A cholinergic mechanism orchestrating task-dependent computation across the cortex

This study demonstrates that the basal forebrain cholinergic system orchestrates task-dependent cortical dynamics by spatiotemporally multiplexing diverse signals and causally driving evidence accumulation and decision-making processes during specific cognitive tasks.

The Gist

Imagine your brain is a massive, bustling city with thousands of neighborhoods (cortical regions) working together to get things done. Sometimes, the city needs to handle a simple task, like walking down a familiar street. Other times, it needs to solve a complex puzzle, like navigating a maze while counting invisible landmarks.

For a long time, scientists wondered: How does the brain switch gears so quickly? How does it reorganize its entire network to handle a simple task one moment and a complex one the next?

This paper suggests the answer lies with a tiny, invisible "traffic controller" called acetylcholine.

The Two Tasks: The "Tower" Game vs. The "Signpost" Game

To figure this out, the researchers taught mice to play two different video games in a virtual reality maze.

1. The Signpost Game (Simple): Imagine you are driving, and there is a giant, obvious signpost pointing to the exit. You just follow the sign. You don't need to think hard; you just react.
2. The Tower Game (Complex): Now, imagine the signpost is gone. Instead, little towers flash on the walls as you drive. Sometimes there are more on the left, sometimes on the right. You have to count them in your head, keep a running tally, and then decide which way to turn. This requires "accumulating evidence"—gathering bits of information over time to make a smart guess.

The mice had to switch between these two games dozens of times, often without knowing which one was coming next.

The Traffic Controller: Acetylcholine

The researchers focused on acetylcholine, a chemical messenger released by a specific group of neurons in the brain's "basal forebrain." Think of these neurons as a central dispatch center that sends out thousands of tiny fiber-optic cables (axons) to every neighborhood in the brain city.

What they found:

• It's not just an "On/Off" switch. When the mice were playing the simple "Signpost" game, the traffic controller sent a steady, low-level signal.
• It's a dynamic conductor. When the mice switched to the complex "Tower" game, the traffic controller went into overdrive. It didn't just turn the volume up; it changed the pattern of the signal.
  • It sent specific instructions to the "counting" neighborhoods.
  • It helped the brain ignore distractions.
  • Most importantly, the chemical signal itself started mimicking the math the mouse was doing. As the mouse counted more towers, the acetylcholine signal ramped up, literally encoding the "decision" in real-time.

The "Light Switch" Experiment

To prove this chemical wasn't just a bystander but the cause of the smart behavior, the researchers used a clever trick. They used light (optogenetics) to temporarily "turn off" the acetylcholine traffic controller while the mice were playing the games.

• In the Signpost Game: The mice were fine. They could still follow the obvious sign.
• In the Tower Game: The mice got confused. Without the traffic controller, they couldn't count the towers or make the right decision. They forgot how to accumulate evidence.

Furthermore, when they turned off the controller, the different neighborhoods of the brain stopped talking to each other effectively. The "city" fell into chaos, and the complex calculation fell apart.

The Big Picture: A Creative Analogy

Think of the brain's cortex as a giant orchestra.

• The musicians are the different brain regions (visual, motor, decision-making).
• The music is the behavior (running, turning, deciding).

In the past, we thought the conductor (acetylcholine) just told the orchestra to "play louder" when the animal was awake or excited.

This paper shows the conductor does something much smarter.
When the orchestra needs to play a simple march (the Signpost game), the conductor gives a simple beat. But when they need to play a complex symphony (the Tower game), the conductor doesn't just speed up the tempo. They rewrite the sheet music in real-time. They tell the violins to play a specific rhythm that matches the math of the problem, ensuring every section of the orchestra knows exactly what to do to solve the puzzle.

Why This Matters

This discovery changes how we understand learning and decision-making. It suggests that our ability to switch between simple habits and complex thinking isn't just about "trying harder." It's about a specific chemical system that reconfigures the entire brain network to support the specific type of thinking we need at that moment.

If this system breaks down (as it might in Alzheimer's or ADHD), the brain might lose its ability to "switch gears," leaving us stuck in simple habits or unable to solve complex problems, even if our basic senses and muscles are working fine.

Technical Summary

1. Problem Statement

Animals frequently switch between different cognitive tasks requiring distinct neural computations, often involving distributed activity across the cortex. While it is known that the basal forebrain cholinergic system modulates cortical activity and is involved in arousal, attention, and sensory processing, the specific circuit mechanisms by which it orchestrates task-dependent dynamics remain unclear. A critical unanswered question is whether acetylcholine (ACh) merely provides a global "contextual" signal (e.g., arousal) or if it actively encodes and drives specific cognitive computations, such as the accumulation of sensory evidence, in a task-specific manner.

2. Methodology

The authors employed a combination of behavioral paradigms, large-scale neural imaging, and causal manipulations in mice:

• Behavioral Paradigm (Task-Switching VR):

  • Mice navigated a virtual T-maze and switched unpredictably between two tasks:
    1. Towers Task (Evidence Accumulation): Mice had to count briefly flashing "towers" on the left or right walls and choose the side with the higher count. This required gradual accumulation of sensory evidence.
    2. Visually Guided Task: Mice ignored the towers and followed a permanent visual guide indicating the correct side.
  • Strategy Classification: Using Generalized Linear Models combined with Hidden Markov Models (GLM-HMM), the authors separated trials into "engaged" (strategy-driven) and "disengaged" (bias/history-driven) states to ensure analysis focused on active cognitive processing.
  • Controls: Pupil diameter (arousal) and running velocity were tracked to control for non-cognitive variables.

• Neural Imaging (Cholinergic Axons):

  • Genetics: ChAT-IRES-Cre mice were crossed with flex-jGCaMP8m mice to express the calcium indicator specifically in basal forebrain cholinergic neurons.
  • Technique: Mesoscale widefield calcium imaging was performed through a cleared skull to record activity from cholinergic axons innervating the entire dorsal cortex simultaneously (~50µm/pixel resolution).

• Causal Manipulation (Optogenetics & Pharmacology):

  • Optogenetic Silencing: In a subset of mice, cholinergic neurons expressed the red-shifted inhibitory opsin Jaws. A red LED (620 nm) was used to illuminate the dorsal cortex during the evidence region of the maze to silence cholinergic axon terminals specifically during task execution.
  • Pharmacology: Systemic administration of scopolamine (a muscarinic antagonist) was used in a separate cohort to confirm findings.
  • Simultaneous Imaging: In silencing experiments, excitatory cortical activity (GCaMP6s in Thy1-GCaMP6s mice) was imaged simultaneously to observe downstream effects on cortical dynamics.

• Analysis:

  • Decoding: Logistic and linear regression decoders were trained on cortical activity to classify task identity, behavioral strategy, choice, and evidence strength.
  • Encoding Models: L2-regularized linear models were used to quantify how much variance in cholinergic activity was explained by sensory, motor, arousal, and cognitive variables.
  • Geometric Analysis: Angles between decoding axes were computed to determine if evidence and choice signals were orthogonal or aligned.

3. Key Contributions & Results

A. Spatiotemporally Heterogeneous and Multiplexed Cholinergic Signals

• Task Dependence: Cholinergic activity was not uniform; it varied significantly across cortical regions and maze positions depending on the task.
• Multiplexing: Cholinergic axons simultaneously encoded multiple variables: task identity, behavioral strategy (engaged vs. disengaged), upcoming choice correctness, sensory evidence, motor output, and arousal.
• Evidence Accumulation Signal: Crucially, cholinergic activity was significantly higher during the evidence accumulation phase of the "Towers" task compared to the visually guided task, even when controlling for running speed and pupil diameter.

B. Cholinergic Input Encodes the Decision Variable

• Evidence-Dependent Choice Coding: In the Towers task, the projection of cholinergic activity onto choice-decoding axes ramped up linearly with the strength of sensory evidence ($\Delta$ towers). This ramping was absent in the Visually Guided task and in "disengaged" strategies.
• Discrete Updates: Single-trial analysis revealed that individual tower stimuli caused step-like increases in the choice variable, indicating that cholinergic activity behaves like a canonical evidence accumulator.
• Geometric Alignment: In the Towers task, the decoding axes for "evidence" and "choice" were non-orthogonal (aligned), meaning sensory pulses drove the decision variable. In the Visually Guided task (incongruent trials), these axes were orthogonal or opposed, confirming that cholinergic signals specifically track the computation of evidence accumulation, not just the choice itself.

C. Causal Role in Evidence Accumulation

• Selective Behavioral Impairment: Optogenetic silencing of cholinergic terminals in the cortex selectively impaired performance in the Towers task (evidence accumulation) but had no effect on the Visually Guided task.
• Reduced Evidence Weight: Silencing reduced the weight of sensory evidence in the mice's decision-making model without altering side biases or trial-history effects.
• Disruption of Cortical Dynamics:
  • Decorrelation: Normally, evidence accumulation is associated with a decorrelation of activity across cortical regions. Silencing ACh restored positive correlations, effectively "reverting" the network to a less distributed state.
  • Loss of Coding: Silencing significantly reduced the accuracy of decoding both "choice" and "evidence" from excitatory cortical activity. The linear relationship between evidence strength and choice information (the decision variable) was abolished.

4. Significance

• Beyond Contextual Modulation: This study challenges the view that acetylcholine acts solely as a global modulator of arousal or attention. Instead, it demonstrates that ACh release is computationally specific, directly encoding and driving the accumulation of sensory evidence.
• Orchestrating Distributed Networks: The basal forebrain cholinergic system acts as a flexible "orchestrator" that reconfigures large-scale cortical dynamics to support specific cognitive demands. It enables the transition from correlated, feedback-dominated states (simple tasks) to decorrelated, feedforward-dominated states (complex accumulation tasks).
• Mechanism for Flexible Cognition: The findings suggest a circuit mechanism where the brain rapidly switches between different computational modes (e.g., simple discrimination vs. evidence accumulation) by modulating the cholinergic tone in a task- and strategy-dependent manner.
• Theoretical Implications: These results necessitate revised theoretical frameworks for neuromodulation, moving from "permissive" models (where ACh simply enables processing) to "participatory" models (where ACh actively shapes the decision variable).

Conclusion

The paper establishes that the basal forebrain cholinergic system is a key node in the distributed network underlying sensory evidence accumulation. By providing spatiotemporally heterogeneous, task-specific signals that directly track the decision variable, acetylcholine orchestrates cortex-wide dynamics to enable flexible cognitive behavior.