Amyloid-β Precursor Protein APP Down-Regulation Alters Actin Cytoskeleton-Interacting Proteins in Endothelial Cells

The amyloid-β precursor protein (APP) is a ubiquitous membrane protein often associated with Alzheimer’s disease (AD) and cerebral amyloid angiopathy (CAA). Despite its role in the development of the pathogenesis, APP exerts several physiological roles that have been mainly investigated in neuronal tissue. To date, the role of APP in vasculature and endothelial cells has not been fully elucidated. In this study, we used molecular and proteomic approaches to identify and investigate major cellular targets of APP down-regulation in endothelial cells. We found that APP is necessary for endothelial cells proliferation, migration and adhesion. The loss of APP alters focal adhesion stability and cell–cell junctions’ expression. Moreover, APP is necessary to mediate endothelial response to the VEGF-A growth factor. Finally, we document that APP propagates exogenous stimuli and mediates cellular response in endothelial cells by modulating the Scr/FAK signaling pathway. Thus, the intact expression and processing of APP is required for normal endothelial function. The identification of molecular mechanisms responsible for vasoprotective properties of endothelial APP may have an impact on clinical efforts to preserve and protect healthy vasculature in patients at risk of the development of cerebrovascular disease and dementia including AD and CAA.


Immunoprecipitation (IP)
HUVEC cells were cultured in normal growth conditions until they reached confluence. Cells were quickly washed twice with ice-cold PBS, lysed in in 1% Triton lysis buffer and spun at 16,000× g for 20 min. 300 µg of cleared lysate was immunoprecipitated for 2 h at 4 • C under gentle rotation with 50 µL/sample of Protein-G DynaBeads (Thermo Fisher Scientific, Waltham, MA, USA), preincubated in 4 µg/sample of anti-VEGFR2 antibody (Cell Signaling #9698). Beads were washed 3 times with PBS, resuspended in 20 µL of 1× loading buffer and boiled for 10 min at 70 • C. Samples were analyzed by western blot as described above using anti-APP and anti-VEGF antibodies.

Wound Healing Scratch Assay
ECs migration was assessed using an in vitro wound healing assay as previously reported [20]. Briefly, cells were silenced for 48 h as previously described. Transfected cells were then harvested and seeded on a 24-mutiwell plate (1 × 10 5 cells/well) and incubated under their normal growth conditions (EGM-2, 10% FBS) until they reached complete confluence (18-24 h). A sterile 1000 µL micropipette tip was used to scrape the confluent monolayer and create the wound. Wells were washed twice with PBS and cells were exposed to EBM-2 and the indicated treatment (0.1% FBS; 10% FBS or 50 ng/mL VEGFa). 2.5 mg/mL ARA-C (Cytosine β-D-arabinofuranoside; Sigma-Aldrich) was added to the wells to suppress cell proliferation. Images of the wound in each well were acquired from 0 h to 8 h, 18 h and 24 h under a phase contrast microscope at 10x magnification. Finally, cells were fixed and stained with the PanReac kit. Results were quantified using ImageJ software, and data are reported as % of scratch closure normalized to siCtrl.

Cell Adhesion Assay
HUVEC were silenced as previously described; after 48 h, cells were harvested and seeded in triplicate in 96-well plates (2 × 10 4 cells/well) pre-coated with fibronectin (3 µg/mL) or collagen I (1 µg/mL) as previously described [23]. For short-term adhesion assays, attached cells were quantified after 1 h; for de-adhesion assays, cells were incubated for 18 h in normal growth conditions. At the end of the assay period, cultures were rinsed with two gentle washes with PBS (with Ca/Mg), fixed for 5 min with 4% paraformaldehyde and stained for 20 min with 0.5% Crystal Violet in distilled water for 20 min. Excess crystal violet was removed, and the cells were washed with water. The plates were dried and the stain extracted by 100 µL of methanol for 5 min. Absorbance was read at 540 nm in a plate reader (Infinite 200 Pro, SpectraFluor, Tecan, Männedorf, Switzerland).

In Vitro Permeability Assay
Cells were silenced for 48 h, harvested and seeded (2 × 10 5 cells/well) on gelatin-coated insert membranes (0.4 µm diameter pores, Corning, New York, USA), and the inserts were placed in 12-multiwell plates and incubated for 48 h in normal growth conditions (EGM-2, 10% FBS). Confluent monolayers were then starved with EBM-2 with 0.1% FBS for 4 h and treated with IL-1β (10 ng/mL) for 6 h [24]. FITC-Dextran (3 kDa, 10 µm) was used as a fluorescent marker of permeability, which was evaluated after 15 min by measuring the fluorescence in a plate reader (Infinite 200 Pro, SpectraFluor, Tecan, Männedorf, Switzerland) at 485/535 nm (excitation/emission). Data are reported as fold change of fluorescence units (RFU), taking as reference the siCtrl sample in the condition of medium EBM-2 with the addition of 0.1% serum (control condition).

LC-MS/MS Proteomic and Bioinformatic Analysis
Control and silenced HUVEC were lysed in 2% Sodium deoxycholate (SDC)/100 mM ammonium bicarbonate, reduced with 5 mM tris-(2-carboxyethyl)-phosphine (TCEP) and alkylated in the dark with 10 mM iodoacetamide (IAA). Protein quantification was assessed using Pierce-BCA protein assay kit. 60 µg of proteins, for each sample, were processed adding trypsin (1:40) and incubated at 37 • C overnight. All reaction mixtures were acidified with 1% formic acid (FA) [25,26]. Digested samples were desalted using OASIS cartridges and reconstituted in 0.1% formic acid in water/acetonitrile (97/3, v/v). LC-MS/MS analyses were performed using a Q-Exactive Plus Orbitrap mass spectrometer (Thermo Fisher Scientific). These experiments were carried out using a DDA setting to select the "top twelve" most-abundant ions for MS/MS analysis. MS data analysis was conducted using a Proteome Discover 2.1 (Thermo Fisher Scientific). Functional annotation analysis was performed using DAVID 6.8 [27] and FunRich 3.1.3 [28]. The default peak-picking settings were used to process the raw MS files in MaxQuant (version 1.6.1.0), Perseus (version 1.6.1.1) and its integrated search engine Andromeda [29,30]. The protein quantification and calculation of statistical significance was carried out using two-way Student-t test and error correction (p value < 0.05) with the method of Benjamini-Hochberg. For further visualization, a Principle component analysis (PCA) [31], a heatmap graphic analysis with clustering tree and, to detect differentially expressed proteins, a volcano plot analysis were performed.

Statistical Analysis
Results are expressed as means ±SEM. Statistical analysis was generated by GraphPad software Prism 7 (San Diego, CA, USA). Student's t-test and Mann-Whitney U-test were applied to determined statistical differences between conditions. p < 0.05 was considered statistically significant.

Loss of APP Affects Endothelial Cells Proliferation, Migration and Cytoskeleton Organization In Vitro
To study the physiological role of APP in ECs, we knocked down APP gene expression in HUVEC cells using an siRNA strategy. We started our experimental approach by selecting three different siRNAs, two targeting the 3'UTR region (siAPP-A and siAPP-B) and one targeting the APP mRNA coding sequence (siAPP-C). All three siRNAs (siAPP-A, siAPP-B and siAPP-C) knocked down significantly APP mRNA and protein expression after 48 h post transfection, as confirmed by RTqPCR analysis and western blot (WB) ( Figure 1A,B and Figure S1A,B). In mammals, the APP gene family includes, besides APP, two genes encoding the APP-like proteins (APLP1 and APLP2) that share similar structural organization and partially overlapping functions [32]. To confirm the specificity of APP knockdown, we checked relative mRNA expression levels of APP homologues genes APLP1 and APLP2 upon silencing with siAPP-A, -B or -C respectively, and observed that siAPP-C was the only siRNA that didn't show any off-target effect ( Figure 1A and Figure S1C,D). Moreover, the siAPP-C was the only siRNA with no effect on cell survival, as both siAPP-A and siAPP-B significantly induced caspase-3 activation after 48 h transfection ( Figure S1E,F). Taking into account these findings, we therefore used siAPP-C as the specific siRNA to knockdown APP (hereinafter referred to as siAPP).
We then investigated the effect of APP silencing on cell proliferation and migration of HUVEC cells. The MTT assay showed a reduction of cell proliferation 24 h after siRNA transfection in ECs silenced for APP (siAPP) compared to control siRNA treated cells (siCtrl) ( Figure 1C). The wound healing scratch assay showed normal migration at 8 h after scratch and a significant reduction of migration rate at 18 and 24 h after scratch ( Figure 1D).
To understand if the migratory defect was due to cytoskeletal disorganization, we labeled control cells (siCtrl) and APP-silenced cells (siAPP) with an anti-APP antibody and F-actin stress fibers marker (Phalloidin). APP immunoreactivity was observed at the cell membrane, in the cytoplasm and in the Golgi of siCtrl HUVEC, and was expectedly lost in siAPP cells. APP-silenced cells appeared bigger and flatter and the phalloidin staining showed an altered organization characterized by shorter and thicker actin stress fibers, suggesting a potential role of APP in controlling assembly, disassembly or rearrangement of cytoskeletal structures ( Figure 1E).
Altogether, our data suggest that APP is involved in ECs migration, proliferation and a correct cytoskeleton organization.

Proteomic Analysis Confirms the Cytoskeleton Organization Defect and Reveals Alteration of Vascular Specific Pathways in APP-Silenced Cells
To investigate further the effect of APP on ECs biology, we performed a proteomic analysis to gain a comprehensive and quantitative description of changes in protein expression that occur in HUVEC cells upon APP silencing. We used a free-label quantification (LFQ) approach to directly compare relative abundance of proteins obtaining an unbiased insight into protein expression changes. To identify and quantify proteins, liquid chromatography followed by mass spectrometry (LC-MS) was performed on control HUVEC (siCtrl) and HUVEC silenced for APP (siAPP). Cells were silenced for APP for 48 h and samples were collected 24 h after siRNA transfection, maintaining the same experimental conditions of the results previously described.
Protein identification and quantification was performed with MaxQuant using the Uniprot_Homo sapiens (proteome:up000005640) database (search parameters were 20 ppm tolerance on peptides, 0.02 Da on fragments, and less than 1% false discovery rate, FDR). We identified 1225 and 1396 proteins for three biological replicates of siCtrl and siAPP, respectively ( Figure S2A). To gain insight into the global similarities and differences between the six groups (siCtrl1, siCtrl2, siCtrl3 and siAPP1, siAPP2, siAPP3), we first performed an unsupervised clustering analysis that showed a differential clustering between siCtrl (siCtrl1, siCtrl2, siCtrl3) and siAPP (siAPP1, siAPP2, siAPP3) samples ( Figure S2B). The principle component analysis (PCA) of all replicates of siCtrl and siAPP samples confirmed the clear proteomic differentiation between the two datasets ( Figure S2C).
We then performed a functional comparison between the siCtrl common dataset and siAPP common dataset using FunRich software to analyze differences (p value < 0.001) in cellular components and biological processes (Figure 2A,B and Figure S2D). We observed a reduction, in the expression of proteins involved in membrane (GO:0016020) and cytoskeleton (GO:0005856) cellular components ( Figure S2D). In line with the previously described functional data, we observed a decrease of proteins associated with cell proliferation (GO:0008283) in the siAPP sample compared to the control ( Figure 2B). We also found an evident reduction of proteins involved in cytoskeleton organization (GO:0030036) in the silenced sample ( Figure 2B), confirming the defect in actin cytoskeleton of siAPP HUVEC cells previously described ( Figure 1E).
The differential abundance of the 1158 common proteins was illustrated by volcano plot. A two-way Student-t test (p value < 0.05) was employed using Perseus software (v 1.6.1.1) to define the proteins that were differentially regulated in siAPP and siCtrl samples. The following criteria were applied: siCtrl-siAPP ≥ ±1.5, permutation-based FDR value set at 0.05. Proteins with a p-value less than 0.05 were considered statistically significant. We quantified 68 up-regulated and 105 down-regulated proteins in the siAPP HUVEC compared to the control group (siCtrl) ( Figure 2C, Table S1). We identified a significant differential expression of proteins mainly involved in actin cytoskeleton organization (GO:0030036), in cell adhesion (GO:0007155) and in cell-cell junction organization (GO:0045216). In particular, we observed a down-regulation of actin-binding proteins involved in cytoskeleton regulation and dynamics such as radixin (RADI), α-actinin1 (ACTN1) and DPLI-1 and of proteins involved in promoting endothelial cell adhesion and migration such as integrins (ITGA2; ITGB1), and the integrin-activator kindlin3 (FERMT3) (Figure 2C,D).
We finally assessed differential expression of proteins involved in angiogenesis and vascular development and observed that endothelial markers such as VEGFR-2 (KDR) and VE-cadherin (CADH5) were not affected by the APP knockdown ( Figure 2E). However, therefore was an up-regulation of endothelial activation markers such as ICAM-1, s-100 (S100), stabilin-1 (STAB1) and Fibronectin (FINC), indicating a role of APP in vascular inflammation ( Figure 2E).
Collectively, these data confirm the presence of proliferation, migration and cytoskeleton organization defects observed in siAPP HUVEC and identify a significative change in ECs anchoring to the extracellular matrix, as well as in the interaction between adjacent ECs. This suggests the involvement of APP in vascular stability, integrity, and responsiveness to exogenous stimuli.
Collectively, these data confirm the presence of proliferation, migration and cytoskeleton organization defects observed in siAPP HUVEC and identify a significative change in ECs anchoring to the extracellular matrix, as well as in the interaction between adjacent ECs. This suggests the involvement of APP in vascular stability, integrity, and responsiveness to exogenous stimuli.  between siCtrl common dataset and siAPP common dataset using FunRich software reveals differences in cell proliferation, cell adhesion and cytoskeleton organization (p-value < 0.001). A fold change is calculated to show a functional comparison and reveals a reduction in the percentage of proteins involved in different biological processes; (C) The Volcano plot was constructed to show the significantly differentially expressed proteins between two datasets. Proteins with statistically significant differential expression (siCtrl-siAPP ≥ ±1.5, FDR < 0.05) are located in the top right and left quadrants. The orange points represent significantly upregulated proteins in siAPP and the blue points represent significantly downregulated proteins in siAPP; (D) The heat maps show proteins with statistically significant differential expression; (E) The heat map shows the differential expression of proteins present in siAPP and siCtrl common database and specifically involved in vascular development, angiogenesis and ECs proliferation. Only proteins with red asterisk show statistical significance.

APP Silencing Affects Endothelial Cells Adhesion and Promotes a Reorganization of Focal Adhesions Complex
To confirm our proteomics results, we further investigated the role of APP on HUVEC cell-matrix adhesion and focal adhesions organization. We first assessed siAPP cell-matrix adhesion by investigating cell attachment on two different integrin's substrates, collagen I and fibronectin. HUVECs use predominantly α1β1 and α2β1 to adhere to collagen I and α5β1, α4β1 and αVβ3 to adhere to fibronectin [33]. Loss of APP prevented HUVEC cell attachment on both substrates ( Figure 3A,B).  Moreover, loss of APP led to a reduction of protein expression of integrin β1, integrin β3 and the integrin activator kindlin-3 (FERMT3) in siAPP cells, validating our proteomic results ( Figure 3C).
Integrin-dependent cell adhesion is predominantly mediated by focal adhesion proteins that link the integrins domains to the actin cytoskeleton to form the adhesion complex [34]. Thus, the down-regulation of integrin β expression might result in an altered organization of other focal adhesion components. To test this, we analyzed the expression levels of paxillin and vinculin, important components of focal adhesions [35]. Western blot analysis showed reduced protein expression levels of paxillin, but not vinculin in siAPP HUVEC ( Figure 3D and Figure S4). However, while siCtrl cells showed a clear localization of vinculin to the cell membrane, siAPP cells presented a less organized expression ( Figure 3E, arrows).
Taken together these results demonstrate that loss of APP interferes with the integrin β-mediated cell adhesion by modulating focal adhesion proteins expression and organization.

APP Controls Endothelial Barrier Function
Integrin-mediated cell-matrix adhesion, as well as endothelial cell junctions-mediated cell-cell adhesion, is essential for maintenance of endothelial monolayer barrier function [36][37][38].
Our proteomic data showed a reduced abundance of proteins involved in cell-cell junction organization, indicating a possible defect in permeability in HUVEC cells lacking APP. To test this, we performed an in vitro permeability assay using confluent monolayers of siCtrl and siAPP HUVECs. We observed a significant increase of cellular permeability in HUVECs lacking APP under basal conditions (CTRL). To stress the system, we treated cells with interleukin-1 beta (IL-1 β), a pleotropic cytokine known to induce endothelial permeability [39]. IL-1β (10 ng/mL, 6 h) induced a significant increase in paracellular flux of FITCH-dextran in both siCtrl and siAPP cells; however, siAPP HUVEC cells showed an almost 4 fold-change increase when compared to siCtrl ( Figure 4A). These results suggest a higher susceptibility to pro-inflammatory stimuli of ECs in the absence of APP.
Endothelial cell junctions, such as adherens junctions and tight junctions, have a critical role in maintaining vascular integrity. Modifications in the expression and organization of these cellular components increase endothelial permeability and vascular fragility in vivo [40,41]. We, therefore, assessed the expression and organization of the tight junction proteins ZO-1 and claudin5 and of the adherens junction proteins VE-cadherin and β-catenin in siAPP confluent HUVEC.
Immunostaining for ZO-1 and claudin5 revealed a reduction in tight junctions and ZO-1 localization at cell-cell contacts of siAPP HUVEC ( Figure 4B, asterisk and white arrows). We also observed a reduction in the intensity of β-catenin staining, but no obvious defect in VE-cadherin expression and distribution ( Figure 4B, white arrows). The reduction in the total amount of ZO-1, claudin5 and β-catenin protein levels was further confirmed by western blot analysis ( Figure 4C and Figure S4). mRNA expression levels were not affected by loss of APP, suggesting a defect in proteins stability ( Figure 4D). Although VE-cadherin expression was unaffected by the silencing, the reduction of β-catenin protein levels indicated a defect in VE-cadherin activity. Altogether these data suggest that APP is involved in modulating the stability and localization of endothelial cell junctions.

APP Modulates Endothelial Response to Pro-Angiogenic Stimuli
Integrins and monolayer integrity are essential for vascular tissue homeostasis and endothelial responsiveness to angiogenic factors. We hypothesized that loss of APP, and the consequent deregulation of integrin expression would reduce endothelial cell ability to respond to extracellular stimuli including growth factors, thus negatively affecting angiogenesis.
To investigate the effect of APP silencing on angiogenesis we stimulated siCtrl and siAPP HUVECs with VEGF and measured VEGF-mediated cell migration and cell proliferation. The wound healing scratch assay showed that APP silencing significantly inhibited cell migration in response to VEGF at 8 h after scratch. Interestingly, silenced cells migrated normally in presence of 10% FBS, indicating that the defect observed was due to a dysregulation of VEGF-mediated signaling ( Figure 5A). Moreover, silencing of APP also reduced the cell proliferation in response to VEGF treatment ( Figure 5B).  (E) Quantification of the immunoblots in c), The bar graphs represent the relative level of p-VEGFR2 Y1175, Y951 or Y1059 over total VEGFR2 in siCtrl and siAPP samples as determined by band density analysis (mean ± SEM, n = 4 replicates, ** p < 0.01, *** p < 0.001).
We then investigated the ability of siAPP HUVEC seeded on Matrigel layer to form capillary-like tube structures in response to VEGF (50 ng/mL, 8 h). VEGF significantly promoted tube formation in siCtrl sample, while the formation of web/net-like structures was significantly inhibited in siAPP cells ( Figure 5C).
To further support the hypothesis that the loss of APP affects the VEGF-mediated signaling, we analyzed activation of the VEGF_receptor-2 (VEGFR2) in response to VEGF. We observed that APP silencing significantly suppressed VEGFR2 phosphorylation following VEGF stimulation at its major phosphorylation sites, located, respectively, in the kinase insert domain (Y951), in catalytic domain (Y1054/1059), as well as in the carboxy-terminal domain (Y1175). Interestingly, we did not observe a decrease of VEGFR2 total protein and mRNA expression ( Figure S3A,B). Notably, VEGFR2 was unable to immunoprecipitate APP, suggesting that VEGFR2 and APP do not interact or form a complex ( Figure S3C). Taken together, these results suggest that APP controls ECs response to VEGF by indirectly modulating VEGFR2 activation.

APP Modulates Src/FAK Signaling
We next investigated the activation of VEGF-VEGFR2 downstream signaling in APP-silenced cells. HUVEC cells were transfected with siCtrl or siAPP for 48 h and then stimulated with VEGF (50 ng/mL). VEGFR2 phosphorylation at Y1175 site promotes ECs proliferation and migration through ERK pathway activation. Consistently with VEGFR2 Y1175 reduced activation, we found a significant reduction in p-ERK in siAPP HUVEC treated with VEGF, while total ERK protein and mRNA levels were unchanged ( Figure 6A,B and Figure S3B).
VEGFR2 phosphorylation by VEGF also activates the Src/FAK pathway to promote ECs migration and adhesion. However, Src can be also activated by integrins binding to extracellular matrix, making the Src/FAK pathway a meeting point between growth factors-mediated and integrins-mediated ECs proliferation, migration and survival.
We found that phosphorylation of Src at Y416 site, as well as phosphorylation of FAK at one of the Src-target sites (Y396) were significantly reduced in APP-silenced HUVEC when exposed to VEGF ( Figure 6C,E). Total Src protein levels, but not mRNA levels were reduced by APP knockdown ( Figure 6D and Figure S3B). FAK total protein expression was unchanged, while mRNA levels were slightly up regulated by APP silencing, probably due to a compensatory mechanism ( Figure 6F and Figure S3B).
Notably, phosphorylation of ERK, Src and FAK was significantly reduced also at basal conditions in siAPP HUVEC, suggesting that APP regulates focal adhesion components phosphorylation independently of VEGFR2 activation ( Figure 6G).
Taken together these data strongly indicate that APP controls Src/FAK pathway.

Discussion
The amyloid precursor protein (APP) is a membrane bound protein present in multiple cell types, including ECs. However, its functional role in the vasculature is still unknown. Previous studies suggest that APP may exert a protective role by regulating eNOS expression in cerebral

Discussion
The amyloid precursor protein (APP) is a membrane bound protein present in multiple cell types, including ECs. However, its functional role in the vasculature is still unknown. Previous studies suggest that APP may exert a protective role by regulating eNOS expression in cerebral vasculature. Yet, the phenotype of ECs lacking this protein is poorly known [42]. In the present study, we show that APP has an important role in endothelial cell function by modulating expression of cytoskeleton-interacting proteins and mediating endothelial cell responses to extracellular stimulation. Indeed, the loss of APP in ECs resulted in a defective actin cytoskeleton organization with a consequent reduction of cell adhesion, migration, proliferation and barrier function, and in inhibition of response to the pro-angiogenic growth factor VEGF-A.
Proteomic results showed that cellular compartments most strongly affected by the loss of APP are the plasma membrane and cytoskeleton. In particular, proteins involved in cell adhesion (FERMT3; ITGA2; ITGB1) and actin cytoskeleton organization (RADI; DPLI-1; ACTN1) were downregulated, suggesting that APP acts in ECs to maintain cellular functionality. Furthermore, protein markers of endothelial activation and dysfunction were upregulated (ICAM-1, S100, STAB1 and FINC), indicating that APP is instrumental for physiological endothelial-dependent vascular homeostasis.
We further demonstrated that the loss of APP resulted in a reduced expression of integrin-β1 and integrin-β3. Moreover, both proteomic and molecular results showed a down-regulation of the integrin-activator kindlin3 (FERMT3), indicating that APP not only controls expression and/or the stability of integrins in ECs expression, but also their activation.
Integrins activity is essential for ECs homeostasis [33,43,44]. As part of the focal adhesion complex, integrins provide not only a mechanical linkage between the intracellular actin cytoskeleton and extracellular matrix, but also mediate bidirectional signaling "outside-in" and "inside-out" through the cell membrane. Kindlins participate in linking integrins to the actin cytoskeleton, mediating the "inside-out" signal from the cytoplasm to the extracellular integrin domain [45,46]. Interestingly, kindlins have been shown to modulate APP metabolism and a genome-wide study on Alzheimer's related genes identified kindlins as a genetic risk factors [47].
Dysfunction of integrin-mediated signaling was confirmed by reduction of paxillin expression and loss of Src and FAK activation in siAPP cells. Paxillin and Src/FAK mediate the "outside-in" signal and their activation is critical for the stabilization of cell adhesion and the promotion of cell migration, proliferation and survival. In response to ligand-integrin binding, FAK auto-phosphorylates (Y397) and promotes Src activation that, in turn, phosphorylates FAK (Y576/577), promoting its kinase activity and its interaction with paxillin. Paxillin regulates gene expression through MAPK cascade and it is responsible to changes in shape and reorganization of the actin cytoskeleton by binding to downstream kinases and adaptors including the actin-binding protein vinculin [48]. Accordingly, in our study the reduction of paxillin expression coincides with an altered distribution of vinculin in siAPP cells.
Deregulation of the integrin-mediated signaling can also be responsible for the loss of endothelial barrier function. A recent study showed that the loss of integrin β1 and the consequent disruption of integrin β1-matrix interaction increase cerebral microvascular endothelial cell monolayer permeability in vivo and in vitro through reorganization of tight junction proteins via altered F-actin conformation [43]. Accordingly, in siAPP HUVECs we observed a decreased ZO-1 and claudin5 expression, associated with a reduction of integrin β1 expression and a consequent increase of endothelial permeability. Furthermore, in agreement with the notion that ZO-1 works as a major cytoskeleton organizer in ECs by controlling F-actin fiber distributions and regulating the tensile force acting on VE-cadherin [41], we also observed a decreased β catenin expression and altered VE-cadherin localization Finally, we showed that the reduced expression of the focal adhesion complex and its altered interaction with actin-cytoskeleton, affects the ability of ECs to respond to the VEGF growth factor. We observed a reduced activation of VEGR2 receptor in response to exogenous VEGF in siAPP HUVEC and consequent inhibition of VEGF-induced endothelial cell migration and proliferation and loss of ECs angiogenic potential. We speculate that APP modulation of VEGF/VEGFR2 signaling is mediated by APP-integrin interaction since APP can interact with integrin-β1 [13,49], however further investigations are needed to elucidate this mechanism.
Here, we observed that siAPP HUVEC express lower levels of integrin-β3 and this coincides with lower levels of activation of the receptor. It is well known that integrin-β3 promotes VEGF/VEGFR2 signaling and that integrin KO models result in inhibition of angiogenesis and endothelial dysfunction [44,50,51]. However, the reduced response to VEGF might also be due to the alteration of shape and disorganization of the actin cytoskeleton in absence of APP. Indeed, the correct phosphorylation and stability at the cell membrane of the VEGF2 receptor is mediated by the correct interaction with cytoskeleton and cell surface proteins [52][53][54].
The downstream signaling of integrins and growth factors converge on the Src/FAK pathway. We showed that Src/FAK signaling was suppressed upon VEGF treatment in absence of APP. However, we also observed an inhibition of p-Src, and p-FAK in siAPP cells in basal condition, suggesting that the effect of APP on Src/FAK signaling is VEGF-independent.
Taken together, our results indicate that APP is necessary for the focal adhesion complex expression/stability. The interaction with cytoskeleton-interacting proteins such integrins, is essential to maintain ECs proliferation, adhesion and angiogenesis. Loss of APP expression results in altered cellular response to environmental stimuli and leads to endothelial dysfunction (Figure 7).

Conclusions
While the pathological outcome of APP dysregulation is mainly associated to its overexpression, recent studies showed that APP expression levels tend to decrease with age [59,60], and APP-knockout mice show age-dependent cognitive deficit and impaired locomotor activity [61,62]. Moreover, senescent brains showed an increase of amyloidogenic processing mediated by BACE1 [63], indicating that amyloid-β peptides accumulation is not necessarily associated with APP overexpression. Furthermore, clinical trials targeting amyloid-β have failed to reverse the cognitive

Conclusions
While the pathological outcome of APP dysregulation is mainly associated to its overexpression, recent studies showed that APP expression levels tend to decrease with age [59,60], and APP-knockout mice show age-dependent cognitive deficit and impaired locomotor activity [61,62]. Moreover, senescent brains showed an increase of amyloidogenic processing mediated by BACE1 [63], indicating that amyloid-β peptides accumulation is not necessarily associated with APP overexpression. Furthermore, clinical trials targeting amyloid-β have failed to reverse the cognitive loss, and in some cases resulted in vascular severe adverse events (i.e., bapineuzumab), indicating the importance of physiological levels of amyloid-β peptides and highlighting the urgency to better understand the function of APP and its cleavage products in cell homeostasis.
In this study, we show that a correct expression level of APP may be necessary for the correct functionality of ECs. Although the use of HUVEC as model is not representative of all endothelial cell types found in an organism, they are an excellent model for the study of vascular endothelium properties and the main biological pathways involved in endothelium function. More studies addressing the APP function specifically in cerebrovascular endothelium will be needed.
A deep understanding of APP function in maintaining vascular homeostasis might shed light on new therapeutic targets and provide a new perspective on treatment options of neurodegenerative diseases.