Fig. 4: The two thalamic cell types are engaged by different task input statistics.
From: Thalamic circuits for independent control of prefrontal signal and noise
a, Example tagged MDGRIK4 neuron recorded in the task. b, Tagged MDGRIK4 neurons are more likely to be conflict-preferring (n = 17 neurons from 3 mice; P = 0.0042; binomial). c, Example tagged MDD2 neuron recorded in the task. d, Tagged MDD2 neurons are more likely to be conflict-non-preferring (n = 20 neurons from 3 mice; P = 4.0 × 10−5; binomial). e, MDGRIK4 suppression recapitulates generic MD suppression (n = 20 sessions from 4 GRIK4-cre mice; Wilcoxon signed-rank). f, MDD2 inactivation enhances performance accuracy on trials with high cueing conflict (n = 20 sessions from 4 D2-cre mice; Wilcoxon signed-rank). g, Expanded neural model with two MD cell types. h, The two-cell-type model captures experimental data (n = 2,000 trials, chi-squared). MD+, MD intact; G−, without GRIK4; D−, without D2. i, Stimulus configuration for sparseness-driven uncertainty. j, Performance accuracy is modulated by cueing sparseness, and optical MD deafferentiation diminishes performance on trials with higher cueing sparseness (n = 25 sessions, 4 mice; *P = 0.0222, ***P = 1.02 × 10−4; chi-squared). k, MDGRIK4 inactivation improves performance accuracy on both high and low signal trials (n = 20 sessions from 4 GRIK4-cre mice; Wilcoxon signed-rank). l, Optical MDD2 inactivation recapitulates optical generic MD deafferentiation (n = 20 sessions from 4 D2-cre mice; Wilcoxon signed-rank). All statistical tests are two-tailed. Box plot parameters as in Fig. 1. Data are mean ± s.e.m. for h, j.
