A comprehensive and standardized pipeline for automated profiling of higher cognition in mice

Abstract

In rodent behavior research, observer-independent methods, such as the IntelliCage, enhance data collection in a social, and thus stress-reduced, environment. The IntelliCage system allows experimenters to create cognitive challenges for mice motivated by rewards. Given the extensive and diverse data from IntelliCage, there is a high demand for automated analysis. Here, we introduce IntelliR, a free and standardized pipeline for analyzing IntelliCage data, including a cognition index for performance comparison across challenges. IntelliR supports the automatic analysis of three challenges that cover spatial, episodic-like, and working memory with their reversal tests and can also be adapted for other designs. Results from three cohorts of adult female C57B6 mice showed improved task proficiency over time. To validate cognitive impairment detection, we used adult female NexCreERT2xRosa26-eGFP-DTA mice after neuron ablation in cortex and hippocampus, in which we observed reduced learning capabilities. IntelliR integrates easily into research, improving time management and reproducibility.

Keywords: CP: neuroscience; IntelliCage; automated phenotyping; behavior; cognitive domains; cognitive flexibility; episodic-like memory; reversal learning; spatial memory; working memory.

Graphical abstract

Figure 1

Descriptive overview of the IntelliCage setup and programmed challenges (A) IntelliCage system for cognitive challenges of up to 16 mice. Conditioning corners allow access to water bottles and are equipped with sensors to detect presence and track behavior of mice within the corners. (B) Graphical description of the three basic behavior readouts provided by the system: presence of mice within the tube leading to the water bottles (visit), mice touching the door blocking the access to a water bottle with their noses (nose poke), and, after this, the door opens and the mice have access to water bottles and are able to drink (licks). (C–E) Graphical outline of the programmed challenges representative for one group. (C) On day 1 (d1), mice have free access for 24 h to all water bottles in the corners (blue circles) to habituate to the new environment and task to get water. On d2, water access is only available in the dark phase (active phase) and one specific corner per group (place learning). On d3, water access is only in the dark phase and the corner diagonally opposite of the previously correct corner (reversal learning). On d4, during dark phase, correct corner for water access switches every 3 h between the two previously correct corners (multiple reversal learning). (D) After one day with free access to water for 24 h (extinction, d5), challenge 2 starts with access to water from 10p.m. to 12p.m. in the first of the previously learned corners (episodic like memory–acquisition). On d7, the settings from d6 are repeated (episodic like memory–retrieval). On d8, water access shifts to 8p.m.–10p.m. and to the corner diagonally opposite of the formerly correct corner (episodic like memory reversal–acquisition). On d9, the settings from d8 are repeated (episodic like memory reversal – retrieval). (E) After another day with free access to water for 24 h (extinction, d10), challenge 3 tests for working memory via the patrolling paradigm. On d11 and d12, water access is granted during active phase and correct corners switch clockwise after each correct drinking attempt (nose poke). The first correct corner is defined by the first drinking attempt of the mice. On d13 and d14, correct corners switch counterclockwise after each correct drinking attempt. First correct corner is defined by the last correct drinking attempt during clockwise patrolling.

Figure 2

Learning performance of three independent cohorts of healthy female mice in individual cognitive tasks Cumulative successes at each drinking attempt were plotted for individual mice (faint lines), and group performances calculated by linear regression (bold lines; 95% CI depicted as a shade) were compared to the average by-chance performance (dashed line) through estimated marginal means contrasts. All three independent cohorts of mice successfully learned all cognitive tasks as indicated by significant above-chance performance (all p < 0.0001). Only a small number of individual mice failed to learn some of the tests as indicated by close to or below chance performance. Global comparisons were calculated using analysis of covariance (ANCOVA), with specific group comparisons calculated through estimated marginal means contrasts.

Figure 3

Histopathological consequences of tamoxifen induced DTA expression in hippocampal pyramidal neurons (A) Fluorojade C and DAPI staining of male diphtheria toxin A (DTA) mice showed prominent neuronal degeneration in the cornu ammonis region 1 week after 3× tamoxifen injections as compared to DTA mice injected with 3× corn oil. Scale bar corresponds to 50 micrometers. (B) Iba1 and DAPI staining of hippocampal sections from mice presented in (A) show clear microgliosis upon tamoxifen induction. (C) Microgliosis and prominent hippocampal atrophy in tamoxifen-treated DTA mice that were used to validate the IntelliR pipeline. Eight mice per group (50%) were randomly selected and perfused for histological examination after the IntelliCage based phenotyping. (D) Quantitative assessment of hippocampal atrophy reveals a prominent shrinkage of the cornu ammonis subregions (p < 0.001), in which pyramidal cells have been ablated with tamoxifen, as well as a minor but significant atrophy of the dentate gyrus. (E) Densitometric analysis of the microglia/macrophage marker Iba1 shows a significant increase in microglia/macrophage density within the CA1 (p = 0.0395) and CA3 (p = 0.0249) but not the dentate gyrus of tamoxifen-induced DTA mice.

Figure 4

Learning performance of the DTA cohort To test if the cognitive challenges are sensitive enough to detect hippocampal dysfunction, the learning performance of DTA mice was tested after induction of hippocampal pyramidal cell death by tamoxifen and compared to corn-oil-treated healthy littermate controls. Cumulative successes at each drinking attempt were plotted for individual mice (faint lines) and group performances, calculated by linear regression (bold lines; 95% CI depicted as gray shadows), were compared to the average by chance performance (dashed line) and between groups through estimated marginal means contrasts. While control mice showed a significant above-chance performance in all tests (p < 0.0001), DTA mice failed to learn clockwise patrolling during day 1 and showed significantly slower learning rates (p < 0.05) than controls in all tests except place learning and episodic-like memory retrieval. Global comparisons were calculated using analysis of covariance (ANCOVA), with specific group comparisons calculated through estimated marginal means contrasts.

Figure 5

Cognitive performance of tamoxifen-induced (red) versus corn-oil-treated (black) DTA mice (A) Place errors are plotted to evaluate the spatial component of all cognitive tests. DTA mice show significant deficits during the first day of clockwise (p = 0.005) and counterclockwise patrolling (p = 0.0009) respectively. (B and C) Challenge errors (B) and direction errors (C) are presented for all challenges involving a reversal task to evaluate extinction of the previous task and cognitive flexibility. In (B), we show that there was a significantly lower challenge error for the DTA animals during MRL (p < 0.001), while their challenge error was significantly higher during ELM day 3 (p = 0.0216). The dotted lines indicate the average by-chance performance (3/4 corners are incorrect). (D) Time errors are plotted to assess learning of the temporal component of the episodic-like memory test. Average by-chance performance is indicated by the dotted line (22/24 h are incorrect). A significant change across days was observed (p = 0.0002). (E) To assess the overall cognitive performance of mice across all challenges, the cognition indices are plotted for all challenges. Taking the different types of errors into account, the cognition index shows that within each challenge, mice significantly improved their cognitive performance with time (all p < 0.0001, repeated measures ANOVA for challenge 1 and Friedman’s two-way test for challenges 2 and 3). Furthermore, the cognition index shows a significantly worse cognitive performance in the patrolling task in mice upon hippocampal pyramidal cell ablation (p = 0.0002, Friedman’s two-way test). Data presented as mean ± standard deviation. Pairwise comparisons in (A–D) calculated as described in the method section.

Figure 6

Monitoring basic activity-related mouse behavior using IntelliR Total number of corner visits (A), exploratory visits (B), and repetitive events (C) per test day (left graphs) and averages over all test days (right graphs) were used to evaluate general activity (A), exploratory behavior (B), and repetitive behavior (C). No differences were observed between tamoxifen-treated (red) and corn-oil-treated (black) DTA mice. Data presented as mean ± standard deviation. Welch’s corrected two-sided t tests were used to compare the averaged parameters.