In AppNL-G-F/NL-G-F mice, age-dependent cortical Aβ amyloidosis began by 2 months and saturated around 7 months of age (Additional file 1: Fig. S1) [12]. These mice also developed Aβ amyloidosis in the hippocampal and subcortical regions [12]. Despite aggressive Aβ amyloidosis in AppNL-G-F/NL-G-F mice, neuroinflammatory responses such as astrocytosis and microgliosis were not intense at the age of 6–9 months, whereas greater reactive gliosis was observed in cortical and hippocampal regions, as well as in subcortical regions, at the age of 15–18 months (Additional file 1: Fig. S1) [12, 18]. By contrast, Aβ plaques and neuroinflammatory responses were negligible even at 18 months of age in AppNL/NL mice, despite elevation of the Aβ level in the brain [12, 18]. Based on this neuropathological information, we carried out behavioral assays to capture cognitive (BM and CFC tasks) and emotional (EPM task) alterations in App-KI mice over the course of aging (Additional file 1: Fig. S1). In the experimental design, we noted that the same group of mice (Group 4) was repeatedly tested at 4, 6, and 8 months of age in the BM task (Additional file 1: Fig. S1).
App
NL-G-F/NL-G-F mice exhibit anxiolytic-like behavior, whereas App
NL/NL mice exhibit anxiogenic-like behavior, in comparison with control WT mice
Anxiety-related behaviors were assessed using the EPM task, in which increased exploration of open arms indicates anxiolytic-like behavior [27, 28]. In 6–9-month-old AppNL-G-F/NL-G-F mice, the amount of time on (Fig. 1a; F[2, 21] = 4.35, p = 0.026, post hoc, WT vs. AppNL-G-F/NL-G-F, p = 0.565) and entries into (Fig. 1b; F[2, 21] = 2.22, p = 0.133) open arms during the 10-min test were slightly increased in comparison with WT mice, although these differences were not statistically significant with our sample size. The average total number of arm entries (Fig. 1c; F[2, 21] = 1.95, p = 0.167) and the distance travelled during the 10-min test (Additional file 2: Fig. S2a and b; F[2, 21] = 0.27, p = 0.766) were also slightly increased in AppNL-G-F/NL-G-F mice, although these differences were not statistically significant. By contrast, AppNL/NL mice exhibited similar levels of the amount of time on (Fig. 1a; post hoc, WT vs. AppNL/NL, p = 0.170) and entries into (Fig. 1b) open arms to those of WT mice, with no alterations in general exploratory activity (Fig. 1c, Additional file 2: Fig. S2b).
However, the patterns of exploration in AppNL-G-F/NL-G-F mice differed from those observed in WT mice. When we analyzed the time course of open arm exploration by scoring the percentage of time spent on the arms in each 2-min interval (Fig. 1d), WT mice exhibited significant reductions in the time spent on the open arms as the test progressed (F[4, 28] = 3.75, p = 0.014), consistent with a previous report [28,29,30]. By contrast, AppNL-G-F/NL-G-F mice persistently explored the open arms throughout the test (Fig. 1d; F[4, 28] = 0.68, p = 0.610). At a later time point, AppNL-G-F/NL-G-F mice spent significantly more time on the open arms than WT mice (Fig. 1d; Time 8–10, F[2, 21] = 6.11, p = 0.008, post hoc, WT vs. AppNL-G-F/NL-G-F, p = 0.022). In contrast, similar to WT mice, AppNL/NL mice exhibited a decrease in open arm exploration as the test progressed (Fig. 1d; F[4, 28] = 2.69, p = 0.051). These results suggest that AppNL-G-F/NL-G-F mice have altered responses to aversive situations, such as open spaces.
Previous studies demonstrated that laboratory rodents exhibited a significant reduction of open arm exploration when re-exposed to the EPM [28, 30, 31]. This suggests that prior test experience caused a qualitative shift in emotional state, and the acquisition of a phobic state rather than an unconditioned anxiety response. To investigate whether prior test experience could alter anxiety-related behavior, we re-tested the App-KI and WT mice in the same EPM paradigm.
As reported previously, WT mice exhibited robust avoidance responses to the open arms in the second trial of our EPM task, reflected by reduced percentages of time spent on and entries into the arms. By contrast, 6–9-month-old AppNL-G-F/NL-G-F mice spent significantly more time on the open arms (Fig. 1e; F[2, 21] = 5.30, p = 0.014, post hoc, WT vs. AppNL-G-F/NL-G-F, p = 0.034) and entered them more frequently (Fig. 1f; F[2, 21] = 6.11, p = 0.008, post hoc, WT vs. AppNL-G-F/NL-G-F, p = 0.006) during the 10-min test period than WT mice. The time course analysis also revealed a persistent and durable exploration of open arms in AppNL-G-F/NL-G-F mice (Fig. 1h; F[4, 28] = 0.207, p = 0.932). These mice showed slightly higher preference toward the open arms in comparison with WT mice at each time point, although the differences were not statistically significant (Fig. 1h; Time 0–2, F[2, 21] = 1.33, p = 0.286; Time 2–4, F[2, 21] = 2.44, p = 0.111; Time 4–6, F[2, 21] = 3.05, p = 0.069; Time 6–8, F[2, 21] = 1.35, p = 0.280; Time 8–10, F[2, 21] = 1.59, p = 0.228). AppNL/NL and WT mice engaged in similar levels of open arm exploration (Fig. 1e, f and h). As with the case of the first trial, AppNL-G-F/NL-G-F mice exhibited slight increases in the average total number of arm entries (Fig. 1g; F[2, 21] = 1.53, p = 0.240) and the distance travelled during the 10-min test (Additional file 2: Fig. S2c and d; F[2, 21] = 1.22, p = 0.316), although these differences were not statistically significant with our sample size.
Taken together, these results suggest that 6–9-month-old AppNL-G-F/NL-G-F mice exhibit robust anxiolytic-like behavior, even after they have habituated to a test environment.
To investigate whether the observed anxiolytic-like behavior in AppNL-G-F/NL-G-F mice was maintained during aging, we performed the same EPM task at 15–18 months of age. In the first trial, AppNL-G-F/NL-G-F mice showed a tendency to spent more time on (Fig. 1i; F[2, 30] = 6.78, p = 0.004, post hoc, WT vs. AppNL-G-F/NL-G-F, p = 0.401) and a slightly higher frequency of entries into (Fig. 1j; F[2, 30] = 7.01, p = 0.003, post hoc, WT vs. AppNL-G-F/NL-G-F, p = 0.700) the open arms during the 10-min test than WT mice, although these differences were not statistically significant. By contrast, AppNL/NL mice tended to spend less time in the open arms (Fig. 1i; post hoc, WT vs. AppNL/NL, p = 0.054) and entered them significantly less frequently (Fig. 1j; post hoc, WT vs. AppNL/NL, p = 0.021) than WT mice, suggesting anxiogenic-like behavior in AppNL/NL mice. General exploratory activity was slightly increased in AppNL-G-F/NL-G-F mice, although the difference was not statistically significant with our sample size (Fig. 1k; F[2, 30] = 1.07, p = 0.356; Additional file 2: Fig. S2e and f; F[2, 30] = 0.18, p = 0.836).
As observed at 6–9 months of age, 15–18-month-old WT and AppNL/NL mice exhibited clear avoidance of the open arms as the test progressed (Fig. 1l; WT, F[4, 44] = 8.96, p < 0.001; AppNL/NL, F[4, 36] = 4.15, p = 0.007), whereas AppNL-G-F/NL-G-F mice did not exhibit a significant change in open arm exploration during the test (F[4, 40] = 1.83, p = 0.141). At an early time point, AppNL/NL mice spent significantly less time on the open arms than WT mice (Fig. 1l; Time 0–2, F[2, 30] = 4.13, p = 0.026, post hoc, WT vs. AppNL/NL, p = 0.021). By contrast, AppNL-G-F/NL-G-F mice exhibited slightly higher open arm exploration in the latter half of the test in comparison with WT mice, but the differences were not statistically significant (Fig. 1l; Time 4–6, F[2, 30] = 5.30, p = 0.011, post hoc, WT vs. AppNL-G-F/NL-G-F, p = 0.466; Time 6–8, F[2, 30] = 5.74, p = 0.008, post hoc, WT vs. AppNL-G-F/NL-G-F, p = 0.079; Time 8–10, F[2, 30] = 7.94, p = 0.002, post hoc, WT vs. AppNL-G-F/NL-G-F, p = 0.125). These results suggest that 15–18-month-old AppNL-G-F/NL-G-F mice exhibit alterations in the habituation process to aversive stimuli.
In the second trial, 15–18-month-old AppNL-G-F/NL-G-F mice spent significantly more time on the open arms (Fig. 1m; F[2, 30] = 13.87, p < 0.001, post hoc, WT vs. AppNL-G-F/NL-G-F, p < 0.001) and entered them more often (Fig. 1n; F[2, 30] = 9.37, p < 0.001, post hoc, WT vs. AppNL-G-F/NL-G-F, p = 0.009) than WT mice. The time course analysis also revealed a persistent and durable exploration of open arms in AppNL-G-F/NL-G-F mice (Fig. 1p; F[4, 40] = 0.74, p = 0.570). AppNL-G-F/NL-G-F mice spent more time on the open arms from the beginning of the test (Fig. 1p; Time 0–2, F[2, 30] = 5.13, p = 0.012, post hoc, WT vs. AppNL-G-F/NL-G-F, p = 0.053; Time 2–4, F[2, 30] = 3.31, p = 0.050; Time 4–6, F[2, 30] = 3.51, p = 0.043, post hoc, WT vs. AppNL-G-F/NL-G-F, p = 0.157) and particularly at later time points (Time 6–8, F[2, 30] = 7.26, p = 0.003, post hoc, WT vs. AppNL-G-F/NL-G-F, p = 0.005; Time 8–10, F[2, 30] = 15.14, p < 0.001, post hoc, WT vs. AppNL-G-F/NL-G-F, p = 0.003). In addition, AppNL-G-F/NL-G-F mice exhibited a significant increase in the total number of arm entries in comparison with WT mice (Fig. 1o; F[2, 30] = 7.85, p = 0.002, post hoc, WT vs. AppNL-G-F/NL-G-F, p = 0.021). We also measured the distance travelled during the test (Additional file 2: Fig. S2g and h) and noticed that AppNL-G-F/NL-G-F mice moved longer than WT mice, though the difference was not statistically significant with our sample size (F[2, 30] = 3.76, p = 0.035, post hoc, WT vs. AppNL-G-F/NL-G-F, p = 0.154). In contrast to the first trial, AppNL/NL and WT mice exhibited similar levels of open arm exploration (Fig. 1m and n), presumably due to habituation of WT mice to the test environment.
Taken together, these results suggest that 15–18-month-old AppNL-G-F/NL-G-F mice exhibit robust anxiolytic-like behaviors, with increases in general exploratory activity, whereas AppNL/NL mice displayed unconditioned anxious phenotypes in comparison with WT mice.
App
NL-G-F/NL-G-F and App
NL/NL mice exhibit normal learning and memory of contextual fear up to 15–18 months of age in comparison with WT mice
The CFC task is a commonly used procedure for inducing learned fear, which is believed to be hippocampal-dependent [32, 33]. In this paradigm, a particular context as a conditioned stimulus evokes fear through association with an aversive event, such as a footshock [34]. Conditioned fear responses are impaired in both human patients and mouse models of AD [35,36,37,38].
At 6–9 months of age, the velocities of both AppNL-G-F/NL-G-F and AppNL/NL mice during administration of each footshock were comparable to those of WT mice (Fig. 2a; first, F[2, 18] = 0.32, p = 0.732; second, F[2, 18] = 0.71, p = 0.506; third, F[2, 18] = 1.30, p = 0.297). In addition, AppNL-G-F/NL-G-F, AppNL/NL, and WT mice exhibited the same levels of the freezing response upon subsequent presentation of footshocks during conditioning (Fig. 2b; genotype, F[2, 18] = 0.19, p = 0.830; time, F[1.6, 28.7] = 18.02, p < 0.001). To determine whether there were any locomotor deficits that could have confounded the outcome, we compared the distance travelled during the pre-shock period (the 3-min period prior to the first footshock) among genotypes (Additional file 3: Fig. S3a and b). At these ages, AppNL-G-F/NL-G-F mice seemed to be less active than WT mice during the pre-shock period, although the difference was not statistically significant with our sample size (Additional file 3: Fig. S3b; F[2, 18] = 2.99, p = 0.076). These results suggest that all genotypes were capable of detecting and responding to footshock stimuli at similar levels.
In the context test, min-by-min scoring of the percentage of freezing behavior revealed that all genotypes exhibited similar increases in the response as the test progressed (Fig. 2c; genotype, F[2, 18] = 10.25, p = 0.371; time, F[2.6, 46.4] = 21.08, p < 0.001). Moreover, levels of the freezing response during the total 5-min period were comparable among all genotypes (Fig. 2d; F[2, 18] = 1.05, p = 0.371). These results suggest that both AppNL-G-F/NL-G-F and AppNL/NL mice can learn and memorize the association between cues in the experimental chamber and footshock as effectively as WT mice.
At 15–18 months of age, AppNL-G-F/NL-G-F mice exhibited significantly higher shock reactivity than WT mice, as revealed by an increased velocity during the second and third footshocks (Fig. 2e; first, F[2, 19] = 0.85, p = 0.444; second, F[2, 19] = 7.36, p = 0.004, post hoc, WT vs. AppNL-G-F/NL-G-F, p = 0.007; third, F[2, 19] = 10.82, p < 0.001, post hoc, WT vs. AppNL-G-F/NL-G-F, p = 0.008). This result suggests that 15–18-month-old AppNL-G-F/NL-G-F mice have heightened sensitivity to painful stimuli. During conditioning, both AppNL-G-F/NL-G-F and AppNL/NL mice exhibited levels of freezing upon subsequent presentation of footshocks similar to those of WT mice (Fig. 2f; genotype, F[2, 19] = 0.0, p = 0.994; time, F[1.9, 36.5] = 84.15, p < 0.001). We also found that AppNL/NL mice moved significantly less than WT mice during the pre-shock period (Additional file 3: Fig. S3c and d; F[2, 19] = 5.13, p = 0.017, post hoc, WT vs. AppNL/NL, p = 0.016). However, a slight reduction in locomotor activity in AppNL/NL mice does not significantly affect the behavioral outcomes of the CFC task in AppNL/NL mice, since these mice can exhibit similar levels of shock reactivity and freezing behavior with WT mice (Fig. 2e and f). Locomotor activity during the pre-shock period was also slightly decreased in AppNL-G-F/NL-G-F mice, but the difference was not statistically significant with our sample size (Additional file 3: Fig. S3d; post hoc, WT vs. AppNL-G-F/NL-G-F, p = 0.092).
In the context test, the min-by-min data for freezing behavior revealed that the time course of the freezing response was similar among all genotypes (Fig. 2g; genotype, F[2, 19] = 0.23, p = 0.799; time, F[4, 76] = 5.06, p = 0.001). During the total 5-min period of the test, both AppNL-G-F/NL-G-F and AppNL/NL mice exhibited levels of freezing behavior similar to those of WT mice (Fig. 2h; F[2, 19] = 0.23, p = 0.800).
Taken together, these results suggest that both AppNL-G-F/NL-G-F and AppNL/NL mice have intact learning and memory of contextual fear, even at 15–18 months of age.
App
NL-G-F/NL-G-F mice exhibit alterations in spatial learning ability, with intact memory, in the BM task at 8 months of age
The BM task is a spatial memory task that requires animals to learn the location of an escape hole using spatial cues, and is therefore thought to be hippocampal-dependent [39, 40]. This task is commonly used for assessment of memory deficits in animal models of AD [41,42,43]. In our experiments, mice were asked to acquire the spatial location of a target hole that was connected to a dark escape box during the acquisition phase (Fig. 3a [left]). One day after the fifth session of the acquisition phase, a probe test was conducted without an escape box to investigate whether mice had learned the location of the target hole by extra-maze cues (Fig. 3a [middle]). To further assess cognitive flexibility, mice were subjected to the reversal learning task (five sessions) 1 day after the probe test (Fig. 3a [right]). And as mentioned in the experimental design above, the same group of mice was repeatedly tested at 4, 6, and 8 months of age in this BM task (Additional file 1: Fig. S1).
We found that AppNL/NL, AppNL-G-F/NL-G-F, and WT mice performed equally well in acquisition of the target hole in the BM at the ages of 4 months (Fig. 3b; F[2, 20] = 3.12, p = 0.066, Fig. 3c; F[2, 20] = 0.48, p = 0.625, Fig. 3d; F[2, 20] = 1.10, p = 0.353) and 6 months (Fig. 3g; F[2, 20] = 1.27, p = 0.303, Fig. 3h; F[2, 20] = 2.80, p = 0.085, Fig. 3i; F[2, 20] = 0.24, p = 0.788). The number of errors (Fig. 3b; F[4, 80] = 23.21, p < 0.001, Fig. 3g; F[2.7, 54.8] = 10.07, p < 0.001), latency (Fig. 3c; F[1.9, 37.6] = 29.26, p < 0.001, Fig. 3h; F[1.7, 33.1] = 8.06, p = 0.002), and distance (Fig. 3d; F[2.3, 45.3] = 23.18, p < 0.001, Fig. 3i; F[2.8, 56.5] = 7.11, p = 0.001) to reach the target hole significantly decreased as the session progressed, suggesting that all genotypes had similar learning ability.
In the probe test, all genotypes exhibited similar levels of preference toward the target quadrant that contained the target hole and the two adjacent holes at both 4 months (Fig. 3e; genotype, F[2, 20] = 1.06, p = 0.365; quadrant, F[2.0, 40.3] = 49.06, p < 0.001) and 6 months of age (Fig. 3j; genotype, F[2, 20] = 1.56, p = 0.235; quadrant, F[1.6, 32.0] = 30.84, p < 0.001). A percentage of time spent in the target quadrant for each genotype was significantly higher than chance level (25%) at both 4 months (Fig. 3e; WT, t(14) = 6.59, p < 0.001; AppNL/NL, t(14) = 5.88, p < 0.001; AppNL-G-F/NL-G-F, t(12) = 3.28, p = 0.007) and 6 months of age (Fig. 3j; WT, t(14) = 3.91, p = 0.002; AppNL/NL, t(14) = 4.07, p = 0.001; AppNL-G-F/NL-G-F, t(12) = 3.22, p = 0.007). Moreover, all genotypes exhibited similar levels of exploration of the holes in the target quadrant, with no differences in general exploratory activity, at both 4 months (Fig. 3f; (left) F[2, 20] = 0.05, p = 0.949; (right) F[2, 20] = 1.03, p = 0.375) and 6 months of age (Fig. 3k; (left) F[2, 20] = 0.67, p = 0.524; (right) F[2, 20] = 0.75, p = 0.487). These results suggest that both AppNL-G-F/NL-G-F and AppNL/NL mice had intact spatial learning and memory at 4 and 6 months of age.
At 8 months of age, AppNL-G-F/NL-G-F mice exhibited a significant increase in the number of errors (Fig. 3l; F[2, 18] = 5.34, p = 0.015, post hoc, WT vs. AppNL-G-F/NL-G-F, p = 0.015), latency (Fig. 3m; F[2, 18] = 10.28, p = 0.001, post hoc, WT vs. AppNL-G-F/NL-G-F, p < 0.001), and distance (Fig. 3n; F[2, 18] = 6.24, p = 0.009, post hoc, WT vs. AppNL-G-F/NL-G-F, p = 0.016) in comparison with WT mice. However, AppNL-G-F/NL-G-F mice still exhibited a significant decrease in the number of errors (Fig. 3l; F[2.5, 45.2] = 11.47, p < 0.001), latency (Fig. 3m; F[1.4, 25.2] = 6.91, p = 0.008), and distance (Fig. 3n; F[2.1, 38.1] = 8.35, p = 0.001), and were able to solve the task proficiently (at levels comparable to those of WT mice) by the fifth training session. These results suggest that 8-month-old AppNL-G-F/NL-G-F mice have subtle alterations in their ability to learn the spatial location of the target hole.
In the probe test, all genotypes exhibited similar levels of preference toward the target quadrant (Fig. 3o; genotype, F[2, 18] = 1.36, p = 0.283; quadrant, F[1.8, 31.9] = 38.63, p < 0.001). The percentages of time spent in the target quadrant were significantly higher above chance level for WT and AppNL/NL mice (Fig. 3o; WT, t(14) = 6.00, p < 0.001; AppNL/NL, t(12) = 5.07, p < 0.001), but not for AppNL-G-F/NL-G-F mice (t(10) = 2.11, p = 0.062), with our sample size. The percentage of hole exploration in the target quadrant (Fig. 3p; (left) F[2, 18] = 3.35, p = 0.058) and the total number of hole visits (Fig. 3p; (right) F[2, 18] = 0.35, p = 0.712) were similar among all genotypes.
Taken together, these results suggest that, at 8 months of age, AppNL-G-F/NL-G-F mice exhibit reduced spatial learning ability in comparison with WT mice, but still retain normal spatial memory.
Both App
NL-G-F/NL-G-F and App
NL/NL mice exhibit normal flexibility in a reversal learning task up to 8 months of age
Reversal learning, a way to model some aspects of higher-order cognitive functions in rodents [44, 45], requires cognitive flexibility and impulse control, and thus taps into components of human executive function [46, 47]. Previous studies demonstrated that transgenic mouse models of Aβ amyloidosis are impaired in reversal learning [48,49,50]. To assess reversal learning using the BM, we moved the target hole to the opposite position 1 day after the probe test (Fig. 3a [right]).
We found that both AppNL-G-F/NL-G-F and AppNL/NL mice exhibited similar levels of performance in the reversal learning task in comparison with WT mice at 4 months (Fig. 4a; F[2, 20] = 0.35, p = 0.711, Fig. 4b; F[2, 20] = 0.87, p = 0.434, Fig. 4c; F[2, 20] = 0.32, p = 0.733), 6 months (Fig. 4d; F[2, 20] = 0.38, p = 0.690, Fig. 4e; F[2, 20] = 6.31, p = 0.008, Fig. 4f; F[2, 20] = 1.73, p = 0.202), and 8 months of age (Fig. 4g; F[2, 18] = 0.66, p = 0.530, Fig. 4h; F[2, 18] = 3.00, p = 0.075, Fig. 4i; F[2, 18] = 1.41, p = 0.269).
The number of errors (Fig. 4a; F[1.8, 36.4] = 28.55, p < 0.001, Fig. 4d; F[2.5, 49.8] = 30.91, p < 0.001, Fig. 4g; F[2.2, 40.0] = 28.20, p < 0.001), latency (Fig. 4b; F[1.1, 22.9] = 13.28, p = 0.001, Fig. 4e; F[1.3, 25.6] = 24.78, p < 0.001, Fig. 4h; F[2.1, 38.6] = 29.80, p < 0.001) and distance (Fig. 4c; F[1.5, 29.1] = 20.47, p < 0.001, Fig. 4f; F[1.4, 28.6] = 21.32, p < 0.001, Fig. 4i; F[2.4, 42.7] = 21.93, p < 0.001) to the new target hole were progressively reduced in all genotypes and at all ages, suggesting that both AppNL-G-F/NL-G-F and AppNL/NL mice could adjust their response to find the new location as effectively as WT mice.
We noticed that 6-month-old AppNL/NL mice spent more time to reach the new target hole than WT mice at the first session of the reversal phase (Fig. 4e; genotype × session, F[2.6, 25.6] = 3.47, p = 0.037, simple main effect on first session, p < 0.001, post hoc, WT vs. AppNL/NL, p < 0.001). However, no significant change in the latency was detected at the second session (simple main effect on second session, p = 0.833). Moreover, no such difference was observed in the 8-month-old AppNL/NL mice (Fig. 4h).
Taken together, these results suggest that both AppNL-G-F/NL-G-F and AppNL/NL mice exhibit normal cognitive flexibility in the reversal learning task up to 8 months of age.