# Asymmetric Division: Numb and the Neuroblast

> In the fly, the mother cell puts Numb on one side and aims its spindle at it; the daughter that inherits Numb cannot hear Notch, so it takes the Notch-off fate of that division. Neural stem cells keep themselves and make differentiating cells by dividing asymmetrically. You lay out the four-cell fly bristle lineage, show with timed Notch loss that Notch acts twice, see which sister hears Notch at each division, follow Numb's crescent into one daughter, and use a double mutant to show that Numb is a veto, not an instruction. Then you build the neuroblast's polarity and spindle machinery step by step, and see why a rigged division can compound into overgrowth.

Course: Developmental Neuroscience (Making neurons). Level: advanced university and graduate. About 27 minutes.
Lesson page (link this one when citing): https://stemhop.com/lessons/asymmetric-division/

Builds on: Proneural Genes and Lateral Inhibition · The Ventricular Zone: Radial Glia as Stem Cells · Tools for Following Cells and Genes

## What you'll learn

- Distinguish symmetric from asymmetric division, explain how asymmetric division lets a stem cell keep itself while making differentiating cells, and tell an asymmetry built inside the cell (a segregated determinant) from one supplied by the surroundings (a niche).
- Use timed loss of Notch to show that Notch acts both in picking the fly sensory organ precursor and in its progeny, and name the daughter that hears Notch at each division.
- Explain how Numb, inherited by one daughter of a fly sensory organ precursor, blocks Notch there, predict the cells formed when Numb is lost or present in both daughters of the first division, and use the Notch; numb double mutant to tell a veto from an instruction.
- Describe how apical Par proteins, Inscuteable, Pins, Gαi and Mud orient the neuroblast division, how aPKC keeps determinants off the apical cortex, how Miranda carries Prospero and Brat, and how Partner of Numb (Pon) localizes Numb, so that all three reach the ganglion mother cell.
- Explain with a simple compounding model why fly neuroblasts that fail to segregate these proteins can overgrow, and state what is less certain about Numb and Numblike in the mammalian cortex.

**A stem cell has to do two jobs at once: stay a stem cell, and make cells that differentiate.**

Fly neural stem cells manage both by dividing asymmetrically: before the cell splits, it hands fate-determining proteins to only one daughter. The clearest case is the protein Numb. A fly embryo with no working copy of the gene *numb* still builds its small sense organs, but the sensory neurons in them are replaced by support cells. The protein Numb, it turned out, goes to only one of two sister cells, and there it blocks [Notch](https://stemhop.com/lessons/notch-signaling/), a receptor you have met before.

## Two ways to make sisters differ

A stem cell, or a progenitor (a dividing cell that will give rise to differentiated cells), can divide in two ways. In a **symmetric** division the daughters are alike: two progenitors, which enlarges the pool, or two cells that go on to differentiate, which spends it. In an **asymmetric** division the daughters take different fates, often because one stays a progenitor and the other moves toward a fate. A tissue that must keep its stem cells and still produce differentiated cells needs asymmetric divisions, or a balanced mix of the two symmetric kinds.

A division can become asymmetric in two ways.

1. **From inside.** Before dividing, the mother cell gathers a protein, called a determinant, on one side of its cortex (the layer just under the membrane) and lines its spindle (the array of microtubules that pulls the chromosomes apart) up so the furrow, where the cell pinches in two, separates that side from the other. One daughter inherits the determinant.
2. **From outside.** The daughters start alike, but only one stays in contact with a niche, a set of neighbouring cells that sends signals to remain a stem cell. The daughter pushed out of the niche differentiates.

*Figure (AI-generated illustration).* A division can become asymmetric from inside, when the mother cell gathers a determinant on one side and one daughter inherits it, or from outside, when only one daughter stays in contact with the niche. Image: two panels compare asymmetric division. In the first a mother cell with a gold crescent on one side divides so only one daughter inherits it; in the second two alike daughters form next to niche cells, one stays in contact and keeps stem identity while the other is pushed out and differentiates.

This lesson is about the first kind, the asymmetry built inside the mother cell.

## The layout

Most arguments in this lesson happen on one map, a fly bristle, which the first film draws along with the second map, the neuroblast, met later.

**Map 1: a fly bristle.** A bristle is a small sense organ of four cells: a neuron, a sheath cell that wraps it, and two outer support cells that build the hair and its socket. In the classic scheme these experiments used, all four come from one sensory organ precursor (SOP) in three divisions. (A later revision of the adult lineage, under Going deeper, adds a glial cell.)

- The SOP divides into an **outer precursor** (pIIa) and an **inner precursor** (pIIb).
- The outer precursor divides into a **hair cell**, which builds the shaft (the hair itself), and a **socket cell**.
- The inner precursor divides into a **neuron** and a **sheath cell**.

At each of these divisions the two sisters sit side by side, and one of them hears Notch (its Notch signalling is on) while the other does not. Call the fate a daughter takes when it hears Notch its **Notch-on fate**, and the other its **Notch-off fate**. Which daughter is which is the first thing to find out.

The bristle, and the neuroblast after it, raise four puzzles.

1. **How can two sisters born side by side differ?**
2. **What does the inherited protein do** in the daughter that gets it?
3. **How does the mother put it on one side?**
4. **What happens when the sorting fails?**

> **The one idea to carry through this lesson.** The mother cell puts Numb on one side and aims its spindle at it; the daughter that inherits Numb cannot hear Notch, so it takes the Notch-off fate of that division.

## Notch splits the lineage at every division

**Film transcript: Four Cells and One Switch**

### Film transcript: Four Cells and One Switch

At each division of the bristle lineage, one sister hears Notch. But where in the lineage does Notch act? Notch is a receptor that signals only between touching cells. A ligand called Delta, on the neighbour, activates it. When it is activated, it is cut. Its inner part can then switch genes on. In 1990, Volker Hartenstein and James Posakony used a temperature-sensitive allele of Notch. The mutant protein works at a low temperature and fails at a higher one. So warming the animal removes Notch for a chosen window. They warmed fly pupae in short pulses. First, pulses in the first 14 hours after the pupa formed. That is before the bristle precursors divide.

These gave extra precursors in place of skin cells. Normally the new precursor signals to its neighbours through Notch, and they stay skin cells. Without Notch, they become precursors too. Each extra precursor formed a normal bristle. Now warm the pupae from 14 to 20 hours. That is around the time the precursors divide. Notch now fails at the divisions of the lineage. What cells should the sense organs contain?

The sense organs were made only of neurons. There was no shaft and no socket. Walk it down the map. With no Notch, the first division gives two inner precursors. Each inner precursor then gives two neurons, not a neuron and a sheath cell. So the neuron is a Notch-off fate. The outer branch, which makes the shaft and socket, needs Notch. Notch is used once to pick the precursor. It is used again to split its progeny.

In 1996, Ming Guo, Lily Jan and Yuh Nung Jan went further. Notch works as a two-way switch at each division. That fills in the map. At the first division, the inner precursor is the Notch-off daughter. The outer precursor hears Notch. At the outer precursor's division, the hair cell is Notch-off. The socket cell hears Notch. At the inner precursor's division, the neuron is Notch-off. The sheath cell hears Notch. So at every division, one daughter hears Notch. The sisters sit side by side, and either could signal. What stops one from hearing? Notch is a switch used at every division. One daughter hears it, and one does not.

From the film

- In the fly bristle lineage, Notch acts twice: once to pick the precursor and again to split its progeny.
- At every division one daughter hears Notch and the other, the Notch-off daughter, does not.
- The neuron is a Notch-off fate; the outer branch that makes the shaft and socket needs Notch.

Notch is a receptor that signals only between touching cells: a ligand such as Delta, on the surface of one cell, activates Notch on its neighbour. When Notch is activated it is cut, so that its inner part can switch genes on. Notch also picks the precursor in the first place. In a patch of skin cells that could each become a sensory organ precursor (SOP), the cell that becomes one signals to its neighbours through Notch, and they stay skin cells.

To see where in the bristle lineage Notch acts, in 1990 Volker Hartenstein and James Posakony switched it off at chosen times with a **[temperature-sensitive](https://stemhop.com/methods/temperature-sensitive-mutants/) allele** of *Notch*: a mutant protein that works at a low temperature and fails at a higher one, so warming the animal removes Notch for a chosen window. They warmed fly pupae (the stage between larva and adult) for short pulses at different hours of pupal life. Pulses between 0 and 14 hours after the pupa formed, before the bristle precursors divided, gave extra precursors in place of skin cells, each forming a normal bristle.

**Predict.** Now warm the pupae from 14 to 20 hours, around the time the precursors divide, so that Notch fails at the divisions of the lineage. Using map 1, what cells should the sense organs contain, and what does that say about which fates need Notch?

> **Check your prediction.** The sense organs were made only of neurons, with no shaft or socket. Walk it down map 1. With no Notch, the SOP gives two inner precursors instead of one of each. Each inner precursor then gives two neurons instead of a neuron and a sheath cell. So the neuron is a Notch-off fate, and the outer branch, which makes the shaft and socket, needs Notch. Notch is used once to pick the precursor, and again to split its progeny.

In 1996 Ming Guo, Lily Jan and Yuh Nung Jan showed that Notch works as a two-way switch at each division of the sense organ lineage. That fills in map 1.

| Division | Notch-off daughter | Notch-on daughter |
| --- | --- | --- |
| SOP | inner precursor | outer precursor |
| Outer precursor | hair cell | socket cell |
| Inner precursor | neuron | sheath cell |

The two sisters sit side by side, and either could signal to the other. What stops one of them from hearing?

## A crescent in one daughter

**Film transcript: A Crescent in One Daughter**

### Film transcript: A Crescent in One Daughter

Two sisters sit side by side, and only one hears Notch. What makes the difference? The answer began with a mutant fly. In 1989, Tadashi Uemura and colleagues studied embryos lacking the gene numb. They worked in the laboratory of Lily and Yuh Nung Jan. In these embryos, the sensory neurons were transformed into the support cells of the same lineage. They isolated the gene. That showed numb chooses between sister fates. But it did not show how. In 1994, Michelle Rhyu, Lily Jan and Yuh Nung Jan stained dividing precursors with an antibody to Numb. In this method, an antibody marks the protein, and a DNA dye shows the stage of mitosis. Numb was at the membrane.

It lay in a crescent on one half of the cell. When the cell divided, Numb went into one daughter only. In the scheme they used, that daughter is the inner precursor. Its sister makes the hair and socket. Now two tests. First, remove numb from the precursor and its descendants. What cells form?

Four outer support cells, and no neuron or sheath. Second, make extra Numb only during the precursor's division, so both daughters inherit it. What forms now? The change goes the opposite way. Both daughters take the inner branch. Each then divides as an inner precursor does. You get neurons and sheath cells, but no hair or socket. So there is no bristle.

Opposite changes from loss and gain are strong evidence that a protein decides between two fates. Numb decides between the two branches. Now line this up with the Notch map. Numb goes into the inner precursor. That is the Notch-off daughter. So does Numb tell the cell what to become? Or does it only stop the cell from hearing Notch? Numb forms a crescent and goes into one daughter: the one that does not hear Notch.

From the film

- Numb forms a crescent on one side of the dividing precursor and goes into one daughter only, the Notch-off daughter.
- Losing numb turns the neuron and sheath branch into outer support cells, and giving both daughters Numb does the opposite.
- Opposite changes from loss and gain are strong evidence that Numb decides between the two fates.

The answer began with the mutant from the opening. In 1989 Tadashi Uemura and colleagues in the laboratory of Lily and Yuh Nung Jan reported that in *numb* mutant embryos the sensory neurons become support cells of the same lineage, and they isolated the gene. That showed *numb* chooses between sister fates, but not how.

In 1994 Michelle Rhyu, Lily Jan and Yuh Nung Jan stained dividing sensory organ precursors (SOPs) with an antibody to Numb. The protein was associated with the membrane and lay in a crescent on one half of the cell, and when the cell divided it went into one daughter only. In the scheme they used, that daughter is the inner precursor.

**Predict.** Two tests. First, remove *numb* from the SOP and its descendants. Second, make extra Numb during the SOP division only, so that both daughters inherit it and later divisions run normally. What cells form in each case?

> **Check your prediction.** Without Numb, the lineage made four outer support cells and no neuron or sheath. With extra Numb, the change went the opposite way: both daughters became inner precursors, and each divided as an inner precursor does, giving neurons and sheath cells but no hair or socket, and so no bristle.
>
> Opposite changes from loss and gain are strong evidence that a protein decides between two fates. Numb decides between the two branches.

Does Numb tell the cell what to become, or does it only stop the cell from hearing Notch?

## A veto, not an instruction

**Film transcript: A Veto, Not an Instruction**

### Film transcript: A Veto, Not an Instruction

Numb goes into the daughter that does not hear Notch. Is Numb an instruction, or a veto? You now hold two facts. At each division, Notch decides between the daughters. And Numb goes into one daughter only. Put them side by side. At the first division, Numb goes to the inner precursor, which does not hear Notch. At the outer precursor's division, Numb goes to the hair cell. At the inner precursor's, it goes to the neuron. At every division, the Numb daughter is the Notch-off daughter. Two readings fit that. Perhaps Numb is an instruction. It tells the cell what to become. A cell without Numb cannot take that fate. Or perhaps Numb is a veto. It carries no message of its own.

It only blocks Notch. Then a cell with Numb takes whatever fate a cell without Notch would take. Both predict the numb mutant, where every daughter hears Notch. To tell them apart, take Notch away too. Which single mutant does the double mutant resemble? That reading is called epistasis analysis. The test case is MP2, a precursor in the fly's central nervous system. It makes two neurons, dMP2 and vMP2. Here the Notch signal comes from cells outside the lineage, not from the sister. Notch mutants make two dMP2 neurons. Numb mutants make two vMP2 neurons. What should a fly lacking both make, if Numb is an instruction? And if it is a veto?

If Numb were an instruction, the double mutant could make no dMP2. A veto has nothing to block once Notch is gone. So the double mutant should look like the Notch mutant. In 1996, Eric Spana and Chris Doe found two dMP2 neurons.

Numb is not needed to make dMP2. It is needed only to stop Notch. That earns the rule. The daughter that inherits Numb cannot hear Notch, so it takes the Notch-off fate of that division. The rule names a side, not a fate. The Notch-off fate may be the inner precursor, the hair cell or dMP2. In these lineages, both sisters go on to differentiate. So Notch is not the stem cell signal here. Numb needs its phosphotyrosine-binding domain to block Notch. It does not need that domain to form its crescent. Reaching one side and vetoing Notch are separate jobs. Numb is a veto, not an instruction. Without Notch, there is nothing to veto.

From the film

- Numb is a veto, not an instruction: it carries no message of its own and only blocks Notch.
- In the Notch; numb double mutant the cell makes the Notch-mutant fate, because there is nothing left to veto.
- The rule names a side, not a fate: the Notch-off fate may differ from lineage to lineage.

Put the two facts together on the Notch table: at each division Numb goes into the Notch-off daughter, the inner precursor at the first division, then the hair cell, then the neuron. Two readings fit that match.

- **Numb is an instruction.** It tells the cell what to become, and a cell without Numb cannot take the Numb daughter’s fate.
- **Numb is a veto.** It carries no message of its own. It only blocks Notch, so a cell with Numb takes whatever fate a cell without Notch would take.

Both readings predict the *numb* mutant, in which every daughter hears Notch. To tell them apart, take Notch away as well, and ask which single mutant the **double mutant** resembles. (This reading of a double mutant is called epistasis analysis.)

**Predict.** MP2, a precursor in the fly’s central nervous system (CNS), divides once into two different neurons, dMP2 and vMP2. Here the Notch signal comes from cells outside the lineage, not from the sister. *Notch* mutants make two dMP2 neurons. *numb* mutants make two vMP2 neurons. What should a fly lacking both make if Numb is an instruction? What if it is a veto?

> **Check your prediction.** If Numb were an instruction to become dMP2, a cell without Numb could not become dMP2. The double mutant has no Numb at all, so it should make no dMP2. If Numb is a veto, there is nothing left to veto once Notch is gone. Removing Numb should then change nothing, and the double mutant should look like the *Notch* mutant: two dMP2.
>
> In 1996, in the same issue of *Neuron* as the Guo study, Eric Spana and Chris Doe reported that Delta-Notch signalling was needed for vMP2, that Numb went into dMP2, and that flies lacking both Notch and *numb* made two dMP2 neurons. Numb is not needed to make dMP2. It is needed only to stop Notch.

That earns the rule at the top of this lesson: **Numb is a veto on Notch**, a private bias on how one cell answers a signal from outside it.

**What the rule covers.** Choices between two sisters in the fly: the sensory organ precursor (SOP) lineage, MP2 and the division of the ganglion mother cell (met below). Read it as naming a side, not a fate: the Notch-off side is the inner precursor at one division, the hair cell at the next and dMP2 in the CNS. In these lineages both sisters go on to differentiate, so Notch is not “the stem cell signal” here. Only in some progenitors, such as type II neuroblasts in the fly larva and radial glia in vertebrates, does the Notch-on side mean staying a progenitor. The rule needs a Notch signal on offer, from the sister in the bristle or from outside the lineage in MP2; with no Notch, Numb changes nothing. It does not describe the neuroblast’s other cargo, Prospero and Brat, which limit the small daughter’s potential by other means. Nor does it cover every job Numb does in mammals.

Guo and colleagues showed that Numb negatively regulates Notch, probably by binding Notch through Numb’s phosphotyrosine-binding (PTB) domain (see [protein-interaction modules](https://stemhop.com/lessons/protein-interaction-modules/)). Numb needs that domain to block Notch but not to form its crescent (Frise and colleagues, 1996), so getting to one side and vetoing Notch are separate jobs. How the veto works is still argued; in the sense organ the leading model runs through [endocytosis](https://stemhop.com/lessons/endocytosis-and-signaling-endosomes/).

## How the neuroblast picks a side

**Film transcript: How the Neuroblast Picks a Side**

### Film transcript: How the Neuroblast Picks a Side

The bristle showed what an inherited protein does. But how does a cell put it on one side? The fly neuroblast is where that machinery was worked out. A clue came in 1995. Prospero is a transcription factor, a protein that works in the nucleus. Yet Jürgen Knoblich and the Jans found it at the membrane in dividing cells, in the same crescent as Numb. Its position followed one centrosome, but it did not need microtubules. So it was not carried into place along the spindle. That pointed to a separate system in the cortex that marks one side of the cell. It works in four steps. Step one. Mark the top. Bazooka, the fly's Par3, gathers at the top cortex.

Par6 joins it. So does a kinase called aPKC. Together they define the top side. Step two. Aim the spindle. Inscuteable joins the top complex. Inscuteable recruits a protein called Pins. Pins works with a G protein subunit, Gαi, and a protein called Mud. Together they anchor the motor dynein at the top. Dynein pulls on microtubules from one spindle pole. So the spindle lines up from top to bottom. The furrow will cut between top and bottom. Inscuteable is needed both to aim the spindle and to place Numb and Prospero correctly.

Step three. Keep the cargo off the top. At the top, aPKC phosphorylates Lgl, Numb and Miranda. Phosphorylated proteins let go of the cortex. And a P K C works only at the top. So Numb and Miranda stay bound only at the bottom cortex. L G L is a helper here, not cargo. Step four. Carry the cargo. The adaptor Miranda carries Prospero, and the translational repressor Brat. Numb uses its own adaptor, Partner of Numb. The cell divides. The small daughter, the ganglion mother cell, inherits the cargo. Prospero enters its nucleus and shuts down genes for self-renewal and division. Now remove Miranda. Where do Prospero and Brat go? What happens over many divisions?

They lose their crescent at the bottom cortex. Where they end up depends on the stage and the lineage. Larval miranda mutant tissue overgrows when transplanted. Faulty sorting can let small daughters keep dividing like neuroblasts.

The sensory organ precursor divides within its epithelium. The neuroblast divides across it. But both mark one side and aim the spindle at it. Mark the top, aim the spindle, keep the cargo off the top, and load the small daughter.

From the film

- In the fly neuroblast, the top cortex is marked, the spindle is aimed at it, and determinants are kept off the top.
- aPKC phosphorylates Numb and Miranda at the top so they let go, leaving them bound only at the bottom cortex.
- The small daughter, the ganglion mother cell, inherits the cargo, and Prospero then shuts down genes for self-renewal and division.

The bristle showed what the inherited protein does. How does a cell put it on one side? The fly neuroblast is where that machinery was worked out.

**Map 2: a neuroblast.** The fly’s CNS is built by neuroblasts, stem cells that leave the neural epithelium and divide again and again. In most of them (type I neuroblasts), each division is unequal in size and fate: a large apical daughter stays a neuroblast, and a small basal daughter, the ganglion mother cell (GMC), divides once more to make two neurons or glia. Division after division, the neuroblast stays one neuroblast and makes one GMC each time.

A clue came in 1995. Prospero is a transcription factor (a protein that switches genes on or off in the nucleus), yet Jürgen Knoblich, Lily Jan and Yuh Nung Jan found that in dividing cells it gathers at the membrane and sits in the same crescent as Numb (two other groups reported Prospero’s segregation in neuroblasts the same year). Its localization followed the position of one centrosome but did not need microtubules, so it was not carried into place along the spindle. That pointed to a separate system in the cortex that marks one side of the cell. The machinery works in four steps.

| Step | Proteins | Job |
| --- | --- | --- |
| 1. Mark the top | Bazooka (the fly Par3), Par6 and the kinase aPKC (an enzyme that adds phosphate groups) | define the apical side |
| 2. Aim the spindle | Inscuteable, which joins the apical complex and recruits Pins; Pins with the G protein subunit Gαi and the protein Mud | anchor the motor dynein at the apical cortex, where it pulls on microtubules from one spindle pole |
| 3. Keep the cargo off the top | aPKC phosphorylates Lgl, Numb and Miranda | phosphorylated proteins let go of the cortex; aPKC works only at the top, so Numb and Miranda stay bound only at the basal cortex (Lgl is a helper here, not cargo) |
| 4. Carry the cargo | Miranda carries Prospero and the translational repressor Brat; Numb uses its own adaptor, Partner of Numb (Pon) | load the small daughter from the basal cortex |

Step 2 lines the spindle up along the apical-basal axis, so the furrow cuts between top and bottom and the basal cargo goes to one daughter. Inscuteable is needed both to orient the spindle and to place Numb and Prospero correctly (Kraut and colleagues, 1996). Step 3 is a [phosphorylation switch](https://stemhop.com/lessons/phosphorylation-switch/), found in 2003 when Jörg Betschinger, Karl Mechtler and Jürgen Knoblich showed that aPKC phosphorylates Lgl and releases it from the apical membrane. After the division, Prospero enters the GMC’s nucleus and shuts down genes for self-renewal and division.

*Figure (AI-generated illustration).* In a type I fly neuroblast, apical polarity and spindle-orientation machinery align the division along the apical–basal axis. Basal Numb and Miranda, carrying Prospero and Brat, segregate into the smaller ganglion mother cell (GMC). Prospero then enters the GMC nucleus and represses genes for self-renewal and division. Image: three stages of a type I fly neuroblast division. Before cleavage, apical polarity proteins lie opposite basal Numb and Miranda carrying Prospero and Brat, with the spindle aligned along the apical–basal axis. Division gives a large neuroblast and a smaller basal ganglion mother cell that inherits the cargo. In a final GMC close-up, Prospero enters the nucleus, where self-renewal and division genes are repressed.

**Predict.** A neuroblast lacks Miranda. Where do Prospero and Brat go, and what happens over many divisions?

> **Check your prediction.** Without Miranda, Prospero and Brat lose their crescent at the basal cortex, and where they end up depends on the stage and the lineage. If the small daughter does not get them, it can fail to stop self-renewing and keep dividing like a neuroblast. Larval brain tissue from *miranda* mutants overgrows when transplanted, as the next section shows. Losing Miranda need not change the spindle, because Miranda carries cargo and is not part of the system that aims the spindle.

The sensory organ precursor (SOP) and the neuroblast divide along different axes, but both mark one side and aim the spindle at it. The rule from the bristle carries over too. Numb rides into the GMC, and when the GMC divides, Numb again goes to one sibling and Notch decides between the two, as in MP2.

## When the division is rigged

**Film transcript: When the Division Is Rigged**

### Film transcript: When the Division Is Rigged

What happens when a neuroblast fails to sort its cargo? And why can one small error grow so large? In 2005, Emmanuel Caussinus and Cayetano Gonzalez ran a transplant test. They put larval brain tissue from pins, miranda, numb or prospero mutants into adult flies. The tissue grew to more than a hundred times its starting size. It invaded other tissues and killed its hosts.

In 2006, Jörg Betschinger, Karl Mechtler and Jürgen Knoblich studied Brat and Prospero. In brat or prospero mutants, both daughters grew like neuroblasts and formed brain tumours. Later work showed the two guard different cells. Brat tumours arise mainly in type II lineages, from intermediate progenitors that fail to mature. Prospero tumours arise from ganglion mother cells. But how can a small leak at every division grow more than a hundredfold? Take one neuroblast that divides once per round. With perfect sorting, each small daughter becomes a ganglion mother cell. With leaky sorting, each small daughter has a chance, f, of staying a neuroblast.

Every neuroblast divides in the next round. Run 50 rounds. How many neuroblasts with perfect sorting? How many if one small daughter in ten stays a neuroblast? With perfect sorting, there is still one neuroblast. Its output just adds up, round by round. With the leak, the neuroblasts multiply every round, on average. The count compounds. With one in ten leaking, the average passes a hundred neuroblasts after 50 rounds.

These round counts are illustrative, not measured. The model counts neuroblasts, not tissue size. It is not a fit to the transplant data. It shows only that a small leak at each division, compounded, can multiply the dividing cells more than a hundredfold. In mammals, the picture is less clear. Mice have Numb and a second relative, Numblike. Removing both early in the nervous system used up progenitors. Removing both in the dorsal forebrain, studied across the later peak of neurogenesis, made progenitors overproliferate. The two studies differ in where and when Numb was removed. Which difference matters is not settled. Numb also has other jobs in these cells.

So a mouse phenotype is not automatically a Notch veto. One question points onward. Numb makes two sisters different. It does not say which of many neuron types they become. That depends on when a cell was born, the subject of the next lesson. A rigged division leaks self-renewal into the wrong daughter. Compounded, a small leak can multiply the dividing cells.

From the film

- In fly mutants where sorting fails, neuroblast-like daughters can form tumours.
- A small leak of self-renewal at every division, compounded over rounds, can multiply the dividing cells.
- In mammals the picture is less clear, and a mouse phenotype is not automatically a Notch veto.

If the cargo is not sorted, the small daughter may not stop being a stem cell. In 2005 Emmanuel Caussinus and Cayetano Gonzalez transplanted larval brain tissue from flies mutant for *pins*, *miranda*, *numb* or *prospero* into adult flies. It grew to more than a hundred times its starting size, into invasive, lethal overgrowth.

A few larval neuroblasts (type II) do not make ganglion mother cells (GMCs) directly. They bud off intermediate progenitors, which divide several times, each time making a GMC. In 2006 Betschinger, Mechtler and Knoblich showed that Brat and Prospero together keep the GMC from self-renewing in larval neuroblasts; in *brat* or *prospero* mutants both daughters grew like neuroblasts and formed larval brain tumours. Later work showed that the two genes guard different cells: *brat* tumours arise mainly in type II lineages, from intermediate progenitors that fail to mature, whereas *prospero* tumours arise from GMCs.

How does one sorting error grow a hundredfold? A simple model, not a prediction, shows the logic. Take one neuroblast that divides once per round. With perfect sorting, each small daughter becomes a GMC. With leaky sorting, each small daughter has a chance *f* of staying a neuroblast. Every neuroblast divides every round, and none dies or stops.

**Predict.** Start with one neuroblast and run 50 rounds. How many neuroblasts are there with perfect sorting? How many, on average, if one small daughter in ten (*f* = 0.10) stays a neuroblast?

> **Check your prediction.** With perfect sorting there is still one neuroblast. Its output grows by adding: over 50 rounds it makes 50 GMCs, which divide into 100 neurons or glia if all survive. With a leak, every round multiplies the average number of neuroblasts by (1 + *f*), so the count compounds.
>
> | Leak per round (*f*) | Average neuroblasts after 25 rounds | After 50 rounds |
> | --- | --- | --- |
> | 0 | 1 | 1 |
> | 0.02 | 1.6 | 2.7 |
> | 0.05 | 3.4 | 11.5 |
> | 0.10 | 10.8 | 117 |
> | 0.20 | 95 | 9,100 |
>
> A leak of one in ten gives, on average, more than a hundred neuroblasts after 50 rounds. Single lineages vary around these averages.

The round counts are illustrative, not measured numbers of neuroblast divisions, and the transplants grew in adult hosts, outside any larval schedule. The model counts neuroblasts, not tissue size, and a mutant in which both daughters behave like neuroblasts leaks far more than any row in the table. So the model is not a fit to the Caussinus and Gonzalez data. It shows only that a leak that is small at each division, compounded, can multiply the dividing cells more than a hundredfold.

## In mammals

Vertebrates are less clear. In 1996 Weimin Zhong and colleagues found a mouse Numb at the apical membrane of dividing cortical progenitors; depending on the cleavage plane, it could go to one daughter or both. Mice have a second relative, Numblike. Removing both gave opposite results in two studies that differ in where and when Numb was removed.

| Study | Where | When | Result |
| --- | --- | --- | --- |
| Petersen and colleagues, 2002 | nervous system | early in neurogenesis | progenitors used up |
| Li and colleagues, 2003 | dorsal forebrain | from embryonic day 9.5 (E9.5), across the later peak of neurogenesis | progenitors overproliferate |

In mammals the plane of division predicts fate less reliably than in the fly, and Numb also works in the junctions between progenitors and in endocytosis, so a mouse phenotype is not automatically a Notch veto. Whether these phenotypes reflect Notch antagonism at all, or other Numb functions that differ by stage, is open, and which difference between the two studies explains their results is not settled.

## Putting it together

The four puzzles have one answer, the rule this lesson started with. The heat pulses showed that Notch acts twice, first picking the precursor and then in its progeny, and Guo’s experiments showed it splitting every division; the crescent showed Numb going into the Notch-off daughter; and the double mutant showed that Numb is a veto, not an instruction. The neuroblast showed how a cell marks a side, aims the spindle and loads the small daughter. And the transplants showed what happens when that sorting fails: self-renewal leaks into the daughter that should differentiate, and the leak compounds.

One question points onward. Numb versus Notch is a yes-or-no choice: it makes two sisters different, but it does not say which of many neuron types a GMC’s progeny become. That depends on when the ganglion mother cell (GMC) was born, the subject of [birth order and fate](https://stemhop.com/lessons/birth-order-and-layers/).

### Design problem

In the fly bristle lineage, predict the cells and the bristle you would see if (a) there is no Notch signalling in the sensory organ precursor (SOP) and its descendants, so every daughter is on the Notch-off side, (b) there is no Numb, so every daughter is on the Notch-on side, or (c) Numb is inherited by both daughters of the first division.

> **One way to do it.** (a) Every daughter takes the Notch-off choice, so each division gives two inner-branch cells: neurons, no hair or socket, and no bristle, as in the heat-pulse experiment.
>
> (b) Nothing blocks Notch in either daughter, so the lineage makes four outer support cells (sockets, since Numb also picks the hair cell at the next division) and no neuron or sheath, the result Rhyu and colleagues reported.
>
> (c) Neither daughter of the first division can hear Notch, so both become inner precursors and give neurons and sheath cells but no hair or socket, and so no bristle: the extra-Numb result.

## Going deeper

### The niche

The niche answers the question of how to maintain stem cells from the other side. A stem cell that loses its niche tends to differentiate, and if the niche itself is lost, the stem cells it held are not replaced. In the vertebrate neural tube, the cerebrospinal fluid on one side and the basement membrane on the other are thought to provide a niche for the progenitors of [the ventricular zone](https://stemhop.com/lessons/radial-glia-progenitors/).

The fly ovary shows a niche clearly. Each ovariole, one of the tubes in which eggs form, holds two or three germ line stem cells, which sit against a small group of somatic cells. In 2000 Ting Xie and Allan Spradling showed that these somatic cells act as a niche, with the cap cells a key part: stem cells lost to differentiation were efficiently replaced, so the niche could be repopulated.

### Two axes of polarity

The sensory organ precursor (SOP) divides within the plane of the epithelium, along the head-to-tail axis, with Pins and Numb on the anterior side and the Par complex on the posterior side. The neuroblast, which has left the epithelium, divides along the apical-basal axis, at right angles to the epithelial plane, with Pins and the Par complex together at the apical side and Numb opposite. So the same molecules work along different axes, and Pins sits opposite the Par proteins in one cell and with them in the other.

### How the veto works

In the endocytosis model for the fly sense organ, Numb works with alpha-adaptin and with Sanpodo, a membrane protein required for Notch signalling in these sibling pairs. Once Notch and Sanpodo have been taken in from the surface of the cell that inherits Numb, they are kept from recycling back to it, which lowers Notch at that cell’s surface. How much of the effect is direct binding to Notch remains debated.

### Notch and self-renewal in neuroblasts

In larval neuroblasts the daughter that keeps Notch active keeps progenitor character most clearly in type II lineages. Loss of Notch in type I neuroblasts has much milder effects, and embryonic neuroblasts do not depend on Notch to keep dividing the way type II neuroblasts do.

### Revising the bristle lineage

Later [live imaging](https://stemhop.com/methods/live-cell-imaging/) of the adult bristle revised the lineage: the inner precursor (pIIb) first divides into a glial cell and a cell called pIIIb, which divides again to make the neuron and sheath. So the adult lineage has four divisions and five cells. The map in this lesson leaves out the glial branch, and its inner precursor division stands for the division of pIIIb.

### Speed and passaging of the transplant tumours

The transplanted tissue from the 2005 experiment killed its hosts within about two weeks and could be moved to new hosts for years. Two other laboratories reported *brat* overgrowth in 2006, the same year as the Betschinger study.

---

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