Endothelium-specific deletion of amyloid-β precursor protein exacerbates endothelial dysfunction induced by aging

The physiological function of amyloid precursor protein (APP) in the control of endothelial function during aging is unclear. Aortas of young (4-6 months old) and aged (23-26 months old) wild-type (WT) and endothelium-specific APP-deficient (eAPP−/−) mice were used to study aging-induced changes in vascular phenotype. Unexpectedly, aging significantly increased protein expression of APP in aortas of WT mice but not in aortas of eAPP−/− mice thereby demonstrating selective upregulation APP expression in vascular endothelium of aged aortas. Most notably, endothelial dysfunction (impairment of endothelium-dependent relaxations) induced by aging was significantly exacerbated in aged eAPP−/− mice aortas as compared to age-matched WT mice. Consistent with this observations, endothelial nitric oxide synthase (eNOS) protein expression was significantly decreased in aged eAPP−/− mice as compared to age matched WT mice. In addition, protein expression of cyclooxygenase 2 and release of prostaglandins were significantly increased in both aged WT and eAPP−/− mice. Notably, treatment with cyclooxygenase inhibitor, indomethacin, normalized endothelium-dependent relaxations in aged WT mice, but not in aged eAPP−/− mice. In aggregate, our findings support the concept that aging-induced upregulation of APP in vascular endothelium is an adaptive response designed to protect and preserve expression and function of eNOS.


INTRODUCTION
Vascular endothelium releases several factors involved in the local regulation of vascular tone and modulation of pro-inflammatory molecule production [1]. Among them are endothelium-derived relaxing factors in particular nitric oxide (NO) as well as endotheliumderived contracting factors such as prostaglandins and superoxide anion [2,3]. Constitutive expression of endothelial nitric oxide synthase (eNOS) is mainly responsible for NO production in endothelial cells of conduit arteries [4,5]. On the other hand, cyclooxygenase (COX) is a rate-limiting enzyme in the biosynthesis of prostaglandins. COX1 is constitutively expressed and plays an important role in vascular homeostasis, while COX2 is considered an inducible enzyme in mouse conduit arteries [6].
Aging of blood vessels are major contributors to the development of cardiovascular disease [7]. Existing evidence obtained in human and animal studies suggest that the earliest detectable vascular phenotype in aging involves the development of endothelial dysfunction, which is commonly characterized by the reduced endothelial production of NO and by the increased production of COX-derived vasoconstrictor factors [8][9][10]. Moreover, clinical studies have reported that endothelium-dependent vasodilatation progressively declines with age and occurs earlier in men than in women [11,12].
Amyloid-β precursor protein (APP) is an evolutionarily conserved protein and is implicated in the development of Alzheimer's disease [13][14][15]. However, the exact function of APP outside the central nervous system, AGING particularly in peripheral tissues, remains poorly understood [16][17][18]. We and others have shown that APP is expressed in cultured endothelial cells as well as in vascular wall of wild-type mice [19][20][21][22][23][24]. Under physiological conditions, APP is constitutively cleaved at the cell surface by α-secretase via non-amyloidogenic processing pathway. Alpha-processing of APP causes release of soluble APP alpha (sAPPα) ectodomain in the lumen of blood vessel wall [25][26][27][28][29][30]. Recently, we showed that endothelium-specific inactivation of APP (eAPP −/− ) caused endothelial dysfunction in mice cerebral arteries [29]. However, no previous studies tested the effects of aging on vascular APP expression in systemic large conduit arteries. In the present study, we advance the hypothesis that APP exerts vascular protective effects on endothelial function during aging.

Characterization of mice
Body weight was increased in both aged wild-type (WT) and eAPP −/− mice (P<0.05 vs. respective young mice; Table 1). Blood pressure measurements revealed that systolic blood pressure (SBP), mean blood pressure (MBP), and diastolic blood pressure (DBP) were unchanged in young and aged WT and eAPP −/− mice (P>0.05; Table 1). Blood glucose was also not different between eAPP −/− mice and their WT irrespective of age (Table 1). Lipid profile was unchanged in plasma of young WT and eAPP −/− mice (P>0.05; Table 1). Aging caused a slight but significant increase in circulating levels of cholesterol in WT mice (P<0.05 vs. young WT; Table 1). In contrast, levels of triglycerides were decreased in aged eAPP −/− mice (P<0.05 vs. young eAPP −/− mice; Table 1). Plasma levels of norepinephrine were significantly increased in aged WT and eAPP −/− mice (P<0.05 vs. respective young mice; Table 1). Aging significantly increased plasma levels of sAPPα in both WT and eAPP −/− mice however, levels of sAPPα remained significantly decreased in eAPP −/− mice as compared to age-matched WT mice (P<0.05; Table 1). However, plasma levels of amyloid-β 1-40 (Aβ 1-40 ) were unaltered in young and aged WT and eAPP −/− mice (P>0.05; Table 1).

Vascular contraction responses
Contractions to KCl were significantly increased in aged WT and eAPP −/− mice aortas as compared to young mice (P<0.05; Table 2) while there were no differences between age-matched WT and eAPP −/− mice (P>0.05; Table 2).
Sensitivity to prostaglandin F 2α (PGF 2α )-induced contractions was slightly lower in aortas of young and aged eAPP −/− mice (P<0.05 as compared to their respective WT mice; Figures 2A, 2B, respectively; Table 2). In contrast, PGF 2α potency was significantly increased in aged WT and eAPP −/− mice (P<0.05 vs. respective young mice; Table 2), and efficacy of PGF 2α was significantly decreased in aged WT and eAPP −/− mice (P<0.05 vs. respective young mice; Table 2). While no significant differences between age-matched WT and eAPP −/− mice were observed, the potency of phenylephrine was reduced in aged WT and eAPP −/− mice (P<0.05 vs. respective young mice; Table 2; Figure 2C, 2D). Furthermore, the efficacy was significantly reduced in aged WT mice (P<0.05 vs. young WT mice; Table 2) while it tended to be reduced in aged eAPP −/− mice (P=0.088; Table 2).

Vascular endothelial function
Acetylcholine (ACh)-induced endothelium-dependent relaxations were unaltered in young mice with endothelial-specific APP deletion (P>0.05 vs. young WT littermates; Figure 3A and Table 3). Aging caused a significant impairment to ACh in WT mice (P<0.05 vs. young WT mice; Table 3) while the efficacy was unaltered (Table 3). Furthermore, relaxations to ACh were further decreased in aged eAPP −/− mice (P<0.05 vs. aged WT mice; Figure 3B and Table 3).
Incubation of aortas with indomethacin did not affect relaxations to ACh in young mice (P>0.05; Figure 3C). In contrast, indomethacin significantly improved endothelium-dependent relaxations to ACh in both aged    AGING WT and eAPP −/− mice (P<0.05 vs. aortas without indomethacin; Table 3). However, the efficacy is still significantly reduced in aged eAPP −/− mice in the presence of indomethacin (P<0.05 vs. aged WT mice with indomethacin; Figure 3D and Table 3). Blockade of NOS with N ω -nitro L-arginine methyl ester (L-NAME) in the presence of indomethacin completely blocks relaxations to ACh in both young and aged WT and eAPP −/− aortas (Supplementary Figure 3).

Ex-vivo production of prostaglandins in aorta
We next determined whether age-related changes in vascular prostaglandins contribute to the observed decline in endothelial function of aortas, we measured release of prostaglandins in conditioned medium. Thromboxane B 2 (TXB 2 ) and prostaglandin E 2 (PGE 2 ) productions were significantly increased in both aged WT and eAPP −/− mice (P<0.05 vs. young mice; Figure  4A, 4B) while production of PGF 2α was significantly increased only in aged WT mice (P<0.05 vs. young WT; Figure 4C). In contrast, production of 6-keto PGF 1α (a stable metabolite of prostaglandin I 2 ) was not different among young and aged WT and eAPP −/− mice (P>0.05; Figure 4D).
We also performed western blot analyses for protein expression of COX isoforms in aortas. COX1 expression was not different between WT and eAPP −/− mice irrespective of age (P>0.05; Figure 5A). In contrast, protein expression of COX2 was significantly increased in both aged WT and eAPP −/− mice (P<0.05 vs. young mice; Figure 5B). Moreover, the increase in COX2 was significantly reduced in aged eAPP −/− mice (P<0.05 vs. aged WT mice; Figure 5B).  mice; Figure 6). Furthermore, eNOS protein tended to be reduced in aged WT mice but it did not reach statistical significance ( Figure 6).

Superoxide anion
To investigate whether endothelial dysfunction in aging was caused by increased superoxide anion production in aorta, we performed HPLC-based assay of 2-hydroxyethidium. Levels of superoxide anion were significant-ly enhanced in aged eAPP −/− mice (P<0.05 vs. young eAPP −/− mice; Figure 7) while superoxide anion levels were unaltered in aged WT mice (P>0.05 vs. young WT mice; Figure 7).

Intracellular cGMP levels
Under basal conditions, cyclic guanosine 3',5'monophosphate (cGMP) were unchanged in aortas of WT and eAPP −/− mice irrespective of age (P>0.05;  Western blot results are the relative densitometry compared with β-actin protein (n=10 per group). All results are representing box plots with whiskers showing the median, 25 th to 75 th percentiles, and min-max range. * P<0.05 versus young mice of same strain (two-way ANOVA followed by Tukey's HSD test). AGING Figure 8). ACh stimulated cGMP levels were significantly increased in young and aged WT and eAPP −/− mice (P<0.05 vs. basal levels; Figure 8). However, ACh stimulated cGMP levels were significantly decreased in aged WT and eAPP −/− mice (P<0.05 young mice in the presence of ACh; Figure 8). In contrast, cGMP levels did not differ between WT and eAPP −/− mice groups under basal and stimulated conditions (P>0.05; Figure 8).

Ex-vivo studies with cytokines
We next attempted in different experimental conditions to provide insight into the effects of APP deletion on  eNOS expression in vascular endothelium. Young WT and eAPP −/− mice aortas were used to examine the effects of pro-inflammatory cytokines cocktail on expression of APP and eNOS. Western blot analysis revealed that APP protein was significantly increased in young WT mice treated with cytokines (P<0.05 vs. PBS treatment; Figure 9A) while APP expression was not significantly affected in young eAPP −/− mice (P>0.05; Figure 9A). In contrast, eNOS expression was significantly reduced in young eAPP −/− but not in young WT aortas after cytokines treatment (P<0.05 vs. PBS treatment; Figure 9B). Expressions of iNOS and COX2 were equally increased in both groups of mice (P<0.05 vs. PBS treatment; Supplementary Figure 4).

DISCUSSION
There are several major findings in this study. First, protein expression of endothelial APP and production of sAPPα in endothelium of aged WT mice are increased as compared to aged eAPP −/− mice. Second, impairment of endothelium-dependent relaxations to acetylcholine induced by aging is exacerbated in eAPP −/− mice aortas. Third, eNOS expression was decreased whereas levels of superoxide anion were enhanced in aortas of aged eAPP −/− mice. Fourth, pro-inflammatory cytokines increased APP protein expression only in isolated aortas of young WT mice. In contrast, pro-inflammatory cytokines decreased expression of eNOS exclusively in young eAPP −/− mice. Taken together, the results suggest that during aging endothelial APP exerts vascular protective effects. These effects appear to be mediated by preservation of eNOS and endothelium-dependent vasodilator function.
In the present study we demonstrated for the first time that APP expression and sAPPα release were increased in aged wild-type mice aortas. Importantly, these ageinduced changes in expression and processing of APP were not observed in the aorta of eAPP −/− mice thus demonstrating that endothelial cells are a major site of increased APP expression and production of sAPPα. Of note, Austin and colleagues reported that APP expression is increased mainly in intimal layer of both human and apolipoprotein E-deficient mice atherosclerotic aortas [21]. Importantly, protein expression of mature α-secretase ADAM10 was unchanged in aged aortas. Whether enhanced α-processing of APP could be explained by the increased enzymatic activity of ADAM10, or some other α-secretases, remains to be determined [27].
Prior study reported that hypertension causes shift towards amyloidogenic APP β-processing and development of cerebral amyloid angiopathy [31]. However, we detect neither blood pressure increases nor changes of aortic protein expression of β-secretase, BACE1, and production of Aβ 1-40 in aged wild-type and eAPP −/− mice indicating that amyloidogenic processing of APP in aorta is not affected by aging. Thus, our study demonstrates that upregulation of APP expression in aged wild-type mice is not induced by alterations of arterial blood pressure.
It is not clear what are the exact mechanisms responsible for increased expression and α-processing of APP in large conduit arteries of aged WT mice. Relevant to our observations, it is well known that aging is characterized by a state of chronic, low-grade inflammation [32,33]. Moreover, expression of inflammatory cytokines such as TNFα, INFγ, and/or IL-1ß have been reported to be elevated in arteries of aged mice [34][35][36][37]. We speculate that the presence of proinflammatory stimuli is the most likely explanation for increased expression and non-amyloidogenic processing of APP. Consistent with this hypothesis, TNFα and IL-1ß increase APP expression and secretion of sAPPα from endothelial cells [19][20][21]38]. In the present study, we also demonstrated that ex-vivo treatment of isolated aortas with cocktail of inflammatory cytokines led to increased APP expression in young WT mice but not in eAPP −/− mice again indicating that endothelium is a major source of APP. The concentrations of cytokines used for ex-vivo studies were higher than published serum cytokines levels detected in-vivo in young control mice [39]. Whether aging significantly increases circulating levels of cytokines tested in our study remains to be determined. Nevertheless, it is conceivable that the observed upregulation of APP represents an adaptative/protective response of endothelial cells to aging-induced inflammation [40,41].
Although plasma levels of sAPPα were lower in eAPP −/− mice as compared to WT mice, aging increased sAPPα levels not only in WT mice but also in eAPP −/− mice. This indicates that other cell types besides endothelial cells must contribute to elevation of circulating sAPPα. It is well known that APP is also expressed in platelets and cleaved by nonamyloidogenic APP processing thereby releasing sAPPα upon activation [38,42,43]. Notably, platelets count is significantly increased in aged wild-type C57BL/6J mice [44]. Increased release of sAPPα from platelets may help explain elevated circulating levels of sAPPα in aged eAPP −/− mice.
Circulating levels norepinephrine were increased in both aged WT and eAPP −/− mice despite normal blood pressure. This may explain the observed reduced responsiveness to phenylephrine in aging. Chronic exposure of the peripheral vasculature to high circulating levels of norepinephrine causes compensatory downregulation of reactivity to phenylephrine. Indeed, similar observations were made in healthy human subjects showing that α-adrenergic vasocontraction responsiveness to increased circulating norepinephrine levels is reduced with age in healthy men [45].
Aging also alters arachidonic acid metabolism in endothelial cells and these alterations can profoundly affect vasomotor function [46,47]. Human and animal studies have established that enhanced production of vasoconstrictor prostaglandins can impair endotheliumdependent relaxations during aging [8,10]. In subjects older than 60 years, the treatment with COX-inhibitor, indomethacin, potentiates endothelium-dependent vasodilation to ACh thus demonstrating that vasoconstrictor prostaglandins contribute to impairment of endothelium-dependent relaxations in humans [10]. In line with these studies, we presented evidence that indomethacin significantly improved endotheliumdependent relaxations in response to ACh in both aged WT and eAPP −/− mice aortas. Consistent with these observations, production of TXA 2 and PGE 2 was significantly increased in both aged WT and eAPP −/− mice. The reason why the production of PGF 2α was increased only in aged WT littermates is unclear and remains to be determined. However, this could be related to different expression levels of COX2 between aged WT and eAPP −/− mice (see below). In contrast, the production of 6-keto PGF 1α remained unchanged in aged conduit arteries demonstrating that the metabolism of prostaglandin I 2 , which causes direct vasodilation in smooth muscle cells, is unaffected.
Interestingly, our study revealed that protein expression of COX2 was significantly increased in both aged wildtype and eAPP −/− mice while protein expression of COX1 remained unchanged. Although COX1 primarily contributes to basal vascular production of prostaglandins, COX2 also contributes to the production of prostaglandins [48]. Indeed, it has been reported that COX2 is expressed in endothelial cells and that inhibition of eNOS unmasks the ability of ACh to elicit endothelium-dependent contractions which are sensitive to inhibition of COX2 [49]. Thus, increased production of vasoconstrictor prostaglandins by COX2 appears to be responsible for reduced endothelium-dependent relaxations in aged mice. Notably, this phenomenon was not affected by genetic deletion of APP in endothelium.
To determine the reason why endothelium-dependent relaxations to ACh were still impaired in aged eAPP −/− AGING mice despite treatment with indomethacin, we studied eNOS protein expression. Interestingly, eNOS protein was decreased in the aorta of aged eAPP −/− mice but not in young eAPP −/− mice. In contrast, expression of eNOS is unchanged in aged WT mice aortas and this is in line with the previous reports [50,51]. Moreover, eNOS expression is not altered in conduit and resistance arteries of different mouse models of cardiovascular disease despite presence of inflammatory cytokines [52][53][54]. These observations are confirmed and further extended by our ex-vivo studies of isolated aortas treated with cytokines cocktail showing that eNOS expression was downregulated in young eAPP −/− mice, but not in young WT mice. From these findings we concluded that during aging, the presence of APP in aortic vascular endothelium is essential for normal expression and function of eNOS.
Superoxide anion levels were increased in aged eAPP -/aortas but not in aged WT aortas. Increased ROS generation and/or reduced antioxidant capacity could contribute to the increased levels of superoxide anion. However, insufficient NO production in aged eAPP -/mice can also lead to enhanced production of superoxide anion thereby further exacerbating impairment of endothelium-dependent relaxations ( Figure 10). Indeed, several studies provided evidence that NO inactivated superoxide anion by rapid nonenzymatic reaction with superoxide anion [55][56][57][58].
We were unable to observe any changes in superoxide anion production in conduit arteries derived from aged WT mice under basal conditions and this is in line with existing literature [37,51,59]. In contrast, several studies reported increased levels of superoxide anion in aged wild-type mice arteries [50,60,61]. The reason for this discrepancy is unclear. One possible explanation is that vascular superoxide production is increased in very old (31 months old) wild-type mice [37,61].
Elevation of aortic cGMP levels induced by ACh are decreased in both aged WT and eAPP −/− mice as compared to young mice aortas while basal cGMP levels were not different between WT and eAPP −/− mice. As mentioned earlier, the endothelium is a major source not only of NO but of prostaglandins as well [3]. Relevant to our observations, endogenous prostaglandins have been shown to decrease cGMP levels and to induce contractions in response to ACh in the aorta [62,63] thus suggesting that exaggerated production of prostaglandins during aging can impair ACh-stimulated production of cGMP. At the present time we do not have an explanation as to why levels of cGMP were not decreased in aged eAPP −/− mice.
In aggregate, the present study demonstrates that endothelial dysfunction (loss of NO) is more pronounced in aged eAPP −/− mice than in aged WT littermates. We also provide evidence that the absence of APP in endothelial cells causes downregulation of eNOS expression in eAPP −/− mice. Our findings support the concept that up-regulation of endothelial APP and production of sAPPα are adaptive mechanisms designed to protect blood vessel wall from detrimental effects of inflammation associated with aging. In this regard, preservation of NO production in endothelium of aged aortas appears to be critically important function of endothelial APP.

Experimental animals
APP flox/flox ;Tie2-Cre − mice (wild-type [WT] littermates) and APP flox/flox ;Tie2-Cre + (eAPP −/− ) mice were generated and genotyped in our laboratory as described previously [29]. All mice were maintained on 12h/12h light/dark cycle and on standard chow with free access to drinking water. Institutional Animal Care and Use Committee of the Mayo Clinic reviewed and approved the experimental design. In addition, the protocols complied with the guidelines of ARRIVE and NIH for care and use of laboratory animals. Young male mice (4-6 months old) and aged male mice

Glucose and lipid levels
Blood was collected through cardiac puncture and was transferred to a tube containing EDTA. Glucose was determined immediately using Accu Check ® (Roche Diagnostics, Indianapolis, IN). Thereafter, blood samples were centrifuged for 10 min at 2000 rpm (4° C) and plasma was obtained and stored at -80° C. Plasma levels of cholesterol, HDL, and triglyceride were measured as described previously [24].

Determination of plasma norepinephrine levels
Blood was placed in a tube containing EGTA with reduced glutathione. After centrifugation, plasma was stored at -80° C until assayed. After extraction with activated alumina norepinephrine levels were determined by HPLC with electrochemical detection as described [65].

Determination of ex-vivo release of sAPPα and Aβ 1-40
Ten millimeters long aortas were opened longitudinally and were incubated in minimal essential medium (MEM) in the presence of bovine serum albumin (0.1%), penicillin (100 U/mL) and streptomycin (100 µg/mL; GIBCO ® Thermo Fischer Scientific Inc., Waltham, MA) for 24 hours at 37° C in CO 2 incubator. Thereafter, the cultured medium was collected and stored at -80° C until assayed. Levels sAPPα and Aβ  in conditioned medium were determined using the ELISA kits as described above and all results were normalized against protein levels of aortas.

Determination of ex-vivo release of prostaglandins
Ten millimeters long thoracic aortas were opened longitudinally and were incubated in modified MEM in the presence of 0.1% BSA, 100 U/mL penicillin and 100 µg/mL streptomycin for 24 hours (37° C in CO 2 incubator). Thereafter, the cultured medium was collected and stored at -80° C. Commercially available colorimetric ELISA kits (Cayman, Ann Arbor, MI) were used for measurements of TXB 2 (a stable metabolite of TXA 2 ), PGE 2 , PGF 2α , and 6-keto PGF 1α AGING concentrations in the medium. All results were normalized against aortic protein levels.

Determination of intracellular superoxide anions
Intracellular levels of superoxide anion were quantified in fresh aortic rings using 2-hydroxyethidium standard by HPLC-based fluorescence detection and normalized against tissue protein levels as described in detail in previous studies [67,68].

Determination of cGMP
Twelve millimeters long aortas were incubated for 2 hours at 37° C in MEM containing bovine serum albumin (0.1%), penicillin (100 U/mL) and streptomycin (100 µg/mL). A phosphodiesterase inhibitor 3-isobutyl-1-methylxanthine (100 µmol/L; Sigma) was also added to prevent degradation of cyclic nucleotides. At the end of incubation time, ACh (10 µM) was added for 2 minutes. Thereafter, all samples were immediately placed in liquid N 2 and stored at -80° C until assayed. Colorimetric cGMP ELISA immunoassay (Cell Biolabs Inc., San Diego, CA) was used to determine cGMP levels and the results were normalized against aortic protein concentrations.

Western blot analysis
Dissected aortas were homogenized in lysis buffer using glass grinder and pestle 20 as previously described [68]. After centrifugation, protein levels were determined in supernatants using DC protein assay kit (BioRad, Hercules, CA). For each sample, 50 µg protein was loaded on 7.5% or 10% TGX SDS-PAGE gels (BioRad  Figure 5). In addition, C57BL/6J mouse heart was used as positive control for COX1, and whole cell lysate RAW 264.7 (cat# ab7187, Abcam) was used to identify COX2 and ADAM10 bands (Supplementary Figure 6). All blots were reprobed with β-actin antibody (1:50,000, A5316, Sigma). Densitometry analyses were performed in Odyssey Fc imaging system with Image Studio ™ 5.2 software (Li-Cor Biotechnology, Lincoln, NE).

Statistical analysis
All results are presented as mean ± standard error of mean (SEM; n indicates the number of mice used per experiment). The efficacy and potency of the drugs were determined using LabChart Pro Dose Response Module (ADInstruments Inc., Colorado Springs, CO). Concentration-response curves were compared by analysis of variance (ANOVA) for repeated measurements followed by Bonferroni's correction as described [29]. Two-way ANOVA followed by Tukey's HSD test and unpaired t-test were used for multiple and single comparison, respectively (JMP Pro 14.1 software; SAS Institute, Cary, NC). Differences of P<0.05 were considered statistically significant.

AUTHOR CONTRIBUTIONS
Livius d'Uscio: experimental design, data collection and analysis, and manuscript writing. Zvonimir Katusic: experimental design, data interpretation, and manuscript writing.

CONFLICTS OF INTEREST
The authors declare that they have no conflicts of interest.