Caspase cleavage of members of the amyloid precursor family of proteins

Authors


Address correspondence and reprint requests to Dale E. Bredesen, Buck Institute for Age Research, 8001 Redwood Boulevard, Novato, CA94945–1400, Tel. 415-209-2090, Fax: 415-209-2230, E-mail: dbredesen@buckinstitute.org

Abstract

The synapse loss and neuronal cell death characteristic of Alzheimer's disease (AD) are believed to result in large part from the neurotoxic effects of β-amyloid peptide (Aβ), a 40–42 amino acid peptide(s) derived proteolytically from β-amyloid precursor protein (APP). However, APP is also cleaved intracellularly to generate a second cytotoxic peptide, C31, and this cleavage event occurs in vivo as well as in vitro and preferentially in the brains of AD patients (Lu et al. 2000). Here we show that APPC31 is toxic to neurons in primary culture, and that like APP, the APP family members APLP1 and possibly APLP2 are cleaved by caspases at their C-termini. The carboxy-terminal peptide derived from caspase cleavage of APLP1 shows a degree of neurotoxicity comparable to APPC31. Our results suggest that even though APLP1 and APLP2 cannot generate Aβ, they may potentially contribute to the pathology of AD by generating peptide fragments whose toxicity is comparable to that of APPC31.

Abbreviations used

amyloid β-peptide

AD

Alzheimer's disease

APLPa

AAP-like protein 1

APP

amyloid precursor protein

AraC

cytosine arabinoside

BAF

Boc-aspartyl-fluoromethylketone

DMSO

dimethylsulfoxide

DP

delivery peptide

EthD-1

ethidium homodimer-1

FITC

fluorescein isothiocyanate

GFAP

glial fibrillary acidic protein

MTT

3-(4,5-dimethylthiazal-2-yl)-2,5-diphenyltetrazolium bromide

PBS

phosphate-buffered saline

RT

room temperature

TBS

Tris-buffered saline.

Cell death in the CNS occurs extensively in development, during normal aging and in some pathological states associated with degeneration of specific subsets of neurons. The majority of cell deaths in the developing nervous system occur by the activation of programmed cell death, and neural death in at least some disease states may involve programmed cell death (Bredesen 1995; Sperandio et al. 2000; Yuan and Yankner 2000). Elucidating the molecular mechanisms that initiate and control pathological cell death in the CNS should help in the development of interventions that may prevent or ameliorate degenerative CNS diseases.

The loss of hippocampal neurons is one of the prominent features of Alzheimer's disease (AD). The pathological hallmark of AD is the formation of senile plaques and neurofibrillary tangles in the brain, accompanied by substantial neuronal and synaptic loss in the neocortex. APP is a ubiquitously expressed membrane-spanning glycoprotein that is cleaved during its normal metabolism to generate the amyloid-β protein (Aβ), a 40–42 amino acid peptide that is the main constituent of senile plaques. The deposition of Aβ may account for the enhanced susceptibility of hippocampal and cortical neurons to premature death, since exposure of cultured human neuronal and non-neuronal cells to amyloidogenic Aβ peptide induces the activation of apoptotic cell death pathways (Cotman and Anderson 1995; La Ferla et al. 1995).

Work from this and other laboratories has shown that, in addition to the cleavages that result in the formation of Aβ, APP can be cleaved at its C-terminus by caspases, a family of cysteine proteases central to the execution of apoptosis (Gervais et al. 1999; LeBlanc et al. 1999; Pellegrini et al. 1999; Lu et al. 2000). In addition, work from several laboratories (Gervais et al. 1999; LeBlanc et al. 1999) has raised the possibility that the cleavage of APP precedes and may favor the intramembrane cleavage that leads to the generation of Aβ. This hypothesis, however, has not been verified in other more recent studies (Soriano et al. 2001). Independent from its effect on the generation of Aβ, recent work has demonstrated that the C-terminal fragment released by intracellular caspase cleavage of APP (APPC31) is cytotoxic (Lu et al. 2000; Dumanchin-Njock et al. 2001). Evidence for cleavage of APP at the C-terminal caspase site, D664, was obtained from brains of patients with AD, but not control patients (Lu et al. 2000). Taken together, these data suggest that the cleavage of the C-terminal portion of APP may play an important role in the neural toxicity observed in AD pathogenesis by generating a pro-apoptotic C-terminal fragment, and possibly by increasing the production of the toxic Aβ peptide. If this is indeed the case, then the accumulation of Aβ at neuronal terminals could provide the trigger for a feedback loop of toxicity by inducing the initial activation of caspases. It is possible that even low levels of activated caspases may suffice to cleave APP molecules intracellularly at the C-terminal caspase site. Cytotoxic peptides would then be released as a consequence of C-terminal cleavage of APP, further amplifying the activation of the caspase cascade. Consistent with this idea, mice expressing an APP transgene carrying two point mutations linked to autosomal forms of familial AD develop neurological symptoms and synapse loss in the absence of significant Aβ accumulation or amyloid plaque formation (Mucke et al. 2000).

Here we extend our original studies on APPC31: first, because the initial studies documented toxicity in cell lines but did not evaluate toxicity in primary neuronal cultures, we assessed such neurotoxicity. Second, we addressed the possibility that APP-like protein 1 (APLP1) and APP-like protein 2 (APLP2) may undergo cleavage events similar to that which generates APPC31.

APLP1 and APLP2 are members of the APP family of proteins. However, the sites required for γ and β-secretase cleavage of APP are not conserved in either APLP1 or APLP2. These molecules therefore do not have the capacity to generate β-amyloid-like peptides. However, the C-terminal caspase cleavage site that allows for the generation of APPC31 is conserved in both APLP1 and APLP2. This observation hinted at the possibility that either one or both of the APP family members are cleaved by caspases to generate potentially toxic peptides.

We provide evidence here that APPC31 induces apoptosis in primary neuronal cultures and that, like APP, APLP1 and APLP2 can be cleaved by caspases at their C-termini. Even though the homology between the peptide released by caspase cleavage of APLP1 (APLP1C31) and APPC31 is relatively low and is mainly restricted to a C-terminal YENPTY motif, APLP1C31 induces apoptosis in primary neuronal cultures with an LC50 similar to that of APPC31.

The present work supports the notion that, in addition to the well-established role of APP in the pathogenesis of AD, APLP1 and possibly APLP2 may release C-terminal toxic peptides and thus play a role in the neurotoxicity associated with AD.

Materials and methods

Cells and reagents

Hippocampal or cortical neurons derived from 17-day-old rat embryos were plated in modified minimum essential media (MEM-PAK) supplemented with 5% horse serum. Three days later, the cultures were treated with 10 µm cytosine arabinoside (AraC). Twenty-four hours later the cells were treated with peptide conjugates as indicated in the figures and incubated for an additional 24 or 48 h. Where indicated, cells were incubated in the presence of 50 µm of the general caspase inhibitor BOC-Asp(Ome)-FMK (BAF) for 30 min prior to addition of peptides or with an equivalent volume of dimethylsulfoxide (DMSO) and then maintained in the presence of the same concentration of the inhibitor for the duration of the experiment by adding fresh BAF at 12-h intervals.

All peptide conjugates were prepared as 1 mm and 10 mm stock solutions in water, aliquoted and frozen at − 80°C. Aliquots were thawed in ice and used promptly.

Peptide delivery into cells

Peptides were synthesized and purified at the Stanford University Protein and Nucleic Acid (PAN) Facility. All peptide stocks were solubilized in water at 1 or 10 mm concentration. The delivery peptide derived from the Drosophila Antennapedia homeodomain (C-RQIKIWFQNRRMKWKK; Dorn et al. 1999), also called penetratin (Nakagawa et al. 2000), was cross-linked via an N-terminal Cys–Cys bond to the 31 amino acid peptide generated by caspase cleavage of APP (DP–APPC31) or APLP1 (DP–APLP1C31), to fluorescein isothiocyanate (DP–FITC) or to itself (DP) at the Stanford University PAN facility. Cargo peptides are released from the carrier by reduction of the disulfide bond in the intracellular environment. We have observed no toxicity in transduction experiments using the conjugate of the DP to itself at the concentrations assayed. Additional controls are indicated in the figures and text.

Cell death assessment

Primary hippocampal or cortical neuronal cultures or 293 cells transduced with the different delivery peptide–peptide conjugates were assayed for viability at 24 and 48 h after transduction by trypan blue exclusion as described (Lu et al. 2000) or by conversion of 3-[4,5-dimethylthiazol-2-yl]-2,5-diphenyltetrazolium bromide; thiazolyl blue (MTT, Sigma, St Louis, MO, USA) to insoluble formazan in metabolically active cells and by the LIVE/DEAD assay (Molecular Probes, Eugene, OR, USA). This assay distinguishes live cells by the presence of intracellular esterase activity, which results in the conversion of the non-fluorescent cell permeant calcein–AM to the intensely green fluorescent calcein. Calcein is retained within live cells. Ethidium homodimer-1 (EthD-1) enters cells with damaged membranes and becomes intensely fluorescent when bonding to nucleic acids. EthD-1 is excluded by the intact plasma membrane of live cells. Media were removed and replaced by 4 µm EthD-1 and 2 µm calcein in phosphate-buffered saline (PBS). Images were taken 30 min after treatment. The morphology of nuclei in the cultures was examined by staining with 0.1 µg/mL Hoechst 33342. LC50 values were calculated as 100*[(T−T0)/T0] = −50 and by straight line interpolation.

In vitro protein synthesis and caspase cleavage

In vitro transcription and translation used the Promega Coupled kit (Promega, Madison, WI, USA). The constructs encoding wild-type APP or the APP D664A mutant (pC-FL-APP and pC-APPD664A), APP truncated at D664 (APP C31), pC-APLP1 and pC-APLP2 were translated and the protein products were used to assess caspase cleavage. Cleavage with caspases −3, − 6, − 7 and − 8 was done and assessed as described (Ellerby et al. 1999).

Immunostaining and image analysis

Hippocampal cultures were fixed in 4% paraformaldehyde in 1 × PBS for 20 min at room temperature (RT). Cells were then rinsed in 1 × PBS and then washed once in 1× Tris-buffered saline (TBS) followed by blocking in 10% donkey serum (Jackson ImmunoResearch Laboratories, West Grove, PA, USA) with 0.1% Triton X-100 in 1 × TBS for 1 h at RT. Cultures were incubated overnight in the presence of rabbit anti-GFAP (Sigma) at 1 : 800 dilution and mouse anti-NeuN (Chemicon, Temecula, CA, USA) at 1 : 100 at 4°C. Negative controls were incubated in 2 mg/mL rabbit and mouse pre-immune IgGs (Sigma). All primary antibodies were diluted in 1 × TBS containing 10% donkey serum. Cultures were washed for 90 min in four changes of 1 × TBS and incubated in the presence of donkey anti-rabbit IgG conjugated to Cy3 and donkey anti-mouse IgG conjugated to FITC (Jackson ImmunoResearch Laboratories, Inc., West Grove, PA, USA), at 1 : 250 and 1 : 400, respectively, in 1 × TBS containing 1% donkey serum for 1 h at RT. Cells were washed for 90 min in four changes of 1 × TBS and mounted in VectaShield-DAPI mounting medium (Vector Laboratories, Burlingame, CA, USA). Low magnification images were acquired using Nikon Eclipse-800 microscope and Optronics MagnaFire camera and software, and analyzed using Compix Simple PCI software. The total surface area corresponding to red and green fluorescence in each confocal image was determined by image analysis using Simple PCI software (Compix, Inc., Philadelphia, PA, USA). In each experiment, areas of positive immunoreactivity were identified in the control sample and used to define representative ranges of values for red and green pixels. Using these defined parameters, randomly chosen fields from each sample in a given experiment were collected and the images were analyzed by the PCI image analysis software, which calculated the total area covered by pixels matching the defined intensities for each color component. The values corresponding to total green and red fluorescent areas were averaged.

Generation of the APP-Neo antibody

An antibody that recognizes specifically the epitope generated by cleavage of APP at D664 by caspases was generated at ResGen (Invitrogen Corp., Carlsbad, CA, USA). Briefly, rabbits were immunized with the peptide 657CIHHGVVEVD664, which includes the nine amino acids immediately preceding the caspase cleavage site at position 664 in APP695, coupled to KLH. Antisera from three bleeds over a 10-week period were pooled and affinity-purified in three successive steps:

  • 1Peptide antigen was immobilized on an activated support. Antisera were passed through the column and then washed. After washing, the bound antibodies were eluted by a pH gradient.
  • 2The eluate from (1) was depleted of immunoglobulins that recognize the intact APP molecule by adsorption to a bridging peptide that encompasses the caspase cleavage site (TSIHHGVVEVDAAVTPEE).
  • 3The flowthrough from (2) was affinity-purified on the immobilized immunogenic peptide. After washing, specific antibodies were eluted by a pH gradient, collected and stored in borate buffer.

The ELISA titer for this preparation was < 1 : 142 000 (< 5 ng/mL) against the immunizing peptide (corresponding to the ‘novel’ C-terminus of APP, an epitope that is generated only after caspase cleavage) versus > 1 : 70 (> 10 mg/mL) against the bridging peptide that corresponds to the intact APP sequence across the caspase cleavage site at D664.

Immunohistochemistry

Human hippocampi obtained from AD or age-matched control patients (Harvard Brain Tissue Resource Center, Belmont, MA, USA) fixed with 4% paraformaldehyde were embedded in paraffin. Seven-micrometer microtome sections were deparaffinized in xylene, rehydrated in 100, 95, 80 and 70% ethanol, and washed in 1 × TBS for 15 min at room temperature. A 3% H2O2 solution in methanol was used to neutralize endogenous peroxidase-like activity. Microwave antigen retrieval was performed in 10 mm citrate buffer (pH 6.0) for 5 min at 440 Watts. Slides were allowed to cool to room temperature and were washed in 1 × TBS for 15 min. Samples were blocked in 10% normal horse serum in 1 × TBS for 1 h at room temperature. Primary rabbit IgG to APP-Neo was applied at a dilution of 1 : 10000 in 1% bovine serum albumin (BSA) in 1 × TBS; sections were incubated overnight at 4°C. Rabbit pre-immune IgG (Sigma) diluted to 1 µg/mL in the 1% BSA in 1 × TBS were used as a negative control. Sections were washed for 30 min in three changes of 1 × TBS; biotinylated horse anti-rabbit IgG (Vector Laboratories) was applied at a dilution of 1 : 250 for 1 h at room temperature. Peroxidase-based ABC Elite kit (Vector Laboratories) was used according to the manufacturer's instructions followed by a 30-min wash in three changes of 1 × TBS. A liquid DAB kit (Vector Laboratories) was used for the detection; color development was monitored under the microscope. Sections were washed in 1XTBS, briefly counterstained in aqueous hematoxylin, dehydrated, cleared, and mounted in Permount (Fisher Scientific, Pittsburgh, PA, USA). Images were acquired using Nikon Eclipse-800 microscope and the Optronics MagnaFire camera and software. Low magnification images were acquired using Nikon SMZ-U dissecting microscope and the CoolSnap camera and software.

Results

C31 induces death of rat hippocampal neurons in primary culture

Previous work demonstrated that the C-terminal fragment released by caspase cleavage of APP is toxic in transformed cell lines (Lu et al. 2000). To assess the cytotoxicity of APPC31 in a system more closely related to the mammalian brain, we performed experiments to determine whether APPC31 induced apoptosis in primary cultures of hippocampal neurons. Given that transfection efficiencies in neuronal primary cultures are relatively low, we used a protein transduction method. This approach allows for the introduction of polypeptides into cells with an efficiency close to 100%, and utilizes relatively stress-free conditions (Schwarze et al. 2000). We chose to use the Drosophila melanogaster Antennapedia homeodomain-derived DP, also called penetratin, linked by an N-terminal disulfide bond to the APP-derived C31 peptide or to control peptides. This delivery peptide has been used successfully in similar experiments in the past (Dorn et al. 1999; Gallouzi and Steitz 2001). Disulfide linkage was chosen over other types of covalent bond in order to allow the APPC31 peptide to be released inside the cells, in association with reduction of the S–S bond in the intracellular environment.

To assess the efficiency of transduction, primary neuronal cultures were transduced with penetratin conjugated to FITC (DP–FITC) and analyzed by confocal microscopy. The DP–FITC peptide was internalized in > 95% of the cells in the culture (Fig. 1a). We then transduced hippocampal neuronal cultures derived from 17-day-old rat embryos with 10 µm DP–APPC31 peptide or DP control. A marked decrease in viability in the DP–APPC31-transduced cultures, but not in the control cultures, was observed 24 h after transduction. The cells treated with the DP–APPC31 peptide showed prominent cytoplasmic shrinkage and an almost complete disappearance of the neuritic network (not shown). Fluorescence microscopic examination of the same cultures showed a profound reduction in the number of viable cells (cells capable of calcein retention in their cytoplasm), and a proportional increase in the number of cells with damaged membranes permeable to ethidium homodimer (EthD-1) (Fig. 1b). Essentially identical results were obtained for cortical neuronal cultures (not shown). Also, incubation of primary hippocampal neurons in the presence of increasing concentrations of DP–APPC31, but not DP alone, significantly reduced the number of viable cells (capable of converting MTT into insoluble formazan) present in the cultures (Fig. 1c).

Figure 1.

DP–APPC31 decreases the number of viable cells in primary neuronal cultures. (a) Left panels: Confocal images of HEK293 cells transduced with DP–APPC31 or with DP alone, fixed and stained with an anti-APPC31 antibody followed by anti-rabbit FITC. Right panels: Confocal images of HEK293 cells transduced with unlabelled or FITC-labeled DP. (b) Primary cultures of hippocampal neurons transduced with 10 μm of the indicated peptides were subjected to the LIVE/DEAD assay 24 h later as described in Materials and Methods. Green-fluorescent calcein is retained within living cells, while EthD-1 becomes red fluorescent when bound to nucleic acids of cells with damaged membranes. (c) Primary cultures of hippocampal neurons transduced with the indicated concentrations of DP–C31 or DP alone were assayed for viability 36 h later by the MTT assay.

Immunocytochemical examination of hippocampal cultures using antibodies specific for a neuronal marker, neuron-specific nuclear protein (NeuN), and a glial marker, glial fibrillary acidic protein (GFAP), revealed a marked decrease in the number of NeuN-immunoreactive cells present in the cultures that had been treated with 10 µm DP–APPC31 compared to cultures treated with vehicle (Fig. 2a versus Fig.2b). A slight reduction in the number of NeuN-immunoreactive cells was seen in cultures treated with delivery peptide alone (Fig. 2c). The decrease in the number of NeuN-immunoreactive cells present in cultures treated with 10 µm DP–APPC31 was prevented by the addition of the broad-spectrum caspase inhibitor BAF (Fig. 2d). To determine precisely the extent of cell death induced by APP-C31 and to evaluate the possible cell-type specifity of the peptide's toxicity we performed experiments similar to those shown in Fig. 2 and quantitated the total area of NeuN and GFAP immunoreactivity of representative fields using confocal microscopy and digital image analysis.

Figure 2.

DP–APP31 toxicity in primary neuronal cultures is dependent on the activation of caspases. Confocal images of primary hippocampal cultures pre-incubated in 50 μm BAF or in an identical volume of DMSO and then treated with: (a) water; (b) 10 μm DP–APPC31; (Cc) 10 μm DP; (d) 10 μm DP–APPC31 in the presence of 50 μm BAF. Twenty-four hours later, the cultures were fixed and immunostained with antibodies specific for NeuN and GFAP to identify neuronal and glial populations, respectively.

C31 induces programmed cell death in both neuronal and glial cells

The morphology of the cells that survived transduction with DP was suggestive of glial origin (Fig. 2). To investigate whether this was due to a greater sensitivity of neurons than glial cells to DP–APPC31-induced death, 3-day-old rat hippocampal cultures were exposed to increasing concentrations of DP–APPC31 or control DP peptide and fixed 48 hafter transduction. The fixed cultures were then immunostained with antibodies specific for GFAP (red) and NeuN (green). A quantitative assessment of the total area of red and green fluorescence present in low-magnification confocal images of representative fields obtained from three independent experiments was performed using a digital image analysis system (SimplePCI, Compix, Inc, Philadelphia, PA, USA). We found that both the neuronal and the glial population were significantly reduced in cultures treated with 10 µm DP–APPC31 when compared to untreated or control peptide-treated cultures (Fig. 3). Transduction with higher concentrations of DP–APPC31 (25 µm) were required to reduce the viability of the neuronal population further, while no further toxicity was observed for glial cells at the concentrations assayed. Incubation in the presence of the broad caspase inhibitor, Boc-aspartyl-fluoromethylketone (BAF), delayed the toxicity resulting from transduction with DP–APPC31, arguing that C31-induced neuronal death in primary cultures is caspase-mediated.

Figure 3.

Neuronal and glial cells are comparably susceptible to DP–APPC31 toxicity. Confocal images of primary hippocampal cultures transduced with (a), 0; (b), 2.5 μm ; (c) 10 μm; (d), 25 μm DP–APPC31, (e) with DP alone, or (f) with 10 μm DP–APPC31 in the presence of 50 μm BAF. Cultures were fixed and immunostained with anti-NeuN and anti-GFAP 30 h after treatment. A representative experiment is shown. (g) Total green (NeuN) and red (GFAP) fluorescent areas in primary hippocampal cultures treated as in (a–f) were determined by image analysis as described in Materials and Methods using the PCI image analysis software (Compix). Values are shown as mean ± SD (n = 3).

To resolve the discrepancy in the LC50 values obtained by the MTT assay for cell viability (Fig. 1c, 7.2 µm) and by quantitative image analysis (Fig. 3, 3.75 µm), the extent ofcell death induced by transduction of DP–APPC31 in neuronal cultures was further examined by the trypan blue exclusion method. The LC50 value obtained by trypan blue exclusion for neuronal cultures transduced with DP–APPC31 was 3.75 µm, in agreement with the value obtained by quantitative image analysis (Fig. 4a). Given that 3-day-old cultures of primary neurons were used in all experiments, it is conceivable that the higher LC50 value obtained using the MTT assay was due to variability in the proportion of glial cells present in different batches of primary neurons at the time of plating.

Figure 4.

DP–APPC31-induced death is apoptotic in nature. (a) Primary hippocampal cultures were transduced with the indicated concentrations of DP–APPC31 or DP alone. Thirty hours later, the viability of the cultures was determined by the trypan blue exclusion method. Values are shown as mean ± SD. (b) Experiments similar to those in (a) were performed using the indicated concentrations of DP–APPC31 or DP alone. Cultures were stained with 0.1 μg/mL Hoechst 33342 30 h after transduction and the percentage of condensed, fragmented nuclei were determined for five randomly chosen fields in each sample. Values are shown as mean ± SD (n = 3).

Finally, we quantitated the number of apoptotic nuclei in cultures transduced with different concentrations of DP–APPC31 by Hoechst 33342 staining. An increase in the percentage of condensed, fragmented nuclei present in neuronal cultures was observed when increasingly high concentrations of DP–APPC31 were used for transduction (Fig. 4b). This observation, together with the finding that APPC31 toxicity was delayed by caspase inhibitors, suggests that the cellular death induced by the C31 peptide was apoptotic in nature.

Caspase-cleaved APP in the brains of patients with AD and control, non-AD patients

To document the generation of C31 peptides in cultured cells and tissues, we generated an antibody capable of recognizing exclusively the novel epitope that arises by caspase cleavage of APP at its C terminus (APP-Neo). The method utilized for the generation of this antibody has been described previously (Gervais et al. 1999). We examined sections from hippocampi obtained from AD or age-matched control subjects by immunohistochemistry using the APP-Neo antibody. Hippocampal sections from AD brains showed that APP-Neo immunoreactivity, indicative of cleavage of APP at its C-terminus, is intense anteriorly in the polymorphic layer, reduced in the stratum granulosum, decreased in CA4-CA2 and absent from the stratum moleculare (Fig. 5a). APP-Neo staining was less intense at more posterior levels (Fig. 5b), but could be detected as dense deposits and in efferent fibers near CA3. Staining was abolished if the primary antibody was pre-adsorbed with the immunizing peptide (Fig. 5c) but not if it was pre-adsorbed with a peptide that encompasses the immunizing peptide sequences and the first five N-terminal amino acids of the C31 peptide, past the caspase cleavage site (bridge peptide; Fig. 5d). We observed that specific APP-Neo immunostaining occurred in the hippocampus of a 90-year-old without AD as well (i.e. control brain), but to a lesser degree, staining was low to moderate in cells and fibers of the polymorphic layer and stratum granulosum, declining in CA4–CA2 and absent from the stratum moleculare (Fig. 5e). In contrast to the AD brains, no APP-Neo staining could be detected at more posterior levels in the hippocampus (Fig. 5f). Staining was abolished by pre-adsorption with the immunogenic peptide (Fig. 5g) but not by preadsorption with bridge peptide (Fig. 5h).

Figure 5.

Truncation of APP at the C-terminal caspase site is increased in human AD brain. An antibody that specifically recognizes the novel C-terminus of APP generated by caspase cleavage (APP-Neo) but not the intact APP molecule (see Materials and methods) was used to stain hippocampal sections of (a–d and i) a 77-year-old with Alzheimer's disease and (e–h and j) a normal 90-year-old control patient. (a) In the anterior hippocampus of the 77-year-old AD patient, APP-Neo staining is intense in the polymorphic layer (P), decreased in CA4, further reduced in stratum granulosum (G) and absent in stratum moleculare (M). (b) At more posterior levels, less intense APP-Neo staining is seen in efferent fibers near CA3 (clear arrow) and sometimes in dense accumulations (black arrow). (c) DAB staining is abolished if the APP-Neo antiserum is pre-absorbed with the immunogen (shown), with immunogen plus a peptide encompassing the cleavage site and the first four amino acids of APPC31 (bridging peptide), or is replaced by buffer (not shown). (d) Immunostaining is unaffected by pre-absorbing the APP-Neo antiserum with the bridging peptide alone. (e) In the hippocampus of a normal 90 y.o., APP-Neo staining intensity is low to moderate in the polymorphic layer (P), low in CA4-CA3, even less in stratum granulosum (G), and absent in stratum moleculare (M). (f) No DAB immunostaining is seen when the APP-Neo antiserum has been preabsorbed with immunogen, or (g) with immunogen plus bridging peptide. (h) APP-Neo immunostaining is unaffected when the primary antiserum is preabsorbed with bridging peptide alone. Magnification (a–d) × 7; (e-h) × 6. Hematoxylin counterstain. Higher magnification of the 77-year-old AD hippocampus (i) shows substantial APP-Neo staining in individual cells (clear arrows) of the granular layer (G) and in most neurons (large black arrows) in the polymorphic layer (P). Extracellular deposits of APP-Neo surround these neurons, former processes (thin arrows), and vacant areas (stars), which may have formerly held neurons. The normal 90-year-old hippocampus is shown without hematoxylin counterstain (j) to reveal light APP-Neo staining in isolated granular layer cells (clear arrow), polymorphic layer neurons (small black arrow), and extracellular deposits along processes (thin arrows). Use of immunogen-absorbed APP-Neo antibody in controls eliminated DAB staining, but not the lipofuscin pigment in many neurons of this elderly individual (large black arrows). Magnification (i and j) × 200.

Closer examination revealed that the pattern of staining intensity reflected uneven distribution of APP-Neo immunoreactivity. In the AD hippocampus, cytoplasmic staining was seen in occasional granular layer neurons and most polymorphic neuronal cell bodies. Extracellular deposits of APP-Neo surrounded immunoreactive neurons, neuronal processes, and vacant areas which may have formerly held neurons (Fig. 5i). Immunostaining in hippocami of normal controls was similar but less intense. APP-Neo staining occurred in isolated granule cells, polymorphic layer neurons, and small foci along cellular processes (Fig. 5j). Extracellular deposits were rarely clustered around polymorphic layer neurons or open areas of tissue. Control sections incubated in immunogen-absorbed APP-Neo antibody helped identify and confirm specific staining, and to differentiate it from lipofuscin pigment in older individuals.

APLP1 and APLP2 may be cleaved by caspases

Three members of the APP family of proteins exist: APP, APLP1, and APLP2. Even though the overall similarity of the APP family C-termini is not high, the caspase cleavage site that is required for the generation of APPC31 is completely conserved in all three members. If the DEVD sequences in APLP1 and APLP2 can function as caspase cleavage sites, both proteins could potentially generate C-terminal peptides. It should be noted, however, that the P1′ position in APP is Ala (VEVDA), whereas the P1′ position in APLP1 is Pro (VEVDP) as it is in APLP2. Caspases tend to prefer less bulky residues such as Gly, Ala, or Ser, in the P1′ position, rather than more bulky residues such as Pro (Stennicke et al. 2000). Therefore, at least in theory, the VEVD site in APP should be more readily cleaved by caspases than the sites in APLP1 and APLP2. To determine whether APLP1 and APLP2 can be cleaved by caspases, we assayed a panel of recombinant caspases for their abilities to cleave 35S-labelled, in vitro transcribed/translated APP, APLP1 and APLP2. The results shown in Fig. 6(a), indicate that APP can be cleaved by caspases-3 and -6, but not by caspase-8. APLP1, on the other hand, was cleaved in vitro only by caspase-3 (Fig. 6b), not by caspase-6 or -8. Like APLP1, APLP2 may be cleaved in vitro by caspase-3 only, but with very low efficiency, if at all (Fig. 6b). The 35S-Met-labelled C31 peptide product of the cleavage of APP by caspase-3 and -6 (Fig. 6a) was detected as a ∼ 4 kDa band. However, we could not detect the homologous peptide generated by caspase-3 cleavage of APLP1. Our inability to detect APLP1C31 could be due to the fact that only one methionine (of a total of two in APPC31) is conserved in APLP1C31.

Figure 6.

APLP1 and APLP2 are cleaved by caspases in vitro. (a) Full-length APP and APPΔC31 were in vitro-transcribed and translated in the presence of 35S-Met. The reactions were then subjected to cleavage by the indicated caspases in vitro. Arrows indicate cleavage products. (b) Full-length APLP1 and APLP2 were in vitro-transcribed and translated in the presence of 35S-Met. The reactions were then subjected to caspase cleavage as in (a). Untreated full-length APP and APPΔC31 were included as references. (c) FL-APP or APPΔC31 were expressed from the CMV promoter in HEK293 cells untreated or treated with staurosporine in the presence or absence of BAF. Lysates were electrophoresed, blotted and reacted with an antibody specific for APP (5A3). Arrows indicate full-length and cleaved forms of APP. (d) APLP1-FLAG and APLP2-FLAG were expressed from the CMV promoter in HEK293 cells treated and processed as in (b). Blots were reacted with anti-FLAG antibody (M2, Sigma).

To determine whether cleavage of APLP1 and APLP2 can occur in cultured cells, we transfected CMV-driven constructs expressing N-terminally FLAG-tagged APLP1 and APLP2 or a full-length APP construct in 293 cells and activated the caspase cascade by treatment with staurosporine. Both APP and APLP1 were cleaved in staurosporine-treated 293 cells (Figs 6c and d) and in both cases, cleavage was prevented by incubation of the cells in the presence of BAF. We were able to detect both the full-length and the truncated forms of APP (Fig. 6c, arrows). Full-length APLP1 appeared to be completely degraded in 293 cells treated with staurosporine, but not when BAF was present (Fig. 6d). No cleavage products of FLAG-APLP1 could be detected in these cultures. Both in vitro and in transfected 293 cells, we found that caspases could cleave APP and APLP1 at more than one site. No evidence was found for the cleavage of FLAG-APLP2 in transfected 293 cells (Fig. 6d).

To determine whether APLP1 is effectively cleaved at position 664, we took advantage of the selective reactivity of the APP-Neo antibody. Given that the five amino acids that constitute the novel C-termini in cleaved APLP1 and APLP2 are relatively conserved, epitopes could be generated that might be recognized selectively by APP-Neo after caspase cleavage. To determine whether APP-Neo-immunoreactive epitopes are generated by caspases in APP, APLP1 and APLP2, we incubated unlabelled, in vitro transcribed/translated full-length APP (APP695), APPD664A (a mutant of APP in which the D residue at position 664 has been replaced by A), and full-length APLP1 and APLP2 in the presence of recombinant caspases. The products of the reactions were separated on polyacrylamide gels and immunoblotted with APP-Neo antibody. As a control, we performed immunoblots using lysates from 293 cells transfected with full-length APP695, with an APP construct lacking the APP C-terminal 31 amino acids (APPΔC31), or with APP695 and treated with 10 µm staurosporine, in the presence or absence of 50 µm BAF. As expected, APP-Neo-immunoreactive bands were detected only in lysates from 293 cells expressing APPΔC31 and in lysates from cells expressing APP695 and treated with staurosporine in the absence of BAF (Fig. 7a). Also, an APP-Neo-immunoreactive epitope was detected in immunoblots of in vitro transcribed/translated full-length APP that had been incubated in the presence of recombinant caspase-3 (Fig. 7a), but not caspase-7 or -8 (Fig. 7a). Likewise, in vitro transcribed/translated APLP1 and APLP2 yielded APP-Neo-immunoreactive cleavage products only when incubated in the presence of recombinant caspase-3 (Fig. 7b). Even though the caspase-cleaved APLP2 species was detected as a very faint 35S-labeled band in in vitro cleavage assays, the higher sensitivity of the APP-Neo antibody enabled us to detect the low-abundance, cleaved species in western blot assays. Providing a control for the specificity of the reaction, a mutant form of APP that cannot be cleaved by caspases, APPD664A, did not yield detectable APP-Neo-immunoreactive products after incubation with recombinant caspase-3, -7 or -8.

Figure 7.

Caspase cleavage of APP, APLP1 and APLP2 generates an APP-Neo-immunoreactive epitope. (a) Left panel: Full-length APP or APPΔC31 were expressed from the CMV promoter in HEK293 cells untreated (mock) or treated with staurosporine in the presence or absence of BAF. Lysates were subjected to immunoblotting using the APP-Neo antibody as described in the text. Right panel: In vitro transcribed/translated 35S-Met labeled full-length APP was subject to cleavage by caspases as indicated and immunoblotted using APP-Neo. (b) In vitro-transcribed/translated 35S-labelled APLP1 (left panel) and APLP2 (right panel) were subjected to cleavage with the indicated caspases in vitro, blotted and reacted with APP-Neo.

APLP1C31 induces death in primary hippocampal cultures

The results shown above suggest that APLP1 can be cleaved by caspase-3 at the aspartic acid residue at position 620. If this event occurs in vivo, APLP1 would have the potential to generate a pro-apoptotic C-terminal peptide homologous to APPC31. To determine whether the peptide generated by caspase cleavage of APLP1 is toxic, we generated a fusion of APLP1C31 to the Antennapedia delivery peptide (DP–APLP1C31) and assayed it in protein transduction experiments similar to those presented in Figs 1–4 using rat embryonic hippocampal cultures. Three-day-old rat hippocampal cultures were exposed to increasing concentrations of DP–APLP1C31 or control peptide, fixed 36 h after transduction and immunostained with antibodies specific for GFAP and NeuN. A quantitative assessment of the relative areas of red (GFAP) and green (NeuN) fluorescence present in low-magnification confocal images was performed using a digital image analysis system (SimplePCI, Compix, Inc.). We found that the NeuN-immunoreactive population was markedly reduced in cultures treated with 10 µm DP–APLP1C31 (Fig. 8a). At higher concentrations of DP–APLP1C31 (25 µm), the viability of the neuronal population was reduced further. Surprisingly, we saw only a modest decline in the viability of glial cells, which may have been due to a relatively higher sensitivity of neurons to APLP1C31 toxicity. Incubation in the presence of the broad caspase inhibitor, BAF, delayed DP–APLP1C31 toxicity, arguing that cell death induced by APLP1C31 depends on caspase activity.

Figure 8.

Transduction of DP–APLP1C31 induces death in primary hippocampal cultures. (a) Primary hippocampal cultures preincubated in 50 μm BAF or in an identical volume of DMSO were treated with increasing concentrations of the indicated peptides in the presence or in the absence of BAF. Twenty-four hours later, the cultures were fixed and immunostained with antibodies specific for NeuN and GFAP to identify neuronal and glial populations, respectively. Total green (NeuN) and red (GFAP) fluorescent areas in the cultures were determined by image analysis as described in Materials and Methods using the PCI image analysis software (Compix). The values shown are the average of two independent experiments. (b) Primary hippocampal cultures were transduced with the indicated concentrations of DP–APLP1C31 or DP alone. Thirty hours later, the viability of the cultures was determined by the trypan blue exclusion method. Values are shown as mean ± SD (n = 5).

The extent of cell death induced by transduction of DP–APPC31 in neuronal cultures was further examined by thetrypan blue exclusion method. As shown in Fig. 8(b), a dose-dependent reduction in the viability of the cultures was observed at increasing concentrations of transduced DP–APLP1C31 but not of control DP peptide. The LC50 value obtained for neuronal cultures transduced with DP–APPC31 was 4 µm, while the value obtained by image analysis was approximately 5 µm (Fig. 8a).

Discussion

Our data indicate that the APP fragment that is generated by caspase cleavage of the APP C-terminus at Asp664 is toxic to hippocampal and cortical neurons in primary culture. The presence of this peptide in primary neuronal cultures triggers the activation of programmed cell death, as demonstrated by the condensation and fragmentation of nuclei in transduced cells and by the ability of the general caspase inhibitor BAF to delay the death process. This finding is compatible with our earlier finding that caspase-8 and caspase-9, but not caspase-3, were required for C31-induced cell death (Lu et al., submitted for publication). The biochemical pathway(s) leading from C31 to caspase-8 and -9 and apoptosis activation is not yet known. However, it is compatible with the previous finding of caspase-9 activation in synaptosomal preparations from the brains of patients with AD, but not from control patients (Lu et al. 2000). It is important to add, however, that there is no evidence that apoptosis, as classically defined (Kerr et al. 1972), represents the predominant mode of cell death in neurons in AD. One possible explanation for this apparent discrepancy is that, whereas immature neurons may be induced to undergo apoptosis readily, more mature neurons are more resistant to at least some pro-apoptotic insults (Yakovlev et al. 2001). Thus, insults demonstrated to be pro-apoptotic in neurons in primary culture, such as Aβ, APP-C31 and APLP1-C31, may turn out to induce non-apoptotic cell death in mature neurons in vivo.

To document the caspase-mediated cleavage of APP, which provides indirect evidence of the release of the toxic C31 peptide, we used an antibody that selectively recognizes the neo-C-terminus generated by cleavage of APP at position 664 (APP-Neo). Immunohistochemical examination of brain tissues from AD patients and age-matched controls revealed a pronounced accumulation of the caspase-cleaved form of APP in aggregates adjacent to neuronal cell bodies and in the perikaryal cytoplasm of some neurons in AD, but to a lesser extent in the normal, age-matched control sections.

The evidence, taken as a whole, suggests that APP is cleaved both in cultured cells and in vivo, releasing not only the Aβ peptides, but also APPC31, a neurotoxic peptide. Thus, the C31 peptide is a good candidate to play a role in the death of neurons associated with AD. It should be added that recent work from the d'Adamio Laboratory has shown that the APPC57 peptide, which results from γ-secretase cleavage, may also be cytotoxic (Passer et al. 2000). However, it is not yet clear whether generation of C31 is required for C57 toxicity, as was previously demonstrated for C100 (Lu et al. 2000).

The APP-related gene products, APLP1 and APLP2, demonstrate putative caspase cleavage sites that liberate C31 fragments: for APLP1, the P4-P1′ positions are VEVDP, and for APLP2, the P4-P1′ positions are VEVDP, while in APP, the P4-P1′ positions are VEVDA. These sequences, like those in APP, fit well with previously described caspase cleavage sites for the initiator/apical caspases such as caspase-8 and caspase-9, except at the P1′ position. We have shown here that APLP1, and to a lesser extent APLP2, may be cleaved by caspases. The APLP2 cleavage efficiency is likely to be lower than that for APLP1, given that caspase cleavage products were only detected by immunoblot with a cleavage-specific antibody. The observation that caspase-cleaved APLP1 and APLP2 were recognized by APP-Neo in immunoblots raises the intriguing possibility that potential neo-epitopes derived from APLP1 and APLP2, in addition to APP, could be generated in brain tissues and detected by APP-Neo (Fig. 5). The APLP1C31 peptide is 52% identical and 82% similar to APPC31, while the predicted APLP2C31 is 71% identical and 83% similar to APPC31. In the case of APLP1C31, however, even though identity is mainly restricted to the YENPTY motif, the peptide released by cleavage of APLP1 C-terminus shows a level of toxicity in primary neuronal cultures comparable to that of APPC31. This observation hints at the possibility that toxicity may be mediated by interference, or at least modulation, of the endocytic signal that regulates APP turnover and Aβ secretion (Perez et al. 1999).

We propose that low levels of caspase activation may occur in neurons, particularly in neuronal terminals exposed to mild stress, such as the accumulation of insoluble Aβ peptides. These low levels of caspase activation may be sufficient to cleave APP family members present in synaptic terminals but not be sustained or high enough to propagate the apoptotic signal to the soma. A considerable number of experiments reported in the literature seem to support this idea (Mattson et al. 1998; Chan and Mattson 1999) and also suggest that apoptotic signals in neurons may be continuously ‘dampened’ by pro-survival signals that protect them from stress-induced death (Bergmann et al. 2002; Mattson 2000). Locally, transiently activated caspases could thus support the generation of toxic C-terminal peptides. Even though the mechanism underlying the toxicity of these C-terminal peptides is still unknown, recent work has suggested that they may be involved in the modulation of transcriptional activity of APP-binding proteins (Cao and Sudhof 2001; Cupers et al. 2001; Kimberly et al. 2001). Our results indicate that the toxicity of C-terminal peptides derived from members of the APP family is dependent on the activity of caspases, which suggests that a ‘feedback’ mechanism of caspase activation may exist. We hypothesize that strong pro-survival signals operate in neurons that inhibit this process for prolonged periods of time but that may be eventually overridden by the sustained activation of caspases, eventually leading to neuronal death. It is also possible that the classic apoptotic pathway is not completed in neurons in vivo, and that the mechanisms responsible for the eventual demise of the cell may involve stress-response-related, non-apoptotic pathways.

Acknowledgements

Supported by the National Institute of Neurological Disorders and Stroke (NS35155 and NS33376 to D.E.B); by the National Institute on Aging (AG05131 to E.H.K); by the UCSD Alzheimer's Disease Research Center and the Joseph Drown Foundation.

Ancillary