Behavior Differentially Shapes Spontaneous Cortical Network Dynamics Across Frequencies

The brain is a massive, distributed engine. It manages multiple tasks simultaneously, from maintaining posture to processing sensory input. It must shift between states of intense focus and quiet rest. Scientists understand that these processes are supported by large-scale networks of interacting brain regions. However, a fundamental tension remains. How can the brain use the same physical architecture to support both slow, sweeping changes in global state and the lightning-fast reconfigurations required for immediate behavior?

Current neuroimaging often treats the brain as a collection of static maps. Researchers use functional MRI (fMRI) to track slow changes in blood oxygenation. They also use electrophysiology to track rapid electrical fluctuations. Both methods look for "functional connectivity" (FC), which is the statistical correlation between the activity of distant brain regions. The problem is that these signals operate on vastly different timescales. It has been unclear whether the "maps" seen in slow hemodynamic (blood-flow related) signals are simply low-resolution shadows of fast neural activity. Or, perhaps, the brain employs different dynamical principles depending on the frequency of the signal.

The mismatch between maps and motion

The prevailing assumption in neuroscience is that the spatial organization of the brain is a constant. This is the "who talks to whom" aspect of the brain. If you look at a map of resting-state networks, you expect to see the same landmarks. This should hold true whether you watch a slow wave of blood flow or a rapid burst of electrical activity. But this assumption masks a distinction between structure and dynamics.

Standard connectivity analyses often conflate these two concepts. A high correlation between two brain regions might exist because they are physically wired together. Alternatively, it might exist because they both belong to a long-lasting "state" that persists for several seconds. Previous studies have struggled to decouple these. If we only look at time-averaged connectivity, we might conclude that networks are stable. We could miss the fact that the way those networks are actually expressed changes radically depending on the frequency.

Simultaneous imaging of voltage and blood

To resolve this, the researchers developed a platform to see both sides at once. They used high-frame-rate wide-field imaging in awake mice. This allowed them to capture two distinct biological signatures simultaneously:

1. Neuronal Voltage: Using a voltage-sensitive fluorescent protein (JEDI-1P-Kv), the authors recorded membrane potential fluctuations. They reached speeds of up to 200 frames per second.
2. Hemodynamics: They tracked a red reference channel (mCherry) as a proxy for blood flow. This mirrors the slow signals seen in fMRI.

By recording these signals in the same animal, the authors avoided many confounding variables. They then decomposed the brain activity into "Co-activation Patterns" (CAPs). CAPs are low-dimensional spatial templates. They act as the "building blocks" of brain activity by representing recurring patterns of regional activation. By using k-means clustering (a method to group similar data points), the researchers treated brain activity as a sequence of discrete, recognizable states.

Stable scaffolds, divergent drivers

The core finding is a dissociation between spatial architecture and temporal execution. The authors find that "static" functional connectivity is remarkably preserved across all frequencies .

This includes everything from the slowest infraslow waves to rapid gamma oscillations. The "map" is indeed constant.

However, the dynamics—how the brain moves through its repertoire of CAPs—diverge based on frequency. The researchers report two distinct modes of behavioral modulation:

• Persistence-driven (Slow): In low-frequency and hemodynamic signals, behavior shapes the brain by changing the "dwell time" (how long a state lasts). When a mouse enters a prolonged rest or movement period, network states become more stable and persist longer .

• Selection-driven (Fast): In higher-frequency neural activity, behavior shapes the brain by changing "occupancy" (how often a state is chosen). Instead of making a state last longer, behavior causes the brain to switch between different network motifs more rapidly .

The authors quantify this by showing that while the spatial structure of CAPs remains consistent across mice and sessions, the number of transitions between states increases with frequency.

This causes the average dwell time to decrease as frequency rises .

Limitations of the frequency decomposition

This study has several technical caveats. First, the authors acknowledge that trial durations were relatively short. This was limited by the storage capacity of the high-speed imaging system. Short trials may limit the ability to capture very slow dynamics.

Second, the use of mCherry as a hemodynamic proxy is a compromise. It is not a direct measurement. It is a composite signal influenced by the absorption of oxygenated and deoxygenated hemoglobin. While the authors validated this through modeling, it lacks the pure specificity of dedicated diffuse reflectance methods.

Finally, the process of band-pass filtering (isolating specific frequency ranges) involves temporal smoothing. This might blur the edges of discrete state transitions. This could create a "continuous" appearance for what are actually sharp jumps in neural state.

The verdict: A layered dynamical system

The research suggests the brain operates as a layered dynamical system. Frequency acts as the selector for how information is integrated. The study confirms that static functional connectivity is a reliable but incomplete metric. It tells us the "geometry" of the brain's communication channels. However, it says nothing about the "traffic" flowing through them.

For researchers, this work provides a vital distinction. If you study slow hemodynamic signals, you are primarily observing the brain's tendency toward stability and state persistence. If you want to see the brain's capacity for rapid, flexible reconfiguration, you must look to faster neural oscillations. The "map" is the same, but the driving style is entirely different.

Figures from the paper