mTOR and AMPK: Reading Nutrients and Energy

The drug that protects a transplanted kidney from rejection came out of soil collected on Easter Island in 1964–65. Working out what it hits inside a cell uncovered the enzyme that decides whether a cell grows.

Growth is expensive. Making protein uses amino acids and a large share of a cell’s ATP, so a cell that starts building without supplies runs short of both. Before it commits, a cell reads two gauges. One, built around the kinase mTOR, reads growth factors and amino acids. The other, AMPK, reads energy.

The story: an antifungal from island soil works only with a partner protein

From December 1964 to February 1965 a Canadian medical expedition worked on Rapa Nui (Easter Island). One member, the Montreal bacteriologist Georges Nogrady, collected soil from across the island. At Ayerst Research Laboratories in Montreal, a strain of the bacterium Streptomyces hygroscopicus grown from one sample was found to kill fungi. In 1972 the Ayerst microbiologist Suren Sehgal isolated the compound responsible and named it rapamycin, after the island.

Why it also stops cells from growing was answered in yeast. In Michael Hall’s laboratory in Basel, Joseph Heitman spread yeast on plates containing rapamycin and kept the rare mutants that grew anyway. Most had mutations in the gene for a small protein called FPR1, the yeast version of your FKBP12. So rapamycin does nothing alone. It must first bind FKBP12, and the pair is the poison.

The remaining mutants had changes in one of two new genes, which the group named TOR1 and TOR2, for “target of rapamycin”. The screen was published in 1991. In 1994 and 1995 several laboratories found the mammalian protein that the FKBP12–rapamycin pair binds. The name that lasted is mTOR.

mTOR is a large protein kinase. Beside its kinase domain is a small domain called FRB. Rapamycin sits between FKBP12 and FRB like glue between two blocks, and the bound FKBP12 narrows the path that substrates take to the active site. The drug does not compete with ATP. It is an allosteric inhibitor, meaning one that acts from a site other than the active site.

Yeast has two TOR genes and you have one, but in both the kinase works inside two different complexes. mTORC1, built around a partner protein called Raptor, controls growth and is the complex rapamycin blocks quickly. mTORC2, built around a partner called Rictor, is the kinase that gives Akt its second activating phosphate (the previous lesson), and it resists short rapamycin treatment.

The idea: mTORC1 switches on only when two questions get a yes

In 1996 Hall’s group reported that yeast without working TOR behave as if starved, even in rich medium: they sharply cut back the starting of new proteins and enter a resting state. Growth, they argued, is actively controlled, and TOR is the controller. What must mTORC1 hear before it gives the order? Two things, read by separate systems.

Has the body said grow? A growth factor such as insulin switches on Akt. Akt phosphorylates TSC2, one protein of the TSC complex, and that takes the complex out of action. The TSC complex is a GAP, the GTPase off-switch you met in the lesson on GTPase switches, and its target is a small GTPase called Rheb. With the GAP out of action, Rheb stays loaded with GTP, and Rheb-GTP is the direct activator of mTORC1. In the summer of 2003 several laboratories, Kun-Liang Guan’s among them, reported within weeks of one another that TSC2 is a GAP for Rheb and that Rheb acts just upstream of mTOR.

Are there amino acids to build with? The expected answer was a receptor for amino acids. In 2008 Yasemin Sancak, in David Sabatini’s laboratory, and independently Guan’s laboratory, found the link in another family of small GTPases, the Rags. Sancak’s paper carried a surprise: the Rags do not raise mTORC1’s kinase activity. They move it. In cells starved of amino acids, mTORC1 is spread through the cytoplasm. Add amino acids back and it gathers on one compartment, which the same laboratory identified in 2010 as the surface of the lysosome, the cell’s acidic digestive compartment. Rheb is anchored to internal membranes by a lipid tail, and the lysosome surface is where the two are thought to meet.

Rheb-GTP can activate only an mTORC1 that has been brought to it. Growth factors set whether Rheb is on. Amino acids set whether mTORC1 is delivered. This is an AND gate: output only when both inputs are present.

In 2011 Roberto Zoncu and colleagues reported that part of the amino acid reading is taken inside the lysosome and requires the proton pump in its membrane. Sensor proteins found since then include Sestrin2 for leucine and CASTOR1 for arginine in the cytoplasm, and the transporter SLC38A9 in the lysosomal membrane. All report to the Rags. Which sensor matters most in which tissue is not settled, and neither is where the working pool of Rheb sits.

When both answers are yes, mTORC1 phosphorylates two proteins that control the first step of translation. S6 kinase is switched on. 4E-BP1, which holds back eIF4E, the factor that grips the cap at the start of an mRNA, lets go. mTORC1 also promotes lipid synthesis and holds off autophagy, the process in which a cell digests its own contents in lysosomes to recover building blocks.

The idea: AMPK reads AMP, because AMP moves most

AMPK answers a third question: is there enough ATP? You met it in the biochemistry course as a fuel gauge. Here is why the gauge reads AMP and not ATP.

A healthy cell holds ATP nearly constant, so ATP is a poor alarm. The enzyme adenylate kinase interconverts the three adenine nucleotides and keeps them near equilibrium:

2 ADP ⇌ ATP + AMP, so [AMP] = K × [ADP]² / [ATP]

Divide by [ATP] and you have the rule that matters:

[AMP] / [ATP] = K × ( [ADP] / [ATP] )²

The AMP:ATP ratio goes as the square of the ADP:ATP ratio. Try round numbers, taking K as 0.5, which lies within the range of reported values. At rest, ATP is 5.0 mM and ADP is 0.5 mM, so AMP is 0.5 × 0.25 / 5.0 = 0.025 mM. Now let work use up a tenth of the ATP. ATP is 4.5 mM, ADP settles near 0.93 mM, and AMP is 0.5 × 0.86 / 4.5 = 0.096 mM. ATP fell by 10 percent and AMP rose almost fourfold. Think of a low-fuel light: the needle has barely moved and the light is on.

AMPK has three subunits. Alpha carries the kinase, beta holds the complex together, and gamma carries the sensor: sites where AMP, ADP and ATP compete to bind, so gamma reads their ratio. The main switch is one threonine in the activation loop of alpha, threonine 172, phosphorylated by an upstream kinase called LKB1. That phosphate raises activity a hundredfold or more. LKB1 is active all the time. What AMP changes is AMPK itself: with AMP on gamma, the phosphate is harder for phosphatases to remove and, by most accounts, easier for LKB1 to add, and the kinase gets a further allosteric push.

Once on, AMPK cuts spending and raises income. It phosphorylates and inactivates acetyl-CoA carboxylase, which stops fat synthesis and, as malonyl-CoA falls, lets fatty acids into mitochondria to be burned. It inactivates HMG-CoA reductase, the rate-setting enzyme of cholesterol synthesis. In muscle it brings more GLUT4 glucose transporters to the cell surface.

Two gauges, one decision: energy can veto growth

AMPK phosphorylates TSC2 at sites different from Akt’s, and here the phosphate makes the TSC complex more active, which drains Rheb of GTP. AMPK also phosphorylates Raptor, which inhibits mTORC1 directly.

Both gauges also write on ULK1, the kinase that starts autophagy. mTORC1 adds a phosphate that keeps it off. In the model built on work published in 2011, AMPK adds phosphates at other sites that switch it on. Later studies complicate AMPK’s side of this, so hold it loosely.

You can now predict three states.

• Fed. Akt silences TSC, Rheb holds GTP, the Rags hold mTORC1 on the lysosome and AMPK is quiet. Protein and lipid synthesis run. Autophagy is off.
• No amino acids. Rheb may still hold GTP, but the Rags release mTORC1 and it leaves the lysosome. Translation slows and autophagy starts, which returns amino acids to the cell.
• No energy. AMP rises and AMPK switches on. Fat and cholesterol synthesis stop, fat burning and glucose uptake rise, and mTORC1 is shut off through TSC2 and Raptor even when growth factors and amino acids are present.

Design problem

A cell should build new protein only when three things are true at once: a growth factor is present, amino acids are plentiful and ATP is plentiful. Design the logic using the parts in this lesson. Say where in the cell each condition is read, and name one mutation that would let the cell ignore each condition.

When a gauge is drugged or broken

Rapamycin. T cells need mTORC1 to grow and divide when interleukin-2, a signaling protein of the immune system, tells them to. Rapamycin, under the drug name sirolimus, blocks that step, and it was approved in the United States in 1999 to prevent rejection of transplanted kidneys. Every other tissue uses mTORC1 too, so mouth ulcers and raised blood sugar are among its side effects.

Tuberous sclerosis. People born with a faulty copy of TSC1 or TSC2 develop benign growths in the brain, kidneys, skin and other organs. In cells without a working TSC brake, Rheb stays on and mTORC1 no longer waits for a growth factor. Drugs of the rapamycin family are used to shrink some of these growths.

Metformin. This is the most widely prescribed drug for type 2 diabetes. What is established: it lowers blood glucose mainly by reducing the liver’s glucose output, and in cells it partly inhibits complex I of the mitochondrial respiratory chain, which raises AMP and switches on AMPK. What is still argued: whether AMPK is the reason blood glucose falls. Mice lacking AMPK in the liver still respond to the drug, and other targets have been proposed in the liver and in the gut.