Expression and Processing of Amyloid Precursor Protein in Vascular Endothelium

Abstract

Amyloid precursor protein (APP) is evolutionary conserved protein expressed in endothelial cells of cerebral and peripheral arteries. In this review, we discuss mechanisms responsible for expression and proteolytic cleavage of APP in endothelial cells. We focus on physiological and pathological implications of APP expression in vascular endothelium.

Amyloid precursor protein (APP) together with amyloid precursor-like proteins (APLPs) APLP1 and APLP2 constitute the family of transmembrane glycoproteins highly expressed in the human brain, kidney, and platelets (64, 87, 110). APP is the only member of the family encoding amyloid-β (Aβ) peptides, which are considered major culprits in pathogenesis of Alzheimer's disease (AD) (73, 159, 162). The biological properties of APP have been the subject of intense investigation for almost 30 years (65, 87, 153, 172). Strong conservation of APP during evolution from invertebrates to humans implies that it serves an important cellular function(s) (128). Relevant to this review, APP is expressed in vascular endothelium of cerebral, coronary, and peripheral blood vessels (6, 42, 64, 94). In addition, expression and activity of enzymes responsible for proteolytic cleavage of APP (FIGURE 1) has been detected in endothelial cells of large conduit arteries, resistance arteries, and microvessels. However, vascular function of APP and its cleavage products is incompletely understood.

FIGURE 1.

Schematic diagram of non-amyloidogenic and amyloidogenic processing of APP₇₇₀ in endothelial cells

Left: in the non-amyloidogenic pathway, APP₇₇₀ is cleaved by α-secretase, yielding secretion of sAPPα ectodomain into the lumen of blood vessel wall and release of carboxyl-terminal fragment C83. C83 is further cleaved by γ-secretase within the transmembrane domain, yielding p3 and APP intracellular domain (AICD) peptide. Right: during amyloidogenic processing β-secretase cleaves APP₇₇₀ at the NH₂ terminus of the amyloid-β (Aβ) domain, thus generating sAPPβ and carboxyl-terminal fragment C99. The latter is cleaved by γ-secretase, thus producing Aβ peptides and AICD. Please note that APP₇₇₀ is a full-length protein containing Kunitz-type serine protease inhibitor (KPI) domain in tandem with an OX-2 antigen (OX2) domain in its extracellular region.

APP in Endothelial Cells

Three isoforms of APPs generated by alternative splicing of exons 7 and 8 have been identified: APP₆₉₅, APP₇₅₁, and APP₇₇₀ (number refers to the number of amino acids). APP₆₉₅ is predominantly expressed in neurons and is lacking two exons (87). APP₇₅₁ and APP₇₇₀ possess Kunitz-type serine protease inhibitor (KPI) domain encoded by exon 7, whereas APP₇₇₀ contains an additional immunoregulatory OX-2 antigen domain (OX2) encoded by exon 8 in their extracellular region and are expressed in brain and peripheral tissues (64, 147, 173). Endothelial cells express APP₇₅₁ and APP₇₇₀ (61, 94). Intriguingly, expression of these isoforms is higher in endothelial cells of cerebral blood vessels compared with peripheral arteries (94). Most importantly, in endothelial cells, APP could be processed by non-amyloidogenic and amyloidogenic pathway (FIGURE 1) (7, 61, 74, 94, 95, 199).

Soluble APPα

Under physiological conditions, APP is predominantly cleaved at the cell surface by α-secretase via non-amyloidogenic processing, thereby resulting in the secretion of soluble APPα (sAPPα) ectodomain into the lumen of blood vessel wall and abluminal release of carboxyl-terminal fragment C83 (95) (FIGURE 1; d'Uscio LV, Katusic ZS, unpublished observations). Within the transmembrane domain, C83 is further cleaved by the γ-secretase complex consisting of at least four proteins [presenilin-1 (PS1), nicastrin, anterior pharynx-defective 1, and presenilin enhancer 2], yielding an APP intracellular domain (AICD) peptide and a 3-kDa carboxyl-terminal fragment known as p3 (FIGURE 1). Thus α-secretase cleaves APP within the Aβ sequence, thereby preventing generation of Aβ (58). Several isoforms of a disintegrin and metalloproteinase (ADAM) are responsible for α-secretase activity: ADAM9, ADAM10, and ADAM17. In human cerebrovascular endothelium, ADAM10 (but not ADAM9 or ADAM17) appears to be the most relevant constitutively active α-secretase, even though all three isoforms are expressed in endothelial cells (78, 100).

It has been suggested that sAPPα possesses physiological functions in neuronal cells such as neuroprotection and synaptogenesis (96). The exact receptor(s) activated by sAPPα are not known, but existing evidence suggests that APP family members as well as putative cell surface receptors such as insulin-like growth factor 1 and/or insulin receptors may interact with sAPPα (67, 86, 121). Furthermore, sAPPα can induce production of cyclic guanosine 3′,5′-monophosphate (cGMP) in neuronal cells (10, 84); however, the exact signaling pathway responsible for this effect is unknown. Inadequate sAPPα levels can lead to increased amyloidogenic processing of APP, thus increasing susceptibility to oxidative stress in AD (86, 98, 132, 135). Indeed, overexpression of sAPPα protects the cells from excessive amyloidogenic processing of APP and accumulation of cytotoxic Aβ peptides (132). To our knowledge, there is no information available in current literature regarding the effects of sAPPα in endothelial cells.

Soluble APPβ

β-Site APP-cleaving enzyme 1 (BACE1 or β-secretase) is widely expressed in neuronal and nonneuronal cells, including endothelial cells (7, 52). During amyloidogenic APP processing, the BACE1 cleaves APP at the NH₂ terminus of the Aβ domain, thus generating soluble APPβ (sAPPβ) and carboxyl-terminal fragment C99 (129) (FIGURE 1). The latter is further cleaved by the γ-secretase generating Aβ peptides, most commonly Aβ_1-40 and Aβ_1–42, and APP intracellular domain AICD (93, 115). Under ischemic conditions, increase in cholesterol levels in the lipid membrane enhances enzymatic activity of BACE1 in brain endothelial cells (16, 17, 107). Although sAPPβ shares the same sequence as sAPPα, apart from the last 16 COOH-terminal amino acids, its neuroprotective effect is much less potent (27). In some studies, however, it has been reported that sAPPβ exerts a pro-apoptotic effect in neuronal cells (129). The functional relevance of sAPPβ production by vascular endothelium is unknown.

Aβ

Aβ is a short 4-KDa hydrophobic peptide consisting of 36–43 amino acids, and it is generated by amyloidogenic processing of APP. The main species of Aβ released from endothelial cells and platelets is Aβ_1–40, which contributes to formation of vascular Aβ deposits, whereas the predominant form in neuronal plaques is Aβ_1–42 (79, 154, 165). Moreover, elevated local concentrations of Aβ_1–40 and Aβ_1–42 are reported to be similarly neurotoxic, but they clearly differ in their toxicity to endothelial cells. Aβ_1–40 was found to be less toxic to human aortic endothelial cells compared with Aβ_1–42 or Aβ_25–35 (44, 169).

AICD

AICD is generated by both non-amyloidogenic and amyloidogenic processing of APP (FIGURE 1) and plays an important role as a transcriptional regulator (136, 163). During evolution, this small fragment (6.5 kDa) was highly conserved from fly to human, suggesting a significant role in physiology and/or preservation of homeostasis under pathological conditions (11). In neurons, AICD is very labile and has a rapid turnover, with a half-life of ∼5–30 min (39, 92). However, AICD can be stabilized through binding to several adaptor proteins, e.g., Fe65, and subsequently translocate to the nucleus and modulate gene expression (127). Whether AICD serves an important physiological function in endothelial cells remains to be determined in future studies.

p3

The p3 peptide produced exclusively during non-amyloidogenic processing of APP corresponds to residues 17–42 at the carboxyl end of Aβ. There is no evidence for the physiological vascular function of p3. Of note, studies in zebrafish indicate that p3 does not contribute to vascular defects detected in APP-deficient embryos (113).

Functions of APP in the Cardiovascular System

APP as Coagulation Inhibitor

Several reports have shown that APP is abundantly expressed in platelets and that it is secreted in the circulation upon platelet activation, thus contributing to ∼90% of circulating sAPPα/sAPPβ (19, 36, 62, 95, 110, 183). The predominant forms of APP in platelets are APP₇₅₁ and APP₇₇₀, whereas APP₆₉₅, which is the most abundant in neuronal tissue, is nearly undetectable in platelets (19, 62, 95, 111). The KPI domain present in both APP₇₅₁ and APP₇₇₀ isoforms has been reported to be an inhibitor of activated coagulation enzymes such as platelet coagulation factors IXa and XIa, which are known to be involved in prothrombin activation (158, 166). Upon platelet activation, sAPP is secreted from α-granules and inhibits coagulation factors, thus preventing thrombosis in cerebral, coronary, and peripheral arteries (105, 191, 193). Indeed, studies in APP-deficient mice revealed that thrombosis was accelerated by genetic inactivation of APP (193). Of note, activation of protein kinase C stimulates sAPPα release by platelets (164). Thus sAPPα released from platelets may function in both cerebral and peripheral vascular system as an anticoagulant and hence makes an important contribution to the regulation of hemostasis (35, 105, 164, 165). Production and release of sAPPα from endothelium also contributes to protection against coagulation (see below) (31, 78, 157).

APP in Angiogenesis

Angiogenesis is dependent on proliferation, migration, growth, and differentiation of endothelial cells. APP is highly expressed in the endothelium during fetal life, suggesting an important role for APP and/or its metabolites in early angiogenesis (134). The residues 1–16 of Aβ are essential for vascular activity of Aβ because impaired angiogenesis observed in APP-deficient zebrafish could be restored by Aβ treatment (113). Additionally, angiogenesis in the brain is reduced by β- or γ-secretase inhibitors (138). Several studies reported that low nanomolar concentrations of either Aβ_1–40 or Aβ_1–42 promote angiogenesis by increase of growth, migration, and tube branching in cultured cerebral and peripheral endothelial cells (15, 22, 23). In contrast, supraphysiological concentrations (micromolar) of Aβ_1–40 and Aβ_1–42 impair angiogenesis and accelerate senescence of endothelial cells in vitro (54, 77, 117, 145).

It is well established that vascular endothelial growth factor (VEGF) plays a fundamental role in normal and abnormal angiogenesis (60). Indeed, increased expression of VEGF-A and increased cerebral microvessel density have been detected in the brain of different APP transgenic mouse models overexpressing mutated APP and in brains of AD patients (13, 32, 176). Increased local concentration of Aβ appears to contribute to increased angiogenesis in murine models of AD and in patients suffering from the disease. In contrast, angiotensin II-induced hypertension in APP/PS1 transgenic mice overexpressing mutated human APP and PS1 proteins further increased cerebral vascular Aβ deposits accompanied by a reduction in cerebral microvessel density and a decrease in VEGF-A expression in the brain (32). Moreover, cell culture studies reveal that Aβ also inhibits VEGF-induced tyrosine phosphorylation of VEGF receptor 2 as well as VEGF-stimulated phosphorylations of Akt and endothelial nitric oxide (NO) synthase (eNOS) in endothelial cells, indicating that high concentrations of Aβ act as a VEGF antagonist (32, 77, 145).

Role of NO and Prostacyclin in Expression and Processing of Endothelial APP

Role of NO

There are only few studies regarding regulation of APP processing in endothelial cells (summarized in Table 1). Recently, it was recognized that, in human brain microvessel endothelial cells, a complete loss of endothelial NO increases protein expressions of APP and BACE1 as well as production and release of Aβ peptides (7). These observations were confirmed in cerebral microvessels of young eNOS-deficient mice (7) as well as in cultured neurons (103). Moreover, NO-mediated suppression of APP and BACE1 protein expression is mediated by the cGMP signaling pathway (7). Supplementation of NO increases cGMP levels and in turn attenuates increased expression of APP and BACE1 in cerebral microvessels of eNOS-deficient mice, suggesting that the NO/cGMP pathway plays an important role in modulating expression and processing of APP (8). Importantly, cerebral blood flow is unchanged in eNOS-deficient mice (4), indicating that changes in NO rather than alteration in blood flow is responsible for increased expression and amyloidogenic processing of APP. Consistent with this conclusion, Aβ level is also increased in cerebral arteries and arterioles in aged heterozygous eNOS-deficient mice (171). These observations may help to explain the contribution of endothelial dysfunction caused by the loss of endothelial NO production to development of AD pathology, including cerebral amyloid angiopathy (9, 171).

Table 1.

Modulators of APP processing in endothelial cells

		APP Processing Enzymes and Products
Effectors/Stimuli	Cell Types	APP	Aβ	Mixed sAPPα/β	sAPPα	sAPPβ	α-secretase	β-secretase	Molecular Mechanisms	References
Nitric oxide	HBMEC	↓	↓					↓	eNOS/cGMP activation	7, 8
Iloprost	HBMEC	↑	↔		↑		↑	↔	Via cAMP and PPARδ activation	78
Cicaprost, forskolin	HBMEC	↑					↑		cAMP activation	78
GW501516	HBMEC	↔			↑		↑	↔	SIRT1 activation	78
Thrombin	HUVEC			↑					PKC activation	31
Interleukin-1	HBMEC, MBEC, HUVEC	↑		↑					Increased APP mRNA via PKC activation	20, 61, 66, 95
Glucose	HUVEC	↑	↑						?	26
TNF-α	MBEC		↓ (intra-cellular)						Decreased mRNA and protein synthesis of cellular prion protein	196
Heat shock	HUVEC	↑							Increased mRNA and protein synthesis	30
Oxygen-glucose deprivation	RBMEC	↑	↑					↑	Via hypoxia inducible factor-1	16, 17
Hypoxia	MBEC	↑	↑		↑	↑↑			Increased protein synthesis of APP	126

HBMEC, human brain microvessel endothelial cells; HUVEC, human umbilical vein endothelial cells; MBEC, mouse brain endothelial cell line; RBMEC, rat brain microvessel endothelial cells; GW501516, peroxisome proliferator-activated receptors δ (PPARδ) agonist; eNOS, endothelial nitric oxide synthase; cGMP, cyclic guanosine 3′,5′-monophosphate; cAMP, cyclic adenosine 3′,5′-monophosphate; SIRT1, sirtuin 1; PKC, protein kinase C; TNF-α, tumor necrosis factor-α; ↑, upregulation; ↓, downregulation; ↔, no effect.

Blank fields in this table represent that this parameter was not tested. No data are available regarding the effects of effectors/stimuli on APP-processing products AICD and p3 in endothelial cells.

Given the anti-inflammatory effect of NO under physiological conditions (46), decreased bioavailability of NO can increase local concentration of inflammatory factors and thrombin, which are present in eNOS-deficient mice brains and in microvessels isolated from AD patient brain cortexes (9, 68, 70, 171). Thrombin and interleukin-1 can, in turn, further exacerbate expression of APP in endothelial cells via activation of protein kinase C (31, 66) (Table 1). Moreover, tumor necrosis factor-α decreases expression of cellular prion protein in brain microvessel endothelial cells and thus reduces Aβ clearance from the brain (196).

Role of Prostacyclin

Endothelium-derived prostacyclin (PGI₂) is a potent vasodilator and anticoagulant and plays an essential role in preservation of cerebrovascular homeostasis under pathological conditions (59, 124, 160, 186). The actions of PGI₂ are mediated by activation of PGI₂ receptor (IP receptor) as well as peroxisome proliferator-activated receptor δ (PPARδ). A recent study shows that activation of IP receptor and cyclic adenosine 3′,5′-monophosphate (cAMP) signaling pathway by PGI₂ stimulates expression of APP in brain microvessel endothelial cells (78) (Table 1). Remarkably, PGI₂ increases expression and activity of non-amyloidogenic processing enzyme α-secretase ADAM10 and secretion of sAPPα. The amyloidogenic processing pathway is unaffected by PGI₂ in brain microvessel endothelial cells (Table 1). Furthermore, deletion of PPARδ abolished the iloprost-induced ADAM10 activation and release of sAPPα in brain microvessel endothelial cells, thus demonstrating that PGI₂-induced generation of sAPPα is also dependent on activation of PPARδ (78).

Collectively, cell culture and knockout mice studies indicate that both NO and PGI₂ possess differential roles in expression and processing of APP in cerebral arteries. NO suppresses amyloidogenic processing of APP, whereas activation of PGI₂ stimulates non-amyloidogenic processing of APP. It appears that both NO and PGI₂ exert vasoprotective effects on APP metabolism by minimizing production of Aβ peptides and by promoting generation of sAPPα, respectively. Of note, enhanced production of PGI₂ in eNOS-deficient mice is considered a compensatory mechanism designed to maintain normal vascular function during NO deficiency (146, 168). Consistent with this concept, it appears likely that higher concentrations of PGI₂ promote non-amyloidogenic processing of APP, thereby antagonizing excessive production of Aβ generated by elevated activity of β-secretase in eNOS-deficient mice.

APP Processing and Endothelial Dysfunction

Effects of Overexpression of Wild-Type APP

Increased APP expression and Aβ production in neuronal cells has been shown to be induced by high levels of stress hormones such as catecholamines, glucocorticoids, and central administration of angiotensin II (71, 106, 200). In endothelial cells, high glucose increases expression and amyloidogenic processing of APP (26). In addition, cell surface localization of APP is increased during ischemia, cellular stress, or inflammation in endothelial cells, indicating that increased APP expression may serve as a homeostatic response to stress (17, 30, 125, 126) (Table 1). It is interesting to note that shear stress induced by unidirectional pulsatile laminar flow can increase transcription of APP in human vascular endothelial cells, further supporting the idea that APP is a stress response protein (51). In addition, overexpression of wild-type APP₆₉₅ in neuronal tissue exerts a neuroprotective effect during ischemic injury induced by middle cerebral artery occlusion (34). In contrast, APP knockout mice experience high mortality in response to cerebral ischemia, suggesting that loss of APP or APP cleavage fragments increases vulnerability to ischemia (97). The exact mechanisms underlying protective function of APP are incompletely understood and remain to be determined.

Effects of Overexpression of Mutated APP

A number of transgenic mouse lines overexpressing APP have been developed to study the pathogenesis of AD and associated endothelial function. For example, transgenic mice overexpressing double Swedish mutation of human APP at K670N/M671L (Tg2576) display elevated Aβ levels both in brain and in plasma (42, 89). An increased oxidative stress has been observed in these young mice, thereby causing endothelial dysfunction and reduced bioavailability of NO in cerebral and peripheral arteries (42, 139–141, 156). This effect is mediated by the increased production of superoxide anion via the activation of NADPH oxidase (42, 75, 140). Furthermore, scavenger receptor CD36, an endothelial membrane glycoprotein that binds Aβ, is required for the oxidative stress induced by Aβ (120, 141). Indeed, genetic deletion of CD36 rescues endothelial function from oxidative stress observed in Tg2576 mice cerebral arteries, demonstrating that CD36 inactivation is vasoprotective (141).

Intracarotid administration of Aβ_1–40 (but not Aβ_1–42) inhibits the increase in cerebral blood flow caused by endothelium-dependent vasodilators in wild-type mice, indicating that elevated Aβ_1–40 levels alone are sufficient to impair endothelial function observed in Tg2576 (130, 141, 142). On the other hand, endothelial dysfunction is less pronounced in transgenic mice expressing human neuronal APP harboring the Swedish K670N/M671L and Dutch/Iowa E693Q/D694N mutations (Tg-SwDI) that display elevated Aβ levels in brain but not in plasma (45). Interestingly, administration of exogenous Aβ_1–40 further worsens endothelial dysfunction in Tg-SwDI but not in Tg2576 mice (which already have high circulating Aβ_1–40 levels), again suggesting that circulating Aβ_1–40 levels contribute to the alterations of endothelial function in cerebral arteries (142).

Increased formation of peroxynitrite, which is generated from the reaction of superoxide anion with NO, can easily oxidize tetrahydrobiopterin, an essential cofactor for enzymatic activity of eNOS (122). The consequences of high concentration of peroxynitrite are decreased levels of tetrahydrobiopterin in cerebral and peripheral arteries and consequent uncoupling of eNOS increase eNOS-derived superoxide anion production leading to impairment of endothelium-dependent relaxations (41, 42, 139, 155, 156, 181). In vivo treatment with tetrahydrobiopterin prevents eNOS uncoupling and decreases superoxide anion concentration in cerebral microvessels of Tg2576 mice (156), thereby suggesting that uncoupling of eNOS contributes to oxidative stress induced by Aβ peptides.

Effects of Exogenous Aβ on Cultured Endothelial Cells

Aβ possess both pro-oxidant and antioxidant properties (5). Aβ generation at lower physiological concentrations (∼0.1 nM) has been shown to exhibit antioxidant properties by prevention of autoxidation of lipoproteins in CSF and low-density lipoprotein in plasma (99). Aβ_1–40 possesses the highest antioxidant capacity compared with other Aβ isoforms. Neuroprotective properties at low nanomolar concentrations have been described in prior studies (5, 99, 188).

In vitro studies in endothelial cell culture employed supraphysiological concentrations of Aβ (micromolar range; summarized in Table 2). For example, long-term exposure to high Aβ concentrations promotes Ca²⁺ influx into endothelial cells, thereby causing injury and apoptosis (194). Furthermore, long-term Aβ treatment can increase generation of reactive oxygen species, thus causing oxidative modifications of proteins and lipid membranes, leading to reduced endothelial function (Table 2). Aβ peptides can also cause expression of inflammatory genes and proteins (68, 184), which in turn can increase thrombin expression in cerebral endothelial cells (69, 198). Interestingly, one study shows that short-term treatment with low concentration of Aβ_1–40 stimulated production of kinins and cGMP in endothelial cells (190) (Table 2). There is evidence that Aβ interacts with the receptor for advanced glycation end products (RAGE) that critically regulates Aβ transport into brain vascular endothelial cells and can trigger intracellular signaling (24, 25, 109). Indeed, low concentration of Aβ increases intracellular production of hydrogen peroxide (25), which is known to activate soluble guanylate cyclase and production of cGMP in endothelial cells (18, 43, 91).

Table 2.

Effects of Aβ on cultured endothelial cells

Cell Types	Aβ Type	Pre-Aggregated Aβ Used	Concentration Range Tested	Incubation Time	Minimum Concentration Required	Molecular Mechanisms of Aβ	References
HBMEC	Aβ1–42	Yes	0.0125–1.25 μM	1 h	0.125 μM	Upregulation of C-C chemokine receptor type 5 via JNK, ERK, PI3K activation through RAGE	109
HBMEC	Aβ1–42	Yes	0.001–10 μM	24 h	0.01 μM	Increased production of hydrogen peroxide through RAGE	25
HBEC	Aβ1–40	Yes	1–20 μM	2, 4, 8, 12 h	5 μM	Increase in inflammatory markers MCP-1, IL-8, IL-6, GRO via AP-1 transcription	184
MCEC	Aβ25–35	Yes	2.5–40 μM	24 h	20 μM	Decreased heat shock protein 90, p-Akt, telomerase reverse transcriptase, cyclin-dependent kinase 4	29
	Aβ1–40
MCEC	Aβ25–35	—	25 μM	24 h	25 μM	Activation of AP-1/pro-apoptotic protein Bim and Smac release	197
MCEC	Aβ25–35	—	0.01–50 μM	24 h	1 μM	Apoptosis via activation of caspase-8/caspase-3, mitochondrial dysfunction	194
MCEC	Aβ25–35	Yes	20 μM	24 h	20 μM	Activation of apoptosis signal-regulating kinase 1 leading Bax increase via phosphorylations of p53 and p38-MAPK	80
Rat brain EC	Aβ25–35	—	47–380 μM	72 h	188 μM	Cell toxicity via increased LDH release and decreased glucose consumption	149
Mouse brain EC line	Aβ1–40	—	0.3 μM	30 min	0.3 μM	PARP activation via oxidative-nitrosative stress, increase in intracellular Ca2+	144
BAEC	Aβ1–42	No	0.05–2.2 μM	1–5 min	0.22 μM	Increased Ca2+-influx	12
BAEC, RMCEC	Aβ1–40	—	0.1–1 μM	30 min	0.1 μM	Stimulation of cGMP production, release of kinins	190
BAEC	Aβ25–35	—	1 μM	24 h	1 μM	Impaired activation of K+-channels and reduced nitrite production	28
BAEC	Aβ25–35	—	0.1–1000 nM	24 h	1 nM	Reduced nitrite production	150
	Aβ1–40	—	0.1–1000 nM	24 h	10 nM
BAEC	Aβ25–35	Yes	1 μM	24 h	1 μM	Production of reactive oxygen species, blockade of agonist-stimulated eNOS phosphorylation at Ser1179, decreased nitric oxide production	104
	Aβ1–42		5 μM	24 h	5 μM
HAEC	Aβ25–35	—	Up to 40 μM	30 min to 24 h	5 μM	Apoptosis, necrosis via reactive oxygen species, increase in intracellular Ca2+	169
	Aβ1–42		Up to 32 μM		4 μM
Porcine PAEC	Aβ25–35	Yes	1–50 μM	18 h	20 μM	Apoptosis, increased free radicals production, dysregulation of Ca2+ homeostasis, impaired glucose uptake	14
				24 h	5 μM

EC, endothelial cells; HBMEC, human brain microvessel endothelial cells; HBEC, human brain endothelial cells; MCEC, mouse cerebral endothelial cells; BAEC, bovine aortic endothelial cells; RMCEC, rat microvascular coronary endothelial cells; HAEC, human aortic endothelial cells; PAEC, pulmonary aortic endothelial cells; Aβ, amyloid-β; RAGE, receptor for advanced glycation end products; Akt, protein kinase B; MAPK, mitogen-activated protein kinase; AP-1, activator protein 1; JNK, c-Jun NH₂-terminal kinase; ERK, extracellular signal-regulated kinase; PI3K, phosphoinositide 3-kinases; MCP-1, monocyte chemoattractant protein-1; IL-6, interleukin-6; IL-8, interleukin-8; GRO, growth-regulated oncogene; LDH, lactate dehydrogenase; eNOS, endothelial nitric oxide synthase; cGMP, cyclic guanosine 3′,5′-monophosphate; -, unknown or not tested.

In brain endothelial cells, mitochondria are very important sensors of oxidative stress because they are the primary site of reactive oxygen species formation (133). In vitro experiments indicate that stimulation of brain endothelial cells with Aβ peptide causes nitration of the manganese form of superoxide dismutase and mitochondrial dysfunction, resulting in the loss of their ability to support their antioxidant function (2, 3, 55, 194) (Table 2).

Aβ can exist as monomer, dimer, oligomer, protofibril, and/or fibrillar aggregates. However, mouse Aβ is less aggregable and toxic than human Aβ (171). It is important to note that low concentrations of Aβ do not form oligomer or fibrils, and freshly prepared Aβ also contains only a very small percentage of short fibrils (12, 195). In contrast, elevated Aβ concentrations in solutions (≥1 μM) can easily form oligomers and fibrils that have been shown to exert cytotoxicity in vitro and in vivo (44, 82, 83, 194, 197) (Table 2). In addition, both Aβ peptides Aβ_1–40 and Aβ_1–42 can form dimmers or octamers after 24 h of incubation at 37°C (90). In summary, the vascular effects of Aβ are dependent on type of Aβ peptide, aggregation status, incubation time, and concentration of Aβ (Table 2).

Effects of Exogenous Aβ on Isolated Arteries

In vitro vasomotor studies regarding direct effects of Aβ in isolated cerebral and peripheral arteries are summarized in Table 3. Short-term incubation of arteries with high concentrations (micromolar) of Aβ peptides increases the vasoconstrictions to phenylephrine, serotonin, and endothelin-1 (38, 72, 137, 177). The molecular mechanisms responsible for these observations include increased production of prostaglandins via activation of cyclooxygenase (COX-2) as well as increased production of reactive oxygen species. Furthermore, in most studies, incubation of cerebral and peripheral arteries with soluble Aβ_1–40 impairs endothelium-dependent relaxations. Increased reactive oxygen species and decreased eNOS activity are most likely explanations for endothelial dysfunction since free-radical scavenger superoxide dismutase is able to normalize Aβ-induced endothelial dysfunction (53, 130, 177, 179) (Table 3). In contrast, one study was unable to demonstrate increased production of reactive oxygen species induced by Aβ_1–40, suggesting that oxidative stress is not involved in endothelial dysfunction induced by Aβ_1–40 (63). The reason for the discrepancy may be due to the different Aβ preparations. The limitations of short-term in vitro studies (Table 3) are that the concentrations of exogenous Aβ are much higher than circulating Aβ levels reported in transgenic mice overexpressing mutated human APP or in patients with AD.

Table 3.

Effects of Aβ on isolated blood vessels

Blood Vessel Types	Aβ Types	Concentration Range Studied	Incubation Time	Minimal Concentration Required	Cellular and/or Molecular Mechanisms of Aβ	References
Rat cerebral artery	Aβ1–40	0.0001–1 μM	15 min	0.1 μM	Impaired endothelium-dependent relaxation to ACh	150
	Aβ25–35	0.0001–1 μM	15 min	0.01 μM
Mouse cerebral artery	Aβ1–40	0.01–10 μM	30 min	1 μM	Decreased cerebral blood flow to ACh	130, 131
	Aβ1–42	0.01–10 μM	30 min	n/a	No change
Rat basilar artery	Aβ1–40	1 μM	30 min	1 μM	Impaired endothelium-dependent relaxation to substance P	63
Rat penetrating cerebral arterioles	Aβ1–40	0.001–1 μM	—	0.1 μM	Reduction of tone diameter by increased contraction; decreased endothelium-dependent relaxation to ATP	53
	Aβ1–42
Human middle cerebral artery	Aβ1–40	1 μM	5 min	1 μM	Increased contractions to endothelin-1 via COX-2 and p38 mitogen-activated protein kinase activation	137
Human middle cerebral artery	Aβ1–40	2 μM	2–6 h	2 μM	Increased production of PGF2α and PGE2	137
Bovine middle cerebral artery	Aβ1–40	1 μM	15 min	1 μM	Impaired endothelium-dependent relaxation to bradykinin.	179
Rat coronary artery	Aβ1–40	1 μM	15 min	1 μM	Increased production of reactive oxygen species; impaired endothelium-dependent relaxation to ACh	178
Mouse aorta	Aβ25–35	1 μM	24 h	1 μM	Increased vasoconstriction to serotonin; activation of α1-adrenergic receptor	72
Rat aorta	Aβ1–40, Aβ1–42	1 μM	10 min	1 μM	Increased contractions to endothelin-1	38
Rat aorta	Mixed (?) Aβ1–39, Aβ1–40, Aβ1–42	1 μM	15 min	1 μM	Increased production of reactive oxygen species; impaired endothelium-dependent relaxation to ACh; increased contractions to phenylephrine	177
Rat aorta	Aβ1–40	1 μM	30 min	1 μM	Impaired endothelium-dependent relaxation to ACh, decreased eNOS phosphorylation at Ser1177; decreased eNOS activity; increased PKC phosphorylation at Ser660	63

Aβ, amyloid-β; ACh, acetylcholine; PKC, protein kinase C; COX-2, cycloogygenase-2, PGF_2α, prostaglandin F_2α; PGE₂, prostaglandin E₂.

Clearance of Aβ

In healthy humans, the average circulating levels of Aβ are in the picomolar range (85, 154). Moreover, the concentrations of Aβ are about 10-fold higher in cerebrospinal fluid (CSF) than in mice and human plasma (85, 89). Circulating Aβ levels decline in APP transgenic mice as well as in control individuals with aging and AD. This phenomenon is caused by increased Aβ deposition in the brain (85, 89). Therefore, the highest circulating concentration of Aβ measured in AD patient is ∼50 nM (101, 102).

It is important to note that steady-state levels of Aβ peptides are not only regulated by APP processing but also by the clearance of Aβ through the blood-brain barrier (BBB) (89, 118, 151, 161). The BBB is mainly formed by the cerebral vascular endothelial cells that line the capillaries of the brain. The unique phenotype of the BBB is characterized by intercellular tight junctions, adherens junctions, and numerous membrane receptors and transporters allowing trade of substances between systemic circulation and extracellular fluid compartment of the brain (57). Aβ peptide is cleared from the brain by receptor-mediated transport across the endothelium, by enzymatic degradation, and by periarterial drainage along vascular basement membranes (50, 76, 185). Three BBB-related transport proteins for Aβ peptide have been identified: lipoprotein receptor protein-1 (LRP-1), P-glycoprotein (P-gp), and RAGE (33, 49, 195). Reduction of endothelial expression of either LRP-1 or P-gp in BBB is associated with impaired clearance of Aβ peptide across the BBB into the circulation (33, 161). Under inflammatory conditions, antioxidant treatment preserves LRP-1 function and Aβ efflux, suggesting that LRP-1 can operate independently of P-gp in BBB (56). On the other hand, RAGE binds and transports circulating Aβ peptide toxins across the BBB into the brain (49). Chronic vascular insult, such as hypertension and/or inflammation, can activate RAGE in brain vascular endothelial cells, favoring Aβ deposition (24, 109). Moreover, there is recent evidence that Aβ peptide is also catabolized in peripheral tissues and organs such as liver, kidney, gastrointestinal tract, and skin (192), suggesting that brain-derived Aβ peptide is physiologically cleared through the capillary network of these organs.

Implications for Atherosclerosis and Cerebral Amyloid Angiopathy

APP and Atherosclerosis

Atherosclerosis is considered a chronic inflammatory disease, which can be triggered by cardiovascular risk factors such as hypercholesterolemia, aging, and hypertension (112). Endothelium-derived vasoactive factors play an important regulatory role in vascular homeostasis due to the strategic position of the endothelium between the vascular smooth muscle cells (VSMC) and the circulating blood (40, 114). Several studies have demonstrated that APP and Aβ are increased in apolipoprotein E (ApoE)-deficient mice atherosclerotic aortas, in advanced human carotid artery plaques, and in plasma of patients with coronary heart disease (6, 48, 167). The exact cause for increased amyloidogenic processing of APP during progression of atherosclerosis is unknown; however, inflammatory cytokines such as interleukin-1 have been reported to stimulate APP expression in endothelial cells (30, 61, 66). Overexpression of mutated human APP stimulates development of fatty streak lesions (early atherosclerosis) and endothelial dysfunction in the aorta of Tg2576 mice (42, 108). The presence of APP in endothelium is also important for monocytes adhesion to endothelial cells (6). Consistent with these observations, overexpression of mutated human APP accelerated aortic atherosclerotic development in ApoE-deficient mice (180). Conversely, genetic deletion of the APP gene attenuated atherogenesis in ApoE-deficient mice without affecting cholesterol levels (182). However, the major limitation of the study in double APP/ApoE knockout mice is that only histological studies were performed, and plaque formation was reduced only in distal thoracic and abdominal aorta but not in the aortic valves, root, and arch (182). Furthermore, existing literature suggests that atherosclerosis develops much slower in intracranial arteries compared with peripheral arteries (152). A number of reasons could be responsible for vascular heterogeneity in atherosclerotic pathology (reviewed in Ref. 152). In aggregate, these studies suggest that excessive amyloidogenic progressing of APP is pro-atherosclerotic. However, more precise analysis employing endothelial-specific overexpression or deletion of APP is required to define the exact role of endothelial APP in pathogenesis of atherosclerosis.

APP and Cerebral Amyloid Angiopathy

Cerebral amyloid angiopathy (CAA) is manifested by increased deposition of Aβ peptides in the wall of cerebral arteries, and it is a common pathological feature of AD. Indeed, abnormally high local levels of Aβ pep tide can damage the cerebral arteries by deposition and aggregation in the vasculature, thus leading to vascular disintegration and micro-hemorrhage (47, 81, 148, 170). Existing evidence suggests that neuronal cells are a dominant source of Aβ (21, 45); however, endothelial cells and VSMC are also able to generate Aβ (37, 52). In this regard, it is important to note that BACE-1 is predominantly localized on the abluminal membrane of brain endothelial cells, thereby suggesting that Aβ generation in endothelium may considerably contribute to amyloid formation in the blood vessel wall (52). As already mentioned, VSMC also expresses APP protein and all required enzymes (α-, β-, and γ-secretases) for non-amyloidogenic and amyloidogenic processing of APP (37). Oxidative stress can increase BACE1 expression in VSMC and, in turn, release of Aβ_1–40 and Aβ_1–42 (37). There is evidence that Aβ is normally eliminated by the perivascular drainage, thereby preventing deposition of Aβ in the vascular wall (187). Nevertheless, excessively high concentration of Aβ in cerebral blood vessels can lead to a loss of VSMCs, resulting in the vascular distention and weakening of the vascular wall (88, 116, 119, 189). Recent research using animal models suggests that Aβ decreases adhesion of VSMC to the basement membrane and that increased deposition of Aβ in cerebral capillaries may be associated with occlusion and disturbance of cerebral blood flow (123, 175). Genetic deletion of CD36 or treatment with anti-Aβ antibody reduces CAA and micro-hemorrhages in Tg2576 mice cerebral arteries (143, 174). Furthermore, cerebrovascular protection by antioxidant therapy is dependent on a decrease in CAA formation as well as a direct reduction in CAA-induced vasomotor impairment (1, 75). Reduction of Aβ load in endothelial cells (and VSMC) appears to be a promising strategy in the prevention and treatment of CAA.

Future Directions

A number questions regarding expression and function of APP in blood vessels remain to be answered. First, vascular function of APP has to be better defined. Second, cellular signaling and the roles of APP cleavage products under physiological and pathological conditions have to be identified. Third, signal transduction mechanisms responsible for control of APP expression and processing have to be identified. Fourth, heterogeneity in vascular expression and function of APP in different brain regions and different size vessels deserve additional attention. Fifth, contribution of endothelial APP to pathogenesis of vascular and neuronal diseases (including AD) remains to be established. In aggregate, these advances may provide new targets for development of badly needed therapies for prevention and treatment of cerebrovascular disease and dementia.