Differences in Learning Strategy Selection and Object Location Memory Impairments in APP/PS1 Mice

Abstract

This study investigated the learning strategy preferences of 11-month-old APP/PS1 double transgenic (Tg) mice, a well-established murine model of Alzheimer’s disease (AD). APP/PS1 Tg and non-Tg control mice were serially trained in visual and hidden platform tasks in the Morris water maze. APP/PS1 Tg mice performed poorly in visual platform training compared with non-Tg mice but performed as well as non-Tg mice in hidden platform training. Further analysis of their search paths for locating a hidden platform revealed that APP/PS1 Tg mice used more cued/response search patterns than place/spatial search patterns compared with non-Tg mice. Three months later, the object/location recognition memory of APP/PS1 Tg mice was assessed. Although their object recognition memory was intact, their object location memory was impaired. Neuropathological AD features of APP/PS1 transgenic mice were observed in the medial prefrontal cortex, retrosplenial cortex, and hippocampus, key brain regions involved in learning strategy shifts and spatial cognition. These results indicate that distinct search patterns and spatial memory deficits in APP/PS1 Tg mice are key features of AD animal models.

INTRODUCTION

Animal models of Alzheimer’s disease (AD) have been extensively used to enhance our understanding of the molecular and pathological processes that lead to cognitive impairment in AD [1-3]. Transgenic (Tg) AD mice carrying mutant forms of the human amyloid precursor (APP), presenilin (PS), and/or tau genes display AD-related pathological characteristics, including cognitive impairments [4]. Various types of memory tests, particularly the assessment of hippocampal function using well-established behavioral tasks, have been conducted in multiple AD Tg mice [5]. In most relevant studies, spatial memory was assessed using a hidden platform training procedure in the Morris water maze task, while context fear memory was measured in the Pavlovian conditioning setting [5-8]. These behavioral tasks were performed to assess the hippocampal function in AD mice.

We recently developed a training protocol to evaluate how genetically modified mice select learning strategies in a traditional water maze [3, 9]. The two inbred strains, C57BL/6 and DBA/2, commonly used to develop genetically modified mouse models, underwent serially cued/response training at different visual locations and place/spatial training at a hidden platform location in a water maze. Following training, the strategy preference of the mice was assessed by offering a choice between a visible platform located in a new location (cued/response strategy) and a spatial location (place/spatial strategy). Both C57BL/6 and DBA/2 mice exhibited similar escape performance in visible and hidden platform training. However, in the preference test, C57BL/6 mice preferred a place/spatial strategy to a cued/response strategy significantly more often than DBA/2 mice [9]. Numerous studies have shown that spatial learning depends on the hippocampus, whereas cued/responsive learning relies on the striatum [10-12]. Therefore, we measured phosphorylated CREB (pCREB) levels in the hippocampus and striatum of the two inbred mouse strains following place/spatial or cued/response training. Both strains showed equivalent performance in place/spatial and cued/response training. However, hippocampal pCREB was higher in the C57BL/6 than in the DBA/2 after place/spatial training, whereas striatal pCREB was higher in the DBA/2 than in the C57BL/6 after cued/response training [9, 13]. These findings suggest that the hippocampus is more involved in place/spatial training, whereas the striatum is more strongly engaged during cued and response training.

Furthermore, C57BL/6 mice performed significantly better than DBA/2 mice in place/spatial training at the learning strategy-switching task from cued/response to place/spatial training. After completion of the switching task, we measured pCREB levels in the hippocampus and prefrontal cortex. Prefrontal cortical and hippocampal pCREB levels were higher in C57BL/6 mice than in DBA/2 mice [14]. Meanwhile, C57BL/6 mice performed marginally better than DBA/2 mice in cued/response training at the learning strategy-switching task from place/spatial to cued/response training. Hippocampal pCREB levels were higher in C57BL/6 mice than in DBA/2 mice, but no difference between the two was observed in pCREB levels in the prefrontal cortex [14].

Studies using Tg AD mouse models have reported that TgCRND8 and 5XFAD mice do not rely on place/spatial strategies when navigating the hidden escape platform, instead using a cued/response strategy [3, 15, 16]. Our previous study evaluated the strategy selection in 3-month-old 5XFAD mice using the same training procedure that had been previously applied to inbred mice [3]. Like the two inbred mice, 5XFAD and non-Tg mice exhibited comparable escape performances in place/spatial and cued/response training at the learning strategy-switching task from cued/response to place/spatial training. However, compared to non-Tg mice, 5XFAD mice were more likely to use a cued/response strategy during place/spatial training and showed a stronger preference in the strategy preference test. In the learning strategy-switching task from place/spatial to cued/response training, 5XFAD showed no improved performance in the latencies to find a hidden platform relative to non-Tg mice. The latencies to find a visual platform and the strategy preference test did not differ between the two. Therefore, our subsequent study adapted the learning strategy-switching task from cued/response to place/spatial training and the other measures to assess behavioral aspects not captured by escape latency alone to evaluate the cognitive status of the 4-month-old 5XFAD mice. As demonstrated in our 5XFAD study [3], parameters such as escape latency do not adequately reflect search strategies. Hence, we also analyzed the swimming patterns of mice and assessed their search strategies in place/spatial training based on the previous report [15, 17]. 5XFAD and non-Tg mice showed no significant differences in the cued/response and place/spatial training, although 5XFAD mice showed slightly less performance than non-Tg mice in the place/spatial training. However, 5XFAD mice used less spatial strategy in the place/spatial training [16]. Therefore, the present study investigated the performance of 11-month-old APP/PS1 Tg mice at the learning strategy-switching task from cued/response to place/spatial training. To minimize potential transfer effects from prior water maze training, object and location recognition memory were assessed in APP/PS1 Tg and non-Tg mice three months later. We further examined AD-related pathological characteristics in the medial prefrontal cortex (mPFC), retrosplenial cortex (RSC), and hippocampus.

MATERIALS AND METHODS

Animals

APPswe/PSEN1dE9 double transgenic mice (B6; C3-Tg [APPswe, PSEN1dE9] 85Dbo/Mmjax, stock #004462; MMRRC stock #34829-JAX) were purchased from Jackson Laboratory (Bar Harbor, ME, USA), and bred at Korea Institute of Science and Technology (KIST) and Konkuk University. These APP/PS1 mice were maintained in-house on a mixed C57BL/6J and C3H/HeJ genetic background, and hemizygous transgenic mice were bred with wildtype littermates for colony maintenance. Wildtype littermates were used as the non-Tg control group in all experiments. They were housed in a climate-controlled vivarium (temperature, 22°C ±1°C; humidity, 50%±10%). The mice were housed under a 12:12-h light-dark cycle, with lights on at 8:00 a.m. Food and water were provided ad libitum. Genotype confirmation was performed using polymerase chain reaction (PCR) and electrophoresis. Eleven-month-old male non-Tg (n=12) and APP/PS1 (n=12) mice were used for the experiments. All experiments were conducted in accordance with the guidelines of the Institutional Animal Care and Use Committee of Konkuk University (approval no KU20069).

Genotyping

Genotyping was performed as previously described [18]. PCR was performed using 2X Biomix (Meridian Bioscience, Cincinnati, OH, USA), DNA, and primers. The primers used for Non-Tg and APP/PS1 mice were: 5-AAT-AGA-GAA-CGG-CAG-GAG-CA-3, 5-GCC-ATG-AGG-GCA-CTA-ATC-AT-3, 5-CTA-GGC-CAC-AGA-ATT-GAA-AGA-TCT-3, and 5-GTA-GGT-GGA-AAT-TCT-AGC-ATC-ATC-C-3. The control band size was 324 bp, and the transgene showed a band at 608 bp (Supplementary Fig. 1). APP/PS1 mice showed both bands. After PCR was completed, electrophoresis was performed on a 1.5% agarose gel. After electrophoresis, the bands were confirmed using an ImageQuant LAS 500 (GE Healthcare, Chicago, IL, USA).

Learning strategy-switching task

The learning strategy-switching task in the Morris water maze was conducted as previously described [3, 9] with modifications. Non-Tg (n=12) and APP/PS1 mice (n=12) underwent a learning strategy-switching task comprising two training sessions. Cued/response training was conducted for 4 days, followed by place/spatial training for 4 days, with four trials/day (10-min intertrial interval, maximum trial duration of 60 s, with 20 s on the platform at the end of each trial). During cued/response training, a visible platform was used and moved to a different location in the pool between trials. Blank white curtains were drawn around the pool to occlude the extra-maze cues. During place/spatial training, mice underwent each trial, beginning at one of four equidistantly located positions around the perimeter of the pool. The starting position varied across the trials, whereas the location of the platform remained constant throughout the training trials. A submerged hidden platform was used, and the pool was surrounded by a white curtain marked with three different shapes of extra visual cues. The mice were placed in the water facing the wall, and allowed to swim for up to 60 s. The trial ended when a mouse climbed onto the available platform or after 60 s had elapsed. If the mouse could not find the platform during the trial, the experimenter placed it on the platform. All the mice were left on the platform for 20 s and then transferred to a holding cage for a 10-minute intertrial interval. The experimenter observed the search strategy adopted by the mice for every trial during place/spatial training. On day 9, a competition test was administered, during which the visible platform was positioned opposite to its placement on the place/spatial training days. Two trials were assigned starting points equidistant from the two platform locations. Video recordings were analyzed to determine whether the mice swam to a previously hidden platform location before escaping to the visible platform.

Novel object/novel object location recognition tasks

Novel object recognition (NOR) and novel object location recognition (NOL) tasks were conducted as previously described [19], with some modifications. Tests were performed inside a black open-field square apparatus (27 cm×34 cm×26.5 cm) with a constant-masking white noise source at 70 dB. Twenty-three mice underwent the object/location novelty recognition test: non-Tg (n=11) and APP-PS1 mice (n=12). The mice were handled for 5 min daily and placed inside a box for 10 min daily to familiarize themselves with transportation and the empty arena for five consecutive days. The following day, the familiarization phase of the NOR task was conducted. Two identical sample objects were presented at both corners of the apparatus, whereas the mice were placed at the opposite corners. The test phase was initiated 24 h after the familiarization phase, during which the mice were placed again in the arena, and one of the sample objects was replaced with a novel object (a corner on which the novel object was balanced between the mice). During the familiarization phase, mice were allowed to freely explore the arena for 10 min. During the test phase, mice were allowed to explore the arena for 5 min freely, and the time spent exploring each object was recorded for 3 min. The mice were then returned to their home cages. The following day, an NOL task was performed. During the familiarization phase, two identical objects were placed at the two corners of the apparatus. Twenty-four hours later, in the novel object location phase, one of the two objects was relocated to a novel location in the arena, and the mice were again tested in the same manner as before. The object exploration ratio was computed as the ratio of the time spent exploring a specific object to the total exploration time. The preference for a novel object or novel location was calculated as the time spent exploring the novel object or location relative to the total time spent exploring both the novel and the familiar object using the following formula: novel/(novel+familiar)×100%.

Behavioral and statistical analyses

The escape latencies to find the visual or hidden platform, as well as the average proximity of each animal’s position to the visual or hidden platform position, termed the search error [20], were measured. Escape latencies or search errors during cued/response and place/spatial training were analyzed separately using repeated-measures two-way analysis of variance (ANOVA; group×trial session [day]) to evaluate acquisition. The swimming search pattern was classified as either a place/spatial search strategy or a non-spatial search (cued/response) strategy for each mouse in every trial of the place/spatial training, which was conducted after the cued training, according to the criteria adopted by previous studies [3, 15]. Specifically, by the criteria described in previous studies [3, 15], mice were considered to employ a place/spatial search strategy when they persistently searched around the platform within a target quadrant or swam directly to it. In contrast, swimming patterns such as thigmotaxis, random search, or searching at a fixed distance from the pool wall were categorized as non-spatial (cued) strategies. Student’s t-tests were applied to evaluate the differences between APP/PS1 and non-Tg mice. For the competition test, mice were subjected to two trials. Mice designated as using a “place/spatial strategy” visited the location where the platform had been located the previous training days, before escaping to the newly located visible platform. In contrast, mice using the “cued/response strategy” swam directly to the visible platform in its new location. Representative swimming paths of mice using these two strategies are shown in Supplementary Fig. 2. Using the criteria established by McDonald and White [21, 22], mice were classified using a place/spatial strategy if they visited the previous platform location during either trial. Otherwise, mice were classified using a cued/response strategy. A χ² analysis was further employed to evaluate differences in the frequency of strategies during the competition test.

In the NOR/NOL task, because the exploration ratio of the two objects was measured within the same animal during both familiarization and test phases, a one-sample t-test (two-tailed significance) was applied to analyze behavioral data. Comparisons between the two groups, including AD-related pathologic markers, were performed using independent t-tests. Differences were considered significant at p<0.05. Data are expressed as the standard error of the mean (SEM). SPSS Statistics 25 (IBM, Armonk, NY, USA) and the Prism 9 software (GraphPad Software, San Diego, CA, USA) were used for statistical analyses, and graphical figures were created.

Brain preparation

After completing the NOR/NOL task, mouse brains were collected. All mice were deeply anesthetized with isoflurane and perfused with ice-cold 0.01 M phosphate-buffered saline (PBS), followed by 4% paraformaldehyde in 0.01 M PBS. Brains were removed immediately and post-fixed for 48 h in 4% paraformaldehyde in 0.01 M PBS at 4 °C. After fixation, the brain samples were transferred into 30% sucrose in 0.01 M PBS for embedding in a glass jar for 48 h. Samples were rapidly frozen on dry ice and stored at -70°C. The microtome of Tissue-tek^® embedding was used to obtain a coronal section with a thickness of 30 μm. The selected sections were stored in cryoprotectant (30% ethylene glycol, 25% glycerol, 25% 0.1 M phosphate buffer, and 20% distilled water) before staining.

Immunofluorescent histology

Brain sections containing the mPFC, RSC, hippocampus and dorsal striatum were used to evaluate beta-amyloid and neuroinflammation markers (bregma anterior-posterior, from +2.34 mm to +1.94 mm for the prefrontal cortex; from -1.22 mm to -2.46 mm for the retrosplenial cortex and the hippocampus; from -1.22 mm to -1.58 mm for the dorsal striatum). At least six sections were selected per animal for staining. These brain sections were blocked in 10% fetal horse serum (GIBCO, Grand Island, NY, USA) and 3% Triton X-100 in phosphate-buffered saline for 2 h at room temperature, and subsequently incubated with primary guinea pig anti-NeuN (Sigma-Aldrich, St. Louis, MO, USA; 1:500), rabbit anti-Iba-1 antibody (Wako, Osaka, Japan; 1:500), goat anti-GFAP (Abcam, Cambridge, England, UK; 1:1,000), and mouse anti-4G8 antibody (Biolegend, San Diego, CA, USA; 1:500) for 20 h at room temperature. After washing, sections were incubated in an Alexa dye-conjugated secondary antibody solution including Alexa 488 conjugated donkey anti-mouse antibody, Alexa 568-conjugated donkey anti-rabbit antibody, and/or Alexa 633-conjugated donkey anti-goat antibody (Invitrogen, Carlsbad, CA, USA; 1:200). After staining, the sections were mounted with the ProLong Gold Antifade reagent (Invitrogen) on silane-coated glass slides and covered with coverslips. Before observing the sections, the stained sections were kept in the freezer at -20°C. Stained sections were observed under a confocal microscope (LSM 800; Carl Zeiss, Oberkochen, Germany), and the mean fluorescence intensity was finally measured using the ZEISS ZEN microscopy software.

Histological data analysis

Four regions of interest (ROIs)—the mPFC, hippocampus, RSC, and dorsal striatum—were selected for the analysis. These ROIs included all cortical layers in each hemisphere, and the values from both hemispheres of a single section were measured and averaged for quantification. The mean intensities were normalized to the average values in the non-Tg group. Comparisons between the two groups for mean fluorescence intensity were performed using an independent t-test.

RESULTS

11-month-old APP/PS1 mice used the cued/response strategy more than the place/spatial strategy

In the learning strategy switching task, 11-month-old non-Tg control mice and APP/PS1 mice received cued/response training for 4 days, followed by spatial training for 4 days (Fig. 1A). All mice showed decreased latencies throughout the cued/response training (session: F_3,66=33.72, p<0.001), but APP/PS1 mice took longer to locate the visual platform (group: F_1,22=6.22, p<0.05; Fig. 1B). There was no interaction effect of session×group (F_3,66=0.55, p=0.65). The analysis of search errors was consistent with the latency results (session: F_3,66=19.95, p<0.001; group: F_1,22=9.14, p<0.01; interaction: F_3,66=0.56, p=0.65; Fig. 1C). All mice exhibited decreased latencies (session: F_3,66=2.78, p<0.05) and search errors (session: F_3,66=10.87, p<0.01) over the course of the place/spatial training. No between-group effects (latencies: F_1,22=2.62, p=0.10; search errors: F_1,22=2.90, p=0.10) or interaction effects of session×group (latencies: F_3,66=0.36, p=0.79; search errors: F_3,66=0.35, p=0.79) were observed.

Fig. 1.

Performance of 11-month-old APP/PS1 and non-transgenic (Tg) mice in the visual and hidden platform training in the Morris water maze task. (A) Overview of the behavioral experimental schedule. (B, C) Significant differences in latency and search error were observed between non-Tg and APP/PS1 mice during visual platform training but not during hidden platform training. (D) Non-Tg and APP/PS1 mice showed equivalent swimming speed. (E) In the hidden platform training, APP/PS1 mice employed a place/spatial search strategy less frequently than non-Tg mice. (F) No difference in the percentage of place or spatial search strategy was found between the two groups on the competition test. Non-Tg (n=12), APP/PS1 (n=12), Data are expressed as mean±SEM. *p<0.05, **p<0.01.

Swimming speeds (Fig. 1D) in all mice increased throughout cued/response (F_3,66=34.09, p<0.001) and place/spatial training (F_3,66=4.20, p<0.01). No between-group effects were observed (cued/response: F_1,22=2.07, p=0.16; place/spatial: F_1,22=0.30, p=0.59). There was an interaction effect of session×group (cued/response: F_3,66=4.15, p<0.01; place/spatial: F_3,66=0.57, p=0.64). These results indicate no differences between the two groups regarding motor and visual functions.

We subsequently analyzed the search strategy used by mice during place/spatial training following cued/response training (Fig. 1E). APP/PS1 mice employed the cued/response strategy more frequently than control mice during place/spatial training (F_1,22=5.71, p<0.05). However, no significant session effects (F_3,66=1.78, p=0.16) or no interaction effects of session×group (F_3,66=0.76, p=0.52) were observed. In the competition test, 4 out of 12 control mice and 4 out of 12 APP/PS1 mice employed a place strategy. No difference was observed between the control and APP/PS1 mice in terms of strategy selection, as determined by the χ² analyses (χ²₁=0.00, p=1.00).

Object location memory of the 14-month-old APP/PS1 mice was impaired

Object recognition and location memory were subsequently assessed in 14-month-old APP/PS1 mice. During the NOR familiarization phase, both non-Tg and APP mice showed equivalent exploration ratios for the left (L) and right (R) objects (non-Tg: t₁₀=0.12, p=0.90; APP/PS1: t₁₁=1.017, p=0.33; Fig. 2A). In the NOR test phase, the two groups explored the novel object more than the familiar object (non-Tg: t₁₀=-2.99, p<0.01; APP/PS1: t₁₁=-3.26, p<0.01; Fig. 2A). An independent t-test revealed no between-group differences in preference for the novel object (t₂₁=0.31, p=0.76; Fig. 2C). In the NOL familiarization phase, mice in the two groups showed no significant preference for location (non-Tg: t₁₀=0.63, p=0.55; APP/PS1: t₁₁=-0.38, p=0.71; Fig. 2D). At the test phase, non-Tg mice explored the novel location object more than the familiar location object (t₁₀=-3.21, p<0.01) but APP/PS1 mice did not (t₁₁=0.20, p=0.84, Fig. 2D). Independent t-tests revealed significant differences (t_17.1=1.74, p<0.05, one-tailed; Fig. 2E). The total exploration times of the non-Tg and APP/PS1 mice during the familiarization and test phases of the NOR/NOL tasks are presented in Supplementary Fig. 3.

Fig. 2.

Performance of the 14-month-old non-Tg and APP/PS1 mice in the novel object/location recognition task. (A) Overview of the experimental procedure. All test phases (T) were conducted 24 h after the familiarization phase (F). (B) In the novel object recognition task, non-Tg and APP/PS1 mice equally explored the two identical objects placed on the left (L) and right sides (R) of the open-field box in the familiarization phase. (C) Both non-Tg and APP/PS1 mice explored the novel object (N) significantly more than the familiar object (F). No difference was found between the two. (D) In the novel object location task, non-Tg and APP/PS1 mice equally explored the two identical objects on the L and R sides of the open-field box. During the test phase, non-Tg mice preferred the object at a new location (N) over that at a familiar location (F), whereas APP/PS1 mice did not. (E) A significant difference was found in the preference for the novel location between the two. Data are expressed as mean±SEM. *p<0.05, **p<0.01, ^#p<0.05 (one-tail); Non-Tg (n=11), APP/PS1 (n=12).

14-month-old APP/PS1 mice exhibited AD-like pathological characteristics

Immunofluorescent staining using the 4G8 antibody, which reacts with amino acid residues 17~24 of Aβ, confirmed the expression and accumulation of amyloid-β (Aβ) in the brain. Further, the NeuN-positive, Iba-1-positive, and GFAP-positive signals were measured to assess the localization and activation of microglia and astrocytes. These analyses revealed Aß accumulation in the mPFC, RSC, hippocampus, and dorsal striatum of the APP/PS1 mice but not in non-Tg mice, with glial activation clustering around these deposits (Fig. 3A, 4A, 5A, and Supplementary Fig. 4). The intensities of the 4G8-positive, Iba-1-positive, and GFAP-positive signals were significantly increased in the mPFC of APP/PS1 mice compared with those of non-Tg mice (4G8: t₇=6.09, p<0.001; Iba-1: t₇=3.64, p<0.01; GFAP: t₇=11.21, p<0.001) (Fig. 3B~D). However, no difference was observed between the two groups in the intensity of NeuN-positive signals in the mPFC (t₇=0.69, p=0.51; Fig. 3E).

Fig. 3.

Increases in amyloid-β, microglial activation, and astroglial activation in the medial prefrontal cortex of 14-month-old APP/PS1 mice. (A) Representative image showing immunofluorescence staining of the medial prefrontal cortex of the non-Tg and APP/PS1 mice. (B~E) Quantification of 4G8-, Iba-1-, GFAP-, and NeuN-positive signals in the medial prefrontal cortex of the non-Tg and APP/PS1 mice. Intensities of 4G8-, Iba-1-, and GFAP-positive signals in the medial prefrontal cortex of the APP/PS1 mice were higher than those of non-Tg mice. Non-Tg mice (n=5); APP/PS1 mice (n=4). Data are expressed as mean±SEM. Scale bar=200 μm; Scale bar in the lower rectangle=100 μm. **p<0.01, ***p<0.001.

Fig. 4.

Increases in amyloid-β, microglial activation, and astroglial activation in the retrosplenial cortex of 14-month-old APP/PS1 mice. (A) Representative images showing immunofluorescence staining of the retrosplenial cortex of non-Tg and APP/PS1 mice. (B~E) Quantification of 4G8-, Iba-1-, GFAP-, and NeuN-positive signals in the retrosplenial cortex of the non-Tg and APP/PS1 mice. Intensities of 4G8-, Iba-1-, and GFAP-positive signals in the retrosplenial cortex of the APP/PS1 mice were higher than those of non-Tg mice. Non-Tg (n=5); APP/PS1 (n=4). Data are expressed as mean±SEM. Scale bar=200 μm; Scale bar in the lower rectangle=100 μm. p<0.05, **p<0.01, ^#p<0.05 (one-tail).

Fig. 5.

Increases in amyloid-β, microglial activation, and astroglial activation in the hippocampus of 14-month-old APP/PS1 mice. (A) Representative images showing immunofluorescence staining of the hippocampus of non-Tg and APP/PS1 mice. (B~E) Quantification of 4G8-, Iba-1-, GFAP-, and NeuN-positive signals in the hippocampus of non-Tg and APP/PS1 mice. Intensities of 4G8-, Iba-1-, and GFAP-positive signals in the hippocampus of APP/PS1 mice were higher than those of non-Tg mice. Non-Tg (n=5); APP/PS1 (n=4). Data are expressed as mean±SEM. Scale bar=200 μm; Scale bar in the lower rectangle=100 μm. p<0.05, **p<0.01, ^#p<0.05 (one-tail).

The intensities of 4G8-positive, Iba-1-positive, and GFAP-positive signals were significantly higher in the RSC of APP/PS1 mice than in those of non-Tg mice (4G8: t_3.01=2.55, p<0.05, one-tail; Iba-1: t₇=3.34, p<0.01; GFAP: t_3.00=8.006, p<0.01; Fig. 4B~D). No difference between the two groups was found in the intensity of NeuN-positive signals in the RSC (t_3.05=0.58, p=0.60; Fig. 4E).

The intensities of Iba-1-positive and GFAP-positive signals were significantly higher in the hippocampus of APP/PS1 mice than in the mPFC of non-Tg mice (Iba-1: t_3.04=2.47, p<0.05, one-tail; GFAP: t_3.13=5.53, p<0.01; Fig. 5C, D). The intensities of the 4G8-positive signals of the APP/PS1 were slightly higher than those in non-Tg mice, which were not significant (t_3.12 =1.98, p=0.07, one-tail). No difference was observed between the two groups in the intensity of NeuN-positive signals in the hippocampus (t_3.05=0.578, p=0.60) (Fig. 5B, E).

DISCUSSION

Several studies have suggested that different brain structures are engaged during navigational tasks, depending on the type of learning involved [11, 12, 22]. For example, the place/spatial learning strategy in the Morris water maze task is an allocentric navigational approach that relies on spatial information and the hippocampal system. Conversely, a cued/response-learning strategy employs an egocentric navigation approach that requires instrumental learning and involvement of the striatal system. Some prior studies have indicated that both the PFC and the hippocampus play a role in switching between learning cued/response and place/spatial strategies [14, 23-25]

Previously, we examined the learning strategy preferences of 5-month-old 5XFAD mice carrying five familial AD mutations (Swedish, Florida, and London human APP mutations; M146L and L286V PS1 mutations) [3]. In other words, 5XFAD and non-Tg mice serially underwent cued/response training for different visible platform locations and place/spatial training for a single hidden platform location in a water maze. In the competition test, these mice were required to choose between a spatially hidden platform and a visible platform located opposite the hidden platform. 5XFAD and non-Tg mice performed comparably in cued/response and place/spatial training. However, an analysis of the search strategies adopted by mice during place/spatial training revealed that non-Tg mice employed more place/spatial search strategies than cued/response search strategies, whereas 5XFAD mice did not. In the strategy competition test, non-transgenic mice employed significantly more place/spatial search strategies than 5XFAD mice. In our subsequent study, 4-month-old 5XFAD mice employed cued/response search strategies, as 5-month-old 5XFAD mice did during place/spatial training following cued/response training; however, no difference between control and 5XFAD mice was observed in the strategy competition test [16].

Similar to the 5XFAD study, the present study evaluated the learning strategy preferences of 11-month-old APP/PS1 mice using a learning-switching task. In the initial cued/response training, APP/PS1 mice found the visual platform more slowly than non-Tg mice. However, in subsequent place/spatial training, no difference between the two groups was observed; instead, APP/PS1 mice employed more cued/response search strategies than non-Tg mice. In addition, no differences were observed between the two groups in the strategy competition tests. The unexpected difference from the previous results with inbred and 5XFAD mice was the impairment of 11-month-old APP/PS1 mice in cued/response training. Therefore, we examined AD-pathological characteristics in the striatum, a brain region involved critically in cued/response learning [10-12], observing only minor accumulation of the amyloid-β, causing dysfunction of the striatum. However, there are other possibilities to consider. For example, in the cued/response setting of the Morris water maze, initial swimming paths rely on the extra-maze cues, while subsequent swimming paths to the platform depend more on intra-maze cues [26, 27]. This suggests a cooperative interaction between the hippocampus and striatum, even during cued/response learning. Hence, even if only one of these two memory systems is impaired, cued/response learning may be impaired.

In the learning strategy-switching task, from cued/response training to place/spatial training, a behavioral feature commonly observed in 11-month-old APP/PS1 and 4- or 5-month-old 5XFAD mice was the continuous use of the cued/response search strategy in place/spatial training compared to non-Tg mice. The persistent use of a previously learned cued/response strategy despite a change in task demands from a visible platform to a hidden platform requiring spatial navigation may indicate a deficit in cognitive flexibility, specifically in set-shifting, a key executive function often impaired in AD animal models [28]. Consistent with these findings in animal models, AD patients also exhibit impairments in set-shifting abilities and reversal learning, as demonstrated by clinical neuropsychological assessments [29-31]. Cognitive flexibility has been shown to be associated with amyloid-β (Aβ) plaque accumulation in the hippocampus and prefrontal cortex. Additionally, its alterations are highly correlated with the expression levels of Aβ-degrading enzymes, including neprilysin [32]. To reduce the impact of the learning experience in the Morris water maze on subsequent behavioral tasks, the same mice underwent a novel object/location recognition task 3 months later. The object recognition memory of 14-month-old APP/PS1 mice was intact, but their location recognition memory was impaired. Therefore, we investigated AD-related pathological features in the mPFC, RSC, and hippocampus, as these regions play crucial roles in executing behavioral tasks.

The hippocampus is a key neural structure mediating place/spatial learning and object/location recognition memory [33-37]. The mPFC is engaged in shifting learning strategies [14, 23-25], while the RSC is involved in shifting and relating perspectives for spatial cognition, functioning as a link for frontal functions [37-40]. These brain structures exhibit an accumulation of amyloid-β in the 5XFAD mice [3, 37]. Therefore, we examined amyloid-β expression levels, microglial activation, and astroglial activation, observing amyloid accumulation and glial activation in all three structures.

In the present study, 11-month-old non-Tg mice demonstrated poor performance in the learning strategy-switching task, shifting from cued/response training to place/spatial training, compared with the relatively younger non-Tg mice in our previous study [3]. This age-related decline in performance, such as search strategy, in the AD and non-Tg mice [17] may have contributed to the absence of significant differences between APP/PS1 and non-Tg mice in the subsequent strategy competition test, potentially masking genotype-specific impairments. Therefore, it would be desirable to examine age-associated cognitive decline in animal models of neurodegenerative disorders, including AD, by designing two control groups: a young control group and a control group of the same age [20]. The cognitive performance of the two control groups provided a graded measure of age-related cognitive impairment, distinguishing cognitive loss due to disease.

In conclusion, the key findings of this study were poor performance in the initially presented cued/response task and continued reliance on the cued/response search strategy in the subsequently switched place/spatial task, where the place/spatial search strategy must be adapted to locate a hidden platform efficiently. Given that search strategy analysis in the Morris water maze task reveals early spatial navigation deficits of AD Tg mice [17], the search strategy analysis and the performance of the cued/spatial search strategy at the learning strategy-switching task could be instrumental in future studies investigating cognitive function in animal models with abundant Aβ accumulation or other AD features, as well as in evaluating potential interventions.