binding releases GATOR2 to activate mTORC1.
Study page →Identified CASTOR1 as the direct arginine sensor: when arginine binds CASTOR1, it lets go of GATOR2, switching mTORC1 on.
Study page →Extended nutrient sensing beyond amino acids to METABOLITES: SAMTOR reads S-adenosylmethionine (SAM), the cell's methyl-donor currency, linking methionine and one-carbon metabolism to mTORC1.
Study page →SLC38A9 is a component of the lysosomal amino-acid sensing machinery controlling mTORC1.
Study page →Glutamine activates mTORC1 via a Rag-independent, Arf1-dependent route distinct from leucine.
Study page →The sensors were characterised mostly by binding assays and by acute amino-acid withdrawal and re-addition in cell lines. How much of the response of an intact tissue to a meal runs through any individual sensor, and over what concentration range the response is graded in an animal, is largely uncharacterised.
Work through it
Each of these asks you to commit to something — a prediction, a reading of two studies, a design — before it answers. Everything here is in the page, so nothing is lost if you would rather just read it.
Two nutrients, two routes
| D Sestrin2 is a leucine sensor for the mTORC1 pathway | D Metabolism. Differential regulation of mTORC1 by leucine and glutamine | |
| Model system | Mammalian cells | Mammalian cells |
| Perturbation | Mammalian cells and biochemistry: leucine binding to Sestrin2 tested directly, and the consequence for GATOR2 followed. | Mammalian cells: glutamine supplied and withdrawn in cells with the Rag GTPases disabled. |
| Readout | Sestrin2 releases GATOR2 when leucine binds, and mTORC1 activity follows — a defined molecular chain from one amino acid to the kinase. | mTORC1 still responds to glutamine without functioning Rags, through an Arf1-dependent route distinct from the leucine pathway. |
What do both studies support?
That mTORC1 responds to individual amino acids rather than to protein availability in general, and that specific proteins mediate specific nutrients.
Where do they differ?
In how much of nutrient sensing the canonical machinery accounts for. The first fits the standard picture and fills in its missing molecular step; the second shows a nutrient reaching the same kinase without that machinery at all. Read together, they say that 'the amino-acid sensing pathway' is a family of mechanisms rather than one path with details attached.
What experiment would help next?
A systematic comparison of several amino acids in the same cells, with the canonical machinery removed, would show how much of the nutrient response actually depends on it — the current answer is assembled from separate experiments on separate nutrients.
What calling a protein a ‘sensor’ claims
What this evidence supports
- That the protein binds the nutrient directly and that binding changes what it does, in the systems tested
- That removing it blunts the mTORC1 response to that nutrient in cultured cells
What it does not establish
- How much of a tissue's response to a meal runs through that protein
- Over what concentration range the response is graded in an animal, rather than between full withdrawal and full re-addition
- That the set of sensors is complete — each was found by looking, and the looking is not finished
Why?
These proteins were identified by binding assays, structures and acute withdrawal-and-re-addition in cell lines. That design supports a mechanism and leaves the quantitative physiology open; the Atlas carries the gap as an open question rather than as a footnote.
Think
Predict what happens to mTORC1 signalling if you delete Sestrin2, and then if you delete a GATOR1 subunit. Do the two predictions have the same sign?
Write the chain out with each arrow's sign, then remove one box at a time.
Think first, then reveal
They have opposite signs. Sestrin2 inhibits GATOR2, so deleting Sestrin2 leaves GATOR2 free — GATOR1 stays restrained, Rag signalling tends to stay on even without leucine. Deleting a GATOR1 subunit removes the GAP that shuts the Rags down, which also raises signalling, but for the opposite structural reason and with a different dependence on nutrients: the Sestrin2-null cell has lost a leucine-responsive step, while the GATOR1-null cell has lost the off-switch itself. That difference — loss of regulation versus loss of a specific input — is what an experiment measuring only phospho-S6K would miss, and it is why nutrient-response experiments are usually run as withdrawal and re-addition rather than at steady state.
Glutamine has been reported to reach mTORC1 by a route that does not require the Rag GTPases. What would that do to an experiment that used Rag-null cells to ask 'is this response amino-acid dependent'?
Think about what a negative result in that system would and would not tell you.
Think first, then reveal
It would make a negative result ambiguous. If a response survives in Rag-null cells, the standard reading is that it is not amino-acid dependent — but a Rag-independent route means the response could still be nutrient-driven while bypassing the machinery the experiment was designed around. The general point is that a knockout defines a pathway only as sharply as the field's map of alternative routes, and that map keeps changing. In practice this is an argument for testing more than one nutrient and more than one readout before concluding that a signal is nutrient-independent.
Check yourself
Three questions, easiest first. Nothing is recorded and there is no pass mark — the explanation after each answer is the part worth reading.
Question 1 · Warm-upWhich amino acid does Sestrin2 bind?
Sestrin2 is the leucine-binding protein of this set; CASTOR1 handles arginine and SAMTOR reads S-adenosylmethionine, the metabolite downstream of methionine.
Show the answer
B — Leucine. Sestrin2 is the leucine-binding protein of this set; CASTOR1 handles arginine and SAMTOR reads S-adenosylmethionine, the metabolite downstream of methionine.
Question 2 · Step upLeucine binds Sestrin2. What happens next, in order?
Three inhibitory steps in a row. Counting them carefully is the whole skill here, because getting the sign backwards is the most common error in reading this part of the pathway.
Show the answer
A — Sestrin2 releases GATOR2; GATOR2 inhibits GATOR1; the Rag GTPases stay in their active state. Three inhibitory steps in a row. Counting them carefully is the whole skill here, because getting the sign backwards is the most common error in reading this part of the pathway.
Question 3 · HarderA study shows a protein binds leucine in vitro and that deleting it blunts the mTORC1 response to leucine re-addition in cultured cells. What is the honest limit of that evidence?
Withdrawal-and-re-addition is a much sharper stimulus than eating, and one cell type is not a tissue. The experiment supports the mechanism it tested and leaves the quantitative physiology open.
Show the answer
B — It supports a sensing role in that system, but not how much of a tissue's response to a meal runs through the protein, nor over what physiological concentration range it is graded. Withdrawal-and-re-addition is a much sharper stimulus than eating, and one cell type is not a tissue. The experiment supports the mechanism it tested and leaves the quantitative physiology open.
Go deeper
Follow a guided route through the mechanism:
- How does a cell know it has enough raw material to grow?The amino-acid route in full, from the lumen to the kinase.
- What happens, and in what order, when you stop eating?What happens to these inputs when food is withheld.
Concepts introduced in this lesson
Direct sensing versus signalling · Double-negative logic · GAP and GEF activity · Compartment-specific reporting