Abstract
Accumulation of amyloid-β peptide (Aβ) aggregates in synapses may contribute to the profound synaptic loss characteristic of Alzheimer's disease (AD). The origin of synaptic Aβ aggregates remains elusive, but loss of endosomal proteostasis may trigger their formation. In this study, we identified the synaptic compartments where Aβ accumulates, and performed a longitudinal analysis of synaptosomes isolated from brains of TgCRND8 APP transgenic mice of either sex. To evaluate the specific contribution of Aβ-degrading protease endothelin-converting enzyme (ECE-1) to synaptic/endosomal Aβ homeostasis, we analyzed the effect of partial Ece1 KO in brain and complete ECE1 KO in SH-SY5Y cells. Global inhibition of ECE family members was used to further assess their role in preventing synaptic Aβ accumulation. Results showed that, before extracellular amyloid deposition, synapses were burdened with detergent-soluble Aβ monomers, oligomers, and fibrils. Levels of all soluble Aβ species declined thereafter, as Aβ42 turned progressively insoluble and accumulated in Aβ-producing synaptic endosomal vesicles with characteristics of multivesicular bodies. Accordingly, fibrillar Aβ was detected in brain exosomes. ECE-1-deficient mice had significantly increased endogenous synaptosomal Aβ42 levels, and protease inhibitor experiments showed that, in TgCRND8 mice, synaptic Aβ42 became nearly resistant to degradation by ECE-related proteases. Our study supports that Aβ accumulating in synapses is produced locally, within endosomes, and does not require the presence of amyloid plaques. ECE-1 is a determinant factor controlling the accumulation and fibrillization of nascent Aβ in endosomes and, in TgCRND8 mice, Aβ overproduction causes rapid loss of Aβ42 solubility that curtails ECE-mediated degradation.
SIGNIFICANCE STATEMENT Deposition of aggregated Aβ in extracellular plaques is a defining feature of AD. Aβ aggregates also accumulate in synapses and may contribute to the profound synaptic loss and cognitive dysfunction typical of the disease. However, it is not clear whether synaptotoxic Aβ is mainly derived from plaques or if it is produced and aggregated locally, within affected synaptic compartments. Filling this knowledge gap is important for the development of an effective treatment for AD, as extracellular and intrasynaptic pools of Aβ may not be equally modulated by immunotherapies or other therapeutic approaches. In this manuscript, we provide evidence that Aβ aggregates building up in synapses are formed locally, within synaptic endosomes, because of disruptions in nascent Aβ proteostasis.
Introduction
Disruption of endosomal/lysosomal (E/L) function is recognized as a possible precipitating factor for the development of brain β-amyloidosis, a condition caused by the abnormal aggregation of β-amyloid peptide (Aβ) and one of the main neuropathological features of Alzheimer's disease (AD) (Glenner and Wong, 1984; Roychaudhuri et al., 2009). The role that endosomes play in homeostasis of Aβ is twofold. First, Aβ is produced by cleavage of the amyloid precursor protein (APP) within neuronal endosomal vesicles and perturbations in vesicular pH and trafficking can greatly enhance Aβ production (Nixon, 2007; Pacheco-Quinto and Eckman, 2013; Pacheco-Quinto et al., 2019). Second, the E/L pathway is a major catabolic route for Aβ and contains multiple proteases, including cathepsins and endothelin-converting enzymes (ECE) -1 and -2, that prevent buildup of the peptide. Alterations in the endosomal system therefore have the potential to impact both Aβ production and degradation, leading to excess accumulation of intracellular Aβ and its aggregation (Pacheco-Quinto and Eckman, 2013), a required step for the development of amyloid pathology. Indeed, increased intraneuronal Aβ immunoreactivity (Gyure et al., 2001; Mori et al., 2002; Cataldo et al., 2004; Billings et al., 2005; Pensalfini et al., 2014) and signs of E/L dysfunction, such as enlarged and poorly acidified endosomal vesicles, are evident before the onset of extracellular pathology (Cataldo et al., 1994; Nixon, 2017; Lee et al., 2022).
Synaptic dysfunction is another early feature of AD and is considered, in part, to be an Aβ-driven insult. Numerous studies have demonstrated that soluble Aβ aggregates can interfere with normal synaptic function and cause synaptic degeneration (S. Li and Selkoe, 2020). How neurotoxic Aβ aggregates develop and reach synapses in the first place is still unclear, but evidence suggesting a potential endogenous intraneuronal origin is compelling. Neuronal endosomes are the main Aβ-producing compartments in brain, provide a favorable environment for Aβ aggregation, and in synapses, accumulation of Aβ in endosomes is a characteristic ultrastructural finding in AD (Takahashi et al., 2002; Langui et al., 2004). Hence, similar to the inside-out hypothesis for plaque formation (Gouras et al., 2014), progressive synaptic degeneration could be driven by excess accumulation of Aβ produced within synaptic endosomes.
In this study, using a mouse model of AD-like β-amyloidosis, we conducted a longitudinal biochemical analysis of synaptic Aβ accumulation and explored early changes in intraneuronal Aβ proteostasis contributing to the pathologic process. The results of these experiments, paired with additional data from ECE-1-deficient cells and mice, provide new insights on the origin of intrasynaptic Aβ aggregates and demonstrate that loss of endosomal Aβ homeostasis may be a major driver of amyloid pathology in AD.
Materials and Methods
Mice
TgCRND8 mice were a kind gift from Paul Fraser, University of Toronto (Chishti et al., 2001). All experimental mice were obtained by breeding male TgCRND8 mice and female B6C3F1/J hybrid mice (stock #100010, The Jackson Laboratory). Ece1tm1a(KOMP)Wtsi mice, generated by the Knock-out Mouse Project (KOMP) and obtained from the Mutant Mouse Resource & Research Centers at the University of California–Davis (MMRRC:047475-UCD), contain a “knock-out first” reporter-tagged insertion allele. Humanized App knock-in mice (B6(SJL)-Apptm1.1Aduci/J) (Baglietto-Vargas et al., 2021) were obtained from The Jackson Laboratory (RRID:IMSR_JAX:030898) and were used to prepare synaptosomes for establishing background signal in rodent-specific Aβ ELISAs. All mice were housed in ventilated micro-isolator cages with unrestricted access to food and water and were provided with nesting material and Shepherd shacks (Shepherd Specialty Papers). The housing room was under a 12 h/12 h light/dark cycle with controlled temperature and humidity. The use of mice was approved by the Institutional Animal Care and Use Committee at BRInj and followed American Veterinary Medical Association guidelines.
Synaptosomal preparations
Mice were killed by CO2 asphyxiation, brains removed, and cortex and hippocampus quickly dissected. Tissue was placed on ice in cold 0.32 m sucrose/10 mm HEPES buffer containing protease inhibitor cocktail with EDTA (50 mg wet weight per ml) and was homogenized with 10 strokes of a motorized glass-Teflon Dounce homogenizer. From the crude homogenate, an aliquot was kept for measuring total Aβ. The interstitial enriched “extracellular” extract was prepared as the final supernatant from three sequential centrifugation steps at 4°C (1000 × g for 10 min, 16,000 × g for 10 min, 100,000 × g for 1 h). Crude synaptosomes were pelleted at the 16,000 × g centrifugation step and were washed 3 times by cycles of resuspension in homogenization buffer and centrifugation at 16,000 × g for 10 min at 4°C. Washed synaptosomal pellets from 1 ml homogenate were then lysed in 1 ml RIPA buffer (25 mm Tris-HCl, 150 mm NaCl, 1% NP-40, 1% sodium deoxycholate, 0.1% SDS, pH 7.6) containing Halt Protease Inhibitor Cocktail with EDTA (Fisher Scientific) and centrifuged at 16,000 × g for 10 min to obtain the soluble fraction. Insoluble Aβ was extracted from the resulting pellet by addition of 50 µl of 6 m guanidine-HCl (GuHCl). For extraction of total Aβ, crude homogenates were diluted 1:20 in 6 m GuHCl and incubated for 30 min at room temperature. For analysis of endogenous rodent Aβ in Ece1 (+/+) and (+/−) mice, crude synaptosomes were prepared essentially as described above from brains homogenized at 100 mg wet weight per ml. Washed synaptosomal pellets were extracted directly with 1/10 volume 6 m GuHCl (relative to original homogenate volume) for 1 h at room temperature, and extracts were diluted 1:15 in ELISA binding buffer before analysis. For ECE-1 activity assays, synaptosomes were prepared in the absence of EDTA.
To obtain a pure synaptosomal fraction, crude synaptosomal pellets were loaded on to a 0.6-16 m continuous sucrose gradient and spun for 18 h in a swinging bucket rotor at 165,000 × g at rMAX. From the top to bottom, 500 µl fractions were collected, diluted 5 times with 10 mm HEPES, and spun down at 16,000 × g. For the separation of vesicles contained in synaptosomes, pelleted synaptosomes were subjected to osmotic shock by incubation in 10 mm HEPES for 20 min at 4°C. The solution was then passed through a 25-gauge needle and loaded onto a 5%-25% Optiprep (D1556, Millipore Sigma) continuous gradient and spun for 2 h at 140,000 × g at rMAX in a swinging bucket rotor. Aliquots of 500 µl were collected starting from the top.
Thioflavin staining protocol
Brains fixed in 10% neutral buffered formalin were embedded in paraffin and processed into 10-µm-thick sagittal sections. For each mouse, 3 slides, 100 µm apart, were used to assess amyloid plaques. Slides were deparaffinized and rehydrated by incubation at 60°C for 60 min followed by 5 min incubations in xylene (3 times), 100% ethanol, 95% ethanol, 50% ethanol and water. Slides were incubated in 5% thioflavin T in water (T3516, Millipore Sigma), rinsed with water, incubated in 1% acetic acid solution for 15 min, rinsed, air dried, and mounted. Plaques were visualized with a Zeiss Axio Imager Z1 fluorescent microscope, and all plaques present in cortex and hippocampus were counted manually.
Isolation of brain exosomes
Exosome-enriched extracellular vesicles (EVs) were isolated from dissected cortex/hippocampus using previously described methods with few modifications (Perez-Gonzalez et al., 2017). Half brains were mildly dissociated with 20 U/ml papain (#LK00316; Worthington Biochemical) and centrifuged sequentially at 1000 × g (to pellet cell debris), 10,000 × g (to sediment other EVs, such as macrovesicles or ectosomes), followed by a 100,000 × g ultracentrifugation step to pellet extracellular exosomes. Pellets then were resuspended in PBS and loaded on top of a discontinuous sucrose gradient (0.25, 0.6, 0.95, 1.3, 1.65 m sucrose) and centrifuged for 16 h at 200,000 × g. Based on their density, exosomes equilibrate at the 0.6/0.95 m sucrose interlayer (1.08-1.13 × g/ml density) and produce a visible layer. Using a syringe needle, the exosome layer was extracted, diluted with PBS buffer, and centrifuged again at 100,000 × g to pellet exosomes. Exosomes were resolubilized in 30 µl saline buffer to obtain an average protein concentration of 4 mg/ml. The size of the EVs was measured by Dynamic Light Scattering using a Malvern Zetasizer Nano ZS (Malvern Instruments). Preparations were diluted in 25 mm trehalose to obtain three independent measures.
Intracerebroventricular injection
Mice under isoflurane anesthesia (1.5%-5%) were placed in a stereotaxic frame (Stoelting), with body temperature maintained with a heating pad. A dental drill was used to perforate the skull (0.9 mm posterior, 1.5 mm lateral to bregma) and 2 µl of phosphoramidon (118 nmol) or vehicle (0.9% saline), were administered into each lateral ventricle (depth 2.1 mm) at rate of 0.5 µl/minute with a Hamilton glass microsyringe driven by a microsyringe pump (World Precision Instruments). Before surgery, bupivacaine (0.1 ml of a 0.125% solution) was injected subcutaneously at the incision site and meloxicam (2 mg/kg, i.p.) was administered for analgesia. Related data from a subset of the injected mice were reported in a previous publication (Pacheco-Quinto et al., 2019).
Cell culture
Human neuroblastoma SH-SY5Y cells (ATCC CRL-2266, RRID:CVCL_0019, American Type Culture Collection) stably transfected with pcDNA3 (Invitrogen) containing nonmutant human APP695 cDNA (NM_201414.2) were maintained in DMEM supplemented with 10% FBS and penicillin and streptomycin, and routinely passed by trypsinization. To quantify Aβ by ELISA, cells were grown in 12-well plates and treated with 100 µm of phosphoramidon and/or E-64 (Millipore Sigma) for 48 h and extracted with 100 µl of RIPA buffer (25 mm Tris-HCl, 150 mm NaCl, 1% NP-40, 1% sodium deoxycholate, 0.1% SDS, pH 7.6) containing Halt Protease Inhibitor Cocktail with EDTA (Fisher Scientific). After a 10 min incubation on ice, samples were spun at 16,000 × g and supernatant was used to measure RIPA-soluble Aβ42. The remaining pellet was incubated with 20 µl of 6 m GuHCl at room temperature for 30 min and then diluted by addition of 80 µl RIPA. Extracts were further diluted 1:4 in binding buffer before analysis by ELISA.
Generation of ECE1 KO in SH-SY5Y-APP cells
Guide RNA targeting sequence oligos 1 (sgRNA-ECE1ex4-1, 5′-CACCGCTGAGACACAAGCTTCGCTC-3′) and 2 (sgRNA-ECE1ex4-2, 5′-CACCGCCCTGATGGCCACTCACGCT) were individually subcloned into pSpCas9(BB)-2A-Puro (PX459) version 2.0, a gift from Feng Zhang (Ran et al., 2013) (Addgene plasmid #62988; http://n2t.net/addgene:62988; RRID:Addgene_62988). ECE1 targeting sequences were tested by the Off-Spotter software to minimize potential off target effects. SH-SH5Y-APP cells were cotransfected with the vectors using Lipofectamine 3000 transfection reagent (Fisher Scientific), and placed under puromycin (2.5 µg/ml) selection for 4 d. After 4 weeks, individual colonies were isolated and expanded; clones with genomic insertions/deletions resulting in frameshifts were identified by Sanger DNA sequencing (Genewiz). APP expression level was determined by Western blot compared with the parent SH-SY5Y-APP line. Phenotypes of three clones with unique insertions/deletions were assessed by measuring Aβ levels in culture medium and cell lysate after 24 h incubation ± 100 µm phosphoramidon. All clones were insensitive to phosphoramidon treatment, but some had reduced APP expression, compared with the parent line, after clonal selection (data not shown). For all subsequent experiments in this study, we used a clone with APP expression similar to that of the parent line, that had cleavage on the target sites creating a 131 bp deletion in ECE1, resulting in a premature stop codon. Absence of ECE-1 activity in this clone was confirmed using a big ET-1 conversion assay.
ECE-1 enzymatic activity
Cell membrane fractions were prepared as described by Xu et al. (1994). Cell membranes, or mouse brain synaptosomes, were solubilized in 20 mm Tris-HCl containing 250 mm sucrose and 2.5% C12E10 (polyoxyethylene-10-lauryl ether, Millipore Sigma). Protein concentration was measured using Pierce BCA assay kit (Fisher Scientific). ECE-1 activity was quantified using a big ET-1 conversion assay (Emoto and Yanagisawa, 1995). Briefly, 2.5 µg protein were incubated for 20 min at 37°C with 0.1 µm human big endothelin-1 (big ET-1 1-38) peptide (Enzo Life Sciences) in 50 mm acetic acid, 50 mm 2-(N-morpholino)ethanesulfonic acid, 100 mm Tris, pH 6.8 (Fahnoe et al., 2000), containing Halt Protease Inhibitor Cocktail (Fisher Scientific) and 1 µm thiorphan (Cayman Chemical). Parallel reactions were conducted in the presence of 100 µm phosphoramidon. After stopping the reactions by addition of 5 mm EDTA, mature ET-1 (1-21) peptide was measured by sandwich ELISA (Endothelin-1 Quantikine ELISA kit, R&D Systems).
Immunocytochemistry and microscopy
SH-SY5Y-APP cells grown on glass coverslips were fixed with cold methanol at −20°C for 10 min. Cells were then incubated with 5% BSA/PBS containing 0.05% Triton X-100 for 30 min at room temperature, followed by a 4°C overnight incubation with anti-Aβ1-16 antibody MM27-33.1.1 (also known as Ab3, QED Bioscience, catalog #57003) (1:500) for Aβ detection or anti-amyloid fibrils OC antiserum (EMD Millipore, catalog #AB2286, RRID:AB_1977024) (1:500) for detection of fibrillar species. Anti-mouse or anti-rabbit antibodies conjugated with AlexaFluor-488 (Invitrogen) were used for detection. Nuclei were stained with Hoechst 33258 (Anaspec). For OC colocalization studies, cells were fixed with 10% PFA and permeabilized with 0.1% Triton X-100. OC signal was detected with an anti-rabbit AlexaFluor-546 antibody in cells incubated overnight with 0.5 µg/ml Lucifer yellow (Millipore Sigma) to visualize the endocytic pathway, or in cells transiently transfected with CD63-pEGFP C2 (gift from Paul Luzio, Addgene plasmid #62964; http://n2t.net/addgene:62964; RRID:Addgene_62964) using Lipofectamine 3000. For colocalization with a late endosomal marker, dual immunofluorescence was performed with OC antiserum and mouse anti-Rab7 (E9O7E, Cell Signaling catalog #95746, RRID:AB_2800252). Fluorescent signal was visualized with a Zeiss Axio Imager Z1 fluorescent microscope.
For super resolution imaging, we used a Zeiss LSM880 confocal microscope equipped with Airyscan. Fields of interest were visualized with a 63× oil objective (NA 1.4) and the Airyscan Super Resolution mode. Hoechst and AlexaFluor-488 were excited using the 405 and 488 nm laser lines, respectively. Sixteen-bit stacks were acquired at 1000 × 1000 pixel frames with pixel dwell times of 2.06 µs. Acquired stacks were processed through using linear deconvolution of each of the 32 individual elements to enhance their lateral and axial resolution, followed by pixel reassignment of each element back to the center position. Airyscan filtering (Wiener filter associated with deconvolution) was set to the default value of 6.0 for 2D images. Maximal projections of stacks were prepared using Fiji digital imaging software.
ELISA
Aβ was measured by sandwich ELISA as previously described (Scheuner et al., 1996) using antibody pairs that detect monomeric peptides preferentially, or that detect oligomers. For quantification of human or mouse Aβ1-40, Aβ was captured with N-terminal antibody MM27-33.1.1 (also known as Ab3, QED Bioscience, catalog #57003) or M3.2 (Biolegend catalog #805701, RRID:AB_2564982), respectively, and detected with HRP-conjugated Aβ40-specific antibody MM32-13.1.1 (also known as Ab40.1, QED Bioscience, catalog #57002). For quantification of Aβ ending at position 42, both human and mouse Aβ were captured with Aβ42-specific antibody MM26-2.1.3 (also known as Ab42.2, QED Bioscience, catalog #57001) and detected with HRP-conjugated 4G8 (Biolegend, catalog #800720, RRID:AB_2783372). This Aβ42 assay recognizes N-terminally truncated peptides in addition to full-length Aβ1-42. An alternative Aβ42 sandwich ELISA (capture with anti-rodent Aβ 1-16 antibody M3.2 and detection with anti-Aβ42 antibody BC05 (Fujifilm Wako #014-26903) was used to measure endogenous Aβ42 in synaptosomes from Ece1 (+/−) mice.
Aβ oligomers were quantified using a commercial sandwich ELISA that uses N-terminal antibody MOAB-2 for both capture and detection (Biosensis, catalog #BEK-2215) and preferentially detects oligomers (Tai et al., 2013).
Aβ protofibrils were quantified with a sandwich ELISA composed of protofibril-selective monoclonal capture antibody (mAbSL113) and a biotinylated affinity-purified protofibril-selective detection antibody (apAbSL40-4). The assay was completed by addition of a streptavidin-HRP conjugate, and HRP colorimetric substrates. A standard curve of isolated Aβ42 protofibrils ranging from 0.16 to 10 nm (705-45,140 pg/ml) was constructed, and protofibril concentrations in the samples were determined from linear regression of the protofibrils standard absorbances. R2 values for the standard curves were typically >0.99. The antibodies, modifications, and sandwich ELISA were developed at the University of Missouri–St. Louis, the basis of which was reported previously (Colvin et al., 2017; Grover et al., 2023).
For analysis of extracellular-enriched TgCRND8 mouse brain extracts, 100 µl extract were added to wells containing 50 µl ELISA binding buffer (0.02 m sodium phosphate, pH 7.0, containing 0.002 m EDTA, 0.4 m NaCl, 0.2% BSA, 0.05% 3-[(3-cholamidopropyl)dimethylammonio]-1-propanesulfonate, 0.4% Block Ace, 0.05% NaN3). For analysis of synaptosomal Aβ concentration, RIPA extracts were assayed directly, 100 µl per well. Synaptosomal GuHCl extracts from TgCRND8 mice were diluted 1:17 with ELISA binding buffer and 120 µl of the diluted extract were added per well. For total mouse brain Aβ42, crude homogenates extracted with 6 m GuHCl were diluted 1:1000 in ELISA binding buffer and 100 µl were loaded. For quantification of total Aβ42 in exosomes, 3 µl of saline-resuspended exosome samples were combined with 15 µl of 6 m GuHCl and incubated at room temperature for 30 min. Samples were then diluted with 200 µl ELISA binding buffer and 100 µl were loaded per well.
Dot blot
RIPA-extracted brain synaptosomal vesicle preparations were further diluted 1:50 in RIPA buffer, and 100 µl was applied to a nitrocellulose membrane using a dot blot apparatus. For extracellular-enriched fractions, 35 µl of sample diluted in 185 µl RIPA buffer were added per well. For exosomes, 15 µg were diluted in 200 µl of RIPA buffer and loaded. Membranes were blocked for 1 h with 5% milk and incubated overnight with the OC antiserum (1:1000; AB2286 Millipore, Sigma) that recognizes fibrillar structure over a broad size range (Kayed et al., 2007; Glabe, 2008). After washing with 0.05% Tween-20/PBS, membranes were incubated with anti-rabbit HRP conjugated secondary antibody for 1 h, washed again, and developed with ECL reagents (Millipore Sigma). Chemiluminescent signal was acquired with an ImageQuant LAS 4000 (GE Healthcare Life Sciences). Optical density was quantified with ImageJ software (National Institutes of Health) and expressed as arbitrary units.
Western blotting
RIPA buffer extracts were resolved in Novex 10%-20% Tris-tricine gels (Fisher Scientific), transferred to nitrocellulose membranes, and blocked with 5% BSA. Antibodies against AIP1/Alix (Millipore catalog #ABC40, RRID:AB_11213660), C-terminal APP (Sigma-Aldrich catalog #A8717, RRID:AB_258409), LC3B (Sigma-Aldrich catalog #L7543, RRID:AB_796155), Rab5 (Cell Signaling Technology catalog #3547, RRID:AB_2300649), Rab7 (Cell Signaling Technology catalog #12561, RRID:AB_2797956), syntaxin 1A (Cell Signaling Technology catalog #18572, RRID:AB_2798803), synaptophysin (Abcam catalog #ab8049, RRID:AB_2198854), PSD95 (Cell Signaling Technology catalog #3450, RRID:AB_2292883), and CHMP2B (Proteintech catalog #12527-1-AP, RRID:AB_10603358) were used as primary antibodies. HRP-conjugated antibodies were used as secondary antibodies and signal was developed with ECL reagents.
Electron microscopy (EM)
For ultrastructural analysis, samples containing vesicles in suspension were deposited on copper grids (Electron Microscope Sciences) for 5 min followed by washing and contrasting by negative staining using uranyl acetate. Grids were imaged using a JEM 1400 transmission electron microscope (JEOL) operated at 80 kV and images were acquired with an Olympus-SIS Veleta side-mount 2K × 2K digital CCD camera.
Experimental design and statistical analysis
For longitudinal analysis of Aβ species in TgCRND8 mice (data shown in Figs. 3–6), we used a total of 106 mice (58 M/48 F) ranging in age from 3 to 17 weeks old. Because of different brain processing methodology and volume availability, sample number per Aβ measurement varied; exact numbers are stated in all figure legends. For analysis of Aβ accumulation in ECE-1-deficient mice (see Fig. 10), Ece1 (+/+) and (+/−) littermates from 2 whole litters were used, less one mouse used instead to measure ECE-1 activity. Additional mice from a separate litter were used for ECE-1 activity measurements. For intracerebroventricular injection experiments (see Fig. 11), mice were randomized by age and sex to receive either vehicle (saline) or phosphoramidon. On graphs, each dot represents data from a single animal. Littermate WT mice (n = 3), or humanized App knock-in mice, were used to establish specificity of immunoassays and, where applicable, average results are shown on graphs as a gray line. Statistical analysis, data graphing, linear regression, and nonlinear regression were performed with Prism 7 (GraphPad Software). Total Aβ42 and synaptosomal Aβ42 levels were best fit to a sigmoidal dose–response with variable slope, and dotted lines on graphs indicate the 95% CI. For column graphs, column bars represent mean ± SEM. For comparison of three or more groups, data were analyzed by ANOVA followed by Holm–Sidak's multiple comparisons post-tests. Multiplicity-adjusted p values are indicated on graphs. For experiments with two groups, data were analyzed using unpaired Student's t test with significance set at p < 0.05.
Results
Production and accumulation of Aβ in synaptic endosomes
To characterize the accumulation of Aβ species within synapses, we used synaptosomal preparations from TgCRND8 mice, a model that expresses human APP with two familial AD-associated mutations (KM670/671NL + V717F) that cause Aβ42 overproduction and very early-onset amyloid deposition (Chishti et al., 2001). Synaptosomes are artificial bodies formed by shearing neuronal processes during the homogenization of brain tissue in isosmotic buffer (Michaelis et al., 2017). They are composed of presynaptic and postsynaptic elements and have been extensively used to study synaptic physiology (Evans, 2015). Synaptosomes derived from postmortem AD brain have been shown to contain Aβ aggregates (Sokolow et al., 2012), so first we validated our experimental approach using TgCRND8 mice with abundant plaque pathology. As depicted in Figure 1a, material present in crude synaptosomal preparations was separated by sucrose gradient centrifugation and fractions of increasing density were collected. Synaptosome-containing fractions were identified by Western blotting for the synaptic vesicle marker synaptophysin (Fig. 1b). Soluble and membrane-associated Aβ species (henceforth referred to as “soluble”) were extracted from each fraction with RIPA buffer, and RIPA-insoluble material was further extracted with 6 m guanidine-HCl (GuHCl) for measurement of insoluble Aβ (Fig. 1c). RIPA-soluble fibrillar species were quantified by dot blot with the conformation-specific OC antiserum (Kayed et al., 2007; Glabe, 2008) (Fig. 1d). Using this method, we found that fractions containing purified synaptosomes were selectively enriched in APP, insoluble Aβ42, and soluble fibrils. All other fractions were virtually devoid of Aβ42, indicating that the Aβ42 present in crude synaptosomal preparations from plaque-bearing mice is contained within synaptosomes and is found in an aggregated or insoluble conformation.
Accumulation of Aβ species in synaptosomes from TgCRND8 mice. a, Diagram represents the preparation of synaptosomes from fresh mouse brains homogenized in 0.32 m sucrose buffer and purified in a 0.6-1.6 m continuous sucrose gradient. Fractions collected from top to bottom were pelleted and sequentially extracted with RIPA and 6 m GuHCl. b, Western blotting for synaptophysin identified synaptosomes in fractions 13-15, which were also enriched in APP. c, Insoluble Aβ42 (red line), measured by sandwich ELISA in GuHCl extracts, was specifically detected in synaptosomal fractions. Levels of RIPA-soluble Aβ42 monomers (black line) were low in synaptosomes, relative to the insoluble fraction. d, Soluble OC-immunopositive fibrils were only enriched in the synaptophysin-positive fractions. The experiment was repeated 3 times with consistent results.
Next, to characterize the specific sites of synaptic Aβ accumulation, synaptosomal vesicles were released by osmotic shock (Morciano et al., 2005) and separated by density gradient centrifugation. By Western blot, we evaluated the distribution of APP and its metabolites (Fig. 2a) along with several synaptic and endosomal markers. Immunoreactivity for both full-length APP and APP C-terminal fragments (CTFs) closely followed the distribution pattern of early endosomal marker Rab5, consistent with APP processing in Rab5-positive vesicles. In fractions of higher density, we observed a progressive increase in the ratio of CTF-β/CTF-α, coupled with detection of CTF-γ, indications of increased amyloidogenic APP processing. With increased fraction density, there was also increased immunoreactivity for the 16 kDa LC3B protein, a marker of autophagosomes, ALG-2 interacting protein X (Alix), a cytosolic adaptor protein that regulates cargo sorting to multivesicular bodies (MVBs), and charged multivesicular body protein 2B (CHMP2B), a component of the ESCRT-II complex involved in the formation of MVBs (Henne et al., 2013). Insoluble Aβ42 levels were also more concentrated in higher-density fractions, with maximal levels in fraction 24 coinciding with the highest immunoreactivity for endosomal markers Rab5, Rab7, LC3B, Alix, and CHMP2B. These high-density Aβ accumulating fractions were not enriched in markers of synaptic vesicles (synaptophysin), synaptic plasma membranes (syntaxin), or postsynaptic densities (PSD95). For Alix, in addition to full-length protein (100 kDa), we detected a putative degradation product of ∼42 kDa mostly enriched in fraction 24.
Preferential accumulation of intrasynaptic Aβ in Aβ-producing vesicles of endosomal origin. Synaptosomal vesicles were released from purified synaptosomes by osmotic shock and separated in a 5%-25% Optiprep gradient. a, Insoluble Aβ42, measured by sandwich ELISA, preferentially localized to high-density fractions enriched in markers of endosomes (Rab5, Rab7), MVBs (CHMP2B, Alix), and autophagosomes (LC3B). Less Aβ42 accumulated in lower-density fractions enriched in markers of synaptic vesicles (synaptophysin), synaptic membrane (syntaxin 1A), or postsynaptic densities (PSD95). b, Negative EM showed a representative synaptic marker-enriched fraction (#11), contained vesicles of ∼100 nm, while a fraction enriched in endosomal markers (#24) contained large vesicles similar in size to MVBs, autophagosomes, and amphisomes.
Alterations in intrasynaptic Aβ species before and during plaque development. a, Total levels of Aβ42 in TgCRND8 mice (red circles) were best fit to a sigmoidal curve with variable slope (Hill slope of 0.3918 ± 0.1044, R2 = 0.8435, red line, line also shown for reference in b–f). Elevations of total Aβ42 corresponded with the increase in number of plaques per tissue section (black circles). b, c, Levels of synaptosomal RIPA-soluble Aβ42 and Aβ40 showed a progressive decrease from week 3 to week 11, after which levels stabilized. d, Synaptosomal RIPA-soluble oligomeric Aβ showed more complex dynamics, decreasing between weeks 3 and 11, then progressively increasing from week 11 to week 16. e, A time-dependent accumulation of synaptosomal insoluble Aβ42 started at 6-8 weeks of age. f, Synaptosomal OC+ fibrillar species, quantified by image analysis of dot blots, mirrored the dynamics of oligomers, with an initial decrease up to weeks 9-11 followed by an elevation from week 11 to week 16. g, Representative OC dot blot image with numbers indicating age in weeks. b–f, Data were fit to a third polynomial curve (solid black line). Dotted lines indicate the 95% CI. Sample sizes for specific assays were as follows: a, Total Aβ42: n = 55, 28 M/27 F, plaques: n = 13, 6 M/7 F. b, c, RIPA-soluble Aβ42 and Aβ40: n = 66, 35 M/31 F. d, Oligomeric Aβ: n = 64, 31 M/33 F. e, OC+ fibrils: n = 54, 28 M/26 F. f, Insoluble Aβ42: n = 73, 39 M/34 F. Solid gray line indicates average value obtained from nontransgenic mice (n = 3) and represents background in human-specific Aβ40 ELISA (b) and background and/or endogenous rodent Aβ species in all other assays.
The size of vesicles contained within synaptosomes was further explored by negative-stain EM analysis of representative fractions. A synaptic marker-enriched fraction (#11) contained abundant ∼100 nm vesicles, consistent with the diameter of synaptic vesicles. In contrast, within the high-density fraction (#24) strongly enriched in endosomal/autophagic markers, we found larger vesicles consistent with the size of MVBs, autophagosomes, or amphisomes (Fig. 2b).
Early synaptic Aβ accumulation in TgCRND8 mice
Whereas our data support that synaptic endosomes can produce and accumulate insoluble Aβ42, it has been proposed that Aβ found in synapses represents Aβ aggregates that diffuse from plaques (Koffie et al., 2009; Brody et al., 2017). Therefore, to establish how synaptic Aβ accumulates in relation to extracellular pathology, we performed a longitudinal analysis of the levels of multiple Aβ species in crude synaptosomes beginning well before the onset of amyloid plaque deposition.
First, we generated a reference dataset for the progression of amyloid pathology in TgCRND8 mice bred in our colony, by measuring total brain Aβ42 levels extracted directly from brain homogenates with GuHCl. As shown in Figure 3a, total brain Aβ42 levels remained stable from 3 until ∼10 weeks of age, followed by a rapid accumulation phase (>10× increase) between 10 and 13 weeks, consistent with the reported onset of extracellular amyloid plaque formation in this model (Chishti et al., 2001). Thioflavin T staining confirmed that, after 10.5 weeks of age, extracellular amyloid plaques were present in all mice (Fig. 3a).
Following sequential RIPA/GuHCl extraction of synaptosomes, we found that soluble Aβ40 and Aβ42 monomers, Aβ oligomers, and OC-immunoreactive fibrillar species were readily detectable in synaptosomes from 3-week-old mice (Fig. 3b–d,f). All of these soluble Aβ species decreased between weeks 3 and 11, while insoluble Aβ42 progressively built up (Fig. 3e), with a maximum increase in the rate of accumulation between weeks 8 and 12, preceding the onset of plaques by ∼2 weeks. As plaque formation progressed beyond 12 weeks of age, soluble oligomers and fibrils increased (Fig. 3d,f), while levels of soluble Aβ40 and Aβ42 remained stable (Fig. 3b,c).
In interstitial fluid-enriched fractions (Fig. 4), prepared by gently homogenizing brains in isosmotic buffer without detergents and ultracentrifuged to remove EVs (S3 fraction depicted in Fig. 1a), we observed significant declines in all soluble Aβ species between weeks 3 and 11, closely matching those detected in synaptosomal extracts. After the appearance of plaques, extracellular Aβ levels became highly variable, potentially reflecting equilibrium between soluble and plaque-associated Aβ (Cirrito et al., 2003).
Age-dependent changes in extracellular Aβ species. In extracellular/interstitial fluid-enriched fractions of TgCRND8 mice before plaque deposition, there was a progressive and significant decline in levels of (a) Aβ40 (r2 = 0.435, p < 0.0001), (b) Aβ42 (r2 = 0.474, p < 0.0001), (c) Aβ oligomers (r2 = 0.1469, p = 0.0012), and (d) soluble OC+ fibrils (r2 = 0.0879, p = 0.031). e, Dot blot representative of those analyzed for OC quantification with the corresponding ages in weeks. After the onset of plaque deposition, Aβ levels followed different trends and were more variable. Data from pre-plaque mice (3-10.5 weeks old) were fit to a linear regression (solid line). Dotted lines indicate the 95% CIs. Sample sizes were as follows: a, b, Aβ40 and Aβ42: n = 71, 37 M/34 F. c, Oligomeric Aβ: n = 64, 31 M/33 F. d, OC+ fibrils: n = 73, 37 M/36 F.
To further confirm the presence of intrasynaptic Aβ aggregates at a pre-plaque age, we additionally used a sandwich ELISA method that captures and detects with protofibril selective antibodies (Colvin et al., 2017; Grover et al., 2023). Results showed the presence of soluble Aβ protofibrils in synaptosomes from 3-week-old mice (Fig. 5). Like other soluble intrasynaptic Aβ species, the levels of Aβ protofibrils decreased over the time period leading to the buildup of insoluble Aβ42.
Detection of synaptosomal protofibrillar Aβ before the onset of plaque formation. Protofibrillar Aβ42 species, measured by sandwich ELISA in RIPA-extracted synaptosomal preparations (n = 36, 23 M/13F), were detectable in 3-week-old mice. Levels decreased significantly (R2 = 0.3462, p < 0.0001) up to 8 weeks of age, when protofibrils were no longer detectable above average signal obtained with extracts from nontransgenic mice (solid gray line, n = 3).
Age-dependent accumulation of Aβ in exosomes. a, EVs, purified by density floatation, were highly enriched in the exosomal marker Alix in fractions 10-12. The exosomal fractions were also enriched in full-length APP and APP-CTFs. b, By dynamic light scattering, we confirmed a homogeneous population of small EVs in pooled fractions 10-12, with a diameter of ∼80.7 ± 20 nm (mean ± SD). c, Negative-stain EM showed that vesicles had a cup-like shape typical of exosomes. d, Total levels of Aβ42 in GuHCl-extracted exosomes progressively increased, starting at week 11 (n = 19, 12 M/7 F). e, Levels of fibrillar species in RIPA-extracted exosomes also started accumulating after 11 weeks of age (n = 17, 10 M/7 F). f, Representative dot blot examples for OC signal levels in 11-, 14- and 17-week-old brain-derived exosomes.
Accumulation of fibrillar Aβ in exosomes
Exosomes are EVs formed by the inward invagination of endosomal membranes; and, as such, they are carriers of membrane-bound Aβ produced in endosomes/MVBs (Rajendran et al., 2006; Sardar Sinha et al., 2018; Pacheco-Quinto et al., 2019) and may reflect the dynamics of intraneuronal Aβ accumulation. Using an additional set of brains, we isolated exosome-enriched EVs using a validated method that uses mild papain dissociation and sequential centrifugation to separate cellular material from EVs contained in the extracellular fluid. After further purification by density gradient ultracentrifugation (Perez-Gonzalez et al., 2017), exosome-containing fractions were characterized by Western blot, dynamic light scattering, and EM, and then the exosome-associated Aβ42 burden was determined in mice aged 4-17 weeks. In this isolation protocol, exosomes equilibrate at the 0.6/0.95 m sucrose interlayer (fractions 10-12, Fig. 6a), which was enriched with the exosomal marker Alix. Alix-enriched EV fractions were also enriched in APP and APP CTFs and contained particles with an average diameter of ∼80.7 ± 20 nm (mean ± SD), as measured by dynamic light scattering (Fig. 6b). The size range of the isolated EVs was consistent with that of exosomes, and not larger EVs, such as ectosomes (size range 100-1000 nm) or apoptotic bodies (size range 1000-5000 nm). Furthermore, the presence of the characteristic cup-shaped morphology of exosomes was confirmed by EM (Fig. 6c).
Aβ42 (directly extracted with GuHCl) was detectable in exosomes from very young mice and increased substantially between 11 and 17 weeks of age (Fig. 6d). Dot blot analysis of RIPA extracts confirmed the presence of OC-immunoreactive fibrillar species that built up over the same period (Fig. 6e,f), supporting that fibrillar Aβ accumulating in MVBs is eventually secreted in association with exosomes.
Formation and localization of intravesicular Aβ fibrils induced by protease inhibition in SH-SY5Y-APP cells
As our findings suggest that fibrillar Aβ accumulating in synaptic endosomes may have an endogenous origin, we turned to a cell culture model to further evaluate the propensity of Aβ to aggregate in Aβ-producing endosomes, and to determine whether altered proteostasis influenced this process. Using SH-SY5Y human neuroblastoma cells overexpressing WT APP (SH-SY5Y-APP), we have demonstrated that the level and state of aggregation/insolubility of Aβ produced in endosomal vesicles are dynamic and majorly controlled by the activity of phosphoramidon-sensitive proteases (Pacheco-Quinto and Eckman, 2013; Pacheco-Quinto et al., 2019). Phosphoramidon inhibits a subset of zinc metalloproteases, including Aβ-degrading enzymes ECE-1, ECE-2, neprilysin, and angiotensin-converting enzyme (Roques et al., 1993; Turner and Hooper, 2002). It does not inhibit another important Aβ-degrading metalloprotease, insulin-degrading enzyme (Farris et al., 2003). Using more selective inhibitors of neprilysin and angiotensin-converting enzyme, we have shown in previous studies that inhibition of these enzymes does not reproduce the effect of phosphoramidon treatment on Aβ levels in vitro or in vivo. More selective ECE inhibitors do, however, cause similar elevations in Aβ concentration, strongly suggesting that inhibition of ECE-1 and/or ECE-2 is at least partially responsible for the effect (Eckman et al., 2006; Pacheco-Quinto and Eckman, 2013).
Consistent with prior studies, treatment of SH-SY5Y-APP cells with phosphoramidon for 48 h led to intracellular accumulation of Aβ, measured by sandwich ELISA (Fig. 7a,b) and observed by immunocytochemistry (Fig. 7c) using an N-terminal Aβ antibody. The majority of Aβ42 accumulating in intracellular vesicles following phosphoramidon treatment remained in a RIPA-soluble state. To determine whether further disruption of intracellular proteolysis would drive Aβ aggregation, we treated cells with E-64, a cysteine cathepsin inhibitor that causes accumulation of APP CTF-α, -β, -γ, and other APP fragments (Pacheco-Quinto et al., 2019). By increasing accumulation of CTF-β, the immediate Aβ precursor, E-64 treatment can potentially lead to increased Aβ production in endosomal compartments also containing γ-secretase. The combination of E-64 and phosphoramidon treatment resulted in more intense Aβ immunoreactivity and the appearance of intracellular fibrils, detected by immunocytochemistry with OC antiserum (Fig. 7c). No immunostaining was observed in cells treated with E-64 alone, consistent with ELISA results showing no significant increase in intracellular Aβ42 concentration, and furthermore demonstrating that neither the N-terminal Aβ antibody nor OC antiserum cross-reacted with APP CTFs. Results of ELISA analysis of cells extracted sequentially with RIPA followed by GuHCl showed that the combination phosphoramidon plus E-64 treatment increased soluble intracellular Aβ42 concentration ∼30× compared with control cells (Fig. 7a). In the insoluble fraction (Fig. 7b), Aβ42 concentration was drastically increased >1000×, supporting that the OC+ immunofluorescent signal represented fibrillar Aβ.
Intracellular accumulation of fibrillar Aβ in SH-SY5Y-APP cells. Cells were treated for 48 h with 100 μm phosphoramidon (PA) and/or 100 μm E-64 and were extracted sequentially with RIPA followed by GuHCl. a, RIPA-soluble Aβ42 was significantly increased in cells treated with PA or E-64+PA, but not with E-64 alone. Combination E-64+PA treatment caused a further increase in Aβ42, compared with PA alone (F(3,8) = 705.9, p < 0.0001, ANOVA). b, RIPA-insoluble (GuHCl-soluble) Aβ42 was significantly increased only in cells treated with E-64+PA (F(3,8) = 666.4, p < 0.0001, ANOVA). Multiplicity-adjusted p values (Holm–Sidak's) are indicated on the graphs. c, Increased Aβ immunoreactivity with N-terminal Aβ antibody 33.1.1 was evident in PA and E-64+PA-treated cells, but OC-immunopositive fibrils appeared only in cells with the combined treatment. n = 3 for all bar graphs. Error bars indicate mean ± SEM. Scale bar, 20 µm. The experiment was repeated >3 times with consistent results.
The accumulation of OC+ fibrillar Aβ in endosomal vesicles was investigated using three different colocalization methods. First, the endocytic pathway was labeled by incubating cells with Lucifer yellow. Second, CD63+ vesicles, including MVBs, were labeled by transiently transfecting cells with CD63-EGFP. Third, Rab7+ vesicles, including late endosomes, were identified by immunofluorescence. In cells treated with phosphoramidon plus E-64, OC immunofluorescent signal was distributed in the endocytic pathway (Fig. 8a) and partially colocalized with CD63-EGFP fluorescence (Fig. 8b) and Rab7 immunofluorescence (Fig. 8c). Super-resolution Airyscan confocal images showed OC immunofluorescence arranged in a morphology highly reminiscent of intracellular vesicles, suggesting that intravesicular OC staining may localize to the rim of endosomal structures (Fig. 8d).
Association of intracellular Aβ with endosomal membranes. Immunofluorescent staining with OC anti-fibrils antiserum partially colocalized with (a) Lucifer yellow, labeling the endocytic pathway, (b) GFP-CD63, marker of multivesicular endosomes, and (c) Rab7, marker of late endosomes. In overlay figures, OC immunofluorescence is colored cyan blue, endosomal markers red, and nuclei yellow. d, Super-resolution imaging of OC-positive fibrillar Aβ (cyan) visualized using Zeiss Airyscan confocal microscopy shows fluorescent signal presumably localized to the rim of endosomal vesicles. Scale bars: a–c, 10 µm; d, 2 µm.
Intracellular Aβ accumulation in SH-SY5Y-APP cells with genetic deletion of ECE1: reproduction of the effect of phosphoramidon treatment
Without an inhibitor that selectively inhibits ECE-1 versus ECE-2, it has not been possible to determine the relative contribution of each ECE enzyme to Aβ catabolism in our cell model. To further investigate this, we knocked out ECE1 in SH-SY5Y-APP cells and determined the effect on basal Aβ levels and response to phosphoramidon and E-64 treatment. Compared with WT SH-SY5Y-APP cells, ECE1 KO cells had significantly elevated levels of both secreted and RIPA-soluble intracellular Aβ42 and were insensitive to phosphoramidon treatment (Fig. 9a,b). The accumulation of intracellular Aβ in ECE1 KO cells without protease inhibitor treatment was detectable by immunocytochemistry (Fig. 9d); E-64 treatment resulted in a further increase in N-terminal Aβ immunoreactivity and the appearance of OC-positive signal, indicative of Aβ fibrillization. ELISA measurement of RIPA-insoluble (GuHCl-extracted) Aβ42 revealed a 20× increase in insoluble Aβ42 in E-64-treated cells compared with untreated ECE1 KO cells (Fig. 9c), consistent with the immunostaining.
KO of ECE1 in SH-SY5Y-APP cells reproduces the effect of phosphoramidon treatment. WT SH-SY5Y-APP cells and SH-SY5Y-APP ECE1 KO cells were treated with phosphoramidon (PA, 100 µm) or E-64 (100 µm) for 48 h. Conditioned medium was collected for measurement of extracellular Aβ by sandwich ELISA. Cells were sequentially extracted with RIPA followed by GuHCl for measurement of intracellular Aβ. a, b, Basal levels (control [C]) of extracellular (a) and intracellular Aβ42 (b) were elevated in ECE1 KO cells compared with WT SHSY5Y-APP cells and did not increase in response to PA, indicating that ECE1 is the major Aβ-degrading protease in SH-SY5Y cells (extracellular: F(3,8) = 25.42, p = 0.0002, ANOVA; intracellular: F(3,8) = 20.87, p = 0.0004, ANOVA; multiplicity-adjusted p values indicated on the graph). c, E-64 treatment alone of SH-SY5Y-APP ECE1 KO cells produced a significant elevation of RIPA-soluble Aβ42 and elevation of RIPA-insoluble (GuHCl-extracted) Aβ42. Pairwise comparisons between untreated and treated cells made using unpaired t test. d, Intracellular accumulation of Aβ in ECE1 KO cells was confirmed by immunocytochemistry using N-terminal Aβ antibody 33.1.1. Aβ staining was substantially increased in E-64-treated cells. OC-immunopositive fibrillar species were detectable only in cells treated with E-64. Scale bar, 20 µm.
Endogenous synaptic and extracellular Aβ accumulation in mice with partial Ece1 deficiency. a, Ece1 (+/−) mice (n = 3, 2 M/1 F) had a 40% reduction in synaptosomal Ece1 activity compared with Ece1 (+/+) controls (n = 4, 2 M/2 F), as measured with a big ET-1 conversion assay. Gray line indicates activity in the presence of ECE inhibitor phosphoramidon. b, Synaptosomal Aβ42 was significantly elevated in Ece1 (+/−) mice. c, The increase in synaptosomal Aβ40 was not statistically significant. d, e, Both Aβ42 and Aβ40 were significantly elevated in extracellular-enriched extracts. c, e, Gray lines indicate average background signal from humanized App knock-in mice (n = 3) in the rodent-specific Aβ40 ELISA. Background of these samples in the rodent-specific Aβ42 ELISA was zero. Sample sizes for all Aβ measurements were Ece1 (+/+) mice: n = 8, 8 M/0 F, Ece1 (+/−) mice: n = 7, 5 M/2 F, aged 51 weeks. Pairwise comparisons between genotypes were made with unpaired t test.
Sensitivity of synaptosomal Aβ40 and Aβ42 to degradation by ECE-related proteases in TgCRND8 mice, pre- and post-amyloid deposition. Mice were injected intracerebroventricularly with phosphoramidon (PA) or saline vehicle. Crude synaptosomes were isolated 16 h later and extracted sequentially with RIPA and GuHCl for measurement of Aβ by sandwich ELISA. a, RIPA-soluble synaptosomal Aβ40 was elevated similarly in WT mice (WT, n = 6, 5 M/1 F) and in TgCRND8 mice pre-amyloid deposition (n = 6, 2 M/4 F, aged 8-10 weeks) or post-amyloid deposition (n = 7, 3 M/4 F, aged 16-17 weeks) (F(2,16) = 0.33, p = 0.72, ANOVA). b, There was significantly less PA-induced accumulation of RIPA-soluble synaptosomal Aβ42 in young TgCRND8 mice compared with WT mice, and a further reduction in accumulation in TgCRND8 mice after plaque deposition (F(2,15) = 24.75, p < 0.0001, ANOVA). Multiplicity-adjusted p values (Holm–Sidak's) are indicated on the graph. c, PA treatment did not significantly affect RIPA-insoluble (GuHCl-solubilized) Aβ42 in either age group. Data are plotted for each PA-treated group, expressed as a percent of the mean of age-matched vehicle controls which included WT (n = 6, 4 M/2 F), TgCRND8 pre-deposition (n = 6, 2 M/4 F), and TgCRND8 post-deposition (n = 7, 3 M/4 F). Lines indicate mean ± SEM.
Synaptic Aβ accumulation in ECE-1-deficient mice
Homozygous deletion of Ece1 in mice results in an embryonic lethal phenotype (Yanagisawa et al., 1998). Heterozygous Ece1 KO mice, while apparently healthy, have increased endogenous Aβ levels in whole-brain homogenates (Eckman et al., 2003). To determine whether ECE-1 deficiency affects endogenous Aβ accumulation at the synaptic level, we prepared synaptosomes from heterozygous Ece1tm1a(KOMP)Wtsi mice that contain a “knock-out first” reporter-tagged insertion allele. Compared with littermate controls, Ece1 (+/−) mice had a ∼40% reduction in phosphoramidon-sensitive synaptosomal ECE-1 activity (Fig. 10a) and significantly increased endogenous synaptosomal Aβ42 concentration (Fig. 10b). As seen with ECE1 KO cells in vitro, secreted Aβ42 and Aβ40 were also significantly elevated in extracellular (interstitial fluid)-enriched extracts from Ece1 (+/−) mice (Fig. 10d,e).
Loss of Aβ42 proteolysis by ECE-related proteases in TgCRND8 mice
Since genetic deletion of ECE-1 triggers accumulation of endosomal/synaptosomal Aβ in vitro and in vivo, we next investigated whether the ability of ECEs to degrade intrasynaptic Aβ might be compromised in Aβ-overproducing TgCRND8 mice with aging. As previously reported and shown in Figure 11, global inhibition of ECE-related proteases by intracerebroventricular administration of phosphoramidon causes large, rapid increases in RIPA-soluble Aβ levels, in both WT mice and in young (pre-plaque) TgCRND8 mice. Whereas the magnitude of the phosphoramidon-induced increase in synaptosomal Aβ40 was nearly identical, the increase in Aβ42 was comparatively lower in TgCRND8 mice versus WTs (Pacheco-Quinto et al., 2019). To determine whether the apparent resistance of intrasynaptic Aβ42 to degradation by ECE-related proteases was age-dependent and worsened with the development of amyloid pathology, we next compared the effect of phosphoramidon treatment on intrasynaptic Aβ42 levels at two time points: 8-10 weeks, before extracellular amyloid deposition and when insoluble synaptosomal Aβ42 was beginning to build up; and 16-17 weeks, when insoluble synaptosomal Aβ42 concentration was near maximal and stabilized, and mice had abundant amyloid plaques.
As shown in Figure 11a, we saw similar magnitude increases in soluble synaptosomal Aβ40 in WT and in TgCRND8 mice at pre- and post-deposition ages, indicating that phosphoramidon-sensitive Aβ-degrading activity was not compromised by mutant APP overexpression, age, or amyloid accumulation, and that Aβ40 remained in a conformation sensitive to proteolysis by ECE-related enzymes. In contrast, RIPA-soluble synaptosomal Aβ42 levels increased to a lesser extent on phosphoramidon treatment in young APP-overexpressing mice compared with WT mice (Fig. 11b). By 16-17 weeks, phosphoramidon produced only minimal effects on RIPA-soluble Aβ42. The RIPA-insoluble pool of synaptosomal Aβ42 was unaffected by inhibition of ECE-related protease activity at either age in TgCRND8 mice (Fig. 11c).
Discussion
This study provides evidence that Aβ accumulating in synapses during the progression of amyloid pathology is produced locally, within endosomal compartments. The first indication that Aβ originates endogenously at synaptic terminals is the finding that, in subfractionated synaptosomes, insoluble Aβ42 was concentrated in vesicle fractions enriched in APP, CTF-β, and CTF-γ, markers of Aβ biosynthesis. Aβ-containing fractions were also enriched in multiple markers of endosomes, autophagosomes, amphisomes, and more specifically of MVBs. Interestingly, an ∼42 kDa band immunoreactive for Alix was detected both in the fraction containing the highest Aβ level (fraction 24, Fig. 2a) and in exosomes (Fig. 6a) and, by EM, we observed large (>0.5 µm) vesicles in the fraction with highest Aβ burden (Fig. 2b), all additional evidence suggestive of selective enrichment of MVBs in fraction 24. Morphologic characterization of ultrathin cryosections by EM may further aid in identifying the type of endosomal vesicle accumulating Aβ.
As proof that synaptic Aβ42 accumulation does not require endocytosis of plaque derived Aβ, we discovered that aggregated and insoluble Aβ was present in synaptosomes well before the emergence of extracellular amyloid pathology. In weanling mice at 3 weeks of age, synaptosomes were already enriched in soluble monomers, oligomers and fibrillar Aβ. Over the next 2 months, there was a reduction in soluble intrasynaptic Aβ species corresponding with a buildup of insoluble Aβ42, which accumulated precipitously in the weeks preceding amyloid plaque formation (∼10× increased concentration between 7 and 10 weeks). The insoluble Aβ42 content of intrasynaptic vesicles then began to saturate, coinciding with increased secretion of exosome-associated fibrillar Aβ and the emergence and progression of amyloid plaques.
Although our study of intraneuronal Aβ accumulation in vivo was focused on synaptosomes, it is likely that similar Aβ42 accumulation may occur in endosomes in the neuronal soma. Recently, Nixon and colleagues published an immunohistochemical study demonstrating defective acidification of the E/L network in APP-overexpressing mice well before the onset of extracellular plaques (Lee et al., 2022). They further described the formation of fibrillar Aβ-filled membrane blebs of autophagic vacuoles, termed PANTHOS, surrounding neuronal nuclei, causing neuronal loss, and potentially serving as seeds for plaque formation. PANTHOS are reportedly detectable in TgCRND8 mice at ∼8 weeks of age (Lee et al., 2022), corresponding closely to the onset of rapid insoluble synaptosomal Aβ42 accumulation we observed. Thus, our approach, using biochemical extraction and analysis of multiple Aβ species by ELISA and dot blot, complements immunohistochemical evidence of early intraneuronal Aβ42 aggregation preceding plaque formation in AD model mice.
With loss of Aβ42 solubility in synaptic endosomes appearing to be one of the earliest events in the development of amyloid pathology, we questioned whether disturbances in intracellular Aβ clearance pathways could be a contributing factor, focusing on the unique role of ECEs in degrading monomeric Aβ in the same compartments where the peptide is produced (Eckman et al., 2001; Pacheco-Quinto and Eckman, 2013). Comparing the effects of phosphoramidon-mediated global inhibition of ECE-related proteases on intrasynaptic Aβ content in WT mice and in TgCRND8 mice at pre- and post-plaque deposition ages, we found that the efficiency of ECEs to degrade Aβ40 (less prone to aggregation than Aβ42) was similar, indicating that enzymatic activity per se is not compromised in TgCRND8 mice. However, we discovered an age-dependent impairment in Aβ42 removal by ECE-related proteases in TgCRND8 mice. Since ECEs can degrade only monomeric Aβ in the lumen of endosomes, the most likely explanation for these results is that, in synaptic endosomes, Aβ42 specifically becomes resistant to proteolysis as amyloid pathology develops, either by quickly acquiring a nonmonomeric conformation after it is produced or by binding to membranes. This interpretation is supported by additional data showing that the insoluble pool of Aβ42 was not affected by phosphoramidon treatment. Because of the large synaptic endosomal production of Aβ42 in this mutant APP model, it is likely that nascent Aβ42 rapidly aggregates before ECEs can effectively control its accumulation. Mutations linked to familial early-onset AD that increase Aβ42 production and accumulation in endosomes may cause pathology in humans by a similar mechanism.
Using both in vitro and in vivo KO models, our study confirms that, among the phosphoramidon-sensitive metalloproteases, ECE-1 is indeed a major player in intraneuronal Aβ homeostasis. In vivo, ECE-1 activity localizes to synapses and regulates endogenous intrasynaptic Aβ42 concen tration. In addition, in a cell culture model of intracellular Aβ aggregation, ECE-1 activity prevents endosomal Aβ fibrillization, even under conditions of excess Aβ production. These results, however, do not exclude a role for other ECE-related proteases in preventing intraneuronal Aβ accumulation in brain. Like Ece1, Ece2 KO mice have elevated levels of endogenous Aβ in whole-brain extracts (Eckman et al., 2003). Although both enzymes are active in the endosomal pathway and hydrolyze Aβ optimally at pH ∼5.5, their activities in brain are nonredundant, in part because of expression by different neuronal populations (Pacheco-Quinto et al., 2016).
Based on evidence presented in this report and in previous studies, we propose that dysfunction of ECE-1 or ECE-2 is relevant to the pathogenesis of late-onset AD. With the endosomal pathway recognized as a hub for many AD-risk associated genes, any factor altering endosomal trafficking or pH could conceivably impact ECE-mediated Aβ homeostasis and cause intraneuronal/intrasynaptic formation of Aβ aggregates. Direct genetic evidence of altered ECE function in AD is limited, but several reports show modest linkage of ECE1 polymorphisms with AD risk (Funalot et al., 2004; Scacchi et al., 2008; Jin et al., 2009; Hamilton et al., 2012), and a recent study described a rare loss-of-function mutation in ECE2 associated with familial late-onset AD (Liao et al., 2020). A report describing the highly polymorphic nature of CpG-CA repeats in the ECE1c promoter, with distinct repeat length distribution in AD patients compared with controls (Y. Li et al., 2012), is particularly intriguing in context of our data showing that partially decreased ECE-1 expression results in significantly increased endogenous synaptosomal Aβ42 accumulation.
In conclusion, our results provide evidence that pathologic Aβ species affecting synaptic function may form within synaptic endosomes early in the disease process, and perhaps independently from aggregation of extracellular Aβ. One fundamental implication of this finding is that Aβ-targeting immunotherapies may not effectively modulate this pool of intrasynaptic Aβ. Understanding the relative contributions of extracellular and intrasynaptic Aβ aggregates to cognitive dysfunction in AD is critical for the development of more effective therapies.
Footnotes
This work was supported in part by the Vanech Family Foundation to E.A.E. and J.P.-Q.; University of Missouri FastTrack Award to M.R.N.; and National Institutes of Health R15GM119070 to M.R.N. S.S.-D. was supported by Cure Alzheimer's Foundation Grant CAF-211540-02, Swedish Research Council International Postdoctoral Fellowship 637-2013-503/D0050301, and Leon Levy Foundation Fellowship in Neuroscience. We thank Hui Peng (Biomedical Research Institute of New Jersey) for technical support; and Luis Marcelino (Biomedical Research Institute of New Jersey) for help with animal husbandry.
The authors declare no competing financial interests.
- Correspondence should be addressed to Javier Pacheco-Quinto at pacheco-quinto{at}brinj.org or Elizabeth A. Eckman at lizeckman{at}brinj.org
