GTPase Switches: The Timer

Wash a hormone off a cell and its response fades. Give the cell a GTP look-alike that it cannot cut, and the response stays on through repeated washing. What was keeping time?

The receptor is not the clock. The hormone leaves, the receptor lets go, and something else decides how much longer the cell goes on answering. That something is a protein holding a molecule of GTP and slowly cutting it. This lesson follows how that protein was found and what happens when its timer is broken.

The puzzle: a requirement nobody expected

Around 1970, Martin Rodbell’s laboratory at the National Institutes of Health was studying how the hormone glucagon switches on adenylyl cyclase in liver membranes. Adenylyl cyclase is the enzyme that makes cyclic AMP, the small messenger that tells the inside of a cell a hormone has arrived. The glucagon-binding experiments behaved strangely: in plain buffer the hormone stuck to the membranes and would not wash off, but when ATP was in the buffer it came off within seconds, which was odd, since ATP is the enzyme’s fuel. The cause was a contaminant: commercial ATP carries a little GTP. When the group added GTP on purpose, it proved to be required. Without it, glucagon barely switched the cyclase on.

Think about what that implies. GTP was not the fuel, because the cyclase runs on ATP. And a trace was enough, so GTP was not being used up once for every cyclic AMP made. It was being used over and over.

For years almost no other laboratory could reproduce the effect, because most were not working with membranes as clean as Rodbell’s. Suppose the ATP is 1% GTP. Then the “no GTP” tube already holds enough for a full response, the difference between tubes never appears, and adding more GTP only scatters the data.

Rodbell had a second problem. Many different hormone receptors switch on the same enzyme, and he thought it impossible that all of them were built into it. So he proposed a separate component between receptors and enzyme, passing the signal from one to the other. He called it a transducer. Nobody knew what it was.

Here is a UVa moment: the answer came from the pharmacology department at the University of Virginia, in the 1970s, in the laboratory of Alfred Gilman. He used cyc⁻ lymphoma cells, a line of cancer cells in which hormone cannot raise cyclic AMP because the link between receptor and enzyme is missing. Gilman and his postdoctoral fellow Elliott Ross added a soluble extract from normal cells to cyc⁻ membranes. The hormone response came back, and it came back even when the extract’s own cyclase had been inactivated. So the transducer was a separate protein, one that could be moved from one membrane to another. Using that rescue as an assay, Gilman and his colleagues purified Gs, the stimulatory G protein, from rabbit liver membranes. The pure protein carried an alpha and a beta subunit, and a small gamma subunit, unnoticed at first, was added to the picture later. G protein is short for guanine-nucleotide-binding protein.

The idea: a switch with a timer built in

The definition of a G protein is short: on when it holds GTP, off when it holds GDP. GTP carries three phosphate groups in a row, and GDP is the same molecule with the last one removed. Here is the cycle for a G protein that sits under a receptor.

1. A hormone binds a receptor in the membrane.
2. The occupied receptor does one thing. It pulls a spent GDP out of a nearby G protein, and a fresh GTP takes its place. GTP is far more plentiful than GDP in the cell, so it usually wins. Any protein that does this job is a guanine nucleotide exchange factor, or GEF.
3. Holding GTP forces the G protein into a shape that switches on its target, here adenylyl cyclase.
4. The G protein is also a weak enzyme, and the GTP it holds is its substrate. Sooner or later it cuts off the third phosphate. That leaves GDP, and the shape falls back to off.
5. It cannot reload itself. Another occupied receptor must pull out the GDP first.

A GTPase is simply an enzyme that cuts GTP, and step 4 is slow. A GTPase-activating protein, or GAP, binds the G protein and makes it cut faster. Because the G protein drifts freely in the membrane, one occupied receptor can load many of them in succession, and each can switch on a cyclase of its own.

average time on = 1 / k(cut), where k(cut) is the rate constant for cutting the bound GTP

That is why a G protein is a timer as well as a switch. Step 2 turns it on. Step 4 decides how long it stays on, and the hormone and the receptor do not appear in the equation. Say a G alpha subunit has k(cut) = 0.05 per second, a round number chosen for illustration. It stays on for about 20 seconds on average. A GAP that speeds the step 40-fold cuts that to half a second.

GEF: loading rate 2 (× the unaided cutting rate) GAP: speeds cutting 1-fold (1 = no GAP) Break the timer: GTP cannot be cut

Fraction on (sketch)

—

One molecule stays on

—

The off step sets the duration: the GEF turns the switch on, and how fast GTP is cut decides how long it stays on. The picture is a sketch of the logic. Only the 20 s example (an illustrative round number for a G alpha subunit) and the 40-fold speed-up (the lesson’s lower bound for RGS4 and GAIP on Gi alpha-1) come from the lesson.

Now the washout puzzle has an answer. While hormone is present, occupied receptors keep reloading G proteins. When the hormone is gone, nothing reloads them, and each protein that is still on runs out its own clock. Steps downstream, such as the breakdown of cyclic AMP, add delays of their own. This lesson tracks the G protein’s delay. The rule to keep: the off step sets the duration.

Two main flavors: trimeric and small

G proteins come in 2 main flavors, and they share the cycle.

	Trimeric G proteins	Small GTPases such as Ras
Build	Alpha, beta and gamma subunits; alpha holds the nucleotide	One small protein, about 20 to 25 kilodaltons
Loads GTP (GEF)	The occupied receptor itself	SOS
Speeds the off step (GAP)	RGS proteins	Neurofibromin and p120 GAP
Examples	Gs, Gi, Gq	Ras, Rho, Rab, Arf, Ran

Different alpha subunits point at different targets. Gs switches adenylyl cyclase on, Gi turns it down, and Gq switches on a different enzyme, phospholipase C-beta. The GAPs are the RGS proteins, short for regulators of G-protein signaling. RGS4 and a relative called GAIP make a Gi-family G protein cut its GTP at least 40 times faster in vitro. GAIP binds the active, GTP-holding form of a G protein, so it can act only while the switch is on. A cell can therefore shorten a Gi-coupled response simply by making more of them.

Ras is a single small chain with no receptor attached. SOS, short for son of sevenless, is its exchange factor, and growth-factor receptors recruit it to the membrane. Left alone, Ras cuts its GTP slowly. In 1987 Frank McCormick and Marilyn Trahey found the first GAP, a factor that sped normal Ras’s GTP cutting more than 200-fold in vitro and did nothing to the oncogenic mutant. Two years later Alfred Wittinghofer’s group in Heidelberg solved the structure of Ras holding a GTP analogue it could not cut, and its fold matched EF-Tu, a bacterial protein that delivers tRNAs to the ribosome. The same ancient fold has been used twice.

Ras is one of five families that run this same cycle, each with its own GEFs and GAPs:

• Ras: cell growth and proliferation.
• Rho: the actin cytoskeleton and cell shape.
• Rab: vesicle traffic between membranes.
• Arf: the budding of vesicles, also membrane traffic.
• Ran: movement of proteins into and out of the nucleus.

Breaking the timer

If the off step sets the duration, anything that blocks it should lengthen the response, and anything that speeds it should shorten it. Each has been tested.

An uncuttable GTP. In Gpp(NH)p the oxygen bridging the last two phosphates is replaced by NH, so the G protein can bind it and cannot cut it. In 1975 Schramm and Rodbell treated frog red blood cell membranes with isoproterenol, a drug that activates the adrenaline receptor, together with Gpp(NH)p. Adenylyl cyclase switched on and stayed on through extensive washing, sonication and detergent, and propranolol, which blocks the receptor, could no longer reverse it. With isoproterenol alone, diluting the drug dropped activity to nearly nothing.

Cholera toxin. In 1976 and 1977 Joel Moss and Martha Vaughan showed that the toxin is an enzyme. It splits NAD and can attach its ADP-ribose group to the amino acid arginine. In 1978 D. Michael Gill and R. Meren, and in a separate study Cassel and Pfeuffer, showed that the target is the cyclase system’s GTP-binding protein, now called Gs. The group blocks the step that switches Gs off, so Gs stays on and cyclic AMP keeps rising. A cell lining the intestine responds by pumping salt into the gut, water follows the salt, and the result is the watery diarrhea of cholera.

Ras G12V and Q61. Glycine 12 sits where the GAP reaches into the pocket. When it is replaced by valine, Ras ignores the GAP. A GAP works partly by pushing an arginine into the pocket, and the bulkier side chain gets in the way. Glutamine 61 (changed to lysine in the tumor mutant Q61K) helps stabilize the transition state of the cutting reaction, so Q61 mutants cut poorly, with or without a GAP. In both, GTP stays in the pocket far longer than normal, so Ras keeps signaling. A single amino acid change at codon 12 turns up in many different cancers.

The mirror image: S17N. This mutant breaks the other half of the cycle. Asparagine in place of serine 17 lowers Ras’s affinity for GTP 20- to 40-fold while leaving its affinity for GDP largely unchanged, so Ras cannot be loaded. It also holds on to SOS, which soaks up exchange factor that normal Ras in the same cell needs. A mutant that blocks its normal partner this way is called dominant-negative.

Design problem

Predict how long the response lasts, compared with normal, in four cases: an uncuttable GTP analogue; cholera toxin; extra RGS protein; and Ras G12V with normal GAP present. Give one rule that covers all four.

One way to do it

With the uncuttable analogue, the response is far longer than normal. The G protein cannot cut what it holds, so it outlasts the hormone, the washing and the detergent.

With cholera toxin, the response is far longer. Gs cannot switch itself off, so cyclic AMP keeps rising and the intestinal cell keeps pumping salt.

With extra RGS protein acting on a Gi-family subunit, the response is shorter. Each loaded G protein spends less time on, and the hormone and receptor are unchanged.

With Ras G12V and normal GAP, the response is far longer. The GAP is present and the mutant ignores it, so GTP stays in the pocket far longer than normal.

One rule covers all four: the off step sets the duration. Block it, by an analogue, a toxin or a mutation, and the response lengthens. Speed up a Gi-family switch with RGS, and the response shortens. The hormone and the receptor only start the clock.