White Matter Hypoperfusion and Damage in Dementia: Post‐Mortem Assessment

Abstract

Neuroimaging has revealed a range of white matter abnormalities that are common in dementia, some that predict cognitive decline. The abnormalities may result from structural diseases of the cerebral vasculature, such as arteriolosclerosis and amyloid angiopathy, but can also be caused by nonstructural vascular abnormalities (eg, of vascular contractility or permeability), neurovascular instability or extracranial cardiac or vascular disease. Conventional histopathological assessment of the white matter has tended to conflate morphological vascular abnormalities with changes that reflect altered interstitial fluid dynamics or white matter ischemic damage, even though the latter may be of extracranial or nonstructural etiology. However, histopathology is being supplemented by biochemical approaches, including the measurement of proteins involved in the molecular responses to brain ischemia, myelin proteins differentially susceptible to ischemic damage, vessel‐associated proteins that allow rapid measurement of microvessel density, markers of blood–brain barrier dysfunction and axonal injury, and mediators of white matter damage. By combining neuroimaging with histopathology and biochemical analysis, we can provide reproducible, quantitative data on the severity of white matter damage, and information on its etiology and pathogenesis. Together these have the potential to inform and improve treatment, particularly in forms of dementia to which white matter hypoperfusion makes a significant contribution.

Introduction

White matter abnormalities are demonstrable in vivo by computed X‐ray tomography (CT), magnetic resonance imaging (MRI) or positron emission tomography (PET) in most patients with dementia and many of those with mild cognitive impairment (MCI). The abnormalities include infarcts and ischemic leukoencephalopathy, white matter hyperintensities, enlarged perivascular spaces, microbleeds 13, 14, 15, 90, 101, 109, 111, changes in white matter “integrity” on brain diffusion tensor imaging (DTI) 81, 93, 101, 109 and reductions in white matter perfusion and glucose utilization 12, 40, 53. Some of the abnormalities are likely to be secondary to degenerative changes in the cerebral cortex and deep gray matter structures, but others reflect primary damage to the white matter. These include white matter infarcts and regions of ischemic damage; foci of hemorrhage; damage caused by impaired drainage of interstitial fluid, or leakage of neurotoxic molecules from the bloodstream as a result of blood–brain barrier damage; and in several types of dementia, primary degenerative changes in the white matter itself. Post‐mortem histopathological studies have confirmed the high prevalence of structural white matter abnormalities in dementia, the main types of lesion being microinfarcts, markedly enlarged perivascular spaces and regions of non‐cavitating ischemic damage (ischemic leukoencephalopathy) 16, 34, 36, 52. These abnormalities are particularly common and pronounced in vascular dementia (VaD) but are also seen significantly more often in Alzheimer's disease (AD) than control brains 34, 35, 52. In 60% of brains from patients with AD, the deep white matter was reported to show diffuse rarefaction and gliosis 15, thought to reflect ischemic damage 32, 33.

Clinical significance of white matter damage in dementia

White matter ischemic lesions are associated with impaired cognition 37, 96, 98 and are a significant contributor to cognitive decline in AD 37, 51, 96, 98, Parkinson's disease (PD) and dementia with Lewy bodies (DLB) 24, 25, 45, 60. Their contribution to cognitive decline in patients with AD is most significant in the early stages of disease 26; as the disease progresses, the contribution of ischemic lesions to cognitive decline is overwhelmed by the impact of neurodegenerative pathology 26, 34. White matter abnormalities are demonstrable by DTI even before detectable cognitive decline 42, 43, but the precise pathological substrate of these imaging abnormalities is unclear.

In early dementia, cognitive decline correlates with reduced white matter perfusion and glucose utilization 48, 89. In a prospective study of 7983 people aged 55 years and older, 1730 participants whose cognitive function had been monitored over a mean period of 6.5 years had blood flow velocity in the middle cerebral artery measured by transcranial Doppler ultrasonography 89. Those participants with higher blood flow velocity were significantly less likely to show cognitive decline, an association that persisted after adjustment for age, gender and MRI evidence of vascular brain disease (periventricular and subcortical white matter lesions and brain infarcts).

Causes of white matter damage in dementia

The white matter can be damaged by several processes and some of those processes may have multiple possible etiologies. In the context of dementia, the dominant pathogenic process is hypoperfusion. This is often caused by small vessel disease affecting the cerebral vasculature. Arteriolosclerosis and cerebral amyloid angiopathy (CAA) are the main contributors (Figure 1). However, intracranial (and even intracerebral) atherosclerosis, extracranial cardiovascular disease (eg, atrial fibrillation, aortic valve stenosis, atheromatous stenosis of the carotid artery), orthostatic hypotension and carotid sinus hypersensitivity (neurovascular instability) may all cause or contribute to white matter damage 5, 6, 55, 58, 59, 74, 88. Other factors may include impaired drainage of interstitial fluid (see accompanying paper by Weller et al in this mini‐symposium), systemic or local inflammatory processes 31, 84, 92, extravasation of red blood cells leading to the release of iron and possible oxidative damage, and breakdown of the blood–brain barrier as a result of pericyte loss 10 and increased bradykinin production 4 causing leakage of potentially toxic molecules such as thrombin 10, 21, 65 and plasmin 23. The anatomical distribution of white matter abnormalities that can be demonstrated by DTI in MCI and early dementia suggests that some of the abnormalities, for example, in the parahippocampal white matter 109 and fornices 81, are secondary to degenerative changes in the neurons from which the axons in the white matter originate and probably reflect axonal atrophy and degeneration. However, this presumptive pathogenic sequence is not necessarily correct. In their interpretation of DTI changes in the white matter in PD and DLB, Hattori et al 45 proposed the converse: that imaging abnormalities in the parietal and occipital white matter were caused by white matter hypoperfusion that was responsible for cognitive impairment in PD and led to the atrophy of cortical gray matter. In AD, DTI abnormalities are demonstrable at a preclinical stage. In a cohort of 139 people enriched for AD risk factors but not cognitively impaired, DTI revealed higher fractional anisotropy in the cingulum adjacent to the corpus callosum, hippocampal cingulum and lateral fornix in participants who were amyloid positive on PET with [C‐11] Pittsburgh Compound B than in those who were amyloid negative 83.

Figure 1.

Structural disease of small cerebral blood vessels in dementia. A. Arteriolosclerosis in the parietal white matter. The white matter is rarefied and gliotic. B. In another brain with rarefied, gliotic white matter, perivascular spaces are markedly enlarged. C. The overlying cerebral cortex shows severe capillary cerebral amyloid angiopathy (CAA), demonstrated by immunohistochemistry for Aβ42. D. Old hematoma cavity in the subcortical white matter in association with arteriolar CAA.

Most of the scientific literature on abnormalities of the cerebral vasculature in dementia has focused on structural changes to vessel walls, for example, collagenous thickening, deposition of amyloid, loss of smooth muscle cells or loss of pericytes. However, blood flow and blood–brain barrier function are also affected by alterations in vascular contractility and permeability that are not caused by structural changes but are nonetheless likely to contribute to ischemic damage and edema. Aβ peptides were shown to reduce cerebral blood flow, enhance vasoconstriction and impair both functional hyperemia and cerebral autoregulation in mice transgenic for mutant human amyloid‐β precursor protein (hAPP) 71, 72, 73. The level of the vasoconstrictor endothelin‐1 (ET‐1) is significantly elevated in the cerebral cortex in AD 76, in keeping with the increased synthesis of endothelin‐converting enzyme 2 (ECE‐2) 75 and increased activity of ECE‐1 77. Aβ is, like big‐ET‐1 (the biologically inactive precursor of ET‐1), a substrate of ECE‐1 and ‐2 28, 29, 30, both of which can be upregulated in vitro by Aβ peptides: ECE‐1 by exposure of endothelial cells to Aβ40, and ECE‐2 by exposure of neuroblastoma cells to Aβ42 75, 77. Upregulation of ECEs and elevated ET‐1 production in the cerebral cortex in AD may be unfortunate side effects of the accumulation of excessive substrate in the form of Aβ.

The production of another vasoconstrictor is also increased in the cerebral cortex in AD: angiotensin II, which is converted from its inactive precursor angiotensin I by the action of angiotensin‐converting enzyme (ACE) 68. The level and activity of ACE were significantly higher in frontal cortex in AD than control brains and increased with progression of disease as marked by Braak tangle stage 68. ACE level and activity were also upregulated in vitro in neuroblastoma cells when they were exposed to aggregated Aβ42. The upregulation of ACE may also counteract the upregulation of plasma kallikrein in AD 4; plasma kallikrein catalyses the production of bradykinin, which causes arterial and venous dilatation and increased permeability but is cleaved and inactivated by ACE 103. Although neither ET‐1 nor ACE is increased in the white matter in AD 7, reduced perfusion of the white matter in AD may result in part from angiotensin II‐ and ET‐1‐mediated constriction of the perforating arterioles that traverse the cortex but supply blood to the cerebral white matter.

Cerebral hypoperfusion and neurodegeneration

A range of experimental evidence suggests that cerebral hypoperfusion is likely to exacerbate some of the neurodegenerative disease processes in patients with dementia. Most of the evidence relates to AD, a disease in which the initiating abnormality is thought to be the cerebral accumulation of Aβ peptide (particularly Aβ1–42) produced by the actions of β‐ and γ‐secretases on amyloid‐β precursor protein (APP). Transient focal cerebral ischemia caused overexpression of APP mRNA 94, and hypoperfusion and neuronal hypoxia upregulated BACE1 mRNA and protein, increased β‐secretase activity and the production of Aβ peptide 44, 61, 112, 121, 122. Ischemia also induced expression of PSEN1 mRNA in the gerbil hippocampus 100 and hypoxia enhanced the expression of APH‐1a, another component of γ‐secretase complex 61. Conversely, exposure of neuroblastoma cells and primary neuronal cultures to hypoxia and oxidative stress reduced the expression and activity of neprilysin 41, a major Aβ‐degrading enzyme 18, 66, 67, 104, 105. Other studies showed that oxidative modification of neprilysin, leading to the formation of 4‐hydroxynonenal adducts, reduced Aβ catalytic activity and caused Aβ accumulation 107, 108. Oxidative modification also inactivated insulin‐degrading enzyme, another contributor to Aβ degradation 95. Lastly, transient cerebral ischemia caused hyperphosphorylation of tau 113, 114; this was significantly reduced when cyclin‐dependent kinases were inhibited with roscovitine.

In studies of potential effects of cerebral hypoperfusion on the development of Lewy body pathology, transient cerebral ischemia increased the amount of α‐synuclein that could be detected immunohistochemically in the gerbil hippocampus 49, 117. Acute hypoxia increased the level of α‐synuclein in mouse cerebral cortex 118, and glucose deprivation caused the formation of α‐synuclein inclusions in neuroblastoma cells in vitro 11. Unal‐Cevik et al 106 found that transient middle cerebral artery occlusion induced oligomerization of wild‐type α‐synuclein in mouse brain. When we exposed neuroblastoma cells overexpressing human wild‐type α‐synuclein to either glucose deprivation or combined oxygen and glucose deprivation (simulating ischemia), there were significant increases in the levels of total α‐synuclein and α‐synuclein phosphorylated at serine 129 [a modification reported to promote α‐synuclein aggregation and neurotoxicity 20, 22 ] 69.

There is some evidence from studies in patients and from post‐mortem examination of human brain tissue that cerebral hypoperfusion increases AD pathology. White matter hyperinsensities were significantly associated with cortical atrophy in patients with clinically probable AD 17. Hypoxia due to cardiac arrest was found to increase serum amyloid β levels in humans 120. Yuan et al 119 reported that patients with probable severe AD (ie, having a clinical dementia rating score of 3) had a significantly greater prevalence of severe atheromatous stenosis of intracranial or extracranial arteries than did patients with “early” or mild AD (clinical dementia rating score of 0.5 or 1), and Hofman et al 46 found that severe atherosclerosis increased the odds ratio for AD by about threefold (95% confidence interval 1.5–6.0), although this was partly attributable to an association between APOE ε4 and both AD and atherosclerosis. In several post‐mortem studies, the severity of cerebrovascular atherosclerosis correlated with the burden of AD pathology 9, 47, 56, 86, 87, 88, 116. However, other post‐mortem studies did not find an association between the severity of AD pathology and the presence of vascular clinical risk factors, cerebral infarcts or other vascular pathological abnormalities 27, 85, 91.

The association between cerebral hypoperfusion and neurodegenerative pathology in patients with dementia is likely to be bidirectional. As noted above (see Causes of white matter damage in dementia), Aβ upregulates the production of several vasoconstrictors within the brain. It also has anti‐angiogenic activity 79, binds to (and possibly sequesters) vascular endothelial growth factor (VEGF) 115 and interacts with the extracellular domain of VEGF receptor 2 and interferes with VEGF‐mediated signaling 78. Less information is available in relation to other neurodegenerative diseases that cause dementia. However, in post‐mortem samples of occipital cortex from patients DLB, we showed that von Willebrand factor (factor VIII‐related antigen), an excellent surrogate marker of capillary density 7, was reduced and that this was associated with a lower concentration of VEGF 69. VEGF was also reduced in SH‐SY5Y neuroblastoma cells overexpressing wild‐type human α‐synuclein, the extent of reduction correlating closely with the level of α‐synuclein, suggesting that VEGF deficiency secondary to the accumulation of α‐synuclein may lead to reduced microvessel density in this disease.

Current approaches to assessment of white matter damage (and their limitations)

As noted in the Introduction, CT and MRI have proven invaluable in demonstrating ischemic and other types of white matter abnormality in vivo. Newer MRI techniques such as DTI are highly sensitive to cellular and subcellular changes in the white matter that alter the separation and orientation of membranes and the diffusivity of water molecules in different planes 3, 50, 64. These techniques allow the identification of distinct patterns of white matter damage in different types of dementia, even in preclinical stages 80, 81, 93, 101, 109. Neuroimaging is less useful in determining the process responsible for the damage (eg, ischemia, leakage of neurotoxic molecules from the bloodstream, impaired drainage of interstitial fluid) or the underlying etiology (eg, whether ischemic changes result from arteriolosclerosis, CAA, cardiac dysrhythmias or neurovascular instability and postural hypotension), although intravenous administration of gadolinium can be used to assess the integrity of the blood–brain barrier 110. Imaging methods for measuring cerebral blood flow and glucose utilization may add information, particularly if there are regional reductions in blood flow 12, 40, 53, 57, but do not generally enable a pathogenic reduction in blood flow leading to tissue hypoxia to be distinguished from a reduction secondary to tissue damage and lower metabolic demand.

Conventional histopathological approaches to assessing white matter damage have focused mainly on ischemic changes and foci of hemorrhage and have tended to conflate changes that affect vessel walls (atherosclerosis, arteriosclerosis, arteriolosclerosis, CAA) with changes that reflect altered interstitial fluid dynamics or white matter ischemic damage, even though the latter may sometimes be of extracranial etiology (Figure 2) or a reflection of nonstructural vascular abnormalities (eg, of vascular contractility or permeability). In histopathological studies, the severity of white matter damage or of vascular abnormalities has generally been specified (rather than measured) with reference to arbitrary, semi‐quantitative or categorical scales. These studies have provided some insights into the pathogenesis of the damage [eg, its relationship to arteriolosclerosis, severe CAA or to Aβ parenchymal load 19, 26, 62 ] but have been limited by subjectivity, lack of specificity and relative insensitivity. Abnormalities in the subcortical and periventricular white matter can also be demonstrated in formalin‐fixed post‐mortem brain tissue by T2‐weighted MRI, although post‐mortem MRI was reported to be generally less sensitive than histopathological examination for detecting abnormalities 38, 63.

Figure 2.

Ischemic white matter damage in a dementia patient with cardiac valvular disease and recurrent episodes of bradyarrhythmia. A. A coronal slice through the frontal and temporal lobes reveals multiple cavitated infarcts and foci of softening and gray discoloration. There is only mild atherosclerosis. B. In a more anterior slice through the same brain, the white matter appears pitted, with numerous confluent foci of gray discoloration.

Molecular pathological approaches

Another approach to assessing ischemic white matter damage in post‐mortem brain tissue is the measurement of gene transcripts and proteins involved in the molecular responses to brain ischemia. Hypoxia‐inducible factor (HIF) α subunits are the principal sensors of tissue hypoxia 2, 54. Under normal conditions these constitutively expressed proteins undergo proline hydroxylation, which initiates their interaction with E3 ubiquitin ligase and consequent degradation. Hypoxia inhibits hydroxylation of the proline residues and results in persistence of the HIF α subunits, which associate with HIF β to form transcriptionally active HIF‐1 and HIF‐2 heterodimers. In vitro studies suggest that Aβ can also activate HIF‐1 directly 97. Under conditions of chronic or intermittent hypoxia, there is an increase in HIF‐1α transcription 70, 82. Tissue hypoxia also causes upregulation of a large number of other genes. The transcription of some of these, such as VEGF and heme‐oxygenase 1, is upregulated by HIFs. Others, such as neuroglobin 99, are upregulated by hypoxia independently of HIFs. Fernando et al 39 demonstrated an increase in immunohistochemical labeling of white matter for HIF‐1α and HIF‐2α and neuroglobin in deep subcortical white matter lesions in the elderly. In more recent studies, we demonstrated the post‐mortem stability of VEGF in human brain tissue, for up to 72 h at 4°C or even room temperature 69, and used a sandwich enzyme‐linked immunosorbent assay (ELISA) to measure the concentration of the protein in both gray and white matter 7, 69, 102. As noted below, there was a highly significant correlation between VEGF concentration in the white matter and other measures of white matter ischemia.

A further index of the adequacy of ante‐mortem perfusion of the white matter is the ratio of the “distal” periaxonal oligodendrocyte protein, myelin‐associated glycoprotein (MAG) that is highly susceptible to white matter ischaemia, to more “proximal” myelin proteins such as proteolipid protein 1 (PLP1) and myelin basic protein (MBP), which are relatively resistant 1, 7, 8. Unfortunately, over a period of several hours, MBP undergoes post‐mortem degradation, even when stored at 4°C 8. However, both PLP1 and MAG are relatively stable for at least 72 h, and comparison of the level of these two proteins is a useful means of detecting and quantifying the severity of ante‐mortem hypoperfusion. We found that both MAG concentration and the MAG : PLP1 ratio showed highly significantly negative correlation with semi‐quantitative scores of the severity of small vessel disease, in a Bristol cohort (n = 99) and an independently scored Oxford cohort (n = 77) 8 (Figure 3). In a subsequent study, we demonstrated that as the MAG : PLP1 ratio declined, the concentration of VEGF in the white matter increased 7 (Figure 4). The highly significant correlation between these entirely distinct measures of white matter hypoperfusion provides reassurance as to their validity, particularly as the findings were reproduced in two independent cohorts. A further indicator of white matter hypoperfusion was a decline in the concentration of the vasoconstrictor ET‐1, presumably as a protective response to reduce vasoconstriction, increase blood flow and minimize the risk of ischemic damage. In VaD, however, there was a trend toward increased ET‐1, raising the possibility that ET1‐mediated vasoconstriction may contribute to white matter hypoperfusion in this disease 7.

Figure 3.

Myelin‐associated glycoprotein (MAG) : proteolipid protein 1 (PLP1) ratio as a measure of the severity of chronic hypoperfusion of the cerebral white matter. In a study of two independent post‐mortem cohorts from Bristol and Oxford, MAG : PLP1 declined significantly with increasing severity of arteriolosclerotic small vessel disease (SVD) (Spearman's test: Bristol P = 0.0003, Oxford P = 0.0009). Note that severity of SVD was scored on a 4‐point scale in the Bristol cohort and on a 12‐point scale 96 in the Oxford cohort. Reproduced with permission from 8.

Figure 4.

Correlation between two independent measures of white matter hypoperfusion. In both the Bristol and Oxford cohorts, vascular endothelial growth factor (VEGF) concentration in the white matter increased significantly as the myelin‐associated glycoprotein (MAG) : proteolipid protein 1 (PLP1) ratio declined (Spearman's test: Bristol P = 0.0015, Oxford P < 0.0001). Reproduced with permission from 7.

A biochemical approach can also be used to assess the microvascular density in brain tissue. We found good correlation between computer‐assisted morphometric measurements of vessel density (assessed immunohistochemically in paraffin sections) and the concentration of the endothelial marker, von Willebrand factor, in both cerebral cortex and white matter 7, 69. Computer‐assisted morphometry of microvessel density in immunolabeled paraffin sections is excellent for identification of structural abnormalities of the blood vessels but has several limitations. It is relatively labor intensive and time consuming, more affected by sampling variation than are measurements on tissue homogenates (which yield mean values for a much larger volume of tissue), and does not allow direct comparison of microvessel density and biochemical measurements such as VEGF or MAG : PLP1 in the same samples. We found a significant positive correlation between von Willebrand factor level and VEGF concentration in the white matter 7, as might be expected given the pro‐angiogenic actions of VEGF: that is, part of the response to hypoperfusion includes a HIF‐mediated increase in VEGF that leads to an increase in vessel density. Other potential biochemical approaches to the post‐mortem assessment of white matter damage and its pathogenesis include measurement of serum proteins such as albumin to quantify blood–brain barrier dysfunction; of phosphorylated and non‐phosphorylated neurofilament proteins to assess axonal injury; and of a range of potential mediators of damage to oligodendrocytes, myelin, axons and blood vessels.

Conclusions

An ideal method for assessing white matter hypoperfusion and post‐mortem damage would be sensitive and specific, and provide reproducible, quantitative data on severity, as well as information on etiology and pathogenesis. No single method meets all of these objectives. However, by combining in vivo neuroimaging data with post‐mortem histopathology and biochemical analysis of a range of molecular markers, we can now meet most of these objectives and have the potential to make detailed assessment of white matter damage in dementia, to gain insight into physiological and pathological processes involved in the regulation of cerebral perfusion in the human brain and to identify an expanded range of pathogenic processes that contribute to cognitive decline.