Increased Vulnerability of Hippocampal Neurons from Presenilin-1 Mutant Knock-In Mice to Amyloid β-Peptide Tox

Central Roles of Superoxide Production and Caspase Activation

Authors

  • Qing Guo,

  • Lois Sebastian,

  • Bryce L. Sopher,

  • Miles W. Miller,

  • Carol B. Ware,

  • George M. Martin,

  • Mark P. Mattson


  • Abbreviations used : Aβ, β-amyloid peptide ; AD, Alzheimer's disease ; DEVD-CHO, N-acetyl-Asp-Glu-Val-Asp-aldehyde ; DHR, dihydrorhodamine ; HE, hydroethidium ; PBS, phosphate-buffered saline ; PS1, presenilin 1 ; PS2, presenilin 2 ; ROS, reactive oxygen species ; sAPP, soluble β-amyloid precursor protein ; TBARS, thiobarbituric acid-reactive substances ; zVAD-fmk, benzyloxycarbonyl-Val-Ala-Asp-fluoromethyl ketone.

Address correspondence and reprint requests to Dr. M. P. Mattson at 211 Sanders-Brown Building, University of Kentucky, Lexington, KY 40536-0230, U.S.A.

Abstract

Abstract : Many cases of early-onset inherited Alzheimer's disease (AD) are caused by mutations in the presenilin-1 (PS1) gene. Overexpression of PS1 mutations in cultured PC12 cells increases their vulnerability to apoptosis-induced trophic factor withdrawal and oxidative insults. We now report that primary hippocampal neurons from PS1 mutant knock-in mice, which express the human PS1M146V mutation at normal levels, exhibit increased vulnerability to amyloid β-peptide toxicity. The endangering action of mutant PS1 was associated with increased superoxide production, mitochondrial membrane depolarization, and caspase activation. The peroxynitrite-scavenging antioxidant uric acid and the caspase inhibitor benzyloxycarbonyl-Val-Ala-Asp-fluoromethyl ketone protected hippocampal neurons expressing mutant PS1 against cell death induced by amyloid β-peptide. Increase oxidative stress may contribute to the pathogenic action of PS1 mutations, and antioxidants may counteract the adverse property of such AD-linked mutations.

Alzheimer's disease (AD) is characterized by deposition of amyloid β-peptide (Aβ) and neuronal degeneration in brain regions involved in learning and memory processes such as the hippocampus. An increasing amount of evidence suggests that Aβ contributes to the neurodegenerative process in AD, possibly by increasing neuronal vulnerability to apoptosis and/or excitotoxicity (for review, see Yankner, 1996 ; Mattson, 1997). Although most cases of AD are not caused by a specific genetic defect and have a late age of onset, some cases are characterized by an early age of onset and a dominant inheritance pattern. Mutations in the genes encoding presenilin-1 (PS1) on chromosome 14 and presenilin-2 (PS2) on chromosome 1 are responsible for many cases of early-onset inherited AD (for review, see Hardy, 1997 ; Kim et al., 1997 ; Mattson et al., 1998a). Presenilins are integral membrane proteins with six or eight membrane-spanning domains that are expressed in neurons throughout the brain, wherein they appear to be localized mainly in the endoplasmic reticulum (Doan et al., 1996 ; Kovacs et al., 1996 ; Walter et al., 1996 ; Lehmann et al., 1997). Studies of transfected cell lines and tissues from AD patients and transgenic mice suggest a role for increased production of an amyloidogenic form of Aβ (Aβ1-42) in the pathogenic mechanism of PS1 mutations (Borchelt et al., 1996 ; Duff et al., 1996 ; Scheuner et al., 1996). Recent studies have shown that cultured tumor cells overexpressing mutant PS1 or PS2 exhibit increased vulnerability to apoptosis induced by trophic factor withdrawal and exposure to Aβ (Guo et al., 1996, 1997 ; Wolozin et al., 1996 ; Janicki and Monteiro, 1997). It is not known whether presenilin mutations exert a similar apoptosis-enhancing action when expressed at normal levels in primary neurons, nor has the mechanism whereby presenilin mutations endanger neurons been established.

Oxidative stress (Smith et al., 1991 ; Goodman and Mattson, 1994 ; Mecocci et al., 1994 ; Lovell et al., 1997 ; Smith et al., 1997) and activation of apoptotic cell death pathways (Loo et al., 1993 ; Su et al., 1994 ; Kruman et al., 1997 ; Guo et al., 1998a,b) are believed to contribute to the neurodegenerative process in AD. Aβ deposition may contribute to the increased oxidative stress and neuronal degeneration in AD, as suggested by data showing that Aβ induces membrane lipid peroxidation and accumulation of peroxynitrite and hydrogen peroxide in cultured neurons and synaptosomes (Mattson et al., 1993 ; Behl et al., 1994 ; Butterfield et al., 1994 ; Goodman and Mattson, 1994 ; Keller et al., 1997 ; Kruman et al., 1997 ; Mark et al., 1997a). Such oxidative stress may promote neuronal degeneration by impairing membrane ion-motive ATPases and glucose transporters, thereby rendering neurons vulnerable to excitotoxicity (Mark et al., 1995, 1997a,b ; Keller et al., 1997). Aβ can also induce apoptosis in cultured neurons (Loo et al., 1993), which can be prevented by antioxidants (Goodman and Mattson, 1994 ; Kruman et al., 1997), suggesting a key role for oxidative stress in the neurotoxic action of Aβ. It was recently reported that overexpression of Mn-superoxide dismutase in cultured PC12 cells protects those cells against Aβ-induced apoptosis (Keller et al., 1998), demonstrating an important role for superoxide anion radical (O2.-) production in the neurotoxic action of Aβ.

Previous studies suggesting that presenilin mutations promote apoptosis have used transfected tumor cells overexpressing mutant presenilins at high levels (Guo et al., 1996, 1997, 1998a ; Wolozin et al., 1996). Therefore, an important unresolved issue is whether presenilin mutations expressed at normal levels in primary neurons also exhibit a cell death-enhancing action. To address this issue directly we have used gene targeting methods to generate knock-in mice that express the human PS1-M146V mutation (Guo et al., 1999). We now report that primary hippocampal neurons from the PS1M146V knock-in mice exhibit increased vulnerability to Aβ-induced cell death and that the mechanism underlying this adverse effect of the PS1 mutation involves increased superoxide production and caspase activation.

MATERIALS AND METHODS

PS1 mutant knock-in mice

The targeting strategy used to generate PS1 mutant knock-in (PS1M146VKI) mice is detailed elsewhere (Guo et al., 1999). In brief, a genomic DNA clone was isolated from a mouse 129/Svj P1 library by Genome System (St. Louis, MO, U.S.A.) using PCR primers designed to amplify a 180-bp product from the first coding exon of the murine PS1 gene. A 1.2-kb ScaI-HindIII fragment containing exon 5 was subcloned into pAlter-1 (Promega) and mutagenized with a 39-bp mutagenic oligonucleotide designed to introduce the I45V/M146V double mutation and a BestEII restriction site. Introduction of the I145V substitution humanized the only mouse/human polymorphism in exon 5 of the murine PS1 gene and introduced a unique restriction enzyme site (BestEII) that simplified genotyping. The mutagenized DNA and remaining 5′ and 3′ targeting arms were assembled in pZErO-2.1 (InVitrogen)-derived vectors, into which we had inserted additional cloning sites and loxP sites in the appropriate positions. These 5′ and 3′ targeting arms were then subcloned into the pNTK2 targeting vector, and the assembled vector was linearized with PvuI and electroporated into 129/Sv-derived R1 ES cells. Genomic DNA was isolated from 250 clones surviving double selection, and 19 of the clones produced the expected HindIII and BglI polymorphisms. Four of the 19 targeted cell lines (19, 106, 157, and 179) were injected into recipient blastocysts and transferred to foster mothers to produce male chimeras, which were then mated with C57BL/6 females to produce heterozygous PS1M146V mice. Mice derived from line 106 were used in the present study.

Mice were genotyped using a PCR assay using primers (5′-AGGCAGGAAGATCACGTGTTCAAGTAC-3′ and 5′-CACACGCACACTCTGACATGCACAGGC-3′ to amplify genomic DNA sequences flanking exon 5 before digestion of the amplified DNA with the restriction enzyme BstEII. The expected full-length PCR product is 530 bp. The expected product sizes for the wild-type and targeted allele (PS1M146V) following BstEII enzyme digestion are 530 and 350/180 bp, respectively. Expression levels of mRNA and protein from the targeted PS1 mv allele appear to be normal (Guo et al., 1999). Homozygous PS1M146V knock-in mice live for at least 12 months with no overt phenotype and reproduce reliably, indicating that the targeted M146V mutation does not dramatically impair the normal developmental and physiological functions of PS1. These findings suggest that the pathogenic mechanism of PS1 mutations in AD likely involves a gain function and are consistent with recent reports showing that ectopic expression of human PS1 protein (with familial AD mutations) in mice can fully suppress the PS1 null phenotype (Davis et al., 1998 ; Qian et al., 1998). All experiments in the present study were performed on cells from either wild-type or homozygous PS1M146VKI mice.

Hippocampal cell culture methods

Cultures of dissociated hippocampal cells were prepared from postnatal day 1 wild-type and homozygous PS1M146VKI mouse pups using methods similar to those described previously (Furukawa et al., 1997 ; Mattson et al., 1997a). In brief, hippocampi were removed and incubated for 15 min in Ca2+- and Mg2+ -free Hanks' balanced saline solution (GibcoBRL) containing 0.2% papain. Cells were dissociated by trituration and plated into polyethylenimine-coated plastic or glass-bottom culture dishes containing minimum essential medium with Earle's salts supplemented with 10% heat-inactivated fetal bovine serum, 2 mM L-glutamine, 1 mM pyruvate, 20 mM KCl, 10 mM sodium bicarbonate, and 1 mMHEPES (pH 7.2). Following cell attachment (3-6 h postplating), the culture medium was replaced with Neurobasal medium with B27 supplements (GibcoBRL). Experiments were performed in 8-day-old cultures. Immediately before experimental treatment the medium was replaced with Locke's buffer (NaCl, 154 mM ; KCl, 5.6 mM ; CaCl2, 2.3 mM ; MgCl2, 1.0 mM ; NaHCO3, 3.6 mM ; glucose, 5 mM ; and HEPES, 5 mM ; pH 7.2).

Immunoblots and immunocytochemistry

For immunoblot analysis, 50 μg of solubilized proteins was separated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (10% gel) and transferred to a nitrocellulose membrane. The membrane was incubated overnight at 4°C in the presence of 5% nonfat milk and then incubated for 2 h with rabbit polyclonal PS1 antibody (Guo et al., 1997) at a dilution of 1 : 1,000 The membrane was then exposed for 1 h to horseradish peroxidase-conjugated secondary antibody (1:3,000 ; Jackson ImmunoResearch Labs, West Grove, PA, U.S.A.), and immunoreactive protein was visualized using a chemiluminescence-based detection kit according to the manufacturer's protocol (ECL kit ; Amersham Corp., Arlington Heights, IL, U.S.A.). The methods for immunostaining cultured cells were similar to those described previously (Mattson et al., 1997a). In brief, cells were fixed in 4% paraformaldehyde, membranes were permeabilized by exposure for 5 min to 0.2% Triton X-100 in phosphate-buffered saline (PBS), and cells were placed in blocking serum (5% goat serum in PBS) for 30 min. Cells were then exposed to PS1 antibody (1:200) overnight at 4°C, followed by incubation for 1 h with biotinylated goat anti-rabbit secondary antibody (1:200) and 30 min in the presence of fluorescein isothiocyanate-avidin (Vector Labs, Burlingame, CA, U.S.A.). Images of immunofluorescence were acquired using a confocal laser scanning microscope with a 60X oil immersion objective (excitation at 488 nm and emission at 510 nm).

Experimental treatments and quantification of neuronal survival

Aβ1-42 was synthesized by the University of Kentucky Macromolecular Structure facility and was prepared as a 1 mM stock in water 4 h before addition to cultures. Uric acid (Sigma) was prepared as a 100X stock in water. The broad-spectrum caspase inhibitor benzyloxycarbonyl-Val-Ala-Asp-fluoromethyl ketone (zVAD-fmk ; Bachem) was prepared as a 500X stock in dimethyl sulfoxide. The method for quantification of neuron survival in hippocampal cell cultures was described previously (Mattson et al., 1997a). In brief, undamaged neurons in premarked microscope fields were counted before and at indicated time points following exposure to experimental treatments. Neurons with intact neurites and a cell body that was smooth and round or oval in shape were considered viable. Neurons with beaded or fragmented neurites and a cell body that was shrunken and rough in appearance were considered nonviable. For propidium iodide staining cultures were fixed in 4% paraformaldehyde for 10 min, incubated for 5 min in the presence of 0.2% Triton X-100, and incubated for 20 min in the presence of 10 μM propidium iodide in PBS. Images of propidium iodide fluorescence were acquired using a confocal laser scanning microscope with excitation at 488 nm and emission at 510 nm.

Measurements of superoxide levels, mitochondrial reactive oxygen species (ROS), and membrane lipid peroxidation

Levels of intracellular superoxide anion radical were measured using hydroethidium (HE), which is oxidized to fluorescent ethidium cation by superoxide, using methods similar to those described previously (Bindokas et al., 1996). In brief, cells were incubated for 30 min in the presence of 5 μM HE (Molecular Probes), washed twice with Locke's solution, and confocal images of cell-associated HE fluorescence were acquired (excitation at 488 nm and emission at 510 nm). The average pixel intensity in individual cell bodies was determined using Imagespace software (Molecular Dynamics) ; all images were coded and analyzed without knowledge of experimental treatment history of the cultures. The dye dihydrorhodamine (DHR) was used to quantify relative levels of mitochondrial peroxynitrite using methods similar to those described previously (Mattson et al., 1997a ; Keller et al., 1998). DHR localizes to mitochondria and fluoresces when oxidized to the positively charged rhodamine-123 derivative. In brief, cells were incubated for 30 min in the presence of 5 μM DHR and washed three times with Locke's solution, and confocal images of cellular fluorescence were acquired and analyzed as described for HE fluorescence. A thiobarbituric acid-reactive substances (TBARS) fluorescence-based method was used as a measure of membrane lipid peroxidation (Goodman and Mattson, 1996). In brief, at designated times following exposure to Aβ cells were incubated for 30 min at room temperature in the presence of fixative containing 50% HPLC-grade methanol, 10% glacial acetic acid, and 40% (vol/vol) purified water, plus 2 mM EDTA and 38 mM 2-thiobarbituric acid. The cultures were then heated to 85°C for 45 min, fixative was removed, antifade solution was added, and confocal images of cellular fluorescence were aquired and analyzed as described for HE fluorescence.

Measurement of caspase activity

Caspase-3-like protease activity was assessed using a previously described protocol that employs the biotinylated caspase substrate N-acetyl-Asp-Glu-Val-Asp-aldehyde (DEVD-CHO ; Mattson et al., 1998b). In brief, at designated time points following exposure of cultures to Aβ, cultured cells were exposed for 10 min to Locke's solution containing 0.01% digitonin and were then incubated for 20 min in the presence of 10 μg/ml DEVD-biotin (Calbiochem). Cells were washed three times with PBS (2 ml per wash), incubated for 5 min in PBS containing 0.2% Triton X-100, and then incubated for 20 min in the presence of streptavidin-Oregon Green conjugate (Molecular Probes). Cells were washed twice in PBS, and images of cellular fluorescence (corresponding to conjugates of activated caspase-3 with DEVD-biotin) were acquired using a confocal laser scanning microscope. Levels of fluorescence in neuronal somata were quantified as described for ethidium fluorescence.

RESULTS

Levels and subcellular localization of PS1 protein are similar in hippocampal neurons from wild-type and PS1 mutant knock-in mice

To verify that expression of PS1 from the targeted allele in PS1M146VKI mice was normal, we performed immunoblot analyses of hippocampal cells from early postnatal wild-type and homozygous PS1M146VKI mice. Levels of full-length (46-kDa) PS1 and the 17-kDa PS1 C-terminal fragment were essentially identical in hippocampal homogenates from wild-type and PS1M146VKI mice (Fig. 1A). We next performed confocal laser scanning microscope analysis of PS1 immunoreactivity in cultured hippocampal neurons. In neurons from both wild-type and PS1M146VKI mice the PS1 immunoreactivity was localized to patches within the cytosol and was absent from the nucleus (Fig. 1B). PS1 immunoreactivity was present in the cell body and in neurites, where it was distributed in patches throughout their entire length. This distribution of PS1 protein immunoreactivity is consistent with localization in endoplasmic reticulum, as was reported previously (kovacs et al., 1996 ; Walter et al., 1996). We observed no overt differences in PS1 protein localization in wild-type and PS1M146VKI neurons.

Figure 1.

. Levels and subcellular localization of PS1 protein are unchanged in hippocampal neurons from PS1 mutant knock-in mice. A : Immunoblot of cell homogenates from hippocampi of 2-day-old wild-type mice and heterozygous and homozygous PS1M146V knock-in mice. Each lane was loaded with 50 μg of protein. The 46-kDa band corresponds to full-length PS1, and the 17-kDa band corresponds to a C-terminal fragment of PS1. B : Confocal laser scanning microscope images of PS1 immunoreactivity in cultured hippocampal neurons from wild-type mice and homozygous PS1M146VKl mice.

FIG. 1

Hippocampal neurons from PS1M146VKI mice exhibit increased vulnerability to Aβ-induced death

Although previous studies had shown that overexpression of PS1 or PS2 mutations can increase the vulnerability of tumor cell lines to apoptosis induced by Aβ (Guo et al., 1996 ; Wolozin et al., 1996), it was not known whether presenilin mutations similarly endanger primary neurons or whether such endangerment occurs with normall levels of expression of mutant PS1. To address these important issues, we established hippocampal cell cultures from homozygous PS1M146VKI mice and wild-type mice and examined neuronal vulnerability to Aβ toxicity. We did not observe any differences in neuronal growth and survival between wild-type and PS1M146VKI hippocampal neurons under basal culture conditions (data not shown). Exposure of hippocampal cultures from wild-type and PS1M146VKI mice for 4 days to increasing concentrations of Aβ1-42 resulted in concentrations-dependent decreases in neuronal survival. PS1M146VKI hippocampal neurons were significantly more sensitive to death induced by Aβ than were their wild-type counterparts (Fig. 2A). Analysis of the time course of neuronal death following exposure of cultures to Aβ revealed that PS1M146VKI neurons were killed more rapidly than wild-type neurons (Fig. 2B). Morphologically, neuronal death induced by Aβ was characterized by neurite fragmentation and cell body shrinkage (Fig. 3A). Consistent with previous studies of cultured rat hippocampal neurons exposed to Aβ (Kruman et al., 1997), staining of membrane-permeabilized cells with the fluorescent DNA-binding dye propidium iodide revealed nuclear chromatin condensation and fragmentation consistent with an apoptotic mode of cell death (Fig. 3B).

Figure 2.

Increased vulnerability to Aβ toxicity in hippocampal neurons from PS1M146VKI mice. A : Cultures from wild-type mice (WTPS1) and PS1M146VKI mice (M146VKI) were exposed for 72 h to the indicated concentrations of Aβ1-42, and neuron survival was quantified. Data are mean ± SE (bars) values of determinations made in four cultures. **p < 0.05, ***p < 0.01 compared with value for WTPS1 cultures by ANOVA with Scheffé's post hoc tests. B : Cultures were exposed to 20 μM Aβ1-42 for the indicated intervals, and neuron survival was quantified. Data are mean ± SE (bars) values of determinations made in four cultures. **p < 0.01, ***p < 0.001 compared with value for WTPS1 cultures by ANOVA with Scheffé's post hoc tests.

Figure 3.

Increased accumulation of superoxide and mitochondrial ROS and apoptotic cell death in hippocampal neurons from PS1M146VKI mice shown in phase-contrast micrographs (A) and propidium iodide fluorescence (B) in hippocampal neurons from wild-type mice (WITPS1) and PS1 mutant knock-in mice (M146VKI) 48 h following exposure to 20βM Aβ1-42. Note increased shrinkage and damage and nuclear chromatin condensation and fragmentation in neurons from the M146VKI mice. C and D : HE fluorescence (C ; a measure of superoxide levels) and DHR fluorescence (D ; a measure of mitochondrial ROS) 12 h following exposure to 20 μM Aβ1-42. Note increased levels of superoxide and mitochondrial ROS in neurons from the M146VKI mice.

FIG. 2.

FIG. 3.

Evidence that increased superoxide production contributes to the endangering action of mutant PS1

Because oxidative stress and mitochondrial dysfunction have been linked to AD in humans (for review, see Mark et al., 1996 ; Benzi and Moretti, 1997), and because overexpression of Mn-superoxide dismutase protects cultured tumor cells against Aβ-induced apoptosis (Keller et al., 1998), we measured relative levels of superoxide anion following exposure to Aβ in cultured PS1M146VKI and wild-type hippocampal neurons using the probe HE. In wild-type neurons, Aβ induced a two- to threefold increase in levels of HE fluorescence during a 24-h exposure period (Fig. 4A). In contrast, hippocampal neurons from PS1M146VKI mice exhibited significantly greater and more rapid increases in superoxide levels such that HE fluorescence levels were elevated four- to sixfold within 4-12 h of exposure (Fig. 4A). Confocal laser scanning microscope images of HE fluorescence in wild-type and PS1M146VKI hippocampal neurons following exposure to Aβ are shown in Fig. 3C. We next measured relative levels of mitochondrial ROS in hippocampal neurons from wild-type and PS1M146VKI mice using the fluorescent probe DHR, which detects primarily peroxynitrite (Kooy et al., 1994 ; Mattson et al., 1997a). Aβ induced a significantly greater increase in DHR fluorescence in neurons expressing PS1M146V as compared with wild-type neurons (Figs. 3D and 4B Aβ-induced membrane lipid peroxidation, as measured using the TBARS assay, also increased more rapidly and to a greater magnitude in hippocampal neurons from the PS1M146VKI mice (Fig. 4C).

Figure 4.

Aβ-induced accumulation of ROS is enhanced in hippocampal neurons expressing mutant PS1. A : Cultures were exposed to 20 μM Aβ-42 for the indicated intervals, and levels of HE fluorescence, a measure of superoxide levels, were quantified. Data are mean ± SE (bars) values of determinations made in four cultures (20-30 neurons assessed per culture). **p < 0.01, ***p < 0.001 compared with value for WTPS1 cultures exposed to Aβ by ANOVA with Scheffé's post hoc tests. B : Cultures were exposed for 12 h to 20 μM Aβ1-42, and levels of DHR fluorescence, a measure of mitochondrial ROS levels, were quantified. Data are mean ± SE (bars) values of determinations made in four cultures (20-30 neurons assessed per culture). ***p < 0.001 compared with value for WTPS1 cultures exposed to Aβ by ANOVA with Scheffé's post hoc tests. C : Cultures were exposed to 20 μM Aβ1-42 for the indicated intervals, and levels of TBARS fluorescence, a measure of membrane lipid peroxidation, were quantified. Data are mean ± SE (bars) values of determinations made in four cultures (20-30 neurons assessed per culture). ***p < 0.001 compared with value for WTPS1 cultures by ANOVA with Scheffé's post hoc tests.

FIG. 4.

Peroxynitrite, which is formed by the interaction of nitric oxide with superoxide, is increasingly recognized as a key ROS involved in various neurodegenerative conditions, including AD (Smith et al., 1997), and has been shown to play a major role in Aβ neurotoxicity (Hu and el-Fakahany, 1993, Mattson et al., 1997a ; Keller et al., 1998). Because the analyses with HE and DHR suggested increased Aβ-induced superoxide and peroxynitrite production in hippocampal neurons expressing mutant PS1, we determined whether the increased oxidative stress was necessary for the increased vulnerability. To this end, we used the natural antioxidant peroxynitrite scavenger uric acid (Mattson et al., 1997a ; Hooper et al., 1998 ; Keller et al., 1998). Uric acid significantly attenuated Aβ-induced neuronal death in neurons from PS1M146VKI mice (Fig. 5). suggesting a major role for increased ROS levels in the death-enhancing action of the PS1 mutation.

Figure 5.

Uric acid protects hippocampal neurons against the death-enhancing effect of mutant PS1. Cultures were pretreated for 1 h with either vehicle or 200 μM uric acid. Cultures were then exposed to 20 μM Aβ1-42 for 48 h, and neuron survival was quantified. Data are mean ± SE (bars) values of determinations made in four cultures. **p <0.01 compared with value for WTPS1 cultures exposed to Aβ and to Aβ M146VKI cultures pretreated with uric acid and then exposed to Aβ by ANOVA with Scheffé's hoc tests.

FIG. 5.

Evidence that increased caspase activation contributes to the endangering action of mutant PS1

Caspases play a major role in effecting the cell death process in cells undergoing apoptosis (Miller, 1997), and a previous study demonstrated efficacy of a caspase inhibitor in preventing Aβ-induced apoptosis in a cultured tumor cell line (Keller et al., 1998). To establish whether caspases play a role in Aβ neurotoxicity and its enhancement by mutant PS1 in primary hippocampal neurons, we measured levels of activated caspases and determined whether a caspase inhibitor would prevent cell death. Exposure of wild-type hippocampal neurons to Aβ resulted in a relatively slow but progressive increase in caspase activity during a 12-h interval (Fig. 6A). In contrast, a more rapid and much greater increase in caspase activation occurred in response to Aβ in hippocampal neurons from PS1M146VKI mice (Fig. 6A). Pretreatment of hippocampal cultures from PS1M146VKI mice with the broad-spectrum caspase inhibitor zVAD-fmk abolished the death-enhancing effect of the PS1 mutation and largely prevented Aβ-induced cell death (Fig. 6B). These findings suggest that caspase activation is a necessary step in the cell death process activated by Aβ in hippocampal neurons expressing mutant PS1.

Figure 6.

Evidence that increased caspase activation contributes to the cell death-enhancing effect of mutant PS1 in hippocampal neurons. A : Cultures were exposed to vehicle [control (Con)] or 20 μM Aβ1-42 for the indicated intervals, and levels of DEVD fluorescence, a measure of caspase activation, were quantified. Data are mean ± SE (bars) values of determinations made in four cultures (25-40 neurons assessed per culture). **p < 0.01 compared with WTPS1, Con cultures. ***p < 0.001 compared with value for WTPS1 cultures exposed to Aβ. B : Cultures were pretreated for 2 h with vehicle or 100 μM zVAD-fmk (ZVAD) Cultures were then exposed to 20 βM Aβ1-42 for 72 h, and neuron survival was quantified. Data are mean ± SE (bars) values of determinations made in four cultures. *p <0.05, ***p <0.001 compared with value for cultures not pretreated with ZVAD by ANOVA with Scheffé's post hoc tests.

FIG. 6.

DISCUSSION

Our data showing that hippocampal neurons from PS1M146VKI mice exhibit increased vulnerability to Aβ toxicity provide direct evidence that presenilin mutations can endanger a population of neurons vulnerable in humans with AD. Previous studies showing that presenilin mutations increase cell vulnerability to apoptosis (Guo et al., 1996, 1997 ; Wolozing et al., 1996) were performed in cultured tumor cells, which lack many properties, e.g., expression of glutamate receptors, synaptic connections, and high sensitivity to oxidative injury, of the populations of neurons vulnerable in AD. Increased production of superoxide and peroxynitrite appears to play a major role in the neuronal death enhancing action of PS1 mutations in hippocampal neurons. Hippocampal neurons form PS1M146VKI mice exhibited increased superoxide and peroxynitrite levels following exposure to Aβ, as measured using the probes HE and DHR. Associated with increased ROS was increased lipid peroxidation as measured with the TBARS assay. Increased superoxide levels can promote lipid peroxidation by increasing levels of both peroxynitrite and hydroxyl radical (Beckman and Crow, 1993). The ability of uric acid, a natural antioxidant with potent peroxynitrite-scavenging properties (Mattson et al., 1997a ; Hooper et al., 1998 ; Yu et al., 1998), to counteract the death-enhancing action of mutant PS1 argues for a necessary role for increased superoxide and peroxynitrite levels in the endangering action of this mutation. We previously reported that uric acid can inhibit Aβ toxicity in wild-type hippocampal neurons in cell culture (Mattson et al., 1997a ; Keller et al., 1998). The latter studies showed that uric acid suppressed lipid peroxidation and stabilized mitochondrial function following exposure of neurons to Aβ. The specific mechanism whereby PS1 mutations promote increased superoxide production following exposure of neurons to an Aβ is not known. However, the ability of PS1 mutations to perturb cellular calcium homeostasis (Guo et al., 1996, 1997, 1998a) may play a role because sustained elevations of intracellular [Ca2+], such as those induced by Aβ (Mattson et al., 1992, 1993), can induce increased mitochondrial superoxide production (Lafon-Cazal et al., 1993). Further evidence for a role for perturbed calcium homeostasis in the pathogenic action of PS1 mutations comes from studies showing that overexpression of the calcium-binding protein calbindin protects PC12 cells expressing mutant PS1 against Aβ toxicity and suppresses oxyradical production and mitochondrial dysfunction (Guo et al., 1998a). In addition, agents that block calcium release from the endoplasmic reticulum or influx through plasma membrane channels can attenuate Aβ-induced apoptosis in PC12 cells expressing mutant PS1 (Guo et al., 1996).

Caspases are known to play a key role in the effector phase of apoptosis (Miller, 1997). We found that levels of caspase activity following exposure to Aβ were increased in hippocampal neurons from PS1 mutant mice and that the broad-spectrum caspase inhibitor zVAD-fmk protected neurons expressing mutant PS1 against Aβ toxicity. These findings are in line with the increased levels of mitochondrial oxidative stress observed in cells expressing mutant PS1 (Guo et al., 1997, 1998a, present study), because caspase activation is downstream of such mitochondrial alterations in many different paradigms of apoptosis (Kroemer et al., 1997). It was recently reported that PS1 is a substrate for caspases (Kim et al., 1997), although this cleavage does not appear to affect the normal function of PS1 or its effect on APP processing (Brockhaus et al., 1998), and it is unclear whether caspase-mediated cleavage of PS1 plays a role in the cell death process.

Our data from PS1 mutant knock-in mice show that normal levels of expression of mutant PS1 are sufficient to increase the vulnerability of neurons to apoptosis. Prior studies used clonal cell lines in which mutant presenilins were overexpressed at levels manyfold greater than normal (Guo et al., 1996, 1997 ; Wolozin et al., 1996 ; Janicki and Monteiro, 1997). We have shown that levels of expression of PS1M146V protein from the targeted allele are normal and that the mutation does not appear to alter the subcellular distribution of the PS1 protein. This mouse model therefore more closely mimicks human familial AD caused by PS1 mutations than does the transgenic mouse model or transfected cells that overexpress mutant PS1. It has been reported that PS1 null mice die before or shortly after birth with severe vertebral skeletal malformations, CNS neuronal cell loss, and brain and/or spinal cord hemorrhage (Shen et al., 1997 ; Wong et al., 1997). The absence of embryonic lethality in the homozygous PS1M146VKI mice strongly suggests that familial AD mutations do not alter the normal developmental function of PS1, although the possibility of loss of some other function of PS1 that is not involved in development cannot be completely ruled out. Our data are therefore consistent with recent reports showing that ectopic expression of mutant human PS1 protein in mice can protect PS1 null mice against embryonic lethality (Davis et al., 1998 ; Qian et al., 1998). PS1 mutations therefore appear to result in gain of an adverse property of the protein that leads to increased vulnerability to apoptosis.

The relationship(s) between presenilin mutation effects on APP processing (Borchelt et al., 1996 ; Duff et al., 1996 ; Scheuner et al., 1996) and cell vulnerability to apoptosis are unknown. It is well established that PS1 mutations increase production of Aβ1-42 in brain tissues and in cultured cells (Hardy, 1997). The cell deathenhancing action of mutant PS1 does not appear to result from increased Aβ production or cellular release of other cytotoxic factors, because mouse hippocampal neurons expressing mutant PS1 exhibit increased vulnerability to apoptosis yet such rodent cells produce a form of Aβ that does not readily aggregate and is not neurotoxic (Otvos et al., 1993 ; M. P. Mattson, unpublished data). Moreover, medium taken from PC12 cell cultures overexpressing mutant PS1 does not increase the vulnerability of control PC12 cells to apoptosis (M. P. Mattson, unpublished data). However, we cannot rule out a role for altered APP processing in the cell death-enhancing action of presenilin mutations, because α-secretase cleavage is reduced in cultured cells expressing PS1 mutations (Ancolio et al., 1997), and the α-secretase-derived from of soluble β-amyloid precursor protein (sAPPα) protects neurons against apoptosis (Goodman and Mattson, 1994 ; Furukawa et al., 1996 ; Guo et al., 1998c). Increased levels of metabolic/oxidative stress (Gabuzda et al., 1994 ; Gasparini et al., 1997) and increased intracellular calcium levels (Buxbaum et al., 1994 ; Querfurth and Selkoe, 1994) shift β-amyloid precursor protein processing in a manner similar to that observed with PS1 mutations, i.e., increased levels of Aβ and decreased levels of sAPPα. It is therefore reasonable to consider that the altered β-amyloid precursor protein processing is secondary to a primary effect of the PS1 mutations on calcium homeostasis and oxyradical metabolism. By increasing Aβ levels and decreasing sAPPα levels, the altered β-amyloid precursor protein processing may render neurons vulnerable to excitotoxicity and apoptosis (Mattson, 1997).

Studies of postmortem brain tissue from patients with late-onset sporadic AD have provided considerable evidence for increased superoxide production, peroxynitrite formation, and membrane lipid peroxidation in association with degenerating neurons (Good et al., 1996 ; Lovell et al., 1997 ; Sayre et al., 1997 ; Smith et al., 1997 ; Su et al., 1997). Superoxide levels have not been directly measured in AD brains, but levels of Mn-superoxide dismutase and Cu/Zu-superoxide dismutase were reported to be increased in homogenates of AD brain tissue (Bruce et al., 1997), suggesting a response to increased superoxide levels. Moreover, neurons from Down's syndrome patients (an early-onset form of AD involving amyloid deposition) show increased generation of ROS and undergo “spontaneous” apoptosis in cell culture (Busciglio and Yankner, 1995). Aβ may promote oxidative stress and neuronal degeneration in AD as suggested by the ability of Aβ to induce production of superoxide and peroxynitrite and membrane lipid peroxidation in cultured neurons (Mark et al., 1997a ; Mattson et al., 1997a ; present study). Exposure of hippocampal neurons to oxidative insults and Aβ, both in cell culture and in vivo, results in antigenic changes in the microtubule-associated protein tau similar to those seen in the neurofibrillary tangles of AD (Mattson, 1990 ; Stein-Behrens et al., 1994 ; Busciglio et al., 1995 ; Mattson et al., 1997c ; Geula et al., 1998). By enhancing disruption of calcium homeostasis and enhancing oxidative stress in neurons, presenilin mutations may promote neurofibrillary degeneration and apoptosis. Although it remains to be conclusively established that neurons die by apoptosis in AD, the expression of Par-4, a gene linked to neuronal apoptosis, is increased in tangle-bearing neurons of AD patients and in PC12 cells expressing mutant PS1 (Guo et al., 1998b). Par-4 appears to promote apoptosis, in part, by enhancing levels of oxidative stress. Further support for the involvement of oxidative stress in the pathogenesis of AD comes from studies showing that antioxidants such as vitamin E and estrogen (which act by suppressing lipid peroxidation) can protect neurons against Aβ toxicity (Behl et al., 1992 ; Goodman and Mattson, 1994 ; Goodman et al., 1996) and can reduce the risk for developing AD and/or slow its progression (Tang et al., 1996 ; Sano et al., 1997). Our data suggest that early-onset familial AD, caused by PS1 mutations, may also involve increased oxidative stress in neurons and, accordingly, that antioxidant approaches may prove effective in treating patients with familial AD.

Acknowledgements

Acknowledgement : We thank J. Partin and L. Yang for technical assistance. This work was supported by grants to M.P.M. from the National Institutes of Health (AG 14554, AG05144, and NS35253), to C.B.W. from the National Institute on Aging (AG05136), to B.L.S. from the University of Washington Nathan Shock Center for Excellence in the Basic Biology of Aging, and to G.M.M. from the National Institute on Aging (ADRC AG05136).

Ancillary