Phosphorylation-dependent Regulation of the Interaction of Amyloid Precursor Protein with Fe65 Affects the Production of (cid:1) -Amyloid*

Neuronal Fe65 is an adapter protein that interacts with the cytoplasmic domain of the (cid:1) -amyloid precursor protein (APP). Although the interaction has been reported to occur between the second phosphotyrosine interaction domain of Fe65 and the YENPTY motif in the cytoplasmic domain of APP, the regulatory mechanism and biological function of this interaction remain un-known. We report here that (i) a single amino acid mutation at the Thr-668 residue of APP695, located 14 amino acids toward the amino-terminal end from the 682 YENPTY 687 motif, reduced the interaction between members of the Fe65 family of proteins and APP, whereas interaction of APP with the phosphotyrosine interaction domain of other APP binders such as X11-like and mammalian disabled-1 was not influenced by this mutation; (ii) the phosphorylation of APP at Thr-668 diminished the interaction of APP with Fe65 by causing a conformational change in the cytoplasmic domain that contains the Fe65-binding motif, YENPTY; and (iii) the expression of Fe65 slightly suppressed maturation

␤-Amyloid precursor protein (APP) 1 is the precursor of ␤-amyloid (A␤), which is associated with the pathogenesis of Alzheimer's disease (AD) (reviewed in Ref. 1). APP belongs to a member of a family of proteins that includes APP-like proteins (APLPs) 1 and 2. This family of proteins has a membraneassociated receptor-like structure composed of a large extracellular, a single transmembrane, and a short cytoplasmic domain (2)(3)(4)(5)(6). APP is ubiquitously expressed as different isoforms in neuronal and nonneuronal tissues. The protein is cleaved proteolytically to generate A␤ during its association with protein secretory and/or endocytotic pathways (reviewed in Ref. 7). Three functional motifs in the cytoplasmic domain of APP are thought to regulate its rate of secretion, endocytosis, and A␤ production. The amino acid sequence 653 YTSI 656 (human APP695 isoform numbering) forms a characteristic internalization and/or basolateral sorting signal, YXXI (8,9). The motif 667 VTPEER 672 contains Thr-668, the amino acid phosphorylated by neuronal cyclin-dependent kinase (cyclin-dependent kinase 5) in neurons (10), cdc2 kinase in dividing cells (11,12), and glycogen synthase kinase 3␤ and stress-activated protein kinase 1b in vitro (13,14). The phosphorylation of APP regulates neurite extension in differentiating PC12 cells (15). The motif also contains Arg-672, a necessary residue for the metabolism of APP by the nonamyloidogenic pathway (16). The amino acid sequence, 682 YENPTY 687 contains an NPXpY element, which is a typical internalization signal for membraneassociated receptor protein (8,17,18); however, in APP, Tyr-687 is not phosphorylated (17,19). These motifs are thought to function through interaction with cytoplasmic proteins termed APP binders.
Several APP binders have been identified. A microtubuleinteracting protein, PAT1, interacts with the 653 YTSI 656 motif and regulates the basolateral sorting of APP (9). The phosphotyrosine interaction (PI) domains of the proteins Fe65, Fe65like (Fe65L) 1, Fe65L2, X11, X11-like (X11L), X11L2, and mammalian disabled-1 (mDab1) interact with the 682 YENPT-Y 687 motif (20 -27). The UV-damaged DNA-binding protein, which does not contain a PI domain, also interacts with the 682 YENPTY 687 motif (28). However, the role of the 667 VT-PEER 672 motif in the binding of APP to APP binders remains to be determined. Recent analysis of this motif by multidimen-sional NMR spectroscopy showed it to consist of a type I ␤-turn and an amino-terminal helix-capping box structure (29). It was therefore expected that the 667 VTPEER 672 motif would influence the maintenance of the overall structure of the cytoplasmic domain of APP. Furthermore, it is possible that phosphorylation at Thr-668 of APP, which changes the conformation of the cytoplasmic domain, regulates the interaction between the cytoplasmic motifs of APP and APP binders. Therefore, we explored the effects of phosphorylation of Thr-668 on the interaction between APP and APP binders.
In previous studies it has been shown that the 682 YENPT-Y 687 motif of APP is essential for the basic interaction with Fe65 (30). In this study, we demonstrate that this interaction is regulated by phosphorylation at Thr-668. Analysis by CD spectroscopy showed that the structure of the cytoplasmic peptide phosphorylated at Thr-668 differed from that of the nonphosphorylated peptide. Thus, phosphorylation of APP at Thr-668 is likely to alter the conformation of its cytoplasmic domain.
Overexpression of Fe65 decreased the production of A␤ from APP; however, this effect was diminished when a mutation at Thr-668 was introduced. These observations suggest that APP metabolism is regulated by the neuron-specific phosphorylation of APP at Thr-668, which causes a conformational change in the cytoplasmic domain.
Preparation of Proteins and in Vitro Pull-down Assay-Human cDNAs encoding Fe65, Fe65L, and Fe65L2 were isolated in a yeast two-hybrid assay using the cytoplasmic domain of APP as bait. Fe65 cDNA was cloned into the pcDNA3.1(Ϫ)/mycHisA vector (CLONTECH). The cDNAs encoding the second PI (PI2) domains derived from Fe65 (amino acids 540 -665 of GenBank TM /EBI Data Bank accession number O00213), Fe65L1 (amino acids 548 -684 of GenBank TM /EBI Data Bank accession number Q92870), and Fe65L2 (amino acids 287-417 of Gen-Bank TM /EBI Data Bank accession number O95704) and WW domains derived from Fe65 (amino acids 232-288 of GenBank TM /EBI Data Bank accession number O00213) were subcloned into pGEX-4T-1 (Amersham Pharmacia Biotech) to produce GST-fusion proteins. The cDNA encoding the PI domain of mDab1 was cloned by reverse transcriptionpolymerase chain reaction using human brain total RNA (CLONTECH) and primers with the nucleotide sequences 5Ј-gcagtgaagccacttgataaagagg-3Ј (forward) and 5Ј-ctaattcttctcttgcttcaattcat-3Ј (reverse) (genome sequence, DDBJ/EMBL/GenBank TM accession number NM021080). These cDNA encoding the PI domain of X11L were subcloned from pcDNA3-hX11L (26). The cDNA fragments were also subcloned into pGEX-4T-1 (Amersham Pharmacia Biotech) for the production of GSTfusion proteins. These recombinant fusion proteins were affinity-purified as described (26).
Co-immunoprecipitation-HEK293 cells were cotransfected with pcDNA3.1-Fe65myc/HisA encoding human Fe65 tagged with Myc and His epitopes at the carboxyl-terminal and either pcDNA3-FLAGAPP695, pcDNA3-FLAGAPP695T668A, or pcDNA3-FLAGAPP695T668E. The Ala and Glu substitutions at Thr-668 were introduced to produce pcDNA3-FLAGAPP695T668A and -T668E (15). At 48 h after transfection, the cells were harvested, lysed, and centrifuged at 15,000 ϫ g for 10 min at 4°C. The anti-Myc antibody was added to the supernatant, and the sample was incubated on ice for 2 h. Protein G-Sepharose beads were then added, and the samples were incubated for 1 h at 4°C and then centrifuged. The beads were washed with TBST three times, and proteins were eluted by boiling the beads in SDS sample buffer. Proteins were then analyzed by SDS-polyacrylamide gel electrophoresis (6% (w/v) polyacrylamide) and immunoblotted with anti-FLAG antibody. Immunocomplexes were detected using an ECL detection kit (Amersham Pharmacia Biotech).
CD Spectroscopy-CD spectra were recorded at 25°C on a JASCO J-600 spectropolarimeter using a 0.1-cm path length cuvette. The cytoplasmic peptides of APP-(648 -695) with or without phosphate at the Thr-668 residue were dissolved at a concentration of 0.1 mM in 20 mM sodium phosphate buffer (pH 6.0) containing either 0% (v/v) or 47% (v/v) trifluoroethanol (TFE). Each CD spectrum is an average of five scans with a 1.0-nm bandwidth, a time constant of 2 s, and a step resolution of 0.2 nm. The percentage of ␣-helicity was calculated as [] 222 /[] max ϫ 100%, where [] max ϭ Ϫ39500 (1-2.57/n) where n is the number of residues in the peptide (32).
Pulse-Chase Study-Pulse-chase labeling of cells was carried out with [ 35 S]methionine (1 mCi/ml, Amersham Pharmacia Biotech AGQ0080). HEK293 cells were cotransfected with a combination of pcDNA3-FLAGAPP695 and pcDNA3.1-Fe65myc/HisA or pcDNA3-FL-AGAPP695 and pcDNA3.1(Ϫ)myc/HisA. At 48 h after transfection, the cells were labeled metabolically in Dulbecco's modified Eagle's medium with [ 35 S] methionine excluding cold methionine for 15 min followed by a chase period as indicated. The chase was initiated by replacing the labeling medium with medium containing excess unlabeled methionine. APP was immunoprecipitated from cell lysates using the anti-FLAG antibody M2 (Sigma) and separated by SDS-polyacrylamide gel electrophoresis (6% (w/v) polyacrylamide). Radioimmune precipitation buffer (50 mM Tris-HCl (pH 8.0), 150 mM NaCl, 1% Nonidet P-40, 0.5% deoxycholate, and 0.1% SDS) containing 25 g/ml pepstatin, 25 g/ml leupeptin, and 25 g/ml chymostatin was used for cell lysis. sAPP, a large extracellular amino-terminal domain cleaved at ␣and/or ␤-sites, was isolated from medium by immunoprecipitation with anti-FLAG antibody and separated by SDS-polyacrylamide gel electrophoresis (7.5% (w/v) polyacrylamide). Radioactivity in APP and sAPP was analyzed using a Fuji BAS 2000 imaging analyzer and autoradiography.

RESULTS
Role of Thr-668 of APP695 in Interaction with Fe65-The motif, 667 VTPEER 672 (numbering for APP695 isoform), of APP is thought to be important for the regulation of function and Phosphorylation-dependent Interaction of APP with Fe65 metabolism of the protein (11,15,34). Furthermore, phosphorylation of Thr-668 in this motif is believed to regulate the function of the 667 VTPEER 672 motif via alteration of the threedimensional structure of the cytoplasmic domain of APP (29). Many known APP binders such as the Fe65 family, X11, X11L, and mDab1 recognize the 682 YENPTY 687 motif, which lies in the cytoplasmic domain of APP. However, in the case of Fe65, effective binding to APP requires a larger motif that consists of the 682 YENPTY 687 sequence plus several extra amino acids that extend toward the amino terminus (35). Therefore, we first examined whether the phosphorylation site, Thr-668, of APP is involved in the binding of Fe65. It has been reported that PI2 of Fe65, but not PI1 or the WW domain (WW), interacts with the cytoplasmic domain of APP (35). GST-fusion proteins consisting of the PI2 or WW domains of Fe65 were used in pulldown assays with the wild-type form of APP695 or forms that contained amino acid mutations at Thr-668 (Fig. 1a). Glutathione-Sepharose beads bearing GST-fusion proteins were incubated with the lysate of COS7 cells transiently transfected with pcDNA3-FLAGAPP695 carrying alanine or glutamic acid substitutions for Thr-668 of APP695 or with wild-type APP695. Wild-type (T) APP bound to the PI2 domain of Fe65 (Fe65PI2) strongly, but the binding of the Ala-668 (A) and Glu-668 (E) mutants of APP to Fe65PI2 was weak even though these lysates contained the same amount of expressed APP. APP did not interact with the WW domain of Fe65 (Fe65ww) or the GST protein alone (GST) as was expected.
It is well known that the other Fe65 family proteins, Fe65L1 and Fe65L2, as well as other proteins including mDab1 and X11L also interact with the 682 YENPTY 687 motif of APP. Thus we examined whether the Thr-668 residue of APP was involved in the interaction of this protein with the PI domains derived from these APP binders. GST-fusion proteins consisting of the PI2 domain from Fe65L1 (Fe65L1PI2) and Fe65L2 (Fe65L2PI2) and the PI domains derived from mDab1 (mDab1PI) and X11L (X11LPI) were examined for binding activity to APP695 (Fig. 1b). As observed for Fe65, Fe65L1PI2 and Fe65L2PI2 bound to wild-type APP (T), and substitution of Thr-668 with Ala (A) or Glu (E) in APP restricted this binding. We concluded that binding of PI2 of all member of the Fe65 family to APP was decreased by mutation at Thr-668. mDab1PI and X11LPI bound to APP regardless of the presence of a mutation at Thr-668.
The importance of the Thr-668 residue of APP on the interaction with Fe65 was confirmed in vivo using a co-immunoprecipitation assay. Full-length Fe65 protein tagged with Myc/His sequences (Fe65-myc/His) and APP695 tagged with a FLAG sequence (FLAG-APP695) were expressed transiently in HEK293 cells. The detergent-soluble fraction of these cell lysates was subjected to immunoprecipitation using anti-Myc antibody, and the immunoprecipitates were analyzed by Western blot using anti-FLAG (Fig. 2) or anti-Myc (data not shown) antibodies. A large quantity of wild-type APP (T) was recovered in association with Fe65. However, comparatively small quantities of APP carrying Ala (A) and Glu (E) substitutions for Thr-668 were co-immunoprecipitated with Fe65 even though the amount of APP expressed was identical between experiments. Furthermore, APP was not co-immunoprecipitated with anti-Myc antibody when only APP (Ϫ in Fe65-myc) or Fe65myc/ His (Ϫ in APP) were expressed. These results demonstrate that the stable binding of Fe65 family members to APP requires the presence of Thr-668 of APP695. Other APP binders such as X11L and mDab1 did not need Thr-668 to be present for stable binding to APP.
Interaction of APLP2 with Fe65-APP is a member of the APP family of proteins, which consists also of APLP1 and APLP2 (2)(3)(4)(5). We have demonstrated previously that the phosphorylation site corresponding to Thr-668 of APP695 is conserved in the cytoplasmic domain of APLP2 as Thr-736 (numbering for APLP2-763 isoform) but not in the cytoplasmic domain of APLP1 (31). In neurons, Thr-736 of APLP2 is phosphorylated in an identical manner to that of Thr-668 in APP. 2 In addition, the YENPTY motif is also conserved in APLP2. Therefore, we examined the influence of Thr-736 on the binding of Fe65 to APLP2 (Fig. 3). Glutathione-Sepharose beads bearing GST-fusion proteins consisting of Fe65PI2 were incubated with lysate from cells transiently expressing wild-type APLP2 2 K. Iijima, K. Ando, and T. Suzuki, manuscript in preparation.
FIG. 1. Role of Thr-668 of APP695 in the interaction with the PI domains of APP-binders. The role of Thr-668, a phosphorylation site in APP695, in the interaction with the PI domain from APP binders was analyzed using pull-down assays in vitro. Lysates derived from COS7 cells expressing wild-type FLAG-APP695 (T) or FLAG-APP695 containing an alanine (A) or glutamic acid (E) substitution for Thr-668 were incubated with GST-PI domain fusion proteins, and the amount of bound APP was quantified by Western blot analysis with anti-FLAG antibody. a, GST-PI2 domain (Fe65PI2), GST-WW (Fe65WW) domain of Fe65, or GST alone (GST) coupled to glutathione beads were incubated with cell lysate. The expression level of FLAG-APP695 was quantified by Western blot analysis of cell lysate (Lysate). b, GST-PI2 domains of Fe65L1 (Fe65L1PI2) and Fe65L2 (Fe65L2PI2) or GST-PI domains of mDab1 (mDab1PI) and X11L (X11LPI) coupled to glutathione beads were incubated with cell lysates containing equal amounts of APP (Lysate). Lysate derived from cells transfected with vector alone (Ϫ) was used as the negative control. The amount of APP that was bound to the glutathione beads was determined by Western blot analysis with anti-FLAG antibody.
(T) or mutant APLP2 containing an alanine (A) substitution for Thr-724 (the position at Thr-724 of APLP2-751 corresponds to Thr-736 of APLP2-763). Wild-type APLP2 bound to Fe65PI2, but the substitution of Ala for Thr-724 of APLP2 interfered with binding. Both wild-type and mutated protein bound to mDab1PI and did not bind to GST beads alone (GST). We concluded that no significant difference was detected between wild-type and mutated APLP2 in the binding to mDab1PI. The result suggests that regulation of the interaction of APLP2 with Fe65 is similar to that for APP and Fe65 in that it is mediated by the phosphorylation of the corresponding threonine in the cytoplasmic domains.
Phosphorylation at Thr-668 Interferes with the Interaction of Fe65 with the 682 YENPXY 687 Motif of APP-Mutagenesis studies of APP695 indicated that the presence of Thr-668 allowed stable interaction of Fe65 with the cytoplasmic domain of APP and also suggested the possibility that phosphorylation at this site regulated the interaction of Fe65 with the 682 YENPTY 687 motif. To demonstrate this possibility, a competition assay studying the interaction of APP with Fe65 was performed using APP cytoplasmic domain peptides with or without phosphate at the Thr-668 residue as competitors (Fig. 4). Glutathione-Sepharose beads bearing GST-fusion proteins consisting of Fe65PI2 were incubated with the APP cytoplasmic domain peptide, APP-(648 -695), or the phospho-Thr-668 peptide, [pThr-668]APP-(648 -695), prior to incubation with lysate derived from COS7 cells, which expressed FLAG-APP695 transiently. After washing the beads, the amount of APP that had bound to the beads was determined by Western blot analysis using an anti-FLAG antibody (Fig. 4a), and the interaction of APP with Fe65PI2 was quantified (Fig. 4c). The APP cytoplasmic peptide containing nonphosphorylated Thr-668 interfered with the interaction of Fe65 with APP695 more strongly than that containing phospho-Thr-668, although both peptides, which contained the 682 YENPTY 687 motif, involved in recognition of Fe65, had the ability to compete with APP695 for the binding of Fe65PI2. Shorter peptides that consisted of nonphosphorylated Thr-668 (APP663-676) or phosphorylated Thr-668 (pAPP663-676) without the 682 YENPTY 687 motif did not interfere with the interaction of Fe65 with APP (Fig. 4b). These results indicate that the cytoplasmic domain peptide of APP that contains both the 682 YENPY 687 motif and a nonphosphorylated Thr-668 is highly competitive with APP for the binding of Fe65. The amino acid sequence around Thr-668 did not seem essential for recognition by Fe65.
We previously found that the phosphorylation of APP at Thr-668 is specific to neuronal tissue (10). In neurons, ϳ10% of mature APP695 (N-and O-glycosylated) exists in the phosphorylated form and is distributed in the plasma membrane and in neurites. Immature APP695 (N-glycosylated) is localized to the endoplasmic reticulum and early Golgi and exists in the completely nonphosphorylated form. Therefore, we examined the interactions of the different forms of endogenous APP derived from brain with Fe65PI2 (Fig. 5). APP was recovered from the brain lysate of adult rats by immunoprecipitation using the anti-APP antibody, UT-18 (AbAPP). Quantification of the content of APP and APP phosphorylated at Thr-668 (pAPP) was performed by Western blot analysis using UT-18 and the antiphospho-Thr-668 APP antibody, UT-33 (Fig. 5a). As described previously, the phosphorylated form of APP (pAPP) consists of two types of mature APP695 modified with differential Oglycosylation, whereas the nonphosphorylated form contains immature APP695 (10). Aliquots of the same brain extract containing these endogenously phosphorylated and nonphosphorylated forms of APP were incubated with glutathione-Sepharose beads coupled with GST-fusion proteins consisting of Fe65PI2 or X11LPI. The APP that attached to the beads was eluted and analyzed by Western blot using UT-18 and UT-33 (Fig. 5b). Two mature APP and one immature APP species bound to Fe65PI2 and X11LPI but not to GST alone. The immature APP preferentially bound to Fe65PI2, but both mature APP and immature APP bound to X11LPI equally with respect to the physiological ratio of mature APP to immature APP in the brain (compare the left panels in Fig. 5, a and b). Although mature APP did bind to Fe65PI2, the level of association of the phosphorylated form of mature APP with Fe65PI2 was significantly lower compared with that of the nonphosphorylated form. Phosphorylated mature APP showed a stronger association with X11LPI compared with its association with Fe65PI2 (compare right panel with left panel in Fig. 5b). This result indicates that Fe65 interacts preferentially with APP nonphosphorylated at Thr-668; however, the phosphorylated form of APP can still interact with X11L.
Phosphorylation of APP at Thr-668 Causes a Structural Change in the Cytoplasmic Domain of APP-To determine whether phosphorylation of APP at Thr-668 causes a structural change in the cytoplasmic domain of APP, we recorded CD spectra of the APP cytoplasmic domain peptide, APP-(648 -695), with or without phosphate at the Thr-668 residue. As previously reported for experiments performed at 4°C (36), we found that the spectrum of the nonphospho-peptide (Thr) indicated the presence of a random coil structure under conditions of 0% (v/v) TFE at 25°C (Fig. 6a). The phospho-peptide (pThr) also showed an identical spectrum indicating a random coil structure (Fig. 6a). Under different conditions consisting of 47% (v/v) TFE, the CD spectrum of the nonphospho-peptide indicated the presence of an ␣-helix structure (Fig. 6b). The spectrum of the phospho-peptide also indicated an ␣-helix pattern; however, the decrease in ellipticity was less than that for the nonphospho-peptide (Fig. 6b). The percentage of the ␣-helix of the phospho-peptide was 18.76% and that of the nonphosphopeptide was 33.49% under these conditions. This result indicates that, in a hydrophobic environment, the structure of the cytoplasmic domain of APP phosphorylated at Thr-668 is different from the unphosphorylated form.
Effect of the Thr-668 Residue of APP, in the Presence of Fe65, on APP Metabolism and A␤ Release-The interaction of some APP binders with APP is known to modify intracellular APP metabolism and A␤ production (20,24,33,37,38). The effect of the phosphorylation site, Thr-668 of APP, on Fe65-dependent modulation of APP metabolism was analyzed using APP carrying a Glu substitution for Thr-668. We first examined the effect of Fe65 on APP metabolism in HEK293 cells using a pulse-chase study. In HEK293 cells, expression of endogenous Fe65 is under detection level (39). The phosphorylation of immature APP and mature APP was not observed by Western blot analysis (data not shown). When the cells were subject to synchronization at G 2 /M phase of the cell cycle, phosphoryla- The role of phosphorylation of APP695 at Thr-668 in the interaction with the PI2 domain from Fe65 and PI domain from X11L was analyzed with pulldown assays in vitro. a, the expression and phosphorylation state of endogenous APP from rat brain. APP was recovered from rat brain lysate (5 mg of protein) by immunoprecipitation with anti-APP antibody (UT-18), and the total levels of APP and APP phosphorylated at Thr-668 (pAPP) were determined by Western blot analysis with UT-18 and the anti-phospho-Thr-668 antibody, UT-33, respectively. mAPP, mature APP695 subjected to differential O-glycosylation; imAPP, immature APP695. b, rat brain extract containing phosphorylated mAPP, nonphosphorylated mature APP695 subjected to differential O-glycosylation and nonphosphorylated immature APP695 was incubated with GST-Fe65 (Fe65PI2), GST-X11LPI (X11LPI), or GST alone (GST) coupled to glutathione-beads. The amount of APP that was bound to the beads was determined by Western blot analysis with UT-18 (APP) and UT-33 (pAPP). tion of immature APP but not mature APP was detected as reported (11,12). Wild type (T) or the Thr-668Glu (E) mutant of APP was expressed transiently with or without Fe65 in HEK293 cells. The expression of Fe65 does not affect the level of APP expression. The cells were labeled metabolically and followed by a chase period as indicated. Immature APP and mature APP in cell lysates and secreted APP in medium were isolated by immunoprecipitation and quantified, and their relative ratios with respect to total APP levels were calculated (Fig. 7). At 0 h of chase time, almost all the APP was in the immature APP form, and mature APP was almost undetectable (indicated as 1.0 for immature APP plus mature APP at 0 h). The levels of immature APP decreased gradually with time, whereas mature APP levels increased initially but then decreased because of secretion during the chase period. APP mutated at Thr-668 (Fig. 7, a and b) behaved similarly. This indicated that maturation and secretion were not influenced by the mutation of Thr-668 in the absence of Fe65.
Expression of Fe65 slightly delayed the decrease in levels of wild-type immature APP, which indicates that Fe65 tends to slow the maturation of wild-type APP (Fig. 7d). The slight suppression of APP maturation by expression of Fe65 resulted in a slight decrease in the relative levels of mature APP (Fig.  7e). These effects were weakened if Thr-668 was changed to Glu-668 (Fig. 7, g and h). These results suggest that Fe65 delays the maturation of APP by binding to the cytoplasmic domain of immature APP. The mutation of Thr-668 does not affect the secretion of secreted APP in the absence of Fe65 (Fig.  7c). Secreted APP from wild-type APP (T) was suppressed by the expression of Fe65 (Fig. 7f). The suppression of secreted APP secretion by Fe65 expression was also cancelled partially in T668E mutation of APP (Fig. 7i). The mean with S.D. from duplicate assays is shown (n ϭ 2), and consistent results were reproduced in two independent experiments. The autoradiograms of cellular APP are shown (Fig. 7, j and k). It is possible that the mutation of APP at Thr-668 weakened the association of Fe65 with APP and partially cancelled the effect of Fe65 on APP metabolism. The mutation at Thr-668 does not completely inhibit the association with Fe65 ( Figs. 1 and 2). This may be a reason why the mutation did not cancel the Fe65-dependent modification completely (Fig. 7, g-i).
Fe65 regulates mature APP metabolism, which influences the production and secretion of A␤. Wild-type APP695 or a mutant APP695 carrying a substitution of Glu for Thr-668 (T668E) was expressed transiently in HEK293 cells with or without Fe65. The amount of A␤ in the medium was quantified using sandwich enzyme-linked immunosorbent assay (Fig. 8). The secretion of A␤40 and A␤42 from cells expressing wild-type APP was decreased by the presence of Fe65 (Fig. 8a). The suppression of A␤40 and A␤42 production by Fe65 was prevented in part by the T668E mutation of APP (Fig. 8b). Secretion of A␤40 and A␤42 from cells expressing the wild-type and mutant APP is identical in the absence of Fe65 (Fig. 8c). These results indicate that Fe65 suppresses the production of A␤40 and A␤42 in HEK293 cells and that the mutation of APP at Thr-668 prevents this effect by interfering with the interaction of Fe65 with the cytoplasmic domain of APP. DISCUSSION APP and presenilin are suspected causative factors in the pathogenesis of familial AD. Mutations of these genes have been detected in patients with familial AD; however, familial AD makes up only a minority of all cases of AD. These mutations are known to increase production of A␤, which is thought to be the first step in the pathogenesis of AD (reviewed in Refs.  1 and 7). The majority of AD cases are of the sporadic type, and these patients do not carry mutations of these causative genes. Other mechanisms such as prevention of A␤ degradation and/or acceleration of A␤ aggregation, as well as the stimulation of A␤ production, are also believed to be involved in the pathogenesis of sporadic AD.
The cytoplasmic domain of APP contains several motifs and amino acid signals associated with the metabolism and function of APP (8,10,16,17,19). One of these is the Thr-668 residue of APP695. Phosphorylation of this residue in mature APP occurs only in the brain, even though APP is expressed in many tissues, and is believed to be mediated by neuronal protein kinase cyclin-dependent kinase 5 (10), glycogen synthase kinase 3␤ (13), and/or stress-activated protein kinase (14). Therefore, the neuron-specific phosphorylation of APP is thought to play an important role in its function and/or metabolism. It is thought that an intracellular signal may regulate the phosphorylation of APP at Thr-668, because phosphorylation occurs independently from its extracellular domain. 3 We have found that the phosphorylation of APP at Thr-668 is important for the extension of neurites of differentiating PC12 cells after nerve growth factor stimuli (15). The phosphorylation of APP at Thr-668 is also observed in the immature APP of dividing cells, but the detection by Western blot with the phosphorylation state-specific antibody is difficult if cells are not synchronized at the G 2 /M phase of the cell cycle or in nonneuronal tissues, because population of cells in the G 2 /M phase is very minor, and the term of the G 2 /M phase in the cell cycle is extremely limited (11,12). The role of this transient phosphorylation has not been elucidated. In neurons, it was not clear whether the phosphorylation of mature APP affects the regulation of APP metabolism including A␤ production.
It is known that cytoplasmic proteins interact with motifs in the cytoplasmic domain of APP (20 -27  38). However, the molecular mechanism involving the interaction of the cytoplasmic domain of APP with the APP binders has not been well analyzed. The present study demonstrated that for stable interaction of APP with Fe65, both the Thr-668 residue and the Fe65-recognition motif, 682 YENPTY 687 , of APP need to be present. Furthermore, we found that the phosphorylation of APP at Thr-668 weakens the interaction of Fe65 with APP and thought that it results in a lack of association of Fe65 with the 682 YENPTY 687 motif of APP. This is the first demonstration of the mechanism of how the binding of the APP binders to the cytoplasmic domain of APP is regulated.
In previous reports, overexpression of Fe65 enhanced the secretion of A␤ in Madine-Darby canine kidney cells, and overexpression of Fe65L stimulated the maturation of APP and secretion of sAPP in H4 neuroglioma cell lines (37,38). In HEK293 cells, however, we found that overexpression of Fe65 stabilized immature APP and slowed down the maturation of APP slightly, which resulted in lower levels of mature APP and suppressed secretion of secreted APP. This Fe65-dependent effect on APP maturation was prevented by the introduction of a mutation at Thr-668. Immature APP is phosphorylated transiently at the G 2 /M phase of the cell cycle in nonneuronal dividing cells (11,12), whereas immature APP is not phosphorylated in post-mitotic neuronal cells (10). Thus, it is possible to postulate that the phosphorylation of APP at Thr-668 regulates the interaction of Fe65 with mature APP rather than that with immature APP to control APP metabolism in neurons.
We further found that secretion of A␤40 and A␤42 was suppressed by overexpression of Fe65 and that this effect was partially cancelled by mutation of APP at Thr-668. The ability of Fe65 to suppress A␤ generation in HEK293 cells contradicts previous reports in which Fe65 and/or Fe65L stimulated APP maturation, sAPP production, and A␤ secretion in Madine-Darby canine kidney and H4 neuroglioma cells (37,38). The difference may depend on the cell types used. Despite these contradictions, our results obtained with HEK293 cells are consistent with our preliminary observation that, in primary cultured neurons from rodents that expresses human APP, inhibition of phosphorylation of APP suppressed the release of A␤ derived from wild-type APP but not from APP that carried an Ala substitution for Thr-668. 3 This suggests that in neurons, Fe65 can suppress the release of A␤ through its interaction with the cytoplasmic domain of APP. Therefore, we strongly believe that Fe65 regulates the generation of A␤ in an APP phosphorylation-dependent manner in neurons.
Fe65 requires the last 30 amino acids of the cytoplasmic domain of APP that includes a recognition motif, 682 YENPT-Y 687 , for interaction (35). The Thr-668 residue is located 14 amino acids toward the amino-terminal end from the recognition motif, 682 YENPTY 687 . It is important to elucidate how phosphorylation of Thr-668 influences the binding of Fe65 to the 682 YENPTY 687 motif to understand how Fe65 regulates A␤ generation. Recent analysis with multidimensional NMR spectroscopy demonstrated that the amino acid sequence around Thr-668 consists of a type I ␤-turn and an amino-terminal helix-capping box structure (29). Phosphorylation of Thr-668 may disturb this conformation and alter the structure of its carboxyl-terminal sequence including the recognition motif of Fe65. In the present study, the peptides 667 VTPEER 672 containing Thr-668 or phospho-Thr-668 but missing the 682 YENT-PY 687 motif did not compete with the association of Fe65 and APP. This result supports the idea described above that the phosphorylation site Thr-668 is a regulatory site but not a binding site for Fe65.
The amino-terminal helix-capping box is reported to influence the stability of following ␣-helix structures in proteins (40). The 667 VTPEER 672 motif of APP is followed by a nascent helical region (29), and it is possible that phosphorylation at Thr-668 causes some structural change not only in the 667 VT-PEER 672 motif but also in the following nascent helical region. In the hydrophobic environment provided by 47% (v/v) TFE, there was a clear difference in the molar ellipticity of the CD spectrum between the phospho-and nonphospho-cytoplasmic peptides of APP. The percentage of ␣-helix state of the phosphopeptide was less than that of nonphospho-peptide. This result indicates that the phosphorylation state of Thr-668 influences the structure of the cytoplasmic domain of APP in response to the environment. Recent NMR analysis suggest that, in general, regions such as the helix-capping box and the nascent helix region are relatively unstable structures common to early states in the protein folding process (29). Changes to the conformational structure of the cytoplasmic domain of APP in response to the solute environment suggest that the interaction FIG. 8. Effect of Fe65 expression and mutation of APP at Thr-668 on A␤ generation. HEK293 cells were transiently transfected with pcDNA3-FLAGAPP695 or pcDNA3-FLAGAPP695T668E in the presence or absence of pcDNA3.1-Fe65myc/HisA. The culture medium was collected and quantified for A␤40 and A␤42 using sandwich enzyme-linked immunosorbent assay. a, A␤ production from wild-type APP695 (wild type) in the presence (ϩFe65) or absence (ϪFe65) of Fe65 overexpression. b, A␤ production from wild-type APP695 (wild type) or APP695 containing a glutamic acid substitution for Thr-668 (T668E) in the presence of Fe65 overexpression. c, A␤ production from wild-type APP695 (wild type) or APP695 containing a glutamic acid substitution for Thr-668 (T668E) in the absence of Fe65 overexpression. The concentrations of A␤40 and A␤42 are presented as the mean with S.D. (n ϭ 9). Asterisks indicate statistical significance by standard t test (*, p Ͻ 0.005; **, p Ͻ 0.001; ***, p Ͻ 0.00001). of APP binders with the cytoplasmic domain of APP may provide a hydrophobic environment. Other APP binders such as X11L and mDab1, which bind to APP regardless of the phosphorylation state of Thr-668, may also contribute to such a hydrophobic environment. Furthermore, while our manuscript was being reviewed, a report showing that phosphorylation of Thr-668 in the cytoplasmic domain of APP changes its structure was published (41). This observation using NMR analysis strongly supports our idea.
In neurons, mature APP is phosphorylated specifically at Thr-668 (10). We propose that phosphorylation of mature APP regulates the interaction of Fe65 with the cytoplasmic domain of APP and control the metabolism of mature APP in neurons. The production of A␤40 and A␤42 is believed to occur in association with a protein secretory pathway either within or following the trans-Golgi network (reviewed in Ref. 1). Within these compartments, APP is phosphorylated, and O-glycan residues are added (10). The identification of the intracellular location at which APP is phosphorylated and A␤ is generated may support our idea that the generation of A␤ is regulated by controlling the association of Fe65 with the cytoplasmic domain of APP through the phosphorylation of APP. If the phosphorylation level of APP at Thr-668 increases in the brains of patients with AD, less Fe65 will associate with the cytoplasmic domain of APP, which will result in changes to the rate of A␤ production. The potential to control the phosphorylation of APP at Thr-668, with its subsequent influence on the interaction of Fe65 with APP, may provide a useful treatment to suppress A␤ production in AD.