Regulation of Amyloid Precursor Protein Processing by Serotonin Signaling

Proteolytic processing of the amyloid precursor protein (APP) by the β- and γ-secretases releases the amyloid-β peptide (Aβ), which deposits in senile plaques and contributes to the etiology of Alzheimer's disease (AD). The α-secretase cleaves APP in the Aβ peptide sequence to generate soluble APPα (sAPPα). Upregulation of α-secretase activity through the 5-hydroxytryptamine 4 (5-HT4) receptor has been shown to reduce Aβ production, amyloid plaque load and to improve cognitive impairment in transgenic mouse models of AD. Consequently, activation of 5-HT4 receptors following agonist stimulation is considered to be a therapeutic strategy for AD treatment; however, the signaling cascade involved in 5-HT4 receptor-stimulated proteolysis of APP remains to be determined. Here we used chemical and siRNA inhibition to identify the proteins which mediate 5-HT4d receptor-stimulated α-secretase activity in the SH-SY5Y human neuronal cell line. We show that G protein and Src dependent activation of phospholipase C are required for α-secretase activity, while, unexpectedly, adenylyl cyclase and cAMP are not involved. Further elucidation of the signaling pathway indicates that inositol triphosphate phosphorylation and casein kinase 2 activation is also a prerequisite for α-secretase activity. Our findings provide a novel route to explore the treatment of AD through 5-HT4 receptor-induced α-secretase activation.


Introduction
The most common form of dementia in elderly people is Alzheimer's disease (AD), which is pathologically characterized by progressive neuronal loss and deposition of the amyloid-b peptide (Ab) in amyloid plaques. Current therapeutic targets are the band c-secretases, which generate Ab from amyloid precursor protein (APP). Several drugs inhibiting or modulating the activity of these secretases have failed in clinical trials due to severe side effects or to difficulty in delivery through the blood brain barrier [1]. APP is also cleaved by a third secretase called a-secretase. The a-secretase cleaves APP within the Ab peptide sequence, producing a soluble APPa fragment (sAPPa), which precludes Ab generation. Indeed, in vivo overexpression or upregulation of asecretase activity in mice indicate that a-secretase activation leads to a decrease in Ab production and a reduction in the amyloid plaque load in AD mouse models [2,3]. These effects were accompanied by an improvement in the cognitive deficits, providing considerable support for modulation of a-secretase activity as a viable strategy in the fight against AD [2,3].
To specifically target the non-amyloidogenic pathway of APP processing, a fundamental consideration would be to understand the mechanism of a-secretase activation and to determine the signaling cascade of kinases and second messengers that directly regulate a-secretase-mediated proteolysis of APP. These molecules can be directly targeted pharmacologically, but also indirectly via G protein-coupled receptors (GPCR), such as the muscarinic, glutamatergic and serotonergic receptors. In particular, the G protein coupled 5-hydroxytryptamine 4 (5-HT 4 ) receptor is gaining considerable interest as a modulator of a-secretase activity due to its role in memory and learning and regulation of APP processing [4]. Activation of the 5-HT 4 receptor leads to an increase in the population spike amplitude in the hippocampal CA1 region, and this effect persists in a transgenic mouse model of AD [5,6], suggesting that 5-HT 4 receptor-mediated signaling remains functional under these pathological conditions. On the other hand, agonist stimulation of the 5-HT 4 receptor results in increased sAPPa secretion with a concomitant decrease in Ab peptide levels in primary neuronal cultures and an alleviation of amyloid plaque load in AD mouse models [7][8][9]. Such amelioration of disease pathology is correlated with improvements in memory and learning in behavioral paradigms and scopolamineinduced models of cognitive deficit [10][11][12]. Additionally, an increase in acetylcholine release is observed after 5-HT 4 receptor agonist application in vivo [13]. This could be a valuable property when considering 5-HT 4 receptor agonists for AD treatment, which could complement the currently licensed therapy of cholinesterase inhibition for partial symptomatic relief [14].
Despite numerous reports on 5-HT 4 receptor function in memory and learning and its effect on APP processing, the downstream signaling pathway responsible for this 5-HT 4 receptor-mediated effect is still poorly understood. 5-HT 4 receptor stimulation results in an accumulation of cAMP, a second messenger required for protein kinase A (PKA) and exchange protein activated by cAMP (Epac) activation. However, 5-HT 4 receptor-mediated non-amyloidogenic processing of APP occurs independently of PKA activation, but can be regulated by Epac1 activation of Rac1 and Rap signaling in cell lines and primary neurons [15]. The 5-HT 4 receptor is constitutively bound to the Src non-receptor tyrosine kinase, which is required for ERK activation [16]. In addition, 5-HT 4 receptor stimulation in adrenocortical cells and cardiomyocytes results in an increase of calcium influx, which results in activation of voltage-gated calcium channels through PKA [17,18]. It is unknown whether these latter pathways also contribute to a-secretase activation downstream of the 5-HT 4 receptor. Altogether, these studies suggest a complicated picture of the downstream signaling pathways involved in 5-HT 4 receptor stimulation and reveal the importance of delineation of the mechanism of 5-HT 4 receptor-mediated APP proteolysis.
Finally, several metalloproteinases have been proposed as asecretase; however, the identity of 5-HT 4 receptor-induced asecretase activity has not been fully addressed. The disintegrin and metalloprotease ADAM10, a major constitutive a-secretase of APP [19,20], is a feasible candidate [21]. However, ADAM17 is more likely to be the inducible APP a-secretase based on studies which have investigated the regulated ectodomain shedding of other ADAM substrates after protein kinase C (PKC) activation [22]. In support of this is the observation that M1 receptor induced sAPPa release correlates with increased ADAM17 expression levels [23]. Nevertheless, additional metalloproteinases, such as meprin b and membrane-type matrix metalloproteinases, were shown to mediate a-cleavage of APP [24,25].
Here, we specifically determined the intracellular signaling cascade involved in 5-HT 4d receptor stimulation and inducible asecretase activity. We used human SH-SY5Y neuroblastoma cells to analyze APP processing for practical reasons and experimental consistency. Human SH-SY5Y cells can generate sustainable cells with characteristics that resemble the morphology and biochemistry of mature neurons [26]. We present evidence that the G protein-dependent pathway activating Src, phospholipase C (PLC) and casein kinase 2 (CK2) is responsible for the 5-HT 4d receptorstimulated induction of a-secretase activity. Interestingly and in contrast to previous publications, we find that adenylyl cyclase (AC) and cAMP signaling are not required for 5-HT 4d receptormediated a-secretase activity [15]. Furthermore, we analyzed the reported a-secretases as putative mediators of the 5-HT 4d receptor effect on APP shedding using RNAi studies.

Inhibitor treatment and soluble APP analysis (SEAP assay)
SH-SY5Y human neuroblastoma cells (CRL-2266, ATCC) were cultured in DMEM/F12 supplemented with 10% fetal bovine serum (FBS). For analysis of soluble APP secretion, a mix of 1,5 mg plasmid encoding human wild type APP695 linked to Alkaline Phosphatase (AP-APP) at the N-terminus (pEAK12-AP-APP, [29]), 1,35 mg of 5-HT 4d receptor isoform in pcDNA3.1 (pcDNA3.1-5-HT 4d , kindly provided by Joris De Mayer and Jan Schuurkes, Movetis, Turnhout, Belgium) and 0,15 mg of GFP (pmaxFP-Green-N, Amaxa) was prepared in OPTI-MEM and combined with 20 ml of Lipofectamine 2000 (Invitrogen). After 20 minutes incubation at room temperature transfection mix was combined with a trypsinized cell suspension in growth medium containing 10% FBS. After another 15 minutes incubation at room temperature cells were seeded in a 96-well plate at 80.000 cells/well. The next day medium was changed to DMEM/F12 supplemented with 5% dialyzed FBS (10,000 molecular weight cutoff), which is devoid of serotonin otherwise present in undialyzed FBS that causes 5-HT 4 receptor desensitization. After three days, cells were washed and incubated in serum free medium (SFM) for another 24 hours. Next cells were stimulated with 1 mM of the following compounds: prucalopride, 5-HT, PMA and GR113808 in SFM for 24 hours and the conditioned medium was analyzed for secreted AP (SEAP) activity with Great EscAPe SEAP Chemiluminescence Kit 2.0 (Clontech) according to manufacturer's instructions. Luminescence was measured with the EnVisionH multilabel reader (PerkinElmer). For signaling studies dilution curves of inhibitory compounds were made in combination with induction by 1 mM prucalopride or 5-HT. DMSO incubation was used as a control in all experiments carried out. The ratio of individual luminescence counts from the tested conditions to the mean value of DMSO treated cells was plotted as SEAP fold induction. Cells were used for the MTS proliferation assay (CellTiter 96H AQ ueous Non-Radioactive Cell Proliferation Assay) and the LDH cytotoxicity assay (CytoTox 96H Non-Radioactive Cytotoxicity Assay) according to manufacturer's instructions (Promega). Compound dilution curves were performed in the range of the reported effective concentrations (Table 1) and working concentrations were determined in the SEAP assay as those giving significant inhibition of 5-HT 4d receptor-stimulated sAPPa secretion. MTS and LDH assays were used to define working concentrations of the different compounds that were nontoxic to the cells. cAMP assay cAMP levels were assessed using the AlphascreenH cAMP assay kit (PerkinElmer Life Sciences). 2,88?10 6 or 1?10 6 SH-SY5Y cells were seeded in T75 or T25 flasks, respectively. Adherent cells were transfected after 4 hours with Lipofectamine and Plus reagent (Invitrogen) according to the manufacturer's instructions. A mix of 7,5 mg pEAK12-AP-APP, 6 mg pcDNA3.1-5-HT 4d and 1,5 mg pmaxFP-Green-N plasmids (ratio of 5:4:1) was used for transfection in T75 flasks. A mix of 1,25 mg pEAK12-AP-APP, 0,65 mg pcDNA3.1-5-HT 4d receptor and 2,6 mg pcDNAI-Amp-Ga s DN or pcDNA3.1 as an empty vector control (ratio of 2:1:4) was used for transfection in T25 flasks. 3 hours later transfection mixes were replaced with growth medium for 16 hours and cells were treated with medium supplemented with 5% dialyzed FBS and SFM as described under ''inhibitor treatment and soluble APP analysis''.
Then cells were gently dissociated with Versene solution (Invitrogen) to obtain a single cell suspension. Next cells were counted to determine the exact cell number. Equal numbers of cells were combined with the acceptor beads coupled to an anti-cAMP antibody and biotinylated cAMP, both provided in the AlphascreenH cAMP assay kit (PerkinElmer Life Sciences), and a serial dilution of compound. After incubating the cells for one hour, streptavidin-donor beads were added and the cells were permeabilized with 0,3% Tween-20 for 30 minutes, which released intracellular cAMP. The assay is based on competition between endogenously produced cAMP by the stimulated cells and exogenously added biotinylated cAMP. The electron transfer between donor and acceptor beads was measured with the EnVisionH multilabel reader (PerkinElmer). DMSO was diluted to a final concentration of 0,1% and kept equal in all samples to avoid differential effects of different DMSO concentrations on the cells. cAMP concentrations were determined using a standard curve.

Calcium measurements
Calcium imaging was assessed using the Fluo-4 NW calcium assay kit (Invitrogen). SH-SY5Y cells were transfected with pEAK12-AP-APP, pcDNA3.1-5-HT 4d receptor and pmaxFP-Green-N in Optilux black wall clear bottom plates (BD Biosciences) and treated as described under ''inhibitor treatment and soluble APP analysis''. Next cells were loaded with Fluo-4 NW dye mix according to manufacturer's instructions. Binding with calcium ions increases fluorescence of the dye. Baseline fluorescence of the dye was recorded at the steady state, while stimulated calcium release was assessed after automated addition of the compounds at different time points using IN Cell Analyzer 2000 (GE Healthcare). Calcium images were analyzed using the ''Plot Z-axis Profile'' function of ImageJ (NIH). Data are presented as a ratio of fluorescence intensity of Fluo-4 NW at any given time to baseline fluorescence (F/F 0 ).

Construction of mutated cDNA
Mutations in the cDNA of the 5-HT 4d receptor were introduced using the QuickChange II XL site-directed mutagenesis kit from Stratagene. All vector modifications were validated with sequencing using BigDyeH Terminator v3.1 Cycle Sequencing and the ABI PrismH 3100 Genetic Analyzer (Applied Biosystems). Obtained data were analyzed with the Sequence Scanner program and LALIGN tool from ch.embnet.org.

siRNA-mediated knockdown and immunoblotting
Knockdown of the proteins of interest was performed 4 hours after SH-SY5Y cells were transfected with pEAK12-AP-APP, pcDNA3.1-5-HT 4d receptor and pmax-FP-Green-N plasmids. Half of the medium was replaced with transfection mix containing 3 nM target protein siRNA and Lipofectamine RNAiMAX (Invitrogen) and left on the cells overnight. Next we proceeded with the protocol as described under ''inhibitor treatment and soluble APP analysis''. The following siRNAs were used: Stealth RNAi TM siRNAs were used for GNAS HSS104240, GNAQ HSS104237, GNA13 HSS173827, PLCG1 HSS108094, CSNK2A1 HSS175396, ADAM9 HSS189548, MMP9 HSS181135 and BLOCK-iT TM Alexa FluorH Red Fluorescent Oligo as a control (Invitrogen). The siGENOME SMARTpool was used for ADAM10 and siGENOME Non-Targeting siRNA Pool #1 as a control (Dharmacon). The FlexiTube GeneSolution GS6868 SI02664501 was used for ADAM17 and AllStars Negative Control siRNA as a control (QIAgen). Conditioned medium was collected to measure SEAP activity. For detection of Ga s , Ga q , Ga 13 , CK2, ADAM9, 10, APP and b-Actin, cells were lysed in RIPA buffer (50 mM Tris-HCl pH 7.4, 150 mM NaCl, 1% NP-40, 0.5% Sodiumdeoxycholate, 0.1% SDS and Complete protease inhibitor tablets (Roche Applied Science)). For detection of MMP9, conditioned medium was cleared from cell debris by centrifugation at 1500 rpm and concentrated with Ultracel-50 centrifugal filter unit (Millipore) according to manufacturer's instructions. For detection of ADAM17, cells were homogenized in 50 mM Tris-HCl pH 8.0 and 150 mM NaCl, 1 mM batimastat Table 1. Overview of agonists, antagonists and inhibitors used to investigate the proteins contributing to the induction of sAPPa after 5-HT 4d receptor stimulation.

Compound
Target Ag/antag/inh Potency Experimental system Citation Prucalopride 5-HT 4 Ag EC 50 10 nM SH-SY5Y cells [9] 5-HT 5-HT 4 Ag EC 50  and Complete protease inhibitor tablets and centrifuged at 100.000 g for 1 hour at 4uC. Pellets were resuspended in RIPA buffer supplemented with 1 mM batimastat and centrifuged at 100.000 g for 1 hour at 4uC and supernatants containing the membrane fraction were collected. Protein concentrations were determined in each preparation using the Bradford assay (Bio-Rad). Equal amounts of protein were separated with SDS-PAGE in Novex Bis-Tris gels (Invitrogen), transferred to nitrocellulose membranes (Whatman), blocked and probed with antibodies in 3% milk plus 0,1% Tween-20/TBS buffer. Secondary antibody staining was detected with the Renaissance chemiluminescence kit (Perkin Elmer). b-Actin staining was used as a loading control.

Statistical analysis
Differences between multiple means of data with parametric distribution were assessed by ANOVA followed by Tukey-Kramer or Dunnet's post-hoc tests. All experiments were repeated three times. All statistical analysis was performed with GraphPad Prism 5 (GraphPad Software). Optical density of specific immunobands on western blot was quantified using the 1D densitometry package of Aida Image Analyzer v4.27.039.

5-HT 4d receptor-stimulated APP shedding requires G protein signaling and is independent of b-arrestin recruitment
In agreement with previous reports which show that activation of the 5-HT 4d receptor induces APP shedding in CHO cells [7], we find a significant upregulation of SEAP (secreted alkaline phosphatase) activity in SH-SY5Y human neuroblastoma cells transiently transfected with human 5-HT 4d (pcDNA3.1-5-HT 4d ) and human wild type APP695 coupled to SEAP (pEAK12-AP-APP) following treatment with the 5-HT 4 receptor agonists prucalopride or 5-HT. Co-treatment with the 5-HT 4 receptor antagonist GR113808 abolishes the increase in AP-APP shedding, demonstrating the specificity of the effect ( Figure 1A). Using an alphaLISA specific for sAPPa, we previously demonstrated that the increase in SEAP activity reflects an increase in sAPPa release [9].
Signaling pathway activation down-stream of the 5-HT 4 receptor is mediated through assembly and activation of a heterotrimeric G protein complex. Receptor phosphorylation by GPCR-related kinases (GRKs) limits G protein-mediated signaling and facilitates recruitment of b-arrestins, which mediate receptor internalization and turnover and provide a scaffold for the initiation of signals to several kinases [30]. We used 5-HT 4d receptor mutants deficient in either G protein or b-arrestin signaling to distinguish which pathway leads to increased sAPPa secretion upon receptor stimulation. We introduced mutations in the DRY conserved motif, which interrupt coupling to G proteins and prohibit the G protein complex from acquiring an active GTP bound state [31,32]. Expression of the alanine substitution DRY mutant of the 5-HT 4d receptor (pcDNA3.1-5-HT 4d DRY117/ 118AAY) in pEAK12-AP-APP transfected cells resulted in significant downregulation of sAPPa secretion after 5-HT treatment, indicating the putative involvement of G protein signal transduction in 5-HT 4d receptor-stimulated release of sAPPa ( Figure 1B). Recruitment of b-arrestins to the receptor requires phosphorylation of the C terminus, allowing further internalization and possible signal transduction through the scaffolding of down-stream kinases. We generated a 5-HT 4d receptor mutant truncated at amino acid 346 in the C terminus (pcDNA3.1-5-HT 4d D346), which lacks a conserved sequence of serine and threonine residues required for association of b-arrestins with the receptor after phosphorylation by GRKs. Expression of this mutant in pEAK12-AP-APP transfected cells maintained stimulated induction of sAPPa secretion after prucalopride and 5-HT treatment ( Figure 1B). These data suggest that b-arrestins do not contribute to 5-HT 4d receptor-induced APP shedding.
In order to confirm that G proteins participate in a-secretase induction, we co-treated pEAK12-AP-APP transfected cells with inhibitors of Ga s , i.e. CTB and NF449. Inhibition of Ga s signaling indeed resulted in significant decrease of induced sAPPa secretion ( Figures 1C and D). We also tested a Ga s DN mutant, which abolishes all GPCR-mediated G protein-dependent signaling [28]. We found that co-transfection of pcDNAI-Amp-Ga s DN with pEAK12-AP-APP and pcDNA3.1-5HT 4d inhibited induced sAPPa secretion after prucalopride and 5-HT treatment ( Figure 1E), indicating that Ga s activation is involved in 5-HT 4d receptor-stimulated a-secretase activity.
The 5-HT 4 receptors promiscuously activate several G proteins, i.e. Ga s , Ga q and Ga 13 , leading to distinct second messenger generation [33][34][35]. We wondered which G protein subunit is specifically responsible for the effect on sAPPa secretion. Therefore, we performed RNAi mediated knock-down of Ga s , Ga q and Ga 13 in SH-SY5Y cells and analyzed sAPPa secretion upon 5-HT 4d receptor stimulation ( Figure S1A). Knock-down efficiency and specificity of siRNA oligonucleotides was confirmed by western blotting (Figures S1B and C). Surprisingly, our data show that single knock-down of each individual Ga subunit equally abolishes 5-HT 4d receptor-mediated sAPPa secretion ( Figure S1A), suggesting that sAPPa release can be mediated through Ga s , Ga q and Ga 13 . Such an effect could be explained if there is a requirement for the functional activation of the Gbc subunits. We used gallein to inhibit Gbc signaling and found that co-treatment of pEAK12-AP-APP transfected cells with this inhibitor and prucalopride or 5-HT abolished induction of sAPPa secretion ( Figure 1F). Altogether, these studies suggest that Ga and Gbc activation is required for 5-HT 4d receptor-stimulated sAPPa release.

5-HT 4d receptor-stimulated APP shedding does not involve activation of adenylyl cyclase and cAMP
Ga s and cAMP mediate canonical signaling of 5-HT 4 receptors [33]. Therefore, we sought to determine whether, in SH-SY5Y cells, accumulation of cAMP is also necessary for a-secretase activity as previously described for CHO cells [15]. We used the adenylyl cyclase inhibitors SQ22536 and DDA, which potently inhibit increases in cAMP levels ( Figure 2C). Interestingly, we found that these inhibitors do not affect prucalopride or 5-HTstimulated sAPPa release in SH-SY5Y cells (Figures 2A and B). These results suggest that activation of adenylyl cyclase and accumulation of cAMP is not required for 5-HT 4d receptorstimulated APP shedding.

5-HT 4d receptor-stimulated APP shedding requires Src and phospholipase C
Given that 5-HT 4d receptor-stimulated APP shedding does not require an elevation in cAMP levels, we sought to determine whether generation of inositol triphosphate (IP3) is involved in 5-HT 4d receptor-stimulated sAPPa release. This second messenger is produced by PLC and can be activated either directly downstream of Ga q and Gbc or through the Src non-receptor tyrosine kinase (reviewed in [36]). To analyze the contribution of PLC and Src, we co-treated pEAK12-AP-APP and pcDNA3.1-5-HT 4d transfected SH-SY5Y cells with the Src inhibitor Bosutinib or the PLC inhibitor D609 and 5-HT 4 receptor agonists. In both cases, we observed that APP shedding was abolished compared to control treatment (Figures 3A and B).
PLC cleaves phosphatidylinositol 4,5-bisphosphate into IP3 and diacylglycerol, resulting in mobilization of intracellular calcium and activation of several downstream effector proteins including PKC [37]. In addition, several studies suggest that calcium and PKC signaling can activate a-secretase shedding of APP [38,39]. Cotreatment of transiently transfected SH-SY5Y cells with the PKC inhibitor GF109203X did not induce sAPPa secretion after 5-HT 4d receptor stimulation. In contrast, direct activation of PKC with PMA induced sAPPa, but this induction was inhibited with GF109203X showing that the inhibitor was functional ( Figure 3C). Similarly, prucalopride did not significantly alter extracellular calcium influx in SH-SY5Y cells, in contrast to ionomycin and ATP; two positive controls that prove assay functionality ( Figure 3D). Taken together these data suggest that Src and PLC, but not PKC or calcium signaling, contribute to 5-HT 4d receptor-induced APP shedding.
5-HT 4d receptor-stimulated APP shedding requires inositol polyphosphates and casein kinase 2 IP3 can be further phosphorylated by inositol 1,4,5-triphosphate 3-kinase (IP3K) and inositol polyphosphate multikinase (IPMK) to generate inositol polyphosphates (IP4, IP5 and IP6). These molecules recently emerged as versatile second messengers with an increasing number of cellular functions [40]. We tested the IP3K inhibitor and the IPMK inhibitor CGA in transfected SH-SY5Y cells and found that prucalopride or 5-HT-stimulated asecretase activity depends on the generation of these inositol polyphosphates ( Figures 4A and B). The reported literature suggests that IP4 and/or IP6 can activate CK2 in vitro [41]. In  cells, Wnt3a can induce IP5 generation which then activates CK2 [42]. Inhibition of CK2 activity with TBB in pcDNA3.1-5-HT 4d receptor and pEAK12-AP-APP expressing SH-SY5Y cells stimulated with prucalopride or 5-HT led to a decrease in sAPPa down to baseline levels ( Figure 4C), suggesting that CK2 is involved in 5-HT 4d receptor-stimulated APP shedding. In addition, we found that co-transfection of CK2 siRNA completely abolished stimulated APP shedding in SH-SY5Y cells treated with prucalopride or 5-HT ( Figures 4D, E and F). These results demonstrate that 5-HT 4d receptor-stimulated APP shedding requires inositol polyphosphates and CK2.
ADAM9, ADAM10, ADAM17 and MMP9 are not responsible for 5-HT 4d receptor-stimulated a-secretase activity Several enzymes of the ADAM and MMP family, such as ADAM9, 10, 17 and MMP9, are suggested candidate proteins responsible for inducible shedding of APP (reviewed in [43]). To determine the relative contribution of the metalloproteinases in 5-HT 4d receptor-stimulated sAPPa release, we first analyzed expression levels of ADAM9, 10, 17 and MMP9 in SH-SY5Y cells. Expression of ADAM9, 10, 17 and MMP9 was not changed after prucalopride treatment of pEAK12-AP-APP transfected SH-SY5Y cells (Figures S3A and B). To test whether a metalloproteinase would be responsible for induced a-secretase activity, we treated the cells with non-toxic concentrations of the broad spectrum metalloproteinase inhibitor GM6001 (dose response curve shown in Figure S4). Treatment with GM6001 abolished induction of sAPPa secretion ( Figure 5A), confirming that a metalloproteinase is indeed responsible for 5-HT 4d receptorstimulated sAPPa release. To identify the 5-HT 4d receptorstimulated a-secretase, we performed RNAi knock-down of the candidate a-secretases. We found that induction of sAPPa release was preserved after prucalopride treatment and single knock-down of ADAM9, 10, 17 or MMP9 ( Figure 5B). The efficiency of the downregulation was between 85-95% as documented by western blot analysis (Figures 5C and D). These data suggest that ADAM9, 10, 17 or MMP9 are not responsible for 5-HT 4d receptormediated inducible a-secretase activity in SH-SY5Y cells. We also analyzed constitutive sAPPa secretion upon ADAM10 knockdown in non-treated cells and confirmed that ADAM10 acts as the constitutive a-secretase of APP in our experimental conditions (data not shown).
Metalloproteinases are notorious for their functional redundancy between family members. To test whether more than one candidate metalloproteinase could be responsible for induction of a-secretase activity, we treated transfected SH-SY5Y cells with combinations of RNAi directed at ADAM9 and 10, ADAM9 and 17, ADAM10 and 17 ( Figure 6A). We observed no change in sAPPa secretion upon 5-HT 4d receptor stimulation under any of these conditions. Moreover, knock-down of all four candidate  Figure 6B). The levels of C-terminal fragments generated by the cleavage of APP at band b9-sites remained unchanged after the knock-down of ADAM9, 10, 17 and MMP9, suggesting that bsecretase activity was not affected by reduced expression levels of these metalloproteinases (Figures S5A and B). We used western blotting to confirm the efficiency and specificity of RNAi mediated downregulation (Figures 6C and D). Notice also the strong upregulation of MMP9 expression when ADAM9, 10 and 17 are downregulated, while single MMP9 knock-down did not affect 5-HT 4d receptor-induced sAPPa secretion. Altogether, our data suggest that an unidentified GM6001-sensitive metalloproteinase participates in the regulated cleavage of APP upon 5-HT 4d receptor stimulation ( Figure 6B).

Discussion
In this report, we examined the signaling pathway that leads to a-secretase induction after 5-HT 4d receptor stimulation in the human neuroblastoma SH-SY5Y cell line. We present here a previously uncharacterized signaling pathway involved in the mediation of 5-HT 4d receptor-induced a-secretase activity (Figure 7). The characterization of this pathway was based on a combination of pharmacological, siRNA and site-directed mutagenesis experiments. Our data indicate that PLC is essential for asecretase activation following 5-HT 4d stimulation. This effect is dependent on Ga and Gbc recruitment and signaling downstream of the 5-HT 4d receptor. Src tyrosine kinase acts as an intermediate molecule, mediating PLC activation and inositol triphosphate production. The latter is converted by multiple kinases to inositol polyphosphates, which activate CK2. Downstream of CK2, a yet unknown mechanism of a-secretase activation is triggered. The 5-HT 4d receptor-induced a-secretase activity could not be ascribed to any known candidate a-secretase (ADAM9, 10, 17 and MMP9) in the SH-SY5Y cells, which has hampered delineation of the final step regulating 5-HT 4d receptor-stimulated sAPPa release.
We found also that b-arrestin signaling did not contribute to asecretase activity upon 5-HT 4d receptor stimulation as the mutant receptor deficient in b-arrestin recruitment maintained the ability to stimulate sAPPa secretion after agonist treatment ( Figure 1B). Interestingly, b-arrestins have recently emerged as regulators of Ab generation downstream of the b2-adrenergic receptor and GPR3, independently of G protein activation [44,45]. In these studies, b-arrestins appear to bind to the Aph1 subunit of the csecretase complex, affecting complex localization and thereby increasing the catalytic activity of the c-secretase complex. Our work suggests that different signaling pathways regulate aand csecretase activity as we find that G proteins are indispensable for 5-HT 4d receptor-stimulated a-secretase activity, while they are not involved in the c-secretase regulation by GPCRs. Indeed, several molecules that are activated downstream of G proteins are proposed to regulate a-site APP processing, e.g. PKC, PKA, MAPK, ERK and PI3K (reviewed in [46]). Originally the 5-HT 4 receptor was shown to couple to Ga s [17,[33][34][35]47]. We show here that the Ga s , Ga q and Ga 13 subunits are equally required for sustainable a-secretase activity after 5-HT 4d receptor stimulation ( Figure S1A). We did not observe functional compensation between the different Ga subunits, even though protein expression was modulated in a reciprocal manner for Ga s and Ga q ( Figure S2). These results were unexpected and we speculate that parallel signaling initiated by the different Ga subunits or a certain threshold level of G proteins would be required to transduce the agonist-dependent signal. In those views reduction of the level of one of subunit could already abolish the signal. Our results also show that the signal relies on the association with the Gbc subunits as a converging signal transduction mediator ( Figure 1F). Similar observations have been previously made for PLC activation by G proteins in vitro. IP3 production was more efficient in the presence of the G protein trimeric complex than in separate preparations of either Ga q or Gbc alone [48,49]. Interestingly, we also observe a reduction in constitutive sAPPa release after the knock-down of Ga s , Ga q and Ga 13 ( Figure S1A). These data suggest a dominant negative effect of G proteins inhibition on constitutive a-secretase activity, which may be mediated by additional GPCRs besides the 5-HT 4d receptor.
5-HT 4 receptor coupling to different G proteins suggests several possibilities for downstream signal transduction. Several reports describe a PKA-independent and cAMP-dependent a-secretase activation following 5-HT 4d receptor stimulation [7,50,51]. In CHO cells, sAPPa release is regulated by Epac1, which promotes small GTPases Rap1 dependent Rac activation [15]. However, we find that AC and cAMP accumulation are not required for 5-HT 4d receptor-induced APP shedding under our experimental conditions ( Figure 2). Differences in the cellular systems, a human neuronal cell line versus a Chinese hamster ovary cell line, could explain the discrepancy between the studies. We then found that IP3 generation through Src and PLC activation contributes to 5-HT 4d receptor-induced a-secretase activity (Figure 3). PLC is also an important component of the a-secretase activation pathway through Ga q coupled GPCRs, e.g. mGluR1 and mGluR5 [52], M1 and M3 [53], 5-HT 2a and 5-HT 2c [54] and thus a point of convergence for several transduction pathways activating asecretase.
Investigations of the cerebral cortex and cerebellum of ADaffected individuals reveal disturbed G protein signal transduction compared to control patients [55]. In accordance, the phospho-inositide hydrolysis pathway is also altered in AD because of reduced levels of phosphatidylinositide 3-kinase and disturbed agonist and G protein regulation of PLC [56,57]. It is proposed that 5-HT 4 receptor stimulation could counteract such detrimental changes. We show here that the 5-HT 4d receptor indeed induces IP3K and IPMK mediated IP3 conversion to inositol polyphosphates and that these contribute to the non-amyloidogenic pathway of APP processing (Figure 4). This effect is mediated through the activation of CK2, which was recently identified to be downstream of the cholinergic receptors in a pathway of asecretase induction [58]. As activation of the 5-HT 4 receptor can increase acetylcholine levels in the brain [6,13] and we need 24 hours to obtain a significant induction of the a-secretase, an indirect mechanism through upregulation of acetylcholine could play a role. As our cells are of the dopaminergic origin, we think this possibility is rather unlikely. However, we cannot rule out that other indirect mechanisms are playing a role in the 5-HT 4d receptor-mediated a-secretase induction.
To understand the molecular mechanism of a-secretase activation downstream of the 5-HT 4d receptor, we investigated the contribution of ADAM9, 10, 17 and MMP9 in the regulation of APP processing. Previously, regulated a-secretase activity was partially attributed to MMP9, whose expression levels increased after 5-HT 4d receptor stimulation in APP-overexpressing H4 human neuroglioma cells [59]. However, in our cellular system, expression levels of the investigated proteinases do not change ( Figure S3) and specific protein downregulation suggests that a different metalloproteinase, besides the major candidate asecretases ADAM9, 10, 17 and MMP9, contributes to 5-HT 4d receptor-induced sAPPa release (Figures 5 and 6). Indeed, ADAM10 was not responsible for the 5-HT 4d receptor-dependent induction of sAPPa release through the cAMP/Epac pathway [21]. At this moment, we cannot rule out that the remaining protein expression of these four major a-secretases contribute to the preserved inducible a-secretase activity. Our data are consistent with the present view of different proteases contributing to regulated APP processing as previously reported for the M1 receptor, the insulin-like growth factor-1 receptor and the purinergic P2Y2 and P2X7 receptors [23,[60][61][62]. To identify the metalloproteinase(s) responsible for induced a-secretase activity we were reluctant to use differences in susceptibility to GM6001 because we were working with overexpression conditions. This would require a large-scale RNAi knock-down study but is beyond the scope of the current manuscript.
In conclusion, our studies show the complexity of a-secretase regulation upon 5-HT 4d receptor stimulation. Taking into consideration that receptor modulation of signaling pathways depends on the cellular context and that recombinant overexpression and RNA interference may reveal cell type specific results, a relevant physiological system should be used for the confirmation of the identified signaling pathway. Clinical trials of agonists targeting 5-HT 4 serotonergic and M1 muscarinic receptors will provide validation of a-secretase activation as a therapeutic approach for the treatment of AD. We report here that PLC dependent production of IP3 and CK2 activation are important mediators of the 5-HT 4d receptor signaling that enhance the non-amyloidogenic processing of APP. These proteins can also participate in signaling downstream of muscarinic receptors, suggesting the possibility of a common pathway for asecretase activation through GPCRs. Finally, our data will also aid with the development of 5-HT 4 receptor agonists as therapeutics for neurodegenerative or psychiatric disorders and allow for a better understanding of potential risks associated with these drugs. Figure 7. Schematic representation of the proposed 5-HT 4d receptor-stimulated signaling pathway leading to increased sAPPa production. The proteins involved in 5-HT 4d receptormediated non-amyloidogenic APP shedding are shown with green circles while orange circles and red characters indicate proteins or second messengers that were tested but were ineffective in modulating 5-HT 4d receptor-stimulated sAPPa release. The dotted lines with the question marks indicate remaining areas of investigation for further elucidation of the molecular mechanism of a-secretase activation. cAMP-dependent pathway of a-secretase induction was previously reported and is depicted as a plausible way for 5-HT 4d receptormediated sAPPa release [15]. Table S1 (DOCX)