Background
The glymphatic system connects perivascular spaces (PVS) over which cerebrospinal fluid (CSF) surrounding the brain exchanges with the interstitial fluid (ISF) within the parenchyma [
1]. Its essential function is to take away wastes from the brain parenchyma. Detailly, CSF enters brain extracellular space via PVS, exchanges with ISF, and finally efflux via PVS again, taking away brain metabolic wastes. The failure of the glymphatic system is recently proposed as the final common pathway to dementia because it contributes to the aggregation of proteins, activates neuroinflammation and facilitates a positive feedback loop between wastes accumulation and further glymphatic dysfunction, and finally causes brain injury and dementia [
2]. While the association between glymphatic dysfunction and dementia is widely acknowledged, few studies investigated factors related to glymphatic impairment.
Alzheimer’s disease (AD) pathologies, especially Aβ aggregation, were previously demonstrated to be related to glymphatic dysfunction in AD animal models [
3‐
5]. While some studies suggested glymphatic dysfunction as the initial event precedes Aβ aggregation [
3,
6], others found that glymphatic function compromised in the early stages of Aβ deposition in the APP/PS1 model [
5,
7]. One recent study that used an in vivo imaging marker to evaluate the glymphatic function validated the relationship between Aβ accumulation and glymphatic dysfunction [
8]. This study found that the impaired in vivo glymphatic marker preceded Aβ accumulation. Apart from AD pathologies, a damaged glymphatic function is also consistently reported in CSVD-related animal studies [
9‐
11]. Hypertensive rats, the most common model of CSVD, showed suppressed glymphatic function [
9], and the multiple microinfarct murine model also demonstrated global glymphatic pathway impairment [
11]. In vivo studies using imaging markers for CSVD, mostly white matter hyperintensities (WMH) burden, found its spatial correlation with enlarged PVS and other non-invasive glymphatic markers [
12,
13]. The effect of AD pathologies and CSVD on glymphatic function is separately proved in AD and CSVD models, respectively. Few studies have investigated the independent association of CSVD and AD pathologies with glymphatic function in vivo.
AD pathologies co-existing with CSVD are prevalent in aging and the most common pathogenesis for dementia. Of patients diagnosed with AD, > 80% were with CSVD [
14].
Therefore, it is important to reveal the effects of Aβ and CSVD on glymphatic function from cognitively normal to dementia, if Aβ and CSVD independently contribute to glymphatic dysfunction or interactively damages glymphatic function.
To evaluate glymphatic function in vivo, we employed three non-invasive methods. Enlarged PVS is a commonly used approach to evaluate glymphatic function. It provides valuable information about the state of PVS, with severe enlargement indicating glymphatic dysfunction [
15,
16]. The “Diffusion Tensor Imaging Analysis along the Perivascular Space” (DTI-ALPS) method, initially proposed by Taoka et al. [
17], quantifies the diffusion of water within PVS along deep medullary veins and has robust correlations with glymphatic clearance, as determined by dynamic contrast-enhanced imaging [
18]. Sleep, a crucial factor influencing glymphatic function, has been linked to DTI-ALPS [
19,
20]. Furthermore, DTI-ALPS is closely associated with vascular risk factors [
21,
22], which are linked to a decreased driving force for glymphatic function [
9,
23,
24]. This method has also been applied to quantify in vivo glymphatic activity in various neurological disorders, including idiopathic normal pressure hydrocephalus (iNPH) [
25], Parkinson’s disease [
26], and multiple sclerosis [
27] (MS), consistently revealing that reduced DTI-ALPS values are associated with disease severity. Choroid plexus volume, suggested as a driving force behind glymphatic function, has demonstrated associations with disease severity in conditions like frontotemporal dementia (FTD) [
28] and AD [
29] and has been used to predict conversion to Parkinson’s disease dementia [
30].
Therefore, in this study, we aim to investigate (1) the association of CSVD and Aβ with glymphatic markers focused on AD continuum participants from cognitively normal to dementia and (2) whether glymphatic markers mediate the relationship between CSVD/Aβ and cognitive performance. CSVD was defined using WMH, a widely recognized and frequently utilized imaging marker for CSVD in the context of AD [
31,
32]. We speculate that both WMH and Aβ are independently associated with glymphatic impairment and could damage cognitive performance via glymphatic dysfunction.
Discussion
This study found that (1) both CSVD and Aβ burden were independently associated with DTI-ALPS and choroid plexus volume; (2) cognitive performance in most domains was associated with DTI-ALPS and choroid plexus volume; and (3) more importantly, DTI-ALPS mediated the relationship between CSVD/Aβ and cognitive performance.
A series of studies have demonstrated the association between Aβ and glymphatic function [
3,
5,
6,
39]. Our study using in vivo imaging methods found that decreased DTI-ALPS and increased choroid plexus volume were related to increased Aβ aggregation. However, we did not find the correlation between PVS enlargement severity and Aβ aggregation, which was consistent with previous findings [
40,
41]. DTI-ALPS assesses diffusivity along deep medullary veins and is suggested as a marker for evaluating glymphatic clearance function [
17]. Choroid plexus volume plays a crucial role in the production of cerebrospinal fluid (CSF) and may act as a driving force for the glymphatic system [
13]. Enlarged PVS observed on MRI is mostly situated around arterioles, which might cause hydrodynamic changes in CSF inflow and exchange with parenchymal fluid [
42]. The differential association of these markers with Aβ aggregation may suggest distinct contribution of the sub-processes in the waste clearance pathway. Nonetheless, it could also be due to different sensitivities of these MRI methods. Further validations are still needed.
By plotting the changes of DTI-ALPS and choroid plexus volume against amyloid accumulation along AD continuum, we found that Aβ accumulated from CN − to MCI + and flattened afterwards, whereas both DTI-ALPS and choroid plexus volume exhibit consistent changes from CN + to AD + . Notably, the curve representing Aβ accumulation exhibited an earlier onset compared to the changes in DTI-ALPS and choroid plexus volume. These results seem to be inconsistent from previous assumptions that amyloid deposition should be a consequence of glymphatic dysfunction. Indeed, the evidence concerning the association between glymphatic dysfunction and Aβ accumulation has primarily been derived from comparative studies. For example, APP mice showed decreased glymphatic function than wild mice [
5,
43], and late-onset AD participants demonstrated glymphatic impairment compared with normal controls [
39]. However, it is essential to acknowledge that comparing AD and non-AD is not sufficient to conclude that glymphatic dysfunction results in Aβ aggregation or vice versa. The complexities of this relationship and the direction of causality indeed call for further investigation to achieve a more comprehensive understanding. Some studies using animal models with deficiency of AQP-4 [
3,
6], which was suggested as the most important channel for glymphatic function, suggested that glymphatic dysfunction was followed by Aβ aggregation. However, it is important to consider that brain parenchymal Aβ plaques could also drive reduced perivascular localization of AQP-4 [
44,
45], and this would cause impaired glymphatic function. Besides, Aβ distributed in the walls of capillaries, and arteries could impair the motion for fluid movement along perivascular space [
46,
47], therefore decreasing the fluid diffusivity. Moreover, Aβ-related inflammatory reactions might increase resistance to fluid movement in perivascular space [
48]. Two studies also indicated the late-stage involvement of in vivo glymphatic markers [
8,
40]. One of these studies emphasized the relationship between AD pathologies and DTI-ALPS, suggesting that DTI-ALPS may be a later marker than AD pathologies. Another study, which focused on the severity of PVS burden, also did not find evidence to support the role of in vivo glymphatic markers in the early stages of AD pathogenesis [
40]. These collective findings emphasize the complex and multifaceted nature of AD progression, requiring a deeper exploration of the temporal relationships among various factors involved in the disease.
Our finding of the association between CSVD and glymphatic markers in the AD continuum was in line with previous studies focusing on CSVD patients [
13,
18,
21]. Brain-wide network of perivascular pathways made up glymphatic system [
49], so the damage of small cerebral vessels was closely correlated with glymphatic function. Fluid movement along perivascular space depends on the pulsation of the cerebral artery. Studies on cerebral arterial pulsation found that arterial stiffness was associated with glymphatic dysfunction [
23,
50,
51]. Also, venous collagenosis was suggested to be associated with decreased glymphatic clearance function [
52,
53]. Apart from the motion provided by cerebral small vessels, the blood–brain barrier (BBB) broken with increased perivascular inflammation factors and wastes could increase resistance to fluid motion. AQP-4 was also reduced in CSVD rat models [
10,
54]. Taken together, CSVD-related decreased driving force, increased resistance, and loss of water channel may all lead to damaged glymphatic function along the AD continuum.
We further found an association of DTI-ALPS and choroid plexus volume with cognitive performance in each domain. This result responded to the opinion that “glymphatic failure is a final common pathway to dementia” [
2]. Unlike previous studies, we considered the effect of CSVD on cognition that could be mediated by glymphatic function, in addition to the effect of Aβ along the AD continuum. We observed that the relationship of CSVD and Aβ with memory and language was mediated by DTI-ALPS, and the indirect effect coefficient was the largest in the memory domain. This finding corroborated the idea that AD dementia is multifactorial rather than just AD pathologies. Another interesting observation was that though we did not find glymphatic markers as the mediator between CSVD and visual-spatial performance and executive function, the total effects were significant, which meant that CSVD had another way to impact AD progression. This notion further supported the pivotal role of CSVD along the AD continuum [
55,
56].
Our study helps to further understand the role of the in vivo glymphatic index in AD. Firstly, unlike previous studies that only focused on AD pathologies in AD dementia or CSVD on vascular dementia, our study investigated the independent or interactive effects of AD pathologies and CSVD on AD continuum participants, consisting of participants from CN to dementia. The results of the independent effects of both pathologies suggested that controlling Aβ or CSVD could effectively improve glymphatic function. More importantly, our finding from the view of glymphatic function added the mechanism-based evidence for mixed AD dementia (AD pathologies combined with CSVD), in addition to prior studies focusing on their addictive role on whole brain functional or structural connectivity.
Our study had several limitations. Firstly, our evaluation of CSVD was solely based on WMH burden, without considering other important CSVD markers such as lacunes and microbleeds. It is important to acknowledge that WMH can have non-vascular origins, which might potentially introduce some bias into our results. However, it is worth noting that most pathogenic studies have primarily associated WMH with CSVD. Furthermore, the low prevalence of lacunes (12%) and microbleeds (7.5%) among our study participants had a substantial impact on the statistical power of our analysis. Secondly, our study involved participants from 17 different centers, which could introduce variability in DTI-ALPS. Due to the small sample size, it is difficult to compare participants from different centers. Nonetheless, as the acquisition protocol of ADNI has been harmonized by a specialized imaging core, the difference should have been minimized between these SIEMENS scanners. Lastly, our study was cross-sectional in nature. Although our hypothesis was supported by prior evidence, curve fitting along the AD continuum, and mediation analysis, a longitudinal design would be more suitable for a more comprehensive exploration of this question. These limitations should be considered when interpreting our findings, and they also point to opportunities for future research to address these issues and provide a more robust understanding of the relationship between CSVD, AD biomarkers, and glymphatic markers.
In conclusion, our study provided evidence that AD pathology (Aβ) and CSVD were associated with glymphatic dysfunction, which is further related to cognitive impairment.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.