Swedish Alzheimer Mutation Induces Mitochondrial Dysfunction Mediated by HSP60 Mislocalization of Amyloid Precursor Protein (APP) and Beta-Amyloid*

Background: Alzheimer is associated with mitochondrial dysfunction, yet the mechanism leading to APP and beta-amyloid accumulation is unknown. Results: Beta-amyloid and γ-secretase components accumulate in mitochondria via HSP60-mediated interactions. Conclusion: HSP60 mediates accumulation of APP and beta-amyloid in the mitochondria of Alzheimer transgenic and human brains. Significance: This study identifies a molecular player that translocates beta-amyloid and APP to mitochondria, contributing to its dysfunction. Alzheimer disease (AD) is a complex disorder that involves numerous cellular and subcellular alterations including impairments in mitochondrial homeostasis. To better understand the role of mitochondrial dysfunction in the pathogenesis of AD, we analyzed brains from clinically well-characterized human subjects and from the 3xTg-AD mouse model of AD. We find Aβ and critical components of the γ-secretase complex, presenilin-1, -2, and nicastrin, accumulate in the mitochondria. We used a proteomics approach to identify binding partners and show that heat shock protein 60 (HSP60), a molecular chaperone localized to mitochondria and the plasma membrane, specifically associates with APP. We next generated stable neural cell lines expressing human wild-type or Swedish APP, and provide corroborating in vitro evidence that HSP60 mediates translocation of APP to the mitochondria. Viral-mediated shRNA knockdown of HSP60 attenuates APP and Aβ mislocalization to the mitochondria. Our findings identify a novel interaction between APP and HSP60, which accounts for its translocation to the mitochondria.


Alzheimer disease (AD) is a complex disorder that involves numerous cellular and subcellular alterations including impairments in mitochondrial homeostasis.
To better understand the role of mitochondrial dysfunction in the pathogenesis of AD, we analyzed brains from clinically well-characterized human subjects and from the 3xTg-AD mouse model of AD. We find A␤ and critical components of the ␥-secretase complex, presenilin-1, -2, and nicastrin, accumulate in the mitochondria. We used a proteomics approach to identify binding partners and show that heat shock protein 60 (HSP60), a molecular chaperone localized to mitochondria and the plasma membrane, specifically associates with APP. We next generated stable neural cell lines expressing human wild-type or Swedish APP, and provide corroborating in vitro evidence that HSP60 mediates translocation of APP to the mitochondria. Viral-mediated shRNA knockdown of HSP60 attenuates APP and A␤ mislocalization to the mitochondria. Our findings identify a novel interaction between APP and HSP60, which accounts for its translocation to the mitochondria.
The etiology of Alzheimer disease (AD) 3 is not fully understood, though experimental evidence suggests mitochondrial impairment induced by enhanced reactive oxygen species (ROS) occurs early in the disease pathogenesis (1)(2)(3). Supporting mitochondria are a prominent organelle affected in AD are the studies in transgenic mouse models of AD that show altered mitochondrial bioenergetics arise prior to plaque development (4,5). Amyloid beta (A␤), the main component of plaques, accumulates in mitochondria collected from brains of human AD cases and transgenic mouse models of AD (6 -8). Multiple studies investigating the role of A␤ in AD have directly linked A␤ to mitochondrial dysfunction by several possible mechanisms that include affecting cytochrome c oxidase of the respiratory chain, mitochondria membrane potential, ATP production, and fusion/fission interactions (9 -11). Corroborating studies in cells lacking mitochondrial DNA have revealed A␤-induced toxicity is dependent on a functional respiratory chain (12). The molecular processes that contribute to APP and A␤ mislocalization to the mitochondria in AD are undefined.
APP contains a mitochondrial target sequence, although its physiological function in the mitochondria remains unclear (13). APP is imported into this organelle through binding to the transporter outer membrane 40 (TOM40) and transporter inner membrane 23 (TIM23) mitochondrial import proteins, but complete translocation of APP to the mitochondria may be impeded by its acidic C-terminal sequence (13). To determine the effects of APP translocation to the mitochondria and its implication on mitochondrial function, previous investigation revealed that the ␥-secretase complex that cleaves APP to generate A␤ was localized to the mitochondria (14). The mechanism responsible for APP and A␤ mitochondrial localization is undetermined; however, a recent study showed that rescuing activation of mitochondrial permeability pore complex via cyclopilin D deficiency attenuates A␤-induced mitochondrial dysfunction and cognitive deficits. Consequently, these studies reveal mitochondria are a prominent organelle affected by A␤ (15).
Converging evidence from Parkinson disease (PD) and other neurological disorders show molecular chaperones play an integral role in neurodegeneration (16). Corroborating evidence implicates heat shock proteins (HSPs) in metabolism and the aggregation of both A␤ and tau (17,18). Mitochondrial associated HSPs, mtHSP70/Grp75/Mortalin, HSP60, and HSP10, are predominantly found in the mitochondria matrix and play an important role in protein folding. Subcellular studies implicate HSP60 and mtHSP70 in many cellular functions by showing these chaperones localize to the plasma membrane, ER, and endocytic vesicles (19). Alterations in HSPs function cause pleiotropic effects on mitochondria function, such as protein aggregation, and import deficits leading to morphological changes (20). These studies connect HSPs in numerous cellular processes important for maintaining mitochondrial integrity.
In the present study, we sought to identify the molecular processes that contribute to APP and A␤ mislocalization to the mitochondria, and ultimately its dysfunction. Here we show 3xTg-AD mitochondria exhibit decreases in cox IV activity and respiratory capacity accompanied with increased A␤ and ␥-secretase components. To identify the molecular determinants involved in APP and A␤ mislocalization to the mitochondria in the AD brain, we used a proteomics approach to identify APP and A␤ binding partners. We find HSP60 and APP/A␤ form a strong molecular association in mitochondria harvested from both transgenic and human AD subjects. Using a viral transduction strategy, we generated stable neural cell lines expressing human wild-type or Swedish APP, and provide substantiating in vitro evidence that HSP60 mediates translocation of APP to the mitochondria. Knockdown of HSP60 via an shRNA-based approach abrogated APP and A␤ localization to the mitochondria in Swedish APP-transduced cells. Our studies identify a novel interaction between APP and HSP60 that is important for its translocation to the mitochondria in AD.

EXPERIMENTAL PROCEDURES
3xTg-AD mice were previously characterized and maintained on a hybrid C57BL6/129 background (21). Female 12-month-old homozygous 3xTg-AD and age-matched hybrid controls were used for the purposes of these studies. All procedures were performed in accordance with the National Institutes of Health and University of California guidelines.
Human Tissue-Brain tissue collected by the University of California, Irvine Alzheimer Disease Research Center (ADRC) is tested for pathological hallmarks of AD. Human AD brains (n ϭ 8) were processed as described below and compared with APOE allele-, age-, and sex-matched non-demented controls (n ϭ 8). The average post mortem index for non-demented controls was 5.8 and 5.2 for AD specimens.
C17.2 Cells-Generation of these cells was described previously (22). C17.2 cells were passaged every other day by trypsinization and adding a tenth of the cells to a new flask.
Mitochondrial Fractionation-Mouse brains were submerged in 6 ml of mitochondrial homogenization buffer (H-Buffer) (225 mM mannitol, 75 mM sucrose, 10 mM MOPS, 1 mM EGTA, 0.5% BSA, pH 7.2, at 4°C) followed by dounce homogenization on ice. Differential centrifugation was performed on homogenates followed by iodoxinal density gradient subcellular fractionation as performed previously (23). Collected fractions were washed in H-buffer and lysed in TBS with 2% CHAPs. Western blot analysis of GM130 (Golgi), calnexin (ER), GAPDH (cytosolic), and Complex IV subunit IV were assessed to determine the purity of each mitochondria preparation.
Complex Activity Assay-Complex IV activity was assessed from isolated mitochondrial preparations as described previously (2).
Respiration Assay-Respiration rates were determined by using the Clark-type electrode by Hansatech Oxygraph (Norfolk, England) instruments. Freshly prepared mitochondria, 500 g of protein, was used to measure State 3 and 4 oxygen consumption rates in the presence of complex I substrate glutamate and malate. FCCP (p-trifluor-omethoxy carbonyl cyanide phenyl hydrazone) was used to measure uncoupling rate for the maximum respiration rate and oligomycin to inhibit Complex V.
DNP Derivative Assay-Oxyblot/DNP derivative assay (Millipore, Billerica, MA) was performed in accordance to the manufacturer's protocol.
ATP Assay-ATP levels were assessed using the manufacturer's protocol for the ATP lite activity assay (Perkin Elmer, San Jose, CA).
Seahorse XF-24 Extracellular Flux Analyzer-According to the manufacturer's protocol, four groups (APP WT , APP SWE , APP WT hsp60 Ϫ/Ϫ , APP SWE hsp60 Ϫ/Ϫ ) of C17.2 cells plated in Seahorse XF-24 plates at a density of 6 ϫ 10 4 cells/well, with 3 replicates/group were incubated overnight. After initial baseline measurements of OCR and ECAR, oligomycin (1 M), p-trifluoromethoxy carbonyl cyanide phenyl hydrazone (FCCP at 1 M) and Rotenone (1 M)ϩ antimycine(1 M) were sequentially injected to each wells for further OCR and ECAR determination. Total protein was determined from each well via Bio-Rad Bradford assay followed by protein normalization.
ELISA-ELISA analysis was performed using beta-amyloid ELISA kits from WAKO (Wako catalogue #s 298-62401 and 294-62501). ELISA was performed on 100 g of each mitochondrial lysates.
Immunoprecipitation-Immunoprecipitation was performed using Dyna Beads (Invitrogen, Eugene, OR) as described by the manufacturer's protocol. Dyna beads were bound to 4G8, 6E10, or Hsp60. Antibody bound to dyna beads was incubated overnight with rotation in 400 g of mitochondrial lysate or 800 g of cytosolic lysate. Eluted proteins were then loaded on to a 4 -12% gel followed by electrophoresis.
In-gel Tryptic Digest-Protein bands were excised and washed with 1:1 acetonitrile (ACN) and 50 mM ammonium bicarbonate (AMBC) and vortexed and incubated at 50°C for 10 min for a total of three times. Gel pieces were dehydrated for 5 min with ACN and reduced with 10 mM DTT for 30 min. DTT was removed and iodoacetamide (100 mM) in AMBC and incu-HSP60-mediated Mislocalization of APP to the Mitochondria bated at room temperature for 30 min. Gel pieces were dehydrated with ACN for 5 min, removed, and 1ϫ trypsin solution (Promega, Madison, WI) was added to the gel pieces. The following day, the supernatant was removed and 30 l of extraction buffer (50% ACN and 5% formic acid in proteomic grade water) was added and removed every 10 min and placed into a new 1.5 ml tube. The concentrated peptides were desalted with zip tips (Millipore, Billerica, MA) and eluted in 10 l of 50% (w/v) acetonitrile/0.1% trifluoroacetic acid.
Mass Spectrometry-To identify tryptic peptides matrix assisted laser desorption/ionization-time of flight (MALDI/ TOF) mass spectrometry was performed. MS analysis was performed using a 5800 Plus MALDI TOF/TOF (Applied Biosystems, Framingham, MA) at the University of California, Irvine mass spectrometry core. The MS/MS spectra were analyzed using the Paragon algorithm (Protein Pilot software, Applied Biosystems, Carlsbad, CA) against the UniProt SP plus contaminant protein database. The search parameters were adjusted for cystein alkylation with idoacetamide. Additional parameters included a confidence interval Ն 95% that was used for protein identification (Unused ProtScore Ͼ 1.3).
A␤ Preparation-A␤42 oligomers were prepared as described previously (27). Dot blot analysis confirmed A␤42 oligomers, which tested negative for OC (fibril antibody) and positive for A11 (pre-fibrillar oligomers antibody).
Statistics-For experiments involving quantification, the standard error of the mean was determined from at least three independent experiments, with an "n" of one representing one transgenic mouse accompanied by one wild-type littermate control. Effects of genotype were analyzed for significance using two-way ANOVA, followed by Bonferroni's test for all pair-wise comparisons. In all cases, a p value of Յ 0.05 was considered significant.

3xTg-AD Mice Exhibit Mitochondrial Deficits-Dysregu-
lated mitochondrial function has been previously reported in the AD brain and various AD transgenic mouse models (4,28). To elucidate the mechanism for mitochondrial dysfunction in AD, we used 12-month-old 3xTg-AD that have both A␤ and tau pathology (29). Complex IV activity, a critical component of the mitochondrial respiratory chain, was detected in either 3xTg-AD or non-transgenic (nTg) mitochondria by measuring the rate of oxidation of reduced cytochrome c. Complex IV activity was decreased in 3xTg-AD mitochondria compared with age-matched nTg (Fig. 1A). To test whether oxidative phosphorylation was impaired in 3xTg-AD mice, mitochondrial respiration was assessed from freshly isolated mitochondria from the forebrain. The respiratory rate was determined by adding mitochondrial substrates glutamate and malate (state 2), ADP (state 3), followed by oligomycin addition to inhibit complex IV (state 4). 3xTg-AD mitochondria showed decreases FIGURE 1. 3xTg-AD mitochondria exhibit deficits in oxidative phosphorylation and oxidative stress. A, Cox IV activity was assessed by measuring the rate of oxidation of reduced cytochrome c. Cox IV was significantly decreased in 12-month-old 3xTg-AD compared with age-matched nTg (n ϭ 15 for each time point from 3xTg-AD and nTg) (B) State 3 and state 4 respiration was determined by using mitochondria from cortical and hippocampal regions followed by oxygen electrode measurement in the presence of L-malate, L-glutamate, and ADP to initiate state 3 respiration and atractyloside to measure state 4. 3xTg-AD mitochondria exhibit a significant decrease in state 3 and 4 respiration compared with nTg (n ϭ 10) (C) 5 g of protein was subjected to the DNP conversion followed by Western blot analysis. DNP detection was significantly increased in 12-month-old 3xTg-AD mitochondria compared with age-matched controls. (n ϭ 6 for each time point from 3xTg-AD and nTg). For all data points, *, p Ͻ 0.05 3xTg-AD compared with nTg controls. AUGUST 31, 2012 • VOLUME 287 • NUMBER 36 in respiration at both state 3 and 4 compared with nTg controls (Fig. 1B). Mitochondrial dysfunction is associated with both oxidative stress and AD related neuropathology. Consequently, to detect oxidative stress we employed the oxyblot detection system that converts oxidized side chains to 2,4-dinitrophenylhydrazone (DNP), which can be detected by Western blot. Mitochondria from 3xTg-AD revealed a significant increase in oxidative stress compared with nTg controls, as evidenced by increased DNP levels (Fig. 1C).

HSP60-mediated Mislocalization of APP to the Mitochondria
Increased A␤ and ␥-Secretase Components in 3xTg-AD Mitochondria-To determine whether mitochondrial dysfunction was associated with increased A␤ localization to the mitochondria, differential centrifugation followed by density gradient fractionation was performed to obtain pure mitochondrial fractions (23). Biochemical analysis revealed mitochondrial fractions tested positive for mitochondrial proteins Complex IV subunit IV and VDAC, but negative for other organelle markers (e.g. calnexin and GM130) ( Fig. 2C and data not shown). Western blot analysis on mitochondrial fractions showed no significant changes in APP levels in the 3xTg-AD mitochondria compared with non-transgenic controls, however, quantitative analysis by ELISA revealed A␤ 40 and A␤ 42 levels were significantly increased in 3xTg-AD mitochondria fractions at 6 -24 months of age compared with non-pathological 3 month-old animals ( Fig. 2A and data not shown). A recent study suggested 3xTg-AD mice have increased intraneuronal APP and not A␤ (30). To confirm our ELISA results 6E10 immunoprecipitation was performed on 12-month 3xTg-AD pooled cytosolic (400 g) or mitochondrial containing fractions (800 g). Western blot analysis to detect the A␤ monomer by using the 6E10 biotin conjugated antibody revealed increased A␤ monomer (4 kDa) in both 3xTg-AD cytosolic and mitochondrial containing fractions (Fig. 2B). Previous studies have shown that increased mitochondrial ␥-secretase activity promotes A␤ generation by cleaving APP (14), which prompted us to investigate whether ␥-secretase components were increased in the 3xTg-AD mitochondria. Corroborating with prior studies in AD transgenic mouse models, our Western blot results showed ␥-secretase components presenilin-1, -2, and nicastrin were increased in 3xTg-AD mitochondria versus nTg controls (Fig. 2C), thus prompting further studies to investigate the mechanism responsible for A␤ accumulation in 3xTg-AD mitochondria.
APP/A␤ Species Interact with Molecular Chaperone HSP60-A proteomics approach to identify the mechanism responsible for APP and A␤ mislocalization to the mitochondria was performed on cytosolic and mitochondrial fractions to identify APP and A␤ binding proteins. Cytosolic and mitochondrial fractions were pooled from either 3xTg-AD or non-transgenic fractions and subjected to 4G8 (detects endogenous and exogenous APP/A␤) immunoprecipitation, which showed many alterations between APP-binding proteins in 3xTg-AD com-FIGURE 2. 3xTg-AD mitochondria exhibit increased A␤ and ␥-secretase components. A, A␤40 (BA27) and A␤42 (BC05) levels were examined via ELISA in 3xTg-AD mitochondrial fractions. A␤40 and A␤42 levels are significantly increased by 6 months of age and persist to 24 months compared with 3-month-old 3xTg-AD mitochondria. B, 6E10 was immunoprecipitated from either four pooled (100 g each) 12-month 3xTg-AD cytosolic fractions or eight pooled mitochondrial containing fractions (MCF) followed by 6E10 -biotin immunoblot. Short, medium, and long exposure revealed increased A␤ monomer (4 kDa) in both 3ϫ cytosolic and MCF fractions. Western blot analysis also showed an increase in an A␤ trimer at 12 kDa. C, Western blot analysis for PS1, PS2, nicastrin, VDAC, Cox IV, GAPDH, and calnexin was performed on 12-month-old 3xTg-AD and nTg mitochondrial fractions. represents the use of a ladder in the well. Results show increases in ␥-secretase components compared with non-transgenic controls (n ϭ 8). *, p Ͻ 0.01 by one-way ANOVA, 3xTg-AD ␥-secretase protein levels versus nTg.

HSP60-mediated Mislocalization of APP to the Mitochondria
pared with non-transgenic that were analyzed by mass spectrometry (MS) (Fig. 3A). To identify potential proteins that could assist in APP/A␤ mitochondria translocation, gel bands were excised followed by in-gel tryptic digestion and MS. Of the proteins identified by matrix assisted laser desorption ionization time of flight (MALDI-TOF) MS/MS, HSP60 underwent further investigation based on previous results implicating it as an important mitochondrial chaperone. To confirm our MS results, Co-IP analysis was performed on pooled cytosolic and mitochondria fractions from either non-transgenic or 3x-Tg-AD mice. APP/A␤ (4G8) immunoprecipitation followed by HSP60 immunoblot revealed APP and/or A␤ binding was increased in the 3xTg-AD brain and virtually undetectable in non-transgenic animals (Fig. 3B). It is noteworthy that HSP60 undergoes post-translational modification upon mitochondrial import, including the removal of 27 amino acids (31). In addition, mtHSP70, another mitochondrial chaperone interaction was assessed, however the interaction was not as pronounced as the APP/HSP60 interaction (Fig. 3D). To verify the interaction, immunoprecipitation of HSP60 followed by 4G8 detection was performed. Western blot analysis revealed increased APP binding to HSP60 in the 3xTg-AD cytosolic and mitochondrial fractions compared with non-transgenic controls (Fig. 3C). These results reveal increased interactions between mitochondrial chaperone HSP60 and APP in 3xTg-AD.
HSP60 and APP Interaction Exists in Human AD-Tissue collected by the University of California, Irvine Alzheimer Disease Research Center (ADRC) undergoes rigorous cognitive and pathological analysis to validate AD versus non-demented specimens. To determine whether A␤ accumulates in mito- A band approximately at 60 kDa was excised and processed for mass spectrometry. The band was identified as HSP60, which is represented by the MS/MS spectra from a peptide sequence with over a 99% confidence level. The HSP60 peptide sequence with peptides identified with high (green text) to moderate (orange text) confidence. Additional proteins identified, GAPDH, dihydropyrimidenase-2, isocitrate dehydrogenase (spot 1), heat shock protein 60 (2), isocitrate dehydrogenase (3), GAPDH (4), myelin basic protein isoform (5), and tubulin (6). B, co-IP experimentation by pulling down APP/A␤ (4G8) and blotting for HSP60 revealed increased APP/A␤ binding to HSP60 in both cytosolic and mitochondrial fractions pooled from 4 3xTg-AD age matched samples compared with 4-pooled non-transgenic samples. 4G8 immunoblot showed equal levels of APP/A␤ were pulled down. C, HSP60 IP followed by 4G8 immunoblot shows increased APP binding to HSP60 in 12-month 3xTg-AD cytosolic fractions compared with nTg controls. HSP60 immunoblot shows equal amounts of HSP60 were immunoprecipitated. D, co-IP experimentation by pulling down APP/A␤ (4G8) and blotting for mtHSP70 revealed a slight increase in APP/A␤ binding to mtHSP70 in the cytosolic, but not mitochondrial fraction pooled from 4-pooled 3xTg-AD age-matched samples compared with four pooled non-transgenic samples. 4G8 immunoblot showed equal levels of APP/A␤ were pulled down. AUGUST 31, 2012 • VOLUME 287 • NUMBER 36 chondria from AD subjects versus normal, non-demented controls, mitochondrial fractions were prepared followed by biochemical analysis. Our ELISA results showed significant increases in both A␤40 and A␤42 in both the AD cytosolic and mitochondrial fractions compared with non-demented controls (Fig. 4A). To test whether HSP60 interacted with APP and/or A␤ occurred in the AD brain, pooled cytosolic or mitochondrial fractions from either AD or non-demented controls were subjected to immunoprecipitation with 6E10 (human specific antibody for APP/A␤) followed by HSP60 immunoblotting. We found AD mitochondria exhibit increased HSP60 binding to APP/A␤ compared with non-demented controls (Fig. 4B). HSP60 was undetectable in the cytosolic fraction pull down (Fig. 4B). To verify these results, immunoprecipitation of HSP60 was performed from 8 pooled human mitochondrial containing fractions followed by Western blotting with the 6E10 antibody conjugated to biotin to avoid detection of heavy and light chains. Our results revealed increases in HSP60 binding to APP in the AD mitochondria in contrast to non-demented controls (Fig. 4C).

HSP60-mediated Mislocalization of APP to the Mitochondria
Swedish APP Leads to A␤-induced Mitochondrial Dysfunction-To better understand the molecular process associated with A␤ intracellular accumulation, a neural cell line (C17.2) was transduced with lentiviruses containing either wild-type human APP (APP WT ) or familial AD Swedish mutation (APP SWE ) that causes BACE cleavage, thus generating more A␤. To delineate the influence of the Swedish mutation on APP and A␤ mitochondrial translocation, cytosolic lysates, whole cell lysates, and mitochondrial-containing fractions were prepared from either APP WT or APP SWE cells. Western blot analysis on either cytosolic or whole cell lysates showed no significant increases in total APP protein levels between APP WT and APP SWE (Fig. 5A and data not shown). Mitochondrial fractions from the APP SWE mutation showed a significant increase in APP and its derivatives in comparison to APP WT (Fig. 5A). To verify the purity of the fractions, VDAC and Complex IV Subunit IV were used to quantify mitochondrial fractions, while ␤-tubulin was used as a cytosolic loading control. To determine whether oxidative phosphorylation was compromised in APP SWE versus APP WT , an ATP luciferase activity assay was performed on both cell lines and revealed APP SWE cells produced significantly less ATP production compared with APP WT cells (Fig. 5B). To further assess mitochondrial respiration, APP WT and APP SWE lines were treated with antimycin (inhibits flow of electrons from cytochrome b to cytochrome c 1 , complex III), oligomycin (inhibits ATP synthase by blocking its proton channel (Fo subunit), and carbonyl cyanide m-chloro phenyl hydrazone (CCCP, uncoupler). ATP luciferase activity showed APP SWE cells were significantly more susceptible to mitochondrial inhibitors and uncouplers versus APP WT cells. APP SWE cells also exhibited a slight increase in oxidative stress compared with APP WT cells as evidenced by the DNP deriva- . The interaction between APP and HSP60 exists in the human brain. A, ELISA analysis showed increased A␤40/42 in AD brains compared with normal controls, n ϭ 4 for both AD and non-demented controls. *, p Ͻ 0.001 AD mitochondria compared with Normal controls. B, 6E10 was immunoprecipitated from either pooled cytosolic (n ϭ 8) or mitochondrial fractions (n ϭ 8) followed by HSP60 immunoblot. The HSP60 interaction between APP/A␤ was slightly detectable in AD pooled mitochondrial fractions compared with NDC controls. The arrow represents the anticipated molecular weight of HSP60 in the cytosol. C, to confirm the interaction between APP/A␤ and HSP60, HSP60 immunoprecipitation was performed on 8 pooled NDC or AD mitochondrial containing fractions (800 g total) followed by 6E10 biotin Western blot. APP doublet (ϳ75 kDa) binding to HSPs was observed in AD mitochondria compared with NDC. D, 4G8 was immunoprecipitated from either pooled cytosolic (n ϭ 4) or mitochondrial fractions (n ϭ 8) followed by mtHsp70 immunoblot. The mtHsp70 interaction between APP and its derivatives was slightly increased in 3xTg samples compared with nTg controls.

HSP60-mediated Mislocalization of APP to the Mitochondria
tive assay (Fig. 5C). To determine whether the APP SWE mutant resulted in increased APP binding to HSP60, immunoprecipitation using 4G8 was performed on pooled mitochondrialcontaining fractions (n ϭ 6 for each group). Western blot analysis revealed APP SWE cells exhibit increased APP binding to HSP60 in contrast to APP WT (Fig. 5D). To verify these results, immunoprecipitation using HSP60 was conducted on pooled mitochondrial containing fractions followed by Western blotting with 4G8 antibody. Our results corroborated our in vivo results showing increased HSP60 binding to APP in the AD mitochondria.
HSP60 Mediates APP and A␤ Localization to the Mitochondria-To determine the role HSP60 plays in APP and A␤ mitochondrial translocation and its effects on mitochondrial homeostasis, lentiviruses containing either HSP60 shRNA or PLKO (empty control vector) were generated in APP WT and APP SWE cell lines. Western blot analysis revealed HSP60 levels were drastically reduced by shRNAs directed toward HSP60 compared with PLKO controls (Fig. 6A). To test the hypothesis that HSP60 is important for APP and A␤ mislocalization to the mitochondria, mitochondrial fractions were prepared from APP WT and APP SWE cells with or without HSP60 knockdown. Biochemical analysis showed the increased A␤ and APP in the Swedish mitochondria fraction was significantly reduced by HSP60 knockdown compared with cells expressing the Swedish mutation and PLKO (Fig. 6A). HSP60 knockdown had no effect on APP and its derivatives in the APP WT mitochondrial fractions. To determine whether A␤ levels were diminished in the HSP60 knockdown lines, ELISA analysis was performed to detect A␤ 40 and A␤ 42 levels. Our results showed that APP SWE cells have significant increases in mitochondrial A␤ 40 and A␤ 42 compared with APP WT , and that HSP60 knockdown attenuated Swedish-induced mitochondrial A␤ accumulation (Fig.  6B).
To assess whether HSP60 knockdown could revert diminished ATP levels back to APP WT levels, an ATP luciferase assay was performed on APP WT and APP SWE with or without HSP60. Our results showed HSP60 knockdown did not ameliorate the diminished ATP levels observed in APP SWE cells (Fig. 7A). HSP60 knockdown decreased ATP production in WT APP cells, corroborating with previous findings the importance of HSP60 in mitochondrial bioenergetics. To further assess the impact of HSP60 knockdown on mitochondrial function, basal cellular respiration and glycolysis were determined in APP WT and APP SWE with or without HSP60 by using the Seahorse XF-24 metabolic flux analyzer. Oxygen consumption rates (OCR) and extracellular acidification rates (ECAR) were calculated. Basal respiration was established followed by the addition of 1 m oligomcyin (vertical line A), which showed a decline in OCR by ϳ50%, revealing oxygen consumption in these cells utilizes FIGURE 5. In vitro FAD mutation promotes APP/A␤ mitochondrial accumulation and mitochondrial dysfunction. A, mitochondrial containing and cytosolic fractions were prepared and blotted for mitochondrial markers Cox IV and VDAC, while ␤-tubulin was used to verify cytosolic fractions. Western blot analysis for human APP (6E10) revealed a significant increase in APP/A␤ derivatives in SWE (S) mitochondrial fractions compared with WT APP (W). B, ATP levels were assessed with an ATP luciferase assay in both W and S cell lines. ATP production was significantly decreased in the SWE cell line compared with WT. In addition, SWE cells were more sensitive to mitochondrial uncoupler (FCCP), complex III inhibitor (antimycin), and the ATP synthase inhibitor oligomycin compared with WT. C, 5 g of protein from either WT or SWE whole cell lysates and MCF fractions were subjected to the DNP derivative assay followed by Western blotting. SWE cells exhibited increased DNP activity compared with WT cells (n ϭ 8). All data points, *, p Ͻ 0.05 SWE cells compared with WT APP cells. D, co-IP experimentation by pulling down APP/A␤ (4G8) and blotting for HSP60 showed increases in APP/A␤ binding to HSP60 in APP SWE mitochondrialcontaining fractions (MCF) compared with APP WT (n ϭ 6). HSP60 IP followed by 4G8 immunoblot shows increased APP binding to HSP60 in APP SWE cells compared with APP WT . 4G8 and HSP60 immunoblots show equal amounts of HSP60 and 4G8 were immunoprecipitated for each experiment.
both oxidative phosphorylation and glycolysis equally (Fig. 7, B and C). Additionally APP SWE cells exhibit a higher maximal respiratory capacity rate than APP WT cells, whereas, HSP60 knockdown in both APP WT and APP SWE significantly diminished the maximal respiratory rate (Fig. 7B). We also found the Swedish mutation induced A␤ that diminishes ATP levels causes a compensatory mechanism by increasing maximal respiration (Fig. 7B). Furthermore, to determine whether HSP60 knockdown could attenuate a secondary insult, we treated APP WT and APP SWE cells with or without HSP60 knockdown with A␤ oligomers for 48 h. Viability analysis showed APP SWE cells were more susceptible to A␤ oligomer-induced cell death compared with APP WT , however, HSP60 knockdown cells were just as vulnerable to A␤ oligomer-induced cell death in both APP WT and APP SWE cells (Fig. 7D). In addition to its role in maintaining mitochondrial homeostasis, we show HSP60 also mediates APP and A␤ mislocalization to the mitochondria.

DISCUSSION
In the present study, we sought to identify the molecular processes that contribute to APP and A␤ mislocalization to the mitochondria, and ultimately its dysfunction. Here we demonstrate that mitochondria harvested from 3xTg-AD mice are impaired, as evidenced by decreases in state 3 and 4 respiration, lower cox IV activity, and elevated oxidative stress. Our study further shows that 3xTg-AD mitochondria exhibit increased A␤ and ␥-secretase components, providing a platform to investigate the mechanism responsible for APP and A␤ ectopic localization to the mitochondria. To identify the molecular determinants involved in APP and A␤ localization to the mitochondria in the AD brain, we used a proteomics approach to identify APP and A␤ binding partners. Proteomic and biochemical analyses of 3xTg-AD mitochondria identified a novel interaction between APP and HSP60. We found HSP60 and APP/A␤ form a strong molecular association in mitochondria from both transgenic and human AD subjects. APP immunoprecipitated from human AD mitochondria exhibited a stronger propensity to interact with HSP60 versus non-demented controls. Using a viral transduction approach, we generated stable neural cell lines expressing either human wild-type APP or Swedish APP, and provided substantiating in vitro evidence that HSP60 forms a strong molecular association with APP in FIGURE 6. HSP60 knockdown attenuates SWE-induced APP/A␤ mitochondrial translocation. A, lentiviruses containing HSP60 shRNA (clones 95, 96, and 99) were generated followed by neural cell transduction. Western blot analysis shows dramatic reduction of HSP60 levels in all three clones. ␤-Tubulin was used as a loading control. Mitochondrial fractions were prepared from WT or SWE lines with or without HSP60 shRNA followed by Western blot analysis to assess APP/A␤ levels. Increased APP levels in the SWE expressing cell line were returned to WT levels with HSP60 knockdown. B, ELISA analysis performed on mitochondrial fractions showed that the increased A␤ levels in SWE-expressing cells were returned to baseline with HSP60 knockdown. *, p Ͻ 0.05 by two-way ANOVA/Bonferroni post-test compared with WT. an in vitro model of familial AD. Knockdown of HSP60 abrogated APP and A␤ localization to the mitochondria in Swedish APP-expressing cells. This study identifies a novel interaction between APP and HSP60 that is important for APP and A␤ translocation to the mitochondria.
Biochemical analysis of 3xTg-AD showed A␤ and ␥-secretase components accumulated in mitochondrial fractions, corroborating previous studies that found A␤ and APP localized to the mitochondria leading to its dysfunction. 6E10 immunoprecipitation identified A␤ monomers accumulating in 3xTg-AD cytosol and mitochondria, whereas, a ϳ12 kDa band was also detected in the mitochondrial fraction that was not apparent in the cytosolic fraction. The ϳ12 kDa band likely represents an A␤ trimer or C99, that could potentially disrupt mitochondrial function. A recent study found mitochondria from 5XFAD mice accumulate the C99 fragment, resulting in mitochondrial dysfunction (32); consequently, APP mitochondrial localization takes on added significance, since our findings and others have detected increased mitochondrial A␤ and ␥-secretase components in AD patients (14). To determine the mechanism responsible for APP and A␤ mislocalization to the mitochondria, neural cells expressing human wild-type APP or Swedish APP were generated and provided supporting in vitro evidence that APP and A␤ accumulate in the mitochondrial concomitant with decreased ATP levels and elevated oxidative stress. 3xTg-AD mice harbor the APP SWE mutation, which promotes A␤ 42 generation. This explains why we did not observe significant changes in total APP protein levels between the human FIGURE 7. Mitochondrial homeostasis requires HSP60. A, ATP levels were assessed with an ATP luciferase assay in both WT and SWE cells with or without HSP60 knockdown. ATP production was decreased in the SWE cells compared with WT and both WT-and SWE-HSP60 shRNA-transduced cells. SWE-HSP60 shRNA cells were less sensitive to ATP synthase inhibitor oligomycin, however, HSP60 knockdown had no effect on ATP levels when treated with FCCP, rotenone, and antimycin (RA). wild-type and Swedish APP whole cell lysates. However, we did detect increased APP and A␤ in the mitochondria of APP SWE cells versus APP WT . These results indicate that the familial AD Swedish mutation generates more A␤ 42 and induces mitochondrial dysfunction. Corroborating previous reports that showed APP SWE cells were healthy (33), this was justified by cellular respiration analysis that demonstrated both APP SWE and APP WT cells utilize oxidative phosphorylation and glycolysis equally, thus providing an alternate energy source if oxidative phosphorylation is compromised. Unexpectedly, respiration results revealed that the maximal respiration was increased in APP SWE cells, suggesting APP SWE cells compensate for diminished ATP levels. APP SWE -expressing cells also exhibited elevated, but not significant levels of oxidative stress, most likely due to the ability of pre-mitotic C17.2 cells to dilute out reactive oxygen species with each cell division in contrast to post-mitotic neurons. We did find that APP SWE cells were more susceptible to a secondary insult, such as A␤ oligomers. These results add to previous findings that APP and A␤ ectopic localization to the mitochondria cause its dysfunction.
To determine the molecular process involved in APP and A␤ localization to the mitochondria in 3xTg-AD mice, a proteomic strategy was performed identifying HSP60 as an APP/A␤ binding partner. Immunoprecipitation with either 4G8 or 6E10 showed HSP60 forms a strong interaction with APP in both cytosolic and mitochondrial fractions from 3xTg-AD and cells expressing Swedish APP compared with controls. Although HSP60 immunoprecipitation exhibited only a slight increase in the APP/HSP60 interaction in either fraction, these results imply that the rabbit polyclonal HSP60 antibody nonspecifically binds to additional proteins in the fraction that share sequence homology with APP, such as amyloid-like protein 1 or APLP1 or APLP2. APLP1 and APLP2 are similar to APP except they lack the A␤ domain. APLP2 is a more likely candidate at the mitochondria than APLP1 based on subcellular studies showing APLP2 is predominantly intracellular, whereas APLP1 is associated with the plasma membrane (34). The HSP60/APP interaction is stronger with 4G8 and 6E10 antibodies, because these are monoclonal antibodies that target the A␤ epitope. Immunoprecipitation of HSP60 or APP showed APP complexes with HSP60 in AD mitochondria versus non-demented controls. In contrast to 3xTg-AD, the cytosolic interaction was undetectable in both human AD and non-demented controls, proposing cytosolic HSP60 is rapidly degraded by active proteases, since human brains were not subjected to the same rapid freezing protocol performed on mouse brains. This likely scenario proposes mitochondrial HSP60 is in part protected from degradation, since it is compartmentalized in the mitochondria, although mitochondria also possess proteases, thus making the interaction less robust compared with our in vivo and in vitro results. These findings are significant in that they reveal increased binding of HSP60 to APP in the AD brain.
Combining in vivo and in vitro models of AD, we have demonstrated the interaction between HSP60 and APP is increased in the AD brain. To determine the mechanism responsible for APP and A␤ localization to the mitochondria, neural cells expressing the Swedish APP mutation that produces more A␤ and leads to mitochondria dysfunction were used to discern whether HSP60 mediated APP and A␤ localization to this organelle. Knockdown of HSP60 by shRNA attenuated ectopic localization of APP and A␤ to the mitochondria in Swedish APP expressing cells. Human wild-type APP expressing cells were unaffected by HSP60 knockdown, supporting our hypothesis that A␤ induces HSP60-mediated mislocalization of APP and A␤ to the mitochondria. Chaperone interactions are stimulus specific, thus alterations in cellular homeostasis may free the chaperone to associate with additional substrates. HSP60 has been implicated in possessing both anti-and pro-apoptotic roles (35). Under normal physiological conditions HSP60 forms a pro-survival complex with Bcl-2, an anti-apoptotic mediator, and this complex is hindered by nutrient deprivation resulting in cell death (33). Consequently, increases in A␤ may interact with HSP60 directly or indirectly to release HSP60 from Bcl-2, thus freeing HSP60 to mediate APP and A␤ mislocalization to the mitochondria and promote mitochondrial dysfunction. This hypothesis coincides with previous findings that show HSP60 possess pro-apoptotic abilities by initiating the maturation of procaspase-3 to promote apoptosis (36). To test whether HSP60 knockdown would be protective in response to A␤-induced mitochondrial dysfunction, we treated HSP60 knockdown cells with A␤ oligomers, and found HSP60 knockdown cells offered no significant protection. One such explanation is A␤ oligomers cause Bcl-2 to be sequestered by pro-apoptotic Bcl-2 homology 3 domain (BH3) proteins, which bind and suppress Bcl-2 pro-survival function in the absence of HSP60. Accordingly, Bcl-2 would be more prone to inhibitory mechanisms without HSP60, thus corroborating with previous reports that show HSPs are important for maintaining cellular homeostasis. Earlier studies revealed A␤ binds and inactivates complex IV of the respiratory chain (37); therefore, A␤ and APP binding to HSP60 may disrupt its physiological role at the mitochondria leading to the organelles malfunction. Supporting that HSP60 is important for mitochondrial bioenergetics is our findings that HSP60 knockdown caused a greater deficiency in mitochondrial function as evidenced by decreased ATP and respiration. These results indicate that the familial AD Swedish mutation generates more A␤ 42 and induces mitochondrial dysfunction by promoting HSP60-mediated APP/A␤ mislocalization.
In conclusion, previous reports have identified mitochondrial proteins that promote import of APP and A␤ into the mitochondria; however, very little is known to what mediates APP and A␤ localization to the mitochondria during pathological conditions. Our results provide evidence that A␤ induces HSP60-mediated APP/A␤ mislocalization to the mitochondria, ultimately leading to its dysfunction. This finding is significant based on previous studies implicating APP and A␤ mitochondria accumulation in the buildup of oxidative stress, and eventually leads to mitochondrial dysfunction. In conclusion, our findings provide insight into the mechanism behind APP and A␤ mislocalization to the mitochondria that contributes to mitochondrial dysfunction.