Signaling to the Nucleus: CREB, AP-1 and Immediate Early Genes

Add a growth factor to a resting cell and, within minutes, the cell is transcribing a gene called c-fos. Stop every ribosome in the cell first, and the gene still switches on. So whatever switched it on was already made and waiting. This lesson gathers the last step of the pathways in this course: what a signal does to a transcription factor when it arrives.

Other activation besides expression? The factor is already made

A transcription factor is a protein that binds a short, specific DNA sequence near a gene and changes how often that gene is copied into RNA. A cell could control a factor by making more of it. That works, but it is slow, since a gene must first be transcribed and its RNA translated. Prof. Deppmann’s question, “Other activation besides expression?”, asks what else a signal could do. Most pathways end by changing a factor that is already made, and they do it in one of four ways.

What the signal does	Example	Where you met it
Adds a phosphate that lets the factor hold a co-activator	CREB	This lesson
Removes phosphates so the factor can enter the nucleus	NFAT	This lesson
Destroys the factor’s inhibitor	NF-κB, freed when IκB is destroyed	The Toll-like receptor and NF-κB lesson
Makes the messenger itself the factor	STATs, Smads, the cut piece of Notch	The JAK-STAT, TGF-β and Notch lessons

The first row has other members; Elk-1, which ERK phosphorylates in the Ras-to-MAPK lesson, is one.

In all four, the slow step of making the factor is done in advance, so the signal only has to flip it.

CREB: one serine, two messengers

CREB stands for cAMP response element-binding protein. The cAMP response element, or CRE, is an eight-letter DNA site, TGACGTCA, that sits in front of many genes switched on by cAMP. CREB binds this site as a pair, and it is largely there before any signal arrives. Being on the DNA is not enough to switch the gene on.

The mechanism of CREB activation starts where the cAMP and PKA lesson stopped. cAMP frees the catalytic subunits of PKA, and some of them enter the nucleus. There PKA puts a phosphate on one serine of CREB, serine 133. That phosphate is a handle. A co-activator is a protein that does not bind DNA itself but is brought to a gene by a factor that does. The co-activator CBP (CREB-binding protein) has a small domain called KIX that grips CREB far more tightly when the phosphate is on. Once held at the gene, CBP adds acetyl groups to the histones that package the nearby DNA, which loosens them, and it contacts the machinery that brings in RNA polymerase.

Now the second messenger. In the calcium lesson, calcium bound calmodulin, and the pair switched on a family of kinases. One of them, CaMKIV, is found largely in the nucleus, and it phosphorylates CREB on the same serine 133. So CREB is a place where two messengers meet: either one adds the handle. Growth factors reach the same serine as well, through kinases that ERK switches on.

Prof. Deppmann asks you to “draw an Ion channel mechanism for activating CREB”. In a neuron it runs like this. Activity depolarizes the membrane. L-type voltage-gated calcium channels open, and so do NMDA receptors, which are glutamate-gated channels that pass calcium. Calcium enters and binds calmodulin, and calcium-calmodulin switches on CaMKIV. CaMKIV phosphorylates serine 133, CREB holds CBP, and genes with a CRE are transcribed.

Design problem

Design a transcription factor that can be activated by cAMP or calcium. Draw the pathway for each messenger, from the cell surface to the gene. Then say how a cell could require both messengers at once, instead of either.

One way to do it

For “either”, copy CREB: a DNA-binding region that keeps the factor on its site, and one serine that two kinases can reach.

Pathway one: hormone, G-protein-coupled receptor, Gs, adenylyl cyclase, cAMP, PKA into the nucleus, phosphate on the serine, co-activator held, gene on.

Pathway two is the channel route above: depolarization, calcium channel, calmodulin, CaMKIV, phosphate on the same serine.

Either kinase is enough, so the factor computes OR.

For “both”, split the job into two steps that need different messengers. Let calcium control location, as in NFAT: phosphates hide the factor’s nuclear import signal until calcineurin removes them. Let cAMP control activity: once in the nucleus, the factor still needs PKA to phosphorylate the serine that holds the co-activator. Calcium alone puts a factor in the nucleus that cannot recruit help. cAMP alone phosphorylates a factor that is still shut out. Only the two together give transcription, so the factor computes AND.

Basic Leucine Zipper Factors: a zipper that makes pairs

CREB belongs to the bZIP family, short for basic leucine zipper. The name describes two stretches of one long α-helix.

The zipper stretch has a leucine at every seventh position. An α-helix turns once about every 3.6 residues, so every seventh residue lands on nearly the same side, and the leucines form a stripe down the helix. Two such helices lay their stripes together and wind around each other, which hides the leucines from water. That is the zipper, and it holds two proteins together as a pair.

The basic stretch is rich in lysine and arginine, which carry positive charge. Beyond the end of the zipper the two helices part, and each basic stretch lies in the major groove of the DNA and reads half of the site. Picture chopsticks held together at the top, with the DNA gripped between the tips.

Why build a factor that works only as a pair? Because pairs multiply the choices. Each partner reads one half-site, so changing a partner changes which DNA the pair prefers and what it does there. The best-known case is AP-1 (activator protein 1), a pair drawn from the Fos and Jun families. Jun can pair with Jun. Fos does not pair stably with Fos, so it needs a Jun, and a Fos-Jun pair holds DNA more tightly than a Jun-Jun pair. AP-1 prefers a seven-letter site, TGACTCA, one letter shorter than the CRE. With four Fos proteins and three Jun proteins, seven genes give well over a dozen AP-1 pairs.

n proteins that can all pair with one another make n(n + 1) / 2 different pairs

Humans have roughly fifty bZIP proteins, which would give 1,275 pairs if all paired freely. They do not: the residues beside each leucine stripe decide which helices fit together. Prof. Deppmann’s own earlier work includes a cross-species review, with Alvania and Taparowsky (Molecular Biology and Evolution, 2006), that compiles which bZIP proteins pair with which in humans, fruit flies, a plant and yeast.

Calcineurin and NFAT: a phosphatase opens the door

CREB is switched on by adding a phosphate. NFAT, the nuclear factor of activated T cells, is switched on by taking phosphates off. In a resting T cell NFAT stays in the cytoplasm, carrying many phosphates that hide its signal for import into the nucleus. When the T-cell receptor fires, calcium rises and stays high, through the store-operated entry of the calcium lesson. Calcium-calmodulin switches on calcineurin, a protein phosphatase. Calcineurin strips the phosphates from NFAT, the import signal is exposed, and NFAT enters the nucleus. There it binds DNA beside a Fos-Jun pair on genes such as interleukin-2, whose product tells the T cell to divide. When calcium falls, kinases put the phosphates back and NFAT leaves.

Much of this was learned from two drugs. Cyclosporin A and FK506 both shut T cells down, and both were being given to transplant patients before anyone knew what they touched inside a cell. Chemically they are unalike. In 1989 a group at Merck and the group of Stuart Schreiber each reported the small cell protein that FK506 binds, later named FKBP12. Cyclosporin A binds a different one, cyclophilin. In 1991 Schreiber’s group and collaborators reported where the drugs meet: each drug, held by its helper protein, binds and inhibits calcineurin.

A test-tube result leaves a question open. Is that why the T cell shuts down? In 1992 Neil Clipstone and Gerald Crabtree answered it by making T cells produce extra calcineurin. Those cells were more resistant to both drugs, and NFAT-driven transcription went up. If more enzyme means more drug is needed, the enzyme is the target. Jam calcineurin and NFAT keeps its phosphates and stays in the cytoplasm, so the interleukin-2 gene stays quiet.

The first genes to answer

Return to c-fos. An immediate early gene is one whose transcription rises within minutes of a signal and needs no new protein. The test is the one in the opening, done with a ribosome-blocking drug such as cycloheximide: does the mRNA still rise? For c-fos it does, because its control region carries a site for CREB and a site for the factors that ERK switches on through Elk-1. cAMP, calcium and growth factors can each reach the gene through factors already in place.

In 1984 Michael Greenberg and Edward Ziff reported that c-fos transcription rises within minutes of adding growth factors to resting fibroblasts and soon shuts off again. The Fos protein that is made pairs with Jun, and that AP-1 pair switches on a second wave of genes. The second wave does need new protein, namely Fos and other first-wave products, so cycloheximide blocks it. A signal that lasts minutes at the membrane becomes a program that lasts hours.

Prof. Deppmann’s phrase for the practical use is “IEGs to map brain function”. A resting neuron makes very little Fos. In a strongly excited neuron, calcium enters and the route to CREB switches on c-fos, and the Fos protein can be stained in the nucleus an hour or two later. Stain a brain after an animal has done something, and the labeled nuclei show which cells were recently active. Two cautions: a neuron can fire without making Fos, and the stain is a snapshot with timing no finer than hours.

Ion channel to CREB, and memory

Prof. Deppmann teaches that the route from ion channel to CREB is involved in memory formation. A brief change at a synapse can be made by modifying proteins already there. A long-lasting change needs new transcription, and CREB is one of the factors it needs, in the animals where this has been tested. In the sea slug Aplysia, Eric Kandel’s laboratory reported in 1990 that putting decoy CRE DNA into a sensory neuron’s nucleus blocked the long-lasting strengthening of its synapse and spared the short-lasting form. In 1994, fruit flies made to express a blocking form of CREB failed to form long-term memory of an odor, and mice lacking two forms of CREB showed a similar loss, with short-term memory intact in both. Later mouse results varied more: the first mutants still made a third form of CREB, the outcome depended on gene dosage and on how training was spaced, and in CREB-deficient mice the related protein CREM was raised and may partly stand in for it. In Aplysia the messenger is cAMP; in mammalian neurons, calcium entering through NMDA receptors and L-type channels reaches the same serine. None of this shows that CREB is where a memory is stored.