Psilocybin

Psilocybin’s lasting action requires pyramidal cell types and 5-HT2A receptors

This mouse study investigates how psilocybin affects different types of brain cells in the medial frontal cortex (mPFC; decision-making processes and judgement). The research finds that psilocybin increases dendritic spine density in both pyramidal tract (PT) and intratelencephalic (IT) neurons, but only PT neurons are essential for psilocybin's anti-stress effects through 5-HT2A receptor activation.

Authors

Bilash, O. M.
Che, A.
Davoudian, P. A.

Published

April 2, 2025

Nature Communications

individual Study

Links

Read Paper DOI Google Scholar

Abstract

Psilocybin is a serotonergic psychedelic with therapeutic potential for treating mental illnesses. At the cellular level, psychedelics induce structural neural plasticity, exemplified by the drug-evoked growth and remodelling of dendritic spines in cortical pyramidal cells. A key question is how these cellular modifications map onto cell-type-specific circuits to produce the psychedelics’ behavioural actions. Here we use in vivo optical imaging, chemogenetic perturbation and cell-type-specific electrophysiology to investigate the impact of psilocybin on the two main types of pyramidal cells in the mouse medial frontal cortex. We find that a single dose of psilocybin increases the density of dendritic spines in both the subcortical-projecting, pyramidal tract (PT) and intratelencephalic (IT) cell types. Behaviourally, silencing the PT neurons eliminates psilocybin’s ability to ameliorate stress-related phenotypes, whereas silencing IT neurons has no detectable effect. In PT neurons only, psilocybin boosts synaptic calcium transients and elevates firing rates acutely after administration. Targeted knockout of 5-HT2A receptors abolishes psilocybin’s effects on stress-related behaviour and structural plasticity. Collectively, these results identify that a pyramidal cell type and the 5-HT2A receptor in the medial frontal cortex have essential roles in psilocybin’s long-term drug action

Unlocked with Blossom Pro

Research Summary of 'Psilocybin’s lasting action requires pyramidal cell types and 5-HT2A receptors'

Introduction

Psilocybin is a serotonergic psychedelic that produces sustained improvement in depressive symptoms in clinical trials and has been shown in rodents to provoke enduring increases in dendritic spine density and size in cortical pyramidal cells. Neocortical pyramidal neurons are heterogeneous, principally comprising pyramidal tract (PT) neurons that project to subcortical targets and intratelencephalic (IT) neurons that project within the forebrain; these cell types differ in physiology and circuit participation. Prior work demonstrated psychedelic-evoked structural plasticity in pyramidal cells, but it remained unclear which excitatory cell types drive the drugs’ behavioural effects and how receptor mechanisms such as 5-HT2A contribute at the cell-type level. Shao and colleagues set out to determine how a single dose of psilocybin affects PT and IT neurons in the mouse medial frontal cortex, and whether those effects are necessary for psilocybin’s longer-term action on stress-related behaviours. The study combined longitudinal in vivo two-photon structural imaging, awake dendritic calcium imaging, cell type–specific electrophysiology, chemogenetic silencing, and conditional, regionally targeted knockout of the Htr2a gene (encoding 5-HT2A) to map structural, physiological and behavioural consequences to cell types and receptor expression.

Methods

The investigators used adult mice (C57BL/6J and several Cre-driver and floxed Htr2a lines) and applied cell type–specific viral strategies to label and manipulate PT and IT neurons in the medial frontal cortex (ACAd/medial MOs). PT neurons were targeted via retrograde AAV injections into the ipsilateral pons; IT neurons via retrograde injections into the contralateral striatum. For structural studies, sparse EGFP labelling allowed longitudinal two-photon imaging of apical tuft dendrites through cranial windows. Mice received a single i.p. injection of psilocybin (1 mg/kg) or saline and were imaged at baseline and multiple time points up to 65 days to assess spine density, formation and elimination rates; confocal microscopy on fixed tissue was used post hoc to examine spine head width. Acute dendritic and spine calcium signalling were measured in awake, head-fixed animals expressing GCaMP6f in PT or IT neurons; fields were imaged for 10 minutes pre-treatment and again within 1 hour after psilocybin or saline. Automated event detection after neuropil correction and branch-subtraction estimated branch and branch-independent spine calcium transients. Single-unit electrophysiology in awake mice used Neuropixels probes combined with opto-tagging: ChR2 was expressed in PT or IT neurons in Fezf2-CreER or PlexinD1-CreER mice, allowing identification of cell-type-tagged units and assessment of firing changes after psilocybin (1 mg/kg i.p.). Behavioural causality was tested using inhibitory DREADDs (hM4DGi) expressed broadly in PT or IT neurons (tamoxifen-inducible Cre lines Fezf2-CreER and PlexinD1-CreER). Deschloroclozapine (DCZ, 0.1 mg/kg i.p.) was given 15 minutes before psilocybin to silence targeted neurons during drug exposure. Behavioural assays included head-twitch response, learned helplessness (active avoidance after inescapable shocks), tail suspension test, and fear-extinction after chronic restraint stress. To test receptor dependence, the study used Htr2a f/f mice with regionally targeted AAV-CaMKII-Cre to knock out 5-HT2A receptors in the medial frontal cortex, and a PT-targeted strategy combining retrograde Cre with Floxed EGFP in Htr2a f/f for cell-type-targeted knockout. Validation included transcript qPCR and slice electrophysiology. Statistical analyses for imaging used linear mixed effects models to account for nested measurements; electrophysiology and behaviour used standard ANOVA/t-tests as specified.

Results

Longitudinal two-photon imaging showed that a single dose of psilocybin (1 mg/kg i.p.) increased apical tuft spine density in both PT and IT neurons in medial frontal cortex. Quantitatively, on day 1 psilocybin produced a 19±2% increase in PT spine density versus −4±2% for saline, and a 14±2% increase in IT spine density versus −4±1% for saline; the main effect of treatment was P < 0.001 (mixed effects model). The elevated spine density persisted through the final imaging time point at 65 days. For both cell types the increase was driven primarily by a rapid rise in spine formation rate within 1 day; PT neurons also showed a modest decrease in spine elimination. These structural changes were observed in both sexes and did not reflect changes in protrusion length; laminar location within IT neurons did not explain the cell-type differences reported. Behavioural experiments using chemogenetic silencing established a causal role for PT but not IT neurons in psilocybin’s effects on stress-related tests. Psilocybin induced the expected head-twitch response, and this was not altered by DREADD-mediated silencing of either cell type. By contrast, psilocybin’s capacity to reduce escape failures in learned helplessness and to decrease immobility in the tail suspension test was abolished when frontal cortical PT neurons were silenced during drug administration (interaction effects of treatment × DREADD P < 0.001), whereas silencing IT neurons had no detectable impact. Frontal cortical PT neurons were also required for psilocybin-driven facilitation of fear extinction in chronically stressed mice. Acute calcium imaging showed that psilocybin preferentially increased dendritic and spine calcium transients in PT neurons. For dendritic branches, the rate of spontaneous calcium events rose by 23±4% (n = 149 branches, 4 mice) after psilocybin versus 5±2% after saline; IT branches showed no change (interaction P = 0.008). For branch-independent spine transients, PT spines showed a 68±5% increase after psilocybin (n = 2637 spines, 4 mice) compared with 37±6% after saline, whereas IT spines showed small, nonsignificant changes (interaction P < 0.001). These data indicate preferential enhancement of synaptic calcium signalling in PT neurons. Cell type–specific awake electrophysiology revealed that a subset of PT neurons robustly increased firing after psilocybin. Among opto-tagged PT units (n = 90 tagged neurons, 5 mice), 16% displayed substantial post-drug firing increases (14 cells with post-drug mean z-score Z>2), and, on average, PT neurons showed a significant increase in spike rates after psilocybin (P < 0.001). Tagged IT neurons (n = 57, 5 mice) did not show a consistent firing change (P = 1.0). Thus, psilocybin acutely elevates spiking in a subset of PT neurons. Region- and cell-targeted deletion of 5-HT2A receptors established receptor dependence for both behavioural and structural effects. Cre-mediated knockout in medial frontal cortex of Htr2a f/f mice abolished the psilocybin-induced alleviation of stress-related behaviours in learned helplessness and tail suspension tests, while leaving the local head-twitch response intact (constitutive, widespread knockout reduced head-twitches, indicating regional specificity). qPCR and slice recordings validated local receptor loss: GFP+ cells from Cre-injected cortex lacked Htr2a transcript and did not show 5-HT-evoked increases in sEPSCs. Importantly, PT-targeted 5-HT2A knockout prevented psilocybin-induced spine density increases: control PT neurons showed 15±2% spine density increase after psilocybin versus −2±2% for saline, whereas PT neurons lacking 5-HT2A had no psilocybin-evoked increase (−1±2% for psilocybin; interaction P = 0.01). Confocal analyses of fixed tissue on day 3 showed a psilocybin-evoked increase in spine head width in PT apical tufts (0.53±0.01 μm vs 0.50±0.01 μm for saline), an effect absent when 5-HT2A was deleted (interaction P = 0.025).

Discussion

The investigators conclude that psilocybin’s long-term behavioural effects in stress-related paradigms depend on a specific pyramidal cell type—frontal cortical PT neurons—and on local 5-HT2A receptor signalling. Although both PT and IT neurons exhibit durable structural plasticity after psilocybin, only PT neurons showed preferential acute increases in synaptic calcium signalling and an identifiable subset that increased firing, and silencing PT neurons negated psilocybin’s beneficial effects on learned helplessness, tail suspension and fear extinction. Shao and colleagues note that abundant Htr2a transcript in both cell types cannot fully explain the PT selectivity and propose circuit-level explanations. They suggest that biased long-range inputs preferentially targeting PT neurons or disinhibition via suppression of specific interneuron classes (for example, deep-layer somatostatin interneurons) could focus psilocybin’s in vivo action on PT neurons. The authors position this cell-type dissociation as potentially useful for future efforts to separate therapeutic effects from perceptual/psychedelic effects, noting that PT neurons form the cortical subcortical output pathway and have been implicated in state transitions such as waking. Key limitations acknowledged include that the role of IT neuron plasticity remains unresolved—it may be epiphenomenal or relevant for behaviours not assayed here—and that receptor expression alone does not account for the observed cell-type specificity. The authors recommend further studies to dissect circuit and receptor mechanisms and to examine how these findings might inform neuromodulation or precision treatments that aim to harness or augment psychedelic-evoked plasticity.

View full paper sections

MAIN

Psilocybin is a classic psychedelic that has shown promise as a treatment for psychiatric disorders. Clinical trials demonstrated that one or two sessions of psilocybin-assisted therapy attenuate depression symptoms for many weeks. It has been hypothesized that antidepressants may work by forming and strengthening synapses in the prefrontal cortex, which counteracts synaptic dysfunction in depression. Consistent with this framework, recent studies in mice demonstrated that a single dose of psilocybin or related psychedelic drugs leads to sustained increases in the density and size of apical dendritic spines in cortical pyramidal cells. However, neurons are heterogeneous, and it is unclear how psychedelic-evoked neural adaptations manifest in different excitatory cell types. Notably, there are two major, nonoverlapping populations of cortical pyramidal cells: pyramidal tract (PT) and intratelencephalic (IT) neurons. PT and IT neurons have distinct cellular properties and participate in different long-range circuits because they send disparate axonal projections to communicate with different brain regions(Fig.). PT neurons are subcortical projection neurons that send axons to subcerebral destinations including the thalamus and brainstem, and also to ipsilateral cortex and basal ganglia. By contrast, axons of IT neurons stay within the cerebrum, but can project to both ipsilateral and contralateral cortical and striatal locations. These pyramidal cell types constitute a microcircuit motif that is found in most regions in the neocortex, supporting a range of behavioral functions. Impairments of these distinct types of pyramidal cells have been linked to neuropsychiatric disorders. How may PT and IT neurons respond to psilocybin? Classic psychedelics are agonists at serotonin receptors. In response to serotonin, some pyramidal cells elevate spiking activity via 5-HT2A receptors, whereas other pyramidal cells suppress firing via 5-HT1A receptors. It was reported that in mouse brain slices, serotonin-evoked firing occurs in pyramidal cells with commissural projections (IT neurons), but not those with corticopontine projections (PT neurons). Transcript expression in the mouse frontal cortex corroborates this view: although PT and IT neurons both express Htr2a 25 , there is more Htr2a in IT neurons. However, another study performed in anesthetized rats showed that psychedelics can excite midbrain-projecting pyramidal cells, which would constitute PT neurons. Therefore, current literature provides conflicting clues towards how the main pyramidal cell types should contribute to psychedelic drug action. In this study, we measured the acute and long-term impact of psilocybin on PT and IT neurons in the mouse medial frontal cortex in vivo. We found that PT neurons were the pyramidal cell type selectively driven by psilocybin to increase synaptic calcium transients and elevate spiking activity in awake animals. Moreover, although psilocybin evokes structural plasticity in both PT and IT neurons, causal manipulations indicate that frontal cortical PT neurons are needed for psilocybin's effects in stress-related behavioral assays. Using conditional knockout mice, we found that 5-HT2A receptor is required for psilocybin-evoked structural remodeling in PT neurons. The results thus reveal frontal cortical PT neurons and 5-HT2A receptor as essential components mediating psilocybin's long-term drug action in the brain.

CELL-TYPE SPECIFICITY IN THE STRUCTURAL PLASTICITY INDUCED BY PSILOCYBIN

To sparsely express EGFP in PT or IT neurons for dendritic imaging, we injected a low titer of the retrogradely transported AAVretro-hSyn-Cre in the ipsilateral pons or contralateral striatum, and AAV-CAG-FLEX-EGFP in the medial frontal cortex of adult C57BL/6J mice (Fig.Extended Data Fig.). We focused on the cingulate and premotor portion (ACAd/medial MOs) of the medial frontal cortex, because brain-wide c-Fos mapping indicates the region robustly responds to stressand psilocybin. Histology confirmed that EGFP-expressing cell bodies of PT neurons were restricted to deep cortical layers, whereas somata of IT neurons were spread across layers 2/3 and 5 (Fig.), in agreement with the laminar distribution of the cell types. We used two-photon microscopy to image through a chronically implanted glass window while the animal was anesthetized. We visualized the same apical tuft dendrites located at 20 -120 µm below the pial surface over multiple sessions across >2 months (Fig.). At baseline, PT neurons had lower spine density but higher spine head width than IT neurons (Extended Data Fig.)., Similar to b-c for IT neurons. CP, caudoputamen. f-g, Longitudinal two-photon microscopy. h, Example field of view, tracking the same apical tuft dendrites for 65 days after psilocybin. i, Density of dendritic spines in the apical tuft of PT neurons after psilocybin (yellow; 1 mg/kg, i.p.) or saline (gray) across days, expressed as fold-change from baseline in first imaging session (day -3). Mean and s.e.m. across dendrites. j, Spine formation rate determined by number of new and existing spines in consecutive imaging sessions across two-day interval, expressed as difference from baseline in first interval (day -3 to day -1). k, Similar to j for elimination rate. l-n, Similar to i-k for IT neurons after psilocybin (purple) or saline (light purple). There was no cell-type difference in psilocybin's effect on spine density, formation rate, or elimination rate (p-values for interaction effect of treatment × cell type, indicated in plots, mixed effects model). *, p < 0.05. ***, p < 0.001, post hoc with Bonferroni correction for multiple comparisons. Sample size n values are provided in Methods. Statistical analyses are provided in Supplementary Table. For each of the four cell-type and treatment conditions, we tracked and analyzed 1040-1147 spines from 69-85 dendrites in 8-9 mice of both sexes. For statistical tests, mixed effects models were used, which included random effects terms to account for the nested nature of the data where spines are imaged from the same dendrites or same mouse. Details for sample sizes and statistical tests for all experiments are provided in Supplementary Table. One dose of psilocybin (1 mg/kg, i.p.) increased spine density in both pyramidal cell types (PT: 19±2% for psilocybin, -4±2% for saline on day 1; IT: 14±2% for psilocybin, -4±1% for saline; main effect of treatment: P < 0.001, mixed effects model; Fig.; Extended Data Fig.). The elevated number of dendritic spines remained significant in the last imaging session at 65 days for psilocybin relative to control. For both cell types, the higher spine density was driven by an increase in the rate of spine formation within 1 day after psilocybin (Fig., o; Extended Data Fig.), with additionally a smaller decrease in spine elimination rate for PT neurons (Fig.; Extended Data Fig.). The psilocybin-evoked structural remodeling occurred in mice of both sexes (Extended Data Fig.). There was no change detected in spine protrusion length (Extended Data Fig.). Due to the sparse labeling, we could often trace the dendrites back to the cell body. Separately analyzing IT neurons residing in layer 2/3 and layer 5 (Extended Data Fig.) indicated that laminar position is not the reason for the difference observed across cell type. These results replicate our prior finding 7 that psilocybin increases spine density in frontal cortical pyramidal cells, while extending the observation window to show that the change persists for >2 months in mice, which occurs for both the PT and IT subpopulations.

FRONTAL CORTICAL PT NEURONS ARE KEY FOR PSILOCYBIN'S EFFECT ON STRESS-RELATED BEHAVIOR

An important question is whether the frontal cortical cell types are relevant for psilocybin's behavioral effects. To answer this question, we expressed broadly and bilaterally inhibitory DREADDin PT and IT neurons by injecting AAV-hSyn-DIO-hM4DGi-mCherry in adult Fezf2-CreER and PlexinD1-CreER mice (Fig.). These tamoxifen-inducible Cre-driver lines target PT and IT neurons respectively. Control mice were injected with AAV-hSyn-DIO-mCherry. We treated animals with the chemogenetic ligand deschloroclozapine 32 (DCZ; 0.1 mg/kg, i.p.) 15 minutes before injecting psilocybin (1 mg/kg, i.p.) or saline, thereby silencing the respective subsets of pyramidal cells when the drug is active in the brain. We tested four behavioral assays. The head-twitch response is an indicator of hallucinogenic potency of a compound in humansand occurs nearly immediately after the administration of psilocybin in rodents. Psilocybin induced head twitches in our mice as expected, which was not affected by the DREADD-mediated silencing of frontal cortical PT or IT neurons (n = 11-20 mice in each group; Fig.; Extended Data Fig.). Next, learned helplessness is a preclinical paradigm relevant for modeling depression pathophysiology. Mice were exposed to inescapable footshocks during two induction sessions and subsequently tested for avoidance when faced with escapable footshocks during a test session (Fig.). A single dose of psilocybin reduced escape failures, suggesting that drug-treated animals were less affected by the uncontrollable stress (Fig.). This psilocybin-induced relief of the stress-induced phenotype was abolished if frontal cortical PT neurons were silenced during drug administration (interaction effect of treatment and DREADD: P < 0.001, two-factor ANOVA; n = 13-16 mice in each group; Fig.; Extended Data Fig.). Meanwhile, inactivating IT neurons had no effect (n = 11-14 mice in each group; Fig.; Extended Data Fig.). The tail suspension test assesses stress-related escape, in which immobility time serves as an indicator of stress-induced escape behavior (Fig.). Mice treated with psilocybin 24 hours prior to testing showed a significant reduction in immobility time compared to saline-treated animals, an improvement that was likewise abolished specifically by inactivation of frontal cortical PT neurons (interaction effect of treatment and DREADD: P < 0.001, two-factor ANOVA; n = 10-14 mice in each group; Fig.; Extended Data Fig.). Finally, we found that frontal cortical PT neurons are needed for psilocybin-driven facilitation of fear extinction in chronically stressed mice (Extended Data Fig.). Together the behavioral data indicate that PT neurons in the medial frontal cortex are a key part of the brain's circuitry for mediating psilocybin's effect on stress-related behaviors.

PSILOCYBIN ACUTELY ELEVATES DENDRITIC CALCIUM SIGNALING IN PT NEURONS

What are the early events that initiate the psilocybin-induced structural and behavioral adaptations? Calcium is a second messenger that regulates synaptic plasticity in pyramidal cells. There are different plasticity mechanisms that depend on calcium elevations, both globally in dendritic branchesand locally in dendritic spines 36 . To determine whether calcium in dendritic branches and dendritic spines are involved in psilocybin's action, we used twophoton microscopy to image the apical dendrites of pyramidal cells in ACAd/medial MOs of awake, head-fixed mice. We focused on the acute phase of psilocybin action, imaging the same fields of view located at 20 -120 µm below the pial surface for 10 minutes before and within 1 hr after drug injection (Fig.). To visualize calcium transients, we expressed the genetically encoded calcium indicator GCaMP6f in PT or IT neurons by injecting AAVretro-hSyn-Cre in the ipsilateral pons or contralateral striatum respectively, and AAV-CAG-FLEX-GCaMP6f in the medial frontal cortex (Fig.). We used automated proceduresto detect calcium events in regions of interest corresponding to dendritic branches and dendritic spines before and after administering psilocybin (1 mg/kg, i.p.) or saline. For dendritic branches, a single dose of psilocybin increased the rate of spontaneous calcium events in PT neurons (psilocybin: 23±4%; n = 149 branches, 4 mice; saline: 5±2%, n = 140 branches, 4 mice; Fig.; Extended Data Fig.). Conversely, psilocybin did not affect calcium events in dendritic branches of IT neurons (psilocybin: -2±3%; n = 95 branches, 3 mice; saline: 1±3%, n = 90 branches, 3 mice; interaction effect of treatment × cell type: P = 0.008, mixed effects model; Fig.; Extended Data Fig.). For dendritic spines, we analyzed fluorescence signals after subtracting contribution from adjoining dendritic branch using a regression procedureto estimate calcium transients arising from subthreshold synaptic activation. Similar to what we saw for dendritic branches, psilocybin elevated the rate of synaptic calcium events in dendritic spines of PT neurons (psilocybin: 68±5%; n = 2637 spines, 4 mice; saline: 37±6%, n = 2307 spines, 4 mice; Fig.; Extended Data Fig.), but not in IT neurons (psilocybin: 20±3%; n = 2198 spines, 3 mice; saline: 16±2%, n = 2237 spines, 3 mice; interaction effect of treatment × cell type: P < 0.001, mixed-effects model; Fig.Extended Data Fig.). These data show that psilocybin preferentially boosts dendritic and synaptic calcium signaling in PT neurons in the medial frontal cortex.

PSILOCYBIN SELECTIVELY INCREASES FIRING IN A SUBSET OF PT NEURONS

The heightened dendritic calcium signals are likely due to increased dendritic excitability, which can lead to higher spiking activity in PT neurons. Alternatively, it has been shown that some 5-HT1A receptors localize to the axon initial segment, creating a scenario where dendrites can be excitable while firing remains unchanged or suppressed in PT neurons. To disambiguate these possibilities, we used cell-type specific electrophysiology to record from PT and IT neurons in awake, head-fixed mice. To identify the cell type, we expressed channelrhodopsin (ChR2) in PT or IT neurons by injecting AAV-EF1a-double floxed-hChR2(H134R)-EFYP into the medial frontal cortex of adult Fezf2-CreER or PlexinD1-CreER mice (Fig.). We targeted the medial frontal cortex with a high-density Neuropixels probe(Fig.) and isolated single units via spike sorting and quality metrics (Extended Data Fig.). We recorded for 30 minutes, injected psilocybin (1 mg/kg, i.p.) or saline, and then recorded for another 60 minutes. At the end of each recording session, we performed "opto-tagging" by applying trains of brief laser pulses (473 nm, 20 ms) to identify ChR2-expressing cells. The opto-tagged PT and IT neurons were reliably driven by the photostimulation to spike with short latency (Fig.; Extended Data Fig.). A fraction of the opto-tagged PT neurons in Fezf2-CreER mice responded vigorously to psilocybin. Specifically, 16% of the PT neurons substantially increased spiking activity, whereas few cells exhibited decrease after psilocybin or change after saline (psilocybin: 14 cells with post-drug mean z-score Z>2, 2 cells with Z<-2, n = 90 tagged neurons, 5 mice; saline: 2 cell with Z>2 and 3 cell with Z<-2, n = 104 tagged neurons, 6 mice; Fig.). On average, comparing between pre-versus post-drug firing, PT neurons showed significantly higher spike rates after psilocybin (P < 0.001, paired t-test with Bonferroni correction; Fig.). By contrast, there was no notable change in firing activity for IT neurons in PlexinD1-CreER mice after psilocybin administration (psilocybin: 2 cells with Z>2, 0 cells with Z<-2, n = 57 tagged neurons, 5 mice; saline: 2 cells with Z>2 and 0 cells with Z<-2, n = 38 tagged neurons, 5 mice; P = 1.0, paired t-test with Bonferroni correction; Fig.). These results show that psilocybin produces cell type-specific changes in neural dynamics in the medial frontal cortex, highlighted by a set of PT neurons that responded acutely to drug administration by firing vigorously.

-HT2A RECEPTORS IN MEDIAL FRONTAL CORTEX MEDIATE PSILOCYBIN'S ALLEVIATING EFFECT ON STRESSRELATED BEHAVIORS

Our results thus far indicate frontal cortical PT neurons as a target for psilocybin. Does the cell type act through 5-HT2A receptors? Current literature provides conflicting data on whether the 5-HT2A receptor is neededor nonessentialfor the long-term neural and behavioral effects of psychedelics. The discrepancy may stem in part from the use of constitutive knockout animals and antagonist drugs, which can have unwanted effects on neurodevelopment or other receptors. Therefore, here we took a different approach, using a conditional knockout mouse Htr2a f/f for region-and cell-type-targeted deletion of 5-HT2A receptors in adult animals. We first asked if there are 5-HT2A receptors in frontal cortical excitatory cell types. Analysis of Allen Institute's single cell sequencing datarevealed abundant Htr2a transcripts in a proportion of frontal cortical PT and IT neurons (Fig.). Next, we validated Cre-mediated knockout of 5-HT2A receptors in Htr2a f/f mice. Following injection of AAV-CaMKII-GFP-Cre into the medial frontal cortex, at the transcript level, qPCR confirmed the absence of Htr2a transcript in GFP+ cells (control: 2 mice, knockout: 3 mice; Fig.). At the synaptic level, we performed wholecell recordings from GFP+ layer 5 pyramidal cells, which did not exhibit 5-HT-evoked increase in sEPSCs (control: 22 cells from 4 mice, knockout: 23 cells from 4 mice; Fig.; Extended Data Fig.), a 5-HT2A receptor-dependent phenomenon. Leveraging the Htr2a f/f mice, we asked if the 5-HT2A receptor in frontal cortical neurons are needed for psilocybin's effects in the same set of behaviors tested for Fig.. We injected either AAV-hSyn-Cre-P2A-Tomato or AAV-hSyn-EGFP bilaterally and broadly in the medial frontal cortex of Htr2a f/f mice. Animals with the localized knockout of 5-HT2A receptors exhibited the same amount of psilocybin-evoked head-twitch response as controls (n = 6-9 mice in each group; Fig.). The lack of dependence on 5-HT2A receptor for the psilocybin-evoked head-twitch response was specific to local manipulation in the medial frontal cortex, because CaMKII Cre ;Htr2a f/f mice with constitutive and more widespread receptor knockout had markedly fewer head-twitch response than control animals after psilocybin administration (Extended Data Fig.). Notably, the region-specific 5-HT2A receptor knockout was sufficient to render psilocybin ineffective for ameliorating the stress-related phenotypes in learned helplessness (n = 8-13 mice in each group; Fig.) and tail suspension test (n = 8-9 mice in each group; Fig.). We note the caveat that although results from head-twitch response and tail suspension test were clearly interpretable, the response of control animals to psilocybin in learned helplessness did not reach statistical significance, likely due to floor effect from the low baseline rate of escape failures in this Htr2a f/f strain. Collectively, the data show the importance of 5-HT2A receptors in the medial frontal cortex for psilocybin's ameliorative effects on stress-related behavior.

-HT2A RECEPTOR IS REQUIRED FOR PSILOCYBIN-INDUCED STRUCTURAL PLASTICITY IN PT NEURONS

Is the 5-HT2A receptor needed for psilocybin-evoked dendritic remodeling? To answer this question, we performed targeted knockout of 5-HT2A receptors by injecting low titer of AAVretro-hSyn-Cre into the ipsilateral pons and AAV-CAG-FLEX-EGFP into the medial frontal cortex of Htr2a f/f (Fig.). In this viral strategy, the Cre recombinase was needed for dual purposes to express EGFP for visualization and to mediate knockout, therefore the control animals need the same viruses, which are injected into wild type C57BL/6J mice. We used two-photon microscopy to image the same apical tuft dendrites for 4 sessions including before and after treatment with psilocybin (1 mg/kg, i.p.) or saline (Fig.). For each condition (genotype and psilocybin or saline), we tracked and analyzed 445-1008 spines from 31-68 dendrites in 5-7 mice of both sexes. In agreement with our earlier findings, the frontal cortical PT neurons in control animals exhibited increased spine density following a single dose of psilocybin (spine density: 15±2% for psilocybin, -2±2% for saline on day 1). By contrast, the cell type-targeted 5-HT2A receptor knockout abolished psilocybin's effects (spine density: -1±2% for psilocybin, 1±2% for saline on day 1; interaction effect of treatment × genotype: P = 0.01 for spine density, mixed effects model; Fig.; Extended Data Fig.). In vivo two-photon microscopy has spatial resolution close to the limit needed for measuring spine size, motivating us to perform post hoc confocal microscopy in fixed tissues from the same animals to determine psilocybin's impact on spine morphology (n = 68-136 dendrites from 3-7 mice in each group; Fig.). Extracted from day 3 after psilocybin dosing, the confocal data showed a psilocybin-evoked increase of spine head width in apical tufts of frontal cortical PT neurons (0.53±0.01 μm for psilocybin, 0.50±0.01 μm for saline), an effect that was absent when 5-HT2A receptors were selectively deleted (0.50±0.01 μm for psilocybin, 0.51±0.01 μm for saline; interaction effect of treatment × genotype: P = 0.025, two-factor ANOVA; Fig.). These data strongly point to the necessity of 5-HT2A receptors for psilocybin-induced structural neural plasticity..

DISCUSSION

We demonstrate that psilocybin's long-term behavioral effects are dissociated at the level of pyramidal cell types in the frontal cortex. The cell-type specific dissociation may be a mechanism leveraged by novel psychedelic analogs to isolate therapeutic effects from hallucinogenic action. A key finding is that frontal cortical PT neurons are essential for psilocybin's beneficial effects in stress-related phenotypes. The consequence for the structural plasticity in frontal cortical IT neurons is unclear; it may be an epiphenomenon, or the IT cell type may mediate other psilocybin-induced behavioral changes that were not tested in this study. Our results emphasize the importance of 5-HT2A receptors for psilocybin's long-term effects. However, given that many PT and IT neurons in the frontal cortex have abundant Htr2a transcripts, the expression profile cannot fully explain why PT neurons respond preferentially to psilocybin. It is plausible that under in vivo conditions, circuit mechanisms steer psilocybin's action to favor PT neurons. For instance, psilocybin may heighten activity of certain long-range axonal inputs with biased connectivity to frontal cortical PT neurons, such as those from contralateral medial frontal cortexand ventromedial thalamus. Another possibility is that psilocybin may cause disinhibition by suppressing specific GABAergic neurons, such as deeplying somatostatin-expressing interneurons that preferentially inhibit PT neurons. Receptor and circuit mechanisms are not mutually exclusive and their relative contributions to psilocybin's impact on frontal cortical neural dynamics should be determined in future studies. A hallmark of psychedelics is their ability to alter conscious perception. Layer 5 pyramidal cells, including specifically the PT neuron subpopulation, have been implicated in the transition from anesthesia to wakefulness. In the medial frontal cortex, PT neurons represent the subcortical output pathway, sending axons to ipsilateral thalamus and other deep-lying brain regions. There is growing interest to develop new treatments for depression that pair antidepressants with other approaches, such as electroconvulsive or transcranial magnetic stimulation, with a goal to augment neural plasticity and enhance therapeutic outcome. This study delineates the cell types and receptors that underpin psychedelic action, highlighting the neural circuits that may be promising targets for neuromodulation and precision treatment.

ANIMALS

Wild-type C57BL/6J (Stock No. 000664), Fezf2-2A-CreER 1 (B6;129S4-Fezf2 tm1.1(cre/ERT2)Zjh /J, Stock No. 036296), PlexinD1-2A-CreER 1 (B6;129S4-Plxnd1 tm1.1(cre/ERT2)Zjh/J , Stock No. 036296), Thy1 GFP line M 2 (Tg(Thy1-EGFP)MJrs/J, Stock No. 007788), and CaMKIIa Cre (B6.Cg-Tg(Camk2a-cre)T29-1Stl/J, Stock No. 005359) mice were from Jackson Laboratory and bred in our animal facility. Htr2a f/f mice were described in a previous studyand bred in our animal facility. For behavioral and electrophysiological studies involving PT and IT neurons, 5-to 8-week-old homozygous Fezf2-2A-CreER and PlexinD1-2A-CreER mice were used for viral injection, then tested 2 weeks later. For two-photon imaging studies, 5 to 7-week-old C57BL/6J or homozygous Htr2a f/f mice were used for viral injection, then implanted with a glass window and imaged 2-3 weeks later. For validation and behavioral studies involving the Htr2a f/f mice, 5 to 8-week-old homozygous Htr2a f/f mice or littermate controls were used for viral injection, then tested 3 weeks later. Animals were housed in groups with 2 -5 mice per cage in a temperature-controlled room, operating on a normal 12 hr light -12 hr dark cycle (8:00 AM to 8:00 PM for light). Food and water were available ad libitum. Animals were randomly assigned to different experimental groups. Animal care and experimental procedures were approved by the Institutional Animal. All viruses had titers ≥ 7×10 12 vg/mL. The viruses were stored at -80°C. Before stereotaxic injection, they were taken out of the -80°C freezer, thawed on ice, and diluted to the corresponding titer for injection.

SURGERY

Prior to surgery, each mouse was injected with dexamethasone (3 mg/kg, i.m.; DexaJect, #002459, Henry Schein Animal Health) and carprofen (5 mg/kg, s.c.; #024751, Henry Schein Animal Health) for anti-inflammatory and analgesic purposes. At the start of surgery, anesthesia was induced with 2 -3% isoflurane and the mouse was affixed in a stereotaxic apparatus (Model 900, David Kopf Instruments). Anesthesia was maintained with 1 -1.5% isoflurane. Body temperature was maintained at 38°C using a far-infrared warming pad (#RT-0515, Kent Scientific). Petrolatum ophthalmic ointment (#IS4398, Dechra) was applied to cover the eyes. The hair on the head was shaved. The scalp was disinfected by wiping with ethanol pads and povidone-iodine. Small burr holes were made above the targeted brain regions using a handheld dental drill (#HP4-917, Foredom). Adeno-associated virus (AAV) was delivered intracranially into the brain by inserting a borosilicate glass capillary and using an injector (Nanoject II Auto-Nanoliter Injector, Drummond Scientific). Injections were done for the various experiments using different viruses and volumes, as specified in the paragraphs below, using 4.6 nL pulses with 20 s interval between each pulse. To reduce backflow of the virus, we waited 5-10 min after completing an injection at one site before retracting the pipette to move on to the next site. For the medial frontal cortex and striatum, the stereotaxic apparatus was positioned at four sites corresponding to four vertices of a 0.2 mm-wide square centered at the coordinates mentioned below. Throughout the procedure, the brain surface was kept moist with artificial cerebrospinal fluid (aCSF; in mM: 135 NaCl, 5 HEPES, 5 KCl, 1.8 CaCl2, 1 MgCl2; pH: 7.3). After injections, the craniotomies were covered with silicone elastomer (#0318, Smooth-On, Inc.), and the skin was sutured (#1265B, Surgical Specialties Corporation). At the end of surgery, animal was given carprofen (5 mg/kg, s.c.) immediately and then again once on each of the following 3 days. respectively. After 2 weeks, the mouse would undergo a second procedure. An incision was made to remove the skin and the periosteum was cleared. A dental drill was used to make a 0.9 mm craniotomy and a 0.86 mm self-tapping bone screw (#19010-10, Fine Science Tools) was placed through the skull bone into the cerebellum to act as a ground screw and provide further structural support for head-fixation. A custom stainless steel headplate was affixed on the skull using a quick adhesive cement system. The mouse would recover for at least one-week post-surgery prior to commencement of electrophysiological experiments. In both cases, injection of AAVs were made prior to administration of tamoxifen. For validation of the Htr2a f/f mouse line, homozygous Htr2a f/f animals were bilaterally injected with AAV9-CaMKII-HI-GFP-Cre.WPRE.SV40 (441.6 nL, 1:10 diluted in PBS) or AAV8-CaMKIIa-EGFP (441.6 nL, 1:10 diluted in PBS) in the medial frontal cortex (AP: 1.5 mm, ML: +/-0.4 mm, DV: -0.4, -0.6, and -1.2 mm, relative to dura). Incised skin was sutured. Animals would recover for at least 3 weeks prior to sacrifice for transcript or slice electrophysiology experiments. For behavioral experiments involving Htr2a f/f mice, homozygous Htr2a f/f animals were bilaterally injected with AAV9-hSyn-Cre-P2A-Tomato (690 nL, 1:50 diluted in PBS) or AAV9-hSyn-EGFP (690 nL, 1:50 diluted in PBS) in the medial frontal cortex (AP: 1.5 mm, ML: -0.4 mm, DV: -0.4, -0.6, and -1.2 mm, relative to dura). Incised skin was sutured. Animals would recover for at least 3 weeks prior to behavioral experiments.

TAMOXIFEN

Tamoxifen was used for inducible Cre-dependent gene expression in Fezf2-CreER and PlexinD1-CreER mice. Tamoxifen (#T5648, Sigma-Aldrich) was dissolved in corn oil (#C8267, Sigma-Aldrich) at concentration of 20 mg/mL in an ultrasonic bath at 37°C for 1 -4 hr. The solution was then aliquoted into 1 mL tubes, wrapped with aluminum foil, and stored at -20°C. For injections, the tamoxifen aliquots were thawed at 4°C. Each animal was weighed and received tamoxifen (75 mg/kg, i.p.) once every 24 hours for 5 consecutive days. Experiments involving inducible Cre expression were conducted at least 2 weeks after the last dose of tamoxifen to allow time for viral-mediated expression.

HISTOLOGY

Histology was performed to determine the accuracy of injection locations and assess transgene expression. For two-photon imaging and behavioral studies, after completion of experiments, mice were perfused with PBS, followed by paraformaldehyde solution (PFA, 4% (v/v) in PBS). The brains were extracted and further fixed in 4% PFA at 4°C for 12 -24 hr. Subsequently, 40-µm-thick coronal sections were obtained using a vibratome (#VT1000S, Leica) and mounted on slides with glass coverslips. Sections were imaged using a wide-field fluorescence microscope (BZ-X810, Keyence). For electrophysiology, the coronal sections were prepared similarly, mounted on slides using Vectashield containing DAPI (#H-1200-10, Vector Laboratories) and imaged. To locate the Neuropixels probe, we used SHARP-TRACK 5 to align the images of the coronal sections including the DiI tracks with the standardized Allen Common Coordinate Framework 6. Reconstructed probe tracks were visualized within the Allen Common Coordinate Framework using Brainrender 7 .

TWO-PHOTON IMAGING

Two-photon imaging experiments were performed using a Movable Objective Microscope (MOM, Sutter Instrument) equipped with a resonant-galvo scanner (Rapid Multi Region Scanner, Vidrio Technologies) and a water-immersion 20X objective (XLUMPLFLN, 20x/0.95 N.A., Olympus). ScanImage 2020 software 8 was used to control the microscope for image acquisition. To visualize GFP or GCaMP6f-expressing dendrites, a tunable Ti:Sapphire femtosecond laser (Chameleon Ultra II, Coherent) was used as the excitation source. The excitation wavelength was set at 920 nm, and emission was collected behind a 475 -550 nm bandpass filter for fluorescence from GFP or GCaMP6f. The laser power measured at the objective was typically ≤40 mW and varied depending on the imaging depth. When imaging of the same field of view across days, the laser power was kept the same in each imaging session. For structural imaging of dendrites, in each imaging session, the mouse was head fixed and anesthetized with 1% isoflurane through a nose cone. Body temperature was maintained at 37.4°C via a heating pad system (#40-90-8D, FHC) with feedback control from a rectal thermistor probe. Each imaging session lasted 0.5 -1.5 hr. To target the ACAd and medial MOs subregion of the medial prefrontal cortex, we imaged within 400 µm of the midline as determined by first visualizing the sagittal sinus in bright-field imaging. To target apical tuft dendrites, we first imaged 0 -200 µm below the pial surface to identify the apical tuft dendrites and apical trunk, and then select apical tuft dendrites located between 20 -120 µm below the pial surface for longitudinal imaging. Multiple different fields of view were imaged in the same mouse. For each field of view, 10 -40-µm-thick z-stacks were collected with 1 µm steps using 15 Hz bidirectional scanning at 1024 × 1024 pixels with a resolution of 0.11 µm per pixel. Each mouse was imaged at the same fields of view on day -3, -1, 1, 3, 5, 7, 35 and 65 relative to the day of drug administration. On the day of treatment (day 0), no imaging was performed, and the mouse was injected while awake with either psilocybin (1 mg/kg, i.p.; prepared from working solution, which was made fresh monthly from powder; Usona Institute) or saline (10 mL/kg, i.p.). After injection, the mouse was placed in a clean cage, and head twitches were visually inspected for 10 min before returning the mice to their home cage. At the end of imaging session, for the purpose of reconstructing the apical dendritic trees, a z-stack was acquired between 0 -900 µm below the dura with 2 µm steps. For structural imaging of dendrites, 148 dendrites from 17 C57BL/6J mice were imaged for psilocybin (8 males including 5 for PT and 3 for IT neurons; 9 females including 4 for PT and 5 for IT neurons), and 154 dendrites from 16 C57BL/6J mice were imaged for saline (7 males including 4 for PT and 3 for IT neurons; 9 females including 4 for PT and 5 for IT neurons). For structural imaging to test effects of 5-HT2A receptor knockout on dendrites, 117 dendrites from 11 mice were imaged for psilocybin (6 Htr2a f/f mice; 5 C57BL/6J mice), and 80 dendrites from 12 mice were imaged for saline (5 Htr2a f/f mice; 7 C57BL/6J mice). For structural imaging of Thy1 GFP mice, 38 dendrites from 2 Thy1 GFP ;Htr2a WT/WT mice were imaged for psilocybin, 49 dendrites from 5 Thy1 GFP ;Htr2a f/f mice were imaged for psilocybin, and 98 dendrites from 3 Thy1 GFP ;Htr2a f/f mice were imaged for saline. For calcium imaging of dendrites, the mouse was habituated to head-fixation in an acrylic tube under the microscope for 3-4 days, with increasing durations each day, before the day of data collection. To examine the acute effects of psilocybin, we imaged 2 fields of view, each for 10 min to obtain pre-treatment baseline data. Imaging was then paused to inject psilocybin (1 mg/kg, i.p.) or saline (10 mL/kg, i.p.). At 30 min after injection, we imaged those same 2 fields of view again, each for 10 min to acquire posttreatment data. Each animal received both psilocybin and saline, with at least 1 week between imaging sessions and the order of treatment was balanced across subjects. For calcium imaging of dendrites, 8 C57BL/6J mice including 3 males and 5 females were treated with psilocybin (244 dendritic branches including 149 from PT and 95 from IT neurons, with 4835 dendritic spines including 2637 from PT and 2198 from IT neurons) and saline (230 dendritic branches including 140 from PT and 90 from IT neurons, with 4544 dendritic spines including 2307 from PT and 2237 from IT neurons).

ANALYSIS OF THE IMAGING DATA

For structural imaging of dendrites, motion correction was performed using StackReg plug-inin ImageJ. Quantification of structural parameters such as spine head width and spine protrusion length were done according to standardized critera. In brief, a dendritic spine was counted when the protrusion extended for >0.4 µm from the dendritic shaft. The line segment tool in ImageJ was utilized to measure the distances. The spine head width was determined as the width of the widest part of the spine head. Dendritic spine protrusion length referred to the distance from the tip of the head to the base at the shaft. Alterations in spine density, spine head width, and spine protrusion length were calculated as fold change compared to the value measured for each dendritic segment on the first imaging session (day -3). The raw values for spine density, spine head width, and spine protrusion length are provided in Extended Data. Spine formation rate was calculated by determining the number of newly formed dendritic spines between two consecutive imaging sessions (i.e., day -3 and day -1) divided by the total number of dendritic spines counted in the preceding imaging session (i.e., day -3). Similarly, spine elimination rate was calculated by determining the number of missing dendritic spines between two consecutive imaging sessions divided by the total number of dendritic spines counted in the preceding imaging session. To assess the longitudinal alterations in spine formation and elimination rates, we calculated the difference of the spine formation or elimination rate from the baseline rate, which was the spine formation or elimination rate for same dendritic segment before psilocybin and saline injection (between day -3 to day -1). The raw values for spine formation and elimination rates are provided in Extended Data. To divide IT neurons based on laminar position, we treated those with cell bodies residing in depth between 200 -400 µm below the dura as layer 2/3, while those with cell bodies residing in depth between 450 µm to 650 µm as layer 5. For calcium imaging of dendrites, multi-page .tiff image files from one experiment were concatenated and processed with NoRMCorre 13 in MATLAB to correct for non-rigid translational motion. As an overview, processing involved: (1) Regions of interest (ROI) corresponding to dendritic branches and spines were manually traced using an in-house graphical user interface in MATLAB; (2) The average fluorescence trace from each ROI was then processed similar to prior workto exclude background neuropil signal, and converted to fractional change in fluorescence (ΔF/F(t)); (3) deconvolve the fluorescence trace into discrete calcium events. Details for each of these processing steps are described below. Dendritic branch and spine ROIs were manually traced by scrolling through the imaging frames to find putative dendritic segments (i.e., neurite segments with > 10 spiny protrusions showing a correlated pattern of fluorescence transients). First, a given branch ROI would be traced around the dendritic shaft segment using a lasso drawing tool. Next, the putative dendritic spines for that branch segment were captured using a circle drawing tool (typically 0.8 -1.2 µm diameter ROIs). For each ROI, the pixel-wise average was calculated at each data frame to generate a fluorescence time course FROI(t). Since calcium imaging was performed on the same field of view before and after drug injections, a single ROI mask was used to extract calcium signals before and after treatment. All ROI selection was done while blinded to treatment group. Each ROI was then processed to reduce the contribution from background neuropil. Taking each ROI's area and considering a circle with equivalent area that has radius, r, an ROI-specific neuropil mask was created as an annulus with inner radius 2r and outer radius 3r centered on the centroid of the ROI. Neuropil masks excluded pixels belonging to any other dendritic branch or spine ROI. To exclude neuropil mask pixels that may belong to unselected dendritic structures, we calculated the time-average signal for each pixel, taking the median amongst pixels in the mask. Pixels were excluded from the neuropil mask if their time-averaged signal was higher than the median. Finally, the remaining pixels in the neuropil mask were averaged per data frame to generate Fneuropil(t). Each ROI had the fluorescence from its neuropil mask subtracted as follows: where the neuropil correction factor, c, was set to 0.4. Next, the fractional change in fluorescence ΔF/F(t) was calculated for each ROI by normalizing F(t) against its baseline, F0(t), estimated as the 10 th percentile within a two-minute sliding window: For each dendritic spine's ΔF/Fspine(t), we estimated the branch-independent spine activity, ΔF/Fsynaptic(t), by subtracting a scaled version of the fluorescence from the corresponding dendritic branch, ΔF/Fbranch(t), as follows: where the branch scaling factor, α, was computed in an ROI-specific manner using a linear regression of ΔF/Fsynaptic(t) predicted by ΔF/Fbranch(t) forced through the origin. In a previous study, we have calibration to show that with this analysis approach, the majority of the spontaneously occurring calcium transients in dendritic spines can be attributed to synaptic activation. Calcium events were detected using automated procedure for each ΔF/Fspine(t) and ΔF/Fbranch(t) using a deconvolution "peeling" algorithm. The peeling algorithm uses an iterative template-matching procedure to decompose a ΔF/F(t) trace into a series of elementary calcium events. The template for elementary calcium events was set to have an instantaneous onset, an amplitude of 0.3, and a single-exponential decay time constant of 1 s. Briefly, the algorithm searches a given ΔF/F(t) trace for a match to the template calcium event, subtracts it from the trace (i.e., "peeling"), and successively repeats the matching process until no events are found. This event detection process outputs the recorded event times with a temporal resolution by the original imaging frame rate. In this way, it is possible to detect multiple calcium events during the same imaging frame (e.g., for large amplitude transients). For each imaging session, an ROI's calcium event rate was computed by dividing the number of calcium events by the duration of the imaging session. The calcium events were examined further by their binned amplitude (average number of calcium events per frame, among frames with at least one event detected) and frequency (number of imaging frames with at least one event, divided by the total imaging duration). The change in calcium event rate, amplitude, and frequency across treatment injections was computed for each ROI using the post-injection minus pre-injection values divided by the pre-injection values and provided raw values for calcium event rates averaged across dendritic branches in the same field of view. Separately, we have tried analyzing the ΔF/Fspine(t) and ΔF/Fbranch(t) using a different calcium event detection algorithm OASIS 17 , which yielded qualitatively similar results (data not shown).

CONFOCAL IMAGING

After longitudinal two-photon imaging and at 3 days after psilocybin (1 mg/kg, i.p.) or saline (10 mL/kg, i.p.) injection, the mouse was deeply anesthetized with isoflurane and transcardially perfused with PBS followed by paraformaldehyde (PFA, 4% in PBS). The brains were fixed in 4% PFA for 24 hr at 4°C, and then 50-um-thick coronal brain slices were sectioned using a vibratome (VT1000S, Leica) and placed on slides with coverslip with mounting medium (Vector Laboratories #H-1500-10). The brain slices were imaged with a confocal microscope (LSM 710, Zeiss) equipped with a Plan-Apochromat 63x/1.40 N.A. oil objective (zoom 2.5) and 0.37 µm steps at 1024 × 1024 pixels with a resolution of 0.08 µm per pixel to collect the structural imaging data. In total, 204 dendrites from 10 mice were imaged for psilocybin (5 Htr2a f/f mice; 5 C57BL/6J mice), and 207 dendrites from 10 mice were imaged for saline (3 Htr2a f/f mice; 7 C57BL/6J mice).

OVERVIEW OF BEHAVIORAL STUDIES

All behavioral assays were conducted between 10:00 AM and 4:00 PM. For the animals uesd in chemogenetic manipulation, the same mice were tested on all assays. At least 2 weeks were allotted between stress-related assays. Mice were randomized into different groupings for each assay (i.e., the same mouse could be part of the psilocybin group on first assay, and then saline group on the second assay). For studies involving PT neurons, Fezf2-2A-CreER mice were tested on fear extinction, then after the last extinction session 2-3 weeks later on learned helplessness, then 1-2 weeks later on head-twitch response, and finally 3 weeks later on tail suspension. We started, for fear extinction, with 58 Fezf2-2A-CreER mice injected with DREADD or control viruses, including 17 mice for psilocybin:mCherry (9 males, 8 females), 13 mice for saline:mCherry (7 males, 6 females), 13 mice for psilocybin:hM4DGi (8 males, 5 females), and 15 mice for saline:hM4DGi (7 males, 8 females). For learned helplessness, we had 57 Fezf2-2A-CreER mice remaining, including 13 mice for psilocybin:mCherry (8 males, 5 females), 15 mice for saline:mCherry (8 males, 7 females), 16 mice for psilocybin:hM4DGi (9 males, 7 females), and 13 mice for saline:hM4DGi (5 males, 8 females). For head-twitch response, we had 53 Fezf2-2A-CreER mice remaining, 9 mice were tested on both psilocybin and saline with 1-week interval while the rest received psilocybin or saline, including 14 mice for psilocybin:mCherry (6 males, 8 females), 13 mice for saline:mCherry (6 males, 7 females), 20 mice for psilocybin:hM4DGi (12 males, 8 females), and 15 mice for saline:hM4DGi (7 males, 8 females). For tail suspension test, we had 49 Fezf2-2A-CreER mice remaining, including 14 mice for psilocybin:mCherry (7 males, 7 females), 10 mice for saline:mCherry (5 males, 5 females), 13 mice for psilocybin:hM4DGi (6 males, 7 females), and 12 mice for saline:hM4DGi (8 males, 4 females). For studies involving IT neurons, PlexinD1-2A-CreER mice were tested on learned helplessness, then 1-2 weeks later on head-twitch response, and finally 3 weeks later on tail suspension. We started, for learned helplessness, with 47 PlexinD1-2A-CreER mice injected with DREADD or control viruses, including 14 mice for psilocybin:mCherry (7 males, 7 females), 11 mice for saline:mCherry (5 males and 6 females), 11 mice for psilocybin:hM4DGi (6 males, 5 females), and 11 mice for saline:hM4DGi (5 males, 6 females). For head-twitch response, we had 47 PlexinD1-2A-CreER mice remaining, 4 mice were tested on both psilocybin and saline with 1-week interval while the rest received psilocybin or saline, including 15 mice for psilocybin:mCherry (8 males, 7 females), 12 mice for saline:mCherry (6 males, 6 females), 11 mice for psilocybin:hM4DGi (6 males, 5 females), and 13 mice for saline:hM4DGi (7 males, 6 females). For tail suspension test, we had 41 PlexinD1-2A-CreER mice remaining, including 14 mice for psilocybin:mCherry (7 males, 7 females), 9 mice for saline:mCherry (5 males, 4 females), 9 mice for psilocybin:hM4DGi (5 males, 4 females), and 9 mice for saline:hM4DGi (5 males, 4 females). For behavioral studies involving Htr2a f/f mice, separate groups of mice were used for each behavioral test. For learned helplessness, 16 local 5-HT2A receptor knockout mice were tested: 8 mice with saline (4 males, 4 females) and 8 with psilocybin (4 males, 4 females). 24 littermates injected with control virus were tested: 13 with saline (6 males,

HEAD-TWITCH RESPONSE

For each mouse, deschloroclozapine (DCZ; 0.1 mg/kg, i.p.; #HY-42110, MedChemExpress) or saline (10 mL/kg, i.p.) was injected if chemogenetic manipulation was tested, and then psilocybin (1 mg/kg, i.p.) or saline (10 mL/kg, i.p.) was injected 15 min later. Head-twitch response was measured in groups of 2-3 mice, typically with psilocybin-and saline-treated mice tested simultaneously. After the injection, each mouse was immediately placed into its own plexiglass chamber (4'' x 4'' x 4''), which had a transparent lid and was positioned within a sound attenuating cubicle (Med Associates). A high-speed video camera (acA1920, Basler) was mounted overhead above the chambers. We recorded videos for 10 min. Between measurements, the chambers were thoroughly cleaned with 70% ethanol. The videos were scored for head twitches by a different experimenter blinded to the experimental conditions. Previously we showed that head twitches can be quantified using magnetic ear tags, however here we were concerned that the ear tag might interfere with performance in other behavioral assays so opted for video recording.

LEARNED HELPLESSNESS

For learned helplessness, we performed the assay using an active avoidance box with a stainless-steel grid floor and a shuttle box auto door separating the two compartments (8'' x 8'' x 6.29'') inside a sound attenuating cubicle (MED-APA-D1M, Med Associates). On day 1 and day 2, there was one induction session on each day. Each session consisted of 360 inescapable foot shocks delivered at 0.2 mA for 1 -3 s, with a random inter-trial interval ranging from 1 to 15 s. At 10-15 min after the end of the second induction session, DCZ (0.1 mg/kg, i.p.) or saline (10 mL/kg, i.p.) was given (the animals used in chemogenetic manipulation), and then psilocybin (1 mg/kg, i.p.) or saline (10 mL/kg, i.p.) was injected 15 min later. On day 3, one test session was conducted, consisting of 30 escapable foot shocks delivered at 0.2 mA for 10 s, with an inter-trial interval of 30 s. A shock would be terminated early if the mouse moved to the other compartment. Movement of the mouse was captured by beam breaks in the shuttle box. A failure was counted when the mouse failed to escape before the end of a shock. After each induction or testing session, the shuttle box was cleaned with 70% ethanol. Before each testing session, the shuttle box was cleaned with 1% acetic acid solution to provide a different olfactory context.

TAIL SUSPENSION TEST

Animals were tested 24 hours after administration of psilocybin (0.1 mg/kg i.p) or saline (10 mL/kg, i.p.). For chemogenetic manipulation, DCZ (0.1 mg/kg, i.p.) or saline (10 mL/kg, i.p.) was given 15 min before psilocybin or saline administration. Within a tall sound-attenuating cubicle (Med Associates), the setup included a metal bar elevated 30 cm from the floor. An animal was suspended from the metal bar by securing its tail to the bar using removable tape (NC9972972, Fisher Scientific). A small plastic tube was placed around the base of the tail to prevent tail climbing during the session. Videos of the suspended animals were recorded for 6 minutes. The behavioral apparatus was thoroughly cleaned with 70% ethanol before and after each session.

STRESS-INDUCED RESISTANCE TO FEAR EXTINCTION

For chronic restraint stress, we based the procedures on a published study. Mice were restrained inside a cone-shaped plastic bag with openings on both ends (Decapicone, MDC200, Braintree Scientific) for 3 hours each day for 14 consecutive days. The opening corresponding to the rear of the mouse was sealed by tying a wire, leaving the mouse's tail protruding. Restrained animals were secured in an upright position inside an empty cage and monitored frequently. At 24 hr after the end of last restraint session, we began fear conditioning and extinction procedures, which were performed using a near-infrared video fear conditioning system (MED-VFC2-SCT-M, Med Associates). Prior to each session, the mouse was brought to the behavior room for habituation for ~30 min. The fear conditioning system was equipped with stainless-steel grid floor and was controlled by the VideoFreeze software (Med Associates). On day 1 (fear conditioning), the chamber had blank straight walls and stainless-steel grid floor. Surfaces of the chamber were cleaned with 70% ethanol (context A). Each mouse was conditioned individually in a chamber and given 3 minutes to habituate. Subsequently, it received 5 presentations of an auditory tone as the conditioned stimulus (CS; 4 kHz, 80 dB, 30 s duration). Each CS co-terminated with a footshock as the unconditioned stimulus (US; 0.8 mA, 2 s duration). A 90-s intertrial interval separated the CS + US pairings. On day 3 (fear extinction 1), for each mouse, DCZ (0.1 mg/kg, i.p.) or saline (10 mL/kg, i.p.) was injected, and then psilocybin (1 mg/kg, i.p.) or saline (10 mL/kg, i.p.) was injected 15 min later. Then 45 min later, we started test for fear extinction, while the drug is presumably still present in the brain. The chamber had two black IRT acrylic sheets inserted for a sloped roof and stainless-steel grid floor covered with a white smooth floor. Surfaces of the chamber were cleaned with 1% acetic acid (context B). Each mouse was tested individually in a chamber and given 3 minutes to habituate. Subsequently, it received 15 presentations of the CS without the US. A 15-s intertrial interval separated the CS presentations. On day 4 (retention 1), we repeated the test for fear extinction in context B. On day 17 (retention 2), we repeated the test for fear extinction in context B.

IN VIVO ELECTROPHYSIOLOGY

Mice were habituated to head fixation with increasing duration over several days. At least 3 hr before recording, mice were anesthetized with isoflurane and a 2-mmdiameter craniotomy was made over the medial frontal cortex (AP: 1.7 mm, ML: 0.5 mm). Cold (4°C) aCSF was used to irrigate to clear debris and reduce heating during drilling. Care was taken to minimize bleeding and keep the area clear of bone fragments. The dura was removed using a metal pin (#10130-10, Fine Science Tools). A piece of Surgifoam (#1972, Johnson & Johnson) soaked in aCSF was placed above the brain tissue, which was covered with silicon polymer (#0318, Smooth-On, Inc.) to keep the craniotomy moist and clean prior to recording. For drug administration, to avoid inserting a needle during recording session which we found to cause animal to move and therefore compromise recording stability, we used a catheter system described previously. A 22-gauge intravenous catheter system (#B383323, BD Saf-T-Intima Closed IV Catheter Systems) was preloaded with psilocybin or saline and maintained at a neutral pressure. At 1 hr prior to recording, mice were briefly anesthetized with isoflurane and implanted with the intravenous catheter to their intraperitoneal cavity and the catheter was fixed with a drop of Vetbond tissue adhesive (#1469, 3M Vetbond). The mice were then head fixed and the catheter tubing was secured to the mouse holder acrylic tube with tape. Silicon polymer and Surgifoam were removed from the skull and the craniotomy was briefly irrigated with aCSF. A high-density silicon probe (#Neuropixels 1.0, IMEC) with the ground and reference shorted was coated using a 10

ANALYSIS OF IN VIVO ELECTROPHYSIOLOGY DATA

SpikeInterface 21 was used to preprocess, spike sort, and calculate single-unit metrics. Putative single units were initially identified by Kilosort 2.5and were further manually curated in Phy (). Quality and waveform metrics were generated via SpikeInterface. We included units that satisfied all the following quality metrics: present for at least 90% of the recording (presence ratio), ISI violation rate less than 0.5, and amplitude cutoff of less than 0.1. To identify opto-tagged neurons, we created peri-stimulus time histograms by aligning putative single-unit spiking activity to the onset of laser stimulation. We classified opto-tagged neurons via visual inspection considering the latency to spike and reliability of spiking in response to onset of laser stimulation.

ANALYSIS OF SINGLE-CELL TRANSCRIPTOMICS DATA

We accessed the "whole cortex and hippocampus 2020" SmartSeq single cell RNAseq data set made publicly available by the Allen Institute

SLICE ELECTROPHYSIOLOGY

Brain slices were prepared as previously described. Briefly, mice were first anesthetized with chloral hydrate (400 mg/kg, i.p.). After decapitation, brains were

STATISTICS

Supplementary Tableprovides detailed information about the sample sizes and statistical analyses for each experiment. For behavioral studies and confocal imaging, statistical analyses were performed with GraphPad Prism 10. For two-photon imaging experiments, statistical analyses were performed based on mixed effects models using the lme4 package in R. Linear mixed effects models were used to account for repeated measures and within-subject nesting (e.g., multiple spines per branch) in a manner that makes less assumptions about underlying data than the commonly used repeated measures analysis of variance. Details about the models are described below. For two-photon imaging of dendritic structure, analyses were performed while blind to for treatment (psilocybin vs. saline), genotype (Htr2a f/f vs. wild-type), and time (day 1 and day 3) as factors, in addition to all second and higher-order interactions amongst these terms. For two-photon imaging involving Thy1 GFP ; Htr2a f/f mice, two-factor ANOVA was used for the analyses of spine density to test the interaction between treatment (psilocybin vs. saline) and conditions (Thy1 GFP ; Htr2a +/+ :psilocybin vs. Thy1 GFP ; Htr2a f/f :psilocybin vs. Thy1 GFP ; Htr2a f/f :saline) and time (day 1 to day 7). Post hoc t-tests were used to compare Thy1 GFP ; Htr2a +/+ :psilocybin versus Thy1 GFP ; Htr2a f/f :psilocybin, Thy1 GFP ; Htr2a f/f :psilocybin versus Thy1 GFP ; Htr2a f/f :saline, or Thy1 GFP ; Htr2a +/+ :psilocybin and Thy1 GFP ; Htr2a f/f :saline. Bonferroni correction was used for multiple comparisons. For imaging of dendritic calcium signals, blinding procedures were implemented by having one person performed the imaging and scrambled the group names, while another person analyzed the data blind to treatment and cell type information. Data were unblinded after all the analyses were completed. A similar linear mixed effects modeling approach was used to examine three dependent variables: calcium event rate, amplitude, and frequency. Dendritic branch and spine (branch-independent) signals were analyzed in separate models (i.e., six models total). Each model included fixed effects terms for treatment (psilocybin vs. saline), cell type (PT vs. IT), and the interaction term for treatment x cell type. Treatment order (psilocybin before saline vs. psilocybin after saline) was included in the model as a nuisance variable. The variation for repeated measures of mice and dendrites were accounted for by including a random intercept for dendrites nested by field of view nested by mice. Post hoc two-sample ttests were used to contrast psilocybin and saline groups, with and without splitting the sample by cell type. The calcium imaging statistical outputs were processed akin to the structural imaging model outputs as described above (i.e., residuals plots were inspected, fixed effect P values were computed with likelihood ratio tests, and post hoc two-sample t-test P values were Bonferroni-corrected for multiple comparisons). For confocal imaging, two-factor ANOVA was used for the analyses of spine density, spine head width, and spine protrusion length to test the interaction between treatment (psilocybin vs. saline) and genotype (Htr2a f/f vs. wild-type). Post hoc t-tests were used to compare psilocybin:PT neuron-targeted 5-HT2A receptor knockout versus saline: PT neuron-targeted 5-HT2A receptor knockout or psilocybin:wild-type and saline:wild-type. Bonferroni correction was used for multiple comparisons. For behavioral studies, performance was analyzed using software with automated procedures for fear extinction and learned helplessness. For head-twitch response and tail suspension test, video scoring was done by a different experimenter blinded to condition. For PT/IT studies, Two-factor ANOVA and post hoc t-tests were used for head-twitch response, learned helplessness test, and tail suspension test. The same statistical test was used for fear conditioning, extinction and retention to test the interaction between treatment (psilocybin vs. saline) and DREADD (hM4DGi vs. mCherry). Post hoc t-tests were used to compare psilocybin:mCherry versus saline:mCherry or psilocybin:hM4DGi and saline:hM4DGi for different sets of tones in a session. Bonferroni correction was applied for multiple comparisons. Due to a technical issue (faulty USB connection causing VideoFreeze software to crash in the middle of a session), a small subset of data from some mice were not used for the statistical test. For behavioral studies involving Htr2a f/f mice, two-tailed unpaired t-tests were used to For in vivo electrophysiology, we first identified optotagged neurons by constructing 0.1 ms bin peristimulus time histogram plots aligned to laser pulse onset. For comparison of time to first spike latency between PT and IT neurons, we conducted two-sided, twosample Kolmogorov-Smirnov test of time to first spike for all neurons. To compare the baseline firing rates we used a two-sample, independent t-test. For comparing the changes in firing rates before and after administration of saline or psilocybin, we conducted a paired, two-sided t-test using each neuron's baseline mean firing rate in the 30-min period before saline or drug administration (Pre) and the mean firing rate in