Autophagy defects at weaning impair complement-dependent synaptic pruning and induce behavior deficits

Abstract

Autophagy is crucial for synaptic plasticity and the architecture of dendritic spines. However, the role of autophagy in schizophrenia (SCZ) and the mechanisms through which it affects synaptic function remain unclear. In this study, we identified 995 single nucleotide polymorphisms (SNPs) across 19 autophagy-related genes that are associated with SCZ. Gene Set Enrichment Analysis (GSEA) of data from the Gene Expression Omnibus public database revealed defective autophagy in patients with SCZ. Using a maternal immune activation (MIA) rat model, we observed that autophagy was downregulated during the weaning period, and early-life activation of autophagy with rapamycin restored abnormal behaviors and electrophysiological deficits in adult rats. Additionally, inhibition of autophagy with 3-Methyladenine (3-MA) during the weaning period resulted in aberrant behaviors, abnormal electrophysiology, increased spine density, and reduced microglia-mediated synaptic pruning. Furthermore, 3-MA treatment significantly decreased the expression and synaptosomal content of complement, impaired the recognition of C3b and CR3, indicating that autophagy deficiency disrupts complement-mediated synaptic pruning. Our findings provide evidence for a significant association between SCZ and defective autophagy, highlighting a previously underappreciated role of autophagy in regulating the synaptic and behavioral deficits induced by MIA.

Introduction

Schizophrenia (SCZ) is a severe psychiatric illness that affects approximately 1% of the global population [1]. Patients exhibit highly heterogeneous clinical features, including positive symptoms (such as hallucinations, delusions, and thought disorders), negative symptoms (such as apathy and social withdrawal), and cognitive impairments [1]. The onset and progression of SCZ are influenced by the interaction between environmental and genetic factors, but the underlying pathological mechanisms remain largely unclear.

Epidemiological studies have suggested that prenatal maternal infection may play a role in the development of neurodevelopmental disorders in offspring, including SCZ [2]. In rodent models, immune activation at specific gestational time points, particularly between embryonic day (E) 9 and E16, can alter the neurodevelopmental trajectory of offspring, leading to abnormal synaptic development, which is a primary cause of behavioral abnormalities, such as impaired sensory gating [3]. The maternal immune activation (MIA) rodent model effectively simulates the clinical phenotypes and pathophysiological features of SCZ [2].

Autophagy is an evolutionarily conserved process responsible for the degradation and recycling of cellular components. It serves as a critical quality control mechanism in cells, helping to mitigate stress. In the central nervous system, autophagy is constitutively active and plays a neuroprotective role, while also being directly involved in neuronal development, including axon growth and guidance, synaptic pruning, and synaptic plasticity [4, 5]. Dysregulation of autophagy can contribute to the development of neurological disorders [4,5,6]. Enhancing autophagy has been shown to improve synaptic structure, synaptic plasticity, and cognitive function in models of fragile X syndrome, as well as in rodents with drug- or stress-induced behavioral impairments [7,8,9]. Several studies have reported autophagy defects in SCZ, including findings from autopsy and nasal biopsy samples [10]. Gene enrichment analysis of RNA sequencing from the dorsolateral prefrontal cortex has demonstrated significant downregulation of the autophagy pathway in SCZ [11]. These findings prompted us to investigate the association between autophagy and SCZ further and to determine whether interventions targeting the autophagy pathway could reverse behavioral deficits in MIA offspring. Additionally, the specific mechanisms by which autophagy defects contribute to abnormal behaviors and impaired synaptic development in SCZ need to be elucidated.

Material and methods

Genomic variation in autophagy-related genes in SCZ

A list of 232 autophagy-related genes was obtained from the Human Autophagy Database (http://www.autophagy.lu/) and is provided in Supplementary Table S1. Single nucleotide polymorphisms (SNPs) within these autophagy-related genes were analyzed using data from the largest genome-wide association study (GWAS) of SCZ to date, conducted by the Psychiatric Genomics Consortium 3. This study includes 67,390 cases of SCZ or schizoaffective disorder and 94,015 healthy controls. More detailed information can be found at http://www.med.unc.edu/pgc. The ANNOVAR tool [12] was used to annotate the identified SNPs.

GSEA of autophagy-related genes in SCZ

Expression data from brain tissue (specifically the prefrontal cortex) of patients with SCZ were retrieved from SZDB (http://www.szdb.org/) using datasets GSE12649 and GSE62191. Autophagy gene sets were downloaded from the Human Autophagy Database (http://www.autophagy.lu/). The normalized expression data from GSE12649, GSE62191, and the autophagy-related gene sets were uploaded into GSEA 4.3.2 software for analysis. A normalized enrichment score (NES) greater than 1 and a false discovery rate (FDR) of less than 0.25 were used to assess the magnitude of enrichment and statistical significance, respectively [13].

Animals

Sprague–Dawley (SD) rats were supplied by Beijing Vital River (Beijing, China) and maintained at 25 °C under a 12-h light/dark cycle. All animal experiments were performed according the Guide for the Care and Use of Laboratory Animals. This research was approved by the Animal Care and Use Committee of the Henan Key Laboratory of Biological Psychiatry. Based on the reported reduction of a crucial autophagy-related protein in the hippocampus of SCZ patients by Merenlender-Wagner et al. [14] and the autophagy defects in the prefrontal cortex of SCZ patients identified through GSEA in our study, we selected the hippocampus and prefrontal cortex for further analysis in this study.

Animal treatment

The MIA rat model was constructed as previously described. Briefly, pregnant rats were injected intravenously with equal volume of saline or 10 mg/kg polyinosinic:polycytidylic acid (Poly I:C) (Sigma, P0913) on E 9.5. Male offspring from each group were selected according to a previous study [3] and were designated as control offspring and MIA offspring, respectively. For the autophagy agonist experiment, MIA and control offspring were intraperitoneally injected with either 10 mg/kg of rapamycin (Solarbio, R8140) or an equal volume of saline daily for 1 week, starting from postnatal day 21 (PND 21). For the autophagy inhibitor experiment, 3-MA (Aladdin, M129496) was diluted in saline, and male SD rats were intraperitoneally injected with either 3 mg/kg [15] of 3-MA or an equal volume of saline daily for 1 week, starting from PND 21.

Behavioral tests

Behavioral tests were conducted on PND 60. All tests were performed during the dark phase of the light/dark cycle.

Open field test

The rats were placed in the center of the open field device for 13 min, with their exploration recorded during the last 10 min. The obtained data were analyzed as a putative indicator of anxiety-like behavior.

Elevated plus maze test

The rats were positioned on the central platform facing an open arm, and their exploration was recorded for 5 min. The time spent exploring the open arms was analyzed to evaluate anxiety-like behavior.

Y-maze test

The Y maze device consisted of three arms, each marked with different colors: the initial, familiar, and novel arms. Initially, the rats were placed in the initial arm, and allowed to explore the initial and familiar arms for 10 min. After a 2-h interval, the rats were repositioned in the initial arm, and their exploration of all three arms was recorded for 5 min. To assess spatial recognition memory, the time spent exploring the novel arm was analyzed.

Novel object recognition test

The test was performed in the open field device. One day before the test, the rats were acclimatized in the open field for half an hour. In the first phase of the test, two same objects were positioned in the open field, and the exploration was allowed for 10 min. After 2 h, the objects were substituted by new objects (with different materials and shapes), and the exploration was recorded for 10 min. Preference to the new objects was analyzed to determine the memory of each rat.

Social communication test

The device contained three chambers and openings. In the first phase, two empty cages were placed in the side chambers. During the second phase, a stranger rat (Stranger 1) was positioned in one of the cages to evaluate sociability. In the third phase, another stranger rat (Stranger 2) was positioned in the remaining cage to assess social novelty. All rats were placed in the middle chamber and the exploration was recorded for 10 min during each phase. The resulting data were analyzed to determine social communication.

Pre-pulse inhibition (PPI) test

PPI test was carried out using a startle reflex measurement system (MED Associate). The rats were acclimated in the device for 5 min prior to testing. The experiments comprised 5 trials: background, startle stimulation, and 3 combined trials with prepulse stimuli. The background noise was fixed at 65 dB. Before applying the 120 dB 40 ms startle stimulus, an interval of 100 ms was set for the 20 ms prepulse stimulus (75, 80 or 85 dB). Each trial was repeated 10 times, with an interval time of 15 s and a fluctuation rate of ± 8 s (ranging from 7 to 23 s). The percentage of PPI was calculated as [1 − (startle amplitude on pre-pulse trial/startle amplitude on pulse alone)] × 100%.

In vivo electrophysiology

The anterior fontanelle served as the reference point, and the electrode was targeted at the right hippocampus (X = 2.7 mm, Y = 3 mm, Z = 3.6 mm). A hole was drilled above the right hippocampus, and the electrode was implanted into the brain at a rate of 1 mm/min. Seven days post-surgery, local field potentials (LFPs) were collected. The LFP data were amplified with a gain of 5000, filtered with a high-pass cutoff at 250 Hz, sampled at 1 kHz, and recorded for 5 min. Data acquisition was performed using a Cereplex recording system (Blackrock), and the data were analyzed with NeuroExplorer 5.

RNA isolation and expression analysis

Total RNA was isolated using TRIzol (Takara, T9108). Relative mRNA quantification was conducted using the QuantiTect SYBR Green RT-PCR kit (Qiagen, 204245) on a StepOne Real-Time PCR System (ABI Q6 System, Applied Biosystems). The copy number of each target mRNA was normalized to GAPDH.

Western blotting

The tissue homogenates were lysed in RIPA buffer (Beyotime, P0013B) containing 1 mM PMSF (Beyotime, ST506) and kept on ice for 30 min, followed by centrifugation (12,000×g, 15 min, 4 °C). Total protein content was determined using the BCA protein assay kit (Beyotime, P0012). The protein samples (40 µg) were separated through SDS-PAGE, and transferred onto PVDF membranes (Millipore, ISQE00010). The membranes were blocked with 5% non-fat milk for 1 h, and exposed to primary antibodies against Iba1 (GeneTex, GTX100042), PSD95 (Cell Signaling Technology, 36233S), SYN (Abcam, ab8049), Beclin (Proteintech, 66665-1-1g), P62 (Proteintech, 18420-1-AP), LC3 (Proteintech, 14600-1-AP), C1qa (Abcam, EPR14634), CR3 (Boster, BA0590), C3b (Abcam,2B10B9B2), Actin (Boster, MA1000) and GAPDH (Boster, M00227) at a 1:1000 dilution overnight at 4 °C. After rinsing thrice with TBST for 10 min each, the membranes were exposed to the corresponding HRP-conjugated antibody (Proteintech) at room temperature for 1 h. The membranes were rinsed thrice with TBST, the protein samples were visualized on an Amersham Imager 600 system (GE RPN2232) using an Enhanced Chemiluminescence Plus detection kit. The protein bands were analyzed using FluorChem HD2 software, with GAPDH or actin used as an internal control to normalize band intensity.

Golgi-Cox staining

Brain tissue sections (thickness = 150 μm) along the coronal plane were placed on gelatine-covered slides. Golgi-Cox staining was conducted using the FD Rapid GolgiStain™ Kit (FD NeuroTechnologies Inc., USA). Data were recorded using a microscope (Nikon Ci-L) and analyzed with ImageJ software. Pyramidal neurons in the hippocampus and prefrontal cortex were used to analyze the morphology and density of spines.

Immunofluorescent staining

The brain tissue sections (10 µm) were prepared using a frozen slicer (Leica, CM1950). Antigen retrieval was conducted with sodium citrate in a microwave oven, and then rinsed thrice with PBS. The sections were incubated with 0.5% Triton X-100 for 10 min, and washed with PBS. After blocking with 1% BSA at 37 °C for 1 h, the sections were exposed to the primary antibodies overnight. The sections were rinsed with PBS, and then exposed to secondary antibodies (A0568, A0562, A0521 and A0516; Beyotime; dilution: 1:500) for 1 h. Hoechst 33258 staining (Beyotime, C1018) was conducted for 10 min. After rinsing with PBS, the stained sections were visualized using a microscope (Leica, STELLARIS 8). Fluorescence intensity analysis of each image was performed using ImageJ software (NIH, USA). For the quantification of PSD95-positive puncta, serial z-stack confocal images (acquired at 0.5-μm intervals) were used to reconstruct three-dimensional projection images. PSD95-positive puncta and microglia were identified in these three-dimensional projections, and the number of PSD95-positive puncta per microglia was counted.

Synaptosome fractionation

Synaptosomes were isolated from rats according to a previous method [16]. Briefly, the tissues were immersed in 10 volumes of 0.32 M sucrose buffer with 5 mM HEPES (pH 7.4), followed by homogenization with a glass-teflon homogenizer. To separate the nuclear fractions, the obtained homogenates were spun at 800–1200×g. To produce crude synaptosomes, the homogenates were further spun at 15,000×g and layered onto a discontinuous sucrose gradient (1.2 M sucrose, 0.8 M sucrose). After spinning at 150,000×g for 2 h, the purified synaptosomes were extracted. The enrichment of proteins in synaptosomes, normalized to equal total protein amounts, was verified by Western blotting.

3-MA treatment of primary hippocampal neurons

Primary rat hippocampal neurons were isolated from the brains of E18 pups as previously described [17]. On the 16th day in vitro, cells were incubated with either normal medium or medium containing 5 mM 3-MA [17, 18]. After 48 h of incubation, the cells were harvested and lysed in RIPA buffer. The lysates were then subjected to Western blotting as described above.

Statistical analysis

All statistical tests were conducted with SPSS v20.0 software or GraphPad Prism 8.0. Unpaired two-tailed Student’s t-test was used to compare differences between two groups. One-way ANOVA was used for comparation of three groups. Normality was assessed using the Shapiro–Wilk test or Kolmogorov–Smirnov test. If data did not meet normality assumptions, appropriate nonparametric tests (Mann–Whitney U test or Kruskal–Wallis test) were conducted. Statistical significance is indicated by * for p < 0.05 and ** for p < 0.01, respectively. Data are expressed as mean ± SEM.

Results

Autophagy-related genes are associated with SCZ

A total of 232 autophagy-related genes were extracted from the Human Autophagy Database. Using data from the Psychiatric Genomics Consortium 3 SCZ GWAS, p-values for SNPs within these 232 autophagy-related genes were aggregated, identifying 19 significant genes and 995 significant SNPs after correction (Fig. 1A). Detailed information is provided in Supplementary Table S2. Among these, 20 SNPs in 8 genes were located in exonic regions, with 7 being nonsynonymous mutations (rs1317826, rs3810450, rs3810449, rs1149, rs2306899, rs13107325, and rs1193851) across 5 genes (AMBRA1, CAPNS1, EIF4EBP1, NFKB1, and RELA) (Table S3). These findings suggest an association between autophagy and SCZ. Gene expression datasets from SCZ brain samples (GSE12649, GSE62191) were analyzed using GSEA. The results indicated that autophagy-related genes were significantly enriched in healthy controls (FDR q-values were 0.07 and 0.22 for GSE62191 and GSE12649, respectively, both < 0.25), as shown in Fig. 1B. This suggests the downregulated expression of autophagy-related genes in SCZ. One overlapping gene (AMBRA1) was identified between the genetic and expression data.

Fig. 1

The association between autophagy and SCZ. A Enrichment of SCZ GWAS variants in autophagy-related genes. B GSEA of autophagy-related genes by using SCZ GEO database

MIA offspring rats exhibit behavioral deficits

To further investigate the relationship between autophagy and SCZ, we used MIA rat model, which closely mimics the clinical phenotype and pathophysiology of SCZ. Pregnant rats were intravenously injected with 10 mg/kg Poly I:C or saline on E 9.5. As displayed in Fig. 2B, after Poly I:C injection for 3 h, the expression levels of maternal peripheral cytokines (e.g., TNF-α, IL-1β and IL-6) were significantly upregulated, indicating successful activation of maternal immunity (Fig. 2B).

Fig. 2

MIA leads to behavioral abnormalities and autophagy deficiency. A Experimental design. At gestation 9.5 days, 10 mg/kg of Poly I:C or saline were injected into pregnant rats, cytokine was detected 3 h after injection; the male offspring were designated as saline or Poly I:C, followed by weaning at PND21, and behavioral tests were performed at PND56. B IL-1β, IL-6 and TNF-α in maternal serum were increased after Poly I:C injection. C, D MIA led to increased anxiety-like behavior as indicated by reduced exploration of central area in the open field test (C) and reduced open arm exploration in elevate plus maze (D) in rats of Poly I:C group. E, F MIA led to impaired memory as indicated by reduced exploration in novel arm in Y maze test (E) and reduced exploration for novel object in novel object recognition test (F) in rats of Poly I:C group. G MIA led to impaired social communication as indicated by reduced exploration time with Stranger 1 than empty cage in sociability test in rats of Poly I:C group. H MIA led to sensory gate function deficiency as indicated by reduced inhibitory efficiency of PPI in rats of Poly I:C group. I, J LC3-II expression and LC3-II/LC3-I ratio was reduced in rats of Poly I:C group at PND21. K Representative fluorescence images of LC3 in the CA1, DG and CA3 region of hippocampus and in the prefrontal cortex, and statistical analysis of fluorescence intensity showed decreased fluorescence signal of LC3. n = 10–15 for behavioral tests, n = 3 for Western blotting and n = 4 for immunofluorescent staining. *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001. Data are presented as mean ± SEM

A series of behavioral tests were conducted on PND 60. The Open field test and elevated plus maze test were performed to examine anxiety-related behavior in rats. In the Open field test, there was no difference in the total distance traveled between MIA and control offspring. However, MIA offspring spent less time in the center area of the open field (Fig. 2C). In the EPMT, MIA offspring had fewer entries and spent less time in the open arms (Fig. 2D). These findings demonstrate that anxiety levels are elevated in MIA offspring. Cognitive abnormalities were further evaluated by Y-maze test and novel object recognition test. In the Y-maze test, MIA offspring showed fewer entries and spent less time in the novel arm (Fig. 2E). In the novel object recognition test, MIA offspring spent less time exploring the novel object compared with control rats (Fig. 2F). Collectively, these findings demonstrate that MIA induces cognitive deficits in the adult offspring.

Three-chamber test was performed to assess the social behavior of rats. In the social interaction tests, MIA offspring spent less time interacting with an age-matched peer (Strange 1) than control offspring (Fig. 2G). This result indicates that MIA offspring have social interaction deficits. Next, in the social novelty test, two groups spent similar time sniffing the Strange 2 or Strange 1 (Fig. 2G). Therefore, it is speculated that MIA offspring exhibit specific deficits in social behavior.

PPI test was performed to assess the sensory gating function, a core indicator of SCZ. The results of PPI test showed that the inhibitory effect of prepulse was significantly reduced in MIA offspring compared with control offspring, indicating that the sensory gating function is impaired (Fig. 2H).

Altogether, these behavioral test results demonstrate that MIA can increase anxiety levels, induce memory and social behavior deficits, and impair sensory gating function in adult offspring.

Autophagy is down-regulated in MIA offspring rats at weaning stage

Macroautophagy is initiated with the formation of autophagosomes, which depends on the lipidation of LC3-I to produce LC3-II. We measured the levels of LC3-I and LC3-II and the LC3-II/I ratio in the hippocampus and prefrontal cortex of MIA and control offspring at P21, P45 and P60 (corresponding to weaning stage, adolescence and early adulthood, respectively). At P21, the levels of LC3-II and the LC3-II/I ratio were remarkably diminished in the hippocampus and prefrontal cortex of MIA offspring (Fig. 2I, J), indicating reduced autophagosome synthesis during the weaning stage. Next, immunofluorescence staining was conducted to assess the distribution of LC3. The punctate distribution of LC3, indicative of newly formed autophagosomes, was markedly decreased in MIA offspring at P21 compared to control rats, consistent with reduced autophagosome formation at this stage (Fig. 2K).

Activation of autophagy at weaning stage ameliorates abnormal behaviors and reverses the increase in hippocampal oscillations in MIA rats

To clarify the relationship between autophagy defects at weaning stage and abnormal behavior in adulthood, MIA offspring rats were treated with rapamycin during weaning to induce autophagy and analyze its influence on adult behavior (Fig. 3A, B).

Fig. 3

Activation of autophagy at weaning partly rescues behavioral abnormality in MIA offspring. A Experimental design. On PND 21, MIA offspring were intraperitoneally injected with rapamycin or saline for 7 consecutive days, and behavior testing was conducted on the PND 60. B Rapamycin treatment increased LC3-II expression. C, D Rapamycin treatment partly decreased anxiety-like behavior in rats of Poly I:C group, as indicated by the increased exploration of central area in the open field test (C) and increased open arm exploration in elevate plus maze (D). E, F Rapamycin treatment partly improved the impaired memory in rats of Poly I:C group, as indicated by increased exploration in novel arm in Y maze test (E) and increased exploration for novel object in novel object recognition test (F). G Rapamycin treatment partly improved impaired social communication in rats of Poly I:C group, as indicated by increased exploration time with Stranger 1 than empty cage in sociability test. H Rapamycin treatment did not improve sensory gate function deficiency, as indicated by the lack of effect on the inhibitory function of PPI. However, the decreased PPI observed in Poly I rats was restored by saline treatment at PND 21. I Both saline and rapamycin treatment increased the startle response in rats of Poly I:C group. n = 10–12 for each group. *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001. Data are presented as mean ± SEM

Open field test and elevated plus maze test were conducted to evaluate anxiety behavior. In our study, the adult MIA rats treated with rapamycin at weaning period showed increased numbers of entry into center area in open field as well as open arms in elevate plus maze, suggesting the decreased anxiety of rats (Fig. 3C, D). In the Y-maze test and novel object recognition test, MIA offspring with rapamycin treatment spent more time in novel arms as well as increased tendency of preference for the novel object, indicating the enhanced memory function (Fig. 3E, F). In addition, the results of social communication test demonstrated that rapamycin-treated MIA offspring spent more time with Stranger 1 rat than placebo-treated MIA offspring in the social test session (Fig. 3G). The social novelty test was not conducted because MIA offspring did not show deficits in social novelty, as demonstrated in Fig. 2G. Furthermore, the PPI test was performed. We found that the decreased PPI observed in the Poly I group was restored by saline treatment, while rapamycin treatment had no significant effect on sensory gating function (Fig. 3H). Additionally, the startle response in the Poly I group during the PPI test was increased, suggesting that the PPI test may not be appropriate (Fig. 3I).

To further evaluate the effect of rapamycin treatment on neural oscillations, we measured LFPs in the hippocampus. Compared to the saline + vehicle group, the power of Theta, Delta, Alpha, Beta, and Gamma oscillations was significantly increased in the Poly I:C + vehicle group. However, these alterations were effectively reversed by rapamycin treatment (Fig. 4).

Fig. 4

Rapamycin reversed the increase in the Delta, Theta, Alpha, Beta and Gamma band power in the hippocampus of MIA offspring. A Time frequency diagrams in the hippocampus. B Power spectra of local field potential in the hippocampus. C Quantification of the total Delta, Theta, Alpha, Beta, and Gamma band power in the hippocampus. *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001. Data are presented as mean ± SEM

Inhibition of autophagy at weaning stage induces abnormal behavior and increases hippocampal oscillations in adulthood

To further confirm the role of autophagy defects in early development to abnormal behavior in adulthood, SD rats were treated with 3-MA, an inhibitor of autophagy, at weaning period, and its effect on adult behaviors was analyzed (Fig. 5A). The 3-MA-treated rats at weaning stage had significantly decreased expression level of LC3-II, indicating that the autophagy process is impaired (Fig. 5B).

Fig. 5

Autophagy deficiency at weaning period leads to behavioral abnormality. A Experimental design. On PND 21, the offspring of SD rats were intraperitoneally injected with 3 mg/kg 3-MA or saline for 7 consecutive days, and behavior testing was conducted on the PND 60. B 3-MA treatment decreased the expression of LC3-II. C, D 3-MA treatment at PND 21 led to increased anxiety-like behavior as indicated by reduced exploration of central area in the open field test (C) and reduced open arm exploration in elevate plus maze (D). E, F 3-MA treatment at PND 21 led to impaired memory, as indicated by reduced exploration in novel arm in Y maze test (E) and reduced exploration for novel object in novel object recognition test (F). G 3-MA treatment at PND 21 led to sensory gate function deficiency, as indicated by reduced inhibitory efficiency of pre-pulse. H 3-MA treatment at PND 21 led to impaired social communication, as indicated by reduced exploration time with Stranger 1 than empty cage in sociability test and preference for Stranger 1 than Stranger 2 in social novelty test. n = 10 for behavioral test for each group. *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001. Data are presented as mean ± SEM

Open field test and elevated plus maze test were carried out to examine the anxiety behavior. In the open field test, the rats treated with 3-MA showed a decrease in distance moved and spent less time in the center area of the open field, while there was no difference in total distance traveled (Fig. 5C). In the elevated plus maze test, 3-MA-treated rats showed fewer entries and spent less time in the open arms (Fig. 5D). Cognitive behavior was evaluated by Y-maze test and novel object recognition test. In the Y-maze test, 3-MA-treated rats had fewer entries and spent less time in novel arm (Fig. 5E). Compared with the control rats, the rats treated with 3-MA spent less time exploring the novel object (Fig. 5F). The results of PPI test showed that 3-MA treatment at weaning period did not change the inhibitory effect of prepulse in adult rats, indicating that the autophagy defects in early developmental period are not associated with sensory gating function in adulthood (Fig. 5G). In the social communication test, the rats in 3-MA group spent less time with Stranger 1 compared with control rats in the social interaction tests (Fig. 5H). In the social novelty test, the control rats spent more time interacting with Stranger 2 than Stranger 1, while the 3-MA-treated rats spent more time interacting with Stranger 1 than Stranger 2, indicating the impaired social novelty (Fig. 5H).

Additionally, LFPs in the hippocampus were measured to assess the effect of rapamycin treatment on neural oscillations. Compared to the control group, the power of Theta, Delta, Alpha, Beta, and Gamma oscillations was significantly increased in the 3-MA group, effectively simulating the abnormal electrophysiology observed in the MIA rat model (Fig. 6).

Fig. 6

Autophagy deficiency at weaning period leads to the increase in the Delta, Theta, Alpha, Beta and Gamma band power in the hippocampus. A Time frequency diagrams in the hippocampus. B Power spectra of local field potentials in the hippocampus. C Quantification of the total Delta, Theta, Alpha, Beta, and Gamma band power in the hippocampus. *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001. Data are presented as mean ± SEM

Inhibition of autophagy at weaning stage impairs synaptic pruning

Synaptic pruning is a fundamental neurodevelopmental process that refines synaptic connectivity during the first 3 weeks of life. To investigate whether autophagy defects during the weaning period can lead to behavioral abnormalities by affecting synaptic pruning in early brain development, a series of experiments were conducted. Dendritic spines were visualized by Golgi-Cox staining. Our results showed the numbers of dendritic spines in the hippocampus and prefrontal cortex were markedly increased in 3-MA treatment group compared with control group (Fig. 7B–D). However, 3-MA treatment had no significant impact on the expression levels of PSD95 and SYN, the markers of synapse (Fig. 8B, C).

Fig. 7

Autophagy deficiency at weaning period leads to synaptic abnormalities. A Experimental design. On PND 21, the offspring of SD rats were intraperitoneally injected with 3 mg/kg 3-MA or saline for 7 consecutive days, and Golgi-Cox staining was conducted on PND 28. B–D Golgi-Cox staining of pyramidal neurons in the prefrontal cortex (B) and hippocampus (D), and the quantitative analysis of spine (C, E). The rats treated with 3-MA showed increased spine density compared to control rats. The right panels are high magnifications of the boxed area in the left panels (B, D), scale bar = 2 μm. n = 20 segments for each rat, and 3 rats for each group were analyzed. *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001. Data are presented as mean ± SEM

Fig. 8

Autophagy deficiency at weaning period impairs synaptic pruning by microglia. A Schematic illustration of synaptic pruning by microglia. B, C 3-MA treatment at PND 21 had no significant effects on the protein expression of PSD95, SYN and Iba1. D The content of protein Iba1 in synaptic fraction was significantly reduced in the brain of 3-MA rats. E Quantitative analysis of Iba1 in synaptic fraction. F The number of PSD95-positive puncta per microglia was significantly reduced in the rats of 3-MA, as indicated by white arrows and analysis. Scale bar = 10 μm. n = 20 cells for each group. *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001. Data are presented as mean ± SEM

Microglia-mediated synaptic pruning is crucial for the precision of neural circuits (Fig. 8A). The marker Iba1, which indicates microglial presence, showed no change in rats treated with 3-MA (Fig. 8B, C). To determine whether the increased spine density is due to impaired microglia-mediated synaptic pruning, discontinuous sucrose gradient was employed to isolate synaptosomes and their associated membrane components [19]. The results showed that the content of Iba1, a marker of microglia, was decreased in the synaptosomes fraction from the hippocampus of 3-MA-treated rats (Fig. 8D, E). In addition, the co-localization of PSD95- and Iba1-positive immunoreactivities was decreased in 3-MA-treated rats compared to control animals (Fig. 8F). Therefore, we speculate that autophagy defects during early brain development may increase dendritic spine density by reducing microglial synaptic elimination.

Autophagy defects impair complement-mediated synaptic pruning

One key mechanism by which microglia prune synapses is through complement-dependent processes (Fig. 9A). To elucidate the role of autophagy defects in synaptic pruning, the complement system was analyzed. Our results showed that the mRNA level of complement 4 (C4) was significantly reduced in the brain of 3-MA group (Fig. 9B). Western blotting results demonstrated that the expression levels of C1q and C3b were remarkably downregulated in the brain of 3-MA-treated rats (Fig. 9C) and primary neurons after 3-MA treatment (Fig. 9D, E). The results of synapse isolation showed the decreased levels of C1q and multiple C3b in the synaptosomes of 3-MA treated rats (Fig. 9F, G). In addition, the content of CR3, which is responsible for the recognition and binding of synaptic C3b on the microglia (Fig. 9A), was significantly reduced in the synaptic fraction (Fig. 9F, G). This suggests that 3-MA treatment can reduce the recognition of synaptic C3b and microglia CR3.

Fig. 9

Autophagy deficiency at weaning period impairs complement-mediated synaptic pruning by microglia. A Schematic illustration of complement-mediated synaptic pruning by microglia. B, C 3-MA treatment led to the reduced mRNA expression of C4 (B) and reduced protein level of C1q and C3b (C) in the hippocampus. D Schematic illustration of the primary neuron experiment. E 3-MA treatment increased the expression of P62, while decreased the expression of Beclin, C1q and C3b. The expression levels of PSD95 and SYN remained unchanged. F The protein expression levels of C1q, C3b and CR3 in synaptic fraction were significantly reduced in the brain of 3-MA rats. G Quantitative analysis revealed a decrease in the content of C1q, C3b and CR3 in synaptic fraction. *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001. Data are presented as mean ± SEM

Discussion

Our study provides evidence that SCZ is associated with defective autophagy, based on genetic and GSEA data. To our knowledge, this is the first study to reveal decreased autophagy in the hippocampus and prefrontal cortex of MIA offspring during the weaning period. Additionally, we found that activating autophagy with rapamycin in MIA offspring during weaning ameliorates abnormal behaviors and electrophysiological deficits observed in adulthood. Furthermore, our results show that inhibiting autophagy during the weaning period impairs complement-dependent synaptic pruning, increases dendritic spine density, and induces abnormal behaviors in adulthood. These findings suggest that autophagy defects are a key pathological mechanism in SCZ and highlight the importance of autophagy in complement-mediated synaptic pruning during early development.

Growing evidence has suggested the possible role of autophagy in mental disorders [5, 6]. Postmortem studies have reported the decreased expression of ATGs, such as BECN1, ULK2 and ATG3, in the BA22 of postmortem specimens from SCZ patients, which affects signaling pathways, autophagosome formation, lipid modification, and other key steps of autophagy [20,21,22]. Autophagy defects in the hippocampus, particularly in the CA2 region, are associated with social skill deficits in SCZ patients [14]. Barnes et al. conducted transcriptome analyses of the BA22 and prefrontal cortex in SCZ patients and found that dysregulated autophagy was a prominent feature in the BA22 but not in the BA10 region, suggesting that dysregulated autophagy may be specific to certain brain regions [21]. This finding contrasts with our GSEA results, which may be due to differences in detection techniques, brain tissue subregions, or patient clinical characteristics. Further research is needed to clarify the status of autophagy in the prefrontal cortex of SCZ patients. In autism spectrum disorder (ASD) patients, reduced LC3-II (a marker of autophagy) and elevated p62 (an autophagy substrate) indicate impaired autophagy [23, 24]. The causal relationship between autophagy and ASD is supported by studies using autophagy-deficient mice [8, 24, 25]. Although the direct antipsychotic effects of autophagy induction are not fully established, many antipsychotics increase LC3-II expression in vitro through various mechanisms [20, 26, 27]. Rapamycin, an mTOR inhibitor, can induce autophagy and restore autistic behaviors in mice [23, 24]. Compared with previous studies, our results not only demonstrate the association between autophagy defects and behavioral abnormalities but also provide novel evidence that autophagy defects during lactation lead to behavioral abnormalities in MIA offspring in adulthood, and that inducing autophagy can ameliorate these behaviors. It is important to recognize that while rapamycin is commonly used to induce autophagy, it also has complex biological roles, such as immunomodulation. A reciprocal experiment using an autophagy inhibitor may help address this limitation. Besides, further studies using specific reagents targeting the autophagy pathway or disease-associated autophagic proteins are necessary to elucidate the precise role of autophagy.

Numerous studies have indicated that maternal infection is a risk factor for neurodevelopmental disorders (e.g., ASD and SCZ); however, the underlying mechanisms remain unclear [2]. The “two-hit” model of neural development emphasizes that early genetic and/or environmental factors can disrupt the nervous development (the “first hit”) and increase individual’s vulnerability or susceptibility to subsequent environmental stresses (the “two hit”) [28]. We hypothesize that MIA impairs individuals’ ability to adapt to various stresses after birth, resulting in cumulative deficits that manifest as abnormal behaviors in adolescence or early adulthood. The cessation of lactation is a significant physiological event in early postnatal life. Therefore, we speculate that MIA offspring may not respond to weaning events in a timely and accurate manner, potentially acting as a “second hit.”

Autophagy is an evolutionarily conserved system responsible for the degradation of intracellular components to remove harmful molecules and ensure a sufficient nutrient supply. In mammals, autophagy is initiated immediately after birth to support the transition from umbilical cord to milk feeding [29]. Mice deficient in autophagy die within 12 h of birth, whereas normal mice can survive up to 24 h without lactation [29]. It is well established that hunger, malnutrition and psychological stress are the main environmental factors contributing to mental disorders, such as ASD and SCZ [30]. The lactation period is a critical phase for mammals, characterized by the dual challenges of physical hunger and psychological stress. We delayed weaning of MIA offspring and found no significant behavioral abnormalities in MIA offspring, suggesting the importance of weaning events in MIA model (Figure S1). Thus, we speculate that autophagy plays a vital role in managing hunger and stress-induced harmful molecules during this period, as depicted in Fig. 10. However, MIA offspring may fail to effectively initiate autophagy during the weaning period (Fig. 10), which could negatively impact neurodevelopment.

Fig. 10

A proposed model illustrating the role of autophagy during weaning. At weaning, rats experience the dual effects of physical hunger and psychological stress, which induce autophagy to manage hunger and remove stress-induced harmful molecules. Autophagy subsequently promotes complement-dependent synaptic pruning. However, MIA offspring are unable to effectively initiate autophagy during the weaning period, which impairs complement-dependent synaptic pruning and leads to behavioral deficits in MIA rat offspring

During development, excessive synapses need to be eliminated in an activity-dependent manner to refine synaptic circuits [31]. Impaired synaptic pruning is implicated in the pathogenesis of various neurodevelopmental disorders, including ASD and SCZ [32]. Autophagy has been identified as a crucial process in neuronal development, particularly through its regulation of spine elimination [33]. Tang et al. found that spine pruning was reduced in ASD brains and negatively correlated with the expression of LC3-II, suggesting that impaired autophagy may be associated with reduced synaptic pruning [24]. Studies in rodent models have demonstrated that autophagy deficits in neurons or microglia can lead to increased spine density due to impaired synaptic pruning [24, 25, 33]. Here, we reported for the first time that autophagy defects are a key mechanism underlying abnormal spine pruning and behavioral deficits in MIA offspring.

Recent studies implicate the classical complement cascade in neurodevelopment [34, 35]. The complement system participates in synaptic elimination by labeling inactive synapses, which are then recognized and “eaten” by microglia [34, 35]. During this process of synaptic pruning, the complement components C1q and C3 were found to localize at synapse, which are essential for microglia engulfment via the CR3 receptor [36, 37]. Although the role of complements in synaptic pruning has been well studied, the mechanisms that trigger its activation remain largely unclear. Our findings indicate that the expression levels of complement components are affected by modulating autophagy, suggesting that autophagy may be a potential regulator of complement activation. However, previous studies have suggested that complement components function as a warning system, with activation of the complement system potentially triggering autophagic responses to remove unwanted material [38, 39]. Therefore, further studies are needed to clarify the regulatory role of autophagy in complement activation.

In this study, we aimed to elucidate the relationship between autophagy and SCZ using animal models and to explore potential mechanisms from the perspective of synaptic pruning. However, several inconsistencies were identified in our findings. (1) In animal models, PPI deficits are a key biological marker of SCZ. However, our rescue experiments in the MIA model, as well as autophagy inhibition experiments in normal rats, did not confirm a direct link between autophagy defects and PPI impairment. Additional behavioral tests related to SCZ, such as amphetamine challenge in the MIA model or studies using other SCZ-related animal models, are needed to further investigate the relationship between autophagy defects and SCZ. (2) Although some studies have shown conflicting results [40, 41], it is generally believed that SCZ patients exhibit reduced dendritic spine density [42]. Our previous research also demonstrated reduced dendritic spine density in MIA offspring at postnatal days 45 and 65 [43]. Therefore, the hypothesis that autophagy defects impair synaptic pruning, leading to increased dendritic spine density, cannot be directly applied to the MIA model. The absence of an analysis of how autophagy defects affect synaptic pruning in the MIA rat model is a limitation of our study. We do not rule out the possibility that increased dendritic spine density due to autophagy defects may occur at specific stages of SCZ (e.g., in the early stages examined in this study) or in specific patient subgroups. It is also important to consider that autophagy defects may contribute to the pathology of MIA or SCZ through other mechanisms. Further systematic studies are needed to clarify the relationship between autophagy and SCZ and to understand the potential mechanisms by which autophagy is involved in the disease.

Conclusions

Our findings reveal an association between autophagy defects and SCZ and demonstrate that autophagy impairment during weaning disrupts complement-dependent synaptic pruning, contributing to behavioral deficits. From a clinical perspective, autophagy could represent a potential target for early pharmacological intervention in neuropsychiatric disorders.