Amyloid- (cid:1) Protein Precursor (A (cid:1) PP) Intracellular Domain-associated Protein-1 Proteins Bind to A (cid:1) PP and Modulate Its Processing in an Isoform-specific Manner*

The amyloid- (cid:1) protein precursor (A (cid:1) PP) is a type I transmembrane molecule that undergoes several finely regulated cleavage events. The physiopathological rele-vance of A (cid:1) PP derives from the fact that its aberrant processing strongly correlates with the onset of Alzheimer’s disease (AD). AD is a neurodegenerative disorder characterized by neuronal cell death, loss of synapses, and deposition of misfolded protein plaques in the brain; the main constituent of these plaques is the amy-loid- (cid:1) peptide, a 40–42 amino-acid-long protein fragment derived by A (cid:1) PP upon two sequential processing events. Mutations in the genes encoding for A (cid:1) PP and some of the enzymes responsible for its processing are strongly associated with familial forms of early onset AD. Therefore, the elucidation of the mechanisms un-derlying A (cid:1) PP metabolism appears crucial to under-standing the basis for the onset of AD. Apart from A (cid:1) , upon processing of A (cid:1) PP other fragments are generated. The long extracellular domain is released in the extracellular space, whereas the short cytoplasmic tail, named A (cid:1) PP intracellular domain (AID) is released in-tracellularly. AID appears be involved in several cellular processes, apoptosis, calcium homeostasis, and transcriptional regulation. We have recently reported the cloning and characterization of different isoforms of AID associated protein-1 (AIDA-1), a novel AID-binding protein. Here we further analyzed the interaction between several AIDA-1 isoforms and the cytoplasmic tail of A (cid:1) PP. Our data demonstrated that the interaction between the two molecules is regulated by alternative splicing of the AIDA-1 proteins. Furthermore, we pro-vide data supporting a possible function for AIDA-1a as a modulator of A (cid:1) PP processing. decided to eliminate the region downstream of the PTB domain by designing a specific antisense primer at the end of the PTB domain. various deletion mutants have been amplified by PCR and cloned into pcDNA3 or pEYFP-C1 primers used to amplify AIDA-1b and the various deletion mutants are 5 (AIDA-1b kid-ney modified supplemented fetal bovine stably A (cid:1) PP were in the above-described cell culture medium supplemented with 5 (cid:4) g/ml puromycin. Transient transfec- tions using FuGENE 6 Applied Science) at

familial forms of Alzheimer's disease revealed the crucial role played by mutations in genes encoding for A␤PP 1 and presenilins in the pathogenesis of the disease (2)(3)(4)(5)(6)(7)(8)(9). Presenilins are part of the ␥-secretasome, an enzymatic complex responsible for the intramembranous proteolysis of several transmembrane receptors, among which is A␤PP (10 -20). Other enzymes, named ␤and ␣-secretases, cleave A␤PP in its extracellular region releasing soluble N-terminal fragments (21,22). The above-mentioned mutations in presenilins and A␤PP are the genetic basis for familial forms of Alzheimer's disease, and they all result in aberrant processing of A␤PP (10,(23)(24)(25). A lot of scientific interest has been more recently focused on studying potential functions for the intracellular domain of A␤PP (AID), which is released upon cleavage by ␥-secretase. AID has been shown to be a pro-apoptotic molecule (26), to play a role in intracellular calcium homeostasis (27), to inhibit Notch signaling (28), and to be required for the activation and potential transcriptional activity of the adaptor proteins Fe65 (29,30) and Jun N-terminal kinase interacting protein (31). Many proteins have been described to interact with AID modulating its functions or interfering with A␤PP processing and/or internalization (32). These are, for example, X11␣ (33)(34)(35)(36)(37), autosomal recessive hypercholesterolemia (38), Shc (39,40), Growth Factor Receptor-Bound Protein 2 (41), Numb, and Numb-like (28), in addition to the previously mentioned Fe65 (42,43) and Jun N-terminal kinase interacting protein (31, 44 -47). We recently reported the cloning and initial characterization of novel AIDinteracting proteins that we named AID-associated proteins-1 (AIDA-1) (48). Different isoforms of AIDA-1 proteins, deriving from alternative splicing of the same transcript, have been described, and the potential significance of this interaction has been speculated upon but not proved yet. In this paper we show that certain isoforms of AIDA-1 proteins do not bind to A␤PP despite the presence of a phosphotyrosine binding domain in their C-terminal region and that the interaction is prevented by an inhibitory region encoded by exon 14 and expressed in an isoform-specific manner; more importantly, we demonstrate that AIDA-1a, the only A␤PP-binding full-length isoform that we have cloned, can reduce A␤PP processing by inhibiting activity of the ␥-secretase, consequently diminishing A␤ secretion.

MATERIALS AND METHODS
cDNA Constructs-The A␤PP-Gal4 fusion construct was a kind gift of Dr. Tommaso Russo, University of Naples, Italy. Constructs encoding for full-length AIDA-1a, AIDA-1b, and AIDA-1b⌬Ank have been de-scribed previously (48). AIDA-1b⌬Ank-2 was cloned by reverse transcription-PCR using total RNA extracted from the pre-B cell line 697 as a template. To reduce the number of different products that could have been obtained because of the different splicing at the C terminus, we FIG. 1. Exon/intron organization of AIDA-1 isoforms and their expression in the brain. a, schematic representation of the basic AIDA-1 isoforms employed in this paper. Exons have been numbered progressively starting from a putative promoter region found in the genome. The coding region for each protein is represented by a pattern of diagonal lines. The relative position of ankyrin repeats, the sterile ␣ motif (SAM) domains, and the PTB domain is shown at the top of the figure. Exons 14a, 15a, and 27a derive from exons 14, 15, and 27, respectively, upon splicing at internal splicing sites within the exons themselves. The difference in exon composition at the 3Ј-end of the cDNAs results in different C-terminal tails in the proteins. Especially interesting is the presence of a PDZ binding motif at the C terminus of AT13. This feature might allow isoform-specific interactions with PDZ-containing proteins. b, Western blot (WB) detection of anti-AIDA-1-reactive bands in brain lysates. Approximately 30 -40 g of mouse, human, and human Alzheimer's disease brain lysate was analyzed by SDS-PAGE, and the membrane was blotted with anti-AIDA-1 antibody. Lysates of HEK293T cells transfected with different AIDA-1 isoforms were loaded for size comparison. To ensure antibody specificity, the AIDA-1 antibody was also preincubated with 50 g/ml of AIDA-1 peptide or with the same amount of an irrelevant peptide. decided to eliminate the region downstream of the PTB domain by designing a specific antisense primer at the end of the PTB domain. The various deletion mutants have been amplified by PCR and cloned into pcDNA3 (Invitrogen) or pEYFP-C1 (Clontech). The primers used to amplify AIDA-1b⌬Ank-2 and the various deletion mutants are 5Ј-GAGTCCTATCCAAAGAAGAGAAAT-3Ј (AIDA-1b⌬Ank-2 exon 10 sense), 5Ј-GATAATGCCTCTGAGGTAGCAGTT-3Ј (AIDA-1b⌬Ank-2 exon 11 sense), 5Ј-CCTAAACAGCGAACATCCATTGTG-3Ј (AIDA-1b⌬Ank-2 exon 12 sense), 5Ј-ATTGACAAAATAATGAGTTCCATA-3Ј (AIDA-1b⌬Ank-2 exon 14 sense), and 5Ј-TGCTTGTAGTGCTAGCTG-3Ј (AIDA-1b⌬Ank-2 PTB antisense).

Recombinant GST Fusion Protein Expression and GST Pull Down-
The GST fusion constructs used in this report have been described previously (26). GST fusion proteins were expressed using the pGEX system (Amersham Biosciences) in the Escherichia coli strain BL21 (Invitrogen). Proteins were purified using glutathione-Sepharose beads according to the Amersham Biosciences protocol. Whole cell lysates from HEK293T cells transfected with different AIDA-1 constructs were incubated with the same amount of GST fusion proteins bound to the beads and used in GST pull-down experiments, as described previously (48).
Immunocytochemistry-HeLa cells were seeded in 24-well plates on glass coverslips and cotransfected with several AIDA-1 constructs. 24 h after transfection cells were fixed with 3% paraformaldehyde in 1 M phosphate buffer, pH 7.4, and permeabilized with phosphate-buffered saline, 0.2% Triton X-100. Fixed cells were incubated with anti-AIDA-1 antibody (1 g/ml in phosphate-buffered saline) for 40Ј and subsequently with anti-rabbit Alexa Fluor 594 nM (Transduction Laboratories) (1:250 in phosphate-buffered saline, 3% bovine serum albumin) for Various constructs were generated progressively lacking one exon (exons are indicated with small rectangular boxes above the flat sequence line) at the N terminus of AIDA-1b⌬Ank-2 and were used in GST pull-down experiments as described in the legends to Figs. 2 and 3. Samples were run on a SDS-polyacrylamide gel and detected by using an anti-AIDA-1 antibody. WB, Western blot.
Antibodies-The anti-living colors monoclonal antibody was purchased from Clontech. The anti-AIDA-1 and the anti-A␤PP C-terminal antibody were obtained from Zymed Laboratories Inc..The anti-A␤PP N-terminal antibody 22C11 was purchased from Chemicon.
Enzyme-linked Immunosorbent Assay-HEK293T cells stably expressing A␤PP 695 , were transfected with different EYFP-tagged AIDA-1 constructs. 24 h after transfection the culture medium was replaced with fresh medium and incubated for 30Ј. Supernatants were analyzed by using a human amyloid-␤ (1-40) enzyme-linked immunosorbent assay kit (Immuno-biological Laboratories, Fujioka, Japan). The assay was performed according to the manufacturer's protocol. Each value represents the average of three independent transfections and two readings/transfection. Fluorescence Resonance Energy Transfer (FRET)-HEK293T were cotransfected with different combinations of ECFP/EYFP constructs (DNA ratio 3:1) and analyzed for FRET 24 h after transfection. Acquisition was performed on an LSR II instrument (BD Biosciences) using the BD-FacsDiva software. ECFP, EYFP, and FRET fluorescence was detected by using 470/20, 525/50, and 505 LP filters, respectively.

RESULTS
Exon/Intron Complexity of AIDA-1 Proteins-The cloning of several isoforms of AIDA-1 revealed the extraordinary complexity of the gene encoding for these proteins. The AIDA-1 gene maps on human chromosome 12 (12q21-23) and spans over 1 Mb on the chromosome. We could identify only one predicted promoter region, upstream of exon 1. It seems very likely, therefore, that the different isoforms derive from alternative splicing of the same transcript rather than from transcripts generated by different promoters. Although the regions encoding for the PTB domain and the two SAM domains are very well conserved among all the isoforms that we have cloned, all the other parts of the messenger RNA for AIDA-1 are potentially subjected to several splicing events. The different constructs that have been used in this paper are schematically represented in Fig. 1a. AIDA-1b, AIDA-1a, and AIDA-1b⌬Ank are full-length clones that have been described previously (48). AT13 is a partial clone isolated with a yeast two-hybrid screening using AID as bait; AIDA-1b⌬Ank-2 is partial clone of an alternatively spliced form of AIDA-1b⌬Ank that was cloned by reverse transcription-PCR from the pre-B acute lymphoblastic leukemia cell line 697.
Three different C-terminal tails, downstream of the PTB domain, are predicted for AIDA-1a, AIDA-1bs, and AT13. In- terestingly, alternative splicing causes exon 27a-containing isoforms to have a PDZ binding motif at their C-terminals (as predicted by the ScanPro software), potentially allowing them to interact with PDZ-containing proteins (50).
In the N-terminal region, the ankyrin repeats predicted for AIDA-1b are spliced out in AIDA-1b⌬Ank, and a large portion of the N-terminal part of the protein is not present in AIDA-1a. We recently reported (and we have reproduced these data in Fig. 6) how these diversities in the primary sequence correlate with differences in intracellular localization. In overexpression experiments AIDA-1b is excluded from the nucleus, whereas AIDA-1b⌬Ank and AIDA-1a show a diffused localization with some nuclear accumulation (48). Western blot analysis of mouse brain, human brain, and human Alzheimer's disease brain lysates (see Fig 1b) shows the presence of several bands specifically reacting with the anti-AIDA-1 polyclonal antibody. No band of the apparent size of AIDA-1b and AIDA-1b⌬Ank could be detected, indicating that these proteins could be expressed below the level of detection of the available antibody. Several bands have the approximate size of AIDA-1a. Interestingly, even though more samples should be analyzed to give statistical significance to this observation, we noticed differences in AIDA-1 expression between the Alzheimer's disease and the normal human brain. In fact many anti-AIDA-1 reactive bands that are clearly visible in the normal brain sample are not present in the Alzheimer's disease brain.
Some Isoforms of AIDA-1 Do Not Bind to the Cytoplasmic Domain of A␤PP-We have recently shown how overexpression of A␤PP can modify AIDA-1a intracellular localization, causing its accumulation outside of the nucleus, where it is otherwise abundantly found (48). On the other hand, overexpression of AIDA-1b⌬Ank, which shares the same intracellular localization of AIDA-1a, together with A␤PP did not result in any modification in the distribution of these two proteins (data not shown). Even more surprising, the overexpression of AIDA-1b, which is excluded from the nucleus, did not have any detectable effect on the mainly nuclear localization of AID (data not shown).
We then decided to test several AIDA-1 isoforms for binding to A␤PP in GST pull-down experiments. As shown in Fig. 2, some isoforms did not interact with GST-AID. This was extremely surprising because the PTB domain, which is sufficient for the interaction to occur (48), is conserved among the different isoforms. Consequently, we hypothesized that the lack of interaction was because of an inhibitory region contained specifically in certain isoforms.

The Sequence Encoded by Exon 14 Is
Responsible for Inhibiting the Interaction between AIDA-1s and AID-By comparing the predicted protein sequences and the binding ability of these different AIDA-1 isoforms, we identified two possible candidate regions for the binding inhibition: the C-terminal tail, downstream of the PTB domain, and the stretch of amino acids comprised between the ankyrin repeats of AIDA-1b and the first SAM domain (Fig. 2, arrows). We generated EYFP-tagged constructs encoding for the PTB domains and the three different C-terminal tails, and we used them to perform GST pulldown experiments (Fig. 3). All of these fusion proteins interacted specifically with the C-terminal tail of A␤PP fused to GST but not with GST⌬NPTY; a very efficient binding could be detected for the PTB domain also in the absence of any Cterminal tail. Thus we concluded that the C-terminal tails of AIDA-1s are not responsible for inhibiting the binding to A␤PP.
To analyze the other candidate inhibitor region, we generated a series of deletion mutants of AIDA-1b⌬Ank-2 by progressively removing exons from the N-terminal side of the cDNA (Fig. 4). Interaction with the cytoplasmic tail of A␤PP could not be detected for any of these artificial constructs, suggesting that the shorter one (AIDA-1b⌬Ank-2 exon 14) of these fusion proteins still contains the inhibitory sequence.
We decided to use FRET (51) as an alternative approach to confirm the in vitro binding data. FRET was detected only when EYFP-AIDA-1a was cotransfected with ECFP-A␤PP (Fig.  5). Consistently with our GST pull-down data, no interaction was detected when ECFP-A␤PP was co-expressed with EYFP AIDA-1b⌬Ank-2 or AIDA-1b⌬Ank-2 exon 14. None of the above-mentioned EYFP-tagged AIDA-1s gave a positive FRET signal when cotransfected with ECFP-Ncas, a form of A␤PP lacking part of the C-terminal tail (downstream of the caspase-3 cleavage site). Ncas lacks the YENPTY motif, a short amino acid sequence that has been shown to be critical for binding to PTB domain containing proteins. Immunofluorescence experiments in HeLa cells show that the differences in binding ability do not depend on artifactual mistrafficking of the exon 14 containing constructs. EYFP-AIDA-1b⌬Ank2 exon 14 shows a diffused localization with some nuclear accumulation, as expected. In fact, as we have demonstrated previously (48) and as it is shown here in Fig. 6, only AIDA-1 isoforms containing ankyrin repeats are prevented from entering the nucleus. Comparative analysis of the various AIDA-1 constructs employed in these experiments led to the conclusion that the stretch of amino acids encoded by exon 14 is respon-sible for inhibiting the binding of certain AIDA-1 isoforms to A␤PP.
AIDA-1a Can Reduce the Transcriptional Activity of an Artificial A␤PP-Gal4 Fusion Construct-We have recently shown that A␤PP can cause extranuclear retention of AIDA-1a in FIG. 8. Overexpression of AIDA-1a causes a reduction in A␤40 secretion. HEK293T cells stably expressing A␤PP were transfected with the indicated AIDA-1 constructs and control. After 24 h supernatants were harvested, and A␤40 concentration was measured by enzymelinked immunosorbent assay. Each bar represents the average of three independent transfections. The difference in A␤ levels between the samples transfected with AIDA-1a and the controls is ϳ30%. *, p Ͻ 0.05 relative to EYFP vector and p Ͻ 0.01 relative to EYFP AIDA-1b⌬Ank-2 exon 14. overexpression experiments, and we speculated that this might have some implications in vivo for the normal function of AIDA-1a. No effects on intracellular localization of A␤PP or its processing-derived fragments were detected upon overexpression of different forms of AIDA-1. We decided to investigate a possible role for AIDA-1 in A␤PP processing by taking advantage of a fusion construct encoding for the entire A␤PP linked at its C terminus to the Gal4 transcription factor. This experimental approach has been successfully employed to clone and characterize modulators of A␤PP processing (52). This construct can be cotransfected with a reporter plasmid in which the expression of luciferase is driven by a Gal4-dependent promoter. Upon physiological processing, an AID-Gal4 fragment is released and translocates to the nucleus, activating transcription of the reporter gene.
Cotransfection of EYFP-AIDA-1a with A␤PP-Gal4 in HEK293T cells caused a dramatic reduction (70 -80%) in luciferase activity when compared with a control transfection (Fig.  7). This outcome was specifically dependent on the binding of AIDA-1a to A␤PP, because cotransfection with AIDA-1b⌬Ank-2 exon 14 did not elicit any effect on luciferase activity.
AIDA-1a Inhibits A␤40 Secretion through Inhibition of ␥-Secretase Activity-The outcome of the above-described experiment with A␤PP-Gal4 could have several explanations. 1) AIDA-1a might down-regulate the expression of A␤PP in a nonspecific manner; 2) AIDA-1a might modulate A␤PP processing; 3) AIDA-1a could destabilize and/or reduce nuclear translocation of the AID fragment released upon the processing of A␤PP. When we analyzed by Western blot the same samples that we had used for the luciferase assay to verify the expression levels of the transfected proteins, we could not detect any difference that could account for the results seen in the reporter assay (data not shown). We therefore decided to approach the question of whether AIDA-1a could be involved in A␤PP processing by employing an A␤-40-specific enzymelinked immunosorbent assay. HEK293T cells stably expressing A␤PP 695 were transfected with AIDA-1a, AIDA-1b⌬Ank-2 exon 14, or empty vector. 24 -30 h after transfection the culture medium was replaced with fresh medium, and upon conditioning for 30Ј, the supernatants were tested for A␤ concentration. AIDA-1a could cause a statistically significant 30% reduction in A␤ production when compared with the controls (Fig. 8). To better determine how AIDA1a inhibits A␤ production, we have analyzed by Western blot the C-terminal fragments of A␤PP upon transfection of AIDA-1a, AIDA-1b⌬Ank2 exon 14, or empty vector in presence or absence of the ␥-secretase inhibitor compound E (Fig. 9). Densitometric analysis demonstrated a consistent 30% increase of C-terminal fragments when AIDA-1a, but not the controls, was transfected in the absence of inhibitors; this difference was abolished by treatment with compound E proving that neither ␤nor ␣-secretase are affected by overexpression of AIDA-1a. We concluded from these experiments that overexpression of AIDA-1a can down-regulate A␤PP processing by ␥-secretase, thus reducing A␤ secretion. DISCUSSION We have recently reported the cloning of three alternatively spliced forms (AIDA-1b, AIDA-1bDAnk, and AIDA-1a) of a novel protein interacting with the intracellular domain of A␤PP (48). In this paper we further analyzed the possible implications of this interaction providing evidence for a potential role for AIDA-1a on A␤PP processing.
Comparative sequence analysis of the different isoforms of AIDA-1 proteins revealed an unusual complexity of their exon/ intron organization. 26 exons are present in the cDNA sequence of AIDA-1b. Not only can many of these exons be spliced in or out generating different isoforms, but some of them also contain internal splicing sites (i.e. exons 14, 15, and 27). The combination of all these hypothetical splicing events exponentially increases the number of theoretically possible isoforms that can be generated from a single transcript.
We have previously shown how alternative splicing can affect the intracellular localization of AIDA-1 proteins (48). In this work we demonstrated that the binding of AIDA-1s to the cytoplasmic tail of A␤PP can be inhibited by the presence, in the primary sequence of AIDA-1 proteins, of a short stretch of 24 amino acids encoded by exon 1. This observation is particularly noteworthy for two sets of reasons. 1) For the first time, we have shown that the interaction of a PTB domain with its elective binding motif can be inhibited by a portion of the PTB-containing protein itself, and 2) alternative splicing events can regulate the binding of AIDA-1s to A␤PP, by removal of the inhibitory sequence, thus promoting isoform-specific interactions. At this point the mechanisms responsible for the binding inhibition are not fully elucidated. Does the 24amino-acid sequence encoded by exon 14 bind to the PTB domain of AIDA-1s thus preventing interactions with other molecules, or does it simply cause a different folding of the protein that renders the PTB domain not accessible to potential interactors? We have not been able to show any interaction between exon 14-containing isoforms and the PTB domain of AIDA-1 by biochemical approaches (data not shown), and ultimately crystallographic studies will be required to address this question.
In this paper we have also shown data, from enzyme-linked immunosorbent assay, Western blotting, and luciferase assays, supporting a function for AIDA-1a as a modulator of A␤PP processing. Other A␤PP interactors have been previously shown to play roles in altering the cleavage of this transmembrane molecule (35)(36)(37)(53)(54)(55), but the mechanisms through which this effect is achieved have not been defined yet. AIDA-1 has revealed itself as potentially a very interesting molecule in Alzheimer's pathogenesis. Not only, as we have previously shown, can it interact with A␤PP at endogenous levels, but also the binding and the phenotypic effects resulting from it (such as altered processing of A␤PP and change in subcellular localization of AIDA-1a) occur in an isoform-specific manner. The outcome of all these observations is a very complicated pattern of functional interactions between the AIDA-1 family of proteins and A␤PP that will require genetics approaches to be elucidated.