Abstract
Understanding how the brain integrates internal physiological states with external sensory cues to guide behavior is a fundamental question in neuroscience. This process relies on interoceptive predictions, which are internal models that anticipate changes in the body's physiological state based on sensory inputs and prior experiences. Despite recent advances in identifying the neural substrates of interoceptive predictions, the precise neuronal circuits involved remain elusive. In our study, we demonstrate that Dopamine 2 Receptor (D2R+) expressing neurons in the paraventricular nucleus of the thalamus (PVT) play key roles in interoception and interoceptive predictions. Specifically, these neurons are engaged in behaviors leading to physiologically relevant outcomes, with their activity highly dependent on the interoceptive state of the mice and the expected outcome. Furthermore, we show that chronic inhibition of PVT D2R+ neurons impairs the long-term performance of interoceptive-guided motivated behavior. Collectively, our findings provide insights into the role of PVT D2R+ neurons in learning and updating state-dependent predictions by integrating past experiences with current physiological conditions to optimize goal-directed behavior.
Introduction
While external senses like vision and audition are well-studied, interoception—the sensing of internal physiological states such as hunger or thirst—is equally vital for survival. To navigate complex environments, the brain does not merely react to internal signals; it generates interoceptive predictions. These internal models anticipate physiological changes based on environmental cues and prior experience. While regions like the hypothalamus are known to sense homeostatic needs, the circuits responsible for integrating these needs with external sensory information to prioritize behavior have remained unclear. This study investigates the role of Dopamine 2 Receptor-expressing (D2R+) neurons within the paraventricular nucleus of the thalamus (PVT) as a potential hub for this integration.
Methodology
To investigate the function of PVT D2R+ neurons, researchers utilized a multi-modal approach in Drd2-Cre mice:
- Fiber Photometry: Real-time neural activity was recorded using the calcium indicator GCaMP6s while mice performed behavioral tasks.
- Chemogenetics (DREADDs): The inhibitory designer receptor hM4Di was used to selectively and chronically silence PVT D2R+ neurons via administration of Clozapine N-oxide (CNO).
- Linear Foraging Task: Mice were trained to wait in a "trigger zone" before traversing a corridor to a "reward zone" to receive various liquid rewards, including high-calorie milkshake, sucrose, sucralose, or water.
- Functional Linear Mixed Models (FLMM): This statistical framework was applied to photometry data to map neural activity to specific behavioral phases (e.g., approach vs. reward delivery) across trials of varying durations.
Key Insights and Results
State-Dependent Neural Activity
PVT D2R+ neurons exhibit activity that is highly sensitive to the animal's internal state. In hungry mice, these neurons showed intense firing during the approach to a food reward. Conversely, this activity significantly diminished when the mice were sated. Statistical analysis revealed a strong negative correlation between neural activity and trial order ($p < 0.001$), indicating that as the mice reached satiety, the neural response attenuated accordingly.
Encoding Nutritional Utility
The circuit appears specifically tuned to the interoceptive value of an outcome rather than just taste. While the neurons initially responded to both caloric (milkshake) and non-caloric (sucralose) rewards, only caloric rewards maintained state-dependent modulation over time. Furthermore, the neurons were sensitive to reward magnitude, with larger volumes (6µl vs 1µl) eliciting significantly higher response magnitudes ($p < 0.05$).
Prediction and Expectation Ramping
A characteristic "ramping" of neural activity was observed as mice approached the reward zone. Crucially, this ramping was absent during "premature" trials where the mouse moved before the environmental cue. This suggests that PVT D2R+ neurons encode the prediction of a reward based on learned environmental cues rather than the motor act of foraging. These interoceptive predictions emerged slowly over several days of training, suggesting a gradual refinement of the internal model.
Behavioral Impact of Chronic Inhibition
While chronic chemogenetic inhibition of PVT D2R+ neurons did not prevent mice from learning the basic mechanics of the foraging task, it significantly impaired long-term performance ($p < 0.001$). Mice with silenced D2R+ neurons completed fewer trials and exhibited a marked reduction in motivation compared to control groups, highlighting the necessity of this circuit for sustained, goal-directed behavior.
Discussion and Limitations
These findings position PVT D2R+ neurons as a critical "prediction engine" that integrates past experiences with current physiological needs. However, the study notes that fiber photometry captures the aggregate activity of a neuronal population. Future research using single-cell imaging may reveal specialized sub-populations within the D2R+ group. Additionally, while this study focused on hunger, the PVT may play a broader role in other interoceptive domains such as stress or pain. Understanding how these predictive circuits malfunction could provide vital insights into psychiatric conditions characterized by disrupted interoception, including eating disorders and addiction.