Cortical excitability inversely modulates fMRI connectivity via low-frequency neuronal coupling

This study demonstrates that regional cortical excitability inversely modulates large-scale fMRI connectivity through low-frequency (<4 Hz) neuronal coupling, revealing that increased excitability leads to fMRI hypoconnectivity while suppressed excitability causes hyperconnectivity despite opposing changes in local firing rates.

The Gist

The Big Picture: The "Volume Knob" of the Brain

Imagine your brain is a massive, bustling city with millions of people (neurons) talking to each other. Sometimes, these people talk in loud, fast bursts (like a busy market); other times, they hum a slow, quiet tune together.

Scientists have long used a special camera called fMRI to take "photos" of this city. This camera doesn't see the people talking; it sees the traffic (blood flow) that happens when they talk. When two neighborhoods in the city show up on the camera moving in perfect sync, we say they have "functional connectivity." It's like seeing two towns dancing to the same song.

But here's the mystery: What actually makes those towns dance together? Is it because the people are shouting louder? Or is it because they are humming a slow, quiet tune?

This paper answers that question by turning the "volume knobs" on specific parts of the brain in mice and watching what happens to the dance.

---

The Experiment: Turning the Volume Up and Down

The researchers used a clever "remote control" (chemogenetics) to change how active the neurons in the mouse's prefrontal cortex (the brain's "CEO") were. They did three things:

1. Turned the volume UP (Excitation): They made the neurons fire faster and louder.
   • The Analogy: Imagine a town square where everyone suddenly starts shouting, running, and panicking. It's chaotic and loud.
2. Turned the volume DOWN (Silencing): They made the neurons fire slower and quieter.
   • The Analogy: Imagine the town square suddenly going silent. Everyone stops moving and sits quietly.
3. Turned the volume UP by removing the "Cops" (Inhibition): They stopped the neurons that usually calm things down (inhibitory neurons).
   • The Analogy: Imagine the police officers (who usually keep the crowd calm) go on break. The crowd gets wild and loud, even though no one was told to shout.

The Surprise Finding: The Inverse Relationship

The scientists expected that if a neighborhood got "louder" (more active), it would connect better with the rest of the city. They thought: More activity = More connection.

They were wrong.

• When they made the brain LOUDER (Excitation or Inhibition removal): The brain's "dance" with other parts of the city stopped. The fMRI showed less connection (hypoconnectivity).
  • The Metaphor: When the town square gets too chaotic and loud, the people are so focused on their own shouting that they can't hear the neighbors anymore. The connection to the rest of the city breaks.
• When they made the brain QUIETER (Silencing): The brain's "dance" with other parts of the city got stronger. The fMRI showed more connection (hyperconnectivity).
  • The Metaphor: When the town square goes quiet, the people stop shouting and start listening. They can finally hear the slow, rhythmic humming of the rest of the city, and they start dancing in sync with them.

The Rule: More local noise = Less global connection. More local quiet = More global connection.

The Secret Ingredient: The Slow Hum

Why does this happen? The researchers listened to the "music" of the brain using electrodes to find the answer. They discovered that the brain has two types of music:

1. Fast, Loud Music (High Frequencies): This is the shouting and running. It happens when the brain is very active locally.
2. Slow, Deep Hum (Low Frequencies < 4Hz): This is the slow, rhythmic breathing of the city.

The Discovery: The fMRI camera (which takes slow "photos" of blood flow) is blind to the fast, loud music. It only sees the slow, deep hum.

• When the brain gets "loud" (high excitability), the fast music drowns out the slow hum. The slow hum stops, so the fMRI camera sees no connection.
• When the brain gets "quiet" (low excitability), the fast music stops, and the slow hum becomes clear and strong. The fMRI camera sees strong connection.

The "Two-Pillar" Framework

The authors propose a simple way to understand this, like a two-legged stool:

1. Pillar 1: The Inverse Rule. If a brain region is hyperactive (noisy), it disconnects from the rest of the brain. If it is quiet, it connects deeply.
2. Pillar 2: The Slow Hum. The "glue" that holds the brain's big networks together is not the fast shouting, but the slow, low-frequency humming (under 4 Hz).

Why Does This Matter?

This is a huge deal for understanding brain diseases like Autism and Schizophrenia.

• The Old View: We thought these diseases were just "broken connections."
• The New View: Maybe these diseases are caused by the brain being too loud in some areas. If a part of the brain is too excitable (too noisy), it stops "listening" to the rest of the brain, causing the connection to break.

It's like trying to have a conversation at a rock concert. If the band (the local brain area) is playing too loud, you can't hear your friend (the rest of the brain), and the connection is lost. You don't need to fix the friend; you just need to turn down the volume of the band so the conversation can happen again.

Summary

• Loud brain = Disconnected.
• Quiet brain = Connected.
• The connection relies on a slow, deep hum, not the fast shouting.

This paper gives us a new map for understanding how the brain talks to itself and why things go wrong when the volume gets too high.

Technical Summary

1. Problem Statement

The neural mechanisms underlying spontaneous resting-state fMRI connectivity (functional connectivity, FC) remain poorly understood. While stimulus-evoked fMRI activity is linked to high-frequency ($\gamma$-band) neuronal activity, the relationship between spontaneous fMRI connectivity and underlying electrophysiology is debated. Previous studies have associated FC with various frequency bands ($\gamma$, $\delta$, slow oscillations), but a unified causal framework is lacking.

A critical gap exists in understanding how local cortical excitability (the balance of excitation and inhibition, or E/I) shapes large-scale network synchronization. Chemogenetic (DREADD) studies have shown that manipulating local excitability can alter fMRI connectivity, but results have been heterogeneous and often contradictory. Furthermore, the specific electrophysiological coupling (coherence) that drives these fMRI changes has not been causally disentangled from local firing rates.

2. Methodology

The authors employed a multimodal approach combining chemogenetics, multielectrode electrophysiology, and resting-state fMRI in the mouse medial prefrontal cortex (PFC). They aggregated novel experimental data with previously published datasets to create a unified analytical framework testing bidirectional manipulations of cortical excitability.

Experimental Manipulations:

1. $\uparrow$Excitation: Chemogenetic activation of pyramidal neurons (CamkII$\alpha$-hM3Dq) to increase intrinsic excitability and firing rates.
2. $\downarrow$Inhibition: Chemogenetic inhibition of Parvalbumin (PV+) interneurons (PV::Cre-hM4Di) to disinhibit the circuit, thereby increasing net firing rates.
3. Silencing ($\downarrow$Excitability): Chemogenetic suppression of pan-neuronal activity (hSyn-hM4Di) to decrease firing rates.

Data Acquisition & Analysis:

• fMRI: Resting-state scans were performed under both sedated (medetomidine-isoflurane) and awake (head-fixed) conditions. Seed-based connectivity analysis focused on the PFC and its Default Mode Network (DMN) targets (Retrosplenial cortex, Thalamus, Striatum).
• Electrophysiology: Simultaneous Local Field Potential (LFP) and Multi-Unit Activity (MUA) recordings were obtained from the PFC and a projection target (Retrosplenial cortex, Rs).
  • MUA: Used as an operational readout of local circuit-level excitability (population firing rate).
  • LFP: Used to calculate power spectral density and magnitude-squared coherence between regions across frequency bands (Slow <1Hz, $\delta$, $\theta$, $\alpha$, $\beta$, $\gamma$).
• Biophysical Modeling: A minimal three-node spiking network model (PFC $\to$ Rs, plus a third independent input to Rs) was used to simulate the effects of DREADD manipulations on firing rates and interareal coherence.

3. Key Contributions

• Unified Framework: The study aggregates distinct chemogenetic manipulations to demonstrate that cortical excitability is the dominant physiological axis governing fMRI connectivity, rather than the specific cell type or receptor targeted.
• Inverse Relationship: It establishes a robust, bidirectional inverse relationship between local cortical excitability and large-scale fMRI connectivity.
• Frequency Specificity: It identifies low-frequency (< 4 Hz) interareal coherence as the specific electrophysiological correlate of fMRI connectivity, distinguishing it from high-frequency coupling which does not predict fMRI changes.
• Mechanistic Model: It provides a biophysical explanation for how local firing rate changes modulate global network synchronization via shared low-frequency fluctuations.

4. Key Results

A. Electrophysiological Effects of Manipulations

• $\uparrow$Excitation & $\downarrow$Inhibition: Both manipulations resulted in increased local firing rates (MUA) and increased high-$\gamma$ power (a correlate of spiking). However, they produced distinct spectral signatures:
  • $\uparrow$Excitation suppressed rhythmic activity up to $\gamma$ but increased high-$\gamma$.
  • $\downarrow$Inhibition caused a broadband increase in power ($\delta$ to high-$\gamma$).
• Silencing: Reduced local firing rates and broadband power, with a trend toward increased slow ($<1$ Hz) power.

B. fMRI Connectivity Findings

• Inverse Modulation:
  • Increased Excitability ($\uparrow$Excitation & $\downarrow$Inhibition): Both manipulations led to fMRI hypoconnectivity (reduced functional connectivity) between the PFC and DMN nodes (Thalamus, Retrosplenial cortex, Striatum), despite increased local BOLD signal and firing rates.
  • Decreased Excitability (Silencing): Led to fMRI hyperconnectivity (increased functional connectivity) between the PFC and DMN nodes.
• Robustness: These effects were observed in both sedated and awake states and generalized to unimodal sensory cortices (Somatosensory cortex) in re-analyzed data.
• Vascular Control: Control experiments with PCP (which induces large BOLD responses without reducing connectivity) ruled out a purely vascular "ceiling effect" as the cause of hypoconnectivity.

C. Electrophysiological Correlates of fMRI

• Coherence Profiles:
  • $\uparrow$Excitation and $\downarrow$Inhibition both reduced low-frequency (< 4 Hz) coherence between PFC and Rs, despite having opposite effects on high-frequency coherence.
  • Silencing increased low-frequency coherence.
• Predictive Power: A strong correlation was found between low-frequency (< 4 Hz) interareal coherence and the magnitude of fMRI connectivity changes across all conditions. High-frequency coherence (including $\gamma$-envelope) showed no significant correlation with fMRI changes.

D. Biophysical Modeling

• The model successfully reproduced the empirical findings:
  • Increasing local excitability in the PFC node reduced low-frequency interareal coherence and increased high-frequency direct communication.
  • Decreasing excitability enhanced low-frequency coherence.
• The model suggests that fMRI connectivity reflects a shared network-state component (slow, global fluctuations) rather than direct high-frequency communication. Local excitability changes bias the circuit away from this shared slow state, thereby reducing fMRI connectivity.

5. Significance and Implications

• Reinterpreting fMRI: The findings challenge the assumption that higher local neuronal activity always correlates with stronger functional connectivity. Instead, hyperexcitable regions exhibit weaker large-scale integration in fMRI terms.
• Neurophysiological Basis of fMRI: The study provides strong evidence that fMRI connectivity is primarily supported by infraslow and slow oscillations (< 4 Hz), acting as a scaffold for large-scale synchronization.
• Clinical Relevance: This framework offers a parsimonious explanation for the heterogeneous fMRI connectivity patterns (both hyper- and hypoconnectivity) observed in neuropsychiatric disorders linked to E/I imbalance, such as autism and schizophrenia. It suggests that these disorders may involve distinct subgroups with either elevated or reduced cortical excitability.
• Developmental and Pathological Insights: The inverse relationship may explain developmental changes in connectivity (linked to inhibitory maturation) and hyperconnectivity in neurodegenerative or lesional conditions (potentially reflecting reduced firing/silencing rather than compensatory reorganization).

In summary, the paper establishes that cortical excitability acts as a master regulator of large-scale fMRI connectivity, operating through an inverse mechanism where increased local firing disrupts the low-frequency synchrony required for functional integration.