Turning the Signal Off: Arrestins, RGS Proteins and Phosphodiesterases

Adrenaline is still sitting in its receptor, and within minutes the cell’s response to it has faded. Nothing has taken the hormone away. So what switched the receptor off, and how would you prove it?

The idea: every step that starts a signal needs its own way to end

You followed the “on” pathway in the biochemistry course: adrenaline, receptor, G protein, adenylyl cyclase, cyclic AMP, protein kinase A, phosphorylated targets. Each step leaves something behind that would keep signaling if nothing removed it, so the cell has an off switch at each one. From the outside in:

1. The ligand is removed. The hormone falls off the receptor and is carried away.
2. The receptor is uncoupled. It is still occupied but can no longer reach its G protein.
3. The receptor is taken inside. It leaves the surface in a vesicle.
4. The G protein’s off step is sped up. A helper protein makes G-alpha cut its GTP sooner.
5. The messenger is destroyed. A phosphodiesterase breaks the ring of cyclic AMP.
6. The phosphates are taken off. Phosphatases reset the target proteins.

The second is the puzzle in the opening line. A response that fades while the hormone is still bound is called desensitization, and its explanation came from a purification that seemed to be going wrong.

The purer kinase still adds phosphate but no longer switches the receptor off

Robert Lefkowitz’s laboratory at Duke University had one firm clue: phosphate is added to the beta-adrenergic receptor as its response fades. The obvious suspect was protein kinase A, the kinase that cyclic AMP itself switches on.

Jeffrey Benovic, a graduate student in the laboratory, tested the suspect with the kin⁻ line of S49 mouse lymphoma cells, which has no working protein kinase A. He mixed an extract of those cells with hamster lung receptor, rebuilt into lipid vesicles, and radioactive ATP. Phosphorylation of the receptor increased when the agonist isoproterenol was present, and the antagonist alprenolol blocked the increase. So another kinase was at work, and it recognized the receptor’s active shape.

He purified that kinase from cow brain cortex, enriching it 20,000 times. The pure kinase still added phosphate to the receptor, as the crude material had. But a batch purified only 350 times had cut the receptor’s ability to activate its G protein by up to 80%, and the pure enzyme cut it by only 10 to 15%.

Stop and work out what that means. The phosphates are the same and the effect has nearly vanished, so the phosphate cannot be the off switch by itself. Something else in the cruder batch was doing the switching, and each column removed more of it.

The missing piece came from studies of vision. Hermann Kühn, at Jülich in West Germany, had described an abundant retinal protein, then called the 48K protein, that works with rhodopsin kinase to shut off rhodopsin, the light receptor of the rod cell. It was later renamed arrestin, because it arrested rhodopsin’s function. Lefkowitz arranged for Kühn to send some to Duke.

Benovic added it to the reconstituted system. With the pure kinase alone, each receptor carried about 8 phosphates and could still switch on Gs. With kinase and arrestin together, its activity fell to about 59% of the untreated receptor’s, a loss of about 40%. The controls show the mechanism:

• Bovine serum albumin, an unrelated protein, could not replace arrestin.
• Heat-denatured arrestin could not replace it either.
• Arrestin added to a receptor the kinase had never marked did nothing. Activity stayed at 98.1% of the value without arrestin, across 13 measurements.

So there are two separate requirements, a mark and a reader. The kinase writes the mark, and only on an active receptor; enzymes of this kind are now called G-protein-coupled receptor kinases, or GRKs. Arrestin reads the mark and sits where the G protein needs to sit.

The retinal protein was a stand-in. Three years later Martin Lohse, a fellow in the same laboratory, found the version that works outside the retina, now called beta-arrestin. It reduced signaling by the kinase-marked receptor by more than 75% and did not act on rhodopsin.

Design problem

Design a scheme to down-regulate a receptor using PKC and PKA. Say which signals each kinase reports, why this differs from the kinase that marks only the occupied receptor, and what the cell gains by having both.

Arrestin carries the receptor into the cell

Arrestin also throws the third switch on the list. Clathrin is the protein that cages a patch of surface membrane and pulls it inward as a coated pit. In 1996 Benovic’s laboratory, by then at Thomas Jefferson University in Philadelphia, reported with James Keen and Vsevolod Gurevich that beta-arrestin and arrestin-3 bind clathrin directly and tightly. Hybrid arrestins that could not bind the receptor, or could not bind clathrin, carried the receptor into the cell less well. In intact cells the receptor, arrestin and clathrin came together in an agonist-dependent way. So the protein that blocks the G protein’s binding site also pulls the receptor into a coated pit.

A line from the lecture fits here: “GPCRs can be unfaithful to trimeric g proteins.” Once arrestin has taken the G protein’s place, the receptor has a new partner with contacts of its own. Besides clathrin and the adaptor AP-2, arrestin can bind the kinase Src and parts of the ERK and JNK kinase cascades. How much signaling this allows with no G protein at all is still argued.

Going inside does not settle the matter either. A receptor in an endosome, the compartment that receives incoming vesicles, was long assumed to be out of service. In 2013 Roshanak Irannejad, in Marc von Zastrow’s laboratory, found the receptor and its G protein switched on in the early endosome membrane too, which a later lesson takes up.

Constant GnRH shuts the pituitary down

Gonadotropin-releasing hormone, GnRH, is a peptide from the hypothalamus that tells the pituitary gland to release the gonadotropins, the hormones that drive the ovaries and testes. In 1978 Belchetz and colleagues studied rhesus monkeys whose hypothalamic lesions had abolished gonadotropin release. They gave GnRH either continuously or once an hour. Hourly pulses restored gonadotropin secretion, and a constant infusion did not. Clinics use the same effect on purpose: they keep a GnRH agonist present to shut gonadotropin output down.

Why would a steady signal fail where pulses work? One general possibility is that a cell’s off switches build up while a signal is on and need a gap to reverse. Pulses would give the cell that gap, and a constant signal never would.

Which off switch does it in the pituitary? Not the one this lesson has taught. The mammalian GnRH receptor has no C-terminal tail. In 1999 Willars and colleagues showed that this makes it resist phosphorylation in response to agonist, and resist rapid desensitization. It also works through Gq and phospholipase C, so cyclic AMP is the wrong messenger here. A kinase mark and an arrestin cannot be the whole explanation.

Two kinds of candidate remain. A failure at the receptor would show up as fewer receptors on the surface, or as less second messenger made per dose. A failure downstream would leave the receptor’s own output normal while hormone release falls. Measure second messenger output during the shutdown and you can tell the two apart. Published work answers only part of this question.

The other off switches, and one rule for how receptors are switched off

RGS proteins. You met these in the lesson on GTPase switches: an RGS protein binds a G-alpha subunit while it holds GTP and makes it cut the GTP sooner. RGS4 and GAIP speed the cut at least 40 times for Gi alpha-1, and act on other members of the Gi family too, so a cell can shorten a response by making more of them. Marilyn Farquhar’s laboratory asked where in the cell GAIP sits. It is held on a membrane by palmitate, a fatty acid bonded to the protein, and in 1998 the laboratory reported where it sits: on clathrin-coated vesicles, some budding from the surface and some from the Golgi. It was the first GTPase-activating protein found on any intracellular membrane.

Phosphodiesterases. A phosphodiesterase cuts one bond of the cyclic phosphate ring, and cyclic AMP becomes ordinary AMP, which does not switch on protein kinase A. If breakdown is proportional to the amount of messenger present, then

messenger level at steady state = rate of synthesis ÷ rate constant of breakdown

An inhibitor makes the bottom of that fraction smaller, so the messenger climbs higher and lingers longer. It cannot start a signal: if synthesis is zero, the level stays at zero.

Two drugs show this. Caffeine inhibits phosphodiesterases in a test tube. Be careful with the coffee story, though: at the concentration a few cups produce, most of its effect is thought to come from blocking adenosine receptors. Sildenafil is the cleaner case. It inhibits phosphodiesterase 5, which breaks down cyclic GMP, a close relative of cyclic AMP. In the smooth muscle of some blood vessels, nitric oxide switches on cyclic GMP synthesis and the muscle relaxes. Sildenafil makes that signal larger and longer once nitric oxide has started it.

Phosphatases. A protein that protein kinase A has phosphorylated stays switched until a phosphatase removes the phosphate, as you saw in the lesson on the phosphorylation switch.

One rule. Dedicated proteins switch a receptor off by acting on the receptor itself, and where the receptor sits does not tell you whether it is still signaling. The GnRH receptor has no tail to mark and still goes quiet under a constant signal, so the mark-and-reader route is one of several.

Two findings would break this rule. The first is a receptor that goes silent while its hormone is still bound and nothing is acting on it: no mark, no reader, and unchanged coupling to the G protein. The second is a receptor that stops signaling every time it leaves the surface, in every cell you test.