*Barnes Maze*

The Barnes maze test is a common tool to measure spatial learning and memory [38] in AD mouse models, based on the aversion of mice to bright open spaces. We particularly preferred the Barnes maze over the Morris water maze, since it presents a less aversive alternative [79]. The apparatus (Ugo Basile SRL, Gemonio, Italy) consisted of a circular surface (diameter 100 cm) with 20 holes at the edge and an escape box positioned below one of the holes. There were four di fferent visual cues positioned around the maze. The task required the mouse to localize the escape hole and enter the box. Initially, we transported each animal to the center of the maze in an opaque vessel to prevent an orientation before the start of the trial. The procedure was divided into two phases. First, in the acquisition phase, each mouse was subjected to two trials per day for four days (3-min limit per trial; inter-trial interval 15–30 min). The trials ended when either the mouse entered the escape box or when a duration of 180 s was over. On day 5, animals were subjected to a probe trial (90 s). During this phase, the escape box was not available anymore. Latencies to the target hole (acquisition & probe) were automatically detected by camera tracking and corresponding EthoVision XT 13 software.

#### *Social Interaction Test*

This method enables the assessment of sociability and social recognition in mice [39]. For this purpose, a three-chamber cage consisting of a central chamber and two lateral compartments (Noldus, Wageningen, Netherlands) was used. The lateral compartments included sex-matched stimulus mice in separate acrylic rod cages, which allowed social interaction without direct contact. Test animals explored the setup during three consecutive phases. During the first time block of 5 min, the mice were allowed to explore only the middle chamber. As a next step, we opened the dividers to the lateral compartments and placed a stimulus mouse into one of the rod cages (social cage). The second rod cage remained empty. The experimental mouse had a period of 10 min to explore the whole three-chamber cage and to interact with the unknown stimulus mouse. For the next 10 min, we placed an additional unknown stimulus mouse into the second rod cage. The cumulative contact time with the familiar and non-familiar conspecific was automatically detected by camera tracking and corresponding EthoVision XT 13 software.

#### *Paired Associates Learning (PAL) Task*

The ability of visuospatial associative learning was tested in males in the touchscreen PAL (touchscreen and corresponding Abet II Touch 18.7.6 software: Campden Instruments, Loughborough, UK and Lafayette Instrument Company, Lafayette, IN, USA). The task requires a lot of training, but is also a valuable tool in terms of translational cognitive research due to its similarities with the human CANTAB [41,80]. Based on the Bussey-Saksida method, animals initially were habituated to the touchscreen chambers during di fferent pre-training phases. After completion, mice were introduced to the proper PAL task. Here, two objects were shown in two spatial locations on the screen. In each trial, only one correct association of object and location was presented, and the animal had to detect it via nose poke. As a result, a reward was delivered automatically (sugared condensed milk, 7 μL, Hochwald Foods GmbH, Thalfang, Germany). Incorrect responses were followed by an aversive light stimulus (5 s time-out period). After an inter-trial interval (20 s), the next trial was initiated by the mouse. A session ended when either 36 trials were completed, or 60 min ran out. The animals were

food restricted through the whole experiment with the aim of reducing body weights to about 90% of the baseline weight before the test. This should enhance the motivation of the mice to collect the reward after each correct trial. Animal weights were monitored three times a week. For the assessment of the 36 sessions of the PAL task, the parameters' session duration, trials completed, and percentage of correct trials were analyzed. The procedure was highly standardized and the closed touchscreen chambers reduced variability due to the experimenter to a minimum.

#### *Place Learning (PL) and Reversal Learning (RL) Task*

The start of the IntelliCage experiment (IntelliCage and IntelliCagePlus 3.2.8 software: New Behavior, TSE Systems, Bad Homburg, Germany) in female mice was scheduled around the time when male mice entered the proper PAL task. Thus, males and females were largely age-matched during the last phase of the behavioral test battery (27 weeks old, resp. 22 weeks on diet). We chose not to test males in the IntelliCage setup, because males are more prone to show aggressive behavior and hierarchical fighting, potentially resulting in injuries due to the housing of male mice in large groups. The IntelliCage tasks of learning ability cover a broad cognitive spectrum by combining the analysis of spatial memory with operant conditioning [43] and provide the advantage of being both home cage and behavioral test during the time of the experiment. Animals from all experimental groups lived together in the special cage for the period of about 7 weeks. Due to this mixed group housing, the experimental diets were substituted by standard maintenance chow (ad libitum) for the duration of this behavioral test. Each apparatus had the capacity to house and detect up to 16 mice simultaneously. The experiment started with a habituation period of 1 week, followed by a pre-training phase on nose poke behavior in corners for water access for 1–2 weeks. During the following week, the animals were habituated to the two defined drinking sessions per day (5–7 a.m.; 7–9 p.m.). In the PL, only one corner per mouse yielded water access in response to nose pokes during drinking sessions (~2 weeks). Motorized doors, controlled by radio-frequency identification (RFID) transponders, opened when a mouse was detected in its adequate corner. In the RL, a di fferent corner was designated as correct (2 weeks). Visits to the correct corners were analyzed for PL and RL. We did not weigh the animals for the duration of the experiment to avoid interference with the automated behavior recording; instead, we visually observed the mice for any sign of deficiency. The IntelliCage enabled a high throughput cognitive investigation of mice, while stress due to human intervention was reduced to a minimum.
