Human brain disorders, including neurodegenerative and psychiatric disorders (Naz and Siddique, 2020), are the leading cause of disability worldwide, imposing a severe burden on families and society. However, the current diagnostic criteria for these disorders are all based on evaluating clinical symptoms, making it crucial to find objective biological indicators. Extensive research on the pathological mechanisms of neurodegenerative disorders has mainly focused on amyloidosis, pathological accumulation of tau protein, and aggregation of α-synuclein (Dugger and Dickson, 2017). Presently, neuropathological assessment during autopsy remains the sole definitive method for identifying disease types. Neuroimaging studies have revealed structural abnormalities in the brains of individuals with psychiatric disorders (Opel et al., 2020). Establishing a less invasive diagnostic method capable of capturing pathological brain changes would greatly aid in clinical molecular classification and drug discovery.

Extracellular vesicles (EVs) are nanoscale membrane vesicles secreted by cells and widely distributed in various biological fluids. These EVs can be categorized into several subtypes: (1) exosomes (~50–150 nm in diameter), (2) retroviruses (80–150 nm), (3) microvesicles, ectosomes, or microparticles (~100–1,000 nm), and (4) apoptotic EVs and apoptotic bodies (~100–5,000 nm) (Mathieu et al., 2019). EVs harbor a variety of biomolecules, including protein, miRNA, mRNA, and more, which can be transported between different cells, tissues, or organs and participate in various physiological processes. In recent years, studies have found that EVs’ abnormal secretion and cargos are closely related to the pathogenesis of central nervous system (CNS) diseases. What’s more, because of their intercellular communication abilities and the capacity to cross the blood–brain barrier (BBB) (Banks et al., 2020; Morales-Prieto et al., 2022), EVs have emerged as potential tools for breakthroughs in the clinical diagnosis of brain disorders. Recent studies have shown that EVs may play a role in the abnormal accumulation or degeneration of proteins in the brains of patients with neurodegenerative disorders through secretion and transmission between cells (Yuyama et al., 2008; Emmanouilidou et al., 2010; Saman et al., 2012; Nonaka et al., 2013). In the case of psychiatric disorders, transplantation of serum-derived EVs from schizophrenia (SCZ) patients into mice resulted in SCZ-like behavioral and molecular phenotypes (Du et al., 2021). Moreover, peripheral injection of blood-derived EVs from healthy controls (HC) into chronic unpredictable mild stress (CUMS) mice alleviated depression-like behavior (Wei et al., 2020).

Analyzing changes in biomolecules within CNS-derived EVs isolated from peripheral tissues may capture alterations occurring in the brain. The utilization of EVs has the potential to facilitate minimally invasive liquid biopsies for clinical diagnosis and tracking of disease progression. However, due to the heterogeneity of EVs and their diverse origins, measuring molecular changes in the brain using total EVs in the bloodstream poses challenges. Purifying specific EVs derived from the brain may circumvent these limitations and enable the investigation of their cargos as biomarkers of brain disorders. Recent studies have reported the existence of neuron-derived EVs (NDEs), astrocyte-derived EVs (ADEs), and oligodendrocyte-derived EVs (ODEs) in the brain (Zou et al., 2022).

In this review, we focus on these three sources of EVs and compile a comprehensive summary of the historical background and biomarker studies conducted in major neurodegenerative [Alzheimer’s disease (AD) and Parkinson’s disease (PD)] and psychiatric disorders [SCZ, bipolar disorder (BD), depression, and autism spectrum disorder (ASD)]. Furthermore, we discuss the limitations inherent in existing studies and propose future directions for research.

Harding and Stahl (1983) and Pan and Johnstone (1983) made the initial discovery of small vesicles (~50 nm) secreted from reticulocytes, which they named “exosomes.” These exosomes were found to play an important role in the reticulocyte transferrin receptor cycle process. Later, Zitvogel et al. (1998) and Théry et al. (2001) found that dendritic cells could secrete exosomes and activate the T cell immune response. Fevrier et al. (2004) discovered the propagation of the first pathological protein, prion protein (PrP), through exosomes. These findings implied the potential relevance of exosomes in both normal and pathological brain physiology. It also raised the possibility of using exosomes to explore disease pathology and develop diagnostic tools for clinical applications. In accordance with the Minimal information for studies of extracellular vesicles 2018 (MISEV2018) recommendations and considering the imprecise methods for isolating pure exosomes, we will use the term “EVs” to replace “exosomes” in the subsequent description.

Fauré et al. (2006) found that neurons and astrocytes in the rat cortical primary cultures could secrete EVs. Through the LC–MS/MS and western blot analyses, they identified specific proteins present in EVs secreted by astrocytes (highly expressing glutamine aspartate transporter, GLAST) and neurons (highly expressing L1 cell adhesion molecule, L1CAM) in the brain. In 2007, it was confirmed that oligodendrocytes also secrete EVs, which carry abundant proteolipid protein (PLP) and 2’3’-cyclic-nucleotide phosphodiesterase (CNP). These findings laid the groundwork for subsequent studies on NDEs, ADEs, and ODEs (Krämer-Albers et al., 2007).

CNS-derived EVs have been studied in clinical blood samples since 2014. Shi et al. (2014) and Fiandaca et al. (2015) independently established an immunoaffinity capturing protocol to isolate L1CAM-containing EVs (referred to as NDEs) from plasma. L1CAM is a cell adhesion molecule highly expressed in neurons. Two years later, Goetzl et al. (2016b) made progress in ADEs using mouse anti-human GLAST biotinylated antibodies. The study examined the association between cargo proteins of plasma ADEs and AD pathology. The first clinical study on ODEs was published in 2019 (Ohmichi et al., 2019). Since then, more and more researchers have carried out clinical research on these special EVs, as described in the following sections.

To evaluate the potential role of molecules from NDEs, ADEs, and ODEs in clinical diagnosis, we conducted a systematic review of published studies that detected cargos in these specific EVs in the blood of patients with brain disorders. PubMed and Web of Science were searched with the following search builder: ((extracellular vesicle) OR (extracellular vesicles) OR (exosome) OR (exosomes) OR (exosomal) OR (exomeres) OR (microparticles) OR (apoptotic bodies) OR (retroviruses)) AND ((neurodegenerative disease) OR (neuropsychiatric disorder) OR (mental disorder) OR (psychiatry disorder) OR (psychotic disorder) OR (psychiatric disorder) OR (neurodevelopmental disorder) OR (neurological disease) OR (neurological disorder) OR (psychosis) OR (#DISORDER#)). #DISORDER# is one of the six major brain disorders: AD, PD, SCZ, BD, depression, and ASD. The inclusion criteria were the following: (1) published in English before April 2022; (2) assessed EVs derived from neuron/astrocyte/oligodendrocyte; and (3) focused on clinical diagnosis. Studies on animal and cell culture models were excluded. The literature selection followed the standard of Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) (Liberati et al., 2009; Figure 1).

A total of 57 publications that met our inclusion and exclusion criteria were included in our review (Supplementary Table S1). These studies covered AD (number of studies, n = 30), PD (n = 16), SCZ (n = 4), BD (n = 1), depression (n = 5), and cross-disorder studies (AD and PD, n = 1). Notably, there are no NDEs/ADEs/ODEs-related studies on ASD or other brain disorders. Among the included studies, 54 of the 57 papers examined NDEs, while only 10 and 3 studies have investigated ADEs and ODEs, respectively. Regarding the types of molecules analyzed in the included literature, only five papers reported changes in miRNAs within EVs, while the rest explored protein molecules. Two studies did not analyze the cargos of EVs. To evaluate the reproducibility of the reported changes in molecules derived from these specific EVs in the cases compared to controls, we examined the consistency of the reported changes in each disorder by the presence of significant results (p < 0.05) consistent in more than one independent study. We summarized those molecules explored in two or more studies in Table 1, including the related study subjects, sample sizes, and the direction of change. It should be emphasized that, based on the currently included studies, it is not possible to assess the reproducibility of the factors explored in the ADEs/ODEs related studies, as none of these factors were explored in more than two related studies. A comprehensive description of the results from the included literature can be found in Supplementary Table S1.

Despite the extensive biomarker discovery research on NDEs, ADEs, and ODEs, the isolation methods for these EVs have not been thoroughly evaluated. All current methods are based on antibodies, making the specificity of both the antibodies and the antigens crucial. The methods currently used to isolate CNS-derived EVs are summarized in Table 2.

L1CAM is the most commonly used antigen for capturing NDEs. It is a member of the cell adhesion molecules mainly expressed in CNS and has been identified as a marker on the surface of EVs specifically derived from neurons (Fauré et al., 2006; Shi et al., 2014). According to the sorted brain cell RNAseq database,² L1CAM is specifically expressed in neurons in the brain (Zhang et al., 2016). However, L1CAM is not only expressed in the brain but also the intestine, monocytes, testis,³ and some cancer cells (Ganesh et al., 2020). Additionally, a recent study found that L1CAM has soluble fragments in the blood that are not associated with EVs (Norman et al., 2021). These observations have raised concerns about using L1CAM as a specific marker for NDEs.

Besides L1CAM, Neural Cell Adhesion Molecule 1 (NCAM1) (Fiandaca et al., 2015; Goetzl et al., 2015a) and SNAP25 (Ohmichi et al., 2019) have also been used for the capture of NDEs. However, NCAM1 is not neuron-specific and is also highly expressed in oligodendrocytes (Sharma et al., 2015; Zhang et al., 2016) and heart muscle (Ackermann et al., 2017). On the other hand, SNAP25 exhibits good cell type and tissue specificity (Hwang and Lee, 2003), but its abundance in EVs has not been fully validated. To capture SNAP25-binding EVs in serum-free conditions, anti-SNAP25 antibodies are used, followed by the quantification of SNAP25+ EVs using anti-CD81 since CD81 is a marker of EVs.

Most methods for capturing ADEs utilize anti-GLAST antibodies, as GLAST is considered to be predominantly expressed in astrocytes. However, GLAST is also moderately expressed in the heart (Xie et al., 2022), ovary, and placenta (see text footnote 3). Aquaporin 4 (AQP4) and Glial Fibrillary Acidic Protein (GFAP), two classical astrocyte markers, have been tested for capturing ADEs by flow cytometry. Although AQP4 and GFAP exhibit good cell type specificity in the brain, they are also expressed in other organs, such as the lungs (Jaskiewicz et al., 2022) and kidneys (Buniatian et al., 1998).

The isolation of ODEs has been reported in three studies, but the justification for the choice of antibodies used in these studies is unclear. Each study employed distinct antibodies to capture ODEs, including anti-oligodendrocyte-myelin glycoprotein (OMG) (Ohmichi et al., 2019), anti-myelin oligodendrocyte glycoprotein (MOG) (Dutta et al., 2021), and CNP (Yu et al., 2020). CNP is a commonly used marker for mature oligodendrocytes and is mainly expressed in the CNS (Sprinkle, 1989; Verrier et al., 2013). CNP and MOG are both specifically highly expressed in oligodendrocytes, while OMG has been found to be expressed in astrocytes as well (Zhang et al., 2016).

The specificity of these antibodies needs to be further evaluated, as the concern regarding their specificity is not addressed in most of these papers. It is necessary to trace the main source of these EVs, such as the genetic source tracking research conducted for human urinary exosomes (Zhu et al., 2021). A comprehensive comparison of blood CNS-derived EVs with brain EVs or CSF EVs would be valuable. Also, the cargos of EVs can provide insights into their origin. The specificity of L1CAM+ EVs was evaluated (Goetzl et al., 2016b; Mustapic et al., 2017), and they were found to be highly enriched for neural proteins such as neurofilament light chain (NF-Lch) and neuron-specific enolase (NSE). The molecular signatures of blood total EVs, L1CAM+ EVs (NDEs), and epithelial cell adhesion molecule (EpCAM) + EVs (tumor-derived exosomes, another subtype in blood EVs) were compared by protein antibody array and showed distinct protein profiles. Another study reported that several biomarkers, total tau and p-T181-tau, showed a positive correlation in L1CAM+ EVs and CSF, while Aβ42 showed a negative correlation (Jia et al., 2019). Although NCAM+ EVs have not been found to contain these neural-specific cargos, Fiandaca et al. (2015) found that total tau, p-S396-tau, p-T181-tau, and Aβ42 have a similar concentration in NCAM+ EVs as they do in L1CAM+ EVs. This suggests that L1CAM+ EVs and NCAM+ EVs may have the same origin. GLAST+ EVs were found to contain astrocyte-specific proteins GFAP and GluSyn. Taken together, although L1CAM and GLAST are not specific to the brain, L1CAM+ EVs and GLAST+ EVs are enriched in neuron- and astrocyte-specific proteins, respectively. Other antigen-captured EVs have not been evaluated for their cell types or brain specificity.

The most commonly employed method for capturing CNS-derived EVs is precipitation coupled with antibody-based immunoaffinity capture, although some researchers have utilized direct antibody immunoaffinity isolation to capture L1CAM+ EVs and CNP+ EVs. While the direct immunoaffinity method is straightforward, there is a concern that soluble fragments may compromise the purity of EVs. Another approach involves centrifugation (not ultra-centrifugation) to eliminate cell debris and then couple it with antibody immunoaffinity. This procedure also poses the same concern regarding soluble fragments. The precipitation step only captures EVs and polymeric substances (Zarovni et al., 2015), which can help to remove the soluble antigen. In sum, further development is necessary to determine the source of CNS-derived EVs and assess the similarity in content between CNS-derived EVs in blood and EVs in CSF.

The methods used in these studies to identify and characterize EVs included transmission electron microscopy (TEM), nanoparticle tracking analysis (NTA), bicinchoninic acid (BCA) protein quantification, Western Blotting, tunable resistive pulse-sensing (TRPS) analysis, and other. To ensure the purity and integrity of isolated EVs, the International Society for Extracellular Vesicles (ISEV) has issued guidelines for identifying exosomes (Théry et al., 2018). However, many of the studies described in this review did not adhere to these guidelines adequately, making it challenging to determine whether there is an impurity in the isolated EVs that could interfere with the present findings. Therefore, in future studies, researchers should focus on selecting antibodies targeting specific proteins, evaluating the source, and improving the purity of the isolated CNS-derived EVs. This may help establish the association between EVs and the CNS and enhance their clinical diagnostic value.

In this review, we summarized the research progress in the diagnosis of six major brain disorders using the study of NDEs, ADEs, and ODEs. Most of these studies were conducted on neurodegenerative disorders, and many of these CNS-derived EVs’ cargos were found to be significantly different between cases and controls. It is worth mentioning that some of the detected molecules in NDEs, including Aβ42, total tau, p-T181-tau, p-S396-tau, NRGN, synaptotagmin, GAP43, SNAP25, cathepsin D, REST and α-synuclein, have been reproducible across different studies, which is of great significance for the clinical diagnosis of diseases. In addition, these peripherally detected changes in EVs represent biological alterations occurring in the brain in a minimally invasive manner, making CNS-derived EVs ideal diagnostic and therapeutic tools. These findings may drive the development of accurate clinical diagnosis for brain disorders. However, the features of CNS-derived EVs and their isolation methods need to be improved, particularly their specificity.

The specificity of many markers of the CNS-derived EVs has been questioned, and it remains to be carefully investigated whether these EVs and their cargos accurately reflect brain conditions. It is necessary to measure and compare the expression levels of the selective specific antigens in EVs to those from the relevant tissues and cell types, either in animal models or in postmortem or surgery tissues. The development and application of EVs tracer techniques also help to evaluate the specific EVs trajectories precisely. The detailed benchmark for CNS-derived EVs may facilitate clinical application and explain current inconsistent results to a certain extent.

Except that, most current studies of CNS-derived EVs only detected specific biomolecules of interest, and these results have yet to undergo validation by other research teams. A limited number of papers have used genome-wide approaches to investigate the expression pattern of CNS-derived EVs (Serpente et al., 2020; Zou et al., 2020; Anastasi et al., 2021; Arioz et al., 2021; Saeedi et al., 2021; Zhong et al., 2021). These studies have employed omics screening to predict disease status and have revealed new potential protein or miRNA biomarkers. For diseases that lack clear pathological features, such as psychiatric disorders, exploration at the omics level is necessary to identify molecular changes that can aid in clinical diagnosis or molecular typing. Therefore, it is essential to conduct additional omics-related research to discover alterations at the molecular level in CNS-derived EVs for biomarker identification. Machine learning algorithms with large omics data may help uncover better biomarkers than those that have been previously tested. While the majority of current studies concentrate on proteins, it is worth mentioning that RNA levels are also important for disease diagnosis and cannot be ignored in future studies.

In addition, compared with the disease diagnosis, early detection may be clinically more important. In the current literature collected, only a few studies related to AD have considered the pattern of biomarkers in pre-clinical samples. Therefore, researchers should carry out more pre-clinical studies on related diseases or explore the biomarkers identified in the disease’s early stages.

To summarize, it is crucial to validate the findings of specific biomolecules in CNS-derived EVs through collaboration with other research teams. Employing genome-wide approaches, considering RNA levels, conducting pre-clinical studies, and utilizing machine learning algorithms can significantly enhance the discovery of robust and reliable biomarkers for disease diagnosis and classification in CNS-derived EVs.

XW, KL, and HY did the literature collection and research results collation. XW and KL conceived the manuscript. XW did the manuscript writing. KL and CL helped with manuscript writing and finalized the manuscript. All authors contributed to the article and approved the submitted version.

This work was supported by the Startup Foundation for Introducing Talent of Central South University (202044006).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.