Brain and circulating EV proteome signatures in schizophrenia as prognostic markers for age-related dementia

Abstract

Schizophrenia (SZ) is epidemiologically linked to an increased risk of developing age-related dementias (ARD) predominantly characterized by Alzheimer’s disease and vascular dementia. However, the molecular mechanisms underlying this association remain insufficiently elucidated. Extracellular vesicles (EVs) play a critical role in neuropathological processes and offer a promising avenue for identifying shared disease mechanisms and potential circulating markers for patient stratification. Here we used a two-phase systems biology approach integrating discovery-driven proteomics with a targeted validation strategy using data-independent acquisition mass spectrometry (DIA-MS) in a large, independent SZ cohort. First, we analyzed brain-derived EVs (bEVs) from post-mortem SZ and ARD subjects to identify shared molecular signatures. Next, we validated the presence and circulation of these bEV markers in circulating plasma EVs (pEVs) using DIA-MS data. Remarkably, SZ and ARD bEV proteome and peptidome showed overlapping alterations in neuronal connectivity, synaptic integrity, neuroinflammation, and metabolism. Unsupervised clustering analysis of correlated bEV/pEV markers stratified SZ patients into two clusters: high dementia risk and control-like profiles. Collectively, these data emphasize the significance of bEVs as crucial mediators of shared neuropathogenic mechanisms in SZ, and ARD. Furthermore, we identified a set of pEVs markers, including proteins and specific peptides, with a robust and promising bench-to-bedside trajectory that may facilitate the stratification of SZ patients at risk for ARD.

Supplementary Information

The online version contains supplementary material available at 10.1186/s40478-026-02223-z.

Introduction

Dementia is typically the clinical manifestation of an underlying neurodegenerative disease accompanied by a significant decline in cognitive capacities, including memory, executive function, and reasoning [11]. This deterioration profoundly disrupts the ability of individuals to perform routine daily activities, maintain social relationships, and function independently [11].

Individuals diagnosed with psychotic disorders, particularly schizophrenia (SZ), exhibit a significantly elevated risk of developing age-related dementias (ARD) compared to the general population [2]. Epidemiological studies suggest that the prevalence of ARD among individuals with SZ is approximately two to three times greater than in those without the disorder [15], highlighting a potential critical overlap in neuropathological mechanisms between these brain disorders. This elevated risk is attributed to multiple factors, including cognitive impairments associated with psychotic disorders, lifestyle-related variables, and the potential neurodegenerative effects of long-term antipsychotic medication use [41]. Consequently, the factors contributing to the elevated risk of ARD in individuals with psychotic disorders raise important questions regarding the underlying molecular pathways involved [2]. While such factors may increase susceptibility to ARD, it remains unclear whether the underlying molecular mechanisms are shared with those of the most common forms of ARD, such as Alzheimer’s disease (AD) or vascular dementia (VaD), or represent distinct pathways leading to specific dementia subtypes. Furthermore, the symptomatic overlap between psychosis and dementia presents challenges for accurate diagnosis and treatment, emphasizing the necessity for comprehensive assessments and tailored interventions for this clinically vulnerable population [53].

Several hypotheses, including shared genetic risk factors, chronic inflammation, oxidative stress, and neurotransmitter imbalances, have been proposed and are currently under investigation [12, 57]. However, these theories lack robust empirical support, and the complex interplay between psychosis and age-related cognitive decline remains an area in need of active and comprehensive research. Furthermore, the research on these previously hypothesized shared neuropathological factors has yielded limited results thus far, suggesting that subtle molecular mechanisms may contribute at the neuropathological level [2], especially in early stages, to the shared clinical and epidemiological manifestations.

Extracellular vesicles (EVs) are nanoscale lipid-bilayer particles that play critical roles in facilitating intercellular communication throughout the body, including the central nervous system (CNS) [10, 27]. They offer important insights into subtle and intricate molecular mechanisms underlying pathological states, particularly as they can be detected in CNS tissues and possess a notable capacity to circulate in body fluids [17, 19, 21]. In the context of ARD,brain-derived and circulating EVs have been extensively characterized through systems biology approaches by our research group [22, 23] and other colleagues (see [54] for a comprehensive literature review). These studies have provided substantial evidence of the capacity of EVs to contribute to neuropathological and neuroprotective mechanisms at various stages of ARD and have demonstrated that they undergo dynamic changes in their cargo and function throughout the course of disease. These findings have thus highlighted their potential as both biomarkers and therapeutic targets in human ARD.

Despite the extensive research on EVs in neurodegenerative diseases causing ARD, the exploration of their role in mental illnesses remains limited [32]. This research gap is particularly striking given the potential insights EVs could offer into the underlying mechanisms of these disorders [32]. Our previous recent work has made significant strides in addressing this gap, identifying the involvement of brain-derived EVs in the altered molecular connectome associated with SZ, and indicate that EVs may play a significant role in the complex molecular networks underlying mental health disorders, much like their established roles in neurodegenerative diseases [38]. Further research in this area is thus required to deepen our understanding and uncover new therapeutic opportunities for the clinically vulnerable population affected by psychosis and SZ.

In this study, we characterize the proteome of brain-derived EVs (bEVs) isolated from post-mortem brain tissue specimens across independent case-control cohorts of individuals with antemortem diagnoses of SZ or ARD to identify shared molecular mechanisms potentially mediated by EVs. Our analyses revealed significant overlaps in the altered EV protein profiles between both conditions, suggesting common pathogenic pathways. Key proteins and molecular functions were consistently altered in EVs from both ARD and SZ, offering new insights into the molecular basis of these disorders. Moreover, we investigated the circulating capacity of these molecules associated with ARD in plasma EVs (pEVs) from a large, independent cohort of individuals with SZ. This analysis enabled evaluation of EV-associated proteins and peptidomesas a panel of candidate clinical biomarkers for identifying and stratifying individuals affected by certain psychotic disorders who are at elevated risk of developing ARD.

Materials and methods

Reagents

Reagents were sourced from Sigma-Aldrich (St. Louis, MO, USA), unless noted otherwise. High-performance liquid chromatography (HPLC) grade water and acetonitrile (ACN) were purchased from Thermo Fisher Scientific (USA). Sequencing-grade modified trypsin was obtained from Promega (Madison, WI, USA).

Post-Mortem brain tissue samples

Post-mortem brain tissue samples from individuals with SZ and age-matched controls, all with a post-mortem delay (PMD) below 24 h, were obtained at autopsy in the Basque Institute of Legal Medicine in Bilbao, Spain. Samples from SZ subjects and age-matched controls (n=30) were collected from the Brodmann area 9 (BA9)(dorsolateral prefrontal cortex). Samples were immediately stored at -80 °C. Medical records were reviewed retrospectively to identify individuals with SZ diagnoses. All diagnoses were madeantemortem by certified psychiatrists of the Basque Healthcare System (Osakidetza) following Diagnostic and Statistical Manual of Mental Disorders, Fourth and Fifth Editions (DSM-IV and DSM-5), or the International Classification of Diseases, Tenth Revision (ICD-10) criteria. All individuals diagnosed with SZ were undergoing antipsychotic treatment either prior to or at the time of sample collection. Subjects with co-occurring psychiatric or neurological disorders, including substance abuse, were excluded from the study. Age-matched control subjects had no history of psychiatric or neurological conditions. Detailed demographic information for both groups is provided in Table 1. There were no significant differences in the age or PMD of SZ and age-matched controls (Table 1).

Table 1.

Characteristics of the post-mortem brain tissues analyzed. PMD refers to post-mortem delay, expressed in hours. Age and PMD are presented as mean ± standard deviation for each group. Statistical significance was evaluated using the Student’s t-test (p < 0.05). * indicates samples were stored for less than one year. ^# indicates that subjects in this group were under antipsychotic medication

	SZ Cohort				ARD Cohort
	Control* n = 15	SZ*# n = 15	p-value		Control* n = 3	ARD* n = 9	p-value
Age	52.47 ± 6.346	52.80 ± 6.581	0.8887		74.33 ± 3.215	76.11 ± 9.466	0.7626
PMD	15.8 ± 6.361	14.33 ± 5.851	0.5164		22.27 ± 3.421	18.86 ± 7.604	0.4798

Brain tissue samples from ARD donors and age-matched controls were obtained from the Harvard Brain Tissue Resource Center in Boston, USA, and the Newcastle Biobank in the UK. Post-mortem brain tissues of 12 ARD subjects and age-matched controls were collected from the Brodmann area 21 (BA21), with a mean PMD of 22.27 ± 3.4 h in age-matched controls and 18.8 ± 7.6 h in ARD group (Table 1). ARD donors had an average age of 76.11 ± 9.4 years and exhibited early to advanced histopathological features of AD or VaD at the time of death. All ARD cases presented amyloid plaques and neurofibrillary tangles. Half of the AD donors were classified as Braak stage III (asymptomatic AD), representing asymptomatic AD without cognitive symptoms, while the other half were at Braak stage IV (intermediate AD), associated with clinical ARD symptoms. VaD samples showed the presence of some AD pathology although with significant cerebrovascular damage, including small vessel disease predominantly in the temporo-parietal region and evidence of lacunar or larger infarcts [30]. Age-matched controls refer to individuals without ARD-related neuropathology, selected to closely match the ARD donors in age. Detailed donor information, such as age, gender, and post-mortem delay, is provided in Supplemental Table 1. There were no significant differences in PMD and age between age-matched controls and ARD group (Supplemental Table 1).

Brain tissues were collected during autopsy and preserved in liquid nitrogen at –150 °C until further analyses. The selected brain regions were then sectioned into smaller pieces, with larger blood vessels carefully removed. The tissue was washed three times with 1X PBS for 30 min each time. Sample quality was evaluated by assessing the integrity of the isolated bEV proteomes. Proteome integrity of bEVs preparations was assessed by determining the proportion of fully tryptic versus non-tryptic peptides and analyzing the length distribution of non-tryptic peptides. A low abundance of semi-tryptic and short peptides indicates minimal protein degradation, thereby confirming the suitability of the samples for subsequent proteomic analysis (see [9] for an in-depth review). In our study, as illustrated in Supplementary Fig. 1, shown less than 2% of the total peptidomes profiled were non-tryptic. Furthermore, no peptides shorter than seven amino acids were identified in any of the analyzed post-mortem brain tissue cohorts. Informed consent was obtained from all participants or their legally authorized representatives. The use of post-mortem brain tissue adhered to the ethical principles of the Declaration of Helsinki, and all experimental protocols were conducted in strict compliance with the relevant institutional guidelines and the ethical committee of the University Hospital Arnau de Vilanova, Lleida, Spain.

Validation cohort of blood samples

Proteomics data from pEVs obtained from 283 subjects were previously generated by [59] and were used in this study to validate the circulating potential of the markers for ARD risk in SZ. This validation cohort comprised blood samples from 134 individuals with SZ and 149 age-matched healthy controls. Briefly, blood plasma samples (at least 10 mL) were collected from individuals diagnosed with SZ, as well as from matched healthy controls, at the Shanghai Mental Health Center (SMHC). SZ diagnoses were confirmed using the Structured Clinical Interview based on the DSM-IV. All patients with SZ were undergoing antipsychotic treatment at the time of sampling. Informed consent was obtained from all participants or their legal representatives. Further details on the validation blood plasma cohort are provided in an earlier publication [59]. The use of clinical blood plasma samples followed the ethical standards set by the Declaration of Helsinki, and all experimental procedures were carried out in full accordance with the applicable institutional guidelines.

Processing of brain tissues prior to EV isolation

Approximately 80 mg of post-mortem brain tissues from each subject were processed using a detergent-free homogenization buffer composed of 100 mM ammonium acetate (AA) and protease inhibitor cocktail tablets, following previously established protocols [18, 21]. Tissue homogenization was performed using a Bullet Blender homogenizer (Next Advance, NY, USA) with metallic beads (0.9 to 2.0 mm in diameter), which were pre-washed three times with 1X PBS for 30 min. The tissue samples were combined with the beads at a 1:1 weight-to-weight ratio. All steps of the EV isolation process were conducted at 4 °C. The tissue was homogenized in four cycles, each with 300 µL of homogenization buffer and a duration of 5 min. The first two cycles were conducted at medium intensity, followed by two cycles at maximum intensity. After each homogenization cycle, samples were centrifuged at 15,000 × g for 10 min, and the supernatants were pooled.

Brain and blood plasma EV isolation

bEV were enriched from the detergent-free brain homogenates using the Protein Organic Solvent Precipitation (PROSPR-Brain) method [21, 42]. In brief, the homogenates were mixed with four volumes of pre-chilled acetone (– 20 °C), then vortexed and centrifuged at 5000 × g for 1 min. The resulting supernatants, which contained the EV, were concentrated to near dryness using a vacuum concentrator (Concentrator Plus, Eppendorf AG, Hamburg, Germany).

pEV were isolated by sequential ultracentrifugation as previously detailed by [36]. Further details on the bEV isolation protocols followed can be obtained from [59].

Label-free digestion of EV proteomes

bEV were lysed in 16 M urea and 100 mM ammonium bicarbonate buffer (ABB), followed by dilution with HPLC-grade water. EV proteins were reduced with 20 mM dithiothreitol (DTT) at 30 °C for 3 h, then alkylated with 55 mM iodoacetamide (IAA) at room temperature (RT) in the dark for 1 h. The samples were diluted to reduce urea concentration below 1 M with 50 mM ABB. Trypsin digestion was performed overnight at 37 °C, with a 1:20 enzyme-to-protein ratio. Proteolysis was halted by adding 0.5% formic acid (FA). The digested peptides were desalted using C18 Sep-Pak cartridges (Waters, Milford, MA), eluted with 75% ACN and 0.1% FA, dried, and reconstituted in 200 µL of 0.02% ammonium hydroxide in water for subsequent HPLC fractionation.

pEV preparations were lysed using sonication and RIPA lysis buffer as previously detailed [59]. Briefly, pEV proteins were then processed using the Filter-Aided Sample Preparation (FASP) method, which involves denaturation with detergents, followed by ultrafiltration to remove contaminants. After multiple wash steps, proteins were enzymatically digested into peptides directly on the filter using trypsin as detailed above.

EV proteomes fractionation by high-pressure liquid chromatography

bEV digested proteomes from ARD subjects and age-matched controls were subjected to high-pressure liquid chromatography (HPLC) fractionation using a C18 column as previously detailed [20]. A 60-minute gradient was applied, and fractions were collected every minute, followed by concatenation and drying in a vacuum concentrator.

Ultrastructural analysis of brain EV by transmission electron microscopy

Representative EV fractions were characterized by transmission electron microscopy (TEM). Briefly, EV were placed onto Cu-Formvar-carbon grids and incubated at RT for 20 min. The grids were then rinsed with HPLC-grade water, fixed with 1% glutaraldehyde in PBS for 5 min, and stained with uranyl oxalate for 5 min. After staining, the samples were embedded in methyl-cellulose-uranyl-oxalate and air-dried for permanent preservation. Electron micrographs were captured using a JEOL JEM-1010 electron microscope operating at 80 kV. Images were calibrated, contrast-adjusted, and analyzed with ImageJ software (NIH, Bethesda, MD, USA).

Analysis of EV preparations by next-generation proteomics

Label-free quantitative analysis of bEV proteomes from ARD patients and age-matched controls was conducted as previously detailed [50] using liquid chromatography-tandem mass spectrometry (LC-MS/MS). Briefly, peptide separation was carried out on a Dionex UltiMate 3000 UHPLC system, which was coupled to an Orbitrap Elite mass spectrometer (Thermo Fisher, Bremen, Germany). A 60-minute gradient was employed for optimal peptide resolution, and data acquisition was performed in positive ion mode using Xcalibur software. This approach allowed for comprehensive proteomic profiling of brain EV samples.

bEV peptide samples from SZ individuals and paired controls were analyzed as previously described [37, 43] using an EVOSEP liquid chromatographic system (EVOSEP, Odense, Denmark) at a flow rate of 300 nL/min connected to a timsTOF Pro mass spectrometer (Bruker Daltonics, Billerica, MA, USA), and data were acquired using four-dimensional parallel accumulation–serial fragmentation (PASEF) technology.

pEV proteomes were analyzed [59] using a data-independent acquisition (DIA) approach with a Synapt G2-Si quadrupole time-of-flight mass spectrometer equipped with an ion mobility option (Waters Corporation, Milford, MA, USA).

Bioinformatics and data analysis

Proteomic raw data were processed using PEAKS Studio X Pro software (version 10.6, Bioinformatics Solutions). For protein identification, a false discovery rate (FDR) of less than 1% was applied across all samples, with trypsin specified as the enzyme for protein digestion. Carbamidomethylation of cysteine residues was set as a fixed modification. The search parameters included a precursor mass tolerance of 10 ppm and a fragment ion tolerance of 0.05 Da. The data were analyzed against a custom database compiled from all available protein sequences within the human UniProt database (downloaded on February 3, 2023, containing 140,065 protein sequences). Cleavage on at least one trypsin-specific end was required.

Label-free relative quantification was performed using spectral counts, as previously reported [37, 38]. Differential expression analysis was conducted in R (Version 4.4.0) and EdgeR (Version 4.2.1) was used to model spectral count data. Planned pairwise comparisons between SZ and ARD were adjusted for multiple testing using the Benjamini–Hochberg false discovery rate (FDR) procedure (q < 0.05 unless stated otherwise). For Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis, the Bioconductor org.Hs.eg.db package (Version 3.16.0) and the clusterProfiler package (Version 4.6.2) were used.

When non-parametric analysis was required, Kruskal–Wallis test was applied, with statistical significance set at p < 0.05 unless otherwise specified. Additionally, data were independently analyzed using parametric one-way ANOVA followed by Tukey’s test for multiple comparisons, with statistical significance determined at a corrected p < 0.05 unless stated otherwise. The sensitivity of the correlation analyses was evaluated using the pwr package (v1.3-0) in R, employing two-sided tests with a significance level of α = 0.05. Large effect sizes (|r| ≥ 0.8) were analyzed across the observed sample sizes of the ARD subset to estimate the achieved power. Unless otherwise stated, proteins were prioritized as shared candidates if they showed no statistically significant difference in abundance between the SZ and ARD groups (p > 0.05) and exhibited strong concordance across groups, defined as an absolute correlation coefficient |r| ≥ 0.80. For pathway analyses, pathways relevant to neuro degeneration were prioritized based on the number of dysregulated proteins identified. Proteins present in multiple pathways were prioritized for further analysis, while those most prominently represented across ARD groups were selected for comparative evaluation. Hierarchical clustering was performed with the k-means algorithm approach, and within each k-means cluster, rows were ordered using hierarchical clustering based on Euclidean distance and Ward’s linkage method, as implemented in the ComplexHeatmap package (version 2.20.0). To explore unsupervised similarity and potential stratification among individuals based on EV proteomic profiles, and to quantify the independent and interactive effects of diagnosis and neuropathology burden on protein levels, the distance matrix was computed to measure the dissimilarity between individuals, and clustering was conducted iteratively to form a hierarchical tree structure. In addition, we fit linear models in R to assess diagnosis, pathology burden, and their interaction, with disease_status treated as a three-level factor (Ctrl, AD, SCZ): protein_level ~ disease_status × pathology_burden.

Results

Ultrastructural and molecular features of bEV in SZ and ARD

We first performed comparative analyses to examine potential differences in the ultrastructural and distribution profiles of bEV isolated from post-mortem brain samples of individuals with SZ and ARD. Ultrastructural characterization revealed no distinctive features specific to any disease group. The analyzed bEV exhibited a spherical morphology and common structural characteristics consistent with prior descriptions of bEV (Fig. 1a). Importantly, no evidence of contamination was detected in the ultrastructural analysis, underscoring the high quality of the bEV preparations (Fig. 1a). Additionally, no significant differences were obtained in the size distribution of bEV from the disease groups based on TEM micrograph analysis (Fig. 1b-c).

Fig. 1.

Morphological, ultrastructural, and compositional characterization of brain extracellular vesicles (bEVs) from healthy controls (C), individuals with schizophrenia (SZ) and individuals with age-related dementias (ARD). a Representative transmission electron microscopy (TEM) micrographs of bEVs. Scale bar represents 200 nm. b Size distribution of the bEVs measured from TEM micrographs. c Relative frequency distribution of bEVs size based on TEM measurements. d Number of proteins identified in bEVs among the top 100 most common EV proteins listed in ExoCarta and Vesiclepedia for each group. e Total number of proteins identified across all groups, highlighting overlap with the top 100 EV proteins in ExoCarta and Vesiclepedia. f Proportion of the top 25 most common EV proteins that showed significantly different levels in the ARD group compared to the SZ group, versus those without significant differences. Statistical significance was determined with p < 0.05 considered significant. Multiple testing were performed by Tukey’s test. Error bars represent standard deviation

To further evaluate the molecular features and quality of the bEV preparations, we investigated the presence of common molecular markers typically associated with EVs. For this analysis, the bEV markers identified in our data were compared to the top 100 markers curated in the specialized EV databases, ExoCarta and Vesiclepedia. While age- and disease-matched controls exhibited the highest number of matches, no statistically significant differences were observed in the number of markers present in the bEVs of any of the analyzed groups (Fig. 1d-e). However, when comparing the levels of these common EV markers in bEV between SZ and ARD, approximately 40% of the proteins showed differential abundance (Fig. 1f). Further analysis of the differentially regulated vesicle-associated proteins revealed that HSPA8, GAPDH, ENO1, CLTC, and ALDOA were significantly downregulated in bEV from SZ individuals compared to those with ARD (Table 2). Notably, no significant differences were observed in the levels of the remaining common bEV markers, including FLOT1, ANXA2, SDCBP, HSP90AA1, EEF1A1, YWHAZ, EEF2, YWHAE, and ANXA1, which are frequently identified among the top 25 EV markers curated in the specialized ExoCarta and Vesiclepedia repositories (Table 2).

Table 2.

Top 25 most common EVs markers in exocarta and vesiclepedia found in brain extracellular vesicles (bEV) of individuals with age-related dementias (ARD) and individuals with schizophrenia (SZ). Information regarding gene symbol (GS), protein description, protein abundance expressed as spectral count (mean ± standard deviation) and the database (V. vesiclepedia, E. exocarta, B. Both) are included for each 25 most common EVs markers. Significance was assessed by two-way ANOVA, with a minimum significance level p < 0.05

GS	Description	Age-related Dementia	Psychotic Disorder	Database	p-value
ANXA2	Annexin A2	5.52 ± 5.45	6 ± 10.39	B	0.9923
SDCBP	Syntenin-1	0.04 ± 0.11	11 ± 12.12	B	0.9623
HSP90AA1	Heat shock protein HSP 90-alpha	1.93 ± 5.78	16.33 ± 11.68	B	0.9362
FLOT1	Flotillin-1	6.85 ± 5.07	1.67 ± 1.53	V	0.9914
EEF1A1	Elongation factor 1-alpha 1	8.67 ± 4.30	20 ± 7.81	B	0.9598
YWHAZ	14-3-3 protein zeta/delta	117.07 ± 44.47	27 ± 24.33	B	0.1426
EEF2	Elongation factor 2	35.41 ± 32.93	3.33 ± 4.16	E	0.7341
YWHAE	14-3-3 protein epsilon	6.11 ± 18.33	8.33 ± 4.51	B	0.9984
ANXA1	Annexin A1	5.44 ± 3.58	43.67 ± 68.77	V	0.6518
HSPA8	Heat shock cognate 71 kDa protein	375.81 ± 285.76	50.33 ± 45.37	B	< 0.0001
GAPDH	Glyceraldehyde-3-phosphate dehydrogenase	290.19 ± 267.63	29.33 ± 17.21	B	< 0.0001
ENO1	Alpha-enolase	296.04 ± 75.59	53.67 ± 67.93	B	< 0.0001
ALDOA	Fructose-bisphosphate aldolase A	157.56 ± 107.75	33 ± 19.16	E	0.0321
HSP90AB1	Heat shock protein HSP 90-beta	154.30 ± 85.97	12.33 ± 11.59	V	0.0044
CLTC	Clathrin heavy chain 1	172.74 ± 80.21	4.67 ± 4.16	E	0.0028

Converging proteome signatures in bEV of SZ and ARD

Subsequently, to investigate potential molecular neuropathological overlaps in bEV from individuals with ARD and SZ, we conducted a proteome-wide analysis to identify proteins that exhibit consistent dysregulation across these conditions. Notably, we identified 32 bEV-associated proteins that demonstrated significant differential abundance in comparison to their respective age-matched controls, within each clinical group, and did not display significant differences when comparing SZ across ARD phases (Fig. 2a, Supplemental Table 2). Among the 32 proteins that exhibited differences from the control group but maintained similar levels between SZ and each ARD group, 23 showed a significant main effect of diagnosis, with no significant interaction observed. The remaining 9 proteins were associated with pathology burden and/or demonstrated a significant interaction between disease status and pathology burden. These findings, derived from the application of linear models, are detailed in Supplemental Table 2. Moreover, our analysis revealed multiple specific proteins consistently present in bEV of individuals with SZ across the various ARD groups examined (Fig. 2a and Supplementary Tables 3–5). These findings elucidate the existence of a shared molecular signature in temporal lobe bEVs that may be fundamental to the progressive neuropathology in both analyzed neuropsychiatric and neurodegenerative conditions.

Fig. 2.

Converging proteome signatures and shared molecular functions in brain-derived EVs (bEVs) from individuals with schizophrenia (SZ) and age-related dementias (ARD). a Number of bEV proteins exhibiting no significant differences in abundance between SZ and each ARD group, but significantly altered relative to controls. The set size indicates the total number of proteins within each comparison group. b Heatmap illustrating differential abundance of bEV proteins identified as the converging proteome between SZ and ARD, with significant differences compared to controls. Data are presented as log₂ fold change (log₂FC) for each ARD group relative (asymptomatic Alzheimer’s disease (Asympt AD), intermediate AD (Interm AD), or vascular dementia (VaD)) to SZ. Red indicates higher protein abundance in ARD; blue indicates higher abundance in SZ. c Pearson correlation analysis between converging bEV proteins (those shared by SZ and ARD) and the top 20 canonical EV proteins (based on ExoCarta and Vesiclepedia) that showed no significant differences between SZ and ARD. The color scale represents Pearson’s correlation coefficient. d Enriched molecular pathways associated with the converging bEV proteomes shared across SZ and ARD, based on KEGG analysis. Dots represent the number of proteins involved in each pathway with proportional diameter e Subgroup-specific pathway enrichment of converging bEV proteins exclusive to SZ and Asympt AD, Interm AD, or VaD. Dots represent the number of proteins involved in each pathway with proportional diameter. Statistical significance was determined with p < 0.05 considered significant, and significance is codified with asterisks representing * p ≤ 0.05; ** p ≤ 0.01; and *** p ≤ 0.001

Further analysis revealed that several proteins implicated in synaptic integrity and plasticity were consistently altered, including VPS26B, VGF, MAP6, MAP4, and MAP1B, substantiating the hypothesis of common synaptic dysfunction as a shared characteristic in bEVs of SZ and in all the analyzed ARD conditions (Fig. 2b). Proteins involved in neuroinflammation and immune response, such as JCHAIN, IGHM, IGHG3, and C4A, were also commonly dysregulated (Fig. 2b), indicating a converging role of specific immune activation in both pathologies. Additionally, we identified alterations in proteins related to cytoskeletal organization (TUBB6, SPTBN2, DTNA), myelination (CNP), and ion transport (ATP2B2, ATP2B1), further corroborating a shared disruption of neuronal integrity and signaling (Fig. 2b).

Converging proteome signatures associate with specific bEV markers

We subsequently investigated whether the identified converging proteomic signature in bEV from SZ and ARD exhibited a significant association with specific EV markers. As illustrated in Fig. 2c, our analysis demonstrated that several immune-related proteins, including C4A, HPX, JCHAIN, IGHM, and IGHG3, which were consistently dysregulated in both pathologies, displayed a strong and highly significant positive correlation with the EV-associated protein ANXA2 in bEVs. Conversely, most of these immune-related proteins showed a negative association with the classical EV marker FLOT1, although this correlation was not statistically significant.

Of particular note, proteins demonstrating a statistically significant and positive correlation with FLOT1 included the cell structural and synaptic regulation-associated microtubule-associated proteins MAP4 and MAP6 (Fig. 2c). Furthermore, several other proteins identified as convergently dysregulated in SZ and ARD were also significantly associated with specific EV markers, curated as top-ranked EV-associated proteins in the Vesiclepedia and ExoCarta databases as detailed in Fig. 2c.

Proteomic convergences in bEV reflect intersecting pathophysiology

To further investigate the mechanistic impact of the identified converging proteomic signatures on CNS tissues, we conducted comprehensive pathway analyses to elucidate potential biological interactions and functional networks influenced by these dysregulated proteins. These analyses aimed to delineate the molecular pathways through which the shared proteomic alterations may contribute to pathophysiological processes in both pathologies. Remarkably, commonly dysregulated proteins in bEVs of SZ and ARD, such as ATP2B1/2, ANK2, and CTNND2, were implicated in calcium signaling, synaptic vesicle cycle, and axon guidance, highlighting disruptions in neuronal connectivity and synaptic integrity (Fig. 2d). Overlapping pathways also encompassed neuroinflammation, metabolic dysfunction (e.g., cAMP/cGMP signaling and TCA cycle), and cytoskeletal anomalies, suggesting a common framework underlying these disorders (Fig. 2d).

Additionally, in asymptomatic AD, dysregulated proteins common to SZ were associated with synaptic dysfunction, prion disease, motor proteins, and cholesterol metabolism, indicating early neuronal connectivity and lipid disruptions (Fig. 2e). In clinical ARD, altered pathways overlapping with SZ encompassed those associated with AD, glutamate metabolism, and neurodegeneration, reflecting progressive neuronal and metabolic decline (Fig. 2e). VaD-associated bEV, in common with SZ, exhibited enrichment in MAPK signaling, glycolysis, the TCA cycle, and mitochondrial function, thus reinforcing the role of metabolic dysfunction indicated by bEV in cognitive deterioration in both pathologies (Fig. 2e).

Association of converging bEV proteomes with autophagy and signaling

Building on our previous observation on the association between bEVs and autophagic fluxes in ARD [13], we aimed to ascertain whether this association extends to SZ and how it relates to the converging proteomes identified in bEV in this study. Our analysis revealed a distinct trend: bEV containing proteins common to SZ and ARD predominantly exhibited association with autophagic markers across SZ, asymptomatic AD, and intermediate AD (Fig. 3a). Conversely, bEV enriched with signaling molecules and the convergent proteomes were more prevalent in control groups, as shown in Fig. 3a.

Fig. 3.

Molecular signaling shifts associated with converging bEV proteomes in schizophrenia (SZ) and age-related dementias (ARD). a Directional trend analysis of significant positive correlations between converging bEV proteins and EV markers of either autophagy or canonical signaling pathways. Arrows point toward autophagy (left) or signaling (right), reflecting predominant association within each group. b Pearson correlation matrix of converging bEV proteins and autophagy-associated EV markers in SZ, ARD subtypes (asymptomatic Alzheimer’s disease (Asympt AD), intermediate AD (Interm AD), or vascular dementia (VaD)), and control groups. c Pearson correlation matrix of converging bEV proteins and signaling-associated EV markers across the same groups. The color scale represents Pearson’s correlation coefficient (r), with red indicating strong positive and blue indicating strong negative associations. Statistical significance determined by Pearson correlation analysis. Significance is indicated with asterisks representing * p ≤ 0.05; ** p ≤ 0.01; and *** p ≤ 0.001

To further illustrate these associations, we generated scatter plots for the strongest correlations observed across Control, SZ, and clinical AD groups (Supplementary Fig. 2). A post hoc power analysis (R, pwr v1.3-0; two-sided α = 0.05) showed that correlations ≥|0.8| achieved power of 0.772 (for r up to 0.9) and 0.801 (for r > 0.9). Consistent with conventional thresholds (power ≈ 0.80), we therefore restricted inference to these large-effect associations, considering only results with robust statistical power. An in-depth analysis of specific autophagic markers and convergent bEV proteomes was conducted. In intermediate AD and SZ, most convergent proteins exhibited positive correlations. Notably, several of these exhibited statistically significant correlations (r > 0.8) with the autophagy marker sequestosome 1 (SQSTM1) detected in bEV (Fig. 3b). In contrast, these variables predominantly exhibited strong negative correlations (r < – 0.8) in the control group (Fig. 3b). Furthermore, convergent bEVs proteomes in asymptomatic AD, intermediate AD, VaD, and SZ demonstrated strong negative correlations with the EV signaling marker RHOA, whereas these convergent bEV proteomes exhibited strong positive correlations in control groups (Fig. 3c). While other analyzed markers exhibited strong correlations regarding the presence of convergent proteomes and signaling markers in bEV, no definitive trends could be discerned from these analyses. RHOA emerged as the sole marker that yielded clearly interpretable data, as shown in Fig. 3c. Scatter plots of the strongest correlations across the analyzed groups are included in Supplementary Fig. 2. For autophagy, correlations between DENR and SPTBN2 with the marker SQSTM1 are shown in Supplementary Fig. 2A-B. For signaling, correlations between LRRC47 and SPTBN2 with the marker RHOA are shown in Supplementary Fig. 2C-D.

Blood plasma circulation of converging bEV proteomes in SZ

Upon confirming the presence and characterizing the convergent proteomes associated with bEV in this study, we aimed to validate their circulatory capacity of these markers in the blood plasma using a large, independent cohort of SZ and control subjects. This methodological approach was designed to evaluate their potential clinical applicability and to identify prognostic markers that may aid in stratifying SZ individuals at increased risk of developing ARD. Our analysis confirmed that several proteins within the identified convergent bEV proteomes were associated with pEV. Notably, IGHM, IGHG3, C4A, HPX, and JCHAIN exhibited significantly elevated levels in pEV of SZ individuals compared to controls in the validation cohort (Fig. 4a-b). This observation was consistent with their corresponding levels in bEV, as illustrated in Fig. 4c-d. Furthermore, a comprehensive analysis of average protein levels revealed higher concentrations of these markers in bEV than in circulating EVs, and these levels were consistently higher in SZ subjects than controls, as anticipated (Fig. 4a-d).

Fig. 4.

Circulating validation and stratification potential of converging brain extracellular vesicles (bEV) proteins in schizophrenia (SZ). a Average levels of converging proteins—identified as commonly altered in bEV from SZ and ARD—in circulating plasma extracellular vesicles (pEV) of SZ patients and age-matched controls (C). Significance determined by Mann–Whitney test (p < 0.05). b Abundance of each converging protein detected in pEV of SZ patients versus controls. c Cumulative levels of converging protein present in bEV from SZ subjects and controls. d Abundance of each converging protein detected in bEV of SZ patients versus controls. Pearson correlation between protein levels in bEV and pEV for e C4A and f IGHM in SZ patients. Shaded areas indicate 95% confidence intervals. g Unsupervised clustering of SZ and control individuals based on log₂-transformed pEV levels of converging proteins. Three distinct clusters emerged: controls (yellow), SZ cluster 1 (light green, lower expression of convergent proteins associated to low ARD risk), and SZ cluster 2 (purple, higher expression of convergent proteins associated to higher ARD risk). h Relative expression levels in bEV of each converging proteins in SZ individuals across clusters defined in panel G, revealing distinct molecular profiles in high-ARD risk versus low-ARD risk SZ subgroups. i Peptide-level Person correlation analysis between bEV and pEV for the converging proteins. Significance is indicated with asterisks representing * p ≤ 0.05 and ** p ≤ 0.01. Multiple testing were performed by Tukey’s test

Subsequently, we examined the circulation capacity of these convergent bEV proteomes. Notably, C4A and IGHM exhibited highly robust and statistically significant associations between their concentrations in bEV and pEV, notwithstanding the fact that the EV proteomes analyzed belonged to independent cohorts (Fig. 4e, f).

Prognostic ability of Circulating converging bEV proteomes in SZ

Finally, we assessed the prognostic potential of the identified circulating convergent bEV proteomes in the blood plasma of individuals with SZ. To this end, we performed clustering analysis on log₂-transformed protein expression data, comparing SZ patients with controls. The analysis revealed a clear group separation: all controls clustered together, while SZ subjects segregated into two distinct subgroups (Fig. 4g and Supplementary Table 6). One SZ subgroup exhibited markedly elevated levels of the identified pEV proteins, whereas the other showed expression profiles similar to those of the control group (Fig. 4g and Supplementary Table 6). Notably, approximately 20% of SZ patients fell into the high-expression subgroup, overlapping with profiles observed in ARD patients. The remaining SZ individuals displayed pEV protein levels comparable to controls.

To further investigate the potential circulatory role of proteins in pEV that were not strongly correlated with their levels in bEV, we analyzed the association patterns of IGHG3, HPX, and JCHAIN with the highly representative proteins C4A and IGHM. These analyses, presented in Supplementary Fig. 3, revealed strong associations between the levels of these proteins in pEV, suggesting their coordinated expression and potential as circulating biomarkers.

Subsequently, to verify whether this pattern was also present within bEV of SZ subjects, we conducted a second clustering analysis based on the expression levels of the five identified proteins in pEV. The analysis revealed a clear separation of the SZ cohort into two distinct subgroups (Fig. 4h and Supplementary Table 7). These results reinforce the observation that both bEV and circulating EV exhibit differential expression patterns of these proteins, supporting their potential as markers for ARD-related risk stratification within the SZ population.

Identification of the Circulating ARD prognostic peptidome

Recent findings, particularly in the context of neurodegenerative disorders, indicate that concentrating on peptide signatures rather than entire proteins constitutes a more precise approach for the identification of prognostic and stratification biomarkers. In accord with this premise, the present study examined the circulation potential of the peptidome associated with the prognostic proteins identified herein. Table 3 shows that, specific peptides, including those with post-translational modifications, were identified for each protein and were exclusively detected in bEV and pEV from individuals within the high ARD-risk cluster among SZ patients. Moreover, the analysis of the circulating association of these peptides demonstrated a significant correlation between their levels in bEV and pEV (Fig. 4I), thereby supporting their potential as specific proteinaceous prognostic markers.

Table 3.

Classification of schizophrenia (SZ) subjects into clusters based on estimated age-related dementia (ARD) risk estimated from the determination of convergent proteins in plasma extracellular vesicles (pEV). Subjects were grouped using unsupervised clustering of log₂-transformed protein levels of converging proteins detected in pEV. Protein abundance patterns define three groups: healthy controls (C), low ARD-risk SZ cluster (SZ cluster 1), and high ARD-risk SZ cluster (SZ cluster 2). Values represent the range or representative log₂ expression level of each protein per group

	C	SZ cluster 1	SZ cluster 2
IGHM	0	1–2.32	5.13
IGHG3	0–1	0	4.39
C4A	0	0	3.58
HPX	0	0	2.32
JCHAIN	0	0	1

Discussion

This study employed a two-phase approach, first by investigating the role of bEV proteomes in mediating ARD neuropathology in SZ, and then by assessing their circulation in pEV and their prognostic potential for ARD risk in a large, independent clinically characterized cohort (Fig. 5). Our findings revealed that while many bEV markers were shared across pathologies, some vesicle-associated proteins showed disease-specific links. These results align with previous studies comparing bEV markers in neurological disorders, reinforcing the idea that these vesicles reflect both unique and overlapping disease mechanisms [40].

Fig. 5.

Schematic overview of the experimental design and analytical workflow. The diagram illustrates the two-phase experimental approach used in the study. Phase one involved the isolation and proteomic profiling of brain-derived extracellular vesicles (bEV) from post-mortem samples of schizophrenia (SZ) and age-related dementia (ARD) patients, including Alzheimer’s disease (AD) and vascular dementia (VaD). Phase two focused on validating candidate bEV protein markers in plasma-derived EV (pEV) from an independent, large cohort of SZ patients using data-independent acquisition mass spectrometry. The schematic outlines key steps including sample processing, EV isolation, protein extraction, mass spectrometry analysis, and statistical clustering used for biomarker identification and risk stratification

Building upon these findings, our data suggest that bEV may contribute to key pathophysiological mechanisms underlying SZ, AD, and even VaD, particularly those affecting neuronal connectivity, synaptic integrity, neuroinflammation, and metabolic dysfunction. These bEV-mediated neuropathological alterations were linked to the dysregulation of ATP2B1/2, ANK2, and CTNND2. Notably, impaired calcium homeostasis resulting from dysfunctional ATP2B1 and ATP2B2 has been previously identified in SZ through genomic studies [14, 28, 46]. In this study, we confirm the involvement of these ion pumps at the protein level in association with bEV. Given their established role in neurodegeneration, their dysregulation may serve as a relevant mediator of neuropathology in SZ, potentially contributing to neuroinflammation and disease progression [48, 52]. These findings support a mechanistic link between ion pump dysfunction and SZ-related neuropathology, warranting further research.

Consistent with its role in neurodegeneration, ANK2 dysregulation has also been linked to impaired voltage-gated calcium channel function in neurons [55], and genome-wide association studies have implicated ankyrins in SZ [60]. These findings, observed in both SZ and neurodegenerative disorders, support our observation of ANK2 dysregulation in bEV, reinforcing its potential role in disrupted calcium homeostasis in SZ, AD, and VaD. Similarly, our study identifies CTNND2 as a shared molecular mechanism between SZ and AD, consistent with genomic evidence [5, 16]. Furthermore, we demonstrate that CTNND2 dysregulation in bEVs is also implicated in the molecular basis of VaD, reinforcing its broader role in neuropathology across these disorders.

In addition to the molecular mechanisms shared between SZ, AD, and VaD identified in our study, we also observed a distinct neuroinflammation-related signature in bEV, specific to SZ and progressive stages of AD. This signature included the proteins C4A, HPX, JCHAIN, IGHM, and IGHG3. Although previous studies have implicated some of these molecules in AD and SZ, our work provides the first evidence that they form a distinct bEV-specific neuroinflammation signature in SZ and progressive AD, with prognostic implications via their circulation in plasma EV. While we recently identified EV as a potential mechanism for transporting immunoglobulins into the brain in SZ [38], the specific role of these antibody-related proteins in neuroinflammation within this disorder remains to be fully elucidated. Previous research has potentially linked the presence of some of these molecules in the brain to the manifestation of certain positive psychotic symptoms [38, 56]. Based on our findings, this hypothesis could also be extended to certain manifestations of cognitive decline in AD, although its validity remains unclear and would require further exploration.

These immune-related molecules are not only dysregulated in bEV of SZ, similar to AD, but also contribute to the prognostic potential of circulating EVs for assessing ARD risk in SZ patients. Likewise, they have been implicated in AD neuropathology. For instance, C4A is linked to complement system activation, which drives synaptic pruning and neuronal loss [6, 35]—hallmarks of AD pathology. Similarly, HPX and immunoglobulin-related proteins (JCHAIN, IGHM, and IGHG3) are associated with chronic neuroinflammation and oxidative stress, both of which accelerate tau pathology and amyloid-β accumulation. For comprehensive works on this subject, refer to [4, 29, 34, 39]. Notably, recent findings also identify JCHAIN as a new member of the CXCL family, involved in the CXC chemokine receptor system [31]. This discovery suggests that JCHAIN may have previously unrecognized critical roles in neuroinflammation and blood-brain barrier disruption, as observed in other CXC chemokines [61].

In a similar context, prior research has established that bEV are integral to both signaling and proteostasis in brains affected by neurodegenerative conditions [1, 23, 33]. Recent investigations further substantiate that disruptions in protein homeostasis—encompassing protein aggregation, proteasome dysfunction, and impaired autophagy—contribute to the pathology of SZ, reflecting biological processes observed in ARD such as AD [45, 47]. For example, a recent study involving olfactory neuronal cells from SZ patients has identified insoluble, ubiquitinated protein aggregates, akin to the misfolded proteins characteristic of AD [44]. Moreover, impairments in the ubiquitin–proteasome system and the downregulation of autophagy-related genes such as BECN1 and ULK2 have been documented in SZ, paralleling similar deficits in AD where proteostasis failure accelerates neurodegeneration [24, 26]. These findings suggest the existence of convergent intracellular mechanisms, although direct molecular-level connections between SZ and ARD remain inadequately characterized. Building on this foundation, our study demonstrates that while convergent bEV proteomes in age- and group-matched controls are primarily associated with signaling pathways, in SZ and ARD, they are predominantly linked to autophagy, reflecting disease-specific proteostatic disturbances. Although dysfunctional autophagy affecting bEV flow has been previously documented in ARD [7, 23, 58], our data provide novel molecular evidence in SZ, highlighting specific bEV-associated proteins as potential mediators of shared neuropathology.

In the second phase of our study, we leveraged the well-established ability of bEV to circulate systemically [3, 22, 51] to investigate pEV data from a large, independent clinical cohort. Our analysis revealed that several proteins initially identified in bEV were also detectable in pEV of SZ patients. Notably, we observed strong correlations - exceeding 0.90 - between protein levels in bEV levels and those in pEV—despite the sample originating from independent cohorts and being processed using a different classical EV isolation strategy. These results not only validate the dysregulation observed in bEV but underscore the potential of circulating pEV proteins as robust biological markers for SZ-related neuropathology. To our knowledge, this is the first demonstration that SZ-associated bEV dysregulation is mirrored in circulating pEV, and that these markers can be used to stratify SZ patients according to their risk for ARD. Notably, C4A and IGHM exhibited a near direct significant linear correlation between their levels in brain-derived and pEV in SZ patients. Additionally, we confirmed the systemic circulation of bEV-associated proteins commonly dysregulated in both ARD and SZ, including IGHM, IGHG3, C4A, HPX, and JCHAIN, which were significantly elevated in pEV of SZ subjects compared to matched controls. The SZ-to-control ratios for these circulating pEV proteins, established in this study, suggest strong clinical validity for identifying individuals exhibiting the bEV-mediated neuropathological mechanisms described here. More importantly, our unsupervised partition-based clustering analysis identified specific ratios within the circulating bEV proteomes that significantly stratify SZ subjects in the clinical cohort. These findings propose a prognostic capacity of these markers to identify SZ individuals at risk for ARD, with our data indicating that approximately 20% of all SZ subjects in the validation cohort exhibit these risk-associated profiles. This study thus represents the first investigation into potential mechanisms mediating common neuropathology between SZ and ARD in SZ patients. Furthermore, it provides a clinically relevant battery of markers associated with bEVs and pEV proteomes that may facilitate the stratification of individuals at risk. Our findings highlight the importance of considering EV-mediated protein secretion within a broader biological context. Previous studies demonstrated that proteins present in EV not only reflect the state of their parent cells but can also be interpreted alongside measurements in brain tissue and known functions of membrane transporters, providing a coherent framework to understand convergent neuropathological mechanisms across SZ and ARD [8, 13, 25, 49]. This perspective underscores the biological plausibility of EV as indicators of intracellular and extracellular alterations, thereby supporting their potential application as biomarkers.

Several considerations must be acknowledged. First, the relatively small sample size of the ARD cohort may present a limitation. In our study, this limitation was addressed by treating the inclusion of this cohort in the initial study phase as hypothesis-generating, forming the initial stage of a larger research plan, and, more importantly, by establishing robust power calculations that support the reliability of the observed effects. Second, individuals with SZ are typically treated with antipsychotic medications, including in Spain, where part of our post-mortem cohort was collected, and where pharmacological treatment is mandatory for these patients. Consequently, medication exposure is expected in nearly all cases, though regimens, duration, and adherence may vary. This variability may introduce confounding effects on molecular signatures, and although our analyses cannot fully disentangle medication-related influences, acknowledging this factor is essential for accurate interpretation.

Therefore, we assert that further validation is required in an independent cohort of SZ patients affected by ARD, compared to non-demented SZ age-matched controls. If clinically validated, these findings provide clinicians with a valuable tool for stratifying ARD risk in this vulnerable population. Early identification of at-risk individuals may facilitate the implementation of preventive and therapeutic strategies aimed at delaying or mitigating this devastating clinical outcome in SZ patients.

Conclusions

The data obtained in this study provide novel insights into the molecular mechanisms linking SZ, AD, and VaD, highlighting bEV as potential mediators of shared neuropathological processes. Our findings reinforce the role of bEV in neuronal connectivity, synaptic integrity, neuroinflammation, and metabolic dysfunction, supporting the idea that these vesicles contribute to overlapping disease pathways in SZ, AD, and VaD. By extending our investigation to pEV, we demonstrated their potential as biomarkers for disease stratification and ARD risk assessment in clinical populations of SZ patients. The strong correlation between bEV and pEV profiles suggests that pEV could serve as accessible clinical tools for monitoring neuropathological changes. Although further validation in independent cohorts is necessary to establish the clinical relevance of these findings, this study represents a significant step forward in stratifying SZ clinical populations, enabling early interventions and ARD risk monitoring in this at-risk population.

In summary, our study confirms prior evidence that EV reflect neurodegenerative processes, while providing novel insights by (i) establishing shared and disease-specific bEV proteomes in SZ and ARD, (ii) identifying bEV-linked autophagy and neuroinflammation as convergent mechanisms, and (iii) demonstrating that five bEV-derived proteins (IGHM, IGHG3, C4A, HPX, and JCHAIN) are mirrored in circulating pEV, where they show prognostic relevance for stratifying ARD risk in SZ. These findings highlight both the confirmatory and novel aspects of our work, situating it at the interface of molecular neuropathology and clinical translation.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

AA

Ammonium acetate

ARD

Age-related dementias

ACN

Acetonitrile

AD

Alzheimer’s disease

ANOVA

Analysis of Variance

Ba

Broadman area

C

Control

CNS

Central Nervous System

DIA

Data-independent acquisition

DLPFC

Dorsolateral prefrontal cortex

DSM-IV

Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition

DTT

Dithiothreitol

EVs

Extracellular vesicles

FASP

Filter-Aided Sample Preparation

FDR

False discovery rate

HBTRC

Harvard Brain Tissue Resource Center

HPLC

High-performance liquid chromatography

IAA

Iodoacetamide

ICD-10

International Classification of Diseases, 10th Edition

LC–MS/MS

Liquid chromatography—tandem mass spectrometry

NTA

Nanoparticle tracking analysis

PASEF

Parallel accumulation-serial fragmentation

PBS

Phosphate-buffered saline

PLSD

Partial Least Squares Discriminant

PMD

Post-mortem delay

PROSPR

Protein Organic Solvent Precipitation

RT

Room temperature

SMHC

Shanghai Mental Health Center (SHMHC)

SQSTM1

Sequestosome 1

SZ

Schizophrenia

VaD

Vascular dementia

Author contributions

J.A.S.M., M.M. and I.M. contributed to the design of the experimental work, the generation and analysis of data, and the drafting of the manuscript.; J.L., M.F.-A., C.L., F.B. and B.M. contributed to the experimental methodology.; M.G-S., I.B., J.M., L.F.C., A.R.-M. and R.N.K., provided clinical samples, critical clinical insight, and contributed to the edition and revisions of the manuscript; A.S. and X.G-P. contributed to the conceptualization, experimental design, supervision, funding acquisition, analysis of data, and writing and edition of the manuscript. All authors have read and approved the final version of the manuscript for submission.

Funding

Support for this work was provided by the National Institute of Health/Instituto de Salud Carlos III-ISCIII, Spain (PI22/00443 and DTS24/00141; grants co-funded by the European Union); the Ministry of Science and Innovation-MCIN, Spain and the National Research Council/Agencia Estatal de Investigación-AEI, Spain (PID2020-114885RB-C21) funded by MCIN/AEI/10.13039/501100011033; The Spanish Ministry of Science and Innovation with funds from the European Union NextGenerationEU (PRTR-C17.I1); from the Autonomous Community of Catalonia within the framework of the Biotechnology Plan Applied to Health (EVBRAINTARGET-Y7340-ACPPCCOL007) coordinated by the Institute for Bioengineering of Catalonia (IBEC)); the University of Lleida through (X25022 funded by the call Innovative Projects with Strong Potential for Knowledge or Technology Transfer); The Diputació de Lleida, Spain (PIRS22/03 & PIRS23/02); the Catalan Research Council-AGAUR (2023 LLAV 00056; and 2022 DI 100) and the Basque Government (Grant IT1512/22). X.G.-P. acknowledges a Miguel Servet program tenure track contract (CP21/00096) from the ISCIII, awarded on the 2021 call under the Health Strategy Action, co-funded by the European Union (FSE+). A.S. acknowledges a Ramón y Cajal program tenure track contract (RYC2021-030946-I) funded by MCIN/AEI/10.13039/501100011033 and by the “European Union NextGenerationEU/PRTR”. The PhD contract of M.M. is funded by the MCIN-AEI (PR2021-097934); the PhD contract of J.L. is funded by AGAUR, Spain (2022 DI 100) and by the company Algèmica Barcelona S.L.; The PhD contract of J.A.S.M. is funded by AGAUR (2024 FI-2 00054), and also supported by Diputació de Lleida ‘Ajuts al Talent en Investigació Biomèdica”. The PhD contract of I.M. and the fellowship for initiation to research 2024 of M.F. are funded by the Diputació de Lleida and IRBLleida through the IREP program. IRBLLEIDA, J.A.S.M., X.G.-P., and A.S. are co-funded by the CERCA Program/Generalitat de Catalunya. IRBLLEIDA, X.G.-P., and A.S. contracts are co-funded by the CERCA Program/Generalitat de Catalunya. A.R.-M, J.J.M, L.F.C. and X.G.P. are members of the ExoPsyCog Consortium, funded by IKUR-Neurobiosciences —Basque Government.

Data availability

All data supporting the conclusions of this study were made publicly available via the ProteomeXchange consortium in the partner repository PRIDE with the following identifiers: [Identifier PXD015578 proteomics data on ARD bEVs]; [Identifier PXD042732 proteomics data on SZ bEVs]; [Identifier PXD040261 proteomics data on SZ pEVs]. The scripts used for bioinformatic analyses are available on GitHub: [https://github.com/JoseASanchezMilan/SZ_DM](https:/github.com/JoseASanchezMilan/SZ_DM).

Declarations

Ethics approval and consent to participate

Informed consent was obtained from all participants or their legally authorized representatives. The use of post-mortem brain tissue adhered to the ethical principles of the Declaration of Helsinki, and all experimental protocols were conducted in strict compliance with the relevant institutional guidelines and the ethical committee of the University Hospital Arnau de Vilanova, Lleida, Spain. The use of clinical blood plasma samples followed the ethical standards set by the Declaration of Helsinki, and all experimental procedures were carried out in full accordance with the applicable institutional guidelines with approval reference CEIC-3089.

Consent for publication

All authors have read and approved the manuscript for submission in its current form.

Competing interests

J.A.S.M., M.M., I.M., J.L.M., C.L., A.S., and X.G.-P. are co-inventors on the European patent application EP25383001.2, filed by the public, non-profit research organizations University of Lleida (UdL) and the Biomedical Research Institute of Lleida (IRBLLEIDA). This application relates to the clinical translation of the biomarker panel described herein for identifying individuals with psychotic disorders—primarily schizophrenia—who are at risk of age-related dementias. All remaining authors declare no competing interests.