Autism spectrum disorder is increasingly understood as a condition rooted partly in the brain’s failure to properly trim its own wiring during development. During typical growth, the brain massively overproduces synaptic connections and then selectively eliminates the ones that are weak or redundant, a process known as synaptic pruning. In autism, converging evidence from postmortem brain tissue, animal models, and neuroimaging points to a pattern of reduced pruning that leaves too many synapses intact, particularly in regions responsible for social cognition, language, and sensory processing. The story is more nuanced than “too many synapses equals autism,” but the pruning angle has become one of the most productive lines of research in the field.
How the Brain Builds and Then Trims Its Connections
Early brain development is extravagant. In the first years of life, neurons form far more connections than the mature brain will ever use. In the human prefrontal cortex, the region most involved in planning, social reasoning, and impulse control, the peak number of synapses can exceed adult levels by two to three times, and the process of trimming them back stretches well into the third decade of life.1PubMed. Dendritic Spines: Synaptogenesis and Synaptic Pruning for the Developmental Organization of Brain Circuits This is not a design flaw. Overproduction followed by selective elimination is how the brain tailors its circuitry to the environment. Connections that get used are strengthened; those that don’t are tagged for removal. The process is activity-dependent, meaning experience literally shapes which synapses survive.
What makes human pruning unusual compared with other mammals is how long it takes. The extended timeline of synaptic reorganization in the prefrontal cortex has implications for understanding why neuropsychiatric conditions with social and cognitive features tend to emerge during particular windows of childhood and adolescence.2PubMed Central. Extraordinary neoteny of synaptic spines in the human prefrontal cortex If pruning goes wrong during any stretch of this long developmental window, the downstream effects on circuit function can be significant.
What Postmortem Studies Show in Autism
The most direct evidence that pruning is disrupted in autism comes from examining brain tissue. Studies of postmortem samples from people with autism, combined with data from animal models, consistently reveal a particular pattern: dendritic spines, the tiny protrusions on neurons where most excitatory synapses form, are found in higher densities than in neurotypical brains. These excess spines also tend to have immature shapes, suggesting they were never properly refined or eliminated.3PubMed Central. Dendrite and spine modifications in autism and related neurodevelopmental disorders in patients and animal models
At the same time, the dendrites themselves, the branching extensions of neurons that collect incoming signals, tend to be smaller and fewer in number. So the picture is not simply “more of everything.” It is more like a wiring diagram that was never cleaned up: too many connection points on branches that are themselves underdeveloped. This combination of excess spines with immature morphology and reduced dendritic complexity is a recurring finding across multiple brain regions and across both human tissue and genetically engineered mouse models.
The mTOR Pathway and Autophagy
One of the most important molecular discoveries in this area involves a cellular signaling pathway called mTOR. Under normal circumstances, this pathway helps regulate cell growth and a housekeeping process called autophagy, where cells break down and recycle their own damaged or unnecessary components. In the context of developing neurons, autophagy is part of how surplus synapses get dismantled. A landmark study in mice found that when mTOR signaling becomes overactive, autophagy is suppressed, and synaptic pruning stalls. Mice carrying a mutation in a gene called Tsc2, which normally keeps mTOR in check, developed behaviors resembling autism and retained far more dendritic spines than their unaffected littermates.4PubMed Central. Loss of mTOR-dependent macroautophagy causes autistic-like synaptic pruning deficits
This finding was significant because it connected a known genetic risk factor for autism (mutations in the TSC1/TSC2 genes cause tuberous sclerosis, a condition with a high rate of co-occurring autism) to a concrete cellular mechanism. The chain runs from gene to overactive signaling to failed autophagy to unpruned synapses to altered behavior. That kind of mechanistic clarity is rare in autism research, where the genetic landscape is otherwise enormously complex.
Microglia and Their Dual Role
Neurons do not prune themselves alone. Microglia, the brain’s resident immune cells, play a central role in the process. During normal development, microglia physically engulf and digest synapses that have been tagged for removal, guided in part by complement proteins, the same molecular flags the immune system uses to mark pathogens for destruction. When microglial function is disrupted, pruning goes awry.
Interestingly, the disruption can go in either direction. In one recent study using mice lacking a gene called SCN2A, which is one of the most commonly mutated genes in autism, microglia became partially activated and actually pruned too aggressively during certain developmental windows, stripping away synapses that should have been kept.5Molecular Psychiatry. Microglial over-pruning of synapses during development in autism-associated SCN2A-deficient mice and human cerebral organoids This challenges the simple narrative that autism always involves too little pruning. Different genetic mutations can push microglia toward either under-pruning or over-pruning, and the behavioral outcome may depend on which brain regions are affected and when.
Disruptions in microglial signaling also arise from the immune balance between a pregnant mother and her developing fetus. When the maternal immune system is activated during pregnancy, by infection or inflammation, offspring in mouse models show increased spine densities and reduced expression of a key receptor that microglia use to carry out pruning. These offspring also display behaviors relevant to autism.6PubMed. Prenatal infection leads to ASD-like behavior and altered synaptic pruning in the mouse offspring The effects appear to differ by sex, which aligns with the well-known finding that autism is diagnosed far more often in males than females, though the reasons for that disparity are still not fully worked out.
Astrocytes Are Involved Too
Microglia get most of the attention, but they are not the only cells that eat synapses. Astrocytes, the star-shaped support cells that outnumber neurons in many brain regions, also actively engulf and eliminate synapses through their own set of molecular pathways. Mice lacking these astrocyte-mediated pruning pathways fail to properly refine their neural connections during development and retain excess functional synapses.7PubMed Central. Astrocytes mediate synapse elimination through MEGF10 and MERTK pathways This process is strongly dependent on neuronal activity, meaning that experience-driven signals direct astrocytes toward which synapses to clear.
Whether astrocyte-mediated pruning is specifically impaired in autism has not been studied as thoroughly as the microglial pathway, but the overlap in function is hard to ignore. Any model of pruning failure in autism that focuses exclusively on microglia is likely missing part of the picture. The brain’s cleanup crew has multiple members, and disrupting any one of them could leave excess synapses in place.
Brain Overgrowth in Infancy
If pruning fails and too many connections are retained, you might expect the brain to be physically larger than normal, at least during certain developmental windows. That is exactly what neuroimaging studies have found. In a prospective study tracking infants at high familial risk for autism, those who went on to receive a diagnosis at age two showed an unusual pattern: cortical surface area expanded more rapidly than normal between six and twelve months of age, followed by overall brain volume overgrowth between twelve and twenty-four months. The degree of overgrowth was linked to the emergence and severity of social deficits.8PubMed Central. Early brain development in infants at high risk for autism spectrum disorder
This timing is striking. The period of cortical surface expansion corresponds to the height of synapse formation, while the subsequent volume overgrowth overlaps with the window when pruning should be ramping up. It suggests that in children who will later be diagnosed with autism, the overgrowth is not random but tracks with the developmental schedule of synaptic overproduction and the expected onset of elimination. The brain is building circuits on schedule, but the cleanup is delayed or incomplete.
What Excess Synapses Do to Circuits
Retaining too many synapses is not like having extra storage capacity on a hard drive. Extra connections create noise. The brain’s ability to process information depends on a delicate balance between excitatory signals (those that make neurons fire) and inhibitory signals (those that quiet them down). When pruning falls short, the balance tips toward too much excitation, a pattern consistently observed in autism research.9PubMed. Microglia in the pathogenesis of autism spectrum disorders
This excitatory tilt helps explain some of the core experiences associated with autism. A brain with too many active connections in sensory areas may struggle to filter incoming stimuli, contributing to the sensory sensitivities that many autistic individuals report, where ordinary sounds, textures, or lights feel overwhelming. Altered connectivity patterns in sensory cortices and in the limbic system, which processes emotion, have been linked to these atypical sensory experiences.10PubMed Central. Neuroplasticity-Based Approaches to Sensory Processing Alterations in Autism Spectrum Disorder On the social side, the prefrontal cortex needs highly refined circuitry to handle the rapid, ambiguous signals of face-to-face interaction. Noisy, unpruned circuits in that region could make social processing less efficient rather than impossible, which matches the experience many autistic people describe.
Imaging studies of older individuals with autism bear this out at a structural level. In neurotypical development, the density of small-diameter axons in certain brain regions decreases with age as local and short-range connections are pruned away. In autism, that density stays high, consistent with local overconnectivity that persists into adulthood.11PubMed Central. Age-related differences in axon pruning and myelination may alter neural signaling in autism spectrum disorder The result is not just more connections but less refined ones, circuits that never underwent the competitive sharpening that typical development provides.
The variety of pruning abnormalities across different brain regions may also explain why autism presents so differently from person to person. One model proposes that the dysregulation of postnatal pruning produces a wide spectrum of signal-over-noise problems, and the specific profile of which regions are affected and to what degree accounts for much of the clinical heterogeneity seen in autism, from language delays and social difficulties to autonomic nervous system differences and common co-occurring conditions.12PubMed Central. Autism spectrum disorders pathogenesis: Toward a comprehensive model based on neuroanatomic and neurodevelopment considerations
The Genetic Landscape Is Broader Than One Pathway
It would be tidy if all autism-related genes converged on a single pruning mechanism, but the genetics are sprawling. Hundreds of genes have been associated with autism risk, and many of them influence synaptic structure and function through different routes. Fragile X syndrome, the most common single-gene cause of intellectual disability and a frequent co-occurrence with autism, involves disruptions to spine architecture, synaptic plasticity pathways, and local protein production at the synapse.13PubMed Central. A Synaptic Perspective of Fragile X Syndrome and Autism Spectrum Disorders The mTOR-autophagy pathway described earlier is just one thread. Other genes affect complement signaling (the tagging system microglia use), neuronal activity patterns that determine which synapses get flagged, and the structural proteins that hold synapses together.
This diversity is one reason why autism research has moved toward describing the condition as a set of “synaptopathies,” disorders united not by a single genetic cause but by their downstream effect on synapses. The shared endpoint is altered synaptic structure and function, even though the roads leading there are numerous.
Why Schizophrenia Looks Like the Mirror Image
An illuminating comparison is with schizophrenia, which appears to involve the opposite pruning error. Where autism tends toward too little pruning and synaptic overabundance, schizophrenia is associated with excessive pruning and synaptic loss, particularly during adolescence and early adulthood.14Undergraduate Research in Natural and Clinical Science and Technology. Neural Pruning and Autism: The Scientific Connection This framing is inevitably a simplification, since both conditions are heterogeneous, but it highlights how much the brain’s wiring depends on pruning landing in a narrow sweet spot. Too little, and circuits are noisy and poorly differentiated. Too much, and connections needed for normal thought and perception are stripped away. The fact that both endpoints produce serious psychiatric conditions underscores how precisely calibrated pruning needs to be.
Critical Periods and Timing
The brain does not stay equally plastic forever. There are developmental windows, sometimes called critical periods, during which circuits are most amenable to being shaped by experience. One hypothesis proposes that in autism, the window of neuroplasticity that normally extends into adolescence and supports the maturation of language and social abilities closes prematurely, forcing the brain to operate with circuitry that has not been fully refined.15PubMed Central. Autism as the Early Closure of a Neuroplastic Critical Period Normally Seen in Adolescence
If this is correct, it would mean that the problem is not just failed pruning at a single moment but a compressed timeline for the entire process. A brain that stops reorganizing its circuitry too early would retain whatever wiring it had at closure, whether or not that wiring was optimal. This could explain why early intervention tends to produce better outcomes: the earlier you provide enriched, structured input, the more of the remaining plastic window you can leverage before it narrows.
Rapamycin and the Search for Treatments
The mTOR-autophagy discovery opened a tantalizing therapeutic possibility. Rapamycin, a drug that inhibits mTOR and is already approved for other medical uses, was tested in the Tsc2 mutant mice described earlier. In those mice, rapamycin corrected both the excess spine density and the autism-like behaviors. Critically, it did not work in mice whose autophagy machinery had been independently knocked out, confirming that the drug’s effect ran through the autophagy pathway rather than some other action of mTOR inhibition.16Neuron. Loss of mTOR-Dependent Macroautophagy Causes Autistic-like Synaptic Pruning Deficits
This is promising but comes with heavy caveats. The mice had a specific single-gene mutation. Most autism does not arise from TSC mutations, and there is no guarantee that restoring autophagy would help in cases driven by entirely different genetic or environmental factors. Rapamycin also suppresses the immune system broadly, which limits its use as a long-term treatment in children. Researchers are now exploring whether more targeted approaches, drugs that enhance autophagy in neurons without the systemic immune suppression, could be developed. That work is still in early preclinical stages.
Sleep, Synapses, and a Vicious Cycle
Sleep disturbances are remarkably common in autism, affecting a large proportion of autistic children and adults. This is not just an incidental quality-of-life issue. Sleep plays a well-established role in synaptic homeostasis: during deep sleep, the brain scales back synaptic strength and clears metabolic waste, effectively doing some of the maintenance work that overlaps with pruning. The connection has led researchers to propose that synaptic dysfunction may be a shared mechanism linking autism and sleep disturbances, rather than one simply causing the other.10PubMed Central. Neuroplasticity-Based Approaches to Sensory Processing Alterations in Autism Spectrum Disorder
If excess synapses make normal sleep harder to achieve, and poor sleep in turn impairs the brain’s ability to prune and maintain synapses, the result is a feedback loop. Addressing sleep problems in autism could, in theory, support the brain’s residual capacity for synaptic housekeeping. It is a hypothesis more than a proven treatment strategy at this point, but it reframes sleep difficulties in autism as something potentially connected to the underlying neurobiology rather than just a behavioral symptom to manage.
Why Pruning Research Has Not Yet Changed Clinical Practice
Despite how compelling the pruning story is, it has not yet translated into new diagnostic tests or approved therapies for most autistic individuals. The reasons are practical. You cannot measure spine density in a living person’s brain with current clinical tools. Neuroimaging can detect large-scale differences in brain volume and connectivity, but it cannot resolve individual synapses. The genetic landscape is so fragmented, with hundreds of risk genes each contributing a small amount, that targeting any single molecular pathway will likely help only a subset of people.
There is also a values question embedded in the science. Many autistic adults and advocacy organizations emphasize that autism is a neurological difference rather than a disease to be cured, and that research focused narrowly on “fixing” pruning may miss the point. The pruning framework is most useful, in this view, when it helps explain and accommodate autistic experiences, such as sensory overload or the need for predictable environments, rather than when it is used solely to justify pharmacological normalization of brain architecture. That tension shapes which research questions get funded and how findings are communicated to the public, and it is unlikely to resolve neatly anytime soon.