Shank3 knockout diminishes taste-elicited insular suppression and perturbs CTA acquisition

(A) The head-restrained aversive taste learning paradigm. CS: conditioned stimulus, 25 mM saccharin, US: unconditioned stimulus, lithium chloride (LiCl, 0.15M, 2% body weight). CST: conditioning session, T: retrieval test session. (B) Example image showing the GRIN lens placement on top of the anterior insular cortex (AIC) (green: GCaMP8s, magenta: NeuN). Scale bar: 500 μm. (C) Example image showing the calcium signals recorded from individual AIC neurons. Scale bar: 50 μm. (D, E, F) Trial-averaged licks during the first (CST1) and second (CST2) conditioning sessions in the CS-only (WT) group (D), the CTA (WT) group (E), and the CTA (KO) group (F) (CS-only (WT): paired t-test, p = 0.1201; CTA (WT): paired t-test, p = 0.0094; CTA (KO): paired t-test, p = 0.0490). Here and below where behavior data are presented: CS-only (WT), n = 7 animals; CTA (WT), n = 9 animals; CTA (KO), n = 9 animals. (G) The trial-averaged lick ratio of CST2 to CST1 in CTA (WT) and CTA (KO) groups. (WT vs. KO: unpaired two-tailed t-test, p = 0.0123). (H, I) Heatmaps showing z-scored changes in calcium responses (ΔF/F) from neurons that are responsive to tastant delivery during conditioning in CTA (WT) (H, CST1: 189 cells, CST2: 172 cells) and CTA (KO) (I, CST1: 66 cells, CST2: 52 cells) groups. 1: cue onset; 2: spout in place; 3: tastant delivery starts; 4: tastant delivery ends. The 3-second taste delivery is marked by yellow dashed lines. In each heatmap, cells are sorted in descending temporal order based on whether their activity is activated or suppressed. Here and below where calcium imaging data are presented: CS-only (WT), n = 7 animals; CTA (WT), n = 8 animals; CTA (KO), n = 7 animals. (J) Calcium traces averaged across suppressed neurons in the CTA (WT) and CTA (KO) groups during CST1 (left) and CST2 (right). (K) Calcium traces averaged across activated neurons in the CTA (WT) and CTA (KO) groups during CST1 (left) and CST2 (right). (L) Change in the percentage of suppressed neurons from CST1 to CST2 (WT vs. KO, repeated-measures mixed ANOVA, genotype: p = 0.0218, session: p = 0.0393, interaction: 0.7819). (M) Change in the calcium fluorescence (ΔF/F) of suppressed neurons from CST1 to CST2. Dashed lines mark the value 0 (WT vs. KO, repeated-measures mixed ANOVA, genotype: p = 0.0011, session: p = 0.7356, interaction: 0.8672). (N) Change in the percentage of activated neurons from CST1 to CST2 (WT vs. KO, repeated-measures mixed ANOVA, genotype: p = 0.1391, session: p = 0.2590, interaction: 0.9705). (O) Change in the calcium fluorescence (ΔF/F) of activated neurons from CST1 to CST2 (WT vs. KO, repeated-measures mixed ANOVA, genotype: p = 0.3404, session: p = 0.2664, interaction: 0.1295). Error bars in this figure and below indicate s.e. (standard error of the mean).

Shank3 knockout increases the correlated variability in AIC during CTA acquisition

(A, B) Cumulative distribution of taste-elicited coactivity (Pearson coefficients) between neurons during CST1 (A) and CST2 (B). Dashed lines indicate the shuffled data (CST1: WT, n = 6201 pairs, KO, n = 2645 pairs, WT vs. KO, Kolmogorov-Smirnov test, p < 0.0001; CST2: WT, n = 8476 pairs, KO, n = 2847 pairs, WT vs. KO, Kolmogorov-Smirnov test, p < 0.0001). (C) Change in the average taste-elicited coactivity from CST1 to CST2, plotted by animal. The dashed line indicates the chance-level correlation. (WT vs. KO, repeated-measures mixed ANOVA, genotype: p = 0.0142, session: p = 0.4198, interaction: 0.4146). (D) Schematic showing how the signal and noise correlations are computed. (E, F) Change in the signal correlation from CST1 to CST2, plotted by neuronal pair (E) or by animal (F) (E: WT: CST1, n = 6201 pairs, CST2, n = 8476 pairs; KO: CST1, n = 2645 pairs, CST2, n = 2847 pairs, two-way ANOVA, genotype: p = 0.1134, session: p = 0.0009, interaction: 0.5469; F: WT vs. KO, repeated-measures mixed ANOVA, genotype: p = 0.8143, session: p = 0.4007, interaction: 0.8213). (G, H) Change in the correlated variability from CST1 to CST2, plotted by neuronal pair (G) or by animal (H) (G: two-way ANOVA, genotype: p < 0.0001, session: p = 0.2274, interaction: 0.6112; H: WT vs. KO, repeated-measures mixed ANOVA, genotype: p = 0.0617, session: p = 0.4017, interaction: 0.7188). (I) Average taste-elicited coactivity in CST1, grouped by inter-neuronal distance and plotted by animal (WT vs. KO, repeated-measures mixed ANOVA, genotype, p = 0.0050, distance, p = 0.7643, interaction: p = 0.5895).

Extinction of CTA in Shank3 knockout mice is faster than in wild-type littermates

(A, B, C) Trial-averaged licks of saccharin (violet) and water (blue) in the CS-only (WT) (A), CTA (WT) (B), and CTA (KO) (C) groups throughout the test sessions (CS-only (WT): repeated-measures ANOVA, taste, p = 0.2530, session, p = 0.4158, interaction, p = 0.4515; CTA (WT): repeated-measures ANOVA, taste, p < 0.0001, session, p < 0.0001, interaction, p < 0.0001; CTA (KO): repeated-measures ANOVA, taste, p = 0.0002, session, p < 0.0001, interaction, p < 0.0001). (D) Trial-averaged lick ratios of saccharin to water throughout the test sessions in CTA (WT) and CTA (KO) groups (WT vs. KO: repeated-measures mixed ANOVA, genotype, p = 0.0069, session, p < 0.0001, interaction, p = 0.0691, post-hoc Tukey test, WT vs. KO, T1 p = 0.8669, T2 p = 0.0032, T3 p = 0.2705, T4 p = 0.9014, T5 p = 0.9531). (E, F) A fitted logistic curve (top) and its derivative (bottom) describing the relationship between lick ratios and test sessions in an example WT mouse (E) and an example KO mouse (F). (G) Average rate of change (derivative) at each test session. The triangle indicates the data mean. (WT vs. KO: repeated-measures mixed ANOVA, genotype, p = 0.0271, session, p < 0.0001, interaction, p = 0.0003; post-hoc Tukey test, WT vs. KO, T1 p < 0.0001, T2 p = 1.0000, T3 p = 1.000, T4 p = 1.000, T5 p = 1.000).

CTA learning increases the response reliability of individual AIC neurons

(A) Example images showing neurons present in all five retrieval test sessions (labeled green). Scale bar: 20 μm. (B, C, D) Lifetime plot showing significant responses of the same neurons across test sessions (T1-5) in the CS-only (WT) group (B, n = 42 cells), the CTA (WT) group (C, n = 70 cells), and the CTA (KO) group (D, n = 39 cells). Violet labels activated states, and white labels nonactivated states. (E) Percentage of neurons active for 1, 2, 3, 4, and 5 sessions in (B-D) (Chi-square test, p < 0.0001). (F, G, H) Lifetime plot showing the water-responsiveness of the same neurons across test sessions (T1-5) in the CS-only (WT) group (F, n = 43 cells), the CTA (WT) group (G, n = 63 cells), and the CTA (KO) group (H, n = 36 cells). Blue labels activated states, and white labels nonactivated states. (I) Percentage of neurons active for 1, 2, 3, 4, and 5 sessions in (F-H) (Chi-square test: p < 0.0001).

The stimulus selectivity of individual AIC neurons degrades faster in Shank3 knockout mice during CTA extinction

(A) Calcium traces of example neurons activated by saccharin only (left), water only (middle), or both (right). Violet marks the calcium responses to saccharin, and blue marks the responses to water. (B) The percentages of nonselective neurons across test sessions and genotypes. Each data point represents one animal. The colored bar marks the data-averaged value and its standard error. The square indicates the fitted mean in each condition based on the linear mixed-effects model, with session and genotype set as fixed parameters and individual animal as random effect (WT vs. KO, linear mixed-effects model, genotype (in reference to WT): KO, β = −5.146, p = 0.554, session (in reference to T1): T2, β = 5.410, p = 0.438, T3, β = 11.692, p = 0.093, T4, β = 14.433, p = 0.038, T5, β = 8.045, p = 0.248, interaction: KO x T2, β = 16.729, p = 0.103, KO x T3, β = 13.466, p = 0.189, KO x T4, β = −5.420, p = 0.597, KO x T5, β = 21.409, p = 0.037, random effect, variance, 74.187; post hoc t-test, KO, T1 vs. T2, p = 0.0218). (C, D, E) Histograms comparing the distributions of the discriminability index between T1 and T2 (top) and between T1 and T5 (bottom) in the CS-only (WT) group (C, T1: 243 cells, T2: 238 cells, T5: 250 cells, F-test with Bonferroni correction, T1 vs. T2 p = 0.5979, T1 vs. T5 p = 0.2110), the CTA (WT) group (D, T1: 355 cells, T2: 382 cells, T5: 358 cells, F-test with Bonferroni correction, T1 vs. T2 p = 0.6687, T1 vs. T5 p = 0.0001), and the CTA (KO) group (E, T1: 212 cells, T2: 212 cells, T5: 241 cells, F-test with Bonferroni correction, T1 vs. T2 p < 0.0001, T1 vs. T5 p < 0.0001). (F) Comparison of absolute discriminability index across CS-only (WT), CTA (WT), and CTA (KO) groups. Each data point indicates a single-cell value, grouped by animal. Black bars indicate the “predicted” subject-level mean value from the mixed-effects model accounting for both fixed (session, genotype) and random (animal) effects. The colored bars mark the “predicted” marginal means from fixed effects (session and genotype) (linear mixed-effects model, group (in reference to CS-only, T1): CTA (WT), β = 0.1058, p = 0.1538, CTA (KO), β = 0.2442, p = 0.0018; session (in reference to CS-only, T1): T2, β = 0.0271, p = 0.4598, T5, β = −0.0035, p = 0.9239; interaction: CTA (WT) x T2, β = −0.0577, p = 0.2203, CTA (WT) x T5, β = −0.1082, p = 0.0213, CTA (KO) x T2, β = −0.2393, p < 0.0001, CTA (KO) x T5, β = −0.2856, p < 0.0001). Planned post hoc comparisons are tested using linear contrasts on the fitted model with Benjamini-Hochberg correction for multiple comparisons (T2 vs.T1: CTA (WT), p = 0.6455, CTA (KO), p = 0.9742; WT vs. KO: T1, p = 0.6455, T2, p = 0.0301, T3, p = 0.1040).

Learning improves population coding of taste in AIC, but this enhancement rapidly decays in Shank3 knockout mice

(A) Comparison of the Euclidean distance between population responses to water and saccharin between CTA (WT) and CS-only (WT) groups throughout test sessions (linear mixed-effects model, group (in reference to CS-only): CTA, β = 0.992, p = 0.005, session (in reference to T1): T2, β = −0.065, p = 0.760, T3, β = - 0.046, p = 0.829, T4, β = 0.101, p = 0.636, T5, β = 0.013, p = 0.950, interaction: CTA x T2, β = −0.033, p = 0.910, CTA x T3, β = −0.282, p = 0.336, CTA x T4, β = −0.680, p = 0.020, CTA x T5, β = −0.441, p = 0.132, random effect, variance, 0.312). (B) Diagram showing that training and prediction are performed within each test session. (C) Comparison of decoding accuracies of the linear classifiers trained and tested in each session for the CS-only (WT) and the CTA (WT) groups. The dashed line indicates the chance-level predictions (CS-only (WT) vs. CTA (WT): two-way ANOVA, group, p = 0.0001, session, p = 0.6637, interaction, p = 0.3273). (D, E) Confusion matrices showing the prediction performances in T1 (left) and T5 (right) in an example CTA (WT) mouse (D) and an example CS-only (WT) mouse (E). (F) Comparison of the Euclidean distance between population responses to water and saccharin between CTA (WT) and CTA (KO) groups throughout test sessions (linear mixed-effects model, genotype (in reference to WT): KO β = 0.225, p = 0.625, session (in reference to T1): T2, β = −0.099, p = 0.697, T3, β = −0.328, p = 0.195, T4, β = −0.578, p = 0.022, T5, β = −0.427, p = 0.091, interaction: KO x T2, β = −0.901, p = 0.015, KO x T3, β = −0.621, p = 0.094, KO x T4, β = - 0.674, p = 0.069, KO x T5, β = −0.618, p = 0.095, random effect, variance, 0.538; post hoc t-test, KO, T1 vs. T2, p = 0.0089). (G) Comparison of decoding accuracies of the linear classifiers trained and tested in each session for the CTA (WT) and the CTA (KO) groups. (WT vs. KO: two-way ANOVA, genotype, p = 0.0078, session, p = 0.7293, interaction, 0.7757). (H) Confusion matrices showing the prediction performances in T1 (left) and T5 (right) in an example CTA (KO) mouse. (I) Heatmaps of decoding accuracies using the classifiers trained on T1 responses and tested across sessions. Each row indicates an animal. (J) Quantification of the decoding accuracies in (I), comparing CTA (WT) and CTA (KO) groups (linear mixed-effects model, genotype (in reference to WT): KO, β = −1.417, p = 0.846, session (in reference to T1): T2, β = 0.031, p = 0.995, T3, β = −6.286, p = 0.214, T4, β = −7.545, p = 0.136, T5, β = - 10.031, p = 0.047, interaction: KO x T2, β = −17.732, p = 0.017, KO x T3, β = −14.260, p = 0.054, KO x T4, β = −8.013, p = 0.279, KO x T5, β = −8.261, p = 0.264, random effect, variance, 96.479; post hoc t-test, KO, T1 vs. T2, p = 0.0063).

Shank3 knockout hinders CTA acquisition in a single-conditioning paradigm

(A) The modified CTA learning paradigm. CS: conditioned stimulus, 25 mM saccharin, US: unconditioned stimulus, lithium chloride (LiCl, 0.15M, 2% body weight). BL: baseline, CST: conditioning session, T: retrieval test session, S: saccharin, W: water. (B) Trial-averaged licks of water and varying concentrations of saccharin (25 mM, 10 mM, 5 mM, 1 mM) in the retrieval test session (repeated-measures mixed ANOVA, genotype, 0.0049, taste, p < 0.0001, interaction, p = 0.0033). (C) Trial-averaged saccharin licks during the conditioning session (unpaired t-test, p = 0.0547). (D) Trial-averaged water licks during the two baseline sessions (WT: n = 13 animals, KO: n = 13 animals, BL-2, unpaired t-test, p = 0.7402; BL-1, unpaired t-test, p = 0.1662).

Comparison of taste-elicited AIC neuronal responses in the CTA and CS-only groups during conditioning

(A, B) Heatmaps showing z-scored changes in calcium responses (ΔF/F) from all recorded neurons during conditioning in WT (A, CST1: 313 cells, CST2: 363 cells) and KO (B, CST1: 194 cells, CST2: 200 cells) groups. 1: cue onset; 2: spout in place; 3: tastant delivery starts; 4: tastant delivery ends. The 3-second taste delivery is marked by yellow dashed lines. In each heatmap, cells are sorted in descending order based on the onset of their evoked responses. Cells that were responsive to taste delivery (between white horizontal lines) were selected and shown in the main figures 1H and 1I. (C) Percentage of neurons activated by the cue (left), lick movement (middle), and the tastant (right) in the CTA (WT) group (313 cells) during the first conditioning session (CST1). (D) Venn diagram showing the overlaps among cue-, lick-, and tastant-responsive neurons in (C). (E) Calcium traces averaged across activated neurons in the CTA (WT) and CS-only (WT) groups during CST1 (left) and CST2 (right). (F) Calcium traces averaged across suppressed neurons in the CTA (WT) and CS-only (WT) groups during CST1 (left) and CST2 (right). (G) Change in the percentage of activated neurons from CST1 to CST2 (CTA vs. CS-only, repeated-measures mixed ANOVA, group: p = 0.5202, session: p = 0.5460, interaction: 0.5372). (H) Change in the calcium fluorescence (ΔF/F) of activated neurons from CST1 to CST2 (CTA vs. CS-only, repeated-measures mixed ANOVA, group: p = 0.8465, session: p = 0.1287, interaction: 0.1650). (I) Change in the percentage of suppressed neurons from CST1 to CST2 (CTA vs. CS-only, repeated-measures mixed ANOVA, group: p = 0.9889, session: p = 0.5189, interaction: 0.0345. (J) Change in the calcium fluorescence (ΔF/F) of suppressed neurons from CST1 to CST2. Dashed lines mark the value 0 (CTA vs. CS-only, repeated-measures mixed ANOVA, group: p = 0.1166, session: p = 0.6204, interaction: 0.9379).

Shank3 knockout increases baseline coactivity in AIC during CTA acquisition

(A, B) Cumulative distribution of baseline coactivity (Pearson coefficients) between neurons during CST1 (A) and CST2 (B). Dashed lines indicate the shuffled data (CST1: WT, n = 6201 pairs, KO, n = 2645 pairs, WT vs. KO, Kolmogorov-Smirnov test, p < 0.0001; CST2: WT, n = 8476 pairs, KO, n = 2857 pairs, WT vs. KO, Kolmogorov-Smirnov test, p < 0.0001). (C) Change in the average baseline coactivity from CST1 to CST2, plotted by animal. The dashed line indicates the chance-level correlation. (WT vs. KO, repeated-measures mixed ANOVA, genotype, p = 0.0225, session, p = 0.7014, interaction: p = 0.2628). (D) Average baseline coactivity in CST1, grouped by inter-neuronal distance (WT vs. KO, repeated-measures mixed ANOVA, genotype, p = 0.0364, distance, p = 0.2776, interaction: p = 0.1821).

The number and amplitude of AIC neuronal responses to taste remain stable during CTA extinction

(A) Heatmaps showing z-scored changes in calcium responses (ΔF/F) to saccharin from all recorded neurons in the CTA (WT) group throughout retrieval test sessions (T1: 355 cells, T2: 382 cells, T3: 376 cells, T4: 356 cells, T5: 358 cells). (B) Heatmaps showing z-scored changes in calcium responses (ΔF/F) to saccharin from all recorded neurons in the CTA (KO) group throughout retrieval test sessions (T1: 212 cells, T2: 212 cells, T3: 221 cells, T4: 250 cells, T5: 241 cells). (C) The percentage (left) and the ΔF/F (right) of neurons activated by saccharin from T1 to T5 (percentage: WT vs. KO, repeated-measures mixed ANOVA, genotype, p = 0.5560, session, p = 0.0595, interaction, p = 0.2615; ΔF/F: repeated-measures mixed ANOVA, genotype, p = 0.9601, session, p = 0.5807, interaction, p = 0.4845). (D) The percentage (left) and the ΔF/F (right) of neurons suppressed by saccharin from T1 to T5 (percentage: WT vs. KO, repeated-measures mixed ANOVA, genotype, p = 0.4674, session, p = 0.2587, interaction, p = 0.6529; ΔF/F: repeated-measures mixed ANOVA, genotype, p = 0.0201, session, p = 0.0998, interaction, p = 0.8582). (E) The percentage (left) and the ΔF/F (right) of neurons activated by water from T1 to T5 (percentage: WT vs. KO, repeated-measures mixed ANOVA, genotype, p = 0.9612, session, p = 0.5387, interaction, p = 0.8687; ΔF/F: repeated-measures mixed ANOVA, genotype, p = 0.7128, session, p = 0.2653, interaction, p = 0.8600). (F) The percentage (left) and the ΔF/F (right) of neurons suppressed by water from T1 to T5 (percentage: WT vs. KO, repeated-measures mixed ANOVA, genotype, p = 0.2913, session, p = 0.6114, interaction, p = 0.7659; ΔF/F: repeated-measures mixed ANOVA, genotype, p = 0.0088, session, p = 0.7359, interaction, p = 0.3925).