Abstract
Background
Granulocyte invasion into the brain is a pathoanatomical feature differentiating neuromyelitis optica spectrum disorder (NMOSD) from multiple sclerosis (MS). We aimed to determine whether granulocyte activation markers (GAM) in cerebrospinal fluid (CSF) can be used as a biomarker to distinguish NMOSD from MS, and whether levels associate with neurological impairment.
Methods
We quantified CSF levels of five GAM (neutrophil elastase, myeloperoxidase, neutrophil gelatinase-associated lipocalin, matrixmetalloproteinase-8, tissue inhibitor of metalloproteinase-1), as well as a set of inflammatory and tissue-destruction markers, known to be upregulated in NMOSD and MS (neurofilament light chain, glial fibrillary acidic protein, S100B, matrix metalloproteinase-9, intercellular adhesion molecule-1, vascular cellular adhesion molecule-1), in two cohorts of patients with mixed NMOSD and relapsing-remitting multiple sclerosis (RRMS).
Results
In acute NMOSD, GAM and adhesion molecules, but not the other markers, were higher than in RRMS and correlated with actual clinical disability scores. Peak GAM levels occurred at the onset of NMOSD attacks, while they were stably low in MS, allowing to differentiate the two diseases for ≤21 days from onset of clinical exacerbation. Composites of GAM provided area under the curve values of 0.90–0.98 (specificity of 0.76–1.0, sensitivity of 0.87–1.0) to differentiate NMOSD from MS, including all anti-aquaporin-4 protein (aAQP4)-antibody-negative patients who were untreated.
Conclusions
GAM composites represent a novel biomarker to reliably differentiate NMOSD from MS, including in aAQP4− NMOSD. The association of GAM with the degree of concurrent neurological impairment provides evidence for their pathogenic role, in turn suggesting them as potential drug targets in acute NMOSD.
Keywords: multiple sclerosis, neuroimmunology, molecular biology, CSF, clinical neurology
WHAT IS ALREADY KNOWN ABOUT THIS TOPIC
Neuromyelitis optica spectrum disorder (NMOSD) is difficult to differentiate from multiple sclerosis (MS) based on clinical and MRI features, and 20%–40% of patients score negative for the gold-standard diagnostic biomarker, anti-aquaporin-4 protein-antibodies (aAQP4).
Furthermore, this biomarker does not associate with specific clinical disease features.
Granulocyte invasion into brain lesions is a key feature of NMOSD, however the potential of granulocyte activation markers (GAM) to differentiate NMOSD from MS has not been explored.
WHAT THIS STUDY ADDS
Increased cerebrospinal fluid levels of GAM differentiate NMOSD from MS as reliably as aAQP4 in acute stages of disease, including also patients with aAQP4− NMOSD.
Furthermore, levels of GAM, but not of other biomarkers upregulated in NMOSD, correlate with the actual degree of neurological impairment in acute NMOSD.
HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY
GAM are novel biomarkers for NMOSD that may close a diagnostic gap at the time of first clinical exacerbation, when timely initiation of effective therapy is crucial for therapeutic success.
In addition, levels of GAM also associate with clinical disability scores at first presentation.
Together with existing preclinical evidence, the current observations suggest a pathogenic role of GAM in NMOSD and may facilitate the development of novel therapeutic approaches for the treatment of acute-stage NMOSD.
Introduction
Neuromyelitis optica spectrum disorder (NMOSD) and relapsing-remitting multiple sclerosis (RRMS) share clinical and imaging characteristics, which can make it difficult to differentiate them, and hence may delay the initiation and the choice of adequate therapy.1 The detection of auto-antibodies targeting the astrocyte water channel anti-aquaporin-4 protein (aAQP4) has become a pivotal biomarker tool to diagnose NMOSD. However, 20%–40% of patients eventually fulfilling NMOSD criteria remain aAQP4− which makes it even more difficult to establish the accurate diagnosis.2–5 Furthermore, the presence or the titre of aAQP4 is not related to clinical disease characteristics.4 Several biomarkers tend to be more highly elevated in NMOSD than in MS, for example, glial fibrillary acidic protein (GFAP), and S100B (both markers of astrocyte damage),6 7 neurofilament light chain (NfL, a marker of neuroaxonal injury),6 8 chemokine (C-X-C motif) ligand 13 (CXCL13, a B-cell attractant),9–11 intercellular adhesion molecule-1 (ICAM-1) and vascular cellular adhesion molecule-1 (VCAM-1) (both leucocyte adhesion molecules)12 and matrix metalloproteinase-9 (MMP-9, a matrix remodelling gelatinase).13 Yet, all lack the necessary diagnostic specificity due to overlapping concentration ranges in the two diseases, and the relation between levels in cerebrospinal fluid (CSF) or blood with clinical severity remains uncertain.6
In NMOSD, activation of neutrophil granulocytes occurs in blood circulation, and their invasion into inflamed neural tissue is observed in 95% of NMOSD brain tissue specimens, a feature that differentiates it categorically from typical MS lesions; the involvement of granulocyte invasion in lesion formation has recently been demonstrated also in myelin oligodendrocyte glycoprotein antibody-associated disease (MOGAD).14–20 In the course of acute inflammation, granulocytes release a wide range of proteases and other proteins from their granular compartments, granulocyte activation markers (GAM), some of which are cell-specific such as neutrophil elastase (nEla), myeloperoxidase (MPO), matrix metalloproteinase-8 (MMP-8) and neutrophil gelatinase-associated lipocalin (NGAL), or are partially cell-specific, like tissue inhibitor of metalloproteinase-1 (TIMP-1) and MMP-9. We hypothesised that in NMOSD, including in aAQP4− cases, GAM produce a humoral footprint in CSF that allows to differentiate NMOSD from RRMS, and that their levels correlate with clinical severity at the time of CSF sampling to give support for their pathogenetic role in NMOSD, as suggested in preclinical models and neuropathological findings.16 In this case-control study, we quantified levels of GAM in acute and subacute/chronic (s/c) NMOSD and RRMS, along with MMP-9, NfL, GFAP, S100B, ICAM-1, VCAM-1, CXCL13 to define a precision medicine tool for the diagnosis of acute-stage NMOSD.
Methods
Participants and samples
The diagnosis of NMOSD with/without aAQP4, and of RRMS was based on respective standard diagnostic criteria.21 22 Acute disease exacerbation/relapse was defined according to 2017 McDonald criteria.22 Disability was assessed using the Expanded Disability Status Scale (EDSS).23 The discovery cohort from Kyushu University Hospital (Fukuoka, Japan) consisted of 34 patients with NMOSD with 42 CSF samples (2 patients contributed 3 CSF samples, 4 patients contributed 2 CSF samples), and 36 patients with RRMS with 40 CSF samples (4 patients contributed 2 CSF samples); these repeated lumbar punctures were performed following independent disease exacerbations. The validation cohort consisted of 25 patients with NMOSD from Kyushu University Hospital (n=11), Ospedale San Raffaele and Mondino Foundation (Milan and Pavia, Italy) (n=8) and Karolinska University Hospital (Stockholm, Sweden) (n=6) and 46 patients with MS (Kyushu n=18, Karolinska n=28). Two control groups (University Hospital Basel, Switzerland) were used to determine values under physiological and highly inflammatory conditions, respectively: ‘symptomatic controls’ (SC)24 consisted of 25 patients in whom a structural neurological disease was excluded, based on normal findings in clinical and MRI evaluations, normal CSF cell composition and protein content and absence of signs of intrathecal immunoglobulin synthesis. The second control group comprised 15 patients with various types of acute inflammatory neurological disease controls24 (inflammatory neurological disease controls: meningoencephalitis/polyradiculitis due to (a) varicella zoster virus (n=5) and (b) tuberculosis (n=1), (c) viral meningitis (n=3), (d) neuroborreliosis (n=3), (e) eosinophilic encephalitis (n=1), (f) autoimmune encephalitis/myelomeningoradiculitis of unknown cause (n=2)).
Measurement of biomarkers
Standard CSF analyses were performed at each centre independently, while here investigated biomarkers were analysed centrally. Expression levels of 12 markers were determined by ELISA and single molecule array assay in the discovery cohort (table 1). CSF samples with >1 erythrocyte/µL were excluded from the analysis. Sample identities in both cohorts were blinded until all analyses had been completed. All samples and calibrators were assayed in duplicate. An explorative analysis of biomarker levels categorised for sites of origin of samples showed comparable values of GAM (not shown).
Table 1.
Analytical panel of biomarkers
| Cellular source in CSF | Function/Biomarker significance | Analytical platform | LOD (pg/mL) | |||
| Neutrophil elastase (EC 3.4.21.37)* |
Granulocytes | Primary granule | Proteolytic enzyme | EnzChek Elastase Assay Kit (ThermoFisher) |
18.2 | |
| Myeloperoxidase (EC 1.11.2.2)* |
Hypochlorous acid (HOCl) production |
SP-X platform (Simoa) (Quanterix) |
0.01 | |||
| Matrix metalloproteinsae-8 (MMP-8, neutrophil collagenase, EC 3.4.24.34*) |
Secondary granule | MMPs: extracellular matrix modulation NGAL: bacteriostatic by Fe chelation |
0.65 | |||
| Neutrophil gelatinase-associated lipocalin† | 0.33 | |||||
| Matrix metalloproteinsae-9 (MMP-9, gelatinase B, EC 3.4.24.35)* | Tertiary granule | ubiquitous | Quantikine ELISA (R&D Systems) |
50.6 | ||
| Tissue inhibitor of metalloproteinase-1 (TIMP-1)‡ |
Secretory vesicles | MMP activation and inhibition | SP-X platform (Simoa) (Quanterix) |
0.74 | ||
| S100 calcium-binding protein B | In CNS: astrocytes | Cytoplasmatic calcium-binding protein, cell damage marker | ELISA (BioVendor) |
3.77 | ||
| Glial fibrillary acidic protein | Cytoskeleton intermediate filament, cell damage marker | HD-X platform (Simoa) (Quanterix) |
0.21 | |||
| Neurofilament light chain | Neurons | 0.04 | ||||
| Vascular cell adhesion protein-1 (CD106) | Endothelial cells | Adhesion molecules | SP-X platform (Simoa) (Quanterix) |
1.12 | ||
| Intercellular adhesion molecule-The supplemental materiaö1 (CD54) | 0.42 | |||||
| C-X-C motif chemokine 13 | In CNS: macrophage-like cells | B-cell chemoattractant | 0.05 | |||
Values below LOD were imputed as a random value between 0.001 and LOD, drawn from a uniform distribution.
*Nomenclature according to https://enzyme.expasy.org.
†Forms complexes with MMP-9; the only other cellular source is from renal tissue.
‡Other than the name indicates, in granulocytes TIMP-1 is involved in MMP activation.43
CNS, central nervous system; CSF, cerebrospinal fluid; LOD, limit of detection; Simoa, single molecular array.
Statistical analysis
CSF levels of biomarkers are presented as median and IQRs by diagnostic groups and were compared using the Wilcoxon rank-sum test. To determine the capacity of distinguishing between NMOSD and RRMS without the potential confounding effect of corticosteroid pretreatment, we repeated the same analyses in treatment-naïve patients. To investigate the temporal dynamics of biomarker concentrations, we used for each biomarker an individual linear model to describe the levels in NMOSD and RRMS within a 60-day period after acute disease exacerbation; this period was defined in days between the onset of acute disease exacerbation and lumbar puncture. Biomarker levels were log-transformed and served as dependent variable. Diagnosis (RRMS vs NMOSD) and time since disease exacerbation, as well as the interaction between these two variables, were used as independent variables. The interaction indicates whether the temporal dynamics differ between patients with NMOSD and RRMS. Again, sensitivity analyses were run after exclusion of pretreated patients. In accordance with the exploratory nature of these analyses, no correction for multiple testing was performed. Accordingly, p values should not be interpreted as confirmatory but rather as a continuous measure of evidence against the corresponding null-hypothesis.
The correlation between biomarker levels and EDSS score was quantified using Spearman’s rank correlation coefficient. The diagnostic capacity of GAM to differentiate NMOSD from RRMS in acute stages (≤21 days after onset of exacerbations) was determined by a logistic model where the disease type (NMOSD vs RRMS) served as dependent variable, and biomarkers, or composites of biomarkers, as independent variables. The predictions from these models were assessed with receiver operating characteristic (ROC) curves, based on pooled data of discovery and validation cohorts. We performed this analysis with and without time of sampling since disease exacerbation as a covariate, in acute patients and in those without corticosteroid pretreatment. For each model, the area under the curve (AUC), as well as sensitivity, specificity, positive and negative predictive value, based on the optimal cut-off according to the Youden Index, are presented. To test the robustness of the model in terms of replicability and to address the risk of overfitting in function of the numbers of markers and time as a covariate in composite models, we validated them by calculating optimism-corrected AUCs based on 500 bootstrap replicates. All analyses were carried out using the statistical software R (V.4.1.2, The R Foundation for Statistical Computing). The significance level was set at p=0.05.
Results
Demographics of discovery and validation cohort of patients
Table 2 shows the baseline characteristics of patients with NMOSD and RRMS in the discovery and validation cohort; they were stratified into ‘acute’ (≤21 days) and s/c (>21 days) stages, depending on the time between onset of acute clinical symptoms and lumbar puncture. All patients with NMOSD were aAQP4+, except three in the discovery and four in the validation cohort (of those, one became positive during a later attack); of these seven patients, six scored negative for anti-MOG antibodies (one patient was not tested). Patients with NMOSD in both cohorts were older and had higher EDSS scores than patients with RRMS. The majority of CSF samples (76.2% in the discovery and 48.0% in the validation cohort) were from patients with NMOSD on continuous oral, or who had received intravenous corticosteroid therapy before the time of CSF sampling, while for patients with RRMS the corresponding proportions were much smaller (37.5% and 8.7%, respectively).
Table 2.
Demographic and clinical characteristics of patient groups and control persons
| Discovery cohort | Validation cohort | Controls | ||||||||
| NMOSD | RRMS | P value | NMOSD | RRMS | P value | INDC | SC | |||
| Patients male (n, %) | 34 5 (14.7) |
36 8 (22.2) |
0.617 | 25 2 (8.0) |
46 14 (30.4) |
0.062 | 15 11 (73) |
25 10 (40) |
||
| Samples (n) acute: s/c | 42* 20:22 |
40* 17:23 |
25 21:4 |
46 34:12 |
15 – |
25 – |
||||
| Age at first CSF sample (years) mean (SD) | 44.8 (11.7) |
37.4 (12.7) |
0.015 | 47.5 (18.0) |
36.2 (11.4) |
0.004 | 54.7 (19.9) |
45.9 (9) |
||
| Disease duration at sampling median (IQR) | 6.7 (2.0, 12.4) |
6.0 (2.0, 11.0) |
0.784 | 3.6 (0.4, 8.5) |
1.9 (0.1, 9.2) |
0.338 | n/a | n/a | ||
| Interval between clinical attack and lumbar puncture† (days) median (IQR) | 30.0 (8.0, 83.0) |
40.5 (9.8, 93.8) |
0.511 | 9.5 (7.0, 13.8) |
13.5 (8.0, 20.8) |
0.057 | n/a | n/a | ||
| EDSS score median (IQR) | 5.5 (3.0, 7.0) |
3.5 (2.5, 6.0) |
0.015 | 4.5 (3.5, 6.5) |
3.0 (2.0, 3.5) |
<0.001 | n/a | n/a | ||
| aAQP4–, n (%) |
3‡ (8.8) |
n/a | n/a | 4 (16.0) |
n/a | n/a | n/a | n/a | ||
| CSF samples with granulocytosis, (n/N, %) CSF cell count (×/µL) median (IQR) |
9/42 (21.4) 3.0 (1.0, 5.8) |
2/40 (5.0) 3.0 (1.0, 5.0) |
0.063 0.985 |
5/14§ (35.7) 2.5 (1.0, 11.8) |
6/18 (33.3) 4.0 (1.5, 11.0) |
1.0 0.561 |
8 (53.3) 37.7 (16.6, 102.3) |
0 1.0 (0.6, 2.0) |
||
| Samples of patients with treatment prior lumbar puncture (n/N, %) | Corticosteroid | Pulse intravenous¶ | 10 (23.8) |
6 (15.0) |
0.388 | 1 (4) |
0 (0) |
n/a | n/a | n/a |
| Oral | 29 (69.0) |
12 (33.3) |
0.001 | 11 (44.0) |
4 (8.7) |
<0.001 | ||||
| Immunomodulatory | 10** (23.8) |
1 (4.0) |
0.012 | 4** (16.0) |
11 (23.9) |
0.634 | n/a | n/a | ||
*Six patients with NMOSD and four patients with RRMS contributed >1 CSF sample from independent disease exacerbations (see ‘Methods’ section).
†The time between disease exacerbation and lumbar puncture was significantly longer in the discovery versus validation cohorts (NMOSD+RRMS) (days, median (IQR)): 36.0 (10.0, 83.0) vs 11.5 (7.8, 18.2), p<0.001.
‡One patient had a lumbar puncture after start of corticosteroid therapy and scored aAQP4– by cell-based assay but became positive when tested during a later attack. Six of seven aAQP4– patients scored also negative for anti-MOG antibodies, one patient was not tested.
§In the validation set the CSF granulocyte cell count was not reported for 11 NMOSD and 28 RRMS samples.
¶The time range of administration prior lumbar puncture was 1–40 days; 7 patients with NMOSD in the discovery set had received intravenous corticosteroid and oral therapy; 1 and 2 patients in the discovery and validation set, respectively, had received intravenous corticosteroids 9–12 weeks before lumbar puncture and were not used for statistical calculations on corticosteroid effects. Three patients with RRMS in the discovery set had received intravenous corticosteroid and oral therapy.
**In total, 14 patients with NMOSD were on immunomodulatory therapy (azathioprine: 7, tacrolimus: 4; methotrexate: 2; cyclophosphamide: 1) at the time point of lumbar puncture. In addition, 2 patients in the discovery set (patient 1: 45–47 and 1–5 days before lumbar puncture; patient 2: 23–27 days before lumbar puncture) and one in the validation cohort (patient 3: 8 days before lumbar puncture) had received intravenous immunoglobulins.
aAQP4, anti-aquaporin-4 antibody; EDSS, Expanded Disability Status Scale; INDC, inflammatory neurological disease control; n/a, not available; NMOSD, neuromyelitis optica spectrum disorder; RRMS, relapsing-remitting multiple sclerosis; SC, symptomatic control.
Biomarker expression profiles in NMOSD and RRMS in discovery and validation cohorts
In the discovery cohort, GAM levels were higher in (a) NMOSD versus RRMS overall, (b) acute versus s/c NMOSD and (c) acute NMOSD versus acute RRMS. The astrocyte markers GFAP and S100B and adhesion molecules VCAM-1 and ICAM-1 were increased in NMOSD versus RRMS (overall and acute), while in acute versus s/c NMOSD this was only the case for S100B. In contrast, levels of NfL, MMP-9 and CXCL13 were not different for these three comparisons (table 3, online supplemental figure 1A).
Table 3.
Levels of biomarkers, all patients
| Cohort | Discovery | Validation | ||||||||||||||
| Disease group (n) |
NMOSD All (42)* |
RRMS All (40)* |
P value | NMOSD Acute (20) |
NMOSD s/c (22) | P value | RRMS Acute (17) |
P value† | NMOSD All (25) |
RRMS All (46) |
P value | NMOSD Acute (21) |
NMOSD s/c (4) | P value | RRMS Acute (34) |
P value† |
| nEla | 42.0 (15.6, 162.6) |
15.3 (6.5, 43.4) |
0.001 | 114.8 (44.9, 670.8) |
17.1 (10.2, 42.3) |
<0.001 | 15.2 (8.3, 38.9) |
<0.001 | 220.7 (83.7, 435.2.9) |
26.1 (13.8, 139.3) |
0.003 | 220.7 (14.4, 435.2) |
253.3 (112.3, 476.4) |
0.858 | 17.2 (9.6, 100.2) |
0.008 |
| MPO | 3.8 (0.7, 10.7) |
1.2 (0.4, 3.0) |
0.013 | 7.9 (2.8, 73.7) |
1.2 (0.3, 5.2) |
0.003 | 0.6 (0.2, 3.4) |
0.001 | 253.7 (0.0, 706.8) |
15.1 (0.0, 161.2) |
0.016 | 280.9 (0.0, 731.7) |
0.0 (0.0, 63.4) |
0.068 | 33.5 (0.0, 165.8) |
0.005 |
| NGAL | 5528 (2846, 7877) |
3809 (2113, 5137) |
0.001 | 6411 (5173, 17 148) |
4012 (2292, 6302) |
0.010 | 2210 (2087, 4374) |
<0.001 | 3289 (1762, 4642) |
1526 (1130, 2349) |
<0.001 | 3786 (2054, 5263) |
1216 (849.4, 1594) |
0.004 | 1506 (1101, 2570) |
<0.001 |
| MMP-8 | 3.6 (2.5, 9.0) |
2.5 (0.7, 3.8) |
0.004 | 8.7 (3.3, 127.3) |
3.1 (1.9, 4.6) |
0.005 | 2.6 (1.8, 2.7) |
0.001 | 6.2 (2.7, 25.1) |
1.4 (1.1, 3.1) |
<0.001 | 6.2 (2.2, 25.1) |
6.1 (3.9, 16.4) |
0.803 | 1.4 (1.1, 3.1) |
<0.001 |
| MMP-9 | 136.1 (23.9, 1923) |
101.6 (25.4, 494.6) |
0.400 | 270.2 (37.3, 2195) |
36.9 (18.6, 1660) |
0.265 | 174.3 (26.5, 1031) |
0.577 | ||||||||
| TIMP-1 | 55.9 (44.1, 85.7) |
39.2 (35.6, 43.4) |
<0.001 | 62.7 (51.2, 155.8) |
51.6 (41.4, 61.6) |
0.032 | 38.0 (34.3, 39.6) |
<0.001 | 24.2 (18.3, 42.8) |
12.6 (10.7, 18.1) |
<0.001 | 28.8 (21.6, 47.7) |
13.4 (12.9, 14.9) |
0.011 | 13.3 (10.9, 18.3) |
<0.001 |
| GFAP | 14.6 (8.1, 39.3) |
7.5 (5.3, 9.3) |
<0.001 | 34.0 (10.7, 611.2) |
12.2 (7.4, 19.4) |
0.060 | 7485 (5541, 11 321) |
0.001 | ||||||||
| S100B | 219.6 (166.5, 311.9) |
175.8 (135.6, 202.8) |
0.009 | 249.2 (186.0, 496.6) |
199.7 (155.4, 246.6) |
0.030 | 168.3 (153.1, 188.3) |
0.005 | 291.8 (192.4, 448.6) |
184.8 (147.3, 242.3) |
0.003 | 295.3 (192.4, 448.6) |
199.1 (161.8, 528.3) |
0.592 | 184.8 (150.4, 260.1) |
0.006 |
| NfL | 1739 (1168, 5156) |
1980 (1249, 4391) |
0.981 | 1732 (1135, 3697) |
2653 (1254, 6129) |
0.371 | 2436 (1397, 8061) |
0.493 | ||||||||
| ICAM-1 | 3198 (1968, 4196) |
2008 (1513, 2474) |
<0.001 | 3698 (2005, 6225) |
2562 (1880, 3837) |
0.116 | 1956 (1720, 2560) |
0.002 | 1729 (1300, 2528) |
1286 (957.6, 1693) |
0.011 | 2301 (1489, 2612) |
1037 (982.4, 1146) |
0.020 | 1335 (988.7, 1693) |
0.003 |
| VCAM-1 | 252.2 (202.4, 362.4) |
209.3 (161.6, 238.2) |
0.001 | 266.9 (226.2, 464.2) |
230.0 (184.4, 289.5) |
0.087 | 219.8 (173.5, 245.2) |
0.010 | 160.4 (132.5, 267.5) |
107.1 (81.5, 172.9) |
0.008 | 167.5 (140.6, 281.1) |
97.4 (88.5 116.3) |
0.045 | 128.5 (91.0, 172.9) |
0.006 |
| CXCL13 | 7.4 (2.5, 41.2) |
4.8 (2.0, 21.3) |
0.228 | 11.4 (3.9, 47.9) |
3.4 (2.0, 29.2) |
0.149 | 6.7 (2.9, 21.5) |
0.326 | ||||||||
Values are medians (IQR) in pg/mL for nEla, MPO, NGAL, MMP-8, MMP-9, S100B, NfL and ICAM-1, and in pg/mL×103 for TIMP-1, GFAP and VCAM-1. P values ≤0.05 are with green background, p values 0.05–0.10 are with yellow background.
The analysis without CSF samples from repetitive relapses (NMOSD: n=34, RRMS: n=36) yielded the same statistical differences as shown in table 3, except for MPO (p=0.097) and S100B (p=0.083).
*Six patients with NMOSD and four patients with RRMS contributed >1 CSF sample from independent disease exacerbations (see ‘Methods’ section).
†P for comparison of acute NMOSD with acute RRMS.
CSF, cerebrospinal fluid; CXCL13, C-X-C motif chemokine 13; GFAP, glial fibrillar acidic protein; ICAM-1, intercellular adhesion molecule-1; MMP, matrix metalloproteinase; MPO, myeloperoxidase; N, number of sample; nEla, neutrophil elastase; NfL, neurofilament; NGAL, neutrophil gelatinase-associated lipocalin; NMOSD, neuromyelitis optica spectrum disorder; RRMS, relapsing-remitting multiple sclerosis; S100B, S100 calcium-binding protein B; s/c, subacute/chronic; TIMP-1, tissue inhibitor of metalloproteinase-1; VCAM-1, vascular cell adhesion molecule 1.
jnnp-2022-330796supp001.pdf (590.6KB, pdf)
In acute NMOSD, the median expression levels of granulocyte-specific GAM (nEla, MPO, NGAL) and of TIMP-1 were similar compared with INCDs; only MMP-8 was slightly higher in NMOSD versus INDC, while for MMP-9 levels were higher in INCD. All analysed markers were higher in acute NMOSD versus SC except for S100B (online supplemental table 1). In s/c NMOSD, all GAM markers, TIMP-1, ICAM-1 and VCAM-1 were lower compared with INCD (online supplemental figure 1A).
Because many markers analysed in the discovery cohort have not been evaluated in NMOSD, we decided to confirm these results in an independent validation cohort. The findings for GAM were fully confirmed in the validation cohort for the comparison of NMOSD versus RRMS (both ‘all’ and ‘acute’), and in part for the comparison of acute versus s/c NMOSD (only four s/c NMOSD samples available) (table 3). Subsequent analyses were therefore performed in the merged discovery/validation set.
Other than in NMOSD, there were no significant differences between acute and s/c levels of GAM and the other markers in RRMS, apart from NGAL and TIMP-1 being higher in s/c RRMS (p=0.013 and p=0.006, respectively), while all other markers were not different between these disease stages in the merged discovery/validation set.
Impact on biomarker levels by immunomodulatory and corticosteroid therapy prior lumbar puncture
GAM levels of patients under immunomodulatory plus corticosteroid therapy showed a strong overlap compared with those of patients being treated only with corticosteroids, in both acute and s/c phases, suggesting that these compounds used for prevention of further NMOSD relapses have no significant impact on granulocyte activation; in contrast, patients under corticosteroid therapy had lower GAM levels in s/c, and to a lesser extent in acute NMOSD, compared with patients without treatment (online supplemental figure 1B). However, GAM and adhesion molecule levels were only numerically higher without as compared with the combined groups with corticosteroid (overall and intravenous); only for nEla this was significant (online supplemental table 2).
After exclusion of corticosteroid treated patients, the higher levels of GAM and adhesion molecules in NMOSD versus RRMS (‘all’ and ‘acute’) as seen in overall patients (table 3) were confirmed, while those of MMP-9, NfL and CXCL13 were again not different; this was also the case for GFAP (not shown).
Association between biomarker levels disease severity/disability status, aAQP4 status and CSF granulocyte count
Figure 1 shows that CSF levels of GAM in all (with or without therapy) patients with NMOSD were correlated with EDSS scores (rho=0.31–0.46, all p≤0.01). In acute NMOSD, this was also the case for NGAL, MMP-8 and TIMP-1 (rho=0.39–0.50, p<0.001–0.011, but not for nEla and MPO, while in s/c NMOSD only nEla was correlated with the EDSS score (rho=0.41, p=0.036) (online supplemental table 3). GFAP levels, only analysed in the discovery set, did not correlate with the EDSS score in acute NMOSD and RRMS, while this was the case in s/c phase for NMOSD (rho=0.58 (0.21, 0.81), p=0.004), and a referring trend was found for RRMS (rho=0.40 (−0.01, 0.81), p=0.004). In the seven patients with aAQP4− NMOSD, GAM levels were similar to those of patients with aAQP4+ NMOSD (figure 1). Interestingly, there was not a general downregulation of GAM levels across all corticosteroid-treated patients, but instead a random distribution with many patients scoring 1–2 logs above average GAM levels (figure 1). In contrast, S100B, NfL, MMP-9 and CXCL13 were not associated with EDSS scores or had rho values ≤0.29 in NMOSD, with or without corticosteroid pretreatment; furthermore, in RRMS all these biomarkers showed only weak correlation (rho≤0.3) with EDSS scores (not shown).
Figure 1.
Association between clinical disease severity and granulocyte activation markers levels in patients with NMOSD. Biomarker values are in pg/mL. Values on x-axis show the EDSS score at the time point of lumbar puncture. Regression lines show correlations in NMOSD patients without (acute 
; s/c:
) corticosteroid pre-treatment; patients with (acute: 
; s/c:
) corticosteroid pre-treatment. Open symbols designate aAQP4− patients. The Spearman’s correlation analysis of all, acute and s/c cohorts of patients showed significant correlations with the EDSS score; note that intercellular adhesion molecule-1 and vascular cellular adhesion molecule-1 levels correlated as well with EDSS scores (see online supplemental table 3). EDSS, Expanded Disability Status Scale; MMP-8, matrix metalloproteinase 8; MPO, myeloperoxidase; nEla, neutrophil elastase; NGAL, neutrophil gelatinase-associated lipocalin; NMOSD, neuromyelitis optica spectrum disorder; TIMP-1, tissue inhibitor of metalloproteinase-1.
The adhesion molecules ICAM-1 and VCAM-1 were similarly associated with EDSS scores as GAM in all, and partly in acute, NMOSD (online supplemental table 3). Being substrates for proteolytic cleavage from the cell surface by nEla and other granulocyte proteases,25 26 levels of nEla showed a strong correlation with those of ICAM-1 and VCAM-1, while this was not the case for RRMS (online supplemental figure 2); results remained essentially the same when corticosteroid-treated patients were excluded (not shown).
Granulocytes were present in nine (21%) CSF samples of patients with NMOSD of the discovery cohort. The granulocyte CSF cell count showed strong correlation with levels of granulocyte-specific activation markers, while there was only a trend for TIMP-1 and MMP-9, and no correlation with NfL and CXCL13 (online supplemental figure 3). None of these markers correlated with the CSF granulocyte cell count in INDCs or in RRMS.
Temporal dynamics of biomarker levels in relation to time between disease exacerbation and lumbar puncture
To further explore the temporal dynamics of biomarker levels observed by categorical analysis (table 3), we ran a time-dependent model applying a time window of up to 60 days after disease exacerbation (figure 2A,B). Thus, we identified three different kinetic patterns of biomarkers in NMOSD versus RRMS. Pattern 1, characterised by peak levels at NMOSD disease exacerbation with stably low or only slightly increased levels (NGAL, TIMP-1) in RRMS, comprised GAM, GFAP, S100B and adhesion molecules. All these markers discriminated NMOSD from RRMS based on the non-overlapping pointwise 95% CIs within the acute disease stage, that is, ≤21 days after disease exacerbations. In contrast, MMP-9, CXCL13 and NfL did not differ between acute stages of NMOSD and RRMS; the former two decreasing from onset in both conditions (pattern 2) and the latter increasing over time in NMOSD and RRMS (pattern 3).
Figure 2.
Modelled kinetics of biomarker levels in NMOSD and RRMS in function of days after disease exacerbation. Biomarker values are in pg/mL. Values on x-axis show days after disease exacerbations. Dotted lines determine 95% CI, based on all patients. (A) Pattern 1: increased in NMOSD, stably low (nEla, MPO. MMP-8, GFAP, S100B, ICAM-1, VCAM-1), or slightly increasing over time (NGAL, TIMP-1) in RRMS. (B) Pattern 2: increased in both NMOSD and RRMS at disease exacerbation: MMP-9 and CXCL13; pattern 3: stably high in NMOSD and RRMS: NfL. NMOSD: yellow (pooled cohorts), red (discovery cohort only); RRMS: brown (pooled cohorts), green (discovery cohort only). CXCL13, C-X-C motif chemokine 13; GFAP, glial fibrillar acidic protein; ICAM-1, intercellular adhesion molecule-1; MMP, matrix metalloproteinase; MPO, myeloperoxidase; nEla, neutrophil elastase; NGAL, neutrophil gelatinase-associated lipocalin; NfL, neurofilament light chain; NMOSD, neuromyelitis optica spectrum disorder; RRMS, relapsing-remitting multiple sclerosis; TIMP-1, tissue inhibitor of metalloproteinase-1; S100B, S100 calcium-binding protein; VCAM-1, vascular cell adhesion molecule-1.
Efficacy of single and combined biomarkers to differentiate between acute stages of NMOSD and RRMS
We next explored the diagnostic value of single GAM concentrations and of their composites to differentiate NMOSD from RRMS. To simulate a situation of unclear differential diagnosis at first disease exacerbation, we restricted the analyses to patients in acute disease stage who had not been exposed to corticosteroids before CSF sampling. Results were expressed as ROC curves with analyses being performed with and without time since exacerbation as covariate. Introducing the time elapsed from symptom onset to CSF sampling as cofactor did not improve AUC values of single GAM (0.74–0.91 with, and 0.69–0.85 without time as covariate), and had an inconsistent effect on measures of prediction of diagnosis (table 4). The combination of granulocyte-specific GAM (composite 1) alone, or in addition with TIMP-1 (composite 2), as an integrated marker raised AUC values to levels 0.90 and 0.94, respectively, leading to sensitivity and specificity values of 0.87 and 0.81 (composite 1) and 0.87 and 1.0 (composite 2), respectively (figure 3, table 4). Here, the inclusion of time as covariate further improved specificity and sensitivity values of composite 2 to 1.00 and 0.92. Neither the additional inclusion of S100B nor that of adhesion molecules into a larger composite improved the capacity to discriminate between NMOSD and RRMS further (not shown). When the risk of overfitting was addressed by calculating the AUC on 500 bootstrap replicates, these results were confirmed. Accordingly, the optimism-corrected AUCs showed a minimal reduction of both composite 1 and 2 to discriminate between NMOSD and RRMS (table 4). In essence, in untreated patients with NMOSD, with inclusion of time since disease exacerbation, specificity and sensitivity scores of these composites were within the same range as gold-standard live cell-based detection platforms for aAQP4, and better than referring ELISA-based assays27 (table 5). In patients with aAQP4− NMOSD, 71% overall (5/7 with and without corticosteroid or immunomodulatory pretreatment) and 100% (4/4) without corticosteroid pretreatment have been diagnosed based on GAM composite models as NMOSD; the two patients with NMOSD who scored negative in both algorithms had received intravenous corticosteroids or intravenous immunoglobulins 4 and 8 days prior lumbar puncture, respectively (table 6).
Table 4.
ROC analyses of pattern 1 biomarkers (granulocyte-activation markers, S100B, adhesion molecules) to differentiate NMOSD from RRMS of pooled cohorts in acute stages in patients without corticosteroid pretreatment
| Marker | Time | AUC | Youden Index | Specificity | Sensitivity | PPV | NPV | |
| Original (95% CI) | Optimism corrected |
|||||||
| nEla | – | 0.85 (0.75 to 0.95) | 0.81 | 0.14 | 0.64 | 0.93 | 0.48 | 0.96 |
| + | 0.86 (0.72 to 0.99) | 0.72 | 0.11 | 0.79 | 0.85 | 0.55 | 0.94 | |
| MPO | – | 0.78 (0.65 to 0.91) | 0.71 | 0.24 | 0.93 | 0.53 | 0.73 | 0.85 |
| + | 0.74 (0.58 to 0.91) | 0.59 | 0.21 | 0.81 | 0.62 | 0.50 | 0.87 | |
| NGAL | – | 0.85 (0.71 to 0.99) | 0.79 | 0.29 | 0.83 | 0.80 | 0.63 | 0.92 |
| + | 0.91 (0.83 to 0.99) | 0.84 | 0.21 | 0.79 | 0.92 | 0.57 | 0.97 | |
| MMP-8 | – | 0.81 (0.66 to 0.96) | 0.73 | 0.16 | 0.71 | 0.87 | 0.52 | 0.94 |
| + | 0.84 (0.71 to 0.96) | 0.69 | 0.24 | 0.86 | 0.69 | 0.60 | 0.90 | |
| TIMP-1 | – | 0.82 (0.66 to 0.98) | 0.77 | 0.33 | 0.93 | 0.73 | 0.79 | 0.91 |
| + | 0.82 (0.65 to 0.98) | 0.70 | 0.42 | 0.98 | 0.69 | 0.90 | 0.91 | |
| S100B | – | 0.69 (0.49 to 0.89) | 0.61 | 0.35 | 0.95 | 0.53 | 0.80 | 0.85 |
| + | 0.81,(0.67 to 0.94) | 0.70 | 0.34 | 0.98 | 0.54 | 0.88 | 0.87 | |
| ICAM-1 | – | 0.69 (0.51 to 0.88) | 0.62 | 0.42 | 0.98 | 0.47 | 0.88 | 0.84 |
| + | 0.76 (0.60 to 0.92) | 0.63 | 0.34 | 0.95 | 0.46 | 0.75 | 0.85 | |
| VCAM-1 | – | 0.71 (0.54 to 0.89) | 0.66 | 0.34 | 0.86 | 0.53 | 0.57 | 0.84 |
| + | 0.74 (0.58 to 0.90) | 0.63 | 0.28 | 0.79 | 0.62 | 0.47 | 0.87 | |
| nEla, NGAL, MPO, MMP-8; composite 1 |
– | 0.90 (0.79 to 1.00) | 0.84 | 0.22 | 0.81 | 0.87 | 0.62 | 0.94 |
| + | 0.96 (0.90 to 1.00) | 0.75 | 0.08 | 0.76 | 1.00 | 0.57 | 1.00 | |
| nEla, NGAL, MPO, MMP-8, TIMP-1; composite 2 | – | 0.94 (0.84 to 1.00) | 0.89 | 0.52 | 1.00 | 0.87 | 1.00 | 0.95 |
| + | 0.98 (0.94 to 1.00) | 0.79 | 0.48 | 1.00 | 0.92 | 1.00 | 0.98 | |
ROC curves are calculated on parameters estimated in the discovery cohort based on a logistic model. Youden Index as estimated in the pooled data cohort. The corrected AUC was calculated with 500 bootstrap runs.
AUC, area under the curve; MMP, matrix metalloproteinase; MPO, myeloperoxidase; nEla, neutrophil elastase; NGAL, neutrophil gelatinase-associated lipocalin; NMOSD, neuromyelitis optica spectrum disorder; NPV, negative predictive value; PPV, positive predictive value; ROC, receiver operating characteristic; RRMS, relapsing-remitting multiple sclerosis; S100B, S100 calcium-binding protein; TIMP-1, tissue inhibitor of metalloproteinase-1.
Figure 3.
ROC curves for the differentiation between NMOSD and RRMS in patients without corticosteroid pretreatment without (A, B) and with (C, D) time as covariate A B C D ROC curves of individual (A, C) GAM and their composites (B, D) (composite 1=nEla+ MPO+NGAL+MMP-8; composite 2=nEla+MPO+NGAL+MMP-8+TIMP-1). For numerical values of AUC (95% CIs), specificity and sensitivity, see table 4. AUC, area under the curve; MMP-8, matrix metalloproteinase 8; MPO, myeloperoxidase; nEla, neutrophil elastase; NGAL, neutrophil gelatinase-associated lipocalin; NMOSD, neuromyelitis optica spectrum disorder; ROC, receiver operating characteristics; RRMS, relapsing-remitting multiple sclerosis; TIMP-1, tissue inhibitor of metalloproteinase-1.
Table 5.
Comparison of validity measures of biomarker composites and aAQP4 testing to differentiate between acute NMOSD and acute RRMS
| Time included in ROC model |
N | AUC† | Sensitivity† | Specificity | |
|
Composite 1§
nEla, MPO, NGAL, MMP-8 |
No | 57 | 90 (79 to 100) | 87 (62 to 96)‡ | 81 |
| Yes | 55 | 96 (90 to 100) | 100 (77 to 100)‡ | 76 | |
|
Composite 2§
nEla, MPO, NGAL, MMP-8, TIMP-1 |
No | 57 | 94 (84 to 100) | 87 (62 to 96)‡ | 100 |
| Yes | 55 | 98 (94 to 100) | 92 (67 to 99)‡ | 100 | |
| aAQP4 test performance* | T-IIF | 79 (69 to 87) | 100 | ||
| ELISA | 60 (44 to 73) | 97 | |||
| EI-M1/M23 | 90 (78 to 96) | 95 | |||
| Fixed cell-based assay | 94 (82 to 98) | 98 | |||
| Live cell-based assay | 92 (78 to 97) | 100 | |||
*Prain et al.27
†Values are 95% CI.
‡Sensitivity values were calculated by Wilson method.
§Data based on modelling of combined cohorts, with patients without corticosteroid pretreatment.
aAQP4, anti-aquaporin-4 antibody; AUC, area under the curve; EI-M1/M23, Neuroimmune M1/M23 biochip slide; MMP, matrix metalloproteinase; MPO, myeloperoxidase; nEla, neutrophil elastase; NGAL, neutrophil gelatinase-associated lipocalin; NMOSD, neuromyelitis optica spectrum disorder; ROC, receiver operating characteristics; RRMS, relapsing-remitting multiple sclerosis; T-IIF, tissue-based indirect immunofluorescence; TIMP-1, tissue inhibitor of metalloproteinase-1.
Table 6.
Identification of patients with aAQP4− NMOSD by GAM composite algorithms
| Disease stage | Cohort | Pretreatment | Detected by composite | ||
| Corticosteroids | Immunomodulators | 1* | 2* | ||
| Acute | Discovery | No | No | Yes | Yes |
| s/c | Discovery | Intravenous −4 days of lp+oral | No | No | No |
| s/c | Discovery | Oral | Azathioprine | Yes | Yes |
| Acute | Validation | No | No | Yes | Yes |
| Acute | Validation | Intravenous −40 days of lp | IVIG −8 days of lp | No | No |
| s/c | Validation | No | No | Yes | No |
| s/c | Validation | (intravenous/tapering −11 weeks of lp)† | No | Yes | Yes |
*Based on models without time factor.
†This patient was not counted as ‘corticosteroid pretreated’, because the biological activity of the drug is unlikely to be present anymore.
aAQP4, anti-aquaporin-4 antibody; GAM, granulocyte activation markers; IVIG, intravenous immunoglobulins; lp, lumbar puncture; NMOSD, neuromyelitis optica spectrum disorder; s/c, subacute/chronic.
Discussion
Current results demonstrate that GAM produce a humoral footprint in CSF that can be used clinically to differentiate these two diseases with equal sensitivity and specificity as aAQP4 in a setting of first disease exacerbation. Our findings also establish GAM as disease activity marker by the correlation of their levels with clinical severity at NMOSD exacerbation, a feature that distinguishes them from the purely diagnostic capacity of aAQP4.4 Moreover, as GAM are also upregulated in aAQP4− NMOSD, they can close a diagnostic gap for these patients.5 Accordingly, metabolomic approaches have allowed to differentiate with high accuracy aAQP4− NMOSD versus MS based on increased plasma levels of myoinositol and formate in the latter disease; different from our study cohort these results were derived from an out of relapse population and it is not known whether the differentiation between the two disease would apply as well in acute disease.28
The correlations of GAM levels with CSF granulocytosis and acute disability scores strengthen the concept of a pathogenetic link between recruitment and activation of granulocytes, neural tissue damage and development of disability in NMOSD. In this context, it is notable that a significant number of patients with acute and s/c NMOSD had markedly increased GAM concentrations, despite corticosteroid or immunomodulating therapy prior to sampling. On the group level, current results suggest that such therapy has only limited capacity to reduce GAM expression in the course of NMOSD exacerbation.
Most other markers tested here displayed overlapping concentration ranges in acute stages of NMOSD versus RRMS, making them unsuitable for differentiating the two diseases in case of individual exacerbations. Furthermore, their levels did not correlate with disability scores, likely because their modulation reflects downstream effects in the course of the inflammatory response in NMOSD. Nevertheless, generic tissue injury-markers, such as NfL and GFAP, may still be clinically useful, as they allow to monitor disease activity and therapeutic response in NMOSD and may predict its long-term disability course, not the least since blood-based samples allow for longitudinal assessments.6 29 30 Despite not being granulocyte products, the leucocyte adhesion molecules VCAM-1 and ICAM-1 were increased in NMOSD compared with RRMS in present results and as found by others.12 Their increase may be an indirect result of the release of enzymes in the course of granulocyte activation, since both molecules are substrates for proteolysis by elastase and other neutrophil secretory enzymes.25 26
The half-life time and kinetics of GAM under physiological conditions and in disease are unknown and may show incongruent kinetics among them. Accordingly, as the time between start of their release in the course of disease exacerbation and lumbar puncture may vary, the individual levels of GAM did not show a consistent pattern of correlation. These findings have their correlate in a recent study in patients with type 2 diabetes mellitus where increased serum levels of nEla and MPO showed only a moderate (rho=0.56) correlation.31 Hence, the rational for the use of GAM composites, rather than a single marker, for the differentiation between NMOSD and MS is to compensate the variability of their levels at respective times of lumbar puncture. An advantage of the proposed GAM composites is that they rest on an analytical platform, ELISA, that is simple to execute and technically robust and allows to differentiate patients with aAQP4− NMOSD from MS, this within a day of sampling as compared with a 1–2 weeks laboratory turnaround time of gold-standard cell-based assays for aAQP4.32 33 Both aspects are clinically important, as the diagnosis of acute NMOSD necessitates a seamless start of plasma exchange (PEX), to optimise its effectiveness.34–36 PEX may not remain the only therapeutic option as novel immunomodulatory therapies specifically interfering with effector molecules of NMOSD pathogenesis, such as protease and complement inhibitors, may emerge as acute phase therapies. For example, eculizumab is currently registered only as interval therapy for secondary prophylaxis against acute exacerbations of NMOSD. However, this compound is in off-label use in acute phases of haemolytic-uraemic syndrome37 among other diseases that go along with acute complement factor 5 (C5) activation and may also be a therapeutic option in patients with NMOSD with acute exacerbations when PEX provides only limited or no benefit.38 Here, GAM composites may be a valuable biomarker for therapeutic decision making, on the background of the enormous costs of anti-C5-antibody therapies.
The clinical finding of a correlation of GAM with neurological impairment corroborate a large body of evidence for the pathogenic role of granulocytes and their secretory products in preclinical models of NMOSD. Thus, granulocyte depletion preserves blood-brain barrier integrity and reduces lesional damage in in vivo rodent models of NMOSD, while induction of a neutrophilic state by granulocyte colony-stimulating factor led to increased neural damage.16 39 An ex vivo model of NMOSD showed extensive potentiation of complement-mediated spinal cord damage by the addition of elastase,40 which could partly be suppressed by the elastase-inhibitor sivelestat and other inhibitors of neutrophil enzymes.16 39 40 Sivelestat also demonstrated therapeutic effects in a rodent in vivo model of NMOSD, but not in MS-like experimental autoimmune encephalomyelitis.41 This study also observed increased serum levels of nEla in patients with NMOSD and provided a possible explanation why interferon-β seems to induce NMOSD exacerbations in humans, since this cytokine induces the release of nEla in cultured granulocytes.41
Limitations
The diagnostic capacity of GAM was only evaluated in NMOSD versus RRMS, while increasing evidence suggests that granulocytes are also involved in the pathogenesis of MOGAD, that is as well difficult to distinguish in acute stage from RRMS and NMOSD.18 19 In a preliminary report, we have found that patients with MOGAD, similarly to NMOSD, displayed a GAM pattern that differentiated it from RRMS.42 We are currently extending these preliminary data based on a larger cohort of patients with MOGAD, in an attempt to explore possible qualitative and quantitative differences of biomarker profiles between this condition, RRMS and NMOSD. Second, there is a need to expand the database of the capacity of GAM to identify aAQP4− NMOSD, as the number of patients is currently small.
Conclusions
Current findings establish GAM as first biofluid markers of NMOSD reflective of the clinical degree of neurological impairment. Second, they establish GAM as an alternative biomarker to aAQP4 for the differential diagnosis of NMOSD versus RRMS, also comprising aAQP4− disease that shares with typical NMOSD granulocyte activation as a common pathomechanism. Third, together with previous preclinical evidence that inhibition of proteolytic activity of granulocyte-derived enzymes inhibits tissue damage in NMOSD models, this study identifies GAM as potential novel drug targets for acute-stage NMOSD.
Footnotes
Contributors: DL, MW and SS had full access to all data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
DL is responsible for the overall content as guarantor.
Concept and design: DL, SS, MW, RF, MG and JK. Acquisition, analysis or interpretation of data: all authors. Drafting of the manuscript: DL, MW, SS, FP, RF, MG, JO, SM and JK. Critical revision of the manuscript for important intellectual content: all authors. Statistical analysis: DL, MW, SS, FP and JK. Administrative, technical or material support: DL, SS, FP, RF, MG, JL, BE, KF, AO, SM and JK. Supervision: DL, MW, FP and JK.
Funding: This investigation was supported by Swiss National Science Foundation (grant 320030_189140/1), the Health and Labour Sciences Research Grant on Intractable Diseases (Neuroimmunological Diseases) from the Ministry of Health, Labour and Welfare of Japan (20FC1030) and Swedish MRC grant no. 2020-02700, Hjärnfonden.
Competing interests: DL is Chief Medical Officer of GeNeuro. MW received speaker honoraria from Novartis Pharma, Chugai Pharmaceutical, Biogen Japan and Alexion. FP has received research grants from Janssen, Merck KGaA and UCB, and fees for serving on DMC in clinical trials with Chugai, Lundbeck and Roche, and preparation of witness report for Novartis. RF has received speaker fees for teaching and workshops from Biogen, Merck, Novartis, Roche, Teva and Alexion. For educational activities, courses or research, he has received unrestricted grants from Biogen, EMD Serono. JL is an employee of Quanterix. BE has received travel grants for ECTRIMS 2018 from Roche. KF has served on advisory boards and received speaker honoraria from Biogen, Roche and Merck and received research funds from Amicus. TM received speaker honoraria from Biogen Japan, Chugai Pharmaceutical, Alexion Pharmaceuticals, Novartis Pharma and Takeda Pharmaceutical. KM received speaker honoraria from Novartis Pharma, Chugai Pharmaceutical and Nihon Pharmaceutical. NI received grant support from Mitsubishi Tanabe Pharma, Osoegawa Neurology Clinic, Bayer Yakuhin and Japan Blood Products Organization and speaker honoraria from Novartis Pharma, Biogen Japan, Alexion, Mitsubishi Tanabe Pharma, Chugai Pharmaceutical, Teijin Pharma and Eisai. J-IK received research funds from Dainippon Sumitomo Pharma, Daiichi Sankyo, Mitsubishi Tanabe Pharma and Kyowa Kensetsukougyo, and consultancy fees, speaking fees and/or honoraria from Novartis Pharma, Mitsubishi Tanabe Pharma, CSL Behring, Biogen Japan, Teijin Health Care, the Takeda Pharmaceutical, Kyowa Kirin, Ono Pharmaceutical, Alexion Pharmaceuticals, Tsumura, Ricoh, EMC and Eisai. JO served on advisory boards for Roche and Merck. JK received speaker fees, research support, travel support and/or served on advisory boards by the Progressive MS Alliance, Swiss MS Society, Swiss National Research Foundation (320030_189140/1), University of Basel, Biogen, Celgene, Merck, Novartis, Octave Bioscience, Roche, Sanofi. No other disclosures were reported.
Provenance and peer review: Not commissioned; externally peer reviewed.
Supplemental material: This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.
Data availability statement
Data are available on reasonable request.
Ethics statements
Patient consent for publication
Not applicable.
Ethics approval
This study was approved by StockholmRegionala Etikprövningsnämnden i Stockholm 2010-02-16, amended several times, last amendment (including waiver for reconsenting those sampled previously), Etikprövningsmyndigheten (Stockholm avdelning 2 medicin), Dnr 2022-03650-022. San RafaeleIRCCS San Raffaele Hospital Ethical Committee, study acronym BANCA-INSPE, number DSAN 1178/53. BaselEthikkommission beider Basel Ref. Nr. EK: 332/064. Pavia Local Ethics Committee IRCCS San Matteo, Pavia, Italy, project code p-202000395415, Kyushu Ethical Committee of Kyushu University (reference number: 730-04). Participants gave informed consent to participate in the study before taking part.
References
- 1. Kuchling J, Paul F. Visualizing the central nervous system: imaging tools for multiple sclerosis and neuromyelitis optica spectrum disorders. Front Neurol 2020;11(June):450. 10.3389/fneur.2020.00450 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2. Hamid SHM, Whittam D, Mutch K, et al. What proportion of AQP4-igG-negative NMO spectrum disorder patients are MOG-igg positive? A cross sectional study of 132 patients. J Neurol 2017;264:2088–94. 10.1007/s00415-017-8596-7 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3. Jarius S, Paul F, Franciotta D, et al. Cerebrospinal fluid findings in aquaporin-4 antibody positive neuromyelitis optica: results from 211 lumbar punctures. J Neurol Sci 2011;306:82–90. 10.1016/j.jns.2011.03.038 [DOI] [PubMed] [Google Scholar]
- 4. Schmetzer O, Lakin E, Roediger B, et al. Anti-aquaporin 4 igG is not associated with any clinical disease characteristics in neuromyelitis optica spectrum disorder. Front Neurol 2021;12(March):635419. 10.3389/fneur.2021.635419 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5. Juryńczyk M, Weinshenker B, Akman-Demir G, et al. Status of diagnostic approaches to AQP4-igG seronegative NMO and NMO/MS overlap syndromes. J Neurol 2016;263:140–9. 10.1007/s00415-015-7952-8 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6. Watanabe M, Nakamura Y, Michalak Z, et al. Serum GFAP and neurofilament light as biomarkers of disease activity and disability in NMOSD. Neurology 2019;93:e1299–311. 10.1212/WNL.0000000000008160 [DOI] [PubMed] [Google Scholar]
- 7. Wei Y, Chang H, Li X, et al. CSF-S100B is a potential candidate biomarker for neuromyelitis optica spectrum disorders. Biomed Res Int 2018;2018:5381239. 10.1155/2018/5381239 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8. Liu C, Zhao L, Fan P, et al. High serum neurofilament levels among chinese patients with aquaporin-4-igG-seropositive neuromyelitis optica spectrum disorders. J Clin Neurosci 2021;83(January 2019):108–11. 10.1016/j.jocn.2020.11.016 [DOI] [PubMed] [Google Scholar]
- 9. Wang S, Yang T, Wan J, et al. Elevated C-X-C motif ligand 13 and B-cell-activating factor levels in neuromyelitis optica during remission. Brain Behav 2017;7:e00648. 10.1002/brb3.648 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10. Alvarez E, Piccio L, Mikesell RJ, et al. CXCL13 is a biomarker of inflammation in multiple sclerosis, neuromyelitis optica, and other neurological conditions. Mult Scler 2013;19:1204–8. 10.1177/1352458512473362 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. Khademi M, Kockum I, Andersson ML, et al. Cerebrospinal fluid CXCL13 in multiple sclerosis: a suggestive prognostic marker for the disease course. Mult Scler 2011;17:335–43. 10.1177/1352458510389102 [DOI] [PubMed] [Google Scholar]
- 12. Uzawa A, Mori M, Masuda S, et al. Markedly elevated soluble intercellular adhesion molecule 1, soluble vascular cell adhesion molecule 1 levels, and blood-brain barrier breakdown in neuromyelitis optica. Arch Neurol 2011;68:913–7. 10.1001/archneurol.2011.148 [DOI] [PubMed] [Google Scholar]
- 13. Hosokawa T, Nakajima H, Doi Y, et al. Increased serum matrix metalloproteinase-9 in neuromyelitis optica: implication of disruption of blood-brain barrier. J Neuroimmunol 2011;236:81–6. 10.1016/j.jneuroim.2011.04.009 [DOI] [PubMed] [Google Scholar]
- 14. Lucchinetti CF, Mandler RN, McGavern D, et al. A role for humoral mechanisms in the pathogenesis of Devic’s neuromyelitis optica. Brain 2002;125(Pt 7):1450–61. 10.1093/brain/awf151 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15. Lucchinetti CF, Guo Y, Popescu BFG, et al. The pathology of an autoimmune astrocytopathy: lessons learned from neuromyelitis optica. Brain Pathol 2014;24:83–97. 10.1111/bpa.12099 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Winkler A, Wrzos C, Haberl M, et al. Blood-Brain barrier resealing in neuromyelitis optica occurs independently of astrocyte regeneration. J Clin Invest 2021;131:e141694. 10.1172/JCI141694 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17. Takai Y, Misu T, Suzuki H, et al. Staging of astrocytopathy and complement activation in neuromyelitis optica spectrum disorders. Brain 2021;144:2401–15. 10.1093/brain/awab102 [DOI] [PubMed] [Google Scholar]
- 18. Hochmeister S, Gattringer T, Asslaber M, et al. A fulminant case of demyelinating encephalitis with extensive cortical involvement associated with anti-MOG antibodies. Front Neurol 2020;11(February):31. 10.3389/fneur.2020.00031 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19. Höftberger R, Guo Y, Flanagan EP, et al. The pathology of central nervous system inflammatory demyelinating disease accompanying myelin oligodendrocyte glycoprotein autoantibody. Acta Neuropathol 2020;139:875–92. 10.1007/s00401-020-02132-y [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20. Murata H, Kinoshita M, Yasumizu Y, et al. Cell-Free DNA derived from neutrophils triggers type 1 interferon signature in neuromyelitis optica spectrum disorder. Neurol Neuroimmunol Neuroinflamm 2022;9:1–11.:e1149. 10.1212/NXI.0000000000001149 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21. Wingerchuk DM, Banwell B, Bennett JL, et al. International consensus diagnostic criteria for neuromyelitis optica spectrum disorders. Neurology 2015;85:177–89. 10.1212/WNL.0000000000001729 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22. Thompson AJ, Banwell BL, Barkhof F, et al. Diagnosis of multiple sclerosis: 2017 revisions of the mcdonald criteria. Lancet Neurol 2018;17:162–73. 10.1016/S1474-4422(17)30470-2 [DOI] [PubMed] [Google Scholar]
- 23. Neurostatus-UHB ltd c/o university hospital basel switzerland. 2016. Available: www.neurostatus.net
- 24. Teunissen C, Menge T, Altintas A, et al. Consensus definitions and application guidelines for control groups in cerebrospinal fluid biomarker studies in multiple sclerosis. Mult Scler 2013;19:1802–9. 10.1177/1352458513488232 [DOI] [PubMed] [Google Scholar]
- 25. Champagne B, Tremblay P, Cantin A, et al. Proteolytic cleavage of ICAM-1 by human neutrophil elastase. J Immunol 1998;161:6398–405. 10.4049/jimmunol.161.11.6398 [DOI] [PubMed] [Google Scholar]
- 26. Lévesque JP, Takamatsu Y, Nilsson SK, et al. Vascular cell adhesion molecule-1 (CD106) is cleaved by neutrophil proteases in the bone marrow following hematopoietic progenitor cell mobilization by granulocyte colony-stimulating factor. Blood 2001;98:1289–97. 10.1182/blood.v98.5.1289 [DOI] [PubMed] [Google Scholar]
- 27. Prain K, Woodhall M, Vincent A, et al. AQP4 antibody assay sensitivity comparison in the era of the 2015 diagnostic criteria for NMOSD. Front Neurol 2019;10(October):1028. 10.3389/fneur.2019.01028 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28. Yeo T, Probert F, Jurynczyk M, et al. Classifying the antibody-negative NMO syndromes: clinical, imaging, and metabolomic modeling. Neurol Neuroimmunol Neuroinflamm 2019;6:e626. 10.1212/NXI.0000000000000626 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29. Schindler P, Grittner U, Oechtering J, et al. Serum GFAP and NFL as disease severity and prognostic biomarkers in patients with aquaporin-4 antibody-positive neuromyelitis optica spectrum disorder. J Neuroinflammation 2021;18:105. 10.1186/s12974-021-02138-7 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30. Schindler P, Aktas O, Ringelstein M, et al. Glial fibrillary acidic protein as a biomarker in neuromyelitis optica spectrum disorder: a current review. Expert Rev Clin Immunol 2023;19:71–91. 10.1080/1744666X.2023.2148657 Available: from:%20http://www.ncbi.nlm.nih.gov/pubmed/36378751 [DOI] [PubMed] [Google Scholar]
- 31. Alexandrovski M, Suciu S, Alexandrovski J. Joint measurements of leukocyte elastase and myeloperoxidase promote identification of the state of neutrophils in diabetic patients. Biores Open Access 2020;9:190–7. 10.1089/biores.2020.0012 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32. Mayo Clinical Laboratories . Neuromyelitis optica (NMO)/aquaporin-4-igG fluorescence-activated cell sorting (FACS) assay serum [internet]. 2021. Available: https://www.mayocliniclabs.com/test-catalog/Overview/38324
- 33. Oxford University Hospitals. Neuromyelitis Optica Antibodies [Internet] 2021. Available: https://www.ouh.nhs.uk/immunology/diagnostic-tests/tests-catalogue/neuromyelitis-optica-antibodies.aspx [Google Scholar]
- 34. Stiebel-Kalish H, Hellmann MA, Mimouni M, et al. Does time equal vision in the acute treatment of a cohort of AQP4 and MOG optic neuritis? Neurol Neuroimmunol Neuroinflamm 2019;6:e572. 10.1212/NXI.0000000000000572 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35. Kleiter I, Gahlen A, Borisow N, et al. Apheresis therapies for NMOSD attacks: a retrospective study of 207 therapeutic interventions. Neurol Neuroimmunol Neuroinflamm 2018;5:e504. 10.1212/NXI.0000000000000504 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36. Bonnan M, Valentino R, Debeugny S, et al. Short delay to initiate plasma exchange is the strongest predictor of outcome in severe attacks of NMO spectrum disorders. J Neurol Neurosurg Psychiatry 2018;89:346–51. 10.1136/jnnp-2017-316286 [DOI] [PubMed] [Google Scholar]
- 37. Benoit SW, Fukuda T, VandenHeuvel K, et al. Case report: atypical HUS presenting with acute rhabdomyolysis highlights the need for individualized eculizumab dosing. Front Pediatr 2022;10(February):841051.:841051. 10.3389/fped.2022.841051 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38. Chatterton S, Parratt JDE, Ng K. Eculizumab for acute relapse of neuromyelitis optica spectrum disorder: case report [Internet]. Front Neurol 2022;13:951423. 10.3389/fneur.2022.951423 Available: https://www.frontiersin.org/articles/10.3389/fneur.2022.951423/full [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39. Saadoun S, Waters P, MacDonald C, et al. Neutrophil protease inhibition reduces neuromyelitis optica-immunoglobulin G-induced damage in mouse brain. Ann Neurol 2012;71:323–33. 10.1002/ana.22686 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40. Zhang H, Bennett JL, Verkman AS. Ex vivo spinal cord slice model of neuromyelitis optica reveals novel immunopathogenic mechanisms. Ann Neurol 2011;70:943–54. 10.1002/ana.22551 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41. Herges K, de Jong BA, Kolkowitz I, et al. Protective effect of an elastase inhibitor in a neuromyelitis optica-like disease driven by a peptide of myelin oligodendroglial glycoprotein. Mult Scler 2012;18:398–408. 10.1177/1352458512440060 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 42. Leppert D. Potential of neutrophil granulocyte markers in CSF to differentiate NMOSD and MOGAD from MS [internet]. in: MS virtual 2020. 2020. Available: https://touchneurology.com/multiple-sclerosis/conference-hub/david-leppert-msvirtual2020-potential-of-neutrophil-granulocyte-markers-in-csf-to-differentiate-nmosd-and-mogad-from-ms/
- 43. Wang X, Rojas-Quintero J, Wilder J, et al. Tissue inhibitor of metalloproteinase-1 promotes polymorphonuclear neutrophil (PMN) pericellular proteolysis by anchoring matrix metalloproteinase-8 and -9 to PMN surfaces. J Immunol 2019;202:3267–81. 10.4049/jimmunol.1801466 [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
jnnp-2022-330796supp001.pdf (590.6KB, pdf)
Data Availability Statement
Data are available on reasonable request.