Glycogen synthase kinase-3beta regulates presenilin 1 C-terminal fragment levels.

The majority of familial Alzheimer's disease cases have been attributed to mutations in the presenilin 1 (PS1) gene. PS1 is synthesized as an inactive holoprotein that undergoes endoproteolytic processing to generate a functional N- and C-terminal heterodimer (NTF and CTF, respectively). We identified a single residue in PS1, Ser(397), which regulates the CTF levels in a population of dimer that has a rapid turnover. This residue is part of a highly conserved glycogen synthase kinase-3beta (GSK-3beta) consensus phosphorylation site within the loop domain of PS1. Site-directed mutagenesis at the Ser(397) position increased levels of PS1 CTF but not NTF or holoprotein. Similar increases in only CTF levels were seen when cells expressing wild type PS1 were treated with lithium chloride, an inhibitor of GSK-3beta. Both wild type and PS1 S397A CTF displayed a biphasic turnover, reflecting rapidly degraded and stable populations. Rapid turnover was delayed for mutant PS1 S397A, causing increased CTF. These data demonstrate that PS1 NTF.CTF endoproteolytic fragments are generated in excess, that phosphorylation at Ser(397) by GSK-3beta regulates the discard of excess CTF, and that the disposal of surplus NTF is mediated by an independent mechanism. Overall, the results indicate that production of active NTF.CTF dimer is more complex than limited endoproteolysis of PS1 holoprotein and instead involves additional regulatory events.


Ford Kirschenbaum, Shu-Chi Hsu, Barbara Cordell ‡, and Justin V. McCarthy
From Scios Inc., Sunnyvale, California 94085 The majority of familial Alzheimer's disease cases have been attributed to mutations in the presenilin 1 (PS1) gene. PS1 is synthesized as an inactive holoprotein that undergoes endoproteolytic processing to generate a functional N-and C-terminal heterodimer (NTF and CTF, respectively). We identified a single residue in PS1, Ser 397 , which regulates the CTF levels in a population of dimer that has a rapid turnover. This residue is part of a highly conserved glycogen synthase kinase-3␤ (GSK-3␤) consensus phosphorylation site within the loop domain of PS1. Site-directed mutagenesis at the Ser 397 position increased levels of PS1 CTF but not NTF or holoprotein. Similar increases in only CTF levels were seen when cells expressing wild type PS1 were treated with lithium chloride, an inhibitor of GSK-3␤. Both wild type and PS1 S397A CTF displayed a biphasic turnover, reflecting rapidly degraded and stable populations. Rapid turnover was delayed for mutant PS1 S397A, causing increased CTF. These data demonstrate that PS1 NTF⅐CTF endoproteolytic fragments are generated in excess, that phosphorylation at Ser 397 by GSK-3␤ regulates the discard of excess CTF, and that the disposal of surplus NTF is mediated by an independent mechanism. Overall, the results indicate that production of active NTF⅐CTF dimer is more complex than limited endoproteolysis of PS1 holoprotein and instead involves additional regulatory events.
The familial form of Alzheimer's disease has been found to be largely attributed to mutations in the presenilin proteins, presenilin 1 (PS1) 1 and presenilin 2 (PS2). The role of PS proteins in Alzheimer's disease is of interest, because of this causal relationship to the disease (reviewed in Ref. 1). PS are polytopic membrane proteins to which a variety of functions have been ascribed (reviewed in Refs. 2 and 3). Functions linked to PS include protein processing and transport; response to stress, survival, and apoptosis; and intracellular signaling in cell fate determination. PS proteins are proteolytically processed to an N-terminal 30-kDa (NTF) and 20-kDa C-terminal (CTF) fragment that form a functional heterodimeric structure with a 1:1 stoichiometry between NTF and CTF (4 -7). Formation of the PS NTF⅐CTF dimer is under tight regulation where absolute levels are saturable (4, 6 -8). The tight control of PS dimer formation suggests that the concentration of dimer is linked to PS function or to an undefined limiting factor. The physiological mechanisms used to control PS dimer levels are presently unknown.
PS proteins have been shown to be components of high molecular weight protein complexes (9 -11). Uncleaved PS monomers are incorporated into an immature complex of ϳ180 kDa or are degraded by a proteosome-dependent mechanism. Maturation of the complex is associated with an increase in molecular mass to ϳ250 kDa and cleavage of the PS holoprotein to generate the functional NTF⅐CTF dimeric structure. This tight regulation of processed PS dimer is exemplified by overexpression of PS1, which results in a replacement of the endogenous PS1 NTF⅐CTF complex. Furthermore, PS1 replacement inhibits the synthesis of PS2 NTF and CTF fragments and vice versa (4,7,12). The PS heterodimer is stable, unlike PS holoprotein, which is rapidly degraded (6,7,13). Ultimately, the PS heterodimer is polyubiquitinated and marked for degradation by the proteosome. Inhibitors of proteosome activity have been shown to cause an increase in PS NTF and CTF as well as uncleaved holoprotein (7, 14 -16).
Within the ϳ250-kDa high molecular weight complex, PS1 is associated with ␤-catenin (17)(18)(19) and nicastrin (20). The association between PS1 and ␤-catenin occurs via the large hydrophilic loop located between transmembrane domains 6 and 7 of PS1 (12,(21)(22)(23). PS1 also associates with other proteins such as glycogen synthase kinase-3␤ (GSK-3␤), the serine/threonine kinase that negatively regulates ␤-catenin cell survival signaling by tagging this protein for degradation (24). The possibility that GSK-3␤ may also regulate PS1 levels is suggested by the presence of three highly conserved GSK-3␤ consensus phosphorylation sites within the hydrophilic loop domain of PS1 (23). Previously, we reported that phosphorylation of PS1 at one of these GSK-3␤ consensus motifs is critical for the interaction between PS1 and ␤-catenin (23). In this study, we identify a different GSK-3␤ consensus motif present in the PS1 loop domain that, when mutated, selectively alters the turnover of PS1 CTF.

EXPERIMENTAL PROCEDURES
Cell Culture-Human embryonic kidney 293T (293HEK) cells were grown in Dulbecco's modified Eagle's medium-21 containing 10% fetal bovine serum, 1% L-glutamine, and antibiotics (50 units/ml penicillin and 50 g/ml streptomycin). Cells were maintained in a humidified 37°C incubator with 5% CO 2 . Transient transfection was carried out with 293HEK cells using the calcium phosphate precipitation method with the indicated DNA constructs as previously described (23). Experiments with lithium chloride were performed 8 h post-transfection with a 24-h exposure of 10 mM lithium chloride. For experiments with proteosome inhibitors, ALLN and lactacystin were purchased from Sigma.
Expression Vector Construction-A cDNA encoding PS1 was amplified by standard polymerase chain reaction techniques and subcloned into a C-terminal Myc-His 6 -tagged mammalian expression vector, pcDNA3.1/Myc-His(Ϫ) (Invitrogen). Site-directed mutagenesis of the * This work was supported in part by Lilly. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
serine and threonine residues found in PS1, which conform to the (S/T)XXXS GSK-3␤ consensus phosphorylation site were generated by using a two-primer pair method protocol outlined by the QuikChange site-directed mutagenesis kit (Stratagene). Specific primers used in the mutagenesis are described by Kirschenbaum et al. (23). The fidelity of each PS1 mutation and polymerase chain reaction amplification were confirmed by DNA sequence analysis.
Western Blot and Co-immunoprecipitation Analyses-293HEK cells were transiently transfected with the indicated constructs and were harvested 24 -48 h post-transfection. The cell monolayers in 100-mm plates were washed twice with ice-cold phosphate-buffered saline and lysed in 1 ml of lysis buffer (10 mM HEPES, 10 mM KCl, 1% Nonidet P-40, 1 mM dithiothreitol, and protease inhibitor mixture (Complete; Roche Molecular Biochemicals). Cells were lysed on ice for 15 min and then spun at 16,000 ϫ g for 20 min, and the supernatants were collected for analysis. The bicinchoninic acid method (Pierce) was used to normalize each sample such that an equivalent amount of protein was applied to the gel. Samples were resolved on 4 -12% NuPAGE Bis-Tris gels (Novex), transferred to membranes, and blotted with the indicated antibody. Proteins were visualized by Western blot with luminol reagent (Santa Cruz Biotechnology, Inc., Santa Cruz, CA). A tetra-His monoclonal antibody was purchased from Qiagen. PS1 N-terminal and C-terminal domain-specific monoclonal antibodies were prepared. The PS1 monoclonal antibody directed to the C-terminal "loop" domain (3.6.1) was obtained by immunizing mice with a synthetic peptide spanning PS1 residues 309 -331. The monoclonal antibody directed to the N terminus (614.1) was raised a glutathione S-transferase fusion protein expressed in bacteria containing the N-terminal 77 residues of PS1. The rat anti-human PS1 N-terminal antibody was purchased from Chemicon. Nickel bead affinity purification of His-tagged CTF was carried out as previously described (23).
In Vitro Phosphorylation-Assessment of the PS1 consensus phosphorylation motif was carried out in vitro according to the method described by Dong et al. (25) with only minor changes. Briefly, synthetic peptides (30 -50 M) containing the Ser 397 , Ser 401 motif, NH 2 -393 VGKA-SATASGDWN 405 -COOH (wild type), or an altered motif to substitute serine 397 for alanine, NH 2 -393 VGKAAATASGDWN 405 -COOH (S397A), was incubated with 5 units of kinase, 10 M ATP, and 2 Ci of [␥-32 P]ATP at 111 TBq/mmol (Amersham Pharmacia Biotech) for 30 min at 37°C. Incorporation of radiolabeled phosphate was determined by adsorption of the peptide to P81 membranes and scintillation counting. Synthetic PS1 peptides used for in vitro kinase studies were obtained from American Peptide Co. GSK-3␤ (rabbit), produced by recombinant methods, was purchased from Sigma or New England Biolabs.
Pulse-Chase Analysis-293HEK cells were transiently transfected with either mock cDNA, wild type PS1, or PS1 S397A mutant with 2 g of DNA per well of a six-well plate. Twenty-four hours post-transfection, cells were washed twice with phosphate-buffered saline, and growth medium was replaced with methionine-, cysteine-, and serum-free medium and incubated at 37°C for 60 min. After the incubation, the medium was supplemented with 100 Ci/ml [ 35 S]methionine/cysteine (Amersham Pharmacia Biotech) and incubated for 20 or 120 min. Following the pulse of radioactivity, cells were washed several times, placed in complete medium, and incubated at 37°C. Cells were harvested at the indicated times and lysed, and CTF and PS1 holoprotein were immunoprecipitated from the lysates as previously described (23). Immunoprecipitates were resolved on 4 -12% NuPAGE Bis-Tris gels (Novex) and analyzed by a PhosphorImager (Molecular Dynamics, Inc., Sunnyvale, CA). Western blot analysis was used to ensure equivalent levels of exogenous PS1 expression.

PS1 Loop Domain Contains GSK-3␤ Consensus Phosphorylation
Sites-PS1 is synthesized as a 46-kDa holoprotein with six or eight transmembrane-spanning domains. This 46-kDa protein is endoproteolytically processed to generate a NTF⅐CTF heterodimer (4 -7). PS CTF has a large hydrophilic loop extension between transmembrane domains 6 and 7 that has been shown to interact with a number of proteins (12,(21)(22)(23)(24). Alignment of the hydrophilic loop domain of PS1 (residues 263-407) from several species revealed that this domain contains three highly conserved and potential GSK-3␤ consensus phosphorylation sites: TERES 324 , STPES 357 , and SATAS 401 , two of which are completely conserved in all species examined (Ref. 23 and Fig. 1). The sequence (S/T)XXXS is known to be a consensus sequence for GSK-3␤ phosphorylation. Since this motif is highly conserved in PS1, it seems likely that these identified GSK-3␤ consensus phosphorylation sites mediate specific PS1 function(s). Indeed, we have previously reported that the STPES 357 motif is necessary for the physical association between PS1 and ␤-catenin (23). Since phosphorylation by GSK-3␤ is known to target proteins for ubiquitination and proteosome degradation (24), we sought to determine whether GSK-3␤ regulates levels of PS1 via these phosphorylation consensus motifs present in the CTF hydrophilic loop domain.
Mutation at a GSK-3␤ Phosphorylation Site Regulates CTF Levels-Mutant PS1 cDNAs were generated containing single or combinatorial amino acid substitutions of the critical serine and threonine codons of all three (S/T)XXXS consensus sites. Table I summarizes the 13 different PS1 mutants generated, the specific residues mutated, and the nomenclature used to reference each mutant. We previously demonstrated that none of the GSK-3␤ consensus site mutations had an effect on PS1 holoprotein endoproteolysis or saturable accumulation of PS1 NTF⅐CTF dimer (23). In the experiments for this work, subconfluent 293HEK cell cultures were transiently transfected with wild type or mutant PS1, each tagged with a C-terminal His epitope. Expression of PS1 in transfected cell lysates was assessed by Western blot analysis with an anti-His monoclonal antibody. The presence of exogenous 46-kDa PS1 holoprotein and 20-kDa CTF was detected, revealing that some of the site-directed mutants resulted in increased levels of CTF ( Fig.  2A). (Exposure was adjusted for mutant levels and was, therefore, insufficient to readily visualize exogenous wild type CTF.) A common feature of all of the mutants showing increased amounts of CTF was the presence of mutations within the SATAS 401 GSK-3␤ consensus motif. The increase in PS1 CTF cannot be simply explained as an artifact associated with protein overexpression, since levels of the 46-kDa holoprotein were roughly equivalent to wild type and the other mutants, nor can the increase in CTF result from the presence of the His-epitope tag, since all contain the tag, both mutant and wild type PS1.
We also examined NTF levels in the same lysates using an N-terminus-specific antibody by Western blotting (Fig. 2A). PS NTF and CTF are typically seen as a dimer with a 1:1 stoichiometry. It was remarkable that unlike the large increase in CTF seen upon expression of SATAS 401 site mutations, the 30-kDa NTF levels were equivalent to wild type for all mutations within all GSK-3␤ consensus motifs, including those at the SATAS 401 site. To elucidate specific residue requirements within the SATAS 401 motif on the accumulation of PS1 CTF, we individually mutated the serine residues within this site. The two single mutants, PS1 S397A and PS1 S401R, as well as the double mutant (PS1 S397A,S401R) and wild type PS1 were transfected into 293HEK cells, and CTF levels were analyzed by Western blot. The results obtained clearly indicated that the single mutation S397A was uniquely responsible for the elevation of CTF as PS1 CTF levels from S401R expression were identical to those obtained with exogenous wild type PS1 (Fig.  2B).
Since the accumulation of PS1 NTF and CTF derivatives typically occurs in a 1:1 stoichiometry, we examined whether mutating residue S397A affected the PS1 NTF⅐CTF stoichiometry and dimer composition. To do so, we immunoprecipitated the dimer using either a CTF-or NTF-specific antibody and then Western blotted to detect the complementary fragment of the dimer. Wild type and PS1 S397A CTF were immunoprecipitated using an anti-His antibody and then were Western blotted for NTF using the N-terminal antibody, 614.1 (Fig. 2C). Consistent with direct Western blot analysis in Fig. 2, A and B, the immunopurified sample contained increased amounts of PS1 CTF in cells expressing PS1 S397A compared with cells expressing wild type PS1. In contrast, immunopurified samples from both wild type PS1-and PS1 S397A-expressing cells had equivalent amounts of PS1 NTF (Fig. 2C, left panel). The reciprocal experiment was carried out. NTF was immunoprecipitated from cells expressing either wild type or PS1 S397A followed by Western blotting for CTF (Fig. 2C, right panel). Equal amounts of both NTF and CTF were seen for both wild type and mutant PS1. These results suggest that the 1:1 NTF/ CTF stoichiometry was maintained in the mutant and that excess CTF was not affiliated in the dimer structure. Thus, mutation at residue Ser 397 within the SATAS 401 GSK-3␤ consensus motif of PS1 perturbs CTF accumulation independent of alterations in holoprotein, NTF, or the functional dimeric complex.
Phosphorylation of PS1 CTF by GSK-3␤-To further assess the role of GSK-3␤ in regulating PS1 CTF levels, we examined whether inhibition of GSK-3␤ affected CTF accumulation. Although it is reported to have several biological effects, including a role as a neuroprotective agent, lithium chloride specifically inhibits GSK-3␤ in the millimolar range but has no effect on numerous other protein kinases, including cAMP-dependent protein kinase, mitogen-activated protein kinase, and casein kinase II. Lithium thus mimics the effects of physiological GSK-3␤ inhibition. Treatment of cells with lithium chloride should mimic the observed selective accumulation of PS1 CTF when residue Ser 397 is mutated. To test this, subconfluent 293HEK cells were transfected with wild type PS1 and then exposed to 10 mM lithium chloride in complete medium for 24 h. A companion culture was identically manipulated except that lithium chloride was not added. Western blot analysis with a PS1 CTF-specific anti-His antibody or a PS1 NTF-specific antibody, 614.1, revealed that inhibition of GSK-3␤ activity by lithium chloride resulted in a significant 3-fold increase in CTF (Fig. 3A). In contrast, when the same cell lysates were assessed for NTF levels, inhibition of GSK-3␤ was without effect (Fig.  3A).
To further confirm the role of GSK-3␤ phosphorylation of PS1 CTF at residue Ser 397 , we carried out in vitro kinase studies. Synthetic peptides harboring the SATAS 401 motif centered within the sequence were used as substrates for GSK-3␤ phosphorylation. Two peptides were evaluated, one bearing the wild type motif (SATAS 401 ) and one with a mutation at Ser 397 changing the serine residue to an alanine (AATAS 401 ). Recombinant GSK-3␤ was used as the enzyme source. The data in Fig.  3B verify that this sequence is a substrate for GSK-3␤ as phosphorylation was only observed with the peptide bearing the wild type motif. Mutating residue Ser 397 within the SA-TAS 401 motif dramatically reduced GSK-3␤ phosphorylation, indicating that Ser 397 was the recipient residue of the phosphate group. Together these in vitro and in vivo data indicate that GSK-3␤ recognizes the SATAS 401 motif and that this site influences the cellular steady-state level of PS1 CTF.
Proteosome Inhibition Further Increases PS1 S397A CTF Levels-PS1 holoprotein, NTF and CTF have been reported by several groups to be degraded by the proteosome (7,15,16). In these studies, inhibitors of proteosome activity, such as the peptide aldehyde ALLN and lactacystin, were shown to inhibit proteosome-mediated proteolysis of PS1 and its derivatives, and that this leads to an accumulation of PS1 proteins. We were interested in determining whether mutating the SA-TAS 401 GSK-3␤ recognition motif in PS1 affected proteosomemediated turnover of PS1 and, specifically, if CTF accumulation by the PS1 S397A mutant could be further increased by proteosome inhibition.
In the first set of experiments, 293HEK cells transiently expressing PS1 wild type or PS1 S397A were treated with 25 M ALLN. At various times following inhibitor treatment, cells were harvested, lysates were prepared, and PS1 proteins were analyzed by Western blot. Upon treatment with ALLN, a timedependent increase in PS1 holoprotein, NTF, and CTF was observed (Fig. 4A). All three PS1 proteins accumulated in parallel for PS1 wild type and PS1 S397A. The data in Fig. 4A also show that although PS1 S397A had increased basal levels of CTF compared with wild type PS1, mutant CTF level could be further increased over time by ALLN treatment, suggesting that PS1 CTF levels/stability can also be regulated in a GSK-3␤-independent manner. Changes in CTF mobility were noted for both PS1 wild type and PS1 S397A that most likely reflected ubiquitination of the protein.
Since ALLN also inhibits nonproteosomal proteases, the highly specific proteosome inhibitor, lactacystin, was included in these studies to confirm that the accumulated and modified forms of PS1 resulted from proteosome-specific inhibition. Treatment of 293HEK cells transiently expressing wild type PS1 or the PS1 S397A mutant with 10 M lactacystin documented the time-dependent accumulation of PS1 holoprotein, NTF, and CTF (Fig. 4B). Consistent with data from ALLN treatment, lactacystin exposure also resulted in the further accumulation of PS1 S397A CTF. From these studies, we conclude that PS1 S397A was degraded by the proteosomal system in a manner identical to wild type PS1. Because proteosome inhibition can further elevate PS1 S397A CTF levels, this sug- Ϫ PS1 S397A,S401R,T320R,S324A ϩ PS1 S353A,S357A,S397A,S401R,T320R,S324A ϩ a ϩ, increase in PS1 CTF over wild type; Ϫ, no increase.

GSK-3␤ Regulates Presenilin-1 CTF Level
gests that PS1 CTF levels can be regulated by both GSK-3␤dependent and -independent mechanisms. Turnover Kinetics of PS1 Wild Type and S397A CTF-To further examine the biology of elevated PS1 S397A CTF, we compared the metabolic half-life of the mutated PS1 CTF with wild type CTF using pulse-chase kinetic analysis. For this, 293HEK cells transiently expressing mutant or wild type PS1 were radiolabeled for 120 min with [ 35 S]methionine/cysteine, after which the nascent radiolabeled protein was chased for the times indicated over a 45-h period. At each time point, PS1 CTF was immunoprecipitated with an anti-His antibody, analyzed by gel electrophoresis, and quantified. Gels were normalized for equal amounts of 0-h radioactivity loaded and for equivalent levels of exogenous PS1 wild type and PS1 S397A expression. The data in Fig. 5A show that S397A CTF exhibited turnover kinetics similar to wild type CTF. However, close inspection of Fig. 5A revealed a biphasic nature to the degradation profiles for both wild type and mutant PS1 CTF in which one population of CTF was rapidly degraded within the first 10 h and a second population was long lived from 10 to 45 h. Furthermore, PS1 S397A mutant CTF had an extended half-life for only the short-lived population compared with wild type CTF. Additional pulse-chase experiments were performed to specifically examine the rapidly degrading CTF population. In these experiments, the radioactive pulse was decreased to 20 min, and additional early time points were taken in the chase period. Nascent PS1 CTF was analyzed from PS1 wild type and PS1 S397A mutant from transfected cells (Fig. 5B). Reproducibly a biphasic profile was seen for wild type and mutant PS1 CTF degradation. DISCUSSION Since the identification of PS1 as a key protein in development of Alzheimer's disease, there has been extensive research on the structure and function of this protein. Endoproteolysis of the 46-kDa PS1 holoprotein to a heterodimer composed of 20-kDa CTF and 30-kDa NTF has been linked to the stabilization of PS1 (6, 7, 13), to its incorporation into a high molecular weight complex (9, 10), and to its maturation as a functional entity (11). In this mature stable state, the PS1 dimeric struc- Twenty-four hours post-transfection, cell extracts were prepared, and PS1 protein levels were assessed by Western blot with an anti-His monoclonal antibody for fulllength PS1 holoprotein and CTF and with the Chemicon antibody for NTF. Fulllength 46-kDa PS1, 30-kDa NTF, and 20-kDa PS1 CTF are indicated. Essentially identical results were obtained in at least another three independent experiments. C, immunoprecipitation (IP) of PS1 CTF followed by Western blot analysis for associated NTF (left) and immunoprecipitation of PS1 NTF followed by Western blot for associated CTF (right). 293HEK cell lysates were prepared after transient transfection with vector only, PS1 wild type, or S397A containing a C-terminal His epitope tag. PS1 CTF was immunoprecipitated using nickel chromatography, and associated NTF was detected by Western blot using the 614.1 monoclonal antibody. For the reciprocal experiment, lysates were immunoprecipitated with 614.1 NTF-specific antibody, and associated CTF was detected by Western blotting with an anti-His antibody. Fulllength PS1 holoprotein documents expression levels. Essentially identical results were obtained in another two independent experiments.
ture subserves a variety of functions related to protein processing, protein transport, and intracellular signaling.
The present study provides new information on the regulation of the PS1 heterodimeric components. We identified three consensus phosphorylation sites for the serine/threonine kinase GSK-3␤ within the PS1 C-terminal hydrophilic loop domain. These sites have been analyzed with respect to PS1 function. In this study, we demonstrate that one of the GSK-3␤ consensus motifs, SATAS 401 , regulates the level of PS1 CTF. Single residue mutation at serine 397 of this motif was shown to be both dominant and sufficient for elevating the levels of the 20-kDa PS1 CTF severalfold. Additional evidence is provided by the in vitro phosphorylation of a PS1 synthetic peptide harboring the SATAS 401 motif by recombinant GSK-3␤. It is important that a peptide bearing a substitution at serine 397 but preserving serine 401 was not phosphorylated by GSK-3␤ in vitro. It is known that phosphorylation by GSK-3␤ is an ordered process requiring prior phosphorylation of substrate at the second serine of the (S/T)XXXS motif most C-terminal to all other GSK-3␤ motifs (26). This is entirely consistent with our results; the SATAS 401 GSK-3␤ motif is the most C-terminal of the three motifs across the PS1 loop domain, and serine 397, not serine 401, is phosphorylated by GSK-3␤. The kinase responsible for prior phosphate addition at the GSK-3␤ motif of PS1 is not known. For other GSK-3␤ substrates, casein kinase II phosphorylation of the most C-terminal motif in the hierarchy has been identified as preceding and necessary for GSK-3␤ recognition (26). Curiously, PS2, not PS1, CTF has been shown to be phosphorylated by casein kinase II (27). PS1 CTF has been reported to be phosphorylated not only by GSK-3␤ (Ref. 23 and data herein) but by two other serine kinases, protein kinase C (27)(28)(29), and protein kinase A (27,28). Therefore, the non-GSK-3␤-mediated phosphorylation of PS1 CTF may represent additional regulatory signals designed to regulate the functional dimeric structure. However, the non-GSK-3␤ phosphorylation of PS1 CTF appears to only occur upon kinase induction. For example, phorbol ester stimulation of protein kinase C resulted in increased phosphorylation of PS1 CTF, and treatment with protein kinase C inhibitors did not reduce PS1 CTF phosphorylation below basal levels (28,29). These results indicate that protein kinase C phosphorylation sites on PS1 CTF are not fully phosphorylated in an unstimulated state, if at all. Our previous data suggest that the basal phosphorylated state of PS1 CTF may be largely due to GSK-3␤ activity because mutation of all three GSK-3␤ consensus sites reduces basal phosphorylated CTF in vivo (23). The extended SATAS 401 motif, KASATAS 401 , also contains the minimal consensus sequence for protein kinase C, (R/K)XS. However, since treatment of cells with lithium chloride, an inhibitor of GSK-3␤, gave the same phenotype as the Ser 397 mutation (i.e. elevated CTF), it is not likely that protein kinase C is responsible for phosphorylating this motif in vivo or for the increase in PS1 CTF.
The regulation by GSK-3␤ phosphorylation is unique to the CTF fragment, since no concomitant increase in the 30-kDa NTF was obtained. Heretofore, NTF and CTF levels were assumed to be coordinately regulated through the endoproteolytic cleavage, which generates the heterodimeric structure. The resulting 1:1 stoichiometry observed for PS1 NTF and CTF is consistent with this simple model of regulation. Our results indicate that the regulation is more complex. We have shown that GSK-3␤ regulates, in part, CTF levels through phosphorylation at SATAS 401 in the loop domain. Attenuating GSK-3␤ action by mutating serine 397 of PS1 or by treatment with lithium chloride disrupts overall NTF⅐CTF stoichiometry. Hence, NTF and CTF levels can be independently regulated after endoproteolytic cleavage of the PS1 holoprotein. Second, we have observed that CTF fragment is produced in excess and that surplus dimer is degraded early (within ϳ5 h) of its generation, after which the remaining population has an extended half-life (ϳ24 h; Refs. 6 and 7). Turnover of surplus CTF was found to be regulated by GSK-3␤, since PS1 Ser 397 mutants have increased amounts of nascent CTF detected in pulsechase experiments. This CTF surplus persists in the steadystate condition as evidenced by increased amounts upon Western blotting.
PS2 lacks all of the three GSK-3␤ consensus motifs; thus, the regulation of the PS2 NTF⅐CTF complex must be different from PS1. Indeed, PS2 displays a phosphorylation pattern distinct from PS1 involving casein kinase II (27). Furthermore, PS2 has been shown to be present in a separate high molecular complex (9) and to have a number of functions different from PS1 as evidenced by mice nullizygous for PS2 or PS1 (30 -32).
The aberrant levels of PS1 CTF resulting from the inability of GSK-3␤ to phosphorylate serine 397 does not perturb any of the known PS1 functions including the ability to bind ␤-catenin or processing of the ␤-amyloid precursor protein to A␤; nor does it increase cellular propensity to apoptosis (23). 2 Thus, the dysregulation of CTF levels does not appear to negatively impact the overall function of the NTF⅐CTF dimeric structure. This is consistent with our observation that surplus CTF is not physically associated with the dimer. It will be of interest to determine whether excess CTF produced upon inhibition of GSK-3␤ phosphorylation is present in the mature ϳ250-kDa A, influence of lithium chloride inhibition of GSK-3␤ on PS1 CTF levels. 293HEK cells transiently expressing vector only or His-tagged wild type PS1 were treated for with 10 mM LiCl for 24 h or were mock-treated. PS1 NTF and CTF were assessed by Western blot using 614.1 and anti-His antibodies, respectively. Three independent experiments were performed with essentially the same results. B, in vitro phosphorylation of SATAS 401 motif by GSK-3␤. In vitro GSK-3␤ kinase reactions were performed using synthetic peptides with the wild type Ser 397 , Ser 401 GSK-3␤ motif, or the motif mutated at Ser 397 . Reactions, run in duplicate, were incubated for 0 or 30 min, after which incorporation of 32 PO 3 was monitored by a filter binding assay. Essentially identical results were obtained in another two independent experiments. molecular mass complex or in the immature ϳ180-kDa complex described by Yu et al. (9,11) and Capell et al. (10). The regulation of PS1 CTF levels by GSK-3␤ is similar, if not identical, to the regulation of ␤-catenin. GSK-3␤ phosphorylation of ␤-catenin reduces ␤-catenin levels by marking this protein for ubiquitination and proteosomal degradation (reviewed in Ref. 33). The role of GSK-3␤ in PS1 dimer regulation might explain, in part, the localization of PS1 in the ␤-catenin signalsome complex (9,10). Like ␤-catenin, wild type and excess CTF produced by PS1 S397A are degraded by the ubiquitinationproteosomal system (7, 14 -16).
The recent report by Mah et al. (34) describes a novel presenilin-associating protein, ubiquillin, which shares functional similarities to the GSK-3␤ phosphorylation effect we observe. Ubiquillin contains multiple ubiquitin-related domains implicated in targeting proteins to the proteosome, and ubiquillin FIG. 4. Effects of proteosome inhibition of S397A PS1 CTF levels. A, treatment with ALLN. 293HEK cells transiently expressing PS1 wild type (WT) or PS1 S397A mutant were treated with 25 M ALLN for up to 6 h. Cultures were harvested at the times indicated, and lysates were prepared. CTF and full-length PS1 holoprotein were analyzed by Western blot using an anti-His monoclonal antibody and NTF analyzed by Western blot with 614.1 antibody. Ub CTF, putative ubiquitinated CTF. Essentially identical results were obtained in another two independent experiments. B, treatment with lactacystin. 293HEK cells transiently expressing wild type or PS1 S397A mutant were treated with 10 M lactacystin for up to 26 h. Cultures were harvested at the times indicated; lysates were prepared; and PS1 holoprotein, NTF, and CTF were analyzed by Western blot as outlined in A. Essentially identical results were obtained in another two independent experiments.
FIG. 5. Turnover kinetics of Ser 397 PS1 CTF. A, pulse-chase analysis of PS1 wild type (WT) and S397A mutant CTF. 293HEK cells transiently expressing PS1 wild type or PS1 S397A were radiolabeled with 100 Ci/ml 35 S methionine/cysteine for 120 min and then chased in complete medium at 37°C for 0, 10, 20, 35, and 45 h. CTF was immunoprecipitated from the lysates as previously described (23). Levels of wild type and S397A CTF bands from each time point were quantified by scanning the autoradiogram on a PhosphorImager. B, pulse-chase analysis of wild type and S397A mutant CTF; same as in A except that a 20-min pulse and 0-, 2-, 4-, and 18-h chase periods were used. In another three independent experiments, with varying time points, essentially identical biphasic effects were observed. was shown to bind to the C-terminal and loop domains of PS. Exogenous expression of ubiquillin was found to facilitate increased PS expression without altering the stability of PS. Unfortunately, this was only demonstrated for holoprotein, not for NTF or CTF. One might speculate that elimination of GSK-3␤ phosphorylation at Ser 397 of the PS1 loop domain interrupts its association with ubiquillin, thereby yielding higher levels of protein.
Levels of the processed PS1 NTF⅐CTF dimeric structure are tightly regulated (4, 6 -8). Here, we describe that GSK-3␤ can exclusively influence levels of the CTF by phosphorylation of Ser 397 in the PS1 loop domain. Phosphorylation at this site eliminates surplus CTF prior to the appearance of the stable CTF in the long lived dimer. Because the NTF⅐CTF unit is the functional PS1 structure, it is possible that perturbations in the regulation of individual components of the dimer may have pathological consequences that contribute to the development of Alzheimer's disease. In particular, modulation of CTF levels by GSK-3␤ may influence cell survival. The C-terminal portion of PS2, ALG-3, has been shown to act as a dominant negative inhibitor of apoptosis (35)(36)(37), and recently an identical activity for the corresponding PS1 C-terminal fragment has been reported (18).