Fig. 2. Chemogenetic silencing of bilateral OFC11/13 impaired experience- and inference-based value updating.
a Injection area (bilateral OFC, Brodmann’s area 11/13) and PET images showing hM4Di expression in each monkey. Filled and open arrowheads represent the lateral and medial orbital sulcus, respectively. b Shema of OFC11/13 silencing per se. DCZ (100 µg/kg) was systemically injected intramuscularly. c, d Behavioral effects of chemogenetic OFC11/13 silencing on NOVEL and FAMILIAR (d) task performance for each monkey. Data for vehicle injections (cyan) and DCZ injections (red) are shown. The optimal choice rate following DCZ (red) and vehicle injections (cyan) was averaged across two monkeys. Solid lines and shaded area represent the mean and s.e.m, respectively. Averaged optimal choice rate for each phase of reversal (Pre, Pre-reversal; PoE, PoL, early, and late phase of the post-reversal trials; see Methods) for each monkey in the NOVEL (e) and FAMILIAR (g) tasks. In the NOVEL task, a three-way ANOVA (subject × phase × treatment) revealed a significant main effect of treatment (F(1,72) = 12.3, p = 7.9 × 10−4) and a significant interaction between phase and treatment (F(2,72) = 11.0, p = 6.8 × 10−5). Subsequent two-way ANOVAs (subject × treatment) for each phase revealed significant differences for treatment during the PoE (F(1,24) = 30.0, p = 1.2 × 10−5), but not during the Pre (F(1,24) = 0.11, p = 0.74) or the PoL (F(1,24) = 0.25, p = 0.62). Similarly, in the FAMILIAR task, a three-way ANOVA revealed a significant main effect of treatment (F(1,60) = 20.8, p = 2.6 × 10-5) and significant interaction between phase and treatment (F(2,60) = 6.6, p = 2.5 × 10−3). The two-way ANOVAs revealed significant differences for treatment during the PoE (F(1,20) = 17.8, p = 4.2 × 10−4) and the PoL (F(1,20) = 8.9, p = 7.2 × 10−3), but not during the Pre (F(1,20) = 0.32, p = 0.58). Error bars: s.e.m. Estimated learning rates in the NOVEL task with the EXP model (f, two-way ANOVA, subject × treatment, main effect of treatment, F(1,24) = 11.3, p = 2.6 × 10−3; subject, F(1,24) = 0.04, p = 0.84; interaction, F(1,24) = 0.004, p = 0.95)) and in the FAMILIAR task with the INF model (h, treatment, F(1,20) = 7.5, p = 1.3 × 10−2; subject, F(1,20) = 0.04, p = 0.85; interaction, F(1,20) = 0.02, p = 0.90). Asterisks: p < 0.05 for significant main effect of treatment. Note that the learning rates for the NOVEL and FAMILIAR tasks were calculated using different models. Thus, they are not directly comparable. Data were obtained from N = 7 and 6 sessions for each treatment in each monkey for the NOVEL and FAMILIAR tasks, respectively.