Serum sEV miRNA как биомаркеры в Myelin Oligodendrocyte Glycoprotein Antibody-Associated Beacy.

Background

Myelin oligodendrocyte glycoprotein (MOG) antibody-associated disease (MOGAD) is a central nervous system (CNS) demyelinating disease characterized by the presence of a disease-specific antibody MOG-IgG, affecting the brain, spinal cord and optic nerve [ 1 ]. It is distinguished from other CNS demyelinating diseases, such as multiple sclerosis (MS) and neuromyelitis optica spectrum disorder (NMOSD), by its distinctive clinical and pathological features [ 1 – 3 ]. In 2023, the international MOGAD panel proposed diagnostic criteria for MOGAD [ 4 ]. However, diagnosing MOGAD remains challenging, as the criteria require clinical interpretation to exclude other conditions such as MS and NMOSD and rely on heterogeneous assay results. Moreover, inconsistencies in the reproducibility of MOG-IgG detection across different assays and reduced reliability of low-positive results present additional challenges [ 4 , 5 ]. Despite these difficulties, studies on diagnostic or prognostic biomarkers in MOGAD remain relatively scarce, whereas liquid biomarkers have been extensively investigated in MS and NMOSD [ 6 , 7 ].

MicroRNAs (miRNAs), small non-coding RNAs approximately 22 nucleotides in length, play critical roles in gene regulation [ 8 , 9 ]. Small extracellular vesicles (sEVs) less than 200 nm in size, have emerged as significant areas of interest and sEVs encapsulating miRNA, are shed from all cellular types and are abundant in various biological fluids, thus serving as accessible biomarkers [ 9 ]. The protective nature of EVs shields miRNAs from enzymatic degradation in the bloodstream, establishing EV-derived miRNAs as potential biomarkers for various diseases [ 9 , 10 ]. The diagnostic and communicative capabilities of miRNAs extend across a spectrum of conditions, including neurological diseases, due to their ability to traverse the blood–brain barrier within sEVs. Such transport facilitates intercellular communication, particularly between neurons and other neurological cells [ 11 , 12 ]. Recent investigations into CNS demyelinating diseases have progressively highlighted the significant roles of miRNAs. In MS, altered miRNA expression compared to healthy controls (HCs) has been documented, with certain miRNAs associated with disease status and treatment response [ 13 – 16 ]. In NMOSD, studies on miRNAs suggest that miRANs may serve as potential biomarkers in distinguishing NMOSD from MS or HCs, as well as indicators of disease status [ 17 , 18 ]. A previous study reported increased expression levels of certain miRNAs in experimental autoimmune encephalomyelitis (EAE) mice, which were subsequently examined in patients with MOGAD [ 19 ]. To date, comprehensive investigations of miRNA profiles in MOGAD, specifically in comparison to MS, NMOSD, or HCs, remain scarce.

Hence, using serum sEV miRNA analysis, we aimed to elucidate the potential role of miRNAs in [ 1 ] distinguishing MOGAD from MS, NMOSD, and healthy individuals, [ 2 ] understanding the specific pathophysiological mechanisms of MOGAD, and [ 3 ] predicting clinical outcomes through a comprehensive miRNA analysis in MOGAD.

Methods

Study Participants and Biological Samples

We prospectively evaluated and enrolled patients who visited the neurology outpatient clinic and were diagnosed with MOGAD, relapsing–remitting MS, and anti-aquaporin 4 antibody (anti-AQP4)-positive NMOSD, based on the 2023 proposed diagnostic criteria for MOGAD, the 2017 McDonald criteria for MS, the 2015 international consensus diagnostic criteria for NMOSD [ 4 , 20 , 21 ]. Anti-AQP4 was assessed using a cell-based indirect immunofluorescence assay, and MOG-IgG was detected using an in-house live cell-based immunofluorescence assay with an antihuman IgG1-Fc secondary antibody (CBA-immunofluorescence) and CBA flow cytometry analysis (CBA-FACS) provided by Eon laboratories [ 4 , 22 – 24 ]. We collected demographic and clinical data, including age at sampling, age at disease onset, sex, annualized relapse rate (ARR), type of initial clinical presentation (optic neuritis, myelitis, brainstem/cerebellum involvement, or cerebrum involvement), disease status (relapse within 30 days), Expanded Disability Status Scale (EDSS) score at the time of sampling, and treatment history prior to sampling. A total of 20 ml of peripheral venous blood was collected from the antecubital vein into serum-separating tubes from all patients. The same volume of venous blood was also obtained from 12 HCs had no history of acute or chronic disease and had not been taking any medication in the previous 3 months. Serum was obtained by centrifugation at 2500 rpm for 15 min and then stored at − 80 °C until sEV fraction extraction. The study was approved by the Institutional Review Board of Samsung Medical Center (IRB no. 2020–04–119). The study was conducted in accordance with the local legislation and institutional requirements. Written informed consent was obtained from all participants.

Isolation of sEVs from the Serum

To isolate the sEV fraction from the serum, we centrifuged the serum at 3000 g to obtain platelet-free fractions. We then centrifuged it at 10,000 g to remove the large extracellular vesicle fraction. sEV fractions were purified using the miRCURY Exosome Serum/Plasma Kit (Qiagen), following the manufacturer’s instructions with minor modifications. For miRNA analysis, fresh sEV fractions were used on the same day as purification. The sEV fractions were subsequently subjected to small RNA isolation using miRNeasy Micro Kit (Qiagen), following the manufacturer’s manual. The quality of the small RNA was assessed using the Small RNA Analysis Kit on a Bioanalyzer 2100 (Agilent Technologies, Germany).

Characterization of sEVs

We measured the protein concentration of sEVs using a bicinchoninic acid protein assay kit (Thermo Fisher Scientific). The quantity and size distribution of sEVs were determined using nanoparticle tracking analysis (Malvern). In brief, the camera level was set to 16, the slide shutter was set to 1500, and the analysis detection threshold was set to five. Three 30-s videos were recorded for analysis. For western blot analysis, protein samples were mixed with NuPAGE LDS sample buffer (Thermo Fisher Scientific). Proteins were then separated by SDS-PAGE (Bio-Rad) and transferred onto PVDF membranes (Merck Millipore). The membranes were subsequently washed with TBS-T buffer and incubated with primary antibodies including anti-Calnexin (Merck Millipore), anti-TSG101 (Santa Cruz), anti-Alix (Santa Cruz), and anti-CD63 (Santa Cruz). The membranes were then washed with TBS-T buffer again and incubated with peroxidase-conjugated anti-rabbit or anti-mouse secondary antibodies (Merck Millipore) for 1 h at room temperature. Finally, the immunoreactive protein bands were visualized using a Dyne ECL STAR Western Blot Detection Kit (Dyne Bio).

Transmission Electron Microscopy

Cryo-transmission electron microscopy (Cryo-TEM) was performed as previously described [ 25 ]. Briefly, lacey carbon grids (electron microscopy sciences) were used for sEV collection. To preserve the structural integrity of the sEVs, the grids were promptly frozen in liquid nitrogen. Following freezing, a Tecnai Twin transmission electron microscope was used to capture high-resolution images of the sEVs. The microscope was operated at a voltage of 200 kV during image acquisition.

NanoString nCounter Platform

NanoString analysis and data processing were performed by LAS Inc. (Seoul, Korea). A total of 1 ng of small RNA was analyzed using a NanoString nCounter Analysis System (NanoString Technologies, Inc. WA, USA) and with the miRNA Panel Assay Kit, following the manufacturer’s instructions. Each RNA sample was mixed with the Master Mix, and the capture probe set was added. The mixture was incubated in a thermocycler at 65 °C (Bio-Rad Laboratories Inc., Hercules, CA, USA) for 16 h (maximum hybridization time not exceeding 48 h). The samples were subsequently transferred to the preparation station (NanoString Technologies, Seattle, WA) equipped with a prepared nCounter Master Kit and a cartridge. The preparation station had 12 lanes per run, with a processing time of approximately 2.5 to 3 h. The cartridges were then transferred to a Digital Analyzer (NanoString Technologies, Seattle, WA, USA) for analysis. Finally, the cartridges were scanned using a Digital Analyzer at 555 fields of view.

Quantifying Gene Expression and Differentially Expressed Gene Analysis

To quantify gene expression values, background correction and normalization were first performed using default options with control probes in each Codeset. Subsequently, the log2-transformed count data were subjected to a t test to identify differentially expressed genes between the two selected biological conditions. A fold change greater than 1.3 and a p value less than 0.05 were considered indicative of a significant difference. The nSolver Analysis software version 4.0 was employed for background correction, normalization, and differential expression analysis. To compare the expression profiles among samples, an in-house R script was employed for unsupervised clustering of the normalized expression values of several hundred differentially expressed genes [ 26 ]. The gene expression values were visualized using heatmaps, while expression fold changes and p values between the two selected samples were represented in volcano plots, both generated using in-house R scripts.

GO and ROC Analysis

To identify target genes of miRNA, the multiMiR R package was employed. The multiMiR database integrates multiple sources, including DIANA-microT, ElMMo, MicroCosm, miRanda, miRDB, miRTarBase, PITA, TarBase, and TargetScan [ 27 ]. Next, gene ontology (GO) analysis was conducted to explore the functional implications of the differentially expressed miRNAs (DEmiRNAs). Additionally, Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis was performed. The GO terms and KEGG pathway enriched in the predicted target genes were identified using DAVID Bioinformatics [ 28 ]. To assess the diagnostic potential of miRNAs in distinguishing diseases, the pROC package was utilized to perform receiver operating characteristic (ROC) analysis on the normalized miRNA expression profiles of serum sEVs [ 29 ].

Statistical Analysis

Regarding demographic data and clinical characteristics, statistical differences among the groups were analyzed using nonparametric tests: Fisher’s exact test for categorical variables and the Kruskal–Wallis test for continuous variables. Continuous variables were presented as medians with interquartile ranges, while categorical variables were reported as absolute frequencies and percentages. For miRNA expression data, one-way ANOVA with Tukey’s post hoc test was used to compare the groups. A p value of < 0.05 was considered statistically significant. Using the ROC curve analysis, sensitivity, specificity, positive predictive value, and negative predictive value were calculated based on the cut-off value for the diagnosis of MS, NMOSD, and MOGAD. Additionally, the association between clinical features (ARR and EDSS scores) and DEmiRNAs was assessed using Spearman’s correlation analysis in patients with MOGAD, MS, and NMOSD. Multivariate linear regression analysis was conducted in Python using the SciPy (v.1.15.2) and Statsmodels (v.0.14.4) libraries. All statistical analyses were performed using Prism 9 (GraphPad Software) or SPSS Statistics 27.0 (SPSS Inc., Chicago, IL, USA).

Results

Clinical Characteristics of the Participants

We analyzed a cohort of 47 participants, comprising 11 patients with MOGAD (45.5% female), 12 patients with MS (75.0% female), 12 patients with anti-AQP4-positive NMOSD (100% female), and 12 HCs (75% female) (Table  1 , Supplementary file 1). All MOGAD patients, except for one, exhibited a “clear-positive” of MOG-IgG result (score ≥ 2) in CBA-immunofluorescence. One patient showed a “low positive” result with a score of 1.5 (CBA-FACS, MFIr = 4.68). Moreover, all MOGAD patients had the core clinical demyelinating events and supportive clinical or MRI features, thereby meeting the MOGAD diagnostic criteria (Supplementary file 2) [ 4 ]. Notably, males were more frequently represented in the MOGAD group, in contrast to the MS, NMOSD, and HC groups which exhibited a female predominance ( p  = 0.019). In contrast, age at sampling did not significantly differ between groups. Furthermore, no significant differences were observed between the MOGAD, MS, and NMOSD groups in terms of age of disease onset, disease status, initial presentation, ARR, and EDSS score at sampling.

Table 1 Demographic and clinical data of patients with MS, NMOSD, MOGAD, and healthy controls MOGAD ( N  = 11) MS ( N  = 12) NMOSD ( N  = 12) HC ( N  = 12) p value Female , N (%) 5 (45.5) 9 (75.0) 12 (100.0) 9 (75.0) 0.019 Age of disease onset, Y, median(Q1–Q3) 28.6 (25.5–39.1) 26.7 (25.2–29.7) 32.4 (24.3–49.8) NA 0.443 Age at sampling, Y, median(Q1–Q3) 34.6 (30.9–41.9) 29.4 (28.2–37.3) 40.0 (29.0–51.2) 33.5 (30.3–34.7) 0.504 Disease status , N (%)   Relapse within 30 days 1 (9.1) 3 (25.0) 1 (8.3) NA 0.584 Onset type , N (%) 0.604   Optic neuritis 7 (63.6) 6 (50.0) 4 (33.3) NA   Myelitis 2 (18.2) 4 (33.3) 5 (41.7) NA   Brainstem/cerebellum 1 (9.1) 2 (16.7) 3 (25.0) NA   Cerebrum 1 (9.1) 0 (0.0) 0 (0.0) Phenotype during disease course , N (%)   Optic neuritis 10 (90.9) 6 (50.0) 5 (41.7)   Myelitis 5 (45.5) 6 (50.0) 8 (66.6)   Brainstem/cerebellum 1 (9.1) 8 (66.6) 5 (41.7)   Cerebrum 2 (18.2) 3 (25.0) 0 (0.0) ARR, median(Q1–Q3) 0.26 (0.11–0.55) 0.44 (0.20–0.73) 0.12 (0.00–0.60) NA 0.479 EDSS score at sampling median(Q1–Q3) 1.50 (1.00–1.75) 1.75 (0.00–2.00) 2.25 (1.00–3.00) NA 0.395 Disease-modifying therapy at sampling , N (%)   No treatment 3 (27.3) 2 (16.7) 0 0.168   High dose steroid within 30 days 0 1 (8.3) 1(8.3) 1.000   Interferon-ß 0 4 (33.3) 0   Glatiramer acetate 0 2 (16.7) 0   Dimethyl fumarate 0 1 (8.3) 0   Teriflunomide 0 3 (25.0) 0   Azathioprine 5 (45.5) 0 6 (50.0)   Mycophenolate mofetil 3 (27.3) 0 5 (41.7)   Oral prednisolone 0 0 3 (25.0) MOGAD , Myelin oligodendrocyte glycoprotein antibody-associated disease; MS , multiple sclerosis; NMOSD , neuromyelitis optica spectrum disorder; HC , healthy control; Y , year; ARR , annualized relapse rate; EDSS , expanded disability status scale; Q1–Q3 : first quartile-third quartile

Demographic and clinical data of patients with MS, NMOSD, MOGAD, and healthy controls

MOGAD , Myelin oligodendrocyte glycoprotein antibody-associated disease; MS , multiple sclerosis; NMOSD , neuromyelitis optica spectrum disorder; HC , healthy control; Y , year; ARR , annualized relapse rate; EDSS , expanded disability status scale; Q1–Q3 : first quartile-third quartile

Characterization of Serum-Derived sEV

Serum-derived sEVs were isolated from patients with MOGAD, MS, and NMOSD, as well as from HCs. Cryo-transmission electron microscopy (Cryo-TEM) confirmed that these vesicles exhibited typical sEV characteristics, with a predominant size range of 30–200 nm (Fig.  1 A). Western blot analysis verified the enrichment of sEV-specific markers (CD63, HSP70, ALIX, and TSG101) and the absence of endoplasmic reticulum marker calnexin, confirming the purity of the sEV preparations (Fig.  1 B). In addition, nanoparticle tracking analysis (NTA) characterized the size distribution and concentration of vesicles, demonstrating consistency with established sEV profiles (Fig.  1 C).

Characterization of isolated sEV from serum and analysis for miRNA expression in sEV. A Representative cryo-TEM images displaying the morphology of sEVs isolated from serum. Scale bar: 100 nm. B Western blot analysis for ALIX, Calnexin, HSP70, TSG101, and CD63 of sEV. C Nanoparticle tracking analysis of size distributions and concentration measurements of sEV. Venn diagram illustrating the overlapping miRNAs identified in the differential expression analysis when comparing the HC with MS, NMOSD, and MOGAD: D total differentially expressed miRNAs, as well as E the upregulated and F downregulated miRNAs

Differential miRNA Expression Patterns Between Patients and HCs

Comprehensive miRNA profiling using the NanoString platform identified distinct miRNA expression profiles in serum sEVs obtained from patients and HCs (Supplementary file 3). Compared to HCs, DEmiRNAs were identified in each disease group: 77 miRNAs in MOGAD, 16 miRNAs in MS, and 32 miRNAs in NMOSD. Venn diagram analysis revealed four DEmiRNAs shared among all three disease groups, with one upregulated (miR- 548a- 5p) and three downregulated (miR- 576 - 5p, miR- 329 - 3p, and miR- 365b- 5p) DEmiRNAs compared to the HCs (Fig.  1 D–F). A heatmap and volcano plot visualized these DEmiRNAs (Fig.  2 ), demonstrating distinct expression patterns in MOGAD patients compared to HCs and patients with MS and NMOSD.

Comprehensive analysis of DEmiRNAs in sEV from MOGAD patients compared with HCs, MS, and NMOSD patients. Hierarchical clustering of DE miRNAs and heatmap representing the significantly differentially expressed miRNAs and volcano plot displaying the DE miRNAs A between MOGAD and HC, B between MOGAD and MS, and C  between MOGAD and NMOSD. The red dots indicate statistically significant DE miRNAs

MOGAD and HC

In MOGAD patients, 22 miRNAs were significantly upregulated and 55 miRNAs were downregulated compared to HCs (Fig.  2 A). GO analysis revealed significant enrichment of the predicted target genes in cell adhesion molecules, including “homophilic cell adhesion via plasma membrane adhesion molecules”, and “cell adhesion”. Additionally, neuronal component-related pathways such as “nervous system development”, “glutamatergic synapse”, “axon”, and “dendritic spine” showed substantial enrichment, suggesting potential involvement of miRNAs in neuronal processes. Moreover, KEGG enrichment analysis demonstrated significant enrichment of pathways related to “adherens junction”, “mRNA surveillance pathway”, and “cellular senescence” (Fig.  3 A).

GO analysis and KEGG pathway enrichment analysis conducted on the genes targeted by the top 10 DE miRNAs in MOGAD patients compared with HC, MS and NMOSD patients: A between MOGAD and HC, B between MOGAD and MS, and C between MOGAD and NMOSD

MOGAD and MS

The analysis identified 18 upregulated DEmiRNAs and 36 downregulated DEmiRNAs in MOGAD patients compared to MS patients (Fig.  2 B). GO analysis revealed significant enrichment of the predicted target genes in chromatin-related processes, specifically involving “chromatin remodeling” and “chromatin organization”. Moreover, neuronal pathways, including “glutamatergic synapse”, “dendrite”, and “postsynaptic density”, were substantially enriched. Furthermore, KEGG enrichment analysis demonstrated significant enrichment of pathways related to “human T-cell leukemia virus 1 infection”, “PI3 K-Akt signaling pathway”, and “MAPK signaling pathway” (Fig.  3 B).

MOGAD and NMOSD

We observed significant upregulation of 35 miRNAs and downregulation of 52 miRNAs in MOGAD compared to NMOSD (Fig.  2 C). Among the enriched GO pathways, pathways related to post-translational modification of proteins, such as “peptidyl-serine phosphorylation” and “protein phosphorylation”, were significantly enriched. Additionally, the “PML body” was found to be overrepresented. In terms of KEGG pathways, “human cytomegalovirus infection” and “hepatitis B” pathways emerged as significant pathways in our analysis (Fig.  3 C).

Evaluation of miRNAs as a Diagnostic Tool Through ROC Analysis

ROC analysis confirmed the high diagnostic potential of miRNAs, with area under the curve (AUC) values indicating a strong ability to distinguish between diseased and healthy conditions (Fig.  4 ). When additional analyses were conducted using the top three miRNAs with the highest AUC values as a panel in each disease group, the AUC values further increased.

ROC curve analysis with serum sEV miRNAs showing top three miRNAs with the highest AUC values and ROC curve with the three miRNAs combined as a panel. ROC curves for distinguishing A MOGAD from HC, B MOGAD from MS, C MOGAD from NMOSD, D MOGAD from MS/NMOSD, and E MOGAD from MS/NMOSD/HC. * P  < 0.05; ** P  < 0.01

Differentiation of MOGAD from HC, MS, or NMOSD

In the ROC curve analysis, the top three miRNAs with the highest AUC values for discriminating MOGAD patients from HCs were hsa-miR- 924, hsa-miR- 891b, and hsa-miR- 346, with AUC values of 0.902, 0.909, and 0.894, respectively. Importantly, when these three miRNAs were combined as a panel, the AUC value significantly increased to 1.000 (Fig.  4 A). To distinguish MOGAD from MS, hsa-miR- 4536 - 5p, hsa-miR- 4516, and hsa-miR- 346 exhibited the highest AUC values (0.856, 0.841, and 0.841, respectively). When these three miRNAs were combined as a panel, the AUC value significantly increased to 0.939 (Fig.  4 B). Additionally, for differentiating MOGAD from NMOSD, hsa-miR- 1976, hsa-miR- 615 - 5p, and hsa-mir- 562 exhibited the highest AUC values (0.886, 0.902, and 0.962, respectively). When combined as a panel, these three miRNAs significantly improved the AUC value to 1.000 (Fig.  4 C).

Differentiation of MOGAD from Other Groups

To distinguish MOGAD from other CNS demyelinating diseases (MS and NMOSD), hsa-miR- 873 - 5p, hsa-miR- 548y, and hsa-miR- 26b- 5p were identified as key miRNAs with AUC values of 0.833, 0.818, and 0.811, respectively. Notably, these three miRNAs exhibited significantly lower expression levels in MOGAD compared to the other diseases (Fig.  4 D). To distinguish MOGAD from MS, NMOSD, and HCs, hsa-miR- 564, hsa-miR- 346, and hsa-miR- 873 - 5p were identified as discriminatory miRNAs with AUC values of 0.798, 0.816, and 0.818, respectively (Fig.  4 E).

Correlation of miRNAs with EDSS and ARR in MOGAD Patients

In MOGAD patients, increased levels of hsa-miR- 664b- 3p ( ρ  = 0.58, p  = 0.004) were associated with higher EDSS scores, whereas lower expression levels of hsa-miR- 1245a ( ρ  = –0.65, p  < 0.001), miR- 346 ( ρ  = –0.65, p  < 0.001), and miR- 924 ( ρ  = –0.67, p  < 0.001) were also associated with higher EDSS scores (Fig.  5 A). Moreover, a higher ARR was associated with increased expression of hsa-miR- 328 - 5p ( ρ  = 0.56, p  = 0.005) and hsa-miR- 422a ( ρ  = 0.60, p  = 0.002), as well as decreased expression of hsa-miR- 1245a ( ρ  = –0.65, p  < 0.001) and hsa-miR- 548i ( ρ  = –0.69, p  < 0.001) in the MOGAD group (Fig.  5 B). To account for potential confounders such as sex and age, a multivariate regression analysis was performed, which identified significant correlations between EDSS and specific miRNAs after adjusting for age and sex (Table  2 ). Notably, hsa-miR- 664b- 3p demonstrated a positive correlation ( β  = 0.4901, p  = 0.0176), while hsa-miR- 1245 and hsa-miR- 346 exhibited negative correlations ( β  = − 0.4644, p  = 0.0256; β  = − 0.5445, p  = 0.0072, respectively) with EDSS scores. In contrast, after adjustment for age and sex, no significant correlations were observed between ARR and any of the analyzed miRNAs. Additionally, to examine whether relapsed patients exhibited distinct miRNA profiles compared to non-relapsed patients, we analyzed the miRNA content of serum-derived sEVs from both relapsed and non-relapsed patients. We identified several differentially expressed miRNAs in the relapsed patient cohort, suggesting that these molecular signatures may provide valuable insights into the relationship between miRNA expression patterns and disease activity (Supplementary Fig. 1 ). To verify that the miRNAs detected in the EVs are not simply co-isolated miRNAs from the serum that may have been internalized into artifactual EVs, we conducted a comparative analysis examining three distinct miRNAs across samples obtained from three patients. Importantly, miR- 346 demonstrated significant differential expression between serum and EV-derived RNA, confirming that it was not merely co-purified from serum. The additional two miRNAs showed patterns indicating potential differential expression, which may become more distinct with a larger sample size. These results confirm that the miRNAs observed in the EVs were not merely co-isolated serum miRNAs internalized into artifactual EVs, thereby strengthening the validity of our findings (Supplementary Fig. 2 ).

Correlation analysis of DEmiRNAs with EDSS and ARR in MOGAD patients. A Four DEmiRNAs showed correlation with expanded disability status scale (EDSS) score. B Four DEmiRNAs showed correlation with annualized relapse rate (ARR)

Table 2 Multivariate regression analysis of the correlation between serum sEV miRNAs and EDSS or ARR, adjusted for age, and sex β P value EDSS   hsa-miR- 664b- 3p 0.4901 0.0176   hsa-miR- 1245  − 0.4644 0.0256   hsa-miR- 346 –0.5445 0.0072   hsa-miR- 924 –0.4032 0.0565 ARR   hsa-miR- 328 - 5p 0.2381 0.2739   hsa-miR- 422a 0.2955 0.1711   hsa-miR- 1245a –0.0405 0.8544   hsa-miR- 548i –0.1245 0.5714 ARR , Annualized relapse rate; EDSS , expanded disability status

Multivariate regression analysis of the correlation between serum sEV miRNAs and EDSS or ARR, adjusted for age, and sex

ARR , Annualized relapse rate; EDSS , expanded disability status

Discussion

In this study, we utilized serum-derived sEVs to profile disease-specific DEmiRNAs in patients with MOGAD. Our findings indicate that specific DEmiRNAs in MOGAD patients exhibit significant discriminative power against MS, NMOSD, and HCs, achieving high AUC values ranging from 0.9 to 1.0 when utilized in a panel of three miRNAs. Furthermore, certain miRNAs demonstrated strong correlations with clinical indices of disease disability and activity, underscoring their potential roles as both diagnostic and prognostic biomarkers in MOGAD.

We identified four miRNAs including one upregulated (miR- 548a- 5p) and three downregulated (miR- 576 - 5p, miR- 329 - 3p, and miR- 365b- 5p) that were significantly dysregulated across MOGAD, MS, and NMOSD, compared to healthy controls. The upregulation of miR- 548a- 3p has been documented in various autoimmune conditions, including rheumatoid arthritis (RA), psoriasis, and acute graft-versus-host disease. Importantly, recent studies have indicated the potential of miR- 548a- 3p as a biomarker for treatment response in MS, particularly in patients achieving no evidence of disease activity (NEDA- 3) post-treatment with fingolimod [ 30 ]. Among the downregulated miRNAs, the level of miR- 576 - 5p in whole blood was not significantly different among HCs, MS, and NMOSD patients in a previous study, whereas its expression in lymphocytes was reported to be lower in patients with systemic lupus erythematosus (SLE) than in HCs [ 31 , 32 ]. In addition, plasma miRNA- 329 - 3p expression was decreased in patients with type 2 diabetic mellitus compared to HCs, and it has the ability to directly bind to TLR4 and control its expression, thereby regulating the inflammatory status [ 33 ]. Lastly, miR- 365b- 5p has been suggested as a key target in Alzheimer’s disease, regulating the pathogenesis via neuroinflammation, Wnt and oxidative pathway [ 34 , 35 ]. Therefore, the four commonly dysregulated miRNAs in our MOGAD, MS, and NMOSD patients may be associated with the regulation of inflammatory processes rather than disease-specific pathogenesis.

We identified a panel of three DEmiRNAs (hsa-miR- 924, hsa-miR- 891b, and hsa-miR- 346) that could distinguish MOGAD patients from HCs with a high AUC value of 1.000. A previous study on MOGAD reported that the splenocytes and lymph node cells from EAE mice showed increased expression levels of miR- 17, miR- 18a, miR- 20a, and miR- 92a- 1, which was also observed in the peripheral blood lymphocytes in MOGAD patients compared to HCs [ 19 ]. However, this study only confirmed the consistency between MOGAD patients and EAE mice, without providing details on the number of patients, their clinical characteristics, or the MOG-IgG assay. Overall, the miRNA profiles in MOGAD patients remain poorly understood. Furthermore, we identified another panel of three DEmiRNAs including hsa-miR- 4536 - 5p, hsa-miR- 4516, and hsa-miR- 346, which could distinguish MOGAD from MS. Additionally, the third panel of three DEmiRNAs (hsa-miR- 1976, has-miR- 615 - 5p, and hsa-miR- 562) demonstrated the highest AUC values for differentiating MOGAD from NMOSD. In a previous study, hsa-miR- 346 was significantly downregulated in EVs derived from epileptic forebrains of the temporal lobe epilepsy rat model [ 36 ]. Although MOGAD is more frequently associated with cortical encephalitis and seizures compared to MS or NMOSD, only one patient in our cohort exhibited cerebral lesions without seizure. Therefore, it remains to be elucidated whether the downregulation of hsa-miR- 346, which is enriched in pathways linked to GABAergic synapses and cell adhesion molecules, may be linked to epileptogenesis in MOGAD [ 37 ].

Moreover, hsa-miR- 873 - 5p, hsa-miR- 548y, and hsa-miR- 26b- 5p exhibited the high AUC values in differentiating MOGAD from the other two diseases, MS and NMOSD. The pathology of MOGAD is characterized by predominant inflammatory demyelination without astrocyte destruction [ 38 ]. It typically begins with perivenous demyelination and follows a fusion pattern, which distinguishes it from MS, as radial expansion is not observed [ 38 ]. In MOGAD lesions, CD4 + T cells are the predominant inflammatory cell type, in contrast to the CD8 + T cell predominance seen in MS [ 38 , 39 ]. Additionally, unlike NMOSD, perivenous complement deposition is less common in MOGAD [ 38 ]. Therefore, these miRNAs, which distinguish MOGAD from MS and NMOSD, may partially explain its underlying pathophysiology.

Among them, hsa-miR- 873 - 5p exhibited significantly decreased expression in MOGAD patients, with the highest AUC value. Previous studies have suggested that miR- 873 - 5p may play an anti-inflammatory and protective role. For instance, miR- 873 - 5p has been implicated in neuronal apoptosis mechanisms in Alzheimer’s disease by targeting heme oxygenase 1, which also plays an important protective role in the EAE model [ 40 , 41 ]. Notably, miR- 873 - 5p is known to inhibit the NF-κB signaling pathway, whereas activation of this pathway promotes CD4 + T cell differentiation [ 42 , 43 ]. Previous research has suggested that IgGs from patients with MOGAD induce the blood–brain barrier (BBB) disruption via NF-κB signaling [ 44 ]. Therefore, the downregulation of miR- 873 - 5p in MOGAD may be related to the activation of the NF-κB pathway and the enhancement of inflammatory responses, particularly through CD4 + T cell differentiation and BBB disruption. In addition to miR- 873 - 5p, miR- 548y was found to be more downregulated in MOGAD compared to MS and NMOSD. While miR- 548as- 3p has been associated with the NF-κB signaling pathway in non-small cell lung cancer (NSCLC), and miR- 548 has been identified as a differentially expressed miRNA in persistent post-concussive symptoms, the role of miR- 548 in CNS or autoimmune inflammatory disorders remains largely unexplored [ 45 , 46 ]. Further research is needed to elucidate how miR- 548 functions in MOGAD and how it relates its distinct pathophysiology compared to MS and NMOSD. Similarly, miR- 26b- 5p was also downregulated in MOGAD compared to other disease groups. This miRNA has been implicated in various inflammatory responses pathways. Overexpression of miR- 26b- 5p has been shown to enhance cell proliferation, inhibit apoptosis, reduce reactive oxygen species (ROS) levels, and suppress inflammatory mediators [ 47 ]. Moreover, in osteoarthritis, miR- 26b- 5p has been reported to attenuate inflammation by promoting the polarization of M1 macrophages toward the M2 phenotype [ 48 ]. In the context of MOGAD, demyelinating lesions exhibit MOG-dominant myelin loss, particularly in early-stage lesions, with active myelin phagocytosing macrophages [ 2 ]. Another study demonstrated that autoantibodies from MOGAD patients can mediate antibody-dependent cellular phagocytosis, suggesting that macrophages and microglia play a role in disease pathology [ 49 ]. Furthermore, miR- 26b- 5p has been shown to inhibit oligodendrocyte precursor cell differentiation in spinal cord injury [ 50 ]. Therefore, the downregulation of miR- 26b- 5p may be associated with the pathophysiology of MOGAD by impairing oligodendrocyte differentiation and inflammatory regulation.

Lastly, when we analyzed miRNA to differentiate MOGAD from others (MS, NMOSD, and HCs), additional miRNAs were identified; hsa-miR- 564, hsa-miR- 346, and hsa-miR- 873 - 5p in MOGAD showed high AUC values to distinguish from the other groups. Previously, hsa-miR- 564 was reported to promote the maturation, proliferation of monocyte-derived dendritic cells by negatively regulating the expression of TP53 in SLE [ 51 ]. Additionally, hsa-miR- 564 is known to function as a tumor suppressor by inhibiting the PI3 K and MAPK signaling pathways, which may influence inflammatory responses. Therefore, the downregulation of hsa-miR- 564 could potentially contribute to increased inflammation [ 52 , 53 ]. Interestingly, hsa-miR- 346 could also distinguish MOGAD from both HCs and MS, and was significantly decreased in MOGAD, compared to all other groups (MS, NMOSD, and HCs). This finding suggests that hsa-miR- 346 may serve as a disease-specific biomarker for MOGAD. Previously, the downregulation of hsa-miR- 346 in EVs from epileptic forebrains was reported, associated with GABA-ergic mechanism in underlying epilepsy [ 37 ]. The hsa-miR- 346 has been reported to indirectly regulate interleukin- 18 (IL- 18) release, suggesting that miRNAs function as regulators that fine-tune the inflammatory response in RA [ 54 ]. IL- 18 is known to play a pathogenic role by inducing T cell–mediated responses, as evidenced by studies in various autoimmune diseases. However, many aspects of its role remain unclear, and research specifically addressing IL- 18 in the context of MOGAD is scarce [ 55 ].

In the correlation analysis, higher expression levels of miR- 664b- 3p were correlated with higher EDSS scores, whereas lower expression levels of three miRNAs (miR- 1245a, miR- 346 and miR- 924) were correlated with higher EDSS scores in MOGAD. Previously, miR- 664b- 3p was identified as a tumor suppressor in colon cancer, where it inhibits cell proliferation, migration, and invasion while promoting apoptosis by suppressing BUB3 protein expression [ 56 , 57 ]. In NSCLC, miR- 924 functions as a tumor suppressor by inhibiting Wnt/β-catenin pathway [ 58 ]. The Wnt signaling pathway plays diverse roles in immune cell regulation, and its dysregulation has been implicated in autoimmune diseases [ 59 ]. Furthermore, activation of the Wnt/β-catenin pathway has been shown to induce differentiation failure and impair remyelination in MS [ 60 ]. The negative correlation between miR- 924 expression and EDSS score in our study may be partially explained by the activation of Wnt/β-catenin signaling, which could exacerbate disease severity in MOGAD by impairing remyelination. Additionally, higher levels of hsa-miR- 328 - 5p and hsa-miR- 422a, as well as lower levels of hsa-miR- 1245a and hsa-miR- 548i, were correlated with ARR. Another study showed upregulation of hsa-miR- 422a in plasma of patients with MS, identifying its target as cholesterol 7 alpha-hydroxylase (CYP7 A1) [ 61 ]. It was suggested that an increase in miR- 422a expression would inhibit the expression of CYP7 A1, resulting in the accumulation of cholesterol. This accumulation may perturb the myelination process and impair the viability of oligodendrocytes, as seen in the dysmyelinating disorder of Pelizaeus–Merzbacher disease [ 61 , 62 ]. In addition, the expression level of miR- 548i was found to be significantly elevated in subacute sclerosing panencephalitis due to an over-activated immune system response [ 63 ]. After adjusting potential confounders, we observed correlations between EDSS and four specific miRNAs but with no significant correlations between ARR and four specific miRNAs, which suggests that miRNA expression patterns associated with EDSS reflect disease severity independently of demographic confounding factors such as age and sex, whereas miRNA expression patterns related to ARR may be more substantially influenced by these variables (Table  2 ). Given that dysregulated immune mechanisms and disrupted myelination are key pathological features of MOGAD, further research is needed to determine whether these miRNAs contribute to disease pathogenesis and how they may serve as potential biomarkers for disease activity and progression.

Our study has several limitations. First, the number of patients was relatively small, and the study was conducted at a single center. Nevertheless, this is the first study to explore the clinical role of exosomal miRNAs as diagnostic or prognostic biomarkers in patients with MOGAD. Moreover, the number of treatment-naïve MOGAD patients was limited. Given the biological effects of disease-modifying therapies, it is crucial to determine whether these miRNA profiles reflect the untreated state of MOGAD. Therefore, further subanalyses considering clinical heterogeneity are needed in a larger cohort, as factors such as age, sex, disease status, disease duration, and treatment could influence miRNA expression, potentially leading to unintended bias or reduced statistical power [ 14 , 64 ]. To assess whether recent steroid use significantly impacted our main findings, we performed a sensitivity analysis excluding four patients who had received steroids within 4 weeks before serum collection. Upon reanalysis, the exclusion of these patients did not significantly alter the results (Supplementary Figs. 3 and 4 ). Second, the DEmiRNAs observed in patients with MOGAD in our study differed from a previous study, which applied findings from EAE mice to MOGAD patients [ 19 ]. The discrepancy between studies may stem from differences in the biological matrix used (leukocytes or small EVs), analytical methods (NGS vs. NanoString analysis), and clinical characteristics of the study populations. Third, samples were stored at − 80 °C for varying durations after collection, which can significantly impact EV sample quality by causing particle loss, reduced purity, and the formation of artifactual particles through fusion phenomena, although we did not subject the samples to freeze–thaw cycles [ 65 ]. However, fresh analysis is practically challenging, which is why many studies continue to rely on stored samples [ 66 ]. Lastly, only a small portion of serum-derived EVs originates from the CNS; therefore, various methods have been used to identify CNS-derived EVs. However, the isolation of brain-derived EVs has not been fully established, despite attempts using various techniques [ 66 – 68 ]. Most methods for enriching CNS-derived EVs from serum rely on L1 CAM-mediated immunocapture. However, recent studies have advised against using L1 CAM for this purpose, as it is widely expressed outside the brain and also exists in a soluble form [ 69 , 70 ]. In addition, unlike other neurodegenerative diseases, MS, NMOSD and MOGAD are CNS demyelinating disorders in which peripheral immune cells interacts with CNS-resident cells, such as neurons, microglia and astrocytes to actively regulate immunological processes affecting the brain, spinal cord, and optic nerve [ 71 ]. Therefore, serum-derived EVs, which reflect peripheral immune dysregulation and pathophysiology may play a more significant role in CNS demyelinating disorders, where peripheral immune interactions are crucial, compared to neurodegenerative diseases that primarily affect the brain.

A systematic review of miRNA studies in MS highlighted that discrepancies between studies may stem from heterogeneity in clinical characteristics, sample sources, sampling timing, RNA isolation techniques, microarray panel selection, miRNA set composition, sensitivity levels, intrinsic controls, and normalization strategies [ 72 ]. Therefore, it remains to be investigated whether DEmiRNAs identified in this study for MOGAD are reproducible and exhibit consistent functional roles in a large cohort of MOGAD patients with diverse racial backgrounds, age groups, sexes, and clinical manifestations. Further validation studies using polymerase chain reaction (PCR) with a larger number of patients may be beneficial.

In conclusion, our study highlights the potential of sEV-derived miRNAs as promising biomarkers for the diagnosis and monitoring of MOGAD. The identified DEmiRNAs not only distinguish MOGAD from HCs and other diseases, but also correlate with key clinical parameters, offering insights into the underlying disease mechanisms and identifying potential biomarkers. Further validation and investigation of these miRNAs may pave the way for novel diagnostic and therapeutic strategies in MOGAD.

Supplementary Information

Below is the link to the electronic supplementary material.

Supplementary Figure 1 miRNA profiles of serum-derived sEVs from relapsed and non-relapsed patients. (A) Volcano plot representing miRNA expression in relapsed patients relative to non-relapsed patients. (B) Heatmap of miRNA expression analysis in serum-derived sEVs.(PNG 5.12 MB)

High Resolution Image (TIF 11.8 MB)

Supplementary Figure 2 Comparative analysis of miRNA expression in serum and serum-derived sEV. Data are presented as mean ± SEM *, P < 0.05(PNG 1.39 MB)

High Resolution Image (TIF 3.44 MB)

Supplementary Figure 3 Comprehensive analysis of DEmiRNAs sEV from MOGAD patients, compared to HCs, MS, and NMOSD patients, excluding those who received steroids within 4 weeks prior to serum collection. The analysis includes hierarchical clustering of DEmiRNAs, a heatmap showing significantly differentially expressed miRNAs, and a volcano plot displaying DEmiRNAs for comparisons between (A) MOGAD and HC, (B) MOGAD and MS, and (C) MOGAD and NMOSD. Red dots indicate statistically significant DEmiRNAs.(PNG 19.2 MB)

High Resolution Image (TIF 24.9 MB)

Supplementary Figure 4 ROC curve analysis of serum sEV miRNAs highlights the top three miRNAs with the highest AUC values, as well as a combined ROC curve using these three miRNAs as a panel. This analysis includes only patients who had not received steroids within the prior 4 weeks. ROC curves for distinguishing (A) MOGAD from MS, (B) MOGAD from NMOSD, (C) MOGAD from MS/NMOSD, and (D) MOGAD from MS/NMOSD/HC.(PNG 7.68 MB)

High Resolution Image (TIF 10.7 MB)

Supplementary Figure 5 Uncropped image of Western blot analysis with ALIX, Calnexin, HSP70, TSG101, and CD63(PNG 2.34 MB)

High Resolution Image (TIF 5.55 MB)

Supplementary Figure 6 Uncropped Cryo-TEM image indicating a EV.(PNG 18.8 MB)

High Resolution Image (TIF 15.6 MB)

Supplementary file7 (XLSX 17 KB)

Supplementary file8 (XLSX 12 KB)

Supplementary file9 (CSV 207 KB)

Acknowledgements

The authors appreciate all the participants of this study for their donation of blood.

Abbreviations

Myelin oligodendrocyte glycoprotein

Myelin oligodendrocyte glycoprotein (MOG) antibody-associated disease

Central nervous system

Multiple sclerosis

Neuromyelitis optica spectrum disorder

MicroRNAs

Small extracellular vesicles

Healthy controls

Experimental autoimmune encephalomyelitis

Anti-aquaporin 4 antibody

Uorescence assay

CBA flow cytometry analysis

Annualized relapse rate

Expanded disability status scale

Cryo-transmission electron microscopy

Differentially expressed miRNAs

Kyoto Encyclopedia of Genes and Genomes

Receiver operator characteristic

Gene ontology

Differentially expressed miRNAs

Kyoto Encyclopedia of Genes and Genomes

Area under the curve

No evidence of disease activity

Author Contribution

HKK, DJ and JHM conceived of the study, participated in the design of the study, and analysed the data. HKK and HJ drafted the manuscript. HKK performed the statistical analysis. HJ, YHC, ESJ, YC, and YJ acquired and analysed data. HKK performed the analysis of miRNAs and analysed data. HKK, HJ, DJ and JHM analysed data and revised the manuscript for important intellectual content. All authors read and approved the final manuscript. HKK and HJ contributed equally as the first author.

Funding

Open Access funding enabled and organized by SungKyunKwan University This research was supported by the National Research Foundation of Korea (NRF) grant funded by the Korea government (MSIT) (2021R1F1A1049347 to JHM, RS- 2024–00399237 and RS- 2024–00345742 to DJ) and by SMC Research and Development Grant (SMO1220511 to JHM).

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

Data Availability

Anonymized data not published in this article will be made available upon reasonable request from qualified investigators.

Declarations

Ethical Approval and Consent to Participate

A signed informed consent was received from all the participants. This study has been approved from the institutional review board of the Samsung Medical Center (IRB No. 2020–04 - 119).

Consent for Publication

Signed informed consent was received from all the participants including consent for publication.

Competing interests

JHM is funded by and has received research support from the National Research Foundation of Korea (MIST and KHIDI) and SMC Research and Development grant. She has lectured, consulted, and received honoraria from Bayer Healthcare, Merk, Biogen Idec, Sanofi, UCB, Samsung Bioepis, Mitsubishi Tanabe, Celltrion, Roche, Janssen, and AstraZeneca. All other authors declare that they have no competing interests.

Footnotes

The original online version of this article was revised to adjust the order of affiliations and the assignment of those affiliations to each author.

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Hark Kyun Kim and Hyunjin Ju contributed equally.

Change history

6/1/2025

The original online version of this article was revised to adjust the order of affiliations and the assignment of those affiliations to each author.

Contributor Information

Dong-Gyu Jo, Email: jodg@skku.edu.

Ju-Hong Min, Email: juhongm@skku.edu.

References

Associated Data

Supplementary Materials

Supplementary Figure 1 miRNA profiles of serum-derived sEVs from relapsed and non-relapsed patients. (A) Volcano plot representing miRNA expression in relapsed patients relative to non-relapsed patients. (B) Heatmap of miRNA expression analysis in serum-derived sEVs.(PNG 5.12 MB)

High Resolution Image (TIF 11.8 MB)

Supplementary Figure 2 Comparative analysis of miRNA expression in serum and serum-derived sEV. Data are presented as mean ± SEM *, P < 0.05(PNG 1.39 MB)

High Resolution Image (TIF 3.44 MB)

Supplementary Figure 3 Comprehensive analysis of DEmiRNAs sEV from MOGAD patients, compared to HCs, MS, and NMOSD patients, excluding those who received steroids within 4 weeks prior to serum collection. The analysis includes hierarchical clustering of DEmiRNAs, a heatmap showing significantly differentially expressed miRNAs, and a volcano plot displaying DEmiRNAs for comparisons between (A) MOGAD and HC, (B) MOGAD and MS, and (C) MOGAD and NMOSD. Red dots indicate statistically significant DEmiRNAs.(PNG 19.2 MB)

High Resolution Image (TIF 24.9 MB)

Supplementary Figure 4 ROC curve analysis of serum sEV miRNAs highlights the top three miRNAs with the highest AUC values, as well as a combined ROC curve using these three miRNAs as a panel. This analysis includes only patients who had not received steroids within the prior 4 weeks. ROC curves for distinguishing (A) MOGAD from MS, (B) MOGAD from NMOSD, (C) MOGAD from MS/NMOSD, and (D) MOGAD from MS/NMOSD/HC.(PNG 7.68 MB)

High Resolution Image (TIF 10.7 MB)

Supplementary Figure 5 Uncropped image of Western blot analysis with ALIX, Calnexin, HSP70, TSG101, and CD63(PNG 2.34 MB)

High Resolution Image (TIF 5.55 MB)

Supplementary Figure 6 Uncropped Cryo-TEM image indicating a EV.(PNG 18.8 MB)

High Resolution Image (TIF 15.6 MB)

Supplementary file7 (XLSX 17 KB)

Supplementary file8 (XLSX 12 KB)

Supplementary file9 (CSV 207 KB)

Data Availability Statement

Anonymized data not published in this article will be made available upon reasonable request from qualified investigators.