EVs are grouped by biogenesis route. Exosomes (40–160 nm) are endosomal-derived. are formed by the inward budding of the plasma membrane to produce early endosomes and MVBs. As MVBs fuse with the plasma membrane, ILVs are released from MVBs as fully developed exosomes. Conversely, microvesicles (50–1000 nm) or ectosomes, are produced by direct outward budding of the plasma membrane (Figure 1) (20, 23). Exosomes and microvesicles may have some overlap in their size range but have different origins, structures, and surface markers (24–26). Exosomes express tetraspanins (CD9, CD63, CD81), ALIX, TSG101, and heat shock proteins (Hsp60, Hsp90) as established identifiers (27, 28). Annexin A1 is emerging as a candidate distinguishing marker for microvesicles (29). Nevertheless, a universal classification system of EVs has not yet been established and thus more studies are required to identify definitive subtype-specific markers.

EVs play key roles in the transfer of cargo between cells, either as direct delivery vehicles for functional cargo that is transferred to a recipient cell or via activation of surface receptors on recipient cells that initiate intracellular signaling (30, 31). The transported cargo is proteins (surface/cytosolic), DNA, mRNA, microRNAs, amino acids, and metabolites (32, 33). These molecules take part in some different processes like immune system adjustment, inflammatory reaction, angiogenesis, and tissue repairments (34–38). Among these exosomal components, exosomal microRNAs are of particular significance as regulators and possible biomarkers for autoimmune and inflammatory diseases (35, 39, 40). EVs show advantageous properties for biomedical applications, such as structural stability, low immunogenicity, and high biocompatibility. These characteristics support their role as a next generation vector for target drugs delivery (41, 42). Recently, studies employing more refined purification methods have called into question earlier findings about the presence of double-stranded DNA and histones in EVs. The current evidence suggests that canonical small EVs do not contain dsDNA or DNA-bound histones, which contradicts earlier assumptions (42, 43). And this also highlights the critical importance of clearly defining the specific composition of EVs in both scientific studies and clinical applications.

EVs that are released by immune cells are related to antigen presentation, T cell activation, and regulation of immune tolerance (44, 45). In addition to immune cells, EVs from nonimmune cells are also involved in immunity. Tumor-derived EVs can induce apoptosis of immune cells and inhibit T cell activation, which helps tumor cells to evade immune attack (46). Inflammatory conditions improve the immunosuppressive properties of mesenchymal stem cell-derived EVs, which control inflammation by favoring Treg differentiation and altering macrophage polarization (47–49). These observations highlight the context-dependent immunoregulatory nature of EVs, which are derived from cells and are surrounded by a microenvironment. A specific EVs’ role in immune mediated eye disease has yet to be investigated. Future research should aim for mechanistic studies with standardized protocols to investigate their potential for treatment and diagnostics.

Exosomes are nano-sized extracellular vesicles distributed through vastly complex body fluids, which makes high-yield exosome isolation challenging (50). Although ultracentrifugation is considered the “gold standard” for exosome separation owing to its high processing capacity, it has notable limitations. The primary challenge is the high content of protein aggregates and lipoproteins present in the sample. Since this method cannot completely separate exosomes from other components, these impurities significantly compromise the accurate quantification of exosomes and subsequent analyses (51). These limitations have driven the development of diverse separation techniques to meet higher exosome purification demands, including traditional methods such as ultrafiltration, size exclusion chromatography, polymer precipitation, and immunoaffinity approaches, alongside emerging methods like microfluidics, DNA aptamer affinity, fluid flow-based separation, thermophoresis, and lipid recognition separation, each with its own unique advantages and disadvantages (52).

A mechanism is the immunostimulatory activity of EVs in GO. Serum-derived exosomes from the GO and GD display higher levels of insulin-like growth factor-1(IGF-1) receptor and heat shock protein 60 (HSP60), these components relate to autoimmune reactions. Due to the lipid bilayer structure of exosomes, GD-EXO can directly fuse with peripheral blood mononuclear cells (PBMCs) or be uptaken by PBMCs through binding to Toll-like receptors (TLRs) on the cell surface via the HSP60 protein it carries. Research has shown that GD-EXO can be taken up by PBMCs from healthy donors and undergo binding with PBMCs, while increasing the proportion of CD11c⁺TLR2⁺ and CD11c⁺TLR3⁺ dendritic cells. Furthermore, antigens and pathogenic factors within the exosomes are transported into PBMCs, activating MyD88 and TRIF, leading to enhanced NF-κB p65 phosphorylation. This ultimately promotes the release of IL-6 and IL-1β, inducing inflammatory responses and contributing to the pathogennesis of GD. However, due to the relatively small number of patients with early-onset GO outpatient who were not receiving treatment, this study regrettably could not proceed with the subsequent GO-EXO functional research. Subsequently, subject to patient availability and experimental conditions permitting, more in-depth research on the role of GO patients’ exosomes in immune and inflammatory responses need to be conducted in future research (53).

Exosomal miRNAs, particularly miR-144-3p, enhance fibrotic and inflammatory responses in OFs, suggesting shared pathological pathways in fibrosis induction across different etiologies. Li et al. demonstrated that miR-144-3p was significantly upregulated in both Pla-Exos and PBMCs during the active phase of GO. Subsequently, miR-144-3p mimics and control mimics were transfected into Thy-1+ OFs, followed by assessment of relevant target mRNAs (inflammatory molecules, HAS1, HAS2, HA) and cell proliferation. Results demonstrated that Pla-Exos derived from active-phase GO patients activated Thy-1+ fibroblasts, inducing expression of inflammatory cytokines and pro-fibrotic markers. Furthermore, these exosomes inhibited proliferation of Thy-1+ OFs. Compared to normal human Pla-Exos, active-phase GO Pla-Exos significantly upregulated the expression and secretion of pro-inflammatory factors (e.g., IL-1β, IL-6, TNF-α, CXCL2, and RANTES) in Thy-1+ OFs. As key mediators of inflammation and fibrosis, the upregulation of these cytokines and chemokines promotes the recruitment of immune cells to orbital connective tissue and enhances hyaluronic acid accumulation in the extracellular matrix, thereby perpetuating inflammation and fibrosis in GO (57).

Given the variable responsiveness of GO patients to intravenous glucocorticoid (ivGC) therapy, studies have explored exosomes as biomarkers to predict the treatment outcomes. Analysis revealed that miR-885-3p as a potential biomarker indicating therapeutic response. In GO patients who showed significant symptom improvement after IVGC treatment, plasma-derived exosomes (SI-exo) were taken up by OFs, transferring miR-885-3p from the EVs into the OFs. Within OFs, miR-885-3p directly binds the 3’-UTR region of the AKT2 gene, achieving targeted suppression of AKT2 and inhibits the activation and phosphorylation levels of AKT. This, in turn, dampened activation of the pro-inflammatory AKT/NF-κB signaling pathway. As this axis was suppressed, glucocorticoid receptor (GR) expression GR-dependent glucocorticoid response element (GRE) activity increased, while inflammatory molecules including IL-1 and ICAM-1 were reduced. Together, these changes appear to enhance glucocorticoid responsiveness in the peripheral vascular layer, which may explain the improved clinical response to IVGC therapy. On this basis, miR-885-3p has been proposed as a potential biomarker for predicting treatment efficacy. However, the current evidence is largely derived from small, single-center studies and lacks robust validation in larger cohorts. For instance, the aforementioned study included only 17 patients, limiting statistical power and generalizability. In addition, the absence of key performance measures, such as sensitivity, specificity, and ROC curves, makes it difficult to judge diagnostic performance objectively. Future work should combine machine-learning based biomarker screening with ROC-based validation, and address potential confounders using multivariable regression (58–60). Reproducibility and clinical translation are also constrained by the lack of standardized protocols for exosome isolation and quantification (55). Therefore, despite encouraging mechanistic insights, substantial obstacles remain before such biomarkers can be adopted in practice. Large-scale, multicenter studies with standardized operating procedures and rigorous performance evaluation will be essential to move these candidates toward clinical use.

Tear derived exosomes from GO patients have increased concentrations and contain pro-inflammatory proteins such as caspase-3, complement C4A and apolipoprotein A-IV. These proteins indicate pathological changes of orbital tissues, which implies that tear EVs can be used as a noninvasive indicator for the disease monitoring. In addition, the increased concentrations of IL-18 and IL-1 in tear EVs suggest that they may be involved in local immune dysregulation and apoptosis. In Shi’s study, results demonstrated that patients with active/severe GO (defined as CAS ≥ 3, with cohort CAS values concentrated within a narrow range of approximately 3.6 ± 0.3) exhibited significantly upregulated IL-1 and IL-18 in tear exosomes, consistently validated in serum. This suggests that these inflammatory cytokine loads align with a “high-activity inflammatory phenotype”. However, this study did not perform direct statistical correlation between molecular levels and CAS on an individual case basis: most samples were pooled, and the cohort primarily comprised active-phase patients with high CAS, making it difficult to assess whether these factors vary along the CAS gradient. Therefore, to establish robust evidence linking ‘molecular load directly to clinical phenotype’ as a non-invasive biomarker, future studies should quantify EV-IL-1/IL-18 in individual patients without pooling samples. Spearman correlation and ROC analysis should then be performed to evaluate their discriminatory performance in distinguishing active/inactive phases (or different CAS grades) (56).

Exosomes derived from the tears of patients suffering from GO cause OFs to produce a variety of inflammatory mediators. Proteomic analysis showed upregulation of vitamin D binding protein, chitinase 3-like protein 1, and matrix metalloproteinase-9, implicating EVs in tissue remodeling and immune cell recruitment (54). Together they point to many different ways that EVs are involved in GO. Their participation covers up the immune system activation, fibrotic remodeling as well as therapeutic responses modification (Figure 2). The recognition of particular exosomal miRNAs and proteins as potential biomarkers gives a promising approach to improving the early detection, following the course of disease and tailoring individual treatment options for GO patients.

Although there is great progress, there are some significant problems that need to be paid attention to in the studies of EVs in GO. Firstly, EV populations are heterogeneous, while the inconsistency of isolation and purification procedures makes it difficult to achieve reproducibility of results and comparisons across studies. Lots of studies use bulk plasma or tear-derived EVs without strict subtype classification, which might make it hard to spot disease-linked EV subsets. Second, the functional validation of identified EV cargoes, e.g., miRNAs or proteins, often lacks mechanistic depth. Some miRNAs like miR-144-3p are regarded as biomarkers or potential therapeutics. But direct causal connections to clinical results stay tentative and want in vivo confirmation. Third, the vast majority of existing data are derived from small, single-center studies with a limited number of patients, and few studies have long-term follow-up. Lastly, only a handful of studies have looked at whether the EV profiles are different between active vs. inactive GO stages and between treatment responders and non-responders. To truly harness the potential of EVs for both diagnosis and therapy in GO, these methodological and clinical gaps must be addressed.

Exosomes isolated from the serum of patients with refractory GD can activate peripheral immune cells. These exosomes induce PBMCs to release more pro-inflammatory cytokines, indicating heightened inflammatory responses. Analysis of microRNA within these exosomes shows increased amounts of specific miRNAs, such as let-7g-3p and miR-339-5p. These microRNAs can also activate the NF-kB, IL-17 signals necessary for the autoimmunity and inflammation of GD (61).

Proteomic analysis by data-independent acquisition (DIA) method found 208 differentially expressed proteins in exosomes of GD patients, including MAP1S, VAMP8 and CXCL7, involved in immune activation and apoptosis are upregulated (62). CircRNA profiling showed that substantial enrichment of hsa_circRNA_000102 in the plasma exosomes of patients with GD. The circular RNA is positively correlated to thyroid-stimulating hormone receptor antibody levels, and may affect disease progression via a circRNA-miRNA-mRNA regulatory network (63).

Thyroid follicular cells secrete exosomes with TSHR protein. These vesicles bind to the GD specific monoclonal antibody M22, acting as decoys and reducing receptor over stimulation and downstream cyclic adenosine monophosphate signaling (65). Environmental toxins (e.g., DDT) disrupt normal TSHR transport or exosome coating processes, potentially triggering autoimmune responses in susceptible individuals (66). The levels of plasma microvesicles and platelet-derived microparticles are increased in GD and Hashimoto’s thyroiditis patients. These vesicles contain inflammatory markers and keep a partial boost after treatment, suggesting their role in monitoring disease activity and immune system dysregulation (67, 68).

The levels of plasma microvesicles and platelet-derived microparticles are elevated in patients with GD and HT and remain high before and after treatment. These particles carry inflammatory markers and may indicate disease activity or underlying immune dysregulation. In addition, some exosomal miRNAs and proteins show differential expression between GD and papillary thyroid carcinoma (PTC), including miR-375-3p and miR-7-5p, suggesting potential utility in distinguishing thyroid diseases (64). Overall, the current evidences support an ideal role for EVs in GD pathogenesis rather than viewing them as byproducts of immune activation. Beyond GD, EVs have also shown diagnostic potential across multiple disease settings, highlighting their promise as biomarkers, and therapeutic targets (69). Future works should confirm these signatures in larger cohorts and develop standardized protocols to support clinical translation.