Altered lipidomic and metabolomic status in cerebrospinal fluid of children with myelin oligodendrocyte glycoprotein antibody-associated disorder

Abstract

Myelin oligodendrocyte glycoprotein antibody-associated disorder (MOGAD) is a group of acquired demyelinating syndromes affecting the central nervous system. MOGAD-associated bioactive molecules remain elusive. A retrospective case-control study was performed to characterize the biochemical and immunological profiles of cerebrospinal fluid (CSF) in MOGAD. Thirteen patients with MOGAD (onset age: 2–14 years, 6 females) and five patients with epilepsy, serving as controls, were enrolled. Liquid chromatography with tandem mass spectrometry was used for lipidomic and metabolomic analyses using CSF samples collected at disease onset (n = 5). The MS/MS system detected a total of 7527 molecules in lipidomic and 17,526 molecules in metabolomic analyses of CSF. Among them, 162 (0.02 %) lipophilic molecules were detected at levels that differed from those of controls. Among the 549 (0.03 %) hydrophilic molecules that were differentially presented, pyridoxine, ribitol, and isethionate levels were significantly lower in patients with MOGAD. Both lipidomic and metabolomic analyses, discriminated CSF samples of patients with MOGAD from controls using Uniform Manifold Approximation and Projection. In summary, CSF samples from children with MOGAD exhibit distinctive lipidomic and metabolomic profiles. These findings provide evidence for the diagnostic potential of CSF-based lipidomic and metabolomic analyses for childhood-onset MOGAD.

Highlights

• •MOGAD shows wide phenotypic spectrum in children.
• •Serum MOG-IgG correlates with CSF IL-10 and IL-17A.
• •HPLC identifies distinct molecular CSF profiles in MOGAD.
• •UMAP differentiates the molecular profiles of MOGAD and control CSFs.

1. Introduction

The pathogenic role of serum antibodies against myelin oligodendrocyte glycoprotein (MOG) has gained increasing interest over the past 10 years [1]. The phenotypic spectrum of neuroinflammation in the central nervous system (CNS) associated with MOG-IgG includes a wide variety of symptoms, collectively known as MOG antibody-associated disease (MOGAD) [2]. Owing to the ease at which cell-based assays can be applied in antibody detection [3], a growing number of patients with MOGAD have been reported to show either or a combination of optic neuritis, acute disseminated encephalomyelitis (ADEM), transverse myelitis, and cortical encephalitis at the onset [4].

The seroprevalence of MOG-IgG is higher in the pediatric population than in adults (30–40 % vs. 6.5–20 %) [3,[5], [6], [7]]. Emerging evidence suggests that children are more susceptible to insults from infectious and environmental exposures, resulting in overactivity of the innate immune response and an increased risk of certain autoimmune diseases [8,9]. Furthermore, pediatric-onset MOGAD shows greater heterogeneity in clinical presentations than adult patients [5,10]. Thus, the diagnosis of a specific phenotype at initial presentation can be challenging, as neurological symptoms may change over time. Although 70 % of patients with pediatric-onset MOGAD show a monophasic course [7], some demonstrate relapsing courses with multiple clinical attacks resembling multiple sclerosis and aquaporin-4 (AQP4)-IgG-seropositive neuromyelitis optica spectrum disorder (NMOSD) [[11], [12], [13]].

The pathological effects of MOG-IgG remain to be elucidated. MOG, exclusively expressed on the outer lamellae of oligodendrocyte membranes [14], was initially identified as a candidate antigenic target of demyelinating antibodies in experimental autoimmune encephalomyelitis [15,16]. Subsequent studies using experimental models of autoimmune encephalomyelitis have verified the substantial role of MOG-IgG in T cell and B cell responses [17]. Thus, it is widely accepted that MOG-IgG is associated with heterogeneous, age-dependent clinical features of acquired demyelinating syndromes [18].

Several reports have focused on the presence of lipids that reflect the hyperactive status of the innate and acquired immune systems during the acute phase of systemic inflammation [9] and demyelinating diseases [[19], [20], [21], [22]]. These data suggest that pediatric-onset MOGAD has a disease-specific signature of bioactive molecules in cerebrospinal fluid (CSF). We hypothesize that metabolites in the CSF act as damage-associated molecular patterns (DAMPs) [23], contributing to the initiation and perpetuation of the intrinsic immune response in the pathophysiology of MOGAD [24]. We herein present the key findings in the lipidomic and metabolomic profiles of CSF for children with MOGAD.

2. Material and methods

2.1. Patients

We recruited 13 children (age 2–14 years; 6 females [46 %]) who presented with the first episode of acquired demyelinating syndrome between 2016 and 2020. The enrolled patients were admitted to two hospitals (Fukuoka Children's Hospital and Kyushu University Hospital), both of which provide tertiary level medical services for children in Fukuoka and surrounding areas in Japan. All serum samples tested positive for MOG-Abs. Demographic information (age at the onset and sex) and clinical data (phenotypic presentation, laboratory data, and MRI findings) were retrospectively evaluated for each child. Relapses and outcomes, evaluated according to the Expanded Disability Status Scale (EDSS) score [25] at the last follow-up examination were reported by the attending pediatricians. We applied the criteria proposed by the International MOGAD Panel for the diagnosis of MOGAD [4], and the final clinical diagnosis of demyelinating phenotypes was described based on the EU Pediatric MOG Consortium consensus [11]. Due to the difficulty of obtaining CSF samples from healthy pediatric controls, age-matched patients with epilepsy (n = 5, age 4–9 years, 3 females [60 %]) were recruited as non-inflammatory neurological disease controls for lipidomic and metabolomic analyses (Supplementary Fig. 1). All patients with epilepsy were ruled out for structural, immune, metabolic, and infectious etiologies.

2.2. Assays for MOG-Abs and cytokines

All CSF and serum samples were frozen immediately after collection at −30 °C until use. We tested MOG-IgG positivity in serum samples obtained at neurological onset using a fixed immunofluorescence cell-based assay (CBA; #Cat. FA 1156-1005-50, Euroimmun, Lübeck, Germany) by visual observation of fluorescein-labelled binding MOG-IgG [26]. For MOG-IgG-positive patients, the antibody titer was quantitatively evaluated in duplicate using an enzyme-linked immunosorbent assay (ELISA) (#Cat. AS-55153-H, AnaSpec, Fremont, CA, U.S.A.) according to the manufacturer's protocol [27]. For quantification of cytokines in concurrent serum and CSF samples, we utilized an ultrasensitive S-PLEX assay that can detect cytokines present at fg/mL levels, analyzed on a QuickPlex SQ 120 instrument (#Cat. K15396S, Meso Scale Discovery, Rockville, MD, U.S.A.). The electrochemiluminescence technology employs SULFO-TAG labels that emit light upon electrochemical stimulation initiated at the electrode surfaces of MULTI-ARRAY microplates [28]. The tested cytokines included interleukin (IL)-1β, IL-2, IL-4, IL-6, IL-10, IL-12p70, IL-17A, interferon (IFN)-γ, and tumor necrosis factor (TNF)-α. These tests were performed by trained investigators who had no access to the clinical data (blinded condition).

2.3. Lipidomics

A lipidomic analysis was performed as previously described [9]. In brief, CSF samples obtained at the acute stage were thawed, centrifuged, aliquoted, and stored at −30 °C prior to the analysis. After lipid extraction using the Bligh and Dyer method [29], samples were injected into a high-performance liquid chromatography (HPLC) system (Agilent 1260; Agilent Technologies, Palo Alto, CA, U.S.A.) with a Poroshell 120 EC-C18 column (Agilent Technologies) after the addition of three internal standards: 1,3(d5)-dieicosanoyl-2-(11Z-eicosenoyl)-glycerol, 1,3(d5)-dinonadecanoyl-glycerol, and 1,3(d5)-ditetradecanoyl-glycerol. Quality control samples were prepared by pooling samples and analyzed to verify the measurement accuracy [30]. Tandem mass spectrometry (MS/MS) analyses were performed with electrospray ionization quadrupole time-of-flight MS 6545 (Agilent Technologies), and a single run was conducted to collect both positive and negative ionized lipid species (scan range 100–1700 m/z). The retention time (RT), m/z values, and peak areas were calculated using a marker analysis (LSI Medience, Japan) [9]. Errors were corrected using internal standards and noise peaks were deleted accordingly. Lipid identification was performed by comparing RT and m/z with the in-house standard dataset. When a previously known compound matched these profiles, then the molecule was classified as high-confidence molecule.

2.4. Metabolomics

The samples from the same cohort used for lipidomics were prepared with acetonitrile as previously described [9]. A liquid chromatography-mass spectrometry (LC-MS) analysis was performed using an HPLC system (Agilent 1260; Agilent Technologies) equipped with a C18 column (2 μm, 2.0 mm i.d. × 50 mm, CAPCELL PAK C18F; Shieseido, Tokyo, Japan), and coupled with electrospray ionization quadrupole time-of-flight MS 6545 (Agilent Technologies). We operated the MS device on a single run in positive and negative scan modes (m/z range 60–1200) with a capillary voltage of 3500 V.

2.5. Phosphatidylcholines and oxidized phosphatidylcholines

Among the detected phosphatidylcholines (PCs) from the CSF samples, we carefully reviewed the chromatogram and MS data to identify the presence of oxidized phosphatidylcholine (OxPC) species. Molecular species derived from PCs carrying functional groups of hydroxides (-OH), epoxides (=O), and hydroperoxides (-OOH) were profiled as OxPCs [31,32]. The MassHunter software program (https://www.agilent.com/en/products/software-informatics/mass-spectrometry-software) was used for data acquisition, and those with RT and m/z values identical to the theoretical value of the corresponding PCs on the MS/MS spectra were defined as oxidized PCs [33].

2.6. Statistical analysis

The original MS data were converted to the comma-separated value (CSV) format using a CSV Data Converter (Agilent Technologies). All peak intensities were subjected to Student's t-test. Two molecules showing differences in RT values ≤ 0.3 min and m/z values < 0.015 (for those with an m/z of <500) or <0.02 (m/z ≥ 500) were considered to have the same chemical moiety. The Mann-Whitney U test and Kruskal-Wallis rank sum test were used to compare numerical variables. Two-tailed p-values of <0.05 were considered to indicate statistical significance. Pearson's correlation coefficient was used to assess the relationship between various cytokines and serum MOG concentrations. Statistical analyses were performed using R (ver. 4.3.2; https://cran.r-project.org/). After conversion of the peak signals to Z-score, we performed hierarchical clustering and dimension-reducing procedures with Uniform Manifold Approximation and Projection (UMAP) on R [34,35].

2.7. Standard protocol Approvals, registrations, and patient consent

This study was conducted in accordance with the institutional guidelines for clinical studies of human specimens. The study protocol was approved by the Ethics Committees of Kyushu University (21046-00, 22258-00) and Fukuoka Children's Hospital (2022-131). Informed consent was obtained from all patients or their guardians.

3. Results

3.1. Patients demographics and clinical parameters

This study included 13 patients who tested positive for MOG-IgG. Table 1 presents their demographic features. The median age was 7.7 years (interquartile range [IQR] 6.5–11.1), and seven males (54 %) were enrolled. The clinical phenotypes of MOGAD included ADEM (n = 6, 43 %), optic neuritis (ON: n = 3, 23 %), and encephalitis (En: n = 4, 31 %). The most frequent neurological symptoms in the acute stage included impaired consciousness (62 %), cranial nerve dysfunction (38 %), limb weakness (23 %), and autonomic dysfunction (31 %). CSF data showed a median white blood cell count of 31/μL (IQR 10–74), and three patients (23 %) had positive oligoclonal bands. The most frequent MRI lesions were located in the cerebral white matter in 10 (77 %) patients, followed by the optic nerve in 7 (54 %), and basal ganglia in 6 (46 %). The median duration of follow-up was 3 years (range 2.0–8.2). Three patients (23 %) showed a relapsing course and 10 (77 %) showed a good prognosis without neurological sequelae at the final observation (Supplementary Table 1). These observations are consistent with previous studies [4,11].

Table 1.

Demographics, clinical features and treatments of 13 patients with MOGAD.

Age at onset, median and IQR years	7.7, 6.5–11.1
Female sex	6 (46 %)
Preceding infection	10 (77 %)
Clinical phenotypes
Acute demyelinating encephalomyelitis	6 (46 %)
Optic neuritis	3 (23 %)
Encephalitis	4 (31 %)
Clinical and laboratory findings at onset
Impaired consciousness	8 (62 %)
Neck, face, extraocular, or bulbar weakness	5 (38 %)
Limb weakness	3 (23 %)
Non-specific sensory symptoms	2 (15 %)
Autonomic dysfunction	4 (31 %)
Seizure	2 (15 %)
Cerebrospinal fluid results
White blood cells, median and IQR K/μL	31, 10–74
Protein concentration, median and IQR mg/dL	39, 29–62
Myelin basic protein, median and IQR pg/mL	107, 52–196
IgG index, median and IQR	0.59, 0.42–0.67
Positive oligoclonal band	3 (23 %)
Lesions on magnetic resonance imaging
Cortical gray matter	4 (31 %)
Subcortical white matter	10 (77 %)
Basal ganglia	6 (46 %)
Optic nerve	7 (54 %)
Brainstem	4 (31 %)
Cerebellum	2 (15 %)
Spinal cord	2 (15 %)
Treatmentacourse
Intravenous methylprednisolone	13 (100 %)
Intravenous immunoglobulin	5 (38 %)
Plasmapheresis	3 (23 %)
Number of demyelinating events, median and range	1, 1–9
Duration of follow-up, median and IQR years	2.7, 1.4–4.3
EDSS score at the last visit, median and range	0, 0–1.5
Neurological sequalae (%)	3, 23

EDSS, Expanded Disability Status Scale; IQR, interquartile range; MOGAD, myelin oligodendrocyte glycoprotein-associated disease.

^aEach patient received one of the treatments during the observation period.

3.2. Cytokine profiles

CSF cytokine levels were compared between patients with MOGAD and controls (Fig. 1A). A significant increase in IL-10 levels was observed in the CSF samples of patients with MOGAD relative to those of the controls (median 681 fg/mL [IQR 135–1510] vs. 34 fg/mL [10−298], respectively; p = 0.026). No significant differences were noted in the other cytokines analyzed. In the comparative analyses of cytokines among the ADEM, En, and ON subgroups, IL-6 was significantly higher in the sera of patients with ADEM (p = 0.036) than in those with ON. However, no such difference was observed for cytokines in the CSF of these patients (Supplementary Fig. 3).

Fig. 1.

Cytokine profiles in CSF and serum samples.

(A) Cytokine profiles in CSF from patients with MOGAD (n = 13) and DC (n = 5). ∗p < 0.05, ∗∗<0.01 (Mann-Whitney U test).

(B) A correlation heatmap for the indicated cytokines in patients with MOGAD and DC. Color codes indicate the correlation coefficient (Pearson correlation test, n = 18 each). Note that the CSF (c) and serum (s) samples form clusters in the correlation matrix. Serum IFN-γ (sIFNγ) is negatively correlated with most of the CSF cytokines.

(C) A correlation map of the MOG-IgG level (serum) with cytokine levels in CSF (upper) and serum (lower) in patients with MOGAD. Lines indicate significant correlation (p < 0.05, n = 11, Pearson correlation test) between the MOG-IgG level and the cytokines indicated. Values represent their respective correlation coefficients. CSF, cerebrospinal fluid; DC, disease control; IL, interleukin; MOG, myelin oligodendrocyte glycoprotein; MOGAD, MOG antibody-associated disease.

Multiple correlation analyses showed that cytokines were well intercorrelated with each other for both serum and CSF samples (Fig. 1B). Interestingly, serum IFN-γ levels were inversely correlated with most of the CSF cytokines. A further analysis revealed a significant positive correlation between serum MOG-IgG concentrations and serum levels of IL-10, IL-2, CSF levels of IL-1β, IL-2, IL-4, IL-10, IL-12p70, IL-17A, IFN-γ, and TNF-α, whereas serum MOG-IgG levels showed significant negative correlation with CSF levels of IFN-γ (Fig. 1C, Supplementary Fig. 4). Serum MOG-IgG levels observed the highest positive correlation with CSF levels of IL-10 (r = 0.64), IL-17A (r = 0.63), and IL-2 (r = 0.65).

3.3. Lipidomics-based profiling of CSF in MOGAD samples

We analyzed CSF samples collected from 5 children with MOGAD (Q1, Q2, F1, F4, and F6), and the results were compared to those of 5 children with epilepsy, representing non-inflammatory neurological disorders. A lipidomic analysis detected 7527 unique molecules, 38 of which were classified as high-confidence molecules based on their RTs (0.16–22.04 min) with liquid chromatography and m/z values (50.35–1698.98) in MS (Fig. 2A, Supplementary Fig. 1). Among them, 162 (0.02 %) lipophilic molecules were detected in patients with MOGAD at different levels (MOGAD/control ratio >1.5 or <0.7) from those in controls (Supplementary Fig. 5A). Among the 38 high-confidence molecules identified which consisted mostly of phosphatidylcholine (PC) species (Supplementary Fig. 6), we observed no significant difference when compared to the controls.

Fig. 2.

An overview of lipidomic (A–C) and metabolomic (D–F) profiling.

(A) A summary of MS/MS data from a lipidomic analysis of 5 patients with MOGAD (F1, F4, F6, Q1 and Q2) and 5 controls (C1–C5). X-axis shows the signal intensities converted to Z scores, and Y-axis the m/z values (detected range: 50.35–1698.98).

(B) A UMAP plot for a total of 7527 lipids identified in CSF samples (n = 5 [MOGAD] and 5 [DC]). Color dots represent 302 molecules with p < 0.05 (Student's t-test).

(C) A heatmap for hierarchical clustering of lipids with p < 0.05 (Student's t-test; MOGAD [n = 5] vs. DC [n = 5]). Dendrogram shows the hierarchical clustering of samples (row) and detected lipid molecules (column).

(D) A summary of MS/MS data from the metabolomic analysis (X-axis: signal intensity in Z-score; Y-axis: m/z 60.02–1198.30).

(E) UMAP for 17,526 hydrophilic molecules. 1031 molecules with p < 0.05 are shown as red dots.

(F) Heatmap for hydrophilic molecules with p < 0.05 (Student's t-test; MOGAD [n = 5] vs. DC [n = 5]). DC, disease control; MOGAD, MOG antibody-associated disease; MS/MS, mass spectrometry; UMAP, Uniform Manifold Approximation and Projection.

After scaling the relative signal intensity of each molecule using Z-scores, we applied the UMAP algorithm, which yielded two seemingly different clusters from all detected lipid molecules (Fig. 2B, Supplementary Fig. 7A). To further delineate the characteristic lipid profiles associated with MOGAD, we attempted hierarchical clustering of molecules with statistical significance. Indeed, we were able to identify subsets of lipid molecules with different profiles compared with the controls (Fig. 2C). Collectively, these results support the concept that pediatric patients with MOGAD had altered CSF lipid profiles, which might be involved in the development process or present as a consequence of demyelinating events.

3.4. Oxidized phosphatidylcholines

From the 12 species of PC identified in this study, an extensive analysis detected 5 oxidized forms of PCs (OxPCs). These molecules were designated PC(32:1)(=O), PC(38:4)(-OH), PC(38:7)(-OH), PC(38:7)(-OOH), and PC(40:6)(-OH) (Supplementary Fig. 8A). We excluded the sample from Patient Q2 because concentrations of multiple CSF OxPC species were outliers relative to the average values obtained from the four patients with MOGAD, in order to mitigate the risk of a false positive result in this underpowered study. Despite this manipulation, we did not find significant differences in the concentrations of OxPCs between MOGAD and disease controls (Supplementary Fig. 8B).

3.5. Metabolomics-based profiling of CSF in MOGAD samples

A metabolomic analysis detected a total of 17,526 (m/z: 60.02–1198.30; RTs: 0.19–35.95 min) hydrophilic molecules (Fig. 2D). Among these, we identified 549 (0.03 %) metabolites that were differentially expressed in the MOGAD and control groups (Supplementary Fig. 5B). Overall, the UMAP analysis did not segregate the 17,526 molecules into distinctive clusters, as observed in lipidomic profiling (Fig. 2E, Supplementary Fig. 7B). We confirmed that the 549 molecules with p values of <0.05 showed a distinct pattern of presence in CSF from those of controls (Fig. 2F).

Further evaluation revealed 128 high-confidence molecules belonging to various categories (Supplementary Fig. 6). Levels of pyridoxine (MOGAD/control ratio, 0.42; p = 0.006), isethionate (0.66; p = 0.015), and ribitol (0.60; p = 0.024) were significantly decreased, while N6-methyllysine (2.22; p = 0.002) was significantly increased in comparison to the controls (Fig. 3).

Fig. 3.

Novel biomarkers identified in the metabolomic analysis.

Among the 128 high-confidence molecules detected, N6-methyllysine showed a significantly higher level (MOGAD/control ratio = 2.22), whereas levels of pyridoxine (0.42), isethionate (0.66), and ribitol (0.60) were significantly decreased. ∗p < 0.05, ∗∗<0.01 (Student's t-test; n = 5 [MOGAD] and 5 [DC]). DC, disease control; MOGAD, MOG antibody-associated disease.

3.6. Correlation between cytokines and lipidomic analyses

To further explore the dynamics of molecular interactions among the biomarkers detected in MOGAD, we examined the possible correlation between cytokines and lipidomic profiles. Although serum MOG levels were significantly correlated with a few CSF cytokines, a UMAP analysis of the CSF cytokine profiles failed to differentiate patients with MOGAD from those with disease controls (Fig. 4A). In contrast, CSF samples from patients with MOGAD and controls were separated on UMAP when 162 lipophilic and 549 hydrophilic molecules (p < 0.05) were applied for their molecular profiles (Fig. 4B). These results further suggest altered lipid and metabolomic profiles in pediatric patients with MOGAD. To identify possible diagnostic markers of MOGAD, we drew heatmaps for the top 20 upregulated and downregulated lipids associated with IL-10 and IL-17A levels in the CSF (Supplementary Fig. 9). These two cytokines were selected because they were the most relevant to the serum levels of MOG-IgG in this study. Contrary to our prediction, however, none of the IL-10- and IL-17A-associated molecules appeared to indicate the diagnosis of MOGAD.

Fig. 4.

Diagnostic utility of cytokine, lipidomic, and metabolomic profiles.

(A) A UMAP algorithm was applied to the cytokine profiles of CSF samples from patients with MOGAD (n = 13) and DC (n = 5). Note that patients with MOGAD and DC can hardly be discriminated from each other based on their distribution patterns.

(B) UMAP plots for patients with MOGAD and DC using the lipidomic (left) and metabolomic (right) profiling data. Quantitated values of selected molecules with p values of <0.05 were applied in the UMAP algorithm. DC, disease control; MOGAD, MOG antibody-associated disease; UMAP, Uniform Manifold Approximation and Projection.

4. Discussion

In this study, we present data from a comprehensive lipidomic and metabolomic analyses combined with cytokine measurements, aiming to achieve a multidimensional understanding of the pathogenic mechanisms involved in children with MOGAD. Cytokine profiling showed that IL-10 levels increased significantly in the CSF samples of MOGAD, and identified CSF IL-17A and IL-10 as cytokines with the highest correlation coefficient with serum MOG-IgG levels. In the lipidomic analysis, we were unable to identify discriminant lipids distinguishing patients with MOGAD from controls among the known 162 molecules. However, this study revealed significantly different subsets of lipids that were identified as targets for future investigation. Taken together, the preliminary evidence indicates that MOG-IgG disease displays distinct CSF lipid profiles. Lipidomic studies on acquired demyelinating syndromes are limited in number and have almost exclusively focused on multiple sclerosis [19,36]. Thus, this preliminary study targeting MOGAD is unique in identifying potential biomarkers, as well as in exploring the underlying pathological mechanisms associated with pediatric MOG-IgG.

The lipidomic analysis of neurodegenerative and neuroinflammatory disorders is an area of dynamic and ongoing research because lipids play a crucial role in the structure and function of myelin in the CNS [37,38]. MOGAD is considered an autoimmune entity in which the immune system targets MOG, a protein located on the surface of myelin. Disruption of lipid metabolism can affect the integrity of myelin, which is primarily composed of lipids, including phosphatidylcholine, sphingomyelin, phosphatidylethanolamine, ceramides, and sulfatides [39]. Thus, lipidomic analyses are ideal for investigating this condition. The degraded lipid products might function as DAMPs [23], initiating cascades in the intrinsic immune response resulting in subsequent demyelination, which might serve as promising biomarkers [36]. In addition, altered lipid profiles can influence the T-cell function and cytokine expression by interacting with key signaling pathways [40]. While most studies assessed the lipidome profile using serum samples, this study evaluated the CSF sample because of its close proximity to the CNS and the probability of tracking the myelin degradation products as a result of demyelination [19]. However, the analysis of CSF is a challenging task because of its low total lipid content, which is approximately 1/300 of the plasma level [20,41]. High-sensitivity MS-based approaches have been developed, and only a limited number of lipidomic studies on CSF have been reported, mainly in the field of dementia and multiple sclerosis [19,20,42].

In the present study, the most abundant lipid molecules detected were PCs, a heterogeneous group of glycerophospholipids [41]. However, no significant difference was observed among the PCs detected in patients with MOGAD relative to those in controls. An untargeted lipidomic analysis conducted on the CSF of patients with multiple sclerosis (age 16–61 years) showed marked reductions in PC(28:0), PC(28:1), PC(35:4), PC(36:1), PC(36:8), PC(37:6), and a few other lipid molecules [43]. Other studies have reported alterations in CSF levels of sphingomyelin, triglycerides, fatty acids, and ceramides [19,43,44]. These results indicate that adult patients with multiple sclerosis show different lipidomic profiles from those of children with acquired demyelinating syndromes, including MOGAD. Alternatively, the lipid metabolism involved in the pathophysiology of MOGAD might be different from that of MS.

We extended our exploration to include the metabolic profiles of the hydrophilic molecules in the CSF. The levels of pyridoxine, isethionate, and ribitol decreased, and that of N6-methyllysine increased, showing statistical significance. Isethionate is related to taurine metabolism [45], and pyridoxine (vitamin B6) acts as a cofactor in cysteine metabolism, a precursor of taurine [46]. Recent studies have highlighted that pyridoxine regulates immune signaling by preventing the overactivation of NLRP3 inflammasome and NF-κB pathways, leading to the reduction of pro-inflammatory cytokine levels [47]. Ribitol, in turn, is involved in the metabolism of riboflavin (vitamin B2) [48]. Thus, the metabolomic profiles in the CSF suggest altered vitamin metabolism in MOGAD. The sample storage temperature and duration before the analysis might affect the detection of these metabolites. A future validation study using orthogonal methods, such as targeted MS, is required to verify this novel finding. This could reveal potential targets for interventions in the management of MOGAD.

In agreement with a previous report, we detected a significant increase in IL-10 levels, an important mediator of regulatory T cells [49]. However, we did not observe an increase in Th17 related cytokines (IL-6 and IL-17A), probably due to the small number of samples. A correlation analysis confirmed a generally good intercorrelation among cytokines. Intriguingly, CSF cytokines showed higher correlation coefficients with serum MOG-IgG levels than those in the serum cytokines. For example, IL-10, IL-17A, and IL-2 in with CSF exhibited the highest correlation coefficients (0.63–0.65) among those tested. IL-17A plays a role in blood-brain barrier breakdown through the recruitment of inflammatory cells to damaged regions in multiple sclerosis [50]. Our data support the concept that the activation of innate immunity leads to an elevated IL-17A signal, which provokes further activation of the acquired immune system in patients with MOGAD [51]. The inverse correlation observed between serum IFN-γ and multiple CSF cytokines suggests an involvement of specific immune signaling, potentially mediated by the blood-brain barrier and the immunomodulatory effects of IFN-γ [52]. A biostatistical method using UMAP showed that lipidomics and metabolomics were more efficient in differentiating MOGAD from disease control than cytokine profiling. Thus, the accumulation of lipidomic and cytokine profiling data may provide useful clues for the discovery of potential biomarkers. From this perspective, future studies are warranted to determine the specific metabolites from the list of the most upregulated and downregulated molecules in CSF of patients with MOGAD.

5. Limitations

The present study was associated with several limitations. Initially, we conducted biochemical analyses for a limited cohort of patients with MOGAD (n = 5) and controls (n = 5). Nonetheless, using a comprehensive unsupervised approach, we identified 162 lipophilic and 549 hydrophilic molecules associated with pediatric MOGAD in this initial exploratory study. Validation studies with a larger sample size are therefore essential to confirm these preliminary findings. Longitudinal assessment of molecular perturbation in one individual will provide further insight into the correlation of the concentrations in their CSF with disease activity. Second, we were unable to identify specific OxPCs that may serve as molecular markers for the active condition of MOGAD [21]. More advanced methods, such as multiple reaction monitoring [9] will help determine the chemical moieties of OxPCs associated with MOGAD. Future studies employing complementary high-resolution mass spectrometry for a comprehensive lipidomic analysis may enable the better identification of unknown compounds [33]. Third, the use of samples from patients with epilepsy serving as ‘non-inflammatory’ controls even after careful exclusion of immune, metabolic, and infectious etiologies may not be optimal. Recent research suggests that neuroinflammatory mechanisms contribute to the onset of epilepsy by promoting the neuronal excitability and disrupting the blood-brain barrier [53]. Thus, future studies would benefit significantly from the inclusion of healthy pediatric controls. Lastly, the cytokine analysis was restricted to only 9 proteins in this study. These results might offer a biased interpretation of the neuroimmunological pathways involved in the pathogenesis of MOGAD. Extensive measurement of other molecules, such as chemokines and complement factors, may contribute to the understanding of cytokine-lipidomic-metabolomic interactions. Such efforts will provide molecular insights into the crosslinks between neuroinflammation and demyelination in MOGAD [54].

6. Conclusion

Our study identified distinct cytokine, lipidomic and metabolomic profiles in CSF of patients with childhood-onset MOGAD. Although high-confidence lipid molecules were not detected, a metabolomic analysis revealed an altered vitamin metabolism. Both lipidomic and metabolomic profiling successfully differentiated the molecular signature of CSF between patients with MOGAD and disease controls. Our preliminary findings suggest that lipidomic and metabolomic profiling is a promising method for the molecular diagnosis of childhood-onset MOGAD. Further research is required to unveil the unique pathogenesis of acquired demyelinating syndromes in childhood.

CRediT authorship contribution statement

Pin Fee Chong: Conceptualization, Formal analysis, Funding acquisition, Writing – original draft. Kenta Kajiwara: Data curation, Investigation, Writing – review & editing. Yuki Ueno: Data curation, Investigation, Writing – review & editing. Satoshi Akamine: Formal analysis, Methodology, Writing – review & editing. Hiroyuki Torisu: Funding acquisition, Supervision, Writing – review & editing. Ryutaro Kira: Funding acquisition, Supervision, Writing – review & editing. Shouichi Ohga: Methodology, Project administration, Writing – review & editing. Yasunari Sakai: Conceptualization, Funding acquisition, Methodology, Project administration, Supervision, Writing – original draft.

Funding sources

This study was supported in part by JSPS Kakenhi (grant numbers JP19K10613 and JP25K11060 to PFC); research grants from the Ministry of Health, Labour, and Welfare of Japan (grant numbers JP22HA1003 and JP25HA1002 to PFC, RK, HT, and JP20FC1054 to YS), the Morinaga Foundation for Health & Nutrition to PFC, AMED (grant numbers JP20ek109411 and JP20wm0325002h); The Japan Epilepsy Research Foundation; and Kawano Masanori Memorial Public Interest Incorporated Foundation for the Promotion of Pediatrics (YS).

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have influenced the work reported in this study.

Appendix A. Supplementary data

The following is the Supplementary data to this article:

Data availability

Data will be made available on request.