*Immunohistochemical Analysis*

A stepwise rehydration of the brain sections was conducted, followed by a heat-induced antigen retrieval in 10 mM citrate bu ffer (pH 6.0) including 0.05% Tween-20 (Sigma-Aldrich, Taufkirchen, Germany). After rinsing, sections were incubated for 5 min in 0.6% H2O2 (Sigma-Aldrich) in PBS (0.1 M; pH = 7.3) in order to block endogenous peroxidases. Sections were rinsed and incubated for 30 min in PBS containing 1% bovine serum albumin (PBS-B) and 5% normal goa<sup>t</sup> serum (NGS, Sigma-Aldrich) to prevent unspecific binding of the antibody. After subsequent rinsing, sections were incubated overnight at 4 ◦C in PBS-B containing 1% NGS and the primary antibody (anti-human Aβ 82E1 mouse IgG MoAb 1:1000, IBL international, Hamburg, Germany). Rinsing was followed by an incubation with goa<sup>t</sup> anti-mouse IgG H&L Biotin (1:1000, Abcam, Berlin, Germany) in PBS-B containing 1% NGS for one hour. Sections were rinsed followed by a 1-h incubation with avidin-biotin conjugate in PBS (ABC; Vectastain Elite ABC HRP Kit, Linaris, Dossenheim, Germany). After another rinsing step, sections were

treated with 3,3-diaminobenzidine tetrahydrochloride (DAB; Sigma-Aldrich) in water (0.2 mg/mL; pH = 7.6) for 10 min. The immunostaining was then developed by adding 50 μL H2O2 to a final concentration of 0.006%, incubating for another 10 min. The reaction was stopped by rinsing in ice-cold distilled water followed by a counterstaining using Mayer's hematoxylin (Morphisto, Frankfurt am Main, Germany). Sections were finally dehydrated and covered with Pertex (Medite). We digitized appropriate sections using a Nicon Eclipse Ni-E microscope (Nikon Instruments Europe BV, Amsterdam, Netherlands). Whole brain images were taken at a final magnification of 100x and the area occupied by plaques in several regions of interest (ROI; Table A2) was analyzed using the color segmentation plugin (Daniel Sage, Biomedical Imaging Group, EPFL, http://bigwww.epfl.ch/sage/soft/colorsegmentation/) for ImageJ software (National Institute of Health, Bethesda, MD, USA). Only animals of the first cohort were immunohistochemically investigated in this study.


**Table A2.** Regions of interest (ROI) in different cortical and hippocampal areas.

## *Preclinical Quality Parameters*

Several aspects were considered to ensure the quality of the applied methodologies and resulting data. These points are in accordance with initiatives such as EQIPD ("European quality in preclinical data"; https://quality-preclinical-data.eu/). The aim of EQIPD is broadly to implement various quality improving measures in order to enhance the reproducibility of preclinical data [63]. In the present study, we performed a power calculation to estimate the needed group size (http://www.biomath.info/power/). The resulting total amount of 112 animals was tested in two consecutive cohorts. Nine animals were lost during the course of the whole study. In terms of translatability, we have decided to include both male and female animals in the experiments, since Alzheimer's disease affects both sexes in the clinical context, with a higher rate in women than in men [48]. In general, female animals are largely underrepresented in neuroscience research [82]. Randomization was applied at several stages along the study course. Mice were initially allocated to the home cages according to a random list (https://www.random.org/) and target holes in the Barnes maze were set randomly. Besides, drinking corners in the IntelliCages as well as the stimulus mice in the social interaction test were also assigned randomly. A within-cage randomization between groups was not applicable in this case because every mouse matched strictly to its adequate experimental diet. All animals were regularly pre-handled and transferred to the experimental rooms at least half an hour before behavioral analysis. Blinding of the experimenter in order to prevent detection bias was not performed here, because in all behavioral tests automated outcome assessment was applied (via EthoVision XT, IntelliCage and Touchsreen software). However, blinding was performed during the immunohistochemical analysis. Here, a second experimenter marked the ROI in the images without being aware of animal ID or experimental group. Furthermore, an automated animal managemen<sup>t</sup> software as well as an electronic lab-book were used throughout the study. Standard operating procedures had been written prior to the experimental procedures.
