Amyloid β-induced Changes in Nitric Oxide Production and Mitochondrial Activity Lead to Apoptosis*

Increasing evidence suggests an important role of mitochondrial dysfunction in the pathogenesis of Alzheimer's disease. Thus, we investigated the effects of acute and chronic exposure to increasing concentrations of amyloid β (Aβ) on mitochondrial function and nitric oxide (NO) production in vitro and in vivo. Our data demonstrate that PC12 cells and human embryonic kidney cells bearing the Swedish double mutation in the amyloid precursor protein gene (APPsw), exhibiting substantial Aβ levels, have increased NO levels and reduced ATP levels. The inhibition of intracellular Aβ production by a functional γ-secretase inhibitor normalizes NO and ATP levels, indicating a direct involvement of Aβ in these processes. Extracellular treatment of PC12 cells with comparable Aβ concentrations only leads to weak changes, demonstrating the important role of intracellular Aβ. In 3-month-old APP transgenic (tg) mice, which exhibit no plaques but already detectable Aβ levels in the brain, reduced ATP levels can also be observed showing the in vivo relevance of our findings. Moreover, we could demonstrate that APP is present in the mitochondria of APPsw PC12 cells. This presence might be directly involved in the impairment of cytochrome c oxidase activity and depletion of ATP levels in APPsw PC12 cells. In addition, APPsw human embryonic kidney cells, which produce 20-fold increased Aβ levels compared with APPsw PC12 cells, and APP tg mice already show a significantly decreased mitochondrial membrane potential under basal conditions. We suggest a hypothetical sequence of pathogenic steps linking mutant APP expression and amyloid production with enhanced NO production and mitochondrial dysfunction finally leading to cell death.

Increasing evidence suggests an important role of mitochondrial dysfunction in the pathogenesis of Alzheimer's disease. Thus, we investigated the effects of acute and chronic exposure to increasing concentrations of amyloid ␤ (A␤) on mitochondrial function and nitric oxide (NO) production in vitro and in vivo. Our data demonstrate that PC12 cells and human embryonic kidney cells bearing the Swedish double mutation in the amyloid precursor protein gene (APPsw), exhibiting substantial A␤ levels, have increased NO levels and reduced ATP levels. The inhibition of intracellular A␤ production by a functional ␥-secretase inhibitor normalizes NO and ATP levels, indicating a direct involvement of A␤ in these processes. Extracellular treatment of PC12 cells with comparable A␤ concentrations only leads to weak changes, demonstrating the important role of intracellular A␤. In 3-month-old APP transgenic (tg) mice, which exhibit no plaques but already detectable A␤ levels in the brain, reduced ATP levels can also be observed showing the in vivo relevance of our findings. Moreover, we could demonstrate that APP is present in the mitochondria of APPsw PC12 cells. This presence might be directly involved in the impairment of cytochrome c oxidase activity and depletion of ATP levels in APPsw PC12 cells. In addition, APPsw human embryonic kidney cells, which produce 20-fold increased A␤ levels compared with APPsw PC12 cells, and APP tg mice already show a significantly decreased mitochondrial membrane potential under basal conditions. We suggest a hypothetical sequence of pathogenic steps linking mutant APP expression and amyloid production with enhanced NO production and mitochondrial dysfunction finally leading to cell death.
Alzheimer's disease (AD) 1 is the most common neurodegenerative disorder (1) marked by progressive loss of memory and cognitive ability. The pathology of AD is characterized by the presence of amyloid plaques (2) and intracellular neurofibrillary tangles and pronounced cell death. The amyloid plaque is composed of amyloid ␤ (A␤) peptide (3), which is derived from the amyloid precursor protein (APP) through an initial ␤-secretase cleavage followed by an intramembraneous cut of ␥-secretase (4,5). Autosomal dominant forms are caused by mutations in APP, presenilin 1, and presenilin 2, mainly associated with the early onset of AD (6). The Swedish double mutation in the APP gene (K670M/N671L) results in a 6 -8-fold increased A␤ production compared with human wild type APP cells (APPwt) (7,8). We have previously shown that the APPsw mutation enhances the vulnerability to secondary insults, e.g. oxidative stress, finally leading to apoptotic cell death through the activation of the c-Jun N-terminal kinase and caspases 3 and 9 (9). The latter observation provided evidence that mitochondriamediated apoptosis might play an important role in these processes.
Intrinsic apoptotic pathway via mitochondria is regulated by members of the Bcl-2 family (10). They are mainly localized in the outer mitochondrial membrane. Bcl-2 and Bcl-x L inhibit apoptosis, whereas other members such as Bax, Bak, Bid, Bik, and Bim are proapoptotic (11). Their most frequently reported mode of action is the regulation of cytochrome c release from the intermembrane space (12). Once released from the mitochondria, cytochrome c interacts with Apaf 1 and procaspase 9 to activate caspase 3, finally leading to cell death. Smac (second mitochondria-derived activator of caspase) is another mitochondrial intermembrane protein that is released in the cytosol during apoptosis (13). Smac interacts with several inhibitors of apoptosis, thereby relieving the inhibitory effect of inhibitors of apoptosis on caspases (14). As an early event in the apoptotic pathway, typically, a rapid reduction of the mitochondrial membrane potential (⌿ m ) takes place (15), which reflects a block of the respiratory chain finally leading to the reduction of ATP levels.
Some previous findings suggest that neurotoxicity of A␤ (16,17) seems to be mediated via oxidative stress probably leading to mitochondrial damage, e.g. A␤ inhibited cytochrome c oxidase (COX) activity in isolated brain mitochondria (18,19), and in neuronal cultures, A␤ caused a loss of activity of all mitochondrial respiratory chain complexes (20 -23). Moreover, activity changes in mitochondrial enzymes including pyruvate dehydrogenase and ␣-ketoglutarate dehydrogenase have been described in AD brain (24). In addition, patients with AD showed impaired cytochrome c oxidase activity in the central nervous system and even in other tissues including platelets (19,(25)(26)(27). Interestingly, p 0 cells repopulated with mitochondria from AD patients also showed a reduced cytochrome c oxidase activity (28). Thus, the defects in mitochondrial energy metabolism that lead to increased production of reactive oxygen species (ROS) seem to underlie the pathology of AD (29). Mitochondria represent not only the major source of ROS but also the main target for ROS damage. Nitric oxide (NO) and its derivative (reactive nitrogen species), also belonging to the group of ROS, are known to inhibit the mitochondrial respiration (30). Hereby, NO itself causes a selective reversible inhibition of cytochrome c oxidase (31), whereas RNS inactivate multiple respiratory chain complexes and ATP synthase (32). In accordance, both oxidative and nitrosative stress seem to represent early events in the pathogenesis of AD (33)(34)(35)(36).
On the basis of these findings, we set out to investigate the precise mechanisms underlying the action of A␤ and/or expression of mutant APP on mitochondrial function in multiple experimental designs mimicking different in vivo situations that are discussed to occur in AD patients: 1) in cell lines overexpressing different levels of A␤ (APPsw PC12 cells exhibiting low physiological concentrations of A␤ within picomolar range and APPsw HEK cells expressing A␤ levels within low nanomolar range) to study dose-dependent effects of A␤ in an in vitro setting characterized by chronic A␤ production due to increased APP processing; 2) after extracellular A␤ treatment to distinguish chronic from acute and/or extracellular actions from effects of intracellular A␤ or APP-processing products on mitochondria; and 3) in a secondary insult model to examine the additional impact of oxidative stress. Finally, we checked the in vivo relevance of our in vitro findings by studying mitochondrial function in brain cells from APP transgenic (tg) mice.

EXPERIMENTAL PROCEDURES
Materials-Rhodamine 123 and tetramethylrhodamineethylester (TMRE) were purchased from Molecular Probes. 4,5-Diaminofluorescein diacetate was obtained from Calbiochem. ViaLight HT kit was purchased from Cambrex. Cytochrome c oxidase assay kit, hydrogen peroxide, rotenone, thenoyltrifluoroacetone, antimycin, sodium azide (NaN 3 ), and oligomycin were obtained from Sigma. L-NAME was purchased from Cayman. N -propyl-L-arginine and 1400W were purchased from Biotrend. A␤ 1-42 was supplied by Bachem. DAPT was obtained from Merck Biosciences. A␤  was dissolved in Tris-buffered saline (pH 7.4) at a concentration of 1 mM and stored at Ϫ20°C. The stock solution was diluted in Tris-buffered saline to the desired concentrations and incubated at 37°C for 24 h to have aged preparations of A␤  .
The transfected cell lines APPwt HEK, APPsw HEK, and control HEK were cultured in Dulbecco's modified Eagle's medium supplemented with 10% heat-inactivated fetal calf serum, 50 units/ml penicillin, 50 g/ml streptomycin, and 400 g/ml G418 at 37°C in a humidified incubator containing 5% CO 2 . The A␤ levels of APP-transfected PC12 cells and HEK cells are shown in Table I.
Detection of A␤ levels-For the detection of secreted A␤ 1-40 , a specific sandwich enzyme-linked immunosorbent assay (ELISA) employing monoclonal antibodies was used. The ELISA was performed according to the instructions given in the Abeta-ELISA kit by BIOSOURCE. The assay principle is that of a standard sandwich ELISA, which utilizes a monoclonal mouse anti-human Abeta1-16 capture antibody, a cleavage site-specific rabbit anti-human Abeta1-40 C-terminal detection antibody, and anti-rabbit IgG peroxidase-conjugated secondary antibody.
Transgenic Animal Brain Tissue-Female C57BL/6 mice bearing the human Swedish and London mutations in the 751 amino acid form of human amyloid precursor protein (tg APP) under the control of a murine Thy-1 promoter at an age of 3 months and non-transgenic littermate animals were used for the experiments (38). APP tg mice exhibited the onset of A␤ plaques at an age of 6 months, but intracellular A␤ load was already detectable at the age of 3 months (39).
Mice were sacrificed by decapitation, and brains were quickly dissected on ice (method modified after Stoll et al. (40)). After removing the cerebellum, the tissue was minced into 2 ml of medium I (138 NaCl, 5.4 KCl, 0.17 Na 2 HPO 4 , 0.22 K 2 PO 4 , 5.5 glucose, and 58.4 sucrose, all in mmol/liter, pH 7.35) with a scalpel and further dissociated by trituration through a nylon mesh (pore diameter 1 mm) with a Pasteur pipette. The resulting suspension was filtered by gravity through a fresh nylon mesh with a pore diameter of 102 M, and the dissociated cell aggregates were washed twice with medium II (110 NaCl, 5.3 KCl, 1.8 CaCl 2 ⅐H 2 O, 1 MgCl 2 ⅐6 H 2 O, 25 glucose, 70 sucrose, and 20 HEPES, all in mmol/l, pH 7.4) by centrifugation (400 ϫ g for 3 min at 4°C). 100 l of the suspension were used for protein determination. After centrifugation, cells were resuspended in 6 ml of Dulbecco's modified Eagle's medium and 500 l/well were distributed on a 24-well plate for the measurement of mitochondrial membrane potential. For the measurement of ATP levels, 100 l/well were distributed on a white 96-well plate. The preparations of APP tg mice and non-transgenic littermate mice (overcross design) were made under the same conditions and in parallel.
Quantification of Apoptosis by Flow Cytometry-Apoptosis was determined by propidium iodide staining and fluorescence-activated cell sorter analysis as described previously (37). PC12 cells and HEK cells were lysed in buffer (0.1% sodium citrate and 0.1% Triton X-100) containing 50 g/ml propidium iodide. Samples were analyzed by flow cytometry (FACSCalibur) using Cell Quest software (BD Biosciences). Cells with a lower DNA content showing less propidium iodide staining than G 1 have been defined as apoptotic cells (sub-G 1 peak).
Determination of Intracellular Nitric Oxide Levels-PC12 cells and HEK cells were plated the day before at a density of 2 ϫ 10 5 cells/well in a 24-well plate. To measure the NO levels, the fluorescence dye 4,5-diaminofluorescein diacetate was used in a concentration of 10 M for 30 min (41). The cells were washed twice with Hanks' balanced salt solution, and the fluorescence was determined with a fluorescence reader (Victor® multilabel counter) at 490/535 nm.
NO levels were also determined after a 48-h incubation in the absence or in the presence of the NO synthase inhibitors 20 mM L-NAME, 1 mM N -propyl-L-arginine, and 1 mM 1400W.
Determination of the Mitochondrial Membrane Potential ( m )-PC12 cells and HEK cells were plated the day before at a density of 2 ϫ 10 5 cells/well in a 24-well plate. PC12 cells were incubated with H 2 O 2 (0.5 mM) for different periods of time. The mitochondrial membrane potential was measured using the fluorescence dye Rhodamine 123 (42). Transmembrane distribution of the dye depends on the mitochondrial membrane potential. The dye was added to the cell culture medium in a concentration of 0.4 M for 15 min. The cells were washed twice with Hanks' balanced salt solution, and the fluorescence was determined with a fluorescence reader (Victor multilabel counter) at 490/535 nm.
The mitochondrial membrane potential of dissociated neurons was also measured with Rhodamine 123 in a concentration of 0.4 M for 15 min. For detailed information regarding the preparation, see "Transgenic Animal Brain Tissue." To test acute and fast changes in m , the fluorescence dye TMRE (43) was used in a concentration of 0.4 M for 15 min. TMRE exhibits a characteristic increase in fluorescence at 490/590 nm after challenging mitochondria with membrane potential-decreasing drugs (44). The mitochondrial membrane potential was recorded, and then the complex inhibitors (2 M rotenone, 10 M thenoyltrifluoroacetone, 2 M antimycin, 10 M oligomycin, and 10 mM sodium azide) were added.
Determination of ATP Levels with a Bioluminescence Assay (ViaLight HT)-PC12 cells and HEK cells were plated the day before at a density of 2 ϫ 10 4 cells/well in a white 96-well plate. PC12 cells were incubated for different periods of time with H 2 O 2 (0.1 mM).
The ATP levels of dissociated neurons were also measured with the bioluminescence assay. For detailed information regarding this preparation, see "Transgenic Animal Brain Tissue." The kit is based upon the bioluminescent measurement of ATP (45).
The bioluminescent method utilizes an enzyme, luciferase, which catalyzes the formation of light from ATP and luciferin. The emitted light is linearly related to the ATP concentration and is measured using a luminometer.
Confocal Microscopy-The amount of mitochondria was measured using the cell-permanent mitochondrion-selective dye Mitotracker Red. This probe can accumulate in active mitochondria and then react with accessible thiol groups of proteins and peptides to form fluorescent aldehydic-fixable conjugates. PC12 cells were plated the day before at a density of 2 ϫ 10 5 cells/chamber slide. Cells were incubated for 30 min at 37°C with Mitotracker Red or for 15 min with Rhodamine 123. Cells were washed twice with phosphate-buffered saline, fixed with 2% paraformaldehyde for 15 min, and then washed twice with phosphatebuffered saline. The samples were embedded in Mowiol and analyzed using a laser-scanning confocal microscope.
Isolation of Cytosolic and Mitochondrial Fractions-Cytosolic and mitochondrial fractions were isolated by digitonin permeabilization (46). 5 ϫ 10 6 cells were washed with ice-cold phosphate-buffered saline, and cells were resuspended for 15 min on ice in permeabilization buffer containing 75 mM NaCl, 1 mM NaH 2 PO 4 , 250 mM sucrose, 1 mM phenylmethylsulfonyl fluoride, additional protease inhibitors, and 0.05% digitonin. Following a centrifugation step at 800 ϫ g at 4°C for 10 min, the supernatant was separated from the pellet consisting of cellular debris. The crude mitochondrial pellet was collected by centrifugation at 13,000 ϫ g at 4°C for 10 min. The supernatant containing cytoplasmic proteins was stored at Ϫ20°C for further investigation. The pellet consisting of mitochondria was dissolved in phosphate-buffered saline for the cytochrome c oxidase assay and resuspended in 0.1% Triton X-100 and mechanically lysed for the Western blot. The total protein content was determined by the method of Lowry (Bio-Rad).
Determination of Cytochrome c Oxidase Activity in Isolated Mitochondria with Cytochrome c Oxidase Assay Kit (47)-The colorimetric assay is based on the observation of the decrease in absorbance at 550 nm of ferrocytochrome c caused by its oxidation to ferricytochrome c by cytochrome c oxidase. The cytochrome c oxidase assay was performed according to the instructions given in the kit.
Western Blot-After determination of the total protein content by the method of Lowry, the cytosolic and mitochondrial fraction was mixed with 4ϫ Laemmli sample buffer and denatured for 10 min at 95°C. An equal amount (10 -20 g) of protein was loaded per lane on acrylamide gels and examined by SDS-PAGE. Dependent on the protein to be detected, we performed glycine or Tricine Western blots with acrylamide amounts of 10 -18% at 90 V for 2-3 h. For proteins with a molecular mass Ͻ30 kDa, Tricine Western blots were more convenient. The proteins were transferred onto polyvinylidene difluoride membranes (Amersham Biosciences) at 25 V for 90 min. After this procedure, membranes were saturated with 5% nonfat dry milk in Tris-buffered saline with Tween 20 for 1 h after antibody exposure. The polyvinylidene difluoride membranes then were exposed to the following antibodies overnight: rabbit anti-Bcl-x L (Cell Signaling); rabbit anti-Bax (Cell Signaling); rabbit anti-apoptosis-inducing factor (AIF) (Chemicon); rabbit anti-Smac (Chemicon); rabbit anti-cytochrome c (BD Biosciences), mouse anti-COX4 (BD Biosciences); goat anti-actin (Santa Cruz Biotechnology); rabbit anti-calreticulin (Chemicon); mouse anti-Na ϩ /K ϩ -ATPase (Chemicon); rabbit anti-nitrotyrosine (Merck Biosciences); and mouse anti-human APP W02 (Abeta; W02 antibody is directed against amino acids 4 -10 of the human A␤ sequence, thereby detecting the mature and immature APP695). After treatment for 1 h with the corresponding horseradish peroxidase-coupled secondary antibodies (Merck Biosciences), the protein bands were detected by ECL reagent (Amersham Biosciences). After detection, the membranes were treated with stripping buffer (100 mM glycine, pH 2.5) for 2 h after reprobing with different antibodies.
Statistical Analysis-The data are given as the mean Ϯ S.E. For statistical comparison, Student's t test and one-way ANOVA followed by Tukey's post hoc test, repeated measures ANOVA, or two-way ANOVA were used. p values Ͻ 0.05 were considered statistically significant.

APP Expression and A␤ Levels of APP-transfected PC12 Cells
and HEK Cells-Increasing knowledge was obtained that A␤ is produced intracellularly (48) and that intraneuronal accumulation of A␤ precedes plaque formation in APP tg mice bearing the Swedish double mutation (49,50). Thus, intracellular accumulation of A␤ might be a primary step in the neurotoxicity cascade of A␤ in vivo and in vitro. APP containing the Swedish mutation resulted in an overall increase of A␤ including both A␤ 1-40 and A␤  .
In our cell model, the APPsw mutation resulted in a markedly increased A␤ secretion of PC12 cells (0.20 nmol/liter) as well as in HEK cells (5.42 nmol/liter) compared with APPwt PC12 cells and APPwt HEK cells. APPsw PC12 cells secreted low A␤ levels, reflecting the physiological situation in vivo during cellular metabolism (7), whereas human APP was equally expressed in APPwt and APPsw PC12 cells and was increased 2-fold compared with the endogenous APP expression of control PC12 cells (37). Of note, HEK cells bearing the APPsw mutation exhibited an ϳ20-fold increased A␤ secretion compared with APPsw-bearing PC12 cells (Table I), a scenario that might occur in familial AD and that could explain the more drastic effects in some assay designs. Furthermore, this might indicate that, in APPsw HEK cells, the secretion pathway is more pronounced than in PC12 cells. The endogenous human APP expression of control HEK cells was increased 2-fold compared with APPsw PC12 cells. Moreover, the APP expression of APPwt and APPsw HEK cells was increased 3-fold compared with control HEK cells and 6-fold compared with APPsw PC12 cells. Using the different cell lines allowed us to study the dose-dependent effects of A␤ on those stably transfected cells with APPwt or APPsw, both representing a model of chronic A␤ stress.
Basal Apoptosis Was Increased in APPsw HEK Cells but Not in APPsw PC12 Cells-Apoptotic cell death plays a very important role in the pathogenesis of AD (51). Thus, we investigated basal apoptosis using propidium iodide staining. In APPsw PC12 cells, we observed no increased basal apoptosis (Fig. 1A). Thus, APPsw PC12 cells were able to compensate for the consequences of the increased A␤ levels. However, APPsw HEK cells, which have high A␤ levels, showed significantly increased levels of apoptotic cells under base-line conditions (Fig. 1B). Additionally, the treatment of PC12 cells with extracellular A␤ 1-42 at concentrations of 10 nM and higher resulted also in increased apoptotic cell death (Fig. 1C). The effect on the induction of apoptosis was less pronounced after extracellular A␤ treatment than in APPsw HEK cells. Even the A␤ 1-42 peptide was used, which is more toxic than the A␤ 1-40 peptide.
Basal NO Levels Were Increased in APPsw Cells-Excessive generation of nitric oxide has been implicated in the pathogenesis of AD (52). Thus, we measured intracellular NO levels using the fluorescence dye 4,5-diaminofluorescein diacetate in APP-transfected PC12 cells and HEK cells. Our data demonstrated that NO levels are significantly enhanced in the following order: control cells Ͻ APPwt Ͻ APPsw (Fig. 1, E and F  cellularly with A␤ 1-42 (Fig. 1G). Interestingly, NO levels were also enhanced after exposure to A␤ 1-42 with a maximal effect already at 1 nM but to a lesser extent than in APPsw PC12 cells. Thus, the extracellularly applied A␤ cannot completely mimic the physiological situation of APPsw transfection. NO Synthase (NOS) Activity Was Enhanced in APPsw Cells-NO was synthesized from L-arginine by three isoforms of NOS, two of which (endothelial NOS and neuronal NOS) were constitutively expressed, whereas the third (inducible NOS) was induced during inflammation (53). There may also be a mitochondrial isoform (54 -56). To study NO synthase activity, we used L-NAME, an unselective NO synthase inhibitor. Interestingly, L-NAME was able to reduce the NO levels of APPsw cells to the levels of control cells (Fig. 1E), showing an enhanced activity of NOS in APPsw cells. In agreement with this finding, a rise in nitrosated tyrosines could be detected (Fig. 1D), which paralleled elevated NO levels (Fig. 1E). To elucidate which isoform was responsible for the enhanced NO levels in APPsw cells, we used selective NO synthase inhibitors for the neuronal NOS and inducible NOS isoforms (57,58). The decrease of NO levels in APPsw PC12 cells after exposure to the neuronal NOS inhibitor N -propyl-L-arginine was ϳ9.4% compared with 4.2% in control cells and 3.7% in APPwt cells. After exposure to the inducible NOS inhibitor 1400W, the reduction of NO levels was ϳ13.7% in APPsw PC12 cells compared with 3.3% in control cells and 4.6% in APPwt cells (data not shown). Thus, both isoforms seemed to be involved in the A␤-induced NO increase in APPsw PC12 cells. We cannot rule out the involvement of the endothelial NOS isoform in A␤induced NO increase, but this experiment could not be performed due to the lack of commercially available selective endothelial NOS inhibitors.
Amyloid ␤ Leads to Mitochondrial Damage-Because mitochondria play a key role in the cell death decision, we investigated the effect of the APPsw mutation on mitochondrial function in PC12 cells and HEK cells. Additionally, we have also analyzed the effects of extracellular A␤ 1-42 on mitochondrial function in PC12 cells.
Mitochondrial membrane potential is a very important marker for the function of mitochondria (15). A decrease in mitochondrial membrane potential has been related to cell death in different cell types. Interestingly, APPsw-bearing PC12 cells showed a hyperpolarized mitochondrial membrane potential compared with control cells and APPwt PC12 cells ( Fig. 2A). However, APPsw HEK cells, which have 20-fold increased A␤ levels, showed a significantly decreased mitochondrial membrane potential in comparison with control HEK cells and APPwt HEK cells (Fig. 2B). These findings are in accordance with the levels of apoptotic cells, which were unchanged in APPsw PC12 cells but significantly increased in APPsw HEK cells. Treating PC12 cells with extracellular A␤ 1-42 resulted also in a significant depolarization of mitochondrial membrane potential (Fig. 2C) but to a lesser degree than in APPsw HEK cells. Thus, it appears to be important to distinguish between acute and chronic A␤ effects on the one hand and between low and high dose-dependent effects of A␤ on mitochondrial function on the other hand.
Mitochondria are the major source of ATP. Interestingly, not only APPsw PC12 cells but also APPwt PC12 cells show reduced ATP levels in comparison with control PC12 cells (Fig.  2D). HEK cells bearing the APPsw mutation also have significantly decreased ATP levels in the following order: APPsw Ͻ APPwt Ͻ control cells (Fig. 2E). The ATP reduction is stronger in APPsw HEK cells than in APPsw PC12 cells, probably indicating a dose-dependent effect of A␤ on ATP levels.
In addition, the extracellular A␤-treated PC12 cells showed reduced ATP levels (Fig. 2F). As already observed with NO measurement, the maximal effect was already seen after exposure to 1 nM A␤  . Again, the ATP reduction was stronger under physiological conditions using APPsw PC12 cells compared with acute treatment with extracellular A␤ (37 versus 13%).

The Inhibition of ␥-Secretase Leads to a Reduction of NO Levels and Stabilization of Mitochondrial Function in APP-
transfected Cells-To further demonstrate that mitochondrial damage is due to intracellular A␤ and not due to overexpression of APP, APP-transfected PC12 and HEK cells were treated with DAPT, a functional ␥-secretase inhibitor that reduces intracellular A␤ levels without affecting APP expression (59 -61). In PC12 and HEK cells, DAPT does not show neurotoxicity by itself. We observed no decrease in MTT reduction after 24, 48, and 72 h of DAPT incubation (data not shown). However, we could show a strong reduction of secreted A␤ 1-40 levels already after 24 h of DAPT incubation in APPsw PC12 cells and after 48 h in APPsw HEK cells (Table I). Thus, the reduction of secreted A␤ 1-40 below the level of control cells was very drastic in APPsw HEK cells, which secreted 20-fold increased A␤ levels compared with APPsw PC12 cells. A similar decrease in secreted A␤ was also seen in APPwt cells (data not shown). As already shown by others (62), DAPT also reduces significantly intracellular A␤ production in both cell lines. Of note, the reduction of A␤ levels by the ␥-secretase inhibitor DAPT lead to a normalization of NO (Fig. 3A) and ATP levels (Fig. 3B) as well as mitochondrial membrane potential (Fig. 3C) in APPwt and APPsw PC12 and HEK cells, respectively. Thus, the increased NO production and mitochondrial damage in APP-transfected cells were triggered by the production and accumulation of A␤.
Our data suggested that mainly the PC12 cell model represents a very suitable approach to elucidate AD-specific cell death pathways under physiological conditions that are the most relevant for the in vivo situation in man because PC12 cells bearing the APPsw mutation secrete A␤ levels within the picomolar range. For this reason, we made all of the following experiments with this cell model.

APPsw PC12 Cells Have an Increased Amount of Mitochondria and Show Presence of APP in Mitochondria-
To exclude the possibility that ATP levels are reduced in APP-transfected PC12 cells due to a loss of mitochondria, we determined the amount of mitochondria in APP-transfected cells using the mitochondrial dye Mitotracker Red and confirmed co-localization with the mitochondrial membrane potential dye Rhodamine 123 (Fig. 4, A-C). Mitotracker Red selectively stains mitochondria independently from the mitochondrial membrane potential. Interestingly, APPsw PC12 cells contained more mitochondria than APPwt and control PC12 cells (Fig. 4, D-G). Thus, the drop of ATP levels in APPsw PC12 cells was due to mitochondrial dysfunction and not due to the loss of mitochondria. Moreover, the up-regulation of mitochondria seems not to be able to compensate for mitochondrial dysfunction with regard to ATP levels.
Recently, it has been shown that APP is also targeted to mitochondria (63) in addition to its localization in the endoplasmic reticulum and the plasma membrane. In this study, transiently transfected cells were used (63). Interestingly, we also observed the presence of APP in mitochondria of stably transfected APPwt and APPsw PC12 cells (Fig. 5). The purity of subcellular fractions was established by Western blot analysis of various fractions using antibodies to Na ϩ /K ϩ -ATPase (plasma membrane-specific), cytochrome c oxidase subunit 4 (mitochondria-specific), and calreticulin (endoplasmic reticulum-specific) as markers. Further investigations have to be done to further elucidate the role of APP in the mitochondrion.
PC12 Cells Bearing the Swedish APP Mutation Showed an Impairment of the Mitochondrial Respiratory Chain-NO is known to inhibit the mitochondrial respiratory chain (30). Therefore, we analyzed the COX activity. Interestingly, APPsw PC12 cells showed a significantly reduced cytochrome c oxidase activity compared with APPwt and control PC12 cells (Fig. 6A). Moreover, we measured the mitochondrial membrane potential with the fluorescence dye TMRE after stimulation with different complex inhibitors. Here, we showed that respiratory chain complexes II, III, IV, and F 0 F 1 -ATPase in APPsw PC12 cells are more vulnerable against mitochondrial membrane potential changes than control PC12 cells, indicating an impaired mitochondrial respiratory chain (Table II). Additionally, the complexes IV and F 0 F 1 -ATPase of APPwt PC12 cells are more sensitive than control PC12 cells but are as sensitive as APPsw PC12 cells. Thus, these two enzyme complexes are especially vulnerable to already very low A␤ concentrations.
Oxidative Stress Induces Mitochondrial Damage in PC12 Cells Bearing the Swedish APP Mutation-Oxidative stress might be involved in the pathogenesis of Alzheimer's disease (64). Thus, we investigated the effect of a secondary insult, hydrogen peroxide, which increases cell death in APPsw PC12 cells (9), on mitochondrial membrane potential and ATP levels. Interestingly, APPsw-bearing PC12 cells showed a significantly decreased mitochondrial membrane potential after exposure to hydrogen peroxide in comparison with APPwt and control PC12 cells (Fig. 6B). Moreover, the ATP reduction after hydrogen peroxide exposure was more pronounced in APPsw and APPwt PC12 cells compared with control PC12 cells, even in the recovery phase during 2, 4, and 6 h of H 2 O 2 exposure in a time-dependent manner (Fig. 6C). Thus, mutant cells have reduced capability to maintain mitochondrial membrane potential and ATP levels after oxidative stress and constantly showed an ATP deficiency.
The Impact of Mutant APP on Members of the Bcl-2 Family and on the Release of Mitochondrial Factors-The Bcl-2 family proteins are important regulators of apoptosis (10). They are proapoptotic or antiapoptotic in nature. That means that they induce or inhibit the release of mitochondrial proteins in the cytosol including cytochrome c, Smac, and AIF. By Western blot analysis, we investigated the proteins of the Bcl-2 family as well as Smac and AIF. We have previously shown that a time-dependent release of cytochrome c was observed after treatment with hydrogen peroxide, reaching a maximum after 6 h (9). After 4 h of treatment with hydrogen peroxide, there was an enhanced release of Smac observed in APPsw PC12 cells (Fig. 7). AIF, a protein that causes nuclear condensation and DNA fragmentation, was not released by the mitochondria after 6 h of hydrogen peroxide incubation (Fig. 7); however, we observed a higher AIF expression in the mitochondria of APPsw PC12 cells at this time point. Interestingly, we found cells. The purity of subcellular fractions was evaluated by Western blot analysis of various subcellular fractions using antibodies to Na ϩ /K ϩ -ATPase (plasma membrane-specific), COX4 (mitochondria-specific), and calreticulin (endoplasmic reticulum-specific) as markers. Actin serves as a marker for equal loading. AIF release after 24 h of hydrogen peroxide incubation (data not shown).
Concerning the Bcl-2 family, we found a reduced amount of the cytosolic anti-apoptotic protein Bcl-x L in APPsw cells under base-line conditions and after exposure to hydrogen peroxide but only minor changes in Bax protein content (Fig. 7). Thus, the Bcl-x L /Bax ratio was decreased in APPsw PC12 cells under basal conditions (control cells, 1.0; APPwt cells, 0.86; APPsw cells, 0.49), which might possibly explain the enhanced vulnerability to cell death after oxidative stress.
APP tg Mice Showed a Decreased Mitochondrial Membrane Potential and Reduced ATP Levels-To understand the in vivo relevance of our in vitro findings, we analyzed mitochondrial membrane potential and ATP levels in dissociated neurons of 3-month-old APP tg mice. Interestingly, the mitochondrial membrane potential in APP tg mice was significantly decreased under basal conditions compared with littermate non-tg control mice (Fig. 8). Additionally, dissociated neurons of tg APP mice showed reduced basal ATP levels (Fig. 8). DISCUSSION AD is associated with multiple lesions including changes in energy metabolism and altered mitochondrial structure and/or function in the brain. However, the precise mechanisms of mitochondrial pathology in AD are not clear.
Our study clearly demonstrates that A␤ induces mitochondrial adaptation and failure in a very vulnerable and dose-dependent pattern, which additionally overreacts to secondary insult, and that NO plays an important role in these processes. Several other studies already indicate that A␤ impairs mitochondrial function (18,20,65). However, nearly all of the studies used synthetic A␤ peptides in the micromolar range, many orders of magnitude over physiological levels, and cells or even isolated mitochondria were exposed only to extracellular A␤. By contrast, our PC12 cell model represents a very valuable approach to investigate AD-specific cell death pathways by studying A␤ levels within the high picomolar range (7). This cell model attempts to mimic physiological conditions in man that are relevant for sporadic AD patients (66,67). The comparison between transfected PC12 cells and HEK cells that exhibit A␤ levels within the low nanomolar range further offered the possibility to compare effects of A␤ on mitochondrial  function in a dose-dependent way. As a result, the latter cell model may reflect increased cellular stress evoked by high pathological levels of A␤, a situation that might be present in familial AD patients (66). Moreover, these two cell models allowed us to study not only effects induced by extracellular A␤ but also the effects of intracellularly generated A␤ and/or other products of the APP-processing pathway, e.g. APP intracellular domain, on mitochondrial function. This is an important advantage over extracellular A␤ treatment, because strong evidence indicates the important role of the intracellular biology of A␤ in the pathogenesis of AD. Our data suggested that NO plays an important role in A␤-induced mitochondrial dysfunction and cell death and that intracellularly produced A␤ is specifically relevant. The mitochondrial respiratory chain is sensitive to both NO-and peroxynitrite-mediated damage (30 -32). In accordance with these findings, APPsw PC12 cells showed reduced cytochrome c oxidase activity that is probably due to a direct inhibitory effect of NO, whereas the complexes II, III, and V in APPsw cells seemed to be impaired by peroxynitrite. Importantly, APPwt PC12 cells also showed an impairment of cytochrome c oxidase and complex V. Thus, these two complexes are especially susceptible to already low A␤ levels. Recently, it has been shown that APP and A␤ are targeted to mitochondria (63,68). In accordance, we observed the presence of APP in the mitochondria of APPwt and APPsw PC12 cells. This presence of APP in mitochondria might be involved in the impairment of the mi- tochondrial respiratory chain. Thus, mitochondrial dysfunction can be induced on the one hand via enhanced NO levels and on the other hand because of a direct effect of A␤ on mitochondria (69), although we can not exclude additional effects of APP itself or other intracellular APP-processing products such as APP intracellular domain on mitochondria.
The reduced cytochrome c oxidase activity in APPsw PC12 cells is consistent with the observation that basal ATP levels of APPsw PC12 cells were significantly reduced compared with control cells. Interestingly, APPwt PC12 cells also show reduced ATP levels but to a lesser extent than APPsw PC12 cells. We suggest that the very low A␤ levels in APPwt PC12 cells are sufficient to reduce ATP levels but that this depletion of ATP does not induce cell death per se, indicating that APPsw PC12 cells are able to compensate for this deficit under normal conditions. Only after treatment with a secondary insult (in this case oxidative stress), APPsw PC12 cells were no longer able to maintain cellular function and cell death was increased. Afterward, extracellular A␤ treatment showed only weak effects on ATP reduction, suggesting again that A␤ exhibits its effects on mitochondrial function mostly by intracellular effects. In addition, the ATP reduction in APPsw HEK cells was stronger than in APPsw PC12 cells, showing a dose-dependent effect of A␤ on ATP levels. To investigate whether APP or A␤ is mainly responsible for the mitochondrial damage, the intracellular A␤ production was inhibited by the ␥-secretase inhibitor DAPT, which is known to reduce intracellular A␤ levels without affecting APP expression (59 -61). Our results showed that the DAPT-mediated inhibition of A␤ production lead to a stabilization of mitochondrial function with regard to both a pathogenic hyperpolarization and depolarization of mitochondrial membrane potential in APPsw PC12 cells and APPsw HEK cells, respectively, and to a decrease in NO levels in APPsw PC12 as well as in HEK cells. This finding clearly indicated that increased NO production and mitochondrial dysfunction resulted from increased A␤ levels in APP-transfected PC12 and HEK cells. Our model further strengthened the thesis that A␤ mainly mediates the toxic intracellular effects. However, mitochondrial function did not recover completely to normal levels. This might be due to irreparable damage of the mitochondria by long term expression of A␤ or to APP still persisting in the mitochondrion that may directly interfere with mitochondrial respiratory chain. Very importantly, in 3-month-old APP tg mice, reduced ATP levels can also be observed under basal conditions showing the in vivo relevance of our findings. Of note, the 3 month-old mice exhibited no plaques but already detectable A␤ levels in the brain (39), probably emphasizing again the very important role of intracellular A␤ in ATP reduction. In 12-month-old amyloid plaque-bearing transgenic mice bearing the Swedish APP mutation, reduced cytochrome c oxidase activity and reduced ATP levels were also found (63). Thus, A␤ impairs the energy metabolism of mitochondria in different AD models, possibly as a very early event in the pathogenesis of the disease, as has been shown for AD patients (70).
We could further exclude the possibility that the reduction of ATP levels is due to a reduced number of mitochondria in APPsw PC12 cells, because APPsw transfection rather increased the amount of mitochondria, possibly due to an adaptation process, in response to the increase in NO and reduction of ATP, respectively. This is in accordance with other findings demonstrating a stimulation of the mitochondrial biogenesis by NO (71). Nevertheless, the increased number of mitochondria was not sufficient to compensate the reduced ATP levels in APPsw PC12 cells.
Based on our data and the knowledge regarding the effects of NO on mitochondrial function (30), we proposed a hypothetical sequence of events linking A␤, NO production, ATP depletion, and mitochondrial membrane potential with caspase pathway and neuronal cell loss.
Under physiological conditions, the mitochondrial membrane potential was estimated at Ϫ150 to Ϫ180 mV with respect to the cytosol. Following acute stimulation with extracellular A␤, we and other groups (72) found a slight decrease in mitochondrial membrane potential at rather high non-physiological A␤ concentrations. This might have resulted from an inhibition of respiratory chain complexes by increased NO levels leading to a drop in the electron transport and consequently in proton extrusion.
In contrast, APPsw-transfected PC12 cells, which chronically secrete A␤ levels in the picomolar range, showed activation of a defense response in an attempt to maintain cellular function and viability. After an initial decrease of mitochondrial membrane potential, chronic inhibition of the respiratory chain by increased NO levels resulted in hyperpolarization of mitochondrial membrane potential, probably due to an increase in the rate of glycolysis followed by the entering of glycolytic ATP into the mitochondria (73)(74)(75). Consistent with this finding, we observed no increased basal apoptosis in APPsw PC12 cells. Interestingly, APPsw HEK cells, which express chronically high A␤ levels, have no possibility to adapt. This is probably due to the prolonged increase in the generation of superoxide anions in the presence of high A␤ levels resulting in the formation of peroxynitrite and thereby facilitating the opening of the permeability transition pore (76,77). This finally leads to the collapse of mitochondrial membrane potential and the release of mitochondrial proteins including cytochrome c and Smac that in turn activates caspase cascade and cell death. Consistently, we found increased basal apoptotic levels in APPsw HEK cells. A similar situation exists in the brains of APP tg mice at an age of 3 months, which have detectable high A␤ levels in the brain but before the onset of plaque formation. The brain cells of these mice showed a significantly decreased mitochondrial membrane potential. Thus, we have distinct effects of A␤ on mitochondrial function dependent on acute or chronic exposure and on low and high concentration levels of the peptide. These different effects might play a role in sporadic and familial forms of the disease.
However, when APPsw PC12 cells were stressed with an AD-relevant secondary insult, e.g. oxidative stress, the cells were no longer able to compensate for mitochondrial dyshomeostasis and the mitochondrial membrane potential significantly decreased. In addition, the ATP reduction after oxidative stress was stronger in APPsw PC12 cells than in control cells. Thus, the increased A␤ production at physiological levels enhanced the vulnerability against mitochondrial dysfunction after oxidative stress in APPsw PC12 cells, a process that might be also relevant in sporadic AD. As we could show, a shift in the Bcl-x L /Bax ratio toward Bax might be involved in the higher sensitivity of APPsw cells to mitochondrial failure after oxidative stress. Secondary stress insult finally lead to the release of mitochondrial proteins, e.g. Smac and AIF, which play a role within the cell death cascade (78,79). The increased Smac release in APPsw cells might explain the higher caspase 9 and caspase 3 activity of APPsw PC12 cells as previously demonstrated (9). In contrast, AIF seems to play only an important role in the late phase of apoptosis in APPsw PC12 cells.
Thus, one can speculate that, also in humans, increased A␤ accumulation and associated mitochondrial toxicity are contributing factors in the pathogenesis of AD syndromes. At the beginning, the damaging effects of low physiological concentrations of A␤ could be overcome by an adaptive response. However, when age-related secondary stress occurs, pronounced mitochondrial impairment might lead to the induction of cell death processes, whereas in familial AD, high A␤ load might be directly responsible for mitochondrial and cellular dysfunction. In summary, we showed novel and distinct actions of A␤ on mitochondria in vitro and in vivo that may contribute to the pathogenic outcome.