Многогранные биомаркеры предлагают аналогичный профиль патологии CNS в релапсации и прогрессивном MS.

Abbreviations

Alzheimer's Disease

aquaporin 4

B‐cell maturation antigen

chitinase‐3‐like‐1

central nervous system

cerebrospinal fluid

Expanded Disability Status Scale

Estimated Expanded Disability Status Scale

formalin‐fixed paraffin embedded

gadolinium

glial fibrillary acidic protein

horseradish peroxidase

immune‐modulating treatment

interquartile range

medulla oblongata

myelin oligodendrocyte glycoprotein–associated disease

multiple sclerosis

normal‐appearing white matter

neurofilament light chain

neuromyelitis optica spectrum disorder

osteopontin

principal component

principal component analysis

progressive multiple sclerosis

primary progressive

primary progressive multiple sclerosis

perivascular space

people diagnosed with

people diagnosed with multiple sclerosis

people diagnosed with primary progressive multiple sclerosis

people diagnosed with relapsing–remitting multiple sclerosis

relapsing multiple sclerosis

relapsing–remitting

relapsing–remitting multiple sclerosis

soluble CD27

standard deviation

single‐molecule array

synaptosomal‐associated protein‐25

secondary progressive multiple sclerosis

Introduction

Multiple sclerosis (MS) is an immune‐mediated demyelinating disease of the central nervous system (CNS) with distinct clinical phenotypes: relapsing MS (RMS)—including relapsing–remitting MS (RRMS) and relapse‐onset MS followed by gradual worsening: secondary progressive MS (SPMS)—and primary progressive MS (PPMS) [ 1 ]. Recent views regard MS as a continuum, with similar disease mechanisms acting differently across the lifespan [ 2 , 3 ].

At one end of this continuum, MS presents with acute inflammatory activity characterized by relapses and/or new gadolinium (GAD)‐enhancing MRI lesions. This acute inflammation involves increased blood–brain barrier (BBB) permeability, circulating lymphocyte invasion, demyelination, and axonal damage [ 2 , 3 ], most prominent in early RRMS but continuing less frequently in progressive MS [ 4 ].

On the other end, PPMS features chronic smoldering inflammation compartmentalized behind a closed BBB, causing demyelination and neurodegeneration [ 2 ]. This involves persistently activated microglia at lesion borders [ 5 ], resident lymphocytes [ 6 ], and cortical demyelination with meningeal lymphocyte infiltrates [ 7 ]. Clinically, these processes likely contribute to disability progression independent of relapse activity (PIRA) and affect MS from its earliest stages [ 8 ]. Additionally, neurodegenerative processes—including age‐related neurodegeneration, iron‐related neurotoxicity, mitochondrial dysfunction, and diffuse injury—contribute to disease progression [ 2 ].

This involvement of acute inflammatory attacks and chronic smoldering disease [ 9 ] distinguishes MS from other inflammatory demyelinating disorders like aquaporin 4 (AQP4)‐related neuromyelitis optica spectrum disorder (NMOSD) and myelin oligodendrocyte glycoprotein–associated disease (MOGAD), which primarily feature inflammatory attacks without progressive components [ 10 , 11 ]. Conversely, progressive neurodegenerative diseases like Alzheimer's disease (AD) exhibit inflammatory components without acute attacks or chronic intrathecal inflammation [ 12 ].

Recent research has identified various biomarkers in MS correlating with distinct disease mechanisms [ 13 ], allowing for their identification and quantification in pwMS [ 14 ]. If RRMS and PPMS are indeed different disease entities, we hypothesize that differences in disease mechanisms between RRMS and PPMS are reflected in their biomarker profiles. To test this, we selected biomarkers for distinct pathological components: B‐cell maturation antigen (BCMA) and soluble CD27 (sCD27) to represent the adaptive immune response, with BCMA marking B‐cell activation and sCD27 indicating both B‐ and T‐cell activation [ 15 , 16 , 17 , 18 ], osteopontin (OPN) reflecting innate immunity [ 19 ], and chitinase‐3‐like‐1 (CHI3L1) indicating chronic inflammation and glial activation within the CNS [ 20 ]. Biomarkers representing tissue damage included synaptosomal‐associated protein‐25 (SNAP‐25) for synapse decay and gray matter neurodegeneration [ 21 ], glial fibrillary acidic protein (GFAP) for astrocytic pathology often associated with reactive gliosis [ 22 ], and neurofilament light chain (NfL) [ 23 ] for neuroaxonal damage.

This study leverages fluid biomarkers and brain tissue pathology to investigate biologically relevant components in RRMS and PPMS and to define their profile in comparison to other CNS diseases: NMOSD, MOGAD, and AD.

Method

Subjects

Serum and CSF samples of people diagnosed with (pw) PPMS, RRMS, AD, NMOSD, and MOGAD were collected from Erasmus Medical Center biobanks. Samples from pwPPMS were collected first. Then, samples from pwRRMS were collected, ensuring no current immune‐modulating treatment (IMT) and similar sex and age distribution to the PPMS cohort. Samples of pwNMOSD, pwMOGAD, and pwAD were also collected.

Diagnoses of PPMS, RRMS, AD, NMOSD, and MOGAD were made by board‐certified neurologists utilizing the most recent diagnostic criteria [ 10 , 24 , 25 , 26 ]. Permission for these biobanks was granted by the Erasmus Medical Center medical ethics committee (MEC‐2016‐069 and MEC‐2014‐033).

Data on sex, date of birth, disease duration, treatment, relapses, MRI activity, results of CSF analysis, and Expanded Disability Status Scale (EDSS) within 1 year of sampling were extracted from medical records. If EDSS was not recorded, an EDSS was estimated (eEDSS) based on patient history and neurological examination, following Kurtzke's guidelines [ 27 ]. Disease duration was defined as the time since first symptoms. Recent disease activity was defined as clinical relapse and/or radiological disease activity (GAD‐enhancing lesions on MRI or new lesions compared to any previous scan) in the 90 days before the sample. Expanding T2 lesions were not taken into consideration due to heterogeneity in scanning protocols and uncertain interpretation as inflammatory events in the context of recent literature on slowly expanding lesions and paramagnetic rim lesions [ 28 ]. PwMS without clinical or radiological activity, or without available radiological data, were considered non‐active MS.

Post‐mortem analysis of PPMS and RMS (RRMS or SPMS) was performed on brain donors from the Netherlands Brain Bank, selected based on the availability of B‐ and T‐cell staining and brainstem and subcortical normal‐appearing white matter (NAWM) [ 6 , 29 ]. Permission for the donor program was granted by the ethics committee of Free University Medical Center (2009/148). The disease course was recorded as registered by the treating neurologist [ 5 ].

Biomarker Selection and Measurements

Serum and CSF samples were frozen at −80°C, stored in the local Erasmus Medical Center biobank, and thereafter shipped on dry ice to the Neuroimmunology Department, University of California, San Francisco, for analysis.

Biomarker concentrations were determined using single molecular array (Simoa) or electrochemiluminescence immunoassay. Samples were measured in duplicates per manufacturers' instructions. Samples with an intrarun coefficient of variation > 20% were excluded. Measurements were done by a blinded lab assistant. Full details are in Supplemental Methods.

Post‐Mortem Pathology Analyses

B‐ and T‐lymphocyte presence (CD20, CD138, CD3, respectively) was assessed in standard dissected tissue blocks from the medulla oblongata (MO) and subcortical white matter lesions. Immunohistochemistry of lymphocytes was performed as previously described [ 6 , 29 ]. Axonal density was assessed by Bielschowsky silver staining in the pyramidal tract at the MO as previously described [ 6 ]. B‐cell presence was scored as CD20 + /CD138 + cells present/absent. The amount of T cells was quantified by the amount of CD3 + cells/μm 2 . Axonal density was quantified using a grid count. Full details are in Supporting Methods .

Statistical Analyses

Baseline characteristics between diagnoses were compared using the Kruskal–Wallis test for continuous outcomes and the chi‐squared test for frequency outcomes. Correlations were calculated with Spearman's rank correlation coefficient. If necessary, biomarker outcomes underwent log transformation to establish normal distribution. ANOVAs with age and sex as covariates were performed to test for differences between diagnoses, using Bonferroni's method to adjust for multiple testing. Linear regression including covariates age and sex was performed to test correlations between MS disease severity (EDSS in quartile groups) and biomarkers. Scaled (based on correlation matrix) principal component analyses (PCAs) were performed to test if different biomarkers resulted in clusters that matched the different diagnoses. K‐means clustering analysis was applied to quantify and visualize the clustering patterns identified by PCA. Differences in B‐cell presence were tested using the chi‐squared test, T‐cell amounts using the Wilcoxon test (MO), or negative binomial mixed model (lesions), and axonal density using the Wilcoxon test.

Analyses were performed using IBM SPSS Statistics V.28, GraphPad Prism V.10.1.1, and RStudio V.4.4.1 (packages are described in Supplemental Methods).

Results

Participant Characteristics

Serum and CSF were collected from pwPPMS ( n  = 104), pwRRMS ( n  = 38), and pwAD ( n  = 22) and sera from pwNMOSD ( n  = 10) and pwMOGAD ( n  = 10). From pwPPMS, 97 serum and 29 CSF samples were collected, with 22 “paired” (serum and CSF of the same patient); from pwRRMS, 30 serum and 30 CSF samples (22 paired), matched in sex and age as closely as possible, all not on IMT, were collected. For pwAD, 20 serum and 20 CSF samples were collected, with 18 paired. Ten serum samples each were collected from pwNMOSD and pwMOGAD. Samples were drawn between October 2002 and February 2022. Baseline comparisons revealed diagnosis‐related differences: pwAD were older, pwPPMS had longer disease duration, less clinical and/or radiological disease activity and higher (e)EDSS than pwRRMS, and pwNMOSD and pwMOGAD were more frequently treated with corticosteroids (Table  1 ). In the NMOSD group, two individuals received rituximab more than 2 years before sampling. In the PPMS group, one individual underwent autologous hematopoietic stem cell transplantation 7 years prior to sampling, and another participant received mitoxantrone, with the last infusion administered 10 months before sampling. No other long‐acting or cell‐depleting IMTs were administered to participants. Among 127 serum samples from pwMS, 6 lacked MRI scans and 38 had uncertain radiological activity (any GAD‐enhancing T1 lesions or new T2 lesions compared to any previous scan) due to missing scans and/or GAD administration. Both groups were classified as radiologically non‐active MS. Of 83 participants with available MRI scans before serum sample collection, 43% ( n  = 36) had a scan within 90 days before sampling. Similarly, of 59 CSF samples from pwMS, 1 patient had no recent MRI and 9 had uncertain radiological activity. Of the remaining 49 participants with available MRI scans before CSF sample, in n =31 (63%), these were acquired within 90 days of sampling.

TABLE 1 Baseline characteristics. NMOSD ( n  = 10) MOGAD ( n  = 10) RRMS ( n  = 38) PPMS ( n  = 104) AD ( n  = 22)
p

p (MS) Age at serum (years. mdn. IQR)
46.50
(35.1–57.8)

36.1
(28.0–52.0)
41.5 (38.4–49.7) ( n  = 30) 52.5 (43.6–59.0) ( n  = 97)
65.5 (61.3–67.4)
( n  = 20)

p  < 0.001
p  < 0.001
Age at CSF
(years. mdn. IQR)
39.7 (35.7–47.6) ( n  = 30) 49.3 (43.9–55.6) ( n  = 29)
64.0 (59.3–67.5)
( n  = 20)

p  < 0.001
p  < 0.001 Sex ( n . % F) 8 (80) 5 (50) 27 (71.1) 54 (51.9) 16 (72.7)
p  = 0.081
p  = 0.041 (e)EDSS at serum sample
a
(mdn. IQR)
3.5
(2.25–4.625)

1.5
(1.0–2.25)

2.0 (1.0–2.5)
( n  = 30)

6.0 (3.5–6.5)
( n  = 97)

p  < 0.001
p  < 0.001 (e)EDSS at CSF sample
a
(mdn. IQR)
2.0 (1.0–2.0)
( n  = 30)

3.5 (3.0–5.0)
( n  = 29)

p  < 0.001 = Disease duration at serum (years. mdn. IQR)
1.0
(0.3–9.7)

2.9
(0.3–14.9)

1.2 (0.2–6.0)
( n  = 30)

6.3 (3.5–15.4)
( n  = 97)

2.4 (1.7–6.6)
( n  = 20)

p  < 0.001
p  < 0.001
Disease duration at CSF
(years. mdn. IQR)

0.8 (0.2–3.0)
( n  = 30)

4.3 (2.8–6.5)
( n  = 29)

3.1 (1.7–6.5)
( n  = 20)

p  < 0.001
p  < 0.001 Any disease activity in 90 days before serum sample
b
( n , % active) 2 (20%) 2 (20%)
12 (40%)
( n  = 30)

9 (9.3%)
( n  = 97)

p  = 0.009
p  < 0.001 Any disease activity in 90 days before CSF sample
b
( n , % active)
15 (50%)
( n  = 30)

3 (10.3%)
( n  = 29)

p  < 0.001 = Any IMT or CS within 4w of serum sample
a
( n , % treated)
2 (20%) CS;
No IMT

3 (30%) CS;
No IMT

1 CS (3.3%);
2 IMT (6.6%)
( n  = 30)

2 CS (2.1%);
4 IMT (4.2%)
( n  = 97)

p  = 0.060
p  = 0.477
Any IMT or CS within 4w of CSF sample
a


( n , % treated)

1 CS (3.3%);
No IMT
( n  = 30)

1 CS (3.4%);
No IMT
( n  = 29)

p  = 0.981 =
Note: Disease duration: from first symptoms. p value: by Kruskal–Wallis test (multiple groups) or Mann–Whitney U (two MS groups) or chi‐squared test. Abbreviations: (e)EDSS, (estimated) Expanded Disability Status Scale; AD, Alzheimer's disease; CS, corticosteroids; CSF, cerebrospinal fluid; IMT, immunomodulatory treatment (glatiramer acetate or interferon bèta); IQR, interquartile range; mdn, median; MOGAD, myelin oligodendrocyte glycoprotein associated disorder; NMOSD, neuromyelitis optica spectrum disorder; PPMS, primary progressive multiple sclerosis; RRMS, relapsing‐remitting multiple sclerosis.
a
pwAD: removed from comparison.
b
Disease activity for pwMS: relapses, gadolinium enhancing lesions on MRI, or new T2 lesions on MRI. For pwMOGAD and pwNMOSD: only relapses. pwAD: removed from comparison.

Baseline characteristics.

46.50

(35.1–57.8)

36.1

(28.0–52.0)

65.5 (61.3–67.4)

( n  = 20)

Age at CSF

(years. mdn. IQR)

64.0 (59.3–67.5)

( n  = 20)

3.5

(2.25–4.625)

1.5

(1.0–2.25)

2.0 (1.0–2.5)

( n  = 30)

6.0 (3.5–6.5)

( n  = 97)

2.0 (1.0–2.0)

( n  = 30)

3.5 (3.0–5.0)

( n  = 29)

1.0

(0.3–9.7)

2.9

(0.3–14.9)

1.2 (0.2–6.0)

( n  = 30)

6.3 (3.5–15.4)

( n  = 97)

2.4 (1.7–6.6)

( n  = 20)

Disease duration at CSF

(years. mdn. IQR)

0.8 (0.2–3.0)

( n  = 30)

4.3 (2.8–6.5)

( n  = 29)

3.1 (1.7–6.5)

( n  = 20)

12 (40%)

( n  = 30)

9 (9.3%)

( n  = 97)

15 (50%)

( n  = 30)

3 (10.3%)

( n  = 29)

2 (20%) CS;

No IMT

3 (30%) CS;

No IMT

1 CS (3.3%);

2 IMT (6.6%)

( n  = 30)

2 CS (2.1%);

4 IMT (4.2%)

( n  = 97)

Any IMT or CS within 4w of CSF sample
a

( n , % treated)

1 CS (3.3%);

No IMT

( n  = 30)

1 CS (3.4%);

No IMT

( n  = 29)

Note: Disease duration: from first symptoms. p value: by Kruskal–Wallis test (multiple groups) or Mann–Whitney U (two MS groups) or chi‐squared test.

Abbreviations: (e)EDSS, (estimated) Expanded Disability Status Scale; AD, Alzheimer's disease; CS, corticosteroids; CSF, cerebrospinal fluid; IMT, immunomodulatory treatment (glatiramer acetate or interferon bèta); IQR, interquartile range; mdn, median; MOGAD, myelin oligodendrocyte glycoprotein associated disorder; NMOSD, neuromyelitis optica spectrum disorder; PPMS, primary progressive multiple sclerosis; RRMS, relapsing‐remitting multiple sclerosis.

pwAD: removed from comparison.

Disease activity for pwMS: relapses, gadolinium enhancing lesions on MRI, or new T2 lesions on MRI. For pwMOGAD and pwNMOSD: only relapses. pwAD: removed from comparison.

Distribution of Serum and CSF Concentrations of Soluble Biomarkers

Investigated biomarkers are shown in Figure  1A . For the markers representing CNS tissue damage/gliosis, positive correlations were found between serum and CSF concentrations of GFAP and NfL (Figure  1B,C , other markers in Figure  S1 ). Markers representing CNS tissue damage/gliosis (GFAP, NfL, and SNAP25) positively correlated with age, except for CSF NfL (Figure  S2 ). Innate inflammatory markers CHI3L1 (serum and CSF) and OPN (CSF) also positively correlated with age. Adaptive immune markers BCMA and sCD27 in CSF negatively correlated with age (Figure  S2 ). Disease‐associated outliers included a 36‐year‐old male with tumefactive RRMS (very high serum NfL) and a 63‐year‐old male with a severe NMOSD attack (very high serum sCD27, OPN, GFAP, and NfL). No differences were found in any biomarkers between sexes (Figure  S3 ).

Investigated biomarkers and correlation between serum and CSF concentrations of GFAP and NfL. (A) Visualization of the investigated immune activation and tissue damage biomarkers. Figure created with Biorender.com . (B) Correlation between concentrations of GFAP in serum ( y ‐axis) and CSF ( x ‐axis) in pg/mL. (C) Correlation between concentrations of NfL in serum ( y ‐axis) and CSF ( x ‐axis) in pg/ml. Data of pwRRMS ( n  = 22 paired samples) are denoted by green dots, data of pwPPMS ( n  = 22 paired samples) are denoted by blue dots ( n  = 22) and data of pwAD ( n  = 18 paired samples) are denoted by purple dots. Statistical significance was tested using Spearman's rank correlation coefficient. p values of < 0.05 were considered statistically significant.

Individual Serum and CSF Soluble Biomarkers Differ Between MS and Other Disease Groups, but not Between RRMS and PPMS

Single serum and CSF biomarker comparisons revealed no differences between pwPPMS and pwRRMS (Figure  2A–M ). However, biomarker concentrations differed between pwAD and pwMS: pwAD had higher serum CHI3L1 levels than pwPPMS ( p  = 0.018, Figure  2L ), lower CSF sCD27 levels ( p  < 0.001 compared to pwPPMS and pwRRMS, Figure  2F ), higher CSF OPN levels ( p  = 0.011 compared to pwPPMS and p  = 0.010 compared to pwRRMS, Figure  2J ), higher serum GFAP levels ( p  = 0.020 compared to both pwPPMS and pwRRMS, Figure  2C ), and higher CSF SNAP25 levels than pwPPMS (p < 0.001, Figure  2K ). In serum, pwNMOSD had higher OPN levels than pwRRMS ( p  = 0.013, Figure  2I ). Thus, these single biomarkers reflect some differences between MS and non‐MS diagnoses, but not between RRMS and PPMS.

Concentrations of individual immune activation and tissue damage/gliosis biomarkers in serum and CSF in the different diagnosis groups and principal component analyses of all serum and all CSF biomarkers. The left axis of these panels show log‐transformed concentrations of (A) serum and (B) CSF BCMA, (C) serum and (D) CSF GFAP, (E) serum and (F) CSF sCD27, (G) serum and (H) CSF NfL, (I) serum and (J) CSF OPN (left y‐axis), (K) CSF SNAP25, (L) serum and (M) CSF CHI3L1, stratified for the diagnostic groups. The right axis shows the corresponding estimated marginal means ANOVA with 95% confidence margins, with age and sex as co‐variates. Bonferroni's method to adjust for multiple testing. Statistically significant differences were indicated with an asterisk (*). (N) PCA of all serum biomarkers in pwRRMS, pwPPMS, and pwAD, (O) the corresponding scree plot and (R) the corresponding Eigenvalues. (P) PCA of all CSF biomarkers in pwRRMS, pwPPMS, and pwAD, (Q) the corresponding scree plot and (S) the corresponding Eigenvalues.

Serum and CSF Biomarker Profiles Differ Between MS and AD, but not Between RRMS and PPMS

Since single biomarkers did not differentiate between RR and PPMS diagnoses, we investigated the distribution of composite biomarker profiles. Due to small numbers and high variance, pwNMOSD and pwMOGAD were excluded. PCA of serum biomarkers (explaining ~50% variance) and CSF biomarkers (explaining ~60% variance) showed almost complete overlap between pwRRMS and pwPPMS, with less overlap for pwAD (Figure  2N–S , Figure  S4 ). K‐means clustering analysis ( k  = 4) confirmed most serum biomarkers of pwRRMS and pwPPMS clustered in cluster 1, while most AD samples were contained in cluster 4 (characterized by relatively high NfL and GFAP). Similarly, CSF biomarkers of pwRRMS and pwPPMS were abundant in cluster 3 (relatively high inflammation biomarkers, lower tissue damage/gliosis biomarkers), while the contribution of AD samples to cluster 1 was substantially larger (relatively low in inflammation biomarkers and high in tissue damage/gliosis biomarkers) (Figure  S5 ). Interestingly, the more intermediate cluster 4 contributed proportionally more to pwPPMS and pwAD compared to pwRRMS, representing an enrichment of individuals with less pronounced inflammatory activity and more damage/gliosis‐associated events in the PPMS group compared to RRMS group.

RR‐Onset and PPMS Brain Donors Display a Similar Lymphocyte Distribution and Pyramidal‐Tract Axonal Density

To further investigate differences in inflammatory and degenerative disease mechanisms, we analyzed post‐mortem brain donor data from 45 PPMS‐diagnosed donors and 86 donors diagnosed with RRMS at onset (at tissue collection, n  = 7 with RRMS; and n  = 79 with SPMS) (Table  S2 ). These analyses reflect MS in its final, often progressive stages. Therefore, the RR‐onset MS group includes patients who typically exhibited progressive pathology at the time of death. Clinical characteristics were described previously by Luchetti et al. [ 5 ] Analysis after CD20 staining (Figure  3A ) showed similar CD20 + B‐lymphocyte distribution between RR‐onset MS versus PPMS donors in the NAWM of brain parenchyma, perivascular space, meninges, and in active and mixed active/inactive lesions ( p  = 0.584, p  = 0.142, p  = 1, p  = 0.481 and p  = 0.349, Figure  3B–E , respectively). This aligns with our results on the B‐lymphocyte activation marker BCMA in serum and CSF, showing similar levels between pwRRMS and pwPPMS. Analysis after CD138 staining (Figure  3F ) demonstrated a similar distribution of CD138 + plasma cells between brain donors with RR‐onset MS and PPMS in the NAWM of brain parenchyma, perivascular space, meninges, and in active and mixed active/inactive lesions ( p  = 0.976, p  = 0.785, p  = 0.205, p  = 0.562 and p  = 0.897, Figure  3G–J , respectively). Similarly, analysis after CD3 staining (Figure  3K ) revealed a similar density of CD3 + T‐lymphocytes in NAWM and in active and mixed active/inactive lesions ( p  = 0.238, p  = 0.566 and p  = 0.094, respectively, Figure  3L,M ). This aligns with our findings on the T‐lymphocyte activation marker sCD27 in serum and CSF, showing similar levels between pwRRMS and pwPPMS. Lastly, quantification of axonal density in the MO of pwRR‐onset MS versus pwPPMS after Bielschowsky staining (Figure  3N ) showed similar axonal density between both diagnosis groups ( p  = 0.532, Figure  3O ). Microglial activity in this post‐mortem brain donor cohort has been previously reported by Luchetti et al. and did not show any differences between pwPPMS and pwRR‐onset MS either [ 5 , 30 ].

Stratification for RR‐onset MS or PPMS diagnosis in the NBB MS cohort. (A) Immunohistochemical staining for CD20 in NAWM (bar = 50 μm) with quantifications in (B) parenchyma, (C) perivascular space (PVS), (D) meninges, and (E) active and mixed active/inactive lesions. p values were calculated with the Pearson chi squared test. (F) Immunohistochemical staining for CD138 in NAWM (bar = 50 μm) with quantifications in (G) parenchyma, (H) PVS, (I) meninges, and (J) active and mixed active/inactive lesions. (K) Immunohistochemical staining for CD3 in NAWM (bar = 50 μm with quantifications in (L) NAWM and (M) active and mixed active/inactive lesions. p values of NAWM were calculated with the Wilcoxon rank sum test, for lesion types with pairwise comparison using negative binomial generalized linear mixed models (GLMMs). p values were calculated with the Pearson chi‐squared test. (N) Bielschowsky staining with (O) quantification of axonal density in the medulla oblongata (bar = 50 μm). p values were calculated with the Wilcoxon rank sum test. p values of < 0.05 were considered statistically significant.

Immune Activation Biomarkers Correlate With Clinical and Laboratory Markers of Inflammation and Serum NfL Associates With Disease Activity in MS

Since soluble biomarkers and post‐mortem findings did not differentiate between RRMS and PPMS, we analyzed their associations with disease activity. All pwMS were stratified based on disease activity within 90 days before sampling, defined as a documented clinical relapse and/or documented radiological disease activity (any GAD‐enhancing lesions on MRI or new lesions compared to a previous scan with a median interval of 7.9 [interquartile range (IQR) 1.1–31.2] months before serum sample and a median interval of 11.9 [IQR 0.8–31.2] months before CSF sample) (Table  1 ). Serum NfL was significantly higher in those with documented active disease ( p  = 0.007), while no other differences were found (Figure  S6 ). Acknowledging missing data, a sensitivity analysis including only participants with sufficient recent clinical and/or MRI data available to determine activity ( N  = 39) showed a similar numerical trend with higher serum NfL in participants with disease activity (Figure  S7B , mean 64.6 pg/mL (standard deviation, SD, 161.1) vs. 20.6 pg/mL (SD 27.6), p  = 0.09). In this subgroup, although not different in the entire cohort, serum OPN was also higher in participants with activity (Figure  S7A , mean 2707.8 (SD 1402.0) vs. 1817.1 (SD 949.6) p  = 0.031). Accordingly, serum OPN was lower in participants without documented relapses or disease activity on MRI within 90 days before sampling, compared to participants without recent relapses but lacking recent MRI data (mean 1817.1 pg/mL (SD 949.6) vs. 3329.0 pg/mL (SD 4577.87), p  = 0.010) while other biomarkers were evenly distributed (data not shown). We further examined the time between last recorded disease activity and sampling. Within the active subgroup, CSF BCMA negatively correlated with time since last disease activity ( p  = 0.031, Figure  4A ), with a similar trend for sCD27 ( p  = 0.12, Figure  4D ). In line with their biological interpretation, CSF BCMA and sCD27 correlated positively with CSF IgG index (both p  = < 0.0001, Figure  4B,E ) and CSF leukocyte count ( p  = 0.0093 and p  < 0.0001, Figure  4C,F , respectively). CSF CHI3L1 also positively correlated with IgG index ( p  = 0.038, Figure  S9 ). Apart from these correlations between immune activation biomarkers and inflammatory parameters, a positive correlation was also found between CSF SNAP25 and time since the last disease activity event ( p  = 0.0083, Figure  S8 ).

Immune activation biomarkers and disease activity; tissue damage biomarkers and disease severity in patients with MS. (A) Correlation between concentration of CSF BCMA and time since last disease activity in days, (B) CSF IgG‐index and (C) CSF leukocyte count. (D) Correlation between concentration of CSF sCD27 and time since last disease activity in days, (E) CSF IgG‐index and (F) CSF leukocyte count. In Figures A–F, pwRRMS are depicted in green, pwPPMS in blue, concentrations are in pg/mL and leukocyte count in *10 6 /L. (G) Correlation between log transformed concentration of GFAP in serum ( y ‐axis) and MS severity as expressed by (e)EDSS score statistically divided in quartiles ( x ‐axis), in (H) this is shown for serum NfL, in (I) for CSF GFAP, in (J) for CSF NfL and in (K) for CSF SNAP‐25. Concentrations in pg/mL. Leukocyte count in *10 6 /L. A linear regression with age and sex as covariate was performed to test any correlation between (e)EDSS and the biomarkers. p values of < 0.05 were considered statistically significant.

Tissue Damage and Gliosis Biomarkers Correlate With MS‐Associated Disability

Lastly, we analyzed correlations between tissue damage/gliosis biomarkers and MS disease severity, measured by (e)EDSS. A higher (e)EDSS score, adjusted for age and sex, correlated with higher serum GFAP and CSF GFAP and NfL ( p  < 0.001, p  = 0.03 and p  = 0.05, Figure  4G,I,J , respectively). This supports the interpretation that these markers reflect the extent of tissue damage and gliosis in MS. Among immune activation markers, the microglial activation marker OPN in serum associated with higher (e)EDSS ( p  = 0.043, Figure  S10 ), while CSF markers showed no association.

Discussion

This study revealed similar biological profiles of CNS pathology between RRMS and PPMS, as indicated by fluid biomarkers of inflammation and tissue damage, which were supported by post‐mortem findings, showing comparable lymphocyte involvement and axonal density in RR‐onset MS and PPMS. Soluble biomarkers distinguished MS from the neurodegenerative, progressive AD and somewhat from the inflammatory, non‐progressive NMOSD. In the MS cohort, active disease was associated with higher serum NfL levels. Immune activation biomarkers correlated with inflammatory parameters like the IgG index and intrathecal leukocyte count, while tissue damage biomarkers correlated with MS severity. This suggests that immune activation biomarkers are closely associated with clinical disease activity, whereas tissue damage biomarkers correlate more strongly with disease progression in MS. Since biological components of MS are similar between RRMS and PPMS, a thorough evaluation of MS‐associated inflammation and tissue damage likely offers a more biologically meaningful classification of MS patients.

None of the analyzed fluid biomarkers differentiated between pwRRMS and pwPPMS, unlike earlier studies of CSF CHI3L1 [ 20 ] and serum or CSF GFAP [ 30 , 31 , 32 ]. These discrepancies may be due to higher inflammation levels in RRMS cohorts from other studies, the relatively older age of pwRRMS in our cohort, or a lack of age and sex adjustments in some earlier studies [ 31 , 32 ]. Our study included older pwRRMS, aimed at minimizing age‐related biases when comparing them to pwPPMS, and adjusted for age in comparisons. Despite the older age, disease course was classified as RRMS at the time of sampling by the treating neurologist.

The analyzed biomarkers revealed, as expected, substantial differences between pwMS and pwAD. Notably, pwAD had higher CSF SNAP25 and OPN, higher serum GFAP, and lower CSF sCD27. Although direct comparisons between pwMS and pwAD are rare, high CSF SNAP25 and OPN and high serum GFAP have been previously reported in AD [ 30 , 33 , 34 ]. Additionally, CSF SNAP25 has been shown to predict progression to AD in non‐demented individuals [ 35 ]. These findings underscore that, while both diseases involve the innate immune system, their tissue injury mechanisms are distinct.

Our post‐mortem tissue analysis showed no differences between pwPPMS and RRMS disease onset, aligning with de Boer et al.'s findings of similar pathological distribution across MS clinical phenotypes [ 36 ], and Luchetti et al.'s results showing no differences in active lesions, microglia/macrophage morphology, and reactive site load (aggregates of HLA + microglia/macrophages) between brain donors with a PPMS‐ or RRMS‐onset diagnosis [ 5 ]. These findings support the view of MS as a continuum rather than distinct phenotypes [ 2 , 3 ].

Serum and CSF biomarkers correlated with clinical and laboratory correlates of inflammation and tissue damage, confirming that they reflect the degree of innate and adaptive immune‐mediated inflammatory events in both RRMS and PPMS. This supports previous findings linking CSF sCD27 and BCMA to disease activity [ 15 , 18 ], and serum and CSF GFAP and NfL to EDSS scores [ 22 , 37 , 38 , 39 ]. Consistent with earlier studies [ 40 , 41 , 42 , 43 ], age was a key factor in predicting MS disease mechanisms, with older age associating with less intrathecal activity of the adaptive immune system but more intrathecal activity of the innate immune system and CNS tissue damage biomarkers.

This study has several limitations. First, the retrospective data collection from medical charts means some participants may have had recent disease activity that could have been missed due to missing MRI scans within 90 days of sampling. We also lacked data on measures of PIRA, which could have affected comparisons between active and non‐active MS, but not between the diagnosed MS phenotypes. Second, age differences between diagnosis groups raise uncertainty about whether observed differences are due to disease course or demographics. Third, the absence of healthy controls means the relevance of biomarkers and the biological mechanisms they represent can only be interpreted in relative comparison with the different diagnoses. Lastly, the sample size was small for NMOSD‐ and MOGAD‐diagnosed patients. Although these patients were included to represent the non‐progressive end of the inflammatory disease spectrum, the limited number of samples restricted our analysis.

Strengths of this study include data‐driven PCA and K‐means clustering analyses, the examination of both CSF and serum biomarkers, and post‐mortem tissue analysis of pwMS. It is among the first to compare pathological components across MS and other neuroinflammatory and neurodegenerative diseases.

In conclusion, this study shows serum and CSF soluble biomarker profiles, along with post‐mortem findings, do not differentiate between RRMS and PPMS diagnoses, supporting that these phenotypes are biologically similar. Evaluating disease mechanisms at play for a person with MS may offer more precise classifications and targeted therapies. Future research should use prospectively collected data and larger cohorts, including healthy controls, to deepen our understanding of MS and refine disease classification.

Author Contributions

Katelijn M. Blok: writing – review and editing, conceptualization, investigation, formal analysis, writing – original draft, visualization. Romy A. M. Klein Kranenbarg: writing – review and editing, conceptualization, formal analysis, writing – original draft, visualization. Kirtana Ananth: writing – review and editing, conceptualization, investigation. Hendrik J. Engelenburg: writing – review and editing, conceptualization, investigation, formal analysis. Aletta van den Bosch: writing – review and editing, conceptualization, investigation. Lucia A. A. Giannini: writing – review and editing, investigation. Janet de Beukelaar: writing – review and editing, conceptualization, supervision, resources. Harro Seelaar: writing – review and editing, resources. Inge Huitinga: writing – review and editing, conceptualization, resources. Ari Green: writing – review and editing, resources. Beatrijs Wokke: writing – review and editing, conceptualization, supervision, resources. Ahmed Abdelhak: writing – review and editing, conceptualization, supervision, investigation. Joost Smolders: writing – review and editing, conceptualization, supervision, resources.

Conflicts of Interest

J.S. received lecture and/or consultancy fees from Biogen, Merck, Novartis, Roche, and Sanofi. The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflicts of interest.

Supporting information

Data S1.