CircRNA-miRNA-mRNA networks in plasma extracellular vesicles as biomarkers for first-onset schizophrenia

Abstract

Background

The circRNA-miRNA-mRNA networks of extracellular vesicles (EVs) in first-onset schizophrenia (FOS) have not been reported yet. Here, we constructed circRNA-miRNA-mRNA networks of EVs, and examined their diagnostic efficiency in FOS.

Methods

The expression levels of circRNAs, miRNAs and mRNAs in EVs derived from 10 FOS patients and 10 healthy controls (HC) were determined by high-throughput sequencing. The circRNA-miRNA-mRNA networks was constructed based on the overlapped miRNAs between differentially expressed (DE) miRNAs and circRNA-targted miRNAs, and overlapped mRNAs between DE-mRNAs and miRNA-targeted mRNAs. Gene expression levels were validated using quantitative real-time PCR in 31 FOS and 31 HC cases. Receiver operating characteristic (ROC) curve analysis was performed to examine the diagnostic efficacy. Correlation analysis was performed using Pearson’s or Spearman’s correlation coefficient.

Results

There were 26,194 DE-circRNAs, 22 DE-miRNAs, and 2637 DE-mRNAs in plasma EVs of FOS patients. Then, the circRNA-miRNA-mRNA networks consisting of 9 circRNA, 6 miRNA and 16 mRNA, were constructed. Three network (chr15:93496587–93499879+—hsa-miR-20b-5p—ANKH; chr7:40037093–40087476+—hsa-miR-22-3p—C5orf24; and chr19:17883266–17883550+—hsa-miR-502-3p—B4GALT5) were selected for further investigation. The expression levels of 9 genes in validation data were consistent with the results of the high-throughput sequencing. The area under the ROC curve (AUC) of the circRNA-miRNA-mRNA network was higher than that of circRNA, miRNA or mRNA alone in plasma EVs, and the AUC of mRNAs in plasma EVs was higher than that of mRNAs in peripheral blood. The expression levels of chr15:93496587–93,499,879+, chr7:40037093–40,087,476+, hsa-miR-22-3p and B4GALT5 were correlated with the PANSS score.

Conclusion

We constructed the circRNA-miRNA-mRNA networks of plasma EVs in FOS, demonstrating their potential as a biomarker for FOS.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12888-025-07073-y.

Introduction

Schizophrenia (SZ) is a serious mental illness characterized by positive symptoms, negative symptoms and cognitive impairment. It can cause impairments in social and occupational function, bringing a heavy burden to the families of patients and the whole society [1]. According to the 2019 China Mental Health Survey, the lifetime prevalence of SZ in China has been on the rise in recent years [2]. Early diagnosis and timely intervention are very helpful to improve the condition [3]. However, the current diagnosis of SZ is based on clinical interviews with physicians. Therefore, the search for reliable biomarkers of SZ is crucial for the effective diagnosis and treatment of the disease.

Extracellular vesicles (EVs) were once considered a disposal system for cellular waste; however, their physiological functions have recently been demonstrated in various fields. EVs are membranous vesicles that encase biomolecules such as proteins and nucleic acids (e.g., mRNA, miRNA, and circRNA). Almost all cells, including neurons, can secrete EVs, which enter blood, cerebrospinal fluid, saliva, urine and other body fluids to play a remote regulatory role [4]. EVs have the following characteristics: they contain different components, have a lipid bimolecular layer that protects internal biomolecules from degradation, are detectable in almost all body fluids, and depend on the internal environment [5]. Therefore, they may be accessible biomarkers of neurological dysfunction.

Many recent studies have shown that EVs are promising biomarkers for psychiatric disorders such as depression and bipolar disorder [6, 7]. Although EVs-derived circRNAs, miRNAs, proteins, and metabolites have been reported in SZ [8–11], the exploration of circRNA-miRNA-mRNA networks is still lacking.

In this study, we detected the expression levels of circRNAs, miRNAs, and mRNAs in plasma EVs of patients with first-onset schizophrenia (FOS), which were newly diagnosed and had no history of drug treatment, and healthy controls (HC) by high-throughput sequencing, and we constructed an circRNA-miRNA-mRNA networks in EVs using bioinformatics method. The expression levels of three circRNA-miRNA-mRNA networks were verified using quantitative real-time PCR, and the receiver operating characteristic (ROC) curve analysis and clinical data correlation were conducted (Fig. 1).

Fig. 1.

Flowchart of the study

Materials and methods

Subjects

This study was approved by the Research Ethics Committee of the Research Ethics Committee of the First Hospital of Shanxi Medical University (2019-K039). It was conducted in accordance with the Declaration of Helsinki, and informed content to participate (Supplementary File: Consent to Participate Declaration) was obtained from all participants was informed. Patients with FOS, who had no history of drug treatment were recruited from the the First Hospital of Shanxi Medical University and Shanxi Province Mental Health Center. The criteria was described previously [12] as following: (1) diagnosed as SZ by two psychiatrists according to the Diagnostic and Statistical Manual of Mental Disorders Fifth Edition (DSM-5); (2) the first onset of SZ and had no history of antipsychotic drug treatment; (3) Aged 16–60 years; (4) The han nationality; (5) Right-handed; (6) PANSS ≥ 60; (7) Patients with other mental disorders, neurological disease, organic brain disease and substance dependence or abuse were excluded. (8) Women who are pregnant or breastfeeding were excluded.

The HC subjects were recruited from community, and the criteria was as following: (1) Aged 16–60 years; (2) The han nationality; (3) Right-handed; (4) Individuals with history or current presence of mental illness, neurological disease, organic brain disease, serious unstable physical illnesses and substance dependence or abuse were excluded.

We first recruited 10 FOS and 10 HC whose age, sex and education levels were matched with that of FOS, and performed RNA high-throughput sequencing of plasma EVs to the discover of differentially expressed circRNAs, miRNAs, and mRNAs, defined as discovery sets. Subsequently, 31 cases of FOS and 31 HC were recruited to verify the expression levels of circRNAs, miRNAs and mRNAs, defined as validation sets. The sex, age, and education level showed no significant difference between the HC group and the FOS groups in the validation set.

EVs isolation and identification

In the morning on an empty stomach, 10 mL of peripheral blood from each participant was extracted into an EDTA anticoagulant tube, and approximately 5 ml of plasma was isolated for EVs extraction using the exoEasy Maxi Kit (Qiagen). The diameter of EVs was determined by Nanosight particle tracking analysis (NTA). The morphology of EVs was observed using electron microscopy. CD63, CD81, and TSG101, the marker proteins of EVs, were detected using western blot analysis.

RNA sequencing

High throughput sequencing was provided by Shanghai Cloud Sequence Biotechnology Co., LTD. Transcriptome Sequencing was performed to detect the expression level of circRNA and mRNA. Experimental process was as follow: rRNAs are removed from total RNA using the Ribo-Zero rRNA Removal Kit (Illumina, USA) according to instructions.The TruSeq Stranded Total RNA Library Prep Kit (Illumina, USA) was used to pretreat RNA and construct sequencing libraries. The BioAnalyzer 2100 instrument (Agilent Technologies, USA) was used for library quality control and quantification. According to Illumina sequencing instructions, the 10pM library was denatured into single-stranded DNA molecules, captured on Illumina flowcell, and expanded in situ into clusters. The 150 cycle sequencing was performed in PE mode on the Illumina Novaseq 6000 sequence. The analysis process is as follows: After sequencing by Illumina sequencer, double-ended reads were harvested. Q30 was used for quality control, cutadapt software (v1.9.3) was used to remove the connectors, and low quality reads were removed and high quality reads were obtained. circRNA and mRNA were analyzed by different methods. circRNA: High quality reads were compared to the reference genome/transcriptome using STAR software (v2.5.1b), and circular RNA was detected and identified using DCC software (v0.4.4). circBase database and Circ2Traits were used to annotate the identified circular RNAs. The edgeR software (v3.16.5) was used for data standardization and circRNA screening. mRNA: Using hisat2 software (v2.1), high quality reads were compared to the human reference genome (UCSC HG19). Then, guided by the gtf gene annotation file, the FPKM of gene-level mRNA was obtained using the cuffdiff software. (Fragments per kilobase of exon per million fragments mapped) value, which is used as mRNA expression profile, and the multiples change and p-value between two/group samples are calculated to screen differentially expressed mRNA (fold change ≥ 2, p value <0.05).

miRNA Sequencing was performed to detect the expression level of miRNA. Experimental process was as follow: The miRNA sequencing library was prepared using the total RNA of each sample, including the following steps: 1) 3’ connector connection; 2) 5’ connector connection; 3) cDNA synthesis; 4) PCR amplification; 5) Recovery of ~ 150 bp PCR amplification fragments (corresponding to ~ 22 nt miRNAs); The libraries were denatured into single-stranded DNA molecules, captured in Illumina flow cells, amplified in situ into clusters, and sequenced at 50 cycles on the Illumina Novaseq 6000 sequencer according to the vendor’s operating instructions. The analysis process is as follows: after sequencing by Illumina sequencer, image analysis and base identification, the original reads after quality control were harvested. First, use Q30 for quality control. cutadapt software (v1.9.3) was used to decalculate original reads, remove low-quality reads, and retain reads with length > = 15 nt. The decalculated reads were obtained. trimmed reads of all samples were then combined to predict new miRNA using miRDeep2 software (v2.0.0.5). trimmed reads of each sample were compared to the merged pre-miRNAs database (miRBase v22 pre-miRNAs + newly predicted pre-miRNAs) using Novoalign software (v3.02.12). A maximum of one mismatch is allowed. The number of tags on each mature miRNA was statistically compared as the original expression level of the miRNA, and the TPM (tag counts per million aligned miRNAs) method was used for standardization. The Fold change between two samples was calculated, and the difference table was screened to miRNAs. The P-value and Fold Change between the two groups of samples (with duplicates) were calculated, and differentially expressed miRNAs were screened (fold change ≥ 1.5, p value <0.05).

circRNA–miRNA–mRNA network construction

CircRNA-miRNA-mRNA networks were constructed based on differentially expressed (DE) circRNAs, miRNAs and mRNAs. First, circRNA-miRNA and miRNA-mRNA networks were constructed. The target miRNAs of DE circRNAs were predicted using TargetScan and miRanda. The target mRNAs of DE miRNAs were predicted using TargetScan and miRDB. Then Venn maps were used to analyze the intersection between DEcircRNA targeted miRNA and DEmiRNA, DEmiRNA targeted mRNA and DEmRNA. Venn diagrams were generated using Venny 2.1.0 (https://bioinfogp.cnb.csic.es/tools/venny/index.html). According to overlapped down- or upregulated miRNA, up- or downregulated circRNA were recognized reversely. Down- or upregulated miRNA were recognized according to overlapped up- or downregulated mRNAs. Alluvial plots created using an online tool (https://www.bioinformatics.com.cn) were used to present the circRNA-miRNA, miRNA-mRNA network.

The circRNA-miRNA-mRNA networks were constructed based on circRNA-miRNA and miRNA-mRNA networks. The intersection of miRNAs between the circRNA–miRNA and miRNA–mRNA networks was taken, and the interacting circRNAs and mRNAs were combined to construct the circRNA–miRNA–mRNA networks. Cytoscape was used for visualization of the network.

Gene expression level determination

Trizol (Invitrogen life technologies) was added to EVs for the extraction of total RNA. Reverse transcription process of circRNA and mRNA was as follow: firstly RNA, N6 primer (ThermoFisher) and dNTP Mix (HyTest Ltd) were mixed and incubated in 65℃ for 5 min and in ice for 2 min, subsequently, the SuperScript™ III Reverse Transcriptase (ThermoFisher), RNase Ihibitor (Epicentre), DTT and Buffer were added, mix well and keep at 37℃ for 1 min 50℃ for 60 min, 70℃ for 15 min. Reverse transcription of miRNA use SuperScriptTM III Reverse Transcriptase (ThermoFisher), The reverse transcription reaction procedure is: 16℃ 30 min, 42℃ 40 min, 85℃ 5 min. Quantitative real-time (qRT)-PCR was performed using qPCR SYBR Green master mix (GenSeq). The primers used are listed in the Table S1, and ACTB was used as an internal control for circRNAs and mRNAs, and U6 was used as internal control for miRNAs. The qPCR reaction is carried out according to the following procedure: 95℃, 10 min; 95℃, 10 s, 60℃, 60 s (collecting fluorescence), 40 PCR cycles and establish the melting curve of the PCR product after the amplification reaction is completed. Data were analyzed using the 2 - △△ CT method.

Diagnostic efficacy analysis

The diagnostic efficacy of circRNA, miRNA, mRNA and circRNA-miRNA-mRNA networks was examined by ROC curve analysis. ROC curves was performed using the OmicStudio tools at https://www.omicstudio.cn/tool/58.

Statistical analysis

Statistical analysis was performed using SPSS version 23.0. Demographic information and quantitative real-time PCR, data are presented as mean ± standard error (SE). Statistical analyses of age and education level and gene expression level which were normally distributed were performed using student’s t-test. Statistical analysis of gender was performed using the Chi-square test. After controlling for sex, age, and years of education, statistical analysis of gene expression level was performed using the general linear model followed by a univariate test. Correlation analysis between gene expression levels and PANSS scores was based on bivariate correlation analysis, and controlling for gender, age, and years of education was based on partial correlation analysis.

Results

DE circRNAs, MiRNAs and mRNAs in plasma EVS

A total of 10 FOS patients and 10 age-, sex-, and education-matched HC subjects were recruited. The mean value of the PANSS score was 75.6 for the 10 FOS patients. The average mean duration of the illness was 57.6 weeks (Table S2).

The EVs derived from plasma were identified by NTA, electron microscopy and western blotting. The predominant vesicle size was between 60 and 90 nm with cup-shaped structures (Fig. S1a, b). And CD63, CD81, and TSG101 Western blot of the vesicles verified that the isolated particles were EVs (Fig. S1c-e).

RNA sequencing was performed on plasma EVs from 10 FOS patients and 10 HC subjects to identify differentially expressed circRNAs, miRNAs and mRNAs in EVs derived from FOS patients. There were 11,860 up-regulated and 14,334 down-regulated circRNAs (Fig. 2a). And there are 14 up-regulated and 8 down-regulated miRNAs (Fig. 2b), 690 up-regulated and 1947 down-regulated mRNAs (Fig. 2c) [12]. Differently expressed profiles could distinguish between the samples from FOS patients and HCs (Fig. 2d-f). Subsequently, we conducted GO functional analysis of the DEmRNAs and found that the DEmRNAs in BP were enriched in the growth, development, differentiation and function of neurons, axons, dendrites and synapses (Fig. 2g). The KEGG pathway analysis revealed that the DEmRNAs were enriched in apoptosis, axon guidance, and TNF signaling pathways (Fig. 2h).

Fig. 2.

Identification of DE-circRNAs, DE-miRNAs and DE-mRNAs in EVs. a-c: Volcano plots of DE-circRNAs (a), DE-miRNAs (b) and DE-mRNAs (c) in SZ EVs. d-f: Heatmap of DE-circRNAs (d), DE-miRNAs (e) and DE-mRNAs (f) in SZ EVs. g: Bubble plot of BP of DEmRNAs. h. Bubble plot of KEGG pathway analysis DEmRNAs

Construction circRNA-miRNA-mRNA networks

CircRNA-miRNA-mRNA networks were constructed based on the DE- circRNAs, miRNAs and mRNAs. First, circRNA-miRNA and miRNA-mRNA networks were constructed. There were 1539 up-regulated circRNAs targeted miRNAs. The Venn map identified 3 down-regulated miRNAs (Fig. 3a). According to 3 down-regulated miRNAs, 4 up-regulated circRNAs were recognized reversely (Fig. 3b). There were 1831 down-regulated circRNAs targeted miRNAs. The Venn map identified 8 overlapped up-regulated miRNAs between down-regulated circRNAs targeted miRNAs and up-regulated miRNAs (Fig. 3c). Based on these eight up-regulated miRNAs, 15 down regulated circRNAs were recognized (Fig. 3d).

Fig. 3.

Construction of a circRNA-miRNA-mRNA network. a. Venn map of overlapping miRNAs between the up-regulated circRNA targeted miRNA and down-regulated miRNA. b. alluvial plot of miRNA (down-regulated)-circRNA (up-regulated). c. Venn diagram of overlapping miRNAs between the down-regulated circRNA targeted miRNA and up-regulated miRNA. d. alluvial plot of miRNA (up-regulated)-circRNA (down-regulated). e. Venn diagram of overlapping mRNAs between the down-regulated miRNA targeted mRNA and up-regulated mRNA. f. alluvial plot of mRNA (up-regulated)-miRNA (down-regulated). g.Venn diagram of overlapping mRNAs between the up-regulated miRNA targeted mRNA and down-regulated mRNA. h. alluvial plot of mRNA (down-regulated)- miRNA (up-regulated). i. circRNA (down-regulated)-miRNA (up-regulated)-mRNA(down-regulated) regulatory network. j. circRNA (up-regulated)-miRNA (down-regulated)-mRNA (up-regulated) regulatory network

The target mRNAs of DEmiRNAs were identified. There were 87 down-regulated miRNAs targeted mRNAs, and 2 up-regulated mRNAs were identified in Venn map (Fig. 3e). Based on these 2 up-regulated mRNAs, 1 down-regulated miRNA was recognized (Fig. 3f). There were 174 up-regulated miRNAs targeted mRNAs, and 27 down-regulated mRNAs were identified in the Venn map (Fig. 3g). Base on these 27 down-regulated mRNAs, eight up-regulated miRNAs were recognized (Fig. 3h).

The circRNA-miRNA-mRNA networks were constructed according to the circRNA-miRNA and miRNA-mRNA networks. We took the intersection of the 8 up-regulated miRNAs in Figs. 3d and 8 up-regulated miRNAs in Fig. 3h, and found 5 overlapped up-regulated miRNA. By combing the interacted circRNAs and mRNAs, we constructed the network including 8 down-regulated circRNAs, 5 up-regulated miRNAs and 14 down-regulated mRNAs, which included 29 edges (Fig. 3i). In the same way, the network including 1 up-regulated circRNA, 1 down-regulated miRNA and 2 up-regulated mRNAs was constructed which included 3 edges (Fig. 3j).

Next, the GO and KEGG pathway enrichment analysis were performed with mRNAs in the circRNA-miRNA-mRNA networks (Fig. S2). The mRNAs in BP were enriched in the regulation of relaxation of muscle, regulation of calcium ion transmembrane transporter activity, cellular response to epinephrine stimulus, cyclic nucleotide catabolic process, cyclic nucleotide catabolic process (Fig. S2a). The mRNAs in CC were enriched in voltage-gated calcium channel complex, calcium channel cpmplex, and cation channel complex, et al. (Fig. S2b). The mRNAs in MF were enriched in 3’,5’−cyclic − AMP phosphodiesterase activity, cAMP binding, 3’,5’−cyclic − nucleotide phosphodiesterase activity, cyclic − nucleotide phosphodiesterase activity (Fig. S2c). As for the the KEGG analysis, the mRNAs were enriched in morphine addiction, cAMP signaling pathway, parathyroid hormone synthesis, secretion and action, and cholinergic synapse (Fig. S2d).

CircRNA in the circRNA-miRNA-mRNA network

The details of nine circRNAs in the circRNA-miRNA-mRNA network were shown in the Fig. S3a, among which the following three were included by circBase: chr15:93496587–93,499,879+, chr7:40037093–40,087,476+, chr19:17883266–17,883,550+. chr7:40037092–40,087,476+ (circBase ID: hsa_circ_0005485) arose from the the exon3 -exon7 of CDK13 gene, and formed the mature sequence with a length of 729 bp (Fig. S3b). chr19:17883265–17,883,550+ (circBase ID: hsa_circ_0050042) was formed from the exon 9 and exon 10 of FCHO1, and the length of the mature sequence was 196 bp (Fig. S3c). chr15:93496586–93,499,879 (hsa_circ_0036984) was transcribed from the the exon 14, exon 15 and exon 16 of CDH2 gene, and the length of mature sequence was 498 bp (Fig. S3d). The three circRNAs were selected for further study.

Verification of the expression levels of circRNAs, miRNAs and mRNAs

We expanded the sample size (HC 31 vs. FOS 31). The two groups showed no significant differences in age, gender, education level, and the mean value of the PANSS score was 80.1 for the 31 FOS patients, and the average mean duration of the illness was 60.2 weeks (Table S2). The aim was to verify the expression levels of chr15:93496587–93,499,879+, chr7:40037093–40,087,476+, chr19:17883266–17,883,550+, which were included by circBase, and their downstream miRNAs and mRNAs (hsa-miR-20b-5p and ANKH, hsa-miR-22-3p and C5orf24, hsa-miR-502-3p and B4GALT5) in plasma EVs. In FOS patients, the expression levels of chr15:93496587–93,499,879 + and ANKH increased, the expression level of hsa-miR-20b-5p decreased. Furthermore, in FOS patients, the expression level of chr7:40037093–40,087,476 + and C5orf24 decreased, the expression level of hsa-miR-22-3p increased. Additionally, in FOS patients, the expression level of chr19:17883266–17,883,550 + and B4GALT5 decreased, the expression level of hsa-miR-502-3p increased in FOS. The differences in gene expression between FOS and HC were significant before and after controlling for age, sex and years of education (Table 1). These results were consistent with the high-throughput sequencing results of 10 HC vs. 10 FOS.

Table 1.

Expression levels of circRNAs, miRNAs and mRNA

	FOS (n = 31)	HC (n = 31)	F	P	Adjusted F	Adjusted P
chr15:93496587–93,499,879+	0.232 ± 0.056	0.029 ± 0.009	31.091	0.001	10.273	0.002
chr7:40037093–40,087,476+	0.291 ± 0.093	2.499 ± 0.720	15.359	0.005	7.548	0.008
chr19:17883266–17,883,550+	0.462 ± 0.118	5.719 ± 1.041	37.472	0.000	25.797	0.000
hsa-miR-20b-5p	0.841 ± 0.064	38.050 ± 3.916	41.789	0.000	80.999	0.000
hsa-miR-22-3p	1.576 ± 0.137	0.516 ± 0.024	37.760	0.000	57.15	0.000
hsa-miR-502-3p	2.067 ± 0.145	1.011 ± 0.116	2.420	0.000	28.992	0.000
ANKH	0.260 ± 0.105	0.006 ± 0.002	11.836	0.022	4.086	0.048
C5orf24	0.058 ± 0.033	1.830 ± 0.516	27.237	0.002	10.042	0.002
B4GALT5	0.003 ± 0.001	0.019 ± 0.004	37.288	0.001	14.728	0.000

Note: Data are presented as mean ± standard error(SE), Adjusted F and Adjusted P indicates the F value and P value controlled for gender, age, and years of education

ROC curve analysis of circRNAs, miRNAs and mRNAs in EVs

The ROC curve analysis was performed in order to examine the diagnostic performances of circRNAs, miRNAs, mRNAs and circRNA-miRNA-mRNA network. A larger the value of the area under the ROC curve (AUC) indicated better the diagnostic value.

First, we examined the diagnostic performances of circRNAs, miRNAs, mRNAs. ROC curve showed AUC values of 0.9303 for ANKH (specificity:0.839, sensitivity:0.871), 1 for hsa-miR-20ba-5p (specificity:1.000, sensitivity:1.000), 0.8606 for chr15:93496587–93499879+ (specificity:0.903, sensitivity:0.742 (Fig. 4a), 0.9532 for C5orf24(specificity:0.839, sensitivity:0.968), 0.9854 for hsa-miR-22-3p (specificity:0.935, sensitivity:1.000), 0.8418 for chr7:40037093–40087476+ (specificity:0.742, sensitivity:0.806) (Fig. 4b), 0.8356 for B4GALT5 (specificity:0.839, sensitivity:0.710), 0.8762 for hsa-miR-502-3p (specificity:0.871, sensitivity:0.839), 0.9469 for chr19:17883266–17883550+ (specificity:0.871, sensitivity:0.871) (Fig. 4c). The AUC values of chr15:93496587–93499879+ -- hsa-miR-20b-5p -- ANKH, chr7:40037093–40087476+ -- hsa-miR-22-3p -- C5orf24, and chr19:17883266–17883550+ -- hsa-miR-502-3p -- B4GALT5 were 1 (specificity:1.000, sensitivity:1.000) (Fig. 4d), 0.9979 (specificity:1.000, sensitivity:0.968) (Fig. 4e), and 0.9605 (specificity:0.968, sensitivity:0.903) (Fig. 4f), respectively. The AUC value was higher for the circRNA-miRNA-mRNA networks than circRNAs, miRNAs, mRNAs alone.

Fig. 4.

ROC curves of circRNAs, miRNAs, mRNAs in plasma EVs or peripheral blood of SZ. a. chr15:93496587–93,499,879+, hsa-miR-20b-5p, ANKH in plasma EVs of SZ. b: chr7:40037093–40,087,476+, hsa-miR-22-3p, C5orf24 in plasma EVs of SZ. c. chr19:17883266–17,883,550+, hsa-miR-502-3p, B4GALT5 in plasma EVs of SZ. d. chr15:93496587–93,499,879+ -- hsa-miR-20b-5p -- ANKH network in plasma EVs of SZ. e. chr7:40037093–40,087,476+ -- hsa-miR-22-3p–C5orf24 network in plasma EVs of SZ. f. chr19:17883266–17,883,550+ -- hsa-miR-502-3p -- B4GALT5 in plasma EVs of SZ. g. ANKH in peripheral blood of SZ. h. C5orf24 in peripheral blood of SZ. i. B4GALT5 in peripheral blood of SZ

Finally, we examined the diagnostic performances of three mRNAs: ANKH, C5orf24 and B4GALT5 in peripheral blood of SZ patiens. We downloaded the GSE27383 dataset, which was acquired from the Gene Expression Omnibus (GEO) database and contained information of PBMCs from 43 SZ and 29 HC. We found that the AUC values of ANKH, C5orf24, and B4GALT5 in peripheral blood were 0.6022 (specificity:0.690, sensitivity:0.558), 0.5341 (specificity:0.828, sensitivity:0.349) and 0.5389 (specificity:0.759, sensitivity:0.465), respectively (Fig. 4g-i), which were much lower than the values in plasma EVs.

Expression level of circRNAs, miRNAs, and mRNAs in EVs correlate with the PANSS scores

To explore the relationship between genes in EVs and clinical data in SZ, the correlation between gene expression levels and PANSS scores was examined. The expression levels of chr15:93496587–93,499,879+, chr7:40037093–40,087,476+, and B4GALT5 were positively correlated with the PANSS scores regardless of whether they were controlled for age, sex and years of education. The hsa-miR-22-3p expression levels was negatively correlated with the PANSS scores regardless of whether they controlled for age, sex and years of education. The expression levels of chr19:17883266–17,883,550 + were positively correlated with the PANSS scores after controlling for age, sex and years of education. There were no correlations between the PANSS scores and gene expression levels of has-miR-20b-5p, hsa-miR-502-3p, ANKH, C5orf24 (Table 2).

Table 2.

Correlation analysis between circRNAs, miRNAs, mRNAs and PANSS scores

	Correlation Coefficient	p-value	Adjusted Correlation Coefficient	Adjuested p-value
chr15:93496587–93,499,879+	0.370	0.042*	0.442	0.018*
chr7:40037093–40,087,476+	0.590	0.000*	0.604	0.001*
chr19:17883266–17,883,550+	0.350	0.052	0.385	0.043*
hsa-miR-20b-5p	0.110	0.560	0.073	0.713
hsa-miR-22-3p	-0.480	0.006*	-0.457	0.015*
hsa-miR-502-3p	0.032	0.860	0.006	0.974
ANKH	0.099	0.600	0.012	0.950
C5orf24	0.210	0.260	0.276	0.155
B4GALT5	0.390	0.029*	0.511	0.005*

Note: Adjusted Correlation Coefficient and Adjusted P-value indicates the Correlation Coefficient and P-value controlled for gender, age, and years of education

Discussion

It has been reported that the contents of EVs, such as miRNAs, circRNAs, and metabolites, have the potential to be used as biomarkers for the diagnosis of SZ. Tan et al. detected circRNA expression in plasma of 5 SZ patients and 5 HC patients by high-throughput sequencing, and found 44 DEcircRNAs, comprising 38 up-regulated and 6 down-regulated. In addition, six SZ patients and six HC patients were used for quantitative real-time PCR validation of eight circRNAs with the high fold change and enriched levels, four of which were consistent with the results of high-throughput sequencing. The GO and KEGG analyses suggest that DEcircRNAs play a potential role in pathogenesis, especially in metabolic processes, stress responses, and histone ubiquitination [8]. Du et al. performed the first genome-wide miRNA expression profile analysis of serum-derived EVs from 49 patients with first-onset, drug-free SZ and 46 control patients. Several differently expressed blood miRNAs in EVs, including hsa-miR-206 and hsa-miR-144-3p, were subsequently validated using quantitative real-time PCR in 100 SZ patients (comprising 57 first-onset drug-free patients and 43 chronically treated patients) and 100 HC subjects. The expression of downstream mRNA BDNF of hsa-miR-206 was also shown to be decreased in peripheral blood. These results support the notion that miRNAs in EVs play an important role in the pathophysiology of SZ and are promising biomarkers for SZ [9]. In addition, the levels of proteins [10] and metabolites [11] in peripheral blood EVs from SZ have also been reported, suggesting their potential as biomarkers. However, the mRNAs and circRNA-miRNA-mRNA networks in SZ EVs have not yet been reported. In this study, we examined the expression levels of circRNAs, miRNAs and mRNAs in plasma EVs derived from 10 FOS and 10 HC subjects, and we found that there were 26,194 DEcircRNAs, 22 DEmiRNAs, 2637 DEmRNAs (Fig. 2). We compared the DEcircRNAs and miRNAs found in this study with those reported in the studies, and found relatively little overlap. This may be due to differences in subjects (FOS VS SZ) or samples (plasma VS serum). We constructed circRNA-miRNA-mRNA networks in plasma EVs of SZ, which included 8 down-regulated circRNAs, 5 up-regulated miRNAs, 14 down-regulated mRNAs and 1 up-regulated circRNA, 1 down-regulated miRNA, 2 up-regulated mRNAs (Fig. 3i-j).Among 9 circRNAs in the circRNA-miRNA-mRNA network, 3 circRNAs were included in the circBase database, namely chr15:93496587–93,499,879+, chr7:40037093–40,087,476 + and chr19:17883266–17,883,550+, with the circRNA ID being hsa_circ_0036984, hsa_circ_0005485 and hsa_circ_0050042, respectively (Fig. S3). We expanded the sample size (31HC and 31 FOS) and verified the expression level of 3 circRNA-miRNA-mRNA networks (chr15:93496587–93499879+ -- hsa-miR-20b-5p – ANKH, chr7:40037093–40087476+ -- hsa-miR-22-3p–C5orf24, and chr19:17883266–17883550+ -- hsa-miR-502-3p -- B4GALT5). The expression trend was consistent with that observed in the high-throughput sequencing (Table 1).

EVs were implicated in several pathogenesis associated with SZ, such as neurodevelopment, synaptic function, neuroinflammation, and mitochondrial dysfunction [5, 13]. Transplantation of blood exosomes from patients with SZ into mice leads to behavioral abnormalities in mice and induces dysregulation of mRNA expression in the prefrontal cortex and hippocampus, with differentially expressed mRNA enriched in synaptic transmission, neurodevelopment pathways that are closely related to SZ [14]. The increase of miR-137 and a decrease of COX6A2 was observed in the peripheral blood exosomes of early psychosis patients which lead to the decrease of mitochondrial autophagy markers and the accumulation of damaged mitochondria, further exacerbating oxidative stress and small albumin interneurons (PVI) damage, suggesting that exosomes may be involved in the process of SZ by regulating mitochondrial function [15]. Exosomal glial fibrillary acid protein(GFAP) concentration was significantly higherin plasma obtained from SZ patients, and elevated GFAP expression was associated with increased astrocyte activation and inflammation in the CNS [10]. In this study, the DEmRNAs were enriched in neuron growth and development, differentiation, axonogenesis, dendritic spine development, synapse organization et al. processes (Fig. 2g), and in KEGG pathway analysis, the DEmRNAs were enriched in apoptosis, axon guidance and TNF signaling pathways (Fig. 2h) indicating the important role of EVs in SZ. In 3 circRNA, 3 miRNA and 3 mRNA whose expression levels were verified by quantitative real-time PCR in expanded sample size (Table 1), the functions of 3 circRNAs (chr15:93496587–93499879+, chr7:40037093–40087476+, and chr19:17883266–17883550+) have not been reported, the miRNAs and mRNAs are associated with neuronal growth and development and neuropsychiatric diseases. hsa-miR-20b-5p is involved in Alzheimer disease pathways and neuronal function [16]. hsa-miR-502-3p in plasma could distinguish early stage Alzheimer disease patients from non-demented subjects [17]. has-miR-22-3p in peripheral blood has been suggested to be a biomarker for SZ [18]. Regarding mRNAs, mutations in ANKH have been shown to cause at least two inherited diseases, namely familial calcium pyrophosphate deposition disease (CPPD), an autosomal dominant crystal-associated arthropathy, and cranial metaphyseal dysplasia (CMD) [19]. B4GALT5 is proven to be essential for the neuronal generation and myelin formation [20]. C5orf24 has been linked to PTSD and Parkinson’s disease [21, 22]. Above all, none of these circRNAs, miRNAs and mRNAs, except for has-miR-22-3p, have been reported to have a direct relationship with SZ. Our results indicated that these circRNAs, miRNAs, and mRNAs in EVs may play important role in SZ pathogenesis.

In addition, pathway studio (https://mammalcedfx.pathwaystudio.com) was used to analyze the relationship among 3 mRNAs and EVs exosomes and SZ. ANKH and B4GALT5 act on EVs exosomes through multiple small molecules and proteins, and EVs exosomes are associated with SZ through multiple proteins. However, there is no clear association between C5orf24 and EVs exosomes (Fig. S4). The results of this study provide an association among C5orf24 and EVs exosomes and SZ (Table 1).

Next, we performed ROC curve on circRNAs, miRNAs, and mRNAs. The ROC of ANKH, C5orf24, and B4GALT5 in the peripheral blood and EVs of SZ were created, and we found that ANKH, C5orf24, and B4GALT5 in peripheral blood could not be used as diagnostic markers (Fig. 4g-i). However, the AUC values of ANKH, C5orf24, and B4GALT5 in EVs were higher than those in the peripheral blood. Compared with peripheral blood mRNAs, mRNAs in EVs can be used as a biomarkers for SZ (Fig. 4a-c). Li Min et al. reported that the diagnostic efficacy of EVs miRNA was superior to that of free plasma miRNA in the early colon [23]. These results indicate that EVs contents have better diagnostic performances than those of plasma. In addition, the AUC value of the circRNA-miRNA-mRNA networks was higher than that of circRNAs, miRNAs and mRNAs alone (Fig. 4d-f), suggesting that multimolecular networks are better candidates for diagnostic markers than individual molecules. In addition, the gene expression levels of chr15:93496587–93,499,879+, chr7:40037093–40,087,476+, chr19:17883266–17,883,550+, hsa-miR-22-3p, B4GALT5 were correlated with PANSS scores (Table 2), implying the important role of genes derived from EVs in SZ.

There are some limitations to this study. First, when matching the sample, we only considered gender, age and education level, and did not include other factors such as childhood experience and social support, BMI and inflammatory response, lifestyle, medication history (in chronic patients), or comorbid conditions. In future studies, we will control these variables to strengthen the validity of the findings. Second, an expanded population sample (31 FOS vs. 31HC) was used to verify the expression level of the circRNA-miRNA-mRNA network. However, the size of the sample size is limited. In future studies, we will expand the sample size and collect SZ patients with different levels of illness to verify the expression levels and diagnostic efficacy of circRNAs, miRNAs, and mRNAs in plasma EVs of SZ with different disease levels. At the same time, we will continue to pay attention to the research on EVs in the SZ of other ethnic groups, and verify the expression of candidate genes in this study to test the generalization of the model’s applicability. In additon, this study is cross-sectional. In the future, we will conduct longitudinal studies to determine whether these biomarkers can predict disease progression or treatment response. Third, the mechanism of action of circRNA-miRNA-mRNA in plasma EVs in SZ needs to be further explored. In future studies, the binding and regulatory effects of circRNAs on miRNAs and miRNAs on mRNAs in the ceRNA network need to be further verified, and their regulation function in SZ needed to be elucidated. At last, more alternative methods should be used to confirm the findings in this study, and multi-omics studies should be applied to explore the molecular mechanisms by which EVs regulated SZ and improve diagnostic accuracy of EVs as biomarkers.

In conclusion, we constructed a circRNA-miRNA-mRNA networks in plasma EVs derived from FOS patients and HC subjects; expanded the sample size to verify the expression of 3 circRNAs, 3 miRNAs, 3 mRNAs; and examined the diagnostic performances, and we found that the circRNA-miRNA-mRNA network has the potential to be used as a biomarker for FOS.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Abbreviations

SZ

Schizophrenia

EVs

Extracellular vesicles

FOS

First-onset schizophrenia

HC

Healthy control

ROC

Receiver operating characteristic

DE

Differentially expressed

qPCR

Quantitative real-time -PCR

SE

Standard error

AUC

Area under the ROC curve

GEO

Gene expression omnibus

Author contributions

Xinzhe Du, Xinrong Li, Yao Gao, Yong Xu and Sha Liu design the study and wrote the manuscirpt. Xinzhe Du, Xiaodong Hu, Long Cheng, Xiaohua Cao and Zhiyong Ren recruited the subjects. Xinzhe Du, Wei Hu, Yao Gao, Junxia Li, Xiao Wang, Wentao Zhao and Xiaohua Cao analyzed the data. Wei Hu, Hongbao Cao and Yu Zhang revised the manuscript.

Funding

This work was supported by the National Natural Science Foundation of China (82271546 to Sha Liu and 82371511 to Yong Xu); the Fundamental Research Program of Shanxi Province (202203021212038 to Xinzhe Du); Shanxi Medical University School-level Doctoral Initiation Fund Project (XD 1904); Special Fund for Science and Technology Innovation Teams of Shanxi Province (202304051001049 and 202204051001027); Fund Program for the Scientific Activities of Selected Returned Overseas Professionals in Shanxi Province (20240041); Shanxi Pronvince Higher Education “Billion Project” Science and Technology Guidance Project (BYJL062); Science and technology innovation project of higher education in Shanxi Province (2020L0204).

Data availability

The datasets used during the current study are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

This study was approved by the Research Ethics Committee of the Research Ethics Committee of the First Hospital of Shanxi Medical University in accordance with the Declaration of Helsinki (2019-K039), and Consent to Participate Declaration were obtained from all participants was informed.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.