Donepezil modulates amyloid precursor protein endocytosis and reduction by up-regulation of SNX33 expression in primary cortical neurons

Donepezil, a therapeutic drug for Alzheimer’s disease, ameliorates cognitive dysfunction through selective inhibition of acetylcholinesterase. However, recent studies have also reported off-target effects of donepezil that likely contribute to its therapeutic effects. In this study, we investigated the (i) role of donepezil in amyloid precursor protein (APP) processing and (ii) involvement of sorting nexin protein 33 (SNX33), a member of the sorting nexin protein family, in this processing. Results showed that donepezil induces an increase in SNX33 expression in primary cortical neurons. The secretion of sAPPα in culture media increased, whereas the expression of full-length APP in the cell lysate remained unchanged. Exposure of cortical cultures to donepezil led to a decrease in amyloid β (Aβ) protein levels in a concentration- and time-dependent manner. This decrease was not affected by concomitant treatment with acetylcholine receptor antagonists. SNX33 knockdown by target-specific morpholino oligos inhibited the effects of donepezil. Donepezil treatment increased cell membrane surface expression of APP in SNX33 expression-dependent manner. These results suggested that donepezil decreases the level of Aβ by increasing SNX33 expression and APP cleavage by α-secretase in cortical neurons.

by α-secretase occurs at the plasma membrane, whereas β-secretase cleavage occurs mainly in the endosome 5 . It has been believed that the balance between the two cleavage pathways likely determines the production level of the Aβ protein. Several studies on endocytic transport factors have shown that endocytic transport is a critical factor determining the balance between the two pathways [6][7][8][9] .
Many aspects of the regulatory mechanism governing APP endocytosis are still unclear. However, we are beginning to understand the factors involved in endocytic trafficking and their roles. The sorting nexin protein (SNX) family comprises endosomal trafficking proteins that share a common phox-homology (PX) domain 10,11 . SNX33, along with SNX9 and SNX18, forms a subfamily characterised by the presence of Src homology 3 (SH3), PX and Bin/amphiphysin/Rvs domains 10 . SNX33 is a recent addition to the subfamily and modifies APP processing. In addition to increasing APP cleavage by α-secretase, SNX33 reduces APP endocytosis and thereby likely decreases shedding by β-secretase 12 . SNX33 also inhibits the function of dynamin, a key component of endocytosis 13 , by directly binding to dynamin through its SH3 domain. Dynamin mutants and dominant negative forms, in turn, increase APP cleavage by α-secretase, reduce APP endocytosis and decrease shedding by β-secretase 7,8,12 . These results suggest that SNX33 modifies processing of APP in the direction of the non-amyloidogenic pathway via the inhibition of APP endocytosis 5,12 .
Donepezil is an acetylcholinesterase (AChE) inhibitor designed for AD treatment. Donepezil moderately improves cognitive function, with relatively low side effects compared with other AD drugs 14,15 . Studies have shown that in addition to being an AChE inhibitor, donepezil protects against neurotoxic insults and up-regulates nAChRs, properties likely involved in its therapeutic effects [15][16][17][18] . Connections between the cholinergic system and α-secretase-mediated APP processing is gradually emerging. Several studies have shown that AChE inhibitors modify APP processing in favour of the non-amyloidogenic pathway in various systems, from SH-SY5Y cells to aged AD model mice [19][20][21][22][23][24] , but the underlying mechanism remains unclear.
In this study, we examined the effects of donepezil on APP processing in primary cortical neurons. We also identified SNX33 as a downstream factor up-regulated by donepezil treatment and examined its role in donepezil-dependent modification of APP processing.

Results
Donepezil up-regulates SNX33 expression. In this study, to examine the mechanism involved in the neuroprotective effects observed by donepezil and galantamine treatment in primary cortical neurons, we initially screened for genes whose expression levels are up-regulated by these two drugs. Among candidate genes identified by microarray analysis of transcripts obtained from primary cortical neurons treated with donepezil or galantamine for 24 h, we focused on the up-regulation of SNX33 expression because the SNX family is believed to be involved in endomembrane transport of various transmembrane proteins, including APP 11 . In particular, SNX33 modulates APP endocytosis and APP cleavage by α-secretase, and the rate of APP endocytosis is a major control factor in APP cleavage by α-secretase 12 . qPCR examination of whether donepezil treatment under conditions previously been associated with neuroprotection up-regulates SNX33 transcription in primary cortical neurons demonstrated that donepezil-treated cells show a concentration-dependent increase in SNX33 mRNA levels compared with untreated control cells (Fig. 1a). In addition, western blotting performed to examine SNX33 expression showed that donepezil-treated cells have significantly increased SNX33 expression compared with untreated control cells (Fig. 1b,c). These results confirmed that donepezil treatment up-regulates SNX33 expression in primary cortical neurons.
SNX33 is up-regulated in primary cortical neurons. Treatment with galantamine and rivastigmine resulted in a slight increase in SNX33 expression, but statistical analysis did not support the significance of this increase (Fig. 2a,b). Treatment with tacrine, nicotine and memantine did not result in an increase in SNX33 expression. Therefore, under the treatment condition used in this study, we observed an increase in SNX33 expression with donepezil treatment. In subsequent experiments, we focused on the effects of donepezil treatment on primary cortical neurons.
Donepezil treatment modulates App processing in primary cortical neurons. Donepezil-treated cells showed a significant increase in sAPPα levels compared with untreated control cells (Fig. 3a,b). Consistent with the role of SNX33 in regulating the endocytotic transport of APP (the precursor of sAPPα) but not its expression 12 , we did not observe a significant increase in APP levels between donepezil-treated cells and untreated control cells (Fig. 3a,c). Further examination of the effects of donepezil on endocytic APP processing by observing Aβ levels in the culture media of primary culture neurons by enzyme-linked immunosorbent assay (ELISA) using Aβ40-and Aβ42-specific antibodies showed that the Aβ40 level in the culture media decreased significantly in a concentration-and time-dependent manner in donepezil-treated cells (Fig. 4a,b). The Aβ42 level also decreased significantly but after 48 h (Fig. 4c,d).

Nicotinic and muscarinic AChR antagonists do not affect donepezil-induced modulation of APP
shedding. Our previous studies have shown that donepezil treatment up-regulates phosphoinositide 3-kinase (PI3K) signalling through activation of nAChRs [25][26][27] . Because PI3K signalling is involved in a multitude of cellular responses, including regulation of the expression of various downstream genes, donepezil might up-regulate SNX33 transcription, and thereby modulate APP cleavage via the nAChR-PI3K signalling pathway. Analysis of the effects of nicotinic and muscarinic AChR antagonists on donepezil-dependent decrease of Aβ40 and Aβ42 levels in the culture media by ELISA showed that simultaneous treatment with donepezil and either mecamylamine or scopolamine does not significantly alter the Aβ40 and Aβ42 levels in the culture media (Fig. 5a,b). Therefore, regulation of SNX33 expression by donepezil might depend on mechanisms independent of the nAChR-PI3K pathway.
www.nature.com/scientificreports www.nature.com/scientificreports/ Target-specific morpholino oligos can suppress SNX33 expression in primary cortical cells. Primary neurons transfected with SNX33 morpholino oligos (MOs) showed significantly lower SNX33 expression than cells transfected with control MOs (Fig. 6a,b), suggesting that target-specific MOs can inhibit SNX33 expression. Cells transfected with control MOs showed a significant increase in SNX33 expression by donepezil treatment, but in cells transfected with SNX33 MOs, donepezil treatment did not significantly increase SNX33 expression (Fig. 6c,d). These results indicated that transfection of target-specific MOs effectively suppresses gene expression of SNX33 in primary cortical cultures and prompted us to investigate the role of SNX33 in donepezil-dependent modulation of APP processing.

SNX33 is required for modulation of APP processing by donepezil treatment. In control
MO-transfected cells, donepezil treatment resulted in a significant increase in the sAPPα levels ( Fig. 7a,b). However, in SNX33 MO-transfected cells, the increase in sAPPα levels was not significant. Because this lack of increase in sAPPα levels may be owing to decrease in APP levels, we examined APP levels in control and SNX33 MO-transfected cells. Results showed no significant change in APP levels between control and SNX33 MO-transfected cells or between donepezil-treated and untreated cells (Fig. 7c,d). In addition, although donepezil treatment led to a significant decrease in Aβ40 and Aβ42 levels in control MO-transfected primary cortical neurons (Fig. 7e,f), it did not significantly decrease Aβ40 and Aβ42 levels in SNX33 MO-transfected cells.  (1) for 24 h. The relative mRNA level of SNX33/β-actin was determined by RT-qPCR. *P < 0.05 compared with control cells. Data are expressed as mean ± SEM of n = 4 independent observations. (b,c) Cells were treated with 10 μM donepezil for 48 h. Cell lysate was subjected to western blotting. Representative images of western blotting analysis (b) and quantification of SNX33/β-actin ratio (c) are shown. ***P < 0.001 compared with control cells. Data are expressed as mean ± SEM of n = 3 independent observations. SNX33, sorting nexin protein 33; mRNA, messenger ribonucleic acid; RT-qPCR, real-time quantitative polymerase chain reaction.
the cell surface showed that in control MO-transfected cells, APP levels on the cell surface increased with donepezil treatment (Fig. 8a,b). However, we did not observe an increase in APP levels on the cell surface in SNX33 MO-transfected cells. (Fig. 8a,b).

Discussion
In this study, we identified SNX33 as a downstream factor up-regulated by donepezil treatment. We also investigated the effects of donepezil treatment on APP processing in primary cortical neurons and role of SNX33 in donepezil treatment-induced effects.
Our results are consistent with the notion that donepezil modifies APP processing via up-regulation of SNX33 expression. Donepezil up-regulated sAPPα and surface APP levels but not total APP levels. Further, donepezil significantly decreased Aβ levels in culture media. Because cleavage by α-secretase, which generates sAPPα, occurs within the Aβ sequence and precludes Aβ generation, the effects of donepezil are consistent with the regulation of APP processing by donepezil. The effects of donepezil were reversed upon inhibition of the expression of SNX33, whose function is to modify APP processing through regulation of endocytotic transport of APP 12 . These results are consistent with an important role SNX33 plays in donepezil-dependent modification of APP processing.
According to a straightforward model, donepezil treatment increases SNX33 expression, which reduces the rate of APP endocytosis, and leads to increased APP levels at the cell membrane surface, where APP is cleaved by α-secretase, thus decreasing Aβ generation. However, Zimmerman et al. 20 showed that donepezil treatment increases the cell membrane levels of constitutive α-secretases of APP, ADAM10 and ADAM17 20 . Ebsen and colleagues showed that the SH3 domain of SNX33 interacts with ADAM10 28 . This raises the possibility that the donepezil-dependent increase in sAPPα and decrease in Aβ are owing to the SNX33-dependent increase of ADAM10 and ADAM17 levels and their activity or stability. Further studies are required to uncover how donepezil regulates APP processing via up-regulation of SNX33 expression.
In this study, we were unable to identify the mechanism responsible for the up-regulation of SNX33 expression. Our previous studies have reported that donepezil activates the PI3K signalling pathway through nAChR activation and that donepezil regulates at least one of its downstream effector genes, α7-nAChR, through the nAChR-PI3K signalling pathway 25,29 . Our initial assumption was that SNX33 expression is up-regulated via the nAChR-PI3K signalling pathway. However, we observed that neither nicotinic nor muscarinic AChR antagonists can reverse the effects of donepezil treatment. These results are consistent with those of a previous study, in which the effects of atropine and bungarotoxin (muscarinic and nicotinic receptor antagonists, respectively) were tested on the donepezil-dependent increase of sAPPα in SH-SY5Y cells 20 . Conversely, the Aβ-decreasing effect of galantamine, an allosteric potentiating ligand of nAChRs 30 , is blocked by methyllycaconitine (an α7-nAChR antagonist) 21,22 , bungarotoxin, atropine and mecamylamine 21 , suggesting that AChRs play a role in galantamine-dependent modulation of APP processing. It is still unclear whether rivastigmine modifies APP processing through AChRs, but stimulation of muscarinic AChRs has been shown to increase sAPPα release 19 . www.nature.com/scientificreports www.nature.com/scientificreports/ Further analysis is required to understand the mechanism leading to SNX33 up-regulation and uncover similarities, if any, between donepezil-and galantamine-dependent modifications of APP processing. Such mechanism may become a therapeutic target to control Aβ levels via regulation of APP processing.
The effects of AD drugs on suppression of Aβ production have also been tested in vivo. Dong and colleagues studied a Tg2576 AD mouse model and showed that donepezil treatment at higher concentrations decreases soluble Aβ protein level and Aβ accumulation compared with lower concentrations 31 . A similar decrease in soluble Aβ protein level, along with improvements in cognitive performance, was observed upon donepezil administration to aged AD model mice with fully developed amyloid pathology 32 . Wu et al. 24 showed that galantamine treatment reduces the total area of amyloid load within the hippocampus in aged APP/PS1 mice 24 . In addition, rivastigmine treatment increases sAPPα and decreases Aβ40 levels in primary embryonic rat cortical cultures 23 . However, the mechanism altering sAPPα and Aβ levels remains unknown. It should also be noted that one study reported that donepezil administration to human patients does not result in a significant decrease in brain Aβ accumulation 33 . On the contrary, a large-scale study has shown that donepezil delays hippocampal atrophy in patients suspected to have prodromal AD, which supports the neuroprotective effect of donepezil 34 . These results indicate that donepezil, galantamine and rivastigmine seem to share the potential to modulate APP processing, although the underlying mechanism is still unclear. Our study might contribute to a further understanding of the phenomenon. Because endocytic transport is among the earliest to display neuropathological abnormalities during early-stage AD and is believed to precede Aβ deposition 35,36 , a new AD therapeutic strategy towards control of Figure 3. Effects of donepezil on non-amyloidogenic processing of APP in primary cortical cells. Cells were treated with 1 μM donepezil for 72 h. Cell lysate and culture media were subjected to western blotting. Representative images of western blotting analysis (a) and quantification of sAPPα (b) and APP/β-actin ratio (c) are shown. The grouping of blots was cropped from different parts of the same gel. *P < 0.05 compared with control cells. n.s., not significant. Data are expressed as mean ± SEM of n = 6 independent observations. APP, amyloid precursor protein.  www.nature.com/scientificreports www.nature.com/scientificreports/ APP processing via action of SNX family proteins could be promising. Whether there are common mechanisms for APP modification by donepezil, galantamine and rivastigmine is yet to be examined. It would be interesting to investigate whether the expression of other SNX family proteins is up-regulated by galantamine and rivastigmine treatment and whether such proteins are responsible for a decrease in Aβ production.

conclusion
In conclusion, we identified a downstream factor of donepezil, SNX33, an endosomal trafficking protein known to regulate processing of APP. Our results indicate that donepezil regulates endocytic trafficking of APP via up-regulation of SNX33 expression.
introduction of morpholino oligos. We delivered 1 mM MOs; Genetools, Philomath, OR, USA) into the cells using the Endo Porter delivery reagent. The sequence of antisense MOs was as follows: 5′-CATGGCTAAGTTATCTGTTAGGTCA-3′. Standard Control Oligo, Classic was used as a negative control. The sequence of control MOs was as follows: 5′-CCTCTTACCTCAGTTACAATTTATA-3′. www.nature.com/scientificreports www.nature.com/scientificreports/ preparation of cell membrane surface proteins. We prepared cell membrane surface proteins using the Pierce ® Cell Surface Protein Isolation Kit (Thermo Fisher Scientific, MA, USA) according to the manufacturer's instructions. Briefly, cells were washed twice with cold PBS. Then, the cell membrane surface was biotinylated by incubating the cells with PBS containing Sulfo-NHS-SS-Biotin. Next, cells were harvested with a cell scraper into a conical tube and centrifuged. The supernatant was then discarded, and cells were pipetted with Tris-buffered saline and centrifuged. The supernatant was again discarded. Cells were then lysed, sonicated and centrifuged at 15,000 rpm at 4 °C for 2 min. After normalisation of protein concentrations, we applied the supernatant to NeutrAvidin Agarose gel columns and incubated it for 1 h. Gels were washed thrice, and biotinylated surface proteins stuck to the gel were eluted by incubating the gel with a SDS-polyacrylamide gel electrophoresis sample buffer containing dithiothreitol for 1 h. Finally, total lysates and eluted surface proteins were subjected to western blotting. Statistical analysis. Data were expressed as mean ± standard error of the mean (SEM). The statistical significance of between-group differences was determined with the unpaired t-test or one-way analysis of variance, followed by Tukey's test, using Prism software (GraphPad Software, San Diego, CA, USA). P < 0.05 was considered to be statistically significant.