All animal surgical procedures and usage protocols in this study were strictly conducted in accordance with the Guidelines of the Chinese Animal Welfare and Ethics Committee (GB/T3589-2018), and the experiment was approved by the Biomedical Ethics Committee of Peking University (approval number: LA2022325). The experimental animals were purchased from Beijing Vital River Laboratory Animal Technology Co., Ltd. SPF-grade female C57BL/6J mice (6–8 weeks old, weighing 20 ± 2 g) were randomly divided into 2 groups: the model group and control group. The mice were housed under controlled laboratory conditions (room temperature: 20 ± 1°C; light/dark cycle: 12 h) with free access to food and water. All behavioral tests were performed during the light phase.

To further ensure compliance with animal welfare and ethical guidelines, predefined humane endpoints were implemented throughout the experiment: these criteria included body weight loss of 20% or more relative to the baseline body weight, inability to independently access food or water for more than 24 consecutive hours, severe motor dysfunction such as persistent paralysis or convulsive seizures, and abnormal respiratory patterns like rapid shallow breathing or pre-asphyxial signs; the body weight, behavioral status, and physiological performance of the mice were monitored daily, and any mouse meeting any of these criteria was immediately euthanized via an overdose of sodium pentobarbital (240 mg/kg) followed by cervical dislocation to minimize animal suffering.

Mice were randomly divided into two groups to explore the pathological changes of MK-801-induced cognitive dysfunction in SZ mice, and the scheme diagram is shown in Figure 1. The mice in the model group received continuous intraperitoneal injection (i.p.) of MK-801 (Sigma-Aldrich Corporation) for two weeks, with a dose of 2 mg/kg/day; while the mice in the control group received continuous i.p. of saline for two weeks, with a dose of 2 mg/kg/day [18].

One day after the completion of the open-field test, a novel object recognition experiment was conducted (provided by the Beijing Key Laboratory of Magnetic Resonance Imaging and Technology). The experimental box for the mice had a size of 50 cm × 50 cm × 40 cm. During the familiarization phase, two identical wooden objects were placed in the center of the experimental box, equidistant from the walls. The mice were allowed to freely explore the box for ten minutes. In the testing phase on the following day, one of the objects in the experimental box was replaced with a different shape, and the mice were allowed to freely explore for ten minutes. The mice’s exploration behavior towards the novel and familiar objects was recorded using a camera installed on the top of the experimental box, and the recognition index (RI) was analyzed using computer-based animal behavior analysis software (provided by Beijing Zhong Shi Di Chuang Technology Development Co., Ltd.). The formula for calculating the RI was RI = TN / (TN + TF), where TN represents the time spent exploring the novel object, and TF represents the time spent exploring the familiar object. After each trial, the experimental box was cleaned with a 45% ethanol solution [19].

After the novel object recognition evaluation, a PPI experiment was carried out to assess the sensorimotor gating (SG) of the mice, whose functional damage is manifested in the attention deficit and thought disorders of schizophrenic patients. The Xeye Startle system (provided by Beijing Zhong Shi Di Chuang Technology Development Co., Ltd.) was used in the experiment. One day before the test, the mice were placed in the behavioral testing laboratory to adapt to the environment. Prior to the experiment, the background noise of the experimental box was set to 70 dB, and the startling pulse was set to 120 dB. The pre-pulse intensity was set to 3 dB above the background noise level (PPI3), 6 dB above the background noise level (PPI6), and 12 dB above the background noise level (PPI12). The single experiment was divided into four stages (Blocks 1–4), each appearing sequentially. Blocks 1 and 4 consisted of 6 individual startling pulses. Blocks 2 and 3 consisted of 6 separate startling pulses, 5 PPI3+startling pulse stimulus patterns, 5 PPI6+startling pulse stimulus patterns, 5 PPI12+startling pulse stimulus patterns, and 5 white noise stimulus patterns with no startling pulse stimulus. The pre-pulse duration was 20 ms, and the startling pulse duration was 40 ms, with a time interval of 100 ms (from start to start) between the pre-pulse and the startling pulse. Each stimulus mode appeared randomly, and the time interval between each mode was randomly set between 5 and 23 s. At the beginning of the experiment, the mice were fixed in a transparent glass box and allowed to adapt to the background noise for 5 minutes before sequentially undergoing the four experimental stages. The recorded peak amplitude (mv) was used to determine the amplitude of the startle response, where the average response amplitude in trials with only startling pulses (S) and in trials with both pre-pulse and startling pulses (PPI_S) was calculated. The formula for calculating the PPI rate (i.e., PPI%) of the startle response amplitude was PPI% = 100 × (S – PPI_S) / S [20].

Mice were anesthetized with 3% isoflurane for induction and 1–2% isoflurane for maintenance via inhalation during surgical procedures, with continuous monitoring of respiratory rate and reflexes. The scalp was incised with a razor blade to expose the skull, and the bregma was marked. The head was horizontally fixed to a stereotaxic instrument, with the bregma as the origin. According to the stereotaxic atlas of the mouse brain, the location of the caudate nucleus was identified and marked (0.7 mm anterior, 1.7 mm right, and 3 mm deep). Using a skull drill, the corresponding position on the skull was drilled, and a fine glass capillary needle was used to draw 10 mmol/L Lucifer Yellow solution and fixed to a microinjection pump. The infusion volume was set at 1 μL, with an injection speed of 0.2 μL/min. The solution was then injected into the corresponding brain region using the microinjection pump. After the completion of the injection, the needle was kept in place for 10 min before slowly withdrawing to avoid liquid reflux. After diffusion for 1 hour, brain perfusion was performed, and the mouse brain was removed and fixed overnight in 4% paraformaldehyde at 4°C. The next day, the mouse brain was removed and sectioned into slanting sagittal sections: the brain was placed in a slicing mold, and a slanted sagittal position was made at a 15° angle along the direction connecting the thalamus and caudate nucleus, with a thickness of 1 mm. The observation surface was placed on a confocal dish for imaging using a confocal laser scanning microscope (CLSM) and was primarily observed with a stereoscope. For fluorescence imaging, the imaging settings for CLSM were as follows: Lucifer Yellow excitation light at 488 nm and emission light at 510–550 nm; myelin sheath reflection excitation light at 633 nm and emission light at 630–640 nm. The panorama function of the Leica DIVE confocal microscope was used to store m × n images during the scanning process, and the LSCM Merging function was used to reconstruct the fluorescence images to obtain a complete image of the specified region of the sample [21].

After the behavioral experiments, mice were decapitated, and their brains were placed in PBS buffer. Under a small animal surgical microscope, the caudate nucleus and thalamus tissues were rapidly dissected using surgical blades and forceps and stored in liquid nitrogen for further experiments. High-resolution mass spectrometry data were collected using the Q Exactive mass spectrometer (Thermo Fisher Scientific, USA) in both positive and negative ion modes to improve metabolite coverage. Data processing was performed using Compound Discoverer 3.0 software (Thermo Fisher Scientific, USA), and metabolite identification was based on the mzCloud (Thermo Fisher Scientific, USA) and BGI library databases (BGI Genomics, Shenzhen, China). Differential metabolites between the model group, treatment group, and control group were screened using the R software package metaX. Preprocessing steps included: 1) Normalization of data to obtain relative peak areas by Probabilistic Quotient Normalization (PQN); 2) Correction of batch effects using quality control (QC)-based robust LOESS signal correction; 3) Removal of metabolites with a coefficient of variation greater than 30% in QC samples. A principal component analysis (PCA) plot was generated to reflect the distribution of QC samples, assessing instrument stability and data reproducibility. For discriminant analysis, partial least squares-discriminant analysis (PLS-DA) was employed to screen for differences between groups, utilizing PLS regression to model the relationship between metabolite expression and sample categories. Metabolic pathway enrichment analysis of differential metabolites based on the KEGG database was analyzed as well [22].

LFB myelin staining was used to investigate the demyelination in the mouse brain. After behavioral experiments, the mice were perfused, and the brains were fixed in 4% paraformaldehyde at 4°C for 24 hours. Subsequently, the brains were dehydrated in gradient sucrose solutions with concentrations of 15%, 20%, and 30% overnight, embedded in freezing embedding medium, and sectioned using a cryostat microtome. The oblique sagittal sections were obtained by rotating 15° outward from the midline of the cerebellum. The sections had a thickness of 20 μm and were mounted on slides, then incubated at 37°C for 12 hours. After sectioning, the brain slices were stored at –20°C.

Before staining, the embedding medium surrounding the tissue was washed away with distilled water. The tissue was hydrated in 95% ethanol, rinsed with distilled water, and stained with 0.1% LFB staining solution at room temperature. Excess stain was removed by rinsing the slides in 95% ethanol, followed by a rinse with distilled water. The differentiation solution, composed of 0.05% lithium carbonate and 70% ethanol, was used for differentiation until the gray and white matter were clearly distinguished. After rinsing with distilled water, the slides were mounted. Images were acquired by optical microscopy and analyzed with CAM-MS software (V3.1, CNC Software Inc.). Myelin sheaths appeared blue, while other structures remained mostly unstained. The internal capsule area was analyzed by a blinded observer, and the mean optical density (OD) value of the internal capsule was analyzed as well [23].

The mice were decapitated, and the brains were placed in PBS buffer. Under a surgical microscope, the internal capsule region connecting the caudate nucleus and thalamus was quickly dissected using a scalpel and forceps. Tissue samples of approximately 1 mm³ in size were collected and placed in a fixative solution specifically for electron microscopy, and tissues were fixed in 2.5% glutaraldehyde in 0.1 M phosphate buffer (pH 7.4) for 24 h at 4°C. On the following day, the tissue was fixed with 1% osmium tetroxide at room temperature for 2 hours in a darkroom. The tissue was dehydrated using a series of graded alcohols: 30%, 50%, 70%, 80%, 90%, and two rounds of 100%, with each step lasting 20 min. Then, the tissue was dehydrated twice with 100% acetone for 15 min each time. The tissue was embedded in epoxy resin to prepare resin blocks. Ultrathin sections (80 nm thick) were cut using an ultramicrotome and collected on 150-mesh copper grids with a carbon support film. The copper grids with tissue sections were stained with 2% uranyl acetate saturated in ethanol for 8 min in a darkroom. After staining, the grids were washed three times with 70% ethanol and then rinsed three times with ultrapure water. The sections were further stained with 2.6% lead citrate solution for 8 min, followed by three washes with ultrapure water. Excess water was removed using filter paper, and the copper grids with tissue sections were placed in a copper grid box to air dry overnight. Then, ultrathin sections were collected on copper grids, and the ultrastructural alterations of the myelin sheath were detected by transmission electron microscopy (TEM, HT7700, Hitachi). The thickness of the myelin sheath and myelinated axon was measured. Besides, the inner and outer diameters of myelinated axons quantified by ImageJ were used as the parameters to calculate area and number of myelinated axons (Quantity ratio), G-ratio, and myelin sheath thickness [24].

The experimental data were statistically analyzed using R software. GraphPad Prism version 9.5 for Windows was used to create statistical charts. The experimental results were presented as mean ± standard deviation. Data normality was assessed using Shapiro-Wilk tests, and for data that followed a normal distribution, a t-test was used to analyze the differences among the groups. For non-normally distributed data, the Mann-Whitney U test was used to compare the differences among the groups. A significance level of P < 0.05 was considered statistically significant for intergroup differences.

To evaluate the cognitive function and SG ability of the MK-801-induced SZ model mice, novel object recognition test and PPI test were performed, and the related behavioral indicators were statistically analyzed. The novel object recognition test is a classic method to assess the recognition memory function of rodents, and the movement trajectory, cognitive index, and exploration times of novel objects can comprehensively reflect the cognitive status of mice.

As shown in Figure 2A, the movement trajectory of SZ model mice in the novel object recognition test was relatively scattered, with no obvious preference for the novel object area; while the Sham group mice showed a concentrated trajectory, and spent more time exploring the novel object area. Quantitative analysis showed that the cognitive index of the SZ model mice was significantly lower than that of the Sham group (P < 0.01, Figure 2B), indicating that the SZ model mice had obvious recognition memory impairment, which was consistent with the core cognitive dysfunction phenotype of SZ. In terms of the number of explorations of novel objects, the SZ model group showed a significant decrease compared with the control group (P < 0.01, Figure 2C). The reduction in the number of explorations of novel objects reflects the decrease in the exploration willingness of SZ mice to novel stimuli, which is closely related to the negative symptoms, such as apathy in SZ.

The PPI test is an important method to evaluate the SG function of the brain, and the impairment of PPI is one of the typical neurobehavioral characteristics of SZ. To exclude the interference of basic auditory function on the test results, the auditory response of mice without stimulation was first detected. The results showed that there was no significant difference in the auditory response intensity between the SZ model group and the Sham group under the condition of no stimulation (P > 0.05, Figure 2D), indicating that the basic auditory function of SZ model mice was not affected, and the subsequent PPI index changes were not caused by auditory dysfunction. Similarly, when stimulated by startle stimuli, the response intensity did not reveal a significant difference (P > 0.05, Figure 2E). The PPI rate of startle reflex, the core index of the PPI test, was significantly reduced in the SZ model group compared with the Sham group (all P < 0.01, Figure 2F). The decrease in PPI rate indicates that the SG function of SZ model mice is damaged, and the brain cannot effectively filter irrelevant sensory information, which is consistent with the information processing disorder of SZ patients. In summary, the SZ model mice induced by MK-801 showed significant cognitive dysfunction and SG impairment, which fully reproduced the core behavioral phenotypes of SZ, providing a reliable behavioral basis for subsequent studies on the pathological mechanism of SZ.

Based on the compartmentalized ISF drainage pathway, we further detected the ISF reflux level in SZ model mice to clarify its distribution characteristics in multiple brain regions and correlation with behavioral phenotypes. The ISF drainage function is closely related to the clearance of neurometabolic wastes, and it is expected to be a key link connecting the pathophysiological mechanism.

To investigate changes in regional ISF drainage between brain regions, a stereotactic injector was used to inject the fluorescent tracer Lucifer Yellow into the caudate nucleus to explore ISF drainage changes. Figure 3A shows the comparison of the fluorescence diffusion area towards the posterior thalamus between the model group and the control group after 1 hour of tracer diffusion. Figure 3B shows the tracer diffusion situation captured by the camera under oblique sagittal sections. In the model group, the tracer crossed the barrier structure and refluxed into the posterior thalamus area, whereas in the control group, the tracer from the caudate nucleus did not reflux into the posterior thalamus area. The model group exhibited significantly larger diffusion area in the thalamus region (P < 0.01, Figures 3C–D), compared to the control group. This suggests that the model group of mice experienced disruption in ISF drainage between the caudate nucleus and the thalamus. In conclusion, the SZ model mice induced by MK-801 have significant ISF reflux abnormality in multiple core brain regions, and this abnormality is closely correlated with the core behavioral phenotypes of SZ.

Small-molecule neurometabolic substances reside in ISF, closely linked to cellular metabolism, secretory processes, and extracellular information transfer. Therefore, ISF drainage maintains the bridge of the metabolic connectome [25]. Consequently, we first investigated the effect of ISF decompartmentalization on region-specific neuro-metabolism within the thalamus and caudate nucleus ISF, under the scenario of spontaneous ISF reflux in SZ mice. To further elucidate the impact of MK-801-induced ISF drainage disruption on neuro-metabolic substance distribution, we analyzed metabolic profiles via global analysis (Figure 4A). Metabolite classification showed lipids and compounds with biological roles, such as amino acids, alcohols, imidazole, phenols and derivatives, pyridine and derivatives, engaging in amino acid metabolism, lipid, nucleotide, and carbohydrate metabolism.

PLS-DA (Figure 4B) demonstrates that the sham group shows a clearer separation trend, underscoring significant differences in metabolic profiles. The volcano plots (Figure 4C, Table S1) identify numerous differentially expressed metabolites, highlighting the altered metabolic landscape in MK-801-treated mice. The heatmap (Figure 4D) displays distinct metabolite expression patterns, further confirming that MK-801-induced ISF drainage disorder disrupts the distribution of neuro-metabolic substances. Collectively, these results show that MK-801-mediated ISF drainage dysfunction leads to profound disturbances in neuro-metabolic substance distribution, impacting multiple metabolic profiles.

To clarify the metabolic regulatory networks underlying ISF reflux-induced brain region dysfunction, we performed KEGG pathway enrichment analysis on differential metabolites between MK-801 and Sham groups in the caudate nucleus and thalamus (Figure 4E), aiming to identify pathway-level metabolic abnormalities associated with SZ. For the caudate nucleus (left panel of Figure 4E), multiple pathways were significantly enriched (P < 0.05) among differential metabolites: Pathways included tryptophan metabolism, protein digestion and absorption, and biosynthesis of amino acids. For the thalamus (right panel of Figure 4E), the enriched pathways differed from those in the caudate nucleus, reflecting region-specific metabolic dysregulation, and the most significantly enriched pathways (P < 0.01) included tyrosine metabolism, pyrimidine metabolism, and biosynthesis of amino acids. In summary, the caudate nucleus and thalamus of MK-801-induced SZ mice exhibit distinct, region-specific metabolic pathway dysregulation, and this regional specificity of metabolic dysfunction may give rise to the functional characteristics of each brain region, further linking ISF reflux-induced metabolic disorders to the core behavioral phenotypes of SZ.

The internal capsule—a bundle of dense myelinated fibers between the thalamus and caudate nucleus—serves as a critical structural barrier that restricts directional ISF drainage between these two brain regions. To assess the integrity of this barrier in SZ mice, we performed LFB myelin staining targeting the internal capsule. Morphological observations revealed overt disruption of myelin integrity in the internal capsule area of MK-801-treated mice, accompanied by reduced myelin sheath density (Figure 5A). Quantitative analyses further confirmed these structural abnormalities: the proportion of myelin structure area in the MK-801 group was significantly lower than that in the Sham group (P < 0.01; Figure 5B). Correspondingly, the mean OD in the MK-801 group was also markedly decreased compared to the Sham group (P < 0.01; Figure 5C). These data collectively indicate severe myelin loss in the internal capsule of SZ mice, which compromises the structural basis of the ISF drainage barrier.

To characterize myelin pathology at the ultrastructural level, we used TEM to examine the internal capsule. Low-magnification TEM images showed that the density and structural integrity of myelin bundles in the internal capsule were substantially disrupted in the MK-801 group (Figure 5D). Quantitative TEM analyses demonstrated that both the quantity ratio (P <0.01; Figure 5E) and area ratio (P < 0.01; Figure 5F) of myelinated fiber bundles in the model group were significantly reduced relative to the control group. High-magnification TEM further revealed prominent demyelination in the dense myelin sheath structure of the internal capsule in MK-801-treated mice (Figure 5G). Consistent with structural damage, the myelin sheath thickness in the model group was significantly increased (P <0.01; Figure 5H), while the G-ratio (the ratio of axon diameter to total fiber diameter, an indicator of myelin compaction) was significantly decreased (P <0.01; Figure 5I) compared to the Sham group. These ultrastructural findings align with LFB staining results, confirming profound myelin dysfunction in the internal capsule.

Taken together, the structural and quantitative impairment of myelin in the internal capsule disrupts the ISF drainage barrier between the caudate nucleus and thalamus—a key pathological event linking myelination defects to aberrant ISF distribution in SZ.