GSK3β Activity Modifies the Localization and Function of Presenilin 1*

Presenilin 1, a causative gene product of familial Alzheimer disease, has been reported to be localized mainly in the endoplasmic reticulum and Golgi membranes. However, endogenous Presenilin 1 also localizes at the plasma membrane as a biologically active molecule. Presenilin 1 interacts with N-cadherin/β-catenin to form a trimeric complex at the synaptic site through its loop domain, whose serine residues (serine 353 and 357) can be phosphorylated by glycogen synthase kinase 3β. Here, we demonstrate that cell-surface expression of Presenilin 1/γ-secretase is enhanced by N-cadherin-based cell-cell contact. Physical interaction between Presenilin 1 and N-cadherin/β-catenin plays an important role in this process. Glycogen synthase kinase 3β-mediated phosphorylation of Presenilin 1 reduces its binding to N-cadherin, thereby down-regulating its cell-surface expression. Moreover, reduction of the Presenilin 1·N-cadherin·β-catenin complex formation leads to an impaired activation of contact-mediated phosphatidylinositol 3-kinase/Akt cell survival signaling. Furthermore, phosphorylation of Presenilin 1 hinders ϵ-cleavage of N-cadherin, whereas ϵ-cleavage of APP remained unchanged. This is the first report that clarifies the regulatory mechanism of Presenilin 1/γ-secretase with respect to its subcellular distribution and its differential substrate cleavage. Because the cleavage of various membrane proteins by Presenilin 1/γ-cleavage is involved in cellular signaling, glycogen synthase kinase 3β-mediated phosphorylation of Presenilin 1 should be deeply associated with signaling functions. Our findings indicate that the abnormal activation of glycogen synthase kinase 3β can reduce neuronal viability and synaptic plasticity via modulating Presenilin 1/N-cadherin/β-catenin interaction and thus have important implications in the pathophysiology of Alzheimer disease.

Pathological features of Alzheimer disease (AD) 3 are characterized by neurofibrillary tangles and amyloid plaques. Amyloid plaques are composed of amyloid ␤ peptide (A␤), which is derived from the sequential cleavages of amyloid precursor protein (APP), whereas neurofibrillary tangles are characterized by the accumulation of hyperphosphorylated Tau. A␤ peptides are yielded by the intramembranous cleavage of the APP C-terminal fragment by Presenilin 1 (PS1)/␥-secretase (1). Interestingly, PS1⅐␥-secretase complex is also involved in the ⑀-cleavage of various other membrane proteins (2)(3)(4)(5).
On the other hand, neurofibrillary tangle is composed of phosphorylated Tau. Glycogen synthase kinase 3␤ (GSK3␤) plays an important role in Tau phosphorylation (6). GSK3␤ is expressed in a variety of tissues with the highest level in the brain, where it localizes especially in neurons (7). Recent study showed that PS1 is an unprimed substrate of GSK3␤ (8) and serine residues in the PS1 loop domain can be phosphorylated (9). Abnormal increases in GSK3␤ level and activity have been associated with AD pathology (6,10), suggesting that GSK3␤mediated phosphorylation of PS1 might also be involved in AD pathophysiology.
Classically, PS1 has been reported to be localized mainly in the endoplasmic reticulum and Golgi membranes (11).
Recently, it is also demonstrated that endogenous PS1 localizes at the plasma membrane as an active molecule (12). Actually, in neurons, PS1 binds to ␤-catenin and N-(and E-) cadherin through a large hydrophilic loop region (13) at the synapse. Moreover, N-cadherin is cleaved by PS1/␥-secretase in response to N-methyl-D-aspartic acid-type receptor stimulation (14,15). N-cadherin is essential for synaptic contact (16) and regulates synaptogenesis and dendritic spine morphology (17,18). In addition, N-cadherin-based cell-cell adhesion activates PI3K/Akt cell survival signaling by recruiting PI3K to the N-cadherin adhesion complex (19). Actually, PS1 facilitates this process by promoting cadherin/PI3K association in ␥-secretase activity-independent manner (20). Collectively, PS1/N-cadherin interaction at the synapse seems to be neuroprotective by facilitating PI3K/Akt survival signaling. However, the cellular mechanism that determines PS1/N-cadherin binding has never been elucidated so far.
Here, we demonstrate that cell-surface distribution of PS1 is enhanced in the presence of N-cadherin-based cell-cell contact.
The physical interaction between PS1 and N-cadherin/␤-catenin plays an important role in cell-surface expression of PS1/␥secretase. Moreover, GSK3␤-mediated phosphorylation of PS1 reduces its binding to N-cadherin, thereby down-regulating its cell-surface expression. Reduced PS1/N-cadherin/␤-catenin interaction leads to the inhibition of contact-mediated PI3K/ Akt cell survival signal activation as well as the differential regulation of substrate cleavage by PS1/␥-secretase. These results demonstrated that the redistribution of PS1 to the plasma membrane is regulated by its phosphorylation and that the distributional change modifies the PI3K/Akt signal as well as substrate cleavage.
CHO cells, stably expressing Swedish (K670N/M671L) mutant human APP695 (APPSw-CHO cells) were obtained as follows: the entire coding sequence of human APP695 was subcloned in a mammalian expression vector pME/sf. Swedish mutations (K670M/N770L) (24) were introduced by site-directed mutagenesis. CHO cells were transfected with the APP cDNA together with pSVbsr plasmid and the cells were selected for resistance against blasticidin (10 g/ml). Cells were maintained in Dulbecco's modified Eagle's medium/ F-12 (Invitrogen) supplemented with 10% fetal bovine serum. CHO cells stably expressing both Swedish mutant APP and human N-cadherin (APPsw/Ncad-CHO cells) were obtained as follows: APPsw-CHO cells were transfected with N-cadherin/pcDNA3.1(ϩ). Cells were selected by 800 g/ml G418 and the establishment of stably transfected clones was verified by Western blot.
Cell Treatment by Reagents-For the induction of S9AGSK3␤, S9A-tet cells were treated with 1 g/ml tetracycline containing medium. siRNA constructs were transfected into HEK293 cells using Lipofectamine 2000, according to the manufacturer's instruction. Cells were analyzed 24 h after siRNA transfection. For calcium switch experiments, confluent MEF PS Ϫ/Ϫ cells were incubated in serum-free medium for 4 h, treated with 4 mM EGTA for 40 min and then switched to serum-free, calcium-containing medium for the times shown (20). For stimulation of N-cadherin cleavage, MEF PS Ϫ/Ϫ cells were treated by 10 M ionomycin dissolved in Opti-MEM for 30 min.
Western Blot and Immunoprecipitation-Preparation of protein samples, the Western blot, and immunoprecipitation analysis were carried out as described elsewhere (25). For some experiments, cells were fractionated as previously described (25).
Biotinylation of Cell-surface Proteins-Confluent cells grown in 10-cm dish were washed three times with ice-cold phosphate-buffered saline, and suspended in the solution containing 0.5 mg of Sulfo-NHS-LC-Biotin (Pierce)/ml of phosphate-buffered saline for 30 min. Cells were then washed three times with phosphate-buffered saline and biotinylated proteins were precipitated by 30 l of streptoavidin-agarose (Invitrogen) from equal amounts of cell lysates. Precipitated biotinylated proteins were then subjected to Western blot analysis.
Immunostaining-The samples for immunostaining were prepared as described elsewhere (25). Samples were examined using a laser scanning confocal microscopy, LSM 510 META (Zeiss), or a fluorescence microscopy, Axiovert 200 (Zeiss).

Cell-surface Expression of PS1 Is
Enhanced by the Cadherin-mediated Cell-Cell Contact-To examine the relationship between the cadherin-based cell-cell contact and subcellular distribution of PS1, we established CHO cell lines stably expressing both Swedish mutations of human APP695 and human N-cadherin (APPsw/Ncad-CHO cells). CHO cells are suitable for the analysis of exogenously introduced N-cadherin, because it barely expresses endogenous cadherin species. APPsw/Ncad-CHO cells were compared with CHO cells stably expressing Swedish mutation of APP695 only (APPSw-CHO cells) by immunostaining (Fig. 1). In APPsw/Ncad-CHO cells, N-cadherin immunoreactivity was prominent at the sites of cell contact (Fig. 1A). PS1 was also seen at the cell-cell contact sites (Fig. 1B), where it co-localized with N-cadherin (Fig. 1C). The outline of APPsw/Ncad-CHO cells was clearly visualized by PS1 staining (Fig. 1, B and C, compared with the perinuclear PS1 staining in Fig. 1E), indicating that N-cadherin expression redistributed the PS1 subcellular localization to the plasma membranes. Next, we compared PS1 and ␤-catenin distribution in the presence or absence of N-cadherin. In APPSw-CHO cells, ␤-catenin and PS1 immunoreactivity were mainly seen at the perinuclear area ( Fig. 1, D-F). Stable expression of N-cadherin recruited both ␤-catenin ( Fig.  1G) and PS1 (Fig. 1H) to cell-cell contact sites (compare with the perinuclear PS1 staining in Fig. 1E), showing co-localization of both proteins (Fig. 1I). Thus, stable expression of N-cadherin led to the redistribution of PS1 to the plasma membrane, especially to cell-cell contact sites, where N-cadherin, PS1, and ␤-catenin all colocalized.
To confirm that the cadherin-based adhesion promotes cellsurface expression of PS1/␥-secretase, we transiently expressed human N-cadherin into CHO cells and examined the level of cell-surface PS1/␥-secretase by the biotinylation assay (Fig. 1J). Transient expression of N-cadherin enhanced both cell-surface expression levels of nicastrin and PS1, whereas the total levels of these proteins were comparable, demonstrating that cadherin-based adhesion enhances the cell-surface expression level of PS1/␥-secretase (Fig. 1J). We further verified this finding by RNA interference (Fig. 1K) using HEK293 cells, which endogenously express human N-cadherin, ␤-catenin, and PS1. 24 h after N-cadherin knockdown, cell-surface expression levels of both PS1 and nicastrin were reduced, whereas the total levels of these proteins remained unchanged (Fig. 1K, fourth lane), compared with control (Fig. 1K, second lane). The protein level of ␤-catenin was reduced in the background of N-cadherin knockdown, indicating that ␤-catenin lost its stability in the absence of N-cadherin (Fig. 1K, fourth lane). We also tested the effect of ␤-catenin knockdown on cell-surface expression of ␥-secretase components. Interestingly, 24 h after ␤-catenin knockdown, cell-surface nicastrin was reduced, without changing the cellular level of N-cadherin (Fig. 1K, third lane), although ␤-catenin knockdown had less impact on cell-surface PS1 distribution compared with N-cadherin knockdown (compare Fig. 1K, third and fourth lanes). Collectively, these experiments indicate that cadherin-based cell-cell adhesion promotes the expression of PS1/␥-secretase at the cell surface.
The above experiments suggested that activation of GSK3␤ would reduce the PS1/N-cadherin interaction via PS1 phosphorylation in the loop domain. Accordingly, the immunoprecipitation assay revealed that transient transfection of constitutively active GSK3␤ (S9A mutant) (22) into HEK293 cells reduced PS1/N-cadherin interaction (Fig. 2D). We further wished to confirm the effect of GSK3␤ activation in neuronal cells. We established human neuroblastoma SH-SY5Y cell lines (S9A-tet cells), in which expression of constitutively active GSK3␤ (S9AGSK3␤) can be induced by tetracycline treatment (tet-on). Tetracycline (1 g/ml) induced S9AGSK3␤ expression 24 h after treatment (Fig. 2E, left). Phosphorylation of ␤-catenin, a representative target of GSK3␤, was increased, whereas the total level of ␤-catenin was decreased (Fig. 2E,  left), indicating that expression of S9AGSK3␤ enhanced its activity as a phosphokinase, leading to phosphorylation and degradation of ␤-catenin. Using these cell lines, we then analyzed the phosphorylation status of PS1 CTF by immunoprecipitation assay. Lysates of S9A-tet cells with or without tetracycline treatment for 24 h were immunoprecipitated by anti-phosphoserine antibody, followed by immunoblotting using anti-PS1 CTF. PS1 CTF, pulled-down by anti-phosphoserine was increased after tet-on, indicating that PS1 phosphorylation by GSK3␤ was enhanced after tetracycline treatment (Fig. 2E, right). Then we examined PS1/N-cadherin interaction, comparing before and 24 h after tet-on. Tetracycline treatment inhibited the PS1/N-cadherin interaction dramatically (Fig. 2F). Collectively, these findings demonstrated that GSK3␤ activity regulates binding of PS1 to N-cadherin.
GSK3␤ Activation Down-regulates PI3K/Akt Cell Survival Signaling-It has been reported that N-cadherin-based adhesion initiates PI3K-dependent activation of Akt, thereby up-  (9). B, wtPS1, deletion mutants (PS1⌬340 -375 or ⌬340 -350) or pseudo-phosphorylation mutants (S353D or S357D PS1) were transfected into HEK293 cells. 24 h after transfection, cells were lysed and lysates immunoprecipitated with anti-PS1 NTF antibody, followed by the Western blot analysis. Control precipitation was done with normal rabbit IgG (first lane). wtPS1 robustly associated with both N-cadherin and ␤-catenin (third lane). None of the deletion mutants significantly associated with N-cadherin or ␤-catenin (fourth and fifth lanes). The association between pseudo-phosphorylation mutants (S353D PS1, sixth lane and S357D PS1, seventh lane) and N-cadherin/␤-catenin was reduced, compared with wtPS1 (third lane), whereas the expression level of each PS1⅐N-cadherin⅐␤-catenin complex component was similar in the total cell lysate (Lys). The bottom is the loading control, represented by the ␤-actin bands. C, either wtPS1 or one of thepseudo-phosphorylation mutants (S353D or S357D PS1) was transfected into MEF PS Ϫ/Ϫ cells, followed by immunoprecipitation, using anti-PS1 NTF antibody and Western blot analysis. Control precipitation was done with normal rabbit IgG (first lane). wtPS1 bound both N-cadherin and ␤-catenin (second lane). The association between pseudo-phosphorylation mutants (S353D PS1, third lane, and S357D PS1, fourth lane) and N-cadherin/␤-catenin was reduced, compared with wtPS1. The expression level of each PS1/N-cadherin⅐␤-catenin complex component was similar in the total cell lysate (Lys). The bottom is the loading control, represented by the ␤-actin bands. D, either control GFP (second lane) or constitutively active (S9A) GSK3␤ (third lane) was transfected into HEK293 cells, followed by immunoprecipitation, using anti-PS1 NTF antibody. Control precipitation was done with normal rabbit IgG (first lane). Note that PS1-bound N-cadherin is reduced under the expression of S9AGSK3␤ (third lane), whereas the total level of N-cadherin did not change significantly (Lys). The bottom is the loading control, represented by the ␤-actin bands. E, left, S9A-tet cells were treated with 1 g/ml tetracycline for 24 h. Cells were lysed after treatment and analyzed by anti-phospho-␤-catenin (Ser 33 /Ser 37 /Thr 41 ) and anti-total ␤-catenin antibodies. The magnitude of GSK3␤ expression was demonstrated by anti-GSK3␤ antibody. The induction of GSK3␤ expression was evident (GSK3␤). Phosphorylation of ␤-catenin was enhanced (phospho-␤-catenin), whereas total ␤-catenin levels were reduced (Total ␤-catenin) 24 h after treatment. The bottom is the loading control, represented by the ␤-actin bands. Right, S9A-tet cells were treated by 1 g/ml tetracycline for 24 h. Cells were lysed after treatment and immunoprecipitated by anti-phosphoserine antibody, followed by the Western blot, using anti-PS1 CTF. The PS1 CTF immunoprecipitated by anti-phosphoserine was increased after tetracycline treatment (Top, PS1 CTF), whereas the total levels of PS1 CTF were comparable between before and after treatment (Lys, PS1 CTF). The bottom is the loading control, represented by the ␤-actin bands. F, S9A-tet cells were treated by 1 g/ml tetracycline for 24 h. Cells were lysed after treatment and the lysates were immunoprecipitated by anti-PS1 NTF antibody. Control precipitation was done with normal rabbit IgG (first lane). PS1/N-cadherin interaction was reduced after tet-on (third lane), compared with before tet-on (second lane). The expression levels of N-cadherin were demonstrated in the total cell lysates (Lys). The bottom is the loading control, represented by the ␤-actin bands.
Next, we wished to determine whether this PS1-mediated activation of PI3K/Akt signaling is affected by the strength of the cadherin-based adhesion. To test this, a calcium switch assay was performed using MEF PS Ϫ/Ϫ cells, because cadherins are known to be calcium-dependent cell-cell adhesion molecules. In the assay, MEF PS Ϫ/Ϫ cells were cultured in the presence of 4 mM EGTA for 40 min for calcium deprivation, after which cells were cultured in the serum-free, calcium containing medium (Fig. 4B, top). Immunostaining results before calcium deprivation showed that N-cadherin immunoreactivity was present along the outline of the cell-cell contact sites (Fig. 4B,  bottom, left). After EGTA treatment, N-cadherin immunoreactivity was observed in a granular pattern, indicating disruption of the cadherin-based cell-cell contacts (Fig. 4B, bottom, middle). Calcium supplement restored the liner N-cadherin immunoreactivity at the junction (Fig. 4B, bottom, right). We utilized the calcium switch assay to examine whether PS1/N-cadherin interaction is necessary for the N-cadherin-mediated transmission of PI3K/Akt survival signaling. Either wtPS1 or PS1⌬340 -350 was transiently transfected into MEF PS Ϫ/Ϫ cells, then followed by the calcium switch assay (Fig. 4C). In the presence of wtPS1, Akt phosphorylation before calcium deprivation was prominent (Fig. 4C, wtPS1, Pre), which was diminished after calcium deprivation (Fig. 4C, wtPS1, time 0) and gradually recovered after calcium supplement (Fig. 4C, wtPS1, 30, and  90). Conversely, after PS1⌬340 -350 transfection, Akt phosphorylation before calcium deprivation was not prominently compared with wtPS1 and remained unchanged throughout the assay (Fig. 4C, PS1⌬340 -350), indicating that the PS1/Ncadherin interaction is necessary for the contact-mediated transmission of PI3K/Akt survival signaling. Next, we examined the effect of GSK3␤ activation on transmission of PI3K/ Akt signaling (Fig. 4D). In the absence of PS1 (Fig. 4D, GFP) or after co-transfection of wtPS1 and S9AGSK3␤ (Fig. 4D,  wtPS1ϩS9AGSK3␤), the phosphorylation state of Akt did not change significantly before and throughout the calcium switch assay, which was in contrast to wtPS1 transfection (Fig. 4D,  wtPS1). Because GSK3␤ affects many molecules, we cannot conclude that the effect of S9A GSK3␤ transfection was solely mediated by PS1 phosphorylation, however, the above data supports the idea that PS1/N-cadherin interaction plays an important role in the transmission of "contact-mediated" PI3K/ Akt signal activation and that GSK3␤-mediated phosphorylation of PS1 may inhibit this process by reducing PS1/N-cadherin interaction.
GSK3␤-mediated Phosphorylation of PS1 Differentially Regulates N-cadherin and APP Cleavage-Because GSK3␤mediated phosphorylation of PS1 changes its subcellular distribution and binding to N-cadherin, we assumed that phosphorylation might also change the substrate specificity of ␥-secretase. To test this, PS1 constructs (wtPS1, pseudo-phosphorylation mutant (S353D or S357D PS1) or PS1⌬340 -350) were transfected into MEF PS Ϫ/Ϫ cells (Fig. 5A). As reported previously (15), N-cadherin is sequentially cleaved by ADAM10  , top), whereas the expression in the endoplasmic reticulum fraction (bottom) was comparable. B, either wtPS1 or one of the pseudo-phosphorylation mutants (S353D or S357D PS1) was transfected into MEF PS Ϫ/Ϫ cells. Control cells were transfected with GFP (first lane). 24 h after transfection, cell-surface proteins were biotinylated, precipitated by streptoavidin-agarose, and analyzed by Western blot. The cell-surface expression of PS1 as well as that of nicastrin (third and fourth lanes) was reduced in the cells transfected with pseudo-phosphorylation mutants, compared with that of wtPS1 (second lane), whereas the expression levels of these proteins in the cell lysate were comparable (Total Nicastrin, Total PS1 NTF). The asterisk designates the glycosylated "mature" Nicastrin. The bottom is the loading control, represented by the ␤-actin bands. and PS1/␥-secretase. Ectodomain shedding by ADAM10 generates a membranous fragment Ncad/CTF1, which is further cleaved by PS1/␥-secretase to produce the cytoplasmic fragment Ncad/CTF2. Immunoblotting by anti-N-cadherin C terminus antibody revealed that the amount of Ncad/CTF1 is reduced after wtPS1 transfection (Fig. 5A, second lane), compared with control GFP transfection (Fig. 5A, first lane), reflecting degradation of Ncad/CTF1 by PS1/␥-secretase. On the contrary, neither pseudo-phosphorylation mutants (S353D and S357D) nor PS1⌬340 -350 transfection effectively reduced Ncad/CTF1 (Fig. 5A, third to fifth lanes), suggesting that phosphorylation of PS1 inhibits N-cadherin cleavage after ectodomain shedding. We then stimulated ectodomain shedding of N-cadherin in these transfected cells by ionomycin treatment (5). Cells were fractionated after ionomycin treatment for 30 min and both membrane and cytosolic fractions were subjected to Western blot assay, using anti-N-cadherin C terminus antibody (Fig. 5B). In the membrane fraction, both full-length N-cadherin and Ncad/CTF1 were observed. The amount of Ncad/CTF1 was reduced in the membrane fraction of cells transfected with wtPS1, compared with other cell lines (Fig. 5B, second  lane). Ncad/CTF2 production was observed only in the cytoplasmic fraction of cells transfected with wtPS1 (Fig. 5B, second lane), indicating that cleavage of Ncad/CTF1 was impaired in the absence of PS1 (Fig. 5B, first lane) or in the presence of pseudo-phosphorylation mutants (Fig. 5B, third and fourth  lane). PS1⌬340 -350 also failed to produce Ncad/CTF2, indicating that PS1-N-cadherin binding is important for cleavage of Ncad/ CTF1 (Fig. 5B, fifth lane).
This led us to ask whether this inhibitory effect of PS1 phosphorylation affects the ␥-secretase activity for other substrates. To test this effect on APP cleavage, we introduced human APP into MEF PS Ϫ/Ϫ cells together with either wtPS1 or PS1 mutant (Fig.  5C), followed by Western blot analysis, using the anti-APP C terminus antibody. Interestingly, all PS1 constructs equally reduced APP CTF␣ and -␤ production (Fig. 5C, second to fifth lanes) compared with control GFP (Fig. 5C, first lane), indicating that the phosphorylation does not significantly affect APP cleavage by ␥-secretase. These results demonstrate that GSK3␤-mediated phosphorylation of PS1 differentially affects N-cadherin and APP cleavage by modulating substrate-enzyme binding and subcellular distribution of ␥-secretase.

DISCUSSION
PS1 has been reported to be localized mainly in the endoplasmic reticulum and Golgi membranes (11), nevertheless, it has also been demonstrated that endogenous PS1 localizes at the plasma membrane as an active molecule (12). The cell surface localization of PS1 is consistent with the observation that it can process many adhesion and receptor molecules (27,28). However, how these differential distributions of PS1/␥-secretase are regulated has never been elucidated.
In this report, we demonstrated that cell-surface expression and its functions of PS1/␥-secretase are regulated by PS1/N-cadherin/␤-catenin interaction (Fig. 3). N-cadherin is an essential adhesion molecule for synaptic contact (16), indicating that expression of PS1/␥-secretase at the synaptic membrane is also regulated by cadherin-based synaptic contact. Importantly, N-cadherin is cleaved by PS1/␥-secretase to disrupt both synaptic contact and PS1⅐N-cadherin⅐␤-catenin complex in response to N-methyl-D-aspartic acid-type FIGURE 4. GSK3␤ activation down-regulates PI3K/Akt cell survival signaling. A, the ability of pseudo-phosphorylation mutants to facilitate PI3K/Akt signaling was tested by introducing wtPS1 or one of pseudo-phosphorylation mutants (S353D, S357D, or S353D/S357D double mutant PS1) for 24 h, followed by immunoblotting (bottom). Transfection of wtPS1 enhanced Akt phosphorylation (second lane). All the pseudo-phosphorylation mutants had reduced activity to enhance Akt phosphorylation (third to fifth lanes), compared with wtPS1 (second lane). B, time course of the calcium switch assay (top). MEF PS Ϫ/Ϫ cells were treated with 4 mM EGTA for 40 min for calcium deprivation. After calcium deprivation, cells were cultured in calcium containing serum-free medium. The effect of the calcium switch assay on N-cadherin-based cell-cell contact is shown (bottom). Before EGTA treatment, N-cadherin in MEF PS Ϫ/Ϫ was seen at the sites of cell contacts as liner immunoreactivity (bottom, left, arrows). After EGTA treatment, N-cadherin concentration at the junction became weak and showed granular immunoreactivity (bottom, middle, arrowheads). 60 min after calcium supplement, N-cadherin concentration at the cell-cell contact sites was restored and appeared as liner structures (bottom, right, arrows). C, either wtPS1 or PS1⌬340 -350 was transfected into MEF PS Ϫ/Ϫ cells. 24 h after transfection, cells were subjected to the calcium switch assay. Under wtPS1 transfection, Akt phosphorylation before calcium deprivation was prominent (wtPS1, Pre), which was diminished after EGTA treatment (wtPS1, 0) and gradually recovered after calcium supplement (wtPS1, 30, 90). Conversely, under PS1⌬340 -350 transfection, Akt phosphorylation before calcium deprivation was not prominent and remained unchanged throughout the assay (PS1⌬340 -350). D, wtPS1 was transfected into MEF PS Ϫ/Ϫ cells in the presence or absence of S9A GSK3␤. Control cells were transfected with GFP. 24 h after transfection, cells were subjected to the calcium switch assay. In the absence of PS1 (GFP) or after co-transfection of wtPS1 and S9AGSK3␤ (wtPS1ϩS9AGSK3␤), the phosphorylation state of Akt did not change significantly before and throughout the calcium switch assay. Conversely, wtPS1 transfection led to prominent Akt phosphorylation before the calcium switch assay (wtPS1, Pre), which is diminished after EGTA treatment (wtPS1, 0) and gradually recovered after calcium supplement (wtPS1, 30, 60).
receptor stimulation (14,15). Thus, the subcellular distribution of PS1/␥-secretase in neurons would be altered dynamically in the course of synaptic remodeling, which would also change processing of various adhesion and receptor molecules.
In the present report, we showed that the increase of GSK3␤ activity reduces PS1⅐N-cadherin⅐␤-catenin complex formation possibly via direct phosphorylation of the PS1 loop domain (Fig. 2). Because activation of GSK3␤ also enhances phosphorylation and degradation of ␤-catenin (Fig. 2E, left), reduction of the ␤-catenin level may also be involved in the process. In accordance with the finding that PS1⅐N-cadherin⅐␤-catenin complex formation enhances cell-surface expression of PS1/␥-secretase (Fig.  3A), the pseudo-phosphorylated form of PS1, which has less ability to form the PS1⅐N-cadherin⅐␤-catenin complex, is reduced from the cell surface (Fig. 3B). To our knowledge, this is the first report that demonstrates the cellular mechanism that modulates the subcellular distribution of PS1/␥-secretase.
Another important finding in this study is that GSK3␤ activation modulates PS1 functions via two independent ways (i) down-regulates cadherin-mediated activation of PI3K/Akt signaling (Fig. 4), and (ii) regulates its cleavage functions (Fig. 5). A previous report (19) demonstrated that N-cadherin-based contact recruits PI3K to the cellsurface N-cadherin complex, activating PI3K/Akt cell survival signaling. PS1 plays an important role in this process by facilitating PI3K/Ncadherin binding and finally leading to activation of PI3K/Akt signaling and the inhibition of GSK3␤ (20) (illustrated in Fig. 6, left). According to our present findings, abnormal activation of GSK3␤ leads to phosphorylation of PS1 in the loop domain, reducing its binding to N-cadherin. A decrease of N-cadherin/PS1 interaction leads to inactivation of PI3K/Akt signaling. At the same time, because PI3K/Akt signaling is an important inhibitory mechanism of GSK3␤ (20), inactivation of this signal could lead to further activation of GSK3␤, constituting a vicious circle (Fig. 6, right). Collectively, PS1 may act as a molecular switch that links cell-cell adhesion to the cell survival signal. GSK3␤-mediated phosphorylation of PS1 would separate PS1 from N-cadherin, thereby "switching off" the linkage between cell-cell contact and survival signal.
In addition, we have shown that GSK3␤-mediated phosphorylation of PS1 differentially regulates N-cadherin and APP cleavage by ␥-secretase (Fig. 5). PS1⅐␥-secretase complex is also involved in ⑀-cleavage of various membrane proteins (2)(3)(4)(5), however, an important, but an unanswered question is that how PS1/␥-secretase activity is modulated in terms of substrate specificity. Recently, TMP-2 was identified to be a modulator of ␥-site (but not ⑀-site) cleavage, negatively regulating A␤ production (29). As shown in the present study, ⑀-cleavage of N-cadherin is down-regulated by phosphorylation of PS1, which is in contrast to APP cleavage. Our data clearly demonstrated the cellular mechanism involved in substrate specificity of ⑀-cleavage by PS1/␥-secretase. Whether ⑀-cleavage of other substrates is affected by the phosphorylation should be investigated in the future.
Recently, a strain of PS1 knock-in mice in which most of the hydrophilic loop sequence was deleted from the endogenous PS1 gene (thus, cannot be associated with N-cadherin/␤-catenin) was created (30). Surprisingly, the homozygous mice exhibit drastically reduced ␥-secretase cleavage at the A␤40, but not the A␤42, site. In addition, it was reported that inhibition of GSK3␣ blocked the production of A␤ peptides by interfering with APP cleavage at the ␥-secretase step, but did not inhibit Notch processing (31). Thus, although GSK3␤-mediated PS1 phosphorylation seems to have less impact on ⑀-cleavage of APP compared with N-cadherin cleavage, whether the phosphorylation could affect ␥-cleavage of APP is an important question to be answered.
AD begins with an impairment of memory, which is caused by a disturbance of hippocampal synaptic function (32). In addition, cognitive decline in AD is correlated to the degree of synaptic loss (33), suggesting that the pathological alteration in the metabolism of synaptic protein is primarily involved in AD pathophysiology. It is known that abnormal increases in GSK3␤ level and activity have been associated with AD pathophysiology (6,10). Taken together, our studies propose a causal link that may connect abnormal activation of GSK3␤ to the synaptic dysfunction in the following two ways. 1) Inhibition of cadherin-mediated PI3K/Akt signal transmission, thereby downregulating cell-survival signaling leading to neurodegeneration. Because synaptic plasticity should involve the process in which certain synapses survive and some others degenerate, dysregulation of this "contact-mediated survival signal" should hinder synaptic plasticity as well. 2) Inhibition of N-cadherin ⑀-cleavage, thereby reducing the production of Ncad/CTF2. Because Ncad/CTF2 carries various signals transmitted from the cellsurface to the nucleus (14,34), inhibition of Ncad/CTF2 production under abnormal activation of GSK3␤ should have negative impact on neuronal plasticity or its viability. Interestingly, FAD-linked mutations of PS1 have been shown to inhibit ⑀-cleavage of N-cadherin (14,34). Thus, from the viewpoint of Ncad/CTF2 production, FAD-linked mutation and phosphorylation of PS1 act in a similar way. According to the previous report, conditional transgenic mice overexpressing GSK3␤ in the adult brain show decreased nuclear ␤-catenin, abnormally phosphorylated Tau and clear evidence of neurodegeneration (35), indicating the possibility that altered N-cadherin metabolism or PS1/N-cadherin interaction could lead to neurodegeneration. Whether these mechanisms are actually involved in neurodegeneration in vivo especially in the case of "sporadic" Alzheimer disease should be elucidated in the future study.