Introduction

Perivascular spaces (PVS) are fluid filled spaces surrounding brain vessels¹. PVS can be pathologically enlarged and are then visible on T2-weighted MRI as hyperintense elongated stripes. They have been related to aging, hypertension, stroke, and inconsistently to cognitive impairment^1,2. Furthermore, enlarged PVS (EPVS) are a common hallmark of cerebral small vessel disease (CSVD)¹. Sporadic CSVD is driven by changes in the vessel’s wall due to arteriosclerosis – mainly in deep brain regions such as the BG - and accumulation of Amyloid-β (Aβ) in cerebral amyloid angiopathy (CAA), found in cortical and leptomeningeal vessels.
EPVS have been implicated in the transport of toxic proteins out of the brain (perivascular clearance) and it has been hypothesized that their enlargement occurs when clearance is dysfunctional². Pinpointing the anatomy of the EPVS and their relationship to measures of blood flow can help shed light on mechanisms of disease and on the routes of perivascular clearance. For example, it remains unclear if PVS enlarge around arteries or veins in the BG and whether increased pulsatility in the vessels is a driver of enlargement^3,4.
This work addresses these knowledge-gaps in a cohort of cognitively normal elderly controls and CSVD patients. We combined in vivo ultra-high-resolution 7T MRI and histopathology to investigate whether PVS enlarge around arteries, veins, or both. Furthermore, we assessed whether pulsatility-index (PI) and blood flow velocity (MeanV=V_mean), measured by 7T phase-contrast MRI in the arteries of the BG, is associated with the respective EPVS measures.

Methods

Forty-one participants (72.21y±5.59; mean±SD) were recruited by the German center for neurodegenerative diseases (DZNE) and the University Clinic of Magdeburg⁵. Each participant underwent a 3T (Siemens Verio) and, after inclusion in the study, a 7T MRI scan (Siemens Healthineers). Based on the 3T MRI, five CAA-cases were selected according to the Boston criteria 2.0⁶. Twelve non-CAA CSVD showed microhemorrhages, a defining CSVD marker, in deep brain regions or mixed locations. Furthermore, twenty-four non demented healthy elderly controls without microhemorrhages and without other severe markers of CSVD were included. A 3T T2-weighted turbo-spin-echo-sequence (voxel-size: 1x1x2 mm) was used to visualize the EPVS. Arteries were detected on 7T time-of-flight-angiography (ToF) (voxel-size: 0.28mm isotropic), veins on quantitative susceptibility mapping (QSM), reconstructed from a 7T T2*-weighted 3D-gradient-echo-pulse-sequence (voxel-size: 0.35x0.35x1.5 mm), as previously described⁵. All segmentations were manually conducted in the axial plane within the BG-region of interest, using ITK-snap. To assess the spatial relation of EPVS and arteries versus veins, the 7T QSM (venography) and 3T T2-weighted-sequence (EPVS) were registered to the 7T-ToF (angiography) with a rigid registration that utilized a multi-resolution mutual information model. For each case, we calculated Dice-scores and Hausdorff-distances between EPVS and arteries, as well as EPVS and veins, to quantify their spatial relation. Additionally, EPVS number, mean EPVS volume, and total EPVS volume per case was extracted. Furthermore, single EPVS were associated with arteries or veins, based on the overlap ratio of the vessel with the EPVS. Overlapping vessel segments were approximated by fitting an ellipsoid (www.scikit-image.org) (Figure 1A).
To calculate PI and V_mean in the small penetrating vessels of the BG, we used a previously described 2D phase-contrast (PC) sequence at 7T (voxel-size = 0.3x0.3x2 mm), in combination with multi-step image processing⁷.
Six autopsy cases (two representing each subgroup) were included from an autopsy-cohort of the Massachusetts General Hospital². Tissue blocks of the BG were sampled from formalin fixed hemispheres, embedded in paraffin, and 6µm-thick adjacent sections were cut. Adjacent sections were stained with luxol-fast-blue (LHE) and underwent immunohistochemistry for smooth muscle actin (SMA). EPVS were identified on LHE-stained sections, and the corresponding vessel characterized as vein or artery on the adjacent SMA-stained section (Figure 1B).

Results

On the MRI, the Dice-score between EPVS and arteries was significantly higher than the one between EPVS and veins (Kruskal-Wallis H-Test: Mdn_vein=0.01, Mdn_artery=0.033, W=847, p<.0001; Figure 2A). The Hausdorff-distance between EPVS and arteries was significantly lower than between EPVS and veins (Kruskal-Wallis H-Test: Mdn_vein=0.921mm, Mdn_artery=0.351mm, W=-861, p<.0001; Figure 2B). Finally, we observed significantly more EPVS overlapping with arteries than with veins (Kruskal-Wallis H-Test: Mdn_vein=10, Mdn_artery=34, W=861, p<.0001; Figure 2C). Non-CAA CSVD patients had a higher number and higher mean EPVS volume compared to the other subgroups. In the histolopathology, we observed that the frequency of the EPVS related to arteries (254/264, 96%) was higher than that of EPVS related to veins (24/264, 4%) (Chi-Square test: n=619, χ²(1)=34.0, p<.001; Figure 2D).
PI in the perforating arteries of the BG tended to be positively associated to total number of EPVS (Spearman's correlation: ρ=0.29, p=.080), and to total EPVS-volume (Spearman's correlation: ρ=0.31, p=.063). No relationship was found between EPVS-measures and V_mean.

Discussion

In the BG, our results show a closer spatial proximity of EPVS to arteries than to veins in high-resolution MRI (Figure 3D). In histopathology, EPVS were more frequently located around arteries. These findings are in line with previous works reporting spatial correlation of EPVS with arteries within centrum semiovale^2,4, but no association with veins³. Conversely, the relationship between PI, V_mean and EPVS-measures showed no conclusive results. Our findings contribute to close knowledge-gaps about the development of EPVS and potentially help to decipher the routes of perivascular clearance in the human brain.