Novel cadherin-related membrane proteins, Alcadeins, enhance the X11-like protein-mediated stabilization of amyloid beta-protein precursor metabolism.

Previously we found that X11-like protein (X11L) associates with amyloid beta-protein precursor (APP). X11L stabilizes APP metabolism and suppresses the secretion of the amyloid beta-protein (Abeta) that are the pathogenic agents of Alzheimer's disease (AD). Here we found that Alcadein (Alc), a novel membrane protein family that contains cadherin motifs and originally reported as calsyntenins, also interacted with X11L. Alc was abundant in the brain and occurred in the same areas of the brain as X11L. X11L could simultaneously associate with APP and Alc, resulting in the formation of a tripartite complex in brain. The tripartite complex stabilized intracellular APP metabolism and enhanced the X11L-mediated suppression of Abeta secretion that is due to the retardation of intracellular APP maturation. X11L and Alc also formed another complex with C99, a carboxyl-terminal fragment of APP cleaved at the beta-site (CTFbeta). The formation of the Alc.X11L.C99 complex inhibited the interaction of C99 with presenilin, which strongly suppressed the gamma-cleavage of C99. In AD patient brains, Alc and APP were particularly colocalized in dystrophic neurites in senile plaques. Deficiencies in the X11L-mediated interaction between Alc and APP and/or CTFbeta enhanced the production of Abeta, which may be related to the development or progression of AD.

homologues of LIN-10 in Caenorhabditis elegans (18,19) and dX11L in Drosophila melanogaster (20). They are neuron-specific adaptor proteins composed of a large amino-terminal region, a central phosphotyrosine interaction (PI) domain, and two carboxyl-terminal PDZ domains. The 681 GYENPTY 687 motif of APPcyt (numbering is based on the APP695 isoform) interacts with the PI domain of X11L (14).
To understand how APP metabolism is regulated, including A␤ generation, it is of interest to isolate other proteins that interact with X11L. This will help elucidate the molecular mechanism by which X11L regulates APP metabolism.
In the present study, we isolated novel human cDNAs encoding Alcadeins (Alcs), which are type I transmembrane proteins that interact with the PI domain of X11L. These genes are conserved in a wide variety of species, including D. melanogaster and C. elegans. We found that the cytoplasmic domain of Alc bound to the PI domain of X11L and initiated the formation of a tripartite complex comprised of Alc, X11L, and APP. This complex stabilized intracellular APP metabolism and significantly suppressed A␤ production by slowing APP maturation. We found that Alc and X11L also formed a tripartite complex with C99, a CTF␤. This complex inhibited the interaction of C99 with PS1, a component of ␥-secretase that generates A␤ from C99. We could recover endogenous X11L⅐APP⅐Alc tripartite complexes from normal mouse brains, and normal mouse neurons displayed remarkable colocalization of these proteins. In AD brains, APP was found to colocalize with Alc in the dystrophic neurites of senile plaques. These observations indicate that APP exists in protein complexes composed of X11L and Alc that regulate APP metabolism, including A␤ production in neurons, and that this regulatory mechanism may be perturbed in AD.

MATERIALS AND METHODS
cDNA Cloning of Human Alc␣1 and Plasmid Construction of Alc␣ Family cDNAs-The yeast two-hybrid system used in this study has been described previously (14). Briefly, the human brain MATCH-MAKER cDNA library (Clontech) was screened with bait composed of cDNA encoding part of the human X11L (hX11L) protein (amino acids 129 -555). This yielded a cDNA clone (X11L-binding protein clone number 31 (XB31␣)) that contained a partial open reading frame of Alc␣1. Using this clone, a clone encoding the full-length Alc␣1 protein was isolated from a human brain gt11 cDNA library (Clontech). This cDNA was designated as human Alc␣1 (GenBank TM accession number AF438482, registered as XB31␣). The full-length hAlc␣1 open reading frame was subcloned into the HindIII and XbaI sites of pcDNA3 (Invitrogen). We identified two similar cDNA clones in the HUGE (Human Unidentified Gene-Encoded Large Proteins data base analyzed by the KAZUSA cDNA Project) Protein Database (21). The HUGE cDNA clones were denoted as human Alc␣2 (accession number KIAA0911) and Alc␤ (accession number KIAA0726). We also found human Alc␥ (accession number NM022131) in the GenBank TM /EBI Data Bank. Alc␤ cDNA, a generous gift from Dr. Nagase (KAZUSA DNA Research Institute, Chiba, Japan), was recloned into pcDNA3 at HindIII and XbaI sites to produce pcDNA3-hAlc␤. The human Alc␣1 and Alc␤ cDNAs were also recloned via an XbaI site into pcDNA3 with a FLAG tag on the amino-terminal end of the insert so that pcDNA3-FLAG-hAlc␣1 and pcDNA3-FLAG-hAlc␤ could be produced.
Protein Interaction Assays in Yeast-The MATCHMAKER two-hybrid system (Clontech) with pGBT9 and pGAD424 was used as described previously (22). ␤-Galactosidase activity in the yeast two-hybrid system was measured in a liquid assay using o-nitrophenyl galactopyranoside and was expressed in Miller units.
Coimmunoprecipitation and Western Blot Assays-COS7 and human embryonic kidney 293 (HEK293) cells (ϳ1 ϫ 10 7 cells) were transfected as described previously (14) with the indicated amounts of various combinations of the pcDNA3-hX11L, pcDNA3-hAlc␣1, pcDNA3-hAlc␤, pcDNA3APP695, and pcDNA3-PS1 plasmids. In another experiment, pcDNA3.1-HA-hX11, in which human X11 cDNA with an attached HA tag at the 5Ј-end was inserted into the NheI and EcoRV sites of pcDNA3.1, was transfected instead of pcDNA3-hX11L. pcDNA3-FLAG-hX11L and pcDNA3-FLAG-hX11, in which human X11L and X11 cDNA with an attached FLAG tag at the 5Ј-end, were also used in coimmunoprecipitation assay. Cells were harvested and lysed for 1 h on ice in CHAPS buffer, which consists of phosphatebuffered saline (PBS; 140 mM NaCl and 10 mM sodium phosphate (pH 7.4)) containing 10 mM CHAPS, 5 g/ml chymostatin, 5 g/ml leupeptin, 5 g/ml pepstatin A, 1 mM Na 3 VO 4 , and 1 mM NaF. After centrifugation at 12,000 ϫ g for 10 min at 4°C, affinity-purified antibodies (indicated IgG amounts) or serum was added to the CHAPS lysate supernatant. The immunoprecipitates were then subjected to Western blot analysis using specific antibodies.
Subcellular Fractionation of Mouse Brain Proteins-The subcellular fractionation of mouse brains was performed as described previously (24,25). Briefly, five brains of 8-week-old male C57BL/6 mice were homogenized on ice in 5 volumes of buffer (10 mM HEPES (pH 7.4), 0.32 M sucrose, 5 g/ml chymostatin, 5 g/ml leupeptin, and 5 g/ml pepstatin) by 10 strokes of a loose-fitting (0.12-m clearance) Teflon homogenizer. The postnuclei supernatants were further centrifuged at 100,000 ϫ g for 1 h. The resulting precipitate (membrane fraction) was resuspended and separated by centrifugation for 115 min at 41,000 rpm in a Beckman SW41 rotor at 4°C on a 0-28% (w/v) iodixanol density gradient. The supernatants were then fractionated into 13 tubes (0.8 ml/tube), and the fractions (20 l) were analyzed by Western blot analysis with the indicated antibodies. Fraction 8 was solubilized by adding an equal amount of 2ϫ CHAPS buffer consisting of 2ϫ PBS (280 mM NaCl and 20 mM sodium phosphate (pH 7.4)) plus 20 mM CHAPS. The mixture was allowed to rotate for 30 min at 4°C and then centrifuged at 100,000 ϫ g for 60 min. The resulting supernatant was used in the coimmunoprecipitation analysis of endogenous proteins.
Pulse-Chase Study-Pulse-chase labeling of cells was performed with L-[ 35 S] in vitro Cell Labeling Mix (0.4 mCi/ml; AGQ 0080, Amersham Biosciences). HEK293 cells were transfected with the indicated combinations of the pcDNA3APP695, pcDNA3-hX11L, pcDNA3-FLAG-hAlc␣1, and pcDNA3 vector plasmids (transfection was performed with 9 g of plasmids in total). After a 48-h transfection, the cells were metabolically labeled for 15 min. This was followed by a 0 -8-h chase period, which was initiated by replacing the labeling medium containing excess amounts of unlabeled methionine. The cells were then lysed and subjected to immunoprecipitation with G369 (anti-APP cytoplasmic domain antibody) and protein G-Sepharose. The recovered APP was separated by SDS-PAGE (6% (w/v) polyacrylamide) and analyzed by Fuji BAS 2000 imaging analyzer.
Quantification of ␤-Amyloid-C99 protein was used as an intracellular ␥-secretase substrate (26). The construct containing Kozak sequence, first methionine, signal peptide sequence of APP, and CTF␤ coding region was cloned into pcDNA3.1(ϩ) at NheI and ApaI sites and designated as pcDNA3.1APP-C99. The C99 protein is expressed as an artificial CTF␤ once its signal sequence is cleaved off by signal peptidase (26).
Neuro-2a cells (ϳ2 ϫ 10 6 cells) were transiently transfected with the indicated combinations of pcDNA3APP695 or pcDNA3.1APP-C99, pcDNA3-hX11L, pcDNA3-FLAG-hAlc␣1, and pcDNA3 vector plasmids (transfection was performed with 9 g of plasmids in total). Cells were supplied with fresh growth medium 5 h after the start of transfection, and conditioned medium was collected 48 h after the medium replacement. A␤40 and A␤42 were quantified with sandwich ELISA by using three types of A␤-specific monoclonal antibodies (27). Intracellular A␤40 and A␤42 were also extracted (28) and quantified by the sandwich ELISA. Briefly, the cells were lysed by sonication in 40 l of PBS containing 6 M guanidine chloride and centrifuged at 20,000 ϫ g for 15 min at 4°C. The resulting supernatant was diluted up to 12-fold by adding PBS and used in the ELISA. The single factor analysis of variance test followed by Tukey multiple comparisons was used to analyze differences among groups of data. Data were presented as means Ϯ S.E.
Immunological Staining of Mouse Brain Sections-Experimental procedures were conducted in compliance with the guidelines of the Animal Studies Committee of Hokkaido University. Adult C57BL/6 mice (6 weeks, male) were perfused at 4°C for 20 min with 4% (w/v) paraformaldehyde in 0.2 M sodium phosphate buffer, pH 7.5. The brains were excised and postfixed with the same fixative at 4°C overnight followed by treatment with 30% (w/v) sucrose in PBS for 2-3 days at 4°C. Brains were embedded into OCT compound (Miles Scientific), and frozen sagittal sections (20 m) were prepared. The sections were incubated with 0.1% (v/v) Triton X-100 in PBS for 5 min at room temperature followed by treatment with 0.3% (v/v) H 2 O 2 in PBS for 5 min at room temperature to quench endogenous peroxidase activity. The sections were blocked with 3% (w/v) bovine serum albumin in PBS for 10 min at room temperature and incubated for 3 h at 4°C with antibodies in 1% (w/v) bovine serum albumin in PBS. After washing, the sections were further incubated with goat anti-mouse IgG coupled with Alexa Fluor 488 or goat anti-rabbit IgG coupled with Alexa Fluor 568 in 1% (w/v) bovine serum albumin for 1 h at room temperature. Sections were viewed using the confocal laser scanning microscope LSM510 (Carl Zeiss).
Immunological Staining of AD Brains-We used paraffin sections of frontal and temporal cortices from five AD subjects. Tissues were fixed with Kryofix (a mixture of ethanol, polyethylene glycol, and water; Merck) for 1-7 days and embedded in paraffin. Dewaxed serial tissue sections (cut into 4-m sections) were immunostained with the ABC elite kit (Vector Laboratories, Burlingame, CA). Some sections were pretreated with a microwave antigen retrieval method for 10 min in 10 mM citrate buffer, pH 6.0 (29). Sections were incubated with antibodies specific for Alc␣ (UT83, 0.8 g/ml), APP (22C11, 0.5 g/ml), and A␤ (4G8, 1:1000 dilution). The peroxidase activity was visualized with diaminobenzidine-H 2 O 2 solution. For control analyses, tissue sections were incubated with anti-Alc␣ antibody UT83 in the presence of the antigen peptide (40 nM) or non-immune rabbit IgG (0.8 g/ml).
For double immunofluorescence analyses, dewaxed sections were incubated with a mixture of UT83 and 22C11 or of UT83 and 4G8 at the same dilutions as above followed by incubation with a mixture of fluorescein isothiocyanate-tagged goat anti-rabbit IgG (Jackson Immu-noResearch Laboratories, West Grove, PA; 1:30 dilution) and Cy3tagged goat anti-mouse IgG (Jackson ImmunoResearch Laboratories, 1:50 dilution). Prior to immunolabeling, the autofluorescence of the lipofuscin granules was blocked with Sudan black B staining (30).

Isolation, Identification, and Characterization of Novel Cadherin-related Membrane Protein
Genes-Yeast two-hybrid screening of a human brain cDNA library with cDNA encoding a part of the hX11L protein resulted in the isolation of a novel cDNA denoted XB31␣ (GenBank TM accession number AF438482). The cDNA encoded a type I transmembrane protein composed of 971 amino acids. In this paper, this protein has been entitled human Alcadein (Alzheimer-related cadherin-like protein) ␣1 or Alc␣1 (Fig. 1A). A thorough search of the human cDNA data base and genome data bases revealed three similar genes. These encode 981-, 968-, and 955-amino acid proteins, and we denoted these proteins as Alc␣2, Alc␤, and Alc␥, respectively (Fig. 1A). Alc␣2 is identical to human Alc␣1 except for 10 additional amino acids between the Alc␣1 residues at positions 71 and 72 (Fig. 1A, pink box). This extra sequence in Alc␣2 is derived from one exon, and both proteins are spliced variants of the XB31␣ (Alc␣) gene. The Alc␤ and Alc␥ proteins are both ϳ50% homologous to Alc␣. They are encoded by different genes and belong to the same gene family to which Alc␣ belongs. Further data base examinations revealed that there are Alc-like genes in D. melanogaster (Gen-Bank TM accession number AAF59384) and C. elegans (Gen-Bank TM accession number NP495189) as well.
The Alc family members contain two cadherin motifs, a putative Ca 2ϩ -binding sequence in their amino-terminal halves, and a single cytoplasmic domain composed of ϳ110 amino acids (Fig. 1A). This cytoplasmic domain includes the X11L-binding site (see Fig. 2), which is highly conserved among different species, and an acidic region at the carboxyl-terminal end of the X11L-binding site (Fig. 1, A and B).
Expression and Distribution of Alc␣-X11L is specifically expressed in neural tissue (14). We examined which human tissues express Alc␣ and Alc␤ by Northern blot analysis, and we found strong expression of the ϳ5-kb Alc␣ transcript in the brain and weaker expression in the heart, while a probe specific for Alc␤ revealed high levels of a 4.4-kb transcript in the brain (data not shown, please see Supplemental Fig. 1).
Expression of Alc protein in mouse brain tissue was examined by Western blot analysis. The anti-Alc␣ antibody UT83 recognizes two endogenous proteins in mouse brain lysates (Fig. 1C) as well as recombinant human Alc␣1 but not recombinant Alc␤ (data not shown, please see Supplemental Fig. 2). The antigen peptide to which UT83 was raised competed with the endogenous proteins for binding to UT83 (Fig. 1C, ϩ). Thus, the UT83 antibody specifically recognizes two endogenous Alc␣ proteins in the mouse brain. When UT83 was used to analyze Alc␣ expression in a variety of murine tissues, the highest levels were found in the brain, and small amounts were found in the lung (Fig. 1D, upper panel). Assessment of X11L expression with the UT30 X11L-specific antibody confirmed its brainspecific distribution (Fig. 1D, upper panel). These observations suggest that, like X11L, Alc␣ is abundant in neurons. This colocalization in the brain supports the notion that X11L and Alc␣ interact physiologically. This idea was further supported when we examined Alc␣ and X11L expression in various parts of the brain by dissecting adult mouse brains into various component regions and blotting them with UT83 and UT30. Relatively high levels of Alc␣ were detected in the cerebral cortex, striatum, hippocampus, and thalamus; moderate levels were found in the cerebellum; and low levels were observed in the olfactory bulb, midbrain, and pons ( Fig. 1D, lower panel). Expression in the sciatic nerve fiber was not observed. X11L was distributed in a very similar pattern (Fig. 1D, lower panel). Thus, X11L and Alc␣ are both largely expressed in neural tissues apart from peripheral nervous system.
Alc and X11L Interaction-We investigated the interactions between Alc and X11L by coimmunoprecipitation assays from COS7 cells expressing X11L together with Alc␣1 or Alc␤ followed by Western blot analysis. The immunoprecipitates resulting from incubating the lysates with anti-Alc␣ (UT83) or anti-Alc␤ (BS7) antibodies included X11L as detected by the X11L antibody (UT30) (data not shown, please see Supplemental Fig. 3). Conversely the immunoprecipitates generated by the anti-X11L antibody contained Alc␣1 and Alc␤. In contrast, non-immune antibody (control IgG) immunoprecipitates did not contain these proteins. Thus, both the Alc␣ and Alc␤ proteins bind to X11L within cells.
We identified the Alc␣-and Alc␤-binding domain of X11L by assessing the in vitro binding of various X11L domain constructs to the putative cytoplasmic domains of Alc␣ and Alc␤. Cell lysates containing X11L protein constructs were incubated with beads coupled to the cytoplasmic domains of human Alc␣ or Alc␤ fused to GST. GST alone was used as a negative control. The pull-downs were subjected to Western blot analysis with the anti-X11L amino-terminal antibody UT29. Full-length X11L and its construct encoding the X11L amino-terminal domain attached to the PI domain (N ϩ PI) bound the GST-Alc␣ and GST-Alc␤ fusion proteins (data not shown, please see Supplemental Fig. 4). However, the amino-terminal domain alone did not bind either protein. GST alone also did not bind to any of the X11L constructs. Thus, as with APP (14), the cytoplasmic domains of Alc␣ and Alc␤ bind the PI domain of X11L.
We also identified the region in the cytoplasmic domain of Alc␣ (Alc␣CTF) that is required for the interaction with X11L by yeast two-hybrid analysis. Truncated Alc␣1CTF constructs lacking the amino-terminal regions (CTF1, ⌬871-902; CTF2, ⌬871-908) or carboxyl-terminal regions (CTF3, ⌬903-971; CTF4, ⌬909 -971) were examined for interaction with the N ϩ PI domain (amino acids 129 -555) of X11L. ␤-Galactosidase activity was measured in a liquid assay and calculated as Miller units. The CTF1 and CTF4 constructs, which have the conserved 904 NPMETY 909 (numbering is based on the hAlc␣1 isoform) sequence in common, bound X11L, but CTF2 and CTF3, which both lack the NPMETY sequence, did not interact with X11L (Fig. 2). The NPXXXY sequence is a modified version of the NPXY motif of the cytoplasmic domain of APP that binds X11L (14). When we introduced mutations into the first two amino acids of the NPMETY sequence generating AAMETY, denoted as NP-AA, Alc␣CTF did not interact with X11L. However, modification of NPMETY to NPMETA (Y-A) had no effect on the ability of the protein to bind to X11L (Fig.  2). Thus, it appears that the first two amino acids of the NPMETY motif are essential for the association of Alc␣ with X11L. This is supported by the mutational analyses of the site used by APP to bind X11 as it was shown that the end tyrosine residue of the NPTY motif of APP is not important for binding to X11 (31). Moreover, the first two Asp and Pro amino acids are conserved in all the Alc proteins, whereas the end Tyr residue is not conserved in Alc␤ or the D. melanogaster and C. elegans Alc proteins (Fig. 1B).
Alc, X11s, and APP Form a Tripartite Complex-The cytoplasmic domains of Alc and APP both bind to the PI domain of X11L. We thus investigated whether Alc and APP compete for binding to X11L or act cooperatively. APP695 (the human APP isoform composed of 695 amino acids) and X11L were expressed together in HEK293 cells in the presence or absence of hAlc␣1 or FLAG-tagged hAlc␤, and APP was immunoprecipitated from cell lysates with the anti-APP cytoplasmic domain antibody G369. The immunoprecipitates were analyzed by Western blot FIG. 1. Comparison of structures and cytoplasmic amino acid sequences of the Alc family proteins and protein expression of Alc␣ and X11L. A, schematic representation of structural features of the human Alc proteins. The pink box in the hAlc␣2 sequence indicates the amino acids derived from an exon of the hAlc␣ gene. Yellow boxes include the predicted cadherin motif, while the black box within the human Alc␤ sequence represents the leucine zipper motif. The green box includes a highly conserved region containing the X11Lbinding sequence. The red box indicates an acidic region. The plasma membrane is indicated as a purple ladder-back structure. The underlines indicate the regions recognized by the specific antibodies UT83 and BS7. The numbers indicate amino acid (a.a.) residues. B, amino acid sequences of the cytoplasmic domains of Alc family proteins. The arrow reveals the NP motif that is the X11L-binding site (blue). Gaps produced by the alignment are indicated by a hyphen in the sequence. The numbers indicate amino acid position. C, detection of Alc proteins in the mouse brain by the UT83 Alc␣ antibody. Mouse brain lysates (50 g of protein) were subjected to SDS-PAGE (6% (w/v) polyacrylamide), transferred to a membrane, and probed with UT83 (0.16 g/ml IgG) in the presence (ϩ) or absence (Ϫ) of the antigen peptide used to raise it (40 nM). UT83 specifically detected two Alc␣ proteins. The numbers refer to the molecular masses (kDa) of the protein standards. D, analysis of the distribution of Alc␣ and X11L in various murine tissues and brain regions by Western blot analysis. Mouse lysates (50 g of protein) of tissues (upper panel) and brain regions (lower panel) were subjected to SDS-PAGE (6% (w/v) polyacrylamide), transferred to the membrane, and probed with the anti-Alc␣ (UT83) and anti-X11L (UT30) antibodies. Br, brain; Ht, heart; Lu, lung; Li, liver; Kid, kidney; Mus, muscle; OB, olfactory bulb, CC, cerebral cortex; ST, striatum; Hip, hippocampus; Ce, cerebellum; Mid, mid brain; Th, thalamus; Sci, sciatic nerve. analysis with anti-APP (G369), anti-X11L (UT29), anti-Alc (UT83), and anti-FLAG (M2) antibodies. As reported previously (14), X11L was coimmunoprecipitated with the anti-APP antibody in the absence of Alc. Surprisingly, in the presence of Alc␣1 or FLAG-Alc␤, higher levels of X11L were recovered in the APP immunoprecipitates (Fig. 3A), and the two Alc proteins were found to be bound to X11L⅐APP complexes (Fig. 3A). However, when APP and FLAG-Alc␣1 were coexpressed in the absence of X11L, the G369 antibody did not coimmunoprecipitate Alc. The identical result was obtained when 22C11 anti-APP extracellular domain antibody was used (data not shown, please see Supplemental Fig. 5). Moreover M2 antibody did not coimmunoprecipitate APP (Fig. 3B). Notably, in the cells that coexpress X11L and APP, the levels of immature APP (lower band of APP, N-glycosylated form) were increased (Fig. 3, A  and B). This effect is related to the fact that X11L stabilizes APP metabolism (see Fig. 5). These findings clearly indicate that APP and Alc form a complex by interacting with X11L and that the binding of APP and X11L is greatly stabilized by Alc.
We examined whether X11 is also able to mediate the formation of a tripartite APP⅐X11⅐Alc complex. Thus, Alc␣1, HAtagged X11, and APP695 were expressed in COS7 cells. When Alc␣1 was immunoprecipitated with the anti-Alc␣ antibody in the presence of X11, APP was coprecipitated (Fig. 3C). Moreover, when APP was immunoprecipitated with the anti-APP antibody 22C11, Alc␣ was recovered along with X11. The binding of APP and X11 is greatly stabilized by Alc as is X11L. Thus, both X11 and X11L can contribute to formation of a tripartite complex comprised of APP⅐X11⅐Alc or APP⅐X11L⅐Alc.
It is usually documented that the PI domain does not interact at the same time with two proteins possessing a similar interaction motif (32). Nevertheless the stoichiometry of X11L bound to APP seems to be increased by addition of Alc (Fig. 3A), but is the APP⅐X11s⅐Alc complex stoichiometry 1:1:1 or 1:2 (or more):1? Thus, we performed coimmunoprecipitation of APP and Alc with the M2 antibody from cells expressing APP, Alc␣1, and FLAG-X11L or FLAG-X11 (Fig. 3D). If the stoichiometry of X11L bound to APP is increased by Alc, the amount of APP recovered by coimmunoprecipitation with anti-FLAG antibody is expected to decrease in the presence of Alc␣1. However, FLAG-X11L and FLAG-X11 coprecipitated more APP in the presence of Alc␣1 than in the absence of Alc␣1. This result suggests that the stoichiometry of X11s bound to APP does not increase by the addition of Alc. The most simple explanation is that the population of X11s bound to APP is increased in cells by the expression of Alc. Alc may increase the chance that APP will interact with X11L or decrease the dissociation constant of the APP⅐X11L complex. Furthermore our preliminary result derived from studies to determine APP-and Alc-binding sites by introducing amino acid substitution in the PI domain of X11L supports the interaction of APP and Alc with different sites of the single PI domain of X11L. 2 However, we cannot rule out the possibility that more than one X11L molecule is bound to APP because both X11s could form homoand/or heterodimers when the respective proteins were overexpressed in cells. 3 We examined whether APP, Alc, and X11L form similar tripartite complexes in the brain. When we separated the postnuclear fraction from the brains of five mice into cytoplasmic and membrane fractions and subjected both fractions to immunoprecipitation assays, we found that APP and Alc␣ (which are both membrane proteins) were not recovered in the cytoplasmic fraction (data not shown). However, the majority of X11L was recovered in the cytoplasmic fraction, although a moderate amount of X11L was also recovered in the membrane fraction (data not shown). To examine formation of these proteins into tripartite complexes in the brain, the membrane fraction was further fractionated by iodixanol density gradient centrifugation, and these fractions were subjected to Western blot analysis with antibodies specific for APP, X11L, Alc␣, the endoplasmic reticulum protein protein-disulfide isomerase, the Golgi-resident protein GM-130, the synaptic vesicle protein SYT, the vesicle transport motor protein KHC, and PS1. The vesicles bearing APP, APPCTF, X11L, and Alc␣ all sedimented in the same medium density membranous protein fractions 7-9, which were a bit heavier than the SYT-containing vesicles in fractions 9 and 10 ( Fig. 4A). APP, KHC, and PS1 were observed in the same fractions, confirming earlier observations (33,34). These data suggest that APP, X11L, and Alc␣ could form tripartite complexes on organelles composed of medium density membranes. To test this, the proteins in fraction 8 were solubilized with CHAPS buffer, immunoprecipitated with anti-APP (G369) or non-immune control antibodies, and subjected to Western blot analysis with antibodies specific for APP, X11L, Alc␣, SYT, and rabbit IgG antibodies. Alc␣ and X11L were coimmunoprecipitated by anti-APP but not by non-immune antibody (Fig. 4B). Also SYT was not coimmunoprecipitated by anti-APP antibody. Thus, APP and/or APPCTF, Alc␣, and X11L form a tripartite complex in vivo, probably on medium density membrane organelles that contain cargo proteins including PS and that differ from the synaptic vesicles that contain SYT.
We next investigated whether APP, X11L, and Alc colocalize in neurons. Sagittal sections of the hippocampus from adult mouse brains were double stained with antibodies specific for 2 Y. Araki and T. Suzuki, unpublished observation. 3 A. Sumioka and T. Suzuki, unpublished observation. APP and X11L, Alc␣ and X11L, or APP and Alc␣. The three proteins largely colocalized (Fig. 4C). Strong X11L immunoreactivity was detected in the CA3 pyramidal cell somata and proximal dendrites where APP and Alc␣ were also localized (Fig. 4C). Colocalization was observed in the large pyramidal neurons of the cerebral cortex (data not shown). High power views of the cell bodies of these neurons confirmed the colocalization of APP, Alc␣, and X11L (Fig. 4C). Thus, neurons expressing X11L also express Alc␣ and APP, and the three proteins colocalize in these cells.
Alc Enhances the X11L-mediated Stabilization of APP Metabolism and Suppression of A␤ and sAPP Secretion-X11 stabilizes intracellular APP metabolism (35,36), and we previously reported that X11L suppresses the secretion of A␤40 (14), indicating that X11L also stabilizes APP metabolism. Since Alc and APP form a complex with X11L through cytoplasmic interactions, and this complex formation strengthens the interaction between APP and X11L (Fig. 3), we speculated that Alc may enhance the X11L-mediated stabilization of APP metabolism and thereby further suppress A␤ and sAPP production. To test this, we assessed the intracellular APP metabolism in HEK293 cells that express APP with or without X11L in the presence or absence of Alc␣1 by pulse-chase assay (Fig. 5). The metabolically radiolabeled APP was recovered by immunoprecipitation at the indicated chase periods and separated by SDS-PAGE. The levels of immature APP were quantified by autoradiography and calculated with respect to the 0 h levels (1.0) (Fig. 5B). When APP was expressed alone, the immature APP levels decreased gradually with time due to the maturation of APP and the secretion of large extracellular domain truncated at ␣or ␤-site (sAPP) during the chase period. As expected, X11L coexpression slightly delayed this decrease in immature APP levels, indicating that X11L stabilizes APP metabolism. When Alc␣1 was also coexpressed, the effect of X11L was greatly enhanced. However, the increased stabilization of APP metabolism due to Alc␣1 was not observed if X11L was not coexpressed.
We next investigated the production of A␤ and sAPP in Neuro-2a cells expressing APP with or without X11L in the presence or absence of Alc␣1. The amount of A␤ in medium secreted from the cells was quantified using sandwich ELISA. As we previously demonstrated (14), coexpression of X11L suppresses the secretion of A␤40 (Fig. 5C). X11L did tend to suppress A␤42 secretion, but it was not significant statistically. When X11L and Alc␣1 were coexpressed, the effect of X11L was remarkably enhanced. However, enhanced suppression of A␤40 secretion was not observed if only Alc␣1 was expressed. Alc␣1 and X11L coexpression did not significantly effect A␤42 secretion. Thus, Alc clearly enhances the X11L-mediated inhiwith pcDNA3APP695 (2 g), pcDNA3-HA-hX11 (0.25 g), and pcDNA3-hAlc␣1 (6.75 g) in various combinations. To standardize the plasmid concentration, adequate amounts of pcDNA3 vector (Ϫ) were added (to yield 9 g of plasmid in total). The cell lysates were immunoprecipitated with anti-Alc␣ antibody (UT-83, upper panel) or anti-APP antibody (22C11, middle panel). The crude lysate (Lysate, lower panel) and immunoprecipitates (IP) were then analyzed by Western blotting with anti-Alc␣ (UT83), anti-HA (12CA5), and anti-APP (G369) antibodies. D, Alc stabilizes APP binding to X11L and X11. COS7 cells (ϳ1 ϫ 10 7 cells) were cotransfected with pcDNA3APP695 (2 g), pcDNA3-Alc␣1 (6.75 g), and pcDNA3-FLAG-hX11L or pcDNA3-FLAG-hX11 (0.25 g) in various combinations. To standardize the plasmid concentration, adequate amounts of pcDNA3 vector (Ϫ) were added (to yield 9 g of plasmid in total). The cell lysates were immunoprecipitated with anti-FLAG antibody (M2). The immunoprecipitates were analyzed by Western blotting with anti-FLAG (M2, first panel), anti-APP (G369, second and third panels), and anti-Alc␣ (UT83, fourth panel) antibodies. The third panel shows a longer exposure (exp) of film than that of the second panel.

FIG. 3. Alc enhances APP-X11L binding, and the three proteins form a tripartite complex.
A, role of Alc in the binding of X11L to APP. HEK293 cells (ϳ1 ϫ 10 7 cells) were transiently cotransfected with pcDNA3APP695 (2 g) and pcDNA3-hX11L (0.25 g) with or without pcDNA3-hAlc␣1 (6.75 g) or pcDNA3FLAG-hAlc␤ (6.75 g) as indicated. To standardize the plasmid concentrations, the pcDNA3 vector (Ϫ) was added (to yield 9 g of plasmid in total). The cells were lysed, and APP in the cell lysate was recovered by immunoprecipitation with the anti-APP cytoplasmic domain antibody G369. The immunoprecipitates were then analyzed by Western blotting with anti-APP (G369), anti-X11L (UT29), anti-Alc␣ (UT83), and anti-FLAG (M2) antibodies. B, role of X11L in the formation of the APP⅐X11L⅐Alc tripartite complex. HEK293 cells (ϳ1 ϫ 10 7 cells) were transiently cotransfected with pcDNA3APP695 (3 g) and pcDNA3-FLAG-hAlc␣1 (3 g) in the presence or absence of pcDNA3-hX11L (3 g). To standardize the plasmid amounts, 3 g of pcDNA3 vector was added in the absence of pcDNA3-hX11L (Ϫ). The cell lysates were immunoprecipitated with the anti-APP antibody G369 (upper three panels) or with the anti-FLAG antibody M2 (lower three panels). The immunoprecipitates were analyzed by Western blot analysis with anti-APP (G369), anti-X11L (UT29), and anti-FLAG (M2) antibodies. C, X11 can complex with APP and Alc␣1. COS7 cells (ϳ1 ϫ 10 7 cells) were cotransfected bition of A␤, at least A␤40, secretion from APP, which correlates with the finding that Alc stabilizes the interaction between X11L and APP (Fig. 3A). The amount of sAPP in the medium secreted from the cells was detected by Western blot analysis with anti-APP extracellular domain antibody 22C11 (Fig. 5D). X11L suppressed sAPP secretion, and Alc␣1 en-hanced this effect. It is likely that the decreased A␤40 and sAPP secretion by cells expressing APP, X11L, and Alc is due to the slow-down of intracellular APP maturation caused by their tripartite complex formation. Because we could not quantify the A␤ levels in the cell lysates, we performed another study with cells expressing C99 protein (see Fig. 6).  ). B, presence of APP⅐X11L⅐Alc tripartite complexes in the brain. Fraction 8 in A was lysed, and APP was recovered by immunoprecipitation with anti-APP (G369) antibody. As a control, non-immune antibody was used instead of G369. Crude lysate (Lysate) and the immunoprecipitates (IP) were subjected to Western blot analysis with antibodies specific for APP and APPCTF (G369), X11L (mint2), Alc␣ (UT83), SYT (clone no. 41), and rabbit IgG antibodies. (H) indicates heavy chain of IgG. C, APP, Alc␣, and X11L colocalize in mouse neurons. Sagittal sections of an adult mouse brain were double stained with antibodies against APP (22C11, red) and X11L (mint2, green) (panels 1-3), Alc␣ (UT83, red) and X11L (mint2, green) (panels 4 -6), or APP (LN27, red) and Alc␣ (UT83, green) (panels 7-9). The merging of the signals (yellow) indicates the colocalization of the two proteins being assessed (panels 3, 6, and 9). CA3 region of the hippocampus is shown as are high power views of the cell bodies of the large pyramidal neurons of the cerebral cortex. The scale bar indicates 10 m.

X11L and Alc Form a Tripartite Complex with C99 and
Suppress the ␥-Cleavage of C99 -As we have shown in Fig. 4, APP not only colocalized with X11L and Alc␣ in the brain, the tripartite complex also colocalized with PS1. This suggests that X11L and Alc␣ could suppress the ␥-cleavage of the CTF of APP. We investigated this possibility by cotransfecting Neuro-2a cells with C99 instead of full-length APP with or without X11L and in the presence or absence of Alc␣1. The stability of C99 was examined by Western blot analysis, which revealed that X11L coexpression caused the intracellular C99 to accumulate (Fig. 6A). This suggests that X11L protects the CTF from ␥-cleavage. This effect was enhanced when Alc␣1 was also coexpressed (Fig. 6A). However, the stabilization of intracellular C99 was not observed if Alc␣ was coexpressed in the absence of X11L.
When the A␤ in the medium from the transfected Neuro-2a cells described above was quantified, it was found that X11L coexpression with C99 suppresses the secretion of both A␤40 and A␤42 into medium (Fig. 6B). Moreover, when Alc␣1 was also coexpressed, this effect was remarkably enhanced (Fig.  6B). We could not quantify the level of intracellular A␤ derived from full-length APP. Thus, we quantified the A␤ in cells expressing C99. When the intracellular levels of A␤ in the transfected Neuro-2a cells were examined, the same effects of X11L and Alc␣1 expression were observed for intracellular A␤40 generation (Fig. 6C). However, expression of X11L and Alc␣1 did not have a significant effect on the intracellular A␤42 levels, although the expression of both X11L and Alc␣1 did tend to suppress the amount of intracellular A␤42 (Fig. 6C). Thus, we concluded that X11L could suppress the ␥-cleavage of CTF␤, and Alc enhances this effect.
PS is essential for ␥-secretase activity (37). We investigated whether X11L inhibits the interaction of CTF␤ with PS1 and whether Alc enhances this inhibitory activity. HEK293 cells were cotransfected with C99 and PS1 in the presence or absence of X11L and with or without Alc␣1. These cells were subjected to immunoprecipitation assays with the anti-APP antibody G369, which recognizes C99, followed by Western blotting to determine the interaction of C99 with PS1. A previous report showed that the interaction between CTF␤ and PS is detectable in the presence of an aspartic protease inhibitor N-acetyl-leucyl-norleucinal (38), and thus we cultured the transfected cells in the presence or absence of N-acetyl-leucylnorleucinal. We observed that, in the presence of N-acetylleucyl-norleucinal, the anti-APP antibody coprecipitated PS1, confirming that C99 can interact with PS1 as expected (Fig.  6D). In the presence of X11L, the anti-APP antibody recovered a smaller amount of PS1, indicating that X11L suppresses the interaction between C99 and PS1. Surprisingly coexpression of Alc␣1 abolished the interaction between C99 and PS1. Thus, it appears that X11L blocks the association of CTF␤ with PS1 and inhibits the ␥-cleavage of CTF␤, thereby suppressing A␤ generation. Furthermore Alc enhances this effect of X11L by forming a stable complex comprised of CTF␤, X11L, and Alc. The mechanism by which CTF␤ processing is regulated through its interactions with X11L and Alc is depicted schematically in Fig. 6E.
Localization of Alc in the AD Brain-To investigate the localization of Alc␣ in AD patient brains, we immunolabeled serial paraffin-embedded tissue sections with antibody specific for Alc␣. In ethanol (Kryofix)-fixed sections (Fig. 7, panel 1) that had not been subjected to the antigen retrieval method, the anti-Alc␣ antibody consistently detected intense Alc␣ immunolabeling of dystrophic neurites in the neuritic plaques together with faint neuronal labeling. The APP labeling pattern was similar (Fig. 7, panel 2). Labeling was not observed in sections stained with antigen-preabsorbed UT83 (data not shown) or non-immune rabbit IgG used at the same concentration as UT83 (Fig. 7, panel 3). To confirm the colo- FIG. 5. Alc␣ enhances the X11L-mediated stabilization of APP metabolism and suppression of A␤ and sAPP secretion. A and B, effect of coexpressing Alc␣1 and X11L on APP metabolism. HEK293 cells (ϳ1 ϫ 10 7 cells) were transiently transfected with pcDNA3APP695 (2 g) in the presence or absence of pcDNA3-hX11L (0.25 g) and with or without pcDNA3FLAG-hAlc␣1 (6.75 g). To standardize the plasmid amounts, pcDNA3 vector was added (to yield 9 g of plasmid in total). After a 48-h transfection, the cells were metabolically pulse-labeled with [ 35 S]methionine for 15 min and chased for the indicated times. APP was immunoprecipitated from cell lysates with the anti-APP cytoplasmic domain antibody G369, separated by SDS-PAGE (6% (w/v) polyacrylamide), detected by autoradiography (A), and quantified using a Fuji BAS 2000 analyzer. The relative ratios of the levels of immature APP to the maximum level of immature APP at 0 h (the latter level was assigned a reference value of 1.0) were calculated (B). mAPP, mature APP695; imAPP, immature APP695. C, effect of coexpressing Alc␣1 and X11L on A␤ secretion. Neuro-2a cells (ϳ1 ϫ 10 7 cells) were transiently transfected with pcDNA3APP695 (3 g) with or without pcDNA3-hX11L (0.3 g) and in the presence or absence of pcDNA3-FLAG-hAlc␣1 (5.7 g). To standardize the plasmid amounts, pcDNA3 vector was added (to yield 9 g of plasmid in total). The culture medium was collected and assessed for A␤40 and A␤42 levels using a sandwich ELISA. The concentrations of A␤40 and A␤42 are presented as means with S.E. (n ϭ 6). The data were analyzed by one-way analysis of variance followed by the Tukey test (**, p Ͻ 0.01; ***, p Ͻ 0.001). D, effect of coexpressing Alc␣1 and X11L on sAPP secretion. Neuro-2a cells (ϳ1 ϫ 10 7 cells) were transiently transfected with pcDNA3APP695 (3 g) with or without pcDNA3-hX11L (0.3 g) and in the presence or absence of pcDNA3-FLAG-hAlc␣1 (5.7 g). To standardize the plasmid amounts, pcDNA3 vector (Ϫ) was added (to yield 9 g of plasmid in total). The culture medium (2 ml) was collected, and sAPP was recovered by immunoprecipitation with anti-APP extracellular domain antibody (22C11) as described previously (4). The immunoprecipitates (Medium) and cell lysates (Cell, ϳ50 g of protein) were analyzed by Western blotting with 22C11. The levels of sAPP (a relative ratio) were normalized to the amount of intracellular full-length mature and immature APP (APP FL ) and indicated relative to the level of lane 1, which was assigned a reference value of 1.0. Results are the average of duplicate assays, and error bars are indicated. calization of Alc␣ and APP, paraffin-embedded tissue sections from AD brains were double labeled for Alc␣ and APP (Fig. 7, panels 4 -6). In a high power view of a neuritic plaque, most of the Alc␣ immunofluorescence (panel 4) colocalized with that of APP (panel 5). The merged view is shown in panel 6. Furthermore double labeling of Alc␣ and A␤ revealed Alc␣-positive neurites around the amyloid core of plaques (Fig. 7, panel 7). Unfortunately antibody specific for X11L (mint2) did not work on the paraffin-embedded tissue sections (data not shown). These observations suggest that in AD, Alc␣ and APP accumulate in dystrophic neurites around the amyloid core of plaques. DISCUSSION The production, secretion, and aggregation of A␤ may cause neural cell death, resulting in the onset of AD. However, the cellular mechanisms involved in this neural cell death remain to be elucidated (2). The mechanisms involved in the development of familial AD involve mutations in APP and PS that appear to increase A␤ production and cause the early onset of FIG. 6. The Alc␣⅐X11L⅐C99 tripartite complex inhibits A␤ generation from C99 and blocks the association between C99 and PS1. A, effect of X11L and Alc on C99 cleavage. Neuro-2a cells (ϳ1 ϫ 10 7 cells) were transiently transfected with pcDNA3-APPC99 (3 g) in the presence or absence of pcDNA3-hX11L (0.3 g) and with and without pcDNA3-hAlc␣1 (5.7 g). To standardize the plasmid concentrations, pcDNA3 vector (Ϫ) was added (to yield 9 g of plasmid in total). The cells were lysed, and C99 was analyzed by Western blotting with anti-APP (G369) antibody. The lower panel indicates a shorter exposure (exp) of the film. B and C, effect of X11L and Alc on the generation of A␤ from C99. A␤40 (left panel) and A␤42 (right panel) in the medium (B) and the lysate (C) of the cells in A were quantified by a sandwich ELISA. The A␤40 and A␤42 concentrations are presented as means with S.E. (n ϭ 6). The data were analyzed by one-way analysis of variance followed by the Tukey test (*, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001). D, effect of X11L and Alc␣1 on the interaction between C99 and PS1. HEK293 cells (ϳ1 ϫ 10 7 cells) were transiently transfected with pcDNA3-APPC99 (2 g) and pcDNA3-PS1 (1 g) in the presence or absence of pcDNA3-hX11L (2 g) and with or without pcDNA3-hAlc␣1 (4 g). To standardize the plasmid amounts, pcDNA3 vector (Ϫ) was added (to yield 9 g of plasmid in total). The cells were cultured for 24 h in the presence (ϩ) or absence (Ϫ) of N-acetyl-leucyl-norleucinal (LLnL) (10 M) and then lysed. C99 in the cell lysate was recovered by immunoprecipitation with the G369 anti-APP antibody. The immunoprecipitates were analyzed by Western blotting with antibodies specific for APP (G369), the full-length PS1 (PS1-NTF), X11L (mint2), and Alc␣ (UT83). E, schematic diagram showing how the tripartite complex composed of CTF␤, X11s (X11L and X11), and Alc can block the ␥-cleavage of CTF␤ by PS. EC1 and EC2 in Alcadein indicate cadherin motifs 1 and 2, respectively. Although the stoichiometry in the complex APP⅐X11L/X11⅐Alc is drawn as 1:1:1 for the sake of convenience, further analysis is needed to reveal the substance of the complex.
dementia (1)(2)(3). However, the majority of AD cases are not associated with these mutations, and these patients are therefore classified as SAD patients. SAD also differs from familial AD by a late onset of dementia. The pathogenesis of SAD thus must involve alternative mechanisms that increase A␤ production (39), prevent A␤ degradation (5), and/or accelerate A␤ aggregation (6,40). To elucidate these pathogenic mechanisms, numerous investigators in this field, including our group, have focused on understanding the role that APPcyt plays in regulating APP metabolism, including A␤ production (39). APPcyt contains at least three functional motifs, namely, 653 YTSI 656 , 667 VTPEER 672 , and 681 GYENPTY 687 (human APP695 isoform numbering). Mutations in these motifs alter the mechanisms that regulate APP intracellular trafficking and/or metabolism (4,8,9). Several cytoplasmic proteins, through interacting with the 681 GYENPTY 687 motif, are known to modify APP metabolism and consequently affect A␤ production (14,35,36,(41)(42)(43)(44)(45)(46). One of these is the adaptor protein X11L. We found that when X11L binds to the 681 GYENPTY 687 motif in APPcyt, the production of A␤ is suppressed (14). However, the molecular mechanisms that regulate the effect of X11L on APP metabolism are as yet unclear, and thus we screened a human brain cDNA library for proteins that bind to X11L. This led to the identification of the novel cadherin-related membrane protein family, Alc.
The first member of the Alc family we isolated was human Alc␣1. On the basis of its sequence, we also identified the cDNAs for human Alc␣2, Alc␤, and Alc␥ in the cDNA and genome data bases. Homology searching also revealed Alc-like genes in D. melanogaster and C. elegans. These proteins share two cadherin motifs and a putative Ca 2ϩ -binding site in the amino-terminal ectodomain, a laminin G domain in the middle ectodomain, and an X11L-binding site and an acidic region in the cytoplasmic domain. It is likely that these proteins belong to the same family and play identical roles in neural function. However, the roles played by these putative Alc domains (apart from the X11L-binding site) in the metabolism and/or function of APP require further investigation. Supporting the notion that the Alc proteins participate in neural function(s) is that a chicken protein that is homologous to Alc has been reported recently to be a postsynaptic membrane protein that may play a role in postsynaptic Ca 2ϩ signaling (47). It is possible that Alc may transmit unidentified extracellular information through an as yet unknown mechanism or that it may serve as a receptor, together with APP, of cargo proteins in membrane transport vesicles (39). Supporting the first possibility, we demonstrated in the present study that Alc couples with APP through cytoplasmic interactions bridged by X11L or X11. AP-Pcyt is thought to transmit some extracellular signals into nucleus by the mechanism of regulated intracellular proteolysis by coupling with adaptor proteins such as FE65 (48). Alc and X11s may moderate this signaling process by regulating APP processing by ␥-secretase complex including PS. Supporting the possible cargo protein receptor function of Alc is our demonstration that APP, X11L, and Alc␣ were recovered together with KHC in identical subcellular fractions. However, there is no direct evidence that APP and Alc operate as receptors of cargo proteins in membrane transport vesicles (39).
We found that the coupling of APP with Alc through X11L significantly stabilized APP metabolism and enhanced the suppression of A␤ production. This effect was due to the suppression of APP maturation, resulting in the suppression of the first cleavage of APP at the ␣and ␤-sites. Normally the majority of APP is subjected to non-amyloidogenic processing by ␣and ␥-secretases that does not generate A␤ (2), although a small proportion of APP is processed into A␤ by an intracellular amyloidogenic pathway in which the protein is cleaved by ␤and ␥-secretases. We found that X11L could also associate with the carboxyl-terminal fragments of APP that are metabolic products of APP cleavage at the ␣or ␤-site. Thus, we examined whether X11L and Alc could also suppress the ␥-cleavage of C99/CTF␤. We found that the X11L-mediated association of Alc with C99 enhanced the suppressive effects of X11L on A␤ generation from C99. Moreover, since a recent report indicates that PS is an essential component of ␥-secretase (35), we examined whether the interaction between C99/CTF␤ and PS1 is inhibited by formation of the C99⅐X11L⅐Alc complex. We found that X11L blocked PS from interacting with C99/CTF␤ and that Alc enhanced this effect. The stable tripartite complex formed by C99/CTF␤, X11L, and Alc may block the access of the ␥-secretase complex to CTF␤. These observations suggest that the development of drugs that up-regulate X11L and Alc function in AD patients and thereby down-regulate CTF␥ cleavage may be useful in the treatment of AD.
Both APP and Alc recognized the PI domain of X11L, but the binding of these two proteins to PI was cooperative, not competitive. The association of Alc with X11L may induce some conformational change of the PI domain of X11L and result in the stable interaction of APP with X11L. The detailed analysis of the mechanism for interaction among three proteins, including the determination of the stoichiometry of proteins in the complex, is under consideration.
We found that the NPXXXY motif in the cytoplasmic domain of Alc was responsible for the interaction between Alc and the PI domain of X11L. The first and second residues (Asn and Pro) were essential for these interactions, unlike the end Tyr residue. Supporting this is that the Tyr residue is not conserved in all of the Alc family molecules. Moreover the NPXY motif in APP, which is responsible for the interaction between APP and X11L, does not require conservation of the end Tyr residue (31). Thus, while the Tyr residue may be important, perhaps as a possible phosphorylation site, it is not required for interaction with X11L. APP immunolabeling is a sensitive method for detecting disturbances in axonal transport (49,50) and consistently identifies dystrophic neurites in the senile plaques in both nondemented and AD brains (51,52). In this study, we observed that Alc␣ accumulated in the plaque neurites along with APP. However, X11L was reported to localize in the plaque cores (53), although the result requires confirmation. In contrast, X11L, Alc, and APP were found to colocalize largely in normal murine neurons. It is likely that deficient interactions between Alc and APP, by implying axonal transport, generate greater amounts of A␤ in dystrophic neurons of AD patients.
Our studies and those of others reveal that various cytoplasmic and membrane proteins such as X11L and Alc interact with APPcyt to control A␤ production. This information may prove highly useful for the development of novel therapeutic drugs that suppress A␤ production in SAD cases.