Transient enlargement of brain ventricles during relapsing-remitting multiple sclerosis and experimental autoimmune encephalomyelitis

The brain ventricles are part of the fluid compartments bridging the CNS with the periphery. Using MRI, we previously observed a pronounced increase in ventricle volume (VV) in the experimental autoimmune encephalomyelitis (EAE) model of multiple sclerosis (MS). Here, we examined VV changes in EAE and MS patients in longitudinal studies with frequent serial MRI scans. EAE mice underwent serial MRI for up to 2 months, with gadolinium contrast as a proxy of inflammation, confirmed by histopathology. We performed a time-series analysis of clinical and MRI data from a prior clinical trial in which RRMS patients underwent monthly MRI scans over 1 year. VV increased dramatically during preonset EAE, resolving upon clinical remission. VV changes coincided with blood-brain barrier disruption and inflammation. VV was normal at the termination of the experiment, when mice were still symptomatic. The majority of relapsing-remitting MS (RRMS) patients showed dynamic VV fluctuations. Patients with contracting VV had lower disease severity and a shorter duration. These changes demonstrate that VV does not necessarily expand irreversibly in MS but, over short time scales, can expand and contract. Frequent monitoring of VV in patients will be essential to disentangle the disease-related processes driving short-term VV oscillations from persistent expansion resulting from atrophy.

Supplemental Figure 1: Correlations between ventricle volume and brain volume. (A) Relative ventricle volume (VV) (upper panels) at timepoints 0, 3, 6, 9 and 12 months, for MS patients with contracting (left, orange) and non-contracting (right, green) VV (normalized to timepoint 0; data is the same as shown in Fig. 6B, C in the main manuscript). The corresponding brain volumes (BV) at these timepoints are shown in the lower panels. There was no statistically significant change in BV over time, considering either the contracting or non-contracting cohorts (p=0.875, p=0.977, respectively, 2-factor ANOVA) or the entire cohort (p=0.1167, Wilcoxon test). (B) There was a significant correlation between VV and BV at each of the investigated timepoints (fit with 95% confidence interval for the entire cohort, black; fit for the VV contracting and non-contracting MS patients, orange, green, respectively). The correlations were significant considering the entire cohort, but not the contracting and non-contracting patients separately (Pearson correlation coefficients and p-values are listed in Supplemental Table 2).   Table 2 MRI methods MRI was performed on a 9.4 Tesla MR animal scanner (Biospec 94/20 USR, Bruker Biospin), using a birdcage radiofrequency (RF) coil tailored for mouse brain imaging and ideal for T 1 brain mapping (71). A RARE sequence was used for anatomical imaging (TR=3 sec, TE=14.3 ms, FOV=16.5x16.5 mm², matrix=512x512, number of slices=15, slice thickness=500 µm). Coronal MR images of the entire brain were acquired in 2 min 6 sec. The contrast agent gadolinium-diethylenetriamine-pentaacetate (300 nmol/g Gd-DTPA, Magnevist, Bayer-Schering) was infused via an intravenous cannula over 2 min, during each imaging session. A RARE sequence with variable repetition time (VTR) was used to generate T 1 maps pre-and post-contrast administration (RARE-VTR: TE=11.53 ms, matrix=128×85, slices/thickness=11/500 µm, VTR=0.38, 0.55, 0.94, 1.48, 2.40 and 7.00 sec, echo train length=4)

Analysis of mouse MRI data
In the mouse, tissues external to the brain were eliminated from MR images using the brain extraction tool (BET) of FSL. One mouse brain MR data set was used as reference for the whole study.
A ventricle mask was generated from this data set using the FMRIB's Automated Segmentation Tool (FAST). Following this step, brain images from all time points for each individual mouse were registered to the reference image using the FMRIB Linear Image Registration Tool (FLIRT), to yield a transformation matrix. The transformation matrix for each individual was then applied to the ventricle mask, to determine the absolute ventricle volume in µl for each mouse at each time point.
The transformed ventricle masks were overlaid against the original MR images, and anatomical errors in the segmentation were manually corrected, using the software ITK-SNAP (www.itksnap.org) (75), to yield corrected ventricle volumes. Consistency of the manual correction steps was confirmed by performing replicate corrections of randomly selected individuals and time points. Relative ventricle volume changes were calculated as the ratio of the volume at each time point to the preimmunization volumes, and expressed as percent change.
Data from the T 1 maps was extracted and post-processed using MATLAB (The MathWorks, Inc.), MIJ and ANTs Toolkit (76, 77). T 1 -weighted images (from the RARE-VTR with the longest TR) and T 1 -maps were filtered using adaptive non-local means denoising (78). 3D intra-subject registration (rigid, affine, elastic) was done to register scans from each time point onto the first time point (preimmunization). This was performed on the pre-contrast RARE-VTR T 1 -mapping image with the longest TR, using a cross correlation similarity method. The warping fields were then applied to the T 1 -maps. ROIs delineating the brain and cerebellum were manually defined for all slices from the first time point then applied to the entire series. The average T 1 relaxation time was computed for each

Supplemental Methods
ROI and the corresponding delta-T 1 maps calculated as described (79). The delta-T 1 (difference between pre-and post-contrast T 1 ) was calculated and plotted over time.

Analysis of human MRI data
Quantification of ventricle volumes (VV) and brain volumes (BV) for the MS patient cohort was done using the FSL v5.0, for each time point and each individual patient. Tissues external to the brain were removed from the T 1 -weighted images using the Brain Extraction Tool (BET). T 1 -weighted images were registered onto T 2 -weighted images using FMRIB's Linear Image Registration Tool (FLIRT), and subsequently registered to the MNI standard brain, from which a ventricle mask was applied.
Ventricle segmentation masks were then generated from the T 2 -weighted images. The raw Immunohistological staining Cryosections were dried for 30 min at room temperature, washed in 0.1M PBS, and incubated with protein blocking solution (0.5 % triton + 10 % normal goat / rabbit serum), and avidin-biotin blocking, as needed. Sections were then incubated overnight at 4°C with primary antibodies against GFAP (rabbit anti-mouse, 1:500, Abcam), CD3 (polyclonal rabbit anti-human, 1:500, DAKO), and F4/80 (biotin rat anti-mouse, 1:200, Abcam) followed by goat anti-rabbit-Alexa488 (1:500, Invitrogen) or streptavidin-Pe-Cy5 (1:500, Invitrogen), and subsequently stained with the nuclear stain DAPI. Images were acquired on a fluorescence microscope (Olympus BX-51). For Fluoro-Jade staining, sections were incubated in 0.06 % potassium permanganate for 20 min at room temperature, then rinsed in distilled water. The sections were then incubated in a solution of 0.001 % Fluoro-Jade B (Chemicon) in 0.1 % acetic acid for 20 min at room temperature in the dark, then rinsed 3 times in distilled water and subsequently stained with DAPI. Fluoro-Jade stained slides were scanned using a Panoramic MIDI II slide scanner (3D Histech). Images were visualized using Case Viewer v.2.3 (3D Histech). Obvious tissue processing artifacts were manually removed, and intensity of Fluoro-Jade staining quantified in the whole section using ImageJ v1.51 (NIH).