The INO80E at 16p11.2 locus increases risk of schizophrenia in humans and induces schizophrenia-like phenotypes in mice

Summary

Background

Chromosome 16p11.2 is one of the most significant loci in the genome-wide association studies (GWAS) of schizophrenia. Despite several integrative analyses and functional genomics studies having been carried out to identify possible risk genes, their impacts in the pathogenesis of schizophrenia remain to be fully characterized.

Methods

We performed expression quantitative trait loci (eQTL) and summary-data-based Mendelian randomization (SMR) analyses to identify schizophrenia risk genes in the 16p11.2 GWAS locus. We constructed a murine model with dysregulated expression of risk gene in the medial prefrontal cortex (mPFC) using stereotaxic injection of adeno-associated virus (AAV), followed by behavioural assessments, dendritic spine analyses and RNA sequencing.

Findings

We identified significant associations between elevated INO80E mRNA expression in the frontal cortex and risk of schizophrenia. The mice overexpressing Ino80e in mPFC (Ino80e-OE) exhibited schizophrenia-like behaviours, including increased anxiety behaviour, anhedonia, and impaired prepulse inhibition (PPI) when compared with control group. The neuronal sparse labelling assay showed that the density of stubby spines in the pyramidal neurons of mPFC was significantly increased in Ino80e-OE mice compared with control mice. Transcriptomic analysis in the mPFC revealed significant alterations in the mRNA levels of schizophrenia-related genes and processes related to synapses upon overexpressing Ino80e.

Interpretation

Our results suggest that upregulation of the Ino80e gene in mPFC may induce schizophrenia-like behaviours in mice, further supporting the hypothesis that INO80E is an authentic risk gene.

Funding

This project received support from the National Key Research and Development Program of China, National Natural Science Foundation of China, Key Research and Development Projects of Hunan Provincial Science and Technology Department, Science and Technology Innovation team of Hunan Province, etc.

Research in context.

Evidence before this study

Chromosome 16p11.2 is a leading locus in the GWAS of schizophrenia, and Chang et al. previously showed that the schizophrenia risk variants at 16p11.2 regulated the mRNA expression of multiple genes. Integrative analyses combining gene expression in human brains also highlighted several genes linked to risk of schizophrenia, and functional studies implicated potential molecular mechanisms mediated by some of those risk genes in brains.

Added value of this study

Our study reveals that INO80E is a principal schizophrenia risk gene at the 16p11.2 GWAS locus, and overexpressing Ino80e in the mPFC of mice mimics several phenotypes of schizophrenia on the behavioural, cellular and molecular levels.

Implications of all the available evidence

Our findings identify an authentic risk gene that may play crucial roles in the pathogenesis of schizophrenia, and mice overexpressing Ino80e in the mPFC might be a solid animal model in the characterization of schizophrenia aetiology.

Introduction

Schizophrenia is a serious psychiatric disorder with an about 1% lifetime prevalence, and the lifespan of patients is usually shortened by 15–20 years.¹ In the past decades, tremendous progress has been made in characterizing the pathogenesis of schizophrenia, such as genetic components, neurodevelopmental hypotheses, and synaptic abnormalities in the brain.² The prefrontal cortex has been identified as a critical brain region in the pathogenesis of schizophrenia,³ and the density of dendritic spines in the prefrontal cortical pyramidal neurons of schizophrenia patients is significantly decreased compared with controls.⁴ Recent genome-wide association studies (GWAS) also highlighted the involvement of the prefrontal cortex and pyramidal neurons in schizophrenia.⁵ Accumulating evidence suggests that schizophrenia risk genes may play predominant roles in the development of its endophenotypes,⁶ such as impaired prepulse inhibition (PPI) and altered dendritic spines.7, 8, 9 For example, animal models driven by schizophrenia risk genes could exhibit impaired PPI10, 11, 12 and alterations in the density and morphology of dendritic spines.13, 14, 15

Single nucleotide polymorphisms (SNPs) in the 16p11.2 locus exhibited significant associations with the risk of schizophrenia,⁵ and Chang et al. previously identified a functional variant rs4420550 in this area regulating the mRNA expression of a cluster of genes (including MAPK3 and INO80E).¹⁶ In addition, rare microduplications at 16p11.2 also increased the risk of schizophrenia by up to 14.5-folds.¹⁷ Through integration of genetic, transcriptomic and clinical data, Vysotskiy et al. prioritized three schizophrenia risk genes (INO80E, TMEM219, YPEL3) within the 16p11.2 region, and using conditional analysis they confirmed that upregulated INO80E played a pivotal role in driving schizophrenia.¹⁸ Similarly, analysing transcriptomic data from mid-gestational human brains, Walker et al. emphasized the significance of INO80E as an important schizophrenia risk gene in the 16p11.2 locus.¹⁹ INO80E encodes a component of the ATP-dependent chromatin remodelling INO80 complex^20,21 which modulates cortical neurogenesis.²² Repressing Ino80e in mouse embryonic stem cells resulted in the significant differential expression of 1276 genes, including key pluripotency genes such as Oct4, Nanog, Klf4, Esrrb, Tcl1, Tbx3, and Foxd3.²³ Mouse Genome Database showed that heterozygous Ino80e knockout mice may display abnormal locomotor activation and increased body weight.²⁴

In this study, we carried out summary-data-based Mendelian randomization (SMR) analysis using PGC3 schizophrenia GWAS summary statistics and the largest brain cortex eQTL dataset to confirm that upregulated cortical Ino80e expression is linked with an increased risk of schizophrenia. We then overexpressed Ino80e in the mPFC of mice using stereotaxic injection technology, and performed behavioural paradigm evaluations, neuronal sparse labelling assay and transcriptomics analysis to elucidate the potential roles of INO80E in schizophrenia risk.

Methods

Public expression quantitative trait loci (eQTL) analyses in brain tissues

We used two human eQTL datasets containing SNP genotype and RNA sequencing data of brain dorsolateral prefrontal cortex (DLPFC) tissues to examine the effects of rs3814883 on INO80E mRNA expression: (1) BrainSeq Phase 1²⁵ and (2) Genotype-Tissue Expression project (GTEx).²⁶

BrainSeq Phase 1 (http://eqtl.brainseq.org/phase1/eqtl/) contained eQTL associations from 227 unaffected controls (age >17) of Europeans and African Americans, which were analysed using MatrixEQTL R package²⁷ based on the formula: log₂(RPKM+1) ∼ SNP + sex + SNP PCs + expression PCs. From GTEx dataset (https://www.gtexportal.org/), we retrieved the eQTL summary statistics from 175 unaffected controls of Europeans, which were calculated using linear regression between SNPs and normalized expression matrices, covarying top five genotyping PCs, sex, genotyping platforms, and other covariates identified by the Probabilistic Estimation of Expression Residuals (PEER) method.²⁶

Summary-data-based Mendelian randomization (SMR) analysis

SMR (v1.03) was applied to investigate putative causal relationships between genes and risk of schizophrenia via gene expression.^28,29 Since only the genotyping chip, number of cases and controls with breakdown by sex, and number of genotyped SNPs for each dataset are provided in their Supplementary materials and this information is insufficient to conduct additional statistical tests for group differences (controls versus cases) to assess comparability, we utilized summary statistics from the largest schizophrenia GWAS comprising 53,386 cases and 77,258 controls of European ancestry for further SMR analysis.⁵ The cases were collected from multiple countries, including the UK, Sweden, Norway, Spain, Turkey, Serbia, the Netherlands, Germany, Denmark, France, Ireland, Italy, Estonia, Russia, Austria, Bulgaria, Finland, Portugal, Iceland, the USA and Canada. As described in the previous study,⁵ the diagnostic processes were similar and consistent between studies, and the final diagnosis is a best-estimate consensus lifetime diagnosis based on multiple sources of information such as clinical evaluation by research psychiatrists, different types of semi-structured interviews made by trained research nurses and psychiatrists, medical records, course of the disease and data from multiple informants. Diagnosis was made in accordance with the Diagnostic and Statistical Manual of Mental Disorders-Version IV (DSM-IV) or International Classification of Diseases, 10th Revision (ICD-10) criteria. All patients met the diagnostic criteria for schizophrenia, schizoaffective or schizophreniform disorder. Unrelated control subjects were recruited from the same geographic areas as the patients in each study.

The eQTL summary statistics was from BrainMeta v2 including cis-eQTL information from 2865 brain cortex samples in 2443 European individuals.³⁰ Briefly, this study used the SNP genotype and RNA sequencing data in cortical tissues from seven cohorts, including BrainGVEX (N = 384), the Lieber Institute for Brain Development (N = 128), the CommonMind Consortium (N = 469) and the CommonMind Consortium’s National Institute of Mental Health Human Brain Collection Core (N = 129), Mount Sinai Brain Bank (N = 635), Mayo Clinic (N = 288) and ROSMAP (N = 832). The discrepancy between the number of brain cortex samples (N = 2865) and the number of individuals (N = 2443) is because the RNA sequencing data from the Mount Sinai Brain Bank cohort were derived from four different brain cortex regions: BM10 (Brodmann area 10, part of the frontopolar prefrontal cortex), BM22 (Brodmann area 22, part of the superior temporal gyrus), BM36 (Brodmann area 36, part of the fusiform gyrus), and BM44 (Brodmann area 44, the opercular part of the inferior frontal gyrus)³⁰

In the SMR analysis, cis-eQTL genetic variants served as the instrumental variables (IVs), gene expression as the exposure, schizophrenia as the outcome, and all the default parameters (e.g., keeping eQTLs with P_eQTL < 5 × 10⁻⁸ and minor allele frequency >0.01, pruning the SNPs which is located in strong linkage disequilibrium (LD) [r² > 0.9] with the top associated eQTL, and removing the SNPs which are in low LD or not in LD [r² < 0.05] with the top associated eQTL) were adopted. We performed heterogeneity test in dependent instruments (HEIDI) to filter out results with P_HEIDI < 0.01.²⁹ A total of 16,278 genes were retrieved from the BrainMeta v2 eQTL dataset, and genes with P_SMR ≤ 3.00 × 10⁻⁶ (Bonferroni correction, ∼0.05/16278) and P_HEIDI > 0.01 were considered as candidate genes associated with schizophrenia genetic risk.

The INO80E gene expression pattern in the developing human brain and cortical organoids

In order to examine the expression pattern of INO80E in the developing human brain and cortical organoids, we referred to the expression trajectories and single cell expression of INO80E using public database (Gene Expression in Cortical Organoids: http://solo.bmap.ucla.edu/shiny/GECO/ and CoDEx Viewer: http://solo.bmap.ucla.edu/shiny/webapp/).31, 32, 33 In brief, the time of cultured human cortical organoids differentiated from iPSC lines was up to 694 days in vitro, and a total of 62 samples (from 4 individuals, 5 iPSC lines (4 male and 1 female)) at 13 time points were selected for RNA sequencing using Truseq stranded RNA RiboZero Gold (Illumina), and the gene expression was normalized using CQN (without quantile normalization, sqn = FALSE) (v.1.28.0).³¹ RNA sequencing data of 196 samples from 24 individuals (9 female and 15 male) selected from BrainSpan was used for describing the gene expression trajectory at each developmental stage (from 10 postconceptional weeks (PCW) to 19 Months (M)), and the gene expression was quantified using the union exon model in featureCounts.^31,32 The CoDEx Viewer containing the RNA sequencing data of 40,000 cells in human neocortical development during mid-gestation (gestation week GW17 to 18), and gene expression was normalized using Seurat-v2.0.1.³³

Animals

Considering that cyclical alteration of physiology hormones in female mice might affect their behaviours and neurochemical indexes, we selected male mice for behavioural analyses. A total of 54 C57BL/6J wild-type male mice with no specific pathogen infection (6 weeks old and weight: 18 ± 3 g) purchased from GemPharmatech Co., Ltd. (Nanjing, China), and 4–5 mice per a cage (length: 375 cm, width: 160 cm, height: 180 cm, made of high-temperature resistant polycarbonate PC imported from Germany, can withstand high-pressure steam sterilization 121 °C 15 min, the wire spacing and diameter of metallic cover grid respectively is 8 mm and 1.6–4.0 mm) were group-housed with sawdust for 1–2 weeks to adapt to the environment and reduce the stress reactions, and the number of test mice were labelled by cutting the toes of mice. The feeding environment is a clean room special for mice, and equipped with a constant power shadowless lighting system, a 12-h light/dark circle (lights on at 08:00, lights off at 20:00, working illuminance: 100–200 LX) and with controlled temperature (22 ± 1 °C) and relative humidity (50–60%), and the noise is less than 60 dB. The mice were permitted free access to food and water, and its health state were supervised throughout the experimental period.

Ethics statement

All animal experiments were designed and conducted in this study following the guidelines established by the National Advisory Committee for Laboratory Animal Research and were approved by the Ethics Committee of the Kunming Institute of Zoology, Chinese Academy of Sciences (IACUC-RE-2023-12-009). The humane endpoints are set as the test mice are infected or weight loss 10%.

Plasmids construction

We constructed the mouse Ino80e gene expressing vector (pAAV-CAG-mouse Ino80e-HA-P₂A-mcherry) to produce Ino80e-OE viruses. Briefly, the DNA fragment 1, which contains the mouse Ino80e gene infused with an HA tag at the C-terminal, the DNA fragment 2 containing HA-P₂A-mCherry sequences. These fragments were then recombined into the pAAV-CAG-tdTomato vector (#59462, Addgene, RRID: Addgene_59462) at the Kpn I and HindIII restriction enzyme sites using the ClonExpress Ultra One Step Cloning Kit V2 (#C116-01, Vazyme). The recombinant vector (pAAV-CAG-Ino80e-HA-P₂A-mcherry) expresses the target protein Ino80e-HA and the red fluorescent protein mCherry, serving as an indicator for infected brain tissues. The empty vector (pAAV-CAG-tdTomato) is used for producing control viruses. The constructed and control vectors were verified by Sanger sequencing. The amplified primers used were listed as follows:

Ino80e-HA-F1: caaagaattggatccggtaccATGGGGATACCCACCCTCC;

Ino80e-HA-R1: TGCCagcgtaatctggaacatcgtatgg;

P₂A-mCherry-F2: tgttccagattacgctGGCAGCGGCGCCACCAAC;

P₂A-mCherry-R2: aggttgattatcgataagcttTTACTTGTACAGCTCGTCCATGCC.

AAVs packaging and purification

The AAVs packaging and purification were conducted as previously described.³⁴ Briefly, the recombinant or control vectors (10 μg) respectively mixed with packaging plasmids (10 μg AAV-DJ and 20 μg pHelper), and were co-transfected into HEK293T cells using 100 μL 1 mg/mL polyethylenimine (PEI). After 72 h, the HEK293T cells containing target AAVs were collected, resuspended in cell lysis buffer (0.15M NaCl; 20 mM Tris–HCl) and then subjected to four freeze–thaw cycles between liquid nitrogen and 37 °C water for releasing the target AAVs. The 0.5% sodium deoxycholate and 50 U/mL benzonase (#E1014, Sigma) was added to the cell lysis buffer at 37 °C for 30 min, and then centrifuged at 8000×g for 90 min at 4 °C. The supernatant containing AAVs was collected and gently transferred to the 40 PA Seal Tube (#345321A, himac) filled with 15%, 25%, 40%, 54% iodixanol (#D1556, Sigma) step gradient, and then centrifuged at 60,000×g for 3 h at 18 °C. The target AAVs containing 40% iodixanol were collected for further purification using an Amicon ultrafiltration tube (#UFC910096, Millipore), the purified target AAVs were stored at −80 °C. The virus titre was detected by RT-qPCR as our previous study.³⁴

Stereotaxic injection

To ensure an adequate sample size for our animal studies, we utilized our previous behavioural data, published studies, and the formula (Sample size = 2 SD² (Z_α/2 + Zβ)²/d²) mentioned by Jaykaran Charan and N. D. Kantharia to calculate the sample size for behavioural tests.³⁵ First, the formula for calculating sample size can be described as follows: Sample size = 2 SD² (1.96 + 0.842)²/d². This is based on Z^α/2 = Z_0.05/2 = Z_0.025 = 1.96 (from Z table) at type 1 error of 5%, and Z^β = Z_0.20 = 0.842 (from Z table) at 80% power. The Standard deviation (SD) is derived from previous studies or pilot study, and the effect size (d) represents difference between mean values. Second, we used our lab’s previous behavioural data to calculate the SD and referenced prior studies to determine the effect size (d).36, 37, 38 We observed a statistically significant difference of approximately 20% in mean values between groups and set 20% of the mean values as the effect size (d) to calculate the sample size for this behavioural study. Third, using our lab’s previous behavioural data, the calculated sample size ranged from 5 to 177, and we deemed that 10–20 per group in four behavioural tests (Open field test, Y maze, Sugar preference test, and Prepulse inhibition) is reasonable. To adhere to the 3R principle of experimental animal ethics and based on sample sizes used in previous studies, we decided to set the sample size at 15 mice per group. For other animal studies related to molecular pathology, we based the sample size on common practice and 3R principle (reduction, replacement, refinement) for laboratory animals, using 3–5 animals per group. Therefore, 54 C57BL/6J wild-type male mice were used in this study. Among them, 6 mice were used for Western blot detection, 30 mice were used for behavioural analyses, 10 mice were used for neuronal sparse labelling assay, and 8 mice were used for RNA-sequencing analyses and quantitative PCR. In each experimental group, the mice were randomly assigned a number by a researcher #1 using RAND function in Excel. Those mice were respectively injected with the control viruses or Ino80e-OE viruses using stereotaxic injection by another researcher #2, and the rest of the experimental conditions were kept the same. Briefly, the mice were anesthetized with isoflurane gas and secured horizontally in a stereotaxic apparatus (RWD Life Science Co., Ltd.), maintaining continuous anaesthesia. Following skull surface exposure, adjustments were made to ensure a horizontal position. Using coordinates corresponding to the mPFC (AP: +1.8 mm; ML: ±0.4 mm; DV: - 2.3 mm), boreholes were drilled with a skull drill. Subsequently, a microinjection syringe inhaling AAVs (titre = 5 × 10¹² VG/mL) was affixed to the stereotaxic apparatus, slowly descending to the target brain region and injecting 200 nL of AAV on each side at a rate of 2 nL/s. Following completion of the injections, absorbable surgical sutures were used to close the incision on the scalp. The wound was disinfected, and measures were taken to maintain warmth until the mice regained consciousness, at which point they were returned to their cages. After injecting AAVs, those mice were group-housed by researcher #3 who was blinded to mice genotype (i.e., whether the mice received control viruses or Ino80e-OE viruses), and researcher #3 also carried out the following experiments. Researcher #2 (who performed AAV injection) did not participate in the following experiments.

Western blotting

In order to ensure Ino80e was overexpressed in the mPFC of Ino80e-OE group mice, we used fluorescence microscope to check whether the mCherry protein (co-expressing with Ino80e) was expressed in the dissected mPFC tissues after AAVs injection 4 weeks, and three samples in each group (controls and Ino80e-OE) were selected for further validation by Western blot. The mice mPFC tissues which labelled with red fluorescent protein were isolated and homogenized using RIPA lysis buffer added with protease and phosphatase enzyme inhibitors (#89901, Thermo Fisher). Protein concentrations were measured using the BCA protein assay kit (#23225, Thermo Fisher). The protein samples were mixed with 5 × protein loading buffer (#P0015, Biotechnology), denatured at 95 °C for 10 min, separated on a 10% SDS-PAGE gels, and then transferred onto the polyvinylidene difluoride membranes (#03010040001, Millipore). Following blocking with 5% BSA, the transfer membrane was incubated overnight at 4 °C with primary antibodies diluted in 1 × TBST buffer containing 5% BSA. The next day, the transfer membrane underwent a 1-h incubation with the corresponding secondary antibodies at room temperature. The Tanon ECL chemiluminescence reagent (#180–5001W, Tanon) and Tanon 5200 imaging system were used to capture the image of the blotting reaction. The antibodies used in this study were listed as follows: the primary antibodies were Rabbit anti-Ino80e (1:2000, #HPA043146, Sigma–Aldrich, RRID: AB_10794923), Rabbit anti-Ino80e (1:2000, #ab122837, Abcam, RRID: AB_11129131), Mouse anti-HA (1:2000, #901513, Biolegend, RRID: AB_2565335) and Rabbit anti-Tubulin antibody (1:4000–20000, #100941-1-AP, Proteintech, RRID: AB_2210695), and the secondary antibodies were HRP-labelled Goat anti-Mouse (1:10000, #E-AB-1001, Elabscience, RRID: AB_2715613) and HRP-labelled Goat anti-Rabbit (1:10000, #E-AB-1003, Elabscience, RRID: AB_2921220). As the ab122837 has been discontinued, we used the second primary antibody HPA043146 to validate the overexpression of Ino80e in the mPFC of mice. The validation results of primary antibodies were presented in the Reagent Validation file.

Behaviour paradigm evaluations

After AAVs injection 4 weeks, the aforementioned mice (control = 15, Ino80e-OE group = 15) were evaluated by behavioural tests. In order to minimize the effect of humans on murine behaviours, researcher #3 took the test mice individually from cages to touch for 2–3 min daily in the behavioural test room (temperature: 22 ± 1 °C, working illuminance:100–200 LX, noise: less than 60 dB), which also allowed the rest of the mice in the cage to adapt to stress of cohort removal. This process lasted a week until the behavioural test started. All behavioural tests were conducted between 9:00 am and 5:00 pm by researcher #3 who is blinded to the mice genotype group. The researcher and operating computer were out of the behavioural room during the behavioural tests, and about 15–20 LX light intensity was uniformly spilled over behavioural apparatus. The finished mouse was put in a new cage near the original cage to further minimize cohort removal stress. A time interval of 2–5 days between different behavioural tests was used to minimize the cross effect. The behavioural recording and data analysis system were supported by Shanghai Xinsoft Information Technology, and the parameters were fixed during the same behavioural test. After behavioural tests, the mPFC tissues were checked with fluorescence microscope to ensure that the red fluorescence proteins (mCherry) were expressed in all test mice.

Open field test (OFT)

The open field test can be used to evaluate anxiety-like behaviour of animal models.³⁹ The bottom of the floor of open field apparatus (length: 40 cm; width: 40 cm; height: 40 cm) was equally divided into 9 squares, with the centre square defined as the central region and the other 8 squares defined as peripheral regions. We referred to the review about anxiety-like behaviours summarized by Prut and Belzung,³⁹ in which each test mouse, facing the wall, was placed in the same corner of the open field arena and allowed 5 min of free exploration. The whole test was finished in a day, and the activity tracks of all test mice were recorded by a video camera and SuperMaze software. The apparatus was cleaned with a 70% ethanol solution and water after each trial. The test mice being in a state of immobility in the open field task is defined as rest, and the total distance, average speed, central distance, time in the peripheral area, and rest time were used to evaluate the anxiety-like behaviours of the test mice.

Y-maze test

The Y-maze test is often used to study spatial working memory in mice.⁴⁰ The Y-maze, comprising three arms set at a 120° angle, was conducted following previous study.⁴¹ The test mouse was placed in the same arm and permitted to freely explore three arms for 8 min. The activity tracks of all test mice were recorded by a video camera and analysed by SuperMaze software. The apparatus was cleaned with a 70% ethanol solution and water after each trial. The whole test was finished in a day. Spontaneous alternation was defined as entering three different arms consecutively. The spontaneous alternation percent was calculated as the ratio of spontaneous alternation number to total number of arm entries minus 2.

Sucrose preference test (SPT)

The sucrose preference test capitalizes on rodents' natural inclination for sweet flavours to investigate hedonic responses, enabling the assessment of negative-like symptoms of schizophrenia, such as anhedonia.⁴² The sucrose preference test was conducted as previously described.³⁴ Individual mice were housed in cages equipped with two water bottles, one containing purified water and the other containing 1% sucrose solution. The position of two bottles were alternated every 24 h. After 48 h, the mice underwent a 24-h water deprivation period. At the onset of the test, two bottles respectively filled with pure water and 1% sucrose solution were accurately weighed and placed back into the cages carefully. The test spanned 12 h, and the two water bottles were swapped position every 6 h. The whole test was finished in four days, the weight of purified water or 1% sucrose solution water in each test mice were recorded at every specific point for further water or sucrose consumption calculation. The weight of the two bottles was measured again after the test. Sucrose preference percent was calculated as the ratio of sucrose consumption to total consumption (containing sucrose consumption and water consumption).

Prepulse inhibition (PPI)

The PPI reflects the function of sensorimotor gating, a process typically deficient in schizophrenia patients. To evaluate potential impairment in PPI in Ino80e-OE mice, we conducted an acoustic startle reflex test as previously described.¹² The background noise was maintained at 62 dB throughout the experiment. The test mouse was situated in cleaned plastic restrainer and positioned on the gravity-sensing platform inside a soundproof chamber, for a 5-min pre-adaptation period. Initially, ten startle pulse (120 dB, 40 ms) were administered at random intervals, averaging 15 s. Subsequently, five different trials were conducted: (I) Only 74 dB, 78 dB, or 86 dB prepulse stimulation with random intervals; (II) Only 120 dB startle pulses; (III–V) Respectively, 74 dB, 78 dB, and 86 dB prepulse stimulation followed by 120 dB startle pulses stimulation at 100 ms intervals. Each trial was repeated ten times with random intervals, and the average interval was 15 s. The startle responses of the test mouse to pulse stimulation were recorded using an acoustic startle reflex measurement system (#XR-XZ208, XinRuan). All mice were tested in two consecutive days. The PPI percent, calculated as (1 - (average startle amplitude to prepulse)/ (average startle amplitude to shock stimulation)) × 100%, was used to evaluate the sensorimotor gating function of the mice being tested.

Neuronal sparse labelling assay

To evaluate the impact of Ino80e overexpression on the density of dendritic spines in mPFC pyramidal neurons, we conducted neuronal sparse labelling assay following previous studies.^34,43 Briefly, the 100 nL sparse labelling AAVs (titre = 6 × 10¹² VG/mL, NCSP-YFP-2E5,#BC-SL001, Brain Case) were mixed with 100 nL tdTomato AAVs (5 × 10¹² VG/mL) or Ino80e-OE AAVs (5 × 10¹² VG/mL) and randomly injected into the mPFC (AP: +1.8 mm; ML: ±0.4 mm; DV: - 2.3 mm) using stereotaxic brain injection. Four weeks after surgery, the mice (control = 5, Ino80e-OE group = 5) without any stimulations were anesthetized with intraperitoneal ketamine injection (150 mg/kg), and sequentially perfused with cooled PBS and 4% paraformaldehyde (PFA) through the heart. The dissected brain tissues were fixed in 4% PFA at 4 °C overnight, dehydrated in a 30% sucrose solution, and embedded in OCT. Coronal sections (50 μm) containing the mPFC were generated using a frozen sectioning machine. These sections were sealed with an anti-quenching sealant VECTASHIELD® Antifade Mounting Medium (#H-1000-1, Vector). The secondary or tertiary dendrites of pyramidal neurons in the mPFC were randomly acquired on a confocal laser microscope (A1R HD25, ZeissNikon, Z-stack image scanning (64 images, 0.5 μm intervals, 1024 × 1024 pixels resolution)) with an oil immersion 60× objectives and 2× digital zoom. Dendritic spine length measurements were conducted from the maximum intensity projection using Z-stack in Image-J software, while the classification and counting of spines were analysed using the NeuronStudio software. The dendritic spine types were classified as mushroom spines (head-neck ratio more than 1.5), thin spines (head-neck ratio less than 1.5, length/head diameter more than 2), and stubby spines (head-neck ratio less than 1.5, length/head diameter less than 2). The density of different dendritic spine morphologies was quantified as the number of spines per 10 μm of dendrite length. The analysis was performed double-blinded.

Transcriptomic analysis

The total RNA samples were extracted from the mPFC tissues of C57BL/6 J wild-type male mice which randomly injected with control viruses or Ino80e-OE viruses for 4 weeks, and the control mice (n = 4) and Ino80e-OE mice (n = 4) were without any stimulations. TRIzol™ Reagent (#15596026, Invitrogen) was used to extract the RNA, which was then used for RNA sequencing. The 150-bp paired-end reads Fastq files were produced using Ilumina NovaSeg 6000 platform. The fastq data files were retrieved for sequencing quality and trim reads examination using trimmomatic-0.36,⁴⁴ yielding clear paired-end reads. The clean reads were aligned to the GRCm39 mouse genome using STAR.⁴⁵ To quantify mRNA expression levels, we employed FeatureCounts with gene annotations from the Ensembl build GRCm39.⁴⁶ Genes with an average TPM (transcripts per million) < 0.1 were excluded from subsequent differential expression analyses. DESeq2 (R package) was used to analyse the differences in gene expression between control and Ino80e-OE mice.⁴⁷ Genes with false discovery rate (FDR) corrected P-value less than 0.05 were identified as differentially expressed genes (DEGs).

Real time quantitative PCR

To assess the upregulation of the C4b gene in the mPFC tissues of Ino80e-OE mice (n = 4) compared with controls (n = 4), we used the RNA samples which used for transcriptomics analysis to conduct RT-qPCR. Reverse transcription of RNA into cDNA used RevertAid First Strand cDNA Synthesis Kit (#K1622, Thermo Fisher), and the cDNA was further used as temples for C4b and Actb gene mRNA expression detection. RT-qPCR was performed using NovoStart®SYBR qPCR SuperMix Plus (#E099-01A, Novoprotein) and LightCycler 96/LC96 (Roche), and the amplification primers were as follows: C4b-F: 5′-GCTGTGGAATACGCATGGGA-3`; C4b-R: 5′-CTCGGGGATAATAGCAGGCG-3`; Actb-F: 5′-TCCTCGGTATGGAATCTTGC-3`; Actb-R: 5′-GAGTATTTACGCTCAGGTGGG-3`. The expression of C4b gene was normalized to Actb gene’s expression, and relative expression was calculated using the 2^−ΔΔCt method.⁴⁸

SynGO analysis

We used the SynGO annotation tool (SynGO release: 20210225) for enrichment analysis of DEGs involved in synaptic-related processes, including biological processes and cellular components.⁴⁹ The background gene was set as “brain expressed”, the evidence filters were set to “No filters, include all annotations”, and the Min. genes count per term in GSEA settings was set as “3”. For more detailed information, please refer to the SynGO website at https://syngoportal.org/help.

The DEGs expression in brains of schizophrenia patients and controls

We downloaded the gene expression data from PsychENCODE and used it to explore whether the expression of DEGs was dysregulated in the DLPFC of schizophrenia patients compared with controls. Briefly, the expression data of brain DLPFC tissues (schizophrenia patients: 559, controls: 936) were quantified using RNA sequencing. For more detailed information about PsychENCODE, please refer to the PsychENCODE website (http://resource.psychencode.org) and the original paper.⁵⁰

Statistical analysis

All statistical analyses were performed using GraphPad Prism v8, and the normality and equal variances of data were assessed before conducting statistical analysis. The normality was examined using the Shapiro–Wilk test, and the equal variances of data were assessed using the Brown–Forsythe test. The normality and equal variance were satisfied in this study, and t-test (two-tailed) and two-way ANOVA test (multiple comparisons) were selected for statistical analysis in behavioural paradigm evaluation, neuronal sparse labelling assay and RT-qPCR based on the data types (Supplementary Table S1). If the test mice incompletely finished all four behavioural tests, they were excluded from further analyses, and the behavioural data of 22 test mice (control = 11, Ino80e-OE group = 11) were included in the final analyses (Supplementary Table S2). The data analysis was conducted by another researcher #1 who did not participate in the aforementioned experiments. Data are presented as the mean ± SEM. P value less than 0.05 was set as statistically significant.

Role of funders

The funders did not participate in study design, data collection, data analyses, interpretation, or in writing of the report.

Results

The schizophrenia GWAS risk SNPs at 16p11.2 are associated with mRNA levels of INO80E in DLPFC tissues

In the latest schizophrenia GWAS of European populations, including 53,386 cases and 77,258 controls,⁵ SNPs within 16p11.2 locus exhibited genome-wide significant associations, and rs3814883 was the lead SNP (P = 1.58 × 10⁻¹⁴) (Fig. 1a). We investigated the relationship between rs3814883 and mRNA expressions of 16p11.2 genes in the DLPFC tissues using the BrainSeq eQTL dataset (n = 227 controls),²⁵ and the risk allele [C] was significantly associated with higher mRNA levels of INO80E (P = 3.59 × 10⁻²⁰) and MAPK3 (P = 6.69 × 10⁻⁸) (Fig. 1b), but did not show any evidence of association with the expression of TAOK2, HIRIP3, TMEM219, DOC2A, KCTD13 or ASPHD1. To validate these observations, we also examined the eQTL associations in the DLPFC tissues of GTEx dataset (n = 175 controls),²⁶ and rs3814883 was again significantly associated with mRNA levels of INO80E (P = 3.00 × 10⁻⁶) and MAPK3 (P = 1.63 × 10⁻⁵) (Fig. 1b), nevertheless, rs3814883 also showed nominal eQTL associations with the expression of TAOK2 and KCTD13, despite the effect sizes were much smaller (Supplementary Fig. S1). These findings in both BrainSeq and GTEx eQTL datasets suggest that INO80E and MAPK3 are likely the predominant schizophrenia risk genes, and the risk allele [C] at rs3814883 consistently predicted higher mRNA levels of INO80E and MAPK3.

Fig. 1.

INO80E was a schizophrenia risk gene. (a) LocusZoom plot for 16p11.2 in the schizophrenia GWAS of European populations. (b) Association of rs3814883 with mRNA expression of INO80E and MAPK3 in BrainSeq and GTEx DLPFC RNA sequencing samples. Error bars depict mean ± s.d. (c) Manhattan plot for the SMR analyses integrating schizophrenia GWAS summary statistics with BrainMeta V2 cortex eQTL dataset.

Integrative analyses confirm INO80E as a schizophrenia risk gene

To further depict the genetic risk architecture of the 16p11.2 locus and to prioritize schizophrenia risk genes, we applied summary-data-based Mendelian randomization (SMR)^28,29 to identify putative causal relationships between SNPs and schizophrenia via gene expression by integrating the schizophrenia GWAS results (53,386 cases and 77,258 controls of Europeans)⁵ with brain cortex eQTL summary statistics from BrainMeta resource (2865 tissue samples from 2443 European individuals).³⁰ In this integrative analysis, elevated mRNA levels of INO80E were significantly associated with increased genetic risk of schizophrenia (P_SMR = 1.26 × 10⁻¹¹, Fig. 1c) and passed the heterogeneity in dependent instruments (HEIDI) test (P_HEIDI = 0.211). MAPK3 was also significantly associated with schizophrenia in the SMR analyses (P_SMR = 5.92 × 10⁻¹⁰), however, it did not pass the heterogeneity test (P_HEIDI = 8.71 × 10⁻³) and was consequently excluded from further analyses. Thus, INO80E emerged as the sole gene displaying significant associations with risk of schizophrenia in the 16p11.2 GWAS locus, and increased INO80E expression likely contributed to the schizophrenia risk. Across the whole genome, INO80E was also one of the top genes (ranked as 8th according to statistical P_SMR, Fig. 1c), and other significant risk genes included FURIN (P_SMR = 5.36 × 10⁻¹⁷, P_HEIDI = 0.668), CNNM2 (P_SMR = 3.98 × 10⁻¹³, P_HEIDI = 0.0158), PCCB (P_SMR = 8.44 × 10⁻¹², P_HEIDI = 0.906), NOTCH4 (P_SMR = 6.87 × 10⁻⁹, P_HEIDI = 0.0106), and GLT8D1 (P_SMR = 1.45 × 10⁻⁸, P_HEIDI = 0.0620) etc (Supplementary Table S3), which confirms previous findings.^51,52

Overexpressing Ino80e in the mPFC of mice leads to schizophrenia-like behaviours

We examined the expression pattern of INO80E in human cortical organoids and developing brain, showing that the gene reached its peak level and gradually decreased to a stable level before birth, and single-cell data from the developing human neocortex showed widespread expression of INO80E in various brain cell types (Supplementary Fig. S2),³¹ supporting its role in cortical function. The DLPFC is considered as a critical brain region for schizophrenia pathogenesis,³ and corresponds to the rodent mPFC associated with schizophrenia-related behavioural abnormalities.⁵³

We therefore utilized AAV stereotactic injection to overexpress Ino80e (i.e., Ino80e-OE) in the mPFC of wild-type mice (7–8 weeks old), and carried out four evaluations for schizophrenia-related behavioural paradigms, including the open field test, Y-maze test, sucrose preference test and acoustic startle reflex test (Fig. 2). The timeline of AAV stereotaxic injection and behavioural tests is presented in Fig. 2a. Initially, we confirmed the site of AAV stereotaxic injection labelled with red fluorescent protein and found that AAV were restricted to the targeted brain region (Fig. 2b). Brain tissues infected with AAV were isolated, confirming the Ino80e protein was overexpressed in Ino80e-OE mice (Fig. 2c and Supplementary Figs. S3–S5).

Fig. 2.

The mice overexpressing the Ino80e gene in the mPFC exhibited schizophrenia-like behaviours and PPI impaired. (a) Timeline of the behavioural tests. (b) Schematic illustration of stereotaxic injection of AAV-Ino80e or AAV-tdTomato in the mPFC (left), representative image showing expression of red fluorescent protein at the injection sites (middle), and coordinates indicating the injection sites in the brain (right). (c) Efficiency of Ino80e gene overexpression in the mPFC. n = 3 mice per group. (d) Overview of the open field test. (e) Representative movement heatmap traces of the Ino80e-OE and control mice during the open field test. (f) The total distance, average speed, centre distance, time in the peripheral area, and rest time of the Ino80e-OE and control mice were measured during the open field test. Ino80e-OE mice showed less centre distance, more time in the peripheral area, and more rest time than control mice. (g) Overview of the Y-maze test. (h)Ino80e-OE mice did not show a notable difference in spontaneous alternation compared with control mice. (i) Overview of the sucrose preference test. (j)Ino80e-OE mice showed a lower sucrose preference compared with control mice. (k) Overview of the PPI test. (l and m) There is no difference in the startle amplitude of the mice to a 120 dB pulse stimulus between Ino80e-OE mice and control mice. However, Ino80e-OE mice showed less PPI than control mice in response to a 120 dB pulse stimulus at 78 dB and 86 dB prepulse levels. Statistical tests: Two-tailed t-tests (c, f, h, j, l) and two-way ANOVA tests (m). n (control) = 11, n (Ino80e-OE) = 11, ∗P < 0.05, ∗∗P < 0.01, ∗∗∗P < 0.001, ∗∗∗∗P < 0.0001. Error bars indicate the standard error of the mean (s.e.m.). Abbreviations: OFT (Open Field Test), NOR (Object Recognition Test), SPT (Sucrose Preference), PPI (Prepulse Inhibition Test), Ino80e-OE (Overexpression of Ino80e), AP (Anteroposterior), ML (Mediolateral), DV (Dorsoventral), PrL (Prelimbic Cortex). See Supplementary Tables S1 and S2 for statistical analysis.

All test mice were confirmed to be without any obvious health problems before the behavioural tests. During the open field test, no significant differences were observed in the total travelled distance and average movement speed between Ino80e-OE and control mice (total travelled distance: Control, 22725 ± 1602 mm, Ino80e-OE, 19806 ± 981 mm, df = 20, t = 1.554, P = 0.136, two tailed t-test; average movement speed: Control, 37.88 ± 2.67 mm/s, Ino80e-OE, 33.02 ± 1.63 mm/s, df = 20, t = 1.551, P = 0.137, two tailed t-test; Fig. 2d–f). This indicates that the locomotion ability of Ino80e-OE mice was unimpaired. However, Ino80e-OE mice exhibited a significant decrease in travelled distance within the centre area compared with control mice, along with a significant increase in time spent in the peripheral area and the rest time (travelled distance in centre area: Control, 1248 ± 127 mm, Ino80e-OE, 894.9 ± 84.5 mm, df = 20, t = 2.311, P = 0.032, two tailed t-test; time in peripheral area: Control, 573.2 ± 2.29 s, Ino80e-OE, 582.5 ± 1.8 s, df = 20, t = 3.194, P = 0.005, two tailed t-test; rest time: Control, 4.194 ± 0.9 s, Ino80e-OE, 10.74 ± 1.96 s, df = 20, t = 3.042, P = 0.006, two tailed t-test; Fig. 2f). These findings indicate that the anxiety-like behaviour of Ino80e-OE mice travelling around the peripheral area was increased, and there was a decreased desire to explore the novel environment. Spatial working memory of the test mice was evaluated by Y-maze test, and the result showed that the alternation of test mice did not differ between Ino80e-OE and control mice (Control, 57.82 ± 2.37%, Ino80e-OE, 63.18 ± 1.07%, df = 20, t = 2.066, P = 0.052, two tailed t-test; Fig. 2g–h). In addition, we found that the sucrose preference of Ino80e-OE mice was significantly decreased compared with control mice (Control, 84.45 ± 1.25%, Ino80e-OE, 69.44 ± 3.46%, df = 20, t = 4.075, P = 0.0006, two tailed t-test; Fig. 2i and j), which is an indicator of anhedonia reflecting one of the schizophrenia negative symptoms.

The sensorimotor gating function, reflected by PPI, is impaired in schizophrenia patients.^8,9 We conducted an acoustic startle reflex test to evaluate whether overexpression of Ino80e affects PPI. The amplitude of the startle reflex response to the 120 dB startle stimulation showed no significant difference between Ino80e-OE and control mice (Control, 1086 ± 153.8 a.u., Ino80e-OE, 1199 ± 130.19 a.u., df = 20, t = 0.56, P = 0.581, two tailed t-test; Fig. 2k and l), indicating that the hearing of the mice was normal. We continued to assess PPI by measuring the reduction in the amplitude of the acoustic startle response after exposure to prepulse (74 dB, 78 dB, 86 dB). Obviously, the inhibitory effect in response to the 120 dB startle stimulation in Ino80e-OE mice was significantly attenuated when exposed to prepulse at 78 dB and 86 dB (percent of PPI-74 db: Control, 20.87 ± 5.38, Ino80e-OE, 10.12 ± 5.53, df = 60, t = 1.31, P = 0.479, two-way ANOVA test; percent of PPI-78 db: Control, 63.53 ± 5.63, Ino80e-OE, 31.39 ± 6.54, df = 60, t = 3.92, P = 0.0007, two-way ANOVA test; percent of PPI-86 db: Control, 67.81 ± 5.64; Ino80e-OE, 33.60 ± 6.00, df = 60, t = 4.17, P = 0.0003, two-way ANOVA test; Fig. 2k and m), indicating that the sensorimotor gating function of Ino80e-OE mice was disrupted. The detailed statistical analysis of all data presented in Fig. 2 was listed in Supplementary Table S1, and the whole processed data obtained from four behavioural tests was listed in Supplementary Table S2.

Overexpressing Ino80e in the mPFC of mice affects the density of dendritic spines

Many studies have shown that the density of dendritic spines is decreased in the prefrontal cortical pyramidal neurons of schizophrenia patients.^4,54,55 We evaluated the impact of Ino80e overexpression in the mPFC of mice on the density of dendritic spines using neuronal sparse labelling assay (Fig. 3). The timeline of AAV stereotaxic injection and neuronal sparse labelling assay is presented in Fig. 3a. We administered sparse labelling AAV mixed with Ino80e-OE AAV or control AAV into the mPFC region of experimental mice using stereotaxic brain injection (Fig. 3b). Our findings revealed no alteration in total dendritic spines density between Ino80e-OE and control mice (total spines per 10 μm: Control, 6.638 ± 0.158, Ino80e-OE, 6.655 ± 0.127, df = 8, t = 0.088, P = 0.682, two tailed t-test; Fig. 3c and d). We then assessed the spine density of various types and found that the density of stubby spines in mPFC pyramidal neurons was significantly increased in Ino80e-OE mice compared with control mice, whereas the density of mushroom spines was slightly decreased in Ino80e-OE mice (stubby spines per 10 μm: Control, 0.984 ± 0.0496, Ino80e-OE, 1.579 ± 0.0977, df = 24, t = 3.804, P = 0.0026, two-way ANOVA test; mushroom spines per 10 μm: Control, 2.768 ± 0.120, Ino80e-OE, 2.379 ± 0.0933, df = 24, t = 2.413, P = 0.0697, two-way ANOVA test; Fig. 3c and e). These results demonstrated that overexpressing the Ino80e gene in the mPFC of mice affected the density of dendritic spines. The detailed statistical analysis of all data presented in Fig. 3 was listed in Supplementary Table S1, and the number of neurons analysed in neuronal sparse labelling assay was listed in Supplementary Table S4.

Fig. 3.

The density of dendritic spines in the mPFC of mice was affected by overexpression of the Ino80e gene in vivo. (a) Timeline of the neuronal sparse labelling assay. (b) This schematic illustrates the stereotaxic injection of sparsely labelled viruses mixed with AAV-Ino80e or AAV-tdTomato in the mPFC (left and middle). It also shows the viral load of AAVs injected into the brain (right). (c–e) Morphological analysis of dendritic spine density in the mPFC of test mice. Scale bars represent 5 μm. (d) There was no difference in total density. However, (e)Ino80e-OE mice showed a lower density of mushroom spines and a higher density of stubby spines compared with control mice. Statistical test: Two-tailed t-tests (d) and two-way ANOVA test (e), Ino80e-OE mice (n = 5) and control mice (n = 5). ∗P < 0.05, ∗∗P < 0.01, ∗∗∗P < 0.001. Error bars indicate the standard error of the mean. See Supplementary Tables S1 and S4 for statistical analysis.

Overexpressing Ino80e in the mPFC of mice alters expression of schizophrenia-related genes

Wang et al., identified 1276 differentially expressed genes in mouse embryonic stem cells upon repression of Ino80e.²³ In our study, we overexpressed Ino80e in the mPFC of mice, and transcriptomic analysis by RNA sequencing revealed 28 differentially expressed genes in Ino80e-OE mice compared with control mice (P_adj < 0.05) (Fig. 4a–d and Supplementary Table S5). Literature review of these genes found that four of them (C4b, Tfrc, Lars2, Klf6) were genetically associated with schizophrenia.56, 57, 58 In addition, eleven genes (C4b, Tfrc, Ywhaq, Btbd2, Ltbp3, Nptx1, Lrrn3, Rps6ka4, Islr, Skiv2l2, Hsp90b1) exhibited differential expression (P < 0.05) in the DLPFC of schizophrenia patients in PsychENCODE (http://resource.psychencode.org/),⁵⁰ and six of them (C4b, Tfrc, Ywhaq, Btbd2, Ltbp3, Lrrn3) demonstrated consistent directional effects between results in PsychENCODE (schizophrenia cases versus controls) and our in house RNA sequencing analysis (Ino80e-OE versus control) (Supplementary Table S6). Notably, the schizophrenia risk gene C4b, upregulated in schizophrenia patients compared with controls (P = 2.07 × 10⁻¹¹), exhibited the most significant upregulation after overexpressing Ino80e (RNA sequencing P_adj = 3.78 × 10⁻⁶, Fig. 4c and d) and was verified by RT-qPCR (Fig. 4e).

Fig. 4.

Overexpressing the Ino80e gene in the mPFC of mice regulated the expression of schizophrenia-related genes and synapse-related biological processes. (a) Timeline of transcriptomics analysis. (b) Schematic illustration of the stereotaxic injection location. (c) Volcano plot illustrating the DEGs in the Ino80e-OE mice compared with control mice (P_adj < 0.05). (d) Heat map showing the expression levels of the twenty-eight DEGs, with 16 upregulated and 12 downregulated, between Ino80e-OE (n = 4) and control mice (n = 4). (e) The mRNA expression of C4b gene was upregulated in the mPFC of Ino80e-OE mice compared with control mice using RT-qPCR. (f–h) SynGO annotations showed that cellular components (such as postsynaptic ribosome, presynaptic ribosome, and synapse) and biological processes (translation at the postsynapse, translation at the presynapse, and processes in the synapse) were significantly enriched among DEGs. See Supplementary Table S1 for statistical analysis.

We further conducted synGO analysis using the SYNGO website (https://syngoportal.org/) and observed significant enrichment of cellular components, including postsynaptic ribosome, presynaptic ribosome, and synapse, among the 28 differentially expressed genes (Fig. 4f and h). Three biological processes (translation at postsynapse, translation at presynapse, and process in the synapse) were also significantly enriched (Fig. 4g and h). These results indicate that Ino80e may influence the regulation of schizophrenia-relevant gene and synapse-related biological processes.

Discussion

As a component of the chromatin remodelling complex, INO80E intricately regulates transcriptional control during cortical development.²² Our analyses underscore a noteworthy upregulation of INO80E in the DLPFC, emphasizing its relevance to susceptibility in schizophrenia. We employed stereotactic injection and AAV-mediated overexpression of Ino80e in the murine mPFC to reveal the potential impacts underlying schizophrenia pathophysiology. We observed that Ino80e-OE mice spent more time navigating the peripheral area compared with control mice, indicating Ino80e-OE mice exhibited anxiety-like behaviour. Given the known role of the mPFC in modulating reward-seeking behaviour by regulating the dynamic interactions with specific distant subcortical regions,⁵⁹ our findings revealed that Ino80e-OE mice exhibited an unwillingness to distinguish between the sucrose water and pure water, suggesting a negative-like symptom (anhedonia). We further revealed that the sensorimotor gating function of Ino80e-OE mice was defective. The mPFC is an essential brain region for sensorimotor gating,⁶⁰ and either decreases or increases activity in the mPFC has been linked to reduced PPI,⁶¹ indicating that the neural circuits connected with the mPFC are impaired by overexpressing the Ino80e gene.

Neuronal dendritic spines, serving as fundamental structures for synaptic formation, typically receive input from an excitatory synapse, and their stability is closely associated with synaptic function.^62,63 Dendritic spine abnormalities are associated with schizophrenia, for example, the density of dendritic spines on prefrontal cortical pyramidal neurons was significantly decreased in schizophrenia patients.⁴ The dendritic spines are mainly classified into mushroom, stubby, and thin types based on their head size and neck length. To investigate the impact of overexpressing Ino80e on the density of dendritic spines, we employed a neuronal sparse labelling assay. Although the density of total spines remained unchanged, the density of stubby spines on prefrontal cortical pyramidal neurons exhibited significant alterations in Ino80e-OE mice compared with control mice. The types of dendritic spine are often related to synaptic function, for instance, head size is correlated with postsynaptic density area, the number of postsynaptic receptors, and the quantity of presynaptic vesicles (indicators of synaptic strength).64, 65, 66 The neck of a spine can filter membrane potential, with its length related to the degree of electrical isolation,⁶⁷ and the neck can also regulate the synaptic strength, as studies have shown that the amplitude of excitatory postsynaptic potentials is inversely proportional to the length of the dendritic spines neck, which means that the longer neck of the dendritic spine leads to weaker electrical signals. However, following synaptic stimulation, the neck can be shortened and synaptic efficacy can be increased.⁶⁸ Consequently, mushroom spines are thought to form stronger and more stable synaptic connections,⁶⁹ and thin spines play an important role in synaptic plasticity by enlarging its head when stimulated.⁷⁰ Stubby spine usually lacks a neck and is mainly present in the developmental stage, it does not adequately respond to dynamic changes in synaptic transmission, possibly due to the disappearance of the mushroom.⁷¹ Importantly, in schizophrenia relevant animal models, a reduction in the density of mushroom spines and an increase in stubby spines have been observed in the prefrontal cortex,⁷² which is consistent with our results. However, an increased spine density has been observed in the frontal cortex of Roman-High Avoidance rat strain (RHA-I), a schizophrenia-relevant genetic animal model (a stop codon results in deletion of the metabotropic glutamate receptor 2 (mGluR 2) protein), while the stubby and mushroom spines showed no differences, and their further analyses identified 203 differentially expressed genes that involved in synaptic function.^73,74 Therefore, we further used the 28 DEGs to conduct synGO analysis, and found that increased expression of Ino80e may also affect the regulation of synaptic-related biological processes. More interestingly, transcriptome results demonstrate that the expression of C4b was significantly increased in Ino80e-OE mice. C4B is also a schizophrenia risk gene,⁵⁸ and encodes a subtype of complement component 4 that mediates synaptic pruning.¹⁵ This suggests that upregulation of C4b gene expression may be a crucial factor contributing to the slightly decrease in the density of mushroom spines in Ino80e-OE mice.

We acknowledge several limitations in this study. First, although we observed that the behavioural phenotypes in Ino80e-OE mice were obviously changed compared with controls, the limitations cannot be ignored that the aforementioned abnormal behaviours in Ino80e-OE mice also need to be verified through overexpressing Ino80e in other brain regions or specific neural cells. Second, the Ino80e-OE mice did not exhibit cognitive defects in the Y-maze test, it does not necessarily imply the absence of cognitive defects in other aspects in the Ino80e-OE mice. To comprehensively assess the cognitive abilities of mice, more sophisticated cognitive tests may be warranted. Third, due to the complexity of linkage disequilibrium and regulatory mechanisms, we cannot completely exclude the possibility that other genes (such as TAOK2 and KCTD13) in 16p11.2 region also conferred or co-conferred risk of schizophrenia, and the higher expression of INO80E may increase risk of a broad variety of human disorders (e.g., 16p11.2 duplication syndrome, autism, depression),^75,76 but rather than solely schizophrenia. Fourth, the mechanisms underlying the upregulation of C4b gene expression resulting from overexpression of the Ino80e gene require further investigation. Despite observing a correlation of 0.69 between the expression of these genes in the RNA sequencing data, further investigation is needed to determine whether Ino80e overexpression enhances the activity of the chromatin remodelling INO80 complex, leading to the upregulation of C4b gene expression. The intriguing question of whether the schizophrenia-like behaviours and sensorimotor gating defects in this study arise from changes in neuronal connectivity or disruptions in neural circuits due to increased C4 expression remains to be explored. Fifth, in the SMR analysis, we used summary statistics from the largest schizophrenia GWAS of European ancestry. However, there may be potential for confounding in the group comparisons, as we are uncertain about the representativeness of the cohorts included in the SMR analysis. Sixth, theoretically, a larger sample size increases the likelihood of detecting differences between groups; however, considering the costs (such as time and resources) associated with the sample size of mice, as well as animal ethics considerations, we must calculate the sample size accurately. During the behavioural tests, four mice in each group were not included in the statistical analysis, which may impact our behavioural results and necessitates further validation in the future. Finally, we realized that the 16p11.2 duplication syndrome, which is associated with schizophrenia and features many genes from the region overexpressed, may be included in the GWAS and eQTL samples we utilized, and thus potentially biased the results. Despite the prevalence of 16p11.2 duplications is relatively low (∼0.26% in schizophrenics and ∼0.025% in general populations),⁷⁷ and the estimated impact of 16p11.2 duplications on the overall results is likely to be minimal, we are still cautious in the interpretation of the results.

In summary, we have characterized the critical functional implications of the schizophrenia risk gene, INO80E, on sensorimotor gating function, density of dendritic spines, and biological processes associated with synapses. We have also provided a potential mechanistic insight, dependent on the upregulation of the INO80E gene, contributing to the pathophysiology of schizophrenia.

Contributors

Q.H.T., S.W.L. and M.L. oversaw the project, designed the project's research program and drafted the first version of the manuscript. B.H., M.Y.Y., C.Y.Z. and Z.S. carried out the primary experiments and data analyses and ensured the assays were double blind. Specifically, B.H., M.Y.Y., C.Y.Z. and Z.S. carried out the primary experiments and data analyses. Specifically, during animal experiments, M.Y.Y. was responsible for the blind randomization of mice, B.H. was responsible for all mice behavioural experiments, C.Y.Z. was responsible for statistical analysis of data after behavioural experiments; S.W.L. and C.Y.Z. were responsible for all bioinformatics analysis. In the subsequent experimental process, M.Y.Y. and B.H. were responsible for the verification of the basic data in the study. In addition, S.W.L., Q.H.T. and Z.S. were responsible for the examination of all raw data of the study. L.W. and X.M.L. assisted in the conduct and analysis of the experiment. All authors read and approved the final manuscript.

Data sharing statement

The data used or produced in this study can be accessed from the corresponding authors for reasonable request. The raw RNA-Seq data in the mPFC tissue of Ino80e-OE and control mice has been deposited in the Gene Expression Omnibus repository (accession number GSE276239).

GTEx: https://www.gtexportal.org/home/.

SYNGO website: https://syngoportal.org/.

PsychENCODE: http://resource.psychencode.org/.

BrainMeta: https://yanglab.westlake.edu.cn/software/smr/#DataResource.

PGC3 Schizophrenia GWAS: https://figshare.com/articles/dataset/scz2022/19426775.

Gene Expression in Cortical Organoids: http://solo.bmap.ucla.edu/shiny/GECO/.

CoDEx Viewer: http://solo.bmap.ucla.edu/shiny/webapp/.

Declaration of interests

The authors have no conflicts of interest to declare.