Osteopontin levels are associated with late-time lower regional brain volumes in multiple sclerosis

Osteopontin (OPN) is a proinflammatory marker produced by systemic immune and central nervous system (CNS) resident cells. We examined, if the level of OPN in the cerebrospinal fluid (CSF) and blood is associated with late-time regional brain volumes and white matter (WM) lesion load in MS. Concentrations of OPN in blood and CSF were related to MRI findings 10.1 ± 2.0 years later in 46 patients with MS. OPN concentration was measured by ELISA, while regional brain volumes and lesion load was assessed by magnetic resonance imaging (MRI) using 3D MPRAGE sequence and automated MR volumetry. OPN measured in the CSF was associated with several regional brain volumes and WM lesion load measured 10.1 ± 2.0 years later. CSF OPN concentration correlated with long-term enlargement of lateral- and inferior lateral ventricles and the elevation of gross CSF volume, in conjunction with the reduction of several cortical/subcortical gray matter and WM volumes. Serum OPN showed no long-term association with regional brain volumes. OPN measured from the CSF but not from the serum was associated with lower regional brain volumes measured a decade later, indicating the primary role of inflammation within the CNS in developing long-term brain related alterations.

www.nature.com/scientificreports/ secondary progressive MS (SPMS) patients 19 , moreover, higher OPN levels in CSF were measured in patients with active disease, as compared to patients with stable disease 17 . The highest OPN concentration in CSF was measured in RRSM patients 17 . OPN levels in the CSF correlate with development of microstructural abnormalities and functional connectivity within 10 years 20 . Therefore, we here examined the long-term effect of OPN on regional brain volumes in patients with MS.

Results
40 reliably segmented structures were extracted from Freesurfer's segmentation output, omitting brainstem. Figure 1 shows the results of automatic segmentation on a randomly selected subject. Multiple linear regression models included age, gender, and estimated total intracranial volume as variables of no interest. Dependent variables were the segmented brain structures and OPN was included as independent variable of interest in separate models. Storage time was also included in the initial models, but was removed from the final ones, as storage time was not proved to be a significant predictor in any of the tested models. OPN measured from serum samples showed no associations with the segmented volumes, regardless of the time of collection [median OPN concentration 12.52(9.0-16.7) ng/mL at the time of CSF collection and 9.8(6.2-18.8) ng/mL at the time of MRI].

Discussion
OPN measured from the CSF was associated with reduced brain volume in several regions within 10.1 ± 2.0 years, indicating that the level of CSF OPN was associated with regional brain volumesmeasured a decade later. Results showed that an elevated CSF OPN concentration predicted the late-time enlargement of lateral-and inferior lateral ventricles and the elevation of gross CSF volume, in conjunction with the reduction of several cortical and subcortical GM volumes. Tortorella et al. conducted a cross-sectional study on patients with clinically isolated syndrome (CIS), measuring OPN concentrations from CSF and gross brain volumetry, including total GM volume, peripheral GM volume, total brain volume, ventricular volume and manually assessed corpus callosal index. Their results showed that OPN levels were only weakly associated with corpus callosum index 21 . In our longitudinal cohort, CIS patients were not included, and we did not observe such associations in our MS populations. Moreover, cross-sectional statistical analyses yielded no significant associations at all, and all significant correlations surviving FDR correction corresponded to OPN concentration from CSF and MRI acquired 10.1 ± 2.0 years later. The suggestion that OPN may be associated with WM damage 21 is well supported by our previous study in the same cohort, demonstrating that CSF OPN levels are related to wide-spread WM alterations localized to the normal appearing white matter (NAWM) of left superior and inferior longitudinal fasciculi, external capsule, forceps minor (genu of corpus callosum) and anterior corona radiata, indicating myelin loss and axonal degenerations 20 . These previous and the current data may thus suggest that patients with higher OPN CSF levels developed more extensive WM damage accompanied by the association with reduced WM and GM www.nature.com/scientificreports/ regional volumes. Indeed, GM atrophy is strongly associated with WM injury in MS patients, particularly with injury to association fibers 22 . It is conceivable that soluble biomarkers, specific for late-time regional brain volumes, alone or in combination with MRI biomarkers, may be clinically valuable in prognostic evaluation at the beginning of MS disease. Brain atrophy has clinically relevant impact on MS pathogenesis: higher atrophy rate leads to the worsening of expanded disability status scale (EDSS) and progression to disability 23 . Recently, several phase III trials defined brain atrophy as an outcome in both relapsing and progressive MS, and a number of disease modifying treatments significantly reduced atrophy rate 24 . If confirmed in other independent cohorts, CSF OPN concentration may be a potential marker for screening patients for high risk of accelerated atrophy rate in the long-term. Despite the fact that due to study design, we cannot state that the observed association with smaller regional brain volumes (and larger ventricles) indicate atrophy, our study may still indicate that in the development of brain volume losses reported earlier 8,9 , OPN produced within the CNS plays an important role. This also emphasizes the role of inflammation within the CNS compartment in the evolution of atrophy.

Methods
Subjects. Forty-six patients with clinically definitive MS (32 females, age range at MRI: 20-68 years) have participated in the study. Serum and CSF samples were collected 10.1 ± 2.0 years before MRI and aliquots were kept at − 80 °C until further processing. A new serum sample was taken at the day of MRI acquisition. All patients participating in the study had MS fulfilling the 2017 modified McDonald diagnostic criteria 25 . In case of relapsing MS, the MRI measurements were taken in the remission phase. Most of the patients were on chronic disease modifying treatment ( Table 1). 11% of the patients had primary progressive MS, and at the time of MRI, 67% had relapsing and 22% secondary progressive MS. During the follow-up period (10.1 ± 2.0 years), the median number of relapses was 3 (IQR:2-4), and EDSS has increased in the study population (p = 0.034, Wilcoxon Signed Rank Test).
The study was conducted according to the World Medical Association Declaration of Helsinki and approved by the Regional Ethical Committee of the University of Pecs (7068-PTE 2018). All patients signed written informed consent prior to study procedures.

Measurement of osteopontin in serum and CSF.
After centrifugation, supernatants were stored at −80 °C until further processing. For quantitative detection of OPN concentrations in the serum and CSF samples, a commercially available sandwich enzyme-linked immunosorbent assay (ELISA) kit was used (Human  Table 2 Volumetric analysis of the T1-weigted MR images. 3D T1 images were fed into volumetric segmentation performed with FreeSurfer v6.0. Details of the procedures are described in previous publications 26,27 . Each dataset was checked within the processing stream to verify the following stages: Talairach transform, skull strip, white matter-and pial surface segmentation, as described in Freesurfer's Recommended Reconstruction Workflow. The white matter hypointensities labels were corrected by hand for all subjects to avoid the mis-segmentation of white matter lesions (T1 black holes) as grey matter. The final volumetric results from Freesurfer were fed into statistical analysis.
White matter lesion segmentation. LST toolbox version 3.0.0 (Lesion Segmentation Toolbox, https:// www. stati stical-model ling. de/ lst. html) was used to automatically segment cerebral white matter lesions on 3D FLAIR images using the lesion prediction algorithm 28 .
Statistical analysis. All statistical analyses were performed using SPSS (IBM Corp., Version 25.0. Armonk, NY). For volumetric analysis, multiple linear regression models were employed with the volumes of the segmented brain structures as dependent variable and OPN concentration, age, gender and estimated total intracranial volume as independent variables. The assumptions of multiple linear regression were satisfied, as judged by testing for linearity, independence of errors, outliers, normality assumptions of the residuals, homoscedasticity and multi-collinearity. Significance level was set at p < 0.05. Given the large number of segmented structures in the volumetric analysis, multiple comparisons correction with Benjamini-Hochberg procedure was applied with a conservative q = 5%.

Limitations
The present study bears limitations inherent to the study design. The main limitation is the lack of initial (baseline) MRI measurements. Without the baseline measurement we cannot state that the association with lower regional brain volumes (and larger ventricles) correspond to brain atrophy per se.