Cytoplasmic Fragment of Alcadein α Generated by Regulated Intramembrane Proteolysis Enhances Amyloid β-Protein Precursor (APP) Transport into the Late Secretory Pathway and Facilitates APP Cleavage*

Background: Alcadein α (Alcα) forms a ternary complex with APP and X11L. Results: Transport into the nerve terminus and metabolism of APP were facilitated in Alcα CTF transgenic mice, along with an increase in Aβ. Conclusion: Alcα ICD, a product of γ-secretase cleavage of Alcα CTF, enhanced APP trafficking from the ternary complex into a late secretory pathway. Significance: Novel function of Alcadein α results from regulated intramembrane proteolysis. The neural type I membrane protein Alcadein α (Alcα), is primarily cleaved by amyloid β-protein precursor (APP) α-secretase to generate a membrane-associated carboxyl-terminal fragment (Alcα CTF), which is further cleaved by γ-secretase to secrete p3-Alcα peptides and generate an intracellular cytoplasmic domain fragment (Alcα ICD) in the late secretory pathway. By association with the neural adaptor protein X11L (X11-like), Alcα and APP form a ternary complex that suppresses the cleavage of both Alcα and APP by regulating the transport of these membrane proteins into the late secretory pathway where secretases are active. However, it has not been revealed how Alcα and APP are directed from the ternary complex formed largely in the Golgi into the late secretory pathway to reach a nerve terminus. Using a novel transgenic mouse line expressing excess amounts of human Alcα CTF (hAlcα CTF) in neurons, we found that expression of hAlcα CTF induced excess production of hAlcα ICD, which facilitated APP transport into the nerve terminus and enhanced APP metabolism, including Aβ generation. In vitro cell studies also demonstrated that excess expression of Alcα ICD released both APP and Alcα from the ternary complex. These results indicate that regulated intramembrane proteolysis of Alcα by γ-secretase regulates APP trafficking and the production of Aβ in vivo.

The neural type I membrane protein Alcadein ␣ (Alc␣), is primarily cleaved by amyloid ␤-protein precursor (APP) ␣-secretase to generate a membrane-associated carboxyl-terminal fragment (Alc␣ CTF), which is further cleaved by ␥-secretase to secrete p3-Alc␣ peptides and generate an intracellular cytoplasmic domain fragment (Alc␣ ICD) in the late secretory pathway. By association with the neural adaptor protein X11L (X11-like), Alc␣ and APP form a ternary complex that suppresses the cleavage of both Alc␣ and APP by regulating the transport of these membrane proteins into the late secretory pathway where secretases are active. However, it has not been revealed how Alc␣ and APP are directed from the ternary complex formed largely in the Golgi into the late secretory pathway to reach a nerve terminus. Using a novel transgenic mouse line expressing excess amounts of human Alc␣ CTF (hAlc␣ CTF) in neurons, we found that expression of hAlc␣ CTF induced excess production of hAlc␣ ICD, which facilitated APP transport into the nerve terminus and enhanced APP metabolism, including A␤ generation. In vitro cell studies also demonstrated that excess expression of Alc␣ ICD released both APP and Alc␣ from the ternary complex. These results indicate that regulated intramembrane proteolysis of Alc␣ by ␥-secretase regulates APP trafficking and the production of A␤ in vivo.
Alcadein (Alc) 2 is a brain abundant type I membrane protein family comprised of Alc␣, Alc␤, and Alc␥ (1), which are also identified as the Ca 2ϩ -binding protein calsyntenin (2,3). They share two cadherin repeats, a concanavalin A-like lectin/glucanase superfamily domain in their amino-terminal extracellular region, an acidic domain, a kinesin light chain-binding WD motif and an X11L (X11-like)-binding NP sequence in their carboxyl-terminal cytoplasmic region (1,4). Originally, we isolated Alc␣ as an X11L-interacting molecule (1,5). X11L is a neuron-specific cytoplasmic adaptor protein and was also identified as a binding partner of amyloid ␤-protein precursor (APP) (6). Both Alc␣ and APP interact with the phosphotyrosine interaction/phosphotyrosine-binding domain of X11L and form a ternary complex comprised of Alc␣, X11L, and APP (1,5).
In neurons, APP695, a neuron-specific isoform, undergoes N-glycosylation in the endoplasmic reticulum, producing immature APP (7). The immature APP is transported to the Golgi and further modified by O-glycosylation to form mature APP (mAPP). The mAPP enters into the late secretory pathway and localizes to the plasma membrane, whereas some mAPP also enter endosomal recycling pathways. During the late secretory pathway, APP is subject to consecutive cleavages (8). Alc␣/ calsyntenin-1 is also subject to intracellular trafficking and metabolism and participates in neural functions, similar to APP (4, 9 -11). Calsyntenins have also been reported to mediate exit of APP from the TGN (12).
It is well known that APP undergoes primary proteolytic cleavage at juxtamembrane ␣or ␤-sites by ␣or ␤-secretase and that membrane-associated APP C-terminal fragments (APP CTFs) are further cleaved at ␥/⑀-sites by ␥-secretase (8). When APP is cleaved by a combination of ␣and ␥-secretases, a metabolically labile p3 peptide is generated, whereas neuro-toxic amyloid ␤ (A␤) peptide is generated when APP is cleaved by ␤and ␥-secretases. The A␤ peptide is known as a causative molecule of Alzheimer disease (13,14). Alc␣ is also subject to proteolytic cleavage by ␣-secretase and remains a membraneassociated C-terminal fragment (Alc␣ CTF), which is further cleaved by ␥-secretase to secrete p3-Alc␣ and generate an intracellular domain fragment (Alc␣ ICD) (10,15).
X11L associates with both APP and Alc␣ in the Golgi and also in the late secretory pathway (16,17). In the Golgi, X11L is thought to regulate the formation of APP and Alc␣ cargo vesicles (17). Formation of the ternary complex composed of APP, X11L, and Alc␣ also regulates the entry of APP into lipid rafts where ␤-secretase is active (16). Additionally, X11L is thought to regulate ␥-cleavage of APP CTFs directly (18).
A␤ production is suppressed when APP is expressed with X11L, and we reported that suppressed A␤ production by X11L was further enhanced when full-length Alc␣ was also coexpressed (1,5,6) due to ternary complex formation. In vitro, binding of APP to X11L is stabilized when Alc␣ is coexpressed, and this enhanced interaction of APP with X11L mediated by Alc␣ is thought to further stabilize APP metabolism as well as regulate intracellular APP trafficking because the cleavage of APP by secretases occurs in the late secretory pathway. In vivo, X11L/X11␤ transgenic mice expressing human APP suppressed amyloidogenic processing of APP (19). Furthermore, X11L gene knock-out (X11L-KO) mice showed enhanced generation of endogenous A␤ in the brain (20), and human APP transgenic mice lacking the X11L gene exhibited enhanced amyloid plaque formation in the brain (21). However, the role of Alc␣ metabolites in APP metabolism and A␤ generation remain unclear in vivo. To determine the function of Alc␣ CTF and its intracellular metabolic fragment Alc␣ ICD, we generated a transgenic mouse line expressing human Alc␣-CTF under the control of the PDGF-␤ promoter and examined its effect on the metabolism of APP in vivo.

EXPERIMENTAL PROCEDURES
Generation of hAlc␣ CTF Transgenic Mouse Lines-cDNA encoding human Alc␣ CTF (amino acids 817 to 971 of the hAlc␣1 isoform) (1, 10) was connected to the signal sequence (amino acids 1 to 28 of hAlc␣1). The construct was inserted into a vector with a 5Ј-PDGF-␤ promoter and a 3Ј-SV40 poly(A) tail to produce the TgPDhAlc␣CTF plasmid. For DNA microinjection, linearized DNA was prepared by digestion with restriction enzymes, as illustrated (see Fig. 1A). Mice were purchased from CLEA-Japan (Tokyo, Japan), and all of the animal studies were conducted in compliance with the guidelines of the Animal Studies Committee of Hokkaido University. Linearized DNA (SalI/NotI fragment) was micro-injected into fertilized eggs produced by mating between BDF1 mice (F1 hybrid of C57BL/6 and DBA/2 mice) according to standard procedures (22). In brief, BDF1 females (6 -8 weeks of age) that had been super-ovulated by injection of pregnant mare serum gonadotropin (serotropin, Asuka Pharmaceutical Co.) and human chorionic gonadotropin (Asuka Pharmaceutical Co.) were mated with males of the same strain. Pronuclear stage embryos were collected from pregnant females, and DNA fragments were injected into the male pronuclei of the zygotes. The embryos were then cultured in potassium simplex optimized medium at 37°C in an atmosphere of 5% CO 2 and 95% humidity. The surviving embryos were transplanted into the oviducts of pseudopregnant females (Multi Cross Hybrid (Institute of Cancer Research), 8 -12 weeks of age). Transgenic founders were identified by PCR and Southern blot analysis of genomic DNA extracted from tail biopsies. Genotyping PCR was performed using a set of sense (Alc␣ CTF, 5Ј-cct gac cat cac cgt caa cc-3Ј) and antisense (SV40, 5Ј-cac ctc ccc ctg aac ctg aa-3Ј) primers with ExTaq DNA polymerase (Takara-bio). For Southern blot analysis, genomic DNA was digested with EcoRI, separated by agarose gel electrophoresis, and transferred onto a Nylon membrane. The membrane was incubated with a probe prepared from the DNA sequence of an injected construct, and signals were detected by CDP-Star (GE Healthcare). Founder mice were backcrossed with C57BL/6 mice, and offspring were backcrossed over 10 times. Non-transgenic littermates derived from the same crosses were used as controls.
Immunohistochemistry-Mouse brain sections (25-m thick) were prepared as described (20), and the sections were stored at Ϫ30°C until use. Frozen sections were washed with PBS for 20 min and then incubated in PBS containing 1% (v/v) Triton X-100 for 20 min to permeabilize the membranes. Tissue sections were then incubated with PBS containing 0.3% (v/v) H 2 O 2 for 10 min to inactivate endogenous peroxidase activity and washed three times with PBS. The sections were blocked with PBS containing 5% (v/v) normal goat serum for 1 h at room temperature and then incubated with no. 958 antibody (serum diluted to 1:8,000) for ϳ8 h at 4°C. After three washes with PBS, sections were incubated with an anti-rabbit IgG peroxidaselinked species-specific whole antibody (GE Healthcare, dilution 1:500) for 30 min at room temperature. The signal was visualized with a VECTASTAIN kit (Vector Laboratories) following the manufacturer's protocol.
Subcellular Fractionation-The cytosolic fraction of mouse brain tissue was prepared as described (23) and subjected to immunoprecipitation with no. 958 antibody. The synaptosome fraction was prepared as described (24) with some modifications. In brief, the cerebral cortex and hippocampus regions of 12-month-old Tg54 and non-Tg mice were homogenized in buffer A (10 mM HEPES (pH 7.4) containing 0.32 M sucrose, 5 g/ml chymostatin, and 5 g/ml leupeptin), and this fraction was used as total lysate. The lysate was then centrifuged at 1,000 ϫ g for 10 min, and a clarified supernatant (S1 fraction) was recovered. The S1 fraction was further centrifuged at 13,800 ϫ g for 20 min, and the supernatant (S2 fraction) and pellet (P2 fraction) were used for assays. The P2 fraction was suspended again in buffer A and overlaid on a discontinuous sucrose gradient prepared with 1.2 M, 1 M, and 0.85 M sucrose solution and centrifuged at 82,500 ϫ g for 2 h with an SW41 rotor (Beckman Coulter). After centrifugation, the layer between 1.0 M and 1.2 M sucrose was collected and resuspended in 6 mM Tris-HCl (pH 8.0) buffer containing 0.5% (v/v) Triton X-100 to prepare the synaptosome fraction.
Antibodies, Immunoprecipitation, and Immunoblot Analysis-A rabbit polyclonal anti-Alc␣ carboxyl-terminal domain antibody, no. 958, was raised against a synthesized peptide, 948 GEQGDPQNATRQQQL 962 , of human Alc␣1. IgG was affinity purified with antigen-coupled beads and used for analyses. This antibody specifically recognizes Alc␣ of human and mouse almost equivalently but does not show cross-reactivity with Alc␤ and Alc␥ of human (see Fig. 2A) or mouse (data not shown).
Immunoprecipitation was performed as described (1) using a previously described plasmid for FLAG-X11L (6). For the APP-GFP construct, human APP695 cDNA was cloned into pEGFP-N3 (Clontech) between the HindIII/NotI sites. For the hAlc␣ ICD construct, cDNA encoding the cytoplasmic region of hAlc␣ from 868 to 971 amino acids was cloned into pcDNA3.1. Amino acid position 868 of Alc␣1 is the major N terminus of Alc␣ ICD, which is generated by ⑀-cleavage of Alc␣ CTF by ␥-secretase (15). Brain and cell extracts used in immunoblot analyses were prepared by homogenizing samples in eight volumes of RIPA buffer containing 5 g/ml chymostatin and 5 g/ml leupeptin. The lysates were centrifuged, and supernatants were used for immunoprecipitation analysis. The procedures used for immunoblotting and the identification of APP CTFs were described previously (25).
Immunofluorescence Microscopy Analysis and Quantification of Fluorescence Intensity-The procedures in immunofluorescence experiments shown in Fig. 5 were followed as described previously (27). The images were analyzed by a fluorescence microscope (Keyence, BZ-700X). Fluorescence quantification was performed using the imaging software ImageJ (http://imagej.nih.gov/ij/, National Institutes of Health). The images were acquired with a resolution of 640 ϫ 480 pixels using a 100ϫ objective lens, which covered the whole cell region. The region of interest that contained the cell body-neurite junction area was selected at a resolution of 132 ϫ 48 pixels, in which the junction was set to a middle region manually using a transparent image obtained by transmission light. The fluorescence intensities in the proximal region of neurite ("neurite," right half of a selected image) and the hillock of cell body ("cell body," left half of a selected image) were then measured. The fluorescence intensities where no cell region was observed in the right and left half images were also measured as background. The background intensity was subtracted from the respective intensity of cell body and neurite, and the values were indicated as a ratio to average intensity of neurite plus cell body.

Generation of Human Alc␣ Carboxyl-terminal Fragment
Transgenic Mice-A cDNA sequence encoding the carboxylterminal region of human Alc␣1 together with the signal peptide sequence was expressed under the regulation of the mouse PDGF-␤ promoter (Fig. 1A). The signal sequence was cleaved correctly by signal peptidase when expressed in HEK293 cells, producing the human Alc␣ carboxyl-terminal fragment (hAlc␣ CTF) with the correct amino-terminal sequence. hAlc␣ CTF then underwent intramembrane cleavage by ␥-secretase to generate p3-Alc␣ with the same amino acid sequence observed in vivo (10). MALDI-TOF/MS analysis of p3-Alc␣ secreated from HEK293 cells showed that the p3-Alc␣35 (amino acid sequence indicated in B) was generated as a major p3-Alc␣ form along with several minor species (Fig. 1B). The result indicates that exogenously expressed hAlc␣ CTF was processed in a manner similar to that of endogenously generated Alc␣ CTF.
We generated six transgenic mouse founder lines using this construct (Fig. 1C). Genomic Southern blot analysis of the six lines shows a few to several hundred copies of exogenous gene incorporated into the genome (Fig. 1D). The protein expression level of hAlc␣ CTF in the brain region, including the cerebral cortex and hippocampus was analyzed in the three lines with the highest trans-gene dose by immunoblotting with an anti-Alc␣ antibody (Fig. 1E). Of these founders, line no. 54, which showed a 4-fold increase in hAlc␣ CTF expression compared with endogenous levels of the protein, was chosen for use in further analyses. Tg54 was subjected to genetic back-cross with C57BL/6 over 10 generations.
Expression of hAlc␣ CTF in Tg54 Mouse Brain-Alcadein family proteins Alc␣, Alc␤, and Alc␥ share a similar structure in the cytoplasmic region. Therefore, we developed an antibody raised against the cytoplasmic sequence of Alc␣. This antibody, designated no. 958, recognizes specifically Alc␣ and its CTF but not Alc␤ or Alc␥. An anti-FLAG antibody was used to confirm the levels of expression in cells expressing these FLAG-tagged constructs ( Fig. 2A).
Using this antibody, we analyzed the expression of hAlc␣ CTF in 3-month-old Tg54 (Tg) and non-Tg (N) mice (Fig. 2, B  and C). Under the regulation of the PDGF-␤ promoter, increased expression of hAlc␣ CTF was observed in the cerebral cortex, hippocampus, and olfactory bulb in Tg compared with non-Tg mice (Fig. 2B). Immunohistochemical analysis of endogenous Alc␣ and hAlc␣ CTF expression agreed well with the results of immunoblot analysis. An immuno-reactive signal was observed throughout the cerebral cortex (Fig. 2C, panels a and d), in pyramidal cells in CA1 to CA3 along with granule cells in the dentate gyrus of the hippocampus (Fig. 2C, panels b and e), and in layers of mitral and granule cells of the olfactory bulb (Fig. 2C, panels c and f). These signals were enhanced in Tg54 mouse brains (Fig. 2C, panels d-f). Immunoreactivity of tissue staining is specific because this antibody does not react to the tissue staining of Alc␣-KO mouse brain (data not shown). Taken together, we concluded that the Tg line expressing hAlc␣ CTF, Tg54, was successfully established. Tg54 mice exhibit normal growth and fertility (data not shown).
We next measured the amounts of p3-Alc␣, which is generated by ␥-secretase cleavage of Alc␣ CTF (illustrated in Fig. 2F). In brains of Tg54 mice, the total amount of p3-Alc␣ greatly increased, whereas in non-Tg mice, it was below the level of detection (Tg, 0.33 Ϯ 0.03 pmol/g tissue; n ϭ 3, Fig. 2D). Corresponding to the increased amount of p3-Alc␣ in brain tissue, a significant amount of hAlc␣ ICD, which was detected as a doublet, was also detected in the cytosolic fraction of Tg54 mouse brain tissue, whereas only very low levels were detected in non-Tg mice (Fig. 2E). These findings indicate that exogenously expressed hAlc␣ CTF is physiologically cleaved by ␥-secretase in the brain to generate p3-Alc␣ along with Alc␣ ICD (10,15).

Facilitation of Intracellular Trafficking and Metabolism of Endogenous Mouse APP in Tg54
Mice-To reveal the effect of Alc␣ CTF on the metabolism of APP in vivo, the amounts of endogenous A␤ 1-40 and A␤ 1-42 in brain lysate isolated from the cerebral cortex and hippocampus were quantified in 12-to 15-month-old Tg54 and non-Tg mice. Unexpectedly, the amounts of both A␤ species were significantly greater in Tg54 than non-Tg mice (TBS insoluble A␤ 1-40 , non-Tg ϭ 0.61 Ϯ 0.04 pmol/g tissue; Tg54 ϭ 0.72 Ϯ 0.02 pmol/g tissue (p ϭ 0.0017); TBS insoluble A␤ 1-42 , non-Tg ϭ 0.18 Ϯ 0.00 pmol/g tissue, Tg54 ϭ 0.22 Ϯ 0.01 pmol/g tissue (p ϭ 0.0105)) (Fig. 3A). The amount of A␤ yielded in the TBS-soluble fraction was too small to detect a significant change. These results suggest that the metabolism of APP is facilitated in the brains of Tg54 mice.
Cleavage of APP in neurons occurs during or after its axonal transport; in other words, in the late secretory pathway. Therefore, the brains of Tg54 and non-Tg mice were fractionated to isolate the synaptosome fraction, which includes membrane vesicles of the late secretory pathway. As shown in Fig. 3B, the amounts of full-length APP, especially mAPP in synaptosomes, was significantly greater in Tg54 mice (synaptosome ϭ 1.24 Ϯ 0.19 (n ϭ 5, p ϭ 0.0458) when the value in non-Tg mice was set   (n ϭ 4). ***, p Ͻ 0.005, Student's t test. #, below detectable levels. E, detection of hAlc␣ ICD in the cytosolic fraction of mice brains. Brain lysates prepared from 2-month-old Tg54 (Tg) and non-Tg (N) mice brains were subjected to immunoprecipitation with no. 958 antibody, and the precipitates were detected by immunoblotting with the same antibody. Arrows indicate hAlc␣ ICD fragments. Numbers on the left side of the panel indicate the molecular mass (kDa). Asterisks indicate IgG heavy and light chains. F, schematic structure of p3-Alc␣ and hAlc␣ ICD. Alc␣ CTF is first cleaved at the ⑀-site by ␥-secretase to release Alc␣ ICD into the cytoplasm. Next, ␥-secretase cleavages reach to the ␥-site to secrete p3-Alc␣ into the extracellular milieu (15). at 1.0). However, total APP, which includes mAPP and immature APP, was not changed significantly in the total lysate of either Tg or non-Tg mice (lysate ϭ 1.06 Ϯ 0.19, n ϭ 5, p ϭ 0.4904). APP CTFs were also significantly more abundant in the synaptosome fraction of Tg54 mice compared with non-Tg mice (1.27 Ϯ 0.09 (n ϭ 7, p ϭ 0.0261) when the value in non-Tg mice was set at 1.0), whereas the amounts were equivalent in total lysate of Tg54 and non-Tg mice (0.92 Ϯ 0.05, n ϭ 7, p ϭ 0.1843). These results suggest that increased amounts of APP undergo enhanced trafficking into the nerve terminus where it is cleaved by primary secretases. Increased CTFs in the late secretory pathway are then further cleaved by ␥-secretase in Tg54 mice, thus facilitating the production of A␤ in the brain. This is not the result of enhanced amyloidogenic processing of APP. A␤ 1-40/1-42 is derived from C99 among three CTF species; C99 (CTF␤), C89 (CTF␤), and C83 (CTF␣). These CTFs are phosphorylated at Thr-668 in brain, and the phosphorylated C99 (pC99) is a major C99 rather than non-phosphorylated form in mouse brain (7). Therefore, an increase of pC99 level can indicate the enhanced amyloidogenic cleavage of APP. However, in Fig. 3B, the ratio of pC99 to total CTF was equivalent in the synaptosome fraction of Tg54 and non-Tg mice brains, although total amounts of APP CTFs increased significantly in the synaptosome fraction of Tg54. Therefore, we understand that the increased A␤ production in Tg54 is due to the facilitated trafficking of APP into the late secretory pathway rather than the enhanced amyloidogenic cleavage of APP. Identification of mature and immature APP and APP CTF species in the brain was described in detail previously (7).
We then considered the possibility that hAlc␣ ICD derived from hAlc␣ CTF could play a role in processing the increased amounts of APP entering the late secretory pathway, because the hAlc␣ ICD was free from membrane association and highly abundant in Tg54 (Fig. 2F). APP and Alc␣ are associated in the Golgi apparatus via interaction with X11 family proteins. Interaction of APP with X11L, a neuron-specific protein, is enhanced by association of Alc␣ with X11L. APP in this ternary complex is thought to suppress further trafficking into the late secretory pathway, thus stabilizing APP for cleavage by secretases. In Tg54 mice, excess expression of hAlc␣ CTF is likely to relieve this suppression. hAlc␣ CTF is then quickly cleaved by ␥-secretase to generate hAlc␣ ICD along with p3-Alc␣ (Fig. 2,  D-E). Thus, hAlc␣ ICD could perform a function different from full-length Alc␣ in stabilizing APP metabolism in the presence of X11L.
To assess this possibility, hAlc␣ ICD was coexpressed in Neuro2a cells, together with GFP-tagged hAPP (hAPP-GFP), full-length hAlc␣ (hAlc␣ FL), and FLAG-tagged X11L (FLAG-X11L), which form a ternary complex (Fig. 4). In the absence of hAlc␣ ICD, co-immunoprecipitation with an anti-FLAG antibody recovered FLAG-X11L together with hAPP-GFP and hAlc␣ FL (Fig. 4A, lane 5). However, increased expression of hAlc␣ ICD decreased the recovery of both hAPP-GFP and hAlc␣ FL significantly (Fig. 4A, lanes 6 -8, and 4B), although hAPP-GFP, hAlc␣ FL, and FLAG-X11L are expressed at equal levels (Fig. 4A, lanes 1-4). These results indicate that Alc␣ ICD performs a novel function that releases both APP and Alc␣ from X11L and is different from the function of Alc␣ FL. Thus, it is possible to conclude that the increased levels of APP and its metabolic fragments APP CTFs and A␤ in the synaptosome fraction of the brains of Tg54 mice is due to enhanced release and anterograde trafficking of APP to the nerve terminus from the cell body. In summary, Alc␣ ICD regulates the trafficking of APP and Alc␣ itself. This may be a novel function involved in regulated intramembrane proteolysis by ␥-secretase (28,29).
Alc␣ ICD Regulates the Trafficking of APP into Neurites-To analyze the function of Alc␣ ICD on the trafficking of APP into neurites, we used Neuro2a cells that exhibit the extensions of neurites under a normal culture condition. In neuronal cells expressing EGFP-APP, APP appears the proximal region of extending neurites because APP is subject to an anterograde transport into the late secretory pathway (4). Therefore, we analyzed quantitatively the fluorescence signals of EGFP-APP in the proximal region of neurites (Fig. 5).
When Alc␣ CTF was expressed in Neuro2a cells expressing EGFP-APP and HA-X11L, fluorescence signal of APP appears in the proximal neurite, whereas the signal reduced significantly in the presence of 10 mM DAPT (␥-secretase inhibitor: (3,5-difluorophenylacetyl)-Ala-Phg-OBu t ) (Fig. 5, B and C, upper panels). Cells treated with DAPT showed the increased Alc␣ CTF levels (Fig. 5D), suggesting the production of Alc␣ ICD by ␥-secretase cleavage of Alc␣ CTF was inhibited.
When Alc␣ ICD was further expressed in the cells along with Alc␣ CTF expression, EGFP-APP fluorescence in the proximal region was not decreased by DAPT treatment (Fig. 5, B and C,  lower panels). These observations indicate that Alc␣ ICD, but not Alc␣ CTF, facilitates the trafficking of APP into the late secretory pathway.

DISCUSSION
Because it was documented that the generation of A␤ from APP and the formation of neurotoxic A␤ oligomers in neurons are closely linked to the etiology of Alzheimer disease, many efforts have been made to identify the modulator of APP metabolism, including A␤ generation and clearance (14). With regard to APP trafficking and metabolism, we and others have reported previously that X11 family proteins (X11s) are regulators of APP metabolism (6, 16, 20, 30 -33). Excess expression of X11s suppresses APP processing, including A␤ generation and APP trafficking, via binding of X11L to APP. Furthermore, X11s function in the regulation of ␥-cleavage of APP CTF directly or indirectly (18) or of ␤-cleavage of APP indirectly (16). Importantly, X11s-KO mice showed a significant increase in endogenous A␤ generation in the brain and increased formation of amyloid plaques was observed in the brains of hAPP-Tg/ X11s-defective mice, indicating that X11s play an important role in APP regulation at various stages of APP metabolism and trafficking in vivo (16,20,21,34,35).
This effect is further enhanced in the presence of Alc␣ by the formation of a metabolically stable ternary APP⅐X11L⅐Alc␣ complex. This suggests that along with X11L, Alc␣ may be a key molecule for APP metabolism and trafficking in vivo. Because our previous in vitro study indicated that not only Alc␣, but also Alc␣ CTF, could form a ternary complex together with X11L and APP in cultured cells (1), we hypothesized that membraneassociated full-length Alc␣ and Alc␣ CTF would suppress the production of A␤ in the presence of X11L.
However, we observed that the brains of Tg54 mice expressing excess amounts of hAlc␣ CTF showed a significant increase in A␤ due to enhanced transport of APP into the late secretory pathway. Moreover, our analysis showed remarkable production of p3-Alc␣ and Alc␣ ICD, which are the products of Alc␣ CTF following cleavage by ␥-secretase. This observation indicates that in vivo, introduced Alc␣ CTF is quickly cleaved by ␥-secretase to release Alc␣ ICD, which is free from membrane association. Indeed, in vitro experiments showed that Alc␣ ICD performs a novel function in releasing APP and Alc␣ from X11L, different from the function of membrane-associated Alc␣ and Alc␣ CTF.
Strikingly, both full-length APP and Alc␣ are released from X11L by excess expression of Alc␣ ICD in cultured cells. These studies in vivo and in vitro strongly indicate that Alc␣ ICD has different binding properties than full-length Alc␣ and may be a  ). B, quantification of APP and Alc␣ recovered by co-immunoprecipitation assay. Protein band densities of hAPP-GFP (open circle) and hAlc␣ FL (filled circle) co-immunoprecipitated with FLAG-X11L were quantified and are shown relative to the value obtained in the absence of hAlc␣ ICD expression (0, set to 1.0). The horizontal axis represents the amount of pcDNA3-hAlc␣ ICD (g) co-transfected with pcDNA3-hAPP-GFP (1 g), pcDNA3-hAlc␣ (1 g), and pcDNA3-FLAG-X11L (0.5 g). Results are given as the means Ϯ S.E. Asterisks indicate statistical significance as determined by Student's t test (n ϭ 3; *, p Ͻ 0.05).
key molecule for the regulation of APP trafficking and metabolism. As shown in Figs. 2E and 4, we detected Alc␣ ICD as a doublet. Interestingly, only the slower migrating band interacted with X11L in co-immunoprecipitation assays (Fig. 4). Although we determined that molecular weight of the faster migrating protein band was the expected size of hAlc␣ ICD (data not shown), we have yet to identify the modification of slower-migrating protein band and determine why it interacts differently with X11L.
Several previous reports indicated that stalling of APP transport increases the production of A␤ because of the increased probability of association with secretases (36 -38). The source of A␤ is still controversial. A recent report suggested that majority of intracellular APP is transported into lysosome from Golgi to generate A␤ (39). In contrast, it has been thought that cell surface APP is a significant source of A␤ (40,41). Our observation that the increased A␤ in Tg mice brain is due to the increased synaptosomal APP level may be consistent with a hypothesis that cell surface APP contributes to the increased production of A␤. We revealed here that increased trafficking of APP into the late secretory pathway facilitated APP metabolism, including A␤ generation, which was regulated by Alc␣ ICD derived from Alc␣ CTF by ␥-secretase cleavage. Although the precise molecular mechanism is still under investigation, it is clear that dysfunctional APP intracellular trafficking is closely related to the aberrant production of A␤. Moreover, the role of Alc␣ ICD in the regulation of membrane protein trafficking is a novel function of regulated intramembrane proteolysis.