Single-nucleus transcriptomics reveals cell type-specific and time-dependent effects of psilocybin and ketamine on gene expression
This mouse study used single-nucleus RNA sequencing to examine how psilocybin and ketamine change gene expression in the medial frontal cortex over time. It found cell type-specific effects, with psilocybin altering genes linked to synaptic plasticity, metabolism and glial function, and its transcriptional response peaking earlier and later than ketamine’s.
Abstract
There is growing interest to investigate classic psychedelics and ketamine as therapeutics for mental illnesses. Previous studies have demonstrated that one dose of psilocybin or ketamine leads to persisting neural and behavioral changes. The durability of these effects suggests that there are likely alterations in gene expression at the transcriptional level. In this study, we performed single-nucleus RNA sequencing of the dorsal medial frontal cortex of male and female mice. Samples were collected at 1, 2, 4, 24, or 72 hours after psilocybin or ketamine administration and from control animals. At baseline, major subtypes of excitatory and GABAergic neurons selectively express particular serotonin receptor transcripts. The psilocybin-evoked differentially expressed genes in excitatory neurons are involved in synaptic plasticity, distinct from genes enriched in GABAergic neurons, which contribute to mitochondrial function and cellular metabolism, and non-neuronal glial cells. The effect of psilocybin on gene expression is time-dependent, including an early phase at 1 hour followed by a late phase at 72 hours of transcriptional response after administration, and differs from the changes following ketamine administration, which peaks at 2 – 4 hours. Collectively, the results provide a resource for understanding the cell type-specific and time-dependent changes in gene expression induced by psilocybin and ketamine in the mouse medial frontal cortex, which may underpin the drug’s long-term effects on neural circuits and behavior.
Research Summary of 'Single-nucleus transcriptomics reveals cell type-specific and time-dependent effects of psilocybin and ketamine on gene expression'
βBlossom's Take
Introduction
Psilocybin and ketamine are both of interest as potential treatments for mental illness, and earlier clinical and preclinical work has suggested that each can produce benefits that outlast their acute effects. The biological basis of these longer-lasting effects remains uncertain, but previous studies have implicated structural plasticity in the prefrontal cortex and, by extension, changes in gene expression. For psilocybin in particular, much less was known about genome-wide transcriptional responses, and earlier studies often relied on targeted methods rather than cell type-resolved transcriptomics. The authors also emphasise that timing matters because the acute and therapeutic effects of these drugs unfold on different timescales. Liao and colleagues therefore set out to create a resource that maps how psilocybin and ketamine alter gene expression across cell types in the mouse dorsal medial frontal cortex over time. Their aim was to use single-nucleus RNA sequencing to define cell type-specific and time-dependent transcriptional responses after a single dose of either drug, and to compare the resulting patterns across early and later post-dose windows. The study was intended as a publicly accessible dataset and analytical resource for understanding how these compounds may influence neural circuits and behaviour.
Methods
The researchers used single-nucleus RNA sequencing in the dorsal medial frontal cortex of C57BL/6J mice, a region encompassing the anterior cingulate cortex and medial premotor cortex. Mice were randomly assigned to receive psilocybin 1 mg/kg intraperitoneally, ketamine 10 mg/kg intraperitoneally, or no injection as controls. Tissue was collected from both hemispheres at 1, 2, 4, 24, or 72 hours after drug administration. The extracted text indicates that 49 animals were initially processed, but 7 were excluded for technical reasons, leaving 42 samples for final analysis. Nuclei were isolated from microdissected tissue and sequenced using the 10x Genomics Chromium Single Cell 3' v3 platform. The authors report an average sequencing depth of 29,562 reads per nucleus before additional quality control, and around 8,049, 8,414, and 7,689 usable reads per nucleus for psilocybin, ketamine, and control samples respectively after filtering. Cell types were annotated using the Allen Brain Atlas MapMyCells tool, and the dataset was organised into 16 major cell groups: 8 excitatory neuron types, 4 GABAergic interneuron types, and 4 non-neuronal cell types. The authors note that the full dataset was made available through an interactive web portal. For differential expression, they used pseudobulk analysis, aggregating raw counts within each sample-cell type combination and testing with pyDESeq2. Genes with fewer than 10 total counts were excluded from a given comparison, and significance was defined using false discovery rate correction at 5%. Gene ontology enrichment analysis was then performed with Enrichr on upregulated differentially expressed genes, using cell type-specific background gene lists rather than the whole genome to reduce misleading enrichment. The text also states that analyses were carried out in standard single-cell software environments including scanpy and scVI, with batch-corrected embeddings and quality control steps such as multiplet removal, low-count filtering, and exclusion of cells with high mitochondrial or ribosomal content.
Results
In control animals, the authors found cell type-specific expression of serotonin receptor transcripts. Htr2a was prominent in excitatory neuron types, including IT and ET neurons, and was also present in deep layer 6b neurons. Among inhibitory neurons, Lamp5 and Pvalb cells expressed Htr2a, whereas Vip cells more clearly expressed Htr2c. Non-neuronal cells showed relatively low serotonin receptor transcript levels. Psilocybin produced distinct transcriptional responses across cell classes. In excitatory neurons, the differentially expressed genes were enriched for synaptic plasticity and excitatory neurotransmission pathways, including chemical synaptic transmission, regulation of neuron projection development, excitatory chemical synaptic transmission, and regulation of synaptic plasticity. Examples highlighted included Mink1, Sipa1l1, Camk1g, Shisa6, Nos1ap, Grin2a, Gria1, and Sorcs3, many of which are linked to dendritic growth, receptor trafficking, or glutamatergic signalling. Across excitatory subtypes, the pattern was biphasic, with an early response at 1 hour and a later response at 72 hours. GABAergic neurons also responded to psilocybin, but their enrichment profile differed from that of excitatory neurons. Pvalb and Sst interneurons showed the largest early responses, with 933 and 662 differentially expressed genes at 1 hour, compared with 154 in Vip cells and 80 in Lamp5 cells. Enriched pathways were related mainly to mitochondrial function and metabolic processes, including aerobic electron transport chain, mitochondrial ATP synthesis, and glycolysis. Representative genes included Cox6c, Ndufs2, Kcnq3, and Amigo1. Non-neuronal cells also displayed strong responses despite low baseline serotonin receptor transcript levels. Astrocytes and oligodendrocytes were the most responsive, with 776 and 789 differentially expressed genes at 1 hour, while oligodendrocyte precursor cells and microglia showed fewer changes. Enriched pathways included neuron projection guidance, axon guidance, axonogenesis, and chemical synaptic transmission. The authors highlighted genes such as Nlgn3, Lrp1, Aldoc, and Lars2, with Lars2 showing a notable later response at 72 hours. When psilocybin and ketamine were compared, both drugs were associated with transcriptional programmes related to synaptic plasticity and excitatory neurotransmission in excitatory neurons, but the time course differed. Psilocybin showed a biphasic pattern, with early and late waves of gene expression, whereas ketamine produced fewer differentially expressed genes that peaked at about 2 to 4 hours after administration. The paper reports that ketamine analyses were also carried out across the 16 annotated cell types, and that convergent biological pathways were observed despite the differing temporal dynamics.
Discussion
The authors interpret the study as showing that single-nucleus transcriptomics can capture cell type-specific and time-dependent gene expression changes after psilocybin and ketamine in the mouse medial frontal cortex. They argue that the dataset provides a resource for studying how these compounds affect frontal cortical cell populations, and that the observed early and late phases of transcriptional response to psilocybin are consistent with its rapid brain exposure and clearance. They suggest that multiple waves of transcription may be a common feature of perturbations and note that the later phase could reflect longer-term processes such as alternative splicing or epigenomic change, although this is presented as speculation rather than demonstrated mechanism. Relative to earlier research, the authors state that many of the pathways they identified are consistent with previous findings on psychedelic- and ketamine-evoked structural plasticity, especially genes involved in synaptic function, glutamatergic transmission, and cytoskeletal regulation. They also relate their results to earlier reports of immediate early gene induction after psychedelic exposure, but note that their snRNA-seq approach detected limited significant changes in genes such as Fos, Jun, Npas4, Arc, Egr1, and Egr2. They suggest that this discrepancy may reflect methodological differences, because nuclear RNA may be more transient and harder to detect than total cellular RNA or protein in this setting. The authors acknowledge several limitations. A major one is that the psilocybin- and ketamine-treated groups were compared with non-injected controls rather than vehicle-injected controls at each time point, so some transcriptional effects could theoretically reflect injection stress. They argue that this is unlikely to explain the main findings because the two drug datasets showed clearly different temporal profiles, and they increased the control cohort size to strengthen the reference dataset. They also note that single-nucleus rather than single-cell sequencing was chosen to avoid damaging cortical neurons with extensive processes, but that this approach may have lower sensitivity for some activity-dependent transcripts. Finally, although sex was treated as a biological variable, the dataset was not powered to analyse sex-specific transcriptional responses rigorously. In terms of implications, the authors present the dataset as a foundation for future studies on the molecular and cellular actions of psilocybin and ketamine, with the broader aim of helping to clarify mechanisms that might guide the development of safer and more effective treatments for mental illness.
View full paper sections
SINGLE-NUCLEUS RNA SEQUENCING OF THE MOUSE MEDIAL FRONTAL CORTEX AT VARIOUS TIMEPOINTS AFTER A SINGLE DOSE OF PSILOCYBIN
In this study, we focus on the mouse dorsal medial frontal cortex, which encompasses the anterior cingulate cortex (ACAd) and the premotor cortex (medial MOs). This region is important for the drug action of psilocybin and ketamine, showing robust c-Fos response in a whole-brain mapping study, corroborating reports of drug-evoked structural plasticity after psilocybin 9 or ketamine. We administered C57BL/6J mice with psilocybin (1 mg/kg, i.p.) or ketamine (10 mg/kg, i.p.), and then collected tissue corresponding to the dorsal medial frontal cortex of both hemispheres via microdissection at 1, 2, 4, 24, or 72 hours after drug administration (Fig.). Additionally, we collected control tissue from animals that received no injection. The tissue was processed through nuclear dissociation, nuclei sorting, and barcoding for snRNA-seq via the 10x Genomics Chromium Single Cell 3' v3 platform. We planned for and processed samples from 49 animals, including 4 for each time point and 9 controls, but excluded 4 animals due to incorrect sequencing depth specification and 3 animals due to low number or low-quality nuclei (see Methods; Supplementary Table). For the remaining 42 samples, we started with 145,593 cells for the psilocybin group, 139,628 cells for the ketamine group, and 67,600 cells for the controls. We targeted the sequencing to 25,000 reads per nucleus for each sample. Initial processing confirmed an average of 29,562 reads per nucleus. We removed cells that did not Vip, vasointestinal peptide-expressing interneurons. Lamp5, Lamp5-expressing interneurons. Pvalb, parvalbuminexpressing interneurons. Sst, somatostatin-expressing interneurons. Astro, astrocytes. OPC, oligodendrocyte progenitor cells. Oligo, oligodendrocytes. Microglia, microglia. Sncg, Sncg-expressing interneurons. Endo, endothelial cells. Other, other annotated cell types. We annotated cells based on the Allen Institute mouse brain taxonomy using the MapMyCells tool, because the standardized cell types enabled interoperability with other datasets (Fig.). As expected, these cell types displayed distinct expression profiles (Fig.). The same cell-type clusters were detected in samples collected at different time points after psilocybin administration, ketamine administration, and in controls (Supplementary Fig.). Most cells were classified with nearperfect confidence (Supplementary Fig.), and similar groupings could be identified using an unsupervised clustering approach (Supplementary Fig.). To facilitate access and exploration of the dataset, we developed an interactive web portal () that allows users to visualize the data and download either selected subsets or the entire dataset for independent analyses. While the resource includes both psilocybin and ketamine conditions, we will focus the subsequent analyses primarily on psilocybin to illustrate the biological insights that can be derived from this resource.
CELL TYPE-SPECIFIC EXPRESSION OF SEROTONIN RECEPTOR TRANSCRIPTS IN THE MOUSE MEDIAL FRONTAL CORTEX
Following administration, psilocybin is rapidly metabolized to psilocin, which enters the brain. Psilocin binds not only to the 5-HT 2A receptor, but also to other serotonin receptor subtypes expressed in the neocortex, including the 5-HT 1A and 5-HT 2C receptors. We visualized the presence of the Htr1a, Htr2a, and Htr2c transcripts which encode the 5-HT 1A , 5-HT 2A , and 5-HT 2C receptors respectively in the control samples in our snRNA-seq dataset (Fig.). The expression patterns of the three transcripts clearly differed, prompting us to quantify how the transcript levels vary across the 16 cell types (Fig.). Because our cell types were annotated based on Allen Institute mouse brain taxonomy, the results can be easily compared with many other existing datasets. To illustrate interoperability, we extracted SMART-Seq single-cell sequencing data from the Allen Institute, focusing on cells residing in the mouse frontal cortex, including the ACA, ALM, ORB, and PL-ILA regions (Fig.). There was strong agreement between the two datasets. The IT and ET excitatory cell types express a high level of Htr2a as well as moderate amounts of Htr1a and Htr2c. Interestingly, there is also an abundance of Htr2a transcripts in the deep-lying layer 6b neurons, which are known for their excitatory effects on apical dendrites and higher-order thalamus. Unlike excitatory neurons, the expression profile is more selective in the GABAergic cell subtypes. The Lamp5 and Pvalb subtypes express Htr2a, whereas the Vip subpopulation contains Htr2c. Although serotonin receptors have been reported in microgliaand astrocytes, the current data suggest that Htr transcript levels are relatively low in the non-neuronal cells. Altogether, these analyses reveal a highly selective pattern of serotonin receptor expression in cortical cell types. Because psilocin is a non-selective serotonergic agonist that binds to a range of receptor subtypes, we anticipate that the drug can directly influence the transcriptional programs of most major excitatory and GABAergic cell types, while indirectly affecting other cell populations.
PSILOCYBIN-INDUCED TRANSCRIPTIONAL CHANGES IN FRONTAL CORTICAL EXCITATORY NEURONS ARE INVOLVED IN SYNAPTIC PLASTICITY
To identify differentially expressed genes (DEGs), we used pseudobulk analysis where sequence reads were summed for each cell type and biological replicate (see Methods). To gain insights into the biological functions associated with the psilocybin-induced transcriptional changes, we performed gene ontology (GO) enrichment analysis on the upregulated DEGs identified in IT neurons across time points. Fig.shows the top 15 overrepresented GO terms when we considered the prevalence of each term averaged across all time points and for all four cell types. The top categories are processes related to synaptic plasticity and excitatory neurotransmission, including chemical synaptic transmission, regulation of neuron projection development, excitatory chemical synaptic transmission, and regulation of neuronal synaptic plasticity. Also notable are pathways related to protein synthesis and intracellular trafficking, such as intracellular protein transport, positive regulation of protein localization to nucleus, establishment of protein localization to organelle, and protein import into nucleus. To illustrate the genes contributing to these enriched pathways, we visualized select upregulated genes including Mink1, Sipa1l1, Camk1g, and Shisa6 (Fig.). Mink1 encodes a kinase that maintains dendrite complexity by regulating actin cytoskeletal dynamics and AMPA receptor trafficking. Sipa1l1 is transcript for the SPAR1 protein, which localizes to the postsynaptic density and complexes with PSD-95 and NMDA receptors and contributes to spine head enlargement. Camk1g encodes a component of a calcium-and calmodulin-dependent protein kinase complex involved in dendritic growth. Select genes were upregulated over more extended durations after psilocybin, such as Shisa6, which contributes to anchoring AMPA receptors into postsynaptic compartments in dendrites. We note that users of the web portal can generate similar expression plots for any genes of interest.
PSEUDOBULK ANALYSIS HAS BEEN
We next examine DEGs in the other excitatory neuron subtypes, including L5 ET, L6b, L6 CT, and L5 NP neurons (Fig.). These excitatory cell populations also exhibit a biphasic pattern of transcriptional changes, with an initial peak at 1 hour and then late response at 72 hours after psilocybin administration (Fig.). Gene ontology analysis revealed similar biological themes, including the same top enriched pathway of chemical synaptic transmission, as well as related yet distinct processes such as glutamatergic synaptic transmission and glutamate receptor signaling pathway (Fig.). We plotted the expression profiles for several upregulated genes (Fig.), such as Nos1ap, which binds neuronal nitric oxide synthase and plays a role in dendrite growth and NMDA receptor-dependent responses. There are also genes with more extended responses including Grin2a and Gria1, which encode the NR2A subunit of NMDA receptor and GluA1 subunit of AMPA receptor, respectively, with clear roles in mediating glutamatergic transmission and excitatory synaptic plasticity. Sorcs3 encodes a neuronal receptor that regulates intracellular protein transportand has also been implicated in synaptic plasticity.
PSILOCYBIN-INDUCED TRANSCRIPTIONAL CHANGES IN FRONTAL CORTICAL GABAERGIC NEURONS CONTRIBUTE TO MITOCHONDRIAL FUNCTION AND METABOLIC PROCESSES
For GABAergic neurons, we characterized the transcriptional responses in Vip, Lamp5, Pvalb, and Sst cell types (Fig.). The magnitude of the response varied across interneuron subtypes. At 1 hour after psilocybin administration, robust responses were detected in Pvalb and Sst interneurons, with 933 and 662 DEGs respectively, but changes in Vip and Lamp5 interneurons were comparatively modest, with 154 and 80 DEGs (Fig.). Unlike excitatory neurons, GO enrichment analysis revealed that DEGs in GABAergic neurons had enriched pathways related to mitochondrial function and metabolic processes, such as aerobic electron transport chain, mitochondrial ATP synthesis coupled electron transport, and canonical glycolysis (Fig.). We plotted select genes associated with the top pathways, including Cox6c, Ndufs2, Kcnq3, and Amigo1 (Fig.). Cox6c and Ndufs2 encode key components of cytochrome c oxidase in the mitochondrial electron transport chainand mitochondrial complex I, respectively. Kcnq3 and Amigo1 encode the subunits of a voltage-gated potassium channel to regulate neuronal excitability. Particularly abundant in Pvalb interneurons, Kcnq3 and Amigo1 were highlighted in the top enriched process of monoatomic cation transmembrane transport, possibly linking the use of metabolic energy to regulate cellular excitability. Overall, the analyses indicate that psilocybin engages transcriptional programs in frontal cortical GABAergic neurons that are associated with mitochondrial function and metabolic processes, distinguishing their response from the synaptic plasticity-related programs observed in excitatory neurons.
PSILOCYBIN-INDUCED TRANSCRIPTIONAL CHANGES IN NON-NEURONAL CELLS
Our analysis of control samples showed negligible levels of serotonin receptor transcripts in non-neuronal cells in the mouse medial frontal cortex (Fig.), yet surprisingly we found that they displayed transcriptional responses to psilocybin. The most responsive non-neuron populations were astrocytes and oligodendrocytes, with 776 and 789 DEGs at 1 hour after administration respectively (Fig.). There were fewer but still notable changes involving 202 and 64 DEGs for oligodendrocyte precursor cells and microglia. GO analysis revealed pathways related to neuronal connectivity, such as neuron projection guidance, axon guidance, axonogenesis, and chemical synaptic transmission (Fig.). As exemplars, we plotted the time courses for Nlgn3, Lrp1, Aldoc, and Lars2 (Fig.). Nlgn3 was upregulated by psilocybin particularly in oligodendrocytes and oligodendrocyte precursor cells, where it plays a crucial role in their differentiation and myelination. Lrp1 encodes a multicargo transporter that interacts with insulin-like growth factor to promote survival and proliferation, including in astrocytes. Aldoc is the transcript for aldolase, which acts as a checkpoint for proliferation of oligodendrocyte precursor cells. Lars2 contributes to mitochondrial regulation and metabolic homeostasis and, interestingly, shows notable response preferentially in the late phase at 72 h. The results demonstrate that psilocybin induces transcriptional responses in non-neuronal cells in the medial frontal cortex, likely indirectly via circuit-level mechanisms.
COMPARING THE TIME-DEPENDENT AND CELL TYPE-SPECIFIC TRANSCRIPTIONAL RESPONSES BETWEEN PSILOCYBIN AND KETAMINE
To facilitate comparisons across cell types and time points, we summarized the number of upregulated and downregulated DEGs following psilocybin administration (Fig.). Psilocybininduced transcriptional responses exhibited a biphasic temporal profile, consisting of an early wave of gene expression at 1 hour followed by a late response at 72 hours after drug administration. This visualization also highlights excitatory neurons as the cell class showing the highest number of DEGs, relative to the more modest number of DEGs in select GABAergic and non-neuronal cell types. Interestingly, when we performed a similar analysis for ketaminetreated animals, a different time course emerged. In contrast to psilocybin, ketamine induced fewer DEGs, which peaked at around 2 and 4 hours after administration (Fig.). For completeness, we performed the differential expression analyses including gene ontology enrichment analyses for ketamine-treated animals as well, for all 16 annotated cell types including the IT excitatory neuron subtypes (Supplementary Fig.), other excitatory neuron subtypes (Supplementary Fig.), inhibitory neuron subtypes (Supplementary Fig.), and non-neuronal cell types (Supplementary Fig.). Despite the different temporal dynamics, we observed examples of convergent biological pathways engaged by both ketamine and psilocybin. For instance, in excitatory neuron subtypes, the top enriched GO terms for the ketamine samples were chemical synaptic transmission, anterograde trans-synaptic signaling, and glutamatergic synaptic transmission (Fig.). These findings indicate that both compounds, despite acting through distinct receptors, induce transcriptional programs associated with synaptic plasticity and excitatory neurotransmission.
DISCUSSION
The present study applied single-nucleus transcriptomics to determine how psilocybin and ketamine alter gene expression in the mouse medial frontal cortex. As a resource, we annotated the dataset according to 8 major excitatory neuron types, 4 GABAergic cell types, and 4 nonneuronal cell types. We developed an interactive web portal () to facilitate data exploration. As examples of the insights enabled by this resource, we show that excitatory neurons, GABAergic neurons, and non-neuronal cells exhibit differentially expressed genes associated with distinct biological processes. We also find that psilocybin's impact is time-dependent with early and late phases of transcriptional responses, which was distinct from ketamine's time course. Together, the dataset provides a resource that unveils the dynamics of transcriptional responses in frontal cortical cell types after a single dose of psilocybin or ketamine. The bolus injection of psilocybin causes rapid elevation of drug concentration in the central nervous system followed by full clearance within a few hours. This pharmacokinetic profile hints at a temporal pattern of transcriptional changes, and here we identified two distinct phases: an early phase at 1 -2 hours followed by a late phase at 72 hours, each marked by a substantial number of differentially expressed genes. Indeed, a couple of studies have reported long-lasting changes in gene expression after the administration of psychedelics for one or few weeks. Our findings echo the framework that there are often multiple waves of transcription following a perturbation. For example, there are early and late response genes in visual cortical cells in dark-reared animals after their first experience of visual stimuli. It is also possible that the longterm changes in gene expression reflect effects of psychedelics on alternative splicing or epigenomic modifications. It is tempting to speculate that the multiple phases of transcriptional responses observed in this study may reflect therapeutic mechanisms for ketamine and psilocybin at different time scales. Notwithstanding the new insights into the cell type specificity and timing of drug-evoked transcriptional responses, several of the differentially expressed genes and pathways identified here are consistent with prior studies. Regulation of transcripts associated with synaptic plasticity and glutamatergic transmission is expected given previous reports of psilocybin-and ketamine-evoked structural plasticity in the medial frontal cortex. Notably, a pioneering study examining the effects of LSD on rat frontal cortex, which profiled about 3,000 genes, reported preferential regulation of genes involved in synaptic function, glutamatergic signaling, and cytoskeletal plasticity. Mitochondrial biogenesis has also been linked to serotonergic signaling, particularly through the activation of 5-HT 2A receptors. It is worth noting that multiple studies have reported increased expression of immediate early genes such as Fos and Arc following psychedelic administration, based on transcript quantificationsor whole-brain immunohistochemistry. While we examined Fos, Jun, Npas4, Arc, Egr1, and Egr2 in our dataset, only Arc and Egr1 in astrocytes showed significant upregulation for psilocybin-treated animals at the 1-hour time point (Supplementary Fig.) and none of these genes reached significance for ketamine at 1 hour (Supplementary Fig.). The discrepancy could be due to methodological differences: our approach measures nuclear RNA, which exhibits more rapid and transient dynamics than total cellular RNA or protein. Previous work comparing activitydependent transcriptional changes across measurement techniques found that such changes are more difficult to detect using snRNA-seq. A weakness of this study is that the psilocybin-and ketamine-treated groups were compared against controls that did not receive an intraperitoneal injection. The injection itself can induce transcriptional responses, particularly in stress-sensitive cell types and brain regions. Ideally, the study would have included vehicle-injected controls collected at each corresponding time point (1, 2, 4, 24, and 72 hours). However, the large number of additional samples required for such a design was prohibitive given the costs. To mitigate this weakness, we increased the size of the control cohort, anticipating the importance of having a robust reference dataset for differential expression analyses. In addition, head-to-head comparison of the psilocybin and ketamine datasets revealed markedly different temporal patterns of transcriptional responses (Fig.). If the injection procedure were a major driver of the observed gene expression changes, we would expect the two datasets to exhibit similar temporal profiles. Therefore, any contribution of the injection protocol is likely modest relative to the drug-specific transcriptional effects. There are limitations for this study. Transcriptomics is a rapidly growing field with numerous approaches for sequencing and analysis. Here, we chose to work with single nuclei instead of single cells, because we were concerned that cortical neurons with extended axons and dendrites may be damaged in protocols needed to dissociate single cells, which can lead to aberrant readout of gene expression. Sequencing with single nuclei may better reflect recent changes in de novo transcription, and has been shown to yield excellent sensitivity and reliable classification of cell types. Moreover, previous studies have identified sex as an important biological variable that can influence the behavioral effects of psychedelics. However, the current dataset was not sufficiently powered to support a rigorous analysis of sexspecific transcriptional response. In summary, this dataset serves as a resource for investigating the transcriptional responses to psilocybin and ketamine across frontal cortical cell types. We anticipate the growing number of studies revealing the molecular and cellular impact of psilocybin and ketamine will eventually reveal mechanisms of action that can accelerate the development of more effective and safer drugs for treating mental illnesses.
ANIMALS.
Animal care and experimental procedures were approved by the Animal Care & Use Committee (IACUC) at Yale University and Cornell University. C57BL/6J (Stock No. 000664) mice were purchased from Jackson Laboratory and habituated in our animal facility at Yale University for 1 week. Animals were housed in same-sex groups of 2-5 mice per cage in a temperature-controlled room, operating in 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. 8-week-old animals were injected with psilocybin (1 mg/kg, i.p, prepared from working solution, which was made fresh monthly from powder; Usona Institute) or ketamine (10 mg/kg, i.p; Zoetis), returned to their home cage, and used 1, 2, 4, 24, or 72 hours after injection. Additional age-matched mice were used with no drug administration as controls. More details for the animals, including their sex and assignment to each treatment group, are included in Supplementary Table.
TISSUE COLLECTION AND GENERATION OF SINGLE-CELL SUSPENSIONS.
For tissue collection, decapitation immediately followed cervical dislocation and brains were immediately dissected from the skull. The dorsal medial frontal cortex, consisting of ACAd and medial MOs, was microdissected in cold RNAse-free 1x PBS, flash frozen in ethanol and dry ice, and stored at -80°C. Samples were shipped overnight on dry ice to the University of Michigan for processing. Nuclei were isolated and processed for single-nucleus sequencing using RNase-free materials and reagents. Fresh buffers were made immediately before nuclei processing. Buffers were made in accordance with recommended protocol (CG000393 Rev A, "Nuclei isolation from adult mouse brain tissue for single cell RNA sequencing", 10x Genomics). Lysis buffer consisted of 10 mM Tris-HCl ph7.4 (Millipore-Sigma T2194), 10 mM NaCl (Millipore-Sigma 59222C), 3 mM MgCl 2 (Millipore-Sigma, M1028), 0.1% Nonidet P40 (Millipore-Sigma 74385), and nuclease-free water (Invitrogen AM9932). Wash buffer consisted of 1% BSA solution (Thermo Fisher Scientific AM2616), 0.2U/µl RNase Inhibitor (Millipore-Sigma 3335399001), and 1x PBS (Gibco, 10010023). Tissue was dounce homogenized in 1 mL lysis buffer and centrifuged at 500g for 5 minutes at 4°C. Supernatant was removed and samples were resuspended in 1mL of wash buffer. Samples were strained through 40 µm cell strainers (Bel-Art H13680-0040) and centrifuged again at 500 g for 5 minutes at 4°C. Supernatant was removed and samples were resuspended in 1mL of wash buffer and passed through 40 (version 0.12.7) and scanpy(version 1.11.5). Filtered feature-barcode matrices for all samples and all drug treatments were imported to AnnData objects and concatenated using anndata.concat() on the intersection of genes. Out of the 49 samples, 7 were excluded to yield 42 samples for final analysis. Sample IDs 39, 40, 41, and 42 were excluded for oversequencing, because we incorrectly ordered for a targeted depth of 250,000 reads per nucleus at the facility. This introduced undesirable technical effects that could not be removed despite subsetting and batch integration attempts. Sample ID 290 was excluded due to low cell number, containing only 968 cells after sequencing compared to the median of 8,216. Sample ID 299 was excluded because it had the second lowest cell count at 3,372, was flagged for a high fraction of reads mapping antisense to genes by the cellranger pipeline, and analysis showed little transcriptional separation among cells. Sample 301 was excluded because its nuclei formed distinct tails extending from robust clusters in the UMAP embedding. Although the exact cause was unclear, this could be due to low-quality nuclei with reduced RNA content. Analysis with the cellranger pipeline confirmed a mean depth of 29,562 reads per nucleus across all samples, consistent with the targeted read depth. Most samples exceeded the target depth of 25,000 reads per nucleus. Nine samples had sequencing depths below 25,000 reads per nucleus, with the lowest being 20,694 reads per nucleus. Thus, all samples exceeded the threshold of 20,000 reads per nucleus recommended by 10x Genomics. We also examined sequencing saturation, which ranged from 35-65% across samples. Then we proceeded to process the data through several standard steps in cellranger. First, only reads that map confidently in the sense orientation to a single gene are retained. Second, for deduplication, reads sharing the same cell barcode and unique molecular identifier (UMI) are collapsed into a single transcript count because they originate from PCR amplification of the same RNA molecule. psilocybin-treated samples had 8482±1086 reads per nucleus (mean ± standard deviation), ketamine-treated samples had 8812±1436 reads per nucleus, and control samples had 8343±602 reads per nucleus. Next, we performed additional filtering for quality control. All genes detected in fewer than 5 unique nuclei or having fewer than 5 total transcript counts were removed. All nuclei with fewer than 3 unique genes detected or having fewer than 3 total transcript counts were removed. Multiplet nuclei were identified and removed with scanpy.pp.scrublet including a batch key identifying individual samples. 13 samples were identified as homogenous, where scrublet identified few or no multiplets. To perform multiplet removal in these samples, the largest multiplet score threshold from the non-homogenous samples was identified and applied to the homogeneous samples. Nuclei with low transcript counts on a per-sample basis (<1000 counts and less than 3 mean absolute deviations below the sample median) were removed. Cells with high mitochondrial or ribosomal counts on a per-sample basis (>20% of counts or more than 3 mean absolute deviations above the sample median) were removed. Ribosomal genes were based on the Ribosomal Protein Gene Database 76 (). The ubiquitously and highly expressed Malat1 gene was removed from analysis. At the end, psilocybin-treated samples contained 8,049±1,115 usable reads per nucleus (mean ± standard deviation), ketamine-treated samples contained 8,414±1,588 usable reads per nucleus, and control samples contained 7,689±826 usable reads per nucleus. Each cell's reads were normalized to counts per million (CPM) with scanpy.pp.normalize_total, and a natural logarithm was taken with scanpy.pp.log1p. This yielded ln(CPM+1), which was used for all downstream expression analysis except pseudobulk. We identified the top 5000 highly variable genes using scanpy.pp.highly_variable_genes with flavor='seurat'. Batch corrected embedding was performed with scVI 78 version 1.4.1 on these highly variable genes from raw RNA counts. The embedding model used default settings, having 10 latent dimensions, one 128-dimensional hidden layer in both the encoder and decoder, 10% dropout rate, and zero-inflated negative binomial likelihood. Nearest neighbors and UMAP was run on the learned batch-corrected embeddings with scanpy default settings. Cell types were annotated using the Allen Brain Atlas MapMyCells toolSncg Gaba and 333 Endo NN, and therefore they were excluded from further analysis. Cells with annotations not included in this list were marked as "Other" and also excluded from further analysis. Marker genes for the remaining cell types were verified with scanpy.tl.rank_genes_groups function and the Wilcoxon rank-sum test, and plotted with scanpy.pl.heatmap.
SEROTONIN RECEPTOR TRANSCRIPT EXPRESSION.
To examine cell type-specific expression patterns of psilocybin's 5-HT receptor targets, we extracted Htr1a, Htr2a, and Htr2c transcript levels from the control samples in our snRNA-seq dataset. Cell type-specific patterns were visualized on a per-cell basis on the dataset's UMAP representation. Group-wise variation among distinct cell subtypes was visualized with violin plots. UMAPs were created using matplotlib and the scanpy UMAP embeddings of the dataset, while violin plots were constructed using scanpy.pl.violin(). The scale of expression was unified across all genes by identifying the maximum expression level. Transcript levels from our snRNA-seq dataset were compared to publicly available single-cell sequencing dataset from the Allen Institute for Brain Science (map.org/atlases-and-data/rnaseq/mouse-whole-cortex-and-hippocampus-smart-seq). Briefly, this independent dataset used SMARTseq-v4 to analyze 76,381 single cell transcriptomes across the mouse cortex and hippocampus, and cluster these cells into transcriptional cell subtypesand 84 Micro_PVM, as identified and annotated already by the Allen Institute. Transcript levels for 5-HT receptors (Htr1a, Htr1b, Htr1d, Htr2a, Htr2b, Htr2c, Htr3a, Htr3b, Htr4, Htr5a, Htr5b, Htr6, Htr7), as well as cell type markers for excitatory (Slc17a7) or inhibitory (Gad1, Vip, Ndnf, Pvalb, Sst) neurons were extracted for each cell subtype. Expression levels were depicted in a dot plot showing log 2 (CPM+1) per cell type, with dot size scaled by the percentage of cells of that type with non-zero transcript reads. Processing and visualization of Allen Institute data were done using MATLAB scripts.
DIFFERENTIAL GENE EXPRESSION USING PSEUDOBULK.
We performed cell type-specific differential expression analysis using pseudobulk techniques to identify genes of interest while minimizing the risk of false discoveries. For pseudobulk, raw transcript counts were sum aggregated within each sample-cell type combination using ADPBulk version 0.1.4. Differential expression testing was performed with pyDESeq2 80 version 0.5.3 with fold change shrinkage. Genes with less than 10 total counts were excluded for the given comparison. Differential expression was measured with log 2 (fold-change) and p-values were adjusted false discovery rate (FDR) correction. Shrinkage was applied to the log 2 (foldchange) values with pyDEseq2. P-value correction was applied with the Benjamini-Hochberg procedure to control the false discovery rate. Genes were identified as significant with an FDR below 5%. plotnine version 0.15.2 was used to create volcano plots. Genes with log 2 (foldchange) or -log 1 0(p-value) beyond the bounds of the plot area were clamped to be displayed on the axis limits. Gene ontology analysis. Gene ontology (GO) analysis was performed using Enrichrand Gene Ontology Biological Processes 2026 database to identify biological processes that were overrepresented for genes that were differentially expressed. For each cell type at each time point, if there were ≥5 upregulated DEGs, then the list of upregulated DEGs was analyzed using Enrichr. It is important to perform this analysis against an appropriate set of background genes. If we used all genes in the genome, the results can be misleading because, for example, excitatory neurons already express many neuronal and synaptic genes at baseline, and they will show up as spuriously enriched if the background is set to all mouse genes. For this reason, we first determine the background genes for each cell type, by taking the output from DESeq2 and filtered the expressed genes, defined as having baseMean > 10 at any time point or drug condition. Compared to the total number of 26,100 genes tested in DESeq2, this step produced cell type-specific background lists with ~8,000 -15,000 genes (specifically, L6 IT: 12693 genes; Computer code to process and analyze the snRNA-seq data is provided as Snakemake scripts, Jupyter notebooks, and R scripts on GitHub ().
Full Text PDF
Full Paper PDF
Create a free account to open full-text PDFs.
Study Details
- Study Typeindividual
- Populationrodents
- Characteristicsbrain measures
- Journal
- Compounds
- Topic
- Authors
- APA Citation
References (9)
Papers cited by this study that are also in Blossom
Goodwin, G. M., Aaronson, S. T., Alvarez, O. et al. · New England Journal of Medicine (2022)
Raison, C. L., Sanacora, G., Woolley, J. D. et al. · JAMA (2023)
Berman, R. M., Cappiello, A., Anand, A. et al. · Biological Psychiatry (2000)
Liao, C., Dua, A. N., Wojtasiewicz, C. et al. · Nature Reviews Neuroscience (2024)
Shao, L-X,, Liao, C., Gregg, I. et al. · Neuron (2021)
Nardou, R., Sawyer, E., Song, Y. J. et al. · Nature (2023)
Madsen, M. K., Fisher, P. M., Burmester, D. et al. · Neuropsychopharmacology (2019)
Savalia, N., Shao, L-X,, Kwan, A. C. · Trends in Neuroscience (2021)
De La, M., Revenga, F., Zhu, B. et al. · Cell Reports (2021)