The study included a cohort of 96 patients comprising 16 with ALS, 26 with focal epilepsy, 23 with CSVD and 31 controls (Table 1). Note that subgroups differed slightly in n-numbers due to a few missing samples. Exact n numbers are given in all figure captions.

Patients were recruited from the Department of Neurology at the Otto-von-Guericke University Magdeburg between 02/2012 and 03/2020. ALS diagnosis was based on the revised El Escorial criteria (Brooks et al., 2000; Carvalho and Swash, 2009). According to these, 3 patients suffered from possible, 8 from probable and 5 from definite ALS within the ALS cohort. Epilepsy patients were diagnosed with focal epilepsy according to the International League Against Epilepsy (ILAE) (Scheffer et al., 2017). The CSVD cohort comprised 11 cases with probable CAA diagnosed according to the modified Boston criteria; four of them fulfilled a concomitant clinical diagnosis of Alzheimer’s disease (AD) according to the NINDCS/ADRDA criteria (Linn et al., 2010; McKhann et al., 1984). Another 12 CSVD patients had HA and diagnosis was based on the existence of deep and mixed, i.e., deep and lobar hemorrhages, detected on T2*-weighted magnetic resonance imaging (MRI) sequences (Pasi et al., 2018; Scheumann et al., 2020).

A hospital-based cohort of neurologic patients, comprising cases with non-specific complaints who underwent lumbar puncture in terms of a diagnostic workup to rule out any neurologic condition served as controls. None of those disease controls suffered from any neurological disorder, especially not from ALS, epilepsy or seizures, CAA or AD, and HA.

In all patients and controls lumbar puncture (LP) and venous puncture (VP) have been conducted for diagnostic workup. For lumbar puncture patients were seated and 9 ml CSF were taken. In all samples, relevant CSF blood contamination was excluded visually/macroscopically and through microscopy of non-centrifugated CSF aliquots. After centrifugation, the pellet was separated to avoid the presence of cells in the final analysis samples. Within 20 min of LP and VP, CSF and serum samples were centrifuged at 4°C, aliquoted and stored at −80°C until brevican and neurocan analysis (see below). In ALS and CSVD, CSF markers of neuroaxonal damage and neurodegeneration were determined immediately after sample collection; they were not available in the epilepsy group. Neurofilament light chain (NF-L) was measured with commercially available ELISA (NF-light^® ELISA, IBL International GmbH, Hamburg, Germany) and total Tau (tTau) was determined either through ELISA (Innotest hTauAg, Innogenetics, Ghent, Belgium (until 12/2019)) or with an automated immunoassay (LUMIPULSE^® G600 II, Fujirebio lnc., Japan (from 01/2020 on)), following the manufacturer’s instructions (see also Schreiber et al. (2018a,b)). CSF Abeta 1-40 was assessed with the Innotest β-Amyloid(1-40) ELISA (until 12/2019) and after 1/2020 also with the automated immunoassay. CSF and serum albumin as well as Immunoglobulin G were determined by rate nephelometry (Nephelometer Immage 800, Beckman Coulter, Fullerton, CA, United States). The CSF albumin/serum albumin ratio (Q_alb x 10^–3) was calculated to assess blood-brain barrier (BBB) integrity, and the IgG-Index (Q_alb/Q_IgG) was calculated to assess intrathecal immunoglobulin synthesis as a predictor for neuroinflammatory processes.

The study was approved by the Ethics Committee of the Otto-von-Guericke-University Magdeburg (Sign 07/17). Patients donating biological material gave written informed consent in accordance with the Declaration of Helsinki. Data and material were handled in a coded fashion guaranteeing patient anonymity.

For the quantitative measurement of total brevican and neurocan we used commercial anti-human brevican und anti-human neurocan ELISA kits, respectively (RayBiotech Norcross, GA, United States) (ELH-BCAN-1 und ELH-NCAN-1). Serum samples were diluted 1:2 for brevican ELISAs and used undiluted for neurocan ELISAs. The dilution factor for CSF was 1:200 to 1:300 for anti-brevican ELISAs (to optimize the dilution for the detection range of the ELISA batches) and 1:10 for anti-neurocan ELISAs. ELISA measurements were essentially conducted according to the manufacturer’s instructions. In brief, pre-diluted CSF and serum samples were incubated in 96-well strip microplates pre-coated with monoclonal mouse anti-brevican (immunogen: aa 23-911) or polyclonal sheep anti-neurocan (immunogen: aa 23-1321). Samples were analyzed as triplicates or duplicates.

To compensate for sensitivity differences in ELISA batches the calibrator concentration of the recombinant brevican or neurocan protein for the standard curve was adjusted according to OD values obtained (highest standard concentrations were set to 3 ng/ml in brevican ELISAs and 25 ng/ml in neurocan ELISAs; standard dilution series: 6 × 1:2 dilution steps).

After 2.5 h incubation, 1:80 diluted biotinylated antibody against the target protein was added and incubated for 1 h, followed by a 45 min incubation with HRP-streptavidin (1:200 for brevican and 1:500 for neurocan) and a 30 min incubation with 3,3′,5,5′-tetramethylbenzidine in the dark. The reaction was stopped by adding 0.2 M H₂SO₄. OD values were measured in a VersaMax™ Tunable Microplate Reader (Molecular Devices, San Jose, United States) at 450 nm and at 650 nm for unspecific background subtraction. According to the manufacturer’s information intra- and inter-assay variability is below 10 and 12%, respectively.

Following primary antibodies were used: rabbit anti-brevican “neo” (Rb399) (custom-made; rat neo-epitope QEAVESE; see Valenzuela et al., 2014; 1:500), polyclonal sheep anti-human/rat brevican (AF4009, R&D Systems, Minneapolis, United States, 1:1000), polyclonal sheep anti-rat/mouse neurocan (AF5800, R&D Systems, Minneapolis, United States, 1:1000), and the antibodies from the ELISA kits (see above).

To prove if the neo-antibody, which was originally produced against the neo-epitope of the rat N-terminal brevican fragment, is also detecting the human brevican fragment neo-epitope EATESE we performed a competition assay with pre-incubation of the primary antibody with a 100-fold molar excess of the peptide Biotin-Ahx-CGQEAVESE containing the immunogen for 1 h at RT. As shown in Supplementary Figure 1, the 60 kDa band detected by this antibody in human CSF samples is clearly diminished in the peptide competition condition, demonstrating its specificity.

As secondary antibodies we used peroxidase-conjugated AffiniPure donkey anti-rabbit IgG (H + L) and peroxidase-conjugated AffiniPure donkey anti-sheep IgG (H + L) (Jackson ImmunoResearch, Cambridgeshire, United Kingdom, 1:2000).

For semiquantitative assessment of brevican and neurocan fragments undiluted CSF samples were incubated with chondroitinase ABC (Sigma-Aldrich, St. Louis, United States) at 1U/mg at 37°C for 30 min to digest chondroitin sulfate side chains and afterward solubilized in 5 × SDS loading buffer (250 mM Tris/HCl, pH 8, 50% glycerol, 10% SDS, 0.25% bromophenol blue, 0.5 M DTT) and boiled at 95°C for 10 min. Samples were separated on 2,2,2-trichloroethanol (TCE)-containing stain-free 5–20% Tris-glycine SDS polyacrylamide gels under reducing conditions. Before Western blotting proteins were activated under UV light for 5 min. Protein transfer onto PVDF membranes (Merck Millipore, Burlington, MA, United States) was performed according to standard protocols. To control for proper protein transfer UV images of the membranes were captured. After blocking for 1 h at RT in 5% horse serum (Sigma-Aldrich, St. Louis, United States) in TBS-T (150 mM sodium chloride, 50 mM Tris, 0.1% (v/v) Tween20, pH 7.6) membranes were incubated in primary antibodies overnight at 4°C. After washing three times with TBS-T for 10 min, secondary antibodies were added and incubated for 60 min at RT. After washing membranes again three times with TBS-T for 10 min, immunodetection was performed using an ECL Chemocam Imager (INTAS Science Imaging Instruments GmbH, Göttingen, Germany). To improve semiquantitative comparability of optical density data from immunoblots we used a standard sample loaded on all gels as a reference to calibrate individual blot data and compensate for gel differences. Quantification of band intensities was done using NHI ImageJ software version 1.52a (US National Institutes of Health, Bethesda, MD, United States).

For the analysis of brevican and neurocan levels non-parametric testing was used due to the sample characteristics, particularly the relatively small sample sizes of the ALS and CSVD groups. Because the overall groups were not matched for age and sex, a rank analysis of covariance was performed on brevican and neurocan levels following the protocol of Quade (1967). To this end, the data were transformed into rank coefficients and sex and age were partialled out of the respective dependent variable (brevican and neurocan levels) using a multiple regression. The resulting residuals of the dependent variable were z-standardized and entered a one-way ANOVA including the factor group (control, ALS, epilepsy, and CSVD). Given a significant group effect, post hoc comparisons were conducted using Tukey’s multiple comparison tests (Tukey’s HSD). For statistical comparisons within the CSVD (HA versus CAA) group, t-tests were calculated on age- and sex-corrected brevican and neurocan levels. For analyses of data uncorrected for sex and age, please see Supplementary Tables 1, 2. For assessment of correlations on the uncorrected data, Spearman’s rank correlation (r_ρ) was used and p values were reported with Bonferroni correction for multiple testing. The level of significance was defined as p ≤ 0.05 and all tests were conducted two-tailed. Results were reported with Cohen’s d and partial eta squared (η_p²) as effect sizes. Statistical analysis was done using GraphPad Prism version 9 (GraphPad Software, Inc., San Diego. CA, United States) and SPSS Version 28 (IBM Corp., Armonk, NY, United States).

The concentrations of total neurocan and total brevican were quantitatively assessed in patients’ body fluids with commercially available ELISAs. For all patient groups the mean values ± SD and the median values of raw data are given in Table 2. These levels are in the same concentration range as documented in previous reports (Minta et al., 2019a; Jonesco et al., 2020). It has to be noted, though, that serum concentrations of neurocan were at the lower detection limit of the assay in many cases. The rank analysis of covariance including the factor group (control, ALS, epilepsy, & CSVD) did not reach statistically significance on both, brevican, F(3,85) = 1.04, p = 0.38, η_p² = 0.04, and neurocan total CSF concentration levels, F(3,83) = 0.03, p = 0.99, η_p² < 0.01; see Figures 2A-D. The effect of group was also not significant for brevican total serum levels, F(3,75) = 0.75, p = 0.53, η_p² = 0.03, and for neurocan total serum levels, F(3,80) = 1.84, p = 0.15, η_p² = 0.07. No significant differences were measured within the CSVD subgroups (CAA versus HA) for brevican and neurocan total CSF concentration levels (all ts ≤ 1.40, ps ≥ 0.18) and total serum concentration levels (all ts ≤ 0.86, ps ≥ 0.40).

Interestingly, the levels of neurocan and brevican in the CSF showed a significant positive correlation (r_ρ = 0.641, p < 0.001) which was observable in all sample groups (Figure 2E). There was also a significant correlation between the serum total neurocan and total brevican concentrations (r_ρ = 0.34, p < 0.05; not shown as a separate plot, but indicated in Figure 3).

To assess the relative abundance of major proteoglycan fragments, i.e., the 60 kDa N-terminal brevican fragment derived from ADAMTS-4 cleavage and the 150 kDa neurocan-C fragment produced independently from ADAMTS activity, we used immunoblots with chondroitinase ABC-digested CSF samples and a reference sample for calibration. Since brevican and neurocan fragment concentrations in the serum samples were too low to be reliably detected on immunoblots, we focused on CSF samples.

To detect correlations of the estimated CNS-derived proteoglycan concentrations in the patients’ body fluids with each other and with established biomarkers we performed a Spearman’s correlation analysis covering relevant diagnostic parameters obtained in the clinical work up (Figure 3). The serum concentrations of brevican and neurocan did not show any correlations with other parameters.

For the total (r_ρ = 0.272, p = 0.01) and cleaved brevican levels (r_ρ = 0.288, p = 0.007) as well as for total neurocan (r_ρ = 0.262, p = 0.014) in the CSF we detected significant positive correlations with the age of the patients, which did not survive Bonferroni correction. Interestingly, CSF proteoglycan levels, except for cleaved brevican, also correlated strongly with the Alzheimer precursor protein fragment Abeta 1-40, a physiological brain-derived extracellular polypeptide (total neurocan r_ρ = 0,747, p < 0.001, total brevican r_ρ = 0,6, p < 0.001, cleaved neurocan r_ρ = 0,56, p < 0.001). For cleaved as well as for cleaved/total brevican in the CSF, we observed significant correlations with CSF albumin and with the albumin quotient (r_ρ = 0.41 and 0.39, respectively, p < 0.01).

Since ALS and CSVD are neurodegenerative disorders, we also explored the correlations with neurodegeneration markers NF-L and total Tau. As shown in Figure 6, there were positive correlations with t-Tau for all neurocan measures analyzed in CSF of the ALS samples (in A for total neurocan r_ρ = 0.82, p < 0.001; in B for cleaved neurocan-C fragment r_ρ = 0.90, p < 0.001; and in C for the ratio of cleaved to total neurocan r_ρ = 0.73, p < 0.01). In the CSVD sample, only total neurocan was significantly correlated with t-Tau values (Figure 6A; r_ρ = 0.52, p < 0.05), and contrary to our expectations we did not find any correlation with NF-L levels in the patients’ CSF samples.

As epilepsy can be caused by neuroinflammatory processes we checked our epilepsy cohort for the occurrence of encephalitis. There were no cases with acute, but 5 cases with putative subacute encephalitis. A comparison between these two subgroups revealed significantly lower amounts of CSF neurocan and neurocan-C in the encephalitis cases (Supplementary Figure 3), however, we did not find a correlation of neurocan levels with the IgG index across groups (Figure 3).

The neural ECM has a relatively low turnover rate (i.e., within days; see Fawcett et al. (2019)), with its biosynthesis and degradation being in equilibrium under healthy conditions keeping homeostasis (Dityatev et al., 2010). Nevertheless, the composition and integrity of the brain ECM can be dynamically adapted to the activity state of neural networks (i.e., within minutes to hours), mostly via the action of ECM-cleaving matrix metalloproteinases (Frischknecht et al., 2014). Thus, strictly controlled proteolytic events yield a more loosened ECM which allows structural changes in the synaptic wiring and functional connectivity alterations. Importantly, dysregulation of these ECM-shaping processes can be a hallmark for several neurological disorders (Gottschall and Howell, 2015), making glycoprotein or proteoglycan fragments in body fluids of patients valuable readouts for the general ECM proteolytic activity. This activity may vary largely with disease stages and severity and may therefore be an additional proxy for the disease progression. For neurocan it was shown that the major proteolytic event yielding the neurocan-C fragment is affected by cytokines like Transforming Growth Factor beta (TGFbeta), Platelet-Derived Growth Factor (PDGF), and Epidermal Growth Factor (EGF) (Asher et al., 2000).

Furthermore, next to the proteolytic activity by itself (both in the brain parenchyma and in the CSF) also the passing rate across the BBB resp. the blood CSF barrier may affect the levels measured in body fluids. This assumption is supported by the observed correlation of cleaved versus total brevican in the CSF with the albumin quotient (Figure 3), arguing for increased barrier leakage driving the occurrence of the proteoglycan fragment (which is in the same size range as albumin with 66 kDa) in the CSF.

Finally, the glymphatic flux rate inside the brain parenchyma is supposedly a major factor in the regulation of the CSF composition. This assumption is supported by our finding that CSF proteoglycan levels largely correlate with the levels of Abeta 1-40, another neural polypeptide found in the extracellular space of the CNS and cleared by the glymphatic system (Iliff et al., 2012), indicating common efflux mechanisms to the CSF. However, cleaved neurocan shows also an effect of the disease group (ALS), pointing to a disease-specific mechanism not explained by physiological clearance rates.

The neural ECM is a dynamic molecular filter for all cellular secretomes in the brain. It undergoes constant homeostatic remodeling and appears disorganized in several pathologies. So far, to our knowledge no other study investigated brevican and neurocan fragments as potential biomarkers for epilepsy, CSVD or ALS. However, in a series of biochemical or proteomic studies a Swedish group investigated brevican and neurocan already in a spectrum of other neurological conditions. In 2018 they demonstrated both brevican and neurocan CSF levels being lower after prophylactic cerebral radiotherapy of lung cancer patients, but only in cases without brain metastases (Fernström et al., 2018). In TBI patients the same group showed that CSF levels of brevican and neurocan as well as of their binding partners tenascin-C and tenascin-R change over time after the injury – again arguing for a disease stage dependence of ECM modifications - and that lower values correlated with a favorable outcome for the TBI patients (Minta et al., 2019a). For neurocan they found a reduction in the serum of TBI cases. In a follow-up study, they demonstrated that mostly N-terminal brevican peptides (present on the 60 kDa fragment investigated here) were reduced in TBI as compared to idiopathic normal pressure hydrocephalus (iNPH) CSF samples as contrast group, and that ADAMTS-like proteolytic activity in the CSF of TBI patients was increased (Minta et al., 2021b). Interestingly, in their study the rise in ADAMTS-like activity correlated also with matrix metalloprotease (MMPs 1, 2, 3, 10) concentrations in CSF, suggesting a general increase in extracellular proteolytic activity. In iNPH patients the same group compared lumbar and ventricular CSF concentrations of proteoglycans and MMPs demonstrating that neither brevican nor neurocan differed between CSF compartments, however, MMPs −1, −2, −10 and the tissue inhibitor of metalloproteases TIMP-1 were found increased in lumbar as compared to ventricular CSF (Minta et al., 2021c).

For different types of dementias specific correlations with neural proteoglycans have been reported: In AD patients the CSF and serum levels of brevican or neurocan were concurringly found to be unchanged as compared to non-demented controls and there was no correlation with the Abeta_42/40 values but a significant correlation between brevican, neurocan, and the brevican-interacting ECM glycoprotein tenascin-R (rho = 0.68–0.77, p < 0.05) (Begcevic et al., 2018; Brinkmalm et al., 2018; Minta et al., 2019b,2021a; Jonesco et al., 2020).

Fragments produced by proteolytic cleavage of CNS proteoglycans may not only be an indicator for the deconstruction of the intact ECM, but they could provide a gain of function by acting as matrikines. This term describes bioactive proteolytic (or glycosidic) fragments exerting signaling functions which may be different from the role of the uncleaved ECM components (Ricard-Blum and Salza, 2014; Fontanil et al., 2021).

Although transport modalities for brain-derived matrikines into the CSF and into the serum are still fully unclear, to study them under pathophysiological conditions could be of interest both from a clinical but also from a cell biological perspective.

Further experiments are necessary to confirm these findings. in an independent sample and to clarify the clinical value.

The statistical power from the relatively small numbers of individuals involved was felt to be too low to draw firm conclusions with regard to the suitability and disease specificity as bona fide biomarkers, however, our results are promising and deserve further validation in larger cohorts to unravel the full potential in the clinical work up, not least to come closer to the pathophysiological mechanisms behind our observations.

Taken together, our experiments suggest that – beyond association with clinical phenotypes – the CSF levels of the neural proteoglycans investigated here seem to be differentially regulated: while total brevican and neurocan as well as the C-terminal fragment of neurocan appear to be largely controlled by the general glymphatic flux, the 60kDa brevican fragment seems to be rather controlled by the BBB function.

From our work we conclude that proteolytic cleavage products of brain-derived perineuronal ECM molecules, such as neurocan fragments, may allow insights into the integrity of the brain’s extracellular environment, being potent CSF indicators for the actual proteolytic ECM fragmentation activity in the CNS tissue, especially if normalized to the concentration of total core proteoglycans. In the clinic, their correlation with established markers for neurodegeneration or BBB leakage may pave the way to a more differentiated diagnosis of neurological disorders, based on body fluids. Our results suggest to include the ratio of cleaved to total neurocan into informative biomarker panels to improve the discriminatory power between disorders.