Lipids as Anchors and Addresses

Ras has no stretch of amino acids that crosses a membrane, yet most of it sits on the inner face of the plasma membrane. What holds it there, and why does the cell care where a protein sits?

Ras is a small GTP-binding switch that relays a growth signal inward from a receptor. The answer to the first question is a fatty tail. The answer to the second is that a membrane works as a floor to search on and as a wall that carries labels. This lesson follows both: lipids that anchor proteins, and lipids that tell proteins where they are.

The idea: location is part of the message

Location, location, location. Picture a worker who must find one machine. Searching a whole warehouse takes a long time, and searching a single floor takes less. A membrane is that floor. A protein held on its surface wanders in two dimensions instead of three, so it meets its partners far more often than it would drifting through the cytosol, the watery interior of the cell. Ras needs this. It must meet SOS, the exchange factor that loads it with GTP, and later Raf, the kinase it switches on. SOS is recruited to the membrane by the adaptor Grb2, and Raf is recruited by active Ras.

The inner face of the plasma membrane also differs from the outer face and from the membrane of the Golgi, the stack of flattened sacs that sorts new proteins. The inner face is rich in negatively charged lipids, and a protein can use that charge as a landmark. The lessons on lipids and on membrane structure and function cover the chemistry. You need only one fact from them here: a long hydrocarbon chain is comfortable buried in a bilayer and uncomfortable in water.

That leaves the puzzle. How does Ras mostly hang out at the membrane? Something must be attached to it.

The anchors: lipids can be attached

Cells attach lipids to proteins with dedicated enzymes, and the chemistry of the bond sets the rules.

Anchor	Size	Where it goes	Enzyme	Can it come off?
Myristate	14-carbon fatty acid	N-terminal glycine, through an amide bond	N-myristoyltransferase	No
Palmitate	16-carbon fatty acid	A cysteine, through a thioester	Palmitoyl acyltransferases (PATs)	Yes, by thioesterases
Farnesyl	15-carbon isoprenoid	Cysteine of a C-terminal CAAX box, through a thioether	Farnesyltransferase	No
Geranylgeranyl	20-carbon isoprenoid	Same CAAX cysteine, when the last residue favors it	Geranylgeranyltransferase I	No

Myristate goes on a glycine that becomes the first residue once the starting methionine is clipped off, usually while the protein is still being made. Src-family kinases carry one. Prenyl groups are built from five-carbon isoprene units, three for farnesyl and four for geranylgeranyl. A CAAX box is a cysteine, two aliphatic amino acids, and a last residue that picks the enzyme. After the lipid is on, the last three residues are clipped off and the new end is methylated, which hides its charge. H-Ras, N-Ras and K-Ras all end in a CAAX box.

The bond decides the behavior. Amide and thioether bonds are essentially permanent, so myristate and the prenyl groups stay for the life of the protein. A thioester is easily hydrolyzed, and cells have thioesterases that do it. Palmitate is the one anchor that can be put on and taken off.

One anchor is not enough

A single myristate or farnesyl is a weak grip. It buries one short chain in the bilayer, and chains pop out again. The protein stays tied to the lipid, but the lipid does not stay in the membrane, so a protein held this way visits membranes and leaves them. Cells add a second signal, of one of two kinds:

• A second lipid. The Src-family kinases Fyn and Lck carry myristate plus palmitate. H-Ras and N-Ras carry farnesyl plus palmitate.
• A patch of positive charge. K-Ras4B carries a farnesyl plus a run of lysines, which are attracted to the negative lipids of the inner face.

Binding free energies add, which means affinities multiply:

ΔG(total) ≈ ΔG(first anchor) + ΔG(second signal), so K(total) ≈ K(first anchor) × K(second signal)

Each extra 1.4 kcal/mol at body temperature makes binding about ten times tighter. One hook holds a coat until someone bumps it, and two hooks hold it.

The second signal is also a handle for control. Remove it and the protein drops off. In K-Ras4B, protein kinase C phosphorylates Ser181, next to the lysine patch. The phosphate adds negative charge beside a positive patch, the attraction to the membrane weakens, and K-Ras4B is released. This is a farnesyl-electrostatic switch. A carrier protein, PDEδ, has a hydrophobic pocket that holds a farnesyl group, and it helps farnesylated Ras proteins cross the cytosol between membranes without their tails touching water.

For H-Ras the handle is palmitate, and a thioester that can come off turns the anchor into a cycle. A palmitoyl acyltransferase at the Golgi adds palmitate to H-Ras. The protein, now held tightly, rides membrane vesicles to the plasma membrane. A thioesterase removes the palmitate, and it can do so on any membrane. The grip weakens, and H-Ras drifts from membrane to membrane until it reaches the Golgi again. It’s a kinetic bilayer trap: palmitoylation traps the protein on whatever membrane it is sitting on, transport carries it to the plasma membrane, and depalmitoylation resets it. No binding site specific to the plasma membrane is needed. The rates of the cycle do the sorting.

Some models place palmitoylated proteins in lipid rafts, small cholesterol-rich patches of the bilayer. Whether such patches exist in living cells, and how large and long-lived they are, is debated, so treat rafts as a hypothesis.

How can we exploit this to treat cancer?

If Ras must carry a farnesyl to work, block the enzyme that attaches it. That was the idea behind farnesyltransferase inhibitors (FTIs). Ras genes are mutated in roughly one in five human cancers, and a Ras that cannot reach the membrane cannot signal. One FTI, tipifarnib, reached Phase III clinical trials.

The drugs worked as drugs and still disappointed against the Ras that matters most. K-Ras is by far the most commonly mutated of the three Ras genes, and H-Ras is the least. When farnesylation is blocked, K-Ras and N-Ras can be picked up by geranylgeranyltransferase I, which puts a geranylgeranyl group on the same cysteine. The protein keeps its anchor and keeps signaling. H-Ras is not rescued this way. The logic has not died: tipifarnib has been tested in cancers driven by mutant H-Ras, and the FTI lonafarnib is approved for progeria, a disease in which a farnesylated protein of the nuclear envelope is the culprit.

Lipids as addresses: written by PI3-kinase, read by PH domains, erased by PTEN

Anchors hold a protein on a membrane. Addresses say which membrane and which spot. The labels are phosphoinositides: phosphatidylinositol (PI) lipids whose inositol ring faces the cytosol and can carry phosphates at positions 3, 4 and 5. Cells make seven of these combinations, and they collect in different membranes: PI4P on the Golgi, PI3P on early endosomes, PI(4,5)P2 (PIP2) on the plasma membrane. Phosphates can be added to the ring and removed again quickly, which is what a signal needs.

The first hint that a signal can rewrite a lipid came by accident in the early 1950s. Mabel and Lowell Hokin were following radioactive phosphate into the RNA of pigeon pancreas slices stimulated with acetylcholine. The counts kept vanishing through the purification. Instead of fixing the leak, they analyzed the discarded material, and it was phospholipid. Their 1953 paper lists Mabel Hokin first and has a title that leads with enzyme secretion and names the phospholipids only at the end. Nobody had a place to file the result. In 1975 Robert Michell argued that this lipid turnover was tied to receptors. Where that lipid goes next belongs to a later lesson. Here the question is what a phosphate on one position does.

The label that matters here sits on a place nobody had mapped. Lewis Cantley’s group was studying a lipid kinase activity that came down with a viral oncoprotein such as v-Src or polyoma middle-T antigen. The product ran about a millimetre short of the known 4-phosphate lipid on a thin-layer plate. With a collaborator, Peter Downes, they worked out which position of the inositol ring carried the phosphate, and it was the 3 position. In 1988 they showed the kinase made phosphatidylinositol-3-phosphate, a lipid with no precedent.

In stimulated cells the enzyme, PI3-kinase, adds a phosphate to PIP2 and makes PIP3. The two fatty acid tails stay buried in the bilayer, so PIP3 stays in the membrane. It spreads sideways only a short way before phosphatases remove it, so it marks the patch under the receptor that switched the kinase on.

Proteins read the label with a PH domain, a fold of roughly 100 amino acids, some of which have a pocket for a phosphoinositide head group. Akt carries one that grips PIP3. Before PIP3 appears, Akt drifts in the cytosol. Once PIP3 is there, Akt is pulled onto that patch and no other. What Akt does when it arrives belongs to a later lesson.

You can watch this happen. The method behind “Visualizing PIP3” is a PH domain fused to GFP, which glows green wherever the domain goes. In a resting cell the glow is spread through the cytosol. After a growth factor, it gathers in a patch of membrane.

PTEN runs the reaction backwards. It removes the 3-phosphate from PIP3 and leaves PIP2. In a cell with broken PTEN, the patch stays marked and PH-domain proteins keep arriving. Loss of PTEN is common in human cancers.

Design problem

Build a reporter that lights up the plasma membrane only where and when PI3-kinase is active. What would you fuse to GFP, and what control shows the signal is specific?

One way to do it

Fuse GFP to a PH domain that binds PIP3 tightly and selectively. The PH domain of Akt also binds PI(3,4)P2, so a domain that prefers PIP3, such as that of GRP1, is a cleaner choice. In a resting cell with little PIP3, the glow is spread through the cytosol. When PI3-kinase is switched on, PIP3 builds up in the membrane under the receptor and pulls the reporter into a bright patch there.

Three controls show the signal is specific. First, the same fusion carrying a mutation in the lipid-binding pocket should stay in the cytosol after stimulation, which shows recruitment needs the lipid. Second, pretreating cells with a PI3-kinase inhibitor such as wortmannin should block the recruitment, which shows the lipid comes from PI3-kinase. Third, cells lacking PTEN should show the reporter at the membrane even at rest, which shows the reporter follows the PIP3 level, not the stimulus.

Also image a red fluorescent protein that is free in the cytosol and report the green-to-red ratio at the membrane, since a thick membrane ruffle can look bright for no chemical reason.