# Brains in a Dish: Stem Cells and Organoids

> The cells bring the program and the embryo brings the surroundings, so an organoid keeps the first and loses most of the second. Skin cells are reset to induced pluripotent stem cells, grown into cerebral organoids, measured against fetal cortex, and fused into regional assembloids. Afterwards you can predict which features of early human cortical development an organoid reproduces and which it cannot, and say what an experiment needs before it blames a mutation for an organoid phenotype.

Course: Developmental Neuroscience (Rebuilding, evolving, and going wrong). Level: advanced university and graduate. About 30 minutes.
Lesson page (link this one when citing): https://stemhop.com/lessons/organoids/

Builds on: The Ventricular Zone: Radial Glia as Stem Cells · Tangential Migration: Interneurons and the Long Way Round · When Building Goes Wrong: Microcephaly, Zika and Lissencephaly

## What you'll learn

- Describe how cerebral organoids are made and which features of early human cortical development they reproduce.
- Trace the line of evidence from nuclear transfer in frogs to induced pluripotent stem cells made with four transcription factors.
- Explain how fusing differently patterned organoids lets interneuron migration be studied, and what a patient line added.
- State what organoids cannot show and how single-cell comparisons with fetal cortex measured their fidelity.
- Use the rule that the cells bring the program and the embryo the surroundings to predict where an organoid will be faithful, and list the controls needed before a mutation is blamed for an organoid phenotype.

**In 2013 a laboratory in Vienna grew brain-like tissue in a dish from stem cells made from the skin cells of a person with severe microcephaly, a brain far smaller than normal. The tissue held zones of dividing cortical progenitors, and those zones were smaller than in tissue grown from control cells.**

Human stem cells, given room and a little guidance, organize themselves into tissue that resembles the early brain. Because a skin cell can be reset to a stem cell with a few transcription factors, that tissue can carry a patient’s own genome, which makes disease questions testable. The harder question is what an [organoid](https://stemhop.com/methods/ips-cells-and-organoids/) gets right and what it gets wrong. This lesson gives you one rule for predicting the answer, tests it, and ends by asking what it takes to blame the mutation in that opening result.

## What an organoid is

An **[organoid](https://stemhop.com/methods/ips-cells-and-organoids/)** is a small piece of tissue, grown in a dish from stem cells, that arranges itself at least partly like an organ. The brain organoids in this lesson start from **pluripotent stem cells**, cells that can give rise to every cell type of the body (some other organoids, such as gut ones, start from adult tissue stem cells).

The key idea is **self-organization**: no one places the cells. In 2008 Mototsugu Eiraku, Yoshiki Sasai and colleagues found that floating aggregates of mouse **embryonic stem cells** (pluripotent cells from an early embryo), and in the same paper human ones, formed polarized cortical neuroepithelium: a sheet of cells with a definite inside and outside. Its **apical** side faces a fluid-filled cavity, dividing progenitors sit near that side, and young neurons lie further out.

Madeline Lancaster, working in Jürgen Knoblich’s laboratory in Vienna, built a different human method on the same idea, one that adds little patterning and lets several brain regions form. Their 2013 recipe has four steps.

1. **Aggregate.** Human pluripotent stem cells are gathered into balls called embryoid bodies.
2. **Steer.** The balls are steered toward neural tissue.
3. **Embed.** Each ball is embedded in a droplet of Matrigel, a gel from a mouse tumor that resembles basement membrane, rich in laminin and collagen IV.
4. **Spin.** The droplets move to a spinning bioreactor, where moving medium improves the supply of oxygen and nutrients.

The tissue grew to a few millimetres across, up to about 4 mm, and they called it a **cerebral organoid**.

*Figure (AI-generated illustration).* The 2013 cerebral organoid recipe: gather human pluripotent stem cells into balls, steer them toward neural tissue, embed each in a droplet of Matrigel, a gel that resembles basement membrane, and move the droplets to a spinning bioreactor, where moving medium improves the supply of oxygen and nutrients. The cells organize the rest themselves. Image: four steps in a row. Loose stem cells gather into a ball in a dish; the ball's outer cells turn neural in medium; the ball is embedded in a dome-shaped gel droplet; and the droplets float in a stirred flask where moving medium circulates around them.

**Predict.** Nothing in the dish tells these cells where they are in a body. Which of these would you expect to appear on their own: polarized progenitor zones around a cavity; outer-radial-glia-like cells, a progenitor type abundant in human cortex; the six layers of mature cortex; blood vessels?

> **Check your prediction.** The first two appear. Progenitor zones form around cavities, and some dividing cells take the shape of outer radial glia. The last two do not. Cortical layering is at best partial, and standard organoids have no blood vessels.
>
> The two that appear are things the cells do among themselves. Blood vessels need something from outside the tissue. Layers are less clear-cut: how much of the shortfall comes from missing surroundings, and how much from the culture itself, is not settled.

That split is the rule for the whole lesson.

> **The one idea to carry through this lesson.** The cells bring the program; the embryo brings the surroundings. An organoid keeps the program and loses most of the surroundings. Trust it most for what the cells do on their own, and test it hardest wherever development needs something from outside the tissue.

**What the rule covers.**

- **Which features.** Early and mid-fetal cortical development: progenitor zones, broad cell classes, a sequence of neuron generation that in some protocols resembles the fetal one, and disease mechanisms inside the cells themselves. Each still has to be benchmarked.
- **What “surroundings” means.** Anything the cells in the dish do not make for themselves: blood vessels and the oxygen they carry; microglia, the brain’s immune cells; the meninges; sensory and thalamic input; distant targets; other brain regions, such as the ventral forebrain; and the embryo’s mechanical and positional frame.
- **A first guess, not a verdict.** The dish adds stress of its own, such as low oxygen and unusual metabolism, so even program features have to be measured. The rule tells you where to look first, not what you will find.

## Where the cells come from

To grow a patient’s cortex in a dish you need cells that can still become cortex, and a skin cell cannot. The method rests on sending a skin cell back to an embryonic state, so the first question is whether that is possible at all.

Embryonic stem cells divide indefinitely in culture; mouse lines were made in 1981, human lines in 1998.

**Is a specialized cell’s genome still complete?** As cells specialize, they might discard genes they no longer use, or they might keep every gene and silence most. Only the second would allow a reset. **[Nuclear transfer](https://stemhop.com/methods/microinjection/)** tells the two apart: replace an egg’s nucleus with another cell’s, and see whether the egg develops. In 1962 John Gurdon, at Oxford, put nuclei from the intestinal lining of feeding *Xenopus laevis* tadpoles into eggs whose own nucleus had been removed or destroyed. Most transfers failed, but a small fraction became normal feeding tadpoles, nervous system included. One clean success answers the question: the gut-cell nucleus still held the instructions for a whole animal. The low success rate left one doubt, that the successes came from rare unspecialized cells; later work settled it (see Going deeper). What in the egg does the resetting was still unknown.

**Predict.** Suppose the egg resets a nucleus with proteins. Which kind of protein would you try first, and how would you find the few that matter among dozens of candidates?

> **Check your prediction.** **Transcription factors**, proteins that bind DNA and switch genes on or off, are the obvious first guess, because a cell’s identity is set by which genes are on. Kazutoshi Takahashi and Shinya Yamanaka, at Kyoto University, picked 24 genes active in embryonic stem cells and delivered all of them with retroviruses into mouse fibroblasts (connective-tissue cells). The cells carried a reporter: a stem-cell gene, *Fbx15*, joined to a gene for resistance to the drug G418, so only cells that switched on part of the stem-cell program survived. Colonies grew. Leaving out one factor at a time showed which ones the colonies needed, and four remained: Oct3/4 (also called Oct4), Sox2, Klf4 and c-Myc. In 2006 they called the cells, made from both embryonic and adult mouse fibroblasts, **induced pluripotent stem cells**, or **iPS cells**.

In November 2007 two groups made human [iPS cells](https://stemhop.com/methods/ips-cells-and-organoids/): Takahashi, Yamanaka and colleagues, with the same four factors on adult human skin fibroblasts, and Junying Yu, James Thomson and colleagues, with OCT4, SOX2, NANOG and LIN28. Gurdon and Yamanaka shared the 2012 Nobel Prize in Physiology or Medicine.

**What reprogramming gives an organoid.** Three things.

1. **The donor’s genome.** The mutation comes with the rest of that person’s genetic background, plus any new mutations the line picks up during reprogramming and culture.
2. **An unlimited supply.** A line can be grown and frozen, and the same cells used again and again.
3. **A reset clock.** Reprogramming resets many marks of the donor’s age, such as telomere length and the DNA-methylation clock, though some memory of the source tissue can remain. The cells then go through neural differentiation from a pluripotent start, which is why organoids model early development, not the patient’s adult brain.

## What the cells build on their own

Inside a cerebral [organoid](https://stemhop.com/methods/ips-cells-and-organoids/), cortex-like regions form around fluid-filled cavities. From the cavity outward:

1. **Radial glia**, the progenitors of the cortex, line the cavity in a layer called the ventricular zone. Their nuclei move toward the apical surface and away again as the cells cycle, a shuttle called interkinetic nuclear migration.
2. **Intermediate progenitors**, which divide a few more times to make neurons, sit beyond them.
3. **Early neurons** lie further out.
4. **Outer-radial-glia-like cells** are scattered among them: dividing cells away from the apical surface, with a basal process and no apical one. That is the shape of the outer radial glia that are abundant in human cortex (see [human brain expansion](https://stemhop.com/lessons/human-brain-expansion/)). In the 2013 organoids such cells were present, but a distinct outer zone like the fetal one was not well formed; later protocols, for example that of Qian and colleagues in 2016, made a clearer one.

*Figure (AI-generated illustration).* Inside a cerebral organoid, cortex-like zones form around fluid-filled cavities with no one placing the cells: radial glia line the cavity and shuttle their nuclei as they cycle, intermediate progenitors sit beyond them, early neurons lie further out, and outer-radial-glia-like cells, with a basal process and no apical one, are scattered among them. In the 2013 organoids a distinct outer zone like the fetal one was not well formed. Image: an inset shows a whole organoid with several fluid-filled cavities, each ringed by cells, and a box marks one ring. The enlarged view shows the tissue around that cavity: tall radial glia line the cavity, with nuclei at different heights and one dividing at the cavity surface; intermediate progenitors lie beyond them; early neurons lie further out; and two outer-radial-glia-like cells, each with a single process pointing outward and none reaching the cavity, sit among the neurons.

The pace is human too. Human organoids develop over months rather than weeks. In one long-term study of directed cortical organoids (Gordon and colleagues, 2021), gene expression shifted toward early postnatal patterns only around days 250 to 300. The timing seems to come largely from the cells themselves, though it varies with protocol and cell type.

Every item on that list arises by self-organization of the cells, within conditions the experimenter supplies: medium, a gel and, in directed protocols, patterning factors. By the rule, this is where an organoid should be most faithful. The next section measures it.

## What the dish misses, measured

**Film transcript: Measured Against Fetal Cortex**

### Film transcript: Measured Against Fetal Cortex

An organoid has no blood. Every bit of oxygen must diffuse in from its surface. What happens at the center? The cells use oxygen as it moves inward. So it runs out a short way in. Take a simple model, with assumed rates of oxygen use. In a large organoid, oxygen reaches only about 0.1 to 0.6 millimetres in. A large organoid is up to about 4 millimetres across. In this model, a third or more of it has no oxygen. That is why cells in the center of a large organoid often die. In adult cortex, capillaries sit only tens of micrometres apart. Faster spinning helps only partly. It keeps the surface supplied, but under these assumptions it cannot feed a deep core.

Shortening the path helps. Slice the organoid, or grow it where air meets liquid. Blood is surroundings. The rule flags this failure, and the estimate shows how large it is. Now a finer test, cell by cell. Do the organoid's cells match fetal cortex? Single-cell RNA sequencing counts the messenger RNAs in thousands of separate cells. Cells with similar profiles are grouped into clusters. Then each organoid cluster is compared with each fetal one. In 2020, Aparna Bhaduri and colleagues compared hundreds of thousands of cells this way. The broad classes matched: radial glia, intermediate progenitors and neurons. Those are program features, as the rule predicts.

Finer subtypes blurred. An organoid cluster often resembled several fetal subtypes about equally. Markers of particular subtypes, such as SATB2 in upper-layer neurons, were much rarer. Here is the twist. Organoid cells expressed stress genes at higher levels. Transplanted into mouse cortex, those stress genes fell. Some cells came to resemble fetal subtypes more closely. So the dish itself seems to add problems. The rule says where to look first. It is not a verdict. Every feature still has to be measured.

From the film

- With no blood, oxygen must diffuse in from the surface, so in a simple model of a large organoid the core runs out of oxygen and its cells often die.
- Compared cell by cell with fetal cortex, organoids matched the broad classes of cells, such as radial glia, intermediate progenitors and neurons, but finer subtypes were blurred.
- Organoid cells expressed stress genes at higher levels, and those fell after transplant into mouse cortex, which suggests the culture environment contributes.
- The program-and-surroundings rule says where to look first; it is not a verdict, so every feature still has to be measured.

Two measurements test the rule, one physical and one cell by cell.

### No blood, so oxygen must diffuse

Standard [organoids](https://stemhop.com/methods/ips-cells-and-organoids/) have no blood vessels. Oxygen enters only from the surface, and the cells use it up as it diffuses inward. In adult cortex, by comparison, capillaries sit a few tens of micrometres apart, so no cell is far from blood.

**Predict.** A cerebral organoid can be up to about 4 mm across. Roughly how far in would you expect oxygen to reach, and what happens at the center? Would faster spinning fix it?

> **Check your prediction.** Take a simple model: a ball of tissue with no vessels, using oxygen at a uniform rate, with both the surface oxygen level and the rate of use assumed (the formula and ranges are in Going deeper). In this model:
>
> - A ball keeps oxygen at its center only if it is about 0.4 to 1.9 mm across or smaller.
> - In a 4 mm organoid, oxygen reaches only about 0.1 to 0.6 mm in, so the oxygen-free core is about a third to four fifths of its volume.
>
> That is why cells in the center of a large organoid often die. These are model numbers, not measurements of any one organoid.
>
> Faster spinning helps only partly. Moving medium keeps the surface close to the oxygen level of the medium, but it cannot push the level higher, so under these assumptions it cannot feed a 4 mm core. Shortening the path can: slicing organoids, or growing them at an air–liquid interface.

Blood is surroundings, which is where the rule says to look first; the estimate shows how large the failure is.

### Cell by cell against fetal cortex

Single-cell RNA sequencing lets organoid cells be compared with fetal cortex of known age (see [tools for development](https://stemhop.com/lessons/tools-for-development/)) in three steps.

1. Count each cell’s messenger RNAs, gene by gene.
2. Group cells with similar profiles into clusters, separately for the organoid and for fetal cortex.
3. Correlate each organoid cluster’s profile with each fetal cluster’s. A cluster with one clear best match has a clear identity. A cluster that matches several about equally has a blurred identity at that level of detail.

In 2015 J. Gray Camp, Barbara Treutlein and colleagues found that organoid cortex used gene programs similar to those of fetal neocortex. In 2019 Silvia Velasco, Paola Arlotta and colleagues showed that a directed cortical protocol gave similar cell compositions from organoid to organoid.

In 2020 Aparna Bhaduri, Madeline Andrews, Arnold Kriegstein and colleagues graded more finely. They compared about 235,000 cells from 37 organoids with about 189,000 cells from developing human cortex.

- **Broad classes matched.** Radial glia, intermediate progenitors, neurons and interneurons were all there. Radial glia, intermediate progenitors and neurons are program features, and the rule predicts them. Interneurons are a different case: they need ventral forebrain tissue, which a directed dorsal organoid lacks and which less directed protocols can sometimes form on their own.
- **Subtypes blurred.** Organoid clusters often correlated about equally with several fetal subtypes, and markers of particular subtypes, such as SATB2 in upper-layer neurons, were much rarer. Fine subtypes depend partly on the program and partly on signals from outside.
- **The dish added stress.** Organoid cells expressed genes of glycolysis and endoplasmic reticulum stress, signs of cells under strain, at higher levels. When organoid cells were transplanted into mouse cortex, those stress genes fell and some cells came to resemble primary subtypes more closely, which suggests that the culture environment contributes.

That last result is the “first guess, not a verdict” caveat at work. Later protocols narrowed some gaps; the grade depends on protocol, age and cell type.

### The other missing pieces

Most of the remaining limits are missing surroundings; the last is partly a matter of how long cultures can run.

- **Microglia.** Most protocols lack them, because microglia come from the yolk sac, outside the neural tube. Unguided organoids can make some mesoderm, and microglia-like cells sometimes arise in them (Ormel and colleagues, 2018), which is why the surroundings are defined as what the cells in the dish do not make for themselves, not simply what the embryo supplies from outside.
- **Input and targets.** A single organoid has no sensory inputs and no distant targets, so long-range wiring cannot form normally, and neither can survival that depends on distant targets, in which neurons are made in excess and compete for limited survival signals from those targets (see [the neurotrophic hypothesis](https://stemhop.com/lessons/neurotrophic-hypothesis/)). Local partners can still affect survival.
- **Layers.** At best partial, as above.
- **Age.** Most organoids model early and mid-fetal stages, not a mature brain.

## Putting back the surroundings

**Film transcript: Putting Back the Surroundings, and Blaming a Mutation**

### Film transcript: Putting Back the Surroundings, and Blaming a Mutation

Most cortical interneurons are born in another region, the ventral forebrain. So how could a cortical organoid get them? Supply the region. In 2017, Sergiu Pașca's group at Stanford grew the two regions separately. In 2017, Sergio Pashka's group at Stanford grew the two regions separately. They labelled the ventral cells, then let one spheroid of each kind fuse. Predict. Which way will the labelled cells move? And how?

They moved from the ventral partner into the cortical one, far more often than the reverse. They moved in jumps. A leading process extends. A swelling forms ahead of the nucleus. Then the nucleus follows. Fetal interneurons move the same way. Once inside, the cells received excitatory synaptic input.

Then came cells from people with Timothy syndrome, caused by a mutation in a calcium channel gene. Patient interneurons made more jumps, but shorter ones. Overall, they moved less efficiently. Fused to control cortex, patient interneurons still showed the defect. So it lay at least partly in the interneurons themselves. Blocking L-type calcium channels improved their migration. Those two experiments, not the channel's location, point to a defect in the cells' own program. An assembloid can show it. A host restores more. Human organoids were transplanted into newborn rat cortex. As the rats matured, the grafts gained a blood supply and responded to whisker deflection.

In 2013, Lancaster and colleagues grew organoids from a person with microcephaly, a brain far smaller than normal. The patient carried mutations in a gene called CDK5RAP2. The patient's organoids had smaller progenitor zones. Is the mutation to blame? Predict. Name three other explanations, and a control that tests each.

The rest of that person's genes. Several donors help, but only moving the gene itself really tests this. A quirk of one cell line. Several clones from the patient test that. A bad batch. Grow patient and control side by side, more than once. These controls strengthen the comparison. Then move the gene itself. Put back into patient tissue, CDK5RAP2 kept more cells as radial glia. Reduced in control tissue by two different hairpin RNAs, it let cells shift toward neurons.

Why was the organoid a good bet? The protein works at the centrosome, inside the dividing progenitor. The rescue and knockdown, done in organoid tissue, made it a result. If a gene acted through blood vessels or microglia, a normal-looking organoid would not clear it. Use the program to choose where to start. Then test the organoid, hardest where the surroundings matter, and measure every feature, because the dish adds stress of its own.

From the film

- Fusing a ventral spheroid to a cortical one supplies the missing source of interneurons, which then move into the cortical tissue in jumps.
- In Timothy syndrome assembloids the interneuron migration defect persisted when fused to control cortex, which supports an intrinsic contribution from the interneurons themselves.
- Before blaming a mutation for an organoid phenotype, control for genetic background, cell-line quirks and batch, then put the gene back and take it away.
- Transplanting human organoids into newborn rat cortex supplied a blood supply and responses to whisker deflection.

If an [organoid](https://stemhop.com/methods/ips-cells-and-organoids/) fails where development needs something from outside, the remedy to try is to supply that thing, and then measure what is restored. This section covers two ways, fusing in another region and putting the organoid into a host; the design problem below adds missing cell types.

### Fusing regions: assembloids

In an unguided organoid, regions form where they form, which makes traffic between two regions hard to study. Most cortical inhibitory interneurons, for example, are born in the ventral forebrain and migrate tangentially into cortex (see [tangential migration](https://stemhop.com/lessons/tangential-migration/)). A cortical organoid on its own lacks their source.

In 2017 Fikri Birey, Jimena Andersen, Christopher Makinson, Sergiu Pașca and colleagues at Stanford made the two regions separately. They patterned human pluripotent stem cells into cortical spheroids (dorsal forebrain) and subpallial spheroids (ventral forebrain), labelled the ventral ones, and placed one of each together until they fused. (Pașca’s group calls its directed organoids spheroids.)

**Predict.** Which way will the labelled cells move, and how?

> **Check your prediction.** Labelled cells moved from the ventral partner into the cortical one far more often than in the reverse direction. They moved in jumps: a leading process extended, a swelling formed ahead of the nucleus, and the nucleus followed. Interneurons in fetal forebrain move in the same saltatory way. Once inside, the migrated cells received excitatory synaptic input.

*Figure (AI-generated illustration).* A labelled ventral (subpallial) spheroid fused to a dorsal (cortical) one. Labelled cells cross into the cortical tissue in jumps: leading process, swelling, then nucleus. Image: two fused spheroids, a ventral one with gold labelled cells and a cortical one, with a broad arrow of cells crossing from ventral to cortical. An inset shows one cell's jump in three steps: process extends, swelling forms, nucleus follows.

The team then made spheroids from people with Timothy syndrome, caused by a gain-of-function mutation, one that makes the channel pass more calcium, in *CACNA1C*, the gene for the L-type calcium channel Cav1.2. Patient interneurons made more jumps, but shorter ones, and moved less efficiently overall. Patient ventral spheroids fused to control cortical spheroids still showed the defect, so it lay at least partly in the interneurons themselves. Blocking L-type channels improved their migration. Those two experiments, not the channel’s place in the membrane, support an intrinsic contribution to the defect, one an assembloid can reveal.

Pașca’s group later named such fused preparations assembloids. Cells in them are labelled and altered with tools from [hacking the cell](https://stemhop.com/lessons/hacking-the-cell/).

### Transplanting into a host

A host animal supplies other missing pieces, such as the mouse cortex that lowered the stress genes in Bhaduri’s transplant. In 2022 Pașca’s group reported human cortical organoids transplanted into the cortex of newborn rats, where, as the rats matured, they gained a blood supply and came to respond to whisker deflection.

Fusion puts back a second region; [transplantation](https://stemhop.com/methods/tissue-grafting-and-transplantation/) puts back blood and sensory input, among others. Each restores only part of what a single organoid lacks, and combining parts that a single organoid cannot make in the right place is one direction the field is taking.

## Blaming a mutation

Back to the opening result. [Organoids](https://stemhop.com/methods/ips-cells-and-organoids/) from a person with severe microcephaly and mutations in *CDK5RAP2*, a gene for a protein of the centrosome (the structure that organizes a dividing cell’s spindle), had smaller progenitor zones, and more of their cells became neurons early. The disorder itself belongs to [microcephaly and malformations](https://stemhop.com/lessons/microcephaly-and-malformations/). The question here is narrower: what does it take to blame the gene?

**Predict.** One person’s organoids are small. List three explanations other than the mutation, and one experiment that tests each.

> **Check your prediction.** | Other explanation | Control that reduces or tests it |
> | --- | --- |
> | The rest of that person’s genetic background | Several patient and control donors where possible; then correcting the mutation in the patient’s line, or creating it in a control line (below) |
> | A quirk of one cell line, including new mutations picked up in reprogramming or culture | Several independent iPS clones from the patient, checked for genome changes |
> | A bad batch of culture | Patient and control lines grown side by side in the same batches, more than once, with batch counted in the analysis |
>
> Watch the starting conditions too. In the 2013 work, patient cells had to be seeded at a higher number to make embryoid bodies of comparable size, so final size alone was a weak readout.

Spreading the comparison still leaves the gene only correlated with the result. To blame it, move the gene in both directions, as Lancaster and colleagues did.

1. **Put it back.** They put *CDK5RAP2* back into patient tissue by [electroporation](https://stemhop.com/methods/misexpression-gain-of-function/), an electric pulse that lets DNA into cells, and found more radial glia among the cells that received it.
2. **Take it away.** They reduced the gene’s expression in control tissue with two different short hairpin RNAs, which make cells destroy the gene’s messenger RNA, and the shift toward neurons appeared. Using two different ones lowers the chance that an off-target effect, a hairpin silencing some other gene, explains the result.
3. **Edit it.** Today you would also correct the mutation with genome editing, or create it in a control line, so the two lines differ mainly at that gene.

Then test the proposed mechanism. A pulse of labelled nucleotide, followed later by staining, shows whether more progenitors leave the cycle and become neurons early.

Why was an organoid a fair test here? The rule made it a good bet: the CDK5RAP2 protein works at the centrosome, inside the dividing progenitor. The rescue and knockdown, done in organoid tissue, turned the bet into a result; the protein’s location alone would not have.

## Putting it together

| Experiment | Preparation | Observed result | What it supports | What it does not establish |
| --- | --- | --- | --- | --- |
| Floating stem-cell aggregates (2008) | mouse and human embryonic stem cells | polarized cortical neuroepithelium around a cavity | cells self-organize into early cortex-like tissue | a whole cortex |
| Cerebral organoids (2013) | human pluripotent stem cells, gel, spinning culture | several regions; radial glia, intermediate progenitors, outer-radial-glia-like cells, early neurons | the cells carry the program for early cortical zones | layers, blood supply, mature stages |
| Patient against control (2013) | organoids from a person with *CDK5RAP2* mutations | smaller progenitor zones, early neurons | a candidate phenotype | that the gene causes it, from this alone |
| Rescue and two hairpins (2013) | patient tissue given the gene; control tissue with two shRNAs | more radial glia with the gene; shift to neurons without it | loss of the gene drives early neuron birth | a whole-brain explanation of microcephaly |
| Single-cell comparisons (2015, 2019, 2020) | organoid and fetal cortex cells; in 2020, also organoid cells moved into mouse cortex | similar gene programs (2015); similar composition across organoids (2019); broad classes matched, subtypes blurred, stress genes high, falling after transplant (2020) | fidelity depends on the feature; the culture environment contributes | which part of the blur comes from culture and which from missing input |
| Fused spheroids (2017) | dorsal plus labelled ventral spheroid | ventral-to-cortical, saltatory migration; synaptic input | a missing region can be supplied | long-range wiring |
| Timothy syndrome assembloids (2017) | patient ventral spheroids, also fused to control cortex | more but shorter jumps; less efficient; improved by blocking L-type channels | an intrinsic contribution to the migration defect | the whole disorder |
| Transplant into newborn rat cortex (2022) | human cortical organoids in a host | blood supply; responses to whisker deflection | a host can supply blood and input | that the tissue matches a human brain |

At every step the rule pointed to where to look first; the experiments, not the rule, gave the answers.

### Design problem

A family carries a mutation in a gene expressed mainly in microglia, and the children have a cortical malformation. A laboratory grows standard cerebral [organoids](https://stemhop.com/methods/ips-cells-and-organoids/) from the patients and from controls and finds no difference. Does that clear the gene? Design what you would do next, and say what each step can and cannot show.

> **One way to do it.** It does not clear the gene. If the mutation acts through microglia, and most protocols lack them, a normal-looking organoid is what you would expect whether or not the gene matters. The gene is expressed mainly, not only, in microglia, so also check its expression in the neural cells.
>
> 1. **Check the preparation.** Stain for microglia-like cells before reading any result. Unguided organoids sometimes contain a few, but a protocol that makes none cannot test this gene.
> 2. **Supply the missing cells.** Add microglia-like cells made from the same patient and control lines, if your laboratory can make them, and grow patient and control organoids with each.
> 3. **Cross the parts.** Patient microglia with control tissue, and control microglia with patient tissue. If the defect follows the patient microglia whatever the tissue, they contribute to it, the same logic that placed part of the Timothy syndrome defect in the interneurons. Check that the added microglia match in number and state.
> 4. **Keep the earlier controls.** Several clones and several control donors, side by side; then correct the mutation, or create it in a control line.
>
> Even a positive result leaves the rest of the surroundings missing, such as blood, input and the embryo’s frame, so it shows that microglia carrying the mutation can change the tissue, not that this is the whole cause in the children.

## Going deeper

### Nuclear transfer in frogs

In 1952 Robert Briggs and Thomas King first did [nuclear transfer](https://stemhop.com/methods/microinjection/) in the leopard frog *Rana pipiens*, using nuclei from blastula cells (the hollow ball of cells an early embryo forms). Some of the eggs developed into tadpoles. In their later experiments, nuclei taken from older embryos supported normal development less often, which looked like evidence that nuclei lose potential as cells specialize. Gurdon’s 1962 transfers from the gut of feeding *Xenopus* tadpoles, cells that had already specialized, pushed the test further.

That left the doubt about rare unspecialized gut cells. Gurdon’s own 1970s transfers from adult frog skin cells reached the tadpole stage. The cleanest answer came in mice in 2002, when Konrad Hochedlinger and Rudolf Jaenisch used nuclei from mature lymphocytes. Those cells have rearranged their antibody or T-cell receptor genes, which marks a donor as fully specialized. The transferred eggs gave embryonic stem cell lines, and those lines gave mice whose cells carried the rearrangements.

*Figure (AI-generated illustration).* Gurdon put nuclei from the intestinal lining of feeding Xenopus tadpoles into eggs whose own nucleus was removed or destroyed. Most transfers failed, but a small fraction became normal feeding tadpoles. Image: a feeding Xenopus tadpole supplies an intestinal epithelial cell. The cell's nucleus is transferred into an egg whose own nucleus was removed or destroyed. Most transfers fail, but a small fraction develop into normal feeding tadpoles.

### How good the first iPS cells were

The first lines were imperfect: their gene expression differed from embryonic stem cells, and in the tests reported they gave no live chimeric mice (mice built partly from injected cells). In 2007 Keisuke Okita, Tomoko Ichisaka and Yamanaka selected cells with a stricter reporter, *Nanog*, and obtained lines that contributed to adult mice and their germ line. The laboratories of Rudolf Jaenisch and Konrad Hochedlinger reported similar results.

### Clonal progenitors before organoids

Single progenitors from embryonic rat forebrain, grown in tiny wells in the late 1980s, made only neurons, only glia or both (see [radial glia and progenitors](https://stemhop.com/lessons/radial-glia-progenitors/)). Dissociation lets progenitors be studied one cell at a time, but it removes their tissue and changes the signals they receive. [Organoids](https://stemhop.com/methods/ips-cells-and-organoids/) keep more of the three-dimensional context.

### The oxygen estimate

Treat an organoid as a ball with no blood vessels, at steady state: uniform tissue, a fixed oxygen concentration at its surface, and oxygen used at a uniform rate. The center first runs out of oxygen once the radius exceeds R = √(6·D·C₀/Q), where D is how fast oxygen diffuses in tissue (about 2 × 10⁻⁹ m²/s), C₀ is the oxygen concentration at the surface, and Q is the rate of use. C₀ is taken as 100 to 200 µM: the upper end is near air-equilibrated medium at 37 °C, and the lower end allows for less oxygen at the surface. Adult brain uses about 3.5 mL of oxygen per 100 g per minute, about 0.026 mol per cubic metre per second. Because immature tissue relies more on glycolysis, Q was taken anywhere from that value down to 10 percent of it. Both ranges are assumptions, not measurements of organoids.

With those numbers the largest ball that keeps oxygen at its center is about 0.4 to 1.9 mm across. That is a threshold for running out of oxygen, not for cell survival. In a larger ball the formula changes, because the core uses none: for an oxygen-free core of radius b, C₀ = Q(R² − 3b² + 2b³/R)/(6D). For a 4 mm ball this gives an oxygenated shell about 0.13 to 0.62 mm thick. The core fraction follows from geometry: everything deeper than a shell of thickness d is a sphere of radius 2 − d mm, so the thinnest shell leaves (1.87/2)³, about 82 percent, and the thickest leaves (1.38/2)³, about 33 percent.

### Electrical activity and the “mini-brain” label

Organoids are electrically active. In 2019 Cleber Trujillo, Alysson Muotri and colleagues recorded cortical organoids for months on electrode arrays and reported network oscillations that grew more complex with time; they reported that some features of this activity resembled features of electroencephalogram (EEG) recordings from preterm infants. A New York Times article that year carried the headline “Organoids Are Not Brains. How Are They Making Brain Waves?” Coordinated firing in tissue with no inputs is a long way from a working brain, and the word “mini-brain” hides that distance.

### Ethics

The ethical questions remain: consent from cell donors, the welfare of animals that receive organoid transplants, and how complex such tissue should become.

### Who built the first assembloids

Two other groups, Joshua Bagley with Knoblich and Yangfei Xiang with In-Hyun Park, reported fused regional organoids in the same months as Birey and colleagues.

---

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