Top-down regulation of ingestive behavior fragmentation

This study identifies a top-down neural circuit projecting from the dorsal subiculum to the mammillary body that acts as a bistable switch to gate the duration of feeding bouts, thereby regulating the fragmentation of ingestive behavior independent of homeostatic state.

The Gist

The Big Picture: Why Don't We Eat Like Pac-Man?

Imagine a video game character like Pac-Man. If he were hungry, he would just eat dots continuously until he was full, never stopping to look around or check his surroundings.

But real animals (and humans) don't eat like that. Even when we are starving, we eat in bouts: we take a few bites, pause to look around, take a sip of water, check our phone, and then go back to eating. We "fragment" our eating into little chunks.

Scientists have long known that we do this, but they didn't know how our brains decide when to stop eating for a second and when to start again. This paper solves that mystery.

The Discovery: The Brain's "Pause Button"

The researchers found a specific "wiring" in the brain that acts like a top-down manager for eating. It doesn't tell you how hungry you are (that's the stomach's job); instead, it decides when to pause your meal to check the environment.

Here is the circuit they found:

1. The Boss (dSub): A part of the hippocampus called the dorsal subiculum. Think of this as the brain's "Strategic Command Center." It knows about the environment, potential dangers, and other tasks you need to do.
2. The Messenger (MB): This center sends a signal to the Mammillary Body (a small structure deep in the brain).
3. The Action: This connection (dSub → MB) is the switch that controls the microstructure of eating.

The Analogy: The "Bistable Light Switch"

To understand how this works, imagine a light switch that has two stable positions: ON (Eating) and OFF (Not Eating).

• The "Low" State (Eating): When the light is dim (low activity in this brain circuit), the animal is in "Eating Mode." It's safe to keep munching.
• The "High" State (Not Eating): When the light is bright (high activity), the animal switches to "Alert Mode." It stops eating to look around, sniff the air, or move.

The Magic of the Circuit:
The brain doesn't just flip this switch randomly. It works like a bistable system (a system that likes to stay in one of two states).

• Small Glitches = Pauses: Sometimes, a tiny bit of "noise" or a small signal bumps the switch up just a little. The light flickers, the animal stops chewing for 2 seconds, looks around, and then the switch snaps back down. Result: A short pause, but the meal continues.
• Big Jumps = Stopping: If the signal is strong enough (like seeing a predator), the switch gets pushed all the way up. The light stays bright. The animal stops eating completely and runs away or hides. Result: The meal is over.

What the Scientists Did (The Experiments)

The team used high-tech tools to prove this:

1. Watching the Brain: They recorded the activity of the "Boss" neurons while mice ate. They saw that every time the mouse started eating, the neurons went quiet (Low State). Every time the mouse stopped, the neurons fired up (High State).
2. The "Remote Control" (Optogenetics): They used light to control these neurons like a remote control.
   • Zapping the "Off" button: When they artificially turned on the signal while the mouse was eating, the mouse stopped eating immediately. The meal got chopped up into tiny, short bites.
   • Zapping the "On" button: When they artificially turned off the signal, the mouse couldn't stop eating. It ate in one giant, continuous block without pausing to look around.
3. The "Threat" Test: They showed the mice a fake predator (a looming shadow on a screen). This made the mice eat in shorter bursts. Crucially, this happened without changing the hunger signal in the stomach. The "Strategic Command" (hippocampus) just decided, "Hey, it's dangerous out there, let's pause more often."

Why This Matters

This discovery changes how we think about hunger.

• Old View: Hunger is a volume knob. The louder the hunger, the more you eat.
• New View: Hunger is just the desire to eat. But the timing of your eating is controlled by a smart manager that balances your hunger against the rest of the world.

The Takeaway:
Your brain has a specific circuit that acts like a safety manager. It constantly asks, "Is it safe to keep eating right now?" If the answer is "maybe not," it triggers a pause. This allows animals to balance the need to eat with the need to stay alive, watch for predators, and interact with the world.

In short: The hippocampus doesn't just tell you to eat; it tells you when to stop and look around.

Technical Summary

1. Problem Statement

While the neural mechanisms underlying the intensity of hunger and homeostatic feeding drives (e.g., AgRP neuron activity) are well-characterized, the mechanisms governing the microstructure of feeding behavior remain unclear. In natural environments, animals do not feed continuously until satiation; instead, feeding occurs in fragmented "bouts" separated by pauses. This fragmentation allows animals to balance internal metabolic needs with external demands (e.g., predator vigilance, social cues). The paper seeks to identify the neural circuit responsible for this moment-to-moment action selection and the "toggling" between feeding and non-feeding states, independent of the overall homeostatic drive.

2. Methodology

The authors employed a multi-modal approach combining anatomical tracing, bidirectional neural activity mapping, optogenetics, and computational modeling:

• Bidirectional Activity Mapping:
  • Used cFos (activity marker) and pPDH (inverse activity marker) to map neural activity across dorsal hippocampal subfields (DG, CA3, CA1, dSub) during fasting and re-feeding.
  • Fiber Photometry: Recorded calcium dynamics (GCaMP6m) in the dorsal subiculum (dSub) and AgRP neurons (Arc) to compare temporal dynamics relative to feeding bouts.
  • Neuropixels 2.0: Performed single-unit recordings in the dSub to characterize population firing rates and individual neuron modulation during feeding.
• Anatomical Tracing:
  • Anterograde: Used rCOMET tracer and light-sheet microscopy with Fluorescence-based Structural Tensor Analysis (STA) to map dSub efferents.
  • Retrograde: Used Cholera Toxin Subunit B (CTB) injected into target regions (Mammillary Body [MB], Anterior Thalamic Nuclei [ATN], Entorhinal Cortex [EC]) to confirm dSub projections.
• Circuit Manipulation (Optogenetics):
  • Photoinhibition: Expressed eOPN3 (inhibitory GPCR) in dSub and targeted specific terminals (MB, ATN, EC) with light to inhibit projections.
  • Photoactivation: Expressed ChR2 in dSub and targeted MB terminals to activate the pathway.
  • Controls: Included lateral hypothalamus (LH) inhibition and AgRP activation to validate technical efficacy.
• Behavioral Assays:
  • Analyzed feeding microstructure (bout duration, inter-bout intervals, latency) under various conditions: fasting, high-fat diet, liquid food, water drinking, nesting (motor control), and exposure to a looming disc (predator threat).
  • Defined "intra-bout interruptions" (brief pauses <4s) vs. "bout terminations."
• Computational Modeling:
  • Developed a stochastic Wilson-Cowan model (interacting Excitatory [E] and Inhibitory [I] populations) to simulate a bistable attractor system.
  • Modeled interruptions as stochastic renewal processes and tested hypotheses regarding how neural perturbations affect the energy threshold between "feeding" and "non-feeding" states.

3. Key Contributions & Results

A. Identification of the dSub-MB Circuit

• Activity Correlation: Unlike AgRP neurons (which decrease activity upon feeding onset), dSub activity showed a unique bimodal distribution: low activity during feeding bouts and high activity during inter-bout intervals.
• Specificity: This activity pattern was specific to ingestive bouts. It was not correlated with general locomotion, jaw movements during nesting, or investigation of novel objects/conspecifics.
• Independence from Homeostasis: The dSub-MB activity pattern tracked feeding bouts regardless of metabolic state (fasted vs. fed), food type (chow, high-fat, liquid), or caloric content. It also tracked water drinking, suggesting it encodes the act of ingestion rather than hunger intensity.

B. Causal Regulation of Feeding Microstructure

• Projection Specificity: The dSub projects primarily to the MB, ATN, and EC.
  • Inhibition: Photoinhibition of the dSub-MB pathway (but not dSub-ATN or dSub-EC) significantly increased feeding bout duration (right-shifted distribution) without changing total food intake.
  • Activation: Photoactivation of the dSub-MB pathway significantly decreased feeding bout duration.
• Mechanism: The manipulation affected bout termination, not initiation. Inhibition made it harder to stop eating; activation made it easier to stop.

C. The Bistable Attractor Mechanism

• Interruptions vs. Terminations: The authors observed that feeding bouts contain brief "intra-bout interruptions" (pauses <4s).
  • Stochastic Perturbations: Small, transient increases in dSub-MB activity during a bout often resulted in a brief pause (resettling into the "feeding" attractor).
  • Gating: Larger or sustained increases in activity pushed the system over an energy threshold, causing a full bout termination (transition to the "non-feeding" attractor).
• Model Validation: The Wilson-Cowan model successfully recapitulated these dynamics.
  • Hypothesis Testing: Optogenetic manipulations did not change the rate of interruptions (renewal process). Instead, they altered the likelihood that an interruption would escalate into a full termination.
  • Threshold Modulation: In the model, inhibitory current (mimicking eOPN3) raised the energy threshold for termination (longer bouts), while excitatory current (mimicking ChR2) lowered it (shorter bouts).

D. Naturalistic Validation (Looming Disc)

• Exposure to a looming disc (simulating a predator) shortened feeding bouts.
• Crucially, this environmental threat did not alter the baseline dSub-MB activity in the feeding state.
• Instead, it increased the frequency of interruptions (the rate of stochastic perturbations) without changing the threshold for termination. This confirmed the model's prediction that external threats can fragment behavior by increasing the rate of perturbations rather than shifting the neural state threshold.

4. Significance

• Top-Down Control: The study identifies a specific top-down circuit (dSub $\to$ MB) that translates cognitive/environmental processing into moment-to-moment action selection. It demonstrates that the hippocampus does not just regulate whether an animal eats (homeostasis) but how it structures the eating behavior in time.
• Action Selection Framework: The findings propose that the hippocampus acts as a "gate" for motivated behaviors, weighing internal drives against external contingencies to determine the stochasticity of behavioral transitions.
• Bistability in Behavior: The work provides a concrete neural substrate for a bistable attractor system in naturalistic behavior, showing how stochastic neural noise can drive transitions between behavioral states.
• Clinical Relevance: This mechanism may explain why hippocampal lesions (e.g., in patient H.M.) lead to a loss of the "stopping signal" for eating, resulting in hyperphagia or continuous eating, distinct from the drive to eat itself.

In summary, the paper elucidates a mechanism where the dorsal subiculum projects to the mammillary body to gate the transition between feeding and non-feeding states, utilizing a bistable system where environmental and internal cues modulate the probability of terminating a feeding bout rather than the drive to initiate it.