Stress Signals: p38 and JNK, p53, Unfolded Proteins and Oxidants

Your cells make a protein called p53 all day long, and half of any newly made batch is destroyed within about twenty minutes. Why build a guardian only to shred it? Because a cell in danger cannot wait for a gene to be read and a protein to be built. If the responder already exists and something is holding it back, the cell only has to let go.

The idea: the responder is made all the time and held back

You have met this design before: HIF is made constantly and destroyed while oxygen is present (the lesson on ubiquitin), β-catenin is destroyed until Wnt arrives, and NF-κB waits in the cytoplasm until IκB is destroyed. Think of a car at a red light with the engine running and a foot on the brake. Lifting the foot is faster than starting the engine.

p53 is a transcription factor, a protein that binds DNA and switches genes on. Its restraint is MDM2, an E3 ubiquitin ligase. MDM2 grips a short helix at the front end of p53, which covers the part of p53 that activates genes and gets p53 tagged with ubiquitin and sent to the proteasome.

One of the genes p53 switches on is the gene for MDM2. p53 orders its own destroyer: negative feedback, as in the lesson on signaling circuits.

DNA damage breaks the pair apart. A break across both strands switches on a kinase called ATM. ATM and the kinases beneath it add phosphates to the stretch of p53 that MDM2 grips, and to MDM2 itself. The grip loosens and destruction slows sharply, so p53 piles up without any new gene being read. It then switches on the gene for p21, a protein that blocks the kinases driving cell division, along with genes for DNA repair. If the damage is severe it also switches on genes that kill the cell, the subject of the next lesson.

A cell without working p53 can keep dividing with broken chromosomes, which is why TP53, the gene for p53, is the most frequently mutated gene in human cancer. Most of those mutations fall in the domain that touches DNA. Some tumors, sarcomas among them, keep p53 intact and carry extra copies of the MDM2 gene instead, so the brake never lifts.

Pulses, not a dial: the timing of a signal carries information

Grind up millions of irradiated cells, measure p53 on a blot, and you see it rise and then fade. The natural reading is a dial: more damage, more p53.

Galit Lahav tested that reading by watching one cell at a time. Working with Uri Alon at the Weizmann Institute, she built human cells in which p53 carried a cyan fluorescent tag and MDM2 a yellow one, irradiated them, and filmed them for hours. In single cells p53 rose and fell in separate pulses. The average height and width of a pulse did not depend on the dose. What rose with the dose was the average number of pulses, and genetically identical cells in one dish gave none, one, two or more.

Why had blots hidden this? The cells pulse out of step, so the peaks of some fill the troughs of others. Average a thousand strobe lights flashing at different moments and you record a steady glow.

Does the count matter? In 2012 Lahav’s laboratory used timed doses of nutlin-3, a small molecule that blocks MDM2, to hold p53 steady at about the height of one natural pulse. In the cell line they studied, irradiated cells left to pulse tended to recover and divide again. Cells held steady more often entered senescence, a permanent exit from division. The protein and its level were the same, and only the timing differed.

Design problem

After a dose of radiation, the amount of p53 in a single cell rises and falls in pulses of about the same size. More damage gives more pulses, not bigger ones. Design a circuit of two proteins that gives pulses and not a steady rise. Then say how a cell could use the number of pulses to choose between repairing its DNA and dying.

The stress kinases: the three-tier plan again

In the lesson on Ras and MAPK you followed three kinases in a row: Raf, MEK, ERK. Stress pathways have co-opted this plan, ending in p38 and JNK. Above p38 sit the MAP2Ks MKK3 and MKK6, above JNK sit MKK4 and MKK7, and above those sit many MAP3Ks, each wired to a different kind of trouble. In the standard route, p38 and JNK, like ERK, stay off until a MAP2K adds two phosphates, to a threonine and a tyrosine one residue apart.

What starts them? Ultraviolet light, heat, a sudden change in salt concentration, oxidants, bacterial products, and the inflammatory signals TNF and interleukin-1. One MAP3K, ASK1, repeats this lesson’s design: it is held off by the small protein thioredoxin until an oxidant makes thioredoxin let go.

p38 was reached from several directions in 1994. At Scripps, Jiahuai Han in Richard Ulevitch’s laboratory cloned a protein that gained phosphate on tyrosine when macrophages met lipopolysaccharide, a piece of bacterial outer wall. It also gained phosphate when cells were put in a solution with too much solute, and its sequence resembled HOG1, the kinase yeast use to survive high salt. At SmithKline Beecham, John C. Lee’s group used labeled versions of their own anti-inflammatory compounds as bait and recovered the same kinase. Two more groups reported it that year as a kinase switched on by stress and by interleukin-1.

Remember two jobs of p38. In immune cells it raises the output of TNF and interleukin-1, partly through a kinase beneath it called MK2. MK2 also phosphorylates Hsp27, a small chaperone that protects a stressed cell’s proteins.

JNK is named for its substrate: c-Jun N-terminal kinase. It was found in the early 1990s as a kinase that ultraviolet light and other stresses switch on. It adds phosphates to serines 63 and 73 of c-Jun, the AP-1 partner from the previous lesson, making c-Jun a stronger activator of genes. When stress persists, JNK also pushes a cell toward death, a link the next lesson takes up. In neurons a MAP3K called DLK feeds JNK, and in 2011 Ghosh and colleagues reported that cultured sensory neurons from mouse embryos lacking DLK resist both the axon degeneration and the cell death that follow loss of nerve growth factor.

A sensor that works by being attacked: Keap1 and Nrf2

Some threats are chemicals: oxidants, and electrophiles, molecules short of electrons that react with DNA and protein. The responder is Nrf2, a transcription factor that switches on genes for protective enzymes, including those that join an electrophile to glutathione so it can be excreted. Nrf2 is made constantly. A protein called Keap1 catches it and presents it to a ubiquitin ligase, so in a calm cell Nrf2 lasts well under an hour.

How does Keap1 learn that an electrophile has arrived? It gets hit by one. A cysteine side chain ends in a sulfur atom that gives up electrons easily, and mouse Keap1 has 25 cysteines among its 624 amino acids, a few of them unusually reactive. An electrophile such as sulforaphane, a compound from broccoli, reacts with them. The current view is that the altered Keap1 can no longer pass Nrf2 on for destruction, so newly made Nrf2 goes to the nucleus instead.

Unlike a receptor that fits its ligand in a shaped pocket, Keap1 reads a chemical property, hunger for electrons, so one protein can report on a large family of unrelated compounds. Which cysteine does the sensing depends on the compound. In mice whose Keap1 has serine in place of cysteine 151, sulforaphane has almost no effect, while some other inducers work as before.

Paul Talalay’s laboratory in Baltimore isolated sulforaphane in 1992, and Ken Itoh and Masayuki Yamamoto in Tsukuba reported Keap1 in 1999. Together the groups then showed that inducers react with Keap1’s cysteines. One medicine, dimethyl fumarate for relapsing multiple sclerosis, is an electrophile that turns on Nrf2, though how much of its benefit comes that way is debated.

A sensor inside the folding compartment: repair first, then death

Proteins headed for the cell surface or for export fold inside the endoplasmic reticulum (ER), helped by chaperones such as BiP. If unfolded chains arrive faster than they can be folded, the cell needs to know. The unfolded protein response does this with three sensors, each spanning the ER membrane with one end inside. The leading accounts: BiP sits on the sensors and is pulled away when unfolded chains need it, and unfolded chains also bind IRE1 directly.

IRE1 was tied to this response twice in 1993, by Kazutoshi Mori in Dallas and by Peter Walter’s laboratory in San Francisco, in yeast screens for mutants that could not report a folding backlog. It looked like an ordinary kinase, and both groups first drew a chain of phosphorylations. When unfolded protein builds up, IRE1 molecules cluster, and clustering switches on an RNA-cutting activity in their cytoplasmic ends. In its best-known job, IRE1 cuts a segment out of one messenger RNA, HAC1 in yeast and XBP1 in animals, and a ligase joins the ends. Only the spliced message yields an active transcription factor, which orders more chaperones. In XBP1 the piece removed is 26 nucleotides, not a multiple of three, so the reading frame shifts and the protein gains a new tail.

PERK is a kinase that phosphorylates eIF2α, a factor ribosomes need to begin translation, so protein synthesis slows and fewer new chains enter the ER. ATF6 works like Notch: under stress it moves to the Golgi, proteases cut it, and the freed piece enters the nucleus to switch on chaperone genes.

All three first repair, by lowering the load and raising folding capacity. If the backlog lasts, the same sensors change their message. PERK’s branch switches on CHOP, a transcription factor that promotes death, and IRE1 recruits an adaptor that feeds ASK1 and JNK. As with p53, duration is the information: a short stress calls for repair, and a long one means repair has failed. Heavy secretors depend on this most: children born without working PERK develop diabetes within months of birth, and mice lacking PERK develop diabetes with too few insulin-producing cells in the pancreas.