Mice are able to perform preying tasks of varying difficulties
To study the predatory pursuit in mice, we customized an interactive platform with closed-loop control (Fig. 2A, see Methods for details). A robotic bait was first placed in the center of the arena to simulate an escaping prey. Subsequently, a habituated hungry mouse was placed at the border of the arena to forage freely for 60 s. The robotic bait was programmed to evade the mice at a predetermined velocity according to the task difficulty level (Fig. 2B). Figure 2C was a flowchart outlining the escape strategy of the robotic bait. Figure 2D showed a representative example of the actual scenario (Movie S3).
To test whether mice were able to perform preying tasks of varying difficulties, three levels of difficulty have been designed for preying tasks (see Methods for details, Fig. 3A). Figure 3B showed that all mice (n = 9/9) could complete the preying tasks within 60 s at low and medium difficulty, while 77.8% of mice (n = 7/9) were able to complete the preying task at the highest difficulty. These results suggested that the difficulty setting was appropriate (Movie S4) and enabled further investigation of preying behavior, particularly preying efficiency.
After repetitive interactions with the robotic bait, the mice were able to complete the preying tasks with a consistent time cost and were deemed experienced. A significant decrease in preying time was found between the inexperienced and the experienced mice at each difficulty level (Fig. 3C, low difficulty, from 35.22 s to 8.33 s, n = 9, P < 0.0001; 3D, medium difficulty, from 18.11 s to 8.11 s, n = 9, P = 0.0299; 3E, high difficulty, from 14.86 s to 7.71 s, n = 7, P = 0.0087, One-way ANOVA). There was also a significant reduction in the moving distance (Fig. 3F, low difficulty, P < 0.0001; 3G, medium difficulty, P = 0.0147; 3H, high difficulty, P = 0.0084, One-way ANOVA). These results suggested that experienced mice had improved preying efficiency, reflected by reduced preying time and moving distance.
In addition, the moving distance of the robotic bait also reduced (Fig. 3F, low difficulty, adjusted P = 0.0195; 3G, medium difficulty, adjusted P = 0.3770; 3H, high difficulty, adjusted P = 0.0188, One-way ANOVA), suggesting that it was more difficult for the programmed strategy to counter experienced mice, consistent with the improvement in the preying efficiency of mice.
Increase in preying efficiency is mainly due to improved pursuit phase
Since the robotic bait would stop escaping when the mouse was close enough, the major contributors to preying time and moving distance in this study would be searching and pursuing. A significant decrease in pursuit time was found between inexperienced and experienced mice at each difficulty level (Fig. 4A, low difficulty, P = 0.0001; medium difficulty, P = 0.0340; high difficulty, P = 0.0067, Paired t-test). There was also a significant reduction in the moving distance at both low and high difficulty levels (Fig. 4B, low difficulty, P = 0.0004; high difficulty, P = 0.0121, Paired t-test). At medium difficulty there was also a very close to statistically significant decrease in moving distance (Fig. 4B, medium difficulty, P = 0.0502, Paired t-test). No such difference was observed in the search phase (Fig. S2A, low difficulty, P = 0.0772, medium difficulty, P = 0.5973, high difficulty, P = 0.4481; S2B, low difficulty, P = 0.0772, medium difficulty, P = 0.4826, high difficulty, P = 0.1734, Paired t-test). In addition, the proportion of time spent in the pursuit phase was significantly lower in experienced mice than in inexperienced mice at all levels of difficulty (Fig. S2C, P < 0.0001, Paired t-test).
The average and maximum velocities of the mice during the pursuit were then compared. There was an increase in the average velocity of the experienced mice (Fig. 4C, low difficulty, P = 0.0027; medium difficulty, P = 0.0515; high difficulty, P = 0.0356, Paired t-test). This was consistent with the improvement of the pursuit phase. However, it was interesting that there was no significant change in maximum velocity at each difficulty (Fig. 4D, low difficulty, P = 0.1289, medium difficulty, P = 0.636, high difficulty, P = 0.9751, Paired t-test). These results suggested that the improvement of pursuit may not attributed to a significant increase in physical ability.
Experienced mice showed deviated pursuit strategy
In an effort to understand the pursuit strategy of mice, we performed a kinematic analysis of the pursuit phase. In both "deviated pursuit" and "parallel navigation" models, we calculated the average residual at each delay from 0 to 1 second (time step was 33 ms) for inexperienced and experienced mice. The minimum residual was considered the representative value of the model (dashed line in Fig. 5A-5B, S4A-S4B).
In deviated pursuit model, we compared the time series of measured bearing angle and predicted bearing angle φi−deviated in both inexperienced and experienced mice. The measured bearing angle in inexperienced mice partially aligned with the predicted value of the deviated pursuit model, primarily occurring in the later half of the pursuit (Fig. S4C, S3A-S3C). In experienced mice, the measured bearing angle was approximately consistent with the predicted value of the deviated pursuit and generally remained around 0 degrees (Fig. 5C, S3A-S3C).
Similarly, the predicted bearing angle φi−parallel in parallel navigation model was calculated. We conducted linear regression for the measured and predicted bearing angles, finding that none of the regression analyses yielded a slope close to the unit line for both inexperienced and experienced mice (Fig. 5D, S4D, experienced mice, -0.05472 ± 0.3959, n = 25, inexperienced mice, 0.06591 ± 0.1763, n = 25). Figure 5E and Fig. S4E showed the trajectories of an experienced and an inexperienced mouse during pursuit, respectively.
In addition, we also tested the pursuit strategy of mice using the proportional navigation model (see Methods for details). Both inexperienced and experienced mice exhibited a navigation constant of less than 1 (Fig. S5, S6, inexperienced mice, 0.2258 ± 0.2112, n = 25, experienced mice, 0.3967 ± 0.3585, n = 25).
Therefore, experienced mice in this study generally employed the deviated pursuit strategy when preying on the robotic bait, aligning their heading with the line of sight.
Experienced mice showed better control of velocity
To explore how experience enhances pursuit strategy, we compared the relative distances between inexperienced and experienced mice and the robotic bait at each difficulty level. A non-smooth difference was found between the inexperienced and experienced mice (Fig. S7A, low difficulty, P = 0.3590; S7B, medium difficulty, P = 0.4790; S7C, high difficulty, P = 0.2440, Augmented Dickey-Fuller test), suggesting that inexperienced and experienced mice had different pursuit characteristics. Experienced mice showed a significant reduction in the number of accelerations during the pursuit at each difficulty level compared to inexperienced mice (Fig. 6A, low difficulty, P = 0.0009; 6B, medium difficulty, P = 0.0470; 6C, high difficulty, P = 0.0488, Paired t-test. Also see Movie S5). These results suggested that experienced mice have fewer accelerations during pursuit.
We then normalized the count of accelerations relative to the total distance traveled, considering that the increased distance covered by the inexperienced mice might explain their higher number of accelerations. As shown in Fig. S8A-S8C, there was no statistically significant difference between experienced and inexperienced mice (low difficulty, P = 0.1309, medium difficulty, P = 0.3034, high difficulty, P = 0.9841, Paired t-test), indicating that the distance traveled during individual accelerations appears relatively consistent in mice.
To better understand the velocity control during pursuit, histograms of velocity (bin = 50 mm/s) were calculated (Fig. 6D-6F). A higher proportion in the low velocity range can be found in inexperienced mice at all difficulty levels, whereas a higher proportion in the high velocity range can be found in experienced mice. Taken together, these results suggested that experienced mice exhibited fewer but higher speed accelerations during pursuit, indicating better control of the velocity gained through experience.