Ubiquitin and Regulated Proteolysis as Signals

Cells in your body make a transcription factor called HIF-1α and, while oxygen is around, destroy it again within minutes. Why build a protein only to throw it away?

A transcription factor is a protein that switches genes on, and HIF-1α switches on the genes a cell needs when oxygen runs short. Phosphorylation, from the previous lesson, can be undone: a phosphatase takes the phosphate off and the protein is as it was. Destroying a protein cannot be undone. This lesson is about that one-way switch and how a signal writes it.

The idea: a switch you cannot flip back

A phosphate is a sticky note. A kinase puts it on, a phosphatase peels it off, and the page underneath is unharmed. That suits a signal that should come and go. Destruction is the shredder. Once a protein is broken into amino acids, the cell has to make a new copy from the gene.

That makes destruction the right tool when a pathway must move in one direction. Destroying an inhibitor frees whatever it was holding, and the pathway cannot slip back by accident. Destroying an activator ends a signal for good: a cell leaves mitosis because cyclin B, a protein that keeps it in mitosis, is destroyed.

The cell has two parts for the job. Ubiquitin is a small protein, 76 amino acids long, that is attached to other proteins as a tag. The proteasome is a barrel-shaped protease that destroys tagged proteins. The earlier lesson on protein turnover and amino acid catabolism covers the proteasome and the discovery of ubiquitin, so here the question is how a signal decides which protein gets tagged. The answer is a degron: a short sequence or chemical change on a protein that tells the tagging system to destroy it.

The tagging machine: where the specificity lives

Think of a parcel relay: a loading dock, a depot, and a local courier who knows the addresses. Ubiquitin travels through three enzymes the same way, called E1, E2 and E3.

E1, the activating enzyme, spends ATP to join the carboxyl end of ubiquitin (glycine 76) to one of its own cysteines through a thioester, an energy-rich bond between a sulfur and a carbonyl carbon. E1 hands the ubiquitin to a cysteine on an E2, the conjugating enzyme. An E3, the ligase, binds the loaded E2 and the target at once, and ubiquitin ends up on a lysine side chain of the target. The bond is an isopeptide bond: an amide formed with the lysine’s side-chain amine instead of the backbone.

The relay fans out toward the target. Human cells have two E1s, about forty E2s and roughly six hundred predicted E3s. One E1 loads ubiquitin for nearly the whole system, and each E3 recognizes its own set of degrons. The specificity lives in the E3.

E3s come in two designs. A RING-type ligase is a scaffold that holds the loaded E2 next to the target, and the ubiquitin passes straight across. A HECT-type ligase first takes the ubiquitin onto a cysteine of its own. The SCF ligases are RING-type, built from a cullin scaffold, a small RING protein that docks the E2, an adaptor called Skp1, and an F-box protein. The F-box protein is the interchangeable part that binds the degron, so swapping it aims the same machine at different proteins.

Once one ubiquitin is on, the E3 can add more, each joined to a lysine of the one before. Ubiquitin has seven lysines, and which one forms the link gives the chain its meaning. A chain joined through lysine 48 and at least four ubiquitins long is the usual signal for the proteasome.

Writing the mark: a phosphate or an oxygen atom

Most proteins do not carry a degron all the time. A signal writes it. Two examples show the logic.

A phosphate that completes a docking site. Some inhibitor proteins carry a short sequence with two serines. The F-box protein β-TrCP binds that sequence only when both serines carry a phosphate. A kinase that a signal has switched on adds the phosphates, β-TrCP brings the inhibitor to an SCF ligase, a lysine-48 chain follows, and the proteasome destroys the inhibitor. IκB, the inhibitor of the transcription factor NF-κB, is the classic case, and the NF-κB lesson covers its pathway. The Wnt lesson shows the same trick on β-catenin, except that there the signal works by stopping the phosphorylation. The reversible phosphate reads the signal, and the irreversible destruction commits the cell.

An oxygen atom that completes a docking site. The second example answers the opening question.

Gregg Semenza found HIF-1 while studying erythropoietin, the kidney hormone that tells the bone marrow to make red blood cells. Its subunit HIF-1α was made all the time and destroyed almost as fast, so only the amount standing at any instant changed with oxygen. Peter Ratcliffe’s group showed in 1999 that the destruction depends on a protein called VHL, and only when oxygen is present. William Kaelin had reached VHL from the other side. VHL is mutated in an inherited cancer syndrome, and cells lacking it made oxygen-regulated messages at high levels in ordinary air. VHL is the substrate receptor of a cullin-RING ligase, the role an F-box protein plays in an SCF. So VHL had to be reading a degron on HIF-1α, and nobody knew what it was.

In 2001 the two laboratories, knowing they were competitors, each found it independently. The degron is a hydroxyl group on one proline of HIF-1α (proline 564 in the human protein; proline 402 is modified too). A short HIF-1α peptide made with hydroxyproline already in place bound VHL tightly with no cell extract, and the same peptide with an ordinary proline did not. The enzymes that add the hydroxyl are the prolyl hydroxylases, or PHDs. A PHD needs iron, a small molecule called 2-oxoglutarate, and one molecule of oxygen, O2. One atom of the oxygen ends up in the new hydroxyl and the other in succinate, which forms when 2-oxoglutarate loses a CO2.

Oxygen is a substrate here, used up like flour in bread, and the pathway needs no separate oxygen receptor. When oxygen falls, the enzyme finishes the reaction less often, fewer prolines are hydroxylated, VHL has less to grip, and HIF-1α is spared. It builds up, pairs with a partner protein called ARNT in the nucleus, and switches on genes such as erythropoietin and VEGF, a growth factor that drives new blood vessels.

A simple model captures the dynamics. A protein P is made at a constant rate s and destroyed at a rate proportional to how much there is, with rate constant k:

dP/dt = s - k × P, so at steady state P = s / k, and the half-life is 0.69 / k

A large k means a low steady state and a short half-life, and the half-life sets how fast the level can change.

Design problem

A transcription factor must be nearly absent when oxygen is plentiful and present in the nucleus within minutes when oxygen falls. Design the control. Why is destroying it continuously better than making it on demand?

Ubiquitin is a code: the kind of mark matters

One ubiquitin on one lysine, called monoubiquitination, seldom sends a protein to the proteasome. It more often changes where the protein goes. A monoubiquitinated growth-factor receptor can be pulled into the cell by endocytosis, in which the membrane folds inward and pinches off a vesicle.

A chain joined through lysine 63 acts as a scaffold. Proteins with ubiquitin-binding domains, which are small folds that grip ubiquitin, clamp onto the chain, so one chain can gather several proteins into a complex. Signaling complexes assemble this way, and so do repair proteins at breaks in DNA.

These marks can be erased. Deubiquitinating enzymes cut ubiquitin off, so such marks behave more like a phosphate than like the shredder.

SUMO is a close relative. A small ubiquitin-like modifier is a protein of about 100 amino acids folded like ubiquitin and attached to lysines by its own E1, E2 and E3. A SUMO mark usually changes what a protein binds or where it sits, and does not decide whether it survives.

A hormone receptor that is a ligase

The plant hormone auxin tells shoots which way to bend and roots where to grow. Botanists looked for its receptor at the cell surface for decades, because animal biology said a receptor catches the hormone, changes shape, and starts a cascade. The receptor that controls auxin’s effect on genes turned out to be part of the destruction machine.

The clue was a mutant. Arabidopsis plants with a broken TIR1 gene respond poorly to auxin. TIR1 is an F-box protein, the interchangeable part of an SCF ligase, and F-box proteins were thought to grab targets and sense nothing. On 26 May 2005, Estelle’s laboratory at Indiana University and, independently, Ottoline Leyser and Stefan Kepinski at the University of York reported in the same issue of Nature that TIR1 binds auxin itself.

Here is how it works. The genes that auxin switches on are held off by repressor proteins of the Aux/IAA family. Without auxin, TIR1 binds one of them poorly. Auxin fits into a pocket at the exact place where TIR1 and the repressor touch, and the grip tightens. The repressor is ubiquitinated and destroyed, and the genes come on. The 2007 crystal structure, from Estelle’s group with Ning Zheng, showed auxin sitting in the joint between the two proteins. The hormone is part of the joint, a molecular glue, and detecting the hormone and carrying out its effect are a single event.

Drug makers now use the same principle on purpose. Lenalidomide, a relative of thalidomide used in multiple myeloma, binds cereblon, the substrate receptor of a cullin-RING ligase. The bound drug changes cereblon’s surface so that it grips two transcription factors, Ikaros and Aiolos, which it would otherwise ignore, and they are ubiquitinated and destroyed. The drug was in clinical use years before anyone knew this. Chemists also build two-headed molecules called PROTACs, with one end that grips a target and one that grips a ligase such as VHL, using a small molecule built around the hydroxyproline that VHL grips in HIF-1α.