Psilocybin Rewires the Cortex Selectively — and Only Where Neurons Fire

For a decade, the headline finding from Yale and Cornell labs has been simple: one dose of psilocybin grows dendritic spines on frontal cortical neurons and the spines last for weeks. That story explained durability but skipped the most clinically relevant question — which inputs land on those new spines? Jiang et al. answer it, and the answer reframes how we should think about psychedelic-assisted therapy.

The plasticity is not promiscuous. It is selective at the level of brain networks and gated by spiking activity at the moment of the experience.

What the Cornell-Yale-Allen team actually mapped

Using monosynaptic rabies tracing — a virus that crosses exactly one synapse retrogradely — Jiang and colleagues mapped, brain-wide, the presynaptic neurons that send axons onto frontal cortical pyramidal neurons in mice that had received either psilocybin (1 mg/kg, intraperitoneal) or saline 24 hours earlier. Tissue clearing + light-sheet microscopy + automated nuclei detection allowed them to count input cells across 65 anatomical regions per brain.

The PT subtype — pyramidal tract neurons that project to the brainstem and other subcortical targets, and which prior work has shown are required for psilocybin's behavioral effects — gained inputs disproportionately from the medial network (retrosplenial cortex, posterior parietal regions, primary somatosensory and visual cortex, thalamic relay nuclei). They simultaneously lost inputs from the infralimbic and medial orbital cortex, anterior insula, CA1, and basolateral amygdala. The IT subtype, which projects within the cortex, mirrored this pattern in reverse.

The retrosplenial cortex is the mouse homolog of the human default mode network. So the rewiring is, in shorthand: DMN-like and sensory inputs onto subcortically-projecting frontal neurons go up; emotional and prefrontal-recurrent inputs come down. A subsequent slice electrophysiology experiment confirmed that the retrosplenial→ACAd synaptic drive is functionally potentiated for at least three days post-dose, with a presynaptic component (decreased paired-pulse ratio).

The single most important experiment is the silencing one. When the retrosplenial cortex was chemogenetically silenced during the window of psilocybin action, the input gain to PT neurons disappeared. Plasticity required that the upstream region be firing while the drug was on board. C-Fos activation across cortex predicted input cell gain (r = 0.44, p = 0.02), tying spiking activity to anatomical rewiring at the brain-wide level.

What this means for therapy

If the rewiring is activity-dependent, then what the patient's brain is doing during the session is part of the mechanism, not background. The "psychological support" element of psychedelic-assisted therapy — set, setting, music, focused attention, the therapist's presence — is no longer plausibly dismissable as ritual. It is the activity pattern that gets crystallized into the next month of cortical wiring.

A few specific implications. First, distraction or dissociation during the dosing window may actively suppress the connections you wanted to potentiate. Second, deliberately engaging certain processes during the experience — interoceptive attention, autobiographical recall, exposure to a target stimulus — may bias which inputs get strengthened. The animal data do not yet validate any specific therapeutic protocol, but they make the central premise of structured psychedelic therapy mechanistically coherent for the first time. Third, for clinicians referring patients to psilocybin programs in Oregon or under Australian AP regulations: ask what happens during the experience, not just before and after.

The flip side: anxiolytic premedication, heavy benzodiazepine pretreatment, or environments that suppress cortical drive may quietly degrade the structural outcome the patient came in for.