Introduction
The cerebral white matter (WM) has become an important focus for investigation of mechanisms in brain ageing and dementia. Age is the single most important risk factor for WM damage, which relates to increased white matter hyperintensities (WMH) on T2- weighted magnetic resonance imaging (MRI) and associated with vascular disease, disability, cognitive impairment and death [
9,
15,
18]. Cerebral hypoperfusion in older age is implicated as a key pathophysiological element, which contributes to diffuse WM changes. Both cerebral small vessel disease and carotid artery disease are risk factors for cerebral WM damage [
4,
9]. WM vascular pathology in brain ageing is observed with a frequency of 71% in non-demented versus 84% in demented subjects [
27]. WM vascular changes are most prominent in vascular dementia [
10]. However, all neurodegenerative dementias bear some degree of vascular pathology ranging from 61% in frontotemporal dementias to 82% in AD [
35]. A significant portion of such pathology is attributed to small vessel disease in the WM [
10]. Previous imaging and pathological studies indicated that the medullary arteries and WM of the frontal lobe are especially susceptible to haemodynamic derangement, leading to more severe WM damage, than for example in the temporal lobe, during ageing and vascular disease [
13,
17]. Obstruction of lumen at the proximal ends of penetrating arteries or becoming coiled or tortuous enroute would reduce blood pressure and impact on perfusion of the WM [
6]. Furthermore, age-related arteriosclerotic changes and segmental loss of vascular smooth muscle cells along lengths of both the medullary and perforating arteries disrupting flow in the distal arteries and affecting perfusion of the deep WM with the creation of an hypoxic environment [
12,
26].
The microvascular network within the WM though less dense than in the neocortex is equally vulnerable. In ageing rats, primates and man endothelial cells of cortical cerebral microvessels shrink with thickening of the basal lamina [
21,
24]. The endothelium is activated as indicated by increased expression of the intercellular adhesion molecule in relation to basement membrane collagen IV (COL4) [
12]. The activation is often accompanied by proliferation of microglia, which release proteases and free radicals to promote damage to components of the extracellular matrix that contain high amounts of collagen [
32]. The microvascular network undergoes severe distortions including tortuosity, coiling and kinking [
6], which increase with age and coincides with leukoaraiosis. Consistent with low tissue oxygen tension, within the deep WM there is induction of hypoxia-inducible factors 1alpha and 2alpha as well as key hypoxia-regulated proteins such as matrix metalloproteinase-7 and neuroglobin [
12]. These observations are consistent with elevated concentrations of the vasoconstrictor endothelin 1, reflecting abnormal regulation of WM perfusion [
5]. However, it is not clear at what threshold point diffuse WM damage implied by the microvascular changes tips over to affect cognitive function.
In this study, we focused on microvessels of the frontal WM as a key component of the gliovascular unit facilitating tissue perfusion [
16]. We concentrated on the frontal lobe because of its contiguity with the centrum semiovale region and relative vulnerability in cerebrovascular disease [
20,
21]. We specifically assessed capillary changes in relation to WM damage across various neurocognitive disorders including Alzheimer’s disease (AD), dementia with Lewy bodies (DLB), Parkinson’s disease with dementia (PDD), Vascular dementia (VaD) as well as post-stroke dementia (PSD).
Discussion
Analyses of different dementias using repeated robust methods of quantification in a large series of samples indicate several novel observations related to microvessels, largely capillaries of the WM compared to those of the cortex in ageing brain. We first report high degree of SVD pathology and capillary abnormalities in the WM in clinically assessed subjects with different neurodegenerative dementia diagnoses including DLB, PDD, Mixed dementias and AD. This was corroborated by the vascular pathology as well as WML scores. Whereas the highest burden of SVD pathology [
29] consisting of severe arteriolosclerosis, WM rarefaction, microinfarction and perivascular spacing was present in VaD and PSD, it was intriguing that subjects diagnosed with neurodegenerative dementias including PDD, DLB and AD exhibited similar SVD changes. Arteriolosclerosis and microinfarcts, which are strongly associated with cognitive impairment [
2,
20,
36], were consistent features in different dementia types. Capillary abnormalities including tortuous or coiled capillaries as well as collapsed string vessels were evident particularly in VaD and PSD cases, suggesting cerebral hypoperfusion or ischaemic changes are likely the cause of microvascular abnormalities. While these microscopic lesions are not apparent on T2-weighted or FLAIR sequences from MRI [
8,
9], they demonstrate they are present in tissue and may contribute to overall dementia diagnoses.
Quantification of microvascular pathology in the WM across the neurodegenerative dementias showed total COL4 immunopositive microvascular densities tended to be decreased by ~ 18% although they were increased by 52% in PSD compared to VaD subjects. We suggest the SVD type of pathological changes, particularly in the vascular dementias, reflects restructuring of the microvascular network by increasing capillary bed to counter hypoperfusion in the WM. Our results also suggest that subjects with more chronic state disease e.g. VaD versus PSD exhibit lower degrees of microvascular plasticity. When capillaries lose their endothelium by chronic cerebral hypoperfusion or ischaemic insults, the basement membrane is not degraded and form functionally useless string vessels that can no longer transport cells or plasma [
6]. Previous results show that 3 to 5 days after ischaemic insults, capillaries lose endothelial cells presenting accordion-like pleating of residual basement membrane and leading to thin acellular strands at 40 days post insult. Since some string vessels were found in ageing controls although to a much lesser extent than in all dementias with extracellular pathology [
6,
23], it suggests nominal vascular remodeling or restructuring also occurs in normal ageing brain. String vessel remnants can provide the skeletal structure for newly forming capillaries by invading capillaries forming new basement membrane within the old, presenting a duplicated basement membrane [
6]. The basement membrane provides the supporting structure of a microvessel composed of several thin layers of insoluble extracellular proteins including collagen (50%) and the framework for growth of endothelial cells [
30] and restoring the gliovascular unit [
16].
Our results showed that both in controls (physiological) and pathological conditions, capillary width was consistently larger by ~ 45% whereas microvascular density was lower by ~ 49% in the WM compared to the cortex. In other words, capillaries in the WM were wider and sparse, whereas capillaries in the cortex were narrower and dense. These observations were verified by a correlation analysis of the two markers, COL4 as well as GLUT-1. The strong positive correlation between microvascular width assessed by COL4 and GLUT-1 confirmed that the size of entire capillary including the vascular lumen is larger in the WM compared to the cortex and this was increased by ~ 20% in dementia states. The robustness of these observations was strengthened by our previous studies [
7] reporting strong correlation between COL4 and GLUT-1 immunostained microvascular length densities (L
v) in the hippocampus (r
2 = 0.687,
P = 0.000). We similarly noted that capillary widths in the WM of the temporal lobe were larger by 17–20% than those in the overlying cortex (
P < 0.01) and in dementia subjects including VaD, Mixed and AD (
P < 0.01) compared to controls (unpublished results). The WM versus cortical or gray matter differences in microvascular densities [
25] and capillary sizes appear specific and likely reflect cellular contents of the regions. Protection of neurons in the cortex requires greater supply of oxygen and nutrients reflecting higher microvascular densities whereas blood flow per unit length of WM capillary is increased. While similar conclusions may be made for GLUT-1 results as those for COL4, these novel findings with respect to dementia suggest there are compensatory mechanisms in the WM to maintain reserves of blood flow within capillaries and ameliorate cerebral hypoperfusion [
11,
16]. It is plausible that capillary dilation with the reflected structural modifications leads to a local increase in the number of erythrocytes of ~ 6 μm in diameter travelling in a single file [
33] as an adaptive mechanism to increase oxygen supply in the chronically hypoxic deep WM [
12]. That WM capillaries also tended to be increased in width in the PSND subjects suggested it was the presence of vascular pathology in the WM that likely instigates widening of the capillaries. However, we cannot refute that the wider vascular width in the cortex evident in AD or Mixed dementia subjects could be due to microangiopathy attributed to amyloid or other proteins, particularly adhering to capillaries [
19].
We emphasise that one of the main limitations of our study is that brain regions other than the frontal lobe were not systematically investigated for SVD pathology. We also did not quantify the densities of string or coiled vessels across all dementias. This is an extremely cumbersome undertaking and we deemed it would not improve the outcomes over the semi-quantitative results presented. While we concentrated on the deep WM of the frontal lobe in accord with our prior hypothesis [
16], we had previously demonstrated the spectrum of SVD pathology in different dementias including VaD, AD and DLB incorporating the temporal lobe and the basal ganglia [
10]. Given that the same neurodegenerative pathologies occur in PDD and other mixed dementias, it is reasonable to suggest similar microvascular or capillary changes occur in the WM of these dementias. Quantification of capillary widths in more regions of the brain can also be quite cumbersome. Still we deem such an undertaking in the future would inform on the relative degrees of microvascular abnormalities in different dementias, as assessment of microvascular abnormalities potentially distinguish dementias with more severe vascular insults (such as VaD and PSD) from other neurodegenerative dementias.
Acknowledgements
We are grateful to the patients, families, and clinical house staff for their cooperation in the investigation of this study. We also appreciate the cooperation of the NBTR directors and staff in assisting us with this study. We are thankful to Janet Slade and Arthur Oakley for the expert technical assistance and for assisting us in managing and screening the cohort.