INTRODUCTION

Spontaneous remyelination partly fails in multiple sclerosis (MS) white matter (WM) lesions. Neuropathological studies¹ suggest that the efficacy of this process might depend on lesion location, being particularly reduced in the vicinity of ventricular cerebrospinal fluid (CSF). However, to date this evidence has not been confirmed by in-vivo data. A recent study² has demonstrated that it is possible to measure remyelination in vivo with positron emission tomography (PET) using the [11C]PiB radioligand. Here, we used this technique to explore the relationship between the extent of remyelination and the distance from ventricular CSF in a group of MS patients

METHODS

Nineteen patients with MS and eight healthy controls (HC) underwent a MRI protocol using a 3T Siemens MAGNETOM Trio MRI scanner equipped with a 32-channel head coil, and a 90 min dynamic [11C]PiB PET exam acquired on a high resolution research tomograph HRRT, at baseline and, only for patients, after 2 to 4 months. MRI protocol included a 3D-T1 MPRAGE (1x1x1.1mm) and a T2-TSE (0.9x0.9x3mm). T2 hyperintense lesions were manually segmented and aligned to the T1 space. After performing a “lesion-filling” procedure³, WM, cortical grey matter, and CSF were segmented in all subjects using Freesurfer. Distance map from ventricular CSF into WM were calculated with FSL. PET images were quantified as previously described⁴, obtaining parametric maps of distribution volume ratio (DVR). For each patient, the T1 images acquired at the two time-points were aligned to each other to create a “halfway space” which was normalized to the MNI standard space using FNIRT. Derived transformations were then combined to align all images (DVR, distance map and lesion mask) to the standard space, passing through the halfway space. On the DVR images of patients, each lesional voxel was classified as demyelinated if its DVR value was < 1 standard deviation below the mean DVR value of all the voxels in HC that were localized at the same distance from the CSF². Individual maps of demyelinated voxels at baseline, as well as maps of dynamically demyelinating voxels (defined as lesional voxels classified as myelinated at baseline and demyelinated at follow-up), and of remyelinating voxels (defined as lesional voxels classified as demyelinated at baseline and normally myelinated at follow-up), were generated (Figure 1). From these individual maps, we computed three probability maps: i) probability map of demyelinated voxels at baseline; ii) conditional probability map of dynamically demyelinating voxels obtained by normalizing the probability of a voxel to be demyelinated at follow-up by its probability to be normally myelinated at baseline iii) conditional probability map of dynamically remyelinating voxels obtained by normalizing the probability of a voxel to be myelinated at follow-up by its probability to be demyelinated at baseline. At the single subject level, the relationship between the probability of remyelination and the distance from ventricles was modelled using a segmented model composed of two linear segments with different slopes. To minimize partial volume effect, the first 3 mm around the ventricles were excluded from this analysis. The relationship between the probability of voxels to remyelinate near ventricular CSF and cortical thickness (CTh) was measured with Pearson correlation coefficient.

RESULTS

Both the probability of voxels to be demyelinated at baseline and to dynamically demyelinate over the follow-up were higher in brain regions close to ventricular CSF compared to more distant regions (Figure 2A-B). The opposite effect was observed for the probability of voxel to remyelinate (Figure 2C). The model based on single subject analysis showed that the mean probability of voxels to remyelinate, which was 0.08±0.09 close to the ventricles, increased of 0.04±0.03 for each mm of distance from CSF up to a distance of 14.12±7.00 mm, where the gradient almost reached a plateau. An example of the model fit is presented in Figure 3. A higher probability of voxels to remyelinate close to the CSF significantly correlated with a higher CTh (r=0.56, p=0.013) (Figure 4).

DISCUSSION & CONCLUSIONS

We demonstrated for the first time in vivo that spontaneous remyelination is significantly reduced in periventricular WM lesions in patients with MS, and becomes progressively more extensive with increasing distance from ventricles. These data suggest that the presence of potentially cytotoxic soluble mediators in the CSF could exert an inhibitory effect on the remyelination process around the ventricles⁵. This hypothesis is further supported by the positive correlation we found in our cohort between periventricular remyelination and CTh, suggesting that similar factors could influence tissue damage and repair in areas close to the CSF both in WM and in cortex.

Acknowledgements

No acknowledgement found.