One in four fetal neurons carries scrambled DNA, and most are gone after birth

Before a baby is born, roughly one in four neurons in the developing cerebral cortex carries a badly scrambled genome, with large stretches of several chromosomes duplicated or missing. By infancy, most of those cells are gone.

That is the central finding of a team at Boston Children’s Hospital and Harvard Medical School that read the DNA of individual neurons, one cell at a time, from fetuses, children and adults. It points to a genetic quality check that the brain runs on its own cells around birth, and it raises the possibility that when this check fails, it contributes to autism, epilepsy and other neurodevelopmental conditions.

The numbers surprised the researchers themselves. In brains sampled before birth, an average of 25.9% of cortical neurons, and as many as 46% in one fetus, had what the researchers call widespread copy number variants: gains or losses of DNA segments spread across more than one chromosome. In brains sampled after birth, the share fell to about 6.2%, a drop of more than fourfold.

“We were totally unprepared for how common these were in early development,” said Christopher Walsh, a neurologist and geneticist at Boston Children’s Hospital and co-senior author of the paper. The result was so surprising that the team spent years testing whether it was real before publishing it.

A copy number variant is exactly what the name suggests. Most genes come in two copies, one from each parent. A cell with a variant has one copy or three copies of a stretch of DNA instead of two, which changes how much protein those genes produce. Neurons are especially sensitive to that kind of dosage change, and inherited copy number variants are a known cause of autism, schizophrenia, epilepsy and intellectual disability.

How they read the genome of a single neuron

The study used postmortem brain tissue from tissue banks, mostly the NIH NeuroBioBank. Prenatal samples came from fetuses between 14 and 25 weeks of gestation. Postnatal samples ranged from infants to people aged 90 or older. To be sure they were looking only at mature neurons, the researchers cut out the cortical plate, the layer where neurons settle after they stop dividing, and confirmed that 98% of the cells they isolated were neurons.

Sequencing a whole genome from a single cell is expensive, so the team designed a cheaper approach that reads each neuron’s DNA at very low depth, roughly the equivalent of a low-resolution photograph of the genome. That is enough to spot large duplications and deletions, though not small mutations. They used it on 1,526 neurons from 16 people and kept only the 1,248 genomes that passed strict quality filters designed to stop dying or degraded cells from passing as mutated ones.

They then repeated the work with a second, independent method that reads DNA about ten times more deeply, on 831 more neurons from 12 people. It gave the same picture: 33.1% of neurons with widespread variants before birth and 7.2% after, a 4.6-fold drop.

A third line of evidence came from the microscope. In fetal brain slices, about 6% of nuclei in the cortical plate sat next to micronuclei, small pouches of misplaced chromosome fragments that are a classic sign of cell-division errors. The genetic material in those micronuclei faded with age, on the same timeline as the scrambled neurons.

Why the flawed neurons disappear

The developing brain overproduces neurons and then kills a large share of them through programmed cell death. Estimates for the human cortex range from 20% to 50% of neurons lost during development, with the biggest wave, up to about 30%, around birth. Biologists have long known this happens, but not why particular cells are chosen.

The new data offer an answer. The team analyzed the activity of genes in 44,861 individual brain cells and found that neurons carrying many copy number variants had abnormally low activity in genes that build and maintain synapses, the connections between neurons. A subset of these cells also showed high activity in cell-death programs. Young neurons compete to wire into circuits, and those that fail to form connections usually die.

“This competition for synapses has been described for years, but without the benefit of a genomic understanding,” Walsh said. “We didn’t realize that this struggle for survival, this Darwinian competition, was nature’s quality control mechanism in the brain.”

The surviving neurons carry a telling signature. Regions of the genome that hold autism-related genes and genes for the receiving side of synapses were rarely disrupted in neurons from older brains, which suggests that cells with damage there were the ones removed. “It suggests that the neurons that survive may be the ones that have kept the most intact genomes during development,” said Christos Proukakis, a neurologist at University College London who was not involved in the study, speaking to Nature’s news team.

How this compares with what was known

Genetic chaos early in development is not new. Up to 73% of early human embryos have cells with large chromosome errors, and up to 30% of the progenitor cells that give rise to neurons show similar damage. Walsh pointed to fertility clinics, which routinely see embryos full of abnormal cells that still grow into healthy children. What was unclear was whether that chaos survives into mature neurons, and earlier studies of adult brains gave contradictory answers.

This study suggests both camps were partly right. Damaged neurons are common before birth and become rare after it. Once the perinatal cull is over, the share stays fairly stable: about 7.1% of neurons with widespread variants in infants up to age two and about 4.8% in people over 60. A slower removal of neurons with smaller variants continues for life. Pooling their data with six earlier studies, the authors estimate that this slow trickle removes about 13% of neurons over 60 years, close to the roughly 10% decline in neuron numbers measured by counting cells in aging brains.

The damage seems to start before neurons are even born. Breakpoints in the DNA cluster in regions that are copied early when a cell divides, the pattern expected from errors during cell division in progenitor cells rather than in mature neurons, which no longer divide.

What the study doesn’t settle

No one watched a neuron die in this study. Every brain was sampled once, after death, so the elimination is inferred by comparing the share of damaged neurons in different people at different ages. The donor numbers are small: nine prenatal and seven postnatal brains in the main analysis. There were no samples from the third trimester, so the exact timing of the cull, before or after delivery, remains unknown.

The mechanism is still a correlation. Damaged neurons show weak synapse-gene activity and strong cell-death signals, but the study does not prove that poor wiring is what triggers their removal. The link to autism is a hypothesis too: the team has not yet compared the brains of people with autism or epilepsy against those of people without.

Some damaged neurons also survive into old age in healthy brains. The researchers do not know whether that leftover genetic diversity is a harmless byproduct or serves some purpose.

Common questions about brain cells and DNA errors

Does the brain lose neurons before and after birth?

Yes. The developing human cortex makes more neurons than it keeps and removes an estimated 20% to 50% of them through programmed cell death, with a large wave around birth. This study suggests that neurons with badly damaged genomes are among the first to go.

Do all neurons in the same brain have identical DNA?

No. Every person is a genetic mosaic. In this study, about 64% of prenatal neurons and about 22.5% of neurons in people aged six or older carried at least one copy number variant, a duplicated or missing stretch of DNA.

Is this discovery linked to autism?

Possibly, but it is not proven. Neurons that survive rarely have damage in autism-related genes, which hints that the cull protects those genes. Whether the process works differently in autistic people has not yet been tested.

Could this affect the adult brain?

Slow removal of neurons with smaller DNA errors continues for life. The researchers are now asking whether the same process plays a part in neurodegenerative diseases such as Alzheimer’s.

What comes next

The study was published in Nature on 28 October 2026 and is open access. The Boston Children’s and Harvard team is now studying whether the elimination of flawed neurons differs in the developing brains of people with autism and other neurodevelopmental disorders, and whether the same mechanism contributes to neuron loss in Alzheimer’s disease. “We think that a lot of different diseases may show abnormalities in this process,” Walsh said. “It’s opened up a whole new universe of things to think about.”

Reference: Shao et al., “Perinatal elimination of genetically aberrant human cortical neurons,” Nature, 2026. DOI: 10.1038/s41586-026-11190-3

Tags: neuroscience, brain development, autism, Boston Children's Hospital, genetics, Harvard Medical School