Slow expansion of multiple sclerosis iron rim lesions: pathology and 7 T magnetic resonance imaging

This study is a collaborative project between the Medical University in Vienna and the Neurology Department of Vanderbilt University, Nashville, TN.

The post-mortem study was performed on two partly overlapping samples of 32 MS cases in total, denoted as first and second samples. The first sample (28 cases) served for global characterization of MS WM lesions, their activity state, CD68 expression (microglia/macrophages), and iron accumulation (Figs. 1, 2, 4, 5). It consisted of double-hemispheric or large sections. Thirteen cases were provided by the University of Vanderbilt which received them from the MS Society Tissue Bank (n = 12) and the Rocky Mountain MS Center, Colorado, USA (n = 1). Two cases were collected at the Institute of Neurology, Medical University of Vienna, Austria, and thirteen were from the archive at the Center for Brain Research, Medical University of Vienna, Austria. For the first sample, clinical data and numbers of WM lesions per case are provided in Table 1. The second sample (10 cases) served for the characterization of the inflammatory activation status of iron-laden microglia/macrophages and iron accumulation in astrocytes within iron rims (Fig. 3, supplementary Fig. 1). The second sample consisted of 10 cases, 6 cases from the first sample, and 4 additional cases from the archive at the Center for Brain Research, Vienna (Table 2).

For the in vivo study, ten patients with MS [35] were consecutively enrolled. Patients fulfilled the following inclusion criteria: age >18 years, expanded disability status scale (EDSS) ≤6.5 [25], no steroid therapy during the last 3 months, and no contraindication for 7 T MRI.

They underwent neurological examinations and brain MRIs at the following timepoints: study entry or baseline, years 1, 2, and 3.5. Patients’ demographic and clinical data at the time of the study entry are given in Table 3.

The first sample (Table 1) consisted of 31 formalin-fixed paraffin-embedded tissue blocks comprising double-hemispheric (n = 24) and large (n = 7, tissue size from 5 × 5 × 1 cm up to one hemisphere) blocks. For four cases (MS 2, 13, 26, 27), only large but no double-hemispheric blocks were available. For three cases (MS 6, 10, and 26), two blocks per case were included. The second sample consisted of ten blocks of ten cases. Three large blocks of the first sample were also included in the second sample and seven were additional routine blocks used for the second sample only. Consecutive double-hemispheric or large sections were cut with a tetrander microtome at a thickness of 10 µm and mounted on glass slides. Routine sections were cut at a thickness of 5 microns and mounted on glass slides. Sections were stained for hematoxylin and eosin to exclude confounding pathologies. Luxol fast blue-periodic acid Schiff (LFB-PAS) myelin staining was performed for assessment of WM lesions, including demyelinating activity and remyelination (Fig. 1a). DAB-enhanced Turnbull blue staining for di- and trivalent (total) non-heme iron was done as previously described [21, 29]. Immunohistochemistry for the myelin protein proteolipid protein (PLP) and the lysosomal glycoprotein CD68, indicating activated microglia/macrophages, was performed as described [21]. Sources, pretreatments and dilutions of primary antibodies are listed in Table 4. CD68 stainings were not counterstained to facilitate subsequent digital quantification. Double-labeling of iron with immunohistochemistry for the microglia/macrophage activation markers p22phox, CD86, CD206, and the astrocyte marker GFAP (Fig. 3) was performed on the second sample. Double-labeling of iron with CD68 (Fig. 4n) was performed on MS 26. For double-labelings, Turnbull blue staining was performed as described, but developed with aminoethyl carbazole (p22phox, CD86, CD206, and CD68) under microscopic control [21]. For double-labeling of iron with the astrocyte marker GFAP (Fig. 3), Turnbull staining was developed with DAB according to the protocol used for single labelings. After washing in aqua bidestillata, the sections were steamed for antigen retrieval (Table 4). Blocking of non-specific antibody binding was followed by incubation of the primary antibodies overnight at 4 °C. Appropriate secondary antibodies were applied for 1 h at room temperature. Secondary antibodies were either biotinylated (for p22phox, CD86, CD206, or GFAP) or directly conjugated to alkaline phosphatase (for CD68). If a biotinylated secondary antibody was used, sections were further incubated with avidin-conjugated alkaline phosphatase for 1 h at room temperature. The sections were developed with Fast Blue substrate (Sigma) at 37 °C.

Characterization of WM lesions was performed as described [9, 15]. Active lesions harbored degradation products reactive for LFB and PLP inside macrophages [9] throughout the lesion. Active/inactive lesions showed a confined presence of macrophages with LFB- and PLP-reactive degradation products at the lesion edge (active zones). In this type of lesions, numerous predominantly perivascular macrophages containing PAS-positive lipids [27] were seen in the lesion cores, suggesting more remote demyelinating activity (inactive zones). Slowly expanding and inactive lesions showed sharply demarcated borders [15], inactive demyelinated centers without any degradation products [15], and occasional LFB-positive myelin degradation products in microglia or macrophages at edges of slowly expanding but not inactive lesions (Fig. 1). Remyelinated shadow plaques were identified by their sharply demarcated homogeneous reduction of LFB staining intensity [8] and were classified as shadow plaques if the whole lesional area was remyelinated. Partially remyelinated lesions showed areas of remyelination and complete demyelination [8].

Digital color (RGB) images sized 0.76 × 0.57 mm (0.43 mm²) were taken at constant illumination and microscope conditions [21] from different ROIs in the slides stained for iron and CD68. Three ROIs, at least 1 cm away from any MS lesion, represented the NAWM. A mean value across the three NAWM-ROIs was computed for the analyses. From each WM lesion, three images at the lesion edge and one in the lesion center were taken. For both iron and CD68 stainings, the three edge images were placed with regard to the most pronounced, least pronounced, and in-between CD68 edge expression, regardless of presence of iron. One center image was placed in the lesion center. Images were quantified using custom-written plugins for the ImageJ version 1.43r, which have been used before [21]. After grey-scale conversion of the images, the plugins apply a constant threshold after subtracting the mean grey value from the image, yielding a numerical estimation of CD68 and iron staining (area fraction) (Fig. 2a, b). On the RGB images of the iron stainings, cells with strong iron accumulation displaying microglia or macrophage morphology were manually counted (Fig. 2c). Values of the three edge images and data of multiple lesions of a lesion type per case were averaged to finally represent each lesion type per ROI per case with one data point for plotting and statistical testing (case-based averages) (Fig. 2). In the second sample, the p22phox-iron and CD206-iron double-labelings (Fig. 3) were quantified by manual counting of p22phox or CD206 single-positive cells, iron single-positive cells (on p22phox-iron stainings only), and double-positive cells (supplementary Fig. 2). Total p22phox or CD206-positive cells were determined by adding p22phox or CD206 single-positive cells to the double-labeled cells. For these quantifications, two ROIs were chosen in the edges of rim lesions: one ROI with the lowest, and one ROI with the highest numbers of iron-laden cells found in the whole lesion edge.

Data were analyzed using IBM SPSS^® Statistics for Windows Version 20. Normally distributed metric data are described using mean ± standard deviation (SD). In case of skewed (histological) data, boxplots indicating medians, interquartile ranges (IQR), 1.5 × IQR (whiskers), and outliers (circles and stars) are presented. Nominal data are described using absolute frequencies and percentages. Due to non-normal distribution of histological data, conservative non-parametric Kruskal–Wallis and post-hoc Mann–Whitney U tests were used for comparison of regions of interest and corrected for multiple comparisons according to Bonferroni–Holm. To test the association between categorical variables (i.e., above/below threshold and slowly expanding/inactive lesions), a Chi-square test was performed. A mixed-model ANOVA (fixed within-subject factor ‘time’, fixed between-subject factor ‘rim/non-rim lesions’, random factor ‘patient’) was used to compare changes of logarithmically transformed relative volume fraction data. An unstructured covariance matrix was applied, because it showed the best fit to the data according to the Schwarz’s Bayesian Information Criterion (BIC). Other covariance matrices tested were Compound Symmetry, Autoregressiv 1, and a Diagonal Covariance Matrix. For modelling the random factor ‘patient’, multiple lesions per patient were considered. The reported p values are results of two-sided tests. P values ≤0.05 were considered to indicate significant results.

The 7 T scan was well tolerated by all patients. Two patients complained of transient and short-lasting dizziness upon entering in the scanner. Of all examined MRIs obtained after contrast injection, i.e., at year 1, none showed enhancing lesions. Data of two patients with secondary progressive MS were excluded from the analysis due to movement artifacts. These two patients also could not complete the follow-up because of disability accretion. One patient with secondary progressive MS did not present any rim lesion and was not included in the lesion volume analysis (Patient 10 in Table 3). One patient missed the 3.5-year follow-up examination due to a newly implanted dental brace but was included in the analysis (Patient 1 in Table 3). Data from seven patients were thus included into longitudinal lesion volumetry.