Realistic coupling enables flexible macroscopic traveling waves in the mouse cortex
- Department of Mathematics, University of Michigan, United States
- Gilbert S. Omenn Department of Computational Medicine and Bioinformatics, University of Michigan, United States
Figures
Figure 1 with 1 supplement
Construction of the cortical connectivity.
(a) Visualization of all neurons in the hemispherical dataset we use to build the cortical connectivity. The hemisphere contains about 1 million neurons in total, including about 300,000 cortical neurons. (b) Algorithm of building the connectome: (i) First, we identify neuronal excitatory/inhibitory information based on the spatial transcriptomic data. (ii) We then identify the voxelized position for all the neurons. (iii) Based on the voxelized projection data, we calculate the number of connections between every pair of voxels (see Use of the voxelized connectivity data). (iv) Connections are created randomly between the neurons in each pair of voxels. (v) Repeat for all voxels in regions of interest. (vi) Complete the connectivity. (c) The connectivity matrix shows the number of connections between different cortical regions (in log 10 scale). A full version of the connectivity matrix with annotated regions is shown in Figure 1—figure supplement 1. (d) Visualization of outgoing connections from sampled neurons in the primary visual cortex (VISp, left), primary auditory cortex (AUDp, middle), and the ventral anterior cingulate areas (ACAv, right).
Figure 1—figure supplement 1
Connectivity matrix with detail region legend and degree distribution.
(a) The same connectivity matrix shown in Figure 1c, but with detailed region legend. (b) The out and in-degree distribution of all neurons in scale.
Figure 2 with 2 supplements
Macroscopic traveling waves emerge in response to random layer-4 stimulation through Allen connectivity.
(a) Stimulation protocol: a 10-Hz random Poisson spike train delivers a fixed-amplitude voltage bump to all layer-4 excitatory neurons; no other external input is applied. (b) Two rows of five spatial snapshots of the local-mean (KNN = 300) intracellular voltage, sampled every 5 ms across the window highlighted in (c). A coherent wavefront propagates along the anterior-to-posterior axis. (c) Global mean intracellular voltage of all neurons (dark blue) overlaid on per-region mean voltages (light traces). The red dashed lines mark the snapshot window visualized in (b). (d) Region-sorted raster plot of ~40,000 neurons sampled evenly from major cortical regions, anterior-to-posterior. See Videos for the movie of the simulation.
Figure 2—figure supplement 1
Single-neuron-resolution view of the simulation in Figure 2.
(a) Region-mean voltage traces (offset for clarity), one trace per major cortical region in anterior-to-posterior order. (b) Two rows of five voltage snapshots taken every 5 ms over the window highlighted in (a), showing the wavefront propagation at single-neuron rather than KNN-averaged resolution.
Figure 2—video 1
Macroscopic wave activity emerging in the Allen-connectivity Poisson simulation of Figure 2.
Two paired three-dimensional renderings of the same simulation are animated side-by-side over a 500-ms window of the recorded activity (1 ms cadence, 30 fps playback). Left: 300-nearest-neighbor local-mean intracellular voltage (the smoothed view of Figure 2b). Right: raw single-neuron intracellular voltage with no spatial averaging (the single-cell-resolution view of Figure 2—figure supplement 1), showing how the macroscopic wavefront emerges from the underlying spiking activity. Both panels use the turbo colormap. The trace at the bottom is the global mean voltage of all ~300,000 cortical neurons over the full 1-s recording, with the red dashed band marking the 50 ms snapshot window highlighted in Figure 2b and a vertical cursor tracking the current animation frame. Simulation parameters match Figure 2: Allen connectivity, nS, mV, 10 Hz Poisson spike train delivered to all layer-4 excitatory neurons.
Figure 3 with 1 supplement
Top: schematics of Allen (a), local (b), and uniform connectivity (c); bottom: visualization of outgoing connections from the same 50 randomly selected neurons in the primary visual cortex (VISp) for the three connectivity.
Figure 3—figure supplement 1
Comparison of the three connectivity matrices: (a) Allen, (b) local, and (c) uniform connectivity.
For local connectivity, the diagonal terms represent the connections within the same region. And the off-diagonal terms exist because these regions are physically adjacent to each other.
Figure 4 with 7 supplements
Allen connectivity produces a higher level of macroscopic wave activity than local and uniform connectivity.
(a) Theta-band (4–8 Hz) phase snapshots from three representative simulations using Allen (top), local (middle), and uniform (bottom) connectivity under matched Poisson stimulation, sampled at −10, −5, 0, +5, and +10 ms relative to the peak of phase gradient directionality (PGD). Colors represent the theta-band generalized phase of each neuron, normalized between 0 and 1. (b) Theta-band PGD of the three simulations versus time, aligned to the PGD peak. (c) Per-band comparison of the maximum PGD of the three representative simulations shown in (a), one bar per band per connectivity. Please see Videos for animations and Quantitative measurement of neuronal activity (illustrated in Figure 4—figure supplement 1) for the calculation of PGD.
Figure 4—figure supplement 1
Illustration of the three-dimensional phase-gradient pipeline used to quantify macroscopic wave activity (Quantitative measurement of neuronal activity), shown for a single representative snapshot.
(a) The per-neuron intracellular voltage is linearly interpolated onto the regular CCFv3 Cartesian grid (color: membrane voltage in mV). (b) After band-pass filtering and the Hilbert transform, every grid point carries a generalized phase (color: phase from to ). (c) The spatial gradient of the phase field is computed at each grid point (white arrows); the phase gradient directionality (PGD) summarizes how well aligned these gradient vectors are and is our scalar measure of macroscopic traveling-wave activity.
Figure 4—figure supplement 2
delta-band (0.5–4 Hz) version of Figure 4 (which shows the theta band).
(a) 3 × 5 spatial snapshots of the delta-band generalized phase (rows: Allen/local/uniform; columns: five frames centered on the Allen-phase gradient directionality [PGD] peak); phase is computed on the voltage signal with 300-NN spatial smoothing in the complex (Hilbert) domain, with no Cartesian-grid interpolation. (b) Delta-band PGD(t) over a 100-ms window centered on the Allen peak, all three connectivities. (c) Per-band max PGD bars for the same simulation (identical across the four band figures, included to match the main-text Figure 4 layout).
Figure 4—figure supplement 3
Alpha-band (8–12 Hz) version of Figure 4; layout and conventions as in Figure 4—figure supplement 2.
Figure 4—figure supplement 4
Beta-band (12–30 Hz) version of Figure 4; layout and conventions as in Figure 4—figure supplement 2.
Figure 4—figure supplement 5
Gamma-band (30–100 Hz) version of Figure 4; layout and conventions as in Figure 4—figure supplement 2.
Figure 4—figure supplement 6
Per-band phase gradient directionality (PGD) heatmaps and Allen advantage across the Poisson sweep.
(a) 3 × 5 grid of max-PGD heatmaps over the Poisson sweep; rows are the three connectivities (Allen/local/uniform), columns the five frequency bands (delta, theta, alpha, beta, and gamma). Every heatmap shares the same axes ( on , on ) and the turbo color scale on the right (Max PGD). (b) Per-band Allen-minus-opponent mean PGD gap (mean ± SEM across the 10 × 12 grid), with grouped bars for Allen–Local (orange) and Allen–Uniform (blue) and two-sided significance stars (* ; **; ***); panel (b) occupies the bottom-right slot of the 3 × 5 grid.
Figure 4—video 1
Theta-band traveling waves under Allen, local, and uniform connectivity, corresponding to the three simulations of Figure 4.
The three brain panels show the theta-band (4–8 Hz) generalized phase at single-neuron resolution for three matched simulations that share the same Poisson stimulation protocol but differ only in connectivity: Allen (left), local (center), and uniform (right).
Figure 5 with 2 supplements
Coupling strength, connectivity, and network synchrony jointly shape macroscopic-wave activity.
(a) Schematic of the three dynamic regimes that emerge across the parameter plane for Allen connectivity. At weak-to-intermediate coupling, the network supports coherent macroscopic traveling waves; at medium coupling, increased synaptic drive pushes the network into a disordered (asynchronous irregular) state where wave structure collapses; at strong coupling, the network reorganizes into a globally synchronous regime that again supports wave propagation. Insets show representative phase snapshots from each regime. (b) Per-band max PGD (mean ± SEM across the Poisson grid), plus the across-bands aggregate (‘All’), for Allen, local, and uniform connectivity, with significance stars; full per-band heatmaps in Figure 4—figure supplement 6. (c) Max PGD (alpha band) versus stimulus magnitude, error bars across . (d) Max PGD (alpha band) versus excitatory conductance , error bars across . (e) Kuramoto order parameter versus in the alpha band. (f) Synchrony-PGD scatter, one point per combination, with Pearson per connectivity (Allen , local , uniform in alpha; all bands positive and significant at ). Per-band versions of (c)–(e) are shown in Figure 5—figure supplement 1, and per-band synchrony-PGD scatters in Figure 5—figure supplement 2.
Figure 5—figure supplement 1
Per-band versions of Figure 5c–e.
A 3 × 5 grid combining (a) max PGD versus (averaged over the sweep; supplement to Figure 5c); (b) max PGD versus (averaged over the sweep; supplement to Figure 5d); and (c) maximum Kuramoto synchrony versus (averaged over the sweep; supplement to Figure 5e). Columns are the five frequency bands (delta, theta, alpha, beta, and gamma); error bars are SEM across the orthogonal sweep parameter. The peak–dip–recovery coupling profile of Figure 5d and the matching synchrony dip are visible in every band for Allen and uniform connectivity; local connectivity decays monotonically across all bands.
Figure 5—figure supplement 2
Synchrony-phase gradient directionality (PGD) scatter across all bands.
One point per combination on the 10 × 12 sweep grid, with max Kuramoto synchrony and max PGD, colored by connectivity (Allen/local/uniform). Per-connectivity Pearson values are shown in each panel’s legend. Synchrony and PGD are positively correlated in every band and every connectivity (all ). This extends Figure 5f from the alpha band to all five canonical bands.
Figure 6 with 1 supplement
Macroscopic waves are maintained by the local field potential (LFP) estimate.
(a) Paired snapshots of the local-average voltage (top, clipped to [−65,−50] mV) and the synaptic-LFP estimate (bottom, -scored) across five consecutive 5 ms frames in a representative Allen Poisson simulation. The same propagating wavefront sweeps across the cortex in both modalities. (b) Voltage versus LFP max PGD scatter for Allen connectivity, one point per ; Pearson per band, all . (c) Mean max PGD per band, voltage versus LFP, for Allen connectivity (paired for every band; LFP carries ≈2–3 × higher PGD than the intracellular voltage). (d) LFP-based mean max PGD per band, three connectivities; Allen wins in every band against both local and uniform. (e) LFP max PGD versus for Allen connectivity, one line per band; the same three-regime profile (peak, dip, recovery) seen in voltage is preserved in LFP. See LFP estimation from intracellular voltage for the LFP estimation pipeline.
Figure 6—video 1
The macroscopic wavefront is preserved in the local field potential (LFP) estimate.
Two paired renderings of the same Allen Poisson simulation are animated side-by-side across a 200-ms window of the recorded activity (1 ms cadence, 30 fps playback). Left: 300-nearest-neighbor local-mean intracellular voltage on the turbo colormap with mV. Right: 300-nearest-neighbor local-mean post hoc-reconstructed synaptic-LFP estimate on the diverging RdBu colormap, -scored across the rendered window.
Tables
Appendix 1—table 1
Parameters used in the simulations.
| Name | Notation | Excitatory | Inhibitory | Unit |
|---|---|---|---|---|
| Membrane capacitance | 1 | 1 | μF/cm2 | |
| Area of neuron | 0.29 | 0.22 | mm2 | |
| Leak reversal potential | –70 | –70 | mV | |
| Sodium reversal potential | 50 | 50 | mV | |
| Potassium reversal potential | –90 | –90 | mV | |
| AMPA receptor reversal potential | 0 | 0 | mV | |
| GABA receptor reversal potential | –70 | –70 | mV | |
| Firing threshold | –61.5 | –61.5 | mV | |
| Rising time of AMPA receptor | 0.94 | 0.94 | ms | |
| Decaying time of AMPA receptor | 0.18 | 0.18 | ms | |
| Rising time of GABA receptor | 10.0 | 10.0 | ms | |
| Decaying time of GABA receptor | 0.25 | 0.25 | ms | |
| Conductance of the leak channels | 0.02 | 0.08 | mS/cm2 | |
| Conductance of Na+ channels | 50 | 46 | mS/cm2 | |
| Conductance of K+ channels | 4.8 | 5.1 | mS/cm2 | |
| Conductance of slow K+ channels | 0.13 | 0 | mS/cm2 | |
| Time constant for | 1123 | 824.5 | ms | |
| AMPA receptor conductance | 15–50 | 15–50 | nS | |
| GABA receptor conductance | 1.5–5 | 1.5–5 | nS | |
| Poisson stimulus voltage bump to layer-4 neurons | 5–32.5 | 0 | mV | |
| Poisson stimulation rate to layer-4 neurons | 10 | 0 | Hz | |
| Firing threshold for synaptic output | –20 | –20 | mV | |
| Simulation timestep | 0.02 | 0.02 | ms |
Additional files
Download links
A two-part list of links to download the article, or parts of the article, in various formats.
Downloads (link to download the article as PDF)
Open citations (links to open the citations from this article in various online reference manager services)
Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)
Realistic coupling enables flexible macroscopic traveling waves in the mouse cortex
eLife 14:RP108208.
https://doi.org/10.7554/eLife.108208.3
