# Notch: Signaling by Contact and by Cutting

> Notch is a receptor that signals almost always between cells that touch, and it does so by being cut: the freed inner part of the receptor travels to the nucleus and switches genes on. This lesson traces the three cuts, weighs what a pathway with no second messenger and no amplification is good for, and shows how two equal neighbours use it to take opposite fates. Afterwards you can predict what happens when the receptor, the ligand or the cutting enzyme is missing or stuck on.

Course: Cell Signaling (Signals that build and defend an animal). Level: advanced university and graduate. About 13 minutes.
Lesson page (link this one when citing): https://stemhop.com/lessons/notch-signaling/

Builds on: Ubiquitin and Regulated Proteolysis as Signals · Cytokine Receptors and JAK–STAT · TGF-β and Smads

## What you'll learn

- Explain why Notch signals mainly between cells that touch, given that its ligands are membrane proteins on the neighbour.
- Trace the three cuts that free the inner part of Notch, and state what each cut needs.
- Compare Notch with a kinase cascade and with the STAT and Smad routes in amplification, reversal and reuse of the receptor.
- Predict the result of lateral inhibition when the receptor or the ligand is missing, and name two diseases of the pathway.

**A fruit fly with one broken copy of a gene called *Notch* has nicks in the edges of its wings. An embryo with no working copy builds too much nervous system and too little skin.**

The nicked wing, first described in 1914, gave the gene its name. The embryo tells more. In the part of the embryo’s outer layer that gives rise to the nervous system, each cell chooses between becoming a nerve precursor and becoming skin. Normally only some choose nerve. Without Notch, far too many do. So Notch is not what makes a nerve cell. It is what lets a cell be told “not you”.

The fly gene was sequenced in 1985 by the laboratory of Spyros Artavanis-Tsakonas, and soon after by that of Michael Young. It encodes a large protein that crosses the membrane once, with 36 repeats on its outer part that resemble epidermal growth factor. In the same issue of *Cell*, Iva Greenwald reported the same kind of repeat in *lin-12*, a worm gene whose mutants make one cell take the fate of its neighbour. A fly and a worm use one family of receptors to settle what two touching cells become. The sequence did not say how. The answer has two parts: contact, and cutting.

## The idea: the signal is a touch and a pull

Most signals you have met leave the cell that makes them. Notch’s ligands do not. In flies they are Delta and Serrate; in mammals, the Delta-like and Jagged proteins. Each crosses the membrane of the sending cell and stays anchored there, while the receptor is anchored in the receiving cell. The two can meet only where two cell surfaces meet, so a cell hears Notch from cells it touches. That contact can be made through long, thin cell extensions, and worms have a few secreted ligands, but the main fly and mammalian ligands work only by contact. In one early test, cultured fly cells that do not normally stick together were made to produce either Notch or Delta, and the Notch cells clumped with the Delta cells.

Binding is not enough. In culture, a piece of ligand free in solution generally fails to switch Notch on, while the same piece fixed to a dish works. In the standard account the receptor has to be pulled. After binding, the sending cell draws its ligand back inside by endocytosis, the pinching-in of membrane to form a vesicle, and the tug passes to the receptor it holds.

Just outside the membrane, Notch folds into a compact lock called the negative regulatory region. Three small modules wrap around the spot where the receptor can be cut and hide it from proteases. The pull peels the lock open. At this point the receptor has registered one fact: a neighbour is holding on and pulling.

## Three cuts: how the receptor becomes the message

1. **While the receptor is made.** In mammalian cells a furin-like protease in the Golgi cuts the new receptor once in its outer part. The two pieces stay clasped together and reach the surface that way. This cut prepares the receptor and carries no signal.
2. **After the pull.** An ADAM protease (ADAM10 in mammals) cuts at the newly exposed site, just outside the membrane. The outer part leaves with the ligand, and a stub stays in the membrane.
3. **Inside the membrane.** γ-Secretase, a complex of four proteins whose cutting subunit is presenilin, cuts the stub within its membrane-spanning segment. It accepts a membrane protein only after most of the outer part has been trimmed away, which is why the second cut must come first.

The third cut happens in a strange place. Cutting a peptide bond uses water, and the middle of a membrane is oily, so presenilin keeps its two catalytic aspartates in a pocket that water can reach.

That cut frees the inner part of the receptor, the Notch intracellular domain (NICD), which enters the nucleus. A DNA-binding protein called CSL binds at the target genes and, in the standard account, keeps them off with co-repressors. Live imaging in flies refines this: without signal, CSL binds DNA only briefly, and it stays far longer once NICD and Mastermind arrive. NICD binds CSL, displaces the co-repressors and recruits a co-activator called Mastermind. The genes turn on. Many encode repressors of the Hes family.

*Figure (AI-generated illustration).* The ligand never leaves the sending cell. A pull opens the lock just outside the membrane, the second cut, made there, releases the outer part with the ligand, and the third cut, inside the membrane, frees the receptor's inner part, which goes to the nucleus and acts on target genes with CSL. Image: four steps between a sending cell above and a receiving cell below. The ligand stays anchored in the sending cell and holds the receptor's beaded outer part; the sending cell pulls, the lock just outside the membrane opens and the receptor is cut there; the outer part leaves with the ligand and the stub is cut inside the membrane; the freed inner part travels to the nucleus and joins CSL on a target gene.

How was it shown that the cut-off inner part is the signal? In steps, each answering an objection to the one before.

In 1993 Gary Struhl, Kevin Fitzgerald and Iva Greenwald made flies and worms that produced only the inner part of Notch or LIN-12, with nothing left to bind a ligand. It changed cell fates as an activated receptor does, and in flies it was in the nucleus. Other fly laboratories reported active truncated Notch that year.

But a fragment made in excess can end up anywhere. In 1996 Raphael Kopan, Eric Schroeter, Harold Weintraub and Jeffrey Nye used a mouse Notch1 that lacked its outer part and was still anchored in the membrane. It was cut, its inner part went to the nucleus, and protease inhibitors stopped the cut.

In 1998 Struhl and A. Adachi tested the intact receptor in a living fly, where too little reaches the nucleus to see by staining. They placed a transcription activator inside Notch, so that a [reporter gene](https://stemhop.com/methods/reporter-gene-assay/) turned on only if that part of the receptor reached the nucleus. It did, and only with ligand. Kopan’s laboratory reported ligand-induced release from mouse Notch1 the same year, and two other fly laboratories (Lecourtois and Schweisguth; Kidd, Lieber and Young) reported ligand-dependent processing of Notch.

In 1999 Struhl and Greenwald showed that flies without Presenilin cannot signal through Notch; other laboratories reported the same need in flies and in mouse cells at the same time.

## What this design gives up, and what it buys

Set Notch beside a kinase cascade. There, each active enzyme modifies many substrates, so the signal grows at every step, and phosphatases can reset it. Notch has no second messenger and no kinase relay. Between the membrane and the gene nothing multiplies: one cut receptor gives one Notch intracellular domain (NICD). A cut cannot be undone, so each receptor signals once.

Set it beside the signal transducer and activator of transcription (STAT) and Smad routes from Unit 4. There too a protein travels from the membrane to the nucleus and acts on genes. But those receptors are kinases, or borrow one, and one active receptor can phosphorylate one STAT or Smad after another. In Notch the traveller is a piece of the receptor itself.

What the cell buys is a count. The amount of NICD made follows, one for one, the number of receptors a neighbour has pulled and cut. The signal itself does not diffuse; it reaches only cells in contact. It is also brief: a kinase in the nucleus marks NICD’s tail, an E3 ligase adds ubiquitin, and the proteasome destroys it (see the lesson on ubiquitin and proteolysis), so many direct target genes go quiet soon after the cutting stops.

The price is sensitivity to dose. With no amplification to cover a shortfall, half the normal amount of a component can show. That is a likely reason a fly with one working copy of *Notch* has nicked wings.

## Two equal neighbours, two fates

**Film transcript: Two Equal Cells, Two Fates**

### Film transcript: Two Equal Cells, Two Fates

In the gonad of a worm larva, two cells lie side by side. They are named Z1.ppp and Z4.aaa. In every worm, one becomes the anchor cell. The other becomes a ventral uterine precursor. But which is which varies. Each cell takes the anchor fate in about half of animals. Kill Z1.ppp with a laser, and Z4.aaa becomes the anchor cell. Kill Z4.aaa instead, and Z1.ppp does. Whichever one you kill, the survivor always becomes the anchor cell. So the other fate needs a neighbour. Greenwald's thesis work, with Paul Sternberg and Robert Horvitz, tied the choice to a gene called lin-12. With no lin-12, both cells become anchor cells. With overactive lin-12, neither cell does.

Now take mosaic worms, in which only one of the two cells lacks lin-12. That cell always becomes the anchor cell. What circuit between two equal cells could produce that?

In the gonad of a worm larva two cells, Z1.ppp and Z4.aaa, lie side by side. In every normal hermaphrodite one becomes the anchor cell and the other a ventral uterine precursor, but which is which varies: each takes the anchor fate in about half of animals. Kill either one with a laser and the survivor always becomes the anchor cell, so the other fate needs a neighbour. Greenwald’s thesis work with Paul Sternberg and Robert Horvitz tied the choice to *lin-12*. With no LIN-12, both cells become anchor cells. With overactive LIN-12, neither does.

### Design problem

Two neighbouring cells start out the same. One must become a nerve cell and the other a skin cell, and nothing outside tells them which. Design a circuit between the two cells that turns a small chance difference into opposite fates. Then predict what the tissue looks like if the receptor in your circuit is missing from every cell.

> **One way to do it.** Give each cell a ligand on its surface and a receptor for that ligand. Add two rules. First, signal received through the receptor blocks the nerve program, so a cell that receives enough becomes skin. Second, the same received signal lowers the cell’s own production of ligand.
>
> Now let chance give cell A slightly more ligand. Cell B receives more signal and makes less ligand. Cell A then receives less signal and makes more. Each turn of the loop widens the gap, until A sends without receiving (nerve) and B receives without sending (skin). Two inhibitions in a row make positive feedback, the switch-making loop from the lesson on signaling circuits, here stretched across two cells.
>
> Without the receptor, no cell can be told “not you”. Every cell runs the nerve program, and the tissue has too many nerve cells and too little skin. That is the fly embryo at the top of this lesson. It is also the worm with no LIN-12, where both cells take the same fate.

Real cells are wired much like this, and the arrangement is called lateral inhibition. In the worm, both cells at first make LIN-12 and its ligand, LAG-2. Before the fates are fixed, the future uterine precursor keeps the receptor and loses the ligand, and the future anchor cell does the reverse. In mosaic worms in which only one of the two cells lacks LIN-12, that cell always becomes the anchor cell: the cell that cannot hear ends up as the sender. In the textbook model of the fly, received signal turns on Hes-type repressors, which shut off the genes that drive both the nerve fate and Delta production. Because the signal neither diffuses nor multiplies, each cell compares itself only with the cells it contacts, and the pattern comes out fine-grained: one nerve precursor among cells that become skin. On the fly’s back, cells also reach cells a few diameters away with thin, moving extensions, and live imaging shows that these contacts help set the spacing of bristles.

*Figure (AI-generated illustration).* Lateral inhibition stretches a positive feedback loop across two cells: received signal blocks the receiving cell's nerve programme and its own ligand production, so a small chance lead widens until one cell sends without receiving and its neighbour receives without sending. Because the signal neither diffuses nor multiplies, each cell compares itself only with the cells it touches. Image: three panels of the same two touching cells, A and B. At the start they carry nearly equal numbers of ligands and receptors, with a small chance lead for A. Then A has more ligands and fewer receptors while B has more receptors and fewer ligands, with arrows crossing from A's ligands to B's receptors. Finally A only sends and B only receives, and the two cells are tinted differently.

## When Notch is stuck on, or missing

**Stuck on.** In 1991 the human gene *NOTCH1* was found broken by a chromosome rearrangement in rare cases of T-cell acute lymphoblastic leukaemia, leaving a shortened receptor. In 2004 mutations in *NOTCH1* were reported in more than half of cases. They cluster in two places. Some loosen the lock near the membrane, so the second cut happens without a pull. Others remove the part of the tail that marks Notch intracellular domain (NICD) for destruction, so each NICD lasts longer.

*Figure (AI-generated illustration).* Mutations in NOTCH1 in T-cell acute lymphoblastic leukaemia cluster in two places. Some loosen the lock near the membrane, so the second cut happens without a pull; others remove the part of the tail that marks NICD for destruction, so each NICD lasts longer. Image: three schematic comparisons of human NOTCH1. Normally, a ligand pulls toward the sending cell, opening the extracellular lock; successive cuts release NICD with its destruction-marking tail, and NICD is later degraded. A loosened lock permits cleavage without a ligand pull, while released NICD retains its tail. Removing the tail still permits ligand-triggered release but makes the released NICD persist longer.

**Missing.** Alagille syndrome is a dominant genetic disease. In most cases one copy of *JAG1*, the gene for the ligand Jagged1, is broken. Most affected children have too few bile ducts in the liver, and many have a heart defect; severity varies widely. As in the fly’s wing, one good copy is not enough.

**Shared scissors.** γ-Secretase also cuts the amyloid precursor protein, and one product is the peptide that builds up in the brain in Alzheimer’s disease. A worm made the connection in 1995, when Diane Levitan and Greenwald found that *sel-12*, a gene that helps *lin-12* signal, resembles a human gene whose mutations cause an inherited, early-onset form of the disease. Drugs that block γ-secretase were developed to lower the peptide, and they block the Notch cut as well. In rodents they turned dividing cells of the gut lining into mucus-secreting goblet cells, the change expected when Notch is lost there. A large trial of one such drug, semagacestat, was stopped in 2010 after patients taking it did worse than those on placebo.

---

From STEMhop (https://stemhop.com), free lessons by Christopher Deppmann. Lesson text © the authors. Practice questions: https://stemhop.com/lessons/notch-signaling/
