Abstract
Protocadherins, a family of adhesion molecules with a crucial role in cell–cell interactions, have emerged as key players in neurodevelopmental and psychiatric disorders. In particular, growing evidence links genetic alterations in the protocadherin 9 (PCDH9) gene with autism spectrum disorder and major depressive disorder. Furthermore, Pcdh9 deletion induces neuronal defects in the mouse somatosensory cortex, accompanied by sensorimotor and memory impairment. However, the synaptic and molecular mechanisms of PCDH9 in the brain remain largely unknown, particularly concerning its impact on brain pathology. To address this question, we conducted a comprehensive investigation of PCDH9’s role in the male mouse hippocampus at the ultrastructural, biochemical, transcriptomic, electrophysiological, and network levels. We show that PCDH9 mainly localizes at glutamatergic synapses and its expression peaks in the first week after birth, a crucial time window for synaptogenesis. Strikingly, Pcdh9 KO neurons exhibit oversized presynaptic terminal and postsynaptic density in the CA1. Synapse overgrowth is sustained by the widespread upregulation of synaptic genes, as revealed by single-nucleus RNA-seq (snRNA-seq), and the dysregulation of key drivers of synapse morphogenesis, including the SHANK2/CORTACTIN pathway. At the functional level, these structural and transcriptional abnormalities result in increased excitatory postsynaptic currents (mEPSC) and reduced network activity in the CA1 of Pcdh9 KO mice. In conclusion, our work uncovers Pcdh9’s pivotal role in shaping the morphology and function of CA1 excitatory synapses, thereby modulating glutamatergic transmission within hippocampal circuits.
Significance Statement
Converging evidence indicates that genetic alterations in the protocadherin 9 (PCDH9) gene are associated with autism spectrum disorder and major depressive disorder. However, our understanding of PCDH9’s physiological role and molecular mechanisms in the brain, as well as its connection to synaptic dysfunction and brain pathology, remains limited. Here, we demonstrate that Pcdh9 regulates the transcriptional profile, morphology, and function of glutamatergic synapses in the CA1, thereby tuning hippocampal network activity. Our results elucidate the molecular and synaptic mechanisms of a gene implicated in neurodevelopmental and psychiatric disorders and suggest potential hippocampal alterations contributing to the cognitive deficits associated with these conditions.
Introduction
Synaptic connections among the approximately 90 billion neurons of the human brain must be established during neurodevelopment to give rise to functional neuronal circuitries. Failures during this process, leading to improper neuronal networks, result in neurodevelopmental and neuropsychiatric disorders. Cell adhesion molecules (CAMs) present on neuronal cell surface play a key part in neurodevelopment, regulating cell–cell recognition, cell adhesion, and migration (Hirano and Takeichi, 2012). Among CAMs, the protocadherin (PCDH) family counts >70 single membrane–spanning glycoproteins, representing the largest subgroup of the cadherin superfamily. PCDHs play critical roles in major neurodevelopmental processes, including neurite outgrowth, axon pathfinding, synapse formation, and maturation (Peek et al., 2017). Indeed, growing evidence links genetic alterations in PCDH genes to a wide range of brain pathologies (Flaherty and Maniatis, 2020; Mancini et al., 2020).
Protocadherin 9 (PCDH9) gene has been associated with autism spectrum disorder (ASD) after the identification of de novo and inherited copy number variations in autistic individuals (Marshall et al., 2008; Bucan et al., 2009). Furthermore, decreased PCDH9 transcript levels were found in the lymphoblasts of ASD patients (Luo et al., 2012). Interestingly, a meta-analysis of three genome-wide association studies linked the single nucleotide polymorphism rs9540720 in the PCDH9 gene with major depressive disorder (MDD) and cognitive function impairment (Xiao et al., 2018). rs9540720 is predicted to lower PCDH9 expression, and reduced PCDH9 mRNA levels were found in MDD patients compared with healthy controls, thus leading to the identification of PCDH9 as a susceptibility locus for MDD (Xiao et al., 2018). A recent study also associated PCDH9 with essential tremor (Clark et al., 2022), a common movement disorder that is also characterized by cognitive and neuropsychiatric features (Lombardi et al., 2001).
Pcdh9 KO mice display long-term social and object recognition deficits, hyperactivity, and reduced sensorimotor competence. These behavioral abnormalities are accompanied by reduced cortical thickness of the somatosensory cortex, defective dendritic arborization, and increased spine density of pyramidal neurons in these regions (Bruining et al., 2015). A second study on a distinct Pcdh9 KO line revealed further behavioral impairments including reduced fear extinction, possibly due to defects in Ppp1r1b+ neurons of the basolateral amygdala (Uemura et al., 2022). These investigations proved Pcdh9’s involvement in cognitive, sensory, and emotional functions, thus strengthening the link with neurodevelopmental and psychiatric disorders. However, very little is known about the physiological role and molecular mechanisms of Pcdh9 in the brain, and how its alteration relates to synaptic dysfunction and brain pathology.
Here, we examined Pcdh9 levels and subcellular localization in the mouse brain. Then, we employed a combination of morphological, transcriptional, and electrophysiological approaches to study Pcdh9 function in the hippocampus in vivo. We show that Pcdh9 depletion leads to aberrant pre- and postsynaptic morphology in the CA1, accompanied by dysregulation of synaptic gene expression. At the functional level, these alterations translate into abnormal excitatory transmission and disturbances in the hippocampal network activity. Collectively, our findings establish an important role for Pcdh9 in the formation and function of CA1 glutamatergic synapses.
Materials and Methods
Mice strains
Pcdh9 KO mice were obtained from Dr. Martien Kas’ laboratory (University of Groningen, NL). Mice were maintained in rooms with 12 h light/dark cycles, temperature between 23 and 24°C, and controlled humidity, with food and water provided ad libitum. Mice were housed at a maximum of five animals per cage in individually ventilated cages. All the experiments followed the guidelines established by the Italian Council on Animal Care and were approved by the Italian Government (protocol number 1152/2020-PR, 20/11/2020).
Immunocytochemistry
Primary neurons were prepared from cortices and hippocampi of Sprague Dawley E18 rat brains as previously described (Zapata et al., 2017). Neurons were plated onto coverslips coated overnight with poly-ʟ-lysine at 75,000 neurons per well, and grown in Neurobasal plus medium (Invitrogen) supplemented with 2% B27 plus (Invitrogen), 1% ʟ-glutamine (Invitrogen), 1% penicillin/streptomycin (Invitrogen), and 10 mM glutamate. Cultured neurons at DIV16–18 were washed in PBS and fixed in 4% paraformaldehyde and 10% sucrose for 15 min at room temperature (RT). After blocking/permeabilization in 10% normal goat serum (NGS), 0.1% Triton X-100, in PBS for 15 min at RT, neurons were incubated with primary antibodies (anti-PCDH9 Proteintech 25090-1-AP, 1:300; anti-PSD95 NeuroMab 75-028, 1:300; anti-HOMER1 Synaptic Systems 160004, 1:500; anti-GEPHYRIN Synaptic Systems 147021, 1:500; anti-VGLUT1 Synaptic Systems 135304, 1:200; anti-VGAT Synaptic Systems 131-003, 1:200) in GDB 1× solution (2×: 0.2% gelatin, 0.6% Triton X-100, 33 mM Na2HPO4, 0.9 M NaCl, pH 7.4) overnight at 4°C. After three 10 min washes with high salt buffer (500 mM NaCl, 20 mM NaPO4(2−), pH 7.4), the coverslips were incubated with secondary antibodies (anti-rabbit IgG Alexa Fluor 488–conjugated Invitrogen A11029, 1:300; anti-mouse IgG DyLight–conjugated 649 Jackson ImmunoResearch 211-492-177, 1:300; anti-guinea pig IgG DyLight–conjugated 649 Jackson ImmunoResearch 706-175-148, 1:300) in GDB 1× solution for 1 h at RT. Neurons were washed three times with high salt buffer and incubated with DAPI (1:10,000) for 5 min at RT. After washing with PBS for 5 min at RT, coverslips were mounted with Fluoromount (Thermo Fisher Scientific).
Immunohistochemistry
One brain hemisphere from WT or Pcdh9 KO mice was quickly washed in PBS and then drop fixed in 4% PFA in PBS for 24 h. The hemispheres were then washed for 72 h in PBS and subsequently cryoprotected in 30% sucrose in PBS. The hemispheres were then frozen in OCT and sectioned in 25-µm-thick slices with a cryostat. Sections were then permeabilized using 1 and 0.3% Triton X-100 in PBS for 20 and 30 min, respectively, after which they were blocked with 20% NGS, 1% BSA, and 0.3% Triton X-100 in PBS for 2 h at RT. Sections were incubated with primary antibodies [anti-MAP2 Synaptic Systems 188004, 1:500; anti-SHANK2 Synaptic Systems 162202, 1:1,000; anti-CORTACTIN Abcam Ab81208 (EP1922Y), 1:1,000] at 4°C for 24 h in 20% NGS, 1% BSA, and 0.3% Triton X-100 in PBS, after which sections were washed in 0.3% Triton X-100 in PBS for 30 min at RT. Sections were then incubated with secondary antibodies (anti-guinea pig Alexa Fluor 647 Invitrogen A21450; anti-rabbit Alexa Fluor 488 Invitrogen A21206) in 20% NGS, 1% BSA, and 0.3% Triton X-100 in PBS for 4 h at RT. After further washes in PBS for 10 min at RT, the sections were incubated with DAPI (1:10,000) for 10 min at RT, washed in 0.3% Triton X-100 in PBS for 15 min, and mounted onto glass slides with Mowiol (Merck Millipore).
Image acquisition and analysis
For cultured neurons, fluorescence images were acquired with an LSM800 confocal microscope (Carl Zeiss) and a 63× oil-immersion objective (numerical aperture 1.4) with sequential acquisition setting, at 1,024 × 1,024 pixel resolution. Images were Z-series projections of approximately 6–10 images, taken at depth intervals of 0.75 µm. Colocalization analysis was performed on randomly selected dendrites using Fiji software (Schindelin et al., 2012) with the plugin Jacop (Bolte and Cordelières, 2006). For Manders’ colocalization coefficient, puncta were calculated by thresholding images at an intensity equal to mean + 3×St. Dev of the signals.
For the CA1 region, fluorescence images were acquired with an LSM800 confocal microscope (Carl Zeiss) and a 40× oil-immersion objective with a sequential acquisition setting, at 2,048 × 2,048 pixel resolution. For each animal, images were acquired from two or three individual brain slices. Images were Z-series of approximately 10 images, taken at depth intervals of 0.8 µm. Single stacks were analyzed using Fiji, extracting the mean intensity of SHANK2 and CORTACTIN puncta that had an intensity greater than the mean + 3× SD of the whole image intensity. The Z stacks were analyzed individually and then averaged to generate the values presented.
Membrane-enriched fractions preparation
Cortices and hippocampi were dissected from Pcdh9 WT and KO 2-month-old animals, chopped with a razor blade in cold HEPES/sucrose buffer (4 mM HEPES pH 7.4, 320 mM sucrose) supplemented with protease inhibitor cocktails, incubated 10 min in ice, and homogenized with a glass-teflon homogenizer. After centrifugation at 1,000 × g for 10 min at 4°C, supernatants (S1) corresponding to total homogenate were centrifuged at 10,000 × g for 10 min at 4°C. The resulting supernatants (S2) contained mainly cytosolic components while the pellet (P2) was enriched in membranes. After resuspension in HEPES/sucrose buffer, P2 fractions were centrifuged at 10,000 × g for 15 min at 4°C to purify the preparation. P2 fractions were resuspended in modified RIPA buffer (50 mM Tris-HCl, 200 mM NaCl, 1 mM EDTA, 1% NP40, 1% Triton X-100, pH 7.4) supplemented with protease inhibitor cocktail and then quantified with bicinchoninic acid assay (BCA; Euroclone) prior to SDS-PAGE and Western blotting.
Electron microscopy
Pcdh9 KO and WT male mice aged 2.5 months (two mice per genotype) were deeply anesthetized with isofluoran before transcardial perfusion with EM Buffer (2.5% glutaraldehyde and 2% paraformaldehyde in 0.1 M sodium cacodylate buffer, pH 7.4). Brains were postfixed in EM buffer for 24 h at 4°C. Moreover, 100 μm slices were generated with a vibratome (Leica VT1000S) and the areas of interest were manually dissected. Samples were then postfixed with 2% osmium tetroxide, rinsed, en bloc stained with 1% uranyl acetate, dehydrated using increasing concentrations of ethanol and finally with propylene oxide, and embedded in epoxy resin (Electron Microscopy Sciences). Thin sections (70 nm) were obtained with an EM UC6 ultramicrotome (Leica Microsystems), stained with a saturated solution of uranyl acetate in ethanol 20%, and observed under a Tecnai G2 Spirit transmission electron microscope (FEI). Profiles of excitatory synapse used for quantitative analyses were identified on the basis of three conditions: (1) the presence in their postsynaptic terminal of the electron-dense postsynaptic density (PSD); (2) the presence of a cluster of at least three synaptic vesicles in the presynaptic compartment; and (3) the presence of a defined synaptic cleft. Images were acquired at a final magnification of 37,000× for the morphometric analysis and 13,500× for the analysis of synapse density using a Megaview (Olympus). Morphometric analyses were performed with Fiji software (Schindelin et al., 2012) whereas, to evaluate synapse density, we used a stereological approach, as previously described (Colombo et al., 2021).
Western blots
Samples were loaded on 9% polyacrylamide gels and then transferred onto nitrocellulose membranes (0.22 μm, GE HealthCare). Membranes were blocked in 5% milk and 0.1% TBS-Tween 20 for 1 h at RT and then incubated with the primary antibodies [anti-PCDH9 homemade rat antibody (Asahina et al., 2012), 1:2,000; anti-GLUA1 Cell Signaling Technology 13185S, 1:1,000; anti-GLUA2/3 homemade rabbit antibody, a kind gift from Cecilia Gotti, 1:3,000; anti-GLUN2B Neuromab 75101, 1:1,000; anti-VGLUT1 Synaptic Systems 135303, 1:15,000; anti-GEPHYRIN Thermo Fisher Scientific PA5-19589, 1:2,000; anti-PICK1 Neuromab 75040, 1:2,000; anti-PSD95 Cell Signaling Technology 3450, 1:20,000; anti-GAPDH Santa Cruz Biotechnology, 1:10,000] in 0.1% TBS-Tween 20 overnight at 4°C. After washing, the blots were incubated at room temperature for 1 h with HRP-conjugated anti-rabbit (1:20,000), anti-mouse (1:5,000) or anti-rat (1:5,000) antibodies in 0.1% TBS-Tween-20. Immunoreactive bands on blots were visualized by enhanced chemiluminescence (GE HealthCare). Quantification of band intensity was performed with Fiji software (Schindelin et al., 2012).
Hippocampal nuclei preparation and snRNA-seq sequencing
Hippocampal nuclei from two 2-month-old mice per genotype were prepared using Chromium Nuclei Isolation Kit (10x Genomics, PN-1000494) according to the manufacturer’s instructions, with minor modifications. Briefly, hippocampi were dissected, flash-frozen in liquid nitrogen and immediately processed for nuclear preparation. Frozen hippocampi were dissociated in the lysis buffer with a pestle, passed onto the nuclei isolation column, and resuspended in the debris removal buffer. Pelleted nuclei were washed twice in wash and resuspension buffer and passed twice through a 40 µm cell strainer (Greiner Bio-One EASYstrainer cell sieve 542040). Nuclei integrity was examined at the confocal after DAPI staining, and the number of nuclei was estimated with a cell-counting chamber. A total of 13,200 nuclei per sample were loaded on a chromium instrument (10x Genomics). Libraries were prepared using a Single Cell 3’ Kit with v3 Chemistry following a standard protocol. Library profile and concentration were assessed using a TapeStation (Agilent). Library concentration was also analyzed using a Qubit fluorometer. All the libraries were sequenced using a NovaSeq 6,000 (Illumina).
snRNA-seq analysis
Raw reads were processed with Cell Ranger v7.1. The raw single-cell matrices were subsequently processed with Seurat v. 4.0.44 (Stuart et al., 2019). Only nuclei with a number of detected genes between 500 and 4,000, with <20,000 unique reads/nucleus and <10% reads mapped to mitochondrial DNA were retained for subsequent analyses. Integration of different experiments was achieved using the SCTransform algorithm, while regressing out the effect of unique molecular identifier and mitochondrial counts. Principal component analysis (PCA) was performed as an initial dimensionality reduction step. All the components required to explain a cumulative variance of >90% were retained for downstream processing, corresponding to the first 30 PCA components. The Uniform Manifold Approximation and Projection (UMAP) algorithm was run on top of PCA to end up with a total of two dimensions. These dimensions were subsequently used as input for cluster identification, which was performed with the Louvain algorithm. To optimize cluster resolution identification, the ClusTree algorithm was applied, and the optimal cluster resolution was selected as a compromise between maximization of the SC3 stability index and cluster number. SingleR (Aran et al., 2019) was subsequently applied to log-normalized gene counts at the cluster level, using the CellDex dataset and the celldex::MouseRNAseqData as a reference to perform cell-type identification, which was followed by direct manual curation. Differentially expressed genes (DEG) were identified using a negative binomial generalized linear model. For each cluster, transcripts were considered to be differentially expressed in WT versus Pcdh9 KO cells when the following three conditions were satisfied: an absolute value of the log2FC(KO/WT) of >0.2, at least 10% of the cells expressing the transcript in at least one genotype, and a Bonferroni-adjusted p-value of <0.1.
CA1 clusters were identified based on CA1 marker genes from Cembrowski et al., (2016). For interactions and GO analysis, the DEG between the two genotypes found in CA1 Clusters 5, 6, and 13 were combined in a unique list of CA1 DEG, from which Pcdh9 was excluded. The interactions of the 31 CA1 DEG were explored with STRING (Szklarczyk et al., 2019). GO analysis of the CA1 DEG was performed with g:Profiler (Kolberg et al., 2023) and SynGO (Koopmans et al., 2019). STRING and SynGO analysis were performed on the human homologs of DEG. Genes were annotated as pre- and postsynaptic (Fig. 3G) based on SynGO ontologies. For GSEA (Reimand et al., 2019), the complete list of genes detected in CA1 Clusters 5, 6, and 13 was ranked by log2FC(KO/WT) * -log10p-value. As a gene set, the list of all human genes annotated as synaptic was used (GO:0045202; SynGO portal; Koopmans et al., 2019), after conversion to their mouse homologs using Metascape (Zhou et al., 2019). The genes coding for components of the excitatory and inhibitory PSD (Extended Data Fig. 3-2B) were identified by Uezu et al., (2016).
Behavioral assays
All behavioral tests were performed at a similar time of the day (1–3 h before the light off) to avoid circadian effects. The mice were allowed to habituate to the behavioral room for at least 45 min before each test. Behavioral equipment was cleaned with 70% ethanol after each test session to avoid olfactory cues.
A social recognition test was performed as in a previous study (Bruining et al., 2015; long-term SRE). Before testing, all mice were group-housed. To avoid exposure to odor prior to testing, test and intruder animals were all housed in separate cages. The day before the test, intruder animals were labeled with neutral-smelling water-based markers. Habituation and testing took place in regular cages (29 × 17 × 11 cm) with fresh bedding. On Day 1, 2-month-old males (test animals) were habituated in the test cage for 5 min and initially exposed to an age-matched male conspecific (familiar intruder) for 2 min. Then, test animals were returned to their home cage. On Say 2, after 24 h, the test animal was habituated again in a clean cage with fresh bedding for 5 min and then exposed at the same time with the familiar intruder of Day 1 and to a novel age-matched male (novel intruder) from a different cage than the familiar intruder. Social investigation was defined as the total time that the test mice engaged in social sniffing, anogenital sniffing, and allogrooming. The time spent by the test animal in investigating each intruder was manually scored from video recordings by an observer blind to the animal's genotype.
For novel object recognition, habituation and testing took place in empty transparent plastic cages. The objects used were made of glass or plastic and were of sufficient height to discourage mice from climbing on the objects. In a pilot experiment, mice were exposed for 5 min to different couples of objects in order to select two that fulfilled the following conditions: mice had to show interest in both objects, with no overt intrinsic preference for one object compared with the other. The choice of which object to use as familiar and which one as novel was counterbalanced among the different animals tested. On Day 1 and Day 2, 2-month-old males were habituated in the test cage for 5 min. Then, test animals were returned to their home cage. On Day 3, test animals were exposed to two identical objects (familiar objects) for 5 min and then returned to their home cage. On Day 4, after 24 h, test animals were exposed for 5 min to the familiar object and to a novel object. Object exploration was defined as approaching and sniffing the object or touching it with the forepaws. The time spent by the animals in investigating each object was manually scored from video recordings by an observer blind to the animal genotype.
In vivo magnetic resonance imaging (MRI)
MRI experiments were conducted on a 7 Tesla scanner for rodents, fully equipped for brain MRI/MRS (BioSpec, ParaVision 6.0 Software Bruker BioSpin). A dedicated mouse head coil (four channels) was used as the receiver together with a volume coil as the transmitter. A mixture of Iso-Vet (isoflurane 1–2%; Zootecnica, #104331020) with oxygen was used to anesthetize animals, and breath rate was constantly monitored to regulate the level of anesthesia. Body temperature was maintained through warm water circulating inside the bed. For cerebral anatomy analysis, T2-weighted MR images were acquired with coronal sections across the entire brain of 0.6 mm and an in-plane resolution of 71 µm (FOV, 16 × 14 mm; matrix, 224 × 200) using a fast-spin-echo sequence (TR/TE, 3,500/44 ms; rare factor, 10; average, 10; 11 min of acquisition). T2-weighted images were analyzed using a Matlab toolbox, the Atlas Normalization Toolbox using elastiX (ANTX, https://github.com/ChariteExpMri/antx2; Koch et al., 2019), allowing to extract the cerebral volume of different brain areas following the Allen Mouse Brain Atlas (http://mouse.brain-map.org/). Areas with a cluster of at least five voxels were considered that correspond to a minimum volume of 0.025 mm3. Single brain region volume was then compared with the total brain volume of each mouse. The whole procedure was completed following the guidelines described in Koch et al. (Koch et al., 2019).
qRT-PCR
mRNA was extracted from murine tissues using NucleoZOL Reagent (Macherey-Nagel, 15443435) following the manufacturer's instructions. A total of 1 μg of extracted mRNA was used to synthesize cDNA using SuperScript VILO cDNA Synthesis Kit (Thermo Fisher Scientific). Pcdh9-specific sequences were amplified from cDNA with SYBR Green PCR Master Mix (Applied Biosystems) using an Applied Biosystems 7000 real-time thermocycler. Parallel qPCR reactions were performed on α-actin as a reference gene. Primers were designed with Primer-BLAST online tool (https://www.ncbi.nlm.nih.gov/tools/primer-blast/): Pcdh9 exon 1 (F: ACCATCACCCAACTCTGACG, R: GGAGTGTATCCCACCGCATC), Pcdh9 exon 2 (F: AGACTGCCCTGGTAAGGGTT; R: AAGGAGGCATTCGGTCCAAG), Pcdh9 exon 5 (F: CTCCTGGCTTGGGTCCATAC, R: GTGGCCGCCATTGTTGAAAT), and α-actin (F: AGATGACCCAGATCATGTTTGAGA, R: CCTCGTAGATGGGCACAGTGT). Each reaction was performed in triplicate, and the results were analyzed using the ΔΔCT method.
Patch clamp
P25–P30 Pcdh9 KO and WT male mice were humanely killed, and the brain was rapidly removed and placed in an ice-cold cutting solution containing the following (in mM): 195 sucrose, 10 NaCl, 25 NaHCO3, 2.5 KCl, 1.25 NaH2PO4, 7 MgCl2, 0.5 CaCl2, and 10 glucose (pH 7.3, equilibrated with 95% O2 and 5% CO2). Coronal slices from PFC and hippocampus (thickness, 250–350 μm) were prepared with a vibratome VT1000 S (Leica Microsystems) and then incubated first for 40 min at 37°C and then for 40 min at room temperature in artificial cerebrospinal fluid (aCSF), consisting of the following (in mM): 125 NaCl, 2.5 KCl, 1.25 NaH2PO4, 1 MgCl2, 2 CaCl2, 25 glucose, and 26 NaHCO3 (pH 7.3, equilibrated with 95% O2 and 5% CO2). Slices were transferred to a recording chamber perfused with aCSF at a rate of ∼2 ml/min at room temperature. Whole-cell patch-clamp electrophysiological recordings were performed with a MultiClamp 700B amplifier (Axon CNS Molecular Devices) and using an infrared-differential interference contrast microscope. Patch microelectrodes (borosilicate capillaries with a filament and an outer diameter of 1.5 μm; Sutter Instrument) were prepared with a four-step horizontal puller (Sutter Instrument) and had a resistance of 3–5 MΩ. Miniature excitatory postsynaptic currents (mEPSCs) and miniature inhibitory postsynaptic currents (mIPSCs) were recorded from pyramidal neurons of the hippocampal CA1 region and the PFC (layers 5/6), as previously described (Murru et al., 2021).
Field excitatory postsynaptic potential (fEPSPs) were evoked (0.05 Hz of frequency) and recorded from the stratum radiatum of the hippocampal CA1 stimulating the Schaffer collaterals (SC) using aCSF-filled monopolar glass electrodes. Stimulus strength was adjusted to give a 50% maximal response. fEPSPs were acquired at 20 kHz and filtered at 5 kHz. For paired-pulse experiments, pairs of stimuli were delivered at 50 ms intervals every 20 s (frequency, 0.05 Hz), and the paired-pulse ratio (PPR) was calculated by dividing the slope of the second response for the first one. SC-CA1 long-term potentiation (LTP) was elicited using a classic stimulation protocol (100 stimuli at 100 Hz; Murru et al., 2017). All the analyses were performed offline with Clampfit 10.1 software (Molecular Devices).
Microelectrode array
Extracellular recordings were carried out with HD-MEA system (BioCam X, 3Brain) using Arena HD-MEA chips equipped with 4,096 electrodes (21 × 21 μm2 in size, 42 μm pitch). Hippocampal brain slices were obtained for patch-clamp experiments (400 µm thickness). Slice activity was recorded under continuous perfusion (4.5 ml/min) in aCSF solution supplemented with the potassium channel blocker 4-aminopyridine (100 μM, Hello Bio). Recordings were performed at full-frame resolution (17,855.5 Hz) and analyses of spike rate, burst rate, burst duration, and spikes inside the burst were conducted offline using dedicated Python routines (SpikeInterface; Buccino et al., 2020). When analyzing hippocampal slices, activated regions were manually identified overlapping slice images taken with a stereomicroscope (Zeiss Stemi 305), and the pseudocolor activity map was visualized on the 3Brain software. For the analysis of the entire hippocampus, electrodes with a spiking rate of >0.05 spike/s were considered, and slices with <20 active channels were discarded. The global synchronization index (SI) was calculated for each active channel as previously described (Kreuz et al., 2007). SI allows identifying clusters of locally synchronized neurons and ranges from 0 (random uncoordinated activity) to 1 (full synchronous activity).
Statistical analysis
Statistical comparisons were performed using GraphPad Prism software. The statistical significance of differences between the two groups was calculated using a nonparametric two-tailed Mann–Whitney test. To compare three or more groups, Kruskal–Wallis followed by Dunn's multiple comparisons test was used. Differences were considered significant at *p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001. n.s. indicates not significant. Pairwise comparisons are shown as brackets.
Results
PCDH9 is enriched at glutamatergic synapses and its expression peaks in the postnatal hippocampus
To examine Pcdh9 function in vivo, we first determined its protein levels in the adult mouse brain by Western blot (WB) analysis. PCDH9 is widely present in the brain areas analyzed, with a more prominent expression in the cortex and hippocampus and undetectable in the other organs assessed, supporting a specific function for PCDH9 in the central nervous system (Fig. 1A,B). Then, we prepared hippocampal lysates from mice at different ages to study PCDH9 expression during neurodevelopment. We found that PCDH9 expression sharply increases at P7 and then progressively decreases and remains stable in the adult mouse (Fig. 1C,D). The first three postnatal weeks are a critical time window for synapse formation and maturation (Li et al., 2010); therefore, PCDH9 upregulation in this time window suggests a potential role in synaptogenesis. To study PCDH9 subcellular distribution, we performed cortex fractionation and mainly detected PCDH9 in the membrane-enriched fraction, as expected for a transmembrane protein (Fig. 1E). In order to investigate PCDH9 localization at the synapse, we coimmunostained DIV18 cultures of cortical and hippocampal neurons for PCDH9 and presynaptic (VGLUT1, excitatory; VGAT1, inhibitory) or postsynaptic markers (PSD95, HOMER1, excitatory; GEPHYRIN, inhibitory; Fig. 1F and Extended Data Fig. 1-1). PCDH9 was moderately enriched in the perinuclear region and clearly present in the neurites. Colocalization analysis (Fig. 1G,H) revealed a broad overlap of PCDH9 and VGLUT1 signals; a moderate PCDH9 colocalization with VGAT1, PSD95, and HOMER1; and largely exclusive profiles with GEPHYRIN, indicating that PCDH9 is mainly present at excitatory synapses, predominantly in the presynaptic compartment. Taken together, these results indicate that PCDH9 is widely present in the brain, with a peak of expression in the hippocampus during the first postnatal weeks, and mainly localizes at excitatory synapses.
PCDH9 is enriched at glutamatergic synapses, and its expression peaks in the first postnatal weeks in the hippocampus. A, Representative PCDH9 Western blot of brain regions and body organs from 2-month-old wild-type mice. Two PCDH9 isoforms at ∼130 and ∼180 kDa are detected, as previously reported (Bruining et al., 2015). OB, olfactory bulb; PFC, prefrontal cortex. B, Quantifications of PCDH9 levels from A. Protein levels were normalized to GAPDH. n = 4. Kruskal–Wallis test, no significant differences. n.d., not detected. C, Representative PCDH9 Western blot on hippocampus from wild-type mice at different ages. E18, embryonic day 18; P3–P180, postnatal day 3–180. D, Quantifications of PCDH9 levels from C. Protein levels were normalized to GAPDH. Mean with SEM is represented. n = 3. Kruskal–Wallis test, **p < 0.01. E, Representative PCDH9 Western blot on cortex from 2-month-old wild-type mice. n = 2. S2, supernatant 2 (cytosol-enriched fraction); P2, pellet 2 (membrane-enriched fraction). PSD95 and ACTIN are used as markers of P2 and S2, respectively. F, Representative confocal images of DIV18 rat hippocampal neurons coimmunostained for PCDH9 and synaptic markers VGLUT1, VGAT, PSD95, and GEPHYRIN. Insets show higher magnification of a dendrite. n = 3 independent cultures. G, Manders’ coefficient quantification of PCDH9 puncta colocalizing with presynaptic (top) and postsynaptic (bottom) markers analyzed in F. VGLUT1, VGAT, n = 13 neurons. Mann–Whitney test, ****p < 0.0001. PSD95, HOMER1, GEPHYRIN, n = 17–27 neurons. Kruskal–Wallis test, *p < 0.05. H, Pearson’s coefficient quantification of PCDH9 signal colocalizing with presynaptic (top) and postsynaptic (bottom) markers analyzed in F. VGLUT1, VGAT, n = 13 neurons. Mann–Whitney test, ***p < 0.001. PSD95, HOMER1, GEPHYRIN, n = 17–27 neurons. Postsynaptic markers, Kruskal–Wallis test, ****p < 0.0001. See Extended Data Figure 1-1 for more details.
Figure 1-1
Coimmunostaining of PCDH9 and HOMER1 in cultured neurons. Confocal images of DIV18 rat hippocampal neurons co-immunostained for PCDH9 and excitatory postsynaptic marker HOMER1. Insets show higher magnification of a dendrite. n = 3 independent cultures. Download Figure 1-1, TIF file.
Pcdh9 deletion leads to enlarged presynaptic and postsynaptic compartments in the CA1
To understand Pcdh9 function in vivo, we took advantage of the Pcdh9 KO mouse line (Bruining et al., 2015). As an initial characterization, we performed MRI analysis on Pcdh9 KO and WT brains. We found significant changes in the volume of multiple brain areas (Extended Data Table 2-1), suggesting that Pcdh9 might play a role in neural proliferation, growth, or migration, similar to other protocadherins (Mancini et al., 2020).
Pcdh9 KO mice display impairments in social and object recognition (Bruining et al., 2015). We could reproduce this evidence in our hands, thus confirming the robustness of the memory phenotype (Extended Data Fig. 2-1A,B). Therefore, we decided to focus our study on the hippocampus, a quintessential region for memory (Lisman et al., 2017). Moreover, hippocampal alterations have been largely documented in MDD (Duman, 2014) and ASD (Schumann et al., 2004; Banker et al., 2021), two pathologies associated with PCDH9 in patients (Marshall et al., 2008; Bucan et al., 2009; Xiao et al., 2018). Strikingly, ultrastructural analysis of excitatory synapses of the dorsal CA1 revealed a significant enlargement of the presynaptic terminal in Pcdh9 KO mice compared with WT animals (Fig. 2A,B). As a consequence, the density of presynaptic vesicles was significantly reduced in Pcdh9 KO neurons, while the pool of docked vesicles was unchanged (Fig. 2A,B, Extended Data Fig. 2-2). Importantly, the overgrowth of the presynaptic terminal was paralleled by a concomitant increase in the length and surface of the postsynaptic density (PSD) and a nonsignificant augmentation of the spine head area (Fig. 2A,B), indicating that both pre- and postsynaptic compartments are enlarged in Pcdh9-depleted neurons.
Pcdh9 deletion leads to aberrant presynaptic and postsynaptic compartments in the CA1. A, Representative electron micrographs of pyramidal neuron synapses from the dorsal CA1 of 2-month-old WT and Pcdh9 KO mice. Scale bar, 100 nm. B, Quantifications of synaptic features from A. n = 30–50 synapses from two mice per genotype. The horizontal solid line across the violin plot represents the median, and the two dotted lines indicate the quartiles. Mann–Whitney test, *p < 0.05, **p < 0.01. C, Representative electron micrographs of pyramidal neurons synapses from the vmPFC layers 5–6 of 2-month-old WT and Pcdh9 KO mice. Scale bar, 100 nm. D, Quantifications of synaptic features from C. n = 50–70 synapses from two mice per genotype. The horizontal solid line across the violin plot represents the median, and the two dotted lines indicate the quartiles. Mann–Whitney test, no significant differences. E, Representative western blot (left) and corresponding quantification (right) of synaptic markers levels on hippocampus from 2-month-old WT and Pcdh9 KO mice. n = 8–15 mice per genotype. Mann–Whitney test, no significant differences. F, Representative Western blot (left) and corresponding quantification (right) of synaptic markers levels on cortices from 2-month-old WT and Pcdh9 KO mice. n = 6–9 mice per genotype. Mann–Whitney test, no significant differences. See Extended Data Figures 2-1 and 2-2 and Extended Data Table 2-1 for more details.
Figure 2-1
Pcdh9 deletion causes defects in social and object recognition A Long-term social recognition assessed as time spent sniffing the familiar versus the novel intruder mouse. The horizontal line across the box plot represents the mean. n = 10 mice per genotype. Mann-Whitney test, *p < 0.05. n.s., non significant. B Long-term object recognition assessed as time spent sniffing the familiar versus the novel object. n = 9-10 per genotype. Mann-Whitney test, *p < 0.05. n.s., non significant. Download Figure 2-1, TIF file.
Figure 2-2
Evaluation of docked vesicles in Pcdh9 KO CA1 neurons Quantification of docked vesicles in CA1 pyramidal neurons from 2-month-old WT and Pcdh9 KO mice. The percentage of docked over total vesicles (left) and the number of docked vesicles per 100 nm of pre-synaptic length (right) are shown. n = 30-50 synapses from 2 mice per genotype. The horizontal solid line across the violin plot represents the median, and the two dotted lines indicate the quartiles. Mann-Whitney test, no significant differences. Download Figure 2-2, TIF file.
Table 2-1
MRI volumetric analysis of brain regions in Pcdh9 KO mouse. Download Table 2-1, XLSX file.
To understand whether Pcdh9’s role in synaptic structure was specific to CA1 or more general in the brain, we also examined the prefrontal cortex (PFC), another brain area strongly implicated in MDD and ASD (Hare and Duman, 2020; Mohapatra and Wagner, 2023). Examination of ventromedial PFC (vmPFC) synapses did not display any difference between the two genotypes, suggesting that the observed effect of Pcdh9 KO on synaptic morphology is specific to CA1 (Fig. 2C,D). Altogether, we showed that Pcdh9 depletion leads to an aberrant size increase both in the presynaptic terminal and in the PSD, demonstrating that Pcdh9 is important for establishing the correct morphology of CA1 excitatory synapses.
Pcdh9 deletion induces the dysregulation of key synaptic genes and the broad upregulation of the synaptic transcriptome in the CA1
Then, we wanted to identify the molecular changes underlying the aberrant synaptic structure in the CA1 of Pcdh9 KO mice. To this aim, we first performed a WB analysis of a panel of pre- and postsynaptic proteins. This approach did not reveal any obvious alterations in synapse composition between the two genotypes, either in the hippocampus or in the cortex (Fig. 2E,F).
To gain a deeper insight into the molecular changes occurring in Pcdh9-depleted neurons, we conducted a single-nucleus RNA-seq (snRNA-seq) experiment on 16,833 cells from WT and Pcdh9 KO hippocampi. Nuclei were separated into 46 clusters that were assigned to neuronal and non-neuronal cell types (Fig. 3A). Differentially expressed genes (DEG) analysis between Pcdh9 KO and WT genotypes was implemented for each cluster. Surprisingly, Pcdh9 was identified as an upregulated DEG in Pcdh9 KO neurons across multiple clusters. This was due to the increased transcription of the Pcdh9 locus downstream of the deleted exon 2, as verified by RT-qPCR (Extended Data Fig. 3-1A). No full-length or truncated PCDH9 protein was detected in Pcdh9 KO hippocampal lysates after WB with antibodies targeting the PCDH9 C-terminal region, confirming the full knock-out of the protein (Extended Data Fig. 3-1B).
Pcdh9 deletions induces the dysregulation of key synaptic genes and the broad upregulation of the synaptic transcriptome in the CA1. A, UMAP plots of snRNA-seq data. A total of 16,833 nuclei from WT and Pcdh9 KO hippocampi were grouped into 46 different clusters and colored based on the attributed cell type. n = 2 mice per genotype. B, UMAP plots of snRNA-seq data, with nuclei colored based on the combined expression levels of CA1 markers. The numbers of the clusters identified as CA1 are in blue. C, Heatmap of CA1 markers fold changes among the different clusters. Clusters identified as CA1 populations are enclosed in a rectangle. D, Volcano plots showing the KO/WT fold change and p-value distributions of genes in CA1 Clusters 5 (top), 6 (center), and 13 (bottom). Upregulated and downregulated DEG are colored in green and red, respectively. E, Network visualization of the combined 31 DEG found in CA1 Clusters 5, 6, and 13. Each node represents a gene and is colored based on its function. A line connecting two nodes indicates a physical or functional interaction, as analyzed with STRING. The line thickness indicates the strength of data support. F, The four GO categories (biological processes) identified with g:Profiler on CA1 DEG are represented. The green and red indicate up- and downregulated DEG, respectively. For each GO category, the statistical significance and the associated genes are represented (brown square, inferred from experiments; orange square, inferred from sequences). G, The three synaptic GO categories identified with SynGO on CA1 DEG. For each GO category, the statistical significance and the associated genes are indicated. H, Representative confocal images of CA1 neurons from 3-month-old WT and Pcdh9 KO mice, immunostained for MAP2 (neuronal marker) and CORTACTIN or SHANK2. Scale bar, 20 µm. I, Quantification of CORTACTIN (top) or SHANK2 (bottom) puncta fluorescence intensity from H. n = 8–12 sections from 3–4 mice per genotype. Mann–Whitney test, *p < 0.05. J, The effect of Pcdh9 deletion on the expression levels of nonsynaptic, presynaptic, and postsynaptic genes in Clusters 5 (top left), 6 (top rigth), and 13 (bottom) is shown. The horizontal line across the violin plot represents the median. Kruskal–Wallis test, ****p < 0.0001. See Extended Data Figures 3-1 and 3-2 and Extended Data Tables 3-1, 3-2, 3-3, 3-4, and 3-5 for more details.
Figure 3-1
Evaluation of Pcdh9 mRNA and protein levels in Pcdh9 KO hippocampus A RT-qPCR analysis of the mRNA levels of Pcdh9 exon 1, 2 and 5 in Pcdh9 KO and WT hippocampi. In Pcdh9 KO mouse line, the whole second exon is replaced with a kanamycin/neomycin cassette (Bruining et al., 2015). n = 3 mice per genotype. Pcdh9 mRNA levels were normalized over Gapdh mRNA levels. Unpaired t-test, **p < 0.01, ***p < 0.001. B Western blot analysis of PCDH9 levels in hippocampal lysates from 2-month-old Pcdh9 KO and WT mice. Download Figure 3-1, TIF file.
Figure 3-2
GSEA of synaptic genes and analysis of excitatory and inhibitory PSD genes in snRNA-seq CA1 clusters A GSEA analysis of all human genes annotated as synaptic across ranked gene expression data of cluster 5 (left), 6 (center) and 13 (right). B The effect of Pcdh9 deletion on the expression levels of inhibitory and excitatory PSD genes in cluster 5 (left), 6 (center) and 13 (right) is shown. The horizontal line across the violin plot represents the median. Mann-Whitney test, *p < 0.05, **p < 0.01. Download Figure 3-2, TIF file.
Table 3-1
DEG between Pcdh9 KO and WT genotypes in hippocampal clusters Download Table 3-1, XLSX file.
Table 3-2
List of combined DEG between Pcdh9 KO and WT genotypes in CA1 clusters Download Table 3-2, XLSX file.
Table 3-3
STRING analysis of CA1 DEG genes Download Table 3-3, XLSX file.
Table 3-4
gProfiler analysis of CA1 DEG genes Download Table 3-4, XLSX file.
Table 3-5
synGO analysis of CA1 DEG genes Download Table 3-5, XLSX file.
Taking advantage of established CA1 marker genes (Cembrowski et al., 2016), we identified seven cell clusters containing CA1 excitatory neurons (Fig. 3B,C). Among the CA1 clusters, Pcdh9 disruption led to the differential expression of 15, 21, and 4 genes in Clusters 5, 6, and 13, respectively (Fig. 3D and Extended Data Tables 3-1, 3-2). Importantly, GO analysis of the combined 31 DEG from the three aforementioned CA1 clusters (Fig. 3E and Extended Data Table 3-3) found enriched categories related to synapse and cell junction (Fig. 3F and Extended Data Table 3-4). In agreement, based on the SynGO database for synapse ontology (Koopmans et al., 2019), multiple DEG were annotated as synaptic genes, mostly related to the postsynaptic compartment (Fig. 3G and Extended Data Table 3-5). Notably, multiple upregulated DEG are known to promote spine formation. These include the PSD scaffold Shank2, a critical regulator of spine shape (Sarowar and Grabrucker, 2016), the PSD components Cttn and Csmd2 required for spine and dendrite morphogenesis (MacGillavry et al., 2016; Gutierrez et al., 2019), and the kinases Camk2a and Dapk1 which act in structural synaptic plasticity (Canal et al., 2011; Goodell et al., 2017). It is remarkable that the top upregulated gene across the CA1 clusters is the noncoding RNA Bc1. Bc1 has been previously implicated in the modulation of spine head and PSD size (Briz et al., 2017); therefore, its dysregulation could also be implicated in the impaired synaptic organization of Pcdh9-depleted neurons. To validate our transcriptomic analysis at the protein level, we immunostained hippocampal sections from WT and Pcdh9 KO mice for the protein products of Cttn (CORTACTIN) and Shank2 (SHANK2). Importantly, we found that both CORTACTIN and SHANK2 levels are significantly increased in Pcdh9 KO CA1 neurons compared with WT (Fig. 3H,I), in agreement with Cttn and Shank2 upregulation detected by snRNA-seq.
Despite the number of DEG reaching statistical significance in CA1 clusters being relatively small, we noticed that a large majority of genes showed a moderate upregulation in Pcdh9 KO neurons compared with WT (Fig. 3D). We asked whether this global upregulation was driven by synaptic genes or rather general across all gene categories. Remarkably, pre- and postsynaptic genes showed a KO/WT fold change significantly higher than the rest of the transcriptome in CA1 clusters (Fig. 3J). Accordingly, gene set enrichment analysis (GSEA) revealed the upregulation of synaptic genes across gene expression data of CA1 clusters (Extended Data Fig. 3-2A). In particular, genes coding for proteins localized to excitatory synapses showed a more pronounced upregulation compared with genes related to inhibitory synapses (Extended Data Fig. 3-2B). Overall, our snRNA-seq approach revealed that Pcdh9 ablation leads to the dysregulation of key regulators of synapse size and to a moderate but broad upregulation of the synaptic transcriptome, thus sustaining the overgrowth of the pre- and postsynaptic compartments in CA1 excitatory synapses.
Pcdh9 deletion causes alterations in glutamatergic transmission and defective network activity in the CA1
Next, we asked how the synaptic and molecular alterations induced by Pcdh9 depletion would impact neuronal function. Electrophysiological recordings on pyramidal neurons from dorsal CA1 acute slices revealed a significant increase in the frequency of miniature excitatory postsynaptic currents (mEPSCs) in Pcdh9 KO CA1, which doubled compared with WT (Fig. 4A,B). This alteration was specific to excitatory currents, as the recording of miniature inhibitory postsynaptic currents (mIPSCs) did not reveal any difference between genotypes, despite a trend toward reduced mIPSCs frequency in Pcdh9 KO neurons was observed (Extended Data Fig. 4-1A,B). In agreement with electron microscopy observations, no differences in mEPSCs properties between genotypes were observed in vmPFC slices (Fig. 4C,D). The change in mEPSCs frequency could originate from presynaptic alteration in glutamate release; therefore, we performed paired-pulse ratio (PPR) experiments. No alterations in glutamate release probability were observed in Pcdh9 KO mice (Fig. 4E), suggesting that a postsynaptic rather than a presynaptic defect might underlie the alteration in mEPSCs frequency. Next, we examined whether Pcdh9 KO impaired long-term potentiation (LTP). Field excitatory postsynaptic potentials (fEPSPs) responses were recorded before and after high-frequency stimulation (HFS). No difference was found between the two genotypes, indicating that Pcdh9 deletion does not affect this form of synaptic plasticity (Fig. 4F). In conclusion, we found that Pcdh9 disruption leads to enhanced glutamatergic transmission, likely due to postsynaptic mechanisms.
Pcdh9 deletion leads to increased frequency of EPSCs and reduced network activity in the CA1. A, Representative traces of mEPSCs recorded from WT and Pcdh9 KO CA1 pyramidal neurons. B, Quantification of amplitude, decay time, area, and frequency of mEPSCs from A. Mean with SD is represented. n = 16–17 recorded neurons from five independent animals per genotype. Mann–Whitney test, *p < 0.05. C, Representative traces of mEPSCs recorded from vmPFC pyramidal neurons from WT and Pcdh9 KO mice. D, Quantification of amplitude, decay time, area, and frequency of mEPSCs from C. Mean with SD is represented. n = 21–22 recorded neurons from five independent mice per genotype. The horizontal line across the plot represents the mean. Mann–Whitney test, no significant differences. E, Quantification of paired-pulse ratio experiments in CA1 pyramidal neurons from WT and Pcdh9 KO mice. n = 5 independent mice per genotype. Mann–Whitney test, no significant differences. F, Quantification of CA1 fEPSP slope before and after HFS from WT and Pcdh9 KO mice. n = 5 independent mice per genotype. Mann–Whitney test, no significant differences. G, Representative MEA recordings from WT and Pcdh9 KO hippocampal slices. H, Representative raster plots showing CA1 electrical activity recorded from 100 channels over 2 min (top) and representative traces recorded from a single channel positioned in the CA1 (bottom), showing reduced firing activity in Pcdh9 KO mice. I, Quantification of firing rate, burst activity, number of spikes per burst, and burst duration from MEA recordings for each hippocampal subregion in WT and Pcdh9 KO genotypes. The horizontal line across the box plot represents the mean. n = 4–6 independent mice per genotype. Mann–Whitney test, **p < 0.01. J, Synchronization index of the CA1 subregion in WT and Pcdh9 KO genotypes. The mean with SEM is represented. n = 3–6 independent mice per genotype. Mann–Whitney test, *p < 0.05. See Extended Data Figure 4-1 for more details.
Figure 4-1
CA1 mIPSCs analysis and evaluation of whole hippocampus network activity in Pcdh9 KO mice A Representative traces of mIPSCs recorded from WT and Pcdh9 KO CA1 pyramidal neurons. B Quantification of amplitude, decay time, area and frequency of mIPSCs from (A). Mean with SD is represented. n = 9-13 recorded neurons from 2 independent animals per genotype. Mann-Whitney test, no significant differences. C Quantification of firing rate, burst activity, number of spikes per burst and burst duration from MEA recordings for the whole hippocampus in WT and Pcdh9 KO genotypes. The horizontal line across the box plot represents the mean. n = 4-6 independent mice per genotype. Mann-Whitney test, no significant differences. Download Figure 4-1, TIF file.
Finally, we investigated the hippocampal network activity from Pcdh9 KO and WT hippocampal slices using high-resolution microelectrode array (MEA; Fig. 4G). Quantification of firing rate, burst activity, number of spikes per burst, and burst duration in the whole hippocampus did not reveal any difference between WT and Pcdh9 KO genotypes (Extended Data Fig. 4-1C). However, the multivariate ANOVA test of these combined four features indicated a significant difference between the two genotypes (p = 0.009), suggesting that Pcdh9 ablation might lead to disturbances in the hippocampal circuits. Therefore, we examined the network activity within each of the different hippocampal subregions separately. Remarkably, we found a significant decrease in firing rate and burst activity specifically in the CA1, but not in the CA3 or DG, of Pcdh9 KO animals (Fig. 4H,I). Furthermore, Pcdh9-depleted CA1 displayed a decrease in the synchronization index (SI; Fig. 4J). Altogether, these results indicate that the structural, transcriptional, and functional synaptic alterations induced by Pcdh9 deletion at the neuronal level are accompanied by a reduction in CA1 network activity and synchronization (Fig. 5).
Pcdh9 ablation leads to an aberrant transcriptional profile, morphology and function of CA1 excitatory synapse, and defective CA1 network activity. Pcdh9 deletion causes synaptic dysregulation at multiple levels, transcriptional (key synaptic genes alterations and widespread upregulation of the synaptic transcriptome), morphological (presynaptic terminal and PSD overgrowth), and functional (increased mEPSCs frequency) and results in impaired network activity in the CA1.
Discussion
In this study, we explored the physiological role of Pcdh9 in CA1 synapses. First, we determined that PCDH9 localizes at excitatory synapses, predominantly in the presynaptic compartment, with an expression peak in the first postnatal week. Second, we combined ultrastructural, biochemical, transcriptional, and electrophysiological approaches to characterize excitatory synapses in Pcdh9 KO mice. One of the most striking findings of our work is that Pcdh9 deletion leads to synapse overgrowth in the CA1. Previous studies have implicated other protocadherins in spine generation or elimination (Mancini et al., 2020), but their role in shaping synapse morphology has received less attention. Minor changes in synaptic structural features were reported in Pcdh17 KO and Pcdh19 mosaic mice (Hoshina et al., 2013; Giansante et al., 2023). Here, we show that Pcdh9 ablation causes the aberrant enlargement of both presynaptic and postsynaptic compartments, demonstrating Pcdh9 requirement for the control of synaptic size in the CA1. In rodents, synapses acquire their typical structure with normal-sized synaptic vesicles and postsynaptic thickening at P7 (Rice and Barone, 2000; Li et al., 2010). Our observation that PCDH9 hippocampal levels peak at P7, in line with analogous findings in the somatosensory cortex (Bruining et al., 2015), supports the conclusion that Pcdh9 plays a role in synapse maturation. Therefore, we speculate that the synaptic alterations observed in Pcdh9 KO CA1 originate during postnatal neurodevelopment and are then maintained in the adult brain.
Regarding synaptic density, we did not observe any change in the dorsal CA1 of Pcdh9 KO mice, while previous reports found an increased spine number in the somatosensory cortex of these animals (Bruining et al., 2015) and reduced spine counts in the ventral CA1 of a distinct Pcdh9 KO line (Uemura et al., 2022). Collectively, this suggests that Pcdh9’s role in spine generation is highly specific to the brain regions considered, possibly depending on the particular Pcdh9 expression profile and PCDH9 protein interactors in the different brain areas.
Thanks to our single-nucleus transcriptional analysis, we gained insights into the molecular changes underlying synapse overgrowth in Pcdh9 KO neurons. First, we noticed that most genes showed a mild upregulation in Pcdh9 KO neurons compared with WT. Remarkably, this global trend toward increased transcriptional levels was driven by synaptic genes: taken collectively, pre- and postsynaptic genes were upregulated much more strongly than nonsynaptic genes. This moderate but broad upregulation of synaptic components could sustain the overgrowth of the presynaptic terminal and PSD observed in Pcdh9 KO neurons. Second, in Pcdh9 KO neurons, we identified several upregulated DEG which are known to promote synapse growth, such as Shank2, Cttn, Camk2a, Dapk1, and Csmd2. The PSD scaffold Shank2 recruits the actin-regulator Cttn to control actin dynamics and stimulate synapse morphogenesis (Sala et al., 2001; Steiner et al., 2008; MacGillavry et al., 2016), and Shank2 deficiency leads to a reduction in PSD thickness in the CA1 region (Pappas et al., 2017). Therefore, our results showing the concomitant increase in Shank2 and Cttn levels in Pcdh9 KO CA1, both at the transcriptional and protein levels, strongly point to the SHANK2/CORTACTIN pathway as a primary mechanism driving synapse overgrowth in Pcdh9 KO neurons. Furthermore, Camk2a promotes activity-dependent spine enlargement by recruiting the actin cytoskeleton (Curtis et al., 2023) and Dapk1 functions in structural synaptic plasticity by regulating Camk2a binding to the NMDA receptor (NMDAR) subunit GluN2B (Goodell et al., 2017). Also, The PSD component Csmd2 promotes the development and maintenance of dendrites and synapses (Gutierrez et al., 2019). Therefore, the coordinated upregulation of the Shank2/Cttn and Camk2a/Dapk1 pathways, along with increased Csmd2 gene levels, are likely responsible for the synaptic overgrowth observed in Pcdh9 KO neurons.
Further studies are needed to clarify how Pcdh9 deletion mechanistically leads to the transcriptional dysregulation of synaptic genes. Beyond their role as adhesion molecules, protocadherins also function as signaling hubs by regulating multiple intracellular cascades, including pathways controlling gene expression such as Wnt/ß-catenin (Pancho et al., 2020). Moreover, a recent report unveiled a direct role for PCDH9 in the repression of gene transcription in gastric cancer cells, through the interplay with the DNA methylation machinery (Zhang et al., 2024). Thus, PCDH9 signaling might negatively modulate the transcription of synaptic genes, which could explain their upregulation in Pcdh9-ablated neurons. Proteomic studies would be instrumental in determining PCDH9 interactors and identify the signaling axis dysregulated in Pcdh9 KO neurons.
Our snRNA-seq approach unveiled gene expression alterations in other neuronal and non-neuronal clusters besides CA1. DEG in these clusters only partly overlap with those identified in CA1 neurons, possibly suggesting cell-type specific effects of Pcdh9 deletion. Further research is needed to assess Pcdh9’s role in other hippocampal regions and cell types.
Through patch-clamp recording, we found that Pcdh9 disruption positively modulates excitatory transmission. Considering that neither the pool of synaptic vesicles nor their release probability was increased in Pcdh9 KO mice, we hypothesize that the augmented mEPSCs frequency originates from alterations in the postsynaptic compartment. In agreement, we revealed increased levels of several PSD components and key regulators of glutamatergic transmission, such as Shank2 and Camk2a. The SHANK family proteins function as a scaffold at the PSD of excitatory synapses (Jiang and Ehlers, 2013), and their dysregulation causes alterations in postsynaptic currents (Peça et al., 2011; Duffney et al., 2013; Pappas et al., 2017). Moreover, enhanced CaMK2a activity has been proven to alter glutamatergic transmission by reducing the proportion of silent synapses in the CA1 (Poncer et al., 2002) or modifying AMPA receptor gating properties (Kristensen et al., 2011). Therefore, we believe that the augmented levels of these genes might underlie the increased mEPSCs frequency in Pcdh9-depleted CA1 neurons. Furthermore, we observed a downward trend for mIPSCs frequency in Pcdh9 KO neurons, which could also contribute to the positive modulation of excitatory transmission.
While Pcdh9 depletion increases the frequency of excitatory currents in CA1 neurons, MEA analysis revealed that the resulting CA1 network paradoxically exhibits a global reduction of neuronal activity and synchronization. Although we cannot provide a comprehensive mechanistic explanation for our observations, multiple factors could account for the contrasting electrophysiological changes at the network and single-neuron levels. First, the most strongly and significantly upregulated DEG in Pcdh9 KO CA1 neurons is the noncoding RNA Bc1. Intriguingly, Bc1 KO mice exhibit increased brain activity in the cortex and prolonged synchronized discharges in the hippocampus (Zhong et al., 2009; Briz et al., 2017). Therefore, it is reasonable to hypothesize that Bc1 upregulation might induce an opposite effect, thus contributing to the observed reduction in CA1 network activity and synchronization in Pcdh9 KO mice. Second, protocadherins are recognition molecules essential for neuronal wiring during neurodevelopment (Peek et al., 2017; Mancini et al., 2020). It is then conceivable that Pcdh9 deletion might lead to inappropriate assembly of hippocampal circuitries, resulting in defective network activity. Finally, it is worth underlining that contrasting electrophysiological effects at the single neuron and network levels were previously described for other synaptic proteins (Giansante et al., 2023; Lee et al., 2024). In particular, Giansante and colleagues showed increased neuronal excitability and reduced network activity in a mouse line carrying the mosaic deletion of another protocadherin family member, Pcdh19. This aspect further highlights the complex nature of protocadherin-related synaptic regulation.
Growing evidence from human neuropathological studies and animal models indicates that autism may be conceptualized as a disease of the synapse (Toro et al., 2010), especially of the glutamatergic ones (Moretto et al., 2018). Postmortem research on ASD individuals revealed an increase in the spine density of cortical pyramidal neurons (Hutsler and Zhang, 2010). Furthermore, multiple genes regulating PSD organization have been associated with ASD and intellectual disabilities (Toro et al., 2010; Penzes et al., 2011). To this regard, our findings are of great interest, as they reveal that disruption of the autism-associated gene Pcdh9 leads to aberrantly enlarged CA1 synapses and dysregulation of genes controlling synaptic size. Notably, five DEG identified in this work (Camk2a, Dnmt3a, Galnt9, Pabpc1, and Shank2) are present in the SFARI database for ASD risk genes. Further research is required to examine whether ASD patients present dysmorphic synapses in the CA1.
In summary, our multilevel study sheds light on Pcdh9 as a novel regulator of CA1 excitatory synapse size and function, with relevant implications for hippocampal circuit activity.
Footnotes
We thank the animal facility of the University Milano Bicocca for the maintenance of mice colonies, Carlo Besta Neurological Institute for the use of the transmission electron microscope, and the Experimental Imaging Centre of IRCCS-San Raffaele Hospital (Milano) for the MRI analysis. This work was supported by Fondazione Cariplo, Italy (2019-3438) and Ministero della Salute, Bando della Ricerca Finalizzata 2021 Giovani Ricercatori, project GR-2021-12375009—The role of the glutamatergic neurotransmission in post-partum depression: an integrated molecular-pharmaco-imaging study (DRACULA).
The authors declare no competing financial interests.
F.M.’s present address: Instituto de Neurociencias, Consejo Superior de Investigaciones Científicas, Universidad Miguel Hernández, Alicante, Spain. G.M.’s present address: MeLis, CNRS UMR 5284, INSERMU1314, Institut NeuroMyoGène, Université de Lyon, Université Claude Bernard Lyon 1, Lyon, France. V.D.M.’s present address: Neuroscience Research Center, Charité-Universitätsmedizin Berlin, corporate member of Freie Universität Berlin, Humboldt-Universität zu Berlin and Berlin Institute of Health, Berlin, 10117, Germany
- Correspondence should be addressed to Federico Miozzo at fmiozzo{at}umh.es or Maria Passafaro at maria.passafaro{at}in.cnr.it.
