Abstract
Endocannabinoid signaling exerts a neurodevelopmental regulatory role via CB1 cannabinoid receptors (CB1Rs), which control pyramidal neuron differentiation, migration, and axonal guidance. Here, we investigated the long-lasting consequences of transient prenatal CB1R downregulation within the mouse prefrontal cortex by assessing its impact on gene expression, neuronal electrophysiological properties, and animal behavioral traits. Transient loss of CB1Rs as induced by in-utero small-interference RNA electroporation at Embryonic Day 14.5, when upper-layer neurons are generated, arrested cell migration leading to ectopic neurons that populated deep layers. Whole-cell current–clamp recordings showed that ectopic neurons are less excitable (increased afterhyperpolarization amplitude, decreased sag, lower firing frequency) than deep-layer–native pyramidal neurons. Differentially expressed genes (DEGs), identified by microarray characterization of FACS-sorted electroporated neurons, were significantly enriched in pathways related to cortical development, regulation of cell migration, neurotransmitter secretion, and cytoskeletal organization. Gene set enrichment analysis also supported enrichment in pathways associated with neurodegenerative disorders and synaptic function. The gene expression profile of siCB1R-derived neurons showed DEGs that had been previously associated with intellectual disability, schizophrenia, and autism. Venn diagrams unveiled one common DEG for neuropsychiatric risk databases and CB1R expression manipulation, namely, the transcription factor ZBTB20. Prenatal knockdown of CB1Rs induced long-lasting behavioral alterations in the adult offspring of either sex, with an impairment of social interaction and motor behavior in siCB1R-derived adult mice. Taken together, these findings highlight the role of CB1Rs in controlling the development of pyramidal neurons in the prefrontal cortex and support the contribution of altered endocannabinoid signaling to neuropsychiatric vulnerability.
Significance Statement
Transient loss of function of CB1 cannabinoid receptors (CB1Rs) during prenatal development of the mouse prefrontal cortex leads to pyramidal neuron migration arrest, altered gene expression program, abnormal intrinsic electrophysiological properties, and deficits in social interaction. These findings underscore the critical role of the endocannabinoid system in cortical pyramidal neuron development and highlight similarities between the molecular signature induced by a prenatal transient loss of CB1Rs and the vulnerability to neuropsychiatric disorders. Given that Δ9-tetrahydrocannabinol, the primary psychoactive compound in Cannabis preparations, acts in the brain via CB1Rs, these results also provide important insights into the neuronal alterations associated with prenatal cannabinoid exposure.
Introduction
The prenatal developing brain is extremely sensitive to genetic variants and environmental stressors that can interfere with the tightly regulated neurodevelopmental program (Parenti et al., 2020). Dysregulated neuronal differentiation, maturation, and connectivity set the ground for neurodevelopmental disorders, which can manifest themselves preferentially by psychiatric alterations [e.g., cognitive deficits, social behavior impairments, anxiety, autism spectrum disorder (ASD), schizophrenia (SCZ), and intellectual disability (ID)] or with neurological symptoms (e.g., neuronal hyperexcitability and seizures) in epileptogenic syndromes. The endocannabinoid (ECB) system, mostly via CB1 cannabinoid receptors (CB1Rs), exerts a neurodevelopmental regulatory role controlling neural progenitor cell proliferation and identity, axonal guidance and fasciculation, and neuronal migration (Bara et al., 2021; Galve-Roperh et al., 2022; Rodrigues et al., 2024). Thus, during cortical development, in vivo manipulation of ECB signaling either by conditional ablation of CB1Rs, of ECB-metabolizing enzymes, or by prenatal cannabinoid exposure (PCE) leads to deficits of long-range projection neuron and interneuron development, synaptogenesis, and neuronal circuit architecture (Tortoriello et al., 2014; Vargish et al., 2016; Díaz-Alonso et al., 2017; de Salas-Quiroga et al., 2020). CB1R activity regulates important signaling pathways involved in the control of neuronal cell fate, including the ERK and PI3K/mTORC1 cascades, which, in turn, modify the activity of transcription factors involved in neurogenesis (e.g., PAX6/EOMES and BCL11B/SATB2; Bromberg et al., 2008; Díaz-Alonso et al., 2012, 2015; Alpár et al., 2014). Also, CB1R regulates epigenetic DNA status (e.g., via KMT2a dysregulation) a mechanism that can explain the persistent consequences of PCE and intergenerational transmission of synaptic and behavioral changes (Bara et al., 2021). Specifically, PCE-induced changes in DNA methylation appear to be enriched in ASD-related genes (Shorey-Kendrick et al., 2023). In this regard, the impact of PCE on the offspring is mediated by the phytocannabinoid THC, which, acting on CB1Rs, disrupts the neurodevelopmental functions of the ECB system (Bara et al., 2021; Galve-Roperh et al., 2022; Rodrigues et al., 2024). Consequently, dysregulation of ECB signaling at key time points of prenatal cortical development induces changes in neuronal development, positioning, and maturation, which contribute to psychiatric vulnerability, social, cognitive, and motor alterations and frequently enhanced seizure susceptibility.
A predominant neurobiological mechanism of action of embryonic cannabinoid signaling is its involvement in the cholecystokinin (CCK) expressing interneuron population lineage (Berghuis et al., 2007). Thus, PCE by interfering with CB1R signaling induces CCK basket cell interneuronopathy associated with chronic inhibitory deficits and altered hippocampal high-frequency oscillations that contribute to impaired social behavior and spatial cognitive deficits (Vargish et al., 2016; de Salas-Quiroga et al., 2020). In addition to neuropsychiatric traits induced by interneuron deficits, dysregulated prenatal cannabinoid signaling also induces seizure susceptibility acting on the CCK basket neuronal lineage (Berghuis et al., 2007; Morozov et al., 2009; de Salas-Quiroga et al., 2015, 2020). However, the regulatory role of the ECB system in pyramidal neuron development, controlling the balance between upper- and deep-layer neuronal populations and long-range axon projection (Mulder et al., 2008; de Salas-Quiroga et al., 2015; Bara et al., 2018; Miller et al., 2018; Navarri et al., 2024), shall also be considered. Alterations of upper- and deep-layer pyramidal neuronal balance contribute to the etiopathology of neuropsychiatric disorders like ASD, SCZ, and ID (Willsey et al., 2022).
Here we focused on the study of the impact of transient embryonic CB1R knockdown in the mouse pyramidal neuron PFC development, owing to its crucial role in high-order cognitive functions that are altered in neuropsychiatric disorders (Bicks et al., 2015). Importantly, the expression of the ECB system elements is dynamically regulated in the developing PFC, suggesting specific roles of ECB signaling in cortical maturation (Long et al., 2012; Tseng and Molla, 2025). By conducting gene expression analyses, combined with electrophysiological and behavioral studies, we demonstrate that downregulation of prenatal CB1R signaling leads to social interaction deficits, and a gene expression signature that closely overlaps with reported ID-associated signatures. Our findings contribute to a better understanding of the role of dysregulated ECB signaling in the pathophysiology of neuropsychiatric disorders by a discrete neuronal population.
Materials and Methods
Animals
Experimental procedures were performed following the guidelines and approval of the Animal Welfare Committees of Universidad Complutense de Madrid and the University of Amsterdam, Comunidad de Madrid, the directives of the Spanish Government and the European Commission (RD 118/2021). All efforts were made to minimize the number of animals used and their suffering throughout the experiments. Mouse embryonic tissues were obtained upon timed mating as assessed by vaginal plug observation [Embryonic Day (E)0.5]. Throughout the study, animals had unrestricted access to food and water. They were housed (typically, 4–5 mice per cage) under controlled temperature (range, 20–22°C), humidity (range, 50–55%), inverted light/dark cycle (12/12 h), and with environmental enrichment.
In utero electroporation
Pregnant C57BL/6J wild-type mice (Janvier-Labs) were subject to in utero bilateral electroporation (IUE) at E14.5 using three electrodes, specifically targeting the developing prefrontal region of the brain. This method has demonstrated its efficacy to manipulate this region's neurons and their involvement in neuropsychiatric disorders (Szczurkowska et al., 2016). The electric field was generated by two lateral negative poles and one positive pole placed in the rostral area, at the most frontal portion of the cerebral ventricles. IUE targets dividing progenitor cells and, depending on the timing of the procedure, enables selective targeting of distinct neuronal progenitor waves that will populate specific cortical layers. IUE allows gene expression, silencing, or editing in a sparse population of neurons within a naive environment, thereby facilitating the study of cell-autonomous signaling mechanisms and behavioral regulation. Neural progenitors that will develop into upper-layer 2/3 pyramidal neurons of the prefrontal cortex were electroporated with either CB1R or control siRNA (final concentration 10 μm, referred to as siCB1R and scrambled siCo, respectively; Dharmacon; Díaz-Alonso et al., 2017; Comer et al., 2020). In addition, a GFP expression construct bearing a strong CAG promoter (pCAG-GFP, 0.5 μg/μl) was included to allow proper visualization of electroporated cells, as well as 0.1% Fast Green. Following IUE, pain was controlled by subcutaneous administration of 150 µl of Bupaq (buprenorphine, 0.3 mg/ml) and Meloxidyl (meloxicam, 5 mg/ml). Matched pregnant CD1 animals were used as lactating nurses for the IUE-derived litters to ensure equal care between experimental groups. Lactating females were weight daily to monitor recovery of surgery, and litters were weight weekly after weaning (4 weeks). No significant changes in animals’ weight gain at early postnatal stages (Extended Data Fig. 1-1) confirmed the lack of toxicity exerted by IUE manipulation.
Immunofluorescence and confocal microscopy
Adult mice were transcardially perfused with a 4% paraformaldehyde (PFA) solution, followed by overnight postfixation of their brains in 4% PFA. After treatment with a 30% sucrose solution before freezing, coronal embryonic and postnatal/adult brain slices (14 and 30 μm thick, respectively) were processed as described (Díaz-Alonso et al., 2017). Immunofluorescence was performed, using standard procedures with a rabbit polyclonal ZBTB20 (Sigma-Aldrich, HPA016815, RRID:AB_1858947) or rabbit monoclonal Gria1 (Abcam, ab183797, RRID:AB_2728702) primary antibody. Immunoreactivity was quantified in the ROI area (GFP+) cells using Fiji. For migration studies, GFP fluorescence confocal fluorescence images were acquired (Carl Zeiss LSM510, Leica SP8), and migration analyses were done semiautomatically using Fiji (guide by Labno C, University of Chicago). A 900 × 900 μm squares grid pattern organized in a 10 × 10 fashion was overlaid on the picture, and cells were counted separately for each of the 10 columns. Cell number values were converted to percentages versus total DAPI+ cells.
Cholera toxin axonal labeling studies
Callosal projections of the prefrontal cortex were retrogradely labeled adapting previous methods to the area of interest (De León Reyes et al., 2019). Fluorescent cholera toxin subunit B (CTB) Alexa Fluor 647 (23 nl/injection, Thermo Fisher Scientific) was injected stereotaxically into P30 adult mice in the following coordinates: anteroposterior 0 mm, mediolateral 0.7 mm, and dorsoventral, at three levels 2 mm (22 injections), 2.2 mm (8 injections), and 1.8 mm (8 injections) to ensure optimal labeling. Three days later, mice were perfused for confocal microscopy analyses as described above. Double-labeled CTB+GFP+ neurons were quantified and are provided as the percentage referred to the GFP+ cell population.
Electrophysiological studies
Experiments were performed using brain slices from C57Bl6 mice that were previously in utero electroporated at E14.5 with either CB1R or control siRNA (see above, In utero electroporation). Animals were killed by cervical dislocation followed by decapitation, and their brain was rapidly removed and placed in ice-cold modified artificial cerebrospinal fluid (mACSF), continuously bubbled with 95% O2–5% CO2, pH 7.4. The 300-mm-thick coronal slices of the medial PFC were cut with a vibroslicer (VT1200S, Leica) in ice-cold mACSF and subsequently incubated at 32°C for 30 min in mACSF. During recording, slices were kept submerged at 28–30°C and were continuously superfused with ACSF. Glass recording pipettes were pulled from borosilicate glass (Science Products) and had a resistance of 2–3 MΩ when filled with intracellular pipette solution (in mM: 131.25 K-gluconate, 8.75 KCl, 0.5 EGTA, 10 HEPES, 4 Mg-ATP, and 0.4 Na2-GTP), pH 7.3. Whole-cell current–clamp recordings were performed from the soma of layer 2/3 and layer 5/6 pyramidal neurons of the infralimbic and prelimbic areas of medial PFC. Signals were acquired using a MultiClamp 700B amplifier or an Axopatch 200B amplifier controlled by the Pclamp software (Molecular Devices) and an EPC10 amplifier controlled by the PULSE software (HEKA Electronic). Signals were low-pass filtered at 4 kHz and sampled at 10 kHz. Series resistance ranged from 6 to 15 MΩ and was compensated to ∼55%. Offline analysis was performed with Igor Pro (Wavemetrics).
Electrophysiological properties, including intrinsic passive and active properties of neurons were recorded in a current-clamp mode and were analyzed using scripts written in Igor Pro. Neurons were held at −55 mV by continuous current injection. Resting membrane potential (EREST) was measured immediately after the whole-cell configuration was achieved. Passive properties were estimated from voltages responses (50 sweeps were averaged for accuracy) to a 200 ms hyperpolarizing current injection of −10 pA amplitude (prepulse) and active properties were estimated from voltages responses to a 1 s current injection ranging from −100 pA to +390 pA in 10 pA steps. The membrane time-constant (τm) was determined by fitting the voltage response with a monoexponential function,
Visualization and morphological analysis of neurons
During electrophysiological recordings, neurons were filled with biocytin (2 mg/ml pipette solution, Sigma-Aldrich) to determine neuronal morphology. Slices were fixed in PFA 4% for 30 min. After permeabilization with 2% Triton X-100 for 1 h, slices were incubated for 2 h in an avidin–biotin–peroxidase complex solution (ABC kit, VECTASTAIN, Vector Laboratories). Biocytin-filled neurons were visualized using DAB (3,3′-diaminobenzidine-4 HCl, Sigma-Aldrich) reaction and were imaged with a confocal microscope (Carl Zeiss LSM510, Leica SP8). Confocal Z-stack images were acquired for 3D neuronal reconstruction using a 10×/0.75 objective with a pixel size 0.47 × 0.47 μm and a voxel depth of 2.5 μm. Images were imported into ImageJ (v1.40), and neurons were reconstructed using the Neuron Morpho plugin. To correct tissue shrinkage due to immunohistochemical processing, we applied cvapp (Cannon et al., 1998). The LMeasure software was used to define the parameters necessary to calculate the dendritic complexity index. Sholl analysis was performed using 10-μm-spaced concentric rings using Fiji together with the Simple Neurite Tracer plugin.
Neonatal behavioral evaluation
Neonatal motor assessments are derived from the adaptation of Fox's battery of tests. The surface righting reflex test is a motor ability of newborns to roll over and stand up from their supine position. In this test, Postnatal Day (P)9–P12 pups are placed in a supine position on a cotton sheet and held in that position for 5 s. After release, the time it takes for them to return to a prone position is recorded. As there is no repeat-related learning reported in this test, a duplicate within 5 min is conducted.
Negative geotaxis
Negative geotaxis is employed as a functional characterization parameter for NDD models and brain injury. The negative geotaxis response is analyzed by assessing the ability of the pups to orient themselves upward, using vestibular cues of gravity, when placed on an inclined plane. A plexiglass platform (12.5 × 45 cm W × L) covered with rough cloth paper was used to test the negative geotaxis response. The offspring of the siCo- and siCB1R-derived groups were tested on P8–P13. The pups were placed on a 45° inclined plane and the mean latency to rotate 180° was recorded. Each pup was subjected to two 30 s trials per day. If the maximum length was reached or the pup walked or fell down the ramp during the trial, a duration of 30 s was assigned. Vibrissae response was evaluated by observing the mouse's head response when it is moved vertically and rostrally toward a flat horizontal surface. When the vibrissae system is fully developed, the mouse responds by retracting its head. The vibrissae response and forelimb and hindlimb grasp reflexes were determined between P9 and P12. Neonatal behavioral raw data are included in Extended Data Figure 5-1 and represented in Figure 5. No sex-dependent differences were observed between groups (Extended Data Fig. 5-2).
Adult mouse behavioral characterization
Determinations were performed according to established protocols and as described in previous studies (de Salas-Quiroga et al., 2020; Maroto et al., 2023) by two independent evaluators blind to the mice manipulations. Video recording was performed during the early light phase under dim illumination (<50 lux in the center of the corresponding maze), and their analysis was performed by a different blind evaluator. An open-field test was conducted by placing the mice in the center of an open-field arena measuring 70 × 70 × 40 cm and allowing them to ad libitum explore the area for 10 min. The arena was illuminated with a constant light intensity, and the behavior of the mice was recorded and analyzed. The open-field test is widely used to study the general activity and exploration patterns of animals in a novel environment, as well as to investigate anxiety-like and motor behaviors. The data obtained from this test can provide valuable insights into the overall motor coordination, spatial navigation, and emotional reactivity of the mice.
The novel object recognition (NOR) test was performed the day after the open-field test, which served as habituation to an empty arena. On the day of the NOR test, mice were given an additional 10 min of habituation in the empty arena, followed by 1 h rest period. Afterward, mice were allowed to ad libitum explore the arena with two identical objects for 10 min, followed by another 1 h rest period. Finally, mice were allowed to explore the arena containing one familiar object and one novel object. The discrimination index (DI) was calculated as the difference in exploration time between the novel and familiar objects, divided by the total exploration time
Elevated plus maze was performed as previously described (Maroto et al., 2023). Before the test, the mice were housed in conditions of red light for acclimation. The test consisted of placing the mice in the center of the maze (consisting of two open arms and two closed arms measuring 30 × 7 cm arranged orthogonally at 60 cm from the floor) and allowing them to explore the maze for 5 min. The test was performed under red light conditions. For the Y-maze test, mice were placed in the center of the Y-maze and allowed them to explore one of the three arms for 3 min. After a 5 min intertrial interval, the mice were again placed in the center of the maze and allowed to explore ad libitum all three arms for an additional 3 min. The time spent in each arm was recorded and analyzed.
Spontaneous motor activity was analyzed using an automated actimeter (Acti-Track; Panlab). The device consisted of a 22.5 × 22.5 cm area with 16 infrared beams surrounding it, all connected to a computerized control unit. After a 1 min period of acclimation, activity was recorded for 5 min, and data were collected using the Acti-Track v2.7 software (Panlab).
The three-chamber test was used to assess sociability and preference for social novelty, as previously described (Maroto et al., 2023). The apparatus consisted of a rectangular plexiglas box (60 × 40 × 22 cm, L × W × H) divided into three interconnected chambers, with a light intensity of 15 lx. For habituation, the test mouse was first allowed to explore ad libitum the entire arena, which contained two empty cylindrical cages (15 cm high, 8.5 cm diameter) positioned in the two outer chambers, for 10 min. After this initial exploration, the mouse was returned to its home cage for a 1 h rest period. Following the rest, a familiar mouse of the same strain (C57BL/6J) and sex as the test mouse was placed in one of the previously empty cylindrical cages, and the test mouse was allowed to explore the arena again for 5 min to assess sociability. After another 1 h rest period, a novel, unfamiliar mouse of the same strain and sex was placed in the second cylindrical cage. The test mouse was then allowed to explore the arena ad libitum for 5 min to assess its preference for social novelty. In both sessions, the observer used a stopwatch to manually record how long the mouse spent sniffing each cage. Additionally, the placement of the cages holding the mice was randomized.
The DI for social interaction test was determined as
Adult behavioral raw data are included in Extended Data Figure 5-3 and represented in Figure 5 as indicated in the text. No sex-dependent differences were observed in open field, actitrack, and social interaction changes between groups (Extended Data Fig. 5-4).
Gene expression analyses
FACS cytometry and sorting. Animals were killed by cervical dislocation followed by decapitation. The cerebral cortex was dissected on ice using a fluorescence magnifying glass to identify the electroporated GFP+ region. Subsequently, this tissue was incubated with Accumax (Sigma-Aldrich) at room temperature for half an hour. GFP-expressing cells were sorted on a FACsAria III flow cytometer or Canto 3L (BD Biosciences), and data were analyzed using the FlowJo (10.4.2) software. IUE cells from the total sample were visualized by positive labeling for GFP and propidium iodide. After isolation, GFP+ sorted cells derived from IUE mice were resuspended in a lysis buffer for RNA extraction (Arcturus PicoPure RNA Isolation kit). Given the capacity for single-cell RNA extraction, extractions were performed on samples ranging from a few hundred to over 100,000 sorted cells. RNA concentration and integrity were evaluated using an Agilent 2100 Bioanalyzer and a Qubit 2.0 fluorometer (Thermo Fisher Scientific). Only samples that met quality criteria were selected for transcriptomic analysis. Gene expression profiling was conducted at E17.5, P7, and P30 (n = 14–16 per time point) using Clariom D Pico microarrays (Thermo Fisher Scientific).
Bioinformatics analysis
For transcriptomic analyses, raw data were generated and controlled using the Transcriptome Analysis Console (TAC) of Applied Biosystems. Standard pipelines were used for quality control, and differential gene expression analysis was performed using the “limma” R package in TAC (Ritchie et al., 2015). Differentially expressed genes (DEGs) were defined as a nominal p < 0.05 and FDR < 0.05 (Extended Data Fig. 6-1) and fold change −1.65< or >1.65 for the comparison between siCo and siCB1R groups. The selection of these threshold values aimed to preserve statistically significant alterations in gene expression while avoiding the exclusion of the finer nuances of gene expression changes that actively contribute to biological pathways.
Data were initially filtered to remove predicted transcripts and unlabeled genes. The analysis of gene ontology specific to neurons was conducted using SynGO (Koopmans et al., 2019). Overrepresentation analysis (ORA) and Gene Set Enrichment Analysis (GSEA) were employed to elucidate the relationships between gene expression patterns and biological states (Yu et al., 2012). Genes that passed the filters were imported to WebGestalt (WEB-based Gene SeT AnaLysis Toolkit, http://www.webgestalt.org/), an enrichment analysis web tool, with p value and FDR < 0.05, to generate the enrichment data. For pathway enrichment analysis, we utilized the clusterProfiler R package (Yu et al., 2012) in conjunction with the ReactomePA package (Yu and He, 2016) to evaluate the enrichment of DEGs within Reactome pathways (reactome.org), using Mus musculus-specific annotations provided by the org.Mm.eg.db package (Carlson and Falcon, 2019). The resulting pathway enrichment data were visualized using the ggplot2 package in R (Wickham, 2009), which illustrates the significance and breadth of the enriched pathways. To evaluate the statistical enrichment of genes that were differentially expressed, GO terms (geneontology.org), PantherDB (pantherdb.org), and Kyoto Encyclopedia of Genes and Genomes (KEGG, kegg.jp) databases were employed (Extended Data Fig. 6-2).
DEGs validation by RT-qPCR analysis
To evaluate the differentially expressed transcripts including coding genes and microRNAs, total RNA was extracted using Nucleozol (Macherey-Nagel) and following manufacturer instructions. RNA samples (1–2 µg) were subjected to reverse transcription using the Transcriptor First Strand cDNA Synthesis Kit (Roche Life Science) with random hexamer primers. Real-time (RT)-qPCR was performed in 96- or 384-well plates using the LightCycler Multiplex DNA Master (Roche Life Science) using Fast SYBR Green probe (Applied Biosystems, #4385610) and appropriate primers in a 7900 HT-Fast (Applied Biosystems). Primers are listed in Extended Data Figure 6-3. The 96- or 384-multiwell plates were analyzed in QuantStudio 12K Flex and QuantStudio 7 Flex, respectively (Applied Biosystems). The ΔΔCt method was used to compute the relative expression ratio of the target gene normalized to the reference gene. Determinations were performed at least in triplicate, and two control transcripts were used for normalization (RNA 18s and GAPDH).
Experimental design and statistical analysis
Experimental design was performed according to ARRIVE guidelines. Animal experiments and their derived samples were performed and analyzed in a blinded manner regarding their in utero manipulation (typically, an experimenter prepared the animals and their derived samples, and another experimenter conducted the assays blinded to group allocation). The sample size for each experiment was limited according to the success of IUE-derived litter. In any case, mice from at least three different litters were included together in the analysis, to control litter effects. The number of biological replicates (e.g., number of mice) is provided in the corresponding figure legends. The number of technical replicates (e.g., number of behavioral trials per mouse) is provided in the corresponding Materials and Methods. No data were excluded for the statistical analyses except when, very rarely, there was an obvious technical problem during the measurement. Source data from animal experiments were collected and analyzed as disaggregated for sex. No statistically significant differences were found between male and female mice in the postnatal and adult behavioral analyses of the study. Behavioral raw data and sex-dependent analyses are provided as Extended Data Figures 5-1–5-4, and, likewise, the transcriptomic raw data have been deposited in the public repository Docta UCM (https://hdl.handle.net/20.500.14352/124213) and are available for reutilization upon request following FAIR recommendations.
Statistical analysis
GraphPad Prism 8 (GraphPad Software), RStudio (version 2023.09.1; RStudio team 2020), “ggplot2” package (Wickham, 2009), and TAC (version 4.0.1, Applied Biosystems) were used for all statistical analyses and visual representations. All datasets were tested for normality (Kolmogorov–Smirnov's test) and homoscedasticity (Fisher's F test and Levene's test) prior to analysis. Behavioral differences between treatment groups were determined with ordinary one-way ANOVA with Tukey's multiple-comparison test (adjusted p < 0.05). Neuronal migration, Sholl analysis, and firing frequency were analyzed by a two-way repeated–measure ANOVA. Neuronal properties (active and passive) were analyzed with a two-way ANOVA. The fraction of regular spiking to intrinsic bursting cells was analyzed using Fischer's exact test. Multiple-comparison correction was performed where necessary. All data are represented as mean ± SEM, unless otherwise stated. The statistical test applied to each dataset is indicated in the corresponding figure legend. The precise p values are given in the figures. All datasets are presented as dot plots.
Results
CB1R knockdown alters the intrinsic electrophysiological properties of projection neurons
Embryonic CB1R knockdown was induced by IUE of siCB1R during PFC development, when layer 2/3 projection neurons are generated (E14.5). Neuronal gene expression and electrophysiological characteristics were determined at the indicated time points (Fig. 1A). In addition, adult mice derived from siCB1R IUE were subjected to behavioral analysis to determine the long-term consequences of the prenatal CB1 receptors deficiency, in comparison with siCo mice. CB1R knockdown induced an arrest of radial migration of developing neurons that was evident as early as E17.5 that persisted into adulthood (Fig. 1B,C). Although a fraction of neurons overcame this effect and reached their proper upper-layer destination (57% ± 9 of GFP+ neurons), a significant fraction remained ectopically located in the deep layers of adult siCB1R-derived mice (43% ± 9 of GFP+ neurons; p values bin 4, 8, 9, and 10: 0.0124, 0.0102, 0.0142, and 0.0452). These findings are consistent with previous reports of impaired radial migration in other cortical areas following CB1R knockdown (Díaz-Alonso et al., 2017). To determine whether arrested neurons retain their identity, we analyzed SATB2 expression in GFP-electroporated neurons in the cingulate cortex. Representative images are shown in Figure 2A, top panels. Quantification of GFP+ neurons in layers 2/3 evidenced their reduction by CB1R knockdown when compared with siCo, and this was reflected by the appearance of increased GFP+ neurons in layer 5 of siCB1R-derived cortices, which were almost absent in control conditions (Fig. 2A, bottom panel; p < 0.0001). The analysis of SATB2 expression and the quantification of double SATB2+GFP+ neurons confirmed that all ectopic neurons preserved SATB2 expression (Fig. 2A, bottom panel; p < 0.001). Considering the ectopic layer location of siCB1R-targeted neurons, we analyzed if this was accompanied by an altered axonal projection connectivity. As, typically, upper-layer neurons project mostly to contralateral cortical areas via the corpus callosum, we injected fluorescent CTB in this region of IUE-derived mice. A representative image of CTB labeling in GFP-electroporated neurons is shown at low magnification (Fig. 2B,b1) together with the appropriate callosal filling with CTB (Fig. 2B,b2). We quantified the number of electroporated (GFP+) neuronal somas and double GFP+CTB+ neurons in layers 2/3 (arrowheads) and layer 5/6 of the cingulate cortex (Fig. 2B,b3). siCB1R electroporation resulted in a minor decrease of retrograde callosal labeling into layers 2/3 that was accompanied by a significant increase of CTB+ neurons in layer 5/6 somas (Fig. 2B,b3, p = 0.001). These findings show that CB1R knockdown-induced cell migration arrest evokes contralateral connectivity changes, leading to the emergence of callosal axons wrongly originating from ectopic neurons misplaced in layer 5.
Transient CB1R knockdown impairs neuronal migration. A, Graphical abstract of the experimental design depicting the different time points for analyses performed after in utero electroporation (IUE) at E14.5 with either siRNA control (siCo) or siRNA targeting CB1 receptor (siCB1R) together with a GFP expression plasmid to identify the targeted neurons. B, C, Representative GFP expression in embryonic E17.5 (B) and adult PFC slices (C) are shown. Quantification of GFP+ neurons along cortical bins (p values obtained by two-way ANOVA with Sidak's multiple-comparison test). Extended Data Figure 1-1 provides data of weight gain of IUE-derived litters at postnatal stages.
Figure 1-1
Table with the weight gain of siCB1R and siCo at postnatal stages. Download Figure 1-1, XLSX file.
Characterization of neuronal identity and axonal connectivity changes induced by transient CB1R knockdown. A, Immunofluorescence analysis of SATB2 expression (red) in electroporated GFP+ (green) neurons located in layers 2/3 and layer 5 in the cingulate cortex of P15 siCB1R- and siCo-derived mice (E14.5). Representative images and quantification of SATB2+GFP+ neurons are shown (top and bottom panels, respectively). B, Representative images at low and high magnification showing CTB retrogradely labeled GFP-electroporated neurons. In b1, the left panel provides a coronal brain section mosaic reconstruction with the general view of CTB and GFP fluorescence, and the right panel shows an image demonstrating CTB filling of the corpus callosum. In b2, representative fluorescent images of double-labeled CTB+GFP+ neuronal somas in layers 2/3 are shown (white arrow heads). CTB was injected into the corpus callosum at P30 and analysis performed 3 d later. Quantification of GFP+ somas in the total DAPI population and double-labeled CTB+GFP+ in the electroporated GFP population is shown in bottom panels. Bar sizes, A, b1 25 mm; b2, 200 mm; b3, 15 mm (p values obtained by two-way ANOVA with Sidak's multiple-comparison test).
To evaluate the long-term functional consequences of prenatal CB1R loss of function at the single-cell level, we performed whole-cell current–clamp experiments on acute coronal slices from 8-week-old mice. We characterized passive and active electrophysiological properties of siCB1R-derived ectopic neurons (deep-layer GFP-positive siCB1R–derived neurons) and compared their properties with GFP-labeled upper-layer neurons (siCo and siCB1R-derived neurons) and with native deep-layer pyramidal neurons (unlabeled control neurons derived from siCo slices). Comparison between siCo and siCB1R-derived neurons showed no significant differences in passive properties (resting membrane potential, input resistance, cell capacitance), rheobase, and in single AP properties (AP threshold, AP amplitude, AP half-width, Table 1). We next examined the repetitive AP firing (frequency of AP as a function of the amplitude of the injected current, F/I plot, Fig. 3A). We observed that siCB1R-derived ectopic neurons (located in layers 5/6) have AP firing frequencies comparable with siCB1R-derived layer 2/3 neurons, while they have significantly upper AP frequencies than upper-layer 2/3 siCo-derived neurons (Fig. 3A,B; F(49, 3,917) = 253.0; p < 0.05) and significantly lower AP firing frequencies than native deep-layer siCo–derived neurons (Fig. 3A,B; F(49, 2,497) = 4.013; p < 0.0001). We also observed that siCB1R-derived ectopic neurons have AHP values (16.23 ± 0.76 mV) comparable with the values measured for layer 2/3 neurons [both derived from siCo (16.25 ± 077 mV) and siCB1R slices (18.51 ± 0.54 mV)] and an AHP amplitude larger than those of deep-layer neurons in their proper position (13.49 ± 0.57; Fig. 3C; Table 1; t(80) = 3.15; p = 0.0137). Besides firing AP at different frequencies, another major characteristic that differentiates upper and lower layer neurons is the pronounced sag observed in deep-layer neurons compared with the relative small (or absence of) sag in upper-layer neurons in response to hyperpolarizing current injections (reflecting the activation of HCN channels that contribute to the regulation of neuronal excitability; Shah, 2014). Sag characteristics provide important functional differences between deep and upper cortical neurons (Harnett et al., 2015). Interestingly, neurons arrested in deep layers have a sag (6.32 ± 1.57%) comparable with the small sag observed for upper-layer neurons (siCo, 2.51 ± 0.65%; siCB1R, 5.68 ± 1.09%) which are significantly smaller than the sag measured from native deep-layer neurons (13.91 ± 1.66, Fig. 3D; t(80) = 4.044; p = 0.0007). These findings suggest that ectopic siCB1R knockdown-derived neurons, although located in deep cortical layers, retain some key electrophysiological characteristics of upper-layer neurons, such as AHP amplitude and sag percentage, that influence AP firing properties. However, even though they are less excitable than native neurons, they remain more excitable than upper-layer neurons.
Transient CB1R knockdown impairs neuronal electrophysiological properties. A, Representative examples of AP firing in response to depolarizing current injection of adult (Week 8) layer 2/3 (top) and layer 5/6 (bottom) neurons under control conditions (gray) and siCB1R knockdown-derived slices (blue). B, Average AP firing frequency in response to various depolarizing current injections was quantified in adult layer 2/3 neurons from brain slices (siCo and siCB1R knockdown-derived slices n = 16 per group) or layer 5/6 neurons (control and siCB1R knockdown; n = 28 and 24, respectively; p values obtained by two-way ANOVA with Sidak's multiple-comparison test, layers 5/6 control vs siCB1R comparison is shown above, and layers 2/3 control vs layers 5/6 siCB1R is shown in the bottom part of the panel). C, Average AHP amplitude determined in upper- and deep-layer pyramidal neurons in control conditions (gray) and after siCB1R knockdown (black) at E14.5 (p values obtained by two-way ANOVA with Sidak's multiple-comparison test). D, The percentage of sag was calculated as the % change between the maximum of the voltage response and the voltage at the end of the response (bottom panel; layer 2/3 siCo and siCB1R n = 16 per group; layer 5/6 control and siCB1R n = 28 and 24, respectively; p values obtained by two-way ANOVA with Sidak's multiple-comparison test).
Electrophysiological properties of upper-layer 2/3 and deep-layer 5/6 neurons derived from electroporated medial PFC siCo and siCB1R-derived neurons at E14.5
Examples of reconstructed adult medial PFC neurons filled with biocytin illustrating upper-layer neurons and deep-layer neurons are shown in Figure 4A. Deep-layer neurons can be broadly classified into two categories: tufted neurons (with a broad and complex apical dendritic tree) and slender neurons (with a simpler dendritic tree). siCB1R-derived neurons that display an arrest of migration in the deep layers are morphologically different from both siCo- and siCB1R-derived upper–layer neurons. We also observed that these ectopic neurons have a reduced apical dendritic complexity compared with native deep-layer neurons. Sholl analysis indicated that ectopic neurons have a more complex proximal dendrite (within 200 mm from the soma) than upper-layer neurons and a longer apical dendrite extending from deep layers toward and branching in upper layers. When compared with native deep-layer neurons, the main differences in apical dendritic complexity originated from the distal part (>390 µm away from the soma, Fig. 4B; F(59, 1,534) = 1.46; p = 0.014). Morphological differences between upper-layer and ectopic siCB1R-derived neurons could explain their differential excitability, as dendritic morphology and complexity shape the firing pattern (van der Velden et al., 2012; Harnett et al., 2015). In addition, we characterized the repetitive firing pattern of ectopic neurons and compared their regular spiking versus intrinsic bursting behavior. In control conditions, we quantified 43% of bursting neurons and 57% regular spiking neurons, whereas most neurons that underwent CB1R knockdown are regular spiking neurons (91%; Fig. 4C,D; p = 0.0105). Thus, the electrophysiological properties of improperly migrated neurons preserve some of the excitatory properties of upper-layer 2/3 neurons. To rule out the possibility that the electrophysiological alterations observed in electroporated neurons result from a persistent reduction in CB1R levels, we assessed Cnr1 gene expression levels by analyzing microarray signals and quantifying CB1R transcripts with qPCR at E17.5, P7, and P30. Lower CB1R mRNA levels were only observed by qPCR at E17.5 in siCB1R extracts, but not at P7 or P30 (0.531, 0.9672, and 1.063; p = 0.0419, 0.6759 and 0.7270, respectively; n = 9; p values obtained by unpaired Student's t test); thus CB1R mRNA resumes to normal levels by P7. Taken together, our results show that interfering with CB1R expression during the prenatal period of PFC development results in a permanently altered laminar cortical organization, accompanied by distinct electrophysiological properties.
Transient CB1R knockdown alters the dendritic complexity and spiking pattern of ectopic deep-layer neurons. A, B, Example of reconstructed adult medial PFC neurons illustrating layer 2/3 neurons as well as layer 5/6 neurons in control conditions (siCo) and in adult siCB1R-derived mice (siCB1R). Sholl analysis quantifying the number of intersections of dendritic branches within concentric circles at increasing distances from the cell soma (control n = 15, siCB1R knockdown n = 16; p values obtained by two-way ANOVA). C, Representative burst firing (top) and regular spiking (bottom) pattern of adult layer 5 pyramidal neurons of control-derived medial PFC slices. D, Quantification of bursting (BS) and regular spiking (RS) layer 5/6 neurons in adult medial PFC induced by the transient knockdown of CB1R (control, n = BS:12, RS:16; siCB1R knockdown, n = BS:2, RS: 22; p value obtained by Fisher's exact test).
Embryonic CB1R knockdown in the PFC interferes with social behavior in adulthood
Behavioral characterization of embryonically electroporated siCB1R- and siCo-derived mice was performed at P9–P12 and P60 adult stages. From P9 to P12, we evaluated the righting reflex and the negative geotaxis tests, used as early indirect indicators of neuropsychiatric traits. siCB1R-derived mice showed a higher mean time to rotate 180° than the siCo group (Fig. 5A; F(1,52) = 1 3.2; p = 0.0006) and, similarly, spent more time in the negative geotaxis test (Fig. 5B; F(1,13) = 17.58; p = 0.0065). Actitrack analysis of adult siCB1R mice indicated a hypokinetic phenotype (Extended Data Fig. 5-5), while in the open field, siCB1R mice have decreased ambulation (Fig. 5C; t(52) = 3.423; p = 0.0012). Also, siCB1R mice traveled more distance and expended more time in the center (Fig. 5C; t(50) = 2.324; p = 0.0242 and t(54) = 2.068; p = 0.0434) and less in the periphery (Extended Data Fig. 5-6). In the NOR test, no differences were found, indicating the absence of cognitive impairment (t(36) = 0.7512; p = 0.4574). Also, no differences were evident in the Y and elevated plus maze tests indicating the absence of greater anxiety (Extended Data Fig. 5-7). Social interaction analyzed by the three-chamber test revealed that siCB1R-derived mice had a lower DI for both the novel mouse versus the empty cage (Fig. 5D; t(32) = 3.313; p = 0.0023) and the novel mouse versus the familiar mouse (Fig. 5E; t(53) = 2.134; p = 0.0375). No sex-dependent differences were observed either in postnatal behavior analyses or in social interaction deficits in adulthood (Extended Data Figs. 5-2, 5-4). In summary, transient prenatal loss of CB1R function in developing PFC projection neurons induces long-term deficits in social interaction.
Behavioral characterization of siCB1R and siCo-derived mice. A, B, Negative geotaxis test and righting reflex test were performed in postnatal siCB1R and siCo mice from P9 to P12 (n = 9 and 6 respectively; p values obtained by multiple comparisons). C, Representative mouse trajectories in open field (OF) and quantification of ambulation (total distance traveled, cm), distance in center (cm), time in center (s), and resting time in center (%) in a 10 min OF test (n = 21–30 animals per group, p values obtained by unpaired Student's t test). D, E, Sociability and social novelty preference at P60 were determined in the three-chamber test and DI is indicated (n = 34 and 20, siCB1R and siCo, respectively; p values obtained by unpaired Student's t test). Extended Data Figures 5-5–5-7 show actitrack, open field, Y, and elevated plus maze behavior data of siCo and siCB1R-derived adult mice. Extended Data Figures 5-1 and 5-2 provide neonatal behavior raw data and sex-dependent analyses, while Extended Data Figures 5-3 and 5-4 provide adult behavior raw data and sex-dependent analyses of the different tests performed.
Figure 5-1
Table with the neonatal behavior raw data of the different tests performed. Download Figure 5-1, XLSX file.
Figure 5-2
Table with the sex dependent analyses of postnatal behavior characterization. Download Figure 5-2, XLSX file.
Figure 5-3
Table with the adult behavior raw data of the different tests performed. Download Figure 5-3, XLSX file.
Figure 5-4
Table with the sex dependent analyses of the different tests performed. Download Figure 5-4, XLSX file.
Figure 5-5
Actitrack behavior of siCo and siCB1R-derived adult mice. A, Quantification of activity, distance, locomotion and fast movements in adult siCo and siCB1R derived mice. B, Quantification of slow movements, resting time, stereotyped movements, maximum and mean velocity in the same groups (n = 21-30 animals per group, p values obtained by unpaired Student´s t test). Download Figure 5-5, TIF file.
Figure 5-6
Open field of siCo and siCB1R-derived adult mice. A, Quantification in different areas of global activity, immobility and resting time for adult siCo and siCB1R derived mice (n = 21-30 animals per group, p values obtained by unpaired Student´s t test). Download Figure 5-6, TIF file.
Figure 5-7
Y and elevated plus maze tests of siCo and siCB1R-derived adult mice. A, Y maze quantification of the total distance, distance and time in novel and known zones, and number of entries in the novel and know zone. Raw data distribution from the experimental groups is represented in the estimation (scatter) plot (left hand side) and the disparity in group means is illustrated alongside a 95 percent confidence interval (right hand side). B, Elevated plus maze entries and time in open and closed arms and total distance travelled (n = 21-30 animals per group, p values obtained by unpaired Student´s t test). Download Figure 5-7, TIF file.
Embryonic CB1R knockdown during pyramidal neuron differentiation alters the gene expression program
We characterized gene expression changes induced by embryonic CB1R knockdown at different developmental time points. Upper-layer committed neurons were FACS-sorted at E17.5, and siCB1R and siCo-derived neuronal RNA extracts were analyzed by microarray Clariom D Pico. A total number of 2,062 DEGs were found (fold change −1.65< or >1.65, padjust value < 0.05), of which 1,849 were downregulated and 213 were upregulated (Extended Data Fig. 6-1). Volcano plot illustrates the magnitude of fold change of downregulated and upregulated DEGs at the E17.5 time point (Fig. 6A) and principal component analysis the clustering of samples based on their DEGs expression profiles (Extended Data Fig. 6-4A). Most DEGs belong to multiple complex coding genes (Fig. 6B). ORA of biological processes using Gene Ontology (GO) showed enrichment in genes coding for CNS, forebrain and cerebral cortex development, axon guidance, neuron migration, cortex neuron differentiation, transcription coregulator activity, post- and presynapse, and glutamatergic synapse (Fig. 6C). Top ontology terms according to SynGO (Fig. 6D) indicate a high enrichment of DEGs involved in synaptic organization and function (including pre and postsynaptic categories). Hierarchical clustering heatmap of DEGs showing transcript abundance and the correlations of cluster proportions between siCo and siCB1R groups is shown in Figure 6E. Complementary analysis of DEGs and the distribution of gene sets enriched among DEGs through GSEA are shown in Extended Data Figures 6-4 and 6-2. Validation of selected DEGs was performed by qPCR analyses. We confirmed that CB1R knockdown (p = 0.0419) is associated with decreased levels of transcriptional regulators involved in pyramidal neuron cortical layer identity (Fig. 6F): Bcl11b, which is essential in deep-layer neurogenesis, Auts2 (whose mutations induce autistic traits), and the transcription factor Zbtb20 (involved in upper-layer callosal projection neuronal development and astrogliogenesis; p = 0.0081, 0.0359 and 0.0419, respectively). Moreover, we validated other DEGs, including genes encoding for proteins involved in neuronal activity as Gria1 and Gria4 (glutamate AMPA receptor subunits), Scnm1, Cnrip1, and Nrg1, and other DEGs of interest (Extended Data Fig. 6-5A). Finally, we confirmed by immunofluorescence that fewer ZBTB20-expressing neurons were evident at E17.5 in siCB1R-derived PFC sections compared with siCo (Extended Data Fig. 6-5B).
Transcriptomic analyses of developing neurons at E17.5. A, A volcano plot of DEGs at E17.5. B, Circular diagrams of E17.5 time point of the differential upregulated and downregulated expression genes. C, Enrichment analysis of GO terms (bp, biological process; cc, cellular components; mf, molecular function). D, SynGO synapse ontology terms graphs derived from downregulated DEGs at E17.5. E, Hierarchical clustering heatmap of E17.5 DEGs. Scaled expression log2 values were color-coded as per the figure legend. F, qPCR validation of the downregulation of Cnr1, Bcl11b, Auts2, and Zbtb20 gene expression at E17.5. P values obtained by unpaired Student's t test. Extended Data Figures 6-4–6-6 show complementary transcriptomic analyses and validation of microarray-derived results. Extended Data Figures 6-1–6-3 and 6-7 provide a complete list of DEGs between siCo and siCB1R sorted GFP+ cells at E17.5, P7, and P30; top enriched Syngo synapse terms in E17.5 downregulated DEGs, ORA, GSEA, and GeneMania; qPCR primers employed for validation of transcriptomic analysis identified DEGs; number and names of DEGs list at E17.5, P7, and P30 shown in Venn diagrams for CB1R knockdown and their overlap with the different neuropsychiatric variants databases and THC-treated hiPSC–derived neurons.
Figure 6-1
Table with the DEGs between siCo and siCB1R sorted GFP+ cells at E17.5, P7 and P30. Download Figure 6-1, XLSX file.
Figure 6-2
Table with the top enriched Syngo synapse terms in E17.5 down-regulated DEGs, ORA, GSEA and GeneMania. Download Figure 6-2, XLSX file.
Figure 6-3
qPCR primers employed for validation of transcriptomic analysis identified DEGs. Download Figure 6-3, XLSX file.
Figure 6-4
Complementary transcriptomic analyses. A, PCA analysis of E17.5 samples (Blue, siCo and red, siCB1R). B, KEGG orthology-based annotation system (KOBAS) analyses enriched bubble plot of E17.5 siCB1R DEGs. C, Gene Set Enrichment Analysis (GSEA) analyses of E17.5 siCB1 DEGs for biological processes (BP) and molecular function (MF). D, Top SynGo terms enriched in common genes between E17-P7-P30. E, GSEA enrichment plot of DEGs for GO terms and Reactome. Download Figure 6-4, TIF file.
Figure 6-5
Validation of microarray derived results. A, qPCR validation of the indicated selected DEGs between siCo and siCB1R samples at E17.5 (n = 9 siCo and n = 11 siCB1R, p values obtained by unpaired t test). B, Quantification of Mean fluorescence intensity (MFI) of ZBTB20+ neurons in PFC siCo- and siCB1R-GFP+ derived neurons at E17.5 (n = 3, p values obtained by unpaired Student´s t test). C, Quantification of Mean fluorescence intensity (MFI) of GRIA1+ neurons as above in P60 adult brain sections (n = 3, p values obtained by unpaired Student´s t test). Download Figure 6-5, TIF file.
Figure 6-6
Complementary transcriptomic analyses. A, Table indicating the number of overlapping genes of E17.5 DEGs or THC-treated hIPSC from (Guennewig et al., 2018) with the indicated neuropsychiatric categories. B-C, Venn diagrams of P7 and P30 time DEGs indicating their overlap with ID, ASD, and SCZ databases. Tables (right hand panels) indicate the number of genes overlapping within a single neurodevelopmental alteration. Download Figure 6-6, TIF file.
Figure 6-7
Table with the number and names of DEGs list at E17.5, P7, and P30 shown in Venn diagrams for CB1R knockdown and their overlap with the different neuropsychiatric variants databases and THC-treated hiPSCs-derived neurons. Download Figure 6-7, XLSX file.
We analyzed siCB1R and siCo microarray-derived gene expression data at P7 and P30 to determine the long-term changes induced by the embryonic absence of CB1R signaling. At P7 we found an elevation in the proportion of upregulated versus downregulated genes compared with the E17.5 stage (51 and 1,990 DEGs, respectively; Fig. 7A; Extended Data Fig. 6-1). At P30 stage, the total number of DEGs was strongly reduced with 107 downregulated and 123 upregulated DEGs (Fig. 7A; Extended Data Fig. 6-1). Venn diagrams of the datasets of DEGs at E17.5, P7, and P30 revealed four common transcripts changed at the three points analyzed (Fig. 7B). At P30 we confirmed the upregulation of Gria1 by qPCR and immunofluorescence (Extended Data Fig. 6-5C).
Transcriptomic analysis of postnatal siCB1R-derived projection neurons and comparative analysis with neuropsychiatric disorders gene banks. A, Hierarchical clustering and heatmap of P7 and P30 DEGs (FC 1.65 −1.65). Scaled expression log2 values were color-coded as per the figure legend. B, Venn diagram with the overlapping CB1R-induced DEGs at E17.5, P7, and P30: Cdrt4 (CMT1A Duplicated Region Transcript 4), Hprt1 (hypoxanthine phosphoribosyltransferase 1), Mir7037 (microRNA 7037), and S1pr1 (sphingosine 1 phosphate receptor). C, A Venn diagram comparing the data from the E17.5 analysis of siCB1R DEGs with the datasets of THC-induced hiPSC–derived neuron DEGs from Guennewig et al. (2018) and their comparison with ID-, ASD-, and SCZ-associated genes databases.
Global analysis of transcriptomic signatures upon CB1R embryonic knockdown and its association with neuropsychiatric disorders
To understand the potential global impact of DEGs induced by CB1R knockdown in neuropsychiatric disorders, we compared the DEGS at E17.5 with different databases of genes associated with ASD, SCZ, and ID (Guennewig et al., 2018). As revealed by the overlapping number of DEGs between embryonic CB1R knockdown and neuropsychiatric diseases databases, ID was the most represented category, with 186 common DEGs, quite above ASD and SCZ (Fig. 7C; Extended Data Figs. 6-6A, 6-7). To strengthen the potency of the comparison, we also included DEGs after THC treatment of hiPSC-derived neurons (Guennewig et al., 2018). Among the 69 DEGs shared between siCB1R E17.5 and THC-treated hiPSC-derived neurons, the highest overlap was related again with ID > ASD > SCZ (20, 2, and 0, respectively). The common gene affected at E17.5 by CB1R downregulation and THC-treated neurons, and overlapping with the three neuropsychiatric databases, was ZBTB20 (Fig. 7C). Mutations of its gene are responsible for the Primrose syndrome that is characterized by ID, ASD, and other behavioral disturbances (Cordeddu et al., 2014). Additional Venn diagrams of P7 or P30 DEGs compared with the neuropsychiatric databases are shown in Extended Data Figure 6-6, B and C. Transcriptomic analyses revealed that with the recovery of CB1R levels, the majority of DEGs also quickly adapt and return to basal levels, with remaining aberrant long-term transcription alterations. In summary, these findings highlight the importance of embryonic CB1R signaling in regulating the developing neuronal gene expression program and its possible connection with psychiatric traits of neurodevelopmental disorders.
Discussion
In the present study, we characterized in detail the molecular, neuronal, and behavioral consequences of transient loss of CB1R function during prenatal development of the PFC. siCB1R-induced knockdown induced a migration arrest of upper-layer committed neurons, and as a consequence, deep layers are populated with ectopic neurons that possess distinct electrophysiological properties, induced long-lasting deregulated neurogenic gene expression program, and social interaction deficits in the adult offspring.
PFC altered layer projection neurons in the neuropsychiatric consequences of prenatal deregulated cannabinoid signaling
siCB1R-induced ectopic neurons show distinctive morphology, with reduced apical dendritic complexity and lower AP firing rates. These neurons have greater AHP amplitude and decreased sag, indicating lower excitability. In addition, most siCB1R-derived neurons showed regular spiking pattern rather than bursting spiking characteristics. Sag is mediated by hyperpolarization-activated cation (Ih) currents via HCN channels, and it distinguishes deep from upper cortical neurons (Kalmbach et al., 2018; Moradi Chameh et al., 2021). Deep-layer 5/6 pyramidal neurons, with prominent sag, are typically the most excitable. Sag is influenced by apical dendrite structure (Harnett et al., 2015), which also affects the neurons’ computational abilities (Beaulieu-Laroche et al., 2018; Dembrow et al., 2024). Thus, decreased apical dendritic complexity following CB1R knockdown likely contributes to reduced sag and excitability.
PFC dysfunction is implicated in numerous disorders characterized by deficits in cognitive performance, attention, motivation, and impulse control (Bicks et al., 2015). Particularly, social cognition deficits, as observed in ASD and SCZ, are associated with an altered balance of excitation and inhibition within the PFC. Moreover, the appropriate balance of mPFC projection neuron populations, i.e., extratelencephalic neurons or intratelencephalic neurons and burst spiking or regular spiking neurons, is critical for supporting distinct cognitive function (Anastasiades and Carter, 2021; Leyrer-Jackson et al., 2021). Intratelencephalic neurons are proposed to play a role as temporal integrators involved in sensory processing, while extratelencephalic neurons can contribute to high-order cognitive functions and flexibility (Moberg and Takahashi, 2022). Therefore, altered mPFC neuronal populations can compromise top–down cognitive control required for the discrimination of relevant information. It is thus conceivable that the reduced excitability of ectopic deep-layer neurons and the shift in the proportion of bursting and regular spiking neurons in siCB1R-derived mice may be responsible for the long-term alteration of PFC cognitive functions that lead to the observed deficits in social interaction. Regarding glutamatergic synaptic function, transcriptomic analyses revealed changes in glutamate ionotropic receptor AMPA-type subunits and changes in synaptophysin expression (Extended Data Figs. 6-1, 6-5). Mutations of these receptor genes contribute to the family of ionotropic glutamate receptor-associated neurodevelopmental disorders (Ismail et al., 2022; Wang et al., 2022). Our results align with prior studies reporting PCE-induced alterations of glutamatergic neurotransmission in the PFC, accompanied by cognitive deficits (Mereu et al., 2003; Campolongo et al., 2007; Bara et al., 2018; DeVuono et al., 2024).
Long-term follow–up studies of children exposed to cannabis in utero, have demonstrated a higher incidence of ASD and ID (Corsi et al., 2020). Other studies investigating the consequences of PCE in humans have reported increased attention deficits and externalizing behaviors (Paul et al., 2021; Baranger et al., 2024), while others found reduced internalizing behaviors (Moore et al., 2023). Moreover, a meta-analysis of PCE studies found little evidence of broad psychopathological impact in early childhood, except for attention and externalizing problems (Sorkhou et al., 2024). PCE-induced changes of subcortical areas, including the striatum, amygdala, insula, and ventral tegmental area, may alternatively be responsible for changes in impulsivity and reward sensitivity (Wang et al., 2004; Grewen et al., 2015; Frau et al., 2019). Hence, evidence from studies in children and adults with PCE showing alterations in neuronal connectivity, cognitive and emotional homeostasis, highlights the vulnerability of CB1R-mediated neurodevelopmental processes in projection neurogenesis and maturation to the interference of exogenous cannabinoid compounds.
CB1R regulation of developing projection neuron gene expression and its implication in neuropsychiatric diseases
Genetic or pharmacological CB1R signaling manipulation leads to changes in the developmental neurogenic program involved in pyramidal neurogenesis and circuit establishment (Galve-Roperh et al., 2022; Rodrigues et al., 2024). In embryonic stages, CB1R signaling activity regulates layer-selective neuronal identity by controlling the balance of BCL11B and SATB2 transcriptional activity (Paraíso-Luna et al., 2020). As a result, in vivo CB1R knock-out mice or PCE leads to long projection corticofugal deficits (Mulder et al., 2008; Wu et al., 2010; Díaz-Alonso et al., 2012; de Salas-Quiroga et al., 2015). Here we found that CB1R knockdown induced downregulation of Bcl11b and of other DEGs, Sgrap1 and Robo1, that can participate in ECB projection neuron subcortical axonal pathfinding regulation (Ip et al., 2011; Alpár et al., 2014). Other relevant CB1R-associated neurodevelopmental DEGs are Auts2 and Zbtb20. Mutations of Auts2 are responsible for AUTS2 syndrome, a genetic disorder that causes ID, microcephaly, and neuronal hyperexcitability (Biel et al., 2022). Zbtb20 was not previously associated with CB1R signaling. Not only is Zbtb20 affected by the absence of CB1R, but it is also upregulated in THC-treated neurons (Guennewig et al., 2018) and is present in the three neuropsychiatric disorder databases. ZBTB20 participates in intermediate neural progenitor cell fate, layer 2/3 neuron development, and astrogliogenesis (Tonchev et al., 2016; Medeiros de Araújo et al., 2021), suggesting that ZBTB20 may participate in CB1R neurodevelopmental regulatory role (Galve-Roperh et al., 2022; Rodrigues et al., 2024). Recently, the contribution of CB1R receptors expressed in glutamatergic neurons to autosomal recessive nonsyndromic ID model has been evidenced (Costas-Insua et al., 2024).
Cannabinoid administration studies, depending on the developmental stage analyzed (embryonic, postnatal, or adolescent), have shown that THC, by activating and/or desensitizing CB1Rs, may influence the different neurodevelopmental processes regulated by the ECB system according to the precise neuronal, glial, or non-neural cells that express the CB1Rs (Bara et al., 2021; Galve-Roperh et al., 2022; Rodrigues et al., 2024). Thus, while manipulation of ECB signaling during the prenatal period disrupts cortical neurogenesis and lamination (de Salas-Quiroga et al., 2015; Díaz-Alonso et al., 2017; Bara et al., 2018), altering its developmental role during adolescence leads to impaired maturation of pyramidal neurons, decreased dendritic arborization, cortical thickness, and white matter alterations (Miller et al., 2018; Sánchez-de la Torre et al., 2022; Navarri et al., 2024). Omic studies investigating the consequences of PCE using human iPSCs and organoids revealed profonde alterations of neurogenesis and cellular toxicity (Ao et al., 2020; Notaras et al., 2021). In summary, the gene expression changes in the neurodevelopmental program of PFC projection neurons induced by CB1R knockdown pave the way for an increased risk of neuropsychiatric disorders, with a profile showing greater similarity to ID.
CB1Rs located on projection neurons of the developing prefrontal cortex are required for proper adult social behavior
Previous studies have shown that PCE induces changes in interneuron development, especially of CCK basket cell subpopulation, as well as in long-range projection neurons (Bara et al., 2021; Galve-Roperh et al., 2022; Rodrigues et al., 2024). Embryonic THC administration induces a reduction of hippocampal and cortical CCK basket cells that lead to spatial cognitive impairment and social interaction deficits (Vargish et al., 2016; de Salas-Quiroga et al., 2020). Noteworthy, repeated PCE rapidly induces CB1R downregulation and hence THC embryonic exposure, unless acute, can act as CB1R functional antagonist (de Salas-Quiroga et al., 2015). Although cannabinoid-induced interneuronopathy was proposed to be responsible for sociability deficits and to induce an ASD-like phenotype (Vargish et al., 2016), this assumption was not corroborated with interneuron-selective CB1R conditional knock-out mice. Conditional ablation of CB1R in projection neurons impacts the appropriate balance between deep and upper-layer cortical projection neuron development, and such lamination alterations are known to participate in neuropsychiatric disorders and epilepsy (Willsey et al., 2022). Also, CB1R-null mice display deficits in social interaction and communication, supporting the contribution of a deregulated CB1R signaling to ASD etiopathology (Fyke et al., 2021; Wang et al., 2025). Results shown herein support that selective ablation of CB1R in developing pyramidal neurons contributes to behavioral ASD-like traits and hence challenges the exclusive contribution of interneuron deficits to cannabinoid-induced ASD phenotype. In agreement, perinatal cannabinoid exposure-induced social interaction deficits in adulthood are mediated by altered projection neuron activity and decreased mGluR5 levels (Bara et al., 2018). Although we did not observe potential compensatory changes of other elements of the ECB system (Extended Data Fig. 6-1), their occurrence cannot be completely ruled out. In this regard, the alternative metabotropic cannabinoid receptor CB2R has been shown to participate in the control of social behavior (Komorowska-Müller et al., 2021).
Our transcriptomic analyses show that CB1R knockdown induced changes in gene expression that are closest to ID > ASD > SCZ. Interestingly, similar findings were reported in THC-treated human induced pluripotent stem cell-derived neurons (Guennewig et al., 2018). Moreover, analysis of the overlap between E17.5 siCB1R-induced DEGs and THC–DEGs revealed that ID was similarly the most represented neurodevelopmental disorder. In summary, our findings demonstrate that transient impairment of CB1R signaling disrupts pyramidal neuron migration and induces a gene expression signature consistent with ID and ASD that results in social interaction deficits.
Footnotes
This work was supported by the Instituto de Salud Carlos III and cofunded by the European Union (Grants PI21/00938 and PI18/00941 to I.G.-R.), the Spanish Ministerio de Ciencia, Innovación y Universidades (MICINN/FEDER; Grant PID2021-125118OB-I00 to M.G.), and the Dutch Epilepsy Fundation (EpilepsieNL, Grant WAR17_11 to T.R.W. and P.C.). S.S.-S., D.G.-R., and J.P.-L. were supported by Instituto de Salud Carlos III (PFIS), Fundación Tatiana Pérez de Guzmán, and the Spanish Ministerio de Ciencia, Innovación y Universidades (FPU Program), respectively. We are indebted to Alba Mérida, Franck Richard, Tania Aguado, Tania Gavaldá, Estrella Villanueva, Aníbal Sánchez, David van Oorschot, and Laila Kulsvehagen for their experimental assistance and data processing; to Héctor Montero and Eva Resel for their expert technical assistance and lab managing; and to Val Fernández (IRYCIS) for the bioinformatics guidance.
The authors declare no competing financial interests.
J.P.-L.’s present address: Instituto de Neurociencias (Universidad Miguel Hernández - Consejo Superior de Investigaciones Científicas), Alicante 03550, Spain
- Correspondence should be addressed to Pascal Chameau at p.j.p.chameau{at}uva.nl or Ismael Galve-Roperh at igalvero{at}ucm.es.
