Extended Data Fig. 6: Pontine events characteristics in the ponto-geniculo-hippocampal recordings.

a, Pontine peri-event LFP activity for p-waves (uncoupled, black trace), ripple-coupled (red trace) and theta-coupled (blue trace) PGO waves: averaged time courses (top) and spectrogram averages (bottom). Shaded areas indicate SEM. b, Analogous to panel a, but for LGN activity. c, Peri-event BLP signals averages in the 4–10 Hz LFP range in PBn, LGN and hippocampus. Whereas PGO waves occur in all structures, P-waves display significant power increases only in the pons (P < 0.01, permutation t-test for PBn responses; P > 0.05, permutation t-test for LGN and hippocampal responses). Shaded areas indicate SEM. d, Consistency score for the detection of pontine waves (P-waves and PGO waves) using the recording pairs PBn-LGN and PBn-HC (Methods). Statistics are computed across experiments of the three-structure experimental session. e, Exemplary LFP traces of ongoing LGN, neocortical, that is, primary VCX and HC activity. The dashed line encloses a PGO event. f, Expanded version of the event highlighted in e. g, Statistics of the phase shift (left) and time lag (right) between LGN and visual cortex during PGO events across experimental sessions. Individual dots (right subpanel) correspond to each of the 7 experimental sessions where the triad LGN-VCX-HC were recorded, animals A13, D11 and i09. h, Putative causal interaction between LGN and VCX assessed using transfer entropy (TE) (P > 0.2890 sign test; N = 7 experimental sessions, animals A13, D11 and i09). i, Point-process conditional intensity functions between these events and detected PGO waves in LGN-VCX-HC sessions (asterisks indicate significant coupling, P < 0.01 permutation test against an H₀ distribution of random point proceses of the same rate; 7 experimental sessions; animals A13, D11 and i09). For the visual cortex data, we detected candidate PGO waves as peaks of low-frequency (2–15 Hz) LFP power exceeding an event-related threshold proportional to the standard deviation of the signal (z-scores) in LGN. Theta (2-15 Hz), gamma (25-75 Hz) and ripples (90-190 Hz) were detected independently from the hippocampal LFP traces (N = 7 experimental sessions; animals A13, D11 and I09). Point-process cross-correlation analysis revealed that PGO-associated potentials in LGN significantly co-occur with hippocampal theta and SWR, but not gamma episodes (Panel i). Analogous to our analysis of PBn-LGN PGO waves, albeit more variable, we found a phase lag of visual cortex activity with respect to LGN activity in the PGO frequency band (2-15 Hz) (Panel g). This phase lag was consistent across all sessions and animals where LGN and visual cortex activities were recorded (mean phase lag with 95% circular confidence interval 8.76 ± 8.44 degrees; P < 10⁻² circular test for the significance of the median; N = 7 experimental sessions, animals A13, D11 and I09). The corresponding time delay between LGN and VCX found in this analysis is consistent with previous studies in cats²⁶. We found difficult to detect a putative causal influence of the phase of LGN on the phase of visual cortex during peri-PGO potentials, as hinted by the non-significant differences in the LGN-to-VCX interaction direction as compared to the opposite VCX-to-LGN direction (Panel h; P = 0.2890, sign test for the comparison between TE values associated with the LGN-to-VCX direction and the VCX-to-LGN direction; N = 7 experimental sessions, animals A13, D11 and I09). The results of this causal inference analysis for PBn-LGN and LGN-VCX experimental sessions, were found consistent with the spectral independence-based causal inference analysis introduced in the next subsection of the manuscript in the context of PBn-hippocampus PGO event-related interactions (Supplementary Methods). These results indicated that causal inference might be difficult to detect in LGN-VCX recording pairs due to the highly recurrent connectivity of the macaque visual cortex. In addition, our results indicate that the detected phasic events in our recordings are indeed potentials that co-occur in the brainstem, thalamic nuclei, and neocortex, thus corresponding to PGO waves¹¹.