Oliver's mTOR Atlas Evidence Platform
Mode
Reading level

mTOR Core · Lesson 08 · Intermediate · 25 min

Feedback Regulation

Why inhibiting mTORC1 can make signalling upstream of it go up

The question

Why does a pathway push back on its own input?

What you should be able to do

After this lesson you should be able to

  • Explain how a negative feedback loop can make an inhibitor raise the signal it was meant to lower.
  • Predict what happens to upstream signalling when mTORC1 output is removed, and over what timescale.
  • Critique an experiment that reports ‘mTOR activity’ from a single phospho-readout at a single time point.
  • Distinguish released feedback from off-target complex disruption as explanations of the same metabolic effect.

Research skill Systems thinking

The core idea

Everything so far has run in one direction: inputs converge, mTORC1 responds, outputs change. Real signalling networks are not one-directional, and this one has a well-described loop running backwards. mTORC1 output weakens the growth-factor input that activated it.

The best-characterised route runs through S6K1. Active mTORC1 activates S6K1; S6K1 phosphorylates IRS-1, the adaptor that couples the insulin receptor to PI3K, and reduces its function and its abundance. A second route was found later by phosphoproteomics: mTORC1 phosphorylates Grb10, stabilising it as an inhibitor of insulin and IGF-1 signalling. Two mechanisms, the same sign.

This changes what an inhibitor experiment means. If mTORC1 activity is what suppresses the upstream branch, then removing mTORC1 activity releases that suppression, and signalling through PI3K and AKT can rise rather than fall. That is not an oddity of one cell line; it is the expected behaviour of a negative feedback loop with the loop opened, and it is a large part of why the clinical story of mTOR-directed drugs has been more complicated than the pathway diagram suggests.

So far everything has run one way: signals in, mTORC1 responds, jobs change. Real signalling networks also run backwards.

mTORC1 weakens the very insulin signal that switched it on. It does this through S6K1, which damages the adaptor that connects the insulin receptor to the rest of the pathway.

This changes what a drug experiment means. Block mTORC1 and you also remove that damage — so the signalling above the drug's target can go up.

The mechanism, in outline

mTORC1S6K1 · Grb10IRS-1PI3K / AKTactivatesthe loop is negative: mTORC1 output weakens the very input that switched it onmTORC1 inhibitorremove mTORC1 output andthe brake goes with it

Trace the loop on the left of the figure. PI3K and AKT activate mTORC1 (Lesson 07); mTORC1 activates S6K1; S6K1 acts on IRS-1; weakened IRS-1 means less signal from the insulin receptor into PI3K. Sustained mTORC1 activity therefore produces a cell that responds less well to insulin — cellular insulin resistance arising from the growth pathway itself rather than from anything wrong with the receptor.

Two companion papers established this in 2004 from opposite directions: one showing that loss of the TSC complex, which leaves mTORC1 and S6K1 constitutively active, disrupts insulin–PI3K signalling through IRS proteins; the other showing that inappropriate activation of the Rheb–mTOR–S6K cassette depletes IRS-1 and IRS-2 and produces cellular insulin resistance. Reading both is a good habit: independent groups, converging design, published within months.

The right of the figure is the same loop seen through a drug. Inhibiting mTORC1 in cancer cells relieves the feedback and activates upstream receptor-tyrosine-kinase and AKT signalling. A related route reaches the MAPK pathway: mTORC1 inhibition activates MAPK signalling through a PI3K-dependent feedback loop. Both are ways in which a drug aimed at a growth pathway can leave the cell with more proliferative signalling than before in some compartments.

There is also a whole-animal version of the problem with a different mechanism. In mice, chronic rapamycin disrupts mTORC2 as well and produces insulin resistance — and, importantly, that metabolic effect could be uncoupled from the lifespan extension. Feedback and off-target complex disruption are two distinct explanations for why the same drug produces both benefits and metabolic costs; distinguishing them matters for anyone trying to keep one and drop the other.

Follow the loop. Insulin signalling switches mTORC1 on; mTORC1 switches S6K1 on; S6K1 damages IRS-1; with less IRS-1 the insulin receptor connects less well. A cell that has been building for a while responds to insulin less.

Two groups showed this within months of each other, from opposite directions.

Seen through a drug, the same loop runs in reverse: blocking mTORC1 in cancer cells raises signalling upstream, sometimes through more than one route.

In mice there is a second, separate problem: long-term rapamycin also disturbs mTORC2 and causes insulin resistance — and that effect could be separated from the lifespan benefit.

Interactive model

The loop, and what happens when you open it

mTORC1
Feedback step
IRS-1 / IRS-2PI3KAkt/PKBPRAS40mTORC1S6K1

The negative feedback loop drawn as a loop: the return path along the bottom is S6K1 acting on IRS-1. AKT reaches mTORC1 here through PRAS40, the brake inside the complex; the TSC route from Lesson 04 is left out so the loop stays readable. The second control is not a drug but an experiment — blocking the feedback step itself, which is how the mechanism gets tested rather than assumed.

mTORC1 active · S6K1 active · IRS-1 being degraded · insulin input falling

The loop as it normally runs. Sustained mTORC1 output drives S6K1, S6K1 acts on IRS-1, and the insulin input that started all of this weakens — a cell that has been building for a while listens to insulin less well.

IRS-1 / IRS-2
Scaffold recruiting PI3K to activated insulin/IGF-1 receptors.The adaptor that connects the insulin receptor to PI3K.Serine phosphorylation by S6K1 (and others) triggers IRS-1 degradation — the molecular basis of the best-characterised mTORC1 feedback loop and of rapalog-associated insulin resistance.
PI3K
Class I PI3K phosphorylates PIP2 to PIP3 at the plasma membrane.Makes the membrane signal that pulls Akt in.p110/p85 heterodimer; PIK3CA is one of the most frequently mutated oncogenes in human cancer. Its output is a lipid, not a phosphoprotein — which is why PTEN reverses it.
Akt/PKB
AGC kinase requiring PIP3 recruitment plus two phosphorylations to be fully active.The main 'grow' relay from growth factors.T308 by PDK1 and S473 by mTORC2. Recruitment and activation are separate events — a distinction the older Atlas diagram blurred.
PRAS40
Raptor-binding inhibitor displaced by Akt phosphorylation.A plug in mTORC1's substrate slot.AKT1S1. Competes with substrate for the Raptor TOS-motif site; its displacement is a substrate-access mechanism, not a change in kinase catalytic rate.
mTORC1
mTOR–Raptor–mLST8 (+PRAS40, DEPTOR); switched on only at the lysosome.The growth decision itself.Coincidence detector: nutrients supply location via the Rags, growth factors supply activation via Rheb. Neither alone is sufficient — the single most important idea in the pathway.
S6K1
Ribosomal protein S6 kinase; the standard readout of mTORC1 activity.mTORC1's best-known output kinase.T389 phosphorylation by mTORC1 is rapamycin-sensitive, which is why S6K1 became the field's default assay — and why the field long over-read rapamycin as a complete mTORC1 inhibitor.

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
mTORC1Feedback stepReadoutWhat it means
highintactmTORC1 active · S6K1 active · IRS-1 being degraded · insulin input fallingThe loop as it normally runs. Sustained mTORC1 output drives S6K1, S6K1 acts on IRS-1, and the insulin input that started all of this weakens — a cell that has been building for a while listens to insulin less well.
inhibitedintactmTORC1 inhibited · S6K1 off · IRS-1 restored · AKT signalling risesInhibit mTORC1 and the brake on IRS-1 goes with it, so signalling upstream of the drug's target increases. This is the result that looks like a paradox and is not one: whenever inhibiting X raises Y, check whether Y is upstream of X.
highblockedmTORC1 active · S6K1 active · IRS-1 intact · insulin input maintainedThe experiment that tests the mechanism rather than assuming it: make IRS-1 resistant to S6K1 and the cell keeps its insulin sensitivity while mTORC1 stays on. If insulin resistance persisted here, this loop would not be the explanation.
inhibitedblockedmTORC1 inhibited · S6K1 off · IRS-1 intact · AKT signalling unchangedWith the loop already blocked, inhibiting mTORC1 no longer raises AKT signalling. Comparing this state with the second one is how you would attribute a rise to released feedback rather than to something else the drug did.

How to think about loops without getting lost

A practical rule: whenever you read a result of the form 'inhibiting X increased Y', ask whether Y is upstream of X. If it is, the first hypothesis should be a released feedback loop, not a paradox and not an off-target effect. The second question is the timescale — feedback release is often fast and transient, and a measurement at one hour and a measurement at one week can point in opposite directions. One study in this lesson found biphasic AKT regulation after mTOR kinase inhibition for exactly that reason.

The reverse rule is just as useful. If a paper reports that a treatment raises mTORC1 signalling over days, ask what happened to the upstream branch, because sustained mTORC1 activity should be eroding its own input. A result that shows both high mTORC1 output and undiminished insulin signalling over time is telling you something is unusual about the system, and that is worth understanding rather than skipping.

There is a design consequence too. Because the loop exists, an experiment that measures only one node of the pathway is measuring a moving target. This is one of the reasons phospho-S6 alone is a weak readout for 'mTOR activity' — it reports one substrate of one complex at one moment, in a network arranged to compensate.

A rule that saves a lot of confusion: if inhibiting X raises Y, first ask whether Y is upstream of X. If it is, released feedback is the ordinary explanation.

The second question is time. Feedback release is often fast and temporary, so a measurement after an hour and one after a week can point in opposite directions.

It also means one measurement is a moving target — which is why a single phospho-band is weak evidence for 'mTOR activity'.

Scope of the evidence here

The feedback mechanisms below were characterised in cultured cells, several of them cancer cell lines chosen because their signalling is strong and easy to read. The mouse study is tier C. None of this is a description of what happens in a person taking one of these drugs, and the corpus's human evidence on metabolic effects is a separate literature with its own tiers. Keep the mechanism and the clinical claim in separate boxes.

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.

One half of the discovery of mTORC1's main negative feedback loop.

Study page →

mTORC1 inhibition relieves feedback and activates upstream RTK-PI3K-Akt signalling.

Study page →

mTORC1 inhibition activates the MAPK pathway via a PI3K-dependent feedback loop in cancer.

Study page →

Phosphoproteomics identify Grb10 as an mTORC1 substrate driving negative feedback on insulin/PI3K.

Study page →

In mice, chronic rapamycin also disrupts mTORC2, causing insulin resistance; lifespan extension can be uncoupled from this side effect.

Study page →

How much of the metabolic cost seen with long-term mTOR-directed drugs comes from released feedback and how much from disruption of mTORC2 is not resolved, and the two explanations imply different fixes. The mouse work that separated the metabolic effect from the lifespan effect is tier C and has no equivalent in people.

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.

Compare the evidence

Two groups, two months apart, one loop

D The TSC1-2 tumor suppressor controls insulin-PI3K signaling via regulation of IRS proteinsD Inappropriate activation of the TSC/Rheb/mTOR/S6K cassette induces IRS1/2 depletion, insulin resistance, and cell survival deficiencies
Model systemMammalian cells (TSC1/TSC2-null MEFs, human cells)Mammalian cells
PerturbationMammalian cells including TSC1/TSC2-null fibroblasts, in which mTORC1 and S6K1 activity is constitutively high.Cultured mammalian cells with the Rheb–mTOR–S6K cassette inappropriately activated.
ReadoutIRS-1 protein and insulin–PI3K signalling, both reduced when S6K1 activity is left on.IRS1 and IRS2 depleted, with cellular insulin resistance and impaired downstream signalling as the consequence.
What do both studies support?

That sustained mTORC1–S6K1 activity weakens the insulin input that activates it, through IRS proteins — a negative feedback loop reached from two directions by two groups within months.

Where do they differ?

Less than usual, which is the point of putting them side by side. They differ in how the pathway was forced on — loss of the brake in one, activation of the cassette in the other — and converge on the same readout. Independent designs converging is a different kind of evidence from one result repeated, and it is worth recognising when you see it.

What experiment would help next?

Both are cell work with the pathway pushed hard. The open question is quantitative: how much of the insulin resistance seen with mTOR-directed drugs in animals or people is this loop, and how much is disruption of mTORC2 — two mechanisms that predict different fixes.

Scientific caution

Reading a feedback result

What this evidence supports

  • That mTORC1 output suppresses upstream insulin signalling through S6K1 and Grb10 in cultured cells
  • That inhibiting mTORC1 can therefore raise AKT signalling rather than lower it

What it does not establish

  • That an increase in AKT signalling after mTORC1 inhibition means the drug missed its target
  • That the loop's contribution in a person on one of these drugs is known — the mechanism is cell work and the mouse study is tier C
  • That one phospho-readout at one time point describes a network arranged to compensate
Why?

Most of these mechanisms were characterised in cancer cell lines chosen because their signalling is strong and easy to read. Keep the mechanism and the clinical claim in separate boxes; the corpus's human evidence on metabolic effects is a different literature with its own tiers.

Think

A tumour has an activating change in PI3K. On the diagram, mTORC1 is downstream, so an mTORC1 inhibitor looks like the obvious move. Using this lesson, give two reasons the result might disappoint.

One reason is about the loop; one is about what the drug does and does not block.

Think first, then reveal

First, releasing feedback: removing mTORC1 output removes S6K1-mediated suppression of IRS-1 and Grb10-mediated suppression of insulin/IGF-1 signalling, so AKT signalling can rise — in a cell already driven by an activating PI3K change, that is the branch you least wanted to strengthen. A related route raises MAPK signalling through a PI3K-dependent loop. Second, rapamycin does not block all mTORC1 outputs equally; substrates such as 4E-BP1 are less sensitive than S6K1, so 'mTORC1 inhibited' on a phospho-S6 blot can coexist with continued cap-dependent translation. Both reasons argue for measuring more than one substrate and more than one time point, and both are why combination approaches and active-site inhibitors were pursued.

Chronic rapamycin in mice produces insulin resistance, and the same work found that this could be uncoupled from lifespan extension. Why is ‘uncoupled’ a stronger and more useful word here than ‘unrelated’?

Think about what an uncoupling result licenses you to try next.

Think first, then reveal

'Unrelated' would be a claim about mechanism — that the two outcomes share no causal machinery — which the experiment does not support. 'Uncoupled' is a claim about what was observed: conditions exist in which one appears without the other. That is weaker, and much more useful, because it converts a resignation ('the side effect is the price of the benefit') into a design problem ('find the conditions that separate them'). It is also directly testable, and the Atlas carries it as an open question rather than a settled result. Notice how much of the epistemic content of the finding sits in one word choice.

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-upThrough which substrate does mTORC1 most classically weaken insulin signalling?

Show the answer

B — S6K1, which acts on IRS-1. S6K1 phosphorylates IRS-1 and reduces its function and abundance, so the receptor couples less well to PI3K. Grb10 is the second, later-described route with the same sign.

Question 2 · Step upIn a cancer cell line, adding an mTORC1 inhibitor raises AKT phosphorylation. What is the first hypothesis a careful reader should reach for?

Show the answer

B — AKT is upstream of mTORC1, so removing mTORC1 output released the negative feedback that was suppressing it. Whenever inhibiting X raises Y, check whether Y is upstream of X. If it is, released feedback is the ordinary explanation and should be excluded before anything more exotic.

Question 3 · HarderWhy is phospho-S6 alone a weak readout for ‘mTOR activity’ in an experiment that runs for days?

Show the answer

C — It reports one substrate of one complex at one moment, in a network with feedback that shifts over time and with substrates that differ in drug sensitivity. The readout is real but narrow. Feedback release changes the upstream branch over time, and substrates such as 4E-BP1 respond differently from S6K1 to the same drug — so one band on one day can be consistent with several different states of the network.

Go deeper

Follow a guided route through the mechanism:

Concepts introduced in this lesson

Negative feedback · Feedback release · Cellular insulin resistance · Timescale in signalling experiments