Alzheimer’s Aβ assembly binds sodium pump and blocks endothelial NOS activity via ROS-PKC pathway

Amyloid β-protein (Aβ) may contribute to worsening of Alzheimer’s disease (AD) through vascular dysfunction, but the actual molecular mechanisms remain controversial. Using ex-vivo blood vessels and primary endothelial cells derived from human brain microvessels, we revealed that patient-derived Aβ assemblies, termed amylospheroids (ASPD), exist on the microvascular surface in patient brains and inhibit vasorelaxation through binding to the α3 subunit of sodium, potassium-ATPase (NAKα3) on endothelial cells. Interestingly, NAKα3 also serves as the toxic target of ASPD in neurons. ASPD elicit neurodegeneration through calcium overload, while ASPD suppress vasorelaxation by inhibiting nitric oxide (NO) production. ASPD-NAKα3 interaction on cerebrovascular endothelial cells disturbs the NO release by inactivating endothelial NO synthase through mitochondrial reactive oxygen species and protein kinase C. The findings suggest that ASPD may dually contribute to neuronal and vascular pathologies through binding to NAKα3. Thus, blocking the ASPD-NAKα3 interaction may be a useful target for AD therapy.

Recently, the symptomatic overlap between AD and vascular dementia has been focused, and the vascular biomarkers are expected to improve the clinical diagnosis of AD (Jack et al., 2018). Notably, earlier works from the Nun studies have suggested that symptomatic progression of AD related to Aβ deposition, but not to NFTs, appeared to be significantly modified by the presence of cerebrovascular abnormalities in AD (Snowdon et al., 1997). Therefore, a better understanding of the molecular mechanisms underlying Aβ-related cerebrovascular dysfunction in AD should help us to understand how vascular dysfunction contributes to AD progression, and will open up new therapy.
In studies of the mechanisms of neurodegeneration in AD brains, we purified highly neurotoxic ~30-mer assemblies of Aβ (later termed "amylospheroids" (ASPD)) from human AD brains (Hoshi et al., 2003;Noguchi et al., 2009). We proved that ASPD bind directly to the neuronal isoform of α3 subunit of the sodium pump (sodium, potassium-ATPase α3 (NAKα3)) and cause the death of mature neurons by impairing species (ROS) in mitochondria and activates protein kinase C (PKC). This increases the PKC-phosphorylated inactive form of endothelial nitric oxide (eNOS), and decreases nitric oxide (NO) production. This in turn would suppress the relaxation of blood microvessels, and might cause a reduction of cerebral blood flow and other vascular dysfunctions in AD brains. Thus, we show a new possibility that brain Aβ assemblies accelerate worsening AD pathologies by affecting the cerebrovascular systems via interaction with the sodium pump.

Results
ASPD are present in cerebrovascular vessels of AD brain. We have established ASPD-tertiary-structure dependent antibodies that are highly specific and do not detect other Aβ oligomers recognized by a pan-Aβ oligomer A11 antibody (see details in Table S1 in (Noguchi et al., 2009)). Using one of ASPD-specific antibodies, rpASD1 (Noguchi et al., 2009), we first examined whether ASPD accumulate in AD cerebrovascular vessels. As we have reported (Noguchi et al., 2009), ASPD were widely distributed around senile plaques and neurons in the frontal cortex of AD patients (Fig. 1A). In addition to such brain parenchymal staining, we also detected ASPD in most microvessels (see "*" in Fig. 1A left). A high-power view showed that ASPD staining is also present at the inner endothelial surface of the microvessels (marked by arrows in Fig. 1A right), where it is colocalize with Aβ1-42 and Aβ1-40 stainings (compare arrows in Fig. 1A right with that of Fig. 1B or 1C right). The staining result is consistent with our previous Mass analyses showing that ASPD purified from AD patients' brains are composed of both Aβ1-40 and Aβ1-42 (Noguchi et al., 2009). These results suggest that ASPD are present not only in brain parenchyma but also in the endothelial cells of the microvessels in AD brains. Accordingly, in this work, we focused on the effect of ASPD on brain endothelial cells.

ASPD inhibit relaxation of blood vessels through inhibition of endothelial NAKα3.
Endothelial cells release NO, leading to relaxation of blood vessel smooth muscles (Arnold, Mittal, Katsuki, & Murad, 1977;Furchgott & Zawadzki, 1980;Ignarro, Buga, Wood, Byrns, & Chaudhuri, 1987). Therefore, we examined whether ASPD affect the relaxation response of blood vessels. To this end, we used aortic rings isolated from rats, because the aortic rings are considered to be a most sensitive ex-vivo vascular model available to detect the change in NO-dependent relaxation response among blood vessels including peripheral arteries and cerebral microvessels (Shimokawa & Godo, 2016). According to the established method (Angus & Wright, 2000;Sasahara, Yayama, Matsuzaki, Tsutsui, & Okamoto, 2013), the isolated aortic rings were contracted by treatment with phenylephrine (an adrenergic α1 receptor agonist), and NO-dependent relaxation was induced by carbachol (a muscarinic M3 receptor agonist).
On mature neurons, ASPD affect in a dose-dependent manner, and the effect reaches a plateau at ~40 nM (Ohnishi et al., 2015) (note that ASPD concentrations are shown by using the average mass of ASPD, 128 kDa). We therefore used this plateau concentation of ASPD for the present experiments. Treatment of the isolated aortic rings with 46 nM ASPD for 1 hr inhibited the carbachol-induced relaxation response ( Fig. 2A upper panel) and doubled the ED50 values of carbachol required for relaxation ( Fig. 2A bottom   table). We also present the ED10 values, because the change of blood vessels in the actual brain takes place in a narrow range, as it is directly linked to the blood pressure.
As shown in figure 2A, ASPD also doubled the ED10. When ASPD were incubated with ASPD-specific mouse monoclonal mASD3 antibody, which blocks ASPD binding to neurons (Noguchi et al., 2009;Ohnishi et al., 2015), the increase in ED50 and ED10 was completely abolished (compare black and red circles in Fig. 2A). These results show that ASPD directly suppress the NO-dependent relaxation of the blood vessels, probably through affecting endothelial cells. To eliminate the possibility that ASPD directly affect the smooth muscles of the blood vessels, we confirmed that ASPD did not affect the relaxation response induced by papaverine, which directly relaxes blood vessel smooth muscles in an endothelium-independent manner (Lugnier, Bertrand, & Stoclet, 1972;Martin, Furchgott, Villani, & Jothianandan, 1986). Indeed, 46 nM ASPD did not affect either the papaverine-induced relaxation response of the aortic rings (% maximal relaxation induced by papaverine: 100.9 ± 0.5 and 101.4 ± 0.7% with and without ASPD, respectively; n = 5, P = 0.46) or the time to reach the maximal relaxation (3.4 ± 0.1 and 3.6 ± 0.3 min with and without ASPD, respectively; n = 5, P = 0.69). These results collectively support that ASPD act on endothelial cells.
Next, we elucidated to identify the target protein on endothelial cells to which ASPD bind to inhibit NO release. We speculated that NAKα3 might also serve as an ASPD toxic target in the endothelial cells as it does in mature neurons (Ohnishi et al., 2015). Because NAKα3 is a neuron-specific isoform (Shrivastava, Triller, & Melki, 2018), we first examined whether NAKα3 is present on the endothelial cell surface by immunostaining. We detected patchy NAKα3 staining (green signals indicated by arrows in Fig. 2B) on the vascular lumen surface of the endothelial cells of the isolated aortic rings (red shows a signal of von Willebrand factor (vWF) glycoprotein, an endothelial cytoplasmic marker (Rakocevic et al., 2017)). To confirm the functional involvement of NAKα3 in the suppression of the blood relaxation response, we examined the effect of 100 nM ouabain, a concentration that is enough to inhibit the rodent NAKα3 isoform, but not other rodent isoforms (Noel, Fagoo, & Godfraind, 1990). As shown in figure 2C, this concentration of ouabain sufficiently inhibited the relaxation response and increased both the ED50 and the ED10 of carbachol required for relaxation of the blood vessels (Fig. 2C), as observed in the ASPD treatment ( Fig. 2A).
These results collectively support the idea that ASPD suppress blood vessel relaxation by affecting endothelial cell function through inhibition of NAKα3 pump activity. To further dissect the molecular action of ASPD, we decided to use primary cultures of endothelial cells obtained from human brain microvessels. To further confirm ASPD-NAKα3 interaction on the brain endothelial cells, we examined whether knockdown of NAKα3 expression by small interfering RNA (siRNA) blocks the interaction of ASPD and NAKα3. Western blotting and immunostaining consistently showed that the transfection of ATP1A3 siRNA decreased the amount of NAKα3 to 38 ± 11% (n = 3, P = 0.01 compared with siRNA-nontreated group; Fig. 3E left) and to 45 ± 5% (n = 5, P = 0.02 compared with siRNA-nontreated group; Fig. 3E right), respectively. In correlation with the decrease of NAKa3 level, the ASPD binding to total NAKa3 was decreased to 31 ± 11% on average (n = 5, Fig. 3F and the representative 2D staining image of 100 nM ASPD-treated cells in Fig. S1).
Mock siRNA transfection did not affect either NAKα3 expression (Fig. 3E) or ASPD binding to total NAKα3 (Fig. 3F). The results collectively support that ASPD interact with NAKα3 on the brain endothelial surface.
To elucidate the effect of ASPD on NO production using these human brain endothelial cells, we next determined the ED50 of carbachol required for NO release using diaminofluorescein-FM (DAF-FM; Sekisui Medical, Tokyo, Japan), a fluorescent probe for NO quantification (Kojima et al., 1999). As shown in figure S2, a 5-min treatment with carbachol increased NO release dose-dependently, which reached a plateau at around 100 µM. For further experiments, we treated human primary brain microvessel endothelial cells with carbachol at 1 µM, which was approximately the ED50 required for NO release (Fig. S2).
As shown in figure 4A, ASPD antagonized carbachol-induced NO release from the human brain microvessel endothelial cells in a dose-and time-dependent manner ( Fig. 4A) , i.e., the NO release was decreased more rapidly and more strongly in correlation with the increase in ASPD binding ratio to the endothelial NAKa3 (compare

The mitochondrial ROS/PKC pathway is involved in eNOS-Thr 495
phosphorylation by ASPD in human cerebral endothelial cells. We next clarified how the ASPD-NAKα3 interaction increases eNOS-Thr 495 phosphorylation independently of the usual relaxation mechanisms. Previous studies have shown that eNOS-Thr 495 phosphorylation is mainly regulated by three kinases, protein kinase C (PKC), Rho kinase (ROCK), and AMP-activated protein kinase (AMPK) (Fleming & Busse, 2003;Heiss & Dirsch, 2014). Among the tested inhibitors specific for each kinase, bisindolylmaleimide I (a selective PKC inhibitor) clearly inhibited the ASPDinduced increase in eNOS-P-Thr 495 , but Y-27632 (a ROCK inhibitor) and compound C (an AMPK inhibitor) did not (Fig. 6A). To further confirm the involvement of PKC, we used another inhibitor that works differently: while bisindolylmaleimide I competes at ATP binding site of PKC, calphostin C inhibits the interaction between diacylglycerol and the PKC-regulatory domain (Iida, Kobayashi, Yoshida, & Sano, 1989;Kobayashi, Nakano, Morimoto, & Tamaoki, 1989;Toullec et al., 1991). As shown in figure 6B, calphostin C also completely inhibited the ASPD-induced increase in eNOS-P-Thr 495 .
Taken together, these results show that PKC is a major regulator for the ASPD-induced increase in eNOS-Thr 495 phosphorylation.
The next question is how the ASPD-NAKα3 interaction leads to PKC activation in the endothelial cells. We elucidated three possible activation mechanisms by using specific inhibitors of each mechanism: tempol (a scavenger of ROS), BAPTA-AM (a chelator of intracellular calcium), or U-73122 (an inhibitor of phospholipase C (PLC)). Because all the tested activation mechanisms are known to be associated with auto-phosphorylation at Ser 660 in PKC (Cosentino-Gomes, Rocco-Machado, & Meyer-Fernandes, 2012;Feng & Hannun, 1998), the PKC activation was monitored by the ratio of PKC-Ser 660 phosphorylation (PKC-P-Ser 660 ). As shown in figure 7A, only tempol abolished the increase in PKC-P-Ser 660 associated with PKC activation. Tempol also blocked the increase in eNOS-P-Thr 495 induced by the ASPD-NAKα3 interaction (Fig. 7B). These results consistently show that the ASPD-NAKα3 interaction induces PKC activation through ROS production.
The final question is where this ROS production occurs (Santilli, D'Ardes, & Davi, 2015), which we examined by using an ROS indicator, CellROX (Thermo Fisher Scientific, Waltham, MA). As shown in figure 8, CellROX detected an increase in ROS in the cytoplasm of endothelial cells treated with 35 nM ASPD for 6 hr. This increase was wiped out by pretreating the cells with inhibitors of mitochondrial ROS generation, YCG-063 or mito-tempol, but was not affected by pretreatment with NADPH oxidase inhibitors, VAS2870 or apocynin (Fig. 8). Although vascular ROS is also produced by xanthine oxidase (Santilli et al., 2015), the results in figure 8 supported that mitochondria are the major source of the ROS production induced by the ASPD-NAKα3 interaction. The scheme in figure 9 summarizes and illustrates our new findings in this study.

Discussion
A decrease in bioavailable NO leads to cerebrovascular dysfunctions such as reduced cerebral blood flow (Binnewijzend et al., 2016;Boo & Jo, 2003;O'Brien et al., 1992). We found that the aberrant ASPD-NAKα3 interaction in cerebrovascular endothelial cells disturbs NO release by inactivating eNOS through a new system mediated by mitochondrial ROS production, independently of the physiological relaxation system (Fig. 9). The current study opens up a new possibility that ASPD may contribute not only to neurodegeneration in the brain (Komura et al., 2019;Noguchi et al., 2009;Ohnishi et al., 2015), but also to cerebrovascular changes. This is in line with recent observations that vascular changes in the brain contribute to worsening of AD (Govindpani et al., 2019). Such a decrease of cerebral blood flow is responsible for alternating the physiological neurochemical environment, promoting the development of AD-related neuropathology, such as dysfunction and loss of neurons (Zlokovic, 2005). It is noteworthy that ASPD share the same toxic target, NAKα3, in neurons and in cerebral blood microvessels. Therefore, the ASPD-NAKα3 interaction may be a useful target for AD therapy.
The presence of NAKα3 on endothelial cells has not been well investigated, but our immunostaining data revealed the presence of NAKα3 as clusters approximately 230 nm in diameter on the surface of brain microvessel-derived endothelial cells ( Fig.   2B and 3A). Moreover, ASPD signals were essentially located proximal to the cellsurface NAKα3 clusters on the endothelial cells (Fig. 3D). This result suggests that NAKα3 may exist in a specific subdomain of the plasma membrane of the brain microvessel-derived endothelial cells, such as membrane microdomains (Leo, Hutzler, Ruddiman, Isakson, & Cortese-Krott, 2020), to which ASPD bind. Western blotting and mRNA analyses also supported the expression of NAKα3 in the cerebrovascular microvessels (Fig. 3B). Notably, ASPD-toxicity-neutralizing antibody and siRNA against NAKα3 clearly abolished the effect of ASPD on the cerebrovascular endothelial cells ( Fig. 2A, 3F, 4B, and 5C). Whether NAKα3 is actually present in membrane microdomains or not, and if so, what is its functional role in the physiology of endothelial cells need to await future studies. However, our results collectively demonstrated that ASPD-NAKα3 interaction is responsible for inhibition of eNOS in cerebrovascular endothelial cells.
The treatment of rat endothelial cells with a low concentration of ouabain was reported to increase intracellular calcium (Dong, Komiyama, Nishimura, Masuda, & Takahashi, 2004;Noel et al., 1990). We therefore expected that ASPD-induced PKC activation would occur in an intracellular calcium-dependent manner. However, the intracellular calcium chelator, BAPTA-AM, did not affect PKC activation (Fig. 7A), suggesting that the ASPD-NAKα3 interaction activates PKC through a different pathway. Eventually, we proved that ROS mediates PKC activation by ASPD (Fig. 7A), which was completely abolished by a selective scavenger of mitochondrial ROS production (Fig. 8). The molecular link between NAKα3 inactivation and mitochondrial ROS production remain to be clarified. Interestingly, mitochondrial ROS production was reported to be suppressed by activation of the NAKα3 pump by an NAK-DR region-specific antibody (Yan et al., 2016). These findings together support that NAK activity negatively regulates mitochondrial ROS production.
Suppression of the relaxation response of the blood vessels after Aβ treatment was first reported about two decades ago using rat-derived aortic rings, raising the possibility that Aβ may directly or indirectly reduce NO release Thomas, Thomas, McLendon, Sutton, & Mullan, 1996) . Since then, four papers have shown that the eNOS activity is actually decreased after the treatment of blood endothelial cells derived from umbilical vein, aorta, or basilar artery with high concentrations of Aβ (more than 5 µM) (Chisari, Merlo, Sortino, & Salomone, 2010;Gentile et al., 2004;Lamoke et al., 2015;Suhara et al., 2003). These previous studies suggested a possible link between Aβ and eNOS activity regulation, but it has been unknow whether Aβ directly causes the decrease of eNOS activity of the brain microvessels. Moreover, if Aβ does act directly on the eNOS activity of the brain microvessels, there still remained several questions that need to be clarified, such as what molecular entity of Aβ (Aβ monomers or a certain form of Aβ assemblies) actually works, through which target on the blood vessel Aβ acts, and what molecular mechanism lead to the decrease of eNOS activity. Here, we address these questions using human primary endothelial cells derived from the brain microvessels. We previously demonstrated that ~30-mer Aβ assemblies present in AD brains, ASPD (Hoshi et al., 2003;Noguchi et al., 2009), bind to NAKα3 on the endothelial surface, leading to inhibition of eNOS activity by increasing an inactivated state of eNOS phosphorylated at Thr 495 (Fig. 3, 4, and 5). Thus, by showing that ASPD-specific antibody completely blocked reduction of the eNOS activity observed after ASPD treatment (Fig. 4B), we showed that ASPD directly decreases the eNOS activity by binding to NAKα3. Importantly, while the previous four studies mentioned above observed a decrease in the eNOS-Ser 1177 phosphorylation after Aβ treatment, which is a part of the physiological pathway for regulating eNOS activity, such a decrease in eNOS-Ser 1177 phosphorylation was not detected in the case of ASPD (Fig. 5B). Instead, as described above, we found that ASPD activate PKC, which increases eNOS-Thr 495 phosphorylation, through mitochondrial ROS production (Fig. 6, 7 and 8). PKC has been reported to decrease eNOS-Ser 1177 phosphorylation by activating Ser/Thr protein phosphatase 2A (PP2A) (Michell et al., 2001). Nevertheless, ASPD appeared not to significantly affect the level of eNOS-Ser 1177 phosphorylation (Fig. 5B). Thus, ASPD is likely to regulate the contraction of blood vessels in a way different from the physiological pathway for NO release which was previously reported.
To further clarify the molecular link between cerebrovascular dysfunction and parenchymal neuronal damage in the onset of AD induced by Aβ assemblies, one of the essential questions sequestered in future is to understand the source of the Aβ assemblies in the blood microvessels. In the case of ASPD, we have already shown that ASPD are selectively formed in excitatory neurons and secreted (Komura et al., 2019). Therefore, it is natural to consider that ASPD are delivered to the cerebral blood vessels through apolipoprotein E (ApoE), clusterin, or brain meningeal lymphatics, as reported previously (Beeg et al., 2016;Da Mesquita et al., 2018;Garai, Verghese, Baban, Holtzman, & Frieden, 2014;Nelson, Sagare, & Zlokovic, 2017). However, other possibilities, e.g. formation of ASPD in the cerebral blood microvessels, cannot be excluded, because Aβ precursor protein (APP) is also expressed on the surface of cerebral endothelial cells, even though the APP isoform profiles in endothelial cells are different from those in neurons (Kitazume et al., 2010). Even if ASPD are formed in cerebral blood microvessels, as they are in neurons, we consider that a pathological trigger leading to ASPD formation in endothelial cells should be different from that in neurons owing to the differences in APP isoforms and APP processing between neurons and endothelial cells (Grinberg, Korczyn, & Heinsen, 2012;Kakuda et al., 2017;Lasiecka & Winckler, 2011). We believe that further studies to identify the origin of cerebral vascular ASPD will not only deepen our understanding of ASPD themselves, but also help to understand the origin of other Aβ assemblies that accumulate in cerebral blood microvessels.
Interestingly, cerebral vascular ROS was reported to decrease cerebral blood flow through pericyte constriction of cerebral blood vessels (Nortley et al., 2019). Here, we found that ASPD induce mitochondrial ROS production in cerebral endothelial cells, but because released ROS could diffuse to affect nearby cells, it seems plausible that ASPD-induced ROS production in endothelial cells would also affect neighboring pericytes and block the physiological relaxation of cerebral blood vessels.
Decreased cerebral blood flow is an apparent risk factor for AD development (Zlokovic, 2005). Blocking the cerebrovascular toxicity of ASPD is thus expected to be an effective target to prevent worsening of AD. We have previously found an ASPDbinding tetrapeptide that is similar to the ASPD-binding domain on NAKα3, and surprisingly found that the treatment of ASPD with this peptide completely abolished ASPD-induced neuronal apoptosis by blocking the interaction of ASPD with NAKα3 (Ohnishi et al., 2015). Therefore, we expect that this may lead to a new AD therapeutic strategy based on dual attenuation of both the vascular and neuronal toxicities of ASPD. Synthetic ASPD preparation. ASPD were prepared from in-house-prepared Aβ1-42 as previously described (Ohnishi et al., 2015). Aβ1-42 monomer was aggregated to oligomers by slowly rotating a solution in F12 buffer without L-glutamine and phenol red for 16 hr at 4˚C. ASPD were then isolated from the fraction that passed through 0.22-µm filters, but was retained on 100-kDa MWCO filters (Sartorius, Tokyo, Japan).
Immunohistochemical staining of AD brain and rat aorta. Immunohistochemical staining of autopsied brains from AD patients was performed as previously described (Noguchi et al., 2009). Four µm sections of prefrontal cortex were prepared from the formalin-fixed brains embedded in paraffin wax. The sections for ASPD staining were pretreated with microwave radiation, and those for Aβ1-40 or Aβ1-42 staining were pretreated with formic acid for 5 min. Then, all sections were pretreated with 0.3% H2O2-methanol for 60 min followed by incubation with normal goat serum for 30 min at r.t., and further pretreated with a blocking kit (Vector Laboratories, Burlingame, CA).
These sections were incubated overnight at 4°C with primary antibodies in the presence of normal goat serum in PBS, followed by incubation for 60 min at r.t. with appropriate biotinylated secondary antibodies. Immunoreactivities were detected by the avidinbiotin-peroxidase complex method using a Vectastain ABC kit (Vector Laboratories).
Counterstaining was carried out with Mayer's hematoxylin. Images were viewed by using a light microscope AX80T (Olympus, Tokyo, Japan) and captured with a digital camera DP70 (Olympus).
Paraformaldehyde (PFA)-fixed aortas were prepared from isofluraneanesthetized Wistar rats (7-week-old, male; Japan SLC, Shizuoka, Japan) and embedded in paraffin wax. Then 4 µm sections were prepared and pretreated with 10 mM citrate buffer for 30 min at 95°C, followed by incubation with normal goat serum for 30 min at r.t. These sections were incubated overnight at 4°C with primary antibody against Germany).

ex-vivo relaxation of rat blood vessels.
The ex-vivo vascular study was performed as previously described with some modifications (Sasahara et al., 2013). Wistar rats (7-week-old, male, Japan SLC) were sacrificed by bleeding from the carotid arteries under isoflurane anesthesia, and the aorta was excised and immediately placed in Krebs-Henseleit solution (118.4 mM NaC1, 4.7 mM KC1, 2.5 mM CaCl2, 1.2 mM KH2PO4, 1.2 mM MgSO4, 25.0 mM NaHCO3, and 11.1 mM glucose, 37°C) (K3753; Sigma-Aldrich Japan, Tokyo, Japan). The aorta was cleaned of adherent tissue and cut into 3 mm aortic rings. These rings were vertically fixed under a preload of 1.0 g in a 5 mL organ bath filled with Krebs-Henseleit solution continuously aerated with 95% O2/5% CO2 gas, and allowed to equilibrate for 60 min. After the pretreatment with ASPD or ouabain for 60 min, the aortic rings were constricted by phenylephrine (1 µM Cell culture. Primary cultures of human endothelial cells derived from brain microvessels (ACBRI 376; Cell Systems, Kirkland, WA) were cultured on collagen Icoated dishes with EGM-2MV medium (CC-3202; Lonza Japan, Tokyo, Japan) at 37°C under 5% CO2. The medium of 80% confluent cultures at passage 6 ~ 8 was replaced with fresh medium, and after 24 hr these cultures were employed for the experiments below.
Immunocytochemical staining of NAKα3 and ASPD. Endothelial cells cultured on collagen I-coated cover glass were treated with ASPD and fixed with 4% (w/v) PFA for 20 min at 37 °C after washing with PBS. The fixed cells were rinsed three times with PBS, treated with 2 mg/ml glycine for 10 min at r.t., rinsed three times with PBS, permeabilized with 0.2% (v/v) Triton X-100 for 5 min at r.t., and rinsed three times with PBS. These cells were pretreated with PBS containing 3% (w/v) BSA (Sigma-Aldrich Japan) and 10% (v/v) normal goat serum (Immuno-Biological Laboratories, Gunma, Japan) for 30 min at r.t., and incubated overnight with primary antibody against ASPD (ASPD-specific mASD3 antibody, 0.1 µg/ml) (Noguchi et al., 2009;Ohnishi et al., 2015) and NAKα3 (sc-16051-R, 0.4 µg/ml, Santa Cruz Biotechnology) at 4 °C. After three washes with PBS, the cells were incubated with the appropriate Alexa Fluorconjugated secondary antibody (1:1000, Molecular Probes) with counterstaining by DAPI (1:500, Dojindo Molecular Technologies) for 60 min at r.t. The cells were rinsed three times with PBS and mounted with Prolong Gold anti-fade reagent (Invitrogen, Waltham, MA). Fluorescence images were acquired with a confocal laser-scanning microscope LSM710 (Carl Zeiss), and z-stacked images were taken at 2 µm intervals.
The quantitative analysis was performed with a confocal quantitative image cytometer CQ1 and CQ1 software (Yokogawa Electric Corp., Tokyo, Japan). Note that the anti-NAKα3 antibody selectively reacts with NAKα3, except for the signals around nucleoli (in the case of non-neuronal cells, including endothelial cells, the anti-NAKα3 antibody shows thick and aggregated signals in nuclei due to non-specific binding (Ohnishi et al., 2015)). Quantitative data is shown as densitometric ratios of phosphorylated eNOS (eNOS-P-Ser1177 or eNOS-P-Thr495) to total eNOS, or phosphorylated PKC (PKC-P-Ser 660 ) to total PKC. RT-PCR of ATP1A3 mRNA. Total RNA of cultured endothelial cells was extracted using TRIzol reagent (Thermo Fisher Scientific). RNA (0.5 µg) was reverse-transcribed using ReverTra Ace reaction mixture (TOYOBO, Osaka, Japan) with oligo (dT) primer.

Western blotting of
The reaction mixtures were incubated at 42°C for 20 min, 99°C for 5 min, then 4°C for 5 min to synthesize the first strand of cDNA. The cDNA was then mixed with KOD FX PCR reaction mixture (TOYOBO) with forward and reverse primers for ATP1A3 (forward primer 5'-CGCCGGGACCTGGATGACCTC-3' and reverse primer 5'-CGGATCACCAGGGCTTGCTGG -3' for ATP1A3; the PCR product was detected at 434 bp) (Fransen, Hendrickx, Brutsaert, & Sys, 2001) or GAPDH (forward primer 5'-CAAGGTCATCCATGACAACTTTG-3' and reverse primer 5'-GTCCACCACCCTGTTGCTGTAG-3' for GAPDH; the PCR product was detected at following conditions: initial denaturation at 98°C for 2 min; 30 cycles of denaturation at 98°C for 10 sec, annealing at 55°C for 30 sec, and extension at 68°C for 30 sec; final extension at 68°C for 7 min. PCR products were separated on 1.5% agarose gel and visualized using ethidium bromide. Knockdown of NAKα3. ATP1A3 siRNA (s1724, Thermo Fisher Scientific) was mixed with Lipofectamine 3000 reagent (Thermo Fisher Scientific) according to the manufacturer's protocol. The mixture was then added to the culture medium on 80% confluent endothelial cultures, and after 6 hr incubation, the culture medium containing siRNA was replaced with fresh EGM-2MV medium. At 72 hr after the siRNA mixture treatment, the endothelial cells were treated with ASPD. Then, the total protein was then extracted with RIPA buffer for SDS-PAGE/Western blot (see above "Western blotting of NAKα3, eNOS, and PKC"), or the endothelial cells were fixed with 4% (w/v) PFA for immunocytochemical staining (see above "Immunocytochemical staining of NAKα3 and ASPD"). (492000, Merck-Millipore), or apocynin (178385, Merck-Millipore) was added into the culture medium at 30 min before the ASPD treatment. The fluorescence data were analyzed using CQ1 software.

Measurements of NO
Statistical analyses. All data are expressed as mean ± S.E. We used Statecel2 software (OMS Publication, Tokyo, Japan) for statistical analyses. No data points were excluded from the analysis. Statistical comparisons were performed with the unpaired Welch's ttest between two groups or with one-way analysis of variance (ANOVA) followed by pair-wise comparisons using Scheffé's method. Differences were considered significant at P < 0.05.   A: Carbachol-induced NO release from human primary brain microvessel endothelial cells treated with ASPD (0.3, 3, or 32 nM for the indicated times) (n = 4). B: Preincubation of ASPD with ASPD-specific mASD3 antibody abolished the 32 nM ASPD-induced suppression of NO release (n = 4). In (A,B), data are presented as means ± S.E. *P < 0.05/**P < 0.01 (ANOVA with Scheffé's method).    Representative fluorescence images are shown on the right. Scale bars: 100 µm. Data are presented as means ± S.E. *P < 0.05/**P < 0.01 (ANOVA with Scheffé's method). High-power representative 2D images of immunocytochemical multiple staining of ASPD (ASPD-specific, mASD3 antibody), NAKα3, and nuclei (DAPI) on 30 nM ASPD-treated human primary brain microvessel endothelial cells with or without