Synaptic Pruning, Attractor Dynamics, and the Emergence of Schizophrenia

Excessive synaptic pruning during adolescence has been hypothesized to underlie the emergence of schizophrenia since the early 1980s, shortly after adolescent pruning was first described. The scale of normal pruning is substantial: "You actually lose almost 50% of your excitatory synapses in prefrontal cortex between the ages of 8 and about 30. And the slope is particularly steep during adolescence," said Bruno Averbeck, PhD. In most individuals, this loss likely makes cognition more robust while reducing plasticity, locking in cognitive processes developed during adolescence.¹ Limbic circuits are likely relatively spared because they support memory, which must be maintained with age.

In schizophrenia, excessive pruning may introduce pathologically strong attractor dynamics in prefrontal networks. These dynamics could manifest as fixed delusional ideation that persists despite conversation and even medication. Epidemiologic data show that many psychiatric disorders first emerge during this developmental window, although the research does not yet offer a basis for diagnosing individual patients, Averbeck said.

The model suggests a possible mechanism for several treatments. Ketamine, psilocybin, and likely electroconvulsive therapy—though not antipsychotics—may act partly by adding synapses to the brain. "If you're pruning synapses during adolescence and then you add synapses back, you're sort of bringing the brain back to a slightly more plastic state," Averbeck said, which may allow psychotherapy to work more effectively.

Neuroimaging findings warrant careful interpretation. Pruning removes only a small portion of gray matter volume, so even excessive pruning would alter volume only slightly. Moreover, apparent gray matter thinning in normal development likely reflects myelin encroaching on cortical layers 5 and 6, obscuring the gray–white boundary on MRI segmentation, rather than loss of neuropil.² This complicates the common interpretation of excessive gray matter thinning in psychiatric disorders.

Approximately 30% of patients with schizophrenia are treatment resistant, and the pathophysiologic difference between treatment-resistant patients and responders is unknown. It is also unclear how changing dopamine levels in the striatum affects striatal computation. Future work could measure striatal neural dynamics while manipulating dopamine levels to determine how those changes alter computation and may give rise to delusions, which are thought to be driven largely by excessive dopamine.

Dr Averbeck is a principal investigator at the National Institute of Mental Health.

References

1. Averbeck BB. Pruning recurrent neural networks replicates adolescent changes in working memory and reinforcement learning. Proc Natl Acad Sci U S A. 2022;119(22):e2121331119.

2. Natu VS, Gomez J, Barnett M, et al. Apparent thinning of human visual cortex during childhood is associated with myelination. Proc Natl Acad Sci U S A. 2019;116(41):20750-20759.