Cytokine Receptors and JAK–STAT

Add interferon, a protein that warns cells of a virus, to human cells and a particular set of genes switches on within minutes, even when a drug has stopped the cells from making any new protein. Every part of the path from the cell surface to those genes must therefore be in the cell already, waiting. The people working on it judged the second messengers of Unit 3 a poor fit: a rise in a second messenger is a fairly nonspecific event, and interferon’s action is highly specific. The answer was a receptor that borrows a kinase and a transcription factor that waits in the cytoplasm for one phosphate.

The idea: a receptor with no enzyme borrows one

A cytokine is a small secreted protein that tells other cells what to do, most often cells of the blood and the immune system. Interferons are cytokines. So are the interleukins, which carry messages between white blood cells. Several hormones use receptors of the same family: erythropoietin (EPO), which tells the bone marrow to make red blood cells, and also growth hormone, prolactin and leptin.

Each receptor chain crosses the membrane once. A receptor tyrosine kinase, such as the insulin receptor from the receptor field guide, is its own kinase. A cytokine receptor has no enzyme in it at all. A stretch of its inside tail, near the membrane, holds a separate tyrosine kinase called a JAK (Janus kinase), and the JAK is there before any cytokine arrives. Think of an antenna with no electronics of its own and a radio bolted to its base.

Binding does one job: it sets two receptor chains, and so two JAKs, in the right position against each other. One growth hormone molecule has two different faces, and each face grips one copy of the same receptor chain. For some receptors, the EPO receptor among them, structural evidence suggests the chains are loosely paired beforehand and binding re-sets their geometry. Either way, two JAKs end up close enough to phosphorylate each other.

A JAK has two kinase-like domains side by side. One is a working kinase. The other, the pseudokinase domain, has the kinase fold but little or no catalytic activity, and it restrains the working one. The family is named for Janus, the Roman god with two faces. Mammals have four: JAK1, JAK2, JAK3 and TYK2.

The interferon puzzle: other activation besides expression?

A cell can raise a transcription factor’s activity by making more of it. The inhibitor experiment rules that out here, so the factor must exist already and be activated some other way. Two laboratories went looking, from opposite ends.

James Darnell’s laboratory at Rockefeller University started at the gene. Interferon-responsive genes share a short DNA sequence, and the laboratory identified a protein complex, ISGF3, that binds it. Its proteins wait in the cytoplasm and gain the ability to bind DNA within minutes of interferon reaching the cell. In August 1992 Christian Schindler, Ke Shuai and their colleagues reported that interferon causes proteins of the complex to be phosphorylated on tyrosine, and that staurosporine, a drug that inhibits many kinases, blocks the interferon-induced activation of the complex. Some kinase had to act first. The experiment could not say which.

George Stark and Ian Kerr’s laboratory in London started with genetics. Sandra Pellegrini built human cells in which an interferon-responsive stretch of DNA controlled a marker gene, so that in the right medium only cells deaf to interferon survived. She mutagenized the cells and screened about a hundred million of them. Each mutant line that came through had lost a part of the pathway. To find which part, you put DNA from normal cells into the mutant and look for the rare cell whose response returns, a test called complementation. Pellegrini did this in her own new laboratory at the Institut Pasteur in Paris. In July 1992, two weeks before the Darnell paper, her group reported the gene that rescued one mutant. It encoded TYK2.

The two results fit together: a JAK is required, and the DNA-binding proteins carry phosphotyrosine. Darnell’s laboratory named those proteins STATs, signal transducers and activators of transcription.

The mechanism: one phosphate, and an SH2 domain used twice

Try this before reading on: design a signaling pathway that is activated by receptor oligomerization and tyrosine phosphorylation. (Oligomerization means several chains coming together.) The cell’s version has six steps.

1. The cytokine sets two receptor chains in position.
2. The two JAKs phosphorylate each other on their activation loops, the switch described in the phosphorylation lesson, and become fully active.
3. The JAKs phosphorylate tyrosines along the receptor tails.
4. A STAT from the cytoplasm docks on one of those phosphotyrosines by its SH2 domain, the phosphotyrosine-binding module from the lesson on protein interaction modules.
5. The JAK phosphorylates the docked STAT on a single tyrosine near its C-terminal end (tyrosine 701 in STAT1).
6. Two phosphorylated STATs pair, each SH2 domain holding the other’s phosphotyrosine. The pair enters the nucleus, binds a specific DNA sequence and switches on transcription.

The SH2 domain is used twice. It first holds the STAT on the receptor, next to the JAK. Once the STAT carries its own phosphotyrosine, the same domain holds a partner STAT instead, and two grips hold the pair where one held the STAT to the receptor.

Count the proteins between the receptor complex and the gene: one. With no second messenger and no chain of kinases, this is about the shortest route there is from the cell surface to a gene.

Design problem

Design the shortest possible path from a receptor at the cell surface to a gene in the nucleus. Say what each part must be able to do. Then compare your design with a STAT protein, domain by domain, and say what this short path gives up when you set it beside a three-kinase cascade.

One family, many jobs: four JAKs, seven STATs

Mammals have four JAKs and seven STATs (STAT1, 2, 3, 4, 5A, 5B and 6) to serve dozens of cytokines. The receptor does the sorting. Each receptor tail holds particular JAKs and carries particular docking tyrosines, and the amino acids around each phosphotyrosine decide which STAT’s SH2 domain fits.

Mice with one gene deleted, called knockout mice, show what each STAT is for:

• STAT1: no response to interferons; the mice succumb to infections that normal mice survive.
• STAT4: impaired response to interleukin-12.
• STAT6: impaired response to interleukin-4.
• STAT5A and STAT5B: failed responses to prolactin and growth hormone; the two partly cover for each other.
• STAT3: the embryo dies early, so its jobs were worked out one tissue at a time.

You might expect the JAKs to cover for each other as well, since the four are similar in sequence. Two results say that on at least two receptors they do not.

James Ihle’s laboratory had shown in 1993 that JAK2 is the kinase on the EPO receptor’s tail. In 1998 it reported mouse embryos lacking JAK2. They were anaemic and died at about day 12.5 of gestation, and their fetal livers, where a mouse embryo makes its adult-type red cells, held no red-cell precursors that could be grown in culture. Klaus Pfeffer’s laboratory reported the same result in the same issue. The genes for JAK1, JAK3 and TYK2 were untouched, and none of those kinases took over.

The human case is severe combined immunodeficiency (SCID), in which a child is born with almost no working immune system. In 1993 Warren Leonard’s laboratory reported that boys with the X-linked form carry mutations in the gamma chain of the interleukin-2 receptor. The disease is far worse than losing interleukin-2 alone, because the receptors for interleukins 4, 7, 9, 15 and 21 use the same chain, now called the common gamma chain. Then a patient studied with Rebecca Buckley of Duke University had the disease with a normal gamma chain. In 1995 John O’Shea’s and Leonard’s groups found that patient’s mutation in JAK3, the kinase that sits on the gamma chain. A group in Italy working with O’Shea had reported JAK3 mutations in other SCID patients about two months before.

Switching it off, and what happens when a JAK is stuck on

Two things end the signal. Tyrosine phosphatases remove the phosphates from the JAKs, the receptor tails and the STATs. A STAT that loses its phosphate in the nucleus returns to the cytoplasm, ready for reuse.

The pathway also makes its own brake. Among the genes that STATs switch on are those for the SOCS proteins, short for suppressors of cytokine signaling. A SOCS protein has an SH2 domain, so it goes to the phosphotyrosines on the active receptor and JAK. Some SOCS proteins block the JAK’s catalytic site directly. All of them carry a region that recruits a ubiquitin ligase, which can mark the receptor complex for the proteasome. This is negative feedback as you met it in the signaling circuits lesson: more signal makes more brake, after a delay.

In 2005 several research groups reported the same mutation in most patients with polycythaemia vera, a disease in which the bone marrow makes too many red blood cells: valine 617 of JAK2 changed to phenylalanine. Residue 617 lies in the pseudokinase domain. With that restraint weakened, JAK2 signals from the EPO receptor with little or no EPO present.

Missing JAK2 gives no red cells and overactive JAK2 gives too many, so JAKs became drug targets. Ruxolitinib, which inhibits JAK1 and JAK2, was approved in the United States in 2011 for myelofibrosis, a bone-marrow disease related to polycythaemia vera. Tofacitinib followed in 2012 for rheumatoid arthritis. Its logic came from the SCID patients: a person without JAK3 has an immune defect and little else, so blocking JAK3 should quiet an overactive immune system and spare other tissues. Like imatinib, these drugs sit in the kinase’s ATP pocket. The four JAK pockets are alike, so no drug hits only one, and anaemia from partly blocked JAK2 is a known side effect of several.