All animal experiments were performed in accordance with the European Communities Council Directive (63/2010/EU) with local ethical approval under the UK Animals (Scientific Procedures) Act (1986) and its amended regulations (2012), and under the project licence number PP2213334 compliant with the ARRIVE guidelines 2.0 [35]. The study was exploratory. No power calculations were performed a priori.

Mice were bred at our local animal facility. Heterozygous 5×FAD mice, on a black C57Bl6/J background (B6.Cg Tg (APPSwFlLon, PSEN1*M146L*L286V; 6799Vas/Mmjax, JAX MMRRC Stock# 034848)) were crossed with mice harbouring the Icelandic mutation generated by Clustered Regularly Interspaced Short Palindromic Repeats–CRISPR-associated (CRISPR-Cas) gene editing of a single nucleotide into the murine APP gene at position 673 on a black C57Bl6/J background, termed mAPP^A673T mice. Screening for potential off-target sites confirmed 4 low frequency targets (score 3.5; for comparison, A673T score is 100) with unlikely consequences on the phenotype. These were therefore not confirmed. Crosses were bred from heterozygous 5×FAD (male or female) with heterozygous mAPP^A673T (male or female). Ear biopsies were genotyped for the 5×FAD and the A673T mutation in the murine APP gene by Transnetyx Inc. (Cordova, USA) and yielded heterozygous offspring only. Mice were grouped by sex and according to one of the four genotypes: C57Bl6/J wild type (WT), mAPP^A673T, 5×FAD and 5×FAD × mAPP^A673T. A total of seventy-one male and female mice, 6- to 7-month-old, were included in the study (Table 1). Experimental mice were kept in sex- and genotype-specific litters $\ge$2 in stock box open housing under constant environmental conditions (20–22 °C temperature, 50–65% humidity, an air exchange rate of 17–20 changes per hour, and a 12-h light/dark cycle with lights turned on at 7 am with simulated sunrise/sunset) and ad libitum chow (Special Diet Services, Witham, UK) and water throughout. Mice were provided with corncob bedding, paper strips, and cardboard tubes (DBM, Edinburgh Scotland, UK) as enrichment throughout the experiment. They were kept in the same holding room throughout the study except when they were transferred to the euthanasia room for sacrifice and tissue harvest. Experimenters and care takers were blinded to the genotype of mice during maintenance and tissue collection. Following tissue collection, independent experimenters, also blinded to the genotype of mice, performed immunohistochemistry, ELISAs, and all statistical analyses relating to these measurements.

Brain tissue was harvested from all seventy-one mice (Table 1). All chemicals were purchased from Merck Millipore (Burlington, MA, USA) if not otherwise stated. Mice were euthanised via intraperitoneal injections of a lethal dose of sodium pentobarbital (#08007/4034, Dolethal (200 mg/mL), Covetrus, UK) before undergoing intra-cardiac perfusion with heparinised phosphate-buffered saline (0.1 M PBS with 0.05% (v/w) heparin, pH 7.4 (#9041-08-1, Sigma-Aldrich, Darmstadt, Germany)) for 5 minutes. Skulls were dissected and whole brains retrieved. The right brain hemisphere was dissected, fixed overnight at room temperature in 10% (v/v) neutral-buffered formalin (#HT501128, Merck, Darmstadt, Germany), dehydrated and embedded in paraffin. Sagittal sections were prepared at 5 µm using a rotary microtome (HM 325, Leica Biosystems, Sheffield, UK), and mounted onto glass slides (SuperFrost^TM, Thermo Fisher Scientific, Lutterworth, UK). Sagittal sections were collected from regions at interaural 0.96 to 1.44 mm lateral of midline [36], and three sections were collected on one slide for each mouse and antibody. After brain removal, the left-brain hemisphere was transferred immediately to liquid nitrogen and kept at –80 °C until used for protein extraction, ELISA and immunoblot quantification.

Wax-embedded sagittal sections were stained in a sex-specific way using four immunohistochemistry staining boxes for male and two for female samples. Each box included a balanced number of all four genotypes. All chemicals were purchased from Merck Millipore (Burlington, MA, USA) unless otherwise stated. Sections were stained according to our standard protocol [37] using the VECTASTAIN® ABC-HRP kit (VECTOR laboratories #PK-4000), the ImmPACT DAB substrate (VECTOR laboratories, Newark, CA, USA #SK-4105), and the 6E10 anti-A$\beta$ antibody (Biolegend, San Diego, CA, USA # 803004, diluted 1:1000). Images of hippocampal cornu ammonis (CA1), the dentate gyrus (DG), the visual cortex (CTX), the prefrontal cortex (PFC), and the cerebellum (CB) were taken using a light microscope at a 100$\times$ magnification (Axio Imager M1, Carl Zeiss, Jena, Germany) and saved as TIFF file format. Entire microscopic images were analysed using ilastik (Version 1.4.0.post1, https://www.ilastik.org) [38] and Fiji (Version 2.14.0, https://fiji.sc) [39]. The pixel and object classification tool in ilastik enabled training of the software based on a small subset of samples and then apply them to larger sets of images [38]. Models were trained to segment images into positively stained pixels and unstained background tissue or artefacts, and additionally to specifically recognise extracellular A$\beta$ plaques. Variability in staining across different slices was accounted for by including faint, high and intermediate staining intensity images during the training process. After applying these models to all images, the percentage of positively stained area for the entire image, as well as extracellular plaques characteristics (number, size, and area) were quantified using Fiji. The total stained area (%), plaque count, average plaque size (µm²) and plaque area (µm²) were each analysed.

Wax-embedded sagittal sections were dewaxed and stained as described above using NeuN (Millipore #mAB377 diluted 1:1000) and glial fibrillary acidic protein (GFAP) (ThermoFisher, Waltham, MA, USA #14-9892-82, diluted 1:100) antibodies. Images from CA1, DG, CTX, PFC and CB were taken, and positive area was quantified as described above (percentage of positively stained area).

All chemicals were purchased from Merck Millipore (Burlington, MA, USA) unless otherwise stated. The left hemibrains were pulverized in a liquid nitrogen prechilled stainless steel mortar and pestle (BioPulverizer, BioSpec, Oklahoma, USA) and homogenized with a pestle and hammer. RIPA lysis buffer (Thermo Fisher Scientific, #89900) containing Pierce Protease and Phosphatase Inhibitor Mini Tablets (Thermo Fisher Scientific, # A32959) and 1mM AEBSF (4-(2-aminoethyl)benzenesulfonyl fluoride hydrochloride) (Thermo Fisher Scientific #78431) were added in a ratio of 5:1 (mL buffer to mg wet tissue) and the homogenate was incubated for 30 minutes on ice with occasional agitation. After centrifugation at 19,000 g for 2 hours at 4 °C (Centrifuge 5427 R – Microcentrifuge (Eppendorf, Hamburg, Germany), using the FA-45-48-11 rotor), the supernatant (referred to as the RIPA-soluble supernatant fraction S1) was transferred into new reaction tubes. The residual pellet was homogenised in 5 volumes TBS (pH 7.6) containing 5 M guanidine hydrochloride (GuHCl) and 1mM AEBSF and incubated with mild agitation (11 rotations per minute, Multi Bio RS-24, Biosan, Riga, Latvia) for 16 hours at room temperature. After centrifugation at 15,000 g for 30 minutes at room temperature, the resultant supernatant fractions (referred to as GuHCl fraction, or RIPA-insoluble fraction or S2) was were each transferred into new tubes. AEBSF was added to both S1 and S2 extraction buffers at a 1 mM final concentration to prevent degradation of A$\beta$. S1 and S2 fractions were stored at –20 °C until use. Total protein concentration of S1 and S2 fractions was determined using the bicinchoninic acid (BCA) protein assay (Pierce™ BCA Protein Assay Kit, Thermo Fisher Scientific, #23225) with bovine serum albumin (BSA: 0.125–2.000 mg/mL) as a reference standard.

All ELISAs were conducted according to the manufacturer’s instructions, and each sample was measured in duplicates.

RIPA-soluble S1 was used to measure human A$\beta$40 (Invitrogen #KHB3481), human A$\beta$42 (Invitrogen, Waltham, MA, USA # KHB3441), mouse tau (Invitrogen #KMB7011), mouse synaptosomal associated protein 25kDa (SNAP25, MyBiosource #MBS451917), mouse syntaxin 1A (STX1A, MyBiosource #MBS452386), and mouse synaptophysin (SYP, MyBiosource #MBS453910). First, all S1 samples were diluted to a protein concentration of 4 µg/µL in RIPA (including protease and phosphatase inhibitors + AEBSF). For A$\beta$40 and A$\beta$42, S1 samples were further diluted 1:5 in dilution buffer provided within each kit. All 5×FAD and 5×FAD × mAPP^A673T samples were used, and one WT and one mAPP^A673T sample was included on each plate as a control. For tau, S1 samples at 4 µg/µL in RIPA were used, and quantification was conducted for all 71 mice. For synaptic proteins, S1 samples were further diluted in PBS at 1:2 for STX1A and 1:10 for SYP and SNAP25 and quantification was conducted for 70 mice (1 female WT excluded for SYP/SNAP25 due to sample preparation error). Additionally, A$\beta$40 and A$\beta$42 were quantified in GuHCl S2 fractions using the same kits as above. All S2 samples were first diluted to a protein concentration of 1 µg/µL in TBS (pH 7.6) containing 5M GuHCl (including protease and phosphatase inhibitors + AEBSF) and further diluted 1:1000 for A$\beta$40 or 1:7500 for A$\beta$42 using the dilution buffer provided within each kit. All 5×FAD and 5×FAD × mAPP^A673T samples were used, and one WT and one mAPP^A673T sample was included on each plate as a control.

S1 RIPA-soluble samples were used for immunoblotting (4 µg/µL in RIPA buffer including protease and phosphatase inhibitors + AEBSF). All chemicals were purchased from Merck Millipore (Burlington, MA, USA) if not otherwise stated. In brief, protein extracts were mixed with 4$\times$ Laemmli sample buffer (Bio-Rad Laboratories, Hercules, CA, USA, #1610747) and incubated for 15 minutes at 37 °C. Twenty µg protein per lane was loaded onto stain-free 4–15% gradient glycine gels (Bio-Rad Laboratories #4568086) and a protein standard (Bio-Rad Laboratories # 1610376) was loaded onto each gel as molecular weight (MW) marker. Proteins were separated in Tris-glycine-buffer (192 mM glycine, 25 mM Tris and 0.9% (w/v) SDS) at 100 V for around 2 hours on ice using a Mini-PROTEAN Electrophoresis Cell (Bio-Rad Laboratories). Proteins were transferred to methanol-activated PVDF membranes (Bio-Rad Laboratories #1620177) at 5V for 30 minutes in Towbin transfer buffer (25 mM Tris, 200 mM glycine, 0.1% (w/v) SDS and 20% (v/v) ethanol). Membranes were then blocked for 1 h at RT in blocking solution (4% (w/v) BSA) in TBS-T (TBS with 0.2% (v/v) Tween-20) and incubated overnight at 4 °C in 5 mL primary antibody (Table 2) diluted in blocking solution. The next day, membranes were washed 3 $\times$ 10 minutes in TBS-T and incubated for 1 h at RT in 25 mL secondary antibody (goat anti-mouse IgG, Bio-Rad Laboratories #5178-2504, or goat anti-rabbit IgG, Bio-Rad Laboratories #5196-2504, 1:5000) diluted in blocking solution containing StrepTactin-HRP conjugate (Bio-Rad Laboratories #1610381; 1 µL conjugate per 100 mL blocking solution). After washing 3 $\times$ 10 minutes in TBS-T, membranes were overlaid for 1 min with ECL solution (Bio-Rad Laboratories #1705061). The chemiluminescence signals were detected by the ChemiDoc Imaging System and the Image Lab software (ChemiDoc™ XRS+ Imaging System (Bio-Rad Laboratories #1708255)) and normalised to protein loading signals using Coomassie Blue stain (0.1% Coomassie in 20% acetic acid and H₂O). A mixture of all samples was included on each gel for between-gel normalization.

No a priori exclusion criteria were set. However, some immunohistochemistry (IHC) samples were excluded due to tissue damage during sectioning or lack of staining possibly due to sample preparation errors, and additionally some samples were excluded after immunoblotting due to damage of the gel. Details are specified in the respective sections below. Data were analysed and graphs generated in R (Version 4.4.3, R Core Team, Vienna, Austria) using linear models or generalized linear models and analysed using 2- or 3-Way ANOVA or Wald $\chi$² tests. Where appropriate, post-hoc tests were performed using Bonferroni correction. For 6E10 and NeuN IHC staining, males and females were analysed separately and the effects of brain region, genotype and their interaction on each parameter were assessed. For each analysis, it was first determined whether data met assumptions for normality or if any data transformations were necessary. Data met necessary assumptions after transformation using either simple methods (square root, log) or more advanced methods (Box-Cox or Yeo-Johnson transformation). As IHC was performed over several days, a nuisance factor “Staining Day” was included in statistical models if it had a significant effect on the variable being analysed. Total 6E10-positive area showed a significant nuisance factor effect in both males and females. Meanwhile in the analysis of plaque parameters, NeuN and GFAP positive area, the staining day showed only a weak or non-significant effect and was therefore excluded as a factor. A similar approach was taken for ELISA data, where the effects of sex, genotype and their interaction on protein levels were assessed. Data were first tested for necessary assumptions and transformed if necessary. For A$\beta$ and tau ELISA, data were transformed using either simple or more advanced methods (see above) while synaptic protein data already met assumptions for two-way ANOVA. Due to the large number of samples multiple ELISAs were performed, which in part were from different lots and performed on different days. This was accounted for by inclusion of a nuisance factor where necessary. Nuisance factor was included in S1 A$\beta$42 and analysis of synaptic proteins. Western blot data were analysed using one-way ANOVA with factor genotype following data transformation where necessary (for details, see figure legends). All statistical outcomes are reported based on linear or generalised linear models of transformed data, but figures show untransformed data. Due to a sample preparation error, one sample (female WT) had to be excluded from SYP and SNAP25 ELISA. No other samples or data points were excluded from analysis. For each genotype and sex Pearson correlation matrices were generated from A$\beta$ ELISA and A$\beta$ IHC data and compared visually and statistically using the Jennrich test [40] to determine if the matrices were significantly different from each other. To determine whether the level of soluble or insoluble A$\beta$42/A$\beta$40 affected plaque counts and whether this effect varies between genotypes, generalized linear modelling was used. Negative binominal models were used and nested models (with or without interaction/factors) were compared using likelihood ratio tests to determine significance of each main effect (A$\beta$42/A$\beta$40 ratio and genotype) and interaction. Similarly, linear modelling was applied to determine the effect of A$\beta$42/A$\beta$40 ratio on plaque area and whether this differs between genotypes. All data are presented as mean $\pm$ standard deviation (SD) and alpha was set to p 0.05.

We proceeded to assess via IHC whether the introduction of the Icelandic mutation in a 5×FAD background changed A$\beta$ levels using the monoclonal antibody 6E10. This antibody is widely used in AD research; it recognises APP fragments that contain the A$\beta$ sequence (including full-length A$\beta$40 and A$\beta$42, as well as smaller fragments when used during immunoblotting) and is expected to label both intracellular and extracellular deposits of APP and A$\beta$. Representative micrographs of male 5×FAD and 5×FAD × mAPP^A673T (Fig. 1A), female 5×FAD and 5×FAD × mAPP^A673T (Fig. 1B), as well as WT and mAPP^A673T mice (Supplementary Fig. 1A) reveal uniform and punctate cytosolic staining (Fig. 1A,B & Supplementary Fig. 1A, black arrowheads) and frequent nuclear as well as occasional axonal/ dendritic staining (Fig. 1A,B & Supplementary Fig. 1A, white arrowheads). In WT and mAPP^A673T mice (Supplementary Fig. 1A), there were abundant 6E10-positive neurones across all cortical layers in visual cortex and PFC and especially in the pyramidal cell layer of CA1 and granular cell layer of DG. Fewer 6E10-positive cells were found in other CA1 and DG layers as well as in the hilus. In CB granule cell layer showed widespread cytoplasmic labelling while fewer positive cells were seen in the molecular layer. Additionally, large Purkinje cells were also frequently positive for 6E10 labelling. A similar cytosolic and axonal/ dendritic staining was also seen in 5×FAD and 5×FAD × mAPP^A673T (Fig. 1A,B).

Fig. 1.

A$\beta$ immunohistochemistry using the antibody 6E10. Representative A$\beta$ immunohistochemistry images of brain sections of male (A) and female (B) 5×FAD and 5×FAD × mAPP^A673T mice stained with the antibody 6E10 (Biolegend # 803004, diluted 1:1000). Images from CA1, DG, CTX and PFC were taken using a light microscope at a 100$\times$ magnification. Black arrowheads, cytosolic staining; white arrowheads, axonal/dendritic staining; black arrows, dense core plaques with halo; white arrows, small dense plaques with no/little halo; asterisk, diffuse plaques; scale bars, 100 µm. 6E10 labelling was quantified using ilastik for plaque counts (C,D), plaque size (E,F) and plaque area (G,H) in male and female 5×FAD and 5×FAD × mAPP^A673T mice in four brain regions. Data is shown as individual values, group mean, and S.D. Statistical analysis entailed Wald $\chi$² test (C,D) or two-way ANOVA (E–H) with genotype and region as independent variables. Significance of each factor and the interaction is indicated above each graph. No data transformation was performed for plaque counts (C,D) while size and area were Box-Cox (E,F,H) or square-root transformed (G). Males: 5×FAD: n = 17 (PFC n = 16), 5×FAD × mAPP^A673T: n = 14 (PFC n = 13). Females: 5×FAD: n = 4, 5×FAD × mAPP^A673T: n = 5. Abbreviations: CA1, hippocampal CA1; CTX, visual cortex; DG, dentate gyrus; ns, not significant; PFC, prefrontal cortex; 5×FAD, 5× familial Alzheimer’s disease; APP, amyloid precursor protein; SD, standard deviation.

Extracellular A$\beta$ deposits were absent in WT and mAPP^A673T mice (Supplementary Fig. 1A), but 5×FAD and 5×FAD × mAPP^A673T mice of both sexes showed abundant extracellular A$\beta$ deposits (Fig. 1A,B). These consisted of characteristic A$\beta$ plaques with an intensely labelled core and a fainter diffuse halo (Fig. 1A,B, black arrows). In addition, deposits of smaller, intensely labelled core-only plaques with little to no halo (Fig. 1A,B, asterisk) and less intensely labelled diffuse plaques with no discernible core (Fig. 1A,B & Supplementary Fig. 1A, white arrows) were identified. All three types of plaques were found in hippocampal and cortical areas in 5×FAD and 5×FAD × mAPP^A673T, but none were seen in CB (Fig. 1A,B). Plaque number, size and area were measured to quantify extracellular A$\beta$ deposits. These three parameters differed significantly between the four genotypes, confirming the A$\beta$ plaque pathology phenotype in 5×FAD and 5×FAD × mAPP^A673T male and female crosses (Supplementary Fig. 1B–G, p values 0.001). When the total 6E10 signal was quantified, these genotype differences persisted only in female but not male cohorts (Supplementary Fig. 1H, p not significant in males and Supplementary Fig. 1I, p 0.001 in females). However, while the number of plaques was similar between 5×FAD and 5×FAD × mAPP^A673T male (Fig. 1C) and female mice (Fig. 1D), male 5×FAD × mAPP^A673T had significantly smaller plaques than male 5×FAD (Fig. 1E, F_Genotype(1,114) = 5.24, p = 0.024), but no genotype-related differences were measured for this parameter in female cohorts (Fig. 1F). The plaque area was also similar between genotypes in male (Fig. 1G) and female mice (Fig. 1H). Finally, the number of plaques varied significantly between brain regions in male 5×FAD and 5×FAD × mAPP^A673T males (Fig. 1C, $\chi$²_{Brain Region}(3) = 12.14, p = 0.007), where significantly more plaques were counted in PFC than in CA1 (post-hoc test p = 0.009).

In summary, we have confirmed the A$\beta$ plaque pathology phenotype in 5×FAD and 5×FAD × mAPP^A673T male and female cohorts and show, for males, that the mA673T mutation significantly decreases the size of A$\beta$ plaques in 5×FAD × mAPP^A673T crosses compared to 5×FAD.

Given this significant reduction of A$\beta$ plaque size in male 5×FAD × mAPP^A673T crosses, we further explored, using ELISA, whether this led to changes in soluble and insoluble A$\beta$40 and A$\beta$42 isoforms (Fig. 2). While all 5×FAD and 5×FAD × mAPP^A673T samples were measured, only one WT and one mAPP^A673T samples were included. Both presented with very low signals, or signals below detection thresholds and confirmed the specificity of the ELISA assays for human A$\beta$ (data not shown). Female 5×FAD and 5×FAD × mAPP^A673T crosses had almost twice as much soluble A$\beta$40 than their male counterparts (Fig. 2A, F_sex(1,36) = 12.77, p = 0.001), and this was also the case for soluble A$\beta$42 (Fig. 2B, F_sex(1,35) = 8.35, p = 0.007), while the A$\beta$42/A$\beta$40 ratio was similar between cohorts (Fig. 2C). Similarly, females of both genotypes had more insoluble A$\beta$40 (Fig. 2D, F_sex(1,36) = 7.67, p = 0.008), and A$\beta$42 (Fig. 2E, F_sex(1,36) = 8.02, p = 0.007), but again a similar A$\beta$42/A$\beta$40 ratio (Fig. 2F) compared to their male counterparts. Neither soluble, nor insoluble A$\beta$40 and A$\beta$42 nor their ratios differed between 5×FAD and 5×FAD × mAPP^A673T crosses, but a trend towards reduction for A$\beta$42/A$\beta$40 in S2 was observed for 5×FAD × mAPP^A673T compared to 5×FAD (Fig. 2F, F_Genotype(1,36) = 3.31, p = 0.077).

Fig. 2.

Quantification of soluble and insoluble A$\beta$. Human A$\beta$40 and A$\beta$42 and A$\beta$42/A$\beta$40 ratios were quantified in RIPA-soluble (A–C) and insoluble fractions (D–F) in male and female WT, mAPP^A673T, 5×FAD and 5×FAD × mAPP^A673T mice. Data were analysed using two-way ANOVA with sex and genotype as independent variables. Significance of each factor and the interaction is indicated above each graph. Ratios did not require data transformation and remaining data were transformed using Yeo-Johnson (A,D) or Box-Cox (B,E) transformation. Data is shown as individual values, group mean, and S.D. Males: 5×FAD: n = 17, 5×FAD × mAPP^A673T: n = 14. Females: 5×FAD: n = 4, 5×FAD × mAPP^A673T: n = 5. One wild-type and one mAPP^A673T were included on each ELISA plate as a control (see Methods). Abbreviations: ns, not significant; WT, wild type.

We next investigated A$\beta$, APP, and its metabolites using immunoblotting to assess whether the murine A673T mutation would shift the processing of the human isoforms from the amyloidogenic to the non-amyloidogenic pathway (Fig. 3 and additionally see Supplementary Fig. 2 for uncropped images of the complete cohort). We have used three different anti-APP/A$\beta$ antibodies (Table 2) on male cohorts as only these returned genotype-specific differences for A$\beta$ plaques (Fig. 1).

Fig. 3.

Quantification of APP/A$\beta$ species using immunoblotting. Proteins from RIPA-soluble S1 fractions were separated by SDS-PAGE (20 µg per lane) and labelled with the antibody 2B3 against sAPP$\alpha$ (A), Poly8134 against sAPP$\beta$ (E), and CT695 against CTFs (I). Densitometric quantification of the bands of interest (arrowheads) was conducted using the Image Lab software and normalisation to total protein loading. For antibody 2B3, three bands at 10 kDa (B), 100 kDa (C) and 17 kDa (D) were identified. For the antibody Poly8134, three bands were quantified at 100 kDa (F), the 50 kDa (G) and 17 kDa (H). The third antibody, CT695 revealed four bands at 75 kDa (J), 50 kDa (K), 25 kDa (L) and 19 kDa (M). Data is shown as individual values, group mean, and SD Data were analysed using 1-way ANOVA with genotype as independent variable and significance is indicated above each graph. No data transformation was required. Males: WT: n = 8 (n = 6 for Poly8134 antibody), mAPP^A673T: n = 9, 5×FAD: n = 17 (n = 15 for Poly8134 antibody), 5×FAD × mAPP^A673T: n = 14. Abbreviations: crosses, 5×FADx mAPP^A673T; ns, not significant; *: loading control.

The monoclonal antibody 2B3 is directed against the C-terminus of human sAPP$\alpha$. Applying our immunoblotting protocol to RIPA-soluble S1 fractions, this antibody revealed three bands: two higher molecular weight bands at around 140 and 100 kDa (sAPP$\alpha$-140 and sAPP$\alpha$-100), as well as a 17-kDa fragment (Fig. 3A, see black arrowheads). The levels of these three bands, however, was similar between genotypes (Fig. 3B–D). The second antibody, Poly8134, is polyclonal and directed against APP$\beta$. It too revealed three bands: sAPP$\beta$-100 (MW ~100 kDa), sAPP$\beta$-50 (MW ~50 kDa) and a 17-kDa fragment (Fig. 3E, see black arrowheads), all of which were similar across the four genotypes (Fig. 3F–H). The third antibody, CT695, reacts with CTFs of human APP and revealed four fragments: CTF75 (~75 kDa), CTF50 (~50 kDa), CTF25 (~25 kDa), and CTF19 (19 kDa, Fig. 3I, see black arrowheads). Again, all these four bands were similar in quantity between genotypes (Fig. 3J–M). All three antibodies revealed considerable cross-reactivity for murine and human APP and their metabolites (e.g., similar bands for WT and 5×FAD mice), likely because mouse and human APP differ by only three amino acids [41].

Given the synergetic and reciprocal regulatory effect of A$\beta$ and tau, and their established role in inducing synaptic protein alterations in AD patients and AD mouse models, we have further examined whether the A673T mutation in the murine APP gene changes endogenous tau levels and/or rescue alterations of synaptic proteins. Mouse tau and three synaptic proteins—SYP, SNAP25, and STX1A—were measured using mouse-specific ELISAs (Fig. 4). Tau was similar across genotypes and sexes (Fig. 4A), as were SYP (Fig. 4B) and SNAP25 (Fig. 4C, all F values 1). STX1A however, was different between the 4 genotypes (Fig. 4D, F_Genotype(3,62) = 3.1, p = 0.034), but none of the differences reached statistical significance in post-hoc tests.

Fig. 4.

Quantification of tau and synaptic proteins. (A) Mouse tau, (B) mouse synaptophysin, (C) mouse synaptosomal associated protein 25kDa, and (D) mouse syntaxin 1A were quantified in RIPA-soluble S1 fractions in male and female WT, mAPP^A673T, 5×FAD and 5×FAD × mAPP^A673T mice. Data were analysed using two-way ANOVA with sex and genotype as independent variables. Significance of each factor and the interaction is indicated above each graph. No data transformation was needed. Males: WT: n = 9, mAPP^A673T: n = 9, 5×FAD: n = 17, 5×FAD × mAPP^A673T: n = 14. Females: WT: n = 4 (SYP, SNAP25 n = 3), mAPP^A673T: n = 9, 5×FAD: n = 4, 5×FAD × mAPP^A673T: n = 5. Abbreviations: ns, not significant; SNAP25, synaptosomal associated protein 25kDa; STX1A, syntaxin 1A; SYP, synaptophysin.

Pearson correlations were generated for data from 5×FAD and 5×FAD × mAPP^A673T male and female mice. These correlation matrices included A$\beta$ pathology (IHC and ELISA) and tau quantification (Supplementary Fig. 3, see supporting information). Correlation matrices differed significantly between male 5×FAD and 5×FAD × mAPP^A673T mice (Supplementary Fig. 3A,B, p 0.001, see supporting information). Although differences were obvious between female 5×FAD and 5×FAD × mAPP^A673T, sample sizes were too small to compare correlation matrices statistically (Supplementary Fig. 3C,D, see supporting information). Overall, there was a high degree of correlation for A$\beta$ (IHC with ELISA), especially in 5×FAD males, while almost no correlations were observed between A$\beta$ and tau in either genotype. When only amyloid pathologies are correlated (Fig. 5A–D), we found that male 5×FAD mice showed significant positive correlations between A$\beta$40 and A$\beta$42 levels with plaque counts and plaque area which were almost entirely absent in 5×FAD × mAPP^A673T (Fig. 5A and Fig. 5B, see asterisks for significant correlations). Additionally, 5×FAD males showed significant negative correlations between A$\beta$42/A$\beta$40 ratio in S2 with plaque counts/area. In female mice, the A$\beta$42/A$\beta$40 ratio in S1 fraction correlated significantly with plaque count/area in 5×FAD mice, but this was not the case in 5×FAD × mAPP^A673T (Fig. 5C and Fig. 5D, see asterisks for significant correlations).

Fig. 5.

Correlation matrices and linear modelling for the different A$\beta$ measurements. (A–D) Pearson correlation matrix between A$\beta$40 and A$\beta$42 levels and their ratio (A$\beta$42/A$\beta$40) in soluble and insoluble fractions and plaque measurements (count, size, and area) are displayed for 5×FAD male (A), 5×FAD × mAPP^A673T male (B), 5×FAD female (C) and 5×FAD × mAPP^A673T female (D) mice with blue for positive correlations, red for negative correlations and white where no correlation was seen (* p 0.05). A$\beta$ levels were quantified using ELISA and plaque counts, size and total area were quantified using immunohistochemistry (averaged across brain regions). Data were analysed using Jennrich test to detect differences between matrices. (E–H) Generalized linear modelling to explore the effect of A$\beta$42/A$\beta$42 ratio in S1 and S2 on plaque counts and whether this differed across genotypes. Model-predicted plaque counts depending on A$\beta$42/A$\beta$40 ratio for 5×FAD and 5×FAD × mAPP^A673T male in S1 (E) and S2 (F), as well as 5×FAD and 5×FAD × mAPP^A673T female in S1 (G) and S2 (H) are presented. Dashed lines indicate non-significant effects.

To further explore these differences in correlations, generalised linear modelling was used to determine whether the A$\beta$42/A$\beta$40 ratio in insoluble and soluble fractions would predict plaque count and whether this effect differs between genotypes (Fig. 5E–H). In males, independent of genotype, the A$\beta$42/A$\beta$40 ratio in S1 did not influence plaque count (Fig. 5E). By contrast, in S2, 5×FAD males showed a negative association between A$\beta$42/A$\beta$40 ratio and plaque count (Fig. 5F, p 0.001). The relationship showed a positive direction in 5×FAD × mAPP^A673T (p 0.001), resulting in lower plaque counts in 5×FAD × mAPP^A673T than 5×FAD males when the A$\beta$42/A$\beta$40 ratio is low, with a significant difference between both genotypes for the number of plaques which depended on A$\beta$42/A$\beta$40 (Fig. 5F, likelihood ratio test: $\chi$²(1) = 8.79, p = 0.003). In 5×FAD female mice, increase in A$\beta$42/A$\beta$40 ratio in S1 was associated with a predicted decrease in plaque counts (Fig. 5G, p 0.001). The opposite was the case in 5×FAD × mAPP^A673T females, with increasing A$\beta$42/A$\beta$40 values associated with an increase in plaque counts (Fig. 5G, p = 0.001). This resulted in lower predicted plaque counts in 5×FAD × mAPP^A673T compared to 5×FAD females for low values of A$\beta$42/A$\beta$40, and a significant difference between genotypes in the prediction of plaque count based on the A$\beta$42/A$\beta$40 ratio (Fig. 5G, likelihood ratio test: $\chi$² (1) = 7.52, p = 0.0061). The A$\beta$42/A$\beta$40 ratio in S2 was not significantly associated with plaque counts in females independent of genotype (Fig. 5H). The same patterns were seen when investigating the relationship between A$\beta$42/A$\beta$40 and total plaque area (Supplementary Fig. 4).

Neuronal loss and gliosis associated with A$\beta$ plaque pathologies have been reported for 5×FAD mice as early as 6 months of age [21, 23]. Therefore, neurons and astrocytes were quantified in different regions of the brain using NeuN and GFAP as markers. This was done in 5×FAD, 5×FAD × mAPP^A673T crosses, as well as their control counterparts WT and mAPP^A673T (Fig. 6).

Fig. 6.

NeuN and GFAP immunohistochemistry. (A) Representative NeuN immunohistochemistry images in brains of male WT, mAPP^A673T, 5×FAD and 5×FAD × mAPP^A673T mice stained with the antibody NeuN. Images from CA1, DG, CTX, PFC, and CB were taken using a light microscope at a 100$\times$ magnification. scale bars, 100 µm. NeuN levels were quantified using ilastik as total stained are in male (B) and female (C) mice in five individual brain regions. (D) Representative GFAP immunohistochemistry images in brains of male WT, mAPP^A673T, 5×FAD and 5×FAD × mAPP^A673T mice stained with the antibody GFAP. Images from CA1, DG, CTX, PFC, and CB were taken using a light microscope at a 100$\times$ magnification. scale bars, 100 µm. GFAP levels were quantified using ilastik as total stained area in male (E) and female (F) mice in five individual brain regions. Data are shown as individual values, group mean, and S.D. Statistical analysis entailed two-way ANOVA with genotype and region as independent variables. Significance of each factor and the interaction is indicated above each graph. Data were transformed using square-root transformation. NeuN - Males: WT: n = 9, mAPP^A673T: n = 9, 5×FAD: n = 17 (PFC n = 16), 5×FAD × mAPP^A673T: n = 14 (CTX and PFC n = 13). NeuN - Females: WT: n = 4 (CB and PFC n = 3), mAPP^A673T: n = 9 (PFC n = 7, CB n = 8), 5×FAD: n = 4 (PFC n = 3), 5×FAD × mAPP^A673T: n = 5. GFAP - Males: WT: n = 8 (PFC n = 6, CB, DG n = 7), mAPP^A673T: n = 9 (PFC n = 5, DG n = 8), 5×FAD: n = 17 (CA1, CTX n = 16, CB, PFC n = 15, DG n = 11), 5×FAD × mAPP^A673T: n = 14 (PFC, CTX, n = 13, DG n = 11). GFAP - Females: WT: n = 4 (CB, PFC n = 3), mAPP^A673T: n = 9 (CB, PFC n = 8), 5×FAD: n = 4, 5×FAD × mAPP^A673T: n = 5.

Representative NeuN images from CA1, DG, CTX, PFC, and CB are shown (Fig. 6A). Their quantification revealed significant genotype differences in male (Fig. 6B, F_Genotype(3,222) = 3.72, p = 0.012) and female cohorts (Fig. 6C, F_Genotype(3,84) = 3.00, p = 0.035). In males, the average over the five brain regions confirmed the difference between the genotypes (F_Genotype(3,45) = 2.30, p = 0.090, data not shown) revealing a modest reduction of NeuN in 5×FAD compared to WT (–5.7%), while this reduction was even less pronounced in 5×FAD × mAPP^A673T crosses compared to WT (–2.8%). A similar, although not significant, observation was seen in females (data not shown), where NeuN was reduced in 5×FAD compared to WT (–12.3%), and again the reduction was less pronounced in 5×FAD × mAPP^A673T crosses compared to WT (–7.9%). Additionally, the NeuN signal differed significantly between brain regions both in male (Fig. 6B, F_{Brain Region}(4,222) = 146.05, p 0.001) and female cohorts (Fig. 6C, F_{Brain Region}(4,84) = 90.75, p 0.001). Post-hoc analyses yielded a lower NeuN signal in CA1 compared to all other regions in males (Fig. 6B, all p values 0.001), and females (Fig. 6C, all p values 0.001).

Astrocytes were labelled using GFAP (Fig. 6D), and quantification revealed genotype differences in male (Fig. 6E, F_Genotype(3,194) = 9.57, p = 0.001), and female cohorts (Fig. 6F, F_Genotype(3,86) = 72.11, p 0.001). A significant difference between brain regions was also seen. For example, in males CA1 and DG had more GFAP-labelled astrocytes than CTX, PFC and CB (Fig. 6E, F_{Brain Region}(4, 194) = 106.04, p 0.001), and similar results were observed for female cohorts (Fig. 6F, F_{Brain Region}(4,86) = 54.51, p 0.001). In male mice, post-hoc tests revealed that mAPP^A673T had significantly less GFAP-positive area than 5×FAD × mAPP^A673T (p 0.001) and 5×FAD (p 0.001) mice. In females, post-hoc analysis revealed that WT and mAPP^A673T, both had less GFAP than 5×FAD and 5×FAD × mAPP^A673T in CTX and PFC (all ps 0.001). In male and female mice, post-hoc tests revealed that 5×FAD and 5×FAD × mAPP^A673T had significantly more GFAP-positive area than WT and/or mAPP^A673T (p values 0.001); 5×FAD and 5×FAD × mAPP^A673T mice, however, were not significantly different from each other.

Histopathologically, the mA673T mutation led to a decrease in A$\beta$ plaque size in 5×FAD × mAPP^A673T males compared to 5×FAD. While both A$\beta$40 and A$\beta$42 are found in plaques, however an increased cerebral A$\beta$42/A$\beta$40 ratio is another well-established biomarker of A$\beta$ pathology in patients and 5×FAD mice, due to the greater aggregation propensity of A$\beta$42 [44, 45]. While no overt differences were identified for soluble/insoluble A$\beta$40 and A$\beta$42, we found the way in which their ratio was associated with plaques differed considerably between 5×FAD and 5×FAD × mAPP^A673T; male 5×FAD mice showed significant positive correlations between insoluble A$\beta$40 and A$\beta$42 with plaque counts and area. These were almost entirely absent in 5×FAD × mAPP^A673T. Similarly, in females with heightened soluble and insoluble A$\beta$40 and A$\beta$42 levels ([25], this study), the A$\beta$42/A$\beta$40 ratio in soluble fractions correlated significantly with plaque count/area in 5×FAD mice, but this was not the case in 5×FAD × mAPP^A673T. When modelling these, genotype-differences depended on A$\beta$42/A$\beta$40 ratios; a protective effect (i.e., reduced plaque burden in 5×FAD × mAPP^A673T) was seen at low ratios that disappeared or is reversed at high ratios. These differences suggest a genotype-dependent sensitivity to A$\beta$ accumulation. They would also suggest the strength of the amyloid burden in 5×FAD mice is too aggressive and the protection is too weak to counteract their aggregation propensity.

Only a few publications have addressed the potential protective effects of the Icelandic mutation in AD models in vivo. The first used a knock-in rat model of humanized A673T-APP, K670N/M671L-APP (Swedish mutation) or both, and found a reduction of A$\beta$40 and A$\beta$42 pathology (using ELISA) for A673T-APP compared to wild-type APP but not when the Icelandic mutation was combined with the Swedish mutation [33]. Using immunoblotting, they corroborated an increase in non-amyloidogenic APP metabolites (sAPP$\alpha$) and a decrease in amyloidogenic APP metabolites (sAPP$\beta$ and $\beta$CTF) again for the Icelandic mutation alone, but not when combined with the Swedish mutation. The authors suggested that the Swedish and Icelandic mutations may act independently but the magnitude of the protective effect caused by the Icelandic mutation is smaller than the aggressive pathogenic effect of the Swedish mutation. We have found no differences in APP fragments between genotypes using immunoblotting, confirming a lack of efficacy of mA673T when combined with the Swedish mutation and suggesting no shift in APP processing in 5×FAD mice when the mA673T mutation is introduced on a Swedish/Florida/London background. The second study generated knock-in mice with humanized APP with the Arctic (E693B) and Beyreuther/Iberian (I715F) mutations and compared them to mice also carrying the Icelandic mutation [32]. The protective A673T mutation reduced plaque area in cortex and hippocampus at 8 months of age but at 12 months, only the number of plaques larger than 20 µm was decreased while smaller plaques showed similar levels in both genotypes. They additionally report a decrease in $\beta$CTF at 3 months (where no A$\beta$ pathology is established yet) but it is unclear if this persists at older age, where we also could not see any shift in APP processing. In our male 5×FAD × mAPP^A673T mice, only the plaque size was decreased compared to 5×FAD, suggesting a more aggressive A$\beta$ pathology produced by the Swedish/Florida/London mutations as compared to the Arctic or Beyreuther/Iberian APP mutations. This is also supported by in vitro findings, where it has been shown that the protective effect of the human A673T mutation was specific to the London mutation (V717I) but was absent in the Florida (I716V) and Swedish (KM670/671NL) mutations [30, 43]. Another study inoculated APPswe/PS1dE9 transgenic mice with either recombinant non-mutant human A$\beta$ or human A$\beta$ containing the A673T mutation once at 2 months of age and found no changes in A$\beta$ levels when analysed at 6 months. There was only a rescue in synapse density and spatial memory which remained unexplained [31]. In this model, similar to our 5×FAD mouse, the role of PS1 mutations remain unexplored and individual contributions of these mutations to the amyloid load, and a possible block of the A673T protection are elusive to date. Lastly, a recent study that introduced the mA673T mutation into a tau-transgenic model, L66, reported no modulation of mouse A$\beta$ or human tau pathologies and no rescue of motor and neuropsychiatric behaviour in these mice [46].

5×FAD mice overexpress randomly integrated mutant human A$\beta$, while in mAPP^A673T mice, the Icelandic A673T mutation was generated in the murine APP gene. It has been shown that co-expression of murine APP can alter A$\beta$ pathology in APP23 transgenic mice but not in the much faster A$\beta$-depositing APPPS1 transgenic mice [34]. Moreover, the targeted knock-in of human BACE1 lead to amyloidosis purely based on murine A$\beta$ [47]. On the contrary, Jankowsky and co-workers showed that overexpression of mouse APP did not alter A$\beta$ pathology when expressed on a PS1dE9 background, while it increased A$\beta$ pathology when expressed on a more aggressive APPswe/PS1dE9 background [48]. These data suggest a differential effect of murine A$\beta$ on human A$\beta$ deposition in the different APP mouse models and may explain the mild effects observed in this study.

Several lines of evidence suggest a connection between A$\beta$ and tau in the pathophysiology of AD, with both proteins being abundant and often co-localising at synapses [49, 50, 51, 52, 53]. It is therefore important to quantify tau levels to confirm if they are affected by APP alterations. A study investigating the effect of the Icelandic mutation in an APP/PS1 mouse model of AD reported a decrease in phospho-tau pathology in the A673T-A$\beta$ groups, but this reduction remains unexplained [31]. By contrast, the mA673T mutation did not affect tau levels and was unable to rescue behavioural impairment in a tau-transgenic mouse model [46]. A recent exploratory study in 6 non-AD patients (unconfirmed idiopathic normal pressure hydrocephalus cases) comparing CSF of three APP^A673T carriers to three age- (and sex-) matched control subjects reported that disease-relevant soluble APP-$\beta$ and A$\beta$42 levels were significantly reduced in the CSF of APP^A673T carries. Yet, soluble APP-$\alpha$, total tau and phosphorylated tau (p-tau 181) were not altered [30]. This is in line with our finding that the Icelandic mutation had no effect on tau, as 5×FAD and 5×FAD × mAPP^A673T mice presented with similar tau levels. It is worth mentioning that 5×FAD showed normal tau levels not dissimilar of WT mice and is in line with unchanged total tau levels in 5×FAD compared to WT at 3 months of age [54].

Synapse loss is a key event in AD that strongly correlates with cognitive decline [55, 56]. Additionally, a link between A$\beta$ plaque formation and synaptic dysfunction has been established [57]. The presynaptic proteins SYP and SNAP25 were chosen as established markers for synapse loss in AD and AD mouse models, while STX1A was chosen as negative, non-changing, marker [26, 55, 58]. The expression of the mAPP^A673T mutation in 5×FAD did not alter levels of these three synaptic markers, in line with a recent report investigating the exact same mutation in a tau-based animal model [46]. However, they also were unchanged across all genotypes despite previous reports of a general reduction of synaptic proteins in 5×FAD as early as 6 months [26], most notably a reduction between 30 and 45% for SYP [59, 60, 61]. These discrepancies likely relate to the different quantification methods used (immunoblotting/immunofluorescence vs. ELISA).

Neuronal loss is a further key pathological feature of neurodegenerative disease such as AD [62]. Conflicting findings were reported for neuronal loss in 5×FAD mice. On one hand, stereologically counted neuron numbers were lower in cortical layer 5 starting at 9 months [63] and persisted at 12 months [23], while on the other neuronal loss appeared as early as 6 months in the subiculum [64]. Our analyses based on area stained in microscopic images using the ilastik software returned no significant changes of the NeuN staining in 5×FAD mice in any of the five brain regions analysed, and no effect of the mA673T mutation. Contrary, more GFAP-positive astrocytes were found for in 5×FAD mice, but no protective effect was observed in 5×FAD × mAPP^A673T crosses.