De novo formation of familial Alzheimer’s disease-relevant Aβ strains occurs exclusively in the human brain

Abstract

In familial Alzheimer’s disease (FAD), mutations in the amyloid precursor protein gene (APP) accelerate the aggregation of Aβ peptide, leading to deposition of amyloid plaques within the brain. Conformational variants of Aβ aggregates (“Aβ strains”) possess distinct pathological properties and may contribute to disease heterogeneity. While the Swedish APP mutation produces wild-type (WT) Aβ, the Arctic APP mutation generates E22G-mutant Aβ. To clarify whether different FAD mutations are sufficient to specify the formation of distinct Aβ strains, we performed Aβ seeding studies in App^NL−F knock-in mice. Aβ aggregates purified from Swedish and Arctic FAD human brains induced distinct neuropathological signatures, characterized by differences in the morphology and localization of the seeded Aβ deposits, which were preserved upon serial passage. Neither synthetic E22G-mutant Aβ aggregates nor Aβ aggregates isolated from the brains of transgenic mice producing E22G-mutant Aβ recapitulated the phenotype induced by Arctic FAD brain-derived Aβ aggregates upon injection into App^NL−F mice. Similarly, neither synthetic WT Aβ aggregates nor Aβ aggregates from Swedish APP-mutant transgenic mice reproduced the seeding behavior of Swedish FAD brain-derived Aβ aggregates. This suggests that it is the structure of Aβ aggregates rather than the Aβ sequence that dictates seeding activity. We conclude that de novo formation of FAD-relevant Aβ strains occurs exclusively in the human brain, implying the existence of human brain-specific cofactors that modulate Aβ aggregate structure and seeding activity.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s40478-026-02390-z.

Introduction

A neuropathological hallmark of Alzheimer’s disease (AD) is the presence of extracellular plaques containing aggregated β-amyloid (Aβ) peptide. In AD brains, parenchymal Aβ deposits can exist as several different morphological variants spanning from diffuse inclusions to dense-core plaques that react with amyloid-binding dyes such as Thioflavin S (ThS) [12]. Aβ deposits are also commonly found within the lining of cerebral blood vessels, which is referred to as Aβ cerebral amyloid angiopathy (Aβ-CAA). The Aβ peptide is produced by sequential endoproteolytic cleavage of the amyloid precursor protein (APP) by β-secretase and γ-secretase [84]. Cleavage of APP by γ-secretase is heterogeneous, leading to the production of Aβ C-terminal isoforms with lengths ranging from 38–43 amino acids. Although the role of Aβ in AD pathogenesis remains to be fully resolved, a wealth of information implies that cerebral Aβ deposition is an initiating event [68].

While most AD cases are sporadic, a small proportion (~ 5%) are early-onset familial AD (FAD) cases caused by mutations in the genes that encode APP (APP) or components of the γ-secretase complex. APP mutations act by either increasing production of all Aβ isoforms, increasing relative proportions of the more amyloidogenic Aβ42/43 isoforms, or by generating a mutant Aβ peptide that is more prone to aggregation. For example, the Swedish APP mutation (K670N/M671L), which occurs just N-terminal to the start of the Aβ sequence within APP, generates higher levels of wild-type (WT) Aβ by increasing cleavage of APP by β-secretase [8, 18, 45]. In contrast, the Arctic APP mutation (E693G) occurs within the sequence of Aβ and promotes the formation of toxic Aβ protofibrils containing E22G-mutant Aβ peptide [32, 46, 48]. While the Swedish mutation produces a neuropathological signature similar to that observed in sporadic AD, Aβ deposits in individuals with the Arctic mutation are morphologically distinct and react poorly with amyloid-binding dyes [3, 6, 25, 31, 36, 56]. Transgenic or knock-in mice that express human APP containing one or more FAD-associated mutations are frequently used as models of cerebral Aβ accumulation [65].

Aβ aggregates are “prion-like” in that they can act as seeds that recruit and incorporate monomeric Aβ into a growing fibrillar aggregate [5, 24, 83]. Fragmentation of larger Aβ fibrils into smaller seeds allows Aβ aggregates to spread within and to the brain through the process of templated conformational conversion. Indeed, intracerebral or peripheral injection of susceptible APP-overexpressing mice with pre-formed Aβ aggregates induces cerebral Aβ deposition and the spreading of Aβ pathology [13, 14, 26, 42–44, 73, 81]. Human and mouse brain-derived Aβ seeds are more potent inducers of cerebral Aβ deposition than in vitro-generated Aβ aggregates [72, 73], possibly due to the absence of brain-specific cofactors [49]. The size of Aβ aggregates can also influence seeding behavior, with soluble Aβ assemblies exhibiting greater specific seeding activity than larger insoluble fibrillar species [16, 30].

It has been hypothesized that the clinical and pathological heterogeneity observed in AD patients may be explained by the presence of distinct “strains” of Aβ aggregates within the brain [33]. Each individual protein aggregate strain is associated with a unique structure (fold) of the same protein monomer within a fibrillar assembly [66]. There is considerable evidence that Aβ can misfold and assemble into multiple different aggregate structures, some of which have been shown to induce distinct pathological phenotypes upon injection into an animal model [9–11, 17, 19, 37, 40, 41, 52–54, 59, 60, 72, 77]. Furthermore, cryo-electron microscopy (cryo-EM) studies have revealed that Aβ42 aggregates isolated from sporadic AD brains exist as either type I, type II, or type III structures, or mixtures thereof [78, 85].

How FAD-causing mutations, particularly those that alter the primary sequence of Aβ, affect Aβ strain generation and Aβ seeding activity remains incompletely understood. Atypical types of cerebral Aβ deposits are often present in FAD cases, and there is evidence that Aβ fibrils in FAD brains can be structurally distinct from those found in sporadic AD brains [21, 69, 80]. Previously, it was shown that brain extracts from Arctic and Swedish FAD patients induce distinct Aβ neuropathological signatures upon injection into APP23 transgenic mice, suggesting that E22G-mutant and WT Aβ assemblies may constitute distinct strains of Aβ aggregates [77]. To clarify whether the Arctic and Swedish FAD mutations are sufficient to generate distinct Aβ strains, we performed in vivo seeding studies in App^NL−F knock-in mice using brain-derived and synthetic Aβ aggregates. While Aβ aggregates purified from Arctic and Swedish FAD brains reproducibly induced distinct neuropathological signatures upon injection into mice, neither synthetic Aβ aggregates nor Aβ aggregates produced in the brains of transgenic APP mouse models recapitulated the phenotypes observed following seeding with the corresponding human FAD brain-derived Aβ aggregates. This suggests that APP and Aβ mutations alone are insufficient to generate aggregate structures with seeding properties associated with FAD brain-derived Aβ.

Methods

Human and mouse tissue samples

Human FAD and control (non-neurologic) frozen brain samples were obtained from brain banks located at the Karolinska Institutet (Ethics approval numbers 2011/962–31/1 and 2023–07255-02) or Uppsala University (Ethics approval numbers 2005–103 and 2021–04356). Sample details are provided in Table 1. The Swe1 and Arc1 cases are the same as those used in our previous study [77]. Research on human brain tissue was performed according to an ethics protocol approved by the University of Toronto (00038879: “Use of human tissue for research on neurodegenerative diseases”). For the studies involving transgenic mouse-derived Aβ aggregates, frozen hemibrains from a 21-month-old tg-APP_Swe and a 19-month-old tg-APP_ArcSwe mouse were utilized. All human and mouse brain samples were homogenized in nine volumes of chilled phosphate-buffered saline (PBS) using CK14 soft tissue homogenizing tubes and the MiniLys homogenizer (Bertin Technologies). The 10% (w/v) brain homogenates were then aliquoted and stored at -80 °C. For experiments requiring detergent-extracted brain homogenate, nine volumes of 10% (w/v) brain homogenate were mixed with one volume of 10X detergent buffer (5% (v/v) Nonidet P-40, 5% (w/v) sodium deoxycholate), then incubated on ice for 20 min with vortexing every 5 min. Afterward, the samples were centrifuged at 1,500 × g at 4 °C for 5 min and the supernatant was stored at -80 °C.

Table 1.

FAD patient brain sample details

Sample	FAD type1	Sex	Age (years)	Brain region	Source2	Amount of purified Aβ inoculated per AppNL−F mouse
Ctrl	Healthy control	Male	63	Temporal cortex	Uppsala	–
Swe13	Swedish	Female	61	Temporal cortex	Uppsala	100 ng
Swe2	Swedish	Female	62	Temporal cortex	Karolinska	50 ng
Swe3	Swedish	Male	62	Temporal cortex	Karolinska	50 ng
Arc13	Arctic	Male	64	Temporal cortex	Uppsala	100 ng
Arc2	Arctic	Female	66	Temporal cortex	Karolinska	50 ng

¹Samples are from FAD patients with either the Swedish (K670N/M671L) or Arctic (E693G) mutations in APP, or a healthy age-matched individual

²Samples were obtained from brain banks located at Uppsala University or the Karolinska Institute in Sweden

³The Swe1 and Arc1 cases are the same as those used in our previous study [77]

Purification of Aβ fibrils

Aβ fibrils were purified from human and mouse brain samples according to a previous protocol [73]. Brain homogenate (10% (w/v)) was diluted 1:1 in citrate lysis buffer (20 mM sodium citrate pH 6, 2% (w/v) Triton X-100, 274 mM NaCl, 2 mM ethylenediaminetetraacetic acid (EDTA)) for a total volume of 3 mL and then incubated on ice for 30 min. The brain homogenate was split into two tubes, where 0.6 mL of 60% (w/v) iodixanol (Sigma Aldrich, #D1556) was added to 1.4 mL of brain homogenate. Each sample was used to prepare an iodixanol gradient with the following layers from top to bottom in 5 mL open-top centrifuge tubes (Beckman Coulter, #326819): 2 mL of sample, 850 µL of 30% (w/v) iodixanol, and 850 µL of 35% (w/v) iodixanol. The tubes were capped with mineral oil and the gradients were ultracentrifuged at 60,000 × g for 20 min in an MLS-50 rotor (Beckman Coulter) at 4 °C. The top layer and the lipid-containing layer were removed, then the next two layers were collected from both tubes. The combined sample was diluted 1:1 with citrate buffer (10 mM sodium citrate pH 6, 137 mM NaCl, 1 mM EDTA) and layered into a second pair of 5 mL centrifuge tubes containing iodixanol gradients with the following layers, from top to bottom: 2 mL of brain homogenate, 850 µL of 26% (w/v) iodixanol, and 850 µL of 35% (w/v) iodixanol. The tubes were capped with mineral oil and centrifuged for 40 min at 60,000 × g in an MLS-50 rotor at 4 °C. Afterward, the layer of mineral oil and the top layer were discarded, and the next two layers were collected from both tubes. The combined sample was diluted 1:1 in citrate buffer and divided into 1 mL aliquots in microcentrifuge tubes (Beckman Coulter, #357448), then centrifuged at 21,000 × g for 40 min at 4 °C in a TLA-55 rotor (Beckman Coulter). The supernatant was discarded, and the pellet was resuspended in Tris buffer (10 mM Tris–HCl pH 8.3, 1.71 M NaCl, 1% (w/v) zwittergent 3–14, 1 mM EDTA, 1 mM dithiothreitol (DTT)), then centrifuged at 21,000 × g for 30 min at 4 °C. The supernatant was discarded, and the pellet was resuspended in 50 µL of TMS buffer (50 mM Tris–HCl pH 7.8, 100 mM NaCl, 10 mM MgCl₂). The samples were collected into one tube, then stored overnight at -80 °C. The following day, the sample was treated with 150 units of benzonase (MilliporeSigma, #70,746) for 2 h at 37 °C with shaking at 600 rpm. After centrifuging the sample at 21,000 × g for 30 min at 4 °C, the sample was incubated with proteinase K (PK; Thermo Scientific, #EO0491) at a final concentration of 50 µg/mL for 1 h at 37 °C with shaking at 600 rpm. The reaction was stopped with phenylmethylsulfonyl fluoride (PMSF) at a final concentration of 1 mM and the sample was adjusted to 1.71 M NaCl, then loaded on top of 100 µL of 1 M sucrose in a microcentrifuge tube. The sample was centrifuged at 21,000 × g for 30 min at 4 °C. The supernatant of the resulting sample was discarded, and the pellet was resuspended in 100 µL of 0.1 M sodium acetate. The sample was centrifuged at 21,000 × g for 30 min at 4 °C, then the supernatant was discarded and the pellet was resuspended in 100 µL of dH₂O. Purified Αβ was stored at -80 °C.

Partial purification of Aβ aggregates

Detergent-extracted brain homogenates were digested with PK at a concentration of 50 μg/mL for 1 h at 37 °C with shaking at 600 rpm. For the second passage experiments, 1 mL of detergent-extracted mouse brain homogenate from the initial passage of Arctic and Swedish FAD Aβ aggregates in App^NL−F mice was digested. As a negative control, brain homogenate from an App^NL−F mouse previously injected with dH₂O was processed similarly. For the tg-APP_Swe and tg-APP_ArcSwe experiments, 334 µL of brain homogenate was used. Digestions were stopped by adding PMSF to a concentration of 2 mM and then sarkosyl (Sigma, #61747) was added to a final concentration of 2% (v/v). Samples were ultracentrifuged for 1 h at 4 °C in a TLA-55 rotor and then pellets were resuspended in 300–400 µL inoculation diluent buffer (5% (w/v) bovine serum albumin in PBS), sonicated for 10 min in a water bath sonicator, and then stored at -80 °C until inoculation.

Characterization of purified Aβ aggregates

Analysis of purified Aβ involved diluting the purified samples 1:10 into 1X Bolt LDS Sample Buffer containing 2.5% β-mercaptoethanol, followed by analysis by SDS-PAGE and silver staining (Thermo Scientific, #PI24612). The total Aβ content in purified Aβ samples was quantified using an Aβ(1-x) ELISA (IBL International, #JP27729). The purified samples were diluted 1:4 in dH₂O to a total volume of 20 µL. After adding 40 µL of chilled formic acid, the samples were sonicated for 5 min in a water bath sonicator and then ultracentrifuged at 100,000 × g for 1 h at 4 °C. A portion of the supernatant (10 µL) was diluted 1:20 in neutralization buffer (1 M Tris–HCl, 500 mM sodium phosphate) and then analyzed by ELISA. Because the Aβ(1-x) ELISA did not recognize E22G-mutant Aβ present in the Arctic FAD samples, levels of Aβ in these samples were determined by comparison to known amounts of Aβ from the Swedish FAD samples following separation by SDS-PAGE and silver staining. For negative-stain transmission electron microscopy, purified Aβ samples from the Swe1 and Arc1 FAD cases were placed onto a formvar/copper grid and then excess sample was removed using filter paper. Phosphotungstic acid (1% w/v) was added and incubated for 2 min. Samples were then examined using a Hitachi H7000 transmission electron microscope.

Generation and characterization of synthetic Aβ aggregates

Synthetic WT and E22G-mutant Aβ(1–40) and Aβ(1–42) peptides were purchased from Bachem (catalog numbers H-1194.0500, H-6694.0500, H-1368.0500, and H-6124.0500). The peptides were dissolved in hexafluoroisopropanol (HFIP), aliquoted, and then HFIP was allowed to evaporate overnight at 22 °C. Dried peptides were stored at -80 °C. Peptides were dissolved in 20 μL of DMSO, diluted with 480 µL of phosphate buffer (10 mM sodium phosphate pH 7.4), quantified using a NanoDrop spectrophotometer, and then further diluted using phosphate buffer. To generate Aβ aggregates, diluted Aβ samples were incubated for 72 h at 37 °C with constant shaking (900 rpm). For biochemical characterization, Aβ aggregates were made at a concentration of 5 µM. For seeding studies in App^NL−F mice, Aβ aggregates were formed at a concentration of 50 µM. Aggregates were then aliquoted and stored at -80 °C. PK digestion, conformational stability, and dye-binding assays on WT and E22G-mutant Aβ(1–40) aggregates were performed as previously described [62].

Inoculation of App^NL−F knock-in mice

Homozygous App^NL−F knock-in mice on a C57BL/6 background were provided by Takashi Saito and Takaomi Saido [63]. Mice were maintained on a 12 h light/12 h dark cycle and were given unlimited access to food and water. All experiments used approximately equal numbers of male and female animals. Once the mice reached 5–6 weeks of age, intracerebral inoculations were performed under isoflurane anesthesia. Prior to inoculation, all samples were sonicated in a water bath sonicator for 10 min. Each mouse was injected into the right parietal lobe at a depth of 3 mm with 30 µL of sample using a tuberculin syringe containing a 27-gauge 0.5-inch needle (BD Biosciences, #305945). Mice injected with purified FAD brain-derived Aβ received either 50 or 100 ng of Aβ aggregates. As a control for the FAD inoculations, mice were inoculated with an equivalent volume of sample from a control case that was purified using the same protocol. Mice injected with synthetic Aβ received 1 µg of Aβ aggregates diluted in dH₂O or dH₂O alone as a negative control. Mice injected with partially purified Aβ aggregates received 30 μL of sample prepared as described above. Mice injected with purified Aβ for the second passage experiments received 0.08 ng of Aβ aggregates. Inoculation experiments were performed in accordance with guidelines set by the Canadian Council on Animal Care under an animal use protocol (#4263.23) approved by the University Health Network Animal Care Committee. Inoculated App^NL−F knock-in mice, or age-matched uninoculated control mice, were euthanized by transcardiac perfusion with 0.9% saline at 6 months post-inoculation. Brains were removed and then bisected parasagittally. The left hemispheres of the brains were stored at -80 °C for biochemical experiments while the right hemispheres were fixed in 10% neutral buffered formalin and embedded in paraffin for immunohistochemical staining.

Neuropathological analysis of Aβ deposition

The paraffin-embedded samples were cut into sagittal sections with a thickness of 5 μm, deparaffinized in xylene, then rehydrated through sequential 5-min incubations in 50% xylene in ethanol, 100% ethanol, 95% ethanol, 75% ethanol, 50% ethanol, and water. Prior to staining with Aβ antibodies, the sections were treated with formic acid for 6 min. The blocking step was performed with the mouse-on-mouse kit for slides stained with 12F4 antibody (Vector Laboratories, #BMK-2202). Next, the slides were quenched with 3% (v/v) hydrogen peroxide in PBS for 5 min with agitation, then washed three times with PBS containing 0.05% Tween 20 (PBST) for 5 min. The tissue area was outlined with a PAP pen (Ted Pella Inc., #22309) and blocked for 1 h at 22 °C in 2.5% (v/v) horse serum. The slides were incubated with primary antibody overnight at 4 °C. Immunostaining was performed with the following antibodies: 82E1 antibody for total Aβ (IBL, #10323; 1:2000 dilution), 11A50-B10 antibody for Aβ40 (BioLegend, #805401; 1:500 dilution), 12F4 antibody for Aβ42 (BioLegend, #805501; 1:2000 dilution), or 7–14-4 antibody for Aβ38 (BioLegend, #808601; 1:100 dilution). Dilutions were performed with Dako Antibody Diluent (Agilent, #S0809). The following day, slides were washed three times in PBST for 5 min with agitation. Afterward, they were incubated in secondary antibody for 30 min at 22 °C. The slides were washed three times in PBST and incubated with ImmPACT DAB (3,3’-diaminobenzidine) peroxidase substrate for 1 min (Vector Laboratories, #VECTSK4105). The reaction was stopped with water, then the slides were washed in running water for 10 min. Counterstaining was performed by adding hematoxylin dropwise to the slides and incubating them at 22 °C for 2 min. The hematoxylin was removed, and the reaction was stopped with water. The slides were rinsed under running water for 10 min, then dehydrated in successive 5-min incubations in 50% ethanol, 75% ethanol, 95% ethanol, 100% ethanol, 50% xylene in ethanol, and 100% xylene. The stained slides were mounted with Cytoseal 60 (Thermo Scientific, #8310–4) and then covered with standard coverslips (Fisher, #12544E). Stained slides were scanned using the TissueScope LE120.

For ThS staining, slides were deparaffinized and rehydrated as described above. Afterwards, the slides were incubated with filtered 1% (w/v) ThS (Sigma-Aldrich, #T1892) for 6 min in the dark. The slides were incubated twice with 80% ethanol for 3 min, then incubated with 95% ethanol for 3 min. The slides were washed three times with distilled water for 5 min prior to mounting with ProLong Gold fluorescence mounting medium containing DAPI (Thermo Scientific, #P36941). They were covered with standard coverslips (Fisher Scientfic, #12544E) and their edges were sealed with clear nail polish. The slides were stored in the dark at 4 °C until they were scanned using the Akoya PhenoImager HT.

Dense-core plaques were determined based on their rounded shape and larger size compared to diffuse/granular Aβ deposits. The number of Aβ deposits in the cortex and hippocampus of single sagittal brain sections, taken approximately at the level of the midline, were counted manually following staining for Aβ. Similarly, the number of dense-core Aβ deposits and ThS-reactive Aβ deposits in the subcallosal region were counted manually, as were the number of Aβ-positive blood vessels in the leptomeninges.

For double labeling with Alexa Fluor 594-labelled anti-Aβ antibody D54D2 (Cell Signalling Technology, #35363; 1:500 dilution) and ThS, deparaffinization, rehydration, and blocking with 2.5% (v/v) horse serum were performed as described above. Next, slides were incubated with Alexa Fluor 594-labelled D54D2 diluted in antibody diluent overnight at 4 °C. The following day, the slides were washed three times (5 min each) with PBST, then washed three times (5 min each) with dH₂O. The slides were subsequently incubated with filtered 1% ThS and then washed as described above for ThS staining. The slides were incubated with DAPI (1 µg/mL in PBS) for 5 min in the dark and then washed three times with PBS (5 min each). The slides were mounted using Dako Fluorescence Mounting Medium (Agilent, #S302380-2), then covered with coverslips and scanned as described above for ThS staining.

Protease digestions

PK digestions were performed using detergent-extracted human or mouse brain homogenates. Total protein concentration in each sample was measured using the bicinchoninic acid (BCA) assay. Samples containing 200 µg of total protein were incubated with PK at a final concentration of 50 μg/mL in a final volume of 200 µL at 37 °C for 1 h with shaking at 600 rpm. The digestion was stopped by the addition of 3 µL of 200 mM PMSF. Next, 20 µL of 30% (v/v) sarkosyl was added to the samples prior to ultracentrifugation at 100,000 × g in a TLA-55 rotor at 4 °C for 1 h. The resulting pellets were resuspended in loading buffer (1X Bolt lithium dodecyl sulfate (LDS) Sample Buffer (Thermo Scientific, #B0007) containing 2.5% β-mercaptoethanol) and boiled prior to analysis by SDS-PAGE and immunoblotting. Densitometry for PK-resistant Aβ was performed using ImageJ. Relative signals for PK-resistant Aβ levels were obtained by normalizing the signal of each sample to those present in the brain of a Swe2-inoculated App^NL−F mouse.

Conformational stability assays

Detergent-extracted human or mouse brain homogenate (10 µL) was incubated with 30 µL of diluted GdnHCl to generate final concentrations of 1, 2, 2.5, 3, 3.5, 4, 4.5, 5, 5.5, or 6 M GdnHCl. Samples were incubated at 22 °C for 2 h with shaking at 800 rpm and then diluted to 0.4 M GdnHCl in a total volume of 600 µL using detergent buffer. The samples were digested with PK at a concentration of 20 µg/mL for 1 h at 37 °C with shaking at 600 rpm. The reaction was stopped by the addition of 6 µL of 200 mM PMSF, and then 43 µL of 30% (v/v) sarkosyl was added to each sample. Samples were ultracentrifuged at 100,000 × g for 1 h at 4 °C in a TLA-55 rotor and then pellets were resuspended in loading buffer and analyzed by immunoblotting. Densitometry for levels of residual Aβ was performed using ImageJ. The concentration of GdnHCl at which the percentage of residual Aβ aggregates was 50% ([GdnHCl]₅₀) was calculated using a variable slope (four parameter) model in GraphPad Prism. The top value was fixed at 100 while the bottom value was fixed at 0.

Immunoblotting

Samples were analyzed by SDS-PAGE using Bolt 4–12% Bis–Tris Plus gels (Thermo Scientific, #NW04122BOX). Gels were run at 165 V for 32 min followed by transfer to 0.45 µm Immobilon-P polyvinylidene fluoride (PVDF) membranes (Millipore, #IPVH00010) for 1 h at 35 V. The membranes were incubated in blocking buffer (5% (w/v) skim milk in Tris-buffered saline (TBS) with 0.05% Tween 20 (TBST)) for 1 h. Afterward, the membranes were incubated overnight at 4 °C with the following primary antibodies diluted in blocking buffer: D54D2 for total Aβ (Cell Signaling Technology, #8243; 1:5,000 dilution), 6E10 for total Aβ (BioLegend, #803001; 1:1,000 dilution), 7–14-4 for Aβ38 (BioLegend, #808601; 1:1,000 dilution), D8Q7I antibody for Aβ40 (Cell Signaling Technology, #12990; 1:5,000 dilution), or D9A3A for Aβ42 (Cell Signaling Technology, #14974; 1:5,000 dilution). The following day, membranes were washed three times with TBST for 10 min, then incubated with horseradish peroxidase-conjugated secondary antibody (Bio-Rad, #172–1011 and #172–1019) diluted in blocking buffer for 1 h. After three washes in TBST, the membranes were incubated with SuperSignal West Dura ECL (ThermoFisher, #34076) for 5 min and exposed to HyBlot CL film. Alternatively, images were captured using a LiCor Odyssey Fc system.

Quantification of formic acid-extractable Aβ levels in App^NL−F mice

Total protein concentration in 10% (w/v) brain homogenate samples from App^NL−F mice was quantified using the BCA assay (Thermo Scientific, #23225). Next, 500 µg of total protein was diluted to a final volume of 100 µL using PBS. The diluted samples were combined with 200 µL of chilled 100% formic acid and mixed by vortexing, then sonicated in a water bath sonicator for 5 min. The samples were subsequently ultracentrifuged in a TLA-55 rotor for 1 h at 100,000 × g and 4 °C. The supernatant was removed and diluted 1:20 in neutralization buffer (1 M Tris–HCl, 500 mM sodium phosphate). Aβ levels were then quantified using an Aβ(x-42) ELISA kit (Thermo Scientific, #KHB 3442). For samples that fell below the lower detection limit, 250 µL of the supernatant fraction following formic acid extraction was placed in a speed vacuum concentrator for 1 h. The dried pellets were resuspended in 100 µL of dH₂O, sonicated for 10 min in a water bath sonicator, and then re-analyzed by ELISA.

Statistical analysis

All statistical analysis was performed using GraphPad Prism (version 10.6.1) with a significance threshold of P < 0.05. ELISA data was assumed to be normally distributed but standard deviations were not assumed to be equal. Thus, Welch’s t test was used when comparing 2 samples and a Welch’s ANOVA followed by a Dunnet T3 multiple comparison test was performed when comparing 3 or more samples. Neuropathological data was not assumed to be normally distributed. Thus, a Mann–Whitney test was used when comparing 2 samples and a Kruskal–Wallis test followed by Dunn’s multiple comparison test was performed when comparing 3 or more samples.

Results

Characterization and purification of Aβ aggregates from FAD brains

It has previously been shown that Aβ aggregates from single cases of Swedish and Arctic FAD exhibit distinct Aβ isoform compositions and conformational signatures [77]. To extend these findings, we sourced brain samples from the temporal cortex of three cases of Swedish FAD and two cases of Arctic FAD (Table 1). The Swe1 and Arc1 FAD cases are the same as those used in the previous study [77]. As a control, we used a brain sample from a non-neurologic age-matched control individual. To assess the Aβ isoform composition of the Aβ aggregates in the FAD samples, we generated brain extracts and examined levels of detergent-insoluble, proteinase K (PK)-resistant Aβ by immunoblotting. As expected, PK-resistant Aβ, as detected using an antibody to “total Aβ” that recognizes all C-terminal Aβ isoforms, was found in all five FAD brain extracts but not in the control brain extract (Fig. 1a). Using Aβ isoform-specific antibodies, we found that Aβ38 and Aβ40 aggregates were more prevalent in the Arctic FAD brains whereas Aβ42 aggregates were the dominant species in the Swedish FAD brains (Fig. 1a, b). Protein aggregate strains can be differentiated based on their relative resistance to denaturation with guanidine hydrochloride in a conformational stability assay (CSA) [34]. Thus, we performed CSAs to determine a structural fingerprint of the Aβ aggregates in the 5 FAD brain samples. Consistent with previous results [77], Aβ aggregates present in Swedish FAD brain extracts were more stable than those present in Arctic FAD brain extracts (Additional File 1: Fig. S1a, b).

Fig. 1.

Characterization and purification of Aβ aggregates from FAD brain samples. a Immunoblots of detergent-insoluble, proteinase K (PK)-resistant Aβ species in homogenates from control (Ctrl) and FAD brains. Total Aβ, Aβ42, Aβ40, and Aβ38 were detected using the indicated Aβ antibodies. b Quantification of the relative ratio of PK-resistant Aβ42 and Aβ40 signal to PK-resistant total Aβ signal for the Swedish (n = 3) and Arctic (n = 2) FAD samples. c Schematic of the purification of Aβ aggregates from the temporal cortex of control and FAD brains using density gradient centrifugation and PK digestion. d Silver-stained SDS-PAGE gels of Aβ aggregates purified from control or FAD patient brains. e Immunoblots of Aβ aggregates purified from control or FAD patient brains. Total Aβ was detected using the antibodies D54D2 (Ctrl, Swe2, Swe3, and Arc2 samples) or 6E10 (Swe1 and Arc1 samples). f Representative transmission electron microscopy images of purified Aβ aggregates from the Swe1 and Arc1 FAD brain samples. Scale bar = 200 nm (applies to both images)

Previously, Aβ seeding studies utilizing FAD samples were conducted with crude patient brain extracts [77]. These samples likely contained a mixture of soluble and insoluble Aβ species, which are known to exhibit differences in seeding potential [16, 30]. To mitigate this issue and to enable direct comparison of the seeding behavior of FAD-derived Aβ aggregates, we purified PK-resistant Aβ aggregates from the brains of the 5 FAD samples and the control sample using a protocol that enriches for larger Aβ assemblies such as fibrils (Fig. 1c) [73]. While smaller soluble Aβ assemblies may not be captured by this procedure, the molecular structure of soluble Aβ aggregates isolated from AD brains is identical to that of plaque-associated Aβ fibrils [71]. Both SDS-PAGE followed by silver staining as well as immunoblotting revealed that purified Aβ was detectable in preparations from the 5 FAD cases but not the control case (Fig. 1d, e). Transmission electron microscopy was used to assess the ultrastructure of Aβ aggregates in preparations purified from the brains of the Swe1 and Arc1 FAD cases. For the purified Swedish FAD Aβ aggregates, bundles of intertwined fibrillar species were observed (Fig. 1f). While smaller spherical assemblies predominated in the purified Arctic FAD Aβ aggregates, fibrillar species were also present, although we cannot rule out the possibility that fibril formation occurred due to concentration of Aβ aggregates during the purification process.

Injection of FAD-derived Aβ aggregates into App^NL−F mice

To compare the pathological phenotypes induced by Swedish and Arctic FAD-derived Aβ aggregates, we performed in vivo Aβ seeding assays in homozygous App^{NL−F/NL−F} humanized knock-in mice, hereafter referred to as App^NL−F mice. These mice, which predominantly produce human Aβ42 due to the Iberian/Beyreuther APP mutation (I716F), spontaneously develop cerebral Aβ pathology beginning around 6–9 months of age [61, 63]. However, cerebral Aβ deposition in App^NL−F mice can be induced/accelerated by inoculation with Aβ aggregate seeds [58, 61, 62, 70]. Weanling App^NL−F mice were intracerebrally inoculated with purified Aβ aggregates from control or FAD brains, and then cerebral Aβ pathology was assessed at 6 months post-inoculation, which corresponds to ~ 7.5 months of age (Fig. 2a). At this age, App^NL−F mice only exhibit scant Aβ deposits, which are restricted to the cerebral cortex [61]. In our initial inoculation experiments, which utilized the Swe1 and Arc1 samples, sufficient Aβ was obtained to enable each mouse to be injected with 100 ng of purified Aβ aggregates. In our follow-up studies using the Swe2, Swe3, and Arc2 samples, it was only possible to inject each mouse with 50 ng of purified Aβ.

Fig. 2.

Induction of cerebral Aβ aggregation in App^NL−F mice inoculated with purified FAD Aβ aggregates. a Schematic of in vivo Aβ seeding assay in which purified Aβ aggregates from control and FAD brains were intracerebrally inoculated into App^NL−F mice. b ELISA-based quantification of formic acid-extractable Aβ42 levels in brains from App^NL−F mice inoculated with purified Aβ aggregates from control (n = 8), Swe1 (n = 10), Swe2 (n = 10), Swe3 (n = 9), Arc1 (n = 9) or Arc2 (n = 9) samples. c Pooled ELISA data for Aβ42 levels in brains from App^NL−F mice inoculated with purified Aβ aggregates from Swedish (n = 29) or Arctic (n = 18) FAD samples. d Sex-specific analysis of Aβ42 levels in mice inoculated with purified Aβ aggregates from Swedish (n = 14 males, n = 15 females) or Arctic (n = 9 for both males and females) AD brains. e, f Representative immunoblot (e) and quantification (f) of levels of detergent-insoluble, PK-resistant total Aβ species in the brains of App^NL−F mice inoculated with purified Aβ aggregates from control (n = 8), Swedish (n = 29), or Arctic (n = 18) FAD samples. In panel e, Aβ was detected using the antibody D54D2. All graphs display mean ± S.E.M. Statistical significance was assessed using Welch’s ANOVA followed by Dunnett T3 multiple comparisons test (b, f) or a Welch’s t test (c, d)

To confirm the induction of cerebral Aβ deposition in App^NL−F mice, we measured levels of formic acid-extractable Aβ42 in brain extracts from inoculated animals. Aβ42 levels were more than 100-fold elevated in the brains of mice inoculated with purified FAD Aβ aggregates compared to mice inoculated with material from the control case (Fig. 2b). Aβ42 levels were slightly elevated in one control-inoculated mouse, likely due to the presence of a small quantity of cerebral Aβ aggregates in the aged control human brain. Overall, there was no significant difference in Αβ42 levels in the brains of mice inoculated with Swedish or Arctic FAD Aβ, implying that WT and Arctic-mutant Aβ aggregates stimulate Aβ42 deposition to a similar extent (Fig. 2c). Because the kinetics of spontaneous Aβ deposition in transgenic mice expressing mutant human APP is sex-dependent, with females developing cerebral Aβ pathology more rapidly [7, 20], we analyzed whether the degree of Aβ42 induction differed between male and female animals. However, no sex-specific effects were found in mice inoculated with either Swedish or Arctic FAD Aβ aggregates (Fig. 2d). Levels of PK-resistant Aβ aggregates were significantly higher in the brains of App^NL−F mice inoculated with either Swedish or Arctic FAD Aβ aggregates than in the brains of mice inoculated with the control preparation (Fig. 2e, f). There was no significant difference in the amount of PK-resistant Aβ aggregates between mice inoculated with Swedish versus Arctic FAD Aβ (Fig. 2f). We also performed CSAs on the induced Aβ aggregates in brains from inoculated App^NL−F mice. Unlike for the human brain-derived samples, the stability of Aβ aggregates induced by inoculation with Swedish FAD Aβ was indistinguishable from aggregates induced by inoculation with Arctic FAD Aβ (Additional File 1: Fig. S1c, d).

Next, we conducted neuropathological analysis on brains from FAD Aβ-inoculated App^NL−F mice to determine the extent, localization, and morphology of the induced cerebral Aβ deposits. Aβ deposition was assessed using the antibody 82E1, which recognizes the N-terminal region of Aβ and does not cross-react with APP [22]. We first focused on the hippocampus and cortex of inoculated mice, as these are the two main regions in which induced Aβ pathology manifests upon inoculation of App^NL−F mice with Aβ aggregates [61]. In the hippocampus of both female and male App^NL−F mice, induced Aβ deposits were abundant in animals inoculated with either of the two Arctic FAD Aβ preparations but were largely absent in mice inoculated with the Swedish FAD Aβ preparations (Fig. 3a-c). When data was pooled according to FAD type, mice inoculated with Arctic FAD Aβ exhibited significantly higher numbers of Aβ deposits in the hippocampus than mice inoculated with Swedish FAD Aβ (Fig. 3d). Similarly, mice inoculated with Arctic FAD Aβ displayed significantly higher numbers of cortical Aβ deposits (Fig. 3a, b, e, f). In control-inoculated mice, rare small cortical Aβ deposits (1–3 deposits per section) were found in 5/8 animals (Fig. 3b), which is consistent with the kinetics of spontaneous Aβ deposition in this line [61]. Another commonly observed pathological manifestation in App^NL−F mice inoculated with Aβ aggregates is Aβ-CAA [58, 61, 62]. Elevated levels of Aβ-CAA in the leptomeningeal vessels were observed in all App^NL−F mice inoculated with purified FAD Aβ aggregates compared to mice inoculated with the control preparation, and the extent of Aβ-CAA was significantly higher upon inoculation with Arctic than with Swedish FAD Aβ (Additional File 1: Fig. S2). As expected, the leptomeningeal Aβ-CAA deposits were reactive to ThS (Additional File 1: Fig. S3a, b).

Fig. 3.

FAD Aβ strains deposit in different brain regions in App^NL−F mice. a Representative low magnification images of Aβ deposition in the brains of male App^NL−F mice inoculated with purified Aβ aggregates from control, Swedish, or Arctic FAD samples. Aβ was detected using the antibody 82E1. Scale bar = 200 µm (applies to all images). b Representative high magnification images of Aβ deposition in the hippocampus and cortex of female and male App^NL−F mice inoculated with purified Aβ aggregates from control, Swedish, or Arctic FAD samples. Aβ was detected using the antibody 82E1. Scale bar = 50 µm (applies to all images). c, e Quantification of the number of Aβ deposits in the hippocampus (c) and cortex (e) of male and female App^NL−F mice inoculated with purified Aβ aggregates from control (n = 8), Swe1 (n = 10), Swe2 (n = 10), Swe3 (n = 9), Arc1 (n = 9) or Arc2 (n = 9) samples. d, f Pooled Aβ deposition data for the hippocampus (d) and cortex (f) of male and female App^NL−F mice inoculated with purified Aβ aggregates from Swedish (n = 29) or Arctic (n = 18) FAD samples. All graphs display mean ± S.E.M. Statistical significance was assessed using a Kruskal–Wallis test followed by Dunn’s multiple comparisons test (c, e) or a Mann–Whitney test (d, f)

Induced Aβ deposits were also found within the corpus callosum of inoculated App^NL−F mice (Fig. 4a). However, the morphology of the induced deposits differed depending on the FAD Αβ source. Mice injected with purified Swedish FAD Aβ aggregates exhibited primarily large dense-core Aβ plaques in the corpus callosum whereas mice inoculated with Arctic FAD Aβ aggregates displayed diffuse or smaller, more granular Aβ inclusions (Fig. 4a-c). The dense-core plaques in the corpus callosum of Swedish FAD Aβ-inoculated mice were reactive to ThS whereas significantly fewer ThS-positive inclusions were found in Arctic FAD Aβ-inoculated animals (Fig. 4d, e). Double-labeling experiments confirmed that the dense-core plaques in the corpus callosum of Swedish FAD Aβ-inoculated mice could be detected with both an anti-Aβ antibody and ThS whereas Aβ deposits in the corpus callosum of Arctic FAD Aβ-inoculated mice tended to be ThS-negative (Additional File 1: Fig. S3b). The induced Aβ deposits in the hippocampus and cortex, as well as the induced leptomeningeal Aβ-CAA in FAD Aβ-inoculated App^NL−F mice could be detected with antibodies against Aβ42 but not with antibodies against Aβ40 or Aβ38, confirming that Aβ42 is the principal Aβ species found within the aggregates (Additional File 1: Fig. S4). Collectively, these results demonstrate that while Swedish and Arctic FAD Aβ aggregates stimulate Aβ42 accumulation to a similar extent, they deposit in distinct brain regions and induce morphologically distinct inclusions in App^NL−F mice.

Fig. 4.

FAD Aβ strains induce morphologically distinct Aβ deposits in App^NL−F mice. a Representative images of Aβ deposition in the corpus callosum of male and female App^NL−F mice inoculated with purified Aβ aggregates from control, Swedish, or Arctic FAD samples. Aβ was detected using the antibody 82E1. Scale bar = 20 µm (applies to all images). b Quantification of the number of dense-core Aβ deposits in the corpus callosum of male and female App^NL−F mice inoculated with purified Aβ aggregates from control (n = 8), Swe1 (n = 10), Swe2 (n = 10), Swe3 (n = 9), Arc1 (n = 9) or Arc2 (n = 9) samples. c Pooled data for dense-core Aβ deposits in the corpus callosum of male and female App^NL−F mice inoculated with purified Aβ aggregates from Swedish (n = 29) or Arctic (n = 18) FAD samples. d Representative images of ThS-positive inclusions in the corpus callosum of female App^NL−F mice inoculated with purified Aβ aggregates from Swedish or Arctic FAD samples. Scale bar = 20 µm (applies to all images). e Quantification of the number of ThS-positive Aβ deposits in the corpus callosum of male and female App^NL−F mice inoculated with purified Aβ aggregates from Swedish (n = 29) or Arctic (n = 18) FAD samples. All graphs display mean ± S.E.M. Statistical significance was assessed using a Kruskal–Wallis test followed by Dunn’s multiple comparisons test (b) or a Mann–Whitney test (c, e) serial passage of FAD-derived Aβ aggregates in App^NL−F mice

To determine whether the unique pathological phenotypes induced by FAD Aβ aggregates are stable, we performed serial passaging experiments. Initially, we purified Aβ aggregates from the brains of FAD-inoculated App^NL−F mice using the same protocol we used to purify Aβ aggregates from FAD brains. As low quantities of purified Aβ aggregates were obtained, each App^NL−F mouse was injected with only 0.08 ng of purified passaged Aβ aggregates (Additional File 1: Fig. S5a-c), which is ~ 1000-fold less than the amount of Aβ that was inoculated in the experiments involving aggregates purified directly from FAD brains. While leptomeningeal Aβ-CAA was present in mice injected with passaged Swedish or Arctic FAD Aβ but not in mice inoculated with the passaged buffer sample (Additional File 1: Fig. S5d, e), the overall amount of induced Aβ deposition was minimal. However, in one animal inoculated with passaged Swedish FAD Aβ, we found evidence for dense-core Aβ plaques in the corpus callosum whereas a smaller, more diffuse Aβ plaque was observed in the corpus callosum of a mouse inoculated with passaged Arctic FAD Aβ (Additional File 1: Fig. S5f).

To address the issue of low yield of purified Aβ aggregates from the brains of FAD Aβ-inoculated App^NL−F mice, we employed a partial purification procedure that enriches for Aβ aggregates by isolation of sarkosyl-insoluble, PK-resistant Aβ species [61]. Partially purified Aβ aggregates isolated from the same groups of FAD- or buffer-inoculated App^NL−F mice were injected into groups of young App^NL−F mice (Fig. 5a). Higher, but not significantly different, levels of formic acid-extractable Aβ were found in the brains of mice inoculated with the partially purified passaged Swedish and Arctic FAD Aβ samples compared to the brains of mice inoculated with the passaged control sample (Fig. 5b). Aβ levels in two of the control-inoculated mice were clearly above background, and although levels of leptomeningeal Aβ-CAA were elevated in mice inoculated with the passaged FAD samples, Aβ-CAA was also present in most of the mice inoculated with the passaged control sample (Additional File 1: Fig. S6). This is most likely due to the presence of pre-depositing Aβ seeds in the brains of buffer-inoculated App^NL−F mice. Indeed, formation of Αβ seeds occurs prior to when Aβ deposits become detectable by immunohistochemistry in APP23 transgenic mice [76]. As in the experiments where FAD Aβ aggregates were directly inoculated into App^NL−F mice, a significantly higher degree of hippocampal Aβ deposition was observed in the brains of App^NL−F mice inoculated with passaged Arctic FAD Aβ than in the brains of mice inoculated with passaged Swedish FAD Aβ (Fig. 5c, d). Moreover, half of the mice inoculated with passaged Arctic FAD Aβ had elevated numbers of cortical Aβ deposits compared to mice inoculated with passaged Swedish FAD Aβ (Fig. 5c, e). Additionally, dense-core Aβ plaques in the corpus callosum were selectively present in the brains of mice inoculated with passaged Swedish FAD Aβ aggregates (Fig. 5f). Collectively, these results demonstrate that the distinguishing pathological features following seeding with Swedish versus Arctic FAD Aβ aggregates are maintained upon serial passage in App^NL−F mice.

Fig. 5.

Serial passage of FAD Aβ strains in App^NL−F mice using partially purified Aβ aggregates. a Schematic of second passage experiments in which partially purified Aβ aggregates from the brains of App^NL−F mice previously injected with dH₂O (dH₂O_NL-F) or purified FAD Aβ aggregates (Swe1_NL-F and Arc1_NL-F) were intracerebrally inoculated into App^NL−F mice. b ELISA-based quantification of formic acid-extractable Aβ42 levels in brains from App^NL−F mice inoculated with partially purified dH₂O_NL-F (n = 9), Swe1_NL-F (n = 9), or Arc1_NL-F (n = 8) Aβ aggregates. c Representative images of Aβ42 deposition in the hippocampus and cortex of male and female App^NL−F mice inoculated with partially purified second passage Aβ aggregates. Scale bar = 50 µm (applies to all images). d, e Quantification of the number of Aβ42 deposits in the hippocampus (d) and cortex (e) of male and female App^NL−F mice inoculated with partially purified dH₂O_NL-F (n = 9), Swe1_NL-F (n = 9), or Arc1_NL-F (n = 8) Aβ aggregates. f) Representative images of Aβ42 deposition in the corpus callosum of male and female App^NL−F mice inoculated with partially purified second passage Aβ aggregates. Scale bar = 20 µm (applies to all images). In panels c and f, Aβ42 was detected using the antibody 12F4. All graphs display mean ± S.E.M. Statistical significance was assessed using Welch’s ANOVA followed by Dunnett T3 multiple comparisons test (b) or a Kruskal–Wallis test followed by Dunn’s multiple comparisons test (d, e)

Injection of transgenic mouse-derived Aβ aggregates into App^NL−F mice

Having established that Aβ aggregates isolated from FAD brains induce mutation-specific pathological phenotypes in App^NL−F mice, we next assessed the seeding properties of Aβ aggregates generated in the brains of transgenic mice expressing corresponding FAD-mutant APP. For this purpose, we utilized tg-APP_Swe and tg-APP_ArcSwe mice, which respectively express either Swedish-mutant human APP (and thus overproduce WT human Aβ) or Swedish/Arctic-mutant human APP (and thus overproduce E22G-mutant human Aβ) [38]. As determined by biochemical, structural, and neuropathological analysis, distinct types of Aβ aggregates form spontaneously in the brains of tg-APP_Swe and tg-APP_ArcSwe mice [51, 55, 86]. Aβ aggregates were partially purified from aged tg-APP_Swe and tg-APP_ArcSwe brains and then intracerebrally inoculated into young App^NL−F mice (Fig. 6a, b). Significantly higher levels of formic acid-extractable Aβ42 as well as leptomeningeal Aβ-CAA were found in the brains of mice inoculated with tg-APP_Swe or tg-APP_ArcSwe Aβ aggregates compared to a cohort of age- and sex-matched uninoculated mice, confirming that both samples exhibit Aβ seeding activity (Fig. 6c, Additional File 1: Fig. S7). Unlike what was observed following inoculation with FAD brain-derived Aβ aggregates, levels of hippocampal Aβ deposition were low in both groups of inoculated App^NL−F mice, and fewer cortical Aβ deposits were observed in the brains of mice inoculated with tg-APP_ArcSwe Aβ aggregates than in mice injected with tg-APP_Swe Aβ aggregates (Fig. 6d-f). Furthermore, dense-core Aβ plaques were preferentially found in the corpus callosum of App^NL−F mice that had been injected with tg-APP_ArcSwe Aβ aggregates (Fig. 6d, g).

Fig. 6.

Injection of Aβ aggregates from transgenic mice expressing FAD-mutant APP into App^NL−F mice. a Immunoblot of detergent-insoluble, PK-resistant Aβ species in partially purified brain extracts from aged tg-APP_Swe and tg-APP_ArcSwe mice. Aβ was detected using the antibody 6E10. b Schematic of seeding experiments in which partially purified Aβ aggregates from the brains of tg-APP_Swe or tg-APP_ArcSwe mice were intracerebrally inoculated into App^NL−F mice. c ELISA-based quantification of formic acid-extractable Aβ42 levels in brains from App^NL−F mice inoculated with partially purified tg-APP_Swe (n = 10) or tg-APP_ArcSwe (n = 10) Aβ aggregates as well as age-matched uninoculated mice (n = 9). d Representative images of Aβ deposition in the hippocampus, cortex, and corpus callosum of uninoculated male and female App^NL−F mice as well as App^NL−F mice inoculated with partially purified Aβ aggregates. Scale bars = 50 µm (hippocampus, cortex) or 20 µm (corpus callosum). e–g Quantification of the number of Aβ42 deposits in the hippocampus (e) and cortex (f) as well as the number of dense-core Aβ42 deposits in the corpus callosum (g) of uninoculated male and female App^NL−F mice (n = 9) or App^NL−F mice inoculated with partially purified Aβ aggregates from tg-APP_Swe (n = 10) or tg-APP_ArcSwe (n = 10) brains. All graphs display mean ± S.E.M. Statistical significance was assessed using Welch’s ANOVA followed by Dunnett T3 multiple comparisons test (c) or a Kruskal–Wallis test followed by Dunn’s multiple comparisons test (e–g)

Injection of synthetic Aβ aggregates into App^NL−F mice

Next, we asked whether synthetic Aβ aggregates composed of either WT or E22G-mutant Aβ peptide could reproduce the pathological phenotypes observed with FAD brain-derived Aβ aggregates. Synthetic WT and E22G-mutant Aβ(1–40) and Aβ(1–42) peptides were polymerized in vitro using a protocol that yields PK-resistant Aβ aggregates [62]. As expected based on previous studies [15, 32, 46, 82], E22G-mutant Aβ(1–40) polymerized into aggregates more rapidly than WT Aβ(1–40) (Fig. 7a), and a proportion of the WT and E22G-mutant Aβ(1–40) aggregates generated were resistant to PK digestion (Fig. 7b). In contrast, synthetic E22G-mutant Aβ(1–42) formed aggregates that were sensitive to PK digestion (Additional File 1: Fig. S8). Since PK-resistant Aβ aggregates were used for the studies involving human and transgenic mouse brain-derived Aβ seeds, and because PK-resistant Aβ40 species were abundant in Arctic FAD brains (Fig. 1a, b), we decided to use the WT and E22G-mutant synthetic Aβ(1–40) aggregates for further experiments. The E22G-mutant Aβ(1–40) aggregates exhibited lower binding to Thioflavin T (Fig. 7a), as has been observed previously [15], which could indicate structural differences between the WT and E22G-mutant Aβ(1–40) aggregates. To further address this issue, we compared the fluorescent emission spectra upon reaction of WT and E22G-mutant Aβ(1–40) aggregates with either curcumin or luminescent conjugated oligothiophenes (LCOs), both of which are conformation-sensitive dyes that can discriminate between different Aβ aggregate structures [2, 9, 27, 28, 47, 50, 53, 60]. Subtle differences in the emission spectra for WT and E22G-mutant Αβ(1–40) aggregates were observed upon reaction with curcumin or the LCO pentameric formyl thiophene acetic acid (pFTAA) (Fig. 7c, d). However, a more pronounced difference was observed when the LCO heptameric formyl thiophene acetic acid (hFTAA) was used (Fig. 7e). Furthermore, E22G-mutant Aβ(1–40) aggregates were more resistant to denaturation than WT Aβ(1–40) aggregates (Fig. 7f, g). Collectively, these results suggest that WT and E22G-mutant Aβ(1–40) aggregates are structurally dissimilar. To examine the pathological behavior of the synthetic Aβ aggregates, we performed seeding studies in App^NL−F mice (Fig. 8a). Because synthetic Aβ aggregates are known to exhibit much lower in vivo seeding activity than brain-derived Aβ aggregates [73], each mouse was injected with 1 μg of Aβ, which is 10–20-fold more than the amount used for the injection studies involving purified FAD brain-derived Aβ aggregates. Significantly elevated levels of formic acid-extractable Aβ42 were observed in the brains of mice injected with either WT or E22G-mutant synthetic Aβ aggregates compared to mice injected with dH₂O, indicating that both types of aggregates exhibit seeding activity (Fig. 8b). While the number of cortical Aβ deposits did not differ between the groups of inoculated mice (Fig. 8c, d), a clear increase in the amount of Aβ-CAA in the leptomeninges and cerebellum was present in the mice injected with synthetic WT or E22G-mutant Aβ(1–40) aggregates (Fig. 8e, f). This is similar to the phenotype observed in App^NL−F mice injected with other synthetic Aβ aggregates [62]. Although Aβ deposits in the corpus callosum of App^NL−F mice injected with synthetic Aβ aggregates were rare, mice that received synthetic WT Aβ(1–40) aggregates selectively exhibited dense-core Aβ plaques (Fig. 8g, h).

Fig. 7.

Conformational characterization of synthetic WT and E22G-mutant Aβ(1–40) aggregates. a Thioflavin T (ThT) aggregation curves for synthetic WT and E22G-mutant Aβ(1–40) peptides. Data is mean ± S.E.M. for n = 3 independent replicates. b Silver-stained SDS-PAGE gel of detergent-insoluble Aβ species in synthetic WT and E22G-mutant Aβ(1–40) aggregate preparations that were treated with PK ( +) or left untreated (-). c-e Fluorescent emission spectra for synthetic WT and E22G-mutant Aβ(1–40) aggregates reacted with either curcumin (c), pFTAA (d), or hFTAA (e). The graphs display the mean values for n = 2 independent aggregate preparations. f, g Conformational stability assays for synthetic WT and E22G-mutant Aβ aggregates. Representative immunoblots (f) and quantification (g) of residual levels of detergent-insoluble, PK-resistant Aβ species following treatment with the indicated concentrations of GdnHCl. PK-resistant Aβ was detected using the antibody 6E10. Data is mean ± S.E.M. for n = 3 independent aggregate preparations

Fig. 8.

Injection of synthetic WT and E22G-mutant Aβ(1–40) aggregates into App^NL−F mice. a Schematic of seeding experiments in which synthetic WT or E22G-mutant Aβ(1–40) aggregates were intracerebrally inoculated into App^NL−F mice. b ELISA-based quantification of formic acid-extractable Aβ42 levels in brains from App^NL−F mice inoculated with dH₂O (n = 7), synthetic WT Aβ(1–40) aggregates (n = 8), or synthetic E22G-mutant Aβ(1–40) aggregates (n = 9). c Representative images of Aβ deposits in the cortex of female App^NL−F mice inoculated with synthetic Aβ(1–40) aggregates. Scale bar = 50 µm (applies to both images). d Quantification of the number of cortical Aβ deposits in App^NL−F mice inoculated with either dH₂O (n = 7), WT Aβ(1–40) aggregates (n = 8), or E22G-mutant Aβ(1–40) aggregates (n = 9). e Representative images of Aβ-CAA in the leptomeninges and cerebellum of female App^NL−F mice inoculated with synthetic Aβ(1–40) aggregates. Scale bar = 20 µm (applies to all images). f Quantification of the number of Aβ-positive vessels in the leptomeninges of App^NL−F mice inoculated with either dH₂O (n = 7) or with WT (n = 8) or E22G-mutant (n = 9) Aβ(1–40) aggregates. g Representative images of Aβ deposits in the corpus callosum of male and female App^NL−F mice inoculated with synthetic Aβ(1–40) aggregates. Scale bar = 20 µm (applies to all images). h Quantification of the number of dense-core Aβ deposits in the corpus callosum of App^NL−F mice inoculated with WT (n = 8) or E22G-mutant (n = 9) Aβ(1–40) aggregates. In panels d and f, Aβ was detected using the antibody 82E1. All graphs display mean ± S.E.M. Statistical significance was assessed using Welch’s ANOVA followed by Dunnett T3 multiple comparisons test (b), a Kruskal–Wallis test followed by Dunn’s multiple comparisons test (c, e), or a Mann–Whitney test (g)

Comparison of phenotypes induced by FAD, transgenic mouse, and synthetic Aβ aggregates

Finally, we compared the pathological phenotypes induced by the various types of Aβ aggregates. Levels of formic acid-extractable Aβ42 were on average ~ 30-fold higher in the brains of App^NL−F mice injected with FAD brain-derived Aβ aggregates compared to mice that were inoculated with either transgenic mouse brain-derived Aβ aggregates or synthetic Aβ(1–40) aggregates (Fig. 9a). While levels of induced leptomeningeal Aβ-CAA were comparable across most of the experimental groups (Fig. 9b), the presence of elevated amounts of hippocampal and cortical Aβ deposits was unique to mice injected with Aβ aggregates purified from Arctic FAD brains (Fig. 9c, d, Additional File 1: Fig. S9). Likewise, the elevated numbers of dense-core Aβ plaques in the corpus callosum observed upon inoculation with Aβ aggregates purified from Swedish FAD brains was not recapitulated by injection with either Aβ aggregates from tg-APP_Swe mice or with synthetic WT Aβ(1–40) aggregates (Fig. 9e). Thus, neither transgenic mouse-derived Aβ aggregates nor synthetic Aβ aggregates recapitulated the divergent pathological phenotypes induced by human brain-derived Arctic and Swedish FAD Aβ aggregates.

Fig. 9.

Comparison of induced Aβ deposition in App^NL−F mice inoculated with FAD brain-derived, transgenic mouse brain-derived, or synthetic Aβ aggregates. a Comparison of formic acid-extractable Aβ42 levels, as determined by ELISA, in the brains of inoculated App^NL−F mice. b-e Comparison of the number of Aβ-positive leptomeningeal blood vessels (b), the number of hippocampal Aβ deposits (c), the number of cortical Aβ deposits (d), and the number of dense-core Aβ deposits in the corpus callosum (e) in the brains of inoculated App^NL−F mice. Mice were inoculated with either purified Aβ aggregates from human Swedish or Arctic FAD brains (n = 29 and n = 18, respectively), partially purified Aβ aggregates from the brains of tg-APP_Swe or tg-APP_ArcSwe transgenic AD (TgAD) mice (n = 10 each), or with synthetic Aβ aggregates composed of WT or E22G-mutant Aβ(1–40) (n = 8 and n = 9, respectively). All graphs display mean ± S.E.M. Statistical significance was assessed using Welch’s ANOVA followed by Dunnett T3 multiple comparisons test (a) or a Kruskal–Wallis test followed by Dunn’s multiple comparisons test (b-e)

Discussion

The existence of structurally distinct strains of Aβ aggregates provides a potential explanation for the phenotypic heterogeneity observed within and between cases of familial and sporadic AD [33]. In this study, we investigated whether the Arctic and Swedish FAD mutations reproducibly generate distinct Aβ strains that induce unique pathological phenotypes in an AD mouse model. Protein aggregate strains are defined as unique conformations of the same peptide or protein sequence within a fibrillar assembly. While previous work demonstrated that Aβ aggregates from the Swe1 and Arc1 FAD cases induce distinct pathological phenotypes in an AD mouse model, the simultaneous presence of Aβ aggregates composed of several different Aβ C-terminal isoforms precluded an assessment of whether the differences stemmed from distinct structures of Aβ aggregates versus variable mixtures of Aβ C-terminal isoforms [77]. Using Aβ aggregates purified from the brains of two Arctic FAD cases and three Swedish FAD cases, we observed consistent mutation-specific pathological signatures upon injection of Aβ aggregates into App^NL−F mice. For instance, Arctic FAD Aβ aggregates induced Aβ deposition in additional brain regions compared to Swedish FAD Aβ aggregates, and the morphology of the induced Aβ pathology in the corpus callosum differed depending on the inoculated Aβ species. As App^NL−F mice predominantly produce a single Aβ C-terminal isoform, the conformation of Aβ aggregates, not the relative ratios of different Aβ C-terminal isoforms, underlies the unique pathological seeding phenotypes encoded by the Arctic and Swedish FAD mutations. Because the unique pathological signature induced by Arctic-mutant FAD Aβ aggregates was conserved upon serial passaging in App^NL−F mice, which produce WT Aβ, this further confirms that the structure of Aβ aggregates, not the sequence of Aβ per se, is the major driver of pathological heterogeneity.

Our data suggests that the Arctic and Swedish FAD mutations each specify the formation of a unique strain of Aβ aggregates within the human brain. This conclusion is bolstered by recent structural data on Aβ aggregates obtained using cryo-EM. Two different structures of Aβ42 filaments (type I and type II) have been detected in sarkosyl-insoluble fractions isolated from the brains of sporadic and familial AD patients [78]. While type I and type II Aβ42 filaments predominate in the soluble fraction, rare type III Aβ42 filaments have also been detected [71, 85]. The structure of Aβ40 fibrils isolated from the leptomeninges of sporadic AD patients exhibits a fold different from those observed in Aβ42 filaments [29, 79]. Although the structures of Aβ42 and Aβ40 filaments from the brains of Swedish FAD patients have not yet been determined, they may be expected to resemble those present in sporadic AD brains since the Swedish APP mutation does not alter the sequence of Aβ. While there are regions of overlap with the structures of type I and type II Aβ42 filaments from sporadic AD brains, the structure of Aβ aggregates isolated from the brain of an Arctic FAD patient reveals subtle conformational differences, likely due to the presence of the E22G mutation [80]. Thus, distinct Aβ aggregate structures likely underlie the unique pathological phenotypes in App^NL−F mice following inoculation with either Arctic or Swedish FAD Aβ aggregates.

We found that Aβ aggregates from the brains of tg-APP_Swe and tg-APP_ArcSwe mice, which respectively produce WT and E22G-mutant Aβ, did not reproduce the pathological phenotypes induced by human FAD brain-derived Aβ aggregates. This is not surprising since the pathological phenotype in tg-APP_ArcSwe mice does not resemble human Arctic FAD Aβ pathology, and plaques in the brains of tg-APP_ArcSwe mice react well with Aβ amyloid PET ligands whereas plaques in Arctic FAD patients do not [55, 67, 74]. Interestingly, the structure of Aβ aggregates isolated from the brains of the tg-APP_ArcSwe line resembles the structure of type I Aβ42 filaments from sporadic AD, not the structure of Aβ aggregates from Arctic FAD [80, 86]. This may explain why we observed dense-core Aβ plaques in the brains of App^NL−F mice injected with Aβ aggregates from tg-APP_ArcSwe mice and further supports our conclusion that the structure rather than the sequence of Aβ aggregates dictates their seeding activity and deposition patterns within the brain.

Much lower levels of formic acid-extractable Aβ and cerebral Aβ deposition were present in the brains of App^NL−F mice following inoculation with transgenic mouse brain-derived Aβ versus human FAD brain-derived Aβ. As we did not determine the concentration of Aβ in the partially purified fractions from the transgenic mice, it is possible that inoculation with lower amounts of Aβ aggregates could partially explain this observation. However, inoculation of App^NL−F mice with 100 ng of Aβ aggregates purified from two other AD mouse models also yielded lower levels of formic acid-extractable Aβ42 in the brain compared to mice inoculated with 50–100 ng of Aβ purified from Arctic or Swedish FAD brains (Additional File 1: Fig. S10) [62]. Lower specific seeding activity of transgenic mouse brain-derived versus AD brain-derived Aβ aggregates has also been documented in another study [1]. These differences in seeding activity could potentially be explained by differences in the structures, assembly states, and/or N-terminal truncation of Aβ aggregates formed in mouse versus human brains [57], a differential involvement of immune cells, and/or the existence of species-specific cofactors that modulate Aβ aggregation. Indeed, the structure of Aβ aggregates from the brains of App^NL−G−F knock-in mice, which produce Arctic-mutant Aβ, differs from that of Aβ aggregates isolated from a human Arctic FAD brain [35, 80]. While the precise cellular factors within the human brain that affect the structure and/or assembly state of Aβ aggregates remain to be uncovered, additional densities present in the cryo-EM structures of brain-derived Aβ aggregates implies the existence of cofactors that could alter seeding activity [78, 80].

Although we do not know the structures of Aβ42 aggregates in the brains of App^NL−F mice following inoculation with FAD-derived Aβ, some of the key neuropathological features of the corresponding FAD cases appear to be conserved. The brains of Swedish FAD patients contain cored Aβ plaques like those found in sporadic AD [6, 31], and dense-core plaques that react with ThS were found in App^NL−F mice inoculated with Swedish FAD Aβ. In contrast, the brains of Arctic FAD patients exhibit “ring-like” Aβ plaques that lack a core and are poorly reactive to amyloid-binding dyes such as ThS [3, 25, 36, 56]. While the Aβ deposits in Arctic FAD Aβ-inoculated App^NL−F mice were not ring-like, they were more diffuse in nature and were less reactive to ThS, thus partially mimicking the properties of human Arctic FAD [25]. The differences in Aβ deposit morphology between Swedish and Arctic FAD Aβ-inoculated mice could reflect accumulation of Aβ aggregates of different sizes. For instance, the dense-core plaques in Swedish FAD Aβ-inoculated mice may indicate the presence of larger Aβ fibrils whereas the more diffuse/granular Aβ deposits in Arctic FAD Aβ-inoculated animals may indicate smaller Aβ assemblies. As the Arctic APP mutation has been linked to generation of smaller Aβ aggregate species such as protofibrils [23, 32, 46], this potentially suggests that structural features of the Arctic FAD Aβ strain are conserved upon injection into App^NL−F mice. Since smaller Aβ assemblies may spread more readily within the brain, this could also explain why Aβ deposits were enriched in the hippocampus and cortex of App^NL−F mice inoculated with Arctic FAD Aβ aggregates compared to mice inoculated with Swedish FAD Aβ aggregates. Perhaps the conformation of Aβ aggregates that form spontaneously is governed by the presence of specific cofactors in human brain, but once Aβ aggregates acquire their conformation, their structural properties can be faithfully transmitted by seeding in the absence of said cofactors, as exemplified by the ability of the Swedish and Arctic FAD brain-derived Aβ aggregates to be serially passaged in mice.

Consistent with what has been previously described for synthetic WT Aβ aggregates [73], we found that synthetic Aβ aggregates composed of Arctic (E22G)-mutant Aβ(1–40) peptide are poor inducers of cerebral Aβ deposition compared to brain-derived Aβ aggregates. As the structure of synthetic E22G-mutant Aβ(1–40) aggregates is very different from the structure of Aβ aggregates isolated from Arctic FAD brain [75, 80], the most likely explanation is that the structure of synthetic E22G-mutant Aβ(1–40) aggregates is less seeding-competent than brain-derived Aβ aggregates. This could be related to the absence of appropriate human brain-specific cofactors during aggregation of Aβ in vitro. When combined with the results from the experiments utilizing tg-APP_ArcSwe Aβ aggregates, we conclude that the presence of the Arctic mutation alone is insufficient to specify the formation of Aβ aggregates with the pathological properties of human Arctic FAD brain-derived Aβ aggregates. Whether this conclusion can be applied more broadly to other FAD-causing mutations that occur within the sequence of Aβ will require further investigation. Although synthetic Aβ(1–40) aggregates were much less potent at stimulating Aβ deposition in App^NL−F mice, the induced pathology did partially resemble the Aβ pathology following inoculation with FAD brain-derived Aβ aggregates. Thus, it is conceivable that synthetic Aβ aggregates exist as a “cloud” of conformations [60], only a minor fraction of which structurally resemble the Aβ aggregates present in human FAD brains. Consistent with this theory, synthetic E22G-mutant Aβ aggregates exhibit a high degree of structural polymorphism [15, 48].

There are several limitations to our study. We only performed Aβ seeding experiments in a single AD mouse model. It is possible that the simultaneous presence of several Aβ C-terminal isoforms is required for the full pathological spectrum of a given Aβ strain to manifest. This could partially explain why Arctic FAD Aβ aggregates produced a distinct neuropathological phenotype in App^NL−F mice, which produce Aβ42, than in APP23 transgenic mice, which produce both Aβ40 and Aβ42 [77]. Furthermore, complex interactions between WT and mutant Aβ in FAD cases could also contribute to strain-specific phenotypes [39], which may not be accurately recapitulated in App^NL−F mice that only produce WT Aβ. While tg-APP_ArcSwe mice produce E22G-mutant Aβ, the presence of the Swedish mutation may modulate the properties of Arctic-mutant Aβ aggregates by altering the subcellular localization of APP cleavage and Aβ generation [18]. This could contribute to the differential seeding behavior of Arctic FAD and tg-APP_ArcSwe Aβ aggregates we observed in App^NL−F mice. Lastly, our studies with synthetic E22G-mutant Aβ only utilized a single Aβ C-terminal isoform, whereas several Aβ isoforms can be detected in Aβ aggregates purified from Arctic FAD brains. As synthetic WT Aβ aggregates composed of different Aβ C-terminal isoforms exhibit variable seeding activities in App^NL−F mice [62], it is conceivable that E22G-mutant synthetic Aβ aggregates composed of other Aβ isoforms, or mixtures thereof, may better mimic the seeding properties of Arctic FAD brain-derived Aβ aggregates.

The brains of certain AD mouse models, such as the tg-APP_ArcSwe line, spontaneously develop cerebral Aβ aggregates with structures similar to those found in sporadic AD brains, suggesting that they may constitute powerful tools for uncovering therapeutics for sporadic AD [78, 86, 87]. However, our findings emphasize the importance of using human brain-derived Aβ aggregates to ensure translational relevance when studying FAD-associated Aβ strains. As frozen human brain tissue from FAD patients is not widely available, an urgent priority should be to develop methodology for generating synthetic Aβ aggregates in vitro that exhibit the same seeding behavior as FAD brain-derived Aβ aggregates. It may be possible to force synthetic Aβ to adopt FAD-specific Aβ structures by performing seed amplification assays that utilize human brain-derived Aβ aggregates as the seed [64], or by identifying cofactors that promote spontaneous formation of FAD-specific Aβ strains. Blocking the spreading of Aβ pathology by interfering with Aβ seeding represents a promising therapeutic strategy for halting or slowing disease progression in individuals with FAD-causing mutations. Since small molecules that block seeded protein aggregation can exhibit strain-specific efficacy [4], utilizing Aβ aggregates that accurately mimic brain-derived structures may be paramount to the development of Aβ-directed therapies for FAD.

Author contributions

Conceptualization: A.M., A.R.-R., M.I., J.C.W. Formal analysis and investigation: A.M., A.R.-R., M.M.B., N.R.G.S., H.H.C.L., E.S., S.Z., J.C.W. Resources: D.S., S.S., M.S.P., K.P.R.N., C.G., M.I. Writing—original draft: A.M., J.C.W. Writing—review and editing: A.M., A.R.-R., M.M.B., N.R.G.S., H.H.C.L., E.S., S.Z., D.S., S.S., M.S.P., K.P.R.N., C.G., M.I., J.C.W. Funding acquisition: J.C.W. Supervision: M.S.P., M.I., J.C.W.

Funding

This work was funded by grants to JCW from the Canadian Institutes of Health Research (MOP-136899) and Alzheimer Society of Canada/Brain Canada (16–13).

Data availability

All data generated or analyzed during this study are included in this published article or its supplementary information.

Declarations

Ethics approval and consent to participate

Human frozen brain samples were obtained from brain banks located at the Karolinska Institutet (Ethics approval numbers 2011/962–31/1 and 2023–07255-02) and Uppsala University (Ethics approval numbers 2005–103 and 2021–04356). Informed consent had been obtained for all cases at the point of tissue collection. Research on human brain tissue was performed according to an ethics protocol approved by the University of Toronto (00038879: “Use of human tissue for research on neurodegenerative diseases”). All mouse experiments were performed in accordance with guidelines set by the Canadian Council on Animal Care under an animal use protocol (#4263.23) approved by the University Health Network Animal Care Committee.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.