Abstract
Objective To test the hypothesis that serum levels of glial fibrillary acidic protein (GFAP) and neurofilament light chain (NfL), which are an intermediate astrocyte and neuron filaments, respectively, are clinically useful biomarkers of disease activity and disability in neuromyelitis optica spectrum disorders (NMOSD).
Methods Levels of GFAP and NfL in serum (sGFAP and sNfL, respectively) and in CSF samples were measured in healthy controls (HCs) (n = 49; 49 serum samples), patients with NMOSD (n = 33; 42 CSF and 102 serum samples), and patients with multiple sclerosis (MS) (n = 49; 53 CSF and 91 serum samples) by ultrasensitive single-molecule array assays. Association of sGFAP and sNfL levels with clinical parameters was determined.
Results For both GFAP and NfL, CSF and serum levels were strongly correlated. Both were higher in the serum of patients with NMOSD than in HCs (both p < 0.001). Moreover, sGFAP was higher in NMOSD than in MS (median 207.7 vs 121.1 pg/mL, p < 0.001). In NMOSD, sGFAP concentration increased after recent relapse (540.9 vs 152.9 pg/mL, p < 0.001). Multivariate analyses indicated that sGFAP and sNfL were associated with Expanded Disability Status Scale score in NMOSD (p = 0.026 and p < 0.001, respectively). Higher sGFAP/sNfL quotient at relapse differentiated NMOSD from MS with a sensitivity of 73.0% and a specificity of 75.8%.
Conclusions sGFAP and sNfL are likely to be good biomarkers of disease activity and disability, and the sGFAP/sNfL quotient at relapse is a potential diagnostic marker for NMOSD.
Glossary
- AQP4=
- aquaporin-4;
- CI=
- confidence interval;
- DMD=
- disease-modifying drug;
- EDSS=
- Expanded Disability Status Scale;
- GFAP=
- glial fibrillary acidic protein;
- HCs=
- healthy controls;
- IgG=
- immunoglobulin G;
- IQR=
- interquartile range;
- MS=
- multiple sclerosis;
- Nf=
- neurofilament;
- NfL=
- neurofilament light chain;
- NMOSD=
- neuromyelitis optica spectrum disorders;
- PPMS=
- primary progressive multiple sclerosis;
- ROC=
- receiver operating characteristic;
- RRMS=
- relapsing-remitting multiple sclerosis;
- sGFAP=
- serum glial fibrillary acidic protein;
- sNfL=
- serum neurofilament light chain;
- SPMS=
- secondary progressive multiple sclerosis
Neuromyelitis optica spectrum disorders (NMOSD) are autoimmune inflammatory diseases of the CNS in which autoantibodies targeting aquaporin-4 (AQP4) on perivascular astrocyte endfeet play a major role.1 Glial fibrillary acidic protein (GFAP) is a principal intermediate filament that forms the astrocyte cytoskeleton.2 GFAP concentration in CSF is higher in NMOSD (table e-1 available from Dryad, doi.org/10.5061/dryad.dp75314)3,–,6 and is regarded as a biomarker of astrocyte injury. In addition, increased CSF-GFAP has been reported in multiple sclerosis (MS), especially at progressive stages,7,–,10 and reportedly correlates with higher levels of disability7,–,11 and disability worsening,11,12 suggesting that CSF-GFAP partially reflects reactive astrogliosis (table e-1 available from Dryad). However, no studies have specifically examined whether serum GFAP (sGFAP) levels are associated with clinical parameters in NMOSD.
Neurofilaments (Nf) are structural elements of neurons composed of light (NfL), medium, and heavy Nf chains.13 Neuro-axonal damage results in Nf release, which then appears in CSF and blood.14,15 Although NfL levels are high in CSF from patients with a wide spectrum of neurological diseases, lumbar punctures are relatively invasive procedures that limit longitudinal and clinical application.16 Recently, novel single-molecule array (Simoa; Quanterix, Lexington, MA) assays have been developed that can measure NfL levels in blood with higher sensitivity.16,–,18 This method revealed that serum NfL (sNfL) levels are high in patients with neurodegenerative diseases, trauma, and MS.19,–,22 However, an association between sNfL levels and clinical parameters in NMOSD has yet to be assessed with this sensitive technology.
We therefore measured sGFAP and sNfL in NMOSD and assessed their usefulness as biomarkers.
Methods
Participants
Overall, 33 patients with NMOSD with or without AQP4–immunoglobulin G (IgG) (with: n = 30, 39 CSF and 99 serum samples, 39 paired samples; without: n = 3, 3 CSF and 3 serum samples, 3 paired samples) and 49 patients with MS seronegative for AQP4-IgG (53 CSF and 91 serum samples, 53 paired samples) were enrolled in the study. CSF and serum samples were retrospectively collected and stored at −80°C in the Department of Neurology at Kyushu University Hospital (Fukuoka, Japan). At baseline, among the patients with MS, 38 were in the relapsing-remitting MS (RRMS) stage, 7 had secondary progressive MS (SPMS), and 4 had primary progressive MS (PPMS). Six patients with RRMS converted to SPMS during follow-up. Serum samples were also collected from 49 healthy controls (HCs) (49 samples) whose age and sex were matched to those of the patients with NMOSD (table 1 and tables e-2 and e-3 available from Dryad, doi.org/10.5061/dryad.dp75314). All samples were collected between September 1, 2004, and September 20, 2017. As shown in table e-2 available from Dryad, 58.9% of the CSF samples were obtained within recent relapses (maximum of 2 months between symptom onset of relapse and sampling) because most were collected for diagnostic reasons, whereas serum samples were obtained mainly during remission. Samples from NMOSD and RRMS were collected at comparable intervals after relapses (all samples: NMOSD 4.0 [0.0–15.0] months and RRMS 1.0 [0.0–15.5] months, p = 0.370; samples during remission: NMOSD 11.0 [5.0–29.0] months and RRMS 15.5 [6.0–28.0] months, p = 0.596). In samples collected during remission, 47 (34 from NMOSD and 13 from RRMS) were collected in the early stage of remission (<12 months from onset of recent relapse), while 50 (31 from NMOSD and 19 from RRMS) were collected in the late stage (≥12 months from onset of recent relapse).
Demographic features of participants at baseline
Diagnosis of NMOSD was based on diagnostic criteria published in 2015,23 and that of MS was based on the 2010 McDonald criteria.24 Medical records, laboratory data, and MRI findings were retrospectively reviewed. A relapse was defined according to the 2010 McDonald criteria,24 and disability was evaluated with the Expanded Disability Status Scale (EDSS).25 The number of T2 lesions (including brain and spinal cord lesions) was counted with T2-weighted and/or fluid-attenuated inversion-recovery MRIs.
Standard protocol approvals, registrations, and patient consents
This study was reviewed and approved by the Ethical Committee of Kyushu University. All patients and HCs provided written informed consent.
Measurement of NfL and GFAP
Serum samples were centrifuged at 1,740g for 10 minutes at room temperature, and CSF samples were centrifuged 120g for 5 minutes at 4°C. Both were stored at −80°C within 3 hours of collection. For measuring NfL concentrations in CSF and sera by the Simoa NfL assay, we used the same methodology as that used in a previous study.20
GFAP concentration was measured in serum and CSF with a commercially available Simoa GFAP Discovery Kit. Calibrators, samples, and quality controls (serum 1:4 dilution, CSF 1:40 dilution) were measured in duplicate. The interassay coefficients of variation for 3 native serum samples were 10%, 4%, and 10% for control samples, with mean concentrations of 47.5, 80.7, and 102.5 pg/mL, respectively. The mean intra-assay coefficients of variation for duplicate determinations of concentration were 4.7% in serum and 2.7% in CSF. Eleven CSF samples and 9 serum samples exhibited GFAP values above the highest calibrator (4,000 and 40,000 pg/mL for serum and CSF samples, respectively). These samples were remeasured at higher dilutions. All sample measurements were done blinded, and analysts were unaware of any clinical or diagnostic information.
Statistics
Categorical variables were described by counts and percentages, and continuous and ordinal variables by median and interquartile ranges (IQRs). Demographic features of participants at baseline were compared by use of the Fisher exact test or the Wilcoxon test. To analyze GFAP and NfL levels and the GFAP/NfL quotient, levels of these parameters were log-transformed to meet the normal assumption. However, for clarity, the results are described using the original scales. The associations between CSF and serum levels for NfL and GFAP were tested with linear regression models in which regression coefficients are denoted with β. In the HCs, the associations of sex and age with sNfL or sGFAP levels were assessed by t tests and linear regression models, respectively. A repeated-measure mixed-effect model analysis of covariance with a first-order autoregressive structure on time was used to investigate clinical associations with levels of sNfL, sGFAP, or sGFAP/sNfL. This model considers within-participant or within-relapse correlations from repeated measurements. In these mixed-effect models, regression coefficients (βmult) were back-transformed to the original scale and therefore reflect multiplicative effects. When sGFAP and sNfL levels or the sGFAP/sNfL quotient was compared between groups, analyses were corrected for age and sex. Age, sex, and EDSS score were included in all models for multivariate analysis. Other variables adopted in each multivariate analysis are indicated in the footnote of the corresponding table. A receiver operating characteristic (ROC) curve was generated to estimate the diagnostic accuracy of the sGFAP/sNfL quotient, and sensitivity and specificity of the optimum cutoff point were defined to maximize the area under the ROC curve. All analyses were performed with JMP Pro version 13.0.0 software (SAS Institute, Cary, NC) or statistical software R (version 3.5.0, R Foundation for Statistical Computing, Vienna, Austria). The significance level was set at p < 0.05.
Data availability
Anonymized data not published within this article are available from the corresponding authors on reasonable request from any qualified investigator.
Results
Baseline demographics and disease characteristics
Patients with MS were younger than the HCs and tended to be younger than those with NMOSD (table 1). Patients with NMOSD showed a trend for having higher EDSS scores than those with MS. As expected, oligoclonal IgG bands and CNS T2 lesions were more frequent in patients with MS than in those with NMOSD, while longitudinally extensive spinal cord lesions were more prevalent in patients with NMOSD than in those with MS (table 1).
Correlation between CSF and serum levels for GFAP and NfL
After the data from the patient groups were pooled, analysis of paired CSF and serum samples revealed that median GFAP concentration in serum (167.0 [IQR 101.2–317.6] pg/mL) was 51.5-fold lower than in the CSF (8,601.5 [6,129.3–15,227.2] pg/mL). We found a strong linear correlation between CSF-GFAP and sGFAP levels, with a 10% increase in CSF-GFAP leading to 8.1% and 7.4% increases in sGFAP for NMOSD and MS, respectively (NMOSD: log10[sGFAP] = −0.963 + 0.813 × log10[CSF-GFAP], β = 0.81, 95% confidence interval [CI] 0.74–0.88, p < 0.001; MS: log10[sGFAP] = −0.761 + 0.746 × log10[CSF-GFAP], β = 0.75, 95% CI 0.53–0.97, p < 0.001; figure 1A). This relationship was similar between patients with NMOSD and those with MS (p = 0.558).
(A and B) Correlation between CSF and serum levels for (A) glial fibrillary acidic protein (GFAP) and (B) neurofilament light chain (NfL) in patients with neuromyelitis optica spectrum disorders (NMOSD) or multiple sclerosis (MS). Paired CSF and serum samples were available for 42 patients with NMOSD and 53 patients with MS. (C and D) Correlation of (C) serum GFAP (sGFAP) or (D) serum (sNfL) levels with Expanded Disability Status Scale (EDSS) score in patients with NMOSD or MS. (E and F) Correlation of (E) sGFAP and (F) sNfL levels with months after relapse in patients with NMOSD or MS. Sample data from 1 patient with NMOSD were deemed to be an outlier (sGFAP 152.9 pg/mL, sNfL 24.4 pg/mL, 256 months after relapse) and are not shown in panels E and F. Translucent bands indicate 95% confidence intervals.
After the data from the patient groups were pooled, analysis of paired CSF and serum samples revealed that median NfL concentration in serum (33.3 [22.1–72.3] pg/mL) was 53.7-fold lower than in the CSF (1,786.7 [1,156.9–5,140.5] pg/mL). As with GFAP, the 2 concentrations correlated with each other; a 10% increase in CSF-NfL led to 5.8% and 7.4% increases in sNfL for NMOSD and MS, respectively (NMOSD: log10[sNfL] = −0.404 + 0.595 × log10[CSF-NfL], β = 0.60, 95% CI 0.42–0.77, p < 0.001; MS: log10[sNfL] = −0.929 + 0.747 × log10[CSF-NfL], β = 0.75, 95% CI 0.61–0.88, p < 0.001; figure 1B). The degree of sNfL increase per CSF-NfL increase did not differ significantly between the NMOSD and MS groups (p = 0.161).
Group comparison of sGFAP and sNfL levels
In HCs, median sGFAP and sNfL concentrations were 97.2 (IQR 77.9–141.4) and 20.1 (12.7–24.1) pg/mL, respectively. sGFAP and sNfL levels were positively associated with age (βmult = 1.01, p = 0.009; βmult = 1.03, p < 0.001, respectively) but did not differ between sexes.
GFAP levels
Median sGFAP levels in patients with NMOSD and MS were 207.7 (IQR 117.7–433.7) and 121.1 (89.6–201.0) pg/mL, respectively. sGFAP levels were significantly higher in patients than in HCs (NMOSD: βmult = 3.07, p < 0.001; MS: βmult = 1.44, p = 0.004; figure 2A). Moreover, sGFAP levels were higher in patients with NMOSD than in those with MS (βmult = 2.28, p < 0.001). sGFAP levels were higher in patients with NMOSD in both relapse and remission than they were in HCs (relapse: 540.9 [169.9–1,540.6] pg/mL, βmult = 7.48, p < 0.001; remission: 152.9 [111.6–256.4] pg/mL, βmult = 1.81, p < 0.001), while levels during relapse were markedly higher than they were during remission (βmult = 4.05, p < 0.001; figure 2B). Moreover, sGFAP levels even in the late stage of remission in NMOSD (133.5 [104.2–211.6] pg/mL) were higher than those of HCs (βmult = 1.43, p = 0.003). In patients with MS, sGFAP levels were significantly higher in those with RRMS and SPMS/PPMS than they were in HCs (RRMS: 111.9 [90.2–181.7] pg/mL, βmult = 1.52, p = 0.003; SPMS/PPMS: 186.4 [87.3–235.1] pg/mL, βmult = 1.45, p = 0.008), while levels did not differ between patients with RRMS and those with SPMS/PPMS (p = 0.602). In patients with RRMS, sGFAP levels were higher during relapse (129.8 [92.5–236.1] pg/mL) than during remission (109.4 [89.9–129.3] pg/mL, βmult = 1.25, p = 0.035) or in HCs (βmult = 1.65, p = 0.002). Moreover, sGFAP levels during remission also tended to be higher than those in HCs (βmult = 1.21, p = 0.085).
(A and B) Serum glial fibrillary acidic protein (sGFAP) levels, (C and D) serum neurofilament light chain (sNfL) levels, and (E and F) sGFAP/sNfL were compared between healthy controls (HCs) and patients with neuromyelitis optica spectrum disorders (NMOSD) or multiple sclerosis (MS). In panels A, C, and E, all samples from each group were included. In panels B, D, and F, NMOSD samples are subdivided into relapse and remission phases, while MS samples are subdivided into relapse and remission phases of relapsing-remitting MS (RRMS) and progressive MS, including secondary progressive MS (SPMS)/primary progressive MS (PPMS). Boxes depict median and interquartile range (IQR), and upper/lower whiskers extend from the hinge toward the largest/smallest values but no farther than 1.5 × IQR from the hinge. The p values were obtained with a mixed-effect model adjusted for age and sex. *p < 0.05, **p < 0.01, and ***p < 0.001.
NfL levels
Median sNfL levels in patients were 36.5 (IQR 22.7–56.4) pg/mL for NMOSD and 26.2 (16.8–40.2) pg/mL for MS. These levels were significantly higher in patients than in HCs (NMOSD: βmult = 2.02, p < 0.001; MS: βmult = 1.79, p < 0.001; figure 2C). Although sNfL levels were comparable between both diseases during relapse (NMOSD: 37.0 [22.8–69.6] pg/mL; MS: 30.2 [24.1–91.6] pg/mL; p = 0.955), they were significantly higher in NMOSD than in MS during remission (NMOSD: 36.5 [22.3–52.8] pg/mL; MS: 17.3 [13.5–24.6] pg/mL; βmult = 1.25, p = 0.013; figure 2D). In patients with NMOSD, sNfL levels were higher than they were in HCs during both relapse and remission (relapse: βmult = 2.03, p < 0.001; remission: βmult = 1.99, p < 0.001). Moreover, sNfL levels in the late stage of remission in NMOSD (25.0 [20.1–43.9] pg/mL) were significantly higher than those in HCs and patients with MS (HCs: βmult = 1.43, p = 0.001; MS: 15.5 [13.1–18.8] pg/mL, βmult = 1.43, p = 0.034). In patients with MS, sNfL levels were significantly higher in those with RRMS and SPMS/PPMS than in HCs (RRMS: 24.6 [16.3–39.1] pg/mL, βmult = 1.96, p < 0.001; SPMS/PPMS: 28.6 [24.8–42.1] pg/mL, βmult = 1.68, p < 0.001). Again, we found no difference between RRMS and SPMS/PPMS (p = 0.248). In patients with RRMS, sNfL levels were significantly higher during both relapse and remission phases than they were in HCs (relapse: βmult = 2.49, p < 0.001; remission: βmult = 1.42, p = 0.003), while levels during relapse were higher than those during remission (βmult = 1.81, p < 0.001).
Association of sGFAP levels with clinical parameters
Patients with NMOSD
Using a univariate model, we found that sGFAP levels were positively associated with EDSS score (p = 0.002; figure 1C and table 2) and recent relapse (p < 0.001; table 2) in patients with NMOSD. Higher EDSS score (p = 0.026) and recent relapse (p < 0.001) were confirmed to be independently associated with sGFAP levels via a multivariate model (table 2). We assessed whether sGFAP levels were influenced by different relapse phenotypes. Analysis showed that spinal cord relapses were independently associated with higher sGFAP levels (p < 0.001), while other relapse locations were not (table e-4 available from Dryad, doi.org/10.5061/dryad.dp75314). Immunosuppressive treatment was not an independent factor that influenced the levels of sGFAP (table 2) even if we analyzed the effect of corticosteroid and immunosuppressant separately or if analysis was limited to the remission phase.
Univariate and multivariate models testing the associations between sGFAP levels and clinical parameters in patients with NMOSD
Patients with MS
Similarly, in patients with MS, univariate analyses revealed positive associations between sGFAP and the EDSS score (p = 0.003; figure 1C and table 3) and presence of recent relapse (p = 0.012; table 3). Multivariate analysis showed that EDSS score predicted sGFAP levels in MS (p = 0.049; table 3). The effect of treatment was not detected in MS or RRMS, even if corticosteroids were mentioned in addition to disease-modifying drugs (DMDs).
Univariate and multivariate models testing the associations between sGFAP levels and clinical and MRI variables in patients with MS
Associations of sNfL levels with clinical parameters
Patients with NMOSD
A univariate model revealed that sNfL levels were positively associated with age (p = 0.002; table 4), as seen in HCs, and with EDSS score (p < 0.001; figure 1D and table 4). The multivariate model revealed that age (p = 0.031) and EDSS score (p < 0.001) were independently associated with sNfL (table 4). Relapse site was not associated with sNfL level (data not shown). Immunosuppressive treatment was not an independent factor that influenced sNfL levels (table 4).
Univariate and multivariate models testing the associations between sNfL levels and clinical parameters in patients with NMOSD
Patients with MS
Univariate analysis indicated that in patients with MS, sNfL was higher in those with higher EDSS scores (p = 0.003; figure 1D and table 5) and was associated with recent relapse (p < 0.001; table 5). Multivariate analysis showed that sNfL was positively associated with EDSS score (p = 0.011), recent relapses (p = 0.007), and CNS T2 lesions (p = 0.034). We found that treatment with DMDs was also an independent factor (p = 0.041; table 5), with DMD-treated patients having lower sNfL levels than those who were untreated. When we included treatment with interferon-β or fingolimod instead of DMDs as a factor in the multivariate model, we found that for fingolimod, sNfL levels tended to be lower than they were for no treatment (βmult = 0.67 [IQR 0.45–1.00], p = 0.052), but the same was not true for interferon-β (βmult = 0.89 [0.60–1.32], p = 0.566). No effect of corticosteroids was identified (data not shown).
Univariate and multivariate models testing the associations between sNfL levels and clinical and MRI variables in patients with MS
Difference in sGFAP/sNfL quotient between NMOSD and MS
We evaluated whether the sGFAP/sNfL quotient can differentiate NMOSD from MS. Univariate analysis indicated that the quotient was significantly higher in NMOSD than in MS (p = 0.004; figure 2E). Moreover, multivariate analysis showed that the presence of recent relapse (p = 0.009) and the diagnosis (p = 0.007) were both independent factors for predicting the sGFAP/sNfL quotient (table e-5 available from Dryad, doi.org/10.5061/dryad.dp75314). While the quotient during remission for both NMOSD and MS was comparable to that for HCs, in relapse, it increased in patients with NMOSD and decreased in those with MS (figure 2F). ROC curve analysis (area under the ROC curve 0.781) showed that the sGFAP/sNfL quotient during relapse was useful in differentiating NMOSD from MS (figure e-1 available from Dryad). At a cutoff value of 5.71, the sensitivity and specificity were 73.0% and 75.8%, respectively. When this cutoff value was applied to the samples collected within 3 years from disease onset, the sensitivity and specificity were 66.7% (6 of 9) and 83.3% (5 of 6), respectively. In addition, the sGFAP/sNfL quotients in 2 patients with NMOSD without AQP4-IgG, which is often difficult to differentiate from MS, were both >5.71 (11.56 and 6.76, respectively).
Changes in sGFAP and sNfL levels after relapse
sNfL levels significantly decreased over time after relapse in patients with NMOSD and MS (univariate p = 0.003, multivariate p = 0.002; table e-6 available from Dryad, doi.org/10.5061/dryad.dp75314). Comparing the changes in sGFAP and sNfL levels after relapse showed that the per-month decrease in sGFAP levels was similar between NMOSD and MS (univariate analysis p = 0.946, multivariate analysis p = 0.664; figure 1E). The sNfL levels decreased significantly more slowly in NMOSD than in MS (univariate: p < 0.001; multivariate: p = 0.010; figure 1F and table e-6 available from Dryad). In NMOSD, in samples from remission, sGFAP levels were not associated with the intervals from recent relapses (univariate p = 0.179, multivariate p = 0.141) or the maximum EDSS scores during recent relapses (univariate p = 0.763, multivariate p = 0.654; table e-7 available from Dryad). However, sNfL levels were negatively associated with the intervals from recent relapses (univariate p = 0.038, multivariate p = 0.034) but not with the maximum EDSS score during recent relapses in multivariate models (univariate p = 0.018, multivariate p = 0.128; table e-7 available from Dryad). Conversely, in RRMS, sGFAP levels were associated with maximum EDSS scores during recent relapse (univariate p = 0.002, multivariate p = 0.009) but not with intervals from recent relapses (univariate p = 0.276, multivariate p = 0.450; table e-8 available from Dryad), while sNfL levels were associated with both maximum EDSS scores during recent relapses (univariate p = 0.002, multivariate p = 0.004) and the intervals from recent relapses (p < 0.001 in both univariate and multivariate models; table e-8 available from Dryad).
Relationship between sGFAP and sNfL levels and subsequent relapse
Finally, we investigated whether sGFAP and sNfL levels in the remission phase predicted subsequent relapse activity in patients with NMOSD and RRMS. When sGFAP values were above the 99th percentile of HCs (337.0 pg/mL), the levels were regarded as high on the basis of previous studies.20,22 Of 65 NMOSD samples in remission phase, high sGFAP levels were found in 9 samples. Of 32 RRMS samples, none were higher than the limit of sGFAP. In NMOSD, age at sampling was significantly older in the high level group than the normal level group (58.8 vs 49.5 years, p = 0.033; table e-9 available from Dryad, doi.org/10.5061/dryad.dp75314), and future relapses tended to be more frequently observed within 1 year in the high-level group compared with the normal-level group (57.1% vs 23.1%, p = 0.078) despite their similar treatment status. However, when multivariate models were analyzed, neither sGFAP nor sNfL was associated with next-year relapses in patients with NMOSD (table e-10 available from Dryad). In RRMS, univariate and multivariate analyses indicated that sNfL levels were associated with future relapses (p < 0.001 and p = 0.006, respectively; table e-11 available from Dryad).
Discussion
Here we have shown that for both GFAP and NfL, CSF and serum levels were strongly correlated. In addition, in cases of NMOSD, sGFAP and sNfL were associated with EDSS score, while sGFAP also predicted the presence of recent relapses. In cases of MS, sNfL and sGFAP were associated with EDSS score, while sNfL was also associated with the presence of recent relapse, number of CNS T2 lesions, DMD use, and future relapses. The reduction of sNfL levels after relapse was slower in NMOSD than in RRMS. The sGFAP/sNfL ratio at relapse was significantly greater in NMOSD than in MS.
The strong positive correlation of both sGFAP and sNfL levels with CSF levels suggests that these sensitive measurements will be useful for developing blood-based biomarkers that reflect CNS pathology. Although numerous studies have demonstrated correlations between CSF and serum NfL,20,26,–,29 this is the first study to show the correlation in patients with NMOSD and that the extent of the correlations was similar between NMOSD and MS. Recent mouse studies have shown that NfL in blood is derived from the CNS and reflects CNS pathology,15,30 although peripheral nerve damage may increase sNfL.31,–,33 GFAP has also been reported to be correlated with CSF and serum levels in MS34 and to migrate from the CNS to blood via the glymphatic system.35 Similarly to MS, we also revealed, for the first time, the correlation between CSF and serum levels of GFAP in NMOSD. Taken together, GFAP and NfL levels in blood appear to reflect their respective CSF levels and the pathologic changes to the CNS that occur in diseases such as MS and NMOSD.
Elevated sGFAP during NMOSD relapse is consistent with the primary astrocytopathy that is mediated by AQP4-IgG36 and with increased CSF-GFAP levels in acute relapses.3,–,6 A previous study used ELISA to measure sGFAP levels in patients with neuromyelitis optica but did not detect any differences between relapse and remission phases.37 The limited number of patients in remission and the sensitivity of the assay may have contributed to this finding. Our finding that all GFAP levels, including those determined during remission, were well above the detection threshold of the Simoa assay illustrates the relative benefit of the excellent sensitivity of Simoa technology. The strong associations of sGFAP levels with both EDSS score and the occurrence of recent relapse indicate that sGFAP levels are biomarkers of disability and disease activity in NMOSD, independently of patient age or sex. sGFAP was associated with recent spinal cord relapses but not with clinical disease activity originating from other locations. This may be related to relatively more extensive disruption of spinal cord gray matter astrocytes, where AQP4 is abundantly expressed.36 In MS, sGFAP was significantly associated with EDSS score, which is consistent with previous studies analyzing CSF samples7,–,11 and with a recent study that used the Simoa assay to analyze serum samples.34
The median sGFAP levels during NMOSD remission were lower than those during relapse but still higher than those of HCs even in the late stage of remission. High sGFAP levels in the earlier stages of remission in NMOSD may well reflect the acute astrocytic damage during relapse. However, in traumatic brain injury, elevated sGFAP levels normalize within 6 days after injury.38 Of note, not all patients with NMOSD had high sGFAP levels in remission, but the high sGFAP levels observed in some patients with NMOSD during remission may indicate subclinical disease activity. This is also supported by the observation that sGFAP levels in remission were not significantly associated with intervals from recent relapses and maximum EDSS scores (as a proxy for relapse severity) during acute relapses. Unfortunately, treatment effects on sGFAP levels were not observed in our study. One of the possible reasons is that all patients with NMOSD were treated with corticosteroid and/or conventional oral immunosuppressants but not with molecular-targeted drugs (i.e., anti–interleukin-6 receptor, anti-C5, and anti-CD20 antibodies) because these drugs have not been approved in Japan. Therefore, future studies to assess the effects of these molecular-targeted drugs on sGFAP levels and to identify the underlying mechanisms of sustained high sGFAP levels during remission phase in patients with NMOSD are needed.
Although a study using an electrochemiluminescence assay has reported increased sNfL in patients with NMOSD,39 its association with clinical parameters was unknown until now. In our study, sNfL levels in NMOSD were higher in patients with higher EDSS scores and older age, and sNfL levels after relapse remained higher for a longer time in patients with NMOSD than in those with MS. These findings, along with the fact that sNfL levels in remission were higher in NMOSD than in MS, suggest that neuro-axonal damage might be generally more pronounced in NMOSD than in MS. Wallerian degeneration, retrograde axonal degeneration, and loss of trophic support from astrocytes damaged by AQP4-IgG40 are all mechanisms that might underlie neuro-axonal damage in NMOSD. Such neuro-axonal damage may be more likely to occur in older people with NMOSD.
Our study of Japanese patients with MS has confirmed 2 findings that were reported in European patients with MS: increased sNfL levels in MS, especially during relapse, and the association of sNfL with EDSS score.20,22,27,41,42 Therefore, sNfL appears to be a biomarker of disease activity and disability in MS across different races. Moreover, association of sNfL levels with the number of T2 lesions, which was also observed in Europeans,20,41 was true for Japanese patients with MS. In addition, sNfL levels in people being treated with DMDs were lower than those in untreated patients, which is consistent with previous studies.20,27,42,43 Although the efficacy of fingolimod in decreasing sNfL levels was only marginally significant in our study, likely because of the small sample size, recent longitudinal studies in European patients showed that fingolimod reduces sNfL levels,42,44 suggesting that sNfL levels could reflect responses to treatment. We also found that sNfL during remission was an independent factor predicting future relapse in RRMS. Although our study was not primarily designed to investigate prognosis, sNfL may be a potential biomarker for future clinical disease activity in Japanese patients with MS, as previously demonstrated in European patients.20,22
The difference in sGFAP/sNfL quotient during relapses between NMOSD and MS might reflect distinct disease mechanisms. Unlike NMOSD, in which astrocytopathy is the primary pathology followed by neuro-axonal damage, our study suggests that neuro-axonal damage precedes or dominates astrocytic damage in the active disease phase of MS. In MS pathophysiology, astrocytes are thought to sustain neuro-axonal damage in multiple steps during the inflammatory process.45,46 The sGFAP/sNfL quotient within 2 months of relapse may aid in discriminating NMOSD from MS, especially early in the course of the disease when making clinical and MRI-based diagnoses is frequently difficult.
A limitation of our study is its retrospective design with limited clinical follow-up information. Therefore, as a next step, we need to conduct a well-organized longitudinal prospective study to confirm our findings. However, we believe our results provide the foundation for future longitudinal studies that need to enroll more patients to establish the prognostic value and disease-activity monitoring potential of sGFAP and sNfL in NMOSD. We show that serial, minimally invasive sGFAP and sNfL measurement by sensitive Simoa technology is a feasible and promising means to further characterize and monitor disease activity in patients with NMOSD.
Study funding
Supported by grants from the Japan Society for the Promotion of Science (JSPS) KAKENHI (grants 16H02657, 16K09694, 17K16124, and 18K07529); a Health and Labour Sciences Research Grant on Intractable Diseases [H29-Nanchitou (Nan)-Ippan-043] from the Ministry of Health, Labour, and Welfare, Japan; and the Swiss National Research Foundation (320030_160221).
Disclosure
M. Watanabe received a grant from JSPS KAKENHI (grant 17K16124). Y. Nakamura received a grant and salary from Mitsubishi Tanabe Pharma, Bayer Yakuhin, Ltd, and Japan Blood Products Organization. Z. Michalak reports no disclosures relevant to the manuscript. N. Isobe is supported by a grant from JSPS KAKENHI (grant 18K07529) and received grant support from Mitsubishi Tanabe Pharma, Osoegawa Neurology Clinic, Bayer Yakuhin, Ltd, and Japan Blood Products Organization. C. Barro received travel support from Teva and Novartis. D. Leppert is an employee of Novartis Pharma AG. T. Matsushita received speaker honoraria payments from Mitsubishi Tanabe Pharma, Takeda Pharmaceutical Co, and Biogen Japan. F. Hayashi reports no disclosures relevant to the manuscript. R. Yamasaki received a grant from JSPS KAKENHI (grant 16K09694). J. Kuhle received speaker fees, research support, and travel support from and/or served on advisory boards for European Committee for Treatment and Research in MS (ECTRIMS), Swiss MS Society, Swiss National Research Foundation (320030_160221), University of Basel, Bayer, Biogen, Genzyme, Merck, Novartis, and Teva. J. Kira is supported by grants from JSPS KAKENHI (grant 16H02657) and Health and Labour Sciences Research Grants on Intractable Diseases [H29-Nanchitou (Nan)-Ippan-043] and received consultant fees, speaking fees and/or honoraria from Novartis Pharma, Mitsubishi Tanabe Pharma, Boehringer Ingelheim, Teijin Pharma, Takeda Pharmaceutical Co, Otsuka Pharmaceutical, Astellas Pharma, Pfizer Japan, and Eisai. Go to Neurology.org/N for full disclosures.
Acknowledgment
The authors thank Junji Kishimoto (Center for Clinical and Translational Research, National University Corporation Kyushu University, Fukuoka, Japan) for helpful comments on statistical analyses. They also thank Rachel James, PhD, and Adam Phillips, PhD, from the Edanz Group (edanzediting.com/ac) for editing a draft of this manuscript.
Appendix Authors
Footnotes
Go to Neurology.org/N for full disclosures. Funding information and disclosures deemed relevant by the authors, if any, are provided at the end of the article.
↵* These authors contributed equally to this work.
- Received December 8, 2018.
- Accepted in final form May 2, 2019.
- © 2019 American Academy of Neurology
References
Letters: Rapid online correspondence
- Reader response: Serum GFAP and neurofilament light as biomarkers of disease activity and disability in NMOSD
REQUIREMENTS
If you are uploading a letter concerning an article:
You must have updated your disclosures within six months: http://submit.neurology.org
Your co-authors must send a completed Publishing Agreement Form to Neurology Staff (not necessary for the lead/corresponding author as the form below will suffice) before you upload your comment.
If you are responding to a comment that was written about an article you originally authored:
You (and co-authors) do not need to fill out forms or check disclosures as author forms are still valid
and apply to letter.
Submission specifications:
- Submissions must be < 200 words with < 5 references. Reference 1 must be the article on which you are commenting.
- Submissions should not have more than 5 authors. (Exception: original author replies can include all original authors of the article)
- Submit only on articles published within 6 months of issue date.
- Do not be redundant. Read any comments already posted on the article prior to submission.
- Submitted comments are subject to editing and editor review prior to posting.
You May Also be Interested in
Hemiplegic Migraine Associated With PRRT2 Variations A Clinical and Genetic Study
Dr. Robert Shapiro and Dr. Amynah Pradhan