The Association Between Glymphatic System Dysfunction and Cognitive Impairment in Cerebral Small Vessel Disease

The mechanism of cognitive impairment in patients with cerebral small vessel disease (CSVD) remains unknown. The glymphatic system dysfunction, which has been demonstrated to influence cognitive impairment, can be evaluated by diffusion tensor image analysis along the perivascular space (ALPS index). We explored whether cognitive impairment in CSVD is associated with glymphatic clearance dysfunction. In this study, 133 patients with CSVD were enrolled and underwent neuropsychological test batteries as well as magnetic resonance imaging (MRI). They were then categorized into a CSVD with cognitive impairment (CSVD-CI) group and a cognitively normal CSVD (CSVD-CN) group. The ALPS index and four CSVD markers [white matter lesions (WMLs), cerebral microbleeds (CMBs), lacunes, and perivascular spaces (PVSs)] were also assessed. Univariate analysis showed that the ALPS index was significantly different between the CSVD-CN (n = 50) and CSVD-CI groups (n = 83) (p < 0.001). This difference remained significant (95% CI < 0.001–0.133) after adjusting for six common risk factors (age, education, hypertension, diabetes, smoking, and alcohol abuse) as well as CSVD markers. The ALPS index was independently linearly correlated with global cognitive function, executive function, attention function, and memory after adjusting for the aforementioned six risk factors or CSVD markers. Our results suggest that glymphatic system impairment is independently related to cognitive impairment in patients with CSVD.


INTRODUCTION
Cerebral small vessel disease (CSVD) is common among older people. Cognitive impairment is one of the most important manifestations of CSVD. Vascular cognitive impairment and vascular dementia constitute the second most common cause of cognitive impairment. However, the factors causing cognitive impairment remain unknown. White matter lesions (WMLs) and lacunes, which are classical CSVD markers, are related to cognitive impairment in CSVD and could be used to predict cognitive impairment (Hamilton et al., 2021). However, in the clinic, some CSVD patients with mild WMLs have a severe cognitive impairment, which suggests that there are still some important factors that contribute to cognitive impairment in CSVD aside from traditional imaging markers and other imaging markers might enable predictions of cognitive impairment caused by CSVD.
Commonly, cognitive impairment in patients with CSVD occurs due to cerebral hypoperfusion or because other blood components permeate into the brain through a broken bloodbrain barrier (BBB). Both of these situations are harmful to neurons and neuroglia (Wardlaw et al., 2013a). In addition, many studies have found that amyloid-beta (Aβ) protein is deposited in the brains of patients with CSVD (Ye et al., 2015). In addition, the deposition of toxins leads to neuronal dysfunction or death in the brain and contributes to cognitive decline. In addition to Aβ, other toxins might also accelerate cognitive impairment.
Recently, the glymphatic system was discovered. With the help of Aquaporin 4, cerebrospinal fluid (CSF) flow in the periarterial space enters the brain and becomes interstitial fluid (ISF). It was then drained orderly through the perivenous space, meningeal or olfactory mucosal lymphatics, cervical lymphatic vessel, and finally returned to the peripheral venous. This CSF-ISF exchange system is called the glymphatic system, and its main role includes metabolic product transportation and metabolic waste elimination (Nedergaard and Goldman, 2020). In patients with Alzheimer's disease (AD) and animal models, dysfunction of the glymphatic system contributes to the deposition of Aβ and cognitive impairment (Reeves et al., 2020). CSVD leads to arteriosclerosis and vascular pulsation (Fu et al., 2006), which are the driving forces of the glymphatic system (Nedergaard and Goldman, 2020). In addition, some patients with CSVD have comorbid sleep disorders, and changes in sleep rhythm also impair the function of the glymphatic system (Hablitz et al., 2020;Chong et al., 2022). Our animal study verified the dysfunction of the glymphatic system in a rat model of CSVD (Xue et al., 2020). These studies indicated that damage to the glymphatic system might be another factor that leads to cognitive impairment in patients with CSVD.
It is difficult to evaluate the glymphatic system in the human body directly. In 2017, based on diffusion tensor imaging (DTI), Taoka et al. proposed the DTI analysis along the perivascular space (ALPS) index to evaluate the glymphatic system function. The ALPS index was calculated by the diffusivity at a site beside the lateral ventricle body at which projection and association fibers are orthogonal to the perivascular space (PVS; Taoka et al., 2017). More recently, Lou et al. compared this method with classical enhanced glymphatic magnetic resonance imaging (MRI) by the intrathecal administration of gadolinium, and they confirmed that the ALPS index could reflect glymphatic clearance function (Zhang et al., 2021).
To further explore the reason underlying cognitive impairment in patients with CSVD and identify other new imaging markers of cognitive impairment in patients with CSVD, we studied the relationship of the ALPS index and cognitive impairment in these patients.

Patients
Patients who visited the clinic at the Neurology Department of Huashan Hospital and met the inclusion and exclusion criteria below were prospectively consecutively enrolled in our study from Mar 2015 to Mar 2019.
Subjects were enrolled if they met the following criteria: (1) Fazekas 2-3 WMLs and at least one of the following three markers: lacune, cerebral microbleed (CMB), and PVS markers; (2) aged 50-80 years old. The exclusion criteria were as follows: (1) cerebral infarction with a diameter of infarction core larger than 2 cm; (2) cortical infarction; (3) stroke due to any potential cardioembolic source; (4) lacunar stroke syndrome within 6 months; (5) cerebrovascular stenosis with >50% luminal stenosis suggested by transcranial Doppler ultrasound, computed tomography angiography or magnetic resonance angiography; (6) genetically confirmed hereditary CSVD or suspected hereditary CSVD according to the family history and clinical features; (7) potential inflammatory and immunologically mediated CSVD according to clinical features and immune index; (8) WMLs that might be caused by other reasons except CSVD, such as multiple sclerosis; (9) cognitive impairment that might be caused by other diseases, such as neurodegeneration disease or hydrocephalus; (10) severe depression or other affective disorder that affects cognitive evaluation; (11) language disturbances that impair communication; and (12) MRI contraindications.
All demographic information, vascular risk information, and history of diseases or medicine were recorded. This study was approved by the Institutional Review Board of Huashan Hospital. All participants signed informed consent documents.

Cognitive Function Evaluation
All subjects underwent a neuropsychological test battery. All evaluations were administered by an experienced clinical psychologist who was blinded to the clinical and imaging data. The Montreal Cognitive Assessment (MoCA) was used for global cognitive function evaluation. Memory function was assessed by the total number of five recalls in the Auditory Verbal Learning Test (AVLT-sum), including three immediate recalls, one short delayed recall and one long delayed recall. Executive function was evaluated by the difference time value between the Trail Making Test B and A time [TMT (B-A)] and the total Verbal Fluency Test (VFT) score. Our VFT included animal, fruit, and vegetable categories, in which as many names as possible must be mentioned within 60 s. Attention function was assessed by the Symbol Digit Modalities Test (SDMT), language function was evaluated by the Boston Naming Test (BNT), and visuospatial function was assessed by the copy trial of Rey's Complex Figure  Test (RCFT-C). The cutoff MoCA scores to identify vascular cognitive impairment were 13/14, 19/20, and 24/25 for those who had received 0, 1-6, and 7 or more years of education, respectively (Lu et al., 2011). Based on the MoCA score, the cognitive diagnosis of the patients with CSVD was finally determined by the consensus meetings, which were attended by a neurologist, a clinical psychologist, a neuroradiologist, and a nurse.

Traditional CSVD Marker Evaluation
Cerebral small vessel disease markers were rated by two experienced neurologists who were blinded to the clinical data according to the Standards for Reporting Vascular Changes on Neuroimaging (STRIVE) criteria (Wardlaw et al., 2013b).
White matter lesions were defined as regions that were hyperintense on T2-weighted and T2-FLAIR imaging and were not hypointense on T1-weighted imaging. WMLs were rated visually on axial T2-FLAIR images using the modified Fazekas scale. The severity of WMLs in both the centrum semiovale and periventricular regions was rated as level 1 (mild), level 2 (moderate), and level 3 (severe; Pantoni et al., 2005). Then, the severity ratings of the two areas were summed.
Cerebral microbleeds (CMBs) were defined as hyperintense, round lesions that were smaller than 10 mm on SWI scans. CMBs were counted in the lobar, deep, and infratentorial regions, and then the total CMBs were recorded (Gregoire et al., 2009).
Lacunes were defined as round or approximately round lesions, 3-15 mm in diameter, that were hyperintense on T2weighted imaging, hypointense on T1-weighted imaging, and hypointense with a rim of hyperintensity on T2-FLAIR imaging without hyperintensity on SWI scans. The lacunes were counted in the lobar, deep and infratentorial regions. Then, the total number of lacunes was recorded (Pasi et al., 2017).
Perivascular spaces were defined as linear, round, or ovoid regions of signal with an intensity similar to that of CSF on all sequences and a diameter usually smaller than 3 mm. To distinguish PVSs from small lacunae, PVSs were defined as not having a hyperintense rim on T2-FLAIR images. The severity of PVSs in the centrum semiovale and basal ganglia was rated according to the number of spaces in a slice from the side containing the larger quantity of PVSs: level 0, none; level 1, 1-10; level 2, 11-20; level 3, 21-40; and level 4, > 40 (Potter et al., 2015). Either centrum semiovale or basal ganglia severity >2 was regarded as moderate to severe PVS (Banerjee et al., 2017).

ALPS Index Calculation
The free FMRIB software library (FSL) diffusion toolbox (FDT) was used to calculate the diffusion metric images. The ALPS index was calculated as described in a previous report (Taoka et al., 2017). First, a diffusion tensor at each voxel of the brain was estimated and outputted as a series of images, such as colorcoded fractional anisotropy (FA), mean diffusivity (MD), and diffusivity maps with x-, y-, and z-axes (Dxx, Dyy, and Dzz). These diffusion tensors were used to measure the diffusivity along the PVS direction by means of comparison with those of projection fibers and association fibers in a slice through the body of the lateral ventricle. At this level, the PVS direction was perpendicular to the ventricle wall. As a result, the PVS direction at this level was parallel to the x-axis (right-left), mostly in the axial plane. This direction is also perpendicular to the directions of both the association fibers (mostly on the yaxis) and the projection fibers (mostly on the z-axis). Thus, the diffusivity along the x-axis at areas with projection/association fibers at least partially represents the diffusivity along the PVSs. Then, a 5 mm diameter spherical region of interest (ROI) within the area of the projection fibers, the association fibers, and the subcortical fibers in the left hemisphere was placed on a colorcoded FA map of the plane at the level of the lateral ventricle body. Furthermore, the ALPS index that reflected the activity of the glymphatic system in individual cases was calculated. That is, the ratios of the mean x-axis diffusivity in the area of the projection fiber (Dxproj) and x-axis diffusivity in the area of association fibers (Dxassoc) to the mean y-axis diffusivity in the area of the projection fiber (Dyproj) and z-axis diffusivity in the area of association fibers (Dzassoc) were determined. In brief, ALPS index = mean (Dxproj, Dxassoc)/mean (Dyproj, Dzassoc).
In the area of the projection fibers, the dominant fibers run along the direction of the z-axis, and the x-and y-axes are perpendicular to the dominant fibers. Similarly, in the area of the association fibers, the dominant fibers run along the direction of the y-axis, and both the x-and z-axes are perpendicular to the dominant fibers. The major difference in the behavior of water molecules between the x-axis diffusivity in these areas (Dxproj and Dxassoc) and the diffusivity perpendicular to them (Dyproj and Dzassoc) was the existence of PVSs. Therefore, the ALPS index was calculated by measuring the diffusivity from the composite along the direction of the PVSs in perpendicular association fibers and projection fibers to reflect the function of the glymphatic system.

Statistics
For the statistical analysis, SPSS 18.0 (IBM Inc., Chicago, IL, USA) was used. For categorical variables, the differences between two groups were assessed by the chi-square test or continuity correction test. For continuous variables, the differences between two groups were compared using the Mann-Whitney U-test. Backward stepwise multivariable logistic regression models were used to evaluate variables that were independently associated with cognitive impairment. The area under the receiver operating characteristic curve (AUROC) was used to evaluate the ability of the model. Multivariate regression linear analysis was further performed to determine the factors that were linearly correlated with each neuropsychological test.

RESULTS
A total of 133 patients with CSVD were enrolled in our study. Among the 133 patients with CSVD, 83 had cognitive impairment (CSVD-CI), and 50 were cognitively normal (CSVD-CN). The demographic data are shown in Table 1. The patients with CSVD-CN and CSVD-CI averaged 64.45 ± 7.768 and 66.18 ± 7.575 years of age, respectively, and the length of their education averaged 10.21 ± 4.247 and 8.78 ± 4.749 years, respectively. Additionally, 64.0 and 63.9% of the participants in the CSVD-CN group and the CSVD-CI group, respectively, were male. The prevalence of hypertension in patients in the CSVD-CN and CSVD-CI groups was 87.8 and 92.9%, respectively, and the percentages of patients in the CSVD-CN and CSVD-CI groups with a history of diabetes mellitus were 17.15 and 25.0%, respectively. The smoking rate of patients in the CSVD-CN group was 15.0%, and that of patients in the CSVD-CI group was 23.2%. The alcohol abuse rate was 10% in the CSVD-CN group and 14.0% in the CSVD-CI group. There was no difference between the patients in the CSVD-CN and CSVD-CI groups in age, education, sex, hypertension, diabetes mellitus, smoking rate, or alcohol abuse rate.
As shown in Table 2, the backward stepwise multivariable logistic regression model for cognitive impairment eliminated age, education, hypertension, diabetes, smoking, alcohol abuse, the number of lacunes, the number of CMB, and moderate to severe PVS; in the end, only the ALPS index and WMLs remained in the model. In this model, both the ALPS index (p = 0.005) and WMLs (p = 0.003) were significantly different between the patients in the CSVD-CN and CSVD-CI groups. Furthermore, as shown in Figure 2, we calculated the AUROC of this model combining the ALPS index and WMLs. The AUROC of the model, at 0.780, was greater than the AUROC of the ALPS index alone (0.694) or WMLs alone (0.714).
To further explore the relationship between the ALPS index and cognitive impairment in patients with CSVD, using multivariate regression linear analysis our results showed that after adjusting for vascular risk factors of age, education, hypertension, diabetes mellitus, smoking, and alcohol abuse, the ALPS index was positively related to the MoCA (p = 0.005), score in the AVLT-sum (p = 0.002), the right number of SDMT (p < 0.001), and the VFT score (p = 0.008), and was negatively related to TMT (B-A; p = 0.010) ( Table 3). However, there were no significant linear associations among the BNT (p = 0.814), RCFT-C (p = 0.526), and ALPS index (Table 3). Furthermore, after adjusting for the severity of WMLs, the number of lacunes, the number of CMBs and moderate to severe PVS, in the multivariate regression linear model, as shown in Table 4, the ALPS index was positively related to the MoCA (p = 0.048), AVLT-sum (p = 0.014), and SDMT (p = 0.001) and was negatively related to TMT (B-A; p = 0.041; Table 4). However, there were no significant linear associations among the VFT (p = 0.089), the BNT (p = 0.610), the RCFT-C (p = 0.663), and the ALPS index (Table 4).

DISCUSSION
Our results show that the ALPS index, which reflects the function of the glymphatic system, is related to cognitive impairment in  patients with CSVD after adjusting for vascular risks and four classical CSVD markers, WMLs, lacunes, CMBs, and PVS.
Our results show that the decline in the ALPS index is independently related to cognitive impairment in patients with CSVD. In 2017, the author who proposed the ALPS index found that a decline in the ALPS index was correlated with a decrease in Mini-Mental State Examination (MMSE) scores in patients with AD (Taoka et al., 2017). This result was confirmed by another AD-related study that included only 36 participants (Steward et al., 2021). Using a battery of neuropsychological tests spanning multiple domains, a recent study also showed that the ALPS index in individuals in the Parkinson's disease (PD)-MCI and PDdementia (PDD) groups was lower than that in the cognitively normal PD patient group . More recently, consistent with our results, a CSVD MRI study showed that the decline in the modified ALPS index in patients with CSVD and the modified ALPS index were independently related to MMSE scores in individuals in the CSVD cohort (Zhang et al., 2021).
The decline in the ALPS index with cognitive impairment is associated with the glymphatic system, which is reflected by this index and plays a vital role in the exchange between ISF and CSF in the brain. When ISF-CSF exchange is impaired, metabolic waste clearance, such as Aβ, in the brain is hindered, thus leading to neuronal damage and cognitive impairment (Wardlaw et al., 2020). In addition, the glymphatic system can carry immune cells (Mentis et al., 2021) and inflammatory factors (Carotenuto et al., 2021). As a result, dysfunction of the glymphatic system might result in neuroinflammation and neuronal damage (Carotenuto et al., 2021).
Our results show that both the ALPS index and WMLs are independent markers for CSVD cognitive impairment. The ALPS index calculated from DTI reflects the glymphatic system function, while WMLs are mainly a marker of demyelination. Thus, they represent different types of CSVD pathophysiology. However, the glymphatic system dysfunction is also associated with WMLs. First, the accumulation of toxic products and solutions caused by the glymphatic system dysfunction can damage the axon and lead to WMLs (Carotenuto et al., 2021). Second, arteriosclerosis and vascular stiffness in CSVD not only contribute to ischemia demyelination but also impair the impetus of the glymphatic system. In conclusion, the glymphatic system dysfunction and WMLs rooted in vascular damage have similar causes.
To the best of our knowledge, no study has analyzed the relationship between glymphatic system function and different cognitive domains in CSVD, and no study has assessed whether the ALPS index could be an independent marker to evaluate CSVD cognitive impairment. The present study used a multivariable logistic regression model to explore it.    Aging and hypertension are confirmed CSVD risk factors, and the years of education are vital cognitive assessment scores. A recent study found a connection between diabetes and the ALPS index (Yang et al., 2020). Smoking leads to endothelial damage (Chang et al., 2014), and the latter triggers a series of cerebral small vessel dysfunctions (Wardlaw et al., 2013a). All those factors were added into the multivariable logistic regression model even though they did not reach statistical significance in univariate analysis. Our results show that the ALPS index, after vascular risk or traditional CSVD imaging marker adjustment, is linearly correlated with executive, attention, and memory function but is not correlated with language and visuospatial function.
The following principles may explain why the connection between ALPS index decrease and cognitive decline in CSVD is selective. First, vascular cognitive impairment selectively affects specific cognitive domains rather than homogeneously affecting all domains; numerous studies have demonstrated that executive function, attention, and memory are the main cognitive domains that are impaired in CSVD (Hamilton et al., 2021;Schroeter, 2021). Second, in murine studies in which a tracer was injected into the cerebellomedullary cistern, the tracer arrived in the dorsal and ventral brain asynchronously (Iliff et al., 2013a;Zeppenfeld et al., 2017;Mortensen et al., 2019). This phenomenon suggests that the speed of metabolic clearance based on the glymphatic system varies in different areas. The difference ultimately influences cerebral metabolism. Third, the glymphatic system derives motives from arterial pulsation (Iliff et al., 2013b;Mestre et al., 2018). Differences in arterial remodeling and hemodynamics of superficial and deep arterioles may influence glymphatic circulation in corresponding brain areas (Blanco et al., 2017;Spence, 2019). This difference finally has an effect on cerebral metabolism. Last, the ALPS index data are based only on measurements taken beside the body of the lateral ventricle, which includes a variety of connections to different cortical areas. This may affect the validity of cognitive domain evaluations.
Our study had some limitations. First, even though the emerging ALPS index has been confirmed by a study comparing the ALPS index method to the gold standard method, the ALPS index method is not a direct way to evaluate the glymphatic system and needs to be verified by more studies in the future. Second, this study was a single-center study, and the sample size was relatively small. This might be the main reason why the severity of PVS or the number of lacunes in patients in the CSVD-CN group was not significantly different from that in patients in the CSVD-CI group. Third, considering the relatively small sample size, we did not adjust for all vascular risk factors in the multivariate analysis. This may influence the reliability of our research.

CONCLUSION
In conclusion, our study has shown that a decrease in the ALPS index is independently associated with executive, attention, and memory function impairment in patients with CSVD, and further studies should explore whether the ALPS index could predict cognitive decline in these patients.

DATA AVAILABILITY STATEMENT
The datasets presented in this article are not readily available because, we do not have permission to share them. Requests to access the datasets should be directed to JF, jianhuifu@126.com.

ETHICS STATEMENT
This study was approved by the Institutional Review Board of Huashan Hospital. The patients/participants provided their written informed consent to participate in this study.