Extended Data Fig. 5: Activity, firing rate correlations, and waveform properties of different neuron subclasses.
From: Low-dimensional population dynamics in the brainstem gate REM sleep
(a) Distribution of c1 coefficient across brain regions. For each neuron, c1 quantifies the strength and direction (positive or negative) of its correlation with PC1. The general distribution for c1 across all areas is shown in black. Distributions are represented as KDE plots. Solid lines, distributions for areas whose mean significantly differed from the general c1 distribution (black). Two-sided Mann-Whitney U test with Bonferroni correction, c1: DR, P = 1.04×10−22, U = 36649; PRN, P = 4.03×10−5, U = 110272; all units, n = 1160; DR, n = 134; PRN, n = 155 units. (b) Average firing rates of c1-c2-, c1+c2+ , c1+c2-, and c1-c2+ neurons for each brain state. c1-c2-: RM ANOVA, P = 3.38×10−114, F(2, 472) = 677.04; two-sided paired t-tests with Bonferroni correction, N vs R: P = 3.90×10−39, T(236) = 16.03; N vs W: P = 5.49×10−73, T(236) = -26.77; R vs W: P = 2.13×10−83, T(236) = -30.51; c1+c2+ : RM ANOVA, P = 2.15×10−108, F(2, 568) = 753.54; two-sided paired t-tests, N vs R: P = 4.28×10−71, T(284) = -24.37; N vs W: P = 7.43×10−34, T(284) = 14.030; R vs W: P = 1.44×10−97, T(284) = 32.62; c1+c2-: RM ANOVA, P = 2.68×10−145, F(2, 894) = 997.41; two-sided paired t-tests, N vs R: P = 5.86×10−161, T(447) = -43.14; N vs W: P = 3.58×10−62, T(447) = -19.70; R vs W: P = 2.04×10−84, T(447) = 24.55; c1-c2+ : RM ANOVA, P = 7.27×10−15, F(2,120) = 69.36; two-sided paired t-tests, N vs R: P = 2.31×10−16, T(60) = 11.52; N vs W: P = 1.07×10−6, T(60) = 5.72; R vs W: P = 1.46×10−7, T(60) = -6.24; c1-c2-, n = 237; c1+c2+, n = 285; c1+c2-, n = 448; c1-c2+, n = 61 units. Dots, individual units; error bars, 95% CIs. R, REM; N, NREM; W - Wake. (c) Activity of c1-c2-, c1+c2+, c1+c2-, and c1-c2+ neurons at NREM→REM and REM→Wake transitions. Time point tsig from when on the activity consistently differed from baseline (two-sided t-tests with Bonferroni correction; Methods), c1-c2-: tsig = -55 s, P < 0.00047; c1+c2+: tsig = -45 s, P < 0.0047; c1+c2-: tsig = -5 s, P < 0.000049; c1-c2-, n = 237; c1+c2+, n = 285; c1+c2-, n = 448; c1-c2+ , n = 61 units. Lines, mean across units; shadings, 95% CIs. See Supplementary Table 1 for full statistical details. (d) Cross-correlation between c1-c2- and c1+c2+ neurons during NREM sleep (Methods). Shadings, 95% CIs. Inset, histogram of the time lags in the cross-correlograms of negatively correlated neurons. The average time lag was -1.22 s ± 0.08 s, indicating that the activity of c1-c2- neurons leads that of c1+c2+ neurons; two-sided one sample t-test; P = 3.56×10−52, T(7878) = -15.31; n = 7879 pairs from 6 mice. (e) Peak-to-valley interval and half-width of the spike waveforms for the different neuron subclasses. Peak-to-valley: Kruskal-Wallis test; P = 3.80×10−8, H(3) = 37.39; two-sided Mann-Whitney U test (Bonferroni-corrected), c1+c2- vs c1-c2-: P = 9.37×10−8, U = 39156.5; c1+c2+ vs c1-c2-: P = 5.94×10−7, U = 24629.5; c1-c2+ vs c1-c2-: P = 0.12, U = 5837.5; half-width: Kruskal-Wallis test; P = 5.42×10−3, H(3) = 12.67; two-sided Mann-Whitney U test (Bonferroni-corrected), c1+c2- vs c1-c2-: P = 7.46×10−3, U = 45134.0; c1+c2+ vs c1-c2-: P = 0.01, U = 28390.5; c1+c2- vs c1-c2-: P = 0.0075, U = 45134; c1-c2-, n = 237; c1+c2+ , n = 285; c1+c2-, n = 448; c1-c2+, n = 61 units. *P < 0.05, **P < 0.01, ***P < 0.001. Horizontal lines, mean; error bars, 95% CIs. See Supplementary Table 1 for full statistical details.
