Prostaglandin E2 Stimulates the Production of Amyloid-β Peptides through Internalization of the EP4 Receptor*

Amyloid-β (Aβ) peptides, generated by the proteolysis of β-amyloid precursor protein by β- and γ-secretases, play an important role in the pathogenesis of Alzheimer disease. Inflammation is also important. We recently reported that prostaglandin E2 (PGE2), a strong inducer of inflammation, stimulates the production of Aβ through EP2 and EP4 receptors, and here we have examined the molecular mechanism. Activation of EP2 and EP4 receptors is coupled to an increase in cellular cAMP levels and activation of protein kinase A (PKA). We found that inhibitors of adenylate cyclase and PKA suppress EP2, but not EP4, receptor-mediated stimulation of the Aβ production. In contrast, inhibitors of endocytosis suppressed EP4, but not EP2, receptor-mediated stimulation. Activation of γ-secretase was observed with the activation of EP4 receptors but not EP2 receptors. PGE2-dependent internalization of the EP4 receptor was observed, and cells expressing a mutant EP4 receptor lacking the internalization activity did not exhibit PGE2-stimulated production of Aβ. A physical interaction between the EP4 receptor and PS-1, a catalytic subunit of γ-secretases, was revealed by immunoprecipitation assays. PGE2-induced internalization of PS-1 and co-localization of EP4, PS-1, and Rab7 (a marker of late endosomes and lysosomes) was observed. Co-localization of PS-1 and Rab7 was also observed in the brain of wild-type mice but not of EP4 receptor null mice. These results suggest that PGE2-stimulated production of Aβ involves EP4 receptor-mediated endocytosis of PS-1 followed by activation of the γ-secretase, as well as EP2 receptor-dependent activation of adenylate cyclase and PKA, both of which are important in the inflammation-mediated progression of Alzheimer disease.

the synthesis of prostaglandin E 2 (PGE 2 ), a potent inducer of inflammation and has two subtypes, COX-1 and COX-2. COX-1 is expressed constitutively, whereas expression of COX-2 is induced under inflammatory conditions and is responsible for the progression of inflammation (20 -22). The following evidences of the involvement of PGE 2 (and COX-2) in the progression of AD suggest that they are good targets for the development of AD drugs: (i) Elevated levels of PGE 2 and overexpression of COX-2 have been observed in the brains of AD patients (23)(24)(25); (ii) the extent of COX-2 expression correlates with the amount of A␤ and the degree of progression of AD pathogenesis (26); (iii) transgenic mice constitutively overexpressing COX-2 show aging-dependent neural apoptosis and memory dysfunction (27); (iv) prolonged use of nonsteroidal anti-inflammatory drugs, inhibitors of COX, delays the onset and reduces the risk of AD (28); (v) PGE 2 stimulates the production of reactive oxygen species in microglia cells, resulting in activation of ␤-secretase (29).
We recently reported that PGE 2 stimulates the production of A␤ in human embryonic kidney (HEK) 293 and human neuroblastoma (SH-SY5Y) cells that stably express a form of APP with two mutations (K651N/M652L) (APPsw) that elevate cellular and secreted levels of A␤ (30). Similar results were reported by another group (31). Using agonists and antagonists specific for each of the four PGE 2 receptors (EP 1 , EP 2 , EP 3 , and EP 4 ), we found that EP 4 receptors alone and also both EP 2 and EP 4 receptors are involved in PGE 2 -stimulated production of A␤ in HEK293 or SH-SY5Y cells, respectively (30). Furthermore, experiments with transgenic mice suggest that EP 2 and EP 4 receptors are involved in the production of A␤ in vivo (30). Based on these results, we propose that antagonists of the EP 2 and/or EP 4 receptors may be therapeutically beneficial for the treatment of AD. Understanding the mechanism governing EP 2 and EP 4 receptormediated stimulation of production of A␤ by PGE 2 will be important for such drug development.
Activation of EP 2 and EP 4 receptors causes activation of adenylate cyclase and an increase in the cellular level of cAMP (32). We have shown that an EP 4 receptor agonist or both EP 2 and EP 4 receptor agonists increase the cellular level of cAMP in HEK293 or SH-SY5Y cells, respectively, and that a cAMP analogue, 8-(4-chlorophenylthio)-cAMP (pCPT-cAMP), increases the level of A␤ in HEK293 cells (30). These findings suggest that the cellular level of cAMP is important for PGE 2 -stimulated production of A␤. An increase in the cellular level of cAMP is known to activate protein kinase A (PKA), which is important for cAMP-regulated intracellular signal transduction (33). However, an inhibitor of PKA, N-[2-(p-bromocinnamylamino)ethyl]-5-isoquinoline-sulfonamide (H-89), does not block PGE 2 -stimulated production of A␤ in HEK293 cells (30). Other cAMP-regulated signal transduction factors, such as phosphatidylinositol 3-kinase and Epac (exchange protein directly activated by cAMP), were also shown not to be involved in PGE 2stimulated production of A␤ in HEK293 cells (30). Thus, the mechanism whereby the activation of EP 2 and EP 4 receptors stimulates the production of A␤ has remained unknown. In this study, by using inhibitors of adenylate cyclase and PKA, we found that activation of the EP 2 receptor stimulates production of A␤ through activation of adenylate cyclase and PKA. We also propose that activation of the EP 4 receptor causes its co-internalization with PS-1 (␥-secretase) into endosomes and that this co-internalization is important for EP 4 receptor-mediated stimulation of A␤ production by PGE 2 through the activation of ␥-secretase.
Animals-APP23 transgenic mice, a gift from Dr. M. Staufenbiel, were generated as described previously (34). APPsw/EP 4 Ϫ/Ϫ and APPsw/EP 4 ϩ/ϩ mice were generated as described previously (30). The experiments and procedures described here were carried out in accordance with the Guide for the Care and Use of Laboratory Animals as adopted and promulgated by the National Institute of Health and were approved by the Animal Care Committee of Kumamoto University.
For transient expression, cells were seeded 24 h before transfection in 24-well plates at a density of 1.5 ϫ 10 5 cells/well. Transfections were carried out using Lipofectamine (TM2000) according to the manufacturer's instructions. Cells were used for experiments after a 24-h recovery period. Transfection efficiency was determined in parallel plates by transfection of cells with pEGFP-N1 control vector. Transfection efficiencies were greater than 90% in all experiments. The stable transfectants expressing each gene were selected by immunoblotting or realtime reverse transcription-PCR analyses. Positive clones were maintained in the presence of 200 g/ml G418.
Immunoblotting Analysis-Whole cell extracts were prepared as described previously (36). For detection of CTF␣ and CTF␤, membrane fractions were prepared as described previously (37). The protein concentration of each sample was determined by the Bradford method (38). Samples were applied to polyacrylamide-SDS gels (Tris-Tricine gel for the detection of CTF␣ and CTF␤ or Tris-glycine gel for other proteins) and subjected to electrophoresis, after which proteins were immunoblotted with each antibody.
Sandwich Enzyme-linked Immunosorbent Assay for A␤ and ␥-Secretaseor ␤-Secretase-mediated Peptide Cleavage Assay-Cells were cultured for 24 h, and the conditioned medium was subjected to a sandwich enzyme-linked immunosorbent assay using three types of specific monoclonal antibodies as described previously (35,39).
We measured fluorescence using a plate reader (Fluostar Galaxy) with an excitation wavelength of 355 nm and an emission wavelength of 440 nm (for the ␥-secretase) or 510 nm (for the ␤-secretase).
Immunostaining Microscopy-Cells or mouse brain sections were incubated with antibody against each protein for 30 min before (for HA or EP4) or after (for PS-1 and Rab7) treatment with PGE 2 . Samples were fixed and incubated with the respective secondary antibody. We acquired images with a confocal fluorescence microscope (Olympus FV500).
Co-immunoprecipitation Assay-Immunoprecipitation was carried out as described previously (35), with some modifications. Cells were harvested, lysed with buffer containing 1% CHAPSO, and centrifuged. The antibody against HA or EP 4 was added to the supernatant, and the samples were incubated for 12 h at 4°C with rotation. Dynabeads-Protein G was added and incubated for 2 h at 4°C with rotation. Beads were washed four times, and the proteins were eluted by boiling in SDS sample buffer.
Surface Biotinylation Assay-This assay was carried out as described previously (42) with some modifications. Proteins on the cell surface were biotinylated with a reversible membraneimpermeable derivative of biotin (sulfo-NHS-S-S-biotin). Internalization of proteins was allowed to occur by incubation at 37°C for 1 h. The remaining cell surface biotin was cleaved by glutathione, and the cells were lysed. Biotinylated proteins were precipitated using UltraLink immobilized Neutravidin beads and eluted by boiling in SDS sample buffer.
Statistical Analysis-All values are expressed as the mean Ϯ S.D. Two-way analysis of variance followed by the Tukey test or the Student's t test for unpaired results was used to evaluate differences between more than three groups or between two groups, respectively. Differences were considered to be significant for values of p Ͻ 0.05. 2 Receptor-mediated Stimulation of A␤ Production by PGE 2 -Although primary neurons should be used for this type of experiments, we used an immortalized cell line in this study, because we used stable and transient transfection for the experiments in this study (see below). We confirmed our previous results that H-89 did not block PGE 2 -stimulated production of A␤ in HEK293 cells and found that an inhibitor of adenylate cyclase, SQ22536, also did not block this stimulation (supplemental Fig. S1, A and B), suggesting that an increase in the cellular level of cAMP is not involved in PGE 2stimulated production of A␤ in HEK293 cells. In contrast, in SH-SY5Y cells, both H-89 and SQ22536 decreased the level of A␤ in the presence, but not in the absence, of PGE 2 (supplemental Fig.  S1, C and D). These inhibitors at the concentrations used did not affect cell viability (data not shown). Therefore, the results shown in Fig.  1, C and D, suggest that an increase in the cellular level of cAMP and the resulting activation of PKA are involved in PGE 2 -stimulated production of A␤ in SH-SY5Y cells. In a previous report (30), we used HEK293 cells that express only APPsw stably. In the current study, we used CHO and HEK293 cells that express each EP receptor transiently in addition to the stable expression of APPsw (see below). For these types of cells, 10 nM PGE 2 , which had been used for experiments in the previous study (30), was not enough to stimulate the production of A␤ clearly (supplemental Figs. S7-S10), and thus we used 1 M PGE 2 in all experiments described in this article.

Mechanism for EP
As the EP 2 receptor is involved in PGE 2 -stimulated production of A␤ in SH-SY5Y but not in HEK293 cells (30), the results shown in supplemental Fig. S1 suggest that the  increase in the cellular level of cAMP and resulting activation of PKA contributes to EP 2 receptor-dependent (but not EP 4 receptor-dependent) stimulation of A␤ production by PGE 2 . To confirm this proposition, we examined the effects of these inhibitors on the production of A␤ in CHO cells artificially expressing APPsw and EP 2 or EP 4 receptor tagged with HA. Lack of functional endogenous EP 2 and EP 4 receptors in cells has been reported previously (43). Expression of these receptors was confirmed by immunoblotting (data not shown) and immunostaining (Fig. 3A). Both H-89 and SQ22536 suppressed PGE 2 -stimulated production of A␤ in cells expressing the EP 2 receptor but not in cells expressing the EP 4 receptor (supplemental Fig. S2, A-D), supporting our theory described above. We reported previously that treatment of HEK293 cells with PGE 2 increases ␥-secretase activity in extracts of these cells (30). Here we found that PGE 2 treatment increased ␥-secretase activity in extracts prepared from CHO cells expressing the EP 4 receptor but not in cells expressing the EP 2 receptor and that it did not affect ␤-secretase activity in either of these cell types (supplemental Fig. S2, E and F). Furthermore, pCPT-cAMP did not affect ␥-secretase activity in HEK293 cells (data not shown). On the other hand, treatment of cells with PGE 2 did not affect the expression of ␥-secretase (PS-1-NTF) (supplemental Fig.  S3). These results suggest that the EP 2 receptor mediates PGE 2stimulated production of A␤ through activation of cAMP and PKA without an increase in ␤and ␥-secretase activity and that the EP 4 receptor mediates PGE 2 -stimulated A␤ production through different mechanisms, which involve an activation rather than induction of expression of ␥-secretase.
Mechanism for EP 4 Receptor-mediated Stimulation of A␤ Production by PGE 2 -Agonist-dependent internalization (endocytosis) of the EP 4 receptor (but not the EP 2 receptor) has been described: The binding of PGE 2 to the EP 4 receptor induces formation of vesicles that contain the receptor and the vesicles are trafficked to endosomes (44,45). Thus, we used inhibitors of endocytosis (sucrose and concanavalin A) to test whether agonist-dependent internalization of the EP 4 receptor is involved in PGE 2 -stimulated production of A␤. As shown in Fig. 1, both sucrose and concanavalin A suppressed PGE 2 (1 M)-stimulated production of A␤ and decreased ␥-secretase activity in HEK293 cells. Similar results were obtained with 10 nM PGE 2 in HEK293 cells (supplemental Fig. S7) and in CHO cells expressing the EP 4 receptor but not in those expressing the EP 2 receptor (data not shown). These inhibitors, at the concentrations specified in Fig. 1, did not affect cell viability (data not shown). These results suggest that agonist-dependent internalization of the EP 4 receptor is involved in PGE 2 -stimulated A␤ production.
Agonist-dependent internalization is initiated by the formation of clathrin-coated vesicles, and thus clathrin is essential for this internalization (46). We examined the effect of siRNA targeting the clathrin heavy chain on PGE 2 -stimulated production of A␤ in HEK293 cells. Transfection with siRNA inhibited clathrin expression in the presence and absence of PGE 2 ( Fig.  2A). This siRNA suppressed PGE 2 -stimulated production of A␤ and ␥-secretase activity (Fig. 2, B and C), suggesting that clathrin-dependent vesicle formation at the cell surface and their subsequent internalization are involved in EP 4 receptormediated stimulation of A␤ production by PGE 2 . siRNA did not affect A␤ production in the absence of PGE 2 (Fig. 2B) and its pCPT-cAMP-dependent stimulation (Fig. 2D), suggesting that siRNA specifically affects EP 4 receptor-mediated stimulation of A␤ production by PGE 2 .
By immunostaining we observed PGE 2 (1 M)-dependent internalization of EP 4 receptors (Fig. 3A). In contrast, EP 2 receptors remained localized to the cell surface even after PGE 2 treatment (Fig. 3A). It has been reported that the C-terminal region of the EP 4 receptor is required for its agonist-dependent internalization (45). We confirmed that a mutant form of the EP 4 receptor, which is truncated after Thr-369 (EP 4 -t369) (47), does not exhibit agonist-dependent internalization (Fig. 3A). As shown in Fig. 3, B and C, in contrast to CHO cells expressing the wild-type EP 4 receptor, in CHO cells expressing EP 4 -t369, PGE 2 did not stimulate the production of A␤ and ␥-secretase activity. Similar results were obtained with 10 nM PGE 2 ; however, the effects were not so apparent, and some of them were not statistically significant (supplemental Fig. S8). This suggests that agonist-dependent internalization of the EP 4 receptor is essential for EP 4 receptor-mediated and PGE 2 -dependent stimulation of A␤ production and ␥-secretase activation.
In clathrin-dependent endocytosis, vesicles formed at the cell surface are trafficked first to early endosomes and then to LEL. Rab5 and Rab7 are essential for the traffic to early endosomes and LEL, respectively (48). To examine the role of the traffic in PGE 2 -stimulated A␤ production, the effects of siRNA for Rab5 and Rab7 on the production of A␤ were examined. Each siRNA clearly inhibited the expression of their target protein (Fig. 4, A and D). Furthermore, siRNA for Rab5 or Rab7 suppressed the PGE 2 -stimulated production of A␤ and ␥-secretase activity in HEK293 cells (Fig. 4, B, C, E, and F), suggesting that traffic of vesicles containing the EP 4 receptor to LEL is important for PGE 2 -stimulated A␤ production and ␥-secretase activity.

Contribution of Co-internalization of PS-1 with the EP 4 Receptor to PGE 2 -stimulated Production of A␤-It was
reported recently that ␥-secretase is activated when it is trafficked into endosomes via agonist-dependent internalization of the ␤-adrenergic receptor, which interacts with ␥-secretase (49). Thus, we hypothesized that the EP 4 receptor also interacts with ␥-secretase and that ␥-secretase is trafficked to LEL in a PGE 2 -dependent manner, resulting in activation of ␥-secretase and stimulation of A␤ production. To test this theory, we first examined the interaction between the EP 4 receptor and PS-1 by a co-immunoprecipitation assay. As shown in Fig. 5A, efficient immunoprecipitation of PS-1-NTF with antibody against was dependent on the expression of the HA-tagged EP 4 receptor. On the other hand, PS-1-NTF was not immunoprecipitated with antibody against HA in cells expressing HA-tagged EP 2 receptor (supplemental Fig. S4). These results suggest that the EP 4 receptor can physically and specifically interact with PS-1-NTF (␥-secretase). The physical interaction between EP 4 receptor and PS-1-NTF was also observed in SH-SY5Y cells without artificial overexpression of EP 4 receptor (supplemental Fig. S11A). PS-1 is cleaved to produce PS-1-NTF and PS-1-CTF in cells, and both of PS-1-NTF and PS-1-CTF are included in ␥-secretase complex. The results in Fig. 5 also show that EP 4 -t369 can interact with PS-1-NTF, suggesting that the interaction of PS-1 (␥-secretase) with the EP 4 receptor alone is not sufficient for PGE 2 -stimulated production of A␤ and ␥-secretase activity (see Fig. 3, B and C). It has been reported that a general acceleration of cellular endocytic pathways and agonist-induced endocytosis of some receptors (such as the angiotensin II receptor) does not affect the production of A␤ and ␥-secretase activity (49,50). Thus, internalization of PS-1 (␥-secretase) with the EP 4 receptor seems to enhance production of A␤ and ␥-secretase activity specifically.
Next, we tested co-internalization of PS-1-NTF with the EP 4 receptor using a surface biotinylation assay. Cells were surfacebiotinylated, and after induction of internalization, biotinylated proteins remaining on the cell surface were cleaved by glutathione. The biotinylated proteins (internalized proteins) were precipitated, and the presence of each protein in the precipitates was monitored by immunoblotting. As shown in Fig. 5B, the FIGURE 5. Interaction between the EP 4 receptor and PS-1 and their PGE 2dependent co-internalization. HEK293 cells expressing APPsw were transiently transfected with expression plasmid encoding the EP 4 receptor or EP 4 -t369 or with control vector. Whole cell extracts were immunoprecipitated with antibody against HA. A, whole cell extracts (WCE) and the immunoprecipitates (IP) were analyzed by immunoblotting with antibody against HA or PS-1-NTF as described in the legend for Fig. 2 (n.d., not detectable). B, cells were surface-biotinylated and incubated with or without 1 M PGE 2 . Then, cells were treated with glutathione to cleave biotin from the surface proteins. Biotinylated proteins present in the cell lysates were precipitated with Neutravidin, and the precipitates were analyzed by immunoblotting with HA or PS-1-NTF as described in the legend for Fig. 2. wild-type EP 4 receptor and EP 4 -t369 bands were not apparent after glutathione cleavage, whereas preincubation of cells with PGE 2 (1 M) prior to cleavage gave rise to the wild-type EP 4 receptor band but not the EP 4 -t369 band. This shows that the wild-type EP 4 receptor, but not EP 4 -t369, internalizes in a PGE 2dependent manner. The PS-1-NTF band was also not apparent following cleavage; preincubation with PGE 2 recovered the band in cells expressing wild-type EP 4 receptor but not in cells expressing EP 4 -t369 (Fig. 5B). Similar results were obtained with 10 nM PGE 2 in HEK293 cells (supplemental Fig. S9) and in SH-SY5Y cells without artificial overexpression of EP 4 receptor (supplemental Fig. S11B). We also examined the effect of sucrose and concanavalin A on a surface biotinylation assay for EP 4 receptor and PS-1-NTF. As shown in supplemental Fig. S5, preincubation of cells with PGE 2 prior to cleavage did not restore the band of either the EP 4 receptor or PS-1-NTF in the presence of sucrose or concanavalin A, confirming that these inhibitors suppressed the endocytosis. Similar results were observed for transfection of siRNA for clathrin or Rab5 but not for Rab7 (supplemental Fig. S6). These results suggest that internalization of PS-1 as a result of PGE 2 treatment is dependent on internalization of the EP 4 receptor.
For further confirmation of PGE 2dependent co-internalization of the EP 4 receptor and PS-1, we performed a co-immunostaining assay. Similar to what was observed for the EP 4 receptor, PS-1-NTF was localized to the cell surface in the absence of PGE 2 ; however, strong staining of PS-1-NTF in intracellular components was observed after PGE 2 treatment (Fig. 6A). As shown in Fig. 6A (see merged panel), localizations of the EP 4 receptor and PS-1-NTF were well matched, suggesting that PS-1-NTF co-internalizes with the EP 4 receptor.
To identify the intracellular components where both PS-1-NTF and the EP 4 receptor localize after PGE 2 treatment, we performed a co-immunostaining assay for these factors with Rab7, an LEL marker. As shown in Fig. 6, B and C, intracellular localizations of the EP 4 receptor (and PS-1-NTF) and Rab7 were well matched in PGE 2 -treated cells but not in control cells, suggesting that PS-1 is trafficked into LEL with the EP 4 receptor in a PGE 2 treatmentdependent manner. Results similar to those in Fig. 6 (in HEK293 cells with 1 M PGE 2 ) were obtained with 10 nM PGE 2 in HEK293 cells (supplemental Fig. S10) and in SH-SY5Y cells without artificial overexpression of EP 4 receptor (supplemental Fig. S12).
Finally, we tested the in vivo relevance of our in vitro results using transgenic mice expressing APPsw (APP23) that were crossed to EP 4 Ϫ/Ϫ mice (APPsw/EP 4 Ϫ/Ϫ mice). We had previously reported that the amount of A␤ in the brains of APPsw/ EP 4 Ϫ/Ϫ mice was lower than in APPsw/EP 4 ϩ/ϩ mice (30). As shown in Fig. 7A, the ␥-secretase activity in extracts prepared from the brains of APPsw/EP 4 Ϫ/Ϫ mice was very slightly, but significantly, lower than in those from APPsw/EP 4 ϩ/ϩ mice. This is consistent with the in vitro results, which showed that PGE 2 increases␥-secretaseactivityinextractsinanEP 4 receptordependent manner (supplemental Fig. 2F). There was no significant difference of the expression of PS-1-NTF between APPsw/EP 4 Ϫ/Ϫ and APPsw/EP 4 ϩ/ϩ mice (Fig. 7B), suggesting that activation rather than expression of ␥-secretase is important for the decrease in the amount of A␤ in the brains of APPsw/EP 4

Ϫ/Ϫ
. We previously reported that in HEK293 cells where the EP 4 but not the EP 2 receptor is functional, PGE 2 does not affect the expression and maturation of APP or ␣and ␤-secretase activities (30). In this study, we showed that the expression and maturation of APP (the ratio of the mature form of APP (mAPP) to the immature form of APP (imAPP)) and the ␣and ␤-secretase activities in extracts prepared from the brains of APPsw/EP 4 Ϫ/Ϫ mice were similar to those from APPsw/EP 4 ϩ/ϩ mice ( Fig. 7B and C). (CTFs of APP that are generated by ␣or ␤-secretase (CTF␣ or CTF␤, respectively) are used as an indirect index of secretase activity.) Furthermore, we compared the co-localization of PS-1-NTF with Rab7 in brain sections prepared from these mice. As shown in Fig. 7, D and E, yellows spots (index of the intracellular co-localization of PS-1-NTF with Rab7) were more apparent in brain sections prepared from APPsw/EP 4 ϩ/ϩ mice than in sections from APPsw/EP 4 Ϫ/Ϫ mice, suggesting that even in vivo PS-1 is trafficked to LEL in an EP 4 receptor-dependent manner. The results shown in Fig. 7 suggest that our in vitro results are relevant in vivo and that PGE 2 -dependent traffic of PS-1 into LEL contributes to the observed increase in ␥-secretase activity and resulting stimulation of A␤ production by expression of the EP 4 receptor in vivo.

DISCUSSION
We recently reported the importance of PGE 2 as a factor that connects inflammation and AD; PGE 2 stimulates production of A␤ and this stimulation is mediated by EP 2 and EP 4 receptors. We also suggested the importance of EP 2 and EP 4 receptors in production of A␤ in vivo by showing that the amount of A␤ in the brains of APPsw/EP 2 Ϫ/Ϫ and APPsw/EP 4 Ϫ/Ϫ mice was lower than in the respective control mice. Based on these results, we proposed that antagonists for EP 2 and/or EP 4 receptor would be therapeutically beneficial for AD (30).
To determine the potential drug target (whether it is the EP 2 receptor, the EP 4 receptor, or both), it is important to understand the molecular mechanism for intracellular signal transduction governing EP 2 (or EP 4 ) receptor-mediated stimulation of A␤ production by PGE 2 . Because activation of both EP 2 and EP 4 receptors is coupled to activation of the cAMP-PKA pathway, we speculated that this pathway might be involved in the signal transduction. However, our attempts to prove this conjecture failed in our previous study (30). Furthermore, our finding that both EP 2 knock-out mice and EP 4 receptor knock-out mice showed decreased levels of A␤ in the brain could not be explained by the hypothesis that the cAMP-PKA pathway is responsible for both EP 2 receptor-and EP 4 receptor-mediated signal transduction pathways for PGE 2 stimulated A␤ production. In the current study, using inhibitors for adenylate cyclase and PKA, we showed that the cAMP-PKA pathway is involved in EP 2 -mediated but not EP 4mediated stimulation of A␤ production by PGE 2 . As the EP 4 receptor is not linked to the cAMP-PKA pathway for PGE 2stimulated A␤ production, EP 4 receptor antagonists may be therapeutically beneficial for the treatment of AD. It has been reported that the cAMP-PKA pathway is important for longterm potentiation (51, 52) so EP 2 receptor, but not EP 4 receptor, antagonists may have side effects on the memory system by inhibiting long-term potentiation. On the other hand, deletion of EP 2 receptor in mice was shown to decrease oxidative damage and inhibit ␤-secretase, resulting in a decrease in the level of A␤ in brain (29), or to enhance A␤ phagocytosis in the brain (53). It has also been suggested that PGE 2 -dependent activation of cAMP-PKA pathway causes induction of expression of APP (54,55), indicating that the EP 2 receptor antagonists may be therapeutically beneficial for the treatment of AD. The fact that EP 2 and EP 4 receptors stimulate the production of A␤ through different mechanisms could explain why single knock-outs of either the EP 2 or EP 4 receptor reduce A␤ production in vivo.
Much attention has been paid to agonist-dependent internalization (endocytosis) of receptors, including the EP 4 receptor, because this causes receptor desensitization. Furthermore, this internalization was also recently reported to be important for signal transduction (44). From the current study, we propose that the EP 4 receptor mediates the PGE 2 signal through its cointernalization with PS-1 (␥-secretase). This suggestion is based on the following results: The immunoprecipitation assay The brain sections were prepared from the same mice and subjected to immunostaining as described in the legend for Fig. 6 (D). The ratio of cells with yellow spots (positive cells) to total cells in the brain sections (three sections/brain) was determined. Values are given as means Ϯ S.D. (n ϭ 8). **, p Ͻ 0.01 (E). revealed a physical interaction between the EP 4 receptor and PS-1-NTF; PGE 2 -stimulated production of A␤ was not observed under conditions in which co-internalization of the EP 4 receptor and PS-1 is inhibited (such as in the presence of endocytosis inhibitors, in cells transfected with siRNA for clathrin, and in cells expressing EP 4 -t369); internalization of PS-1-NTF was observed to be dependent on both PGE 2 and expression of the wild-type EP 4 receptor in the surface biotinylation assay. This is the first demonstration of the EP 4 receptor mediating signal transduction through co-internalization with other molecules. Similar mechanisms may be involved in EP 4 receptor-mediated signal transduction for other responses.
We have also concluded that PS-1 (␥-secretase) is trafficked to LEL in a PGE 2 -and EP 4 receptor-dependent manner and that this traffic is important for PGE 2 -stimulated production of A␤ and ␥-secretase activity. This conclusion is based on the following results: PGE 2 -dependent co-localization of PS-1-NTF, EP 4 , and Rab7 was observed in vitro, and co-localization of PS-1-NTFand Rab7 was observed in APPsw/EP 4 ϩ/ϩ mice but not as distinctly in APPsw/EP 4 Ϫ/Ϫ mice; and transfection of siRNA for Rab5 or Rab7 inhibited PGE 2 -stimulated production of A␤ and ␥-secretase activity. A similar mechanism was proposed for ␤-adrenergic receptor-mediated stimulation of production of A␤ and ␥-secretase activity (49). They also showed the enhancement of ␥-secretase activity and elevation of A␤ production in endosomes. It is well known that the pH in endosomes is relatively low and that as a result ␥-secretase is more active in endosomes (56 -58). Thus, the relatively low pH value in endosomes may contribute to EP 4 receptor-mediated stimulation of A␤ production. However, we found that the ␥-secretase activity in extracts decreased when the extracts were prepared from cells cultured under conditions in which the traffic of PS-1 (␥-secretase) into endosomes is inhibited, even though the ␥-secretase assay was carried out at the same pH. Thus, something other than the lower pH of endosomes (such as induction of expression and post-translational modification of ␥-secretase) may also contribute to the stimulation of A␤ production by internalization of PS-1 (␥-secretase) into endosomes. It is also possible that internalization of the EP 4 receptor stimulates the traffic of ␤-secretase to endosomes, resulting in its activation, as is the case for apolipoprotein receptor-2 (59), because ␤-secretase is also more active at lower pH (60). However, we reported previously that the activity of ␤-secretase (estimated from the amounts of CTF␤) is not enhanced by PGE 2 (30). Thus, the EP 4 receptor may not interact with ␤-secretase.
In summary, we have shown that EP 2 and EP 4 receptors mediate stimulation of production of A␤ by PGE 2 through distinct mechanisms. This finding is important for understanding the mechanisms underlying the inflammation-mediated progression of AD and the regulation of vesicle transport of PS-1 (␥-secretase) and for identifying potential targets for the development of drugs to treat AD.