Endoplasmic reticulum stress-induced cysteine protease activation in cortical neurons: effect of an Alzheimer's disease-linked presenilin-1 knock-in mutation.

Endoplasmic reticulum (ER) stress elicits protective responses of chaperone induction and translational suppression and, when unimpeded, leads to caspase-mediated apoptosis. Alzheimer's disease-linked mutations in presenilin-1 (PS-1) reportedly impair ER stress-mediated protective responses and enhance vulnerability to degeneration. We used cleavage site-specific antibodies to characterize the cysteine protease activation responses of primary mouse cortical neurons to ER stress and evaluate the influence of a PS-1 knock-in mutation on these and other stress responses. Two different ER stressors lead to processing of the ER-resident protease procaspase-12, activation of calpain, caspase-3, and caspase-6, and degradation of ER and non-ER protein substrates. Immunocytochemical localization of activated caspase-3 and a cleaved substrate of caspase-6 confirms that caspase activation extends into the cytosol and nucleus. ER stress-induced proteolysis is unchanged in cortical neurons derived from the PS-1 P264L knock-in mouse. Furthermore, the PS-1 genotype does not influence stress-induced increases in chaperones Grp78/BiP and Grp94 or apoptotic neurodegeneration. A similar lack of effect of the PS-1 P264L mutation on the activation of caspases and induction of chaperones is observed in fibroblasts. Finally, the PS-1 knock-in mutation does not alter activation of the protein kinase PKR-like ER kinase (PERK), a trigger for stress-induced translational suppression. These data demonstrate that ER stress in cortical neurons leads to activation of several cysteine proteases within diverse neuronal compartments and indicate that Alzheimer's disease-linked PS-1 mutations do not invariably alter the proteolytic, chaperone induction, translational suppression, and apoptotic responses to ER stress.

The endoplasmic reticulum (ER) 1 is responsible for the syn-thesis, initial post-translational modification, and proper folding of proteins, as well as their sorting and export for delivery to appropriate cellular destinations. A variety of conditions, such as loss of the ER intraluminal oxidative environment or calcium content or the mutation or overexpression of relatively insoluble proteins, cause accumulation of misfolded proteins within the ER (1). Protein misfolding triggers three compensatory responses. One is the unfolded protein response (UPR), involving increased expression of molecular chaperones such as Grp94 and Grp78/BiP that promote proper protein folding (2), as well as SERCA2, an energy-dependent transporter for loading the ER lumen with calcium (3). A second response is the generalized suppression of translation mediated by the serine/ threonine kinase PERK, which phosphorylates and inactivates the translation initiation factor eIF2␣ (4,5). A third is ERassociated degradation, in which misfolded proteins are expelled from the ER and targeted for degradation by cytoplasmic proteasomes (6,7). Although these three protective responses control transiently the accumulation of misfolded proteins within the ER, they can be overcome by sustained ER stress, which leads to apoptosis (8,9). Molecular genetic and pharmacologic experiments have established that cells deficient in the UPR, ER-associated degradation or translational suppression responses are more vulnerable to ER stress-induced apoptosis (10 -12), whereas increased expression of genes involved in the protective responses reduces apoptosis (13,14).
Alzheimer's disease (AD) is a slowly progressive cognitive and behavioral brain disorder characterized by neurodegeneration and abnormal accumulation of protein into neuritic plaques and neurofibrillary tangles. New insights into the pathogenesis of AD have been provided by the linkage of an inherited early onset form of AD to mutations in the presenilins (15,16), two closely related genes (PS-1 and PS-2) encoding polytopic integral membrane proteins that reside in the ER, intermediate compartment and Golgi (17)(18)(19). Several potential pathogenic mechanisms have been proposed for presenilin mutations, including impairments in ER stress signaling and increased sensitivity to stress-induced apoptosis. Evidence has been presented that mutant PS-1 enhances ER stress-induced apoptosis by reducing the UPR and PERK-mediated translational suppression not only in the context of overexpressing cell lines but also for primary mouse neurons bearing an AD-linked PS-1 knock-in mutation (13). Reduced responsiveness to ER stress may be particularly germane to AD, because the principal constituent of the insoluble neuritic plaques, the 42-residue amyloid A␤ protein, is synthesized partly in the ER and intermediate compartment (20,21) and may progressively accumulate there (22,23), where it can trigger ER stress responses and apoptosis (24). Nevertheless, the role of altered ER stress responsiveness as a neuropathogenic mechanism for presenilin mutations is uncertain. Another study failed to demonstrate an involvement of PS-1 or an AD-linked PS-1 mutant in protective responses to ER stress (25). Furthermore, signs of ER stress or aberrant apoptosis have not been reported so far for transgenic mouse lines expressing pathogenic presenilin mutations (25,26). Finally, the influence of mutant PS-1 on ER stress-induced apoptotic signal transduction has not been investigated.
The signaling pathway that initiates ER stress-induced apoptosis has begun to be defined, and it involves activation of cysteine proteases distinct from those that trigger the widely studied mitochondrial and death receptor apoptotic pathways. ER stress-mediated apoptosis is dependent on an ER-resident cysteine protease, caspase-12 (24), for which activation may be regulated by processing of the procaspase-12 zymogen by the calcium-dependent cysteine protease calpain (27) and by zymogen clustering mediated by TRAF2 (28). There are conflicting reports on the role of mitochondrial release of cytochrome c in ER stress-induced apoptosis (29 -31), and little is known about how caspase-12 activity leads to the execution of apoptosis. We describe here the further analysis of the cysteine protease activation responses to ER stress. The study has been carried out in primary cortical neurons and fibroblasts derived from mice, either wild type for PS-1 or homozygous for a pathogenic P264L knock-in mutation in PS-1 (32). After characterizing the activation of calpain and caspases in response to ER stress, we investigated the influences of the mutant PS-1 on proteolysis, apoptosis, and protective responses of UPR and PERK activation.

EXPERIMENTAL PROCEDURES
Antibodies and Other Materials-Dulbecco's modified Eagle's medium, Neurobasal, trypsin, and B27 supplement were from Life Technologies, Inc. Fetal calf serum was from Hyclone. Tunicamycin, thapsigargin, staurosporine, Hoechst 33342, and soybean trypsin inhibitor were from Sigma. Calpeptin was from Calbiochem. Ac-Asp-Glu-Val-Asp-aldehyde was from Enzyme Systems Products. Supplies for immunoblotting were from DuPont. Horseradish peroxidase-coupled secondary antibodies were from Santa Cruz Biotechnology. Biotinylated secondary antibodies and avidin-biotin-peroxidase complex were from Vector. Anti-caspase-12 was a rat monoclonal antibody, generously provided by Drs. T. Nakagawa and J. Yuan (Boston). Anti-caspase-3 was a monoclonal antibody (clone 46) purchased from Transduction Laboratories. Anti-Grp94 and anti-Grp-78/BiP were from Stressgen. Anti-cytochrome c was from PharMingen. Anti-actin (clone 4) was from Roche Molecular Biochemicals. Anti-active caspase-3 was a rabbit antiserum (Ab206) that reacts selectively with the p17 large subunit, but not with procaspase-3, and was prepared and characterized in this laboratory previously (32). Anti-calpain-cleaved ␣-spectrin (Ab38) reacts specifically with the ϳ150-kDa ␣-spectrin derivative formed by either calpain I or II but does not react with ␣-spectrin derivatives produced by caspases or other proteases, and it has been characterized extensively (32,33). Anti-PERK was an affinity-purified rabbit antibody generously provided by Dr. Takashi Kudo (Osaka, Japan). An antiserum reactive specifically with the caspase-6-derived NH 2 -terminal fragment of lamin A (Ab255) was prepared as follows. The hexapeptide corresponding to the COOH terminus of the lamin A fragment (RLVEID) was synthesized along with an NH 2 -terminal cysteine residue by Research Genetics. It was coupled via its cysteine residue to keyhole limpet hemocyanin using the heterobifunctional coupling agent maleimidobenzoyl-N-hydroxysuccinimide ester (Pierce), and the conjugate was used to immunize rabbits and generate the Ab255 antiserum.
Mouse Lines and Cell Culture-Primary cortical neuronal cultures were derived from E17 mice of the 129/CD-1 outbred strain, either wild type for PS-1 or homozygous for the PS-1 P264L knock-in mutation. The latter mouse line was generated by a two-step gene-targeting strategy that first introduced a base change into exon 8 encoding the P264L mutation and subsequently used the Cre-loxP system to remove the drug selection cassette from the upstream intron. This gene-targeting strategy enables the Alzheimer's disease-linked P264L mutation to be introduced into the endogenous mouse PS-1 gene, such that the mutant gene is expressed at normal levels and with endogenous regulatory controls over splicing and expression. The derivation and initial phe-notypic characterization of these mice, including PS-1 gene expression in the brain, has been described previously (32).
Dissociated monolayer cultures of essentially pure mouse cortical neurons were established using the serum-free Neurobasal/B27 medium formulation (34) as described previously (32). Cells were plated onto polyornithine/laminin-coated 6-well plates or 100-mm dishes at 60,000 cells/cm 2 and maintained for 7-14 days in vitro. Embryonic fibroblasts were cultured from PS-1 wild-type and PS-1 P264L homozygous knock-in mice using conventional procedures. Briefly, E17 embryos were decapitated, eviscerated, and minced into small pieces. Cells were dissociated by trypsinization followed by passage through narrowtip pipettes; large clumps were removed, and the remaining cells were plated onto plastic dishes at 5 ϫ 10 4 cells/cm 2 in medium consisting of Dulbecco's modified Eagle's medium, 10% fetal calf serum. Nonadherent cells were discarded after 1 h, and the medium was replaced every second day thereafter. When near confluency, fibroblasts were replated using mild trypsinization. Experiments were performed on secondary fibroblasts grown to ϳ50% confluency.
For immunocytochemistry, cultures were rinsed in phosphate-buffered saline, fixed in ice-cold 4% paraformaldehyde in phosphate-buffered saline for 30 min, permeabilized with 0.05% Triton X-100, and stained by the biotin-avidin-peroxidase technique (33,35). For detecting neuronal apoptosis, fixed cultures were permeabilized and were then stained with Hoechst 33342 at 37°C for 15 min, rinsed, and examined under ultraviolet illumination. Images were captured using a Nikon Diaphot microscope and Spot CCD (charge-coupled device) camera. Quantitative analysis of neuronal death was conducted using neurons cultured for 14 days and trypan blue staining as described previously (32). For each experimental treatment, five cultures of each genotype were evaluated, and for each culture three randomly chosen fields and Ͼ300 cells were evaluated. The apoptotic index was calculated as the number of trypan blue-positive cells/total number of cells.
Subcellular Fractionation-Cultured neurons from two 100-mm dishes were washed twice in ice-cold homogenization buffer consisting of 0.25 M sucrose, 50 mM Pipes-KOH (pH 7.2), 5 mM KCl, 1 mM MgCl 2 , 3 mM 2-mercaptoethanol, and 5 mM EGTA plus protease inhibitors (100 M leupeptin, 2 M pepstatin A, 30 M calpain inhibitor II). Cells were collected by gentle scraping in homogenization buffer and centrifugation and were then broken in a small glass homogenizer with 60 strokes of a loose-fitting glass pestle using homogenization buffer lacking the protease inhibitors. The homogenate was centrifuged for 8 min at 1000 ϫ g to pellet nuclei and unbroken cells, and then the supernatant was centrifuged 15 min at 10,000 ϫ g. The pellet was taken as the crude mitochondrial/synaptosomal fraction. The supernatant was centrifuged for 60 min at 100,000 ϫ g, and the resulting pellet was taken as the ER-enriched crude microsomal fraction, whereas the supernatant was taken as the cytosolic fraction. To analyze the composition of the fractions, 10 g of protein from each fraction was subjected to SDS-PAGE and immunoblotting to detect procaspase-12, cytochrome c, or Grp94.
To evaluate the sensitivity of procaspase-12 to calpain cleavage, a post-mitochondrial supernatant was prepared from mouse brain. Briefly, brains from 9-day-old mice were homogenized in 5 volumes of homogenization buffer and centrifuged as described above for cultured cells. The supernatant from the 10,000 ϫ g centrifugation was taken as the post-mitochondrial fraction. It was incubated 60 min at 37°C in either 5 mM CaCl 2 to activate the endogenous calpain (36) or 5 mM EGTA and was then prepared for SDS-PAGE and immunoblotting. Some reactions contained calpeptin at 30 M or Ac-Asp-Glu-Val-Aspaldehyde at 5 M.
RT-PCR-Levels of Grp78/BiP mRNA were determined by semiquantitative RT-PCR analysis (25) of mRNA isolated from cultured mouse cortical neurons. Cultures were treated for 6 h with vehicle (0.1% Me 2 SO), tunicamycin (0.5 g/ml or 3 g/ml), or thapsigargin (0.2 M or 1 M). Each 100-mm dish was washed twice in Dulbecco's phosphatebuffered saline, and cells were scraped gently into the same buffer and collected by centrifugation. Total RNA was isolated using Trizol reagent (Life Technologies, Inc.). Aliquots of reverse-transcribed RNA were subjected to PCR using primer pairs specific for Grp78/BiP (5Ј-CTGGG-TACATTTGATCTGACTGG-3Ј and 5Ј-GCATCCTGGTGGCTTTCCAG-CCATTC-3Ј and glyceraldehyde-3-phosphate dehydrogenase (5Ј-GAT-GACATCAAGAAGGTGGTGAAG-3Ј and 5Ј-GTGAGGGAGATGCTCA-GTGTTGG-3Ј). To establish a linear amplification range, equivalent aliquots were incubated in a PCR, and individual reactions were removed at 16, 18, 20, and 22 cycles. Based on the results, subsequent PCRs were incubated for 22 cycles. The PCR products were fractionated on agarose gels, stained, and quantified by fluorescent imaging (Typhoon 8600 Variable Mode Imager, Molecular Dynamics).

RESULTS
The blockade of N-linked glycosylation by tunicamycin leads to misfolding of newly synthesized proteins in the ER (2) and elicits the ER stress responses of chaperone induction (the UPR), translational suppression, and caspase-mediated apoptosis in a variety of cultured cells, including neurons (8,9). We examined the effects of tunicamycin on primary cortical neurons cultured from mice that were either wild type for PS-1 or homozygous for the AD-linked point mutation PS-1P264L, which was introduced into the mouse genome by gene targeting (PS-1 knock-in (32)). The PS-1 P264L knock-in mutation is phenotypically active in the mouse, increasing the secretion of the amyloid A␤42 protein from primary cortical neurons, elevating A␤42 levels in the brain, and accelerating amyloid neuropathology in afflicted brain regions (32). Treatment of primary cortical neurons with tunicamycin for 24 h (9) led to apoptosis characterized by perikaryal shrinkage, retraction of processes, chromatin condensation, and fragmentation, features that did not differ qualitatively between neurons of the two PS-1 genotypes (Fig. 1A).
To examine the involvement of particular cysteine proteases in tunicamycin-induced neuronal apoptosis, a panel of antibodies was employed for immunoblot and immunohistochemical detection of protease activation and substrate degradation. The p17 large subunit of caspase-3 is generated by proteolytic processing of the proprotease zymogen (37,38). An antibody that reacts specifically with the activated caspase-3 p17 subunit readily immunostained cortical neurons following tunicamycin treatment but exhibited little neuronal labeling under control conditions (Fig. 1B). Active caspase-3 immunoreactivity extended throughout neuronal perikarya into multiple dendritic processes and their distal branches. Activation of caspase-6 was monitored using a cleavage site-specific antibody (Ab255) reactive with the NH 2 -terminal fragment of lamin A, a nuclear matrix protein and preferential caspase-6 substrate that is cleaved during apoptosis (39,40). Whereas little immunoreactivity for caspase-6-cleaved lamin A was observed in neurons under basal conditions, ER stress produced intense immunolabeling concentrated in neuronal nuclei. These data demonstrate that ER stress in cortical neurons causes the activation of at least two effector caspases and spreads from the ER to the cytosol and nucleus.
ER stress induces apoptosis through a signaling pathway thought to be distinct from the "mitochondrial" and "death receptor" cascades triggered by caspase-9 and -8, respectively, and instead may be initiated by calpain (27), a family of calciumdependent cysteine proteases (41). Subsequently, calpain reportedly cleaves and activates an ER-resident protease caspase-12 (24), triggering apoptosis through a poorly defined mechanism. We investigated this signaling pathway by subcellular fractionation and immunoblotting, and found that the ϳ60-kDa procaspase-12 was concentrated in the ER-enriched microsomal fraction ( Fig. 2A). The fidelity of the fractions was confirmed by enrichment of cytochrome c in the mitochondrial fraction ( Fig. 2A) and of calpain I in the cytosol fraction. 2 To determine whether calpain can process procaspase-12, we incubated post-mitochondrial supernatants from mouse brain with calcium to activate the endogenous calpains I and II (36). Calpain activation diminished the level of procaspase-12 and produced a smaller ϳ35-kDa immunoreactive fragment (Fig.  2B), essentially identical in size to the calpain derivative of procaspase-12 reported previously (27). Calpain activation also caused degradation of ␣-spectrin, detected with a cleavage sitespecific antibody reactive only with the calpain-derived NH 2terminal ϳ150-kDa fragment. The loss of procaspase-12, appearance of the smaller derivative, and cleavage of ␣-spectrin were all blocked by calpeptin, a peptidyl aldehyde calpain in-2 R. Siman and R. W. Neumar, unpublished observations. FIG. 1. ER stress-induced apoptosis and caspase activation in primary mouse cortical neurons. Cortical neuron cultures were derived from mice either wild type for PS-1 or homozygous for a P264L knock-in mutation. Some cultures were treated for 24 h with tunicamycin, and then the neurons were either stained with Hoechst 33342 to visualize nuclear morphology (A) or immunostained with antibodies to active caspase-3 or a caspase 6-derived fragment of the nuclear matrix protein lamin A (B). A, whereas few neurons under basal conditions exhibit chromatin condensation and fragmentation indicative of apoptosis, many neurons of both PS-1 genotypes underwent apoptosis during ER stress. B, immunoreactivity for active caspase-3 was abundant in tunicamycin-treated PS-1 wild-type neurons, in which it filled the perikarya and extended into multiple neuritic processes. Immunoreactivity for caspase 6-cleaved lamin A also became prominent during tunicamycin treatment and was concentrated in neuronal nuclei.
hibitor, but not by Ac-Asp-Glu-Val-Asp-CHO, a peptidyl aldehyde caspase-3 inhibitor. Our results confirm that procaspase-12 is enriched in the ER fraction and is processed by calpain.
Immunoblotting was used to evaluate the processing of procaspase-12 and activation of calpain, caspase-3, and caspase-6 following ER stress to cortical neurons and to determine whether the PS-1P264L knock-in mutation alters apoptotic protease activity. In addition to tunicamycin, ER stress was induced by the sarcoplasmic-endoplasmic reticulum calcium ATPase inhibitor thapsigargin, which depletes intraluminal ER calcium and induces apoptosis (42). As shown in Fig. 3, tunicamycin and thapsigargin treatments for 24 h caused the processing of procaspase-12 and activation of calpain, caspase-3, and caspase-6. Although levels of the ϳ60-kDa procaspase-12 were markedly reduced, smaller immunoreactive fragments could not be detected even with prolonged film exposures or varying durations of ER stress. This suggests that in cortical neurons, either processed caspase-12 is unstable or the proenzyme is processed to fragments undetectable by the antibody being used. Neuronal ER stress led to calpain activation and cytoskeletal protein degradation, manifested by appearance of an ϳ150-kDa NH 2 -terminal fragment of the actinbinding protein ␣-spectrin and detected with a cleavage sitespecific antibody reactive exclusively with this calpain derivative (32,33). In addition, ER stress caused activation of caspase-3, detected by loss of the ϳ32-kDa proenzyme and appearance of the ϳ17-kDa large subunit and of caspase-6, as evidenced by the appearance of an ϳ27-kDa NH 2 -terminal fragment of lamin A generated by caspase-6 cleavage. Processing of procaspase-12 and activation of calpain, caspase-3, and caspase-6 also were observed following treatment with staurosporine (Fig. 3), a prototypical activator of the mitochondrial apoptotic signaling pathway (43).
The stress-induced activation of cysteine proteases from neurons bearing the PS-1P264L mutation was compared with that from neurons wild type for PS-1, and only minor differences were detectable (Fig. 3). The activation responses of calpain, caspase-3, and caspase-6 did not differ, and for the PS-1 mutant neurons only a very small enhancement in procaspase-12 processing was discernable at a submaximal dose of tunicamycin. In accordance with our previous report (32), there was no difference in staurosporine-induced cysteine protease activation between cortical neurons bearing the two PS-1 genotypes. Degenerated, apoptotic cortical neurons accumulated progressively during the 14-day culture period (34) and, consistent with the lack of significant effect of the PS-1 knock-in mutation on neuronal sensitivity to ER stress, there was only a minor, statistically insignificant difference between the two genotypes in the basal amount of apoptosis. Furthermore, the dose dependence for tunicamycin-induced apoptosis and its maximal extent were essentially unchanged (Table I). These data demonstrate that the PS-1 P264L knock-in mutation does not alter appreciably the sensitivity of cortical neurons to ER stressinduced proteolysis or apoptosis.
Pathogenic presenilin mutations reportedly impair the unfolded protein response to ER stress (13), although this effect has not been observed in all studies (25). To examine the influence of the PS-1 P264L knock-in mutation on the UPR in cortical neurons, we measured by immunoblotting the levels of the inducible chaperones Grp94 and Grp78/BiP. Basal levels of Grp94 and Grp78/BiP were similar in neurons that are wild type for PS-1 and those expressing the P264L knock-in mutation. Treatment with tunicamycin for 24 h increased steadystate levels of Grp94 and Grp78/BiP, whereas thapsigargin had a more modest effect on levels of the chaperones, and no increase could be discerned with staurosporine treatment (Fig.  4A). The PS-1 knock-in mutation did not change either the basal levels of these ER chaperones or their increase evoked by tunicamycin or thapsigargin. When chaperone levels were normalized to those of actin for three independent experiments, tunicamycin increased Grp94 levels by 38 (wild type) and 35% (P264L knock-in) and elevated Grp78/BiP levels by 50 (wild type) and 54% (P264L knock-in). To investigate further the effect of mutant PS-1 on the UPR in cortical neurons, RT-PCR was used to evaluate basal and ER stress-induced levels of FIG. 2. Caspase-12 is enriched in microsomes and is processed by calpain. A, subcellular fractions prepared from cultured neurons contained procaspase-12 predominantly in the ER-enriched microsomal fraction (MICRO). Cytochrome c (CYT C), on the other hand, was concentrated in the mitochondrial/synaptosomal fraction (MITO). B, processing of procaspase-12 and calpain-mediated cleavage of spectrin were evaluated in a post-mitochondrial fraction from mouse brain. Calcium activation of calpain led to degradation of the calpain substrate ␣-spectrin, detected with a cleavage site-specific antibody, and to processing of procaspase-12 and formation of a smaller immunoreactive fragment (arrow). These effects of calcium were reduced by the calpain inhibitor calpeptin (CALPEP) but not the caspase inhibitor Ac-Asp-Glu-Val-Asp-CHO (DEVD).

FIG. 3. Cysteine protease activation during ER stress and lack of effect of the PS-1 P264L knock-in mutation.
Cultured mouse cortical neurons were maintained under basal conditions or were treated for 24 h with tunicamycin (ϩ, 0.5 g/ml; ϩϩ, 3 g/ml), thapsigargin (ϩ, 0.2 M; ϩϩ, 1 M), or staurosporine (250 nM) and then evaluated by immunoblotting for processing of procaspase-12 and activation of calpain, caspase-3, and caspase-6. Note that all treatments caused loss of procaspase-12, calpain cleavage of spectrin, loss of procaspase-3, appearance of the active caspase-3 p17 subunit, and caspase 6 cleavage of lamin A. There was no appreciable difference in protease activation or substrate cleavage in neurons bearing the PS-1 P264L knock-in mutation.
Grp78/BiP mRNA. As shown in Fig. 4B, basal levels of Grp78/ BiP mRNA were similar for the two PS-1 genotypes, and treatment with tunicamycin or thapsigargin increased Grp78/BiP mRNA levels. The increase in expression did not differ across the two PS-1 genotypes. Thus, for mouse cortical neurons neither protein nor mRNA levels for molecular chaperones were impacted by the PS-1 P264L knock-in mutation.
The results presented thus far demonstrate that for primary mouse cortical neurons, ER stress responses of cysteine protease activation, chaperone induction, and apoptosis are not altered significantly by an Alzheimer's disease-linked PS-1 P264L knock-in mutation. To determine whether the lack of effect of the mutant PS-1 on ER stress responses is a peculiarity of cortical neurons, we examined embryonic fibroblasts derived from mice of the two PS-1 genotypes. Procaspase-12 and -3 were analyzed following treatment with tunicamycin, thapsigargin, or staurosporine. As shown in Fig. 5, the two ER stressors and staurosporine led to processing of procaspase-12 and -3 in fibroblasts. Activation of these proteases was virtually indistinguishable based on the PS-1 genotype. The basal levels of procaspase-12 and procaspase-3 also were not altered detectably by the PS-1 P264L knock-in mutation. Similarly, the ER stressors activated caspase 6-mediated lamin A degradation that did not differ based on PS-1 genotype. 2 Next, the effect of PS-1P264L on the UPR was evaluated by measuring the levels of Grp94 and Grp78/BiP. Basal levels of the two ER chaperones did not differ as a function of PS-1 genotype. Treatment with tunicamycin and, to a lesser extent, thapsigargin increased levels of the two ER chaperones, and there was no discernible reduction in chaperone levels from PS-1 mutant fibroblasts (Fig. 5). Quantitative analysis from four independent experiments confirmed that the tunicamycin-induced increases in Grp94 and Grp78/BiP were unaltered by the mutant PS-1 (38 and 35% increases, respectively, for PS-1 wild-type cells; 46 and 62% increases, respectively, for PS-1 mutant knock-in fibroblasts). Finally, the influence of PS-1P264L on the translational suppression response to ER stress was assessed in fibroblasts. Stress-induced activation of the protein kinase PERK stimulates phosphorylation and inactivation of the translation initiation factor eIF2␣ and can be evaluated by the activation-induced autophosphorylation of PERK and its mobility shift upon SDS-PAGE (5,25,44). As shown in the immunoblot of Fig. 6, tunicamycin caused a time-dependent decrease in PERK mobility, indicative of activation-induced autophosphorylation. PERK activation was essentially complete by 7 h. Neither the rate nor the maximal amount of PERK activation was changed in PS-1 P264L mutant fibroblasts. Procaspase-12 processing was not initiated until after the peak of PERK autophosphorylation and did not differ in time course or magnitude between the two PS-1 genotypes.

TABLE I Tunicamycin-induced apoptosis as a function of PS-1 genotype
The PS-1 P264L knock-in mutation had little effect on cortical neuronal vulnerability to basal or ER stress-induced apoptosis. Cortical neurons maintained in primary culture for 14 days were treated for 24 h with either Me 2 SO vehicle or varying doses of tunicamycin. The proportion of apoptotic neurons was determined as described under "Experimental Procedures." Note that neither the basal level of apoptosis nor the tunicamycin-induced cell death differed significantly between the two PS-1 genotypes. WT/WT, wild type for PS-1.

. Chaperone induction by ER stress in cortical neurons is not impaired appreciably by the PS-1 P264L knock-in mutation.
A, immunoblot analysis of Grp94, Grp78/BiP, and actin levels from cortical neurons maintained under basal conditions or treated for 24 h with tunicamycin (low, 0.5 g/ml; high, 3 g/ml), thapsigargin (low, 0.2 M; high, 1 M), or staurosporine (250 nM). Note that both doses of tunicamycin increased the levels of Grp94 and Grp78/BiP relative to actin, whether neurons were wild type (WT/WT) for PS-1 or bearing the P264L knock-in mutation. Modest increases were elicited by thapsigargin and none by staurosporine. B, semiquantitative RT-PCR analysis of Grp78/BiP and glyceraldehyde-3-phosphate dehydrogenase (GAPDH) mRNA levels. Basal levels of Grp78/BiP mRNA did not differ appreciably between neurons wild type for PS-1 (WT-C) or those homozygous for the PS-1 P264L knock-in mutation (KI-C). Following 6-h treatments with tunicamycin (low, 0.5 g/ml; high, 3 g/ml) or thapsigargin (low, 0.2 M; high, 1 M), Grp78/BiP mRNA levels increased with little difference between the two PS-1 genotypes.

FIG. 5. ER stress-induced chaperone induction and caspase activation in fibroblasts are unaltered by the PS-1 P264L knock-in mutation.
Cultured mouse embryo fibroblasts were maintained under basal conditions or treated for 24 h with tunicamycin (ϩ, 0.5 g/ml; ϩϩ, 3 g/ml), thapsigargin (ϩϩ, 1 M), or staurosporine (200 nM). Tunicamycin treatment increased the levels of Grp94 and Grp78/ BiP, which were not discernibly different based on PS-1 genotype. This experiment was repeated three times with essentially the same result. Refer to "Results" for the quantitative analysis. Each treatment evoked processing of procaspase-12 and -3, and the PS-1 P264L knock-in mutation caused only a very slight increase in caspase activation.

DISCUSSION
Alzheimer's disease-causing mutations in the presenilins reportedly increase the vulnerability of cultured cells to ER stress-induced apoptosis (13,45), an endangering mechanism of potential relevance to the extensive aging-related neocortical neurodegeneration of AD. This effect is manifested by impaired cellular protective responses to ER stress, including induction of chaperones such as Grp94 and Grp78/BiP and translational suppression initiated by activation of the protein kinase PERK, as well as by increased stress-induced cell death. To investigate further the influence of mutant presenilin on ER stress responsiveness in cortical neurons, we first characterized the activation of cysteine proteases of the caspase and calpain families following ER stress to primary mouse cortical neurons and then compared stress-induced proteolysis between cells derived from PS-1 wild-type and homozygous PS-1 P264L knock-in mice. Our results provide evidence that in cortical neurons, ER stress leads to activation of proteases involved in both the initiation and execution of apoptosis, not only within the ER but also in other neuronal compartments, and leads to the degradation of ER, cytoskeletal, and nuclear protein substrates. The homozygous AD-linked PS-1 P264L mutation does not alter appreciably the ER stress-stimulated activation responses of calpain and caspase-3, -6, or -12 or influence significantly the Grp94 or Grp78/BiP induction, PERK activation, and apoptosis. The lack of effect of mutant PS-1 is observed with either of two ER stressors and in two different cell types. These findings are similar to the lack of effect of presenilin deletion or mutation on UPR and translational suppression observed for several cell lines in another study (25), extending them to primary cortical neurons and to the proteolytic signaling that mediates ER stress-induced apoptosis. Therefore, although certain presenilin mutants impair ER stress responses and enhance cell death, this effect does not occur for neurons expressing all AD-linked PS-1 mutations and so is unlikely to be a critical pathogenic mechanism by which presenilin mutations cause early onset AD.
Our study used the processing of procaspase-12 as a biochemical marker for activation of the ER apoptotic pathway. ER stress-induced apoptosis is dependent on this ER-resident protease (24,27), activation of which likely involves proteolytic processing and oligomeric assembly in a manner similar to all other caspases (46). We confirmed that procaspase-12 is enriched in ER-containing microsomal fractions from the brain and is processed by the family of cytosolic calcium-dependent cysteine proteases, calpain (27). Furthermore, by using a cleavage site-specific antibody that reacts exclusively with a fragment of the cytoskeletal protein spectrin generated by calpain (32,33), we provide evidence for activation of calpain and degradation of a cytoskeletal calpain substrate coincident with processing of procaspase-12 in response to cortical neuronal ER stress. These results support the concept that a rise in cytosolic free calcium concentration and activation of calpain trigger caspase-12 activation and ER apoptotic signaling, although they do not exclude the possibility that other proteases may be involved in activation of caspase-12. Neither calpain-mediated spectrin degradation nor procaspase-12 processing is altered appreciably in cortical neurons by the homozygous PS-1 P264L knock-in mutation, whether the stressor is tunicamycin, which impairs protein folding by preventing N-linked glycosylation, or thapsigargin, which interferes with protein folding by depleting ER calcium stores. This result is somewhat unexpected, given the evidence that mutant presenilins interfere with intracellular calcium homeostasis by enhancing stimulus-induced release of calcium from ER stores (45,47), attenuating capacitative calcium entry (48), and elevating ER calcium stores (49). Any mutation-induced increase in calpain activation could have been readily detected, because the method has demonstrated already that necrotic stimuli evoke much stronger calpain activation than do apoptotic agents 2 (32). The lack of effect of mutant PS-1 on calpain activation reported here may be attributable to differences between studies in presenilin mutation or levels of mutant PS-1 expression. Another possibility stems from the finding that stimulus-induced calpain activation depends not only on the magnitude of a calcium rise but also its source (50). Further study will be required to identify definitively the calcium pools and mechanisms contributing to ER stress-induced calpain activation. Nevertheless, the lack of effect of PS-1 P264L on activation of calpain and caspase-12 by either tunicamycin or thapsigargin indicates that an AD-linked mutant presenilin does not influence the initial triggering mechanism for ER stress-induced apoptotic signaling. This conclusion is substantiated by the lack of effect of the PS-1 knock-in mutation on procaspase-12 processing in a second cell type, the fibroblast, as well as on tunicamycininduced neuronal apoptosis.
ER stress in cortical neurons not only initiates apoptotic signaling via processing of procaspase-12 within the ER but also activates cytosolic and nuclear caspases involved in the execution of apoptosis, all in a manner that is not altered by the pathogenic PS-1 P264L knock-in mutation. Immunocytochemistry with cleavage site-specific antibodies localize activated caspase-3 to the cytosol and caspase-6-cleaved lamin A to the nucleus of primary cortical neurons following ER stress (Fig. 1). The activation of caspase-3 and -6 is verified by immunoblot detection of the stress-induced loss of procaspase-3 and appearance of the p17 large subunit as well as the ϳ27-kDa NH 2terminal fragment of lamin A (Fig. 3). Caspase-3 and -6 are effector caspases characterized by their requirement for processing-induced activation by initiator caspases, abundance of cytosolic, membrane, cytoskeletal, and nuclear protein substrates, and prominent roles in morphological changes that characterize the execution phase of apoptosis (51,52). There are several pathways by which ER stress could activate caspase-3 and -6. One possible route is through the well described release of mitochondrial cytochrome c, formation of the apoptosome, and activation of caspase-9, which in turn processes and activates downstream caspases such as -3 and -6. There are conflicting reports on activation of the mitochondrial pathway in response to ER stress (29 -31), and should mitochondria be involved, the signals linking ER stress to cytochrome c release remain to be identified. An alternative possibility is the direct activation of caspase-3 and -6 by either caspase-12 or calpain. The former route has not been reported FIG. 6. Evidence that ER stress-induced PERK activation is not altered by the PS-1 P264L knock-in mutation. Immunoblot analysis shows that tunicamycin treatment of fibroblasts caused a time-dependent decrease in electrophoretic mobility of PERK, indicative of PERK autophosphorylation, which accompanies activation. A discernible shift in PERK phosphorylation (P-PERK) is evident by 3 h and is essentially complete by 7 h. There is no difference in the time course for PERK activation in fibroblasts derived from the PS-1 P264L knock-in mouse. Tunicamycin also caused a time-dependent loss of procaspase-12 that occurred subsequent to PERK activation and did not differ between the two PS-1 genotypes. thus far, and the effect of calpain on caspase-3 is complex, leading either to processing of the prodomain, which facilitates activation of the caspase (53), or alternatively to inhibition of procaspase processing and caspase-3 activation (54). Currently, we are examining the pathway leading from ER stress to activation of executioner caspases. Whatever the mechanism, neither the loss of procaspase-3 nor the appearance of the activated p17 subunit and the caspase 6-cleaved lamin A is modified in cortical neurons or fibroblasts carrying the PS-1 P264L mutation. When coupled with the lack of effect of PS-1 P264L on the activation responses of calpain or caspase-12 or the dose-dependence for tunicamycin-induced apoptosis, these results indicate that several biochemical and morphological indices of stress-induced cortical neuronal apoptosis are not altered by this pathogenic presenilin mutation.
In addition to unaltered apoptotic protease signaling, cells bearing PS-1 P264L exhibit no appreciable impairment in their UPR or translational suppression response to ER stress. Protein levels for Grp94 and Grp78/BiP exhibit small but significant increases in cortical neurons treated with tunicamycin, effects that are not influenced by PS-1 P264L (Fig. 4). Additionally, Grp78/BiP mRNA levels were analyzed as a robust measure of the UPR, and for primary cortical neurons neither the basal mRNA levels nor their induction by tunicamycin or thapsigargin were altered appreciably by the PS-1 knock-in mutation. This confirms and extends the findings of Sato et al. (25), who demonstrated that neither presenilin deficiency nor overexpression of an AD-linked mutant PS-1 impairs the UPR in several cell lines. The lack of effect of PS-1 P264L on the UPR is not peculiar to cortical neurons, as similar results have been obtained using fibroblasts (Fig. 5). Furthermore, the unimpaired stress-induced mobility shift in PERK (Fig. 6) provides evidence that the PS-1 P264L knock-in mutation does not alter PERK activation required for the translational suppression response. Consequently, whereas some PS-1 mutations interfere with protective responses to ER stress and endanger cells to stress-mediated apoptosis (13,45), these effects are not observed for all AD-linked PS-1 mutations or cell types. It remains possible that presenilin regulation of ER stress responsiveness may differ between cultured mouse primary neurons and those of the adult human brain, and therefore an involvement of impaired ER stress responses in the pathogenesis of AD cannot be completely excluded. Nevertheless, the current findings suggest that impaired ER stress responses and enhanced vulnerability to stress-driven apoptosis are unlikely to be contributory factors.
Given the evidence against a role for impaired ER stress responsiveness and increased vulnerability to stress-induced apoptosis, what alternative pathogenic mechanism might link presenilin mutations to familial AD? Among the potential mechanisms receiving experimental support, mutant presenilins have been reported to increase cell vulnerability to degeneration elicited by a variety of apoptotic and necrotic insults (45,(55)(56)(57). However, cortical neurons derived from the homozygous PS-1 P264L knock-in mouse exhibit morphological and biochemical responses to several apoptotic and necrotic agents indistinguishable from neurons wild type for PS-1 (32). Instead, in the gene-targeted mouse, PS-1 P264L enhances A␤42 secretion by primary cortical neurons, elevates A␤42 levels in the brain, and accelerates the onset of cerebral amyloid deposition and reactive astrocytosis, even as a single mutant allele. In light of these and the present findings, along with the preponderance of evidence suggesting a critical role for A␤42 in AD pathogenesis (58 -60), it is important that future studies define the molecular and cellular mechanisms by which the PS-1 P264L mutation modulates A␤ production.