Human fetal brains purge billions of mutant neurons around birth, study finds

Every human brain starts out as a construction site that builds far more cells than it will eventually keep. During fetal development the cerebral cortex churns out neurons in vast excess, and a large share of those cells are dismantled and cleared away before and shortly after birth through programmed cell death, a process developmental biologists have documented for decades. What has never been settled is whether that pruning is purely a matter of arithmetic, a way of matching supply to the final size of the skull, or whether it also acts as a filter on the quality of the cells themselves.

Copy number variants are duplications or deletions of stretches of chromosomal DNA. Some are harmless, but large chromosome-scale changes are a well-known source of developmental disease, and they are strikingly common in early embryos. The new work notes that up to 73 percent of embryos carry large chromosomal aneuploidies in at least some cells. Reading those variants one neuron at a time, across thousands of cells and across a human lifespan, was until recently too costly and too error prone to attempt at scale.

A paper published in Nature on October 28, 2026 now argues for the second explanation, and it does so with numbers that surprised its own authors. The study, titled Perinatal elimination of genetically aberrant human cortical neurons, comes from a team at Harvard Medical School and Boston Children's Hospital, with senior authors including Peter Park of the Blavatnik Institute and Christopher Walsh, an investigator at the Howard Hughes Medical Institute. It combines two independent single-cell whole-genome sequencing methods to read the DNA of individual neurons recovered from postmortem human brain tissue.

Nature reported on October 28, 2026 that the analysis rests on more than 2,300 nerve cells, drawn from donors who ranged from fetuses at 14 weeks of gestation to nonagenarians. The central claim is stark. A large fraction of the neurons born before birth carry badly damaged genomes, and the brain removes most of them around the moment of delivery.

Key Facts

The headline figure is 25.9 percent. In the paper, published on October 28, 2026, the authors report that an average of 25.9 percent of prenatal cortical neurons, within a range of 12.7 to 46 percent, carried widespread copy number variants involving multiple chromosomes. Roughly one in four neurons in the fetal cortex had gained or lost several chromosomal segments at once, a burden far heavier than anything seen in the adult tissue the team examined.

After birth that proportion collapses. Nature reported on October 28, 2026 that the share of neurons with such abnormalities fell by more than fourfold, and the paper puts the average after birth at 6.2 percent. Across the two sequencing approaches the reduction worked out to about 4.5-fold, a result the authors describe as statistically robust, with a Fisher meta-analysis P value of 3.5e-6.

The dataset is deliberately redundant. The team sequenced 2,357 single neurons in total, using Tn5 transposase based amplification on 1,526 neurons from 16 individuals and primary template amplification on 831 neurons from 12 individuals. Because both methods produced similar results, the finding does not hinge on the quirks of a single chemistry. Harvard Medical School said in its release on October 28, 2026 that the first approach amounts to a cheap, low-resolution read of each genome, with the second serving as the deeper validation.

Genome sequence alone does not explain why the flawed cells disappear. The researchers profiled 44,861 nuclei with single-nucleus ATAC and single-nucleus RNA sequencing and found that neurons with high copy number burdens showed abnormal activity of genes involved in creating and maintaining synapses. The paper also reports that the surviving neurons were depleted for dosage-sensitive genes and for genes implicated in neurodevelopmental disorders, at a hypergeometric test P value below 0.05, which points to selection rather than random loss.

The timing is equally pointed. Neurons with widespread copy number variants were mainly eliminated perinatally, roughly in parallel with periods of programmed cell death, while neurons carrying smaller burdens declined only slowly with postnatal age. Medical Xpress reported on October 28, 2026 that the same flawed cells continue to be cleared throughout adulthood, though at a much lower rate. Walsh said the team was totally unprepared for how common the variants were in early development, and he called the process a Darwinian competition that functions as nature's quality control in the brain.

Analysis

The most persuasive element of the study is not the size of the drop but the convergence behind it. Two sequencing chemistries, two cohorts, and a separate chromatin and expression survey all point the same way, and the authors concede that the original numbers looked so implausible that they spent years checking whether the results were real. Christos Proukakis, a neurologist at University College London who was not involved in the work, told Nature that the neurons which survive may be the ones that kept the most intact genomes during development.

The bigger picture here is that the fetal cortex is not simply overbuilding and then trimming, as textbooks have long described. It is running a genomic inspection line, and the inspection appears tied to whether a young neuron can wire itself into a circuit. Neurons that fail to compete for synapses already had a known death sentence. What the new data add is a reason some of them fail, written into their chromosomes, and a molecular signature in the genes that build and maintain those synapses.

That reframing carries a judgement: purifying selection during human neurodevelopment is now the most economical explanation for the data, and the burden of proof now sits with explanations built on technical artefact. The gene-level result is what makes this more than a correlation. If loss were random, the surviving population would not be systematically depleted of dosage-sensitive genes and of genes tied to neurodevelopmental disorders.

What remains inference rather than measurement is the causal chain in living tissue. The work is built on postmortem samples, so the elimination itself is reconstructed from a snapshot series rather than watched in real time, and the authors say the mechanism may be vulnerable in disease rather than claim to have demonstrated it. The concentration of the effect in the perinatal window, alongside programmed cell death, is consistent with selection but does not by itself prove that synaptic failure is the executioner. Diane Shao, a paediatric neurogeneticist at Boston Children's Hospital and first author on the paper, framed the result more broadly, saying the brain is selecting for genomic integrity and that disrupting that system could be part of what goes wrong in conditions such as autism.

Why It Matters

The immediate relevance is clinical. Copy number variants are already implicated in autism spectrum disorder, schizophrenia, epilepsy and intellectual disability, and the study suggests a mechanism by which a brain could end up with too few healthy neurons or too many marginal ones. If a quality-control checkpoint is meant to clear aberrant cells around birth, then a checkpoint that runs too weakly, too late or not at all would leave a cortex built from a compromised pool, which differs from a mutation acting inside a single cell at the moment of wiring.

There is also a quantitative point that reframes how common damage really is. Park said that between 13 and 46 percent of neurons in the developing brain are riddled with copy number variants, yet the adult cortex is largely composed of cells with intact genomes. Any model of brain development that assumes genomic uniformity across neurons is therefore working from a picture the data do not support, and any effort to interpret somatic mutation in neurological disease must account for the fact that most of it is normally removed.

For the field, the practical consequence is a new set of targets: the genes that distinguish surviving from eliminated neurons, and the synaptic programmes that appear to mediate the choice, become candidates for study in childhood onset conditions. Shao's point is that learning more about this process could be key to understanding disorders that typically appear early in life, which places a developmental cleanup operation at the centre of questions usually asked about mature circuits.

Next Up

The open questions are specific. Which dosage-sensitive genes actually tip a neuron toward elimination, whether the same selection operates in brain regions beyond the cerebral cortex, and whether the process can be watched in living tissue rather than inferred from postmortem series remain unresolved. The study's own numbers also invite a wider survey, since the donors span 14 weeks of gestation to 90 years but remain a finite set.

The result will also be tested against disease tissue. If the checkpoint is disrupted in neurodevelopmental conditions, brains from those conditions should show an altered ratio of aberrant to intact neurons at the same developmental stages. Until then, the study stands as evidence that the human brain does not merely build too much and cut back, but inspects what it has built and discards the defective.