Introduction
White matter hyperintensities (WMHs) on MRI scans are one of the hallmarks of cerebral small vessel disease (CSVD), which are associated with cognitive impairment, dementia, stroke and even death [
1‐
3]. However, the pathological mechanisms leading to WMHs remain unclear. WMHs are the earliest and most frequent observed imaging feature in cerebral autosomal-dominant arteriopathy with subcortical infarcts and leukoencephalopathy (CADASIL) [
4] which is a monogenic CSVD caused by Notch 3 gene mutations [
5]. Unlike sporadic CSVD, the occurrence and progression of WMHs are less dependent on cardiovascular risk factors and age in CADASIL [
6]. Therefore, CADASIL could serve as a unique model for investigating the pathological mechanisms of WMHs.
Accumulating studies of CADASIL provide the evidence that WMHs are strongly related to increased extracellular fluid (ECF). For example, a neuropathological study has demonstrated that WMHs in the anterior temporal lobe of CADASIL reflect fluid accumulation in enlarged perivascular spaces (PVS), together with the loss of WM [
7]. A recent study of CADASIL patients reveals that changes in diffusion MRI signals within WMHs are mainly driven by increased extracellular free water [
8]. Furthermore, ECF are differentially distributed in the WMHs, with higher ECF in specific regions of WMHs such as anterior temporal lobe [
9]. However, the association between increased ECF and white matter (WM) microstructural changes in WMHs has not yet been determined. It has been argued that increased ECF would result in a build-up of substances toxic (e.g., plasma proteins) to the WM microstructure [
10,
11]. Although diffusion tensor imaging (DTI) is a well-established method for quantifying microstructural WM alterations in vivo, the model fitting would be largely influenced by the extracellular water and intravoxel crossing fibers which happen to be seen in WMHs [
12].
Recent diffusion MRI methods applying novel post-processing approaches can be used to achieve more direct in vivo characterization of both intra- and extracellular properties of WM tissue. For example, free water (FW) imaging using bi-tensor model could potentially separate the diffusion signals from extracellular FW and tissue compartment [
13]. Apparent fiber density (AFD) imaging using constrained spherical deconvolution (CSD) allows the resolution of crossing fibers in voxels containing multiple fiber orientations, and it could estimates the fraction of space occupied by a fiber bundle [
14].
In this study, in a cohort of patients with CADASIL, we characterized the ECF (indexed by FW) and fiber density (indexed by AFD) in WMHs using the two advanced diffusion approaches. We hypothesized that increased ECF would be associated with fiber microstructural changes in WMHs. Considering the heterogeneity of ECF distribution in WMHs, we classified FW in individual WMHs into four levels using quartiles and used lesion probability map (LPM) to understand the distribution of different levels of ECF in WMHs. We analyzed the relationships between different levels of ECF and microstructural diffusion metrics (including AFD and FW-corrected DTI metrics). Given that WMHs commonly accompany lacunes and microbleeds, we further explored the correlations between ECF in WMHs with both lacunes and microbleeds.
Discussion
This study applied two advanced diffusion imaging methods to obtain deeper insight into the pathologic underpinnings of WMHs in CADASIL. In line with our hypothesis, we found an association between increased level of ECF and fiber microstructural changes in WMHs. Furthermore, we found that ECF was independently associated with both lacunes and microbleeds.
The etiology of WMHs involves complex mechanisms such as ischemia [
27], blood–brain barrier (BBB) leakage [
28], venous collagenosis [
29] and impaired perivascular drainage [
11]. Most of these mechanisms could lead to interstitial fluid (ISF) accumulation in extracellular spaces and the role of ECF in WMHs has been consistently reported in recent studies. For example, increased aortic stiffness could promote an increase in extracellular water content in WM, eventually leading to the development of WMHs [
30]. While the strong relationship between collagenosis of the deep medullary veins and periventricular WMHs has been confirmed in radiological-pathological correlation studies [
29,
31,
32], our recent study further suggests that their association is mediated by increased ECF [
33]. Moreover, enlarged PVS is hypothesized to represent impaired drainage of ISF from the brain [
34] and the fact that WMHs may grow from dilated PVS also suggests significant contribution of ECF accumulation [
35].
To understand the distribution of different levels of ECF within WMHs in our cohort of CADASIL patients, we classified WMHs into four subregions based on the quartiles of FW and mapped the lesion probability of WMHs at each voxel for each quartile of FW. We found that the distribution of ECF in WMHs varied with its levels. Low-level ECF locates mainly at the deep WMHs in frontal and parietal lobes with a diffusely distributed pattern, whereas high-level ECF preferred to locate at the periventricular WMHs. Traditionally, the regional localization of WMHs such as periventricular and deep WMHs would be influenced by different features of vascular anatomy [
36]. However, this theory seems difficult to explain the regional localization of ECF because a high level of ECF is also found in the deep WMHs of temporal lobe (i.e., anterior temporal lobe) which is a characteristic region in CADASIL. Our finding of increased ECF in WMHs of the anterior temporal lobe is consistent with a previous neuropathological study showing that increased ECF is related with enlarged PVS [
7] which is an imaging marker of impaired ISF drainage [
34]. Therefore, it can be speculated that regional variation of increased ECF in WMHs may reflect the degree of reduced capacity for ISF drainage from the WM.
Under normal conditions, the drainage of ISF and soluble metabolites from brain tissues by the perivascular pathways is necessary to maintain homeostasis in the brain [
37]. A pathological hallmark of CADASIL is the early depositions of granular osmiophilic material (GOM) adjacent to basement membranes of arteries, arterioles, and capillaries [
38]. It has been suggested that GOM plays a key role in hindering perivascular drainage [
38]. Furthermore, the degeneration of vascular smooth muscle cells (VSMCs) in CADASIL reduces vascular tone and amplitude of pulsations for the propagation of perivascular lymphatic drainage, resulting in further accumulation of ISF [
39]. The accumulation of ECF could cause a build-up of harmful substances such as plasma proteins from a leaky BBB [
10], which are toxic to surrounding WM microstructures including myelin and axon. From a pathophysiological perspective, increased ECF possibly occurs first, followed by demyelination and axonal damage within WMHs.
The findings from our study may provide support for the above-mentioned theory. First, microstructural changes (significant correlations between FW and AFD, significant decrease in FAt and ADt, significant increase in MDt and RDt) only existed in WMHs subregions with high-level FW (FWq3 and FWq4), implying that WM microstructural damage starts to rise when ECF reaches above the median value of FW. Second, disease duration had a negative correlation (marginally significant) with AFD but not with FW in FWq3 and FWq4, suggesting that fiber disruption occurred at relatively late disease stages. Besides, in a recent study in a large cohort of older people of similar age, MD showed the best differentiation of WMHs from NAWM, suggesting that altered water mobility in the interstitial space may be an early feature of WM pathology [
40]. Due to the cross-sectional nature of the current study, we could not rule out the possibility that WM microstructural degeneration results in the accumulation of ECF. However, if microstructural degeneration occurred before the event of ECF in WM, the atrophy of WM should be observed, like Wallerian degeneration in ischemic stroke [
41]. Conversely, a neuroimaging study showed that patients with CADASIL had larger WM volume than controls [
42], whereas WM atrophy was rarely reported in CADASIL.
Our findings of microstructural changes (decreased AFD, FAt, ADt and increased MDt, RDt) in WMHs were consistent with previous pathological findings [
43,
44]. Combined with findings of ECF distribution, microstructural damage in periventricular WMHs is more severe than that in deep WMHs.
The diffusion alterations in FAt and MDt have been described in a recent CADASIL study [
8]. To better characterize the microstructural damage related with axonal degeneration and demyelination, we used additional FW-corrected DTI metrics including ADt and RDt. Interestingly, ADt showed an opposite change in WMHs in contrast to AD, whereas other metrics showed the same direction in changes as the conventional DTI metrics. AD may increase with increased extracellular water content and reduced density of WM fibers, which allows faster water molecule movement parallel to axons [
45]. This explanation is supported by our findings of increased FW and decreased AFD. The decreased ADt with increased AD has also been reported in white matter regions heavily affected by age [
46]. Although our observation of decreased ADt and increased RDt in WMHs possibly respectively represent the pathological features of axonal degeneration and demyelination [
47], interpretation of them remains to be challenged by other pathological changes in WMHs like gliosis, neuroinflammation and vascular alterations.
Apart from widespread WMHs, other imaging markers including lacunes and microbleeds appearing later on brain MRI, are also related with cognitive impairments in patients with CADASIL [
48,
49]. We further found that increased ECF in WMHs was independently related to both lacunes and microbleeds, which was similar to our findings from sporadic CSVD [
50]. As previously mentioned, the vessel wall structure is already weak due to VSMCs degeneration in CADASIL. On this basis, the increased ECF could further aggravate the weakness of the vessel wall with compromised autoregulation of cerebral blood flow, thus resulting in insufficient perfusion and vascular rupture. Therefore, WMHs, lacunes and microbleeds may share a common pathogenesis. It should be noted that lacunes and microbleeds can in turn contribute to the increased ECF. ISF clearance in the brain is driven by arterial blood flow, while decreased perfusion caused by lacunes and microbleeds can reduce the efficiency of solute clearance by this pathway [
51,
52].
Several limitations should be considered. First, this is a cross-section study and thus, the relationships between diffusion metrics (e.g., FW and AFD) should be interpreted logically and cautiously. Future follow-up studies can be conducted to prove the effects of ECF on microstructural changes in WM. Second, analysis of PVS could offer complementary evidence about the impaired drainage of ECF in the brain, such as in the anterior temporal lobe. Unfortunately, due to lack of T2-weighted images, we did not perform the analysis of PVS. Third, there is currently limited histopathological evidence to specifically validate diffusion MRI metrics. Although measuring water distribution in postmortem tissue is challenging, future studies using novel methods to make direct comparisons between new diffusion metrics and pathology will be significant to improve the specificity of diffusion markers in CADASIL. Last, the sample size is relatively small. Nevertheless, the effect size in MANOVA test reaches medium to high, indicating the reliability of our findings.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.