Highlights

•

Rabies viral tracing shows how psilocybin reshapes brain connectivity

•

Psilocybin strengthens pathways that route sensory inputs to subcortical regions

•

Psilocybin weakens inputs associated with cortico-cortical feedback loops

•

Manipulating neural activity alters the pattern of psilocybin-induced plasticity

Summary

Psilocybin holds promise as a treatment for mental illnesses. One dose of psilocybin induces structural remodeling of dendritic spines in the medial frontal cortex in mice. The dendritic spines would be innervated by presynaptic neurons, but the sources of these inputs have not been identified. Here, using monosynaptic rabies tracing, we map the brain-wide distribution of inputs to frontal cortical pyramidal neurons. We discover that psilocybin’s effect on connectivity is network specific, strengthening the routing of inputs from perceptual and medial regions (homolog of the default mode network) to subcortical targets while weakening inputs that are part of cortico-cortical recurrent loops. The pattern of synaptic reorganization depends on the drug-evoked spiking activity because silencing a presynaptic region during psilocybin administration disrupts the rewiring. Collectively, the results reveal the impact of psilocybin on the connectivity of large-scale cortical networks and demonstrate neural activity modulation as an approach to sculpt the psychedelic-evoked neural plasticity.

Graphical abstract

Keywords

1. psychedelic
2. neural plasticity
3. brain networks
4. depression
5. rabies
6. mouse

Introduction

Psychedelics are known to alter perceptual and cognitive states acutely after administration.1 In the past decade, psychedelic-based treatments have emerged as a potential therapeutic for mental health conditions.2 Among the various psychedelic compounds, psilocybin has yielded some of the most promising results. Numerous clinical trials have shown that psilocybin treatment with psychological support can relieve symptoms in patients with major depressive disorder or treatment-resistant depression.3,4,5,6,7 Strikingly, the therapeutic benefit has been reported to last for at least 6 weeks after the administration of a single dose.5 Psilocybin may be efficacious for other indications, such as alcohol use disorder.8 Because of the clinical relevance, there is an urgent need to understand the neurobiological mechanisms underlying psychedelic drug action.9

Structural neural plasticity is likely involved in psilocybin’s enduring effects on behavior. Individuals with major depressive disorder have fewer excitatory spine synapses and lower expression of synaptic proteins in the prefrontal cortex.10,11 By contrast, the fast-acting antidepressant ketamine promotes the growth of dendritic spines in frontal cortical pyramidal neurons in rodents.12,13,14 Like ketamine, a single dose of psilocybin leads to the formation of new dendritic spines in the mouse dorsal medial frontal cortex.15 Of note, the elevation in spine number density induced by classic psychedelics is long-lasting, persisting for at least a month.15,16 The durability of the structural remodeling in the brain may explain how psilocybin, which has a short half-life in the body,17 can cause sustained changes in behavior.

Considering that dendritic spines are the postsynaptic sites for excitatory synapses, a key unanswered question is the identity of the presynaptic neurons that provide axonal inputs for the new connections formed after psilocybin administration. As part of the medial network of the mouse,18 the dorsal medial frontal cortex receives long-range inputs from other cortical regions, such as the ventromedial prefrontal cortex, retrosplenial cortex (RSP), posterior parietal cortex (VISrl), and anterior insular cortex, as well as from subcortical regions, such as the mediodorsal and ventromedial nuclei of the thalamus, basolateral amygdala (BLA), and claustrum.18,19,20 Additionally, there are local inputs and inter-hemispheric inputs from the contralateral hemisphere.19,20 Uncovering the identity of the presynaptic neurons that provide inputs to the newly formed spines after drug administration is crucial because it will reveal the specific neural pathways that are modified by psilocybin.

Here, we perform monosynaptic tracing to map the brain-wide distribution of presynaptic cells that send inputs to the mouse dorsal medial frontal cortex. We find that psilocybin administration alters neuronal connectivity in a manner that is highly specific to brain networks. The subcortical-projecting, pyramidal tract (PT) subtype of frontal cortical pyramidal neurons gains inputs almost exclusively from the medial, sensorimotor, and visual-auditory networks but loses inputs from the ventromedial prefrontal cortex and lateral network. By contrast, the intratelencephalic (IT) subtype of frontal cortical pyramidal neurons exhibits a distinct and opposite pattern of reorganization for their presynaptic inputs. Moreover, we show that the acute effect of psilocybin on spiking activity determines if a particular source of presynaptic input would be strengthened subsequently. Collectively, our study provides crucial insights into how psilocybin modifies the connectivity of cortical networks.

Results

Whole-brain tracing of drug-evoked differences in monosynaptic inputs

Monosynaptic tracing can be achieved using a rabies virus that is engineered to be EnvA-pseudotyped and glycoprotein (G) deficient.21 Pseudotyping ensures that the rabies virus can enter only a set of experimenter-specified neurons (“starter cells”) that express the EnvA receptor (TVA). The deletion of G means that the rabies virus cannot spread, except from the starter cells in which G is expressed as a transgene. Therefore, a pseudotyped, G-deleted rabies virus (henceforth referred to as simply “rabies virus”) can transduce the starter cells and spread by crossing retrogradely one synapse to the presynaptic neurons (“input cells”), but would halt there from further spread. In this study, we used two viruses for monosynaptic tracing: an hSyn-driven, Cre-dependent adeno-associated virus (AAV) helper virus, AAV-hSyn-DIO-TVA66T-dTomato-CVS N2c G, and the rabies virus, EnvA-CVS N2cΔG-H2B-EGFP (Figure 1A).

Figure 1 Psilocybin modifies inputs to frontal cortical PTFezf2 neurons in a network-specific pattern

(A) The viruses used in this study.

(B) Experimental timeline.

(C) When the viruses are injected in the dorsal medial frontal cortex of a Fezf2-2A-CreER mouse, the PTFezf2 starter cells would express dTomato and EGFP, while input cells would express EGFP only.

(D) Monosynaptic tracing from frontal cortical PTFezf2 neurons using the Fezf2-2A-CreER mouse.

(E) Images from whole-brain imaging, showing starter cells (red and green) and input cells (green). Depths relative to bregma (red circle) are indicated. Scale bar, 1 mm.

(F) The number of starter cells, input cells in the ipsilateral hemisphere, and input cells in the contralateral hemisphere for Fezf2-2A-CreER mice treated with saline or psilocybin. Bar, mean. Circle, individual animal. Wilcoxon rank-sum test.

(G) Proportion of input cells contributed by presynaptic regions in the frontal cortex (mean ± SEM).

(H) Proportion of input cells contributed by other presynaptic regions (mean ± SEM).

(I) Drug-evoked difference (psilocybin subtracted by saline, relative to saline) in the proportion of input cells for all 65 presynaptic regions for PTFezf2 neurons (mean and 90% confidence interval). The list was sorted based on the drug-evoked difference. Presynaptic regions with >10% or <−15% in drug-evoked difference were color coded according to their network membership.

(J) Schematic showing the presynaptic regions with >10% or <−15% in drug-evoked difference.

(K) Network selectivity analysis, testing against the null hypothesis that the increases and decreases in drug-evoked difference were distributed randomly across the 5 cortical networks. Histogram, distribution of χ2 values for the null hypothesis. Vertical line, observed χ2 value (p = 6 × 10−5).

N = 9 mice for psilocybin and 8 mice for saline.

See also Figures S1, S2, and S3 and Table S1.

The experimental timeline for monosynaptic tracing is shown in Figure 1B. The Cre-dependent AAV helper virus was injected into the dorsal medial frontal cortex of an animal, for example, the Fezf2-2A-CreER mouse, which is an inducible Cre-driver strain targeting the PT subtype of cortical pyramidal cells.22 Subsequently, tamoxifen was administered to induce Cre expression. 2 weeks later, after sufficient time for TVA and G expression in the starter cells, psilocybin (1 mg/kg intraperitoneally [i.p.]) or saline was administered, followed a day later by the injection of rabies virus into the dorsal medial frontal cortex. The rabies virus was injected after drug administration to ensure that the labeled input cells would reflect differences caused by psilocybin and not pre-existing connectivity prior to drug administration. We chose a dose of 1 mg/kg for psilocybin because it evokes structural neural plasticity in the mouse dorsal medial frontal cortex.15 Finally, 1 week would elapse for rabies to spread before the brain was collected for analysis. The anticipated result is that we would find starter cells expressing both dTomato and EGFP, whereas the input cells would express EGFP only (Figure 1C).

We targeted the viral injections to the dorsal medial frontal cortex, encompassing the anterior cingulate cortex (ACAd) and the medial portion of the secondary motor cortex (MOs), because brain-wide c-Fos mapping studies showed that this region responds robustly to stress, ketamine, and psilocybin.23,24 Moreover, psilocybin leads to rapid and persistent structural neural plasticity in this region of the medial frontal cortex.15 We verified that the monosynaptic tracing method worked as expected by performing multiple control experiments involving a modified AAV helper virus that does not express G protein (Figures S1A–S1C), AAV helper virus only without rabies (Figures S1D and S1E), and rabies virus only without AAV helper virus (Figures S1F and S1G).

Figure S1 Control experiments validating monosynaptic tracing in Fezf2-2A-CreER mice, related to Figure 1

(A) Schematic of viral vectors used: Cre-dependent AAV helper (AAV1-hSyn-DIO-TVA66T-dTomato-CVS N2c G), Cre-dependent AAV control without G protein (AAV1-hSyn-DIO-TVA66T-dTomato), and EnvA-pseudotyped G-deleted rabies (EnvA-CVS N2cΔG-H2B-EGFP).

(B and C) Experiment involving the Cre-dependent AAV control and G-deleted rabies virus. Experimental timeline. Representative images showing EGFP (green) and dTomato (red) expression restricted to the injection site around the medial frontal cortex of a Fezf2-2A-CreER mouse, with nuclear staining by DAPI (blue). As expected, the G-deleted rabies virus could not spread due to the lack of G protein.

(D and E) Experiment involving the Cre-dependent AAV virus only. Experimental timeline. Representative images showing dTomato expression (red) at the injection site in the medial frontal cortex of a Fezf2-2A-CreER mouse brain. As expected, there was no green fluorescence because the G-deleted rabies virus was not injected.

(F and G) Experiment involving the G-deleted rabies virus only. Experimental timeline. Representative images showing sections corresponding to the medial frontal cortex of a Fezf2-2A-CreER mouse. As expected, the G-deleted rabies virus could not enter any cell because the neurons lack the TVA receptor.

Scale bars, 1 mm (C, left, E, and G), 100 μm (C, right).

Psilocybin modifies inputs to frontal cortical PTFezf2 neurons in a network-specific pattern

Two major subtypes of pyramidal neurons in the neocortex are the PT and IT neurons (Figures S2A–S2F). These subtypes are non-overlapping populations with distinct morphological characteristics, electrophysiological properties, and long-range projection targets.19,25,26 Previous work indicated that the PT subtype of pyramidal neurons in the dorsal medial frontal cortex is essential for the long-term behavioral effects of psilocybin.27 To determine how psilocybin may alter inputs into frontal cortical PT neurons, we performed monosynaptic tracing using Fezf2-2A-CreER mice (N = 4 male and N = 5 female animals for psilocybin; N = 4 male and N = 4 female animals for saline; Figures 1D, S2G, and S2H). We injected viruses into the right hemisphere; therefore, the right hemisphere is ipsilateral and left hemisphere is contralateral. To determine the brain-wide distribution of input cells, we processed the rabies-traced brains through a pipeline involving tissue clearing, light sheet fluorescence microscopy, and machine learning-based automated detection of nuclei, which enabled us to locate and count all starter and input cells in the mouse brain (Figure S3A). As expected, we observed co-expression of red (dTomato) and green (EGFP) fluorescence from starter cells located in the dorsal medial frontal cortex, whereas green (EGFP) fluorescence from input cells was widespread in the brain (Figure 1E). Video S1 shows images of the dTomato (red) and EGFP (green) expression, highlighting the brain-wide distribution of input cells into the frontal cortical PTFezf2 neurons. Video S2 shows images from the same brain for the dTomato expression (red) and NeuN immunostaining (white), showing the starter PTFezf2 neurons with cell bodies in the dorsal medial frontal cortex and axons projecting ipsilaterally to the striatum as well as out of the cerebrum into the thalamus and pons. The images were aligned to the mouse brain atlas and divided based on the 316 summary structures in the Allen Mouse Brain Common Coordinate Framework version 3.28

Figure S2 Estimates for the proportion of excitatory cells as PT, IT, PTFezf2, and ITPlxnd1 neurons in the mouse dorsal medial frontal cortex, related to Figures 1 and 2

(A) Proportion of cell types extracted from the publicly available spatial transcriptomics data hosted by the Allen Institute (Yao Z et al.67), focusing on the ∼14,900 glutamatergic (Glut) neurons in ACAd. IT, intratelencephalic; CT, corticothalamic; and ET, extratelencephalic.

(B) Proportion of layers 1, 2/3, 5, 6a, and 6b neurons that are IT, CT, ET, or other glutamatergic neurons in ACAd.

(C) Proportion of IT, CT, and ET neurons that reside in layers 1, 2/3, 5, 6a, or 6b of ACAd.

(D–F) Similar to (A)–(C) for MOs.

(G) AAV-mDlx-GFP was injected into the ACAd/medial MOs region of a Fezf2-2A-CreER;Ai75 mouse; therefore, GFP was expressed in GABAergic cells, tdTomato was expressed in PTFezf2 neurons, and NeuN in blue was used to stain all neurons (top row). A similar strategy was applied to a PlexinD1-2A-CreER;Ai75 mouse; therefore, GFP was expressed in GABAergic cells, tdTomato was expressed in ITPlxnd1 neurons, and NeuN in blue was used to stain all neurons (bottom row). Images from fixed coronal sections showing the dorsal medial frontal cortex. Scale bars, 200 μm (left), 50 μm (right).

(H) Quantification of the fraction of all neurons that were GABAergic in ACAd and medial MOs (left) and estimated fractions of excitatory (non-GABAergic) neurons that were PTFezf2 or ITPlxnd1 neurons (right) (mean ± SEM).

Figure S3 The distribution of input and starter cells in PTFezf2 and ITPlxnd1 mice after saline or psilocybin treatment, related to Figures 1 and 2

(A) Workflow for whole-brain imaging and analysis. Following tissue clearing, brains were imaged using a light sheet fluorescence microscope. The images were registered to the Allen Brain Atlas for cell counting analysis. The EGFP-expressing input cells were visualized in MeshView.

(B) Relationship between the total number of input and starter cells in the brain in PTFezf2 mice. Open circle, individual animal.

(C) Similar to (A) for ITPlxnd1 mice.

(D) The number of input cells per starter cell as a function of the number of starter cells in the brain of PTFezf2 mice. Line, exponential decay fit to data from the saline condition. Open circle, individual animal.

(E) Similar to (C) for ITPlxnd1 mice.

(F) The distribution of starter cells in various frontal cortical regions in PTFezf2 mice for the saline and psilocybin groups. Bar, mean. Open circle, individual animal.

(G) Similar to (E) for ITPlxnd1 mice.

(H) Proportion of total input cells in the left and right hemispheres for frontal cortical regions of PTFezf2 and ITPlxnd1 mice in the saline condition (mean ± SEM).

(I) Similar to (G) for all other regions in the brain.

Video S1. Light sheet fluorescence images showing the brain-wide distribution of input cells (green) into the frontal cortical PTFezf2 starter cells (red and green) in a Fezf2-2A-CreER mouse after psilocybin administration, related to Figure 2

Video S2. Light sheet fluorescence images showing the frontal cortical PTFezf2 neurons (red) and NeuN stain (white) in a Fezf2-2A-CreER mouse after psilocybin administration, related to Figure 2

Starter cells were defined as those neurons with colocalized red and green fluorescence within the ipsilateral frontal cortex (primary motor cortex [MOp], MOs, ACAd, anterior cingulate area, ventral part [ACAv], prelimbic area [PL], infralimbic area [ILA], orbital area, lateral part [ORBl], medial orbital frontal cortex [ORBm], and orbital area, ventrolateral part [ORBvl]). We had similar initial conditions for the psilocybin and saline groups, with 2,680 ± 491 and 2,350 ± 359 starter cells, respectively (p = 0.7, Wilcoxon rank-sum test; Figure 1F). As expected, most starter cells were detected in the dorsal medial frontal cortex (94% ± 2% and 95% ± 2% in ACAd, MOs, or PL for the saline and psilocybin groups, respectively; Figure S3F). Input cells were defined as those neurons with green fluorescence anywhere in the brain, except they cannot also have red fluorescence, which would make them starter cells. For mice that received psilocybin, there were 5.1 ± 0.5 × 105 input cells, whereas for mice that received saline, there were 4.7 ± 0.3 × 105 input cells (p = 0.7, Wilcoxon rank-sum test; Figures 1F, S3B, and S3D).

To identify the sources of the presynaptic inputs, we calculated the fraction of input cells residing in each region. Many regions did not provide appreciable input to the dorsal medial frontal cortex, so we analyzed only the 65 regions contributing at least 0.3% of the total inputs in any treatment or cell-type condition (henceforth referred to as “presynaptic regions”). Together these presynaptic regions captured nearly all input cells (saline, 87.6% ± 0.6%; psilocybin, 89.2% ± 0.7%; Figures 1G and 1H). Next, we sorted the presynaptic regions based on the normalized difference of psilocybin’s effect on the input fraction relative to saline, revealing numerous regions with increases, including the primary somatosensory cortex and its associated thalamic nuclei (primary somatosensory area [SSp], ventral posteromedial nucleus of the thalamus [VPM], and posterior complex of the thalamus [PO]), VISrl, primary visual cortex (VISp), MOp, RSP, and others (Figure 1I). Other regions had decreased input fraction following psilocybin administration, such as the ILA, ORBm, ventral agranular insular cortex (AIv), hippocampus (CA1), and BLA. Focusing on the regions with the largest psilocybin-evoked increase (>10%) or decrease (<−15%) in input fraction, we discovered that the strengthened presynaptic regions belong to the medial, visual-auditory, and sensorimotor networks (Figure 1J). This demarcation of cortical networks is based on modules identified in a prior large-scale study of long-range connectivity.18 Regions that contributed fewer inputs after psilocybin administration came from the lateral network, ventromedial prefrontal cortex, and medial nuclei of the thalamus (Figure 1J). To test the statistical likelihood of such a network-specific preference in the input fraction change, we performed a chi-squared test to compare the observed data with shuffled data to find that the network selectivity for psilocybin’s effect was highly significant for PTFezf2 neurons (χ2 = 24.6, degrees of freedom [df] = 4, p = 6 × 10−5, chi-squared test; Figure 1K). The results demonstrate that psilocybin induces a network-specific reorganization of the inputs impinging on the PT subtype of pyramidal neurons in the mouse dorsal medial frontal cortex.

Opposing effects of psilocybin on inputs to the two major pyramidal cell types

Another major subtype of pyramidal neurons is the IT neurons. Unlike the subcortical-projecting PT neurons, the axons of IT neurons stay within the telencephalon to communicate with cortical and striatal locations in both hemispheres. Studies that examined cortical PT and IT neurons showed that the two cell types receive different local and long-range inputs,19,29,30,31 which may underpin their distinct roles during behavior.27,32,33,34 To determine how psilocybin may modify the inputs to frontal cortical IT neurons, we performed monosynaptic tracing using the PlexinD1-2A-CreER mouse, which is an inducible Cre-driver line for IT neurons22 (N = 4 male and N = 4 female animals for psilocybin; N = 4 male and N = 4 female animals for saline; Figures 2A, 2B, S2G, and S2H; Videos S3 and S4).

Figure 2 Opposing effects of psilocybin on presynaptic inputs to the two major pyramidal cell types

(A) Monosynaptic tracing from frontal cortical ITPlxnd1 neurons using the PlexinD1-2A-CreER mouse.

(B) Images from whole-brain imaging, showing starter cells (red and green) and input cells (green). Depths relative to bregma (red circle) are indicated. Scale bar, 1 mm.

(C) The number of starter cells, input cells in the ipsilateral hemisphere, and input cells in the contralateral hemisphere for PlexinD1-2A-CreER mice treated with saline or psilocybin. Bar, mean. Circle, individual animal.

(D) Proportion of input cells contributed by presynaptic regions in the frontal cortex (mean ± SEM).

(E) Proportion of input cells contributed by other presynaptic regions (mean ± SEM).

(F) Drug-evoked difference (psilocybin subtracted by saline, relative to saline) in the proportion of input cells for all 65 presynaptic regions for ITPlxnd1 neurons (mean and 90% confidence interval). The list was sorted based on the drug-evoked difference. Presynaptic regions with >10% or <−15% in drug-evoked difference were color coded according to their network membership.

(G) Source of ipsilateral presynaptic inputs to PTFezf2 and ITPlxnd1 neurons in saline-treated mice. Circle, individual animal. ∗p < 0.05, ∗∗p < 0.01, Wilcoxon rank-sum test.

(H) Drug-evoked difference in the proportion of ipsilateral input cells from the 5 cortical networks and outside neocortex to PTFezf2 and ITPlxnd1 neurons. Error bar, 90% confidence interval. ∗, significant based on 95% confidence intervals.

(I) Drug-evoked difference in the proportion of input cells for all 65 presynaptic regions, for PTFezf2 neurons versus ITPlxnd1 neurons. Circle, individual presynaptic region. r, Pearson correlation coefficient.

N = 8 mice for psilocybin and 8 mice for saline.

See also Figures S2, S3, and S4 and Table S1.

Video S3. Light sheet fluorescence images showing the brain-wide distribution of input cells (green) into the frontal cortical ITPlxnd1 starter cells (red and green) in a PlexinD1-2A-CreER mouse after psilocybin administration, related to Figure 3

Video S4. Light sheet fluorescence images showing the frontal cortical ITPlxnd1 neurons (red) and NeuN stain (white) in a PlexinD1-2A-CreER mouse after psilocybin administration, related to Figure 3

Beginning from 5,495 ± 1,046 and 4,511 ± 696 starter cells (p = 0.6, Wilcoxon rank-sum test; 92% ± 1% and 91% ± 3% located in ACAd, MOs, or PL; Figures 2C and S3G), we counted 5.6 ± 0.5 × 105 and 5.7 ± 0.5 × 105 input cells for the psilocybin and saline groups, respectively (p = 0.9, Wilcoxon rank-sum test; Figures 2C, S3C, and S3E). The same set of 65 presynaptic regions, selected based on the criterion of containing at least 0.3% of the total inputs in any treatment or PT/IT condition, captured nearly all the input cells to the frontal cortical ITPlxnd1 neurons (saline, 89.4% ± 0.6%; psilocybin, 89.8% ± 0.6%). We plotted the input fraction for the psilocybin and saline groups in each of the frontal cortical (Figure 2D) and long-range cortical and subcortical presynaptic regions (Figure 2E). We plotted the difference in input fraction due to psilocybin relative to saline (Figures 2F, S4A, and S4D). Although the specific input cells exhibiting changes were distinct between PTFezf2 and ITPlxnd1 neurons, we found that psilocybin’s effect on the input fraction of frontal cortical ITPlxnd1 neurons is also highly network-specific (χ2 = 15.0, df = 4, p = 0.005, chi-squared test; Figures S4B and S4C). The next analyses would be aimed at delineating the significant differences between PTFezf2 and ITPlxnd1 neurons.

Figure S4 Drug-evoked difference and network selectivity for IT neurons in PlexinD1-2A-CreER mice, related to Figure 2

(A) Drug-evoked difference (psilocybin subtracted by saline, relative to saline) in the proportion of input cells for all 65 presynaptic regions for ITPlxnd1 neurons (mean and 90% confidence interval). The list of presynaptic regions was sorted based on the drug-evoked difference value for ITPlxnd1 neurons. Presynaptic regions with >10% or <−5% in drug-evoked difference were color coded according to their network membership.

(B) Schematic showing the location of the presynaptic regions with >10% or <−5% in drug-evoked difference for ITPlxnd1 neurons. The presynaptic regions are color coded according to their network membership.

(C) Network selectivity analysis, testing against the null hypothesis that the increases and decreases in drug-evoked difference for ITPlxnd1 neurons would be distributed randomly across the 5 cortical networks. Histogram, the distribution of χ2 values for the null hypothesis. Vertical line, the observed χ2 value (p = 0.0047).

(D) Drug-evoked difference (psilocybin subtracted by saline, relative to saline) in the proportion of input cells for all 65 presynaptic regions for PTFezf2 neurons (mean and 90% confidence interval). The list of presynaptic regions was sorted based on the drug-evoked difference value for ITPlxnd1 neurons (i.e., same order as in A).

First, only considering the saline condition, we observed that PTFezf2 and ITPlxnd1 neurons receive long-range inputs from different networks. Frontal cortical ITPlxnd1 neurons received a higher fraction of inputs from the medial network (15.0% ± 0.7% for IT, 11.4% ± 0.6% for PT; p = 0.004, Wilcoxon rank-sum test) and visual-auditory regions (2.7% ± 0.3% for IT, 1.9% ± 0.2% for PT; p = 0.05, Wilcoxon rank-sum test), but had a significantly lower fraction of inputs from outside the neocortex (17.6% ± 0.6% for IT, 20.3% ± 0.7% for PT; p = 0.007, Wilcoxon rank-sum test), compared with their PTFezf2 counterparts (Figures 2G, S3H, and S3I). When we determined the effects of psilocybin, there were also cell-type-specific differences. In response to psilocybin, the ITPlxnd1 neurons differed significantly from the PTFezf2 neurons in that they had reduced input fraction from the sensorimotor network (−4% for IT, 17% for PT; statistically significant based on 95% confidence interval from bootstrapping), but instead gained input fraction from the lateral network (10% for IT and −14% for PT) and the ventromedial prefrontal cortex (8% for IT and −12% for PT; Figure 2H). Both pyramidal cell types lost input fraction from regions outside the neocortex after psilocybin administration, but the effect was less pronounced for IT neurons (−3% for IT and −12% for PT). Psilocybin’s opposing effect on the input fraction to the two cell types was most clearly demonstrated by plotting the drug-induced difference in input fraction for all 65 presynaptic regions, which showed a significant negative correlation (r = −0.58, p = 5 × 10−7, Pearson correlation; Figure 2I). These results indicate that, in the mouse dorsal medial frontal cortex, for those inputs that are strengthened in PT neurons by psilocybin, they tend to be weakened in IT neurons after drug administration, and vice versa. Therefore, psilocybin exerts opposing changes to the synaptic input organization for the two main pyramidal cell subtypes in the mouse dorsal medial frontal cortex.

Psilocybin-evoked rewiring is related to the pattern of drug-induced c-Fos expression

We wanted to know why inputs from certain presynaptic regions are selectively strengthened by psilocybin. Presynaptic regions send axon collaterals that terminate in different layers of the mouse dorsal medial frontal cortex, which may influence psilocybin’s plasticity potential. To explore this possibility, we leveraged the Allen Mouse Connectivity database, which contains images of fluorescent axons following viral injections at hundreds of locations in the mouse brain.35 The analyses suggest that the laminar distribution of long-range axons in the dorsal medial frontal cortex is unlikely to explain why certain presynaptic regions are favored for potentiation by psilocybin (Figure S5).

Figure S5 Analysis of the laminar profile of axonal density from presynaptic regions, related to Figure 3

(A) Representative images of 4 presynaptic regions with increased numbers of input cells following psilocybin treatment, as identified by monosynaptic rabies tracing. Top row, 3D visualization of brain-wide axonal projection emanating from neurons in the presynaptic region. Second row, serial two-photon image showing the section at the injection site (arrowhead). Third row, serial two-photon image showing the section at the dorsal medial frontal cortex. Fourth and fifth rows, magnified view of the framed area in the third row. Red, autofluorescence. Green, fluorophore.

(B) Similar to (A) for 4 presynaptic regions with decreased number of input cells.

(C) The laminar profile of pixel intensity (as an estimate of axonal density) in presynaptic regions corresponding to the examples shown in (A). Red dashed lines, positions that divide layers 1, 2/3, 5, 6a, and 6b.

(D) Similar to (C), corresponding to the presynaptic regions depicted in (B).

(E) Scatterplot showing how drug-evoked difference in the number of input cells to PTFezf2 neurons may relate to the summed axonal density, which was calculated by adding up the pixel intensity across the entire profile. Filled circle, individual presynaptic region.

(F) Similar to (E) for location of peak axonal density, which was determined by finding the laminar position with the highest pixel intensity.

(G) Similar to (E) for relative density of axons in layer 1, which was calculated by dividing the summed axonal density in the portion of the profile demarcated as layer 1 by the summed axonal density across the entire profile.

(H) Similar to (G) for layer 2/3.

Another possibility is that the baseline expression of certain subtypes of 5-HT receptors may predispose a presynaptic region to be sensitive to psilocybin-induced rewiring. Using the in situ hybridization data in the Allen Brain Atlas,36 we correlated the expression levels of the Htr1a, Htr2a, and Htr2c transcripts, which encode the 5-HT1A, 5-HT2A, and 5-HT2C receptors, respectively, with the psilocybin-induced difference in input cells for presynaptic regions in the cortex (Figures 3A–3C). There was no clear relationship for Htr1a and Htr2a. For Htr2c, the correlation approached significance for PTFezf2 neurons (p = 0.08, Pearson correlation), suggesting a potential trend, and reached significance for ITPlxnd1 neurons (p = 0.02, Pearson correlation). These analyses suggest that while receptor expression does not fully explain the psilocybin-evoked changes, certain subtypes, such as Htr2c may partially contribute.

Figure 3 Psilocybin-evoked rewiring correlates with the pattern of drug-evoked c-Fos expression

(A) In situ hybridization (ISH) images for Htr1a, Htr2a, and Htr2c transcripts.

(B) The ISH expression of Htr1a, Htr2a, or Htr2c versus psilocybin-evoked differences in input fraction for PTFezf2 neurons. Circle, individual presynaptic region. The color denotes the cortical network membership. r, Pearson correlation coefficient.

(C) Similar to (B) for ITPlxnd1 neurons.

(D) Experimental timeline for imaging psilocybin’s effect on brain-wide c-Fos expression. Insets, representative fluorescence images in a psilocybin-treated mouse. Scale bar, 1 mm.

(E) Psilocybin-evoked differences in c-Fos expression versus psilocybin-evoked differences in input fraction for PTFezf2 and ITPlxnd1 neurons. Circle, individual presynaptic region. The color denotes the cortical network membership. r, Pearson correlation coefficient.

See also Figure S5 and Table S1.

We found a more robust relationship when examining the psilocybin-evoked activation of the immediate early gene c-Fos. In a previous study,37 our lab injected mice with psilocybin (1 mg/kg i.p.) or saline, collected the brains 2 h later, and then used tissue clearing, c-Fos antibody labeling, and light sheet fluorescence microscopy to visualize c-Fos+ cells in the entire mouse brain (Figure 3D). We compared the drug-evoked difference in c-Fos+ cells with the drug-evoked difference in input cells on a region-by-region basis. For the presynaptic regions in the neocortex, those locations with greater increases in the number of c-Fos+ cells were associated with a higher elevation of input cells for frontal cortical PTFezf2 neurons after psilocybin administration (r = 0.44, p = 0.02, Pearson correlation; Figure 3E). By contrast, psilocybin’s effects on c-Fos activation and input cell gain were negatively correlated for the frontal cortical ITPlxnd1 neurons (r = −0.41, p = 0.03, Pearson correlation; Figure 3E). Altogether, these results indicate that psilocybin-induced c-Fos activation is related to the gain and loss of input cells for frontal cortical PT and IT neurons, respectively. Because c-Fos expression is known to be activity-dependent, this exploratory analysis hints at spiking activity as a potential factor underlying the pattern of synaptic reorganization evoked by psilocybin.

Potentiation of RSP inputs onto PTFezf2 neurons in the medial frontal cortex after psilocybin administration

To delineate the potential mechanisms, we focused on one presynaptic region: the RSP. RSP is known to have dense reciprocal connections with the dorsal medial prefrontal cortex in mice.18,20,38 The RSP is relevant for the neurobiology of psilocybin because it is a core region in the mouse analog of the default mode network,39,40,41 which is thought to be centrally involved in mediating the effects of psychedelics in humans.42,43,44 Our whole-brain rabies tracing results indicated that RSP was a top-ranked region with >10% increase in input fraction to frontal cortical PTFezf2 neurons.

We performed longitudinal two-photon imaging to determine the effect of psilocybin on the density of long-range inputs in the medial frontal cortex (Figures 4A and 4B). To label the boutons along axons originating from neurons in RSP, we injected AAV-hSyn-Synaptophysin-mRuby2 into RSP. Initially, we planned to visualize simultaneously axonal boutons and dendritic spines; therefore, we also injected AAV-CAG-Flex-EGFP into the ACAd/medial MOs of these Fezf2-2A-CreER mice to label PTFezf2 neurons; however, the EGFP-expressing dendrites were too dense to resolve most spines, and thus we restricted the analysis to axonal boutons only (Figure 4C). We tracked a total of 60 and 62 fields of view from 7 and 8 animals for the saline and psilocybin conditions, respectively. Using an automated algorithm to count red puncta from the in vivo images, we observed that psilocybin increased the number of detectable RSP axonal boutons in superficial layers of the dorsal medial frontal cortex at 1 and 3 days following treatment, relative to saline (day 1, 5.5% ± 3.3% for psilocybin versus −8.6% ± 3.9% for saline; p = 0.056, post hoc comparison; day 3, 11.4% ± 4.6% for psilocybin versus −7.6% ± 3.9% for saline; p = 0.007, post hoc comparison; treatment and time interaction, p = 0.0004, linear mixed-effects model; Figure 4D). We note a decline in bouton density prior to treatment, possibly due to photobleaching of the synaptophysin-fused fluorophores, which affected the saline and psilocybin groups similarly (day −1, −4.7% ± 1.3% for psilocybin and −4.8% ± 1.7% for saline; p = 1, post hoc comparison). The imaging results thus corroborate the finding from rabies tracing to show a psilocybin-evoked increase in RSP axonal boutons in the dorsal medial frontal cortex.

Figure 4 Potentiation of RSP inputs onto PTFezf2 neurons in the medial frontal cortex after psilocybin administration

(A) Strategy to express mRuby2 in axonal boutons of RSP neurons and EGFP in frontal cortical PTFezf2 neurons.

(B) Experimental timeline.

(C) Example in vivo two-photon images. Scale bar, 10 μm.

(D) The density of RSP axonal boutons in the medial frontal cortex after psilocybin (magenta; 1 mg/kg i.p.) or saline (black) across days, expressed as the fold change from the baseline (mean ± SEM).

(E) Strategy to express ChR2 in RSP neurons and tdTomato in PTFezf2 neurons.

(F) Post hoc histology. Scale bars, 100 μm (left top), 20 μm (middle top), 10 μm (right top), 500 μm (left bottom).

(G) Experimental timeline.

(H) Successive whole-cell recordings were made from a PTFezf2 neuron and a non-PTFezf2 neuron.

(I) Example optogenetically evoked EPSC in a pair of PTFezf2 and non-PTFezf2 neurons for saline and psilocybin conditions.

(J) Amplitude of the optogenetically evoked EPSC, 24 h after saline administration. Circle, individual cell.

(K) Similar to (J) for psilocybin administration.

(L) Based on data in (J) and (K), the ratio is calculated by dividing the amplitude of a PTFezf2 neuron by the amplitude of its paired non-PTFezf2 neuron. Circle, individual cell pair. Bar, mean ± SEM.

(M–O) Similar to (J)–(L), 3 days after saline or psilocybin administration.

(P) Example optogenetically evoked EPSCs evoked by two brief pulses in a pair of PTFezf2 and non-PTFezf2 neurons, for saline and psilocybin conditions.

(Q) PPR, 24 h after saline administration. Circle, individual cell.

(R) Similar to (Q) for psilocybin administration.

(S and T) Similar to (Q) and (R), 3 days after saline or psilocybin administration.

For imaging, n = 60 fields of view from 7 mice for saline and 62 fields of view from 8 mice for psilocybin. For slice electrophysiology, n = 23–25 cell pairs from 6 mice for 24 h after saline, 26–29 cell pairs from 7 mice for 24 h after psilocybin, 33 cell pairs from 5 mice for 3 days after saline, and 34 cell pairs from 7 mice for 3 days after psilocybin. Linear mixed-effects model. See Table S2. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001.

See also Figure S6.

To evaluate further the effect of psilocybin on RSP inputs and specifically in frontal cortical PTFezf2 neurons, we measured synaptic transmission using slice electrophysiology. We injected AAV-hSyn-hChR2(H134R)-EYFP into RSP to enable photostimulation of axons originating from RSP, which was done in Fezf2-2A-CreER;Ai14 mice where PTFezf2 neurons expressed tdTomato so we could target them for whole-cell recording (Figures 4E and 4F). Acute brain slices containing ACAd/medial MOs were prepared at 1 or 3 days after the administration of psilocybin (1 mg/kg i.p.) or saline (5–7 mice for each condition; Figure 4G). During each recording session, we would pair the recordings by measuring in succession from a PTFezf2 neuron and a non-PTFezf2 neuron (i.e., an adjacent pyramidal cell without tdTomato; Figure 4H). The motivation was to account for the variable ChR2 expression across experiments and the opposing effects of psilocybin for inputs onto PTFezf2 and ITPlxnd1 neurons. A single brief pulse of photostimulation was used to elicit an optogenetically evoked excitatory postsynaptic current (oEPSC; Figure 4I). We found that psilocybin induced cell-type-specific effects on oEPSC amplitude when tested 24 h or 3 days after treatment (treatment and cell-type interaction, p = 3 × 10−6, linear mixed-effects model; Figures 4J, 4K, 4M, and 4N). Specifically, the oEPSC amplitude for PTFezf2 neurons relative to non-PTFezf2 pyramidal cells was higher after psilocybin administration (main effect of treatment, p = 0.02, linear mixed-effects model; Figures 4L and 4O). We also assessed psilocybin’s effects on optogenetically evoked inhibitory postsynaptic currents (oIPSCs; Figures S6A–S6G), the oEPSC-to-oIPSC ratio (Figures S6H–S6J), and monosynaptic oEPSC in the presence of tetrodotoxin (TTX) and 4-AP (Figures S6K–S6O). Two brief pulses of photostimulation spaced by a short interval were applied to measure paired-pulse ratio (PPR; Figure 4P). With this protocol, we likewise observed that psilocybin induced cell-type-specific effects on PPR at both 1 and 3 days after psilocybin administration (treatment and cell-type interaction, p = 1 × 10−9, linear mixed-effects model; Figures 4Q–4T). The reduced PPR in PTFezf2 neurons suggests psilocybin’s effects on the RSP→ACAd excitatory connections may include an increase in presynaptic release probability. Additional electrophysiological data obtained via a different viral labeling strategy provided further evidence for psilocybin’s effects on oEPSC and PPR (Figures S6P–S6W). Overall, these data demonstrate psilocybin-induced potentiation of excitatory synaptic drive from RSP to frontal cortical PTFezf2 neurons, which persisted for at least 3 days after drug exposure.

Figure S6 Additional analyses and data for the slice electrophysiology experiments, related to Figure 4

(A) Example optogenetically evoked inhibitory postsynaptic current (IPSC) in a pair of PTFezf2 and non-PTFezf2 neurons for saline and psilocybin condition.

(B) Amplitude of the optogenetically evoked IPSC, 24 h after saline administration. Circle, individual cell.

(C) Similar to (B) for psilocybin administration.

(D) Based on data in (B) and (C), the ratio is calculated by dividing the amplitude of a PTFezf2 neuron by the amplitude of its paired non-PTFezf2 neuron. Circle, individual cell pair. Bar, mean ± SEM.

(E–G) Similar to (B)–(D), 3 days after saline or psilocybin administration.

(H) Example oEPSC and oIPSC recorded from the same non-PTFezf2 neuron.

(I) Ratio of the oEPSC amplitude divided by the oIPSC amplitude, 24 h or 3 days after saline or psilocybin administration, for PTFezf2 and non-PTFezf2 neurons. Circle, individual cell. Line, mean ± SEM.

(J) Similar to (I) except using a light pulse with higher intensity. This was because when using standard light intensity, EPSCs were reliably evoked, but IPSCs were sometimes absent. This was why after testing with the standard light intensity, we typically repeated the measurements with a higher light intensity, which could evoke stronger IPSCs. However, sometimes EPSCs have abnormal shapes at the higher light intensity, and those cells were excluded from this plot. Considering these caveats, we show both sets of data at the two different light intensities.

(K) Example optogenetically evoked EPSC at baseline and after the application of 1 μM TTX or 1 μM TTX and 100 μM 4-AP. TTX and 4-AP were added to isolate monosynaptic transmission.

(L) Amplitude of the optogenetically evoked EPSC in the different conditions, normalized by baseline. Circle, individual cell.

(M) Amplitude of the optogenetically evoked EPSC in the presence of TTX and 4-AP, 24 h after saline administration. Circle, individual cell.

(N) Similar to (M) for psilocybin administration.

(O) Based on data in (M) and (N), the ratio is calculated by dividing the amplitude of a PTFezf2 neuron by the amplitude of its paired non-PTFezf2 neuron. Circle, individual cell pair. Bar, mean ± SEM.

(P) Viral strategy to express ChR2 in RSP neurons. PTFezf2 neurons express tdTomato because AAV-CAG-Flex-tdTomato was injected into the ACAd/medial MOs region of a Fezf2-2A-CreER mouse.

(Q) Post hoc histology showing the fluorophore expression in medial frontal cortex and RSP. Scale bars, 100 μm (left top), 20 μm (middle top), 10 μm (right top), 500 μm (left bottom).

(R) Experimental timeline. Successive whole-cell recordings were made from a PTFezf2 neuron and a non-PTFezf2 neuron. A 470 nm LED was used to photostimulate the RSP axons in the acute brain slice.

(S) Amplitude of the optogenetically evoked EPSC, 3 days after saline administration. Circle, individual cell.

(T) Similar to (S) for psilocybin administration.

(U) Based on data in (S) and (T), the ratio is calculated by dividing the amplitude of a PTFezf2 neuron by the amplitude of its paired non-PTFezf2 neuron. Circle, individual cell pair. Bar, mean ± SEM.

(V) PPR, 3 days after saline administration. Circle, individual cell.

(W) Similar to (V) for psilocybin administration.

Data in (A)–(O) came from the same experiments to obtain results shown in Figures 4E–4T (i.e., Ai14-based labeling of PTFezf2 neurons). For (A)–(J), n = 13–16 cell pairs from 6 mice for 24 h after saline, 19–22 cell pairs from 7 mice for 24 h after psilocybin, 19–26 cell pairs from 5 mice for 3 days after saline, 23–24 cell pairs from 7 mice for 3 days after psilocybin. For (L), n = 15 cells from 14 animals. For (M)–(O), n = 17 cell pairs from 6 mice for 24 h after saline, 17 cell pairs from 5 mice for 24 h after psilocybin. Data in (P)–(W) came from the different experiments involving viral-mediated tdTomato expression in frontal cortical PTFezf2 neurons. n = 18 cell pairs from 4 mice for 3 days after saline, 26 cell pairs from 5 mice for 3 days after psilocybin.

For (B), (C), (E), and (F), a linear mixed-effects model with fixed effects terms of drug (saline or psilocybin), time (24 h or 3 days), cell type (PTFezf2 or non-PTFezf2), and all interactions, with cell pairs per brain slice per mouse modeled as nested random intercepts. For (D) and (G), linear mixed-effects model with fixed effects terms of drug (saline or psilocybin), time (24 h or 3 days), and interaction, with a random intercept for brain slice and animal. Post hoc pairwise comparisons with Bonferroni correction. For (I) and (J), data were analyzed separately using linear mixed-effects models with fixed effects of drug, time, cell type, and all interactions, and random intercepts for cell pair nested within brain slice and animal (Mouse/Slice/CellPair). Post hoc pairwise comparisons between treatment groups were performed for each cell type and ti