Evidence of endothelial dysfunction in the development of Alzheimer’s disease: Is Alzheimer’s a vascular disorder?

Abstract

The etiology of Alzheimer’s disease (AD) remains unclear. The emerging view is that cerebrovascular dysfunction is a feature not only of cerebrovascular diseases, such as stroke, but also of neurodegenerative conditions, such as AD. In AD, there is impaired structure and function of cerebral blood vessels and cells in the neurovascular unit. These effects are mediated by vascular oxidative stress. Injury to the neurovascular unit alters cerebral blood flow regulation, depletes vascular reserves, disrupts the blood-brain barrier and reduces the brain’s repair capacity. Such injury can exacerbate the cognitive dysfunction exerted by incident ischemia and coexisting neurodegeneration. This article summarises data regarding cardiovascular risk factors, vascular abnormalities and brain endothelial damage in AD. In view of accumulating evidence of vascular pathology in AD, we also review the literature (MEDLINE, EMBASE) for clinical evidence of impaired endothelial function in AD. A total of 15 articles investigating endothelial dysfunction in AD were identified. 10 of these articles showed impaired endothelial function in AD patients. The current literature suggests endothelial dysfunction may be involved in the pathogenesis of AD. This aspect of AD pathology is particularly interesting in view of its potential for therapeutic intervention. Future research on endothelial function in AD should concentrate on population-based analysis and combine multiple methods to evaluate endothelial function.

Introduction

Alzheimer’s disease (AD) is a progressive, neurodegenerative disorder characterised by a decline in cognitive function. Over 23.4 million people suffer from dementia today [1]. This figure is predicted to double every 20 years, reaching 81.1 million by 2040. According to numerous epidemiological studies AD is the most common subtype of dementia [2,3]. However, whilst cerebrovascular disease is considered exclusion criterion for AD, lacunar infarcts have been observed in over 70% of AD patients’ suggesting that the differences between vascular dementia (VaD) and AD are not as definitive as initially postulated [4].

Despite thorough research, the pathogenesis of AD remains unclear. The most accepted hypothesis proposes that deposition of insoluble amyloid β protein (Aβ) fragments causes the neurodegeneration in AD [5]. Aβ is thought to contribute to the accumulation of reactive oxygen species (ROS) resulting in oxidative stress, neuronal damage and cognitive decline.

However, there is growing evidence from studies to suggest that vascular dysfunction plays a central role in the development of AD. Amyloid plaques and neurofibrillary tangles (NFTs) may be the result of hypoxia due to inadequate blood supply [6]. In vitro, Aβ has been shown to be toxic to both cerebral and peripheral endothelium [7,8]. It has been postulated that the neurotoxic effects of Aβ may impair normal endothelial function and cause endothelial-dependent vasoconstriction [9,10].

Epidemiological studies, including the Honolulu-Asia Study [11], Rotterdam Study [12,13], EURODEM [14], and Kungsholmen Project [15] report several vascular risk factors (VRFs) associated with AD-type dementia, all of which cause cerebral hypoperfusion. The association of vascular diseases with AD prevalence strongly suggests that vascular mechanisms play a role in the development and/or progression of AD.

The endothelium is the monolayer of endothelial cells lining the lumen of the vascular beds separating the vascular wall from the circulation and the blood components with both metabolic and mechanical roles. Endothelial dysfunction may be reversible so may represent a therapeutic target.

Aims

In this paper, we review:

I) Evidence of vascular involvement in the pathogenesis of AD.

II) In vivo clinical studies investigating the relationship between AD and endothelial dysfunction.

The neurovascular unit

Neurons and glial cells work in concert with endothelial cells (ECs) and smooth muscle cells to maintain homeostasis of the cerebral microenvironment [16,17]. Thus, the neurovascular unit maintains tight control of the chemical milieu of the brain by regulating local cerebral blood flow (CBF) and molecular transport across the blood brain barrier (BBB). Figure 1 shows the structural arrangement of the neurovascular unit.

Figure 1.

The neurovascular unit. This diagram depicts a cerebral blood vessel and surrounding brain parenchyma that comprise the neurovascular unit. The neurovascular unit encompasses cell-cell and cell-matrix interactions between component endothelial, glial, and neuronal cells. (Image adapted with permission from Iadecola 2004) [268].

Large cerebral arteries branch into smaller pial arteries that run along the surface of the brain across the subarachnoid space [18]. These are composed of an EC layer, a smooth muscle layer and an outer layer of adventitial cells. Penetrating intracerebral arteries are separated from the brain by the Virchow-Robin space [19]. As intracerebral arterioles penetrate deeper into the brain, this space disappears and the vascular basement membrane (VBM) comes into direct contact with the astrocytic end-feet. The VBM separates the endothelium from other cells of the neurovascular unit, including pericytes and astrocytes.

Interestingly, cerebral ECs are not fenestrated but are interconnected by tight junctions forming a continuous cellular structure, the BBB. The BBB prevents passive diffusion of solutes between the blood and brain intracellular fluid and limits uncontrolled entry of neurotoxic substances into the brain [20]. Specialised receptors on the membrane of ECs initiate intracellular signaling cascades in response to agonists that activate specific receptors, or changes in shear stress at the cell surface produced by changes in the rate of blood flow. Gap junctions permit crosstalk between adjacent ECs allowing the transmission of intracellular responses. Once initiated, these cascades trigger the release of potent vasodilator substances such as nitric oxide (NO) and prostacyclin, and vasoconstrictors, such as endothelin and endothelium-derived constrictor factor [21-23]. At low concentrations, reactive oxygen species (ROS) function as vasodilators, however at high concentrations they cause vascular dysfunction [24].

Therefore, neurons, glia, and vascular cells compose a functional unit, which serves to maintain homeostasis of the cerebal microenvironment. Additionally, the neurovascular unit plays an important role in protection against cerebrovascular dysfunction and ischemia [25].

Current hypotheses on Alzheimer’s disease etiology

Much of the literature concerning theories of AD etiology favors one of two hypotheses - the ‘amyloid cascade hypothesis’ (ACH) or the ‘vascular hypothesis’. Whilst the ACH implicates the overproduction and deposition of Aβ fragments as the primary cause of AD [5,26], the vascular hypothesis argues that cerebral hypoperfusion resulting from vascular disease and ageing leads to the neuronal death and cognitive dysfunction in AD (Figure 2).

Figure 2.

Overview of the vascular and amyloid hypotheses of Alzheimer’s disease. The amyloid hypothesis proposes that overexpression of amyloid precursor protein (APP) is the primary cause of Alzheimer’s disease (AD). Mutations in the presenilin 1, presenilin 2 and APP gene increase amyloid deposition and raise the risk of developing AD. Senile plaques are a key pathological feature of AD and in mouse models plaque formation is associated with neuronal damage. The vascular hypothesis suggests vascular dysfunction associated with ageing is the primary trigger of AD. The prevalence of vascular risk factors is greater in the elderly population and vascular disease is strongly associated with risk of AD. Cerebral hypoperfusion resulting from impaired vasoregulation in vascular disorders causes a neuroglial energy crisis resulting in cognitive. Vascular dysfunction promotes the conversion from mild cognitive impairment to AD.

The amyloid cascade hypothesis

Since the classification of presenile dementia by Alois Alzheimer in 1907, senile plaques and NFTs have been considered key pathological hallmarks of AD [27]. The discovery of Aβ as the main constituent of senile plaques and the identification of gene mutations in amyloid precursor protein (APP), the presenilin 1 (PSEN1) gene and the presenilin 2 (PSEN2) gene causing Aβ accumulation and early onset familial AD led to the ACH [28-30]. The ACH theorises that abnormal Aβ aggregation and deposition initiates the pathogenesis of AD and results in neuronal apoptosis, NFT formation and cognitive decline. Due to the similar phenotype between familial AD and late-onset AD, it was assumed that increases in Aβ deposition were the cause of all types of AD. Despite substantial evidence supporting senile plaque and NFT formation as the primary cause of AD, recent research has greeted this theory with more scepticism [31-35].

A number of observations are used to support the ACH. AD patients have amyloid plaques containing degenerating nerve endings; the plaque count in these patients is greater than in normal aging [36,37]. Furthermore, Down’s syndrome patients invariably develop classical AD pathology by age 50. These individuals also overproduce Aβ from birth and amyloid plaques far precede NFTs and other AD lesions [38,39]. APP over-expression mice initially develop diffuse plaques, which develop into fibrillar plaques associated with damage to neurons and microglia [40]. Similarly, Aβ fibrils reproducibly damage cultured neurons and activate brain microglia [41,42]. Blocking Aβ plaque formation prevents this toxicity [43,44]. A rare coding mutation in the APP gene (A673T) which leads to approximately 40% less APP cleavage actually protects against AD in the elderly [45]. Finally, Apolipoprotein ε4 (APOE-ε4), a significant genetic risk factor for AD, is associated with overproduction of amyloid in the brain before AD symptoms become evident. Each APOE-ε4 allele lowers the age at onset in a dose-dependent fashion [46]. Thus, there is compelling evidence in favor of the ACH.

Due to recent evidence, however, the theory that senile plaque accumulation is the primary cause of neurodegeneration in AD has been criticised by many [31-35]. Senile plaques and NFTs may occur as a reactive response to neurodegeneration rather than being its cause. In head trauma patients, similar pathological hallmarks are seen as in AD patients in neuronal perikarya and in dystrophic neurites surrounding Aβ deposits [47]. Moreover, increased numbers of βAPP-immunoreactive neurons have been observed in the entorhinal cortex after head injury [48]. These findings suggest APP over-expression and amyloid deposition could occur as an acute-phase response to neuronal damage [49]. This concept of amyloid production as a neurotrophic response is supported by the observations that APP shares structural similarities with the precursor for epidermal growth factor [50] and cerebellar Aβ-deposits contain acute phase proteins such as complement factors and α-anti-chymotrypsin [51]. Other evidence contradicting the ACH has also been reported. The degree of amyloid deposition in the brain does not correlate with the severity of cognitive dysfunction. In addition, the brains of many individuals with normal cognitive function have abundant senile plaques or NFTs [52,53]. Transgenic mouse models over-expressing APP show no neuronal loss in the CA1 region [54] and show cognitive dysfunction before Aβ deposition [55]. Furthermore, vaccinations against Aβ over 8 weeks do not improve cognition in mice transgenic for APP and presenilin 1 [56]. Therefore, whilst there is evidence in support of the ACH, there are increasing questions about the validity of this theory.

The vascular hypothesis

The vascular hypothesis of AD suggests that pathology begins with cerebral hypoperfusion. This in turn leads to a crisis among glia and neurons, eventually culminating in neurodegeneration and cognitive impairment. In addition, interactions between vascular dysfunction and increased amyloid production that facilitate the pathogenesis of AD will be discussed.

Vascular disease and Alzheimer’s disease

Several vascular diseases are associated with an increased risk of AD, all of which induce cerebral hypoperfusion [11,13,15]. Recently, Kaffashian et al. found that Framingham general cardiovascular disease (CVD) and stroke risk scores correlate with cognitive decline in late middle age and may be effective for assessing risk of cognitive dysfunction and targeting modifiable risk factors [57]. Furthermore, Li et al. showed that VRFs promote conversion from mild cognitive impairment to AD [58]. In the same study, treatment of VRFs reduced the risk of incident AD. Interaction between AD pathology and VRFs may aggravate perfusion abnormalities in AD (Figure 3).

Figure 3.

Interaction between vascular disease and AD pathology. Arrows indicate how cerebrovascular factors influence amyloid processing. Hypoxia resulting from vascular disease (e.g atherosclerosis) causes increased expression of hypoxia-inducible factor (HIF)-1α. HIF-1α binds to the hypoxia-responsive element on amyloid cleaving enzyme BACE1. The resulting upregulation of BACE1 mRNA expression increases the production of Aβ fragments. Hypoperfusion resulting from vascular disease and old age further increases BACE1 activity. In addition, hypoperfusion may contribute to oxygen deficiency and cause a neuroglial energy crisis, further ameliorating amyloid plaque production, CAA and ischaemia. CAA = Cerebral amyloid angiopathy, BACE1 = beta secretase 1, sAPP = secreted form of beta-APP, APP = amyloid precursor protein, Aβ = beta amyloid peptide. Image adapted with permission from Kalaria et al. [59].

Hypertension

High blood pressure (BP) has been established as one of the strongest VRFs for AD [11,14,59,60]. The Honolulu-Asia study followed up 3734 men and found a strong correlation between midlife hypertension and AD [14]. This association was greatest in those not undergoing treatment with antihypertensive drugs. Moreover, neuroimaging and post-mortem analysis of AD patients revealed that increased BP in midlife resulted in pathological changes in blood vessels, brain atrophy and an increase in senile Aβ plaques and NFTs in the neocortex and hippocampus [61]. Kivipelto et al. also found an association with midlife hypertension and AD risk [59]. Recently, Rodrigue et al. found that hypertension interacts with the APOE-ε4 genotype to increase amyloid deposition in cognitively healthy middle-aged and older adults [62]. Furthermore, increased pulse pressure was strongly associated with increased mean cortical amyloid level for subjects with at least one ε4 allele. These findings suggest that hypertension greatly exacerbates neuropathological and cognitive brain changes in patients susceptible to AD.

Angiotensin II (Ang II) is an important vasoactive peptide hormone associated with hypertension. Angiotensin I (Ang I) is converted to Ang II through removal of two C-terminal residues by angiotensin-converting enzyme (ACE), causing vasoconstriction and increased BF. Ang II also activates NADPH oxidase (NOX), leading to increased ROS production and oxidative stress [63,64].

Several studies have reported increases in ACE activity in patients with AD. Increased binding of radioactively labeled ACE-I to ACE in the temporal cortex and elevated neuronal and perivascular ACE immunoreactivity has been observed in AD patients post-mortem [65,66]. Furthermore, Miners et al. found the strength of the increase in perivascular ACE immunoreactivity correlated not only with the level of cerebral amyloid angiopathy (CAA) but also with AD severity [67]. Not only does ACE catalyse the production of Ang II, it also cleaves Aβ [68]. The cleavage Aβ of by ACE may contribute to pathological Aβ misfolding. Further evidence for the role of Ang II and ACE in AD pathogenesis comes from clinical trials of ACE inhibitors (ACE-Is) and angiotensin receptor blockers (ARBs), which lower the risk of developing AD, alleviate cognitive decline and decrease the load of amyloid plaques in the brain [69-71].

Long-lasting increases in BP may worsen the risk of AD by inducing small vessel disease, structural white matter alterations and cerebral hypoperfusion through impairment of normal vascular regulation or atherosclerosis. An increase in systemic BP can result in BBB dysfunction in cerebral microvasculature, which further damages subcortical vessels [72]. In addition, this causes impairment of solute transport to the brain and a decrease in glucose delivery results in expression leading to an increase in amyloid fibrillogenesis [73]. Therefore, hypertension probably increases the risk of AD by disrupting BBB transport, increasing Ang II production, damaging the endothelium and causing production of amyloid and free radicals.

Obesity

Epidemiological research has shown obesity in midlife is a strong risk factor for AD, however this relationship may not be present as ageing progresses further [74,75]. Singh-Manoux et al. analysed 6401 adults aged 39-63 years and found that increased BP and other metabolic factors resulting from excess body weight causes middle-aged and older adults to experience a rapid decline in cognitive skills [76]. Participants who were both obese and metabolically abnormal showed a 22.5% faster decline in memory and thinking skills than control subjects. The increase in prevalence of AD in obese individuals is, in part, due to insulin resistance. Around 80% of individuals with obesity also suffer from insulin resistance [77]. However, many other risk factors associated with obesity also increase the risk of developing AD.

Plasma free fatty acid (FFA) levels are increased in most obese individuals [78]. Moreover, obesity causes inhibition of the enzyme metalloproteinase, which is associated with amyloid clearance [79,80]. In cultured vascular cells, high FFAs stimulate ROS production through protein kinase C-dependent activation of NOX [81]. These findings suggest that obesity increases oxidative stress and amyloid accumulation in the vasculature which may increase the risk of dementia. Kalmijn et al. found increased FFAs caused increased tau and amyloid expression [82]. Furthermore, elevated FFAs initiate the release of compounds such as Ang II and TNF-α, which may exacerbate the development of AD [83].

The relationship between obesity, atherosclerosis and hypercholesterolemia is well established [84,85]. Hofman et al. found a 3-fold increase in risk of developing AD or VaD in patients with atherosclerosis [13]. Atherosclerosis in carotid arteries produced the strongest risk of developing dementia. In a 6 year follow-up study of elderly individuals with arterial stiffness and atherosclerosis, high pulse pressure (low diastolic and high systolic) was associated with AD and VaD [86]. These findings suggest aortic compliance and stroke volume play important roles in AD pathogenesis. Postmortem grading of Circle of Willis atherosclerotic lesions shows that atherosclerosis is more severe in cases with AD [4,87]. Unsurprisingly, atherosclerosis in the Circle of Willis is positively associated with the formation of neuritic plaques and NFTs [88]. Endothelial and perivascular damage are common features of atherosclerosis, which may result in diminished blood flow and inhibition of EDNO [89]. Accordingly, dysregulation of EDNO resulting from chronic brain hypoperfusion causes spatial memory loss and Aβ⁴⁰ deposition in the rat hippocampus [90]. Due the slow onset and progression of atherosclerosis, it is probable that the occlusion of coronary arteries prior to cognitive decline precedes AD symptoms. These results may explain the current findings on human cognitive dysfunction and brain hypoperfusion [91]. In patients with carotid artery stenosis who have undergone a carotid endarterectomy and shown no sign of microembolism following surgery cognition is often improved. If however, microembolic events are identified or brain perfusion is not restored, cognition does not tend to improve [92]. If the theory that vessel occlusion precedes cognitive dysfunction is correct, prevention of atherosclerosis may significantly lessen AD prevalence.

Elevated cholesterol levels also seem to contribute to AD pathology. In vitro, high cholesterol leads to overproduction of Aβ [93]. Aβ was found to inhibit cholesterol trafficking and alter cholesterol homeostasis leading to neurodegeneration and the onset of AD pathology, suggesting Aβ has a role in control of cholesterol transport. In addition, cholesterol secosterol aldehydes initiate Schiffs base-formation resulting in enhanced amyloidogenesis [94,95]. Thus, obesity may contribute to AD via a number of mechanisms including increased cholesterol, atherosclerotic occlusion of cerebral blood vessels and insulin resistance.

Diabetes mellitus

A number of population-based studies have found an association between diabetes mellitus and AD with an incidence of AD as high as 2-5 times higher in patients with type-II diabetes [12,96-98]. Both diabetes and AD are associated with localised amyloid accumulation and both increase in prevalence during ageing. Whilst elevated peripheral insulin levels initially result in increased brain insulin, prolonged hyperinsulinemia reduces brain insulin levels via downregulation of BBB insulin receptors and reduced insulin transport into the brain [99,100].

Insulin receptors are expressed in high concentrations in areas of the brain associated with learning and memory such as the hippocampus and limbic system, which may explain the correlation between hyperinsulinemia and cognitive decline [101]. Hyperinsulinemia initiates central inflammation and Aβ production in healthy individuals [102]. In mice with streptozotocin-induced diabetes, impaired insulin-signaling results in tau hyperphosphorylation and elevated Aβ [103]. Furthermore, rat models of spontaneous diabetes elicit pathological changes similar to that observed in AD [104].

Insulin resistance is a key pathological feature of diabetes and there is growing evidence to suggest a link between insulin resistance and AD. Extensive abnormalities in insulin and insulin-like growth factor type I and II signalling mechanisms have been observed in brains with AD, with markedly reduced expression of genes encoding insulin [105]. Hence, insulin resistance in AD has been referred to as ‘type-3 diabetes’. Insulin resistance is also associated with increased oxidative stress, elevated cortisol and mitochondrial dysfunction, which may contribute to cognitive dysfunction [106]. So, diabetes probably increases the risk of AD by increasing senile plaque and NFT production, inducing an inflammatory response, increasing free radical production and causing disruption of nutrient transport in the BBB.

Cerebrovascular disease (stroke)

Numerous studies have found an association between stroke and AD [4,107,108]. Lesions in the white matter surrounding cerebral ventricles closely resembling ischaemic infarcts have been observed in neuropathological studies of AD [109]. Microvascular lesions are a common pathological feature of AD, with similar numbers of lesions being present as there are NFTs and amyloid plaques [110]. Moreover, subcortical infarcts co-exist with AD pathology and increase the risk of developing dementia 4-fold [111].

It has been postulated that cerebrovascular damage (e.g. microinfarcts) precedes the development of AD symptoms. Patients with AD have more pronounced atherosclerotic lesions in the Circle of Willis than age-matched controls and it has been suggested that ischaemic stroke is the main cause of AD in some individuals [87]. Also, AD is more damaging in the presence of cerebral infarction and autopsy studies show around 70% of patients diagnosed with AD have evidence of coexistent cerebrovascular disease [4,107]. A strong correlation between infarcts and AD has been observed, with microinfarcts restricted to watershed cortical zones [112]. Interestingly, CAA was an important risk factor for watershed microinfarcts, suggesting perturbed hemodynamic factors (e.g. decreased BP) play a role in the genesis of cortical watershed microinfarcts, which may further aggravate degeneration and worsen dementia. CAA occurs in over 80% of patients with AD and only around 10% of control subjects. The high incidence of CAA in AD patients suggests it may contribute to the development of AD. CAA is associated with damage to smooth muscle in the cerebral vasculature and reduced amyloid clearance [113]. Pathological amyloid accumulation may have a role in the development of haemorrhagic stroke in AD [114].

In a large population-based study silent stroke doubled the risk of AD in patients aged 60 to 90 years [115]. Lacunar infarcts are small subcortical infarcts that result from occlusion of a single perforating artery [116]. In individuals suffering from silent subcortical lacunar infarcts, cognitive decline correlates with decreased cerebral glucose metabolism and it was found that hypometabolism was a better indicator of cognitive dysfunction than structural changes (e.g. hippocampal atrophy). In conclusion, ischemia, microinfarcts and lesions in AD may provoke stroke triggering the development or progression of AD by reducing blood flow and invoking amyloid overproduction in the brain. It is plausible that a reduction in CBF resulting from stroke facilitates the conversion of preclinical AD to AD. Using effective intervention strategies to prevent VRFs may reduce the risk of AD.

Morphological changes of the cerebral microvasculature in Alzheimer’s disease

There is increasing recognition that structural changes in cerebral blood vessels contribute to the development and progression of AD. Although the presence of cerebrovascular disease is considered exclusion criterion for AD, cerebral capillary degeneration has been found in almost all AD brains examined postmortem and in cortical biopsy material from pathologically confirmed AD [117,118]. Morphological changes in cerebral microvasculature (capillary atrophy, kinking, focal constriction, VBM thickening) have been found to relate to areas of Aβ deposition [119]. The most notable change appears to be thickening of the VBM. The three central components of the VBM are collagen type 1V, laminin, and heparan sulfate proteoglycan (HSPG). Laminin and HSPG are both located at the endothelial (lamina rara interna) and the astrocytic (lamina rara externa) VBM surfaces [120,121]. Both have a role in cell adhesion and attachment and are potent neurite-promoting factors [122,123]. HSPG is negatively charged allowing it to filter anionic and neutral proteins. Charge is crucial in homeostasis of the renal glomerular barrier and may be important for the BBB as well [124,125]. The functional role of the VBM in the maintenance of the BBB is not well established, however it has been postulated that blocked anionic sites could result in disrupted BBB function [126]. Modification of the lamina rara interna could also result in altered attachment and adhesion of the EC. Pathology of the VBM could be associated with impaired solute transport, resulting in leakage of neurotoxic plasma substances or circulating APP into the brain [127,128]. Therefore, VBM pathology In AD may alter the ECs ability to maintain the BBB.

Vascular insufficiency and neurodegeneration

The damaging effect of VRFs and Aβ on cognition suggests a link between ischemia and neurodegeneration. In AD, CBF is reduced and functional hyperaemia is impaired [129,130]. Furthermore, studies indicate that changes in CBF are a powerful predictor of AD [131,132]. Amyloid plaques and NFTs often coexist with vascular pathology in individuals with dementia [133]. Ischaemic conditions may trigger Aβ accumulation by elevating AβPP/β-secretase expression and facilitating amyloidogenic cleavage leading to increased levels of toxic Aβ. In animal models, amyloid deposition is increased in the cortex in response to ischemia [134,135] and a similar process is thought to occur in humans [136]. Indeed, accumulation of both Aβ^1-40 and Aβ^1-42 has been shown to increase dramatically and consistently after cerebral ischemia [137-139]. It is therefore unsurprising that individuals with cerebrovascular insufficiency or VaD have increased levels of Aβ and senile plaques [140].

Conversely, it has been postulated that the toxic effects of Aβ on cerebral blood vessels may induce cerebral hypoperfusion and increase vulnerability to ischemic damage. In APP over-expression mice focal cerebral ischemia produces more severe infarcts than controls, implicating Aβ as a cause for cerebrovascular dysfunction [141]. Similarly, a higher incidence of cerebrovascular lesions is seen in individuals with AD post-mortem [142]. Paradoxically, VRFs can increase vulnerability to AD without affecting neurodegenerative pathology. Despite a two-fold increase in risk of developing AD in diabetics, no increase in amyloid plaques or tau NFTs is observed. In fact, a higher number of microinfarcts have been identified in these individuals suggesting that microvascular lesions may augment the effects of the neurodegenerative pathology in AD [143]. Since amyloid plaques appear to promote ischemia and vice versa, it is plausible that there is a synergistic relationship between the amyloidogenic and vascular features of AD pathology (Figure 4).

Figure 4.

Proposed synergistic relationship between vascular insufficiency and Alzheimer’s disease (Iadecola, 2010 [269]). Vascular insufficiency resulting from vascular risk factors (e.g vascular dementia) can promote β-amyloid (Aβ) production by increasing amyloid cleaving enzyme β secretase-1 and reducing Aβ clearance. Overexpression of amyloid precursor protein in Alzheimer’s disease leads to the formation of senile Aβ plaques, which may induce ROS-mediated vascular damage and compromised vasoregulation. Thus, vascular insufficiency and β-amyloid production may interact via a positive feedback mechanism exacerbating cognitive decline and increasing the risk of dementia.

The research on NFTs and vascular insufficiency is far less conclusive. Elderly patients with hypertension have increased hippocampal NFTs in comparison to control subjects [144]. Furthermore, focal cerebral ischemia enhances tau phosphorylation in rodent stroke models [145]. However, transient hypoperfusion markedly decreases total tau levels in triple transgenic mice expressing amyloid plaques and NFTs. In summary, there is evidence to suggest ischemia contributes to accumulation of Aβ by reducing its degradation and promoting its synthesis whereas the extent to which ischemia regulates tau phosphorylation remains elusive.

Another important question being raised by researchers in this area is whether or not vascular dysfunction and ischemia aggravate cognitive decline in AD. The presence of vascular lesions contributes significantly to the severity of cognitive impairment in AD [146]. Moreover, cerebrovascular disease has a greater capacity to influence cognitive performance at early stages of AD than at more advanced stages of the disease [147] VRFs and lesions also accelerate the progression of cognitive decline in AD [148]. Finally, a study assessing cerebrovascular reactivity to hypercapnia with transcranial Doppler ultrasonography, a measure of cerebrovascular function, found an association between impaired cerebral microvessels functionality and cognitive decline in patients with AD [149]. In conclusion, there is considerable evidence to suggest that vascular dysfunction is linked to cognitive degeneration in AD.

Vascular dysfunction and oxidative stress in Alzheimer’s disease

Vascular oxidative stress induced by elevated Aβ has emerged as a key pathogenic factor in cerebrovascular dysfunction [24,150]. ROS are highly reactive molecules that often have unpaired electrons in their outer orbital. Increased production of free radicals and ROS can disrupt cell membranes and consequently impair cell function (Figure 5).

Figure 5.

Membrane-related amyloid fibrillogenesis and oxidative stress in the pathogenesis of Alzheimer’s disease. Ageing is associated with reduced tissue antioxidant levels, oxidative lipid damage and enhanced Aβ protein aggregation. These changes are even more pronounced in Alzheimer’s disease. Thiol-mediated antioxidant activity is absent in Apolipoprotein E4 alleles which may allow excess oxidative damage to the lipids in lipoprotein particles. Aβ proteins have a pro-oxidant activity towards the polyunsaturated lipids of cell membranes. Lipid oxidation products may exert direct neurotoxicity or induce structural changes in the His residues of Aβ⁴², enhancing its membrane affinity. Furthermore, these products may facilitate the oligomerization of Aβ⁴⁰ and promote fibrillogenesis. Highly reactive oxygen species may be produced as a result of trauma induced-inflammation or fibrillogenesis. Inflammation may also occur in the presence of vascular disease (e.g atherosclerosis). During these processes, Aβ may form neurotoxic pores or channels in endothelial or neuronal membranes. Image adapted from Axelsen et al. [270].

Vascular oxidative stress has been observed in APP over-expression transgenic mice, without any evidence of oxidative stress in other areas of the brain [151]. In the same transgenic models, antioxidant therapy or over-expression of free radical scavenger superoxide dismutase reverses Aβ-mediated cerebrovascular dysfunction, including alterations in both functional hyperemia and endothelial-dependent vasodilatation [152,153]. Furthermore, substitution of methionine 35 with norleucine, a mutation that blocks the ability of Aβ to generate ROS, abolished Aβ^1-40 vasoactivity suggesting that Aβ^1-40 plays a role in the cerebrovascular alterations observed in AD [154].

Experimental studies suggest that free radicals produced by NOX are responsible for the cerebrovascular alterations induced by Aβ [155,156]. ROS produced by NOX contribute to CVDs associated with vascular oxidative stress including atherosclerosis, hyperhomocysteinaemia and hypertension.

The involvement NOX and oxidative stress in AD is illustrated by an increase in NOX levels or activity in mild cognitively impaired and AD brain tissues [157,158]. Moreover, Aβ induces mitochondrial dysfunction and oxidative stress in astrocytes and microglia through activation of NOX [156,159].

Accumulating evidence suggests ROS impair normal vascular function. In animal models of various disease conditions ROS cause endothelial dysfunction [160]. Reaction of ROS superoxide with NO^- generates peroxynitrite, a potentially deleterious oxidant. Several alterations in the function of endothelium are initiated by this reaction. The most established is endothelium-dependent vasorelaxation, which is impaired by a loss of NO^- bioactivity in the vessel wall [161,162]. ROS and peroxynitrite can also trigger oxidative damage to proteins and nucleic acids, including strand breaks in DNA and base/nucleotide modifications, particularly in sequences with a higher number of guanosine bases [163]. In addition, peroxynitrite formation inhibits the activity of prostacyclin synthase, reducing the ability of blood vessels to dilate [164]. Tetrahydrobiopterin, a critical cofactor for the NO synthases, is a crucial target for oxidation by peroxynitrite and in its absence these enzymes become uncoupled, producing ROS rather than NO and causing further oxidative stress [165]. Therefore, ROS-mediated vascular dysfunction is caused by decreased availability of NO and changes in enzymes essential for normal vascular function.

Aβ also contributes to neurovascular dysfunction through other mechanisms. In addition to causing endothelial dysfunction, Aβ can damage other components of the neurovascular unit. Aβ causes damage and apoptosis of glia and neurons, which may attenuate CBF response to stimuli [166,167]. Moreover, intracortical microvessels and NO containing neurons in AD are deprived of a cholinergic neurogenic control, which probably leads to a compromised ability to adjust cortical perfusion in response to neuronal activation during functional tasks related to cognition [168]. Thus, Aβ-mediated damage to neurovascular coupling is not solely a result of vascular dysfunction but also involves damage to neurons and glia.

Pharmacotherapy

Current drugs associated with a modest reduction in AD symptoms such as cholinesterase inhibitors also exert beneficial effects on cerebral perfusion [169]. Although these drugs decrease synaptic breakdown of acetylcholine (ACh), it is possible that their effect on CBF also plays a role in their protective action. Considering the vast amount of evidence suggesting cerebral hypoperfusion is important in AD development, drug trials for statins, non-steroidal inflammatory drugs (NSAIDs) and antihypertensives in the prevention or treatment of AD were initially promising.

However, many potential treatments have not been as efficacious as expected. Statins are a class of drugs that inhibit 3-hydroxy-3-methylglutaryl coenzyme A (HMG-CoA) reductase. HMG-CoA reductase is the rate-limiting enzyme in the cellular cascade of cholesterol biosynthesis. Statins may also reduce Aβ production and inhibit the activation of Rac1, which is critically involved in the activation of NOX [170]. In addition, statins downregulate AT-1 receptor gene expression mediated by destabilization of AT-1 messenger ribonucleic acid.

Due to the possible link between elevated cholesterol and dementia studies have been carried out to test the effect of statins in AD patients. Despite the fact that observational studies suggest statin therapy exerts a benefi cial effect between normal cognitive performance and profound dementia [171], results from randomised control trials have been disappointing in both prevention [172,173] and treatment [174,175] of AD.

The prevalence of AD is decreased in patients with rheumatoid arthritis, suggesting NSAIDs may exert a protective effect against AD. Different NSAIDs inhibit COX isoforms, and the expression levels of COX-1 and COX-2 change in the different stages of AD pathology [176]. Furthermore, COX-1 is heavily expressed in microglia, which are associated with fi brillar Aβ deposits and some NSAIDs inhibit amyloid production via Y-secretase [177]. Indeed, longitudinal observation studies report a reduction in AD symptoms in patients taking NSAIDs [178,179]. Nevertheless, clinical trials found no association between NSAID treatment and cognitive impairment. It is possible that trials were not sustained for a long enough period to yield effective results. More than 2 years of consistent NSAID treatment is likely to be required in order to decrease the risk of AD [179]. Moreover initiation of treatment should occur several years before the onset of AD [180]. However, it is still unclear whether or not infl ammatory mechanisms are invoking damage in AD or whether they are present to remove the debris from other more important pathological processes.

Recent evidence suggests antihypertensive medication may be protective against AD [69-71,181]. In particular, drugs affecting the production of Ang II may have neuroprotective effects. A UK case-control study found that fewer patients prescribed ARBs and ACE-inhibitors went on to develop probable AD [182].

Patients taking ARBs had the lowest risk of developing AD, which may have been due to a lack of ACE-mediated amyloid degradation when taking these drugs [183]. In addition, treatment with ARBs is associated with less AD-related pathology on autopsy evaluations [184]. Large randomised clinical trials are warranted to further investigate the relationship between antihypertensive drugs and dementia. In conclusion, although there is a relationship between VRFs and AD, drug trials have been less successful than anticipated. These findings suggest that the ability of pharmacological intervention in reversing AD pathology is limited.

Endothelial damage and repair in Alzheimer’s disease

Endothelial damage and Amyloid-β

Numerous factors may contribute to endothelial dysfunction in AD. ECs in the cerebral vasculature may contribute to the formation of amyloid deposits surrounding the cerebral blood vessels. Cerebral ECs express APP-cleaving secretases, which may cleave platelet-retained Aβ into fragments leading to increased cerebrovascular amyloid deposition [185-187]. Therefore, brain ECs in AD may produce Aβ causing increased amyloid accumulation and potentially enhancing CAA.

Wu et al. measured gene levels in brain ECs isolated from frontal cortex autopsies in patients with late stage AD, age-matched controls and healthy young adults [188]. Interestingly, the expression of MEOX2, a homeobox gene that is upregulated during vascular differentiation, is significantly reduced in late-stage AD and is required for optimal clearance of Aβ in mice brains. Restoring MEOX2 gene expression in brain ECs from individuals with AD enhanced angiogenesis, transcriptionally suppressed AFX1 forkhead transcription factor-mediated apoptosis and increased the levels of both a major Aβ clearance receptor at the BBB, the low-density lipoprotein receptor- related protein 1 (LRP), and its receptor-associated protein (RAP). They found that in MEOX2 +/- mice, there was a decrease in CBF and capillary length, and that vessels from these mice had an impaired angiogenic response to hypoxia. These findings suggest that MEOX2 may be playing a role in neurovascular dysfunction and other important changes in the AD brain that could contribute to cognitive decline. Thus, MEOX2 could represent a potential new therapeutic target for AD patients.

Aβ has been found to be toxic to cerebral ECs in vitro by causing damage to nuclear and mitochondrial DNA and activation of caspase-8 and caspase-3, which have essential roles in apoptosis [189]. In the same study, Aβ-mediated damage to ECs was inhibited by zVAD-fmk, a broad-spectrum caspase inhibitor or N -acetyl-cysteine, an antioxidant. Numerous other studies have shown the toxic effects of Aβ on ECs, implicating Aβ as the primary cause for impairment of NO production and mitochondrial dysfunction [190,191]. Interestingly, Aβ appears to be dependent on EC state [192]. Moreover, Aβ^1-42 induces EC degeneration by raising intracellular calcium via calcium permeable Aβ channels [193]. These findings suggest a number of signaling pathways could be involved in EC dysfunction in AD. Moreover, disruption of normal endothelial function may precede Aβ deposition suggesting that endothelial dysfunction may play a central role in AD pathogenesis.

Both TNF-α and Aβ are over-expressed in AD. It is possible that TNF-α and Aβ induced alterations in the ASK1-JNK/p38 apoptotic signalling pathways are responsible for EC apoptosis and damage [194,195]. There is evidence to suggest protein phosphatase 2A is required for ASK1 activation in ECs [195]. During Aβ-mediated EC apoptosis activation of sphingomyelinase releases ceramide which in turn activates PP2A. Ceramide is involved in the expression of NADPH-oxidase in ECs and may contribute to ROS-induced damage to ECs [196]. Thus, TNF-α and amyloid induced changes in apoptotic signalling may contribute to EC damage and vascular dysregulation in AD.

Regenerative capacity of brain endothelial cells

The process of vasculogenesis was initially thought only to occur during embryonic development, but not in postnatal life [197]. However, there is increasing evidence to suggest that bone derived endothelial progenitor cells (EPCs) contribute to neovascularization and neoendothelialization and differentiate into ECs in response to injury [198]. In patients with AD, decreased levels of circulating EPCs have been recorded [199]. Levels of EPCs have been suggested to correlate with endothelial dysfunction in AD [200]. Moreover, AD-associated peptide Aβ⁴⁰ blocks endothelial and vascular regeneration and initiates autophagy [201]. A possible explanation for the inhibitory effect of Aβ on endothelial repair was provided by the observation that Aβ blocks vascular endothelial growth receptor (VEGR)-2 [202]. Therefore, the regenerative capacity of ECs may be compromised in AD, perhaps due to the inhibitory actions of Aβ. Since endothelial dysfunction may contribute to Aβ deposition, it is possible that endothelial damage in AD results in a positive feedback loop whereby more Aβ is produced and further aggravates endothelial injury and impairs repair.

The injection of EPCs to induce neovascularization and restore CBF may be a promising treatment opportunity for AD. Indeed, neovascularization can be successfully stimulated by incorporating EPCs into ischaemic sites in animal models [203] and human myocytes [204].

Review of clinical evidence investigating the relationship between endothelial dysfunction and Alzheimer’s disease

Considering the wide range of research favoring a vascular cause for AD discussed in the first section of this review, we believe an analysis of the in vivo evidence of endothelial dysfunction in AD is timely. In this section, we review the literature investigating the relationship between endothelial dysfunction and AD using specific search terms and selection criteria.

In vivo assessment of endothelial function

Despite the evidence supporting the role of endothelial dysfunction in AD, its routine measurement in the clinical setting is challenging. However, due to recent advances in techniques to assess endothelial dysfunction it may now be possible to determine the extent to which vascular impairment correlates with AD. In recent years, a number of non-invasive techniques have been used to assess endothelial function and morphology as an easier and safer alternative to direct assessment of the coronary arteries [205]. Endothelial responses in peripheral vessels correlate well with coronary artery responses [206].

The capacity of blood vessels to respond to mechanical and pharmacological stimuli in the lumen denotes the ability to self-regulate tone and to adjust blood flow and distribution in response to changes in the local environment. NO is probably the most important vasoactive substance produced by the endothelium.

A number of pharmacological (e.g. bradykinin) and mechanical (e.g. sheer stress) stimuli lead to the production of nitric oxide (NO) [207]. Endothelium derived NO (EDNO) is synthesised from the amino acid L-arginine by the endothelial isoform of NO synthase (eNOS), yielding L-citrulline as a byproduct [208]. Specialised calcium-mediated potassium ion channels in the EC membrane open in response to sheer stress [209-211]. Opening K⁺ channels hyperpolarize the EC, increasing influx of calcium ions (Ca⁺). Calcium activates eNOS which catalyses the production of endothelial NO [212,213]. NO diffuses across the endothelial cell membrane and enters the vascular smooth muscle cells where it activates guanylate cyclase (GC), leading to increased blood flow. The balance between NO and various endothelium derived vasoconstrictors and the sympathetic nervous system maintains blood vessel tone.

Loss of EDNO would be expected to promote a vascular phenotype more prone to vascular disease. A disturbed synthesis and bioavailability of NO might contribute to the onset of dementia by causing impaired endothelial function leading vascular disease and loss of direct neuroprotective effects [214,215].

Flow-mediated dilatation

Flow-mediated dilatation (FMD) measures the vasodilatation response of blood vessels in response to sheer stress. A principal mediator of FMD is EDNO. Brief ischemia in distal tissue induced by a tourniquet results in reactive hyperemia, local vasodilatation and increased blood flow in the proximal and distal vessels. The increased blood flow in the proximal vessel (brachial artery) results in increased sheer stress and an NO mediated vasodilatation (FMD). Brachial artery imaging coupled with reactive hyperemia is one of the most common techniques for measurement of vascular function [216]. This technique involves the occlusion of conduit vessels for ~ 5 min causing reactive hyperemia of the vascular bed. The subsequent dilatation that is expressed as a percentage increase from the baseline diameter is mainly dependent on NO activity, with a low percentage indicating poor endothelial function [217].

Laser Doppler flowmetry

Laser Doppler flowmetry (LDF) is an inexpensive, non-invasive method for measurement of skin microcirculation. Iontophoresis coupled with LDF makes it possible to assess the real-time changes of the skin blood flow after the administration of different vasoactive substances without systemic effects. ACh introduced into the skin by iontophoresis causes endothelium-dependent vasodilatation, which can be compared with the endothelium-independent vasodilator effect of sodium nitroprusside (SNP), an NO donor [218,219]. Sodium nitroprusside reacts with tissue sulfhydryl groups under physiologic conditions to produce NO directly and thereby stimulate SMC relaxation. ACh mediates vasodilatation via endothelial-dependent production of NO and/or prostanoids with a possible accessory role played by endothelium-derived hyperpolarizing factor. A Laser Doppler Flowmeter works by reading the frequency of the oscillation produced by the Doppler frequency shift of the red blood cells (RBCs) in a peripheral tissue and translates the frequency to an intensity oscillation. The apparatus is composed of a low-powered laser and probe that takes readings and sends results to an analyser. The apparatus can penetrate 1-4 mm of non-pigmented tissue. The light emitted and reflected is fed through optical fibers to the analyser-recorder. The output of the LDF is the flux of RBCs, defined as the number of RBCs times their velocity, which determines circulation.

Laser Doppler perfusion imaging

Laser Doppler perfusion imaging (LDPI) is based on the recording of Doppler shift caused by movements of RBCs in the backscattered light of a laser beam that successively scans a certain tissue area [220]. A laser Doppler imager generates a color-coded image of the spatial distribution of tissue perfusion. This allows measurement of tissue blood flow over a set area. Iontophoresis of small charged molecules is used to assess the skin microcirculation [221].

This technique is increasingly being adopted as a clinical measure of endothelial function, solely in the microcirculation rather than in the large conduit arteries. This is useful as the microcirculation may well be the initial site for endothelial damage [222,223].

Strain gauge plethysmography

Forearm blood flow can be assessed using strain gauge plethysmography [224]. ACh interacts with endothelial muscarinic receptors and stimulates the release of NO. Intra-arterial (brachial) infusion of ACh and SNP can be used to assess local endothelial-dependent and independent responses [225]. The rate of forearm swelling during the infusion, measured by strain gauge plethysmography, gives an estimate of the endothelial function.

Pulse wave velocity

Pulse wave velocity (PWV) is a measure of regional arterial stiffness of the arterial area between two measurement sites. PWV is related not only to the elastic modulus of the arterial wall (which represents the intrinsic stiffness of the wall), but also to the arterial geometry and blood density [226]. Although structural changes in the vessel wall are a major component of arterial stiffness (increased collagen and decreased elastin), the endothelium also appears to play an important dynamic role in arterial stiffness [227]. Evidence from both animal [228,229] and human studies [230,231] suggests that the endothelium is an important regulator of arterial stiffness, both functionally and structurally. Various interventions that reduce arterial stiffness may also improve endothelial function, particularly ACE inhibitors and statins.

Serum biomarkers

Many serum biomarkers have been shown to correlate with endothelial function including interleukin-6 (IL-6), tumor-necrosis factor-α (TNF-α), soluble P-selectin and soluble intercellular adhesion molecule-1 [232]. Plasma nitrite can be used an indicator of EDNO generation [233]. In addition, impaired tissue plasminogen activator (t-PA) secretion by the endothelium has been suggested as a useful measure of endothelial function [234]. The sensitivity of t-PA in this setting is apparently greater than that for other markers of endothelial dysfunction, including agonist- or flow-mediated vasodilatation, because it is measurable in some clinical conditions (e.g., hypertension) in which agonist-induced vasodilatation is unaltered. CRP is associated with several processes involved in endothelial dysfunction, these include increasing ET-1 synthesis, down-regulating eNOS, increasing the release of PAI-1 from ECs, and influencing endothelial progenitor cells [235,236]. Cellular adhesion molecules (CAMs) are expressed on the surface of activated ECs, and increased plasma concentrations of soluble CAMs are seen in patients with atherosclerosis [237].

Plasma asymmetric dimethylarginine (ADMA) is increased in patients with vascular disease, as well as in patients with risk factors for vascular disease [238-240]. In individuals with hypercholesterolemia, plasma ADMA levels provide a more accurate indication of endothelial dysfunction than LDL cholesterol levels [240]. ADMA inhibits the production of NO in cultured ECs and isolated blood vessels [241,242]. Hydrolysis of ADMA by the enzyme dimethylaminohydrolase (DDAH) produces l-citruline and methylamine [243]. It has been postulated that DDAH regulates NO synthase activity by controlling the concentration of ADMA [244]. Homocysteine inhibits the activity of endothelial DDAH, causing the accumulation of ADMA and the inhibition of NO synthesis [245]. Thus, plasma ADMA appears to be a promising indicator of endothelial dysfunction.

Endothelial function in Alzheimer’s disease: available evidence

Search strategy

A search of the published literature from January 1966 (MEDLINE) and from January 1989 (EMBASE) was conducted. By way of the advanced search method, the following terms were combined: ‘Alzheimer’s disease’, ‘cognitive impairment’, and ‘cognitive decline’, using the Boolean operator ‘OR’. The Boolean operator ‘AND’ was used to link these to the terms ‘endothelial dysfunction’, ‘vascular reactivity’, ‘flow-mediated dilation’, ‘plethysmography’, ‘doppler flowmetry’, and ‘pulse wave velocity’. Papers published until February 2013 were included. Books, abstracts and conference proceedings were excluded.

Eligibility of studies

This review sought publications with original data on humans suffering from dementia of the Alzheimer’s type. Articles without proper AD classification were excluded. Only articles in English were included, as the authors were familiar with these languages. Only papers investigating human subjects without a history of cardiovascular events or major VRFs were included. Since the endothelium possesses different physiologic functions - i.e. regulation of vessel tone and expression of adhesion molecules - the studies were organised into subheadings. The references in the selected articles were scanned for other relevant articles, using the same criteria.

Results

By searching MEDLINE and EMBASE 568 papers were found. Applying the above-mentioned criteria yielded a total of 15 articles. The main reasons for excluding papers were animal models, reviews and vascular or mixed dementia participants. Main findings of the studies for each subheading are discussed below. A summary of results is provided in Tables 1 and 2.

Table 1.

Selected papers using functional assessment to measure endothelial dysfunction in Alzheimer’s disease

Table 2.

Selected papers using serum biomarkers to measure endothelial function in Alzheimer’s disease

Regulation of vascular tone

Dede et al. evaluated endothelial function in 24 patients with AD-type dementia and 25 age-matched controls using flow-mediated dilatation [246]. Significantly lower FMD was found in the AD group, which indicated endothelial dysfunction (mean FMD: AD group, 3.4±1.7%; control group, 8.4±3.3%; P<0.001). When the correlation between Clinical Dementia Rating (CDR) and Mini-Mental State Examination (MMSE) scores and FMD values of the total sample (n = 49) was examined, FMD had a negative correlation with CDR score (r = -0.605, P<0.001) and a positive correlation with MMSE score (r = 0.560, P<0.001). Algotsson et al. measured skin vessel reactivity of 15 patients with AD and 16 age-matched controls using LDPI [247]. Compared to the controls, the skin vessel vasodilatation of the AD group was significantly impaired with respect to the reactions to ACh and isoprenaline but not SNP in all three tests. Khalil et al. measured microvascular responsiveness in 75 patients with AD and 168 age-matched controls using LDF [248]. The ratio of ACh response to saline (ratio E/S) was determined. Mean±SEM of ratios E/S were 8.8±0.9 for controls and 1.4±0.1 for AD patients, thus endothelial dysfunction was present in AD patients. Ferenc et al. evaluated flow motion pattern in forehead skin using LDF. No significant difference in was seen in occurrence of forehead vasomotion pattern between patients with AD (18%) and age-matched controls (32%) [249]. Occlusion (biological zero) values for control subjects and AD patients were 5.5±2.2 and 4.7±0.9 respectively. McGleenon et al. measured endothelial function in 9 male and 8 female patients with AD and 9 male and 12 female control subjects [250]. The basal FBF values were not significantly different for AD and controls (mean±(SE), 2.89±1.0 and 2.81±1.42 d, respectively).

Arterial stiffness

Hanon et al. measured the relationship between arterial stiffness and cognitive function in 126 AD patients and 80 age-matched controls and found significantly higher PWV in AD patients (13.3±2.9 m/s) than individuals without cognitive impairment (11.5±2.0 m/s; P<0.001) [251]. Dhoat et al. also assessed PWV in 16 patients with late onset-AD and 16 age-matched controls [252]. PWV in the muscular and elastic arteries were not statistically different between the two groups.

Expression of serum biomarkers

Borroni et al. found a significant difference in thrombomodulin levels (mean SD, control = 6.05±1.9; AD = 14.9±9.0; p<0.0001) and soluble E-selectin (sE-Selectin) (control = 30.3±18.2; AD = 50.0±28.5; p<0.001) between 30 patients with mild AD and 30 age-matched controls [253]. There was an inverse correlation between plasmatic TM levels and MMSE scores. Gupta et al. studied serum markers of coagulation and inflammation in patients suffering from dementia [254]. Serum concentrations of vWF, factor VII, IL-6, CRP and FDP were significantly higher in all three dementia groups (including the AD-type group) as compared to controls (p<0.001). Zuliani et al. tested endothelial dysfunction in 60 patients with late onset AD by measuring plasma levels of E-selectin and vascular cell adhesion molecule 1 (VCAM-1) [245]. VCAM-1 levels were significantly increased in patients with late onset AD than controls (control = 562 ng/ml (544-1143); AD = 708 ng/ml (471-676); p<0.03) suggesting endothelial function is impaired in AD [255]. Järemo et al. investigated P-selectin behavior in 23 persons diagnosed with moderate AD and 17 healthy elders without obvious memory problems [256]. Circulating P-selectin was analysed using an ELISA technique and flow cytometry was used to measure surface-bound P-selectin. Soluble P-selectin was augmented in AD (p = 0.019) but platelet membrane-attached P-selectin was not significantly different from controls. AD diagnosis was associated with less surface-bound P-selectin after provocation. Significant results were obtained when 74 μmol/L TRAP-6 (a thrombin-receptor activating peptide) was used as a platelet agonist (p = 0.0008). Selley measured plasma homocysteine, ADMA, and NO (nitrite and nitrate) concentrations in patients with AD and healthy controls [257]. AD patients showed higher plasma ADMA (AD = 1.84±0.08; control = 1.41±0.05 μmol/L, p<0.0001) and homocysteine concentrations, and lower NO concentrations, than controls. Abe et al. measured ADMA concentrations in the cerebrospinal fluid in AD patients and healthy controls [258]. AD patients had significantly lower ADMA concentrations in the cerebrospinal fluid (AD = 0.014±0.006; controls = 0.027±0.008 μmol/L, p = 0.0001). There was a significant correlation between lower cerebrospinal concentrations of ADMA and lower MMSE in patients with AD (r = 0.58, p<0.05). Mulder et al. also measured ADMA concentrations in the cerebrospinal fluid in AD patients and healthy controls [259]. Unlike the study by Abe et al. no significant differences were observed between the 2 groups (AD = 0.066±0.012; control = 0.065±0.010 μmol/L; p = 0.91) [258]. Finally, Arlt et al. measured ADMA concentrations both in plasma and in the cerebrospinal fluid in AD patients and healthy controls [260]. AD patients had higher plasma ADMA concentrations (AD = 0.65±0.11; control = 0.54±0.11 μmol/L; p<0.001), but lower cerebrospinal fluid ADMA concentrations (AD = 0.15±0.03; control = 0.22±0.12 μmol/L; P = 0.022), in comparison to healthy controls. However, these differences were not significant.

Discussion

The evidence reviewed suggests vascular involvement plays a role in the pathogenesis of AD. Morphological changes in the endothelium and the VBM, as well as diminished endothelial repair capacity have been recorded in AD [117]. Vascular abnormalities may be attributed to vascular disorders that frequently develop and progress with increasing age [57]. These vascular alterations probably facilitate several factors that are potentially neurotoxic. In particular, amyloid has received a great amount of interest as it is overproduced in AD and has been found to be neurotoxic [40,41]. In ageing, it is possible that vascular dysfunction may occur early on or even precede the pathological features of AD. As AD is a neurodegenerative disorder in which amyloid accumulation is believed a key event, it may be that an association exists between vascular changes and amyloid production.

Aβ may be produced by endothelial, neuronal and glial cells [187]. It has been postulated that Aβ deposition stimulates the production of ROS, which in turn cause vascular oxidative stress and cerebrovascular dysfunction [261]. In vitro findings are supported by studies indicating that cerebral ischemia causes stress-induced amyloid deposition in highly vascularized cortical brain regions [137,139]. These theories suggesting a possible synergistic relationship between Aβ deposition and vascular dysfunction may provide an explanation for endothelial dysfunction in AD.

In our systematic search, we found that studies investigating a possible relationship between endothelial dysfunction and AD in vivo were limited. The results of published studies show evidence of endothelial dysfunction in AD except for the studies by McGleenon et al., Malt et al., Dhoat et al., Arlt et al. and Mulder et al.. Although the measurement of endothelial dysfunction was quite diverse, elevated levels of serum biomarkers suggestive of endothelial dysfunction (TM, vWF, ADMA etc.), abnormal values of FMD and blood flow abnormalities assessed using Doppler shift were all associated with AD. These findings suggest that, independent of any other VRFs, endothelial function is present in AD, which could indicate a systemic element in AD pathology.

Another hypothesis is that only cerebral vascular function is compromised in AD [149]. The techniques reviewed in this study are all limited in their assessment of the peripheral circulation. However, there is currently no effective method of measuring endothelial function in the cerebral circulation. Moreover, McGleenon et al. found that cerebral vessels and peripheral vessels show similar responses to ACh infusion [250]. Nevertheless, the argument remains that whilst the peripheral circulation remains functional the cerebral circulation may be compromised in AD.

In vivo studies of endothelial function are fraught with difficulty and currently no gold standard exists. Many measurements that are purported to measure endothelial function actually measure “vascular reactivity.” Although many researchers consider FMD the gold standard for clinical research [207], it is not without its limitations [262]. FMD provides little mechanistic data and structural alterations of blood vessels in vascular disease which affect their ability to dilate may affect results. Moreover, lack of standardisation and variations in positioning of the arm cuffs and measurement of vessel diameter make comparing study results challenging [263]. Despite the similarities in vascular responses between the skin microcirculation and larger arteries, whether skin blood-flow studies (Doppler imaging and Doppler flowmetry) are indicative of other vascular beds is unclear and requires further research [264]. Strain gauge plethysmography is an invasive technique involving cannulation of the brachial artery, which is not without some risk. Thus, this is a specialist technique that is not suitable for widespread use. Further studies are required to fully comprehend the link between endothelial function and arterial stiffness. Although most studies suggest that the endothelium regulate stiffness, impaired elastic function itself might have adverse effects on endothelial function, and future work needs to address this issue [265]. Finally, the use of serum biomarkers to measure vascular disease or prognosis is still at an early stage in development.

The studies published so far cannot establish whether endothelial dysfunction in AD is causative or coincidental. It would be ideal to collect measurements before symptoms of AD occur in a population-based design. However, this would require a very large study population. Furthermore, none of the markers described are brain-specific endothelial markers and endothelial function may vary across vascular beds [264,266].

A major limitation in the literature on AD and endothelial dysfunction is the relative paucity of studies and small number of subjects in each study. A much larger population is necessary to prove any clinical significance. It may be difficult to recruit volunteers with AD for invasive studies due to volunteer frailty and lack of independent decision-making capacity in these individuals. Patients with AD often have vascular risk factors (VRFs) and it is challenging to find subjects without accompanying vascular conditions. However, only patients without existing VRFs were included in these investigations. Whilst this eliminates the possibility of the endothelial dysfunction of subjects being caused by conditions other than AD, it is well known that a high proportion of the AD population has vascular conditions, which may be associated with the pathogenesis of AD [57,267]. The evidence, therefore, may not be generalisable to the AD population as a whole. Future research investigating the relationship between AD and endothelial dysfunction should include subjects with and without VRFs to give a more accurate representation of the general AD population.

A final limitation of the reviewed studies was that they focused on serum markers of endothelial dysfunction (e.g. vWF) or on function assessment (e.g. FMD). For the future, it would be ideal to combine both methods to provide as robust an evidence base as possible.

Although this review was not a full systematic review, it aimed to provide a thorough analysis of evidence of endothelial dysfunction in AD. Strict selection criteria were employed with the aim to provide enough information to allow repeatability and enhance the validity of findings. However, it is possible that some of the published literature was not included. Whilst the studies reviewed suggest endothelial dysfunction may be present in AD, more research is required to prove this link.

Conclusion

There is conflicting evidence regarding the pathogenesis of AD. The vascular hypothesis challenges the long established neuron-centric view of AD. It is likely that AD has both neurodegenerative and vascular elements. Based on our findings, we suggest endothelial dysfunction may play a role in AD pathogenesis. More basic and clinical research on the vascular features of AD is required. Further study may have implications for therapeutic intervention with the potential to reverse or prevent endothelial dysfunction and relieve or prevent the symptoms of AD.

Disclosure of conflict of interest

None.