Three papers from the past two weeks, in Nature Neuroscience, Neuron and Nature. All three use cells reprogrammed from a person, grown into something with structure, and all three are past the stage of showing that the model can be built.
One grows human myelin, damages it, and watches it come back. One puts a missing gene back into a model of autism and the circuit recovers. And one answers a question about epilepsy by using the thing a dish cannot do: it cannot have a seizure.
1. Human myelin that breaks and grows back
Myelin is the insulation around nerve fibres. Lose it and signals slow or stop, which is what happens in multiple sclerosis. The body can rebuild it, and in MS that rebuilding eventually fails.
There is no approved drug that promotes new myelin. The authors attribute that partly to something unglamorous: a lack of reliable systems that reproducibly model myelin damage and repair. You cannot screen for a drug that regrows human myelin without somewhere human to watch myelin regrow.
So they engineered a spheroid, a ball of tissue grown from human iPSCs, enriched with two things. Mature, actually myelinating oligodendrocytes, the cells that make myelin. And functionally reactive microglia, the brain's resident immune cells, which clear debris.
Then they injured it, and the full repair sequence ran:
- Myelin fragmented.
- Microglia became reactive and ate the myelin debris, which is a necessary step before anything can be rebuilt.
- New oligodendrocytes were generated and matured.
- Axons were wrapped in newly formed myelin.
They show the newly generated oligodendrocytes are the ones doing the remyelinating, and they counted it properly, using electron microscopy to measure the sheaths. Myelin that grew back was thinner than the original.
What to watch:no drug was found here. This is the platform, offered for screening, and the test of a screening platform is whether something found in it survives contact with a patient, which takes years. It is also worth being precise about the immune side. What they describe is microglia, the brain's own immune cells. MS involves an attack from the wider immune system, and infiltrating immune cells are not part of what this abstract describes.
2. Put the gene back and the circuit recovers
Profound autism is often traceable to a single gene, and SCN2A is among the most penetrant, meaning a mutation in it reliably produces the condition rather than nudging the odds. It encodes a sodium channel, part of how a neuron fires.
The circuit implicated runs from cortex to striatum, two regions that have to talk to each other. Studying that in a human has the obvious problem. So this group built an assembloid: grow a cortical organoid and a striatal organoid separately, then fuse them and let the cortical neurons send projections into the striatal side. You get a human cortico-striatal connection you can record from.
In models carrying the autism-causing variants:
- Long-range cortical projections were impaired.
- Striatal spine density was reduced, meaning fewer of the small protrusions where incoming connections land.
- Excitatory signalling from cortex to striatum was weakened.
By building assembloids with defined combinations of genotype on each side, they could show the deficit is not only a cortical delivery problem. The striatal tissue is independently vulnerable.
Then the part that makes this more than a description. They delivered human SCN2A using a canine adenovirus vector, and it rescued the cellular and circuit deficits. The authors frame that as support for targeted gene replacement in SCN2A-related autism.
What to watch: there is an unresolved oddity in their own data. Despite the weakened connections, the SCN2A-deficient neurons were more intrinsically excitable, and the networks fired more spontaneously. The authors call that paradoxical and suggest it may be a distinctive feature of human models, which is a careful way of saying nobody has reconciled it yet. And rescuing a circuit in an assembloid is a long way from treating a person; they claim support for the approach, not a therapy.
3. A dish cannot have a seizure, which is the point
Tuberous sclerosis complex is a genetic condition that produces cortical tubers, malformed patches of brain, and epilepsy that often resists every drug available. The tubers are thought to start when a cell picks up a second mutation disrupting TSC1 or TSC2, which switches on a growth pathway called mTORC1.
Tubers are full of abnormal glia, the non-neuronal cells of the brain. Nobody could say whether those abnormal glia help cause the disease or are simply what brain tissue looks like after years of seizures. In resected human tissue the two are inseparable, because by the time you have the tissue, the seizures already happened.
An organoid does not seize. That is usually listed as a limitation. Here it is the experimental design: track what mutant progenitor cells become when no seizure activity exists to blame it on.
Using single-cell transcriptomics and repeated rounds of immunostaining, across both organoids and resected tuber tissue from patients, they found that losing TSC2 biases neural progenitors toward becoming enlarged, pro-inflammatory reactive astrocytes, and that it happens cell-autonomously, meaning the mutated cell does it by itself rather than being pushed by its neighbours. Those astrocytes showed:
- Reduced glutamate transporter expression. Astrocytes normally mop up glutamate, the main excitatory signal, so less mopping is a plausible route to an overexcitable network.
- More inflammatory cytokine secretion.
- Higher expression of neurodegeneration risk genes including APOE and CLU.
The conclusion is that reactive astrocytes are a primary consequence of losing TSC2, not a scar left by seizures, which makes glial dysfunction a driver of the disease and the astrocytes a target worth considering.
What to watch: the glutamate transporter result is suggestive of a seizure mechanism, but this model does not seize, so that link remains inference rather than demonstration. No treatment was tested. The organoid is also developmentally young and missing most of what surrounds a real tuber, which is why the cross-check against patient tissue is the load-bearing part of the paper.
The thread
For about fifteen years the standard iPSC paper demonstrated that you could make a given cell type from a patient. None of these three is that paper. Each one uses the model to do something: screen on it, intervene in it, or exploit what it lacks.
That last move is the one worth borrowing. Everyone can list what a dish does not have, and those absences are usually treated as the caveat paragraph. The tuberous sclerosis paper turns a missing feature into the control condition. You cannot ask whether seizures caused the glial damage in a brain that has been seizing for years. You can ask it in something that has never seized at all.
If you are starting out with these cells, the unglamorous part decides whether any of this works, and we wrote that up separately in a beginner's guide to culturing iPSCs.
Worth reading alongside these is a review that ran in Cell Stem Cell in the same two weeks, on immunocompetent organoids, which the authors call immunoids. Organoids have historically lacked immune cells, and the review surveys efforts to build them in and argues for agreed standards on how they are made. It is a review rather than a research paper, so it reports no new result of its own.