The Blood–Cerebrospinal Fluid Barrier May Play a Role in Alzheimer Disease Pathogenesis

See also the article by Choi et al in this issue.

Dr Chiang is an associate professor in the Department of Radiology, Division of Neuroradiology, at Weill Cornell Medical College/NewYork-Presbyterian Hospital. Her research combines quantitative MR and PET imaging techniques with plasma and CSF biomarkers to elucidate the pathophysiology of neurodegenerative diseases. Dr Chiang has been principal investigator and co-investigator on several NIH grants, has served on NIH study sections, and serves on the editorial boards of AJNR and JON.

Alzheimer disease (AD) remains the most common cause of dementia, characterized pathologically by extracellular β-amyloid (Aβ) plaques and aggregates of intracellular hyperphosphorylated tau proteins. However, AD pathogenesis is complex, multifactorial, and incompletely understood, which has complicated the search for effective therapies. Some consider the disease to be a mixed proteinopathy of Aβ and tau and that dysfunctional clearance pathways lead individuals to accumulate these toxic proteins (1), causing subsequent neurodegeneration and cognitive decline.

Perhaps one of the most exciting scientific breakthroughs in the past decade has been the discovery of the glymphatic system (2). The glymphatic system is believed to be a mechanism to clear waste from the brain, based on the flow of cerebrospinal fluid (CSF) into the brain along arteries, rapid exchange with interstitial fluid, outflow along veins, and finally drainage into lymphatic vessels in the neck. As soluble waste proteins and metabolic byproducts accompany this flow of CSF and interstitial fluid, they are cleared out of the central nervous system. CSF, therefore, is believed to play a major role in the brain’s clearance of toxic proteins.

Although breakdown of the blood-brain barrier has garnered much attention in numerous neurologic disorders, including AD, breakdown of the blood-CSF barrier (BCSFB) is a newer concept. CSF is produced and secreted by the choroid plexus (CP), which consists of a single layer of ciliated epithelial cells, fluid-filled connective tissue, and fenestrated blood vessels located in the lateral, third, and fourth ventricles. The fenestrated vessels allow for high permeability, enabling CP epithelial cells to produce CSF from the blood (3). However, tight junctions line the apical surface of the epithelial cells, preventing potentially toxic blood-borne substances from leaking easily into the ventricular system, thereby forming the BCSFB. At the BCSFB, there is an active flow of sodium, chloride, and bicarbonate ions into the ventricular system via ion channels and cotransporters, leading to an osmotic gradient that draws in water (4). With age, inclusions form within the CP epithelial cells, termed Biondi bodies, and there is thickening of the basement membranes, which may lead to decreased CSF secretion by the CP (5).

In this issue of Radiology, Choi et al (6) used various imaging techniques to characterize differences in the CP in patients at different stages of cognitive impairment. The authors retrospectively included 532 patients—78 with subjective cognitive impairment (SCI), 158 with early mild cognitive impairment (MCI), 149 with late MCI, and 147 with AD—who had undergone 3.0-T brain MRI between January 2013 and May 2020. Patients with MCI and AD met appropriate clinical criteria for classification. Patients with SCI did not meet criteria for MCI or AD but reported a subjective cognitive symptom. Patients with early MCI versus those with late MCI were distinguished based on a Clinical Dementia Rating of less than 1.5 versus equal to or greater than 1.5, respectively. Patients were between the ages of 45 and 90 years, and 73% were women.

First, Choi et al (6) investigated differences in CP volumes across disease stages, using automated volumetry software based on the FreeSurfer 6.0 platform to segment the T1-weighted MRI scans. Notably, there was a stepwise increase in CP volume with severity of disease stage, with the SCI group having the smallest CP volumes and AD group having the largest. This increase in CP volume with disease severity was significantly associated with decreased cortical and hippocampal volumes, which are more traditional markers of atrophy in MCI and AD. The authors hypothesize that the large CP volumes in AD could reflect stromal fibrosis, basement membrane thickening, and/or inflammatory changes. Interestingly, 17.7% of the patients had amyloid PET scans, and amyloid deposition levels were not associated with CP volume. On the other hand, CP volume was associated with vascular risk factors, such as hypertension, diabetes, and a history of cardiovascular disease, as well as periventricular white matter hyperintensity volumes.

Next, Choi et al (6) investigated differences in the permeability of the CP vasculature, using dynamic contrast-enhanced (DCE) MRI in a prospective arm of the cohort (25% of patients). DCE MRI is a technique that models the transit of an intravenous bolus of gadolinium-based contrast material through tissue, thereby providing information on tissue perfusion and permeability (7). Typical parameters reported from the model include the fractional plasma volume and the volume transfer constant (K^trans), which is reflective of permeability. K^trans was found to significantly decrease from MCI to AD, with no difference in plasma volume, suggestive of decreased permeability in the AD stage. This decreased permeability could be due to the factors listed earlier, such as vessel wall thickening and stromal fibrosis related to vascular risk factors. However, unlike the stepwise increase in CP volumes observed across stages, the group with the highest K^trans was actually the late MCI group. This raises the question of whether inflammation during the MCI stage could increase permeability and K^trans, which then decreases as that inflammation progresses to degeneration and fibrosis of the vasculature, connective tissue, and CP epithelial cells.

Choi et al then investigated differences in tissue iron levels and calcifications in the CP by using quantitative susceptibility mapping (QSM) (8). One would expect larger CP volumes in AD to reflect increased dystrophic calcifications and iron deposition (9). Surprisingly, there were no significant differences in QSM across groups. Because the signal intensity of calcifications is the opposite of that of iron at QSM, perhaps the effects were cancelled out in voxels that had both.

Finally, the authors found that larger CP volumes were associated with decreased memory, frontal executive function, and visuospatial function. This confirms that improved CP function does not accompany the increase in CP volume; rather, CP volumes reflect pathologic processes, as seen with cognitive decline and AD.

A few limitations deserve mention. First, the analyses were largely retrospective, so less than 20% of the study sample had amyloid PET data. Although patients met clinical criteria for MCI and AD, there is increasing awareness that many of these patients may not have Aβ and tau deposition. As a result, the findings in this study should be confirmed in a prospective cohort with deep phenotyping for the pathologic features of AD. Similarly, only a subset of patients were part of the prospective arm for DCE MRI, limiting sample size. Furthermore, CP volumes were obtained using noncontrast T1-weighted images, so areas of calcification, fibrosis, and possibly CSF were included in the volumes, which could decrease mean K^trans over the region. Contrast material may better delineate functional areas of CP. Finally, although CP volumes increased with disease stage, there was significant overlap among groups, limiting their diagnostic use in distinguishing patients without cognitive symptoms from those with MCI or AD. The predictive value of CP volumes also remains to be seen in longitudinal studies.

Nevertheless, this study contributes substantially to the growing literature on the role of the BCSFB in AD by investigating CP volumes, QSM, and vascular permeability across disease stages. In the future, one can envision therapies that potentiate CP function in older age, leading to preserved CSF secretion and potentially increased glymphatic flow to clear out toxic proteins. Besides producing CSF, the CP is also believed to have a more direct role in clearing Aβ. The transporters, lipoprotein receptor–related proteins 1 and 2, are located on the CP epithelium and are involved in Aβ clearance (5). The BCSFB also secretes the protein transthyretin, which can bind soluble Aβ to facilitate clearance. Indeed, transplanting CP cells into a mouse model of AD was found to decrease Aβ deposition and improve cognition (10). I am eager to see whether inventive ways to improve CP function in humans can replicate these findings. Multimodal targeted imaging approaches to monitor CP function will then be crucial.