Impaired reversal learning in male Shank2-KO mice
We first tested adult (12–24 week-old) male mice (10 WT and 10 Shank2-KO) on Task 1, in which one odor cue was paired with a small amount of water (6 µl) and another with an air puff (100 ms, 3 psi), each with 75% probability (Fig. 3A). The mice were trained in the task for three daily sessions of 400 trials (acquisition phase). We used the anticipatory lick rate during the delay period (1 s) as an index for discrimination between two odor cues throughout the study. The anticipatory lick rate diverged rapidly (in < 100 trials) according to CS during the first session (Fig. 3B). Overall, the anticipatory lick rate decreased gradually within each session, possibly reflecting a gradual decrease in thirst. Nevertheless, both genotypes showed higher anticipatory lick rates in CSRw trials than in CSPn trials throughout the three training sessions (Fig. 3B). We divided each session into four stages (100 trials each) and subjected the LDI (see Methods) to two-way mixed ANOVA (Fig. 3C). We found a significant main effect of training (F(11,198) = 8.014; p = 1.7×10− 11) along with a significant training×genotype interaction effect (F(11,198) = 3.318; p = 3.3×10− 4; main effect of genotype, F(1,198) = 1.193, p = 0.289). Post-hoc Bonferroni tests revealed that the LDI was significantly higher in KO mice than in WT mice in the fourth stage of the first session (stages 1–4; p = 0.211, 0.083, 0.073 and 0.004, respectively; second session, p-values > 0.2; third session, p-values > 0.2). These results indicate that although the initial learning rate was faster in Shank2-KO mice than in WT mice, both genotypes were over-trained to discriminate between CSRw and CSPn.
The mice were then trained until they reached the reversal criterion over 1–5 sessions (400 trials per daily session). The dynamic of lick-rate changes during reversal varied widely across individual mice. The number of trials required to reach the reversal criterion differed significantly between WT and Shank2-KO mice (240.3 ± 35.7 and 570.7 ± 146.6 trials, respectively; t-test, t(18) = 2.190, p = 0.042; Fig. 3D, E). The mice sometimes consumed water in the subsequent trial rather than during the inter-trial interval following a rewarded trial. To rule out the influence of such invalid anticipatory licks, we deleted the trials during which the mice consumed water delivered in the previous trial (215 out of 8115 trials; 2.65%) and determined the number of trials needed to reach the reversal criterion. This analysis also found that there was a significant difference between WT and Shank2-KO mice in the number of trials needed to reach the reversal criterion (240.3 ± 35.7 and 571.2 ± 146.7 trials, respectively; t-test, t(18) = 2.190, p = 0.042).
To test whether reversal learning was influenced by the level of initial training, we examined the relationship between the number of trials needed to reach the reversal criterion and the number of licks in CSRw trials, the number of licks in CSPn trials, the difference in the number of licks between CSRw and CSPn trials, and the LDI during the last acquisition session. We found that there was no significant relationship between these measures and the number of trials needed to reach the reversal criterion (Fig. 3H). This indicates that the difference in reversal learning between WT and Shank2-KO mice cannot be accounted for by different levels of initial learning.
We also found that the LDI during the first reversal session (calculated using trials #53–62) differed significantly between WT and Shank2-KO mice (-0.010 ± 0.184 and 0.449 ± 0.101, respectively; t-test, t(18) = 2.189, p = 0.042; Fig. 3F, G). No significant relationship was found between the LDI and the number of licks in CSRw trials, the number of licks in CSPn trials, the difference in the number of licks between CSRw and CSPn trials, or the LDI during the last acquisition session (Fig. 3I). Together, these results indicate that reversal learning is slower in Shank2-KO mice than in WT mice.
Impaired reversal learning in juvenile male Shank2-KO mice
Given that ASD is a neurodevelopmental disorder and people with ASD may show signs of behavioral inflexibility in childhood [72–77], we tested whether juvenile Shank2-KO mice also show deficits in reversal learning. Toward this end, juvenile (P30–45) male mice (10 WT and 10 Shank2-KO) were tested in Task 1 as described for adult male mice (Fig. 4A). The mice were trained for three daily sessions during the initial acquisition phase; unfortunately, however, the third-session data were lost due to a procedural error. We therefore assessed initial learning based on the first two sessions. The anticipatory lick rate diverged rapidly according to CS in the first session and this difference was maintained in the second session (Fig. 4B). Two-way mixed ANOVA of LDI revealed that there was a significant main effect of training (F(7,126) = 8.400; p = 2.1×10− 8), but no significant main effect of genotype (F(1,126) = 0.601; p = 0.448) or training×genotype interaction effect (F(7,126) = 1.628; p = 0.133; Fig. 4C). Although we could not examine the animals’ behavior during the third training session, the mice showed significantly different anticipatory licking responses between CSRw and CSPn trials at the outset (first 100 trials) of the first reversal session before reversal onset, and their rates did not differ significantly from those of the corresponding adult male mice (Additional file 1: Fig. S1). Moreover, the number of trials needed to reach the reversal onset criterion did not differ significantly from that of the corresponding adult male mice (Additional file 1: Fig. S1). These results indicate that the juvenile WT and Shank2-KO mice were well trained to discriminate between CSRw and CSPn.
Upon CS-US contingency reversal (2–4 daily sessions of 400 trials until the reversal criterion was reached), the number of trials needed to reach the reversal criterion differed significantly between the juvenile male WT and Shank2-KO mice (236.5 ± 50.6 and 634.9 ± 88.2 trials, respectively; t-test, t(19) = 2.189, p = 0.001; Fig. 4D, E; similar results were obtained after deletion of trials in which previously delivered water was consumed; 306 out of 8945 trials; 3.42%; t-test, t(19) = 3.428, p = 0.003). The LDI during the first reversal session (calculated using trials #53–62) also differed significantly between the two genotypes (0.203 ± 0.084 and 0.465 ± 0.079, respectively; t-test, t(18) = 2.282, p = 0.035; Fig. 4F, G). No significant relationship was found between these reversal-learning measures (the number of trials needed to reach the reversal criterion and the LDI) and the number of licks in CSRw trials, the number of licks in CSPn trials, the difference in the number of licks between CSRw and CSPn trials, or the LDI during the last acquisition session (Fig. 4H, I). These results indicate that, as seen for adult males, juvenile male Shank2-KO mice were impaired in the reversal learning task studied herein.
Intact reversal learning in female Shank2-KO mice
Because the prevalence of ASD is strongly male-biased [78], we examined whether adult (12–24 weeks old) female Shank2-KO mice also show reversal learning deficits in Task 1 (Fig. 5A). Adult female WT (n = 12) and Shank2-KO (n = 10) mice showed preferential anticipatory licking in response to CSRw compared to CSPn throughout the initial training sessions (Fig. 5B). Two-way mixed ANOVA of LDI revealed that there was a significant main effect of training (F(11,220) = 6.812; p = 7.7×10− 10), but no significant main effect of genotype (F(1,220) = 0.392; p = 0.538) or training×genotype interaction effect (F(11,220) = 1.611; p = 0.097; Fig. 5C).
Upon CS-US contingency reversal (2–4 daily sessions of 400 trials until reaching the reversal criterion), we found that there was no significant difference in the number of trials needed to reach the reversal criterion between female WT and Shank2-KO mice (387.9 ± 45.0 and 386.9 ± 77.2 trials, respectively; t-test, t(20) = 0.012, p = 0.991; Fig. 5D, E; similar results were obtained after deletion of trials in which previously delivered water was consumed; 347 out of 8598 trials; 4.04%; t-test, t(20) = 0.408, p = 0.687). We also failed to find a significant difference in the LDI (calculated using trials #53–62 of the first reversal session; 0.396 ± 0.100 and 0.502 ± 0.127, respectively; t-test, t(19) = 0.661, p = 0.516; Fig. 5F, G). These reversal-learning measures showed no significant relationship with the number of licks in CSRw trials, the number of licks in CSPn trials, the difference in the number of licks between CSRw and CSPn trials, or the LDI during the last acquisition session (Additional file 1: Fig. S2). These results indicate that the reversal learning of female Shank2-KO mice is intact.
Enhanced eye closure responses in male Shank2-KO mice
Since ASD is associated with atypical sensory responses and heightened anxiety [79–81], we examined whether WT and Shank2-KO mice show differential eye closure responses to air puff. Specifically, we delivered the air puff used in Task 1 (100 ms, 3 psi) without any preceding sensory cue (inter-trial interval, 9–11 s, uniform random distribution) and measured the fraction of eye closure before, during, and after air puff delivery (Fig. 2). Two-way ANOVA revealed that there were significant main effects for genotype (F(1,40) = 10.958, p = 0.004), time (F(2,40) = 24.723, p = 1.0×10− 7), and their interaction (F(2,40) = 3.729, p = 0.033) on eye closure responses (Fig. 6A). Post-hoc Bonferroni tests indicated that the eye closure response was significantly stronger in Shank2-KO than WT mice before (1.5-s time window before air puff onset; p = 0.001) and after (between 2.5 and 4 s since air puff onset; p = 0.005), but not during (1-s time window since air puff onset; p = 0.073) air puff delivery. These results indicate that anticipatory eye closure responses differ between adult male WT and Shank2-KO mice.
The above results raised the possibility that abnormal emotional responses to air puff may negatively affect reversal learning in male Shank2-KO mice. One way to explore this possibility would be to test male Shank2-KO mice using a mild air puff that does not induce an abnormal eye closure response. For this, we examined the relationship between air puff strength and eye closure response by systematically varying the duration and intensity of air puff (15 combinations other than the original one [100 ms, 3 psi]). Eye closure responses during air puff delivery (1-s time period since air puff onset) did not differ significantly between the two genotypes for any combination of air puff duration and intensity (two-way ANOVA followed by post-hoc Bonferroni test, p-values > 0.05; Fig. 6B). However, anticipatory eye closure responses before (1.5-s time window before air puff onset) and after (between 2.5 and 4 s since air puff onset) air puff delivery differed significantly between the two genotypes for some combinations of air puff duration and intensity (Fig. 6B). As expected, mild air puffs induced similarly low levels of anticipatory eye closure responses (Fig. 6A) in male WT and Shank2-KO mice. Based on these results, we chose to use the mildest air puff (5 ms, 3 psi; Fig. 6A and 6B) to further examine the reversal learning of male Shank2-KO mice.
Because female KO mice showed intact reversal learning in Task 1, we examined whether they would also show normal levels of eye close responses. In females, unlike males, Shank2-KO and WT mice showed similar levels of eye closure before, during, and after the delivery of the air puff used in Task 1 (100 ms, 3 psi; two-way ANOVA, main effect of genotype, F(1,38) = 0.057, p = 0.814; main effect of time, F(2,38) = 109.621, p = 1.7×10− 16; genotype⋅time interaction effect, F(2,38) = 0.222, p = 0.802; Fig. 6C). We also failed to find a significant difference in eye closure response between female WT and Shank2-KO mice to mild (5 ms, 3 psi; main effect of genotype, F(1,38) = 2.371, p = 0.140; main effect of time, F(2,38) = 62.549, p = 9.5×10− 13; genotype⋅time interaction effect, F(2,38) = 0.279, p = 0.758; Fig. 6C) and very strong (100 ms, 30 psi; main effect of genotype, F(1,38) = 0.002, p = 0.964; main effect of time, F(2,38) = 47.561, p = 4.5×10− 11; genotype⋅time interaction effect, F(2,38) = 0.274, p = 0.762; Fig. 6C) air puffs. These results indicate that female Shank2-KO mice lack the heightened fear response observed in male Shank2-KO mice.
Intact reversal learning of male Shank2-KO mice in the presence of mild air puff
Using the mildest air puff (5 ms, 3 psi), to which adult male WT and Shank2-KO mice showed similar anticipatory eye closure responses, we tested another group of adult male mice (10 WT and 10 Shank2-KO) for reversal learning (Task 2; Fig. 7A). Both male WT and Shank2-KO mice quickly developed and maintained preferential anticipatory licking responses to CSRw versus CSPn during the initial training sessions. Two-way mixed ANOVA of LDI revealed that there was a significant main effect of training (F(11,187) = 14.010; p = 1.9×10− 19) and genotype (F(1,187) = 14.561; p = 0.001), but no significant interaction effect between the two (F(11,187) = 1.092; p = 0.370; Fig. 7B, C). Thus, both genotypes learned the CS-US contingencies well, but the initial learning was stronger in WT mice than in Shank2-KO mice.
The mice were then trained until they reached the reversal criterion over 2–4 daily sessions (400 trials each). The number of trials needed to reach the reversal criterion did not vary significantly between the male WT and Shank2-KO mice (373.6 ± 117.5 and 379.0 ± 79.9 trials, respectively; t-test, t(18) = 0.038, p = 0.970; Fig. 7D, E; similar results were obtained after deletion of trials in which previously delivered water was consumed; 114 out of 7697 trials; 1.48%; t-test, t(18) = 0.084, p = 0.934). The LDI (calculated using trials #41–50 of the first reversal session) also did not differ significantly between the two genotypes (0.328 ± 0.059 and 0.386 ± 0.153, respectively; t-test, t(18) = 0.348, p = 0.732; Fig. 7F, G). These measures of reversal learning had no significant relationship with the number of licks in CSRw trials, the number of licks in CSPn trials, the difference in the number of licks between CSRw and CSPn trials, or the LDI during the last session of the initial training (Additional file 1: Fig. S3). These results indicate that male Shank2-KO mice exhibit intact reversal learning in Task 2. Similar results were obtained with juvenile (P30–45) male WT and Shank2-KO mice tested on Task 2 (Additional file 1: Fig. S4).
Intact reversal learning of Shank2-KO mice in the absence of aversive outcome
To further confirm that the behavioral flexibility of male Shank2-KO mice is intact in the absence of a strong aversive outcome, we tested another group of adult male WT (n = 10) and Shank2-KO (n = 10) mice in reversal learning using only appetitive outcomes. In Task 3, two different odor cues were paired with the same amount of water (6 µl), but with two different probabilities (80 and 20%; CS80% and CS20%, respectively; Fig. 8A). In both genotypes, anticipatory licking responses were higher to CS80% than CS20% throughout the acquisition session (Fig. 8B). Two-way mixed ANOVA of LDI revealed that there was a significant main effect of training (F(11,198) = 5.117; p = 4.8×10− 7) but no significant main effect of genotype (F(1,198) = 0.447; p = 0.513) or training×genotype interaction effect (F(11,198) = 1.329; p = 0.211; Fig. 8C).
During the reversal training (2–4 daily sessions of 400 trials until the reversal criterion), the number of trials needed to reach the reversal threshold did not differ significantly between the male WT and Shank2-KO mice (594.4 ± 97.1 and 614.2 ± 177.4 trials, respectively; t-test, t(18) = 0.098, p = 0.923; Fig. 8E; similar results were obtained after deletion of trials in which previously delivered water was consumed; 210 out of 10,686 trials; 1.97%; t-test, t(18) = 0.950, p = 0.355). Also, the LDI (calculated using trials #53–64 of the first reversal session) did not differ significantly between the male WT and Shank2-KO mice (0.207 ± 0.061 and 0.108 ± 0.042, respectively; t-test, t(16) = 0.076, p = 0.940; Fig. 8F, G), and these reversal-learning measures had no significant relationship with the number of licks in CS80% trials, the number of licks in CS20% trials, the difference in the number of licks between CS80% and CS20% trials, or the LDI during the last acquisition session (Additional file 1: Figure S5). These results verify that the reversal learning of Shank2-KO mice is intact in the absence of a strong aversive outcome.
Effect of DCS on fear response
Previous studies [60, 67] showed that DCS, a partial agonist of NMDA receptor, rescues the social interaction deficits of Shank2-KO mice. We therefore tested whether DCS would also rescue the behavioral deficits of Shank2-KO mice found in the current study. As repeated administration of DCS causes tachyphylaxis [82–85] and the drug’s half-life in mice is only 23 min [86, 87], it would have been difficult to test the effects of DCS on the above-described reversal learning, which takes a relatively long time (1–5 days of training). Given that our results suggested that enhanced fear is the source of reversal learning deficit in male Shank2-KO mice, and eye close responses can be tested within a short period of time, we examined whether DCS could rescue the abnormal eye closure responses (i.e., fear responses) of adult male Shank2-KO mice. We tested the effects of DCS on eye closure responses to the strong (100 ms, 3 psi), mild (5 ms, 3 psi), and very strong (100 ms, 30 psi) air puffs. For the strong air puff (100 ms, 3 psi), three-way ANOVA revealed that there were significant main effects of genotype (F(1,38) = 4.714, p = 0.043) and time (F(2,38) = 45.503, p = 8.2×10− 11), but no significant interaction between them (F(2,38) = 0.719, p = 0.494). The main effect of drug (F(1,38) = 0.122, p = 0.731) and the other interaction effects were not statistically significant (genotype⋅time, F(2,38) = 0.719, p = 0.494; drug⋅genotype, F(1,38) = 0.130, p = 0.723; drug⋅time, F(2,38) = 0.331, p = 0.720; drug⋅genotype⋅time, F(2,38) = 0.489, p = 0.617; Fig. 9A). Similarly, the main effect of drug and the effects of interactions involving the drug were not statistically significant for the mild (main effect of drug, F(1,38) = 0.020, p = 0.888; drug⋅genotype, F(1,38) = 0.012, p = 0.914; drug⋅time, F(2,38) = 0.050, p = 0.951; drug⋅genotype⋅time, F(2,38) = 0.189, p = 0.829; Fig. 9B) and very strong (main effect of drug, F(1,38) = 0.186, p = 0.671; drug⋅genotype, F(1,38) = 2.888, p = 0.106; drug⋅time, F(2,38) = 0.362, p = 0.700; drug⋅genotype⋅time, F(2,38) = 0.559, p = 0.577; Fig. 9C) air puffs. These results indicate that DCS does not rescue the enhanced fear response of male Shank2-KO mice to the air puff.