An Investigation of Water Diffusivity Changes along the Perivascular Space in Elderly Subjects with Hypertension

BACKGROUND AND PURPOSE: Hypertension may be related to alterations of the glymphatic system, a waste metabolite drainage system in the brain. We aimed to investigate analysis along the perivascular space index changes in elderly subjects with hypertension. MATERIALS AND METHODS: Diffusion-weighted images were acquired from 126 subjects, including 63 subjects with hypertension (25 men and 38 women; mean age, 72.45years) and 63 age- and sex-matched controls (25 men and 38 women; mean age, 72.16years). We calculated the analysis along the perivascular space index as a ratio of the mean of x-axis diffusivities in the projection and association areas to the mean of y-axis diffusivity in the projection area and z-axis diffusivity in the association area. The left, right, and mean analysis along the perivascular space indices of both hemispheres were compared between the hypertension and control groups using a Mann-Whitney U test. The Spearman correlation coef ﬁ cient was used to assess the correlation between the left, right, and mean ALPS indices and blood pressure and pulse pressure. RESULTS: The left ( P ¼ .011) and mean ( P ¼ .024) analysis along the perivascular space indices of the hypertension group were signi ﬁ cantly lower than that of the control group. The left, right, and mean analysis along the perivascular space indices of all subjects were signi ﬁ cantly negatively correlated with blood pressure values ( r ¼ (cid:2) 0.200 to (cid:2) 0.278, P ¼ .002 – 0.046) and pulse pressure values ( r ¼ (cid:2) 0.221 to (cid:2) 0.245, P ¼ – 0.013). CONCLUSIONS: Our results are consistent with a model in which hypertension causes glymphatic dysfunction.

T he glymphatic system is a waste metabolite drainage system in the brain. [1][2][3] It consists of four sequential anatomic segments. First, the cerebrospinal fluid (CSF) flows along the perivascular space surrounding the penetrating arteries. Second, the CSF disperses into the interstitial space and transfers waste metabolites into the interstitial fluid (ISF) facilitated by aquaporin-4 water channels, which form the outer wall of the perivascular space. Third, the ISF-containing waste metabolites flow out of the large-caliber draining vein. 2 Finally, interstitial solutes exit the brain through meningeal lymphatic vessels, together with the venous sinuses. 4 Glymphatic dysfunction leads to the deposition of toxic waste products, such as amyloid b and tau proteins, which contribute to the pathogenesis of Alzheimer disease. [5][6][7] Toxic solute accumulation due to drainage dysfunction has been demonstrated in normal aging, 8 traumatic brain injury, 9 and stroke. 10 Hypertension (HT) is a risk factor for Alzheimer disease in the elderly. 11,12 HT involves pathologic alterations of small cerebral blood vessels and capillaries. 13,14 Mestre et al 13 found that the perivascular pump is less efficient in hypertensive mice due to dynamic changes in the vessel wall. Moreover, Mortensen et al 14 demonstrated suppression of glymphatic activity in spontaneously hypertensive rats, which reduced parenchymal waste transport.
On the basis of these findings, we hypothesized that patients with HT have an impairment of the glymphatic system. Alterations in glymphatic function due to HT in living human brains have not been reported to date because of the challenges encountered during in vivo tracer studies. Taoka et al 15 introduced the analysis along the perivascular space (ALPS) method, which is calculated using DWI as a noninvasive tool to evaluate the glymphatic system of living humans. This method assumes that diffusion plays an essential role in fluid transport in the brain parenchyma. We assessed ALPS index differences between elderly subjects with and without HT and evaluated the association between the ALPS index and blood pressure and pulse pressure.

MATERIALS AND METHODS
This study was approved by the ethics committee of Juntendo University in November 2015 and was conducted according to the principles outlined in the Declaration of Helsinki.

Study Participants
The Bunkyo Health Study is a prospective cohort study that has been running for .10 years. 16 We recruited elderly subjects 65-84 years of age who were living in Bunkyo-ku, an urban area in Tokyo. Our cohort comprised 1629 elderly people. Of these, 160 participants underwent both FLAIR imaging and DWI. All subjects were right-handed and had no history of diabetes (hemoglobin A1c , 6.5%), hyperlipidemia (total cholesterol , 240 mg/dL, low-density lipoprotein , 140 mg/dL, high-density lipoprotein . 40 mg/dL, and triglyceride , 150 mg/dL), or obesity (body mass index , 25 kg/m 2 ; Table 1). Subjects with systolic/diastolic blood pressure of $135/85 mm Hg during their first visit or those who had a history of using antihypertensive drugs were included in the HT group. We included 126 elderly participants, including 63 subjects with HT (25 men and 38 women; mean age, 72.45 [SD, 5.26] years) and 63 age-and sex-matched subjects without HT (controls; 25 men and 38 women; mean age, 72.16 [SD,5.11] years) in this study. The demographic characteristics of the study subjects are presented in Table 1.

MR Imaging Acquisition
We performed FLAIR imaging (TR ¼ 11,000 ms; TE ¼ 100 ms; TI ¼ 2000 ms; section thickness ¼ 5 mm) on a 3T MR imaging scanner (Airis Vento; Hitachi) in all participants. Deep white hyperintensity evaluation using the Fazekas scale based on axial FLAIR imaging was performed by an experienced neuroradiologist. DWI data were acquired on a 3T MR imaging scanner (Magnetom Prisma; Siemens) with a 64-channel head coil. Echoplanar imaging was acquired using a b-value of 1000 s/mm 2 along 64 isotropic diffusion gradients in the anterior-posterior phase-encoding direction with the following parameters: TR ¼ 3300 ms; TE ¼ 70 ms; FOV ¼ 229 Â 229 mm; matrix size ¼ 130 Â 130; resolution ¼ 1.8 Â 1.8 mm; section thickness ¼ 1.8 mm; acquisition time ¼ 7 minutes 29 seconds. Each DWI acquisition was completed with a b ¼ 0 image. We also acquired standard and reverse phase-encoded blipped images with no diffusion weighting (blip-up and blipdown) to correct for magnetic susceptibility-induced distortions related to echo-planar imaging acquisitions.

DWI Processing
DWI data were processed using the FMRIB Software Library, 17 Version 6.0 (FSL; www.fmrib.ox.ac.uk/fsl). Diffusion-weighted data were corrected for susceptibility-induced geometric distortions, eddy current distortions, and intervolume subject motion using the eddy and topup toolboxes. Diffusivity maps of each subject were acquired in the directions of the x-(right-left, Dxx), y-(anterior-posterior, Dyy), and z-axes (inferior-superior, Dzz). 18 Dxx corresponds to the direction of the deep white matter vessels in the periventricular white matter. Considering that the glymphatic system runs along these deep white matter vessels, the Dxx is assumed to partly reflect water diffusivity along the glymphatic system. Fractional anisotropy maps of all study participants were also generated and registered to the FMRIB58_FA standard space (https://fsl.fmrib.ox.ac.uk/fsl/fslwiki/FMRIB58_FA) using FSL's Linear Image Registration Tool (FLIRT; http://www.fmrib.ox.ac. uk/fsl/fslwiki/FLIRT) and nonlinear registration tool (FNIRT; http://fsl.fmrib.ox.ac.uk/fsl/fslwiki/FNIRT).

ROI Placement
For ROI placement, 1 subject (68-year-old female control participant) with minimal head movement (ie, with the smallest sum of squared difference) was selected. Using this subject's color-coded fractional anisotropy map, we manually placed 5mm-diameter square ROIs in the projection and association areas at the level of the ventricle bodies of the left and right hemispheres (Fig 1). In the projection area, dominant fibers run in the z-axis direction, perpendicular to both the x-and y-axes, whereas in the association area, dominant fibers run in the yaxis direction, perpendicular to both the x-and z-axes. The resulting ROIs were then registered to the same fractional anisotropy template. Finally, we manually checked the positions of the ROIs on each participant's fractional anisotropy image. Manual corrections were not performed because all ROIs were correctly placed.

ALPS Index Calculation
The diffusivity values of the x-, y-, and z-axes within the ROIs were obtained for each participant. The ALPS index was calculated as a ratio of the mean of the x-axis diffusivity in the projection area (Dxxproj) and x-axis diffusivity in the association area (Dxxassoc) to the mean of the y-axis diffusivity in the projection area (Dyyproj) and the z-axis diffusivity in the association area (Dzzassoc) as follows: DyyprojþDzzassoc . An ALPS index close to 1.0 reflects minimal diffusion along the perivascular space, whereas higher values indicate greater diffusivity. The left and right ALPS indices and the mean ALPS index of the left and right hemispheres were calculated.

Statistical Analysis
Statistical analysis was performed using SPSS Statistics, Version 27.0 (IBM). The left, right, and mean ALPS indices of the control and HT groups were compared using a nonparametric Mann-Whitney U test. The associations between the left, right, and mean ALPS indices of all subjects and the values of systolic blood pressure, diastolic blood pressure, average blood pressure, and pulse pressure were evaluated using Spearman correlation coefficients. A P value ,.05 was considered statistically significant.

RESULTS
The features of the control and HT groups are summarized in Table 1. The control and HT groups did not differ significantly in age, sex, hemoglobin A1c, total cholesterol, high-density lipoprotein, or triglyceride values. As expected, the HT group had significantly higher (P , .001) systolic, diastolic, and average blood pressures and pulse pressures than the control group. Furthermore, the HT group had significantly higher body mass index (P , .001) and Fazekas scale scores (P , .013) than the control group. In contrast, the low-density lipoprotein value of the control group was significantly higher than that of the HT group (P , .043). Nevertheless, all subjects were without obesity or hyper-lowdensity lipoprotein cholesterolemia. Moreover, we confirmed no white matter hyperintensities in the ROIs.
The left (P = .011) and mean ALPS (P ¼ .024) indices of the HT group were significantly lower than those of the control group. However, the right ALPS index did not differ significantly between the HT and control groups (P ¼ .094). Table 2 shows the left, right, and mean ALPS index values of the HT and control groups. Figure 2 shows the box plots of the left, right, and mean ALPS indices between the HT and control groups. Furthermore, the left, right, and mean ALPS indices of all subjects were significantly negatively correlated with systolic blood pressure values (r ¼ À0.202 to À0.263, P ¼ .003-.023), diastolic blood pressure values (r ¼ À0.200 to À0.218, P ¼ .014-.046), average blood pressure values (r ¼ À0.231 to À0.278, P ¼ .002-.009), and pulse pressure values (r ¼ À0.221 to À0.245, P ¼ .006-.013). The scatterplots in Fig 3 show the correlation between the left, right, and mean ALPS indices and systolic blood pressure, diastolic blood pressure, average blood pressure, and pulse pressure values.

DISCUSSION
We evaluated elderly subjects with HT using the ALPS index, which is a potential biomarker for the assessment of glymphatic activity in the living human brain. Our results demonstrated that left and mean ALPS indices in the subjects with HT were significantly lower relative to controls, though no significant difference was observed between the HT and control groups for the right ALPS index. Our study also found negative correlations between the left, right, and mean ALPS indices and blood pressure, and the left, right, and mean ALPS indices and pulse pressure, which suggest that glymphatic dysfunction is related to HT.
HT increases the risk of cognitive impairment, vascular dementia, and Alzheimer disease in the elderly. 11,12,19,20 The accumulation of toxic solutes such as amyloid b and tau protein in the brain was shown to be an important neuropathologic mechanism of cognitive dysfunction in elderly subjects with HT. [5][6][7] Previous studies have indicated that the accumulation of waste metabolites causes glymphatic dysfunction, which leads to further deposition. 1-3 HT and arteriosclerosis have been shown to lead to pathologic changes in small cerebral blood vessels and capillaries. 13,14 Such abnormalities contribute to reductions in cerebral blood flow and volume, which cause white matter hypoperfusion. Moreover, endothelial cell abnormalities and blood-brain barrier dysfunction may contribute  to white matter impairment. In the present study, although we did not detect any white matter hyperintensities in the ROIs, the HT group had more advanced white matter impairment than the control group. Adler et al 21 demonstrated that blood-brain barrier disruption increases the permeability of the vessel wall and mobilizes inflammatory factors such as macrophages, lymphocytes, and complementary components, which may lead to myelin damage. Yamamoto et al 22 also reported dilation of the perivascular space, accompanied by myelin degeneration and defects in ISF drainage in cerebral autosomal dominant arteriopathy, with subcortical infarcts and leukoencephalopathy associated with extensive smallvessel disease. In addition, Kamagata et al 23 reported an association between arterial stiffness and white matter demyelination using magnetization transfer saturation imaging. Notably, myelin-rich tissue, such as white matter, has been shown to be more sensitive to ISF flow obstruction. 24 Therefore, small-vessel diseases may be associated with alterations of both white matter microstructure and drainage function. Furthermore, in small-vessel disease, because of the difference in the anatomic bifurcation of the left and right carotid arteries, the left carotid artery is considered more susceptible to strong pulse pressure directly from the aortic arch, which increases the likelihood of plaque formations and intima damage becoming more severe. 25 Thus, microangiopathy due to HT in the left cerebral hemisphere may be more marked than that in the right cerebral hemisphere. This possibility may explain why the ALPS indices of the left hemisphere showed a significant difference between the HT and control groups, which was not observed in the right hemisphere.
To the best of our knowledge, this is the first study to show changes in water diffusivity along the perivascular space in living patients with HT. Previously, the suppression of glymphatic function has been observed in spontaneously hypertensive rats using dynamic contrast-enhanced MR imaging, 14 which suggested that hypertensive conditions reduce the effectiveness of arterial pulsation as a driver of CSF-ISF exchange and impair glymphatic activity. Using ex vivo fluorescence imaging, Mestre et al 13 also demonstrated that the perivascular pump is less efficient in hypertensive mice due to changes in vessel wall dynamics. The perivascular pump driven by arterial pulsation is a powerful mechanism for fluid transport within the brain. 26 Therefore, the decreased function of the perivascular pump could lead to a reduction in glymphatic activity. In line with previous studies in hypertensive animals, our findings demonstrated the effect of HT on human glymphatic dysfunction. Our study observed weak correlations between the ALPS index and blood pressure and the ALPS index and pulse pressure. Bewick et al 27 reported that a weak correlation with a low R-value can be statistically significant in a large sample size. We used a relatively large sample size; thus, a weak correlation is still considered clinically relevant. High pulse pressure reflects an estimate of the stiffness of the large central arteries, and high blood pressure results from arteriosclerosis in the peripheral arteries. 28 Such arterial damage may lead to glymphatic dysfunction due to arterial pulsation and perivascular pump changes. 1,29,30 The clearance of waste metabolites is essential for tissue homeostasis and is mediated by the blood-brain barrier and the CSF-ISF exchange pathway. 31 The glymphatic system is a clearance mechanism associated with CSF and ISF dynamics, as observed in many tracing studies. Iliff et al 1 showed that CSF enters the brain along the cortical pial arteries by labeling the CSF via an injection of a fluorescent tracer into the cisterna magna CSF in mice. CSF dynamics have also been observed using intrathecal or intravenous injections of gadolinium-based contrast agents as tracers. 3,32-39 However, the invasive injection of gadolinium-based contrast agents carries risks. Intrathecal administration of high-dose gadolinium-based contrast agents may cause gadolinium deposition in the globus pallidus and dentate nucleus 40,41 or serious gadolinium encephalopathy, which includes nausea, dyspnea, subjective chills, delirium, dysarthria, spastic pain of the lower extremities, limb ataxia, and gaze-evoked nystagmus. [42][43][44] Moreover, the tracking method requires several hours to track the distribution of a tracer in the brain, and monitoring glymphatic activity in real-time is challenging.
In contrast, the ALPS index is a noninvasive tool to assess the glymphatic system using DWI, which has a short acquisition time. As mentioned previously, the ALPS method is based on the hypothesis that diffusion plays an important role in fluid transport in the parenchyma. Fluid transport was believed to be caused mainly by cardiac pulsatility. 13,29 However, Smith et al 45 argued against this theory and suggested that diffusion played an important role in the transport of neurofluids. Several reports using mathematic modeling have further demonstrated the involvement of diffusion and advection in fluid transport in the parenchyma. 46,47 Martinac and Bilston 26 showed that diffusion is most likely the primary mechanism that drives fluid transport into the interstitial space. Taken together, diffusivity along the perivascular space could, at least in part, underlie glymphatic activity. Alterations of the glymphatic system in Alzheimer disease, 15,48 idiopathic normal pressure hydrocephalus, 18 diabetes, 49 and Parkinson disease 50 have been shown in living humans using the ALPS method, which is in line with the findings of studies using gadolinium-based contrast agents as tracers. 3,39,51 Additionally, Zhang et al 52 recently reported that the ALPS index is significantly associated with the glymphatic clearance function calculated by glymphatic MR imaging, following intrathecal administration of gadolinium in patients with small-vessel disease (r ¼ À0.772 to À0.844, P , .001). The report strongly supported the clinical use of the ALPS index as a glymphatic biomarker.
This study has several limitations. First, it consisted of a relatively large number of participants compared with previous studies using the ALPS index; however, we obtained DWI data from only a single institution. Future research should include multisite data to evaluate the clinical utility of the ALPS index. In addition, given that a slight difference in the subject's head position can influence the ALPS index, we encourage investigations evaluating the influence of head motion on the calculation of the ALPS index. Second, we did not consider other physiologic statuses such as perfusion or pulsatile motion of the brain. Finally, Yokota et al 18 showed that the size of the ROI may influence the calculated ALPS index, in which larger ROIs were more effective than smaller ROIs. Thus, further studies evaluating different sizes and shapes of ROIs for calculating the ALPS index are needed.

CONCLUSIONS
Our results are consistent with a model in which HT causes glymphatic dysfunction.