Introduction

Schizophrenia (SCZ) is a severe psychiatric disorder affecting approximately 1% of the global population, leading to an average potential life loss of about 14.5 years1. Early diagnosis and intervention can improve patient outcomes2. However, the diagnosis of SCZ currently relies on the subjective assessment and judgment of psychiatrists and lacks objective biomarkers. Additionally, the efficacy of antipsychotic medications varies among individuals, with approximately 40% of patients exhibiting poor response and being classified as having refractory schizophrenia3. Exosomes have garnered considerable attention due to their distinctive physiological functions4. Furthermore, exosomes participate in several physiological mechanisms closely associated with SCZ, including cell communication5, regulation of synaptic plasticity, and nerve regeneration processes6. Previous studies investigating exosomes in SCZ have primarily examined bulk exosomes rather than individual exosomes. In this study, we employed the proximity barcoding assay (PBA)7, an innovative high-throughput method for single-exosome analysis, to simultaneously profile 256 surface proteins, encompassing markers of exosomes and proteins associated with nervous system, immunity, tumors, cardiovascular diseases, and other conditions8,9, on individual exosomes of antipsychotic responders (RES) and non-responders (NRES).

Methods

The study was approved by the Medical Ethics Committee of the second Xiangya Hospital and written informed consent was obtained from all participants. Patients (n = 50) with first-episode SCZ underwent cognitive function assessments and provided plasma samples at baseline and then received risperidone for a duration of 3 months. Following this treatment period, participants were classified as either RES (reduction rate > 50%) or NRES based on the PANSS. Plasma exosomes from bottom 10 NRES and top 10 RES, as well as 10 matched healthy controls (HCs), were used for PBA. Details of PBA was provided in Supplement methods. To prioritize reliable proteins relevant to SCZ, we filtered the differential expressed proteins (DEPs) using two criteria: first, the protein expression was 1.5 folds or higher in the SCZ (NRES) compared to the HC (RES) group, and second, the protein or gene had been previously associated with schizophrenia in the published literature.

Results

The participants in our study included RES (n = 10), NRES (n = 10), and healthy controls (HC, n = 10). There was no difference in total PANSS scores or PANSS positive symptom scores between the RES and NRES at baseline (Table S2).

We first compared exosomes protein expression among SCZ patients, including both RES and NRES, and HC. Total protein expression and the average expression of proteins on extracellular vesicles (EVs) showed no difference (Fig. S1C–E). In differential protein analysis, 52 surface proteins exhibited significant differences (Table S3). After further filtering with criteria in methods, five DEPs remained: FN1, ITGB3, PECAM1, ITGA6, and CD5 (Fig. S1I). These proteins demonstrated an area under the curve (AUC) of up to 0.805 in ROC analysis for distinguishing SCZ from HC (Fig S2A–E).

Furthermore, we examined protein expression between RES and NRES groups. Both total protein expression and average EV protein expression showed no difference between RES and NRES (Fig. S1F–H). When it comes to individual proteins, 18 exhibited significant differences (Table S4). After further filtering, six DEPs—ITGB3, CD33, CD40, CD36, TENM2, and EGFR—met our criteria (Fig. S1J) with an AUC of up to 0.87 in discriminating between RES and NRES(Fig. S2F–L).

To classify exosomes based on proteomic characteristics, an unsupervised machine learning algorithm, FlowSOM10, was applied to generate exosome clusters. The t-SNE plot in Fig. 1A displays the clustering of exosomes from all samples, revealing 13 identified clusters through FlowSOM. Fig, 1B depicts the biomarkers of each cluster. Next, we analyzed the similarities and differences among the RES, NRES, and HC groups (Fig. 1C, D). Among all clusters, cluster 12 exhibited an increasing trend in proportion in SCZ, from 8.48% in the HC to 13.27% in NRES and 20.16% in RES (P = 0.037 for SCZ and HC, P = 0.003 for NRES and RES, Fig. 1F). DEPs ITGB3, PECAM1, and ITGA6 are predominantly expressed in cluster 12, suggesting an association of cluster 12 with SCZ and antipsychotic effectiveness. The proportion of cluster 11 also showed an increasing trend in SCZ, rising from 3.57% in the HC group to 7.05% in the SCZ group (P = 0.031), with no significant difference between RES and NRES (Fig. 1E). Since DEPs PECAM1, and ITGA6 are also predominantly expressed in cluster 11, cluster 11 may also be associated with SCZ.

Fig. 1: Exosome subpopulation alteration in SCZ patients.
figure 1

A Exosome subpopulations from plasma samples of both SCZ patients and healthy donors were determined via the FlowSOM algorithm, an unsupervised machine learning process. Thirteen subpopulations were shown in t-distributed stochastic neighbor embedding (t-SNE) plot. B Proteomic biomarkers of each subpopulation were shown in heatmap. C The distribution of exosomes among each subpopulation was displayed for each sample group, including HC, RES and NRES groups. D t-SNE plot for each sample with group information. E, F Cluster 11 and 12 featured expressing PECAM1, ITGA6 and ITGB3 increased significantly in SCZ group, furthermore, cluster 12 increased in NRES group compared to RES. Data represent the mean ± SD, Statistical analyses were performed using the t test, n = 10 for HC, RES and NRES, n = 20 for SCZ. *P < 0.05; **P < 0.01.

Full size image

To investigate whether exosomes-expressed these surface proteins were correlated with cognitive functions, we assessed the cognitive functions of individuals with SCZ and HC (Table S5) and performed correlation analysis DEPs and cognitive functions (Fig. 2). For DEPs between SCZ and HC, FN1 was positively correlated with completion time of TMTB (r = −0.67 P = 0.004514) and SCWT-A (r = −0.669 P = 0.004514) after FDR correction. For DEPs between NRES and RES, correlations were not significant after FDR correction.

Fig. 2: Correlation Heatmap of DEPs and cognitive functions.
figure 2

For DEPs between SCZ and HC, PECAM1, ITGB3, ITGA6 and FN1 were negatively correlated with scores on BACS-SC, BVMT-R, and CF; Additionally, PECAM1 was negatively correlated with scores on HVLT-R and positively correlated with the time taken for TMTB and SCWT-A; ITGA6 was negatively correlated with accuracy of CPT-Hearing; FN1 was positively correlated with completion time of CPT-Version, TMTA, TMTB and SCWT; CD5 was positively correlated with completion time of TMTB and SCWT-B. For DEPs between NRES and RES, CD33 was negatively correlated with score of HVLT-R, maze task and BVMT-R; CD40 was negatively correlated with score of HVLT-R, BASC-SC and BVMT-R; TENM2 was negatively correlated with WMS-III parallel and total score; *P < 0.05; **P < 0.01; ***P < 0.001.

Full size image

Discussion

Our results reveal differences in the expression of proteins, including FN1, ITGB3, PECAM1, ITGA6, and CD5, on single exosomes between SCZ patients and healthy controls. Similarly, differences were observed in the expression of proteins ITGB3, CD33, CD40, CD36, TENM2, and EGFR between RES and NRES patients. ROC analysis demonstrated their potential as biomarkers for distinguishing SCZ from healthy controls or NRES from RES. Moreover, the proportion of exosome subpopulations expressing PECAM1, ITGA6, ITGB3, and ITGB1 increased in the total plasma exosomes of SCZ patients, particularly in those who were responsive to treatment (RES). Additionally, DEPs including PECAM1, ITGA6, ITGB3 and FN1 were found to be correlated with cognitive function.

Exosomal surface proteins reflect the characteristics of their respective cells and tissues11, indicating that increased expression of these proteins reflects heightened cellular and tissue activities and communication. The increased expression of FN1 may contribute to the transformation of vascular smooth muscle cells in intimal thickening of atherosclerotic lesions12, while PECAM1 is associated with endothelial cell function13. Morris et al.13 have shown that endothelial dysfunction is caused by inflammation, oxidative stress, and mitochondrial dysfunction, all of which play significant roles in the pathology of SCZ. Previous studies indicate that patients with SCZ face a substantially higher risk of cardiovascular disease (CVD) compared to the general population14. FN1 and PECAM1 may contribute to the heightened risk of CVD in SCZ patients. Both ITGA6 and ITGB3 are integrin components involved in mediating cell-matrix and cell-cell adhesion processes15. Integrin β has been found to play a crucial role in synapses, with β3 influencing synaptic strength by regulating the magnitude size and content of excitatory synaptic transmission15. Additionally, β3 integrins are necessary for homeostatic plasticity, a type of synaptic plasticity16. Simultaneously, ITGA6 may be associated with neuronal migration17. Besides, ITGB1, which has not been directly reported to be associated with SCZ but is found to be highly expressed in SCZ group in the present study, is essential for synapse formation18. This suggests that ITGB3, ITGA6, and ITGB1 could play roles in the neurodevelopmental processes and synapses function of SCZ. CD5 is an immune-associated protein expressed on both T cells and B cells. Printz et al.19 demonstrated the elevation of CD5 + B lymphocytes in SCZ patients, which may be the origin of the highly expressed CD5+ exosomes observed in this study.

We also identified 6 DEPs ITGB3, CD33, CD40, CD36, TENM2, and EGFR between RES and NRES. ITGB3, which was also differentially expressed between SCZ patients and HC, was found to be expressed at lower levels in NRES. As mentioned earlier, ITGB3 influences synaptic function. Synaptic dysfunction mediated by low ITGB3 may result in a weak response to drugs. CD33, CD40, and CD36 are immune-related proteins. Both CD3320 and CD4021 were found to be elevated in SCZ patient, induced by pro-inflammation factors and microglia activation22. Lower levels of CD33 and CD40 found in NRES may indicate suppressed microglial activation, suggesting that microglial may play a role in the prognosis of SCZ. Tomasik et al.23 found that monocyte CD36 levels could serve as a reliable indicator of response to olanzapine treatment. They demonstrated that patients with high CD36 expression on monocytes were less likely to respond to olanzapine, consistent with our finding that NRES patients had higher CD36 levels. TENM2, an adhesion molecule highly expressed in neurons, plays a critical role in neural migration24 and intracellular calcium signaling25 and has been proposed as a SCZ-related gene26. EGFR, the receptor of epidermal growth factor, is distributed across cell populations involved in SCZ neuropathology, including GABAergic and dopaminergic neurons27. Postmortem studies have revealed up-regulation of EGFR protein in the forebrain regions of patients with SCZ28. Animal studies have demonstrated that EGFR activation in the brain leads to dopamine release in the striatum or globus pallidus, resulting in behavioral impairments relevant to SCZ29,30,31. Moreover, epidermal growth factor negatively regulates GABAergic development in the neocortex32.

Our findings suggested that DEPs on exosomes are correlated with cognitive functions, particularly those distinguishing SCZ from HC. Exosomes originating from the brain can traverse the blood–brain barrier and be identified in plasma4. On the other hand, elevated levels of cell adhesion proteins are linked with increased cerebrovascular resistance33, potentially leading to reduced blood flow in watershed areas of the brain, resulting in ischemic damage to regions controlling motor function and attention34.

There were several limitations in our study. Firstly, we analyzed plasma samples from a limited pool of 30 individuals, which may have excluded certain biomarkers and obscured correlations between proteins and cognitive functions after FDR correction. Secondly, we did not validate the identified surface proteins using an independent cohort. This was due to the limited number of NRES in our cohort, making it insufficient to establish an independent cohort for validation. Lastly, Lastly, the pathophysiological function of these exosome-expressed DEPs in schizophrenia remains unclear. Therefore, future studies are needed to validate the results in a larger cohort and explore the specific roles these proteins play in schizophrenia.

In conclusion, we utilized a novel single-exosome analysis method, PBA, to profile the expression features of exosome surface proteins in the plasma of SCZ patients. We identified differentially expressed surface proteins in single exosomes and evaluated their diagnostic and prognostic capacities using ROC analysis. Additionally, we observed alterations in exosome subpopulations among RES patients, NRES patients, and the HC group, identifying two subpopulations characterized by the expression of PECAM1, ITGB3, ITGA6, and ITGB1 associated with SCZ. Further research is warranted to validate these potential biomarkers and elucidate their roles in schizophrenia.