Supplementary Figure 3: Granger causal (GC) relationships (1–25 Hz) across sessions for context exploration and object sampling.
From: Bidirectional prefrontal-hippocampal interactions support context-guided memory
GC was greater from HPC to mPFC during (a) context exploration (Two-way ANOVA, F = 1033.46, df = 1, p = 9.83e-215), while mPFC better predicted HPC activity during instances of (b) object sampling (Two-way ANOVA, F = 2571.27, df = 1, p < 0.0001). Furthermore, maximal GC was within the theta range for both context exploration (9.16 +/− 0.83 Hz) and object sampling periods (9.27 +/− 0.92 Hz), and the causal flow reversed within the HPC-PFC circuit at this frequency as a function of behavioral phase, such that HPC directed mPFC activity during context exploration (GC = 0.3983 +/− 0.07, Wilcoxon rank-sum, z42 = 4.60, p = 4.15e-6) and mPFC engaged HPC during object sampling (GC = 0.3518 +/− 0.04, Wilcoxon rank-sum, z42 = 4.96, p = 7.14e-8). Bayesian Information Criterion (BIC) indicated the best fitting model order was at 40 (+/− 2.60) ms and 26 (+/− 2.50) ms lags for context versus object sampling periods respectively. The HPC lead over mPFC observed during contextual exploration is consistent with previously demonstrated Granger causal predictions of HPC-mPFC functional connectivity13. Shaded plots represent S.E.M. across sessions. Inset bar graph: Comparison between GC directionality for frequency of peak GC (mean +/− standard error).
