The amyloid precursor protein (APP)-cytoplasmic fragment generated by γ-secretase is rapidly degraded but distributes partially in a nuclear fraction of neurones in culture

Authors

  • Philippe Cupers,

    1. Neuronal Cell Biology Group, Center for Human Genetics, Flanders Interuniversitary Institute for Biotechnology and Catholic University of Leuven, Leuven, Belgium
    Search for more papers by this author
  • Isabelle Orlans,

    1. Neuronal Cell Biology Group, Center for Human Genetics, Flanders Interuniversitary Institute for Biotechnology and Catholic University of Leuven, Leuven, Belgium
    Search for more papers by this author
  • Katleen Craessaerts,

    1. Neuronal Cell Biology Group, Center for Human Genetics, Flanders Interuniversitary Institute for Biotechnology and Catholic University of Leuven, Leuven, Belgium
    Search for more papers by this author
  • Wim Annaert,

    1. Neuronal Cell Biology Group, Center for Human Genetics, Flanders Interuniversitary Institute for Biotechnology and Catholic University of Leuven, Leuven, Belgium
    Search for more papers by this author
  • Bart De Strooper

    1. Neuronal Cell Biology Group, Center for Human Genetics, Flanders Interuniversitary Institute for Biotechnology and Catholic University of Leuven, Leuven, Belgium
    Search for more papers by this author

Address correspondence and reprint requests to Wim Annaert and Bart De Strooper, CME, Neuronal Cell Biology Group, Herestraat 49, B-3000 Leuven, Belgium. E-mail: AD@med.kuleuven.ac.be

Abstract

The γ-secretase cleavage is the last step in the generation of the β-amyloid peptide (Aβ) from the amyloid precursor protein (APP). The Aβ precipitates in the amyloid plaques in the brain of Alzheimer's disease patients. The fate of the intracellular APP carboxy-terminal stub generated together with Aβ has been, in contrast, only poorly documented. The analogies between the processing of APP and other transmembrane proteins like SREBP and Notch suggests that this intracellular fragment could have important signalling functions. We demonstrate here that APP-C59 is rapidly degraded (half-life ∼5 min) when overexpressed in baby hamster kidney cells or primary cultures of neurones by a mechanism that is not inhibited by endosomal/lysosomal or proteasome inhibitors. Furthermore, APP-C59 binds to the DNA binding protein Fe65, although this does not increase the half-life of APP-C59. Finally, we demonstrate that a fraction of APP-C59 becomes redistributed to the nuclear detergent-insoluble pellet, in which the transcription factor SP1 is also present. Overall our results reinforce the analogy between Notch and APP processing, and suggest that the APP intracellular domain, like the Notch intracellular domain, could have a role in signalling events from the plasma membrane to the nucleus.

Abbreviations used

β-amyloid peptide

AD

Alzheimer's disease

ADAM

a disintegrin and metalloprotease

APP

amyloid precursor protein

BHK

baby hamster kidney

HRP

horseradish peroxidase

NICD

notch intracellular domain

PBS

phosphate-buffered saline

SDS–PAGE

sodium dodecyl sulfate–polyacrylamide gel electrophoresis

SFV

Semliki Forest virus.

Production and aggregation of the β-amyloid peptide (Aβ) is believed to be a central event in the pathogenesis of Alzheimer's disease (AD) (Selkoe 1999). Aβ is generated from the amyloid precursor protein (APP) by β- and γ-secretases (Haass and Selkoe 1993). The γ-secretase cleavage involves cleavage of a peptide bond in the hydrophobic environment of the phospholipid membrane, and depends critically on the presenilins (De Strooper et al. 1998; Naruse et al. 1998; Song et al. 1999; Wolfe et al. 1999; Esler et al. 2000; Herreman et al. 2000; Li et al. 2000; Zhang et al. 2000). γ-Secretase cleavage of the APP fragment generated by β-secretase (APP-C99) does not only release the Aβ peptide, but also gives rise to an APP carboxy-terminal fragment remaining associated with the cell. As γ-secretase can cleave at two positions in the membrane, these putative cytoplasmic fragments consist of the last 10 or 12 amino acids of the transmembrane domain of APP, together with the 47 amino acids of its cytoplasmic tail (APP-C57 and APP-C59, respectively). Although Aβ release in the medium has been extensively studied, the intracellular fate of APP-C57 and APP-C59 has been poorly documented until now. Recently, Passer et al. have reported the potential identification of these fragments in Jurkat cells and in brain homogenates (Passer et al. 2000); they appear to induce apoptosis (Passer et al. 2000). Another APP cytoplasmic fragment containing the last 31 amino acids and generated by caspase cleavage of APP (Gervais et al. 1999; Pellegrini et al. 1999a; Weidemann et al. 1999) was recently also implicated in apoptosis (Lu et al. 2000). Fragments corresponding to the APP γ-stub have been generated in vitro using a γ-secretase assay (McLendon et al. 2000; Pinnix et al. 2001). Production of this fragment correlated with the production of Aβ and could be decreased with γ-secretase inhibitors. Obviously, although all of these studies provide evidence for the existence of APP-C59 and APP-C57 fragments, they do not provide any explanation of why they are so difficult to detect and what their potential function could be. The further study of the generation and fate of these APP fragments in cells is clearly very important, as it could give us clues towards the physiological function of APP (for a review see De Strooper and Annaert 2000). The analogies between the processing of APP and of other proteins like Notch and SREBP, make it tempting to speculate that the APP intracellular fragments could be involved in signal transduction pathways (Annaert and De Strooper 1999; Brown et al. 2000).

Notch is a large integral membrane protein, involved in cell fate decisions (Artavanis-Tsakonas et al. 1999; Weinmaster 2000). Notch is proteolytically processed, first by furin during its transport to the cell surface (Logeat et al. 1998) and then upon ligand binding at the cell surface, by a metalloprotease of the disintegrin and metalloprotease (ADAM) family. This second endoproteolytical event (S2-cleavage) is a prerequisite for the third cleavage (S3-cleavage) close to or in the Notch transmembrane domain, which releases the Notch intracellular domain (NICD) (Brou et al. 2000; Mumm et al. 2000). The S3-cleavage is presenilin dependent (De Strooper et al. 1999; Struhl and Greenwald 1999), suggesting that similar γ-secretase-like activities are responsible for both Notch and APP cleavage (De Strooper et al. 1998, 1999; Song et al. 1999; Struhl and Greenwald 1999; Herreman et al. 2000; Zhang et al. 2000). The NICD can bind to DNA binding proteins of the CSL family (Jarriault et al. 1995), traffic to the nucleus and activate basic helix–loop–helix transcription factors (Weinmaster 2000). The analogies in the cleavage of APP and Notch are particularly intriguing, and we proposed the hypothesis that the APP-C59 fragment produced after cleavage by the presenilin/γ-secretase could play a similar role in signal transduction as the NICD (Annaert and De Strooper 1999). Further substantiating this hypothesis were the observations that Fe65, a nuclear protein, can bind to the APP cytoplasmic tail, and bind by itself to transcription factors of the CP2, LSF and LBP1 family (Russo et al. 1998; Zambrano et al. 1998), although it is unclear whether APP-C59 can remain in complex with Fe65 after γ-secretase cleavage. Unfortunately, several problems hamper the further study of this question. First, in contrast with Notch, the function of APP remains basically unknown (De Strooper and Annaert 2000). By consequence, no target genes have yet been identified that can serve as reporters in assays evaluating the proteolytic release of the APP intracellular domain and its trafficking to the nucleus. This approach has been crucial for the understanding of proteolytic processing of Notch and the role of its cytoplasmic domain in signalling (Jarriault et al. 1995; Lecourtois and Schweisguth 1998; Schroeter et al. 1998; Struhl and Adachi 1998; Song et al. 1999; Struhl and Greenwald 1999). Secondly, as already mentioned, the cytoplasmic fragment of APP, APP-C59, is extremely difficult to observe even under conditions that allow perfect detection of Aβ. It should be noticed that similar problems with the detection of the NICD has caused controversy for years in that field. Only recent studies have managed to establish definitively that the NICD, even when only present in minute amounts in the cell, exerts important signalling functions (Huppert et al. 2000; Schroeter et al. 1998).

In the current study, we have established experiments to analyse the fate of APP-C59 at the biochemical level in cell culture. We find that simple expression of APP-C99, which is probably the best available substrate for γ-secretase and yields large amounts of Aβ peptide (Lichtenthaler et al. 1999a), is not sufficient to allow detection of APP-C59 with available assays. This implies that any APP-C59 generated under these conditions is rapidly degraded. We substantiated this observation by expressing APP-C59 directly in baby hamster kidney (BHK) cells and primary cultures of neurones. The bulk of this fragment has an exceptionally short half-life, although a small fraction (less than 10%) appears to be more stable. Several protease inhibitors were tested but none could inhibit fully the rapid turn-over of the fragment. Finally we investigated the subcellular distribution of the fragment and provide evidence that it distributes to a nuclear fraction that contains the transcription factor SP1.

Materials and methods

Cell culture

The PS1 knockout mice have been described previously (De Strooper et al. 1998). Primary neuronal cultures were generated from trypsinized brain obtained from 14-day-old embryos. Neurones were resuspended and plated on cell culture dishes (Nunc, Naperville, IL, USA) pre-coated with 1 mg/mL of poly-l-lysine (Sigma, St Louis, MO, USA), and incubated in neurobasal medium (Gibco BRL, Rockville, MD, USA) as described previously (Annaert et al. 1999). These mixed cultures, containing essentially cortical neurones, were used in the next 5 days. The BHK cells were cultured in Dulbecco's modified Eagle's medium (DMEM)/F12 (Gibco BRL) supplemented with 10% fetal calf serum (Gibco BRL). For all experiments, cells were plated on six-well dishes (Nunc) at a confluence of 300 000 cells per well.

Semliki Forest virus constructs

Using APP cDNA cloned in pSP65, we performed a full deletion of the APP ectodomain till the Asp1 amino acid of the Aβ region (just at the β-secretase site). An additionnal AspAla (DA) was inserted by site-directed mutagenesis (Stratagene, La Jolla, CA, USA) between the last amino acid of the signal sequence (Ala) and the first one of the βA region (Asp1), ensuring generation of a correct APP-C99 after signal peptidase cleavage (Lichtenthaler et al. 1999b). This fragment is called APP-C99. The APP-C57 and APP-C59 stubs were generated by PCR. These fragments start at Thr43 or at Ile41, respectively, of the APP sequence (amino acid 1 being the methionine of the Aβ sequence). Both constructs are preceded by an ATG coding for an additional methionine. Production of Semliki Forest virus (SFV) particles was performed as described by Annaert et al. (1999).

Viral infection and metabolic labelling

Mouse primary neuronal cultures in neurobasal medium, or BHK cells in DMEM/F12 medium were incubated with 10-fold diluted SFV expressing the pSFV-1 plasmid bearing APP-WT, APP-C99, APP-C59 or APP-C57. After 1-h infection at 37°C, cells were incubated in neurobasal or DMEM/F12 medium (neurones or BHK cells, respectively) without viruses for 2 h at 37°C. Cells were then metabolically labelled for 4 h at 37°C by incubation in methionine-free minimum essential medium (Gibco BRL), supplemented with the B27 complement mixture and containing 200 µCi [35S]methionine (ICN Biomedical, Costa Mesa, CA, USA). At the end of the incubation, culture media were recovered and cells were washed once with phosphate-buffered saline (PBS), and finally lysed in DIP buffer (20 mm Tris-HCl, pH 7.4, 150 mm NaCl, 1% Triton X-100, 1% sodium deoxycholate and 0.1% sodium dodecyl sulfate). Cell extracts and media were spun for 10 min at 20 800 g in a microfuge (Brinkmann–Eppendorf, Westbury, NY, USA) to remove DNA and cell debris. For pulse–chase experiments, cells were labelled with [35S]methionine for the indicated amount of time, washed once with PBS, and then incubated in neurobasal or DMEM/F12 medium (neurones or BHK cells, respectively) without radioactive label for various time intervals. At the end of the chase, cells were washed and lysed as described above.

Cell extracts and cell media (supernatants) were incubated with the appropriate antibodies and Protein G sepharose. Antiserum B7/7 or B7/9 recognizes epitopes in the first 17 amino acid sequences of the Aβ. These antisera do not recognize the p3 or the carboxy-terminal α-stub (De Strooper et al. 1995). Antiserum B11/4 is a polyclonal rabbit antiserum targeted against the last 20 amino acids of the APP carboxy-terminal domain (De Strooper et al. 1995). Antibody 4G8 is a monoclonal antibody targeted against the amino acids 17–28 of the amyloid region, i.e. the region between the α-secretase site and the membrane. After overnight incubation on a rotating wheel at 4°C, immunoprecipitates were washed four times with DIP buffer and once with PBS/water 1 : 3. Samples were finally solubilized in 25 µL of Nu-Page sample buffer (composition specified in Invitrogen NuPage gels protocol; Carlsbad, CA, USA) and heated at 70°C for 10 min, according to the manufacturer's instructions. Sodium dodecyl sulfate − polyacrylamide gel electrophoresis (SDS–PAGE) was performed on pre-casted 4–12% or 10% acrylamide NuPage Bis-Tricine gels, using reducing conditions and 2-(N-morpholino)ethane sulfonic acid (MES) in the running buffer (Invitrogen). Gels were fixed in methanol/water 1 : 1 for 2 h at room temperature (21–24°C) and dried (type-583; Bio-Rad Laboratories, Hercules, CA, USA). Radiolabelled material was detected using a PhosphorImager (Molecular Dynamics Inc., Sunnyvale, CA, USA), and quantification was performed by ImagQuaNT 4.1. All quantitative data obtained for APPs, APP C-terminal fragments or secreted Aβ were normalized to the signal obtained for the corresponding APP holo-form in order to compensate for variations in expression between cultures dishes or the different APP mutants used.

Transfection of BHK cells with Fe65

Rat Fe65 cDNA, cloned in pSK was generously given by Dr Tommaso Russo and Dr Nicola Zambrano (University of Napoli, Italy). Full Fe65 was cloned using Not1/Apa1 in pSG5 vector. The BHK cells cultured in DMEM/F12 medium supplemented with 10% fetal calf serum were washed and transfected with rat Fe65 cloned in pSG5, using the Fugene transfection technique (Boehringer-Roche, Indianapolis, IN, USA). Seventy-two hours after transfection, cells were incubated in DMEM/F12 and infected with SFV as described above. Cells were further analysed as described above. Anti-Fe65 antibody provided by Dr Buxbaum (Mount Sinaï University, New York, USA) was used for immunoprecipitation, and another anti-Fe65 antibody was provided by Drs Russo and Zambrano (University of Napoli, Italy) was used for western blotting. Blots were revealed using secondary goat anti-rabbit antibody coupled to horseradish peroxidase and Super Signal detection (Pierce, Rockford, IL, USA). For co-immunoprecipitation analysis, cell extracts lysed in 150 mm NaCl, 1 mm sodium vanadate, 50 mm Tris pH 7.4, NP-40 0.5% and protease inhibitors, were immunoprecipitated with anti-Fe65 and then blotted with anti-Fe65 or anti-APP (11/4, affinity purified) antibodies.

Cell fractionation

Neurones were infected using recombinant SFV-APP-C59, as described above. Six hours after infection, cells were washed twice with Tris 50 mm pH 7.4, sucrose 250 mm, EDTA 0.1 mm (buffer A), and harvested. Cells were spun at 180 g for 10 min, resuspended and homogenized in buffer A using a cell cracker (EMBL) with beads of 8.01 mm diameter. Neuronal cell homogenate was then centrifuged at 800 g for 10 min and the remaining pellet of this ‘centrifugation step 1’ was then lysed for 30 min in Tris 50 mm pH 7.4, NaCl 150 mm, EDTA 0.1 mm, 0.1% Triton, spun for 15 min at 20 800 g From this centrifugation, the supernatant, representing soluble nuclear material was called ‘N’, and the insoluble pellet (consisting of DNA and associated proteins) was named ‘P’. The supernatant of ‘centrifugation 1’ was spun for 60 min at 150 000 g; the resulting final supernatant consisted mainly of cytosol and was named ‘C’, whereas the pellet consisted of membranes and was called ‘M’. Equal protein quantities of each fraction (C, M, N and P) were separated by SDS–PAGE (NuPage system), transferred on a nitrocellulose membrane, and western blot was performed using antibodies against APP: polyclonal antibody B19.2 directed against the PS1 amino-terminal stub (Annaert et al. 1999); polyclonal antibody against SP1 (Santa Cruz Biotechnology, Santa Cruz, CA, USA); monoclonal antibody directed against clathrin light chain. All blots were revealed after incubation with a secondary antibody coupled to HRP followed by Super Signal detection (Pierce).

Results

Recently, the identification of the cytoplasmic derivatives APP-C59 and APP-C57 (γ-stubs) generated by γ-secretase cleavage of APP in human brain has been reported (Passer et al. 2000). Because we have been unable to detect these fragments in mouse brain extracts or in primary cultures of mouse neurones, we reasoned that those stubs are only present at extreme low levels in steady state conditions. To investigate the fate of these fragments we therefore decided to use overexpression paradigms that have proven their value when analysing APP processing. We first expressed APP-C99 (β-stub; Fig. 1), the direct precursor of APP-C59/57 (APP γ-stub). Expression of APP-C99 with the SFV system in neurones or BHK cells leads to efficient and high production of Aβ and we therefore anticipated that the APP γ-stubs would also be more easily detected under those conditions. APP-C99 migrates in SDS–NuPAGE at about 10 kDa, as indicated in Fig. 2(a), lane 2 (β-stub). Two other fragments with higher mobility are also observed, one migrating like the APP α-stub (Fig. 2a, lane 2). The second band migrates at an apparent molecular weight of 6.5 kDa, and was always stronger in BHK cells than in neurones. We first supposed that this fragment was the γ-stub because it reacted with antibody B11/4 against the 20 carboxy-terminal antibodies of APP, but not with antibodies B7/9 (amino acids 1–16 of Aβ) or 4G8 (amino acids 17–28 of Aβ) (Fig. 2b). However, a synthetic APP-C59 (synthetic γ-stub) runs faster than this 6.5-kDa fragment, indicating that these two bands might not actually correspond to the same protein (Fig. 2c). We tried to confirm the identity of the 6.5-kDa stub by radiosequencing, but all our attempts failed until now, probably because of N-terminal blocking. However, indirect evidence indicates that this band is not the product of γ-secretase cleavage as in PS1–/– neurones, secretion of Aβ is reduced by 80%, whereas the intensity of the 6.5-kDa stub stays identical in PS1+/+ and PS1–/– neurones (Fig. 2d). We notice also that this 6.5-kDa stub is a quite stable protein fragment, that displays a similar turnover as the α- and β-stub fragments of APP in BHK cells (Fig. 3) and in neurones (data not shown). When APP-C59 was expressed, relatively low levels of protein were observed compared with the signals obtained with holo-APP-WT or APP-C99 (Fig. 2c compared with 2a). In pulse–chase experiments, APP-C59 appears to have an extremely short half-life (about 5 min) in BHK cells as well as in neurones, whereas the endogenous holo-APP that is coprecipitated in this experiment remains relatively stable (Fig. 4a, and quantified in 4c). APP-C59 was not secreted into the extracellular medium as shown in Fig. 4(b). Furthermore, from this experiment we deduce that after 120 min the complete intracellular pool of APP-C59 becomes labelled, confirming the rapid turnover of the fragment (Fig. 4b). We next studied the APP-C59 degradation pathway by metabolic labelling and chase experiments in the presence of various protease inhibitors. All proteasome inhibitors, i.e. lactacystine (100 µm) and the calpain inhibitors I and II (100 µm) or MDL28170 (20 µm and 100 µm) had no or only a very slight stabilizing effect on the degradation of APP-C59 (Fig. 5a). Inhibitors of endosomal and lysosomal proteolytic function like leupeptine (20 µg/mL), methylamine (30 mm) or bafilomycine A1 (1 µm) had no effect on the degradation of APP-C59 (Fig. 5a). Finally, also in PS1 deficient neurones (PS1–/–) the turn-over of APP-C59 was not significantly affected (Fig. 5c) suggesting that γ-secretase was not involved in the degradation of APP-C59 (Fig. 5b). Pinnix et al. (2001) recently demonstrated that adding the metalloprotease inhibitor phenantroline in their in vitro assay greatly enhanced the recovery of the APP-C59 product. Therefore we incubated neurones with increasing concentrations of phenanthroline during metabolic labelling, but no effect on the steady state levels of APP-C59 was observed (Fig. 6). Most likely, the metalloprotease that degrades APP-C59 in cell free systems is not involved in the degradation of this fragment in intact cells.

Figure 1.

 APP constructs used. Schematic representation of all APP constructs used in this study: APP is shown with its amino-terminal signal sequence, the ectodomain, the amyloid peptide region and the carboxy-terminal cytoplasmic tail. Sites of α-, β-, and γ-secretases cleavages are also represented. For APP-C99, a supplementary aspartyl-alanine dipeptidyl sequence has been added between signal sequence and the Asp1 of the amyloid peptide in order to obtain a correct cleavage by signal peptidase (Lichtenthaler et al. 1999b). The APP-C59 begins at the γ40 cleavage site of the APP; this construct is synthesized without signal sequence. See Materials and methods for further details.

Figure 2.

 Analysis of various APP carboxy-terminal stubs after metabolic labelling in BHK cells and in neurones. BHK cells (a and b) and PS1+/+or PS1–/– primary neuronal cultures (c and d) were transduced as described in Materials and methods with SFV bearing the indicated APP constructs. Cells were metabolically labelled with [35S]methionine for 4 h at 37°C, washed, and lysed in DIP buffer. Cell extracts (a, b, c and d) and culture media (d) were immunoprecipitated using the indicated antibodies. The precipitates were separated by SDS–NuPAGE 4–12% (a), or 10% (b, c and d), and detection of radioactive material was performed using PhosphorImaging. Arrows on the gels indicate the position of the various APP fragments. Antibody B11/4 against the APP-cytoplasmic tail was used in panels (a), (b) lane 1, (c) and (d), lane 2, antibody B7/9 recognizes mainly epitopes in the first half of the Aβ region amino (amino acids 1–16 between β- and α-cleavage sites), and was used in panels (b), lane 2 and (d), lane 1. mAb 4G8 reacts with epitopes in the Aβ region between the α-cleavage site and the membrane and was used in panel (b), lane 3. In panel (b), an APP-fragment is observed that does not react with the carboxy terminal antibody B11/4, but does react with B7/9 and mAb 4G8 antibodies. This band was not further studied, but likely represents a caspase-generated fragment as was documented previously by others (Pellegrini et al. 1999b; Weidemann et al. 1999; Lu et al. 2000). Gels show a representative experiment out of four (a) or out of two (b, c and d).

Figure 3.

 Turnover of various APP carboxy-terminal stubs generated from APP-C99. BHK cells were transduced as described in Materials and methods with SFV bearing APP-C99 construct (β-stub), and metabolically labelled at 37°C with [35S]methionine for 2 h followed by a chase of 0–60 min (a). Cell extracts were immunoprecipitated using antibody B11/4 and separated by SDS–PAGE (10%). Detection of radioactive material was performed by PhosphorImager. (b) Quantification of bands from gel (a). Similar results were obtained in neurones, but the 6.5-kDa band was weaker. The figure shows one representative experiment out of two.

Figure 4.

 Pulse–chase analysis of APP-C59. Neurones were transduced as described in Material and methods with SFV bearing APP-C59 construct (γ-stub), and metabolically pulse-labelled at 37°C with [35S]methionine for 10 min followed by a chase of 0–80 min (a), or labelled continuously without chase for 60–240 min (b). Cells extracts (a) and culture media (b) were immunoprecipitated using antibody B11/4, and then separated using SDS–PAGE (10%). Detection of radioactive material was performed by PhosphorImager. The figure shows one representative experiment out of three. The SEM is indicated in (c) for the first four time points.

Figure 5.

 Effect of various inhibitors on stability of APP-C59 in neurones. Neurones were transduced as described in Materials and methods with SFV bearing APP-C59 (γ40-stub) or APP-C57 construct (γ42-stub), and metabolically labelled at 37°C with [35S]methionine for 120 min in the presence of various inhibitors, followed by a chase of 60 min, still in the presence of the inhibitors. Cells extracts (a and b) were immunoprecipitated using antibody B11/4 and separated by SDS–NUPAGE (10%). Detection of radioactive material was performed by PhosphorImager. Bands corresponding to APP-C59 are indicated by an arrow on (a) and (b), and were quantified (c) and graphically represented for each condition as the ratio of intensity after the 60 min chase vs. intensity just after the labelling. The final concentrations of the inhibitors used were as follows: 100 µm lactacystine; 100 µm calpaine inhibitor 1; 100 µm calpain inhibitor 2; 20 µg/mL leupeptine; 30 mm methylamine; and 1 µm bafilomycine A1; 20 µm and 100 µm MDL28170. The figure shows one representative experiment out of two, with the variation between both experiments between 5 and 15%.

Figure 6.

 Effect of phenanthroline on stability of APP-C59 (γ-stub) in neurones. Neurones were transduced as described in Materials and methods with SFV bearing APP-C59 (γ-stub). Cells were metabolically labelled at 37°C with [35S]methionine for 120 min in the presence of various concentrations of phenanthroline. Cells were lysed and extracts were immunoprecipitated using antibody B11/4, and finally separated by SDS–PAGE (10%). Detection of radioactive material was performed by PhosphorImager. Bands corresponding to various APP-cleavage forms are indicated by arrows. Gel shown is one representative experiment out of two.

We next checked whether Fe65, a protein that binds the cytoplasmic tail of holo-APP (Fiore et al. 1995; Guenette et al. 1996; McLoughlin and Miller 1996; Russo et al. 1998) could also bind and possibly stabilize soluble APP-C59. As shown in Fig. 7, Fe65 was highly expressed, and APP-C99 and APP-C59 can be co-immunoprecipitated with Fe65 (Fig. 7). However, the degradation of APP-C59 was not influenced by the presence of Fe65 (Fig. 8). In contrast to APP-C59, Fe65 remains relatively stable during a chase in non-radioactive medium (Fig. 8).

Figure 7.

 Coimmunoprecipitation of APP-C99 and APP-C59 with Fe65. BHK cells transiently expressing Fe65 were infected with SFV bearing APP-C99 or APP-C59. After washing, cells extracts were immunoprecipitated with anti-Fe65 antibody (kindly provided by Dr Buxbaum), separated by SDS–PAGE (10%) and transferred on nitrocellulose. Western blot detection was performed either with an anti-APP carboxy-terminal antibody (B11/4, affinity purified), or with an anti-Fe65 antibody (generously given by Dr Russo). Two main Fe65 bands are observed. Gel shown is representative for one experiment out of two. Very low amounts of APP-C99 were also precipitated in cells that were not expressing Fe65, likely co-immunoprecipitating with endogenous Fe65. No APP-C59 precipitated by endogenous Fe65 was detectable, likely because of the weak level of overexpression of APP-C59, compared with APP-C99.

Figure 8.

 Effect of Fe65 on stability of APP-C59 in BHK cells. BHK cells transiently expressing Fe65 were infected with SFV bearing APP-C59. Cells were metabolically labelled at 37°C with [35S]methionine for 120 min, followed by a chase of 0–60 min. Cells extracts (a and b) were immunoprecipitated with antibody B11/4 for APP-C59, or with anti-Fe65 antibody (kindly provided by Dr Buxbaum) and separated by SDS–PAGE (10%). With this antibody one main Fe65 band is observed, while some weaker bands of lower molecular weight are also observed. Detection of radioactive material was performed using PhosphorImager. (a) Turnover of APP-C59 alone, or C59 co-expressed with Fe65. (b) Bands corresponding to APP-C59 and to the main Fe65 90-kDa band were quantified and represented graphically as the ratio of intensity after 60 min chase vs. intensity just after the pulse. The gel shown is representative for three experiments.

Finally, we tested whether overexpressed APP-C59 was able to translocate to the nucleus. We performed cell fractionation and separated cytosolic (C), membrane (M), nuclear-detergent soluble (N) and nuclear-detergent insoluble (P) fractions. We also detected endogeneous holo-APP distributing essentially in the M fraction with a minor amount in the N fraction, most likely reflecting its association with the ER membranes that are in continuity with the nuclear envelope (Fig. 9a). This was confirmed by the distribution of PS1, which is present at a ratio of about 1 : 1 in fractions M and N (Fig. 9c). We have previously shown that PS1 is present in the nuclear envelope (Annaert et al. 1999). APP-α- and β-stubs, originating from the processing of holo-APP also distribute in the M fraction (Fig. 9b). Interestingly, recombinant APP-C59 is mainly found in the M fraction, indicating that this stub is associated with the membranes of some subcellular compartments (Fig. 9b). Moreover a substantial amount of APP-C59 was also located in the P fraction. This P fraction is characterized by the presence of nuclear insoluble material such as DNA, histones (not shown), and the SP1 transcription factor (Fig. 9d). In contrast, clathrin light chain, used here as a control, is associated with the C and M fraction, in agreement with its 1 : 1 distribution between cytosol and membranes in the intact cell.

Figure 9.

 Subcellular distribution of APP-C59 (γ-stub) in neurones. Neurones were transduced as described in Materials and methods with SFV bearing APP-C59 (γ-stub). After 6-h post-infection, cells were washed and scraped in a 0.25 m sucrose solution and mechanically homogenized. Isolation of cytosolic (C), membrane (M), nuclear (N) and insoluble nuclear (P) fractions was performed as described in Materials and methods. Equivalent protein quantities of each fraction were loaded and separated on 10% SDS–PAGE, transferred on nitrocellulose and blotted with various antibodies, as described in Materials and methods. APP-γ stub is detected in the insoluble nuclear fraction where the transcription factor SP1 is also distributed. Although APP is present in the nuclear membrane fraction (probably associated with endoplasmic reticulum membranes that co-enrich with the nucleus) no signal is observed in the detergent-insoluble nuclear fraction. Gels show one representative experiment out of four.

Discussion

The carboxy-terminal fragment APP-C59, generated together with the Aβ peptide by γ-secretase cleavage of APP, has remained elusive for a long time. We investigate in the current study the fate of this fragment in BHK cells and, importantly, primary cultures of neurones. We confirm that the steady state levels of this fragment are below the detection limits of our current assay systems, even upon the overexpression of the direct precursor of this fragment (APP-C99). Significantly, the Aβ peptide, theoretically produced at equimolar concentrations, is detected under these conditions without any problem. To analyse in more detail the fate of APP-C59, we expressed the fragment directly in cell culture. This was the most rational approach, even if caution is indicated because a synthetic substitute could behave differently from the real product that is predicted to be generated by γ-secretase cleavage from its membrane-bound precursor. Similar short APP peptide fragments have been studied before by others (Lu et al. 2000; Passer et al. 2000). The construct was generated without signal peptide to direct it to the cytoplasm. Accordingly, APP-C59 is not secreted into the medium (Fig. 4b). We found that, compared with the signals obtained with APP-C99 under similar conditions, only relatively low levels of the fragment could be visualized under steady state conditions (Fig. 2c). In a pulse–chase experiment we found that the APP-C59 is rapidly degraded (half-life ∼ 5 min; Fig. 4a, quantified in Fig. 4c). Interestingly, this degradation process was not or only slightly influenced by inhibitors of proteasome or endosomal/lysosomal proteolytic function, the two major protein proteolytic degradation systems in cells. We finally observed that a fraction of APP-C59 is distributed to the nucleus, in the same fraction where the SP1 transcription factor (Suske 1999) is found.

Attempts to detect APP-C59 (endogenously generated or after overexpression of APP-C99) from crude brain extracts or neurones have been relatively unsuccesful until now, probably because the rapid turnover of this fragment preclude efficient detection. Only one report has provided evidence for a protein band reacting with APP antibodies and migrating with a similar molecular weight as predicted for APP-C59 in SDS–PAGE (Passer et al. 2000). In the same study, some fragments corresponding to APP-C59, APP-C56, APP-C55 and APP-C52 could be immunoprecipitated from brain homogenates and characterized by MALDI spectrometry (Passer et al. 2000). However, it can not be excluded that these fragments were generated during the homogenization process. On the other hand, the generation of APP γ-stubs was recently reported in an in vitro assay measuring γ-secretase processing of APP in a cell-free system (McLendon et al. 2000; Pinnix et al. 2001), providing independent evidence that the APP-C59 is a real product of γ-secretase processing of APP. In those experiments the amount of generated APP γ-stub was clearly related to the processing of APP-α- and β-stubs, and no other intermediate stubs were detected. Our current findings demonstrate that such a fragment is rapidly degraded in cells, providing an explanation for the low steady state levels of this protein in vivo. Other possibilities, for instance that this C59 fragment is rapidly degraded because it is expressed artificially, can not be completely ruled out. We think however that the observed process is physiologically relevant, and mimicking the fate of the real APP-γ-stub in vivo. Similar problems in detecting the endogenous intracellular product of γ-secretase processing of Notch, the NICD, have been encountered and solved by using overexpressing truncated versions of Notch (Schroeter et al. 1998). In analogy with these studies, we made use of a truncated version of APP, APP-C99, hoping that this would increase the amounts of APP-C59 in our cell systems. Expression of APP-C99 resulted in a sevenfold increase of secreted Aβ (Lichtenthaler et al. 1999b; Cupers et al.), confirming that this stub is a good substrate for γ-secretase. We observed an intracellular fragment with an apparent molecular weight in SDS–PAGE of 6.5 kDa, but with a slightly slower mobility than the APP-C59 or APP-C57 fragments. This 6.5-kDa fragment reacts with an antibody directed against the last 20 amino acids of the APP cytoplasmic tail, but not with antibodies against epitopes in the Aβ region, compatible with the possibility that this fragment is generated by cleavage in the transmembrane domain of APP. However, besides the difference in electrophoretic migration, the intensity of this stub was not affected by the absence of presenilin 1 (Fig. 2d), conditions under which γ-secretase activity is decreased by 80% (De Strooper et al. 1998). Because we were not able to sequence this fragment, its exact composition remains unclear. However, we think that this fragment is generated by further proteolysis of the APP-C99 or APP-C83 fragments either in the cells or during preparation of the cell extracts. The fact that this fragment displays a very similar turnover as the APP-C99 and APP-C83 fragments agrees with this interpretation (Fig. 3).

The short lifetime of APP-C59 parallels the findings with NICD in Notch signalling (Annaert and De Strooper 1999; Kopan and Goate 2000). Both Notch and APP proteins require a first cleavage in their ectodomain before undergoing a PS1-dependent γ-secretase proteolysis (De Strooper et al. 1998, 1999; Lecourtois and Schweisguth 1998; Schroeter et al. 1998; Struhl and Adachi 1998; Struhl and Greenwald 1999; Brou et al. 2000; Mumm et al. 2000). NICD, corresponding to Notch cytoplasmic tail together with some amino acids of its transmembrane domain (equivalent of the APP-C59), is difficult to detect. NICD becomes clearly identified after coprecipitation with RBP3, a member of the CSL transcription factor family. NICD is also more easily detected after deletion of the PEST sequence in its cytoplasmic tail. Fe65 is a prototype example of an APP cytoplasmic domain binding protein (Borg et al. 1996; Fiore et al. 1995; Guenette et al. 1996; McLoughlin and Miller 1996; Russo et al. 1998) and binds transcription factors of the CP2, LSF and LBP1 families (Zambrano et al. 1998). Binding to APP apparently prevents Fe65 translocating to the nucleus (Minopoli et al. 2000), and γ-secretase processing of APP could thus serve to release Fe65 for transport to the nucleus. While further work is needed, we addressed in the context of the current study two important questions. First we demonstrate that APP-C59 can bind to Fe65 and can be immunoprecipitated in a NP-40 resistant complex, suggesting that they could form a functional complex in the cell, not unlike the binding of NICD and CSL. Second, we show that increasing the pool of available Fe65 does not prevent the rapid degradation of APP-C59 (Fig. 8) in BHK cells. We conclude that Fe65 is apparently not able to prevent degradation of APP-C59. Our experiments do not exclude that other APP binding proteins can stabilize APP-C59, which will require a more comprehensive study in the future. More importantly, it also remains to be established whether the Fe65/APP-C59 complex has a role in gene regulation. An indirect argument for this possibility comes from our observation that APP-C59 distributes into an insoluble nuclear fraction as detected by cell fractionation. This insoluble nuclear fraction consisted of DNA, histones (not shown) and transcription factors, as illustrated by SP1 (Fig. 9). Besides its partial distribution in the P fraction, APP-C59 was also detected in the M fraction. APP-C59 contains an amino acid stretch of hydrophobic residues, and its association with membranes could solely reflect its hydrophobic properties. Interestingly, however, NICD bound to the Su transcription factor is also distributed both in a nuclear and membrane-bound fraction (Kidd et al. 1998).

In conclusion, our data demonstrate that, although APP-C59 is rapidly degraded in cells, a small amount seems to be more stable, and is partially associated with the nuclear fraction that contains transcription factor SP1. These data provide the first experimental evidence for the possibility that the PS1/γ-secretase cleavage of APP could be involved in a signal transduction cascade that is similar to the Notch signalling pathway. It is obvious that to substantiate this hypothesis we will need to identify target genes regulated by this hypothetical pathway. The current work provides us with the experimental base from which we can proceed.

Acknowledgements

Financial support from the Fonds voor Wetenschappelijk Onderzoek (FWO), the Katholieke Universiteit Leuven (KULeuven), The Interuniversitaire attractiepolen (IUAP) and The Vlaams Interuniversitaire Instituut voor Biotechnologie (UIB) is gratefully acknowledged. We would like to thank Drs Russo and Buxbaum for providing Fe65 antibodies. PC and WA are postdoctoral fellows and BDS is a group leader of the FWO.

Ancillary