Fig 1.
Diagram of the PiPaw2.0 automated home cage lever-pulling system.
(A) The PiPaw2.0 system consists of a lever-pulling mechanism attached to the side of a mouse home cage. The setup includes the following components: a motor/encoder to measure lever position and movement while controlling the force applied to the lever, an RFID tag reader to identify individual mice, a water spout for dispensing water rewards, an infrared beam to detect nose pokes to start a trial, and the lever which mice must pull to receive a reward. (B) Schematic representation of the lever’s movement range (0° to 30° from front to back) and the goal range for successful lever pulls. The threshold position for registering a pull is set at 3° to eliminate random lever movements and ensure a genuine pull. The goal range for a successful pull is between 6° and 24°. The duration that the lever must be held within this goal range is dynamically adjusted based on the performance of individual mouse to increase task difficulty and assess motor learning and execution.
Fig 2.
Task performance and motor learning across genotypes.
(A) Analysis of the duration mice spent in the initial, less complex ‘Stage 1’ of the task revealed no significant difference between genotypes (Mann-Whitney test, p = 0.378). (B) Representative lever-pull trajectory plots from 100 consecutive trials of an expert WT mouse with a required hold time of 1 second (left panel). A detailed view of a single trial further illustrated the precision in hold time necessary to meet the task criteria (right panel). (C) Throughout Stage 2, daily success rates showed no significant variation between genotypes or across days in the cage (RM two-way ANOVA genotype p = 0.599 F(1, 22) = 0.2839, days p = 0.477 F(8.331, 179.1) = 0.9523, interaction p = 0.897 F(58, 1247) = 0.7696), suggesting a consistent performance level maintained by all mice. (D) The average daily hold time of WT mice showed an increase in response to the progressively demanding requirements of the task, while zQ175 mice reached a plateau at a significantly lower average hold time (RM two-way ANOVA, genotype p = 0.018 F(1, 22) = 6.432, days p < 0.0001 F(4.369, 96.11) = 9.192, interaction p < 0.0001 F(57, 1254) = 2.513). Plots in A, C, and D show mean±SEM.
Fig 3.
Task engagement, activity, and motivation.
(A) Daily number of trials shows a nonsignificant trend toward a higher number of trials for WT mice compared to zQ175, suggesting similar levels of engagement between the two genotypes (RM two-way ANOVA, genotype p = 0.131 F(1, 22) = 2.420, days p = 0.535 F(8.685, 191.1) = 0.8867, interaction p = 0.789 F(57, 1254) = 0.8440). (B) The average time spent in the chamber per entrance showed no significant variations across days in cage or between genotypes (RM two-way ANOVA, genotype p = 0.390 F(1, 22) = 0.7684, days p = 0.290 F(8.065, 175.9) = 1.219, interaction p = 0.453 F(8.065, 175.9) = 1.219). (C) Analysis of the average number of trials per entrance revealed a divergence later in the testing period (after day 26), where WT mice performed more trials per entrance compared to zQ175 mice, hinting at a differential evolution in task engagement strategies between genotypes (RM two-way ANOVA over entire period of task engagement: genotype p = 0.065 F(1, 22) = 3.748, days p = 0.757 F(6.148, 134.3) = 0.5698, interaction p = 0.002 F(57, 1245) = 1.629). (D) Analysis of the frequency of chamber entries showed an opposite pattern to the average trial per entry, where WT mice entered less frequently after day 26 compared to zQ175 mice (RM two-way ANOVA over entire period of task engagement: genotype p = 0.183 F(1, 22) = 1.890, days p = 0.346 F(8.782, 191.8) = 1.126, interaction p < 0.0001 F(57, 1245) = 1.917). (E) The distribution of bout sizes (histogram; left y-axis log scale) and the corresponding average success rate for each bout size (line plot; right y-axis) uncovered a genotype-dependent difference in bout size distribution, with zQ175 mice showing a tendency for engaging in fewer trials per bout. Additionally, success rates increased with bout size for both genotypes, yet no genotype difference in success rates was observed (RM two-way ANOVA, genotype p = 0.245 F(1, 22) = 1.426, bout size p < 0.0001 F(1.193, 26.24) = 124.3, interaction p < 0.0001 F(13, 286) = 6.476 for distribution; genotype p = 0.906 F(1, 22) = 0.01428, bout size p < 0.0001 F(2.824, 62.13) = 14.53, interaction p = 0.762 F(13, 286) = 0.7004 for success rate). Data in all plots are presented as mean±SEM.
Fig 4.
Modulation of performance during learning.
(A) The distribution of daily hold times for a representative WT (top) and zQ175 (bottom) mouse during early (days 1–3) and late (days 56–58) stages in the cage, complemented by a bimodal fit analysis. Only the late stage in the WT mouse shows a Gaussian component (green line), indicative of more timed responses, a feature absent in their early performance and persistently missing in zQ175 mice at the late stage, whose distribution remains predominantly exponential. The ‘Wexp’ value shows the weight of the exponential component in each distribution. (B) A longitudinal comparison of the average exponential fit weight between WT and zQ175 mice reveals that while WT mice show a decreasing trend, indicating a shift towards longer duration (“timed”) trials, zQ175 mice maintain a consistently high exponential weight, demonstrating no significant change over time (b; RM two-way ANOVA, genotype p = 0.0106 F(1, 22) = 7.795, days p < 0.0001 F(5.262, 115.4) = 7.616, interaction p < 0.0001 F(57, 1250) = 2.834). Data presented as mean±SEM.
Fig 5.
Analysis of movement jerkiness.
(A) Two example traces that illustrate a smooth (top) versus jerky (bottom) lever pull trajectory (black traces) with the corresponding 10Hz high-pass filtered trajectories superimposed (orange traces). The standard deviation (Std) of these high-pass filtered trajectories quantitatively captures the movement’s jerkiness. (B) The daily averages of Std for the high-pass filtered trajectories across WT and zQ175 mice further reveals significant differences in movement jerkiness, with zQ175 mice consistently exhibiting higher levels of jerkiness compared to WT mice, a difference that becomes more pronounced later in the testing period as average hold times increase (RM two-way ANOVA, genotype p = 0.017 F(1, 22) = 6.576, days p = 0.463 F(8.019, 176.4) = 0.9676, interaction p < 0.0001 F(57, 1254) = 2.762). Data presented as mean ± SEM.
Fig 6.
Variability in performance and response to task dynamics.
(A) Upon a change in the required hold time (dashed line), a significant reduction in success rate was similarly observed for both WT and zQ175 mice (RM two-way ANOVA, genotype p = 0.926 F(1, 20) = 0.008736, trial p < 0.0001 F(3.915, 78.31) = 27.99, interaction p = 0.985 F(19, 380) = 0.4237). (B) The analysis of trajectory variability, quantified as DTW distance of consecutive trials, showed a non-significant trend towards greater variability in zQ175 mice, peaking later compared to WT mice (RM two-way ANOVA, genotype p = 0.070 F(1, 20) = 3.660, trial p = 0.0005 F(4.291, 85.82) = 5.351, interaction p = 0.542 F(19, 380) = 0.9329). (C) Similarly, hold time variability, measured by moving standard deviation of hold times, presented a pattern where zQ175 mice exhibited a later peak in variability, suggesting a slower adaptation process (RM two-way ANOVA, genotype p = 0.571 F(1, 20) = 0.3293, trial p = 0.155 F(4.597, 91.94) = 1.668, interaction p = 0.0075 F(19, 380) = 2.011). (D) Further examination of trajectory variability within a 10-trial bout showed no differences between genotypes, with both demonstrating a reduction in variability (RM two-way ANOVA, genotype p = 0.630 F(1, 22) = 0.2383, trial p = 0.0002 F(4.154, 91.39) = 6.081, interaction p = 0.693 F(8, 176) = 0.6970). (E) When focusing on five consecutive successful trials, WT mice showed a significant reduction in trajectory variability, whereas zQ175 mice exhibited only a slight decrease, predominantly in the last trial, hinting at genotype-specific differences in optimizing performance following success (RM two-way ANOVA, genotype p = 0.746 F(1, 22) = 0.1074, trial p = 0.012 F(2.836, 62.38) = 4.045, interaction p = 0.018 F(3, 66) = 3.561). (F) Conversely, in five consecutive failed trials, no significant change in trajectory variability was observed for either genotype (RM two-way ANOVA, genotype p = 0.755 F(1, 22) = 0.09948, trial p = 0.631 F(2.664, 58.61) = 0.5467, interaction p = 0.743 F(3, 66) = 0.4142). Data presented as mean±SEM.
Fig 7.
Inter-mouse influence in a group-housed automated home cage setting.
(A) diagrammatic representation of the interaction motif between pairs termed as follower-influencer, specifying the criteria for identifying these interactions. (B) For comparative analysis, control patterns devoid of an influencer presence were also examined, characterized by a solitary mouse performing two bouts separated by a 5–15 minute timeout. (C) The frequency of follower-influencer motifs identified throughout the study period revealed no significant difference in occurrence based on the genotype of the follower, indicating similar levels of this interactive behavior across genotypes (unpaired t-test, p = 0.850). (D) In cases when an influencer exhibited higher success rates than the follower’s baseline (good influencer), WT but not zQ175 followers displayed a significant increase in success rate (one-sample t-test, theoretical mean 0.0, WT p = 0.014, zQ175 p = 0.366), although no genotype differences were noted when comparing these improvements (unpaired t-test, WT vs. zQ175 p = 0.290). (E) Conversely, when influencers had the same or lower success rates compared to the follower’s baseline (bad influencer), both WT and zQ175 followers showed no significant deviation from their initial success rates (one-sample t-test, theoretical mean 0.0, WT p = 0.766, zQ175 p = 0.695; unpaired t-test, WT vs. zQ175 p = 0.615). (F) The control scenarios, lacking an influencer and merely documenting the variance in success rates across two separated bouts, mirrored the outcomes observed with bad influencers, with no significant changes in success rates for either genotype (one-sample t-test, theoretical mean 0.0, WT p = 0.561, zQ175 p = 0.957; unpaired t-test, WT vs. zQ175 p = 0.851). Data presented as mean±SEM.
Fig 8.
Genotypic and hemispheric differences in striatal plasticity post PiPaw2.0.
(A) Schematic representation outlining the experimental setup, detailing the placement of stimulating and recording electrodes within the dorsolateral striatum. (B) Representative field responses before and after plasticity induction through high-frequency stimulation (HFS) depicted for both WT and zQ175 mice in the left hemisphere. (C) The dynamic changes in response amplitude relative to the baseline, following HFS, illustrated over the course of the plasticity experiment, providing a temporal view of the neural responses in WT and zQ175 mice, in the left hemisphere (contralateral to the lever-pulling paw) and right hemisphere (ipsilateral to the lever-pulling paw). (D) An analysis of the change in response size, averaged over minutes 30–35 post-HFS, revealed a significant hemispheric difference in plasticity induction among WT mice, a difference not observed in zQ175 mice (Two-way ANOVA, genotype p = 0.005 F(1, 41) = 8.493, hemisphere p = 0.0002 F(1, 41) = 17.23, interaction p = 0.064 F(1, 41) = 3.605; Sidak’s multiple comparison left vs. right, WT p = 0.0001, zQ175 p = 0.243). The numbers in parentheses in C and the points in D reflect number of slices from a total of 7 WT (4 male) and 5 zQ175 (3 male) mice. Data presented as mean±SEM.