Pavlovian eyeblink conditioning
To examine the behavioral effects of physical and social enrichment on cerebellar learning, we tested our animals in a delay eyeblink conditioning paradigm. First, standard-housed mice (n=12) and enriched-housed mice (n=16) were trained for ten consecutive days with an interval of 250 ms between the CS and US onsets. Subsequently, both groups were trained for another ten consecutive days with a longer interstimulus interval (ISI) of 500 ms (Figure 1), which is known to be more challenging for mice 36,45.
Normalized eyelid closure - all trials
We evaluated the normalized eyelid closure over all trials (NECall_trials) for the first ten days of training with the ISI of 250 ms (Figure 2A). In standard-housed mice, the NECall_trials started at 0.03 (±0.02) on day 1 and increased to 0.53 (±0.16) on day 10 (Supplementary Table 1). In the enriched-housed mice, NECall_trials increased from 0.03 (±0.02) on day 1 to 0.39 (±0.10) on day 10. All values represent mean ± 95% confidence interval. We found a significant effect of ‘group’ (F(1,26)=9.44, p=0.0049), ‘training day’ (F(9,233)=16.45, p<.0001), and ‘group*training day’ interaction (F(9,233)=2.33, p=0.0159). Post hoc testing and correction for multiple comparisons using Holm-Bonferroni, revealed that enriched-housed mice showed statistically significant lower values for NEC - all trials at training day 4 (p=0.0456), day 5 (p=0.0040), day 6 (p=0.0045) (Figure 2B, Supplementary Table 1). After the switch to the longer interstimulus interval of 500 ms, standard-housed mice had a NECall_trials of 0.32 (±0.11) on day 11 and 0.44 (±0.20) on day 20, whereas enriched-housed mice started with 0.25 (±0.11) on day 11 and ended with 0.27 (±0.11) on day 20. We could not establish a significant effect of ‘group’ (F(1,26)=1.12, p=0.3003), ‘training day’ (F(9,228)=1.50, p=0.1484) or ‘group*training day’ interaction (F(9,228)=0.05, p=0.4042) (Figure 2B, Supplementary Table 2).
Normalized eyelid closure – CR trials
Next, we looked into the normalized eyelid closure in CR positive trials (NEC – CR trials) for the first ten days of training with the 250 ms ISI. In standard-housed mice, the NECCR_trials started at 0.19 (±0.05) on day 1 and increased to 0.58 (±0.16) on day 10 (Supplementary Table 3). In the enriched-housed mice, NECCR_trials increased from 0.17 (±0.05) on day 1 to 0.45 (±0.09) on day 10. All values represent mean ± 95% confidence interval. We found a significant effect for ‘training day’ (F(9,196)=9.62, p=<.0001), and a trend for the effect of ‘group’ (F(1,26)=3.95, p=0.0576) and ‘group*training day’ interaction (F(9,196)=1.68, p=0.0953) during the 250 ms ISI paradigm (Figure 2C, Supplementary Table 3). Post hoc testing and correction for multiple comparisons using Holm-Bonferroni did not show a statistical effect for NEC- CR trials. After the switch to the longer interstimulus interval of 500 ms, standard-housed mice started with NECCR_trials at 0.41 (±0.09) on day 11 and 0.56 (±0.17) on day 20, whereas enriched-housed mice started at 0.35 (±0.12) on day 11 and ended with 0.41 (±0.11) on day 20. A significant effect was established for ‘training day’ (F(9,224)=2.34, p=0.0155), but no effect of ‘group’ (F(1,26)=0.30, p=0.5896) or ‘group*training day’ interaction (F(9,224)=1.15, p=0.3255) (Figure 2C, Supplementary Table 4).
CR percentage
Then, we evaluated the percentage of conditioned responses for the first ten days of training with the 250 ms ISI. In standard-housed mice, the CR percentagestarted at 11.64 (±7.31) on day 1 and increased to 92.34 (±6.35) on day 10 (Figure 2D, Supplementary Table 5). In the enriched-housed mice, CR percentage increased from 15.25 (±8.76) on day 1 to 81.44 (±11.20) on day 10. All values represent mean ± 95% confidence interval. We found a significant effect of ‘group’ (F(1,26)=15.14, p=0.0004), ‘training day’ (F(9,233)=52.05, p<0.0001) and the ‘group*training day’ interaction (F(9,233)=7.83, p<.0001) during the 250 ms ISI paradigm. Post hoc testing and correction for multiple comparisons using Holm-Bonferroni, revealed that enriched-housed mice showed a statistically significant decrease in the CR percentage on training days 3-5 (p= 0.0010) and day 6 (p=0.0014) (Supplementary Table 5). After the switch to the longer interstimulus interval of 500 ms, standard-housed mice reached a CR percentage of 71.21 (±14.20) on day 11 and 76.77 (±27.01) on day 20, whereas enriched-housed mice started with 63.62 (±13.16) on day 11 and ended with 59.54 (±17.21) on day 20 (Figure 2D, Supplementary Table 6). We could not establish a significant effect of ‘group’ (F(1,26)=2.06, p=0.1634), ‘training day’ (F(9,228)=1.80, p=0.0684) nor ‘group* training day’ interaction (F(9,228)=0.78, p=0.6386) (Figure 2D, Supplementary Table 6).
Adaptive timing
Next, we assessed the adaptive timing of eyeblink CRs (Figure 3A), i.e. the latency to CR onset and latency to CR peak. For the first ten days of training with the 250 ms ISI, standard-housed mice started with a latency to CR onset at 185.38 (±119.31) on day 1 and ended at 145.92 (±33.40) on day 10 (Figure 3B, Supplementary Table 7). The enriched-housed mice started with a latency to CR onset at 259.03 (±91.96) on day 1 and ended at 161.43 (±17.17) on day 10. All values represent mean ± 95% confidence interval. The latency after CR onset revealed a significant effect of ‘group’ (F(1,26)=7.23, p=0.0123) and ‘training day’ (F(9,196)=2.76, p=0.0047), but not of ‘group*training day’ interaction (F(9,196)=1.60, p=0.1186) (Figure 3B). Post hoc testing and correction for multiple comparisons using Holm-Bonferroni, revealed a statistical significance on day 3 (p=0.0030) driven by the standard-housed mice (Supplementary Table 7). Standard-housed mice showed overall consistent adaptive timing of CR’s, while enriched-housed mice showed a steeper learning curve. After the switch to the longer 500 ms interstimulus interval, standard-housed mice started with a latency to CR onset at 142.50 (±21.99) on day 11 and ended at 188.36 (±25.72) on day 20 (Figure 3B, Supplementary Table 8). The enriched-housed mice started with a latency to CR onset at 177.11 (±30.66) on day 11 and ended at 185.36 (±31.31) on day 20. The latency to CR onset revealed a significant effect of ‘training day’ (F(9,224)=5.05, p<.0001), but not for ‘group’ (F(1,26)=1.08, p=0.3076) or the ‘group*training day’ interaction (F(9,224)=0.55, p=0.8395) (Supplementary Table 8).
We evaluated the latency to CR peak for the first ten days of training with the 250 ms ISI, standard-housed mice started with a latency to CR peak at 448.60 (±235.23) on day 1 and decreased to 325.84 (±33.41) on day 10 (Figure 3C, Supplementary Table 9). The enriched-housed mice started with a latency to CR peak at 530.54 (±140.86) on day 1 and decreased to 303.70 (±49.79) on day 10. All values represent mean ± 95% confidence interval. We found a significant effect of ‘training day’ (F(9,196)=3.47, p=0.0005) and ‘group*training day’ interaction (F(9,196)=2.00, p=0.0418), but not for ‘group’ (F(1,26)=1.93, p=0.1765) during the 250 ms ISI paradigm. Post hoc testing and correction for multiple comparisons using Holm-Bonferroni, revealed a statistical significance on day 3 (p=0.0030), driven by better timing of enriched-housed mice (Figure 3C, Supplementary Table 9). After the switch to the longer interstimulus interval of 500 ms, standard-housed mice showed a latency to CR peak started at 396.63 (±83.67) on day 11 and 644.17 (±67.70) on day 20, whereas enriched-housed mice started with 377.43 (±72.92) on day 11 and ended with 544.50 (±108.76) on day 20. We established a significant effect for latency to CR peak of ‘training day’ (F(9,224)=10.95, p<.0001), but not for ‘group’ (F(1,26)=0.79, p=0.3810) or the ‘group*training day’ interaction (F(9,224)=0.76, p=0.6514) (Supplementary Table 10).
We also analyzed the “perfectly timed” CR’s defined as CRs that peak in the very close proximity of the expected onset of the US (Day 7-10: 250 ± 50 ms; Day 17-20: 500 ± 100 ms). During the 250 ms ISI paradigm, standard-housed mice showed a percentage of 39.39 (±4.47) perfectly timed CRs, and enriched-housed mice of 57.56 (±4.17). All values represent mean ± 95% confidence interval. The percentage of perfectly timed CRs showed a significant effect for ‘group’ (F(1,26)=5.30, p=0.0296), driven by better timing of enriched-housed mice (Figure 3D, Supplementary Table 11). After the switch to the longer interstimulus interval of 500 ms, standard-housed mice showed a percentage of 28.69 (±4.66) perfectly timed CRs, and enriched-housed mice of 35.92 (±4.65). For this longer interval no significant effect was found on the perfectly timed CRs for ‘group’ (F(1,26)=1.31, p=0.2630) (Supplementary Table 12).
Based on these results, we conclude that environmental enrichment in mice does not enhance learning speed in eyeblink conditioning, but slightly improves the adaptive timing of eyeblink CRs.
Balance beam test
To test motor balance, standard-housed control mice (n=11) and enriched-housed mice (n=16) performed the balance beam test (Fig. 4A). We analyzed the time it took for each mouse to cross the beam, for both a 6- and 12-mm width beam. To determine statistical significance, we used a linear mixed-effect (LME) model with ‘group’, ‘beam width’, and ‘group*beam width’ interaction used as fixed effects and ‘mouse’ as a random effect. For the 12 mm wide beam, we found that standard-housed mice needed on average 6.48 (±1.67) seconds and enriched mice needed on average 6.03 (±1.57) seconds to cross the beam (all values: median ± 95% CI). For the narrower 6 mm beam, it took standard-housed control mice on average 8.70 (±1.65) seconds and enriched mice on average 7.83 (±1.58) to walk from one side to the other (Fig. 4B). With an ANOVA on our LME, we found a significant effect for ‘beam width’ (F(1,77)=4.54, p=0.0362), but no effects for ‘group’ (F(1,25)=0.03, p=0.8654) nor the ‘group*beam width’ interaction (F(1,77)=0.82, p=0.3671) (Supplementary Table 13). We thus conclude that cage enrichment did not lead to major improvements in motor balance.
Grip strength test
Muscle strength can have a major effect on motor functioning. We therefore assessed muscle strength of the forelimbs using the grip strength test in standard-housed mice (n=11) and enriched-housed mice (n=16) (Fig. 4C). We determined the maximal muscle strength (N) for each mouse. To determine statistical significance, we used a linear mixed-effect model with ‘group’ as a fixed effect and ‘mouse’ as a random effect. We found that the peak muscle strength value of standard-housed mice was on average 90 (±4.11) Newtons and that enriched mice could deliver on average 102 (±4.62) Newtons (all values: median ± 95% CI) (Figure 4D). Although the enriched mice tended to be slightly stronger, the effect of ‘group’ was not statistically significant (F(1,25)=2.48, p=0.1281) (Supplementary Table 13). We conclude that enriched housing had no major impact on peak muscle strength of the mouse’s forelimbs.
Accelerating Rotarod
To test motor performance, standard-housed control mice (n=11) and enriched-housed mice (n=16) were trained on the accelerating rotarod (Fig. 4E). We analyzed the latency to fall (s) from the rotating rod for each day for 4 consecutive training days. We used a linear mixed effect model with ‘training day’, ‘group’ and ‘training day * group’ interaction as fixed effects and ‘mouse’ as a random effect.
Standard-housed mice were able to walk on the accelerating rod on day 1 for 108.18 (±48.08) seconds and for 178.68 (±47.04) seconds on day 4 (for both values: mean ± 95% CI) (Fig. 4F). We found a significant effect of ‘day’ within the standard-housed group (ANOVA on LME, main effect: F(3,162)=27.61, p<0.0001; post hoc, day 1-2, p<0.0001; day 2-3, p=0.09; day 3-4, p=0.917; correction for multiple comparison using Bonferroni-Holm method) (Fig. 4F; Supplementary Table 13). In contrast, enriched-housed mice started with values of 192.25 (±33.14) seconds on day 1 and ended with 213.13 (±34.21) seconds on day 4. Herewith, the enriched mice performed better on day 1 than the standard-housed mice did on day 4. For the enriched mice, we also found a significant effect of ‘day’ (ANOVA on LME, F(3,237)=6.785, p=0.0002; post hoc, day 1-2, p=0.0003; day 2-3, p=0.63; day 3-4, p=0.54; correction for multiple comparison using Holm-Bonferroni method) (Fig. 4F; Supplementary Table 13). When comparing the two groups, we found a significant effect for ‘training day’ (F(3,399)=25.28, p<.0001), ‘group’ (F(1,25)=9.72, p=0.0046), and the ‘training day * group’ interaction (F(3,99)=6.65, p=0.0002), whereby enriched mice performed much better than the standard-housed mice. We thus conclude that cage enrichment led to significant improvements in motor performance in mice.
ErasmusLadder
To further assess locomotion patterns, standard-housed mice (n=11) and enriched-housed mice (n=16) were trained for five consecutive days on the ErasmusLadder. Each training day, mice received 42 trials. Each trial was a crossing of the ladder from one shelter box to the other. We quantified the percentage of “correct steps'' within each trial, whereby a correct step was defined as a step of the front-paws from a high rung to the next high rung, irrespectively of the length of the step. A step where a lower rung was touched upon, was considered as a “misstep” (Fig. 5A). Inspection of the correct steps for all training days, revealed a relatively normal distribution for both groups (Fig. 5B, C). We used a linear mixed effect model with ‘group’, ‘day’ and the ‘group*day’ interaction as fixed effects and ‘mouse’ as a random effect.
Standard-housed mice had on average 50.27(±4.31) percent correct steps on day 1 and 64.32 (±3.34) percent correct steps on day 5 (all values: mean ± 95% CI) (Fig. 5C).Enriched-housed mice started with an average percentage correct steps of 63.03 (±4.14) on day 1 and ended with an average percentage correct steps of 68.26 (±3.46) on day 5 (Fig. 5C; Supplementary Tables 16). Herewith, the enriched mice performed on day 1 almost as well as the standard-house mice on day 5. We found a significant main effect of ‘day’ (F(4,5376)=110.99, p=<.0001), ‘group’ (F(1,25)= 3.26, p=0.0833) and ‘group*day’ interaction (F(4,5376)= 17.92, p=<.0001). Post hoc testing revealed a significant difference between groups on day 1 (p = 0.0268), but not for the other days (after correction for multiple comparisons using Holm-Bonferroni) (Fig. 5D; Supplementary Table 16). Together, we conclude that cage enrichment significantly improved locomotion in mice on the ErasmusLadder test.