Serum and CSF cytokine profile impact clinical phenotype in pediatric MOGAD

Abstract

Background

Myelin oligodendrocyte glycoprotein antibody-associated disease (MOGAD) has a broad spectrum of clinical phenotypes, including acute disseminated encephalomyelitis (ADEM), optic neuritis (ON), transverse myelitis (TM), and cerebral cortical encephalitis (CCE). Although these conditions exhibit diverse outcomes and require different treatment options for relapse prevention, no biomarkers have been identified to distinguish the clinical phenotypes. This study aimed to investigate the properties of cytokines and chemokines in the serum and cerebrospinal fluid (CSF) of patients with MOGAD and to determine whether they are associated with the clinical phenotype.

Methods

In this retrospective cohort study, we analyzed cytokines and chemokines in paired serum and CSF samples obtained during the acute phase before treatment from patients with MOG-IgG positivity, classified into five clinical phenotypes (ADEM, CCE, ON, TM, and “others”).

Results

Of the 50 patients diagnosed with MOGAD, 19 (38%) were diagnosed with ADEM, 9 (18%) with CCE, 3 (6%) with ON + TM, 9 (18%) with ON, and 10 (20%) with others. Overall, the CSF levels of Th17-related (including IL-6, CXCL8, G-CSF, IL-17, and IL-21) and B cell-related (including CXCL13, BAFF, and APRIL) cytokines/chemokines were elevated in pediatric patients with MOGAD compared with non-inflammatory neurological disease controls. Phenotype-stratified analyses of ADEM, CCE, and ON demonstrated higher CSF IL-6 and CXCL8 levels in encephalitic phenotypes (ADEM/CCE) than in ON. By contrast, no significant difference in the CSF-to-serum ratio (C/S ratio) for each cytokine was observed among the three phenotypes. Only the CXCL13 C/S ratio correlated with the ΔMOG-IgG C/S ratio.

Conclusion

Pediatric patients with MOGAD showed limited phenotype-specific heterogeneity in their cytokine/chemokine profiles. Nevertheless, leptomeningeal inflammation was more prominent in patients with encephalitic presentations. Additionally, intrathecal MOG-IgG synthesis may be linked to intracranial inflammatory activity, which could inform future prospective studies using stratified treatment approaches.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12974-026-03795-2.

Introduction

Myelin oligodendrocyte glycoprotein antibody-associated disease (MOGAD) is a novel inflammatory demyelinating disease of the central nervous system (CNS) [1]. It has a broad clinical spectrum, including acute disseminated encephalomyelitis (ADEM), optic neuritis (ON), transverse myelitis (TM), and cerebral cortical encephalitis with seizures (CCE) [2]. Moreover, pediatric MOGAD differs from adult-onset MOGAD in that intracranial lesions are more frequently observed in children [2]. Although relapse rates are generally lower in children, recurrent intracranial involvement may lead to cognitive impairment, with cerebral lesions recognized as key predictors of cognitive deficits [3, 4]. In addition, intracranial phenotypes such as ADEM and CCE are associated with an increased risk of early relapse [5]. Given the diverse clinical outcomes, the required relapse prevention therapies may differ, making the stratification of these phenotypes a high clinical priority. The underlying mechanisms responsible for such phenotypic heterogeneity in MOGAD remain unclear, despite that its etiology is linked to a singular anti-MOG antibody.

Recently, MOGAD in adults, the status of MOG-IgG _SERUM/_{cerebrospinal fluid (CSF)} been associated with clinical phenotypes. The MOG-IgG _CSF/MOG-IgG _SERUM ratio in CCE cases was significantly higher than that in other MOGAD phenotypes [6]. Thus, intrathecal production of MOG-IgG has been suggested in patients where MOG antibody titers are significantly higher in CSF compared to serum [6, 7]. The CSF cytokine profile of patients with MOGAD showed elevated levels of Th17-related cytokines (IL-6, IL-8, G-CSF, and GM-CSF) and B-cell-related cytokines (CXCL13, APRIL, BAFF, and CCL19), indicating enhanced leptomeningeal inflammation (LMI) [8, 9]. However, research examining the characteristic cytokine and chemokine trends of each MOGAD phenotype is scarce. Furthermore, reports on this topic are few, especially in children.

This study aimed to explore the characteristics of cytokines and chemokines in serum and CSF that determine the clinical phenotype of Japanese pediatric patients with MOGAD.

Materials and methods

Study design/study population

In this retrospective cohort study, we evaluated pediatric patients with neurological diseases and a suspected autoimmune background in the CNS whose serum and CSF were sent to our laboratory between January 1, 2015 and April 30, 2024. Samples from pediatric patients were referred to our laboratory from multiple institutional collaborators in Japan. Firstly, we extracted patients who met the following criteria: (i) chronological age < 18 years at initial disease onset, (ii) serum and CSF samples collected before any treatment in the acute phase, and (iii) sufficient clinical data available for our analysis. Medical records and samples were obtained from all patients. We contacted each institution via E-mail to request missing information and subsequently requested that they send summaries as needed. In all patients extracted, the median fluorescence intensity (MFI) ratios of MOG-IgG in the serum and CSF were measured concurrently, and patients who tested positive for MOG-IgG in the serum and/or CSF were identified. The patients were evaluated to determine whether they met the diagnostic criteria for MOGAD. Among MOG-IgG-negative patients, those who fulfilled all of the following criteria were classified as non-inflammatory neurological diseases (NIND) group: (i) a CSF cell count < 10/mm³; (ii) no MRI lesions suggesting demyelinating disease or autoimmune encephalitis; and (ⅲ) no evidence of other established neuroinflammatory diseases, including ADEM, multiple sclerosis, NMDA receptor encephalitis, AQP4-IgG-positive neuromyelitis optica spectrum disorder (NMOSD), or febrile infection related epilepsy syndrome and used as the control group.

Diagnosis and phenotypic categorization of MOGAD

MOGAD was diagnosed based on the 2023 International MOGAD diagnostic criteria established by Banwell et al. [2]. Only patients with clear-positive MOG-IgG in sera were included in the study. For patients with CSF-restricted MOG-IgG positivity, eligibility additionally required the fulfillment of supporting clinical and/or MRI features and AQP4-IgG seronegativity [2]. We classified them into five clinical phenotypes (ADEM, CCE, ON, TM, and “others”) by two independent evaluators (E.O. and H.S.) based on current diagnostic recommendations [2]. Patients with ON and/or TM were further confirmed as AQP4 antibody-negative. When information on AQP4 antibody serostatus was unavailable, AQP4-IgG was assessed at our institute using indirect immunofluorescence (IIFT; EUROIMMUN, Lübeck, Germany; NMOSD Screen 1; lot T251229BL) with microscopic evaluation. The “others” group comprised MOG-IgG-positive patients who met the diagnostic definitions of cerebral monofocal or polyfocal deficits, or brainstem/cerebellar involvement. Any disagreement was resolved through discussion between them.

Measurements of CSF cytokines, chemokines, and related molecules

Concentration of cytokines/chemokines in the serum and CSF were determined using commercially available beads-based immunoassays (Bio-Plex Pro Human Chemokine Panel; Bio-Rad, Richmond, California and Luminex Human Discovery Assay; R&D systems, Minneapolis, MN), which can measure the following 12 cytokines, chemokines and related molecules simultaneously: interleukin (IL)-1 receptor antagonist IL-1ra, IL-6, CXCL-8, IL-17 A, IL-21, granulocyte-colony stimulating factor (G-CSF), IFN-γ, IFN-gamma-inducible protein-10 (IP-10 and CXCL-10), macrophage inflammatory protein-1α (CCL2), B cell attracting chemokine-1 (CXCL13), B cell activating factor belonging to the tumor necrosis factor family (BAFF), a proliferation inducing ligand (APRIL). The Bio-Plex Pro Human Chemokine Panel includes antibody conjugated beads for the following cytokines and chemokines: IL-6, CXCL8, IFN-γ, CXCL10, CCL2, and CXCL13. The Luminex Human Discovery Assay includes antibody-conjugated beads for the cytokines and chemokines, IL-1ra, IL-17 A, IL-21, G-CSF, BAFF, and APRIL.

The standard curve method was used to determine the concentration of each analyte. In the Bio-Plex assay, serum and CSF were diluted fourfold and twofold, respectively, whereas both serum and CSF were diluted threefold in the Luminex assay. When cytokine levels were below the detection limits, the values of the minimal detection limit were used for the analyses.

Assay of MOG IgG presence

Cell

A cell line expressing human MOG and a green fluorescent protein (GFP) fusion protein (MOG+ cells) was generated as previously described [10]. First, the full-length human MOG α1 sequence (NM_206809.3) was cloned from a frozen block of human brain tissue obtained from the MS and Parkinson’s Tissue Bank (London). The GFP sequence was then fused to the C-terminal of MOG α1 using a short linker, and the resulting construct was ligated into pcDNA5/FRT/TO (Life Technologies Japan, Tokyo). The MOG-GFP-expressing vector and Flp recombinase-producing pOG44 vector (Life Technologies Japan, Tokyo, Japan) were co-transfected into Flp-In 293 cells and screened with hygromycin B. To generate a control cell line (MOG– cells), intact pcDNA/FRT/TO and pOG44 were co-transfected into Flp-In 293 cells and screened as described above.

Measurement of anti-MOG autoantibody

We used a live cell-based flow cytometry assay to quantify human MOG antibodies. For quantitative evaluation, serum binding to MOG-expressing cells (MOG+ cells) and control cells (MOG− cells) were measured simultaneously following established protocols for reliable and observer-independent detection of MOG-IgG [2, 11]. MOG + and MOG− cells were mixed at a 1:1 ratio and prepared in microtubes to prepare aliquots (2 × 10⁵ cells/tube). After blocking with 10% fetal bovine serum, cell aliquots were incubated with diluted human serum (1:400) or human CSF (1:10) for 1 h. Cells were then incubated with Alexa fluor 647 labeled anti-human IgG (H + L) (Jackson ImmunoResearch, PA, United States) for 15 min and analyzed using a BD FACS Canto II flow cytometer (BD Biosciences, CA, United States). Data were analyzed using the FlowJo software (BD Biosciences, CA, United States). Debris and doublets were excluded, and GFP-positive (MOG+) and GFP-negative (MOG−) cell populations were identified based on FITC-GFP fluorescence (Supplementary Fig. A–D). The anti-MOG antibody binding was quantified as the MFI of APC in the GFP-positive cell population (Supplementary Fig. E). MOG-IgG reactivity was determined by comparing APC fluorescence intensity between MOG + and MOG− cells.

Definition of MOG-IgG positivity

MOG-IgG serostatus was defined using the established the MFI of MOG+ / MOG− ratio (MOG-IgG MFI ratio) > 1.5 as clear positive [2, 11]. To minimize diagnostic uncertainty, only patients with clear-positive MOG-IgG in sera were included in the study, whereas those with low-positive MOG-IgG were excluded. When examining correlations with cytokine levels, we used ΔMOG (MFI of MOG+ minus MFI of MOG−), which allowed for a more sensitive assessment of continuous variation in antibody binding. To further support the validity of the selected cutoff, sera with MOG-IgG MFI ratios between 1.5 and 2.0 were additionally tested by indirect immunofluorescence (IIFT; EUROIMMUN, Lübeck, Germany; NMOSD Screen 1; lot T251229BL) with microscopic evaluation, and all were confirmed to have MOG-IgG titers ≥ 100.

Statistics analysis

Statistical analyses were performed using R software version 4.2.2. For group comparisons, the Fisher’s exact test was used for categorical data.

Two-group comparisons

For each cytokine, the differences between the MOGAD and NIND groups were tested using the Mann–Whitney U test. Because 12 cytokines were tested in parallel, p-values were adjusted using the Benjamini–Hochberg false discovery rate (FDR) procedure. Statistical significance was defined as FDR-adjusted q < 0.05.

Comparisons across ≥ 3 clinical groups

Group differences in cytokine concentrations and log₁₀-transformed CSF-to-serum (C/S) ratios were assessed using the Kruskal–Wallis test. Because 12 cytokines/chemokines were tested in parallel, the p-values were adjusted using the Benjamini–Hochberg FDR procedure. When the FDR-adjusted Kruskal–Wallis test was significant (q < 0.05), pairwise group comparisons were performed using Dunn’s post-hoc test with Bonferroni adjustment. For figures, significance markers reflect Bonferroni-adjusted post-hoc p-values < 0.05.

Correlation analyses

Associations between cytokine/chemokine measures and CSF cell counts or ΔMOG were evaluated using Spearman’s rank correlation. For correlation analyses of the C/S ratios, both variables were log10-transformed. FDR correction (Benjamini–Hochberg) was applied only to the set of 12 correlations between cytokine C/S ratios and the ΔMOG C/S ratio.

In addition to p-values, effect sizes were calculated and reported to facilitate interpretation of the magnitude of the observed effects. Effect sizes are reported as η² and r for Kruskal–Wallis tests and pairwise comparisons, respectively. Representative effect sizes are presented in the main text, and complete effect size results are provided in the Supplementary Data.

Results

Among 626 consecutive patients whose serum and CSF samples were sent to our laboratory from January 1, 2015 to April 30, 2024 (Supplementary Fig. 1), 193 patients were excluded due to either lacking paired serum and CSF samples (n = 77) or receiving immunotherapy before sampling (n = 116). MOG-IgG MFI ratio were measured in both the serum and CSF of 433 patients, and 50 patients were positive for MOG-IgG in their serum and/or CSF and met the 2023 International MOGAD diagnostic criteria. AQP4-IgG was negative in all patients with ON and/or TM presentations (including ON + TM) and in all CSF-restricted MOG-IgG-positive cases (n = 6), and no AQP4-IgG/MOG-IgG double-seropositive patients were identified. Finally, we classified the patients into five phenotypes: ADEM (n = 19), CCE (n = 9), ON + TM (n = 3), ON (n = 9), and “others” (n = 10). Fifty-one patients met the inclusion criteria for the NIND group.

Table 1 shows the characteristics of patients with MOGAD, including 19 (38%) with ADEM, 9 (18%) with CCE, 3 (6%) with ON + TM, 9(18%) with ON, and 10 (20%) with others (four with brainstem and cerebellar deficits, and six with cerebral monofocal or polyfocal deficits). The NIND group comprised 27 patients with acute psychosis, 16 patients with refractory epilepsy, 5 with infection-triggered encephalopathy syndrome, 2 patients with cerebellar ataxia, and 1 patient with dyskinesia. The median interval between the onset of clinical symptoms and sampling in the entire MOGAD and NIND group was 9.5 days (interquartile range [IQR]: 2–19.5) and 8.0 days (IQR: 2.0–30.0 days), respectively. The median interval between serum and CSF sampling in these two groups was 0 day (range: 0–5 days) and 1 day (range: 0–4 days), respectively. Although samples were obtained from seven patients during relapse, all samples were collected during the active phase. The median intervals from the onset of clinical symptoms to sampling in the ADEM, CCE, and ON group were 1.5, 18.5, and 8 days, respectively.

Table 1.

Characteristics of MOGAD phenotype

	ADEM	CCE	ON + TM	ON	Others	NIND
n	19	9	3	9	10	51
Age (year at sample collection)	4.0 [3.0, 8.5]	13.0 [9.0, 13.0]	10.0 [9.5, 10.5]	11.0 [7.0, 13.0]	12.0 [4.8, 13.8]	9.0 [5.0, 13.0]
Sex: Male	11 (58)	6 (67)	2 (67)	4 (44)	2 (20)	26 (51)
Clinical symptoms
Seizure	6 (32)	9 (100)	0 (0)	1 (11)	1 (10)	24 (47)
Headache	6 (32)	6 (67)	1 (50)	1 (11)	4 (40)	6 (12)
MOG-IgG status
Serum restricted positivity	5 (26)	2 (22)	1 (33)	5 (56)	3 (30)	0
CSF restricted positivity	1 (5)	2 (22)	1 (33)	1 (11)	1 (10)	0
Both positivity	13 (68)	5 (56)	1 (33)	3 (33)	6 (60)	0
CSF oligoclonal bands
positive	5 (26)	2 (22)	1 (33)	1 (11)	3 (30)	2 (4)
negative	9 (47)	2 (22)	1 (33)	7 (79)	5 (50)	28 (55)
not tested	3 (16)	0 (0)	0 (0)	0 (0)	1 (10)	21 (41)
CSF cell count (/µl)	19.0 [9.0, 34.5]	57.0 [42.0, 98.0]	52.0 [50.0, 54.0]	2.0 [2.0, 2.0]	12.5 [3.8, 27.0]	1.0 [1.0, 3.0]
CSF protein (mg/dl)	34.0 [26.5, 40.8]	39.0 [35.0, 52.0]	70.5 [52.8, 88.3]	31.0 [23.0, 36.0]	32.0 [24.2, 44.3]	19.0 [16.0, 29.0]
NMDAR Ab positivity	2 (11)	3 (33)	0 (0)	0 (0)	1 (10)	0 (0)
Interval between onset of symptoms and sampling* (days)	1.5 [1.0, 20.0]	18.5 [2.0, 30.8]	28.5 [3.0, 32.0]	8.0 [2.5, 11.5]	3.5 [2.0, 33.8]	8.0 [2.0, 30.0]

Data are median [interquartile range], n (%)

All values are rounded to one decimal place unless otherwise indicated

Percentages are rounded to the nearest whole number

ADEM acute disseminated encephalomyelitis, CCE cerebral cortical encephalitis with seizures, NMOSD neuromyelitis optica spectrum disorder, ON optic neuritis, NIND non-inflammatory neurological diseases, CSF cerebrospinal fluid

^*In the event that the date of serum or cerebrospinal fluid collection differs, the subsequent collection was utilized

First, we compared cytokine and chemokine levels between the MOGAD and NIND groups. In the CSF, the MOGAD group demonstrated significantly higher levels of Th17-related cytokines/chemokines (IL-6 (p < 0.0001, q < 0.0001, r = 0.69), CXCL8 (p < 0.0001, q < 0.0001, r = 0.81), G-CSF(p < 0.0001, q < 0.0001, r = 0.85), IL-17 A(p < 0.0001, q < 0.0001, r = 0.86), and IL-21(p < 0.0001, q < 0.0001, r = 0.86); Fig. 1A), B cell-related cytokines/chemokines (CXCL13 (p < 0.0001, q < 0.0001, r = 0.77), BAFF (p < 0.0001, q < 0.0001, r = 0.82), APRIL(p < 0.0001, q < 0.0001, r = 0.85); Fig. 1B), and IL-1ra (p < 0.0001, q < 0.0001, r = 0.79) (Fig. 1D) than the NIND group. CCL2 levels in the NIND group (p < 0.0001, q < 0.0001, r = 0.47) were higher than those in the MOGAD group. Serum concentrations of Th17-related cytokines/chemokines (IL-6 [p < 0.05, q < 0.05, r = 0.73], CXCL8 [p < 0.05, q < 0.05, r = 0.65], IL-17 A [p < 0.05, q < 0.05, r = 0.48], and IL-21 [p < 0.05, q < 0.05, r = 0.57; Fig. 2A]), B cell-related cytokines/chemokines (CXCL13 [p < 0.05, q < 0.05, r = 0.60], BAFF [p < 0.05, q < 0.05, r = 0.55], and APRIL [p < 0.05, q < 0.05, r = 0.77], Fig. 2B), and Th1-related cytokines/chemokines (IFN-γ[p < 0.05, q < 0.05, r = 0.52], CXCL10 [p < 0.05, q < 0.05, r = 0.21], CCL2 [p < 0.05, q < 0.05, r = 0.53], and IL-1ra [p < 0.05, q < 0.05, r = 0.29; Fig. 2D) were significantly elevated in the MOGAD group. Conversely, no significant differences in CSF Th1-related cytokines (IFN-γ and CXCL10) were observed between the MOGAD and NIND groups (Fig. 1C).

Fig. 1.

Comparison of CSF cytokine/chemokine concentrations within the MOGAD phenotypes (ADEM, CCE and ON) and NIND groups. The concentration of Th17-related (IL-6, CXCL8, G-CSF, IL-21, and IL-17 A, A), B-cell-related (CXCL13, BAFF, and APRIL, B), Th1-related (IFNγ and CXCL10, C), and broad-spectrum (IL-1ra and CCL2, D) cytokines/chemokines in the entire MOGAD group compared with the OND group. The dots in the graph represent CSF cytokine/chemokine concentrations in each patient. Significance was assessed using Benjamini–Hochberg FDR-adjusted p values (q < 0.0001). CSF, cerebrospinal fluid; MOGAD, myelin oligodendrocyte glycoprotein antibody-associated disease; ADEM, acute disseminated encephalomyelitis; ON, optic neuritis; CCE, cerebral cortical encephalitis; BAFF, B cell-activating factor belonging to the tumor necrosis factor family; NIND, non-inflammatory neurological disease

Fig. 2.

Comparison of serum cytokine/chemokine concentrations within the MOGAD phenotypes (ADEM, CCE, and ON) and NIND groups. The concentration of Th17-related (IL-6, CXCL8, G-CSF, IL-21, and IL-17 A, A), B-cell-related (CXCL13, BAFF, and APRIL, B), Th1-related (IFNγ and CXCL10, C), and broad-spectrum (IL-1ra and CCL2, D) cytokines/chemokines in the entire MOGAD group compared with the OND group. Dots in the graph represent serum cytokine/chemokine concentrations in each patient. Significance was assessed using Benjamini–Hochberg FDR-adjusted p values (q < 0.05). CSF, cerebrospinal fluid; MOGAD, myelin oligodendrocyte glycoprotein antibody-associated disease; ADEM, acute disseminated encephalomyelitis; ON, optic neuritis; CCE, cerebral cortical encephalitis; NIND, non-inflammatory neurological disease

Second, we compared the concentrations of cytokines/chemokines in the CSF and serum among the three MOGAD phenotypes, (ADEM, CCE, and ON). The ON + TM group was excluded from the comparative analysis of different clinical phenotypes because the number of patients was insufficient, (n = 3). The “others” group was excluded because it represented a heterogeneous group. CSF concentrations of IL-6 and CXCL8 were higher in the ADEM and CCE groups than in the ON group, (ADEM vs. ON padj < 0.05; CCE vs. ON padj < 0.05 for both cytokines), with substantial overall group effects, (Kruskal–Wallis η²=0.29 for IL-6 and 0.28 for CXCL8), (Fig. 3A). By contrast, no significant differences in serum cytokine concentrations were observed across phenotypes for any of the 12 cytokines, (Fig. 4). We performed a four-group comparison across clinical phenotypes, (ADEM, CCE, and ON) and NIND, (Supplementary Figs. 3 and 4). All three MOGAD phenotypes showed significantly higher CSF levels of CXCL8, G-CSF, IL-17, IL-21, CXCL13, BAFF, APRIL, and IL-1ra than the NIND group, (padj < 0.05). Serum IL-6, BAFF, and APRIL levels were significantly higher in all three phenotypes than in NIND group, (padj < 0.05, for all phenotype–NIND comparisons). In addition, most phenotype-dependent differences were confined to ADEM and/or CCE, including higher CSF IL-6 and serum CXCL8, IL-17, IL-21, IFN, IL-1ra, and CCL2 levels than in NIND group. Serum CXCL13 was an exception, being elevated in ADEM and ON compared with NIND group, (padj < 0.05 for the indicated comparisons). Cytokine differences varied across MOGAD phenotypes.

Fig. 3.

Comparison of CSF cytokine/chemokine concentrations within the MOGAD phenotypes (ADEM, CCE, and ON). The concentration of Th17-related (IL-6, CXCL8, G-CSF, IL-21, and IL-17, A), B-cell-related (CXCL13, BAFF, and APRIL, B), Th1-related (IFNγ and CXCL10, C), and broad-spectrum (IL-1ra and CCL2, D) cytokines/chemokines across MOGAD phenotypes (ADEM, CCE, and ON groups). Dots in the graph represent CSF cytokine/chemokine concentrations in each patient. Statistical significance was assessed using Bonferroni-adjusted post-hoc p-values from Dunn’s test (p_adj < 0.05). CSF, cerebrospinal fluid; MOGAD, myelin oligodendrocyte glycoprotein antibody-associated disease; ADEM, acute disseminated encephalomyelitis; ON, optic neuritis; CCE, cerebral cortical encephalitis

Fig. 4.

Comparison of serum cytokine/chemokine concentrations within the MOGAD phenotypes (ADEM, CCE, and ON). The concentration of Th17-related (IL-6, CXCL8, G-CSF, IL-21, and IL-17, A), B-cell-related (CXCL13, BAFF, and APRIL, B), Th1-related (IFNγ and CXCL10, C), and broad-spectrum (IL-1ra and CCL2, D) cytokines/chemokines across MOGAD phenotypes (ADEM, CCE, and ON groups). The dots in the graph represent serum cytokine/chemokine concentrations in each patient. Statistical significance was assessed using Bonferroni-adjusted post-hoc p-values from Dunn’s test (p_adj < 0.05). CSF, cerebrospinal fluid; MOGAD, myelin oligodendrocyte glycoprotein antibody-associated disease; ADEM, acute disseminated encephalomyelitis; ON, optic neuritis; CCE, cerebral cortical encephalitis

Finally, we evaluated the C/S ratio for each cytokine in individual patients with ADEM, CCE, or ON phenotypes, (Fig. 5). No significant differences were observed between the three groups. Given their significant differences across phenotypes, (post-hoc Dunn’s test), (Fig. 3), we further examined whether CSF IL-6 and CXCL8 correlated with CSF cell counts and CSF ΔMOG. CSF IL-6 and CXCL8 concentrations exhibited a positive correlation with CSF cell counts, (Spearman’s ρ = 0.64, p < 0.05, q < 0.05) and, (Spearman’s ρ = 0.45, p < 0.05, q < 0.05), respectively, (Fig. 6A). Additionally, a positive correlation was identified between CSF IL-6 and CSF ΔMOG, (Spearman’s ρ = 0.35, p < 0.05, q < 0.05), whereas CSF CXCL8 was not significantly correlated with CSF ΔMOG, (Spearman’s ρ = 0.16, p = 0.27, q = 0.27), (Fig. 6A). No significant correlation was identified between serum ΔMOG and CSF ΔMOG, (Spearman’s ρ = 0.21, p = 0.14), (Fig. 6B). We also explored the relationship between the C/S ratios of cytokines/chemokines and the MOG-IgG C/S ratio. A strong positive correlation was observed exclusively for CXCL13, (Spearman’s ρ = 0.53, p < 0.05, q < 0.05), whereas no significant correlations were detected for the other cytokines, (Fig. 6C). We compared CSF cell counts and MOG status between the ADEM, CCE, and ON groups, (Fig. 6D). The ADEM and CCE groups showed significantly higher CSF cell counts than the ON group, (ADEM vs. ON, padj < 0.05; CCE vs. ON, padj < 0.05), with a substantial overall group effect, (Kruskal–Wallis η² = 0.38). Serum ΔMOG was higher in ADEM than in CCE, (ADEM vs. CCE, padj < 0.05; Kruskal–Wallis η² = 0.24), whereas no significant difference was observed in CSF.

Fig. 5.

Comparison of log₁₀-transformed CSF-to-serum ratios of cytokine/chemokine concentrations within the MOGAD phenotypes (ADEM, CCE, and ON). CSF-to-serum ratios were log₁₀-transformed. Panels show Th17-related (IL-6, CXCL8, G-CSF, IL-21, and IL-17, A), B-cell-related (CXCL13, BAFF, and APRIL, B), Th1-related (IFNγ and CXCL10, C), and broad-spectrum (IL-1ra and CCL2, D) cytokines/chemokines in each MOGAD phenotypes (ADEM, CCE, and ON groups). The dots in the graph represent individual patients. Statistical significance was assessed using Bonferroni-adjusted post-hoc p-values from Dunn’s test (p_adj < 0.05). CSF, cerebrospinal fluid; MOGAD, myelin-oligodendrocyte glycoprotein antibody-associated disease; ADEM, acute disseminated encephalomyelitis; ON, optic neuritis; CCE, cerebral cortical encephalitis; Log₁₀ CSF/serum, log₁₀-transformed CSF-to-serum ratio

Fig. 6.

Correlations between cytokines, CSF cell counts, and ΔMOG in MOGAD, including CSF, serum, and CSF/serum ratio associations and phenotype comparisons. A The correlation between CSF cytokines (IL-6 and CXCL8) concentration and CSF cell counts and between CSF cytokines (IL-6 and CXCL8) and CSF ΔMOG. Benjamini–Hochberg FDR correction was applied per outcome across the two cytokines, and significance was defined as q < 0.05. B The correlation between CSF (x-axis) and serum (y-axis) ΔMOG. C The correlation between ΔMOG CSF-to-serum ratio (log₁₀) (x-axis) and CXCL13 CSF-to-serum ratio (log₁₀) (y-axis). P-values were BH-FDR adjusted across 12 screened cytokine/chemokine CSF-to-serum ratios. CXCL13 is shown as the representative significant association; full results are available in the Supplementary Data. D A comparison of CSF cell counts and ΔMOG among the ADEM, CCE, and ON phenotypes. Statistical significance was assessed using Bonferroni-adjusted post-hoc p-values from Dunn’s test (p_adj<0.05). The dots in the graph represent CSF cytokine/chemokine concentrations in each patient. CSF, cerebrospinal fluid; ADEM, acute disseminated encephalomyelitis; ON, optic neuritis; CCE, cerebral cortical encephalitis; OND, other neurological disease; CSF/serum ratio, CSF-to-serum ratio

Discussion

We identified several features of serum and CSF cytokine profiles in patients with MOGAD. First, consistent with previously reported findings, CSF analyses revealed significant upregulation of Th17 and B cell-related chemokines/cytokines in the entire MOGAD phenotype compared with the OND group [8, 9]. Although elevated Th17- and B-cell-related cytokine levels in MOGAD have been reported previously, our study expands these findings by examining phenotype-associated patterns in a pediatric cohort, including encephalitic presentations such as CCE. Our findings suggest limited heterogeneity in the cytokine/chemokine responses across the clinical phenotypes of pediatric MOGAD. Nevertheless, encephalitic phenotypes such as ADEM and CCE generally showed higher CSF IL-6 and CXCL8 levels than ON, suggesting relatively increased intrathecal inflammatory activity. In addition, the CXCL13 C/S ratio showed a positive correlation with ΔMOG C/S ratio, supporting a link between this B cell-related chemokine activity and intrathecal antibody dynamics.

Our findings regarding the Th17 cell (IL-6, CXCL8, G-CSF, IL-17 A, and IL-21) and B cell (CXCL13, BAFF and APRIL)-dominant cytokine/chemokine profiles with MOGAD lend support to previous results [8, 9, 12, 13]. The pathogenicity of MOG antibodies remains unclear. Human MOG antibodies exert cytotoxic effects on MOG-expressing cells through multiple mechanisms, including complement-dependent cytotoxicity and antibody-dependent cellular phagocytosis [10, 14]. Thus, MOG antibodies may play a direct pathogenic role. Although rituximab decreases the annualized relapse rate in MOGAD [15], the efficacy of CD20-mediated B cell depletion in MOGAD is occasionally incomplete, and one-third of patients exhibit relapses after rituximab treatment [16, 17]. Thus, factors other than MOG-IgG and B cells may be involved in the disease activity of MOGAD. A study employing purified MOG-IgG isolated from patients with MOGAD in an adoptive transfer experimental autoimmune encephalomyelitis (EAE) model concluded that the pathogenicity of MOG antibodies is secondary, and the degree of interaction between MOG antibodies and MOG-specific T cells may serve as a critical factor in determining the extent of disease pathology [18]. IL-17, a pleiotropic cytokine produced by Th17 cells, plays a crucial role in the induction of EAE in mice [19–21]. A high ratio of Th17 to Th1 cells within the perivascular space (i.e., the area between the endothelial basement membrane and glial limitans in the CSF) contributes to the disruption of the blood–brain barrier (BBB) in a mouse model [22, 23]. Furthermore, Th17 cells induce the formation of robust tertiary lymphoid tissues within the brain meninges, which are associated with local demyelination during EAE [20]. In humans, AQP4-positive NMOSD has been demonstrated to exhibit a substantial proportion of Th17 cells and elevated levels of IL-17, which play an important role in immune pathogenesis [24].

Whether these cytokines/chemokines are associated with the diverse phenotypes in MOGAD was examined. Clinically, MOGAD encompasses heterogeneous phenotypes with potentially different risks for relapse and neurocognitive outcomes, particularly encephalic presentations [3–5]. Our observation that CSF IL-6 and CXCL8 levels were higher in encephalitic phenotypes than in ON may reflect differences in the degree of LMI. IL-6 is a pleiotropic cytokine that plays a crucial role in both Th17 and B cells.IL-6 has been identified as a pivotal switch factor that directs antigen-activated naïve CD4⁺ T cells toward a pro-inflammatory Th17 cell lineage and inhibits regulatory T cell differentiation [25]. IL-6 also induces B cell differentiation into plasma cells, maintains and promotes plasma cell survival, and maintains antibody production [25, 26]. The observed correlation between CSF IL-6 and CSF MOG-IgG or CSF cells suggests that MOG-reactive cells infiltrating the periphery promote antibody production in the CSF. Thus, an increase in IL-6 can affect both the Th17 shift and the synthesis of MOG antibodies in MOGAD, implying the role of IL-6 as a master disease promoter in MOGAD. In support of this, the efficacy of tocilizumab in patients with treatment-refractory MOGAD has been demonstrated, as evidenced by a reduction in relapse [27]. Anti IL-6 receptor therapy utilizing satralizumab as a preventative strategy in MOGAD is now being enrolled [28].

MOG-IgG in CSF is attracting increasing attention owing to its origin and diagnostic significance. AQP4-IgG levels in the serum and CSF are often correlated in NMOSD, which is consistent with a model in which peripherally produced antibodies gain access to the CNS during BBB disruption [6, 7]. By contrast, serum and CSF MOG-IgG titers exhibit no or only weak correlation [6, 7], which fact was also observed in the present study. This dissociation provides supportive evidence of intrathecal MOG-IgG production. In this study, we used the ΔMOG C/S ratio as a relative measure of CSF versus systemic predominance of MOG-IgG. CXCL13 was the only cytokine whose C/S ratio correlated with the MOG C/S ratio. CSF CXCL13 demonstrates a stronger correlation with intrathecal immunoglobulin production and the presence of CSF plasma cells or plasmablasts than with other B cell-attracting chemokines [29, 30]. Additionally, CXCL13 plays a role in the formation of ectopic lymphoid structure(s) (ELS) and facilitates the migration of Th 17 cells [31]. These observations indicate that CXCL13 contributes to ectopic antibody production in the brain under autoimmune conditions and suggest its possible involvement in the intrathecal production of MOG-IgG in some patients.

ELS in secondary progressive multiple sclerosis (MS) consists of meningeal B-cell aggregates characterized as ectopic follicles with germinal centers. This structure contributed to persistent intrathecal antibody production. The presence of ELS has been associated with early onset of the disease and more severe cortical pathology [32, 33]. Research on MS has demonstrated that despite the absence of ELS, higher LMI and elevated cytokines that promote B cell recruitment and activation have been associated with cortical demyelination and antibody production [34–36]. LMI is defined as the presence of inflammatory cells in the leptomeninges, irrespective of the level of organization of specific cell populations [34]. The pathological spectrum observed in LMI ranges from disorganized aggregates of immune cells to well-organized aggregates of immune cells, as seen in secondary progressive MS [20, 37, 38]. Although MOGAD does not manifest a progressive course, LMI-like structures have been identified in its pathology; the demyelinating areas in the corticomedullary junctions in MOGAD were observed to be accompanied by T cell-dominant inflammatory cell infiltration at the meninges, which can be termed disorganized LMI [39]. The role of these meningeal LMI-like structures in MOGAD pathogenesis remains a subject for future research.

Most meningeal B cells are derived from the skull bone marrow and are distinct from blood-borne B cells [40]. The dura is a site of B-cell residence and is potentially involved in both homeostasis and neuroinflammation [41]. Although blood is considered the source of B cells infiltrating the brain in demyelinating diseases such as ADEM, it is reasonable to assume that B cells derived from the skull bone marrow also play a role in the production of MOG-IgG. Moreover, skull bone marrow reflects inflammatory brain responses with a disease-specific spatial distribution in human neurological disorders [42]. The differential role of blood- and skull bone marrow-derived B cells in MOGAD is an interesting topic to be elucidated in the future.

In our study, we collected serum and CSF samples almost simultaneously during the active phase from all enrolled patients, which enabled us to quantitatively compare antibody titers and inflammation-related molecules in both samples. Prior studies have proposed corrected C/S indices, such as the MOG-IgG-specific antibody index, to infer intrathecal MOG-IgG synthesis, which is an independent predictor of poor outcomes in MOGAD [43, 44]. Cerebral lesions have been reported more frequently in pediatric patients with MOGAD than in adults [2]. Cerebral lesions are also associated with an increased risk of relapse and cognitive deficits, raising concerns regarding the long-term consequences of recurrent intracranial inflammation [4, 5]. A substantial proportion of children with relapsing MOGAD experience cognitive difficulties, which may affect school performance and slow information processing, with a greater impact on verbal functions [3, 45]. Together, these observations indicate that intrathecal MOG-IgG synthesis may be linked to intracranial inflammatory activity and thereby influence clinical outcome in some patients. This highlights the need to focus on objective intrathecal markers, including the C/S ratios of MOG-IgG and cytokines/chemokines, as well as corrected indices of intrathecal synthesis, in future longitudinal studies.

Limitations

This study has some limitations. First, this was a retrospective study with a modest sample size, partly because we included only patients for whom paired serum and CSF samples were available and excluded low-positive cases. Second, the number of NIND controls was limited, reflecting the practical difficulty of obtaining pediatric CSF samples from non-inflammatory conditions; therefore, we report effect sizes in addition to FDR-corrected p-values to facilitate the interpretation of the magnitude of group differences. Third, the phenotype-stratified analyses were constrained by small subgroup sizes and were likely powered to detect only large differences; thus, modest phenotype-related effects may have been missed. Fourth, we did not evaluate longitudinal outcomes or treatment responses; thus, the prognostic and therapeutic implications of the cytokine profiles remain uncertain. Fifth, MOG-IgG serostatus was defined using a flow cytometry-based MOG-IgG MFI ratio, but MFI-based cutoffs are not yet fully standardized; consequently, some low-positive cases may have been misclassified or excluded, potentially limiting generalizability. Finally, although all samples were collected during the acute phase, the interval between symptom onset and sampling varied, and some samples were obtained during relapse.

Conclusion

In pediatric MOGAD, serum and CSF cytokine/chemokine profiles show limited heterogeneity across clinical phenotypes. LMI may help characterize encephalitic presentations relative to other phenotypes. In addition, the CXCL13 C/S ratio was associated with the relative CSF predominance of MOG-IgG, supporting further investigation of objective intrathecal markers for risk stratification and treatment planning in longitudinal cohorts.

Authors’ contributions

EO and HS designed the study, collected, analyzed and interpreted the data, and wrote the manuscript draft.KK contributed to the acquisition of data.TM, SMa, HN, SMi, IH, EO, YA, NY, MS, KM, MF, TI, and IK contributed to the acquisition and provision of patient details.HS obtained funding.All the authors revised the manuscript and finally approved the submitted version.

Funding

This research was supported by JSPS KAKENHI (22K07831 and 25K11116 [HS]).

Data availability

The anonymized patient data are not publicly shared as they are being utilized for the Pediatric Inflammatory and Autoimmune Neurological Disorders Registry study. All data underlying the reported results will be made available after the anonymization of patient information. The data will be made available after the publication of this article. The authors will share the data with qualified investigators whose proposals for data use have been approved by an independent review committee.

Declarations

Ethics approval and consent to participate

This study was conducted in accordance with the ethical standards of the institutional and/or national research committee and the 1964 Declaration of Helsinki and its later amendments or comparable ethical standards. The study was approved by the Ethics Committee and Institutional Review Board of the Tokyo Metropolitan Institute of Medical Science (approval number #15 − 3, 18 − 3, 21 − 2). Informed consent was obtained from the guardians of all the patients.

Consent for publication

Not applicable.

Competing interests

Hiroshi Sakuma received an honorarium from Chugai Pharmaceutical Co. Ltd. The authors declare no conflict of interest.