Evolutionarily conserved fMRI network dynamics in the mouse, macaque, and human brain

Abstract

Evolutionarily relevant networks have been previously described in several mammalian species using time-averaged analyses of fMRI time-series. However, fMRI network activity is highly dynamic and continually evolves over timescales of seconds. Whether the dynamic organization of resting-state fMRI network activity is conserved across mammalian species remains unclear. Using frame-wise clustering of fMRI time-series, we find that intrinsic fMRI network dynamics in awake male macaques and humans is characterized by recurrent transitions between a set of 4 dominant, neuroanatomically homologous fMRI coactivation modes (C-modes), three of which are also plausibly represented in the male rodent brain. Importantly, in all species C-modes exhibit species-invariant dynamic features, including preferred occurrence at specific phases of fMRI global signal fluctuations, and a state transition structure compatible with infraslow coupled oscillator dynamics. Moreover, dominant C-mode occurrence reconstitutes the static organization of the fMRI connectome in all species, and is predictive of ranking of corresponding fMRI connectivity gradients. These results reveal a set of species-invariant principles underlying the dynamic organization of fMRI networks in mammalian species, and offer novel opportunities to relate fMRI network findings across the phylogenetic tree.

Fig. 1. C-mode identification and matching across species.

A CAPs represent transients of infraslow fMRI activity that can be matched in pairs exhibiting opposite coactivation (CAPs, anti-CAPs, left). Detection of CAP anti-CAP pairs (middle) allows for the computation of fMRI Coactivation modes (right). B Detection of CAP and anti-CAP pairs given the highest spatial anticorrelation in the between-CAP similarity matrix. C-modes are built by taking the highest occurring CAP from each pair, and spatially averaging it with its corresponding inverted anti-CAP. C Evolutionarily relevant fMRI networks used for matching. D Vectorized network coactivation profiles for each C-mode (spatially z-scored), extracted from the mean fMRI values of voxels within a network mask. Arrows denote the matching of C-modes across species performed by the Hungarian Algorithm. E Correlation between matched C-modes from humans (HNU dataset) and macaques, F between humans (MSC dataset) and macaques; and G between macaques and mice. aDMN anterior default mode, pDMN posterior default mode, SMN somatomotor, VIS visual, DAT dorsal attention, VAT ventral attention, FPN frontoparietal, LIMB limbic, TH thalamus, HC hippocampus. Source data are provided as a Source Data file.

Fig. 2. C-mode topography in awake humans, macaques, and mice.

A Z-scored C-mode maps (Left) and corresponding normalized network coactivation profiles (Right, mean ± SD of voxels within the network mask). B C-mode occurrence rates (mean ± SEM), and between C-mode comparisons (Kruskal–Wallis test, FDR corrected, n = 30 subjects with 10 sessions, n = 8 animals with 2 sessions, and n = 44 animals with 1 session for humans, macaques, and mice respectively). In humans, data from male and female subjects are labeled separately. aDMN anterior default mode, pDMN posterior default mode, SMN somatomotor, VIS visual, DAT dorsal attention, VAT ventral attention, FPN frontoparietal, LIMB limbic, TH thalamus, HC hippocampus, BF basal forebrain, Cd/Pu caudate/putamen. P values: ns, p > 0.05, ****p < 0.0001. Source data are provided as a Source Data file.

Fig. 3. Infraslow dynamics and formation of C-modes.

Group-level power spectral density (blue, mean ± SEM) of C-Mode to fMRI frame correlation time-series. Red insets denote the mean ± SEM correlation values time-locked to peaks in the C-Mode time-series. Black traces show the spectrum of randomly shuffled surrogate C-mode time-courses. Source data are provided as a Source Data file.

Fig. 4. C-mode occurrence within fMRI global signal infraslow cycles.

A Group-level power spectral density (mean ± SEM) of the GS. B Distribution of GS-phases at the occurrence of each C-mode. Blue and red distributions correspond to GS-phases sampled from the positive and negative C-mode time-courses, respectively. All distributions significantly deviate from circular uniformity (Rayleigh test, one-sided, human, n = 30 subjects with ten sessions p < 0.0001 all C-modes, macaque n = 8 animals with two sessions p < 0.02 all C-modes; mouse, n = 44 animals with one session, p < 0.03 all C-modes; all p values are FDR corrected). Source data are provided as a Source Data file.

Fig. 5. Temporal trajectories of C-modes converge to the most recurring state.

A Persistence (top row) and transition (off-diagonal) probability in humans (left), macaques (middle), and mice (right). Black crosses denote the destination C-mode of preferred directional transitions (Pij > Pji). B Entropy of Markov trajectories (HMT) shows that the C-mode with the highest accessibility are C-modes with higher occurrence rates (i.e., C-mode 1 in humans and C-mode 4 in macaques and mice). Higher entropy indicates lower accessibility of a destination C-mode (column) from a starting C-mode(row). C Quantification of the sum of Entropy of Markov Trajectories for all destinations (columns in the HMT matrices at the single-subject level), and comparison between the means (one-way ANOVA, and Tukey test for multiple comparisons, n = 30 subjects with ten sessions, n = 8 animals with two sessions, and n = 44 animals with one session for humans, macaques, and mice, respectively). The most occurring C-modes (cf. Fig. 2) are also the most accessible ones. Student T-test, two-tailed, FDR corrected, *****p < 0.0001. In humans, data from male and female subjects are labeled separately. Source data are provided as a Source Data file.

Fig. 6. C-modes occurrence rate predicts ranking of functional connectivity gradients.

A Spatial correlation (r) between the principal gradients and each C-mode map. Black dots denote the spatial matching between maps obtained using the Hungarian Algorithm. B Scatter plot of the variance explained by each gradient (lambda) versus the corresponding C-mode occurrence rate. R-square from a linear fit (Pearson’s R, student T-test p < 0.01 uncorrected, two-sided, n = 30 subjects with ten sessions, n = 8 animals with two sessions, and n = 44 animals with one session for humans, macaques, and mice, respectively). The most occurring C-modes account for the most variance in the corresponding gradient axis. Source data are provided as a Source Data file.