Comparison of memory, affective behavior, and neuropathology in APP^NLGF knock-in mice to 5xFAD and APP/PS1 mice

Abstract

Transgenic mouse models of Aβ amyloidosis generated by knock-in of a humanized Aβ sequence can offer some advantages over the transgenic models that overexpress amyloid precursor protein (APP). However, systematic comparison of memory, behavioral, and neuropathological phenotypes between these models has not been well documented. In this study, we compared memory and affective behavior in APP^NLGF mice, an APP knock-in model, to two widely used mouse models of Alzheimer’s disease, 5xFAD and APP/PS1 mice, at 10 months of age. We found that, despite similar deficits in working memory, object recognition, and social recognition memory, APP^NLGF and 5xFAD mice but not APP/PS1 mice show compelling anxiety- and depressive-like behavior, and exhibited a marked impairment of social interaction. We quantified corticolimbic Aβ plaques, which were lowest in APP^NLGF, intermediate in APP/PS1, and highest in 5xFAD mice. Interestingly, analysis of plaque size revealed that plaques were largest in APP/PS1 mice, intermediate in 5xFAD mice, and smallest in APP^NLGF mice. Finally, we observed a significantly higher percentage of the area occupied by plaques in both 5xFAD and APP/PS1 relative to APP^NLGF mice. Overall, our findings suggest that the severity of Aβ neuropathology is not directly correlated with memory and affective behavior impairments between these three transgenic mouse models. Additionally, APP^NLGF may represent a valid mouse model for studying AD comorbid with anxiety and depression.

1. Introduction

Alzheimer’s disease (AD) is a neurodegenerative disorder with progressive cognitive decline, yet over 90% of patients also develop neuropsychiatric symptoms including anxiety and depression during the disease progression (Scheltens et al., 2016; Winblad et al., 2016). The neuropathology of AD includes the accumulation of amyloid-β peptide (Aβ) as extracellular plaques, aggregation of hyperphosphorylated tau protein as intracellular neurofibrillary tangles, and activation of multiple neuroinflammatory pathways that, ultimately, lead to neuronal death (Braak et al., 1991; Hyman et al., 2012; Serrano-Pozo et al., 2011). While the majority of AD research is focused on elucidating the mechanisms underlying cognitive decline, little is known about the molecular pathogenesis of neuropsychiatric symptoms in AD. Further, the relationship between affective symptoms (i.e., anxiety and depression) and cognitive decline is poorly understood.

Mouse models of AD represent essential research tools for dissecting the cellular and molecular pathways underlying AD pathogenesis and testing potential therapeutic approaches (Epis et al., 2010; Puzzo et al., 2015). Several transgenic mouse lines overexpressing amyloid precursor protein (APP), such as APP/PS1 and 5xFAD mice, have been developed and widely used as experimental models for Aβ amyloidosis (Bilkei-Gorzo, 2014; Webster et al., 2014). Both memory and neuropathological changes have been well characterized in these two AD models. In the APP/PS1 mouse line, two genetic mutations are combined to reach elevated Aβ levels: the Swedish mutation, which overexpresses APP and leads to elevated amyloid production (Citron et al., 1992), together with the presenilin-1 L166P mutation, which impairs amyloid protein processing leading to elevated Aβ42 levels (Kurt et al., 2001; Radde et al., 2006). The APP/PS1 strain is one of the most commonly used AD models because of its stable genetic background and the early onset of pathological changes (Radde et al., 2006). Brain amyloidosis in APP/PS1 mice starts at 2-months of age, with declines in working and spatial memories in the early (5-6 months) and late (15 months) stages, respectively (Ferguson et al., 2013; Lagadec et al., 2012). In 2006, an aggressive transgenic line, 5xFAD, was developed with five mutations: the Swedish mutation (K670N/M671L); the Florida (1716V) and London (V717I) APP mutations, which alter APP processing and lead to a higher ratio of amyloidogenic Aβ production; and two presenilin mutations (M146L and L286V) (Oakley et al., 2006). 5xFAD mice show intracellular and extracellular Aβ accumulation at 1.5 and 2 months of age, respectively, followed by selective neuronal death (Devi and Ohno, 2010; Oakley et al., 2006; Kalinin et al., 2012). These neuropathological changes drive cognitive and affective behavioral impairment (Girard et al., 2013; Jawhar et al., 2012). 5xFAD mice develop cognitive deficits by 3 months of age in spatial working memory (Ohno et al., 2007; Urano and Tohda, 2010), followed by associative learning impairment in fear conditioning (Devi and Ohno, 2010; Ohno et al., 2007) and working memory deficits in Y-maze at 6 months of age (Devi and Ohno, 2012; Oakley et al., 2006; Shukla et al., 2013). Although these models are ubiquitous in the field of AD research, the non-physiological overexpression of APP results in extra-production of several APP fragments in addition to Aβ (Balducci and Forloni, 2011; Sasaguri et al., 2017). As a result, it is unclear whether the neuropathological and behavioral effects observed in these AD mouse models are actually caused by Aβ itself or by the other APP fragments. Additionally, the overexpression of APP may disturb its physiological function in these models. In recent years, several new alternative AD models such as APP^NLGF mice have been established that utilize an APP knock-in strategy to induce Aβ pathology by avoiding the non-physiological overexpression of APP in the mouse brain (Nilsson et al., 2014; Saito et al., 2014). In APP^NLGF mice, the murine Aβ sequence has been “humanized” by changing three amino acids that differ between the mouse and human proteins (Saito et al., 2014). In addition, the genetic approach in this model involves the introduction of three familial AD-associated mutations directly into the endogenous mouse APP locus, harboring all three mutations within the Aβ sequence: the Swedish (NL) mutation, which induces Aβ production; the Arctic (G) mutation, which promotes Aβ aggregation (Hashimoto et al., 2011; Tsubuki et al., 2003), and the Beyreuther/Iberian (F) mutation, which exacerbates the pathological phenotype by increasing the Aβ42/Aβ40 ratio (Guardia-Laguarta et al., 2010; Lichtenthaler et al., 1999). In this mouse model, Aβ amyloidosis and neuroinflammation were observed in both cortical and subcortical regions starting at 2 months of age (Masuda et al., 2016; Saito et al., 2014). Although neuronal death is not characteristic of this AD mouse model (Saito et al., 2014; Sasaguri et al., 2017), APP^NLGF mice demonstrate a lower number of hippocampal mushroom spines (Zhang et al., 2015, 2016) as well as cortical dysregulation of neural circuit activities (Nakazono et al., 2017).

Despite extensive characterization of biochemical properties and cognitive deficits in APP^NLGF mice, less is known about behavioral changes in this model. Specifically, only a few recent studies have suggested alterations of anxiety-like behavior in APP^NLGF mice (Sakakibara et al., 2018; Latif-Hernandez et al., 2019; Pervolaraki et al., 2019). Most importantly, prior studies have not attemped to directly compare memory and neuropathology of this knock-in model to more traditional APP overexpressing models such as 5xFAD and APP/PS1. The aim of this study is to characterize memory, affective behavior, and amyloid plaques in APP^NLGF mice and to directly compare these findings to 5xFAD and APP/PS1 mice. Our results demonstrate that, despite comparable memory deficits at 10 months of age on measures of working memory, object recognition, and social recognition, affective behavior (i.e., anxiety and depressive-like behavior and social interaction) are markably different in these three AD mouse models. Furthermore, the number, size, and percentage area occupied by plaques in three key-corticolimbic regions involved in the modulation of memory and emotion (prefrontal cortex, hippocampus and amygdala) differ between these modes. Overall, these results suggest that APP^NLGF mice exhibit not only significant memory deficits, but also marked anxiety-like behavior, depressive-like behavior, and impairment of social interaction, despite having much fewer Aβ plaques in corticolimbic regions than 5xFAD and APP/PS1 mice. This detailed characterization of the behavioral phenotype of APP knock-in mice will aid researchers in the appropriate selection of AD mouse models, especially as it pertains to memory deficits and combined alterations of the affective behavior.

2. Material and Methods

2.1. Animals

Three transgenic AD mouse models were used for this study, namely (i) APP^NLGF homozygous (strain name: Apptm^3.1Tcs/Apptm^3.1Tcs; mutations: APP KM670/671NL (Swedish), APP I716F (Iberian), APP E693G (Arctic); modification: APP: knock-in; genetic background: C57BL/6); (ii) 5xFAD heterozygous (strain name: B6.Cg-Tg (APPSwFlLon,PSEN1*M146L*L286V)6799Vas/Mmjax; mutations: APP KM670/671NL (Swedish), APP I716V (Florida), APP V717I (London), PSEN1 M146L (A>C), PSEN1 L286V; modification: APP: transgenic; PSEN1: transgenic; genetic background: C57BL/6); and (iii) APP/PS1 heterozygous (strain name: APPswePsen1de9; B6.Cg-Tg(APPswe,PSEN1dE9)85Dbo/Mmjax; mutations: APP KM670/671NL (Swedish), PSEN1: deltaE9; modification: APP: transgenic; PSEN1: transgenic; genetic background: C57BL/6J). APP^NLGF mice at approximately 8-10 weeks of age were kindly provided by RIKEN (Japan); breeding cages consisted of male and female heterozygous APP^NLGF mice. 5xFAD, APP/PS1, and C57BL/6 mice at approximately 8-10 weeks of age were purchased from Jackson Laboratories. For these two mouse strains, the breeding pairs were set up by housing male 5xFAD and APP/PS1 mice with female C57BL/6 mice. Genotyping for the offsprings was done at 21-24 days post birth by PCR with APP-NLGF primers (forward: 5′-CTC CTT GTG GCT GGC GGT CAC AC-3′; reverse: 5′-CTA TCG TGG ACC GAG AAT GGT CAT G-3′) for the APP^NLGF strain, hPS1 primers (forward: 5′-GCT TTT TCC AGC TCT CAT TTA CTC-3′; reverse: 5′-AAA AAT GAT GGA ATG CTA ATT GTT-3′) for the 5xFAD strain, and PS1 primers (forward: 5′-AAT AGA GAA CGG CAG GAG CA-3′; reverse: 5′-GCC ATG AGG GCA CTA ATC AT-3′) for the APP/PS1 strain. The offsprings were then separated into cages based on sex with their original littermates housing together until 10 months of age. A total number of 80 mice (40 males and 40 females equally distributed within each group: C57BL/6 wild-type (WT) mice, n=20; APP^NLGF, n=20; 5xFAD, n=20; APP/PS1; n=20) were used for this study. Animals were housed in groups of 5 on a 12-hour light/dark cycle (lights on at 8:00 am) and given food and water ad libitum. All procedures were performed according to NIH guidelines for the treatment of animals and the Current Guide for the Care and Use of Laboratory Animals (2011, 8th edition) under the protocol # IS0000543 approved by the Northwestern University Animal Care and Use Committee.

2.2. Behavioral tests

Mice were handled once a day for five consecutive days before starting the behavioral experiments. A series of behavioral tests was conducted at 10 months of age with at least two days of rest between tests. All testing sessions started at 9 am. Before each test, mice were transferred in the testing room for at least 1 hour before starting the experiments (pre-test acclimation period) to reduce stress levels from exposure to the new environment. Importantly, between each session, the chambers, tools, and instruments were thoroughly cleaned with 70% alcohol to eliminate odors. The testing sequence was as follows: a) locomotor activity and open field; b) novel object recognition; c) spontaneous alternation; d) social interaction and social recognition; e) working memory version of Morris water maze; f) light-dark box; g) tail suspension test.

2.2.1. Locomotor activity and open field (LA and OF)

The apparatus consisted of an evenly illuminated plexiglass box (40 cm × 40 cm × 40 cm) placed on a stable table with overhead video recording. Locomotor activity was defined as the distance traveled (m) during a 10 mins trial (Viana et al., 2013). Animal activity was recorded using an automated tracking system (Any-Maze, Stoelting, Wood Dale, IL). Using this recorded data, anxiety-like behavior was evaluated by the time spent and the number of entries in the periphery (outer 2/3rds) or center (inner third) of the chamber.

2.2.2. Novel object recognition (NOR)

This test was conducted in the same open field chambers as described for LA and OF. Two sets of objects were used that were consistent in height and volume but different in shape and color. Mice were individually habituated to the arena for 10 minutes each day for 3 consecutive days before data acquisition. The experimental session consisted of 3 phases: a) acquisition trial (10 mins); b) inter-trial-interval (ITI; home-cage, 1 hr); c) retention trial (10 mins). During acquisition, mice were videotaped while exploring the arena with two identical objects placed diagonally across from each other. Following the ITI, one of the objects was replaced by a novel object of a similar size but different color and shape in a counterbalanced manner, and animals were placed back in the arena for the retention trial. An experimenter blind to the experimental groups scored the amount of time spent exploring each object during the acquisition and retention trials using two stopwatches. Exploration was defined as touching, leaning on the object, orienting the head towards the object, and sniffing within < 1.0 cm. Climbing on top of the objects was not considered exploration. The time spent exploring each object during the retention trial was used to calculate the discrimination index (DI), or the difference in exploration time expressed as a proportion of the total time spent exploring the two objects.

2.2.3. Spontaneous alternation (SA)

The Y-maze apparatus consisted of three-arms (5 cm wide × 21 cm long × 15.5 cm high) with three different special cues positioned in the top inner part of each arm. The apparatus was placed on a stable table with overhead video recording. Mice were placed in arm “A” (starting point, excluded from the analysis) facing the end of the arm and were allowed to freely explore the apparatus for 5 mins without investigator presence while a camera recorded their movements (Any-Maze, Stoelting, Wood Dale, IL). Spontaneous alternation was defined as discrete and successive entries into each open arm, including events where the animal directly progresses from one arm to the next in consecutive fashion (i.e., ABC, ACB, BAC, BCA, CAB, and CBA) without reentering the two previously visited arms. The spontaneous alternation percentage was calculated by dividing the number of total successful alternations by the total number of possible alternations (i.e., the number of total entries minus two) multiplied by 100.

2.2.4. Social interaction and social recognition (SI and SR)

The chamber (60 cm wide × 60 cm long × 40 cm high) was divided equally into three compartments by using removable dividers with a small opening that allowed the experimental mouse to move from one compartment to an adjacent one. The apparatus was placed on the floor with overhead video recording. This protocol consisted of three days. On day 1, mice were exposed to the chamber for 10 mins (habituation day). On day 2 (24 hrs after habituation session), to induce social interactions, a mouse of the same strain, sex, and age was placed in the test chamber under a cylinder (10 cm circumference, 8 cm high). The experimental mouse started the session in the central compartment and was allowed to explore freely for 10 mins. Mice were naïve to their same-strain partners and the test chamber before starting the test. The amount of time during which the mice engaged in social interactions as well as the number of interactions were scored by an observer blind to the experimental group. On day 3 (24 hrs after social interaction session), mice were re-exposed to the same chamber and two mice, namely (i) one of the two familiar mice used on day 2 (“old mouse”) and (ii) a new conspecific (“new mouse”). The experimental mouse was videotaped for 10 mins; social recognition was measured by the time spent around as well as the number of interactions with the “old” and “new” mice.

2.2.5. Working memory version of Morris water maze (MWM)

For our studies, we adapted a modified version of the classic MWM protocol -also called the “delayed non-matching-to-place task”- which specifically aims to explore working memory in mice (Cho and Jaffard, 1994; Rodriguez et al., 2017). The task was conducted in a 1.8 m diameter, 0.9 m deep tank filled with tap water (23 °C), which was made opaque by non-toxic, white tempera paint. The paradigm consisted in 6 consecutive days with four trials per day. Four spatial cues and an overhead video recording were used in this protoco (Any-Maze, Stoelting, Wood Dale, IL).

The location of the platform and the starting position of the mouse changed each day and remained constant for all the trials for that day. Trials 1 and 2 were conducted between 9 and 11 am, with an ITI of 25 secs. Animals were then placed in the home cage and remained in the testing room for 3 hrs prior to trial 3. Trials 3 and 4 were conducted between 12-2 pm, with an ITI of 25 secs. At the beginning of each trial, the mouse was placed in a starting positing and had a maximum of 60 secs to localize the new position and reach the hidden platform. A trial was ended either when the mouse reached the platform or 60 secs had elapsed. After the trials 2 and 4, the mouse was removed from the tank and placed into a drying cage for 5 mins. Days 1-3 were considered habituation days; thus, data from these is not reported. Importantly, given that the modified MWM protocol evaluates the latency to reach the platform placed in the new position, data from trials 2 and 4 -but not trials 1 and 3- were used to express an index of “working memory.” Testing on days 4-6 was recorded, and the latency to find the platform as well as swim speed were averaged across days for trial 2 and 4.

2.2.6. Light dark box (LD)

The chamber (60 cm wide ×60 cm long ×40 cm high) was divided equally into two compartments by using removable dividers with a small opening that allowed the experimental mouse to move from one compartment to another. One compartment was dark (less than 5 lux), and the another was illuminated with a bright light (300 lux). The chambers were positioned on the table with overhead video recording (Any-Maze, Stoelting, Wood Dale, IL). On testing day, the animal was placed in the light compartment and was allowed to explore both the dark and light compartments for 10 mins. The time spent and number of entries in each compartment was used to measure the levels of anxiety-like behavior.

2.2.7. Tail suspension test (TST)

The TST was performed as previously described (Can et al., 2012; Locci and Pinna, 2019). Briefly, mice were suspended by applying tape (17-cm long stripes) onto the last portion of their tails. A video recorder was placed in front of the apparatus. The duration of each session was 6 mins. Videos were scored by an experimenter blind to the experimental groups by using a stopwatch. Immobility time was taken as an index of depressive-like behavior.

2.3. Tissue preparation

One week after TST, 11-month-old mice were sacrificed by injection of Euthasol® solution and perfused transcardially with 1% heparinized 0.01 M phosphate buffer (PBS) for 2 mins and then 4% paraformaldehyde for 15 mins. Brains were removed and post-fixed at 4 °C using the same fixative with 30% sucrose for 48 hrs. The brains were dissected and embedded in Tissue-Tek embedding medium (Electron Microscopy Sciences, Hatfield, PA, USA), and cut into 40 μm thick sections in the coronal plane using a cryostat (Leica CM 1850 UV, Nussloch, Germany). Selected sections were representative of prefrontal cortex, amygdala, and hippocampal CA1, CA3, and DG.

2.4. Thioflavin-S staining

To analyze fibrillar β-amyloid deposits, floating sections were washed in PBS and then stained in 1% thioflavin-S (Sigma) aqueous solution for 5 mins and differentiated in 70% alcohol for 3 mins (Guntern et al., 1992). Stained sections were mounted onto slides coated with 2% pig skin gelatin (Sigma) with a consistent orientation. Slides were coverslipped with 75% glycerol (Sigma) and allowed to dry overnight at 4 °C. Aβ plaques were analyzed on 5 sections/mouse for each brain region studied (6 mice/group, n=30 sections/group).

2.5. Aβ immunohistochemical staining

To qualitatively validate fibrillar β-amyloid deposits identified by thioflavin-S staining, we also conducted immunofluorescence staining. Slices were washed three times for 15 mins with PBS buffer, and then exposed to blocking solution (1% BSA 1x and 2% donkey serum in PBS buffer) for two hours at room temperature. Sections were then incubated overnight in the primary antibody for Aβ (beta Amyloid Recombinant Rabbit Monoclonal Antibody, Catalog # H31L21, Invitrogen, 1:1000) at 4 °C. After three washes in PBS for 15 mins, sections were incubated in secondary antibody (Goat anti-Rabbit, Alexa Fluor 488, Invitrogen, Catalog # A-11008, 1:500) for 2 hrs at room temperature and finally washed 3 times for 15 mins with PBS. Stained sections were mounted onto slides coated with 2% pig skin gelatin (Sigma) with a consistent orientation. Slides were cover slipped with 75% glycerol (Sigma) and allowed to dry overnight at 4 °C.

2.6. Amyloid plaque analysis

Plaques in the prefrontal cortex, amygdala, and hippocampus (CA1, CA3 and DG subregions) stained by both thioflavin-S (quantification) or Aβ antibody (qualitative evaluation) displayed a similar pattern in the selected brain regions. Stained brain regions were imaged using Nikon Eclipse-Ni-E fluorescent microscope at 10x magnification. The specific brain regions were identified and selected according to well established lab protocols (Dong et al., 2007). Briefly, we chose specific slices for prefrontal cortex (bregma: 2.10 mm; interaural: 5.90 mm), amygdala (bregma: −1.06 mm; interaural: 2.74 mm), and hippocampus (bregma: −2.54 mm; interaural: −1.26 mm) (Franklin and Paxinos, 2008). Histologically, prefrontal cortex was defined by two regions, the small area dorsal to the rhinal sulcus and a larger area located in the medial half of the anterior cortex (Guldin et al., 1981). Amygdala was defined by the region adjacent to the olfactory cortex, and the medial and central striatum nuclei, which includes the bed nuclei of the stria terminalis (Swanson and Petrovich, 1998). Hippocampus was defined by the series of adjacent cortical regions that form a wing shape in the inner region of the temporal lobe. Specifically, the dentate gyrus (DG) is located ventrally in the folds of the hippocampus that lie dorsal to the striatum as the first part of the trisynaptic circuit, and the cornu ammonis (CA) subregion is located in the wing-shaped portion of the hippocampus. CA1 was defined by the region included from the end of the subiculum up to the beginning of CA2 subregion, whereas CA3 was defined by the region included from the end of CA2 up to the insertion into the hilus close to the DG (Amaral, 1993). Amyloid plaques were evaluated in 5 sections from each mouse (6 mice/group) using thioflavin-S staining, and the total plaque number, plaque size, and percentage area occupied by plaques in the selected sections were measured and calculated by using ImageJ software (https://imagej.nih.gov/ij/) for statistical analysis according to previous works published by our group (Dong et al., 2004, 2007).

2.7. Statistical analysis

GraphPad Prism 7 software (San Diego, CA, 2016) was used for statistical analyses. We assessed for significant difference between the experimental groups using a one-way or two-way analysis of variance (ANOVA), followed by Tukey’s post hoc test. Data represent mean values ± S.E.M. No samples were excluded from our experiments.

3. Results

3.1. Comparison of memory deficits in APP^NLGF with 5xFAD and APP/PS1 mice

As reported in detail below, all the three AD models showed robust memory deficits compared to WT-control mice, but no significant differences were found between these transgenic mouse models (Fig. 1).

Figure 1. Impairment of subtypes of memory in APP^NLGF, 5xFAD, and APP/PS1 mice.

As shown in these graphs, all the mouse models of AD showed a marked impaired performance in (A) novel object recognition (NOR), (B) spontaneous alternation (SA), (C) working memory version of Morris water maze (MWM), and (D) social recognition (SR) tests compared to WT mice. Importantly, AD mice did not show differences in swim speed compared to WT mice. In addition, only APP/PS1 mice exhibited changes in the time spent around the old mouse (D). Finally, the number of interactions with new and old conspecific did not significantly differ between the three AD mouse strains and compared to WT mice (D). Data represent the mean ± SEM of 10 mice. *p<0.05, **p<0.01, and ***p<0.001 when compared to the sex-matched, WT-control group. Two-way ANOVA followed by Tukey’s post hoc analysis.

3.1.1. NOR

Two-way ANOVA displayed only a genotype effect on recognition memory measured by NOR (F_3,72=24.77, p<0.0001). Tukey’s post hoc tests showed a significant reduction of the discrimination index in both male (APP^NLGF: −57%, p=0.0003; 5xFAD: −59%, p=0.0001; APP/PS1: −56%, p=0.0004) and female (APP^NLGF: −65%, p<0.0001; 5xFAD: −59%, p<0.0001; APP/PS1: −55%, p=0.0003) AD mice compared to sex-matched WT mice (Fig. 1A).

3.1.2. SA

There was a genotype but not a sex effect on the percentage of spontaneous alternation (F_3,72=18.44, p<0.0001). Post hoc tests indicated that the spontaneous alteration was decreased in both males (APP^NLGF: −27%, p=0.007; 5xFAD: −32%, p=0.0008; APP/PS1: −26%, p=0.013) and females (APP^NLGF: −36%, p<0.0001; 5xFAD: −29%, p=0.002; APP/PS1: −33%, p=0.0004) of all the three AD models compared to the sex-matched WT mice (Fig. 1B).

3.1.3. Working memory version of MWM

Again, we only found a genotype effect for the working memory tested by the working memory version of MWM (F_3,72=25.79, p<0.0001). Further analyses indicated a significant increase in latency to reach the hidden platform in both male (APP^NLGF: +110%, p=0.0008; 5xFAD: +154%, p<0.0001; APP/PS1: +119%, p=0.0002) and female (APP^NLGF: +69%, p=0.011; 5xFAD: +111%, p<0.0001; APP/PS1: +62%, p=0.035) AD mice compared the sex-matched WT mice. Notably, swim speed did not significantly differ between all the AD mouse strains and when compared to sex-matched WT mice, suggesting a specific working memory dysfunction (Fig. 1C).

3.1.4. SR

Finally, we found a genotype but not sex effect in the time spent around the new conspecific in the social recognition paradigm (F_3,72=26.77, p<0.0001). Specifically, both male (APP^NLGF: −43%, p<0.0001; 5xFAD: −39%, p<0.0001; APP/PS1: −48%, p<0.0001) and female (APP^NLGF: −28%, p=0.011; 5xFAD: −25%, p<0.0001; APP/PS1: −43%, p=0.035) AD mice showed a reduction in the time spent around the new conspecific compared to the sex-matched WT mice. Moreover, we observed a genotype effect in the time spent around the old conspecific (F_3,72=7.32, p=0.0002). Post hoc analysis only found a significant effect in APP/PS1 compared to their sex-matched WT mice (males: +35%, p=0.02; females: +85%, p=0.002). There was no significant difference in the number of interactions with the new conspecific as well as the old conspecific between any experimental group (Fig. 1D).

3.2. Comparison of the affective behavior in APP^NLGF to 5xFAD and APP/PS1 mice

We also evaluated potential alterations of the affective behavior in these three AD mouse models. Despite similar memory deficits, we observed different patterns between these three AD models in anxiety-like, depressive-like, and social behavior (Fig. 2).

Figure 2. Alteration of affective behaviors in APP^NLGF, 5xFAD, and APP/PS1 mice.

(A) Both APP^NLGF and 5xFAD mice showed anxiety-like behavior in the open field test (OF). More specifically, APP^NLGF and 5xFAD mice spent less time and entered less in the center of the apparatus compared to WT mice. (B) Both APP^NLGF and 5xFAD mice spent less time and entered less in the light compartment of the light dark (LD) box, suggesting increased levels of anxiety-like behavior. (C) Besides the trend of increased locomotor activity (LA) observed in APP/PS1 mice, LA did not significantly differ among AD mice compared to WT mice. In contrast, both APP^NLGF and 5xFAD were less active compared to APP/PS1 mice. (D) In the tail suspension test (TST), both APP^NLGF and 5xFAD mice showed a significant increase of the total time of immobility, suggesting elevated depressive-like behavior compared to WT mice. (E) APP^NLGF and 5xFAD mice, but not APP/PS1 mice, spent less time and interacted fewer times with the conspecific, suggesting impairerd social behavior during the social interaction (SI) paradigm. Data represent the mean ± SEM of 10 mice. *p<0.05, **p<0.01, and ***p<0.001 when compared to WT-control mice; ^#p<0.05, ^##p<0.01, and ^###p<0.001 when compared to APP/PS1 mice. Two-way ANOVA followed by Tukey’s post hoc analysis.

3.2.1. Anxiety-like behavior and LA

When we tested the mice using the OF paradigm, two-way ANOVA indicated significant effects of genotype (F_3,72=12.11, p<0.0001) and sex (F_1,72=13.74, p=0.004) but not a genotype and sex interaction for the time spent in the center of the apparatus. Moreover, there was only a genotype effect (F_3,72=21.68, p<0.0001) for the number of entries in the center of the arena. Tukey’s post hoc analysis suggested that both male (APP^NLGF: −31%, p=0.031; 5xFAD: −35%, p=0.009) and female (APP^NLGF: −43%, 0.009; 5xFAD: −41%, p=0.013) APP^NLGFand 5xFAD mice -but not APP/PS1 mice-showed a significant reduction of the time spent in the center as compared to the sex-matched WT mice. Likewise, both male (APP^NLGF: −42%, p=0.027; 5xFAD: −40%, p=0.039) and female (APP^NLGF: −51%, p=0.005; 5xFAD: −44%, p=0.022) mice showed a significant decrease of the number of entries. However, when we compared male and females within the same strain, we did not find significant sex differences for both the time spent and number of entries in the center (Fig. 2A).

LD analyses revealed only a genotype effect for the time spent in the light (F_3,72=16.03, p<0.0001) and for the number of entries in the light (F_3,72=18.56, p<0.0001). Specifically, we found that both male and female APP^NLGF and 5xFAD mice, but not APP/PS1 mice, spent less time (APP^NLGF: males −44%, p=0.0035, and females −46%, p=0.0036; 5xFAD: males −40%, p=0.013, and females −41%, p=0.016) and entered fewer times (APP^NLGF: males −36%, p=0.044, and females −43%, p=0.007; 5xFAD: male −52%, p=0.0006, and female −55%, p=0.0001) in the light compartment compared to the sex-matched WT mice. Furthermore, both female APP^NLGF (−46%, p=0.004) and 5xFAD (−40%, p=0.02) mice spent less time in the light compartment compared to female APP/PS1 mice. Finally, both male (−48%, p=0.004) and female (−45%, p=0.039) 5xFAD mice entered fewer times in the light compartment compared to the sex-matched APP/PS1 group. Overall, these results suggests an increase in the anxiety-like behavior in the APP^NLGF and 5xFAD strains (Fig. 2B).

A significant genotype effect (F_3,72=11.55, p<0.0001) was found in locomotor activity. Post hoc analysis highlighted that both male and female APP^NLGF and 5xFAD were less active compared to the sex-matched APP/PS1 mice (APP^NLGF: males −47%, p=0.007, and females −44%, p=0.008; 5xFAD: males −40%, p=0.04, and females −38%, p=0.03,) (Fig. 2C).

3.2.2. TST

Then, we tested the depressive-like behavior by TST. We found a genotype effect (F_3,72=49.03, p<0.0001) for the time of immobility. Post hoc analysis indicated a significant increase of the time of immobility in both APP^NLGF (males: +67%, p=0.0002; females: +64%, p<0.0001) and 5xFAD (males: +109%, p<0.0001; females: +80%, p<0.0001) mice compared to the sex-matched WT mice. Importantly, APP^NLGF (males: +42%, p=0.015; females: +63%, p<0.0001) and 5xFAD (males: +77%, p<0.0001; females: +79%, p<0.0001) mice showed significantly higher levels of depressive-like behavior compared to APP/PS1 mice. Of note, APP/PS1 mice did not show a significant increase in the time of immobility compared to WT-mice (Fig. 2D).

3.2.3. SI

Finally, we found a genotype effect for the time spent (F_3,72=20.42, p<0.0001) and the number of interactions (F_3,72=13.18, p<0.0001) with the sex-matched conspecific in the SI paradigm. Post hoc tests revealed a significant reduction of the time spent around the conspecific for APP^NLGF (males: −48%, p=0.0004; females: −50%, p<0.0001) and 5xFAD (males: −36%, p=0.02; females: −38%, p=0.0012) mice compared to the sex-matched WT mice. Likewise, APP^NLGF (males: −36%, p=0.047; females: −36%, p=0.023) and 5xFAD (males: −46%, p=0.009; females: −40%, p=0.006) mice showed a reduction in the number of interactions with the conspecific compared to the sex-matched WT mice. Again, APP/PS1 did not exhibit any difference compared to control mice. Additionally, we found that APP^NLGF mice spent less time (males: −40%, p=0.03; females: −37%, p=0.03) to the sex-matched APP/PS1 mice. Finally only female 5xFAD interacted significantly less (−36%, p=0.03) with the conspecific compared to the sex-matched APP/PS1 mice (Fig. 2E).

3.3. Comparison of corticolimbic Aβ deposition in APP^NLGF to 5xFAD and APP/PS1 mice

In accordance with our hypotheses, we found significant differences in the number, size, and distribution of Aβ plaques across the corticolimbic system between the three AD models (Fig. 3–7).

Figure 3. Aβ plaque number, size, and relative area (%) in the prefrontal cortex of APP^NLGF, 5xFAD, and APP/PS1 mice.

As shown here, these three different AD mouse models differred in the distribution of prefrontocortical Aβ plaques after staining with both thioflavin-S and Aβ antibody (A). Average plaque number (left graph), plaque size (central graph), and percentage area occupied by plaques (right graph) were evaluated in the prefrontal cortex of these three AD mouse strains (B). The final value for each mouse was obtained by calculating the average of five thioflavin-S stained sections/mouse (6 mice/group, n=30 sections/group). Aβ antibody staining was performed for qualitative validation. The images were captured at 10x magnification. Data represent the mean ± SEM of 6 mice. ***p<0.001 when compared to APP^NLGF mice; ^###p<0.001 when compared to 5xFAD mice. One-way ANOVA followed by Tukey’s post hoc analysis.

Figure 7. Aβ plaque number, size, and relative area (%) in the hippocampal DG of APP^NLGF, 5xFAD, and APP/PS1 mice.

As shown here, these three different AD mouse models differed in the distribution of DG Aβ plaques after staining with both thioflavin-S and Aβ antibody (A). Average plaque number (left graph), plaque size (central graph), and percentage area occupied by plaques (right graph) were evaluated in the DG of these three AD mouse strains (B). The final value for each mouse was obtained by calculating the average of five thioflavin-S stained sections/mouse (6 mice/group, n=30 sections/group). Aβ antibody staining was performed for qualitative validation. The images were captured at 10x magnification. Data represent the mean ± SEM of 6 mice. ***p<0.001 when compared to APP^NLGF mice; ^###p<0.001 when compared to 5xFAD mice. One-way ANOVA followed by Tukey’s post hoc analysis.

3.3.1. Prefrontal cortex

Both 5xFAD (85.88±7.28) and APP/PS1 (50.46±3.45) mice showed a higher amount of Aβ plaques in the prefrontal cortex compared to APP^NLGF mice (13.91±2.83) (F_2,15=53.21, p<0.0001, R²=0.88) (5xFAD vs APP^NLGF: +516%, p<0.0001; APP/PS1 vs APP^NLGF: +262%, p=0.0003; 5xFAD vs APP/PS1: +70%, p=0.0004). Moreover, 5xFAD (7.58±0.82 μm) and APP/PS1 (14.22±0.86 μm) showed significantly larger average plaque size compared to APP^NLGF mice (1.96±0.77 μm) (F_2,15=56.85, p<0.0001, R²=0.88) (5xFAD vs APP^NLGF: +288%, p=0.0005; APP/PS1 vs APP^NLGF: +627%, p<0.0001; APP/PS1 vs 5xFAD: +88%, p=0.0001). Finally, both 5xFAD (3.48±0.68 %) and APP/PS1 (3.39±0.45%) showed an increase of the percentage area occupied by plaques compared to APP^NLGF mice (0.34±0.08%) (F_2,15=14.52, p=0.0003, R²=0.66) (5xFAD vs APP^NLGF: +935%, p=0.0008; APP/PS1 vs APP^NLGF: +910%, p=0.0009) (Fig. 3).

3.3.2. Amygdala

Amygdalae of 5xFAD mice (78.58±5.93) were characterized by a higher number of Aβ plaques in the compared to both APP^NLGF (18.14±3.82) and APP/PS1 mice (28.26±3.28) (F_2,15=51.97, p<0.0001, R²=0.87) (5xFAD vs APP^NLGF: +333%, p<0.0001; APP/PS1 vs APP^NLGF: +178%, p<0.0001). Furthermore, amygdalae of both 5xFAD (7.95±0.87 μm) and APP/PS1 (13.38±0.94 μm) were characterized by a larger average plaque size compared to APP^NLGF mice (1.62±0.42 μm) (F_2,15=56.78, p<0.0001, R²=0.88) (5xFAD vs APP^NLGF: +390%, p=0.0001; APP/PS1 vs APP^NLGF: +725%, pμ0.0001; APP/PS1 vs 5xFAD: +68%, p=0.0005). We also found that the percentage area occupied by plaques was increased in both 5xFAD (3.59±0.58%) and APP/PS1 (1.86±0.26%) mice compared to APP^NLGF mice (0.29±0.03%) (F_2,15=20.43, p<0.0001, R²=0.73) (5xFAD vs APP^NLGF: +1128%, p<0.0001; APP/PS1 vs APP^NLGF: +534%, p=0.024; 5xFAD vs APP/PS1: +48%, p=0.011) (Fig. 4).

Figure 4. Aβ plaque number, size, and relative area (%) in the amygdala of APP^NLGF, 5xFAD, and APP/PS1 mice.

As reported in this figure, we found marked differences in the distribution of amygdalar Aβ plaques after staining with both thioflavin-S and Aβ antibody (A). Average plaque number (left graph), plaque size (central graph), and percentage area occupied by plaques (right graph) were evaluated in the amygdalae of these three AD mouse strains (B). The final value for each mouse was obtained by calculating the average of five thioflavin-S stained sections/mouse (6 mice/group, n=30 sections/group). Aβ antibody staining was performed for qualitative validation. The images were captured at 10x magnification. Data represent the mean ± SEM of 6 mice. *p<0.05, and ***p<0.001 when compared to APP^NLGF mice; ^#p<0.05, and ^###p<0.001 when compared to 5xFAD mice. One-way ANOVA followed by Tukey’s post hoc analysis.

3.3.3. Hippocampal CA1

Similar to our results in the amygdala, we found a higher number of Aβ plaques in 5xFAD mice (80.21±5.13) compared to both APP^NLGF (15.86±2.59) and APP/PS1 mice (20.19±1.17) (F_2,15=112.8, p<0.0001, R²=0.93) (5xFAD vs APP^NLGF: +405%, p<0.0001; APP/PS1 vs APP^NLGF: +297%, p<0.0001). 5xFAD (6.36±0.61 μm) and APP/PS1 (13.33±1.05 μm) mice showed a significantly larger average plaque size compared to APP^NLGF mice (2.82±0.77 μm) (F_2,15=41.39, p<0.0001, R²=0.85) (5xFAD vs APP^NLGF: +126%, p=0.02; APP/PS1 vs APP^NLGF: +373%, p<0.0001; APP/PS1 vs 5xFAD: +109%, p<0.0001). In addition, we observed that the percentage area occupied by plaques in both 5xFAD (2.69±0.33%) and APP/PS1 (1.48±0.16%) mice was markedly increased compared to APP^NLGF mice (0.36±0.05%) (F_2,15=30.45, p<0.0001, R²=0.81) (5xFAD vs APP^NLGF: +640%, p<0.0001; APP/PS1 vs APP_NLGF: +307%, p=0.005; 5xFAD vs APP/PS1: +45%, p=0.003) (Fig. 5).

Figure 5. Aβ plaque number, size, and relative area (%) in the hippocampal CA1 subregion of App^NLGF, 5xFAD, and APP/PS1 mice.

As shown here, these three different AD mouse models differed in the distribution of CA1 Aβ plaques after staining with both thioflavin-S and Aβ antibody (A). Average plaque number (left graph), plaque size (central graph), and percentage area occupied by plaques (right graph) were evaluated in the CA1 of these three AD mouse strains (B). The final value for each mouse was obtained by calculating the average of five thioflavin-S stained sections/mouse (6 mice/group, n=30 sections/group). Aβ antibody staining was performed for qualitative validation. The images were captured at 10x magnification. Data represent the mean ± SEM of 6 mice. *p<0.05, **p<0.01, and ***p<0.001 when compared to APP^NLGF mice; ^##p<0.05, and ^###p<0.001 when compared to 5xFAD mice. One-way ANOVA followed by Tukey’s post hoc analysis.

3.3.4. Hippocampal CA3

Both 5xFAD (55.03±3.21) and APP/PS1 (22.58±1.10) mice demonstrated a higher number of Aβ plaques in the CA3 subregion compared to APP^NLGF mice (11.87±1.82) (F_2,15=102.3, p<0.0001, R²=0.93) (5xFAD vs APP^NLGF: +364%, p<0.0001; APP/PS1 vs APP^NLGF: +90%, p=0.0102; 5xFAD vs APP/PS1: +144%, p<0.0001). In line with findings in other brain regions, 5xFAD (7.18±0.77 μm) and APP/PS1 (16.43±1.18 μm) mice showed a significantly larger average plaque size compared to APP^NLGF mice (1.42±0.56 μm) (F_2,15=75.06, p<0.0001, R²=0.91) (5xFAD vs APP^NLGF: +407%, p=0.0008; APP/PS1 vs APP^NLGF: +1060%, p<0.0001; APP/PS1 vs 5xFAD: +129%, p<0.0001). Finally, both 5xFAD (2.19±0.32%) and APP/PS1 (2.01±0.10%) mice showed an increase of the percentage area occupied by plaques compared to APP^NLGF mice (0.23±0.03%) (F_2,15=32.26, p<0.0001, R²=0.81) (5xFAD vs APP^NLGF: +872%, p<0.0001; APP/PS1 vs APP^NLGF: +791%, p<0.0001) (Fig. 6).

Figure 6. Aβ plaque number, size, and relative area (%) in the hippocampal CA3 subregion of App^NLGF 5xFAD, and APP/PS1 mice.

As shown here, these three different AD mouse models differed in the distribution of CA3 Aβ plaques after staining with both thioflavin-S and Aβ antibody (A). Average plaque number (left graph), plaque size (central graph), and percentage area occupied by plaques (right graph) were evaluated in the CA3 of these three AD mouse strains (B). The final value for each mouse was obtained by calculating the average of five thioflavin-S stained sections/mouse (6 mice/group, n=30 sections/group). Aβ antibody staining was performed for qualitative validation. The images were captured at 10x magnification. Data represent the mean ± SEM of 6 mice. *p<0.05, and ***p<0.001 when compared to APP^NLGF mice; ^###p<0.001 when compared to 5xFAD mice. One-way ANOVA followed by Tukey’s post hoc analysis.

3.3.5. Hippocampal DG

We found a higher number of Aβ plaques in the DG of both 5xFAD (54.88±2.04) and APP/PS1 (28.93±2.60) mice compared to APP^NLGF mice (8.43±1.50) (F_2,15=123.6, p<0.0001, R²=0.94) (5xFAD vs APP^NLGF: +551%, p<0.0001; APP/PS1 vs APP^NLGF: +243%, p<0.0001; 5xFAD vs APP^NLGF: +90%, p<0.0001). Similarly to the results of other brain regions, both 5xFAD (6.59±0.54 μm) and APP/PS1 (14.54±0.87 μm) mice showed a significantly larger average plaque size compared to APP^NLGF mice (1.37±0.42 μm) (F_2,15=108.1, p<0.0001, R²=0.94) (5xFAD vs APP^NLGF: +380%, p=0.0001; APP/PS1 vs APP^NLGF: +960%, p<0.0001; APP/PS1 vs 5xFAD: +121%, p<0.0001). Finally, both 5xFAD (2.03±0.20%) and APP/PS1 (2.41±0.34%) mice showed a marked increase of the percentage area occupied by plaques compared to APP^NLGF mice (0.18±0.02%) (F_2,15=27.77, p<0.0001, R²=0.79) (5xFAD vs APP^NLGF: +1032%, p<0.0001; APP/PS1 vs APP^NLGF: +1240%, p<0.0001, respectively) (Fig. 7).

4. Discussion

In the past two decades, multiple transgenic mouse models overexpressing human APP with familial AD mutations have been generated with the goal of translationally modeling the cognitive deficits and neuropathology observed in humans (Ameen-Ali et al., 2017; Puzzo et al., 2015). Despite valuable results obtained by adopting these AD models, non-physiological overexpression of APP may introduce artifacts by disrupting other genes near the site of transgene insertion or by overwhelming protein homeostasis (Balducci and Forloni, 2011; Sasaguri et al., 2017). In recent years, several APP knock-in mouse models, including APP^NLGF, have been created, permitting recapitulation of AD phenotypes including amyloid pathology, microglial and astrocytic activation, and loss of synapses without overexpressing APP (Saito et al., 2014). To our knowledge, this is the first study to directly compare memory, affective behavior, and amyloid plaques between APP^NLGF, 5xFAD, and APP/PS1 mice. We found that all three AD mouse models at 10 months of age show a similar degree of memory deficits tested by commonly used memory tasks, such as NOR, SA, SR, and the working memory version of MWM (Fig. 1). Although learning and memory dysfunction and changes in affective behavior have been widely studied in both 5xFAD and APP/PS1 mice (summarized in Table 1), much less is known about these changes in APP^NLGF mice (Masuda et al., 2016; Saito et al., 2014; Sakakibara et al., 2018). A few studies have indicated that APP^NLGF mice display reduction of spontaneous alternation in Y-maze starting at 6 months (Saito et al., 2014), impairment of spatial and place reversal learning at 8 months (Masuda et al., 2016; Sakakibara et al., 2018), and poorer performance in the standard Morris water maze task at 24 months (Sakakibara et al., 2019). In contrast, one study showed no memory impairment tested by novel object recognition in APP^NLGF mice at 6 months of age (Whyte et al., 2018). In our study, we confirmed the presence of deficits in spatial memory on the spontaneous alterantion paradigm, working memory on the MWM, and object and social recognition at 10 months of age. Our findings provide additional evidence that the APP^NLGF mouse strain is valid animal model for translational studies of cognitive deficits AD.

Table 1.

Schematic summary of reports on memory and affective behavior in APP^NLGF, 5xFAD, and APP/PS1 at 8-12 months of age.

Strain	Sex	Behavior	Result	Publication
		Memory
APPNLGF	♂	Barnes maze	↓	Sakakibara et al., 2018
APPNLGF	♂+♀	Place preference reversal	↓	Masuda et al., 2016
5xFAD	♂+♀	Novel object recognition	↓	Creighton et al., 2019
5xFAD	♂+♀	Spontaneous alternation	↓	Devi and Ohno, 2015
5xFAD	♂+♀	Morris water maze	↓	Maiti et al., 2020
5xFAD	♀	Social recognition	↓	Kosel et al., 2019
5xFAD	♂	Fear conditioning	↓	Hwang et al., 2017
APP/PS1	♂+♀	Novel object recognition	↓	Webster et al., 2013
APP/PS1	♂+♀	Spontaneous alternation	↓	Wang et al., 2017
APP/PS1	♂	Morris water maze	↓	Li et al., 2019
APP/PS1	♀	Morris water maze	↓	Gallagher et al., 2013
APP/PS1	♂	Social recognition	↓	Cheng et al., 2014
		Affective Behavior
APPNLGF	♂	Elevated plus maze	=	Sakakibara et al., 2018
APPNLGF	♂+♀	Open field	↑	Pervolaraki et al., 2019
APPNLGF	♂+♀	Elevated plus maze	↓	Pervolaraki et al., 2019
5xFAD	♀	Elevated plus maze	↓	Jawhar et al., 2012
5xFAD	♂	Open field	↓	Braun and Feinstein. 2019
5xFAD	♀	Open field, light dark box	=	Flanigan et al., 2014
5xFAD	♀	Social interaction	↓	Kosel et al., 2019
APP/PS1	♂+♀	Open field	=	Radde et al., 2006
APP/PS1	♂+♀	Open field, elevated plus maze	=	Webster et al., 2013
APP/PS1	♂+♀	Elevated plus maze	↓	Verma et al., 2015
APP/PS1	♂	Social interaction	=	Olesen et al., 2016

We also evaluated affective behaviors in APP^NLGF mice and directly compared them to 5xFAD and APP/PS1 mice. Our results showed comparable levels of anxiety and depressive-like behavior as well as social interaction impairment between APP^NLGF and 5xFAD mice, but such behavioral changes were not observed in APP/PS1 mice at the age investigated (Fig. 2). It is well known that affective symptoms are strongly linked to AD; indeed, 90% of AD cases are associated with non-cognitive neuropsychiatric symptoms including anxiety and depression (Keszycki et al., 2019; Shin et al., 2005). Meta-analyses have linked depression to AD (Diniz et al., 2013; Lenoir et al., 2011), suggesting that depression is a significant risk factor for future development of AD (Butters et al., 2008; Diniz et al., 2013). Therefore, it is critical to identify and characterize animal models that permit the study of both memory deficits and changes in affective behavior in AD. While some reports have shown anxiety- and depressive-like behaviors in 5xFAD or APP/PS1 mice alone, very few studies have aimed to characterize affective behaviors in APP^NLGF mice and to directly compare them to other AD models (Table 1). Recently, a few studies have identified alterations of anxiety-like behavior in APP^NLGF mice (Latif-Hernandez et al., 2019; Pervolaraki et al., 2019; Sakakibara et al., 2018). In these studies, Sakakibara and colleagues reported a reduction of anxiety-like behavior in 15-18 month old APP^NLGF mice relative to WT mice (Sakakibara et al., 2018). Moreover, Latif-Hernandez and colleagues observed a slight decrease of anxiety-like behavior in APP^NLGF mice at 3 and 6 months -but not at 10 months- suggesting dynamic changes of affective behavior with AD progression. It is also possible that the different behavioral paradigms used by our study (open field and light-dark box) versus Sakakibara’s group (elevated plus maze) may have contributed to the different results. This notion is supported by Pervolaraki and colleagues’ results who demonstrated an increase of anxiety-like behavior in the open field test and, oppositely, a decrease of anxiety-like behavior in the elevated plus maze test in APP^NLGF mice at 8 month of age (Pervolaraki et al., 2019). In our study we observed an increase of anxiety-like behavior in 5xFAD mice, which is in contrast to other studies that suggest a reduction of anxiety-like behavior in this mouse model using both the elevated plus maze and open field paradigms (Jawhar et al., 2012). However, it is worth noting that in Jawhar and colleagues’ study, 5xFAD mice spent only around 10% of the time in the closed arms, which may suggest an abnormal behavior not necessarily linked to anxiety. Indeed, mice usually spend 70-80% of the time in the closed arms of the elevated plus maze (Griebel et al., 2000; Wahlsten et al., 2003; Brigman et al., 2009). As anxiety decreases, this ratio reduces to 50%, indicating that 5xFAD mice were actively avoiding the closed arms in the aforementioned study. Finally, Flanigan and colleagues tested anxiety-like behavior in female 5xFAD mice and observed no differences by performing both elevated plus maze and light-dark box (Flanigan et al., 2014). All together, the mixed results consistently reported in the literature (Kosel et al., 2020) suggest that the characterization of anxiety-like behavior in AD mouse models may be difficult, potentially due to multiple variables, including age, sex, specific task, and testing environment. Despite this, our data provide additional evidence to support that the APP^NLGF strain is a suitable model for AD research as it relates to memory and affective behavior.

In our study, we did not find sex differences between males and females across strains for the majority of tests. However, two-way ANOVA suggested a sex effect for the time spent in the center of the open field apparatus (anxiety-like behavior) as well as the time spent interacting with the conspecific in the social interaction test (social behavior) at this age. Sex differences in memory and affective behavior have previously been reported at younger ages in wild-type mice, particularly for spatial memory, anxiety-like behavior and social behavior (Donner and Lowry, 2013). These discrepancies may be due to the age differences of the experimental mice. Accordingly, some reports in the literature have suggested the absence of sex differences in middle-aged wild-type mice tested for working memory (Y-maze), object recognition (novel object recognition), and anxiety-like behavior (open field) (Szentes et al., 2019). Moreover, some evidence have suggested that wild-type mice at 12 months of age do not show sex differences in social behavior (Bories et al., 2012).

AD mouse models overexpressing APP generally exhibit altered locomotor activity and stereotypes that have been linked to changes in APP metabolism (Ambree et al., 2009; Cisse et al., 2011; Mori et al., 2013). Our results support this hypothesis as we observed a decrease of locomotion in APP^NLGF mice compared to APP/PS1 mice, further validating the APP^NLGF model. Also notable is that, despite the same genetic approach (APP overexpression), we found differences in the affective behavior between 5xFAD and APP/PS1 mice. Besides the different genetic mutations, the mechanisms underlying this discrepancy need to be further investigated.

Finally, we compared the differences in the number and size of Aβ plaques, and percentage of the area occupied by plaques between these three models at 11 months of age in the prefrontal cortex, amygdala, and hippocampus, which are key regions linked to memory and affective behavior. In line with our hypotheses, we found the highest number of Aβ plaques in the 5xFAD mouse brain, followed by APP/PS1, and lastly in APP^NLGF mice in all brain regions studied (Fig. 3–7); a similar pattern of findings was observed for the percentage area occupied by plaques. On the other hand, we found the biggest plaque size in APP/PS1 mice, and the smallest in APP^NLGF mice. Our results demonstrate that Aβ plaque number, size, and area across brain regions commonly implicated in AD neuropathogenesis differentially contribute to memory deficits and affective behaviors observed across these three models. In APP^NLGF mice, despite a relatively mild Aβ pathology, alterations in cognition and affective behaviors are marked, suggesting that other molecular and cellular mechanisms may contribute to the behavioral phenotypes in this AD mouse line. In fact, both clinical and preclinical studies have shown inconsistent relationships between amyloid plaques and cognitive decline in AD (Robinson et al., 2011; Sabbagh et al., 2010). In contrast, a large number of studies have reported a strong link between neocortical neurofibrillary tangles and cognitive deficits (Giannakopoulos et al., 2009; Robinson et al., 2011; Sabbagh et al., 2010). It is possible that some other unknown factors may also independently affect the affective behavior. Therefore, future studies aimed at characterizing and comparing amyloid oligomer levels, tau and phospho-tau levels, neuroinflammatory markers, and gliosis may help to dissect the mechanisms underlying such behavioral outcomes. Notably, we did not compare and correlate the neuropathological changes between sexes at this points, since we did not find sex differences in the majority of the behavioral outcomes.

Our results provide a novel, detailed comparison of affective and cognitive behavior across both APP overexpressing and APP knock-in mouse models of AD; however, some limitations must be noted. All three models of AD in this study are based solely on Aβ accumulation, which means that we are unable to evaluate the contribution of neurofibrillary tangle pathology in memory deficits and affective behavior. Furthermore, testing additional timepoints (e.g., 3, and 6 months) would have permitted correlations between levels of Aβ and behavioral phenotypes during the onset and progression of disease in these three AD models. This is particularly true for 5xFAD mice, which showed very high levels of Aβ plaques compared to both APP/PS1 and APP^NLGF mice at 10 months. It is possible that the absence of differences in memory impairment between these three AD models at 10 months of age is due to the late stage of the disease. Therefore, we cannot exclude whether differences in memory function may be detected at younger ages and may correspond to different severities of AD pathology. Future studies are necessary to evaluate the dynamic changes of Aβ load with aging and AD development, particularly comparing APP knock-in mice to other AD mouse models. Addtionally, in our study, we only compared fibrillar Aβ plaques by thioflavin-S staining. A comparison of APP, amyloid oligomer, and diffuse plaques would provide more complete characterizations of pathology in these three strains, with additional potential implications for the impact of the APP-amyloid metabolism on memory and behavior in AD models.

Conclusions

Although transgenic mouse models of AD are valuable tools to investigate different aspects of AD pathology, disease progression, and treatment strategies, each mouse model demonstrates both advantages and disadvantages. In this study, we have carefully evaluated and compared memory, affective behavior, and Aβ pathology in three commonly used mouse models of AD (i.e., APP^NLGF, 5xFAD, APP/PS1 mouse strains) at 10 months of age. Our findings demonstrate that the APP^NLGF mouse strain displays robust memory deficits, anxiety- and depressive-like behavior, and impairment of social behavior despite having the lowest Aβ plaque load. Likewise, our results suggest that APP^NLGF mice are a valid preclinical model for the study of AD neuropathology, memory deficits, and alterations of affective behavior.

Highlights.

• APP^NLGF mice, an Alzheimer’s disease model with a human APP knock-in instead of APP-overexpression, display marked memory deficits that are comparable to those observed in 5xFAD and APP/PS1 mice at 10 months of age.

• APP^NLGF and 5xFAD, but not APP/PS1 mice, exhibit marked anxiety- and depressive-like behaviors as well as a reduction of social interaction.

• These three AD mouse models show distinguished corticolimbic Aβ plaques number, size, and percentage of the area occupied by plaques.