Introduction

Schizophrenia is a chronic and severe mental disorder that typically manifests in late adolescence or early adulthood, with an estimated lifetime prevalence of approximately 1% in the general population [1]. The clinical features of schizophrenia are categorized into three primary domains: positive symptoms (e.g., delusions and hallucinations), negative symptoms (e.g., reduced motivation and social withdrawal), and cognitive impairments (e.g., deficits in executive functions and memory) [2].

Schizophrenia is widely regarded as a complex neurodevelopmental disorder arising from the interplay between genetic and environmental factors, although its precise etiology remains elusive [3]. Several hypotheses have been proposed to explain the underlying mechanisms, including the dopamine hypothesis [4], the neuroinflammation hypothesis [5], and the neurodevelopmental hypothesis [6]. Twin studies have demonstrated that genetic factors contribute to approximately 80% of schizophrenia cases [7]. Recent advances in large-scale case studies and high-throughput genomic analyses have identified numerous schizophrenia-associated genetic variations, including single nucleotide polymorphisms, copy number variations, single nucleotide variants, and small insertions or deletions [8]. These genetic findings have provided a deeper understanding of the molecular mechanisms underlying schizophrenia [6]. Notably, many of these genetic loci are located in genes involved in synaptic organization, development, and transmission, suggesting that synaptic dysfunction may play a central role in the pathogenesis of schizophrenia [9, 10]. However, current genetic discoveries account for only a fraction of schizophrenia cases [11, 12], highlighting the polygenic nature of the disorder [8] and the need to identify additional risk genes and genetic variations. Understanding the etiology and pathological basis of schizophrenia remains a critical challenge in the field.

Glucose-6-phosphate dehydrogenase (G6PD) is the rate-limiting enzyme in the pentose phosphate pathway (PPP), which oxidizes glucose-6-phosphate to generate reduced nicotinamide adenine dinucleotide phosphate (NADPH). NADPH is essential for reductive biosynthesis and maintaining cellular redox homeostasis [13]. G6PD is ubiquitously expressed, and genetic defects in the G6PD gene, which are X-linked, cause G6PD deficiency—a condition characterized by acute hemolytic anemia triggered by certain foods (e.g., fava beans), drugs, or infections [14]. As the primary enzyme catalyzing NADPH production, G6PD has significant antioxidant properties and plays a neuroprotective role in various conditions [15], including neurodegenerative diseases [16, 17], cerebral ischemia [18], and aging [19]. Recent evidence suggests that disruptions in the PPP and the resulting oxidative stress, characterized by increased reactive oxygen species due to reduced NADPH levels, may contribute to the pathogenesis of schizophrenia [20]. However, the precise mechanisms remain unclear. Since the 1960s, studies have attempted to link G6PD deficiency to schizophrenia [21,22,23,24,25,26,27,28,29,30,31,32]. For example, Dern et al. reported a higher prevalence of G6PD deficiency in patients with catatonic schizophrenia [32], and subsequent studies found reduced G6PD enzymatic activity in the white blood cells of schizophrenia patients [29]. Despite these findings, direct biological evidence demonstrating a causal relationship between G6PD deficiency and schizophrenia has been lacking.

In this study, we examined the behavioral phenotypes of mice with a conditional knockout (CKO) of G6PD in the cerebral cortex and hippocampus. We found that G6PD CKO mice exhibit schizophrenia-like behaviors, including hyperactivity, impaired social interaction, deficient prepulse inhibition (PPI), and cognitive deficits. Proteomic analysis identified numerous dysregulated synaptic proteins in these mice. Furthermore, we observed synaptic deficits, including reduced presynaptic vesicle density, impaired SNARE complex assembly, and decreased dendritic spine density. Collectively, our findings suggest that G6PD deficiency contributes to schizophrenia-like behaviors by modulating synaptic function in the brain.

Materials and methods

Animals

Exon 5 of the G6PD gene was selected as the target for conditional deletion, and floxed G6PD mice were generated using standard protocols by Biocytogen (Beijing, China). Female G6PDflox/+ mice were backcrossed with wild-type C57BL/6 mice (SLAC Laboratories, Shanghai, China) for at least four generations. Genotyping was performed using primers specific to the floxed and wild-type alleles (G6pd-B1LoxP-F: 5′-CAGTATGATGGAGATTGAAGCAAGAGG-3′ and G6pd-B2LoxP-R: 5′-CAAATTCAGTACGTATGTAGCCCAGG-3′). Amplified products of 249 and 197 bp indicated floxed and wild-type alleles, respectively.

To achieve brain region-specific deletion of G6PD, female G6PDflox/+ mice were crossed with male Emx1-Cre mice [33] (#005628; Jackson Laboratory, Bar Harbor, ME, USA), which express Cre recombinase in cortical progenitors from embryonic day 10.5 (E10.5). Male offspring with the genotype Emx1-Cre;G6PDflox/Y were designated as G6PD CKO mice. To eliminate hormonal effects, only male G6PD CKO mice and male controls (G6PDflox/Y or G6PD+/Y) with 2–4 months old of age were used in this study. Mice were housed under controlled temperature (22 ± 2 °C) and humidity (50 ± 5%) with a 12-h light/dark cycle and free access to food and water. No randomization was used for allocation of animals to experimental groups. No blinding was done for group allocation during the experiments or when assessing the outcome. All animal experiments were approved by the Animal Ethics Committee of Tongji University (TJAA01224103).

Western blotting

Cortical tissue was lysed in RIPA buffer (50 mM Tris·HCl pH 7.4, 150 mM NaCl, 1% Triton X-100, 1% sodium deoxycholate, 0.1% SDS, 5 mM EDTA) supplemented with a protease inhibitor cocktail (#78439, Thermo Fisher Scientific, Carlsbad, CA, USA). Protein concentrations were determined using a bicinchoninic acid (BCA) assay kit (#23227, Thermo Fisher Scientific). Equal amounts of protein were separated by SDS-polyacrylamide gel electrophoresis (SDS-PAGE) and transferred to nitrocellulose membranes. Membranes were incubated overnight at 4 °C with primary antibodies, followed by incubation with HRP-conjugated secondary antibodies (1:5000; Epizyme, Shanghai, China). The following primary antibodies were used: G6PD (1:1000; ab993, Abcam, Waltham, MA, USA), GAPDH (1:2000; LF206, Epizyme), Synaptophysin (1:500; ab14692, Abcam), PSD95 (1:2000; MA1-045, Thermo Fisher Scientific), β-actin (1:2000; LF202, Epizyme), Syntaxin1 (1:1000; sc-12736, Santa Cruz Biotechnology, Dallas, TX, USA), β-tubulin (1:2000; LF204, Epizyme), Erc1 (1:1000; 22211-1-AP, Proteintech, Rosemont, IL, USA), Erc2 (1:1000; 21396-1-AP, Proteintech) and Cplx1 (1:1000; 10246-2-AP, Proteintech). Protein bands were visualized using enhanced chemiluminescence and quantified using ImageJ software.

Measurements of enzymatic activity of G6PD and levels of GSH and GSSG

G6PD enzymatic activity, levels of GSH and GSSG in cortical tissues were measured using specific G6PD activity assay kit (S0189, Beyotime, Shanghai, China) and GSH/GSSG assay kit (S0053, Beyotime) according to the manufacturer’s instructions.

Histological analysis

Adult mice were perfused sequentially with PBS and 4% paraformaldehyde (PFA; Sigma, Saint Louis, MO, USA), and the brains were dissected and fixed in 4% PFA overnight at 4 °C. After cryoprotection in a 30% sucrose solution, brains were sectioned at 20 μm using a cryostat (CM1900, Leica, Deer Park, IL, USA).

For Nissl staining, brain sections were incubated in 0.5% cresyl violet solution (C5042, Sigma) for 5 min, followed by dehydration in ethanol, clearing in xylene, and mounting with neutral balsam (Sinopharm Chemical Reagent, Shanghai, China).

For in situ hybridization, probes for Cux2, RORβ, Sox5, Tle4, Gad1 and Sst were used as previously described [34,35,36]. Brain sections were digested with proteinase K, treated with triethanolamine/acetic anhydride, blocked in prehybridization buffer, and incubated with RNA probes at 65 °C overnight. After washing, the sections were incubated with an anti-Digoxigenin-AP antibody (Roche, Basel, Switzerland) overnight at 4 °C, and mRNA signals were visualized using NBT (nitro-blue tetrazolium chloride) and BCIP (5-bromo-4-chloro-3′-indolyphosphate p-toluidine salt).

For immunofluorescent staining, brain slices were incubated with rabbit anti-S100β antibody (1:200; ab52642, Abcam) at 4 °C overnight. After brief washes with PBS, brain slices were incubated with Alexa Fluor Plus 488-conjugated donkey anti-rabbit antibody (1:500; A32790, Thermo Fisher Scientific) at RT for 3 h before mounted onto slides.

Images were captured using a Nikon Eclipse 80i microscope (Tokyo, Japan).

Proteomic analysis

Cortical tissues from control and G6PD CKO mice were subjected to data-independent acquisition (DIA) proteomic analysis by Shanghai Majorbio Company (Shanghai, China). Proteins were extracted, quantified using the BCA method, and validated via SDS-PAGE. Samples were alkylated, digested with trypsin, and analyzed using an EASY-nLC 1200 liquid chromatography-tandem mass spectrometry (LC-MS/MS) system (Thermo Fisher Scientific, USA). The raw data were processed with Spectronaut software. Volcano plots, heatmaps, and GO/KEGG pathway analyses were generated using the Majorbio Cloud platform (cloud.majorbio.com).

Subcellular fractionation

Cortical subcellular fractions were prepared as previously described [37], with modifications. Briefly, cortical tissues were homogenized in HEPES-buffered sucrose (320 mM sucrose, 4 mM HEPES, 5 mM EDTA, pH 7.4) containing protease inhibitors. Homogenates were centrifuged at 1000 g for 10 min. Supernatants (S1) were centrifuged at 10,000 g for 15 min to obtain mitochondria- and synaptosome-enriched pellets (P2) and supernatants (S2). P2 pellets were resuspended in HEPES-buffered sucrose and loaded onto a discontinuous sucrose gradient (0.8–1.0–1.25 M sucrose). After centrifugation at 100,000 g for 30 min, synaptosomes (Syn) were collected from the 1.0/1.25 M sucrose interface. The Syn layer was solubilized with 1% Triton X-100 in 50 mM HEPES/NaOH (pH 8.0) and centrifuged at 20,000 g for 15 min to yield the presynaptic fraction (Pre, supernatant) and postsynaptic density fraction (PSD, pellet).

Electron microscopy

For ultrastructural analysis, cortical tissue samples (1–2 mm3) were fixed in 2.5% glutaraldehyde overnight at 4 °C. After PBS washes, samples were post-fixed in 1% OsO4 in PBS for 2 h at room temperature, dehydrated in graded ethanol, and embedded in resin. Ultrathin sections (80–100 nm) were cut using an ultramicrotome (EM UC7, Leica) and stained with 2% uranyl acetate and Reynold’s lead citrate. Images were captured at 10,000× magnification using a transmission electron microscope (Tecnai G2-20 S-Twin, FEI, Hillsboro, Oregon, USA). Vesicle counts, normalized to a 100 nm length of the PSD, and PSD lengths were quantified using ImageJ software in a blinded manner.

Assay for SNARE complex assembly

SNARE complex assembly was assessed by analyzing the high-molecular-weight SDS-resistant protein complex via Western blotting without prior boiling of the samples [38, 39]. Cortical tissues were homogenized in SNARE buffer (50 mM HEPES, 2 mM MgCl2, 250 mM sucrose) and mixed with SDS sample buffer. Samples were incubated at room temperature for 20 min or boiled for 5 min to confirm heat sensitivity. Proteins were separated by SDS-PAGE and immunoblotted with an anti-Syntaxin1 antibody (1:1000; sc-12736, Santa Cruz Biotechnology) and HRP-conjugated secondary antibodies (1:5000; Epizyme). The ratio of Syntaxin1 in the SNARE complex to total Syntaxin1 was quantified in non-boiled samples using ImageJ software.

Dendritic spine analysis

Golgi staining was performed using the FD Rapid GolgiStain Kit (PK401, FD NeuroTechnologies, Columbia, MD, USA) [40]. Coronal brain slices (150 μm thick) were prepared using a cryostat (CM1900, Leica) and imaged in Z-stacks using a Precipoint M8 microscope (Nikon). Images were merged using Photoshop software (Adobe Systems, San Jose, CA, USA). For spine density quantification, secondary or tertiary dendritic branches were analyzed in at least 12 neurons per mouse from layers V/VI of the prefrontal cortex. Spine density was calculated as the number of spines per 10 μm of dendrite.

Behavioral tests

Adult male mice (2–4 months old) were used for all behavioral tests. Mice were habituated to the test environment for at least 30 min before testing. Experimenters were blinded to the genotypes.

Open field test

Mice were placed in the lower-left corner of a transparent plexiglass box housed in a soundproof enclosure. They were allowed to explore freely for 30 min, and their movements were tracked. Total distance traveled was recorded using activity monitoring software (Med Associates, St. Albans, VT, USA).

Three-chamber social interaction test

The three-chamber test was conducted as previously described [40]. The apparatus consisted of a plexiglass box (90 × 50 × 30 cm) divided into three chambers with square openings (5 × 5 cm) in the partitions for free movement. Two small wire cages were placed in the lateral chambers. In the first phase, a stranger mouse (S1) was placed in one cage, and an inanimate ball was placed in the other. The test mouse was placed in the middle chamber and allowed to explore for 10 min. Interaction time with S1 or the ball was recorded. In the second phase, the ball was replaced with a new stranger mouse (S2). The test mouse was allowed to explore for another 10 min, and interaction time with S1 or S2 was recorded. Interaction was defined as sniffing within 2 cm of the cage. Preference (%) was calculated as the ratio of interaction time with S1, S2, or the ball to the total interaction time.

Prepulse inhibition (PPI) test

The PPI test was performed as previously described [41]. The test mouse was placed in a plexiglass box inside a soundproof chamber. After a 5-min acclimation period, twelve 100-dB startle pulses (40 ms) were presented to the mouse. In the subsequent 48 trials, the startle tone (100 dB, 40 ms) was presented either alone or following one of three prepulses (65, 73, or 82 dB, 20 ms) with a 100-ms delay, in a randomized order. The interval between trials ranged from 20 to 40 s. The percentage of prepulse inhibition (% PPI) was calculated using the formula: % PPI = 100−[(startle response for prepulse + pulse)/(startle response for pulse alone)] × 100.

T-Maze test

The T-maze test was conducted as described previously [40]. Mice were placed on a restricted feeding schedule prior to testing. Each trial consisted of a sample run and a choice run. In the sample run, the mouse was forced to enter either the left or right arm of the maze (determined by the open door), where a sweet food pellet was placed at the end. After consuming the pellet, the mouse was returned to the starting arm and allowed to choose freely between the two arms during the choice run. The interval between the sample and choice runs was approximately 1 or 3 min. Mice were rewarded with a food pellet for entering the previously unvisited arm. The test was conducted over three consecutive days, with eight trials per day and a 3-min interval between trials. The percentage of correct responses and the latency to find the food pellet were recorded.

Novel object recognition test

The novel object recognition test was performed in a square arena (25 × 25 × 25 cm) within a soundproof box, as previously described [42]. The test consisted of three sessions: 1. Habituation Session: Mice explored the empty arena for 10 min per day over three days (Day 1–3). 2. Training Session: Two identical objects were placed symmetrically in the arena, and mice were allowed to explore for 10 min (Day 4). 3. Test Session: One hour after training, one familiar object was replaced with a novel object, and mice explored the arena for 10 min. The objects used were similar in size and smell but differed in shape and texture. Exploration time was counted when the mouse was sniffing within 2 cm of an object. The percentage of exploration time was calculated as the time spent exploring each object divided by the total exploration time.

Contextual fear conditioning test

The contextual fear conditioning test was conducted as described previously [41]. Mice were placed in a test chamber and allowed to explore freely for 2 min before receiving five foot shocks (1.0 mA, 2 s) with 2-min inter-shock intervals. Two minutes after the final shock, mice were returned to their home cages. Freezing behavior was measured as the time spent immobile during each inter-shock interval. Mice were then reintroduced to the same test chamber at various time points (30 min, 1 day, 4 days, 7 days, and 14 days post-conditioning) without receiving additional shocks, and freezing behavior was monitored for 11 min. Freezing behavior (%) was calculated as the time spent freezing divided by the total observation time.

Dark-light choice test

The dark-light choice test was conducted as described previously [42]. The apparatus consisted of two compartments: a smaller dark area (30 × 20 × 30 cm) with a lid and a larger bright area (30 × 30 × 30 cm) without a lid. Mice were placed in the center of the dark compartment and allowed to explore freely for 5 min. The time spent in the light compartment and the number of transitions between the dark and light compartments were recorded using video tracking software (Med Associates).

Elevated plus-maze test

The elevated plus-maze [43] consisted of two open arms (30 × 5 cm), two closed arms (30 × 5 × 16 cm), and a central platform (5 × 5 cm), elevated 50 cm above the ground. Mice were placed on the central platform facing one of the closed arms and allowed to explore freely for 5 min. The number of entries into the open arms and the time spent in the open arms were recorded.

Sucrose preference test

The sucrose preference test was conducted as previously described [42], with minor modifications. Mice were individually housed and acclimated to two water bottles for 3 days. Following a 24-h water deprivation period, mice were given free access to two bottles: one containing tap water and the other containing a 1% (w/v) sucrose solution. The amount of water and sucrose solution consumed over 12 or 24 h was measured. The sucrose preference (%) was calculated as the volume of sucrose solution consumed divided by the total liquid consumed (sucrose solution + water).

Morris water maze test

The Morris water maze test was performed as described previously [34]. A circular pool (1 m diameter) was divided into four equal quadrants, with a hidden platform located in the center of the target quadrant, submerged approximately 1 cm below the water surface. During the training phase, mice underwent four trials per day for seven consecutive days to locate the hidden platform. On the eighth day, the platform was removed, and mice were allowed to search the pool for 1 min. A tracking software (EthoVision XT 8.0, Noldus Technology, Gelderland, Netherlands) was used to monitor mouse movements. The latency to find the platform and the time spent in the target quadrant were recorded.

Statistical analyses

Each experiment was repeated at least three times. All data were tested for normal distribution and homogeneity of variance and all data met the assumptions of the tests. Data were analyzed using two-tailed Student’s t-test or one-way ANOVA followed by Dunnett’s post hoc test. A p-value of less than 0.05 was considered statistically significant. Results are presented as mean ± SEM. The sample size and p value for each graph was stated in figure legends. No statistical methods were used for sample size estimate.

Results

Generation of G6PD CKO mice

G6PD, the rate-limiting enzyme in the PPP, is constitutively expressed in all cells [44]. In particular, G6PD mRNA, protein, and enzymatic activity are present in neurons across various brain regions [45,46,47,48]. To investigate the role of G6PD in neuropsychiatric behaviors, we generated a conditional G6PD knockout mouse model in which the gene is specifically deleted in the brain. Using standard protocols, we generated a floxed mouse line with exon 5 of the G6PD gene flanked by two loxP sites (G6PDflox/+). To achieve brain-specific deletion, we crossed female G6PDflox/+ mice with male Emx1-Cre mice (#005628; Jackson Laboratory), which express Cre recombinase in cortical progenitors from embryonic day 10.5 (E10.5) [33]. The resulting male offspring (Emx1-Cre;G6PDflox/Y) were designated as G6PD CKO mice (Fig. 1A). To avoid hormonal influences during behavioral testing, only male G6PD CKO mice and male controls (G6PDflox/Y or G6PD+/Y) were used in this study.

Fig. 1: Generation and verification of G6PD CKO mice.
Fig. 1: Generation and verification of G6PD CKO mice.
Full size image

A Strategy for generating G6PD CKO mice: Floxed G6PD mice were crossed with Emx1-Cre mice to delete exon 5 in the cerebral cortex and hippocampus, resulting in G6PD CKO mice. B, C Western blot analysis shows a significant reduction in G6PD protein levels in cortical samples from G6PD CKO mice compared to controls. D G6PD enzymatic activity is significantly decreased in G6PD CKO mice compared to controls. E, F Levels of GSH and GSSG in cortical tissue are comparable between G6PD CKO and control mice. G The ratio of GSH to GSSG is significantly decreased in G6PD CKO mice compared to controls. H Nissl staining reveals no gross morphological changes in the brains of G6PD CKO mice compared to controls. Ctx, cortex; cc, corpus callosum; Hip, hippocampus. I–N In situ hybridization of layer marker genes (Cux2, RORβ, Sox5 and Tle4) and interneuron marker genes (Gad1 and Sst) in the cerebral cortex shows no apparent changes in expression between G6PD CKO and control mice. O Immunofluorescent staining of astrocytic marker gene, S100β, in the cerebral cortex shows no apparent changes in expression between G6PD CKO and control mice. Bar graphs are presented as mean ± SEM. *p < 0.05, ***p < 0.001 (two-tailed Student’s t-test). Scale bars: 500 μm.

The deletion of exon 5 in the G6PD genomic locus is predicted to produce a truncated protein with a frame-shifted C-terminus. To examine the changes of G6PD expression, we examined the mRNA and protein levels of G6PD in the cerebral cortex of adult control and CKO mice. Primers targeting exons 5 and 6 of G6PD revealed a 40–50% reduction in mRNA levels in cortical tissues of G6PD CKO mice compared to controls, as determined by qRT-PCR (data not shown). Western blot analysis further confirmed a substantial reduction in G6PD protein levels in cortical samples from CKO mice relative to controls, using a specific antibody targeting the N-terminal portion (Fig. 1B, C, S2A). Enzymatic activity assays using a commercial kit revealed a marked decline in G6PD activity in the cortex of CKO mice compared to controls (Fig. 1D). The ratio of reduced glutathione (GSH) to oxidized glutathione (GSSG) [GSH:GSSG] is considered as an indicator of cellular oxidative stress [49, 50]. A dramatic decrease of this ratio was observed in the cortical tissue of G6PD CKO mice compared to controls (Fig. 1G), although no changes were found in the levels of GSH or GSSG (Fig. 1E, F). These findings indicate that G6PD retains partial mRNA/protein expression and enzymatic activity in CKO mice, likely due to residual expression in endothelial cells and neurons or glial cells lacking Cre recombinase [33]. Collectively, our data confirm the successful generation of the G6PD CKO mouse model.

Unaltered overall cortical architecture in G6PD CKO mice

To determine whether G6PD deletion affects cortical morphology, we examined the cerebral cortex and hippocampus of G6PD CKO and control mice at postnatal day 60 (P60). Nissl staining revealed comparable general cytoarchitecture and white matter organization (corpus callosum) between the two groups (Fig. 1H). To assess potential alterations in cortical lamination, we performed in situ hybridization (ISH) for layer-specific markers, including Cux2 [51] (layers II-IV), RORβ [52] (layer IV), Sox5 [53] (layers V-VI), and Tle4 [54] (layer VI). No significant changes in marker expression were observed (Fig. 1I–L). Similarly, we used ISH to assess the expression of Gad1 and Sst, both of which are markers for cortical interneurons [55], and found no differences in the distribution of Gad1- or Sst-positive cells between G6PD CKO and control mice (Fig. 1M, N). Additionally, we checked the expression of S100β, an astrocytic marker in the cortex and found no changes between G6PD CKO and control mice (Fig. 1O). These findings indicate that the overall cortical morphology, including lamination, is preserved in G6PD CKO mice.

G6PD CKO mice exhibit schizophrenia-like behaviors

Increased locomotor activity, a common feature of rodent schizophrenia models, is thought to reflect the positive symptoms of schizophrenia [56]. While previous studies reported normal motor function in G6PD-deficient mice [57], the presence of hyperactivity had not been investigated. In the open field test, G6PD CKO mice exhibited significantly increased locomotor activity, as evidenced by a greater total distance traveled compared to controls (Fig. 2A, B). Ambulatory time was comparable between genotypes (data not shown), indicating that G6PD is necessary for normal locomotor activity. To assess negative symptoms, we performed a three-chamber social interaction test [58]. Both control and G6PD CKO mice spent more time with a stranger mouse (S1) than with a ball (Fig. 2C, D). However, when a second stranger mouse (S2) was introduced, control mice spent more time with S2, whereas G6PD CKO mice showed no preference between S1 and S2 (Fig. 2C, E), indicating impaired social interaction. Prepulse inhibition (PPI), a measure of sensorimotor gating, is often deficient in schizophrenia patients [59]. G6PD CKO mice exhibited significantly reduced PPI at the 82 dB prepulse level compared to controls (Fig. 2F, G), suggesting impaired sensorimotor gating.

Fig. 2: G6PD-deficient mice exhibit schizophrenia-like behaviors.
Fig. 2: G6PD-deficient mice exhibit schizophrenia-like behaviors.
Full size image

A Diagram of the open field test. B G6PD CKO mice show increased total distance traveled in the open field test compared to controls. C Diagram of the three-chamber social interaction test. In Phase 1, the test mouse is placed between a ball and a stranger mouse (S1). In Phase 2, the test mouse is placed between the initial stranger mouse (S1) and a second stranger mouse (S2). D G6PD CKO mice spend more time interacting with S1 than the ball, similar to controls. E G6PD CKO mice spend comparable time with S1 and S2, whereas control mice show a preference for S2 over S1. F Diagram of the prepulse inhibition (PPI) test. G PPI is reduced at the 82 dB prepulse level in G6PD CKO mice compared to controls. H Diagram of the T-maze test. I G6PD CKO mice show reduced correct arm entries compared to controls. J Latency to find food is increased in G6PD CKO mice compared to controls. K Diagram of the novel object recognition test. In the training phase, the test mouse is placed between two identical objects (A and B). In the test phase, one familiar object (B) is replaced with a novel object (red B). L G6PD CKO mice show comparable exploration time for objects A and B during the training phase. M G6PD CKO mice spend less time exploring the novel object (B) compared to the familiar object (A), whereas control mice show the opposite preference in the test phase. N Diagram of the contextual fear conditioning test. O Freezing time during the post-shock period is comparable between G6PD CKO and control mice. P Freezing time 30 min after conditioning is reduced in G6PD CKO mice compared to controls. Bar graphs are plotted as mean ± SEM. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001, n.s. = not significant (two-tailed Student’s t-test or one-way ANOVA with Dunnett’s post hoc test).

Cognitive impairments, a hallmark of schizophrenia [60], were assessed using the T-maze and novel object recognition tests [61]. G6PD CKO mice displayed reduced correct arm entries and increased latency in the T-maze at a 3-min interval (Fig. 2H–J), but not at a 1-min interval (data not shown), indicating deficits in spatial working memory. In the novel object recognition test [62], control mice explored a novel object more than a familiar one, whereas G6PD CKO mice failed to discriminate between the two (Fig. 2K–M), suggesting impaired visual episodic memory. In contextual fear conditioning [63], G6PD CKO mice exhibited reduced freezing 30 min after conditioning but showed comparable freezing levels to controls on subsequent days (Fig. 2N–P), indicating short-term memory deficits with intact long-term memory. In summary, G6PD CKO mice exhibit schizophrenia-like behaviors, including hyperactivity, social interaction deficits, impaired PPI, and cognitive impairments, underscoring the importance of G6PD in regulating these behaviors.

G6PD CKO mice display no anxiety- or depression-like behaviors and normal spatial learning and memory in the morris water maze

To examine anxiety-like and depression-like behaviors, as well as spatial learning and memory, we conducted a series of behavioral tests. In the dark-light choice test, G6PD CKO mice spent a comparable amount of time in the light box and crossed between the dark and light boxes as frequently as control mice (Fig. 3A–C). Similarly, in the elevated plus-maze test, G6PD CKO mice spent a similar amount of time in the open arms and entered the open arms at rates comparable to controls (Fig. 3D–F). These findings suggest that anxiety-like behaviors are unaltered in G6PD CKO mice. To evaluate depression-like behaviors, we performed the sucrose preference test, a standard paradigm for assessing anhedonia in rodents [64]. G6PD CKO mice showed no differences in their preference for sucrose solution over water during both 12 and 24-h testing periods compared to controls (Fig. 3G), indicating no changes in depression-like behaviors. We also assessed spatial learning and memory using the Morris water maze. During the learning phase, G6PD CKO mice exhibited comparable latency in finding the hidden platform relative to controls (Fig. 3H, I). On the test day, G6PD CKO mice showed similar latency in locating the platform’s previous position after its removal (Fig. 3J) and spent an equivalent amount of time in the target quadrant compared to controls (Fig. 3K). These results indicate that spatial learning and memory are intact in G6PD CKO mice. Together, these findings suggest that G6PD is not essential for anxiety-like or depression-like behaviors, nor for spatial learning and memory as assessed in the Morris water maze.

Fig. 3: G6PD CKO mice display no anxiety- or depression-like behaviors and normal spatial learning and memory in the Morris water maze.
Fig. 3: G6PD CKO mice display no anxiety- or depression-like behaviors and normal spatial learning and memory in the Morris water maze.
Full size image

A Diagram of the dark-light choice test. B Time spent in the light box is comparable between G6PD CKO and control mice. C Number of transitions between dark and light compartments is comparable between G6PD CKO and control mice. D Diagram of the elevated plus-maze test. E Time spent in the open arms is comparable between G6PD CKO and control mice. F Number of entries into the open arms is comparable between G6PD CKO and control mice. G Sucrose preference is comparable between G6PD CKO and control mice during both 12 and 24-h observation periods. H Diagram of the Morris water maze test. The platform is submerged and hidden from the test mouse. I Latency to find the platform during the learning phase is comparable between G6PD CKO and control mice. J Latency to locate the removed platform’s previous position on the test day is comparable between G6PD CKO and control mice. K Time spent in the target quadrant is comparable between G6PD CKO and control mice. Bar graphs are plotted as mean ± SEM. n.s. = not significant (two-tailed Student’s t-test or one-way ANOVA with Dunnett’s post hoc test).

Dysregulation of synaptic proteins in G6PD CKO mice

To investigate the impact of G6PD deletion on brain function, we performed mass spectrometry-based proteomic analysis on cortical lysates from control and G6PD CKO mice (Fig. 4A). A total of 5745 proteins were identified in each sample (Supplemental Table 1), of which 272 were downregulated and 210 were upregulated in G6PD CKO mice compared to controls (Fig. 4B, p < 0.05). Consistent with our Western blot data (Fig. 1B), G6PD protein levels were significantly reduced in CKO brain lysates (Fig. 4B). Gene Ontology (GO) analysis of downregulated proteins revealed significant enrichment in mitochondrial functions, ATP synthesis, biosynthetic processes, and metabolic and catabolic pathways (Fig. 4C; Supplemental Table 2). Similarly, Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis identified significant enrichment in metabolic pathways, oxidative phosphorylation, carbon metabolism, and thermogenesis (Supplemental Table 3, p < 0.05). In contrast, GO analysis of upregulated proteins showed enrichment in synaptic vesicle fusion, synaptic membrane organization, synaptic transmission, and synaptic vesicle priming (Fig. 4D; Supplemental Table 4, p < 0.05). KEGG analysis also revealed significant pathway enrichment for upregulated proteins in glutamatergic synapse and SNARE interactions in vesicular transport (Supplemental Table 5). Given that genome-wide association studies have implicated synaptic structure and function in schizophrenia [9, 10], we aggregated all dysregulated proteins with GO annotations related to synaptic organization, differentiation, and transmission (Fig. 4E; Supplemental Table 1). The expression of Erc1, Erc2, and Cplx1 out of these synaptic proteins was evaluated using Western blots in G6PD CKO mice compared to controls and their expression changes were consistent with the proteomic analysis findings (Figs. S1, S2D). These results indicate that synaptic protein expression is significantly disrupted in G6PD CKO mice.

Fig. 4: Synaptic proteins are dysregulated in G6PD CKO mice.
Fig. 4: Synaptic proteins are dysregulated in G6PD CKO mice.
Full size image

A Schematic of quantitative proteomic analysis of cortical lysates from G6PD CKO and control mice. A total of 5745 proteins were identified (n = 3 per group). B Volcano plot of differentially expressed proteins. Blue and red dots represent significantly downregulated and upregulated proteins, respectively (p < 0.05). G6PD is among the downregulated proteins. C Top 20 significantly enriched GO terms for downregulated proteins. The rich factor represents the ratio of differentially expressed proteins annotated in a GO term to the total number of proteins annotated. D Top 20 significantly enriched GO terms for upregulated proteins. Eight of these terms are associated with synaptic function (highlighted in red). E Heatmap of dysregulated proteins associated with synaptic function in G6PD CKO and control mice.

G6PD is localized in presynaptic terminals and regulates synaptic function and development

The proteomic analysis revealed dysregulation of synaptic proteins, potentially contributing to the synaptic deficits and behavioral impairments observed in G6PD CKO mice. To determine the subcellular localization of G6PD, we examined cortical neurons using discontinuous sucrose density centrifugation. G6PD was enriched in the presynaptic fraction, marked by Synaptophysin [65], but was absent from the postsynaptic density (PSD) fraction, labeled by PSD95 [66] (Figs. 5A, S2B). While G6PD was previously detected in axons [67], our findings demonstrate its specific localization in presynaptic terminals. Given this localization, we hypothesized that G6PD deficiency might affect synaptic vesicle distribution, which is critical for neurotransmitter release and synaptic transmission. Using electron microscopy, we quantified synaptic vesicles normalized to synapse size (length of PSD). G6PD CKO mice exhibited significantly reduced synaptic vesicle density compared to controls (Fig. 5B–D). Additionally, we observed an increase in PSD length in G6PD CKO mice (Fig. 5B, C, E), suggesting postsynaptic abnormalities.

Fig. 5: G6PD is specifically enriched in presynaptic fractions, and its depletion leads to synaptic deficits.
Fig. 5: G6PD is specifically enriched in presynaptic fractions, and its depletion leads to synaptic deficits.
Full size image

A G6PD is enriched in the presynaptic fraction of cortical lysates. Synaptophysin and PSD95 serve as markers for presynaptic and postsynaptic fractions, respectively. B, C Representative electron micrographs of asymmetric synapses in the cortex of G6PD CKO and control mice. Arrowheads indicate presynaptic vesicles, and PSD areas are highlighted. Scale bar: 200 nm. D Presynaptic vesicle density, normalized to PSD length, is significantly decreased in G6PD CKO mice compared to controls. E PSD length is significantly increased in G6PD CKO mice compared to controls. F Representative Western blots of Syntaxin1-containing SNARE complexes in cortical lysates from G6PD CKO and control mice. Boiled samples serve as negative controls. G Quantification of the ratio of Syntaxin1-containing complexes to total Syntaxin1 shows a significant reduction in G6PD CKO mice compared to controls. H Quantification of monomeric Syntaxin1 in non-boiled samples shows a significant increase in G6PD CKO mice compared to controls. I Quantification of Syntaxin1 in boiled samples shows a significant increase in G6PD CKO mice compared to controls. J Representative images of dendritic spines in cortical pyramidal neurons from G6PD CKO and control mice. Scale bar: 10 μm. K Quantification of spine density (spines per 10 μm dendritic shaft) shows a significant reduction in G6PD CKO mice compared to controls. Bar graphs are plotted as mean ± SEM. *p < 0.05, **p < 0.01, ***p < 0.001 (two-tailed Student’s t-test).

The reduction in synaptic vesicles may result from dysregulation of vesicle-associated proteins, including SNARE proteins, which are essential for vesicle fusion and neurotransmitter release [68, 69]. SNARE proteins form a high-molecular-weight, heat-sensitive, SDS-resistant complex that can be dissociated by boiling with SDS [38, 70]. To assess SNARE complex assembly, we measured the ratio of Syntaxin1 in the SNARE complex to total Syntaxin1 (both complexed and monomeric) in non-boiled cortical homogenates. G6PD CKO mice exhibited a significant reduction in Syntaxin1-positive SNARE complexes compared to controls (Figs. 5F, G, S2C). An increase in monomeric Syntaxin1 was observed in G6PD CKO mice (Figs. 5F, H, S2C). The SNARE complex was completely dissociated after boiling and the total Syntaxin1 in boiled cortical homogenates was significantly increased (Figs. 5F, I, S2C). These findings suggest that G6PD deficiency interferes with SNARE complex assembly, even though Syntaxin1 levels were elevated, in line with the proteomic analysis demonstrating an overall increase in synaptic proteins (Fig. 4E). To further investigate the potential impact of G6PD on synaptic function, we examined dendritic spine morphology using Golgi staining. G6PD CKO mice exhibited a significant reduction in spine density compared to controls (Fig. 5J, K), confirming a postsynaptic deficit. Together, these results demonstrate that G6PD is specifically enriched in presynaptic terminals and plays a critical role in regulating synaptic vesicle density, SNARE complex assembly, and dendritic spine development.

Discussion

In this study, we demonstrated that mice with targeted ablation of G6PD in the cerebral cortex and hippocampus exhibit schizophrenia-like behaviors, including increased locomotor activity, impaired social interaction, reduced prepulse inhibition (PPI), and cognitive deficits (Fig. 2). These findings provide direct evidence of a causal relationship between G6PD deficiency and schizophrenia-like behaviors, suggesting that G6PD loss may contribute to the pathogenesis of schizophrenia. Although no gross morphological changes were observed in the cerebral cortex and hippocampus of G6PD CKO mice, as revealed by Nissl staining and in situ hybridization (ISH) of molecular markers (Fig. 1H–L), we detected significant alterations in synaptic structures, including reduced presynaptic vesicle density and decreased dendritic spine density (Fig. 5B–E, J, K). These findings suggest that subtle morphological changes, such as alterations in axonal projections or synaptic connectivity, may exist in G6PD CKO mice and warrant further investigation.

Previous studies on G6PD-deficient mice with a point mutation that reduces protein expression reported normal motor function and sex-specific differences in cognitive performance [57]. Specifically, female G6PD-deficient mice exhibited impairments in executive function, including working and contextual memory, while male mice did not show such deficits [57]. In contrast, our study revealed a slight but significant increase in locomotor activity in male G6PD CKO mice during the open field test (Fig. 2A, B). Additionally, male G6PD CKO mice displayed impairments in several cognitive tasks, including the T-maze, novel object recognition, and contextual fear conditioning tests, indicating deficits in working and contextual memory (Fig. 2H-P). These discrepancies may be attributed to differences in the mouse models used. While previous studies examined conventional mutant mice with reduced G6PD expression throughout the brain and other organs, our study utilized conditional G6PD knockout mice with brain region-specific deletion in the cerebral cortex and hippocampus. Furthermore, we observed no changes in spatial navigation in male G6PD CKO mice during the Morris water maze test, consistent with earlier findings. To better understand the differences between these models, future studies should include behavioral tests of female G6PD CKO mice.

Despite the observed schizophrenia-like behaviors in male G6PD CKO mice, the changes were significant but not drastic. For instance, PPI was only reduced at the 82 dB prepulse level (Fig. 2F, G), and contextual memory impairment was limited to 30 min after conditioning (Fig. 2N, P). These findings suggest that G6PD may play a minor role in the etiology of schizophrenia. Given that schizophrenia is a polygenic disorder involving numerous risk genetic loci [8], we conclude that G6PD deficiency contributes to schizophrenia-like behaviors in mice, but likely as part of a broader network of genetic and environmental factors.

Our behavioral data indicate cognitive impairments in G6PD CKO mice (Fig. 2H–P), potentially attributed to synaptic dysfunction. Electron microscopy revealed a decrease in presynaptic vesicles (Fig. 5B–D), likely resulting in compromised basal and/or evoked synaptic transmission. The concurrent findings of reduced SNARE-complex levels (Fig. 5F, G) suggest a potential decrease in vesicle release probability or quantity, impacting excitatory postsynaptic potentials (EPSP) and, consequently, long-term potentiation (LTP) which is crucial for learning and memory [71]. These alterations in presynaptic activity are expected to affect postsynaptic cells through synaptic transmission or transsynaptic ligand/receptor interactions, leading to dendritic spine structural modifications [72], as we observed in G6PD CKO mice (Fig. 5J, K). While our morphological and biochemical data may be inconclusive regarding synaptic function changes in G6PD CKO mice, future studies should prioritize electrophysiological experiments.

Our proteomic analysis revealed significant dysregulation of proteins in the cerebral cortex of G6PD CKO mice. Downregulated proteins were enriched in pathways related to mitochondria, ATP synthesis, biosynthetic processes, metabolic processes, and oxidative phosphorylation, as identified through Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) analyses (Fig. 4C; Supplemental Tables 2 and 3). These findings align with the role of G6PD as a rate-limiting enzyme in the pentose phosphate pathway (PPP), which is closely associated with glucose metabolism and biosynthetic processes [73]. Such changes are expected given the central role of G6PD in maintaining cellular redox homeostasis and energy metabolism. In addition, upregulated proteins in G6PD CKO mice were significantly enriched in pathways related to synaptic vesicle fusion, synaptic membrane organization, synaptic transmission, and synaptic vesicle priming (Fig. 4D; Supplemental Table 4). Notably, over 10% of all dysregulated proteins were associated with synaptic function (Fig. 4E). These findings are consistent with recent genetic studies implicating synaptic organization, differentiation, and transmission as critical processes in the pathogenesis of schizophrenia [9, 10]. The majority of synaptic proteins revealed in our proteomic analysis exhibited an increase, while upregulated proteins only comprised around 43.56% (210/482) (Fig. 4B, E). G6PD deficiency in the cerebral cortex induced oxidative stress (Fig. 1G), potentially resulting in widespread oxidative modifications of cellular proteins, including synaptic proteins [74]. These modifications may disrupt the normal localization, structure, and function of synaptic proteins [75], such as SNARE-complex assembly observed in G6PD CKO mice (Fig. 5F, G). Further investigations are required to validate these hypotheses. Our results suggest that G6PD deficiency disrupts synaptic homeostasis, potentially contributing to the observed behavioral abnormalities in G6PD CKO mice.