Glyceraldehyde-3-phosphate Dehydrogenase Aggregates Accelerate Amyloid-β Amyloidogenesis in Alzheimer Disease*

Background: There is currently no strong evidence for a linkage between glyceraldehyde-3-phosphate dehydrogenase (GAPDH) and Alzheimer disease (AD). Results: GAPDH aggregates enhanced amyloid-β peptide (Aβ) amyloidogenesis and augmented Aβ40-induced neurotoxicity, both in vitro and in vivo, concomitant with mitochondrial dysfunction. Conclusion: GAPDH aggregates accelerate Aβ amyloidogenesis. Significance: Aβ amyloidogenesis associated with GAPDH aggregation might underlie AD pathogenesis. Alzheimer disease (AD) is a progressive neurodegenerative disorder characterized by loss of neurons and formation of pathological extracellular deposits induced by amyloid-β peptide (Aβ). Numerous studies have established Aβ amyloidogenesis as a hallmark of AD pathogenesis, particularly with respect to mitochondrial dysfunction. We have previously shown that glycolytic glyceraldehyde-3-phosphate dehydrogenase (GAPDH) forms amyloid-like aggregates upon exposure to oxidative stress and that these aggregates contribute to neuronal cell death. Here, we report that GAPDH aggregates accelerate Aβ amyloidogenesis and subsequent neuronal cell death both in vitro and in vivo. Co-incubation of Aβ40 with small amounts of GAPDH aggregates significantly enhanced Aβ40 amyloidogenesis, as assessed by in vitro thioflavin-T assays. Similarly, structural analyses using Congo red staining, circular dichroism, and atomic force microscopy revealed that GAPDH aggregates induced Aβ40 amyloidogenesis. In PC12 cells, GAPDH aggregates augmented Aβ40-induced cell death, concomitant with disruption of mitochondrial membrane potential. Furthermore, mice injected intracerebroventricularly with Aβ40 co-incubated with GAPDH aggregates exhibited Aβ40-induced pyramidal cell death and gliosis in the hippocampal CA3 region. These observations were accompanied by nuclear translocation of apoptosis-inducing factor and cytosolic release of cytochrome c from mitochondria. Finally, in the 3×Tg-AD mouse model of AD, GAPDH/Aβ co-aggregation and mitochondrial dysfunction were consistently detected in an age-dependent manner, and Aβ aggregate formation was attenuated by GAPDH siRNA treatment. Thus, this study suggests that GAPDH aggregates accelerate Aβ amyloidogenesis, subsequently leading to mitochondrial dysfunction and neuronal cell death in the pathogenesis of AD.

tion were consistently detected in an age-dependent manner, and A␤ aggregate formation was attenuated by GAPDH siRNA treatment. Thus, this study suggests that GAPDH aggregates accelerate A␤ amyloidogenesis, subsequently leading to mitochondrial dysfunction and neuronal cell death in the pathogenesis of AD.
Glyceraldehyde-3-phosphate dehydrogenase (GAPDH) has been identified as a key enzyme in glycolysis. Several reports further reveal that GAPDH has a variety of other functions, including DNA repair (1), transcriptional regulation (2), membrane fusion and transport (3), autophagy (4), and cell death (5)(6)(7)(8). Upon exposure to oxidative stress, GAPDH binds Siah (5) and undergoes nuclear translocation. Nuclear GAPDH then induces p53-dependent transcription of apoptotic genes (9). We have previously reported that GAPDH also forms oxidative stress-induced aggregates, which is followed by the production of amyloid-like fibrils, eventually resulting in cell death (7,10). Similarly, in mice treated with methamphetamine, which causes widespread oxidative stress in the brain, there is evidence of GAPDH aggregation and neuronal cell death (11). Furthermore, GAPDH transgenic mice have enhanced neuronal cell death accompanied by robust aggregation of GAPDH (11).
Abnormal protein aggregation has been suggested as a possible underlying mechanism in the pathogenesis of several neurodegenerative disorders (12,13), and an increasing number of studies implicate GAPDH aggregation in this process (14 -17). Indeed, GAPDH aggregates have been found post mortem in the brains of patients with neurodegenerative disorders, particularly Alzheimer disease (AD). 3 However, details regarding its involvement in the development of AD remain unclear (18 -23).
AD is the most common cause of adult onset dementia, characterized by neuronal cell loss, senile plaques, and extensive gliosis in the cortex and hippocampus (24). There is good evidence suggesting that aggregation of amyloid-␤ peptide (A␤), a major component of senile plaques, occurs both intracellularly and extracellularly and that this aggregation is a primary event in the pathogenesis of AD (25)(26)(27). During aggregation, A␤ preferentially adopts the less soluble ␤-sheet structure, rather than the more soluble random coil and ␣-helix structures, leading to formation of oligomers and fibrils (28).
Mitochondrial dysfunction is a hallmark of A␤-induced neurotoxicity in AD (29). Extracellular A␤ aggregates bind to the plasma membrane, stimulating aberrant Ca 2ϩ influx, which then leads to disruption of mitochondrial membrane potential (30). Intracellular A␤ aggregates have recently been associated with the disruption of mitochondrial membrane potential via interaction with various mitochondrial proteins (31). A␤ aggregates cause massive increases in mitochondrial membrane permeability and stimulate the release of small proapoptotic proteins, including apoptosis-inducing factor (AIF) and cytochrome c (32).
To date, several lines of evidence point to an association between GAPDH and AD (18,20,33,34). In post-mortem brain samples from AD patients, GAPDH has been found in senile plaques, and disulfide-bonded GAPDH aggregates have been found in the detergent-insoluble fraction (17,18). Other than what has been revealed by these studies, little is known about the significance of GAPDH aggregation in the context of AD pathogenesis.
Based on these findings, we hypothesize that GAPDH aggregation might be involved in A␤ amyloidogenesis, thereby contributing to AD pathogenesis. In this study, we first show enhancement of A␤40 amyloidogenesis when co-incubated with GAPDH aggregates in vitro. We then demonstrate that GAPDH aggregates potentiate A␤40-induced neurotoxicity, accompanied by mitochondrial dysfunction in vitro and in vivo.
The triple transgenic mouse model of AD (3ϫTg-AD), which exhibits both intracellular and extracellular age-dependent A␤ aggregation (25), is often used in studies investigating AD pathogenesis (35). Therefore, we also examined the link between GAPDH aggregation and AD pathogenesis in 3ϫTg-AD mice. Our results suggest that there is an interaction between GAPDH and A␤ aggregates in the cortical extracellular deposits (the so-called human senile plaques) and hippocampal CA3 pyramidal neurons of these mice.
Preparation of A␤40 Solution-A lyophilized powder of A␤40 was dissolved in a solution of 0.02% ammonia at a stock concentration of 1 mM by brief stirring. Aliquots were stored at Ϫ80°C prior to use. The reaction solution of A␤40 was diluted with ice-cold PBS at a concentration of 50 M at 4°C and used immediately.
Cloning, Expression, and Purification of Human Recombinant GAPDH-Human GAPDH cDNA was generated, as described previously (7). For bacterial expression, cDNA was cloned into pBAD-HisA (Invitrogen) using the SacI-KpnI sites. The sequence of cloned human GAPDH cDNA was identical to that reported in GenBank TM (GenBank TM accession number M33197). The pBAD-HisA vector carrying human GAPDH cDNA was transformed into the gap(Ϫ) Escherichia coli strain, W3CG (37). Recombinant GAPDH protein was expressed and purified, as described previously (7). Briefly, the transformants were cultured for 2 h at 37°C in M63 minimal medium containing 50 g/ml ampicillin, 15 g/ml tetracycline, and 0.2% (w/v) L-(ϩ)-arabinose. After 24 h, cells expressing recombinant protein were collected by centrifugation (3000 ϫ g for 15 min at 4°C) and resuspended in a lysis buffer containing 50 mM sodium phosphate (pH 8.0), 300 mM NaCl, 30 mM imidazole, 10% glycerol, and 2 mM 2-mercaptoethanol. The suspensions were sonicated on ice and centrifuged at 15,000 ϫ g for 30 min at 4°C. The supernatants were incubated with nickel-nitrilotriacetic acid-agarose resin (50% slurry, Qiagen Japan, Tokyo, Japan) for 2 h at room temperature with shaking. Reduced proteins were loaded directly onto a PD-10 column (GE Health-Care UK Ltd.) equilibrated with G2 buffer containing 50 mM Tris-HCl (pH 8.0), 150 mM NaCl, 1 mM EDTA, and 5% glycerol. Protein concentrations were determined using spectrophotometry at 280 nm (⑀ 0.1% ϭ 1.0).
Congo Red Birefringence-The procedure for Congo red birefringence was carried out, as described previously (11), with minor modifications. Aliquots (40 l) of A␤40 at 50 M with or without 10% GAPDH aggregates were incubated for 48 h (day 2) and added to a 360-l Congo red solution (25 g/ml in PBS). This mixture was incubated for 30 min at 25°C and then centrifuged at 20,400 ϫ g for 30 min at 4°C. The resultant pellets were resuspended in 50 l of sterilized Milli-Q water and then dried on a glass slide. Birefringence was observed with an Eclipse LV100POL microscope equipped with a polarizing stage (Nikon, Tokyo, Japan).
Circular Dichroism (CD)-The far-ultraviolet CD spectrum of A␤40 at 50 M treated with or without GAPDH aggregates at day 2 was measured at 37°C with a spectropolarimeter (model J-820, Jasco, Tokyo, Japan). The path length of the optical quartz cuvette was 1.0 mm at 200 -250 nm. Spectra were obtained as the average of eight successive scans with a bandwidth of 2.0 nm. These spectra are expressed as values subtracted from that of each vehicle. These data are expressed as molar residue ellipticity ().
Atomic Force Microscopy (AFM)-Aliquots (10 l) of A␤40 at 50 M with or without GAPDH aggregates were incubated for 48 h (day 2), sonicated, and spotted onto freshly cleaved mica, incubated for 1 min, rinsed three times with water, and then dried. All measurements were carried out in "tapping mode" under ambient conditions using AFM (Nanoscope, Veeco Instruments Inc., Santa Barbara, CA) with single-beam silicon cantilever probes. We examined three regions of the mica surface to ensure that we obtained an accurate sample of structures on the mica (39).
Mitochondrial Membrane Potential Assay-Mitochondrial membrane potential was measured using rhodamine 123 (40). PC12 cells (3 ϫ 10 5 /cm 2 ) were seeded on glass coverslips coated with 0.1 mg/ml poly-D-lysine and grown for 24 h at 37°C. In this study, PC12 cells were divided into four treatment groups as follows: 1) control, vehicle of DMEM/F-12, 1% fetal bovine serum; 2) aggregated A␤40 (50 M) incubated for 48 h at 37°C in vehicle; 3) aggregated A␤40 (50 M) plus 10% GAPDH aggregates (5 M) in vehicle; and 4) 10% GAPDH aggregates in vehicle. After treatment, the culture medium was replaced with DMEM/F-12 containing rhodamine 123 (5 g/ml), followed by incubation for 10 min at 37°C in a 5% CO 2 -humidified incubator. After two washes with DMEM/F-12 and one wash with PBS, fluorescence from rhodamine 123-loaded cells was captured using a confocal scanning microscope (model C1si-TE2000-E, Nikon; excitation, 488 nm; emission, 525 nm). Raw images (500 ϫ 500 m) were first grayed and then transformed into binary images using Scion imaging software (version 4.0.3., Scion Corp.). The number of binary pixels in the square images was measured automatically. For semi-quantification of fluorescence, five microscopic fields were selected at random, and the number of cells among at least 500 total cells was quantified.
Animals-All animal experimental procedures were approved by the Animal Ethical Committee of Osaka Prefecture University and were performed according to the animal ethical guidelines of Osaka Prefecture University. Male C57BL/6J mice (3 months old, SLC Japan, Shizuoka, Japan) were used for A␤40 intracerebroventricular (i.c.v.) injection studies. A triple transgenic mouse model of AD (3ϫTg-AD), harboring human transgenes of amyloid precursor protein (APP SWE ), presenilin-1 (PS1 M146V ), and tau (Tau P301L ) under the control of Thy1.2 promoters (25), and control mice, were obtained from the Jackson Laboratory (Bar Harbor, ME). Mice of both genders were used at 1, 3, 6, and 9 months of age. All animals were maintained on ad libitum ordinary laboratory chow and tap water under a constant 12-h light/12-h dark cycle.
Surgical Procedure-Mice were anesthetized intraperitoneally with sodium pentobarbital (50 mg/kg, intraperitoneally, Abbott) and positioned in a stereotaxic frame (Narishige). The i.c.v. injections of A␤40 aggregates (50 M) incubated with or without 10% GAPDH aggregates were administered at a volume of 2 l using a Hamilton syringe (0.2 l/min) according to the following coordinates: 0 mm anterior to bregma, 0.8 mm lateral from midline, and 2.2 mm ventral from the skull surface. In this study, animals (n ϭ 10) were divided into four treatment groups as follows: 1) control, vehicle, sterilized PBS; 2) aggregated A␤40 (50 M) incubated for 48 h at 37°C in vehicle; 3) aggregated A␤40 (50 M) plus 10% GAPDH aggregates (5 M) in vehicle; and 4) 10% GAPDH aggregates in vehicle. Control and GAPDH siRNA were also injected in 2-l volumes in the same way.
Hematoxylin and Eosin Staining and Immunohistochemistry-Mice were deeply anesthetized with pentobarbital (200 mg/kg, intraperitoneally) and transcardially perfused with ice-cold PBS followed by 4% paraformaldehyde in PBS. The brain was removed and post-fixed in 4% paraformaldehyde in PBS at 4°C overnight and then routinely processed and embedded in paraffin. For morphological analysis, paraffin-embedded brain tissues were cut (5 m thick), deparaffinized, and stained with hematoxylin and eosin. For immunohistochemical analysis, deparaffinized sections were incubated with 100 g/ml proteinase K (Wako Pure Chemical Industries, Ltd., Osaka, Japan) in 50 mM Tris-HCl buffer (pH 7.4) for 10 min at 37°C. For detection of A␤, additional incubation with 70% formic acid for 5 min was performed. After three washes with PBS, sections were incubated with 10% goat serum in PBS, followed by incubation overnight at 4°C with mouse anti-A␤ monoclonal antibody (6E10, 1:1000), rabbit anti-GAPDH polyclonal antibody (1:500), rabbit anti-AIF polyclonal antibody (1:500), mouse anti-cytochrome c monoclonal antibody (1:500), and rabbit anti-GFAP polyclonal antibody (1:1000). Sections were then incubated for 1 h at room temperature with Alexa 488-or Alexa 568-conjugated secondary antibodies, goat anti-rabbit IgG (1:1000), goat anti-mouse IgG (1:1000), and DAPI (1:2000) for nuclear staining and then coverslipped in Fluorescein Mounting Medium (Dako). These antibodies were detected using a confocal scanning microscope (model C1si-TE2000-E, Nikon). For quantification of AIF nuclear translocation, cells exhibiting indication of AIF nuclear translocation were counted and expressed as a ratio to the total number of nuclei stained with DAPI. For quantification of GFAP, cytochrome c, A␤ aggregates, and GAPDH aggregates, raw images were first grayed and then transformed into binary images using Scion imaging software (version 4.0.3., Scion Corp.). The number of binary pixels in the square images was measured automatically.
Co-immunoprecipitation-Insoluble fractions prepared from mice hippocampi (0.5 mg/ml, 300 l) were incubated with normal mouse IgG (2 g) and 10 l of 50% protein G-Sepharose slurry (Amersham Biosciences) to pre-clear proteins that bind non-specifically to IgG. After centrifugation (15,000 rpm, 1 min), supernatants were incubated with an anti-A␤ monoclonal antibody (4 g) overnight at 4°C, followed by incubation with protein G-Sepharose for 1 h. The beads were washed four times with 1 ml of modified RIPA buffer containing 20 mM Tris-HCl (pH 7.5), 15 mM NaCl, 1 mM EDTA, 1 mM EGTA, 1% Nonidet P-40, 0.1% deoxycholate, 0.1% SDS, 1 mM PMSF, and a protease inhibitor mixture. The beads were added to 40 l of 2.5ϫ sample buffer followed by incubation for 5 min at 100°C and Western blotting.
Data Analysis-Data were expressed as mean Ϯ S.D. or S.E. for in vitro or in vivo studies, respectively, from three or four independent experiments. Statistical analyses were carried out using Student's t test and one-way analysis of variance, followed by Dunnett's multiple range test using GraphPad Prism (version 6, GraphPad Software, Inc.). Results with p values Ͻ0.05 were considered statistically significant. Complete (100%) amyloidogenesis of A␤40 was defined as a saturated value of ThT fluorescence obtained from each sample. Time half-values (t1 ⁄ 2 ) were calculated by sigmoidal curve fitting of the data, using GraphPad Prism (Version 6).

GAPDH Aggregates Accelerate A␤ Amyloidogenesis in
Vitro-To validate whether GAPDH aggregates affect A␤40 amyloidogenesis in vitro, we first performed a ThT fluorescence assay (Fig. 1). When A␤40 was incubated without GAPDH aggregates (black solid line), ThT fluorescence of A␤40 was not detected until day 2 (a lag phase), at which point fluorescence increased from day 3, eventually equilibrating at day 6 ( Fig. 1A). The t1 ⁄ 2 was calculated to be 3.86 Ϯ 0.06 days. A␤40 incubation with 10% GAPDH aggregates resulted in a ThT fluorescence curve that increased hyperbolically without a lag phase, equilibrating at day 2 (Fig. 1A, dotted line); t1 ⁄ 2 was 1.16 Ϯ 0.07 days. The degree of the leftward shift was ϳ3-fold. In the following studies, we examined samples at day 2, when we observed the largest differences between treatment with A␤40 with and without 10% GAPDH aggregates. GAPDH aggregates enhanced ThT fluorescence of A␤40 in a concentration-dependent manner (by ϳ10-, 25-, or 40-fold with the addition of 0.1, 1, or 10% GAPDH aggregates, respectively); ThT fluorescence in the solution of 10% GAPDH aggregates was not observed (Fig. 1B). Furthermore, addition of native GAPDH, instead of aggregated GAPDH, did not enhance ThT fluorescence; there was also no enhancement of fluorescence when A␤40 was incubated alone (Fig. 1C). Similarly, A␤42 and A␤ (25)(26)(27)(28)(29)(30)(31)(32)(33)(34)(35), which are known to be more prone to aggregation compared with A␤40, were incubated with GAPDH aggregates for 1 h to demonstrate augmentation of ThT fluorescence (Fig. 1, D and E). Incubation of A␤40 with 10% aggregates of bovine serum albumin did not affect ThT fluorescence (data not shown). These results suggest that GAPDH aggregation specifically augments A␤ amyloidogenesis.
Structural Changes in A␤40 Incubated with GAPDH Aggregates-We next examined A␤40 birefringence stained with Congo red, considered to be one of the most reliable methods for determination of amyloid structure (41). A␤40 incubated with or without 10% GAPDH aggregates was confirmed to bind Congo red under non-polarized light ( Fig. 2A, panels a  and b). Although A␤40 alone did not show any birefringence under polarized light ( Fig. 2A, panel c), A␤40 incubated with 10% GAPDH aggregates displayed prominent apple-green birefringence ( Fig. 2A, white arrow in panel d).
␤-Sheet secondary structure is a common feature of proteins undergoing amyloidogenesis (42,43). Thus, the amount of pro-tein with ␤-sheet structure in samples of A␤40 incubated with or without GAPDH aggregates was analyzed using CD spectra in the far-ultraviolet region (200 -250 nm, Fig. 2B). The spectra of A␤40 incubated alone and fresh A␤40 (without incubation) had minimums under 200 nm and were characteristic of random coil conformation (Fig. 2B, black solid line and black dashed line, respectively). In contrast, the spectra of A␤40 incubated with 0.1 and 1% GAPDH aggregates had increased negative ellipticities at 209 and 216 nm (Fig. 2B, red and green solid lines), respectively. Furthermore, the spectrum of A␤40 incubated with 10% GAPDH aggregates displayed a minimum at ϳ218 nm, a feature indicative of a predominantly ␤-sheet conformation (Fig. 2B, blue solid line) (44). These results demonstrate that GAPDH aggregates increase the amount of A␤40 with a ␤-sheet structure in a concentration-dependent manner.
We further examined the morphology of A␤40 incubated with or without 0.1, 1, or 10% GAPDH aggregates using AFM. A␤40 incubated without GAPDH aggregates was represented by numerous unassembled species and did not display typical amyloidal fibrillar structure (Fig. 2C, panel a). When A␤40 was incubated with 0.1% GAPDH aggregates, typical amyloid fibrils were observed (Fig. 2C, white arrows in panel b). These fibrils were elongated when A␤40 was incubated with 1% GAPDH aggregates (Fig. 2C, white arrowheads in panel c). When A␤40 was incubated with 10% GAPDH aggregates, we observed assemblies of these elongated amyloid fibrils (Fig. 2C, gray arrowhead in panel d). We did not observe any of these structures in 10% GAPDH aggregates incubated alone (Fig. 2C, panel  e). These data indicate that GAPDH aggregates can substantially alter A␤40 morphology in a concentration-dependent manner. GAPDH Aggregates Potentiate A␤40-induced Cytotoxicity in PC12 Cells, Concomitant with Mitochondrial Dysfunction-We next examined whether augmentation of A␤40 amyloidogenesis by GAPDH aggregates influences PC12 cell viability (Fig.  3A). Cell viability significantly decreased by 28% (p Ͻ 0.05) in cells treated with 50 M A␤ alone compared with vehicletreated controls. Treatment with A␤40 co-incubated with GAPDH aggregates further decreased cell viability in a concentration-dependent manner, decreasing by 58% (p Ͻ 0.05) or 78% (p Ͻ 0.01) when A␤40 was incubated with 1 or 10% GAPDH aggregates, respectively. Treatment with 10% GAPDH aggregates alone did not decrease cell viability, compared with vehicle-treated controls (Fig. 3A).
To investigate mechanisms enhancing A␤40-induced cell death by GAPDH aggregates, we next investigated mitochondrial dysfunction, which is known to play a key role in the pathogenesis of AD (32). Mitochondrial membrane potential was measured as an index of mitochondrial dysfunction (45) using rhodamine 123, a marker of mitochondrial membrane potential (Fig. 3B). Treatment with A␤40 incubated alone did not increase rhodamine fluorescence (Fig. 3B, panels a  and b). In contrast, treatment with A␤40 incubated with 10% GAPDH aggregates significantly increased rhodamine fluorescence by ϳ1.7-fold, indicating disruption of mitochondrial membrane potential (Fig. 3B, panel c). Treatment with 10% GAPDH alone did not change rhodamine fluorescence (Fig. 3B, panel d). These results suggest that GAPDH aggre-   gates enhance A␤40-induced cell death via mitochondrial dysfunction. GAPDH Aggregates Potentiate A␤40-induced Neurotoxicity in Vivo-To examine the effect of GAPDH aggregates on A␤40induced neurotoxicity in vivo, i.c.v. injections of A␤40 incubated with or without 10% GAPDH aggregates were administered (Fig. 4A). In C57BL/6J mice treated with A␤40 incubated alone, pyramidal cells in the hippocampal CA3 region with pyknotic nuclei increased in number to a maximum of ϳ2.5fold (Fig. 4B, panel b), compared with vehicle-treated mice (Fig.  4B, panel a). The number of such pathological cells was counted and graphed (Fig. 4B, right panel). Treatment with A␤40 incubated with 10% GAPDH aggregates significantly increased the number of cells with pyknotic nuclei by ϳ2.2fold, compared with mice treated with A␤40 incubated alone (p Ͻ 0.05, Fig. 4B, panel c). In contrast, when mice were i.c.v.-injected with 10% GAPDH aggregates alone, the resultant number of pathological cells was almost identical to that of the vehicle-treated group (Fig. 4B, panel d). Consistent with previous reports showing the vulnerability of the hippocampal CA3 region to A␤ toxicity (46,47), these pathological changes were not observed in any other brain areas (data not shown).
We next examined astrocyte accumulation using GFAP immunohistochemistry (Fig. 4C), which is considered to be an index of pathological gliosis in AD (48). The amount of detected GFAP in mice treated with A␤40 alone increased significantly by ϳ1.8-fold compared with vehicle-treated mice (p Ͻ 0.05, Fig. 4C, panel b). The amount of detected GFAP after treatment with A␤40 incubated with 10% GAPDH aggregates was further elevated by ϳ2.5-fold (p Ͻ 0.01, Fig. 4C, panels b and c), compared with mice treated with A␤40 alone. We did not observe any difference in results after treatment with vehicle or 10% Aβ40 + 10% GAPDH aggregates 10% GAPDH aggregates FIGURE 3. GAPDH aggregates potentiate A␤40-induced cytotoxicity in PC12 cells, concomitant with mitochondrial dysfunction. A, cell viability of PC12 cells was measured 72 h after treatment with A␤40 incubated with GAPDH aggregates. Data are presented as mean Ϯ S.D. of four independent experiments (**, p Ͻ 0.01, relative to control, as determined by Student's t test; #, p Ͻ 0.05; ##, p Ͻ 0.01, relative to 50 M A␤40, as determined by Dunnett's multiple range test). B, mitochondrial membrane potential was assessed using the fluorescent dye, rhodamine 123 (panels a-d show treatment with vehicle, A␤40 incubated alone, A␤40 incubated with 10% GAPDH aggregates, and 10% GAPDH aggregates alone, respectively). Distribution of the fluorescent signal was measured automatically using Scion imaging software. Data are presented as mean Ϯ S.D. of four independent experiments (*, p Ͻ 0.05, relative to 50 M A␤40, as determined by Student's t test). A.U., arbitrary units. Scale bar, 10 m. GAPDH aggregates alone (Fig. 4C, panels a and d). These immunohistochemistry results are quantified in the graph in Fig. 4C (right panel) Together, these results suggest that GAPDH aggregates might be associated with AD pathogenesis.
GAPDH Aggregates Enhance A␤40-induced Mitochondrial Dysfunction in Vivo-To investigate mechanisms underlying the enhancement of A␤40-induced neurotoxicity by GAPDH aggregates, we performed immunohistochemical studies on AIF and cytochrome c (Fig. 5), which are established markers of mitochondrial dysfunction (49). The number of pyramidal cells in the hippocampal CA3 region exhibiting AIF nuclear translocation was counted (Fig. 5A). AIF was predominantly localized to the cytosol of mice treated with vehicle (Fig. 5A, panels a and  e). Approximately 7% of cells showed AIF nuclear translocation after treatment with A␤40 alone (Fig. 5A, panels b and f). The number of cells displaying AIF nuclear translocation signifi-

FIGURE 4. GAPDH aggregates potentiate A␤40-induced neurotoxicity in vivo.
A, study design and experimental scheme. Solutions containing A␤40 with or without GAPDH aggregates were incubated at 37°C for 48 h and then injected i.c.v. Ten days after i.c.v. injection, mice were sacrificed and subjected to the following experiments. B, hematoxylin and eosin staining of the hippocampal CA3 region (left panels; panels a-d show treatment with vehicle, A␤40 incubated alone, A␤40 incubated with 10% GAPDH aggregates, and 10% GAPDH aggregates alone, respectively). The number of pyknotic cells was counted manually (right panel). Data are presented as mean Ϯ S.E. of four independent experiments (*, p Ͻ 0.05, relative to A␤40 treatment, as determined by Student's t test). Scale bar, 100 m. Magnification of selected portions of images are shown in the insets. Scale bar, 25 m. C, immunohistochemistry (IHC) targeting GFAP in the hippocampal CA3 region (panels a-d show treatment with vehicle, A␤40 incubated alone, A␤40 incubated with 10% GAPDH aggregates, and 10% GAPDH aggregates alone, respectively). Fluorescence intensity was measured automatically using Scion imaging software. Data are presented as mean Ϯ S.E. of four independent experiments (*, p Ͻ 0.05; **, p Ͻ 0.01, relative to vehicle treatment, as determined by one-way analysis of variance, followed by Dunnett's multiple range test; ##, p Ͻ 0.01, relative to A␤40 treatment, as determined by Student's t test). A.U., arbitrary units. Scale bar, 100 m. Magnification of selected portions of images are shown in the insets. Scale bar, 25 m. OCTOBER 23, 2015 • VOLUME 290 • NUMBER 43

JOURNAL OF BIOLOGICAL CHEMISTRY 26079
cantly increased by ϳ18% after treatment with A␤40 incubated with 10% GAPDH aggregates compared with A␤40 alone (Fig.  5A, panels c and g). Treatment with 10% GAPDH aggregates alone did not affect AIF nuclear translocation (Fig. 5A, panels d  and h).
Detection of cytochrome c, the punctate signals of which indicate localization of cytochrome c to mitochondria, was observed in vehicle-treated mice (Fig. 5B, panel a). After treatment with A␤40 alone, the signal had a less punctate pattern and was widely distributed in the cytosol (Fig. 5B, panel b) by

Vehicle Aβ40
Aβ40 + 10% GAPDH aggregates 10% GAPDH aggregates FIGURE 5. GAPDH aggregates enhance A␤40-induced mitochondrial dysfunction in vivo. Solutions containing A␤40 with or without GAPDH aggregates were incubated at 37°C for 48 h and injected i.c.v. Ten days after i.c.v. injection, mice were sacrificed and subjected to the following experiments. A, immunohistochemistry (IHC) targeting AIF was performed (panels a-d show treatment with vehicle, A␤40 incubated alone, A␤40 incubated with 10% GAPDH aggregates, and 10% GAPDH aggregates alone, respectively; corresponding images merged with nuclear staining (ϩ DAPI) are shown in panels e-h, respectively). The number of cells displaying nuclear translocation of AIF was counted manually (upper graph). Data are presented as mean Ϯ S.E. of four independent experiments (**, p Ͻ 0.01, relative to vehicle treatment, as determined by one-way analysis of variance, followed by Dunnett's multiple range test; ##, p Ͻ 0.01, relative to A␤40 treatment, as determined by Student's t test). Scale bars, 20 m. B, immunohistochemistry targeting cytochrome c was performed (panels a-d show treatment with vehicle, A␤40 incubated alone, A␤40 incubated with 10% GAPDH aggregates, and 10% GAPDH aggregates alone, respectively), and the release of cytochrome c to the cytosol was quantified using densitometry (lower graph). Data are presented as mean Ϯ S.E. of four independent experiments (*, p Ͻ 0.05; **, p Ͻ 0.01, relative to vehicle treatment, as determined by one-way analysis of variance, followed by Dunnett's multiple range test; ##, p Ͻ 0.01, relative to A␤40 treatment, as determined by Student's t test). Scale bar, 20 m. ϳ3-fold (p Ͻ 0.05), compared with vehicle-treated mice, indicating the release of cytochrome c from mitochondria. This change in signal distribution was more marked after treatment with A␤40 incubated with 10% GAPDH aggregates (Fig. 5B, panel c) by maximal degree to ϳ5.8-fold (p Ͻ 0.01). There was no change in cytochrome c signal distribution in mice treated with vehicle or 10% GAPDH aggregates alone (Fig. 5B, panel d).
These results suggest that GAPDH aggregates likely enhance A␤40-induced mitochondrial dysfunction in the context of AD.
GAPDH Aggregates Accelerate A␤40 Amyloidogenesis and Potentiate A␤ Toxicity-These results raised the question of whether GAPDH aggregates merely accelerate A␤40 amyloidogenesis or whether they also increase A␤40 toxicity. To clarify this point, A␤40 incubated alone for 6 days and A␤40 incubated with GAPDH aggregates for 2 days with similar levels of A␤40 amyloidogenesis (assessed by ThT fluorescence, Figs. 1A and 6A) were used for i.c.v. injection. In mice treated with A␤40 incubated with GAPDH aggregates for 2 days, cells with pyknotic nuclei were significantly more numerous than in mice treated with A␤40 incubated for 6 days (Fig. 6B). Similarly, the number of GFAP-positive cells was higher in mice treated with A␤40 incubated with GAPDH aggregates for 2 days compared with mice treated with A␤40 incubated for 6 days (Fig. 6C).
These results indicate that GAPDH aggregates both accelerate A␤40 amyloidogenesis and potentiate A␤40 neurotoxicity.

Co-localization of GAPDH Aggregates with A␤ Aggregates in the Cortex and Hippocampal CA3 Region of 3ϫTg-AD Mice-
To further investigate potential association between GAPDH aggregates and A␤ in AD pathogenesis, we used 3ϫTg-AD mice for immunohistochemical analysis of A␤ and GAPDH in the brain. The brain regions analyzed are shown as red squares, indicating the cortex and hippocampal CA3 region (Fig. 7A). In 9-month-old 3ϫTg-AD mice, immunostaining revealed extracellular deposits of A␤ in the cortex, consistent with senile plaques in human AD. These deposits were also positive for GAPDH (Fig. 7B), indicating co-localization of GAPDH and A␤. In contrast to the control mice, in 3-month-old 3ϫTg-AD mice, but not 1-month-old mice, hippocampal CA3 pyramidal neurons exhibited punctate A␤-positive signals, which were more abundant at 6 months of age (Fig. 7C). These findings have been confirmed by several groups to be age-dependent A␤ aggregation (25,50). We also observed punctate GAPDH signaling in control and 3ϫTg-AD mice as early as 1 month of age, which increased at 3 and 6 months of age (Fig. 7C). Furthermore, GAPDH-positive and A␤-positive signals partially co-localized in 3ϫTg-AD mice that were 3 and 6-months old (Fig.  7C). These results were quantified in the graph in Fig. 7D. Thus, these data suggest that GAPDH aggregation occurs prior to A␤ aggregation.
We further investigated whether these GAPDH-and A␤-positive signals were related to mitochondrial dysfunction in 6-month-old 3ϫTg-AD mice (Fig. 7E). Consistent with our data from i.c.v.-injected mice, AIF nuclear translocation and cytochrome c release were observed in pyramidal cells of the hippocampal CA3 region (Fig. 7E). Collectively, these results suggest that in 3ϫTg-AD mice, GAPDH aggregates co-localize with A␤ aggregates both extracellularly and intracellularly and that GAPDH aggregation precedes A␤ aggregation in the hippocampal CA3 region, leading to mitochondrial dysfunction.
GAPDH and A␤ Aggregates Interact and Enhance A␤ Aggregation in 3ϫTg-AD Mice-To examine the interaction between GAPDH and A␤ aggregates, we next performed a co-immunoprecipitation assay using both insoluble fractions from the hippocampi of 3-month-old 3ϫTg-AD mice and an anti-A␤ antibody. The result revealed a clear interaction between GAPDH and A␤ in insoluble fractions (Fig. 8).
Finally, to investigate the pathophysiologic significance of GAPDH aggregates on A␤ aggregation in AD pathogenesis, we conducted experiments using GAPDH silencing and supplementary methods. The 3ϫTg-AD mice were i.c.v. injected with either GAPDH siRNA or GAPDH aggregates, and each insoluble fraction from the hippocampus collected 10 days after injection was subjected to Western blotting using an anti-A␤ antibody (Fig. 9A). In mice treated with control siRNA, numerous A␤-positive signals (i.e. A␤ aggregates) were detected. In contrast, these A␤ aggregates were significantly reduced by approximately half in GAPDH siRNA-treated mice (Fig. 9B). Conversely, A␤ aggregates were increased by ϳ1.2-fold in GAPDH aggregate-injected mice compared with vehicletreated mice (Fig. 9C). These results indicate that GAPDH aggregates enhance A␤ aggregate formation in 3ϫTg-AD mice.

Discussion
This study demonstrates that GAPDH aggregates enhance A␤40 amyloidogenesis, as determined by ThT fluorescence, Congo red birefringence, and far-ultraviolet CD spectra; our AFM studies also revealed that GAPDH aggregates markedly change the morphology of A␤40 (Figs. 1 and 2). Moreover, augmentation of A␤40 amyloidogenesis by GAPDH aggregates potentiated A␤40-induced neurotoxicity, accompanied by mitochondrial dysfunction in vitro and in vivo (Figs. 3-6). Concomitant with mitochondrial dysfunction, GAPDH aggregates were also found to co-localize with extracellular and intracellular A␤ aggregates in aged 3ϫTg-AD mice (Fig. 7). Furthermore, we demonstrated an interaction between GAPDH and A␤ aggregates (Fig. 8), and we showed that GAPDH aggregates enhance A␤ aggregate formation in 3ϫTg-AD mice (Fig. 9). These results support our hypothesis. GAPDH aggregates accelerate A␤ amyloidogenesis and may, at least in part, contribute to AD pathogenesis (Fig. 10, hypothetical working model).
GAPDH aggregates are likely involved in the initiation and progression of AD pathogenesis. It is generally accepted that A␤ amyloidogenesis leads to increases in the oligomers and fibrils that play a central role in the initiation and progression of AD (24). The concentration of A␤ in vivo is reported to be ϳ5 nM (51,52), whereas aggregation in vitro requires a concentration nearly 3 orders of magnitude greater than in vivo values (53). To account for this discrepancy, the "seeding" hypothesis hasbeenproposed,basedonnucleation-dependentproteinpolymerization, (54). The seeding hypothesis proposes that a number of lipids and proteins, such as GM1 ganglioside and A␤42, serve as "seeds" for A␤40 (55). The sigmoidal curve of our ThT fluorescence data, demonstrating that incubation of A␤40 with GAPDH aggregates increases ThT fluorescence in a concentration-dependent manner, is consistent with a nucleation-dependent polymerization model of A␤40 amyloidogenesis (54). We further show that co-incubation with GAPDH aggregates, but not soluble forms of GAPDH, specifically enhanced A␤40 amyloidogenesis (Figs. 1 and 2). GAPDH aggregates also potentiated the amyloidogenesis of A␤42 and A␤(25-35) (Fig. 1, D and E), which are known to aggregate more quickly and act as "seeds" for A␤40, contributing to greater neurotoxicity. However, "seed" has been defined as an aggregation-inducing factor leading to immediate polymerization without any lag time (54). Therefore, GAPDH aggregates cannot be strictly defined as seeds due to their lag time (Fig. 1A). These observations raise the possibility that GAPDH aggregation instead serves as an accelerator of A␤ nucleation, which is the rate-limiting step in A␤ aggregation. n.s.

Aβ40 alone (day 6)
Aβ40 + 10% GAPDH aggregates (day 2) 2 days 6 days FIGURE 6. GAPDH/A␤ co-aggregates are highly toxic compared with A␤40 fibrils. A, A␤40 was incubated either alone for 6 days or with GAPDH aggregates for 2 days before ThT fluorescence was measured. Mice were i.c.v. injected with these incubated A␤40 samples and subjected to experiments at day 10 (as described in Fig. 4). n.s., no significance. B, hematoxylin and eosin staining in the hippocampal CA3 region (panels a and b show images from animals treated with A␤40 incubated alone for 6 days and A␤40 incubated with 10% GAPDH aggregates for 2 days, respectively). The number of pyknotic cells was counted manually. Data are the mean Ϯ S.E. of four independent experiments (*, p Ͻ 0.05, relative to treatment with A␤40 incubated alone for 6 days, Student's t test). This study suggests that GAPDH aggregates enhance A␤40induced cell death in vitro and in vivo (Figs. 3 and 4), as a result of increases in ␤-sheet content in A␤40 aggregates (Figs. 1 and  2). Identification of the species specifically causing neurodegeneration is currently one of the most controversial topics in the field of AD research. It is not clear whether fibrils (38,56) or soluble oligomers (57,58) have more serious toxicity responsible for AD pathology. However, there is a growing number of reports supporting the idea that the disease-associated, aggregation-prone protein readily adopts a ␤-sheet secondary structure (42,43) and that the ability to form ␤-sheet secondary structures is important for determining toxicity (59, 60). Therefore, GAPDH aggregates may facilitate the conversion of A␤ to a more toxic species by increasing ␤-sheet content. In addition to the ␤-sheet contents, the concern regarding polymorphisms affecting A␤ fibril toxicity has been growing. An example of a polymorphism having such an effect is that GM1 ganglioside not only accelerates A␤ aggregation, it also changes A␤ fibril conformation and enhances toxicity (61). Taking this into account, our results show that GAPDH aggregates accelerate A␤40 amyloidogenesis and also potentiate that A␤40 neurotoxicity (Fig. 6) is attributed to the formation of highly toxic materials.
Our results indicate that co-incubation of A␤40 with GAPDH aggregates might have caused significant mitochondrial dysfunction (Figs. 3B and 5). This dysfunction occurred  alongside enhancement of A␤40 neurotoxicity by GAPDH aggregates (Figs. 3A and 4). It has been reported that extracellular and intracellular A␤ aggregation in AD induced disruption of mitochondrial membrane potential (30,31). In this study, i.c.v. injections of A␤40 co-incubated with GAPDH aggregates led to nuclear translocation of AIF and cytoplasmic release of cytochrome c from mitochondria, which was also observed in 6-month-old 3ϫTg-AD mice (Figs. 5 and 7E). AIF nuclear translocation activates endonuclease-G in a caspaseindependent manner, and released cytochrome c initiates caspase-dependent cell death (32). Consequently, these events lead to neuronal cell death (32). Therefore, A␤40 co-incubation with GAPDH aggregates likely specifically triggers mitochondrial dysfunction, because A␤40 aggregates alone did not disrupt mitochondrial membrane potential (Figs. 3B and 5). Moreover, taking into account the fact that control mice did exhibit the mitochondrial dysfunction observed in 3ϫTg mice (Fig.  7E), these apoptotic changes are likely attributable to A␤ toxicity.
Our data from 3ϫTg-AD mice show co-localization of GAPDH aggregates with A␤ aggregates in extracellular deposits in the cortex. This co-localization was also observed in hippocampal CA3 pyramidal neurons and increased in an agedependent manner (Fig. 7, C and D). The formation of A␤/GAPDH co-aggregates was also quantitatively confirmed (Fig. 8). A␤ is produced by endoproteolysis of parental amyloid precursor protein and is secreted into the extracellular space (26). There is a substantial amount of evidence that A␤ can also accumulate intracellularly via several mechanisms, such as intracellular production and reuptake of extracellular A␤ (62). Secretion of GAPDH into the extracellular space has been previously reported (63). Therefore, extracellular and intracellular co-localization of GAPDH and A␤ aggregates both appear plausible.
Interestingly, 3ϫTg-AD mice at 1 month of age showed evidence of GAPDH aggregation without A␤ aggregation in the hippocampal CA3 region (Fig. 7C). GAPDH aggregation, A␤ . GAPDH aggregates interact with A␤ aggregates in the 3؋Tg-AD mouse hippocampus. GAPDH/A␤ co-aggregate formation is shown. The hippocampus of a 3-month-old 3ϫTg-AD mouse was subjected to insoluble fractionation and subsequently immunoprecipitated (IP) with an anti-A␤ antibody. This sample was analyzed by SDS-PAGE and Western blotting (WB) against GAPDH and A␤. Values were calculated as the ratio of GAPDH (immunoprecipitated) band intensity (quantified in arbitrary units (A.U.)) relative to GAPDH (input) band intensity (*, p Ͻ 0.05, relative to the samples immunoprecipitated with control IgG, Student's t test). . GAPDH aggregates potentiate A␤ aggregate formation in the hippocampus of 3؋Tg-AD mice. A, schema indicating experiments following i.c.v. injection of GAPDH siRNA or GAPDH aggregates. 3ϫTg-AD mice were i.c.v. injected with GAPDH siRNA or GAPDH aggregates. Ten days after injection (at 3 months of age), mice were sacrificed and subjected to insoluble fractionation followed by Western blotting. B, effect of GAPDH knockdown on A␤ aggregation in 3ϫTg-AD mice was investigated by Western blotting (WB) against A␤ and H2B. H2B was used as internal standard protein present in the insoluble fraction (**, p Ͻ 0.01, relative to the treatment with control siRNA, Student's t test). C, effect of treatment with GAPDH aggregates on A␤ aggregation in 3ϫTg-AD mice was investigated by Western blotting against A␤ and H2B (*, p Ͻ 0.05, relative to vehicle treatment, Student's t test). A.U., arbitrary units.
aggregation, and co-localization of the two increased in an agedependent manner (Fig. 7, C and D). These results imply that GAPDH aggregates are mainly involved in the initiation of A␤ aggregation. This hypothesis is supported by evidence that GAPDH easily aggregates as a consequence of exposure to oxidative stress (7,11). How do GAPDH aggregates occur in 3ϫTg-AD brain? Considering that control mice showed almost the same patterns of GAPDH aggregates in an age-dependent manner as 3ϫTg-AD mice (Fig. 7, C and D), the formation of GAPDH aggregates in the hippocampal CA3 region seems to be age-dependent. Indeed, redox proteomics analyses of rat brains revealed the presence of oxidatively modified GAPDH associated with changes in redox status during the aging process (64). Therefore, emergence of GAPDH aggregates in 3ϫTg-AD mice might depend on normal brain aging rather than its triple mutant phenotype.
Taking into consideration the fact that several proteins known to co-localize with A␤ deposits, such as ␣-synuclein, prion protein, and cystatin C, regulate (either accelerate or inhibit) A␤ aggregation (65), it is possible that GAPDH also facilitates A␤ aggregation. Furthermore, from the results obtained from experiments using the i.c.v. injection of either GAPDH siRNA or GAPDH aggregates in 3ϫTg-AD mice (Fig.  9), we suggest that GAPDH aggregates play a significant pathophysiological role in A␤ aggregation in AD pathogenesis.
In summary, this study demonstrates that GAPDH aggregates augment A␤40 amyloidogenesis and promote A␤40-induced cell death, accompanied by mitochondrial dysfunction in vitro and in vivo (Fig. 10). Moreover, co-localization of GAPDH/A␤ aggregates was observed in the cortices and hip-pocampi of 3ϫTg-AD mice. These findings suggest that interaction between GAPDH aggregates and A␤40 might be involved, at least in part, in the pathogenesis of AD.