Convergent blood–brain barrier breakdown in schizophrenia and autism spectrum disorders: a systematic review of preclinical animal models

Abstract

Background

Schizophrenia (SCZ) and autism spectrum disorder (ASD) are neurodevelopmental disorders with multifactorial origins involving genetic and environmental risk factors. Both conditions share overlapping pathophysiological mechanisms, including neuroinflammation, synaptic dysfunction, and circuit-level disturbances. Emerging evidence implicates blood-brain barrier (BBB) dysfunction as a potential common element in their pathogenesis. Specifically, BBB disruption is a recurring feature in preclinical models of SCZ and ASD, suggesting its role as a transdiagnostic mechanism in neurodevelopmental disorders. This systematic review aims to evaluate BBB alterations—including permeability, integrity, developmental changes, and tight junction (TJ) protein expression—in rodent models of SCZ- and ASD-like phenotypes.

Methods

Following PRISMA guidelines, we screened experimental studies assessing BBB status in murine models of SCZ and ASD compared to wild type rodents. Inclusion criteria focused on models based on genetic manipulation and environmental insults. Included studies consistently reported BBB alterations across diverse models.

Results

Findings showed disrupted TJ protein expression (claudin-5, occludin, ZO-1), increased permeability, and endothelial dysfunction. Both genetic (e.g., Shank3, 22q11.2 deletion, etc.) and environmental (e.g., maternal immune activation, valproate exposure, etc.) models exhibited BBB abnormalities. Pharmacological and experimental interventions targeting the BBB—such as claudin-5 modulation or β-catenin signaling—ameliorated BBB damage and behavioral phenotypes.

Conclusions

That BBB disruption is a recurring feature in preclinical models of SCZ and ASD suggests its pursuit as a transdiagnostic mechanism in neurodevelopmental disorders. However, as many findings rely on single studies, further replication is essential. Future studies should explore sex differences, critical developmental windows, and therapeutic strategies aimed at restoring BBB function.

Supplementary information

The online version contains supplementary material available at 10.1186/s12987-026-00818-1.

Introduction

Neurodevelopmental disorders such as schizophrenia (SCZ) and autism spectrum disorder (ASD) associate with abnormal brain development and share genetic and phenotypic overlaps [1, 2], including social and cognitive deficits that vary in their trajectory and severity [3–5]. Co-occurrence has been reported [6], along with notable male dominance: SCZ presents earlier and more frequently in males, and ASD shows a higher male-to-female ratio [7]. Both conditions have multifactorial origins involving genetic, epigenetic and environmental risk factors such as early-life stress and substance use [8–11].

SCZ affects ~0.3–0.7% of the population (~24 million people), typically manifesting in late adolescence or early adulthood [12, 13]. It is characterized by positive (e.g., hallucinations, delusions), negative (e.g., apathy, social withdrawal), and cognitive symptoms, including impairments in executive function, attention, and memory. ASD, with a median worldwide prevalence of ~1% [12, 14], emerges earlier and involves deficits in social communication, restricted interests and repetitive behaviors.

Shared pathophysiological features of SCZ and ASD include abnormal synaptic pruning [15], synaptic dysfunction [16], altered neural circuits [17], disrupted multisensory processing [18] supported by genomic and neuroimaging data [19], similar genetic determinants [20] and some convergent structural and functional brain abnormalities correlating with clinical features [21].

The blood-brain barrier (BBB) regulates molecular exchange between the circulation and the brain. These endothelial cells are joined by tight junctions (TJs)—including claudins, occludin, ZO proteins, and junctional adhesion molecules (JAMs) as major TJ proteins—with the neurovascular unit being supported by pericytes, astrocytic end-feet, and the basement membrane [22]. BBB integrity is often compromised in SCZ and ASD [23–26], and brain microvascular endothelial cells in psychotic disorders are altered, resulting in BBB disruption [27]. TJ disruption in postmortem brain tissue is a hallmark of psychiatric disorders [28], and elevated peripheral markers such as S-100β support a role for BBB dysfunction in both conditions [29–31]. Additionally, sex-related differences in BBB function have been described in psychiatric disorders [32, 33]. Given the BBB’s fundamental role in brain development [34, 35], its susceptibility to cholinergic impairments [36] and its dysregulation in SCZ and ASD, it may represent a shared pathophysiological mechanism and therapeutic target.

At the BBB, claudin-5 is the most enriched transmembrane TJ protein and its alteration has been involved in both neurological and neuropsychiatric conditions [37]. The levels of the protein product of the CLDN5 gene, located at chromosome 22q11.21, are reduced by 75% in endothelial cells from SCZ patients with 22q11 deletion syndrome [38]. Altered serum levels of claudin-5, ZO-1, and occludin have been reported in SCZ and ASD [25, 39, 40]. Antipsychotics may upregulate claudin-5 expression in a dose-dependent manner [38], suggesting that their therapeutic effects may be in part related to BBB stabilization.

It is worth noting that current treatments for both disorders remain symptomatic. In SCZ, second-generation antipsychotics are preferred [41, 42], yet ~30% of patients are treatment-resistant [43]. ASD management involves behavioral and supportive therapies, often combined with pharmacological treatment for comorbidities such as irritability, aggression, and emotional dysregulation [44–47]. However, these approaches have limited efficacy and do not address core symptoms, highlighting the need for novel therapeutic strategies that move beyond symptomatic relief and instead target the underlying biological and neurodevelopmental mechanisms. By focusing on pathways or mechanisms such as BBB integrity or functionality, it becomes possible to design interventions that not only alleviate symptoms but also modify disease trajectories, offering the potential for more precise and lasting improvements in outcomes.

Animal models recapitulating SCZ- and ASD-like phenotypes are key for understanding underlying mechanisms driven by genetic and/or environmental factors. These models replicate clinical features and shared biological alterations, including maternal immune activation -MIA- [48], microglial dysfunction [49], neuroinflammation [50], N-methyl-D-aspartate receptor (NMDAR) hypofunction [51], dendritic spine abnormalities [52–54], synaptopathies [55], interneuron dysfunction [56–58], circuit-level disturbances [59, 60] and neurovascular alterations [61, 62].

This systematic review examines BBB alterations—focusing on permeability, integrity, development, and TJ composition—in rodent models of SCZ- and ASD-like phenotypes, considering both genetic and environmental factors. To our knowledge, no recent review has systematically addressed BBB dysfunction in preclinical models of both disorders (Fig. 1).

Fig. 1.

Graphical abstract of this review. Created with Biorender.com. Abbreviations: SHANK3 = SH3 and multiple ankyrin repeat domains 3; Sema3F = Semaphorin 3F; CNTNAP2 = Contactin-associated protein-like 2; CHD8 = chromodomain helicase DNA-binding protein 8; Pten = phosphatase and TENsin homolog deleted on chromosome 10; MIA = maternal immune activation; GBHs = glyphosate-based herbicides; VPA = valproic acid; NMDAR = N-methyl-D-aspartate receptors; PPA = propionic acid

Methods

A systematic review was conducted following the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) statement [63] and a protocol registered in Open Science Framework (OSF) on September 15 2025 (https://osf.io/8f9ks).

Search strategy

A systematic literature search was conducted in two electronic databases: Pubmed (MEDLINE) and Web of Science Core Collection. These databases were selected due to their extensive coverage of biomedical and preclinical research. In Web of Science, all available Core Collection indexes were included in the search.

The search was conducted on March 25, 2025, and no filters or restrictions (e.g., language, article type, species) were applied at the database level. Instead, eligibility criteria were applied during the screening process. Articles included in the analysis focused on alterations of BBB permeability or tight junctions in animal models, including wild type mice or SCZ- or ASD-like models. The syntax was adapted to each database as follows: PubMed (MEDLINE): ((“blood-brain barrier”[Title/Abstract] OR BBB[Title/Abstract] OR “tight junction*”[Title/Abstract] OR claudin*[Title/Abstract] OR occludin[Title/Abstract])) AND ((mouse[Title/Abstract] OR mice[Title/Abstract] OR rat[Title/Abstract] OR animal*[Title/Abstract])) AND ((schizo*[Title/Abstract] OR autism[Title/Abstract] OR “autism spectrum disorder”[Title/Abstract] OR ASD[Title/Abstract])) NOT (review[Publication Type]). Web of Science Core Collection: TS = ((“blood brain barrier” OR BBB OR “tight junction*” OR claudin* OR occludin) AND (mouse OR mice OR rat OR animal*) AND (schizo* OR autism OR “autism spectrum disorder” OR ASD)) NOT TS = (review).

Additionally, a manual cross-referencing search was conducted by screening the reference lists of included articles and relevant reviews. When necessary, corresponding authors were contacted for clarification of data.

Eligibility

Inclusion criteria were selected to systematize the search in the databases so that only articles of interest on BBB or tight junctions’ alterations in SCZ and ASD animal models would be obtained. These inclusion criteria were studies on (i) BBB permeability or expression of tight junctions, (ii) studies involving mice or rats, (iii) studies involving animal models related to SCZ or ASD, and (iv) studies with control comparisons. Exclusion criteria were (i) reviews, systematic reviews and meta-analysis, (ii) articles published before 2015, (iii) studies using human samples or clinical trials, (iv) studies that do not focus on SCZ or ASD animal models and (v) articles not focused on BBB or tight junctions’ disruptions. The restriction to studies published from 2015 onwards was implemented to prioritize recent evidence generated using current methodological standards and experimental approaches, thereby improving the relevance and comparability of findings.

Study selection

Three independent reviewers (CM-M, MF-G and CM-C) screened the titles and abstracts to identify studies that met the inclusion criteria outlined above using Rayyan software [64]. The same researchers then reviewed the eligible full texts. The final list of articles was agreed to by consensus. Disagreements on eligibility were resolved by discussions with an additional reviewer (BC-F).

Data extraction, synthesis, and quality assessment

The following data were independently extracted in triplicate from each study: first author, publication year, reference, animal model, species, genetic background, sex, age, behavioral characterization (when reported), experimental methods, analyses, and results specifically related to BBB permeability or alterations in tight junctions. The methodological quality of each study was assessed using the SYRCLE Risk of Bias tool, an adaptation of the Cochrane Risk of Bias tool for animal studies [65]. Outcome measures were categorized into twelve domains: (1) Experimental model; (2) Species (rat or mouse); (3) Genetic background (e.g., Wistar, C57BL/6J); (4) Sex (male or female); (5) Age category (fetus, pup, juvenile, adolescent, adult); (6) Behavioral characterization (when provided); (7) Target outcome (BBB permeability, tight junction components, or BBB formation markers); (8) Methodology (e.g., Evans Blue assay, immunofluorescence, western blotting); (9) Sample type (whole brain or specific brain regions); (10) Observed effect; (11) Evidence of sexual dimorphism (when reported); (12) Statistical significance (defined as p < 0.05). The synthesis of results was conducted as a structured qualitative synthesis, given the heterogeneity of study designs, models, and outcome measures, which precluded formal meta-analysis.

The methodological quality of each study was assessed using the SYRCLE Risk of Bias tool, an adaptation of the Cochrane Risk of Bias tool for animal studies [65]. This tool evaluates potential sources of bias across several domains, including selection bias (random sequence generation, baseline characteristics, allocation concealment), performance bias (random housing, blinding of caregivers and investigators), detection bias (random outcome assessment, blinding of outcome assessment), and attrition bias (incomplete outcome data). Each domain was classified as “low risk”, “high risk”, or “unclear risk” of bias according to the SYRCLE criteria, depending on whether the methodological criteria were adequately reported, not reported, or insufficiently described.

Results

Data analysis: studies included and excluded

A total of 1988 articles were retrieved from the databases. Of these, 545 duplicate references found by cross-referencing were excluded, and 1421 more were excluded for the following reasons: published before 2015 (568 articles), wrong publication type (327), human or cell studies (233), studies not focused on SCZ or ASD murine models (152), studies that did not report BBB or tight junctions’ disruptions (137) and meta-analysis (4). The remaining 22 studies were included in the final analysis (Fig. 2).

Fig. 2.

PRISMA flow diagram of the reviewing process – systematic selection for inclusion or exclusion. Created with Biorender.com

Risk of bias

All studies were found to have unclear or low risk of bias according to the SYRCLE risk of bias tool (Supplementary Figure S1). The methodological quality of the included studies was assessed using the SYRCLE risk of bias tool (Supplementary Figure S1). Overall, the assessment revealed a predominance of unclear risk of bias across most domains, mainly due to insufficient reporting of methodological details.

At the domain level, random sequence generation was rated as unclear in the majority of studies (20/22), with only 2 studies classified as low risk (He et al., 2024; Kumar et al., 2022). Baseline characteristics were reported more consistently, with 6 studies classified as low risk (He et al., 2024; Yu et al., 2022; Rasile et al., 2022; Crocket et al., 2021; Schiavone et al., 2016; Kumar, 2015a), 1 as high risk (Jaini et al., 2021), and the remaining studies as unclear (15/22). Allocation concealment was rarely described, with 21/22 studies classified as unclear risk and only 1 as low risk (Schiavone et al., 2016).

Regarding performance bias, random housing was predominantly rated as unclear (20/22), with 2 studies classified as high risk (Kim et al., 2025; Schiavone et al., 2016). Blinding of interventions was rated as low risk in 6 studies (Kim et al., 2025; Yu et al., 2022; Rasile et al., 2022; Kim et al., 2019; Greene et al., 2018; Schiavone et al., 2016), while most were classified as unclear, with 1 study rated as high risk (Jaini et al., 2021).

For detection bias, random outcome assessment was largely unclear (19/22), with only 2 studies classified as low risk (Jaini et al., 2021; Schiavone et al., 2016) and 1 as high risk (Kim et al., 2025). Blinded outcome assessment showed a similar pattern, with 6 studies rated as low risk (Kim et al., 2025; O’Neill et al., 2025; Yu et al., 2022; Rasile et al., 2022; Deckmann et al., 2021; Schiavone et al., 2016), and the remainder as unclear (16/22). Incomplete outcome data were almost exclusively classified as unclear (21/22), with only 1 study rated as low risk (Kim et al., 2025).

Overall, these findings indicate that incomplete reporting of key methodological procedures represents the main limitation across studies, particularly those related to random group allocation, blinded group allocation, random housing, and incomplete outcome data.

Taken together, while BBB alterations are consistently reported across SCZ- and ASD-related models, the predominance of unclear risk of bias and methodological heterogeneity highlights limitations in reporting and the direct comparison across studies. Despite these constraints, the consistency of findings across models supports the biological relevance of BBB involvement in these disorders.

Impact of genetic factors on BBB integrity in ASD-like animal models

Across ASD-related genetic models, BBB alterations were observed in association with mutations affecting synaptic, neurodevelopmental, and regulatory pathways.

SHANK3 is a pivotal synaptic scaffolding protein involved in synapse development and functional regulation [66], and mutations in this gene have been strongly associated with ASD [67]. Shank3 rodent models have therefore been extensively utilized to explore ASD pathophysiology [68–70]. While its neuronal role is well established, Kim et al. [71] examined its function in brain endothelial cells (BECs) identifying Shank3 as the only SHANK family member expressed in these cells (eShank3), localized at cell–cell junctions during the neonatal period.

Conditional deletion of Shank3 in BECs increased BBB permeability in several brain regions, including the prefrontal cortex (PFC), cerebral cortex, and cerebellum (Table 1), specifically in male neonates. Although BBB permeability normalized in adulthood, behavioral deficits persisted, suggesting long-term consequences of early BBB dysfunction.

Table 1.

Key findings on blood–brain barrier (BBB) disruption in animal models of autism spectrum disorders (ASD)-like behavior

1) BBB Dysfunction in ASD-like Animal Models. Genetic Contributions
FIRST AUTHOR, DATE AND DOI	ANIMAL MODEL	SPECIE	BACKGROUND	SEX	AGE	BEHAVIOR CHARACTERIZATION [Test]	TARGET	METHODS	SAMPLES	EFFECT	SEXUAL DISMORPHISM	STATISTICAL SIGNIFICANCE
Kim et al. (2025). DOI: 10.1038/s41467-025–56720-1	eShank3-KO (homozygous conditional Shank3 knock-out in brain endotelian cells)	Mouse	C57BL/6 J	Males/females	Pups (PND 0–5) Adulthood (12–19 weeks old)	↓ ultrasonic comunication [USVs] (Pups) ↓ Social behaviour [RSA] (adult) ↑ Repetitive behaviours [GBT; MBT] (adult)  = Anxiety-like behaviours [OFT; D&L] (adult)	a) BBB permeability b) TJs components	a) Sodium fluorescein assay (pups/adult) b) IF - WB in eShank3-deficient BEC lines and primary BECs from eShank3-KO; transmission electron microscopy (pups)	a) Whole brain samples, PFC, cortex, thalamus, cerebellum b) Cell cultures; PFC	a) ↑ BBB permeability: ↑ in whole brain samples, PFC, cortex, cerebellum; = thalamus (pups); = (adult) b) ↓ TJs components: ↓ ZO-1 and ↓ Claudin-5 expression in BECs; altered TJ ultrastructure in the PFC	Male-specific alterations	a) p < 0.05, p < 0.01 (pups); n.s. (adult). b) p < 0.05, p < 0.01
Jagadapillai et al. (2022). DOI: 10.3390/cells11142211	Sema 3 -KO (homozygous conditional Semaphorin 3F knock-out in GABAergic neurons)	Mouse	C57BL/6J	Not specified	Adulthood (12–24 weeks old)	─	a) BBB permeability	a) IHC (CD61); IF (5-HT, fibrinogen, PECAM-1, albumin)	a) Cortex; HPC; Amg	a) ↑ BBB permeability: ↑ CD61 immunoreactivity in cortex and hippocampus, ↑ 5-HT expression,↑ fibrinogen and PECAM-1, and ↑ albumin leakage in cortex, hippocampus, amygdala	─	a) p < 0.05, p < 0.01, p < 0.001
Memis et al. (2022). DOI: 10.3390/jcm11102725	Cntnap2 KO (Homozygous contactin associated protein like 2 knock-out)	Rat	Sprague Dawley	Males/females	Not specified	─	a) BBB permeability b) TJs components	a) Fluorescein Isothiocyanate (FITC)-Dextran Assay and Evans Blue Assay, b) IF (ZO-1)	a-b) brain area not specified	a) ↑ BBB permeability: ↑ in brain sections and homogenates. b) ↓ TJs components: ↓ ZO-1 expression	Not specified	a) p < 0.05, p < 0.01 b) p < 0.01
Yu al. (2022). DOI: 10.1038/s41467-022–28746-2	CHD8 ± (heterozygous chromodomain helicase DNA-binding protein 8 haploinsufficiency)	Mouse	C57BL/6J	Males/females	Adulthood (8–14 weeks old)	↑ Anxiety-like features [OFT; D&L] ↓ Social novelty preference; = sociability [3-Ch]; = social interaction [RIT] ↓ short-term recognition memory [NORT] = stereotypic behaviour [MBT; GBT] = depressive-like behaviour [FST]	a) BBB permeability	a) Targeted metabolomics assay; Evans Blue Assay; Sc-RNAseq	a) Cerebrum; serum; cortex; BECs	a) = BBB permeability: ↑ L-Glutamine (cerebrum, serum) → ↑ Glutamate (cerebrum); = in cortex homogenate = Slc7a5 gene expression	Both males and females alterations	a) p < 0.05, p < 0.01, p < 0.001, n.s. (Evans Blue Assay; gene expression)
Jaini et al. (2021). DOI: 10.1038/s41398-021–01472-x	PtenWT/m3m4 (heterozygous knock-in for phosphatase and tensin homolog gene mutation)	Mouse	C57Bl/6	Males/females	Fetus (GD 17.5) Adolescence (PND 40)	↑ repetitive behaviours [MBT] ↓ social behaviour [3-Ch] (greater alterations in offsping from mutant mothers)	a) BBB permeability	a) IF (albumin); IHC (Glut1; PECAM-1; CD61)	a) Whole brain and perivascular areas in foetus	a) ↑ BBB permeability: ↑ albumin leakage in brain parenchyma; ↓ Glut1; PECAM-1; CD31 protein levels (greater alterations in fetus from mutant mothers)	Males > affected than females	a) p < 0.05; p < 0.01
2) BBB Dysfunction in ASD-like Animal Models. Enviromental Contributions
FIRST AUTHOR, DATE AND DOI	ANIMAL MODEL	SPECIE	BACKGROUND	SEX	AGE	BEHAVIOR CHARACTERIZATION [Test]	TARGET	METHODS	SAMPLES	EFFECT	SEXUAL DISMORPHISM	STATISTICAL SIGNIFICANCE
O’Neill et al. (2025). DOI: 10.1016/j.isci.2024.111548	Postnatal (PND 10 or 8-weeks old) infection with Bordetella pertussis	Mouse	C57BL/6J	Males/females	Juvenile (PND 16–25) for PND 10 infected mice; Adulthood (9–10 weeks old) for PND 10 infected mice; Adulthood (12–18 weeks old) for 8-weeks old infected mice	(In PND 10-infected): ↑ repetitive behaviours [MBT; OF; GBT] ↓ Social interaction, ↓ long-term recognition memory [NORT] (In Adult-infected): =	a) BBB permeability	a) Flow citometry of immune cells; IF (fibrinogen); ELISA and RT-qPCR (CXCL1/IL-17A)	a) Whole brain, cortex, HPC	a) ↑ BBB permeability: ↑ extravasation of immune cells in the brain; ↑ fibrinogen in the brain; ↑ CXCL1 and IL-17A in the brain and cortex/HPC 6–15 days post neonatal infection at PND10	Sex-matched, not specified	a) p < 0.05, p < 0.01, p < 0.001
He et al. (2024).DOI: 10.1016/j.ecoenv.2024.117060	Maternal exposure to GBHs (GD 0 to PND 21)	Mouse	BALB/c	Males	Juvenile-adolescence (PND 24–43)	↓ Social behaviours (sociability, social novelty) [3-Ch] ↑ Repetitive behaviours [GBT; MBT]	b) TJs components	b) WB (ZO-1, claudin-5, occludin); IF (ZO-1); electron microscopy	b) PFC	b) ↓ TJs components: ↓ ZO-1, ↓ occludin, and ↓ Claudin-5 in PFC; altered TJ ultrastructure in the PFC	─	b) p < 0.05
Mehra et al. (2024). DOI: 10.1007/s12035-023–03826-6	Prenatal (GD 12.5) exposure to VPA (400 mg/Kg i.p.)	Rat	Wistar	Not specified	Pups/Juvenile (PND 3–11); Adolescence (PND 32–35)	(pups/juvenile): altered reflex responses [Developmental Reflex Assays]; (Adolescence) ↓social dominance [Tube test] ↑ anhedonic behaviour [SPT]	a) BBB permeability b) TJs components	a) Evans Blue Assay; b) qPCR (claudin-2, claudin-3, claudin-5)	a-b) Cerebellum, cerebrum	a) ↑ BBB permeability: ↑ absorbance in cerebellum and cerebellum. b) ↓ TJs components: ↑ claudin-2 (leaky BBB); ↓ claudin-3 and ↓ claudin-5 in cerebellum and cerebrum	─	a) p < 0.05 b) p < 0.05
Kumar et al. (2023). DOI: 10.32598/bcn.2021.2966.1	Postnatal PPA (250 mg/kg/po) administration on PND 21–23	Rat	Wistar	Males	Adolescence (PND 39–50)	↑ Anxiety-like features [EPM]; ↓ Sociability, ↓Social novelty preference [3-Ch]; ↓ repetitive behaviours/altered WM [Y-maze]	a) BBB permeability	a) Evans Blue Assay; water content measurements.	a) Cerebellum	a) ↑ BBB permeability: ↑ absorbance and water content in cerebellum.	─	a) p < 0.05
Tran et al. (2023). DOI: 10.1016/j.jff.2023.105516	Prenatal (GD 12.5) exposure to VPA (600 mg/Kg i.p.)	Mouse	ICR	Males	Adolescence (PND 36–40)	↑ Anxiety-like behaviours [OFT; D&L] ↑ Repetitive behaviours [GBT] ↓ Social behaviours (sociability, social novelty) [3-Ch]	a) BBB permeability b) TJs components	a) Evans Blue Assay, water content measurements. b) WB (ZO-1, claudin-5, MMP-2)	a) PFC, HPC, cerebellum; whole brain b) PFC, HPC, cerebellum	a) ↑ BBB permeability: ↑ absorbance in PFC, hippocampus, cerebellum (>in hippocampus); ↑ brain water content in whole brain b) ↓ TJs components: ↓ ZO-1, ↓ Claudin-5, and ↑ MMP-2 in PFC, hippocampus, cerebellum	─	a) p < 0.01, p < 0.001 b) p < 0.05, p < 0.01
Kumar et al. (2022). DOI: 10.9758/cpn.2022.20.4.725	Prenatal (GD 12.5) exposure to VPA (500 mg/Kg i.p.)	Rat	Wistar	Males	Adolescence (PND 49–50)	↓ Social behaviours (sociability, social novelty) [3-Ch]	a) BBB permeability	a) Evans Blue Assay, water content measurements.	a) Cerebellum	a) ↑ BBB permeability: ↑ absorbance in homogenates of cerebellum; ↑ brain water content in whole cerebellum	─	a) p < 0.05
Deckmann et al. (2021). DOI: 10.1002/jdn.10137	Prenatal (GD 12.5) exposure to VPA (600 mg/Kg i.p.)	Rat	Wistar	Males	Adolescence (PND 30)	─	a) BBB permeability	a) Evans Blue Assay, confocal microscopy; water content measurements.	a) Whole brain, mPFC, primary somatosensory cortex, Amg, HPC, choroid plexus	a) ↑ BBB permeability: ↑ absorbance in choroid plexus, mPFC (all subdivisions), primary somatosensory cortex (II/III, IV/V) layers; = absorbance in HPC and Amg; ↑ brain water content in whole brain	─	a) p < 0.05; p < 0.01; p = 0.05 (Amg)
Kim et al. (2019). DOI: 10.4062/biomolther.2018.230	Acute exposure to NMDA (50–75 mg/Kg i.p.)	Mouse	ICR	Not specified	Juvenile/Adulthood (3–8 weeks old)	age-dependent (juvenile > adult) ↑ repetitive and nociceptive behaviours [Behaviour observation, Hot plate test, Writhing test]	a) BBB permeability b) TJs components	a) Evans Blue Assay, IF (IgG); b) IF (ZO-1)	a-b) PFC; thalamus	a) = BBB permeability: no Evans Blue extravasation and no IgG in tissues; b) = TJs components: no changes in ZO-1 protein	─	a-b) n.s
Bhandari et al. (2018). DOI: 10.1016/j.jff.2018.05.065	Postnatal (3–4 months old) PPA (1 M) intracerebroventricular administration	Rat	Sprague-Dawley	Males/females	Adulthood (4–5 months old)	(3–4 weels after PPA) ↓ sociability,↓ social novelty preference [Reciprocal social interaction test; 3-Ch; partition test = no sex-based differences]; ↑ repetitive behaviours [GBT; MBT]; Communication deficits [olfactory habituation/dishabituation test]; Motor dysfunction [rotarod; actophotometer]; ↑ anxiety-like behaviours [EPM; OFT]; ↑ depressive-like behaviours [FST]; Altered recognition memory [NORT]; = spatial learning, ↓ reversal learning, [Morris MWM]	a) BBB permeability	a) Evans Blue Assay; ELISA (P-glycoprotein (P-gp)/ABCB1 transporter)	a) Whole brain	a) ↑ BBB permeability: ↑ in whole brain; upregulation of (P-gp)/ABCB1 transporter	Not specified	a) p < 0.05
Kumar & Sharma (2016a). DOI: 10.1016/j.brainresbull.2016.03.013	Prenatal (GD 12.5) exposure to VPA (500 mg/Kg i.p.)	Rat	Wistar	Males	Adolescence (PND 43–50)	↑ Locomotor activity [Actophotometer], ↓ Social behaviours (↓ sociability, ↓ social novelty) [3-Ch], ↑repetitive behaviour [Y-Maze], ↑ anxiety-like behaviours [EPM] exploratory activity, ↓ exploratory activity [hole-board apparatus]	a) BBB permeability	a) Evans Blue Assay, macroscopic inspection	a) Cerebellum	a) ↑ BBB permeability: ↑ in homogenates of cerebellum and in whole cerebellum (macroscopic inspection)	─	a) p < 0.001
Kumar & Sharma (2016b). DOI: 10.1016/j.brainres.2015.10.052	Prenatal (GD 12.5) exposure to VPA (500 mg/Kg i.p.)	Rat	Wistar	Males	Adolescence (PND 43–50)	↑ Locomotor activity [Actophotometer], ↓ Social behaviours (↓ sociability, ↓ social novelty) [3-Ch], ↑repetitive behaviour [Y-Maze], ↑ anxiety-like behaviours [EPM] exploratory activity, ↓ exploratory activity [hole-board apparatus]	a) BBB permeability	a) Evans Blue Assay, macroscopic inspection	a) Cerebellum	a) ↑ BBB permeability: ↑ in homogenates of cerebellum and in whole cerebellum (macroscopic inspection)	─	a) p < 0.001
Kumar al. (2015). DOI: 10.1016/j.neuint.2015.10.007	Prenatal (GD 12.5) exposure to VPA (500 mg/Kg i.p.)	Rat	Wistar	Males	Adolescence (PND 43–50)	↑ Locomotor activity [Actophotometer], ↓ Social behaviours (↓ sociability, ↓ social novelty) [3-Ch], ↑repetitive behaviour [Y-Maze], ↑ anxiety-like behaviours [EPM] exploratory activity, ↓ exploratory activity [hole-board apparatus]	a) BBB permeability	a) Evans Blue Assay, macroscopic inspection	a) Cerebellum	a) ↑ BBB permeability: ↑ in homogenates of cerebellum and in whole cerebellum (macroscopic inspection)	─	a) p < 0.001

Abbreviations: 3-Ch = three-chambers test; 5-HT = serotonin; Amg = amygdala; BBB = blood–brain barrier; BECs = brain endothelial cells; CD61 = Cluster of Differentiation 61; CXCL1 = C-X-C motif chemokine ligand 1; D&L = dark & light box test; EPM = elevated plus-maze; FST = forced swimming test; GBHs = glyphosate-based herbicides; GBT = grooming behavioral test; GD = gestational day; HPC = hippocampus; IF = immunofluorescence; IHC = immunohistochemistry; IL-17A = Interleukin-17A; MBT = marbles-burying test; MMP-2 = Matrix Metalloproteinase-2; mPFC = medial prefrontal cortex; MWM = Morris water maze; NORT = novel object recognition test; NMDA = N-methyl-D-aspartate; OFT = open field test; PECAM-1 = Platelet Endothelial Cell Adhesion Molecule-1; PFC = prefrontal cortex; PND = post-natal day; PPA = propionic acid; qPCR = quantitative poli chain reaction; RIT = reciprocal interaction test; RSA = Round social arena; TJs = tight junctions; USVs = Ultrasonic vocalizations test; VPA = valproic acid; WB = western blot

Mechanistically, eShank3 interacts with TJ proteins including ZO-1 and claudin-5 [72–74]. Its loss disrupted TJ architecture and reduced their expression in both BEC lines and primary cells, with ultrastructural abnormalities confirmed by electron microscopy (Table 1).

Semaphorin 3F (SEMA3F), originally characterized as an axonal guidance cue [75], has also been implicated in ASD. Its deletion in inhibitory interneurons induces ASD-like phenotypes, microglial activation, and oxidative stress [76]. Jagadapillai et al. [23] further demonstrated that interneuron-specific loss of Sema3F disrupts electrophysiological activity and is associated with neuroinflammation, increased nitrosative stress markers, and BBB impairment, including platelet activation, endothelial interactions, and albumin extravasation (Table 1).

Contactin-associated protein-like 2 (CNTNAP2) is involved in synaptic organization and neurotransmission [77], and its mutations have been linked to ASD [78, 79]. Consistent with its behavioral characterization [80, 81], Memis et al. [82] showed that Cntnap2 deletion increases BBB permeability and reduces ZO-1 expression (Table 1), indicating BBB disruption associated with oxidative and nitrosative stress.

CHD8 gene, which encodes the chromodomain helicase DNA-binding protein 8 (CHD8), is among the most strongly associated with ASD [83–87]. Yu et al. [88] used Chd8^± mice as an ASD model to explore host–microbiota interactions. These mice showed ASD-like features including anxiety-like behavior, mild social disturbances, cognitive deficits, and excitatory/inhibitory (E/I) imbalance, but lacked stereotypies and stress-related alterations.

These mice exhibited increased glutamate-related metabolites (Table 1). As glutamate does not cross the BBB [89] and is synthesized from precursors [90], elevated levels could reflect altered metabolism or BBB dysfunction. However, no changes in BBB permeability or transporter expression were detected (Table 1). Instead, increased intestinal transporter expression and microbiota alterations were observed, supporting a gut–brain axis mechanism [91].

The phosphatase and tensin homolog (PTEN) mutations are strongly implicated in ASD [92–94]. Using Pten (Pten^WT/m3m4), Jaini et al. [95] showed that maternal genotype influences offspring phenotype. Embryos from mutant mothers exhibited significantly increased BBB disruption and permeability markers (Table 1), demonstrating that maternal genetic background modulates BBB development.

Impact of environmental factors on BBB integrity in ASD-like animal models

Prenatal exposure to infectious agents has been consistently associated with an elevated risk of ASD in offspring [96–98]. O’Neill et al. [99], showed that neonatal Bordetella pertussis infection increased BBB permeability, neuroinflammation, and ASD-like behaviors.

Glyphosate-based herbicides (GBHs) have also been linked to ASD [100, 101]. He et al. [102] reported that maternal GBH exposure induced ASD-like behaviors, microbiota alterations, and reduced TJ protein expression in the PFC (Table 1).

Prenatal valproic acid (VPA) exposure is a well-established ASD model [103–105]. Across eight studies, VPA consistently induced BBB dysfunction, including increased permeability and TJ dysregulation [106–113] (Table 1). These disruptions occurred concurrently with neuroinflammatory responses, elevated oxidative and nitrosative stress, calcium homeostasis imbalance, glial cell overactivation, and alterations in gut microbial composition. Notably, therapeutic interventions—including pharmacological agents with diverse mechanisms of action (nimodipine: L-type calcium channel antagonist; memantine: NMDA receptor antagonist; agomelatine: melatonergic agonist; minocycline: anti-inflammatory) and natural compounds (resveratrol and fisetin: antioxidants; insect-derived bioactive agents)—exerted broad-spectrum protective effects by modulating oxidative stress, inflammation, mitochondrial bioenergetics, and BBB stability.

Several lines of evidence suggest that early inhibition of N-methyl-D-aspartate (NMDA) receptor activity may ameliorate ADS-like behaviors—including social deficits, repetitive behaviors, anxiety-like responses, and decreased nociceptive thresholds—in adult rats prenatally exposed to VPA [114]. Building on these findings, Kim et al. [115] investigated the behavioral effects of acute NMDA activation in ICR mice, specifically evaluating nociceptive sensitivity and repetitive behaviors. Intraperitoneal administration of NMDA at doses ranging from 50 to 75 mg/kg elicited increased grooming and tail biting/licking behaviors in a dose- and age-dependent manner, with juvenile mice exhibiting greater sensitivity than adults. Despite the behavioral alterations, the authors reported no evidence of BBB disruption (Table 1), concluding that these changes in behavior were independent of BBB impairment.

Propionic acid (PPA), a microbiota-derived metabolite [116–118], also induces ASD-like phenotypes. Both subacute [107] and acute [118] exposure increased BBB permeability and neuroinflammation (Table 1). Increased P-glycoprotein expression correlated with inflammatory markers, and therapeutic interventions mitigated these effects.

Impact of genetic factors on BBB integrity in SCZ-like animal models

Copy number variants such as the 22q11.2 deletion significantly increase SCZ risk [119–121]. Crockett et al. [121] investigated whether the BBB is intrinsically impaired in the context of SCZ linked to 22qDS. BBB-like endothelial cells derived from induced pluripotent stem cells of SCZ patients with 22qDS displayed compromised barrier function, a phenotype that was corroborated in a mouse model harboring the 22q11.2 deletion [Df (h22q11)/+]. In these mice, the deletion led to elevated parenchymal accumulation of serum proteins—specifically IgG and fibrinogen—in the PFC, alongside a significant reduction in claudin-5 expression (Table 2). However, no significant changes were observed in ZO-1 expression or in the ultrastructural morphology of TJs when compared to wild-type controls (Table 2). These findings provide translational evidence for BBB dysfunction linked to 22q11.2 deletion, characterized by selective disruption of claudin-5 without broader alterations in TJ architecture. This BBB impairment was further associated with elevated expression of the pro-inflammatory adhesion molecule ICAM-1 and enhanced leukocyte infiltration and activation. Importantly, these results were also validated in postmortem brain tissue from individuals with 22qDS, reinforcing their clinical relevance.

Table 2.

Key findings on blood–brain barrier (BBB) disruption in animal models of schizophrenia (SCZ)-like behavior

1) BBB Dysfunction in SCZ-like Animal Models. Genetic Contributions
FIRST AUTHOR, DATE AND DOI	ANIMAL MODEL	SPECIE	BACKGROUND	SEX	AGE	BEHAVIOR CHARACTERIZATION [Test]	TARGET	METHODS	SAMPLES	EFFECT	SEXUAL DISMORPHISM	STATISTICAL SIGNIFICANCE
Crockett et al. (2021). DOI: 10.1093/brain/awab055	Df (h22q11)/+	Mouse	C57BL/6	Males/females	Adulthood (4–6 months old)	─	a) BBB permeability b) TJs components	a) IF, serum protein levels (IgG; fibrinogen) b) qPCR, (claudin-5); IF, WB (claudin-5, ZO-1); electron microscopy	a) PFC b) Anterior cortex	a) ↑ BBB permeability: ↑ extravasation of IgG and fibrinogen in the PFC; b) TJs components: ↓ claudin-5 (mRNA and protein); = ZO-1 levels; = TJs ultrastructure	Sex-matched, not specified	a) p ≤ 0.001, p ≤ 0.0001; b) p < 0.05 (claudin-5), n.s. (ZO-1); n.s. TJs ultrastructures
Greene et al. (2018). DOI: 10.1038/mp.2017.156	Inducible claudin-5 knockdown in adult mice	Mouse	C57/BL6J	Not specified	Adulthood (15–19 weeks old)	↓ long-term recognition memory [NORT]; ↓ working memory [T-Maze]; ↑ anxiety-like behaviours [EPM] ↑ side bias [Y-Maze]; ↓ acustic prepulse inhibition [PPI]; =/↑ (trend) depressive-like behaviours [SST, FST, OFT]; =/↓ (trend) social preference and novelty [3-Ch]; =/altered (trend) locomotor activity [RotaRod, OFT]	a) BBB permeability b) TJs components	a) IF (biotin, fibrinogen), MRI; b) qPCR, WB, IF	a) mPFC and dHPC; whole brain b) mPFC and dHPC; whole brain	a) ↑ BBB permeability: ↑ extravasation biotin and fibrinogen in mPFC and dHPC; ↑ extravasation of contrast agent in the whole brain. b) TJs components: ↓ claudin-5 mRNA and protein levels in brain homogenates; = ZO-1, occludin, tricellulin levels (characterization of the mouse model)	Not specified	a) p < 0.05, p < 0.01; b) p < 0.05, p < 0.01, p < 0.001
2) BBB Dysfunction in SCZ-like Animal Models. Enviromental Contributions
FIRST AUTHOR, DATE AND DOI	ANIMAL MODEL	SPECIE	BACKGROUND	SEX	AGE	BEHAVIOR CHARACTERIZATION [Test]	TARGET	METHODS	SAMPLES	EFFECT	SEXUAL DISMORPHISM	STATISTICAL SIGNIFICANCE
Rasile et al. (2022). DOI: 10.15252/embj.2022111192	MIA induced by Poly I:C (2 mg/Kg ip) on GD 9	Mouse	C57BL/6	Males/females	Fetus (GD 9 + 6 hours; E17) Adolescence (PND 30) Adulthood (PND 90)	(Adulthood, PND 90) ↑ anxiety-related behaviours in males [MBT]; ↑ repetitive behaviours in males [Self-grooming test]	a) BBB permeability b) TJs components c) BBB formation markers	a) Evans Blue Assay, albumin extravasation, microscopy evaluation; b) WB, IF (claudin-5, ZO-1, occludin); c) qPCR, WB (Pecam1, CD248, VEGFa, PDGFb, TGF-β, Foxf2)	a) Whole brain b-c) cortex	a) ↑ BBB permeability: ↑absorbance and albumin extravasation on both PND 30/PND 90 (male specific); b) ↓ TJs components: ↓ claudin-5, ↓ ZO-1, = occludin on PND 90 (male specific); c) BBB formation markers: ↓ VEGFa, ↑TGF-β1 (males), ↑ CD248, ↓ VEGFa,↑TGF-β2 (females) on GD9 + 6 hrs; ↑ Pecam1,↑ CD248, ↑ VEGFa, ↑ PDGFb, ↑ TGF-β1, ↑ TGF-β2, ↑ TGF-β3, ↑ Foxf2 at E17 (males); ↓ VEGFa, ↑ TGF-β1 on PND 90.	Male-specific alterations; some mild effect in females	a) p = 0.06 (trend), p < 0.05; b) p < 0.05; p < 0.01; c) p < 0.05; p < 0.01; p < 0.001
Simões et al. (2018). DOI: 10.1016/j.jpsychires.2018.02.007	MIA induced by LPS (0.25 mg/kg i.p.) on GD 15	Rat	Wistar	Fetus/males	Fetus (GD 15–16) adulthood (PND 60) (behaviour, sensitivity to repeated KET exposure during late adolescence on PND 54–60)	(control versus LPS)  = locomotor activity, ↓ habituation memory [OFT]; ↓ recognition memory [NORT]; = acustic prepulse inhibition [PPI] (KET in control versus LPS) similar ↑ of locomotr activity, more ↓ of habituation and recognition memories in LPS group at low KET [OFT, NORT]; more ↓ of acustic prepulse inhibition in LPS group at medium-high KET [PPI]	a) BBB permeability	a) Evans Blue Assay	a) Maternal HPC and cortex; placenta; fetus brain	a) ↑ BBB permeability: ↑ absorbance in maternal HPC and cortex (6–12-24 hours post LPS); ↑ fluorescence in fetus brain (6, 24 hours post LPS); ↑ fluorescence in placenta (6 hours post LPS)	─	a) p < 0.05
Schiavone et al. (2017). DOI: 10.1007/s12035-016–9791-8	Social isolation (1 week from weaning on PND 21)	Rat	Wistar	Males	Juvenile (PND 28)	─	a) BBB permeability b) TJs components c) BBB formation/integrity markers	a) Evans Blue Assay, Quantification of brain interstitial fluid. b) WB (occludin) c) qPCR, IF, WB (ORLs, Vof 21, Vof 16, Leng 8, Vnr1, Trank1, MMP9, MMP2, PV-1)	a) Whole brain. b-c) PFC	a) ↑ BBB permeability: ↑ absorbance and ↑ interstitial fluids in homogenates and whole brain b) TJs components: ↓ occludin in the PFC 3) BBB formation/integrity markers: ↑ mRNA and protein levels (all) in the PFC	─	a) p < 0.05; p < 0.01 b-c) p < 0.05, p < 0.01, p < 0.001

Abbreviations: 3-Ch = three-chambers test; BBB = blood–brain barrier; dHPC = dorsal hippocampus; E = embryonic day; EPM = elevated plus-maze; GD = gestational day; HPC = hippocampus; IF = immunofluorescence; i.p. = intraperitoneal; KET = ketamine; LPS = lipopolysaccharide; MBT = marbles-burying test; MIA = maternal immune activation; mPFC = medial prefrontal cortex; MRI = magnetic resonance imaging; NORT = novel object recognition test; OFT = open field test; PFC = prefrontal cortex; PND = post-natal day; Poly I:C = polyinosinic:polycytidylic acid; PPI = pre-pulse inhibition test; qPCR = quantitative polymerase chain reaction; SCZ = schizophrenia; SST = sucrose splash test; TJs = tight junctions; WB = western blot

Similarly, Greene et al. [38] identified a genetic variant in claudin-5 that is associated with SCZ in individuals with 22qDS, resulting in a marked (~75%) reduction in claudin-5 expression within endothelial cells. At the experimental level, the authors demonstrated that sustained downregulation of claudin-5 in the mPFC and dHPC of adult mice led to behavioral SCZ-like phenotypes, alongside increased BBB permeability (Table 2). Importantly, the expression levels of other TJ-associated proteins—ZO-1, occludin, and tricellulin—remained unchanged following doxycycline-inducible claudin-5 knockdown (Table 2), highlighting the specificity of the observed effects to claudin-5 suppression. Moreover, prolonged claudin-5 depletion resulted in the onset of seizures and mortality within 3–4 weeks, underscoring the essential role of this protein in maintaining normal neurological function. The study also reported that antipsychotic medications increased claudin-5 expression in a dose-dependent manner both in vitro and in vivo. In parallel, postmortem analysis of SCZ patient brain tissues revealed aberrant and discontinuous claudin-5 expression patterns.

Impact of environmental factors on BBB integrity in SCZ-like animal models

Maternal immune activation (MIA), is a major environmental risk factor [48, 122–125]. Simões et al. [126] reported that LPS-induced MIA during late gestation (gestational day (GD) 15) in rats provoked acute inflammatory responses, oxidative stress, and increased BBB permeability within 6 to 24 hours post-administration in fetal brain tissues, placenta, and the hippocampus and cortex of pregnant dams (Table 2). In adulthood, offspring exposed to both MIA and repeated ketamine administration during late adolescence (postnatal days 54–60) exhibited impairments in memory performance and sensory motor gating.

Rasile et al. [127] also utilized a murine model of MIA. A single administration of the viral mimic Poly (I:C) at GD 9—mimicking early prenatal viral infection—elicited sex-dependent neurovascular consequences in the offspring. Specifically, male progeny exhibited increased BBB permeability during adolescence (postnatal day, PND 30) and adulthood (PND 60), accompanied by a marked reduction in TJ proteins, including claudin-5 and ZO-1, but not occludin, by PND 90 (Table 2). Notably, these BBB impairments were restricted to male mice, highlighting a sex-specific vulnerability. Transcriptomic profiling at embryonic day 17 identified widespread alterations in vascular development-related genes, with transforming growth factor-beta 1 showing the most consistent dysregulation (Table 2).

Social withdrawal during early life stages in rodents is a model of psychosocial stress and provides a valuable nonpharmacological tool to study SCZ-like dysfunctions and brain circuits involved [128]. In this context, Schiavone et al. [129] carried out an extensive study investigating early molecular alterations in markers related to BBB breakdown and neuroinflammation/oxidative stress after one week of post-weaning social isolation in Wistar rats. Rats exposed to early social isolation displayed a strong dysregulation in the PFC of genes and their protein products involved in BBB formation and integrity (Table 2). The expression of specific markers of BBB integrity, such as matrix metalloproteinase 2, MMP9, occludin, and plasmalemmal vesicle associated protein-1, was also significantly altered after 1 week of social isolation (Table 2). Consequently, BBB permeability was significantly increased in rats isolated for 1 week with respect to controls (Table 2). These effects were accompanied by a significant increase in interleukin 6 expression, with no differences in NOX2 levels, suggesting that BBB disruption precedes oxidative stress in the brain.

Integrative synthesis: patterns and relative strength of evidence

The collective evidence from the 22 included studies reveals a complex but structured landscape of BBB dysfunction across ASD and SCZ models. A thematic integration identifies three critical patterns:

1. Sex-specific vulnerability. Sex-specific vulnerability represents a recurrent but inconsistently analyzed feature across the included studies. Only a subset of studies (8/22) explicitly included both male and female subjects, and even fewer performed formal sex-stratified analyses or tested sex-by-genotype/environment interactions. In most cases, sex was either pooled or reported only as a demographic characteristic without further analytical stratification, limiting interpretability. Among studies that did evaluate sex-dependent effects, a predominant pattern of increased male vulnerability was observed. This was particularly evident in genetic models such as Shank3 [71] and Pten [95], as well as in environmental models including maternal immune activation (MIA) [129], where BBB disruption and associated behavioral abnormalities were more pronounced in males. In contrast, some models, such as Chd8 haploinsufficiency, did not reveal sex-dependent differences in either behavioral or BBB outcomes [88], suggesting that sex effects may depend on the specific genetic or environmental insult. Importantly, several studies failed to specify the sex of the animals used [23, 38, 114, 130], further limiting the systematic evaluation of sex as a biological variable. Taken together, these findings indicate that while male-biased vulnerability to BBB disruption emerges as a recurring trend, the current literature is insufficiently powered and inconsistently designed to draw definitive conclusions regarding sex-dependent neurovascular mechanisms in ASD and SCZ. This represents a significant methodological gap that should be addressed in future preclinical research through mandatory inclusion of both sexes and formal testing of sex-by-condition interactions.

2. Developmental vulnerability. Temporally, most reviewed studies for both ASD- and SCZ-like models identified BBB alterations across early developmental stages—fetal [95, 126, 127], perinatal [71], juvenile [99, 102, 129], and adolescent [106–113, 127]. This underscores how genetic and environmental early-life stressors may converge to trigger early BBB compromise and behavioral disruptions relevant to ASD and SCZ. Importantly, Shank3 deficiency causes BBB leakage during the neonatal period that normalizes by adulthood [71], indicating a transient window of vulnerability. In contrast, prenatal MIA leads to BBB impairments that manifest during adolescence and persist into adulthood [127], suggesting that some environmental insults may initiate a more chronic or progressive neurovascular decline.

3. Relative strength of evidence. When comparing “insult” types, environmental factors demonstrate a more robust and widespread impact on BBB integrity. While some genetic models show no central BBB disruption (Chd8 [88]) or only selective claudin-5 gene and protein dysregulation (22q11.2 [121]; claudin-5 downregulation [38]), environmental stressors—such as VPA [106–113], PPA [107, 118], B. pertussis infection [99], MIA [126, 127], and social isolation [129]—consistently trigger a more aggressive cascade of neuroinflammation, oxidative stress, and measurable parenchymal extravasation. This suggests that environmental “hits” may be more potent drivers of vascular failure than single genetic predispositions.

4. Molecular Convergence: Despite the heterogeneity of ASD and SCZ, the downregulation of claudin-5 and/or ZO-1 emerges as a shared molecular denominator of BBB compromise across genotypes (Shank3, Cntnap2, 22q11.2) and environmental exposures (VPA, GBH, MIA) [23, 38, 71, 82, 102, 112, 127]. Interestingly, the absence of BBB leakage in models such as Chd8 [88] or acute NMDA activation [115] highlights that neurodevelopmental deficits can sometimes occur independently of vascular disruption. This suggests that BBB failure may act as a specific pathological subtype or a secondary catalyst that significantly exacerbates the phenotypic outcomes.

5. Transdiagnostic convergence and disorder-specific divergence: While both ASD- and SCZ-related models consistently exhibit BBB alterations, a more granular comparison reveals both shared and distinct patterns. At the molecular level, convergence is evident in the recurrent dysregulation of tight junction proteins—particularly claudin-5 and ZO-1—across both disorders [23, 38, 71, 82, 104, 114, 129], suggesting a common pathway of endothelial dysfunction. However, disorder-specific features also emerge. ASD models more frequently exhibit early-life, transient, or developmentally restricted BBB disruption (e.g., Shank3 [71]), often tightly linked to synaptic and neurodevelopmental gene mutations [71, 82, 96] and gut–brain axis alterations [104, 115]. In contrast, SCZ models—particularly those involving 22q11.2 deletion [122] or maternal immune activation [128, 129]—tend to show more persistent BBB impairment associated with neuroinflammatory and immune-related mechanisms [128, 129, 131]. Furthermore, environmental insults appear to exert a stronger and more consistent impact on BBB integrity in SCZ-related models [128, 129, 131] compared to genetic models [38, 122], whereas in ASD both genetic and environmental factors contribute more heterogeneously [71, 82, 96, 104, 108–115, 132]. Together, these findings suggest that BBB dysfunction represents a shared but mechanistically heterogeneous process, with both convergent and disorder-specific pathways contributing to neurodevelopmental pathology.

Discussion

SCZ and ASD are neurodevelopmental psychiatric conditions that share significant genetic and environmental risk factors, as well as overlapping neurobiological mechanisms. The studies included in this systematic review investigated preclinical models involving genetic perturbations of key neurodevelopmental genes, including those regulating synapse structure and function (e.g., Cntnap2, Shank3) [71, 82], axonal growth and migration (e.g., Sema3F, Chd8) [23, 88], cell growth regulation (Pten) [95], and chromosomal structural variants linked to SCZ risk such as the 22q11.2 deletion [121], which is also associated with altered claudin-5 levels [38], a key TJ component of the BBB. Environmental models included exposure to infectious agents (e.g., Bordetella pertussis) [99], chemical agents and drugs (e.g., PPA, glyphosate-based herbicides, VPA) [104, 108–115, 132], MIA [126, 127], and early-life social isolation [129]. Each of these factors contributes to the disruption of neurodevelopmental trajectories through the induction of neuroinflammation, oxidative stress, and mitochondrial dysfunction. While these diverse genetic and environmental insults appear to converge on BBB disruption, a direct comparison between ASD and SCZ models reveals both shared mechanisms and disorder-specific patterns of neurovascular dysfunction—altering its permeability and TJ integrity—it remains to be fully elucidated whether this neurovascular impairment is a primary driver or a secondary consequence of the underlying neuroinflammatory and neural pathology in animal models relevant to SCZ and ASD, suggesting a potentially bidirectional relationship that warrants further mechanistic investigation.

Importantly, a transdiagnostic comparison of ASD- and SCZ-related models highlights both convergence and divergence in BBB pathology. Convergent evidence across both disorders supports a central role for tight junction dysregulation—particularly involving claudin-5 and ZO-1—as a common molecular substrate of BBB impairment [23, 38, 71, 82, 104, 114, 129]. In addition, both conditions show strong links between BBB disruption and neuroinflammatory processes, oxidative stress, and altered neurodevelopmental trajectories [23, 104, 108–115, 128, 129]. However, key differences emerge. ASD models more frequently implicate synaptic and neurodevelopmental genes (e.g., Shank3, Cntnap2, Pten) [71, 82, 96] and gut–brain axis interactions [104, 115], often with BBB alterations occurring during early developmental windows and, in some cases, showing partial recovery [71]. In contrast, SCZ models more prominently involve immune-related mechanisms (e.g., MIA) [128, 129] and structural genomic variants (e.g., 22q11.2 deletion) [122], with evidence suggesting more persistent or progressive BBB dysfunction [129, 131]. These distinctions indicate that, although BBB disruption may represent a transdiagnostic feature, the upstream drivers and temporal dynamics of neurovascular impairment differ between ASD and SCZ.

Regarding sex-specific differences, only 8 of the 22 reviewed studies explicitly included both male and female rodents. While some reported male-specific or more severe effects in males [71, 95, 127]—aligning with the male-biased prevalence of SCZ and ASD [7]—others pooled both sexes without performing an analysis of sexual dimorphism [82, 101, 122, 132]. Yu et al. [88] found no sex-specific effects of Chd8 haploinsufficiency on behavior or BBB permeability in male and female mice. Additionally, certain studies failed to specify the sex of the subjects used [23, 38, 112, 115]. Despite the higher prevalence of both disorders in male subjects compared with their female counterparts, these findings emphasize the need for sex-inclusive research, given growing evidence of sex-specific neurobiological mechanisms in SCZ and ASD and the underrepresentation of females in preclinical research. Notably, the biological basis for this dimorphism at the neurovascular level is still poorly understood; consequently, the fact that several studies pooled both sexes or failed to specify them [23, 38, 112] restricts a comprehensive understanding of these putative sex-specific mechanisms

The gut-brain axis has emerged as a potential and intriguing contributor to neurodevelopmental disorders [133], particularly in ASD. Several clinical studies report consistent microbiota alterations in individuals with ASD [131, 134, 135]. Preclinical models (e.g., Shank3, MIA, VPA) also show that microbiota can influence behavior, gut permeability, and neuroinflammation [136–138]. Interventions like antibiotics, probiotics, and fecal microbiota transplantation (FMT) have shown promising effects on behavior and gastrointestinal (GI) symptoms [136, 138, 139]. Similarly, SCZ has been associated with GI inflammation, immune dysregulation, and altered microbiota [140, 141]. FMT from SCZ patients to germ-free mice induced neurotransmitter imbalances [142], while antibiotics reduced microglial pruning and were linked to lower SCZ incidence [143]. Although probiotics show limited impact on psychiatric symptoms, they may alleviate GI issues [144, 145].

Mechanistically, it has been hypothesized that gut dysbiosis may increase intestinal permeability and systemic inflammation, potentially leading to BBB disruption and abnormal neurodevelopment in SCZ and ASD [146–148]. In line with this hypothesis-generating framework, maternal exposure to GBHs or VPA led to ASD-like behavior, gut dysbiosis, and BBB dysfunction [102, 113]. Postnatal PPA exposure also impaired the BBB [107, 118]. While BBB impairment was not directly evaluated in this study, PPA, a short-chain fatty acid biologically produced by the gut microbiota, has been implicated in increased intestinal permeability when present at elevated concentrations. In contrast, Yu et al. [88] demonstrated that Chd8 haploinsufficiency did not affect the BBB, implicating altered amino acid transport and glutamine accumulation—rather than BBB disruption—as a driver of glutamatergic imbalance and ASD-like behaviors. Interestingly, this work highlights that BBB dysregulation is not a universal requirement for the manifestation of ASD-like phenotypes. Together, these findings suggest a complex and potentially bidirectional interplay between gut dysbiosis, intestinal and BBB permeability, and the neurobiology of neurodevelopmental disorders. However, these pathways should be viewed as complementary mechanisms; it remains difficult to determine if BBB failure is a direct result of gut dysregulation or a downstream effect of the systemic inflammatory response.

Beyond microbiota-related mechanisms, several studies propose alternative, non-mutually exclusive pathways linking BBB dysfunction to SCZ and ASD including: (1) Shank3-dependent regulation of TJ proteins in brain endothelial cells [71]; (2) Sema3F expression in interneurons maintaining BBB integrity [23]; (3) Maternal Pten genotype affecting offspring BBB function [95]; (4) 22q11.2 deletion disrupting claudin-5 expression and BBB integrity [38, 121]; (5) MIA-induced neurovascular disruption through impaired pericyte-endothelial signaling, reduced claudin-5, and microhemorrhages [127]. All together, these results underscore the complexity of the mechanisms that may be involved in neurodevelopmental disruption of BBB function. Considerable efforts are still required to elucidate whether these mechanisms act as primary causes of BBB dysfunction or if BBB impairment is a downstream consequence of the neural dysfunction induced by these pathways.

Interestingly, Kim et al. [115] reported that acute NMDA exposure increased repetitive behavior without compromising BBB integrity, suggesting that transient NMDA elevations alone may not induce sustained BBB breakdown. This aligns with the complex, dose- and time-dependent role of NMDA receptors in ASD and SCZ pathophysiology. Both hypo- and hyperfunction of NMDA receptors have been implicated in the pathophysiology of ASD [149]. NMDA receptor antagonists have been proposed as therapeutic strategies in preclinical models of ASD [150], demonstrating efficacy in counteracting BBB disruption, oxidative stress, neuroinflammation and ASD-like behavioral abnormalities in the VPA model [110]. However, it remains unclear whether therapeutic benefits arise from direct vascular stabilization or are secondary to the reduction of excitotoxic and neuroinflammatory signaling.

The reviewed studies assessed BBB alterations across various developmental stages—fetal [95, 126, 127], perinatal [71], juvenile [99, 102, 129], adolescent [106–113, 127], and adult [23, 38, 71, 88, 115, 118, 121, 126, 127]—highlighting how the timing of BBB disruption may influence long-term outcomes. Most genetic and environmental insults caused persistent BBB dysfunction and lasting behavioral abnormalities. Notably, Kim et al. [71] reported BBB recovery in adult Shank3-KO mice, despite sustained behavioral deficits, suggesting that transient early-life BBB disruption may suffice to drive enduring neurodevelopmental changes. These findings suggests that once neural circuits are miswired during critical windows of development, restoring BBB integrity later in life may not be sufficient to reverse the behavioral phenotype. Similarly, O’Neill et al. [99] observed that neonatal—but not adult—B. pertussis infection induced neuroinflammation, BBB dysfunction, and ASD-like behaviors, underscoring the critical impact of early-life challenges on neuropathological outcomes.

Most studies focused on BBB dysfunction and therapeutic strategies to restore it [99, 102, 106–111, 113, 127, 129] during juvenile and adolescent periods, which are critical windows for social and cognitive maturation and GABAergic circuit refinement [151–153]. This window is also marked by high plasticity and susceptibility to environmental insults [154–156]. GABAergic dysfunction during adolescence and BBB disruption may mutually exacerbate one other, and interventions that support GABAergic maturation during this developmental window and ameliorate BBB dysfunction may represent promising strategies for mitigating symptomatology and mechanisms associated with SCZ and ASD. Notably, Sema3F—expressed in GABAergic interneurons—was implicated in both BBB integrity and neural synchronization in ASD models [23], underscoring the need to explore BBB-GABAergic interactions as therapeutic targets.

In addition to pharmacological interventions, two studies explored novel therapies targeting the BBB. Kim et al. [71] showed that restoring β-catenin signaling in neonatal endothelial cells rescued TJ expression, BBB function, neuronal dysfunction, and social behaviors in Shank3-KO mice. These findings suggest that early intervention aimed at BBB stabilization may counteract key pathogenic mechanisms underlying ASD-related neuropathology. Greene et al. [38] demonstrated that antipsychotics increased claudin-5 expression dose-dependently, suggesting a therapeutic potential for TJ modulation in SCZ. Importantly, although these studies offer compelling mechanistic hypotheses and the possibility of exploring new therapeutic targets, they remain far from clinical application in the near future due to the lack of direct validation in human populations.

The findings from these preclinical models demonstrate certain parallels with human pathology, as BBB integrity is similarly compromised in individuals with SCZ and ASD [23–26]. Specifically, the TJ disruption observed in rodent models mirrors hallmarks identified in postmortem brain tissue [28] and aligns with elevated peripheral markers, such as S-100β, found in patients [29–31]. Notably, alterations in claudin-5, ZO-1, and occludin—key proteins discussed in this review—have been detected in patient serum [25, 38–40]. Furthermore, the modulation of claudin-5 by antipsychotics [38] underscores the value of these models for exploring the BBB as a shared pathophysiological mechanism. Nevertheless, these markers primarily reflect an existing association and do not definitively establish the BBB as the initiating etiological factor. Furthermore, the translational relevance of these findings must be framed with caution.

The role of BBB dysfunction as a potential transdiagnostic hallmark is further supported by neurodegenerative research. In Alzheimer’s disease (AD), BBB disruption triggers inflammatory and immune responses that initiate neurodegenerative pathways [157]. Conversely, BBB breakdown often precedes cognitive decline [158] and may contribute to neuropsychiatric symptoms like agitation, depression, and psychosis [159]. These parallels suggest that BBB impairment may represent a common vulnerability point; however, it remains to be elucidated whether neurovascular dysfunction acts as a primary driver or a secondary consequence of the neurodegenerative process across diverse clinical conditions.

Concluding remarks, limitations and future perspectives

Our findings suggest that both genetic and environmental risk factors implicated in SCZ and ASD pathophysiology are frequently associated with alterations in BBB integrity in preclinical rodent models. The reviewed studies consistently demonstrate alterations in BBB permeability and TJ composition in preclinical rodent models, supporting the notion that BBB dysfunction may represent a transdiagnostic hallmark of neurodevelopmental disorders, albeit one arising from partially distinct molecular and developmental mechanisms in ASD and SCZ (see graphical abstract).

A significant limitation of this review is its exclusive reliance on animal models. While these studies provide essential mechanistic insights—and many genetic and environmental risk factors for SCZ and ASD are indeed shared across species—the translational value of the reported findings must be approached with caution. Given the inherent physiological and developmental differences between rodents and humans, these results require direct validation in human cohorts to confirm their clinical relevance.

Furthermore, a notable limitation of the current review is the paucity of replicated findings; several outcomes were based on single primary studies. Consequently, these results should be interpreted with caution until further independent replications are available.

Moreover, based on the studies reviewed, establishing a clear causal hierarchy remains challenging; it is yet to be determined whether BBB disruption is the initiating etiological factor or a downstream consequence of neuroinflammation and neural dysfunction. These two possibilities are not mutually exclusive and likely form a self-perpetuating pathological loop—a complex, bidirectional relationship that warrants further mechanistic investigation to be fully elucidated.

Given the limited number of studies investigating sex-specific BBB alterations, future research should prioritize the inclusion of both male and female subjects. This is crucial to account for known sex differences in clinical presentation and treatment response in SCZ and ASD, and may help improve diagnostic biomarkers and therapeutic strategies, particularly for females.

In addition, future studies should explore dual-hit paradigms that combine genetic susceptibilities with environmental challenges (e.g., exposure to psychotropics, immune insults, or social stress) at sensitive developmental windows—prenatal, early postnatal, and adolescence. These models can better reflect the complex etiology of neurodevelopmental disorders and help identify critical windows for early intervention. Investigating whether early therapeutic strategies can reverse BBB dysfunction is also vital, as early interventions may prove more effective in a neuroplastic brain.

A promising direction for future research involves omics-based analyses (e.g., proteomics, genomics) in preclinical models treated with either conventional antipsychotics or novel strategies targeting BBB integrity—such as β-catenin signaling modulators or claudin-5 enhancers. These approaches may uncover new neurobiological pathways involved in SCZ and ASD pathogenesis and foster the development of innovative treatments aimed at restoring BBB function.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Abbreviations

22qDS

22q11.2 deletion syndrome

ASD

Autism spectrum disorders

BBB

Blood–brain barrier

BECs

Brain endothelial cells

CHD8 or Chd8

Chromodomain helicase DNA-binding protein 8

CNTNAP2 or Cntnap2

Contactin-associated protein-like 2

CNVs

Copy number variants

FMT

Fecal microbiota transplantation

GBHs

Glyphosate-based herbicides

GD

Gestational day

GI

Gastrointestinal

HPC

Hippocampus

iNOS

Inducible nitric oxide synthase

JAMs

Junctional adhesion molecules

LPS

Lipopolysaccharide

MIA

Maternal immune activation

MMP-2

Matrix Metalloproteinase-2

mPFC

Medial prefrontal cortex

NMDA

N-methyl-D-aspartate

PFC

Prefrontal cortex

P-gp

P-glycoprotein

PND

Post-natal day

poly(I

C) = polyriboinosinic-polyribocytidylic

PPA

Propionic acid

PTEN

The phosphatase and tensin homolog

=Semaphorin 3F

SEMA3F or Sema3F

SCZ

Schizophrenia

TJs

Tight junctions

VPA

Valproic acid

Author contributions

CM-M and MF-G filtered, selected and reviewed the eligible full text under the supervision of CM-C. Disagreement on eligibility was resolved by discussions with BC-F. CM-M and MF-G realized the data extraction under the supervision of CM-C. AG-G, SG-C and AM-M carried out quality assessment. AG-G registered the systematic review in OSF. MC, HS, DSG, and MAD revised the final version of the manuscript. MS-R wrote the manuscript. All authors contributed to the first draft of the manuscript and have revised and approved the final manuscript.

Data availability

No datasets were generated or analyzed during the current study.

Declarations

Ethics approval and consent to participate

Not applicable

Competing interests

The authors declare no competing interests.