We recently reported that the glymphatic pathway is a key contributor to the clearance of soluble amyloid β from the brain interstitium and proposed that the failure of this clearance might contribute to amyloid plaque deposition and AD progression (6
). In light of these findings, there may be great value in the development of a clinical prognostic test for measuring glymphatic pathway function throughout the human brain and evaluating whether suppression of this system contributes to the development and progression of AD. Here we provide proof-of-concept data demonstrating that glymphatic pathway function can be measured using a simple and clinically relevant imaging technique, contrast-enhanced MRI, to visualize brain-wide CSF-ISF exchange. We report that following intrathecal administration, contrast moves rapidly by bulk flow through para-arterial pathways to reach major influx nodes at the pituitary recess, the olfactory bulb, and the pineal recess. Furthermore, parametric and nonparametric data analysis of signal changes (T1 shortening) within different regions of the brain parenchyma clearly demonstrates that contrast movement into and through the interstitium is highly dependent upon the molecular weight of the paramagnetic contrast molecule. From the dynamic series of contrast-enhanced MRI images, we also developed and defined simple kinetic parameters to characterize glymphatic system function that reflect rates of CSF-ISF exchange throughout the brain.
We reported in our prior study that subarachnoid CSF rapidly enters the brain parenchyma along paravascular channels surrounding penetrating arteries (6
). Our present findings demonstrate that external brain surface arteries, such as the basilar artery, communicating arteries of the Circle of Willis, and olfactory arteries, constitute a rapid transport pathway for CSF within the wider subarachnoid space and ultimately the brain proper. Anatomically, these CSF transport pathways likely correspond to the leptomeningeal paravascular sheathes described in electron microscopy studies by Weller and colleagues (9
). Both small (<1 kDa) and large (~200 kDa) molecular weight contrast moved through these para-arterial spaces at comparable rates, suggesting that bulk flow is driving CSF flux transport along these pathways (7
). Analysis of intrathecal fluorescently labeled tracer influx into the rat brain confirmed that the bulk flow pathways surrounding the surface arteries are continuous with those surrounding smaller penetrating arteries and provide a direct route for the rapid influx of CSF into and through the brain interstitium. These para-arterial bulk flow pathways comprise the major component of a brain-wide system that facilitates the clearance of solutes and wastes from the brain interstitium.
The exchange of CSF and ISF between paravascular spaces and the interstitium occurs across perivascular astrocytic end feet, which extend nearly complete coverage over the brain microvasculature (8
). As a result, solutes that lack a specific molecular transport pathway (such as ion transporters or channels) across the end feet must instead pass through the approximately 20-nm cleft between overlapping end foot processes to gain access to the interstitial space. In our prior study in the mouse, we reported that small tracers such as TR-d3 (diameter of hydration [dH
] = 6.1 nm; ref. 14
) pass readily into and through the interstitium, whereas large tracers such as FITC-d2000 (dH
> 32 nm; ref. 14
) remained largely confined to paravascular spaces (6
). In this study, we confirm these findings in the rat using fluorescence-based imaging. In addition, we used intrathecal administration of 2 different sized contrast agents, Gd-DTPA (MW ~1 kDa) and GadoSpin (MW 200 kDa), to demonstrate a similar effect of tracer size by dynamic contrast-enhanced MRI. For example, when TACs for the 2 contrast agents are compared at different anatomical locations, the large molecular weight GadoSpin entered the brain parenchyma at a dramatically lower rate than did the small molecular weight Gd-DTPA (compare red and blue curves, Figure ). Moreover, an independent cluster analysis clearly revealed that while both tracers had ready access to paravascular spaces, the brain volume accessible to Gd-DTPA was markedly larger than that accessed by GadoSpin (compare blue zone 3 in Figure D with that in Figure H). Although we attribute the observed differences in CSF-ISF exchange patterns between Gd-DTPA and GadoSpin to differences in molecular size, we acknowledge that other possibilities exist. For example, it is conceivable that other intrinsic molecular properties, such as shape, electrostatic forces, size (radius of gyration), conformation, or physical/chemical interactions, can contribute to the dissimilar CSF-ISF exchange patterns observed with the 2 contrast molecules. Clearly, more investigations using a range of well-characterized contrast molecules will be necessary to fully clarify this issue.
One apparent discrepancy between fluorescence-based and contrast-enhanced MR-based imaging under the current experimental protocol is the inability to detect the small molecular weight contrast agent (Gd-DTPA) in the entire brain parenchyma after intrathecal injection. This is in contrast to the widespread parenchymal labeling with small molecular weight fluorescent CSF tracers in the mouse (6
). To confirm that these differences did not reflect species differences in CSF-ISF exchange, we conducted ex vivo fluorescence imaging in rat brain slices following intrathecal injection of similarly sized fluorescent CSF tracers (compare FITC-d500 [500 kDa] to GadoSpin [200 kDa] and compare TR-d3 [MW 3 kDa] to Gd-DTPA [MW ~1 kDa]) using an injection protocol that was otherwise identical to that used for the MRI experiments. This fluorescence-based imaging confirmed the permeation of small molecular weight CSF tracer throughout the rat brain parenchyma and restriction of the large molecular weight tracer to the paravascular spaces. We therefore conclude that Gd-DTPA, like TR-d3, does in fact move throughout the brain parenchyma. However, the Gd-DTPA contrast concentrations achieved under the current intrathecal infusion protocol within the broader parenchyma simply were not sufficiently high to induce detectable signal changes (T1 shortening).
Notwithstanding these limitations, the intrinsic 3D nature of MRI permitted the visualization of the entire paravascular CSF-ISF exchange pathway throughout the whole brain, allowing us to define key nodes of CSF influx into the brain at the pituitary recess, olfactory artery, and pineal recess and to assess CSF-ISF exchange at many anatomically distinct locations simultaneously (Figure ). Through the dynamic image acquisitions, the kinetics of CSF tracer influx into and clearance from the paravascular and interstitial spaces could be measured and characterized for a given contrast molecule. For example, TACs generated for the subarachnoid pituitary recess and pineal recess after Gd-DTPA injection exhibited similar influx profiles, whereas efflux from the pineal recess was markedly slowed compared with that of the pituitary recess (compare Figure A and Figure B). Similarly, TACs generated for the cerebellum and aqueduct showed similar influx kinetics, while the clearance of contrast from the aqueduct was prolonged compared with that from the cerebellar region (compare Figure C and Figure E). One possible explanation for the protracted clearance from the pineal recess and the aqueduct is that these anatomical regions constitute primary ISF clearance sites and thus appear to retain contrast (by accumulating it from other regions) for longer durations.
Brain-wide glymphatic pathways of CSF-ISF exchange, assessed by contrast-enhanced MRI in the rat.
Analysis of the dynamic time series data also allowed the essential comparison of CSF-ISF exchange between a small (Gd-DTPA) and a large (GadoSpin) molecular weight contrast agent (compare Supplemental Videos 1 and 2; Supplemental Methods). Here it is particularly important to note that, within the pituitary recess, Gd-DTPA and GadoSpin influx and clearance were virtually identical (compare red and blue curves, Figure A), whereas in the parenchymal pontine nucleus, GadoSpin movement into and through the region was dramatically reduced compared with that of Gd-DTPA (compare red and blue curves, Figure F). Using a second approach, cluster analysis, the spatial distribution pattern of CSF-ISF exchange between the 2 contrast agents can be evaluated at the level of the aqueduct. Comparing either the raw brain volume occupied by zone 3 (corresponding to the most distal compartment) or the normalized average zone 3no. voxels × AUC/zone 1no. voxels × AUC ratio (which reflects tracer penetration from the paravascular space into the brain parenchyma), GadoSpin penetration into and through the brain interstitium was dramatically restricted compared with that of Gd-DTPA (Table ). Thus, dynamic contrast-enhanced MRI following intrathecal contrast agent administration provides a novel and simple approach to characterize both the kinetics and spatial distribution of glymphatic CSF-ISF exchange throughout the whole brain.
In our prior study (6
), we reported that the glymphatic pathway is an important contributor to the clearance of interstitial solutes such as soluble amyloid β, a peptide widely believed to be a critical driver of AD pathogenesis (15
). Based upon those findings, we proposed that the failure of glymphatic pathway function may contribute to the deposition of amyloid β plaques and the progression of AD. A key motivation for evaluating glymphatic pathway function by contrast-enhanced MRI in rats was to lay the experimental groundwork for evaluating glymphatic pathway function in the human brain and in the future assess whether its failure contributes to AD progression in humans. To accomplish this, a safe, minimally invasive imaging approach to measuring glymphatic pathway function was necessary. Contrast-enhanced MR cisternography is currently used clinically to visualize CSF leaks in patients undergoing treatment for spontaneous intracranial hypotension and CSF rhinorrhea (4
), and the Gd-DTPA used in this study is already clinically approved for this purpose. While the current injection route via the cisterna magna and scan time required (>2 hours) is not clinically relevant, work within our group is underway to evaluate the suitability of single-shot intrathecal lumbar injections for evaluating glymphatic pathway function in the brain. We are additionally developing approaches to reduce data collection and scan time. However, this study demonstrates that, in the rat, this potentially clinically acceptable approach can be used to visualize and evaluate glymphatic pathway function, permitting assessment of kinetics and anatomical distribution patterns of paravascular CSF-ISF exchange throughout the whole brain. We propose that this approach may provide the basis for a wholly new prognostic strategy for evaluating AD susceptibility and disease progression in the future.