Divergent Age-Dependent Conformational Rearrangement within Aβ Amyloid Deposits in APP23, APPPS1, and App ^NL-F Mice

We have for several years analyzed amyloid fibril deposits of different proteins and used thiophene-based ligands,²⁷ and other molecular scaffolds such as trans-stilbenes^28,29 for optical assignment of distinct protein aggregates, i.e., amyloid fibril polymorphism on the folding and filament assembly levels.⁷ The notion behind this strategy is that different fibril polymorphs have different molecular structures of their ligand binding sites⁷ and should thereby be distinguishable by fluorescence properties of different dyes binding differently to these structures. Oligo-thiophenes with anionic side chains, LCOs, have been shown to be well-suited for amyloid detection due to high affinity and selectivity.²⁷ The flexible structure of the LCOs allows these molecules to adapt their conformation to the shape of the binding site of the amyloid fibril. This property will alter the conformation and hence conjugation length of the bound LCO and afford different optical outputs depending on the fibril structure from one dye.

It is well established that prions manifest as different strains depending on the structure, i.e., the fibril polymorph, of the PrP amyloids and thereby display different incubation time, tissue tropism, and disease phenotype.³⁰ The optical property of conformational sensitive LCOs was shown to enable the separation of different prion strains in studies of prion-infected transgenic tga20 mice. In other words, the infected tga20 mice displayed different LCO fluorescence from PrP amyloid deposits as a function of inoculated prion strain.¹³

We have in this study revisited our protocol of combining two LCOs to increase the contrast of LCO fluorescence spectra as a marker for Aβ-fibril polymorphism. The two LCO dyes, qFTAA (quadro-formylthiophene acetic acid) and hFTAA (hepta-formylthiophene acetic acid) (FigureB), show distinct spectral properties upon binding to different amyloid fibril structures.¹⁵ The dye qFTAA fluoresces with an emission spectrum peaking at around 500 nm upon binding to tightly packed bundled Aβ-fibrils.³¹ The dye hFTAA binds to both single filamentous and bundled Aβ-fibrils, with red-shifted emission spectra with peaks at 540 and 588 nm.³¹ We, therefore, employed co-staining with qFTAA and hFTAA as surrogate markers for amyloid polymorphism within Aβ amyloid plaque. It is known that Aβ amyloid plaque deposits have different microscopic morphologies when stained by immunohistochemistry and amyloid dyes. Aβ1–40 and Aβ1–42 amyloid fibril structural polymorphism is well documented by high-resolution structural techniques of fibrils formed in vitro,^{32 −36} in purified human^8,37 and mouse brain^38,39 derived amyloid fibrils, and in seeding experiments using brain-derived fibrils as seeds for recombinant Aβ.^40,41 The overall architecture is common among the fibril types. In-register parallel β-strands arranged in β-arches comprise the fibril filament cross-β-sheet structures. However, the fold, sequence arrangement of intermolecular interactions, protofilament packing, and fibril assembly appear dramatically different in the Aβ fibril polymorphs. If and how the Aβ fibril polymorphs are associated with AD onset and progression are currently not established.

It has been discussed that conformational variations, as reported by LCO staining, do not fully agree with the high-resolution cryo-EM structures of Aβ fibrils,⁸ in that LCO staining shows a wide variation of conformations in sAD¹⁸ while cryo-EM structures find one predominant (type I) filament structure in sAD.⁸ Furthermore, fAD also had one predominant polymorph (type II) according to,⁸ where LCOs showed a separation, while still highly variable, depending on the type of fAD.¹⁸ Interestingly App^NL-F and APP23 Aβ-amyloid filaments isolated and imaged by the same Cryo-EM procedure were reported to have the same main structure (type II).^8,39 We therefore herein compared side by side Aβ-amyloid plaque conformational typing by LCO staining of three mouse models (Table S1↗, FigureA).

Co-staining with both qFTAA and hFTAA of plaques from aged (18 Mo) APP23 mice revealed two different fibrillar structural arrangements. Selected regions of interest (ROIs) from the core of the plaque were primarily occupied by qFTAA (blue-shifted spectrum with a peak at 500 nm) indicating tightly packed fibrils (FigureC, blue arrow). The core was surrounded by hFTAA-stained ROIs (red-shifted spectrum with peaks at 540 and 588 nm), proposing different polymorphs of fibrils in the periphery or corona of the plaque (FigureC, orange arrow). We supplemented the hyperspectral microscopy with confocal microscopy, allowing the use of multiple channels to utilize both the antibody and LCO staining at the same time, bringing out more detailed information about how the fibrils are organized in situ in different parts of an individual plaque (FigureD).

We then aimed for pairwise comparisons of Aβ-polymorphic differences between these commonly used mouse models expressing human AβPP. The analysis affords resolution of the organization of structures allowed by optical microscopy (∼1 μm) but with the advantages of selective molecular probing with LCOs and observing intact amyloid structures in their near-native environment using cryosections of the flash-frozen brain (Figure). We compared two AβPP-overexpressing transgenic mouse models (APP23 and APPPS1) but with different Aβ42/Aβ40 ratios (Table S1↗). We also compared AβPP knock-in model App^NL-F exhibiting endogenous AβPP-expression with a humanized AβPP transgene sequence with the overexpressor APP23. Using our established protocol^9,18,19 for the LCO discrimination of Aβ polymorphism as well as antibodies against different epitopes of the Aβ peptide to discriminate the two isoforms we deduced structural differences and how they corresponded to expression and dominating Aβ species.

First, age-matched APP23 and APPPS1 mouse brain sections (18 and 19 months, respectively) were stained with a combination of qFTAA and hFTAA and full fluorescence spectra were collected using hyperspectral epifluorescence microscopy.⁴² Four regions of interest (ROIs), each comprising 5 × 5 pixels (corresponding to ∼1 × 1 μm), from the core (FigureA,B) and 4 ROIs from the corona (FigureA,B) of the plaques were analyzed from each plaque. In total, 15 images comprised 20 plaques for the APP23 mouse, and 18 images contained 54 plaques for APPPS1 were analyzed. Fluorescence intensity ratiometric analyses were performed by division of the fluorescence intensity at 500 nm (qFTAA) with the fluorescence intensity at 540 and 588 nm (hFTAA) for each ROI (I₅₀₀/I₅₄₀ and I₅₀₀/I₅₈₈). We first compared the APP23 and APPPS1 mice. The analysis revealed a higher abundance of qFTAA fluorescence in the plaque cores of the APP23 model compared to that of APPPS1. In both mouse models, the qFTAA fluorescence was higher in the core compared to the corona (FigureC). This demonstrated that different transgenic genotypes have different fibril structures in the plaques depending on transgenic genotype and that the morphology differs between different parts (core and corona) of the same plaque (FigureC).

The results were coherent with our previous data.⁹ Both these models overexpress Aβ but the Aβ42/Aβ40 ratio is different in that APP23 largely generates Aβ40 while APPPS1 has up to 4.3-fold excess of Aβ42⁴³ (Table S1↗). To delineate if total Aβ load or Aβ variant is the dominating denominator of polymorphic structure, we compared the APP23 and APPPS1 mice with App^NL-F mice, again analyzing 19 plaques from 15 images of 18-month-old mice. App^NL-F mice are known to generate almost exclusively Aβ42, whereas the AβPP expression is at endogenous levels (Table S1↗).⁴⁴ The App^NL-F mice exhibited lower qFTAA fluorescence in both core and corona than the other two mouse models with the most striking difference between the three genotypes being observed in the plaque cores (FigureC).

App^NL-F mouse brain had a very low abundance of qFTAA-positive plaques at 18 months. Differential qFTAA/hFTAA staining of plaque core versus corona developed during aging in both APPPS1 and APP23 mice, where increased qFTAA fluorescence of the plaque core as a function of mouse age with a transition at 12 months was reported by us as plaque core maturation.⁹ At very old age (>18 months) we previously observed a drop in plaque core qFTAA positivity due to increased hFTAA staining in APPPS1 mice.⁹ We therefore moved on to analyze the qFTAA/hFTAA ratio over several App^NL-F mice ages, 9–21 months (Figure S1A↗). We did not observe a noticeable trend of altered qFTAA fluorescence with App^NL-F mouse age (FigureA). The I₅₀₀/I₅₄₀ ratio for the App^NL-F mouse was around 3-fold lower compared to that of the previously published aged matched APPPS1 mouse at 21 months (FigureA). Our previously published data on APPPS1 mice using the same method on the contrary showed a clear transition toward a more densely packed amyloid in the core of the plaques at ∼12 months, reflected by the increase in qFTAA fluorescence⁹ (FigureA) and subsequent decrease >18 months,⁹ as discussed above. We therefore here performed a more complete analysis also of APP23 plaque cores as a function of mouse age between 6 and 25 months to include in the comparison with App^NL-F (Figure S1B↗). APP23 plaque cores showed a very distinct transition above 12 Mo (FigureA). APP23 plaque cores showed elevated I₅₀₀/I₅₄₀ ratios being even higher after 18 months than APPPS1 (FigureA) making the discrepancy between App^NL-F mice and APP23 even stronger than for APPPS1 (FigureA). Statistical analysis of the separate age groups showed significant differences between all genotypes, young <13 Mo as well as older mice >18 Mo (FigureB).

As eluted to above, when analyzing the distribution of the qFTAA/hFTAA fluorescence ratio of the plaque cores we observed a trend in fluorescence signal tilting toward low qFTAA and high hFTAA positivity that corresponded to increased Aβ42/Aβ40 ratio. To test this hypothesis, we plotted the Aβ42/Aβ40 ratio versus qFTAA/hFTAA fluorescence ratio for the mouse groups in the study at young (6–13 Mo) and old (>18 Mo) ages. While the qFTAA/hFTAA comparison was limited to plaque cores and the Aβ peptide amyloid content was from brain homogenate^{43 −45} it appeared to correlate as hypothesized. The discrepancy, while significant in young mice, was augmented during aging of the three genotypes (FigureC).

The image analysis above was based on ocular determination of the core and corona. Similar numbers of 5 × 5 pixel ROIs (∼1 × 1 μm) were collected from both core and corona regardless of size or number of plaques in each specific genotype. However, the size and number of plaques differ largely between the mouse models. APP23 mice exhibit large plaques as well as CAA (not analyzed here), while the plaques in APPPS1 mice are smaller and more abundant. The knock-in App^NL-F mice, expressing endogenous amounts of AβPP present both fewer and smaller plaques than the overexpressing models, as can be expected. To get an overall score of qFTAA versus hFTAA positivity of the Aβ amyloid, we performed an unbiased whole image analysis that considers each pixel of the hyperspectral images collected. This analysis method will not report on the region-specific differences of plaque morphology (core and corona) but rather on the overall proportion and variability of amyloid staining within each image. This image analysis (described in Supporting Methods↗ and Figures S2 and S3A–D↗) can be performed at any wavelength ratio and here we used I₅₀₀/I₅₄₀ and I₅₀₀/I₅₈₈ as in the ROI analysis. Our ROI-based results showing that APP23 demonstrated the highest and App^NL-F mice the lowest abundance of qFTAA positivity (FigureC) was confirmed in this unbiased image analysis (FigureA–C). APP23 showed considerably higher I₅₀₀/I₅₄₀ and I₅₀₀/I₅₈₈ ratios and a wide distribution compared to APPPS1 (Figure S3E↗). App^NL-F mice showed the lowest ratios and the narrowest distribution (Figure S3E↗). Density plots comprising the density of hyperspectral image pixels of qFTAA/hFTAA I₅₀₀/I₅₄₀ ratios were generated to visualize and compare genotypes, individuals, and within age groups (FigureA–E). This procedure was performed for all three models at 18–19 months. The mean ratio for aged mice was 0.396 for APP23, 0.255 for APPPS1, and 0.172 for App^NL-F (FiguresC and S4B↗). The density plot of different age groups of App^NL-F mice at the intensity ratio matrix at I₅₀₀/I₅₄₀ also did not show any significant individual differences (FigureD,E) or differences at different ages (cf. FigureD,E), all the age groups are essentially overlapping (Figure S4A↗). These results implied that there were no significant changes in the plaque structure over time.

Aβ fibrils generated in vitro render different packing architectures. Aβ1–42 fibrils are predominantly solitary, albeit clustered, while Aβ1–40 tends to form thick laterally assembled bundles comprising several fibril filaments (FigureA,B). Tightly packed Aβ-fibrils will render a higher qFTAA signal in in vitro experiments^9,31 and Aβ1–42 fibrils are to a lower degree than Aβ1–40 fibrils associated with high qFTAA fluorescence (FigureC). Hence, these in vitro results indicate that the fibrils are arranged differently in vivo in the plaque core region compared to the corona. It also indicates that the Aβ variant composition (Aβ40 vs Aβ42) can influence the plaque structure of the different mouse models. The data support that APP23 plaque cores have more tightly packed fibrils compared to APPPS1 and App^NL-F plaque cores.

To further verify the molecular basis for the Aβ-fibril polymorphism in vivo, co-staining with antibody and LCO was performed, and imaged by confocal microscopy. We here focused on the two mouse model extremes in the qFTAA/hFTAA ratio and Aβ42/Aβ40 ratio, APP23 and App^NL-F. The monoclonal antibody 4G8 (Aβ epitope 18–22) was used as a pan-Aβ detector, while the antibody 12F4 (Aβ epitope 36–42) was used to selectively stain Aβ42 (FigureA). The antibody-LCO co-staining revealed that APP23 plaque cores have qFTAA binding along with hFTAA and 4G8 antibodies (FigureB). The periphery or corona showed only hFTAA and 4G8 binding (FigureB). 12F4 is poorly bound to the outermost part of the corona (FigureB). This co-staining showed that individual APP23 plaque comprised two distinct fibril polymorph regions of core and corona. The plaque core consists predominantly of compact Aβ40 fibrils; meanwhile, the corona consists mainly of diffusely packed Aβ40 fibrils. Very little Aβ42 appeared to be present in the APP23 Aβ-plaque as deduced from immunofluorescence. This could be the result of hidden 12F4 epitopes of Aβ42. Nonetheless, the low amount of Aβ42 within plaque is coherent with previous mass spectrometry data for cored plaques in the similar mouse model APP_swe,⁴⁶ showing exceptionally low Aβ42/Aβ40 ratios, likely undetectable by immunofluorescence herein. Interestingly, the low abundance of Aβ42 compared to Aβ40, positively correlated with higher qFTAA/hFTAA ratios within the core plaque of APP_swe mice⁴⁶ supporting the correlation we have eluted to in this discussion.

For App^NL-F mice, qFTAA positivity was essentially not observable using epifluorescence hyperspectral microscopy (Figures and 3). However, with confocal microscopy and optical sectioning, we detected qFTAA positivity in tiny cores (FigureC) of approximately 80% of the plaques. The small size of the cores is likely the reason for a few areas with elevated I₅₀₀/I₅₄₀ ratios, rendering an overall elevation of the ratio of core versus corona also for App^NL-F mice (FigureC). Notably, the cores do not significantly change with App^NL-F mouse age (FiguresA and 4D,E). This demonstrates that the qFTAA positive part of each plaque is minute in comparison to >18 Mo APP23 where it is a dominating species. Co-staining with antibodies revealed that the plaque corona displayed hFTAA, 4G8, and 12F4 binding but no qFTAA fluorescence (FigureC) and the core was positive for hFTAA, qFTAA, 4G8, and 12F4. Thus, we conclude that the qFTAA-positive tiny plaque cores in App^NL-F are composed of compact Aβ42 fibrils, whereas most Aβ-plaques and the corona consist of diffusely packed Aβ42 fibrils. Hence, since qFTAA can stain App^NL-F cores, essentially devoid of Aβ40, it appears that qFTAA can also bind to subtypes of tightly packed Aβ42 fibril polymorphs. We have previously observed elevated qFTAA fluorescence in transgenic Drosophila with tightly packed Aβ1–42 fibrils expressed in glial cells compared to intraneuronal ring-tangle-like aggregates expressed in neurons.⁴⁷ The staining profiles of aged APP23 and App^NL-F plaque are summarized in FigureD.

The amyloid strain phenomenon and its dependence on fibril conformation have been extensively explored in the context of the prion protein and prion disease, where it is established that the prion structure correlates with disease phenotype.³⁰ Amyloid fibril polymorphism coherent with what is known for prion strains appears evident for Aβ and AD.^48,49 The complexity of Aβ aggregation and the consequent heterogeneity of Aβ amyloid fibrillar structures was recently reviewed.⁵⁰ Aβ-amyloid deposits as both plaques⁵ and cerebral amyloid angiopathy (CAA)⁵¹ and can be found in asymptomatic older individuals. Many individuals never develop AD or other dementias, although they reach old age. It is hence important to delineate the molecular details of the Aβ amyloid fibril polymorphs to better address the progression of Aβ amyloidosis and the distribution of benign and disease-relevant fibril types with molecular diagnostic and therapeutic strategies. The molecular tracers used in vivo in clinical practice today do not readily distinguish between disease-related and nondisease-related amyloid deposition.⁵² Furthermore, not all disease-associated Aβ amyloids are detected by amyloid PET tracers. For example, carriers of the Arctic mutation (E22G), despite extensive Aβ fibril load, do not retain PiB in PET imaging.⁵³ A wide distribution of LCO staining patterns in different patients suggests that the polymorphic patterns of different plaques are very hard to predict.¹⁸ How Aβ amyloids form and evolve over time is important to understand disease progression, diagnostics, and treatment.

In this work, we aimed to further understand the influence of isoform and expression level of the Aβ peptides on amyloid fibril polymorphism in different AβPP mouse models by direct staining and imaging of Aβ plaque in situ. We found that Aβ plaques in the three mouse models included in this study exhibit several different morphologies. The Aβ plaque morphology changes over time in APP23 and APPPS1 mice⁹ but not in App^NL-F mice. App^NL-F mice contain small plaques dominated by hFTAA fluorescence. Plaque formation onset has been reported at 6 weeks of age in APPPS1 mice¹⁷ and at 6 months of age in both APP23¹⁷ and App^NL-F(44) mice. This indicates that the age of the mouse or the age of the plaque cannot exclusively explain the difference in plaque morphology and fibril structure between APP23 and App^NL-F mice. Notably, App^NL-F and APP23 Aβ-amyloid filaments isolated and imaged by cryo-EM were reported to have the same main structure polymorph (type II).^8,31 While it is possible that LCO fluorescence is influenced by factors such as other proteins/glycans/lipids bound to the amyloid fibrils, the pure in vitro data (Figure) strongly support that the LCO fluorescence mainly reports on fibril polymorphism. It can however not be ruled out that variable nonamyloid composition of the different plaque types influence morphology and LCO fluorescence of the proteins within the amyloid deposits. To understand the discrepancy with cryo-EM, it is conceivable that LCO staining reports on higher-order assemblies of filaments with the same filament fold, or that cryo-EM preparation, isolation, and selective particle imaging influence the cryo-EM results. The latter effect has recently been discussed in in situ cryo-EM tomography of App^NL-G-F mice compared with cryo-EM structure determination of isolated ex vivo fibril material.⁵⁴ The structures of fibrils within densely packed plaque cores are likely not resolved in the reported single-particle cryo-EM studies. It is also noteworthy that the type II filament structure reported for 21-month-old APP23 mice was composed of Aβ42,³⁹ suggesting that this structure corresponds to peripheral fibrils of the corona or smaller fibrils within the brain, not being intrinsic parts of amyloid plaque.

Several PS1 mutations found in human fAD act by increasing the release of Aβ peptides from AβPP. However, the Aβ isoform varies between mutations. PS1-A431E results in a high abundance of Aβ peptides but a low Aβ42/Aβ40 ratio (approximately 1/7th for the mutant carrier compared to the average for cases of sporadic AD in the same study).⁵⁵ On the contrary carriers of PS1-E280A in a different study had an almost doubled Aβ42/Aβ40 ratio compared to sporadic AD cases⁵⁶ (both studies used ELISA to deduce the Aβ42/Aβ40 ratio). In other words, carriers of PS1-A431E generate more Aβ40, and PS1-E280A carriers generate more Aβ42 than patients without the mutation. In this sense, APP23 mice are like PS1-A431E while App^NL-F resembles PS1-E280A in terms of dominating aggregated Aβ peptide isoform. In studies of human AD cases, it has previously been shown that Aβ plaques in PS1-E280A carriers display very low qFTAA fluorescence (low intensity at 500 nm) while PS1-A431E carriers generate plaques with high qFTAA signature.¹⁸ This is consistent with the LCO signatures of aged APP23 and App^NL-F mice described here. We hence conclude that time is an important factor for the generation of the tightly packed cored plaques seen in APP23 mice and that this cored structure is promoted by the presence of abundant Aβ40. The process of age-dependent plaque core rearrangement called plaque maturation of APPPS1 and APP23 mice is poorly understood but appears to entail tight lateral packing of multifilamentous fibrils. This process would be very interesting to study by high-resolution methods such as in situ cryo-EM tomography at different ages of mice in conjunction with LCO staining, as was performed in this study. Plaque core maturation towards high qFTAA fluorescence is a feature much less pronounced in App^NL-F mice forming almost exclusively Aβ42 plaque. Hypothetically if App^NL-F mice were to be aged for a very long time >30 Mo it is conceivable that the qFTAA signature would increase. Also, the expression levels generating more Aβ in APP23 and APPPS1 mice than in App^NL-F mice may matter. However, this is not the full explanation. A previous study treating APPPS1 mice with a BACE-1 inhibitor decreased total Aβ production and hence the number and size of Aβ-amyloid plaque in young mice, but the qFTAA/hFTAA ratio was altered by increasing in cortical regions and decreasing in the thalamus, hypothalamus, and hindbrain regions.⁵⁷ No difference was seen in treated versus nontreated 14-month-old APPPS1 mice. These results indicate that the issue is more complex than merely Aβ-concentration.

All animal experiments were conducted in agreement with protocols approved by the local Animal Care and Use Committees, respectively. Animal experiments at Linköping University were approved by the animal ethics committee (#10925-2020, #13028-2021). App^NL-F mice were reared by Takashi Saito and Takaomi Saido laboratories at the RIKEN Center for Brain Science, Tokyo, Japan. APP23 and APPPS1 mice were reared at Mathias Jucker lab at Hertie Institute for Clinical Brain Research, Tübingen, Germany. APP23 mouse tissues were handled as described previously.⁴³ APPPS1 mouse tissues used here are described previously.⁹ This study hence allowed a direct comparison of App^NL-F and APP23 with previous data of APPPS1. Data from a total of 50 brains were included in the study: APPPS1 (n = 19), APP23 (n = 22), and App^NL-F (n = 9).

Flash-frozen mouse brains of transgenic APP23 and APPPS1 and knock-in App^NL-F were used for making brain cryosections of 10 μm. Cryosections were fixed in two consecutive ethanol concentrations of 96% (v/v) and 70% (v/v), 10 min for each concentration at room temperature. Tissue sections were further rehydrated with dH₂O and PBS, pH 7.4, each step having 10 min of incubation time. Following the rehydration steps, tissue sections were incubated with LCOs (2:1 qFTAA and hFTAA, prepared as described below) for 30 min. After incubation with LCOs, tissue sections were washed with PBS 3 times and incubated in PBS for 5 min. Tissue sections were then dried in ambient air and mounted with DAKO fluorescence mounting medium and coverslip #1. The LCOs (qFTAA and hFTAA) were synthesized as described earlier.¹⁵

LCOs were dissolved in 2 mM NaOH in dH₂O to have a stock solution of 1 mg/mL, which gives a qFTAA solution of 1.8 mM and an hFTAA solution of 1.1 mM. LCOs were stored at 4 °C until further use. For double staining of mouse brain sections with LCOs, qFTAA was diluted to 1:10000 and hFTAA 1:1392 and mixed in a ratio of 2:1 correspondingly. This gives a final staining solution of 120 nM qFTAA and 262 nM hFTAA.

For antibody staining, 4G8 (Aβ epitope 18–22) and 12F4 (Aβ epitope 36–42) antibodies were diluted to 1:300 to stain mouse brain sections, followed by Alexa Fluoro 594 diluted to 1:400. For co-staining of antibodies and LCOs, the flash-frozen tissue sections were fixed at 70% (v/v) ethanol at 4 °C for 3 min. Prior to 70% (v/V) ethanol incubation, tissue sections were kept at room temperature for 30 min. Following ethanol incubation, tissue sections were rehydrated in dH₂O for 2 × 2 min and in PBS for 10 min. Later tissue sections were blocked with 5% goat serum in PBS-T (0.1% triton x-100) at room temperature for 1 h. Tissue sections were then incubated with 4G8 and 12F4 primary antibodies overnight at 4 °C. After primary antibody incubation, tissue sections were washed with PBS-T for 3 × 10 min. Following the washing steps, tissue sections were incubated with Alexa Fluoro 594 secondary antibody at room temperature for 1 h. Later tissue sections were washed with PBS for 3 × 10 min. Then tissue sections were subsequently incubated with LCOs and mounted as described above.

For hyperspectral imaging of mouse brain sections, a LEICA DM6000 B microscope was used and equipped with a spectral camera (Applied Spectral Imaging, Israel). A 436 nm long pass excitation filter (436/10 (LP475)) was used for image acquisition. Images were acquired with a 20× objective. Images were acquired with 20× objective except for the mouse groups of App^NL-F at different ages (9, 12, 15, 18, and 21 months). These images were captured with a 40x objective. The age series of App^NL-F mice comprised a total of 9 mice (n = 3 at 9 Mo, n = 1 at 12 Mo, n = 1 at 15 Mo, n = 1 at 18 Mo, and n = 3 at 21 Mo). For the age comparison of APP23, we analyzed a total of 22 mice (n = 5 at 6 Mo, n = 7 at 12 Mo, n = 5 at 18 Mo, and n = 5 at 25 Mo). At 6 Mo mostly intracellular inclusions were present in APP23. For age comparison of APPPS1 data reported by us previously were plotted from,⁹ comprising a total of 19 mice.

Representative images from each genotype were analyzed. From each image, four regions of interest (ROI) were selected from the core and 4 ROIs were selected from the corona of each plaque. Fluorescence intensity at 500 and 540 or 588 nm of the spectra from each ROI was used to generate the intensity ratiometric plots I₅₀₀/I₅₄₀ or I₅₀₀/I₅₈₈ respectively. Representative images from each genotype were used for side-by-side analysis and comparisons in Figures and 4.

Zeiss LSM780 confocal microscope was used to acquire z-stack images of LCO and antibody-co-stained tissue sections. Argon 458, 488, and 514 nm laser lines and DPSS 561-10 laser lines were used to excite the LCOs and Alexa Fluoro 594. Images were acquired with Plan-Apochromat 20x/0.8 M27 objective with a frame size of 1024 × 1024 pixels and scanning area at zoom 1.0.

Using an in-house generated program in RStudio, the ratio between intensity at 500 and 540 nm in each pixel was calculated. Pixels containing only background fluorescence were filtered out by the program (Figure S2A–D↗). Fluorescence emission intensity from hyperspectral imaging data for all pixels on an image at 500, 540, and 588 nm respectively were exported as text files using the spectra view software (Applied Spectral Imaging, Israel). These text files were loaded into RStudio as matrices. Relative filter settings were applied to remove unwanted signals, such as background (dark pixels). The low-wavelength matrix is used to filter out the high-intensity noise because high intensities in the low-wavelength matrix reflect bright noise. On the other hand, the high wavelength matrix is used to filter out the dark background noise. Filtering out the unwanted pixels should be done carefully with a robust reference interval to have reproducibility. Gaussian normal distribution of the pixels was a suitable reference interval for that (Figure S2↗). Calculating the mean value (μ) and using fixed values of standard deviation (σ) from the mean value enabled the upper cutoff limit and lower cutoff limit to be set to the low wavelength (500 nm) and high wavelength matrices (540 nm). The general guidelines describing how to set the cutoff limits for different genotypes based on the image outlook are found in Table S2↗. After applying the filter setting, the ratio between the low- and high-wavelength matrixes is calculated and stored in a matrix. This new calculated ratio matrix is further used to illustrate the data. A heatmap plot is generated to justify the effectiveness of the filter setting (Figure S3B–D↗). Moreover, the ratio matrix is used to generate a violin plot, with features similar to those of the ratiometric plot of ROI, but now on a larger full image scale (Figure S3E↗). Density plots were generated to visualize the ratio pixel distributions for images of each genotype or over the age of the same genotypes. Violin plots and pixel density distribution curves for the emission ratios for all images are found in Figures S3E↗, 4C–E, and S4↗.

In vitro fibril formation, hyperspectral microscopy, and transmission electron microscopy of recombinant Aβ1–40 and Aβ1–42 peptides were performed as described before.³¹ In short, the peptides were purchased from rPeptide, dissolved in 2 mM NaOH, and stored as stocks at −20 °C at a concentration of 1 mg/mL. At the time of fibrillation, the peptides were diluted to a concentration of 10 μM in PBS and fibrillated at 37 °C without shaking. For transmission electron microscopy (TEM), samples were collected at the beginning of the equilibrium phase for fibril formation, as deduced by ThT fluorescence. Carbon-coated copper grids were used to prepare TEM samples of fibrils negative stained with uranyl acetate. TEM images were collected by using a Jeol JEM 1230 microscope at 100 kV using a Gatan CCD camera. For hyperspectral microscopy, fibrils were collected at the end point (25 h). qFTAA and hFTAA were added to final concentrations of 13 and 7 nM respectively and left to sediment by gravity at room temperature overnight. Three microliter samples from the bottom of the tube were placed on superfrost + glass slides and dried in ambient air before mounting with Dako fluorescence mounting medium and coverslip. Images were collected as described for mouse tissue. Fluorescence spectra from ROIs covering the area of 2 representative images were collected, and the ratio between emission intensities at 500 and 540 nm was calculated.