The Transcriptionally Active Amyloid Precursor Protein (APP) Intracellular Domain Is Preferentially Produced from the 695 Isoform of APP in a β-Secretase-dependent Pathway*♦

Amyloidogenic processing of the amyloid precursor protein (APP) by β- and γ-secretases generates several biologically active products, including amyloid-β (Aβ) and the APP intracellular domain (AICD). AICD regulates transcription of several neuronal genes, especially the Aβ-degrading enzyme, neprilysin (NEP). APP exists in several alternatively spliced isoforms, APP695, APP751, and APP770. We have examined whether each isoform can contribute to AICD generation and hence up-regulation of NEP expression. Using SH-SY5Y neuronal cells stably expressing each of the APP isoforms, we observed that only APP695 up-regulated nuclear AICD levels (9-fold) and NEP expression (6-fold). Increased NEP expression was abolished by a β- or γ-secretase inhibitor but not an α-secretase inhibitor. This correlated with a marked increase in both Aβ1–40 and Aβ1–42 in APP695 cells as compared with APP751 or APP770 cells. Similar phenomena were observed in Neuro2a but not HEK293 cells. SH-SY5Y cells expressing the Swedish mutant of APP695 also showed an increase in Aβ levels and NEP expression as compared with wild-type APP695 cells. Chromatin immunoprecipitation revealed that AICD was associated with the NEP promoter in APP695, Neuro2a, and APPSwe cells but not APP751 nor APP770 cells where AICD was replaced by histone deacetylase 1 (HDAC1). AICD occupancy of the NEP promoter was replaced by HDAC1 after treatment of the APP695 cells with a β- but not an α-secretase inhibitor. The increased AICD and NEP levels were significantly reduced in cholesterol-depleted APP695 cells. In conclusion, Aβ and functional AICD appear to be preferentially synthesized through β-secretase action on APP695.

A characteristic feature of Alzheimer disease (AD) 5 is the presence in the brain of extracellular amyloid plaques com-posed of the amyloid ␤-peptide (principally A␤  and A␤ , which is derived from the transmembrane amyloid precursor protein (APP). Hence, for almost two decades, the amyloid cascade hypothesis (1,2) has driven much AD research with a focus on the prevention of A␤ accumulation or the enhancement of its clearance as primary therapeutic strategies. In the amyloidogenic pathway of APP metabolism, A␤ is formed through the sequential actions of ␤and ␥-secretases, whereas the non-amyloidogenic ␣-secretase pathway precludes A␤ formation. Enzymic clearance of A␤ is mediated by several enzymes, of which the metallopeptidase neprilysin (NEP) is a key contributor, and up-regulation of A␤-degrading enzymes is a potential therapeutic strategy (3,4).
Three major isoforms of APP are produced due to the alternative splicing of exons 7 and 8, which encode a 56-amino acid Kunitz-type proteinase inhibitor (KPI) domain and a 19amino acid domain that shares sequence identity with the OX-2 antigen of thymus-derived lymphoid cells, respectively (5). The longest isoform, APP 770 , contains both the KPI and the OX-2 domains, whereas APP 751 contains only the KPI domain. The shortest isoform, APP 695 , lacks both domains. In the brain, APP 695 is expressed at high levels, and the APP 751/770 isoforms are expressed at significantly lower levels, although there are regional differences, and it has been suggested that the balance between the KPI-and non-KPI-containing isoforms may be an important factor influencing A␤ deposition (6). In the AD brain (7)(8)(9) and in response to N-methyl-Daspartate (NMDA) receptor stimulation (10,11), there is an increase in the proportion of KPI-to non-KPI-containing isoforms of APP. This has led to the suggestion that the KPIcontaining isoforms of APP can exert important neuroprotective functions, and thus their up-regulation in the AD brain or in response to excitotoxic insult may be to protect against further neuronal loss (12,13).
A major unmet scientific need in the AD field is still to understand the normal function of APP (14). An added complexity is whether the different APP isoforms have similar or distinct localizations, metabolism, and roles (15). A long standing enigma in APP biology has additionally been the in-terplay between and physiological roles of the different proteolytic products produced from APP, which include the soluble ectodomains (potentially sAPP␣ and sAPP␤ from each of the APP isoforms, the different forms of A␤ and its oligomers, and the APP intracellular domain, AICD). By analogy with the Notch intracellular domain (16), AICD has been proposed to act as a transcriptional regulator through its interaction with the adaptor protein Fe65 and its translocation to the nucleus, forming, together with a histone acetyltransferase (Tip60), a transcriptional complex (17) subsequently referred to as an AICD-Fe65-Tip 60 (AFT) complex (18,19). Such "nuclear transcription factories" involving AICD have been directly visualized in nuclei by immunofluorescence microscopy (20). However, the identification and verification of target genes, such as the NEP gene (21), have been highly contentious (22,23). We have unequivocally shown that AICD binds to the NEP promoters causing transcriptional activation and upregulation of NEP mRNA, protein, and activity (24). Neuronal NEP expression is, on the other hand, repressed by deacetylation of histones, and histone deacetylase inhibitors such as valproic acid can reactivate NEP expression and hence may aid amyloid clearance (24 -26).
In the present study, we have sought to understand the relative contributions of the different APP isoforms to the production of APP metabolites, especially A␤, sAPP␤, AICD, and hence NEP up-regulation. We conclude that the distinct APP isoforms differ markedly in their ability to modulate NEP expression and that functional AICD production appears to be preferentially synthesized through a cholesterol-dependent endocytic pathway involving ␤-secretase action on the neuronal APP 695 isoform. These observations have significant implications for the selective pharmacological manipulation of APP metabolites, especially the A␤ peptide, and hence for the development of AD therapeutics.

EXPERIMENTAL PROCEDURES
Cell Culture, Plasmids, Transfection, Treatments, and Sample Preparation-Human neuroblastoma cells (SH-SY5Y), murine neuroblastoma cells (Neuro2a; N2a), and human embryonic kidney cells (HEK293) were maintained at 37°C in Dulbecco's modified Eagle's medium (DMEM) (Lonza, Basel, Switzerland) containing 10% fetal bovine serum in a humidified atmosphere of 5% CO 2 , 95% air. The cDNA encoding human APP 695 , APP 751 , or APP Swe was inserted into the expression vector pIREShyg, and the cDNA encoding human APP 770 was inserted into pIREShyg2 (Clontech) before being stably transfected into SH-SY5Y, N2a, or HEK293 cells. DNA (30 g) was introduced into the cells by electroporation in 4-mm cuvettes with a pulse of 250 V and 1650 microfarads using the ECM630 electroporator (BTX Harvard Apparatus, Holliston, MA). Selection for cells containing the required construct was performed in normal growth medium with 0.15 mg/ml hygromycin B (Invitrogen, Paisley, UK). Cells were grown to 90 -100% confluency, washed twice with phosphatebuffered saline (PBS; 1.5 mM KH 2 PO 4 , 2.7 mM Na 2 HPO 4 , 150 mM NaCl, pH 7.4), and incubated in 10 ml of serum-free Opti-MEM (Invitrogen) for 24 h. Conditioned medium was harvested, and 5 ml was concentrated to 200 l using 10-kDa cut-off Vivaspin filtration columns (Millipore, Billerica, MA). The remaining 5-ml conditioned media sample from the SH-SY5Y cells was used for analysis of A␤ 1-40 and A␤ 1-42 by ELISA. Cells were washed twice in PBS, harvested, and pelleted by centrifugation. Cell pellets were stored at Ϫ20°C for chromatin immunoprecipitation (ChIP) analysis or used for cell lysate preparation. For cell lysates, cells were incubated in radioimmune precipitation buffer (150 mM NaCl, 1% (v/v) Nonidet P-40, 0.5% (w/v) sodium deoxycholate, 50 mM Tris/ HCl, pH 8.0) with a protease inhibitor mixture (Sigma-Aldrich, Gillingham, Dorset, UK) for 20 min on ice. Lysates were clarified by centrifugation at 13,000 ϫ g for 10 min. For subcellular fractionation of SH-SY5Y cells into nuclear and non-nuclear fractions, the procedure was modified from Ref. 27. Cells were harvested, washed twice in PBS, pelleted, and then resuspended in 200 l of buffer (10 mM Tris/HCl, pH 8.0, 0.5 M NaCl, 1% Triton X-100, 10% glycerol, 1 mM PMSF, Complete protease inhibitor mixture (Roche Diagnostics)) on ice for 20 min. Lysates were homogenized through 22-G needles 10 times and then pelleted at 4°C for 5 min at 8000 ϫ g. The supernatant was collected as the non-nuclear sample. Nuclear pellets (DNA-bound protein) were resuspended in 100 l of buffer and sonicated 3 ϫ 10 s before being pelleted at 4°C for 10 min at 11,000 ϫ g. The protein concentration of the samples was determined using bicinchoninic acid (Sigma Aldrich). The inhibitors of ␣-secretase (TAPI-2) and ␤-secretase (␤IV) were obtained from Calbiochem.
Cholesterol Depletion of Cells-For cholesterol depletion, SH-SY5Y cells stably overexpressing APP 695 were treated with 5 mM methyl-␤-cyclodextrin (Sigma-Aldrich) in DMEM for 1 h. Control cells were treated with water in DMEM for an equal length of time. Cells were grown and harvested, and cell lysates were prepared as described above.
SDS-PAGE and Western Blot Analysis-Samples (30 or 40 g of protein) were resolved on 7-17% polyacrylamide gels or, for AICD detection, 10 -20% Tricine gels (Invitrogen) and transferred to Hybond-P polyvinylidene difluoride membranes (Amersham Biosciences, Little Chalfont, Buckinghamshire, UK). The membranes were then blocked overnight at 4°C in PBS containing 0.1% (v/v) Tween 20 and 5% (w/v) dried milk powder. Media sample membranes were incubated with either anti-sAPP antibody 22C11 (Millipore) or anti-sAPP␣ antibody 6E10 (Covance, Cambridge, UK) at a dilution of 1:4000 or with anti-sAPP␤ antibody 1A9, which recognizes a neoepitope on sAPP␤ formed after BACE1 cleavage of APP (28), at a dilution of 1:2500. Lysate sample membranes were incubated with anti-APP antibody 22C11, anti-CD10 (Novocastra Laboratories, Newcastle, UK), or anti-actin antibody AC15 (Sigma-Aldrich). Nuclear fractions were incubated with anti-amyloid precursor protein C-terminal antibody (Sigma-Aldrich) at a dilution of 1:500. Membranes were then washed in PBS containing 0.05% Tween 20 before incubation with peroxidase-conjugated rabbit anti-mouse or donkey anti-rabbit secondary antibodies (Sigma-Aldrich) at a dilution of 1:4000 before detection with the enhanced chemiluminescence method (Thermo Fischer Scientific) and analyzed using Aida two-dimensional densitometry.
Gene Expression Analysis-Cell RNA was prepared using the RNeasy extraction kit (Qiagen, Crawley, UK) according to the manufacturer's protocol. RNA was treated with DNase I (Invitrogen), and cDNA was prepared using the iScript cDNA kit (Bio-Rad). cDNA was amplified using conventional PCR or real-time PCR as in Zuccato et al. (30). DNA amplified by conventional PCR was analyzed in 2% agarose gels containing ethidium bromide (1 g/ml) and visualized on a Molecular Imager Gel Doc XR system with the Quantity One 4.6.1 program (Bio-Rad). Image densitometry was performed using the Aida Array Analyzer 4.15 software. Real-time PCR was performed in an iCycler thermal cycler with multicolor PCR detection system (Bio-Rad) using SYBR Green (Bio-Rad) incorporation, and expression was reported relative to actin mRNA.
ChIP Analysis-ChIP was performed as described previously (24). Cells were fixed, extracts were sonicated, and primary antibodies were applied following treatment with protein G-Sepharose, decross-linking, and DNA extraction and analysis by real-time PCR. Real-time PCR data are represented as the -fold of enrichment of DNA pulled down with the specific antibody over that immunoprecipitated with IgG. Antibodies used in ChIP experiments were: anti-AICD (BR188) (31), a generous gift from Dr. M. Goedert (Cambridge, UK); anti-A␤ (6E10); and anti-HDAC1 and IgG from Abcam.
Statistics-Results were compared using a one-way analysis of variance to compare sample means with a Bonferroni correction to determine differences between group samples or by an unpaired two-tailed Student's t test with a threshold of p Ͻ 0.05.

Characterization of Cell Lines Expressing the Different APP
Isoforms-To compare the processing of the different isoforms of APP and their effects on gene expression, stable lines of the human neuroblastoma, SH-SY5Y, expressing each of the three isoforms were constructed and characterized by electrophoresis and immunoblotting of media and lysate samples. The levels of APP expression as assessed in lysate preparations were significantly greater in all three isoform-expressing cell lines than in the mock-transfected cell line (Fig. 1, A and B) and were identical in the APP 695 -and APP 751 -expressing cell lines and slightly reduced (by 22%) in the APP 770 cell line as compared with the APP 695 line (Fig. 1, A and B). The amount of sAPP␣ in the conditioned media was significantly increased in all three isoform-expressing cell lines as compared with the mock-transfected line (Fig. 1, A and C). The sAPP␣ levels were similar between the APP 695 and APP 770 cell lines but slightly decreased (by 10%) in the APP 751 as compared with the APP 695 cell line (Fig. 1, A and C). However, the level of sAPP␤ in the medium from the APP 695 -expressing cells was significantly (ϳ3-4-fold) higher than from the APP 751 -or APP 770 -expressing cells (Fig. 1, A and D). ELISA analysis of A␤ 1-40 and A␤ 1-42 levels in conditioned media from the cell lines demonstrated that A␤ 1-40 was increased in all the APP-overexpressing cell lines as compared with the mock-transfected cells, although proportionately more in the APP 695 cells (Fig. 1E). However, A␤ 1-42 in the conditioned media was significantly increased over the mocktransfected cells only in the APP 695 -expressing cells (Fig. 1E). The increase in sAPP␤ and A␤ peptides, particularly A␤ 1-42 , indicates that APP 695 is preferentially metabolized via the ␤-secretase pathway as compared with APP 751 and APP 770 . Fractionation of the cells coupled with immunoblotting for AICD revealed that nuclear AICD is only significantly different from the mock-transfected control in the APP 695 -overexpressing cell line (ϳ9-fold increase), whereas all three overexpressing cell lines possess immunoreactive AICD in the non-nuclear (cytoplasmic) fraction (Fig. 1, F and G).
To explore the generality of this phenomenon in other cell lines, the APP isoforms were also overexpressed in another neuronal line (N2a) and a non-neuronal cell line (HEK293) (supplemental Fig. S1). In N2a cells, sAPP␤ levels were highest in the APP 695 cells (supplemental Fig. S1, A and C), as in SH-SY5Y (Fig. 1D), whereas in HEK cells, the sAPP␤ levels did not differ significantly between the mock-transfected and any of the isoform-overexpressing cell lines (supplemental Fig. S1, D and F).
Neprilysin Expression Is Up-regulated Selectively in APP 695 Cells in a ␥-Secretaseand AICD-dependent Manner-We and others (21,24) have previously shown that expression of the amyloid-degrading enzyme, NEP, is regulated by APP in a mechanism involving ␥-secretase-mediated APP processing producing AICD, which, in turn, transactivates the NEP promoters. Hence, we compared the expression level of NEP in the different APP isoform-expressing cell lines as a marker for AICD-mediated gene transactivation. Initial studies comparing NEP expression in the different SH-SY5Y cell lines by conventional PCR suggested that there was a marked difference in expression levels between the cell lines, with APP 695 having the highest level of NEP mRNA ( Fig. 2A). This was subsequently confirmed by quantitative, real-time PCR where it was seen that NEP expression levels did not differ between the mock-transfected and APP 751 -and APP 770 -expressing cells (Fig. 2B). However, NEP levels were ϳ6-fold higher in the APP 695 -expressing cells than in the mock-transfected cells, and this elevated expression was significantly reduced on treatment with the ␥-secretase inhibitor L685,458 (Fig.  2B). To establish whether this specific up-regulation of NEP in the APP 695 -expressing cells was mediated via AICD and histone acetylation, ChIP with an anti-AICD or an anti-HDAC1 antibody followed by real-time PCR was applied. Only in the APP 695 -expressing cells was any enrichment of AICD seen on the NEP promoter (Fig. 2C). Conversely, the APP 751 -and APP 770 -expressing cells both showed significant Transcriptionally Active AICD DECEMBER 31, 2010 • VOLUME 285 • NUMBER 53 JOURNAL OF BIOLOGICAL CHEMISTRY 41445 enrichment of the promoter with HDAC1 consistent with repression of transcription (Fig. 2C). Because A␤ itself has been proposed to act as a transcription factor, e.g. in activation of the p53 promoter (32), ChIP analysis was also performed with an antibody recognizing A␤ (6E10), but no enrichment of the NEP promoter was observed (Fig. 1C) in any of the cell lines, ruling out this possibility.
The expression of the NEP gene can be controlled through two distinct promoters (33,34), both of which can be active in neuronal cell types, and in particular, in SH-SY5Y cells (24). For completeness, we therefore extended the ChIP analysis to examine interaction of AICD or HDAC1 with both NEP promoters and with a coding region (exon 19) of the gene. Anti-AICD was seen to pull down both NEP promoters only in the   APP 695 -overexpressing cell line; exon 19 was not precipitated (Fig. 3A). Again, treatment of the cells with the ␥-secretase inhibitor L685,458 eliminated AICD interaction with the promoters, and only anti-HDAC1 was now able to pull down the NEP promoters (Fig. 3B) in the APP 695 line. In the other cell lines (mock-transfected and APP 751 -and APP 770 -expressing), HDAC1 was associated with both promoters, but not exon 19, in the absence or presence of L685,458.
Next, the effects of overexpressing the APP isoforms in the N2a and HEK293 cells were compared. In the N2a lines, NEP expression was significantly up-regulated only in the APP 695overexpressing line, which correlated with the ability of anti-AICD, but not anti-HDAC1, to pull down the NEP promoter (Fig. 4, A and C). However, levels of NEP mRNA expression did not differ between mock-transfected and APP-overexpressing lines in the HEK293 cells (Fig. 4B).
NEP Expression Is Blocked by ␤-Secretase but Not ␣-Secretase Inhibition-The effects of ␣or ␤-secretase inhibition were next examined on APP metabolism and NEP expression in the SH-SY5Y cells expressing APP 695 because these cells represented the only line to show significant changes in the level of NEP expression. The levels of sAPP␣ and sAPP␤ in the conditioned medium of these cells treated with either the ␣-secretase inhibitor, TAPI-2, or the ␤-secretase inhibitor, ␤IV, were compared. As expected, treatment with TAPI-2 significantly decreased the level of sAPP␣ in the medium, but the level of sAPP␤ was unchanged (Fig. 5, A and C). Conversely, treatment of the cells with ␤IV virtually eliminated sAPP␤ with a small but significant increase in sAPP␣ observed (Fig. 5, B and C). When the effects of the inhibitors were compared on NEP expression in the APP 695 -expressing cells, only ␤-secretase inhibition significantly reduced NEP expression (Fig. 5D). To supplement these data, ChIP analysis was performed on cells after treatment with TAPI-2 or ␤IV. As before, AICD was found to interact with the NEP promoter in the APP 695 -expressing cells, and this interaction was not affected by treatment with TAPI-2 (Fig. 5E). However, treatment with ␤IV caused loss of AICD and increased binding of HDAC1 to the promoter. In contrast, no increase in HDAC1 binding after ␤IV treatment was seen in the APP 751and APP 770 -expressing cells (data not shown).
AICD-mediated Gene Expression Is Increased from the Swedish Mutant of APP-As the Swedish mutant of APP is preferentially cleaved by BACE1 as compared with the wildtype protein (35), we determined whether there was an increase in functional AICD from this isoform. Wild-type APP 695 (APP 695 ) or Swedish mutant APP 695 (APP Swe ) were expressed in the SH-SY5Y cells at comparable levels (Fig. 6, A  and B). Consistent with previous reports (35,36), the amounts of both A␤ 1-40 and A␤ 1-42 were significantly increased (by 2-3-fold) in the APP Swe -expressing cells as compared with the APP 695 cells (Fig. 6C). The level of NEP expression was also increased significantly (2-fold) in the APP Swe -expressing cells as compared with the APP 695 cells (Fig. 6D), and there was an increased enrichment of AICD on the NEP promoter in the APP Swe cells (Fig. 6E), consistent with the increased flux through the ␤-secretase/AICD pathway.
Cholesterol Depletion Decreases AICD Production and NEP Expression in APP 695 -overexpressing SH-SY5Y Cells-The amyloidogenic pathway in neuronal cells is initiated in the cholesterol-enriched membrane lipid rafts where ␤and ␥-secretases co-localize (37,38). To explore whether the se- lective production of transcriptionally active AICD in APP 695overexpressing SH-SY5Y cells is also a cholesterol-mediated event, the APP 695 cells were subjected to cholesterol depletion by treatment with methyl ␤-cyclodextrin as described previously (37). After treatment, which reduced cholesterol levels by ϳ30%, the levels of AICD and NEP were assessed by immunoblotting. AICD levels were reduced by ϳ50% after treat-ment (Fig. 7, A and B), and NEP levels were reduced by ϳ40% (Fig. 7, C and D).

DISCUSSION
Understanding of the physiology and functions of APP has been highly influenced by an amyloid-centric perspective of the protein despite the heterogeneity of its expression and its Transcriptionally Active AICD DECEMBER 31, 2010 • VOLUME 285 • NUMBER 53 processing into multiple metabolites in discrete cellular compartments. Hence, the overall complexity of the APP metabolic network and its regulation have rarely been addressed in their entirety. No clear-cut functional differences have been ascribed to the three isoforms of the APP protein (695, 751, 770) apart from a protease inhibitory role for the KPI-containing additional domain present in both APP 751 and APP 770 (39). The tissue-specific expression of APP does, however, imply distinct functional and metabolic roles for the isoforms. We have tested this hypothesis by comparing the ability of the APP isoforms to mediate amyloidogenic processing to produce A␤ and AICD, monitoring the latter through its enhancement of specific gene transcription. We have further established that this pathway is predetermined by the nature of the ectodomain processing of APP.
Although all three APP isoforms are potentially amyloidogenic, the levels of the ␤-secretase cleavage product sAPP␤, along with both A␤ 1-42 and A␤ 1-40 , were significantly higher in the SH-SY5Y cells expressing APP 695 than in the APP 751or APP 770 -expressing cells. To our knowledge, this is the first time that preferential amyloidogenic processing of the APP 695 isoform has been demonstrated in neuronal cells. However, it has been reported previously that in both human brain and cerebrospinal fluid, an antibody to the initial part of the A␤ sequence recognized only the soluble forms of APP that were also reactive to antibodies against the KPI domain, leading to the suggestion that A␤ may be generated in vivo in humans specifically from the non-KPI 695 isoform (40,41). The increased expression of the KPI-containing isoforms of APP in the brain of AD patients (7-9) and following excitotoxic, ischemic, or oxidative insult (10 -12) may thus reflect a neuroprotective process to reduce the production of A␤, particularly the more amyloidogenic A␤ 1-42 , from the 695 isoform.
The regulated intramembrane processing of APP by ␥-secretase appears to require prior shedding of its ectodomain by the actions of either the ␣-secretase, mediated by one or more metalloproteinases, or the ␤-secretase BACE1 (42,43). ␥-Secretase-mediated processing of the C-terminal "stub" of APP formed by ␣or ␤-secretase cleavage then generates the AICD and either the fragment p3 (from ␣-secretase action) or A␤ (from ␤-secretase action) (44). Elucidating the physiological roles of AICD has proved elusive, in part be- cause of its lability if present in the cytosolic compartment (45). It is, however, protected against degradation when channeled through endosomal compartments to the nucleus. Cao and Südhof (17) first reported that AICD could stimulate reporter gene activation, and the peptide has been localized in so-called nuclear transcription factories alongside the Notch intracellular domain (18,19). Several target genes for AICD have subsequently emerged, including the amyloid-degrading enzyme NEP, providing a novel feedback mechanism limiting amyloid accumulation (21,24). However, a number of studies have failed to observe AICD-mediated transcriptional activation in a variety of model systems (22,23). These anomalies have never been satisfactorily resolved but may well reflect the use of artificial constructs employing truncated forms of APP rather than the intact protein and/or the use of non-neuronal cell lines. Nevertheless, chromatin immunoprecipitation studies have shown that AICD interacts with the promoters for NEP and that this binding correlates inversely with that of HDAC1 (24). Transcriptional activation of NEP expression therefore now represents a well validated model of AICD functional activity.
Unlike APP 751 and APP 770 , the tissue distribution of APP 695 is much more restricted and, like its processing enzyme BACE1, it is predominantly neuronally localized, suggesting that the ␤-secretase pathway may be a preferred route for APP 695 metabolism in neurons. Some recent evidence would support this viewpoint (20,46), although direct competition between the ␣and ␤-secretase pathways has more commonly been assumed. Using a cell-based Gal4-driven lu-ciferase reporter gene assay for ␥-secretase-mediated cleavage of APP, Hoey et al. (46) showed that treatment of mouse primary cortical neurons with an ␣-secretase inhibitor (TAPI-1) stimulated luciferase activity, whereas a ␤-secretase inhibitor (C3) substantially decreased luciferase activity. This led to the conclusion that in these neurons, the ␤-secretase amyloidogenic pathway of APP metabolism primarily mediates AICDdependent nuclear signaling (46). Subsequently, Goodger et al. (20) demonstrated that by blocking endocytosis or inhibiting ␤-secretase, translocation of AICD to the nucleus was reduced. These two studies are consistent with a preferred ␤-secretase pathway for AICD production, although neither study monitored direct AICD promoter binding nor its effect on endogenous gene expression. In apparent contradiction to these studies, Sala Frigerio et al. (47) have recently reported that ␤-secretase cleavage is not required for the generation of AICD. However, in these studies, cellular AICD levels were detected only by immunoblotting and, again, functional gene responses were not monitored.
We have therefore endeavored to resolve these major ambiguities and establish which APP metabolic pathway(s) is/are responsible for the functional activity of AICD. Furthermore, for the first time, we have compared the abilities of the different APP isoforms to act as substrates for nuclear AICD generation as monitored by cell fractionation studies, gene expression, and ChIP analysis. By comparing SH-SY5Y cells overexpressing each of the APP isoforms, we were initially struck by the marked enhancement of endogenous NEP mRNA expression detected in the APP 695 -expressing cells as Transcriptionally Active AICD DECEMBER 31, 2010 • VOLUME 285 • NUMBER 53 compared with mock-transfected or APP 751 -and APP 770expressing cells. This up-regulation in APP 695 cells was entirely dependent on both ␤-secretase and ␥-secretase activity and reflected direct binding of AICD to the NEP promoters as monitored by ChIP. In contrast, the APP 751 -and APP 770 -expressing cells showed enrichment of HDAC1 binding, but not AICD, on the promoters, consistent with transcriptional repression. We have further established that inhibition of ␤-secretase caused the replacement of AICD by HDAC1 on the NEP promoter in the APP 695 -expressing cells, whereas ␣-secretase inhibition did not affect AICD binding to the NEP promoter, nor the levels of endogenous NEP mRNA expression. All three APP isoforms were able to produce a significant increase in A␤ 1-40 levels as compared with mock-transfected cells, whereas only APP 695 -overexpressing cells produced a significant increase in A␤ 1-42 peptide. Hence, these data establish that the ␤-secretase, amyloidogenic pathway acting on APP 695 is the predominant pathway generating A␤ 1-42 and AICD and mediating gene regulation. These conclusions are further supported by the demonstration that A␤ production and AICD-mediated regulation of NEP expression are additionally enhanced in cells expressing the Swedish mutant APP 695 . Furthermore, another neuronal line, Neuro2a, also showed enhanced AICD occupancy of the NEP promoter and up-regulation of NEP expression only in cells overexpressing the 695 isoform. In contrast, a non-neuronal cell line, HEK293, failed to show this effect, demonstrating neuronal specificity.
Taken together, our observations may well explain the controversy that has surrounded functional activity for AICD because transcriptional effects in model cell lines will depend critically on the nature and integrity of the APP constructs that are used. Furthermore, the total cellular levels of AICD, as monitored by immunoblotting (48), may not reflect the minor pool of AICD generated by ␤-secretase that is functionally active in regulating gene transcription. The mechanism underlying the functional differences between the APP isoforms appears to reflect the preferential transit of APP 695 through a cholesterol-dependent and cholesterol-mediated endocytic pathway, consistent with previous reports implicating lipid raft involvement in the amyloidogenic pathway (48,49). These pathways are summarized in Fig. 8, showing the distinct cleavage of APP (all isoforms) at the plasma membrane by ␣-secretase (50) and showing the preferential cleavage of APP 695 by ␤-secretase in a subpopulation of BACE1-containing endosomes following their co-localization in lipid rafts and endocytosis (48,49). Retrograde transport of FIGURE 8. A model for the endocytosis and nuclear delivery of transcriptionally active AICD. APP 695 is sequestered along with BACE1 and ␥-secretase complexes into lipid raft domains, and processing of the APP occurs following endocytosis where the acidic interior environment favors the catalytic action of the secretases (46), which are aspartic proteinases. The AICD, in combination with Fe65, is delivered to the nucleus by retrograde transport (20), where it can facilitate specific gene transcription, e.g. of the NEP gene (21,24). In contrast, the predominant action of the metalloenzyme ␣-secretase on APP isoforms occurs at the cell surface. The subsequent action of ␥-secretase releases AICD into the cytosol, where it can be degraded by insulin-degrading enzyme (IDE) (41). Ac, acetyl.
AICD to the nucleus is presumed to be stabilized by formation of a protein complex with Fe65 and possibly other proteins (20). In contrast, cytosolic AICD formed at the plasma membrane via the non-amyloidogenic pathway is known to be rapidly metabolized by insulin-degrading enzyme and hence is non-functional (45).
A number of factors may underlie these metabolic differences, reflecting differential compartmentation, distinct secretase/substrate kinetics, and/or the involvement of different adapter proteins modulating secretase actions. Sequestration of the APP isoforms into distinct endosomal populations must precede ␤-secretase action because once the APP ectodomain is removed by ␤-secretase, the residual substrate for ␥-secretase is identical in all cases. This differential compartmentation may reflect the distinct ligand-induced internalization of APP 751 and APP 770 , as compared with APP 695 , through forming complexes with ligands of protease nexin 2 such as the low density lipoprotein receptor-related protein (LRP), which can mediate neurite outgrowth (51)(52)(53). Additionally, APP 695 associates with assembled NMDA receptors, which has led to speculation that APP may function as a regulator of intracellular trafficking mechanisms (54). Such a mechanism could deliver the APP isoforms to different internal compartments, which may contain distinct ␥-secretase protein complexes generating different end products. In conclusion, these data emphasize the need for a much fuller understanding of the APP interactome and the influence of these interactions on the trafficking and metabolism of the different APP isoforms. A␤  and AICD may both contribute to AD pathology (55). Our observations could therefore have novel therapeutic implications because if APP 695 generates predominantly both metabolites, manipulating this isoform may provide a selective advantage.