A sip of water has a different value before and after thirst is quenched. An encounter with the same individual can prompt different behavior depending on past experience. When food enters the digestive tract, the resulting signals return to the brain and influence what happens next.

We study the information-processing rules in neural circuits that underlie these transformations. Starting with the signals entering the brain, we measure neural responses and connectivity, then alter the activity of selected cells or pathways to test their effects on behavior and physiology. Experiments in mice allow us to control sensory input precisely and ask how seemingly abstract quantities such as value, familiarity, and internal state arise from biological processes.

These are the neural circuit algorithms we seek to understand: how strongly a sensory input influences a circuit, how current information is compared with memory, and how brief signals produce responses that persist over time. Our three research themes begin with different phenomena and explore both shared principles and their differences.

An unfamiliar individual can attract our attention, yet our response changes as that individual becomes familiar. How does the brain connect present sensory information with memories of previous encounters? In mice, the tendency to investigate unfamiliar animals more than familiar ones provides an experimental entry point into this question. In the video, the mouse on the right is unfamiliar and the one on the left is familiar.

We identified a circuit in which information from the main olfactory system, which detects a wide range of odors, converges with hippocampal familiarity signals in the anterior medial amygdala (MeAa). By disrupting this sensory pathway, manipulating memory input, and recording neural responses across repeated encounters, we separated the contributions of these signals. Nature | Accepted.

MeAa population activity was greater during the first investigation of an unfamiliar animal, whereas hippocampal neurons projecting to this region responded more strongly to familiar individuals. Suppressing this hippocampal input removed both the novelty bias in MeAa activity and the behavioral preference for unfamiliar animals. Together with circuit tracing, these findings support a model in which hippocampal familiarity signals recruit a local inhibitory circuit, reducing MeAa output for familiar animals.

This output reached Tac2-expressing neurons in the posterolateral bed nucleus of the stria terminalis (plBNST). Suppressing these neurons abolished the preference for investigating unfamiliar animals while preserving general sociability: mice still preferred another animal to an empty enclosure. The work therefore distinguishes interest in another animal from a preference for an unfamiliar one at the circuit level.

These findings show that social novelty is a signal constructed from the relationship between current sensory input and past experience, rather than a fixed property of the sensory input itself. We now ask how repeated encounters build familiarity signals, how memory reshapes responses in individual cells, and how this information guides other social behaviors. The specific circuit mechanism was established in mice. Its generality across species, sensory modalities, and social situations remains a question for future work.

Selected original research

• Kim SR, Kim J and Kim SY. A neural circuit for social novelty preference integrates sensory and memory signals. Nature | Accepted.

A review of the circuit background

• Kim SR and Kim SY. Functional Dissection of Glutamatergic and GABAergic Neurons in the Bed Nucleus of the Stria Terminalis. Molecules and Cells (2021). This review examines how BNST functions differ across cell types and projections. Read the review

When we are thirsty and drink water, thirst begins to ease within seconds, before absorbed water restores fluid balance. What does the brain detect to respond so quickly? Holding water in the mouth, swallowing, and moving it through the digestive tract occur almost together. Identifying the cause requires separating these events.

In mice, we induced or blocked individual components of drinking while recording thirst-related neural activity. These experiments identified mechanical stimulation of the throat, specifically the pharynx, during swallowing as a key sensory source of rapid thirst-neuron inhibition. The mechanosensitive ion channel PIEZO2 made a major contribution to this signal. Signals from sensory afferents traveled through the brainstem and median preoptic nucleus to inhibit thirst neurons in the subfornical organ (SFO). Neuron | Accepted.

The key is how the circuit responds to the temporal pattern of swallowing. Both eating and drinking involve swallowing, yet thirst neurons responded differently in these two situations. Neural recordings, quantitative models, and circuit stimulation revealed a rate-dependent filter: rapid sequences of swallows, characteristic of drinking, produce strong inhibition. Stimulating an intermediate circuit node at drinking-like and eating-like rates reproduced this difference even without ingestion.

We also tested what this inhibition does to behavior. Activating thirst neurons specifically during drinking, thereby opposing their natural inhibition, prolonged drinking bouts. Strengthening inhibition shortened ongoing drinking. These experiments establish a causal role for rapid neural inhibition in controlling when a drinking bout ends.

In a human study conducted with colleagues at Sungkyunkwan University and IBS, participants reported greater thirst relief soon after swallowing water than after holding water in the mouth alone. This supports a rapid swallowing-related component of thirst relief in humans. The specific mechanosensory molecules and neural pathway were established in mice.

The work builds on our earlier Nature study (2020), which identified a feedback circuit through which mechanical signals from the upper digestive tract limit intake. It extends that foundation to a rule linking the timing of sensory input to behavioral control. We now ask which cellular and synaptic processes integrate brief inputs. Sustained gastric states and changes in appetite that outlast stress offer complementary questions: does a state persist because sensory input continues, or because processes within the circuit outlast that input?

Selected original research and related studies

• Kim M et al. A pharynx-to-forebrain circuit transforms swallowing into rapid thirst satiation. Neuron | Accepted.
• Kim DY et al. A neural circuit mechanism for mechanosensory feedback control of ingestion. Nature (2020). Read the paper
  _*featured as Highlight, Nature Reviews Gastroenterology & Hepatology 2020
  _*Selected among Korea’s 100 Outstanding National R&D Achievements in 2021.
  _(Lab news, media coverage, and related lecture videos; related articles: Article 1, Article 2)
• An et al. Lateral Septum Somatostatin Neurons are Activated by Diverse Stressors. Experimental Neurobiology (2022). An earlier study of neuronal responses to stress. Read the paper

Reviews of the field

• Kim, Heo and Kim. Neural signalling of gut mechanosensation in ingestive and digestive processes. Nature Reviews Neuroscience (2022). This review synthesizes evidence on digestive mechanosensation and its roles in ingestion and digestion. Read the review
  _(Lab news)
• Ahn et al. Brain circuits for promoting homeostatic and non-homeostatic appetites. Experimental & Molecular Medicine (2022). This review examines the neural circuit mechanisms through which energy deficits and the taste and smell of food drive appetite. Read the review

Warmth can be welcome when we are cold and aversive when we are hot. Detecting temperature is therefore only part of the problem. The brain must also use the body’s current state to determine the value of that sensation and how it should guide action.

We identified cold-responsive neurons in the mouse parabrachial nucleus that are required for heat production, shivering, cold avoidance, and increased food intake in the cold. Suppressing these neurons also reduced dopamine responses to rewarding cooling in a hot environment. This work identifies a circuit node through which cold sensation helps coordinate physiological adaptation, behavior, and reward responses. Nature Metabolism, 2026.

These findings connect sensory processing with the coordination of multiple responses to meet current needs. Where and how state changes the meaning of a signal remain central questions in our research.

In related work, we showed that inhibitory neurons in the lateral hypothalamus are required for diverse thermoregulatory behaviors. The same warming stimulus suppressed population activity in a cold environment but increased it in a hot environment. Parabrachial input was important for these thermal responses and for thermoregulatory behavior. Neuron, 2022.

Repeatedly tracking individual neurons across conditions then revealed their functional diversity. One ensemble responded to aversive thermal stimuli and food-associated cues; another was active during eating and drinking. Responses during consumption scaled with hunger and nutrient concentration. Behaviorally significant stimuli and actual consumption can therefore be represented differently even within the same defined neuronal population. Cell Reports, 2026.

Does bodily state alter the responses of neurons carrying sensory input, or the way downstream circuits use that information to guide action? By precisely controlling temperature and ingestion, we seek the rules through which need changes sensory value and behavioral choice.

Selected original research

• Jung et al. Parabrachial neurons orchestrate cold-induced homeostatic responses. Nature Metabolism (2026). Read the paper
  _Featured in a News & Views article in Nature Metabolism (2026).
  _Media coverage: ChosunBiz, The Asia Business Daily
• Jung et al. A forebrain neural substrate for behavioral thermoregulation. Neuron (2022). Read the paper
  _Featured in a Preview in Neuron (2022).
  _Lab news, media coverage, and related lecture videos
• Lee et al. Distinct lateral hypothalamic GABAergic ensembles encode motivational salience and value-scaled consumption. Cell Reports (2026). Read the paper

Separate sensory input from state and experience.
We precisely control temperature, ingestion, mechanical stimulation of the digestive tract, and social encounters while varying bodily conditions or prior experience. This allows us to distinguish responses to the stimulus itself from changes associated with state or experience.

Measure what neurons represent during behavior.
We use two-photon calcium imaging and fiber photometry to record activity in individual neurons and neuronal populations. Tracking the same cells across conditions alongside ingestion, movement, body temperature, or muscle activity helps relate neural signals to sensory, behavioral, and physiological events.

Manipulate cells and pathways to test causality.
Optogenetic and chemogenetic approaches allow us to alter the activity of selected neurons or pathways and measure the consequences for behavior and physiology at specific times. We go beyond observing activity that accompanies behavior to distinguish which functions require a circuit from the responses its activation can produce.

Link circuit function to cells and synapses.
We combine circuit tracing, electrophysiology, gene expression analysis, and three-dimensional tissue imaging. By identifying what distinguishes functionally different cells and how their connections influence one another, we aim to understand the biological implementation of information-processing rules.

Explain measured patterns and test predictions experimentally.
We use quantitative models to explain changes in neural activity over time, then test predictions by stimulating circuits at different rates and with different temporal patterns. In our thirst study, this linked recordings, modeling, and perturbations to explain why rapid and slow sequences of swallowing produce different outcomes.

During his doctoral training at Stanford, Sung-Yon Kim studied circuits controlling distinct features of anxiety. This work, published in two Nature papers (Nature 2011, Nature 2013), was recognized with the Society for Neuroscience’s Donald B. Lindsley Prize for outstanding doctoral research. During this period, he also contributed to studies of circuit mechanisms underlying anxiety- and depression-related behaviors in animal models (Nature 2012, Nature 2013, Nature 2015) and coauthored a related review (Brain Research 2013). At Stanford, he also contributed to the development of CLARITY through collaborative work (Nature 2013).

As a postdoctoral researcher at MIT, he led or contributed to projects developing methods for labeling and visualizing molecules in large tissue samples, including whole brains. This period included a review (Trends in Cognitive Sciences 2013) and original research (PNAS 2015, Cell 2015). These experiences laid a foundation for integrating functional and structural circuit analysis in our current research.

After establishing the Neural Circuit Lab at Seoul National University, we developed ZOOM with Professor Yan Lee’s group (Advanced Science 2019). This method expands biological specimens by an adjustable factor, making finer structures accessible to conventional light microscopy. Such method development and collaboration support our efforts to examine behaviorally relevant circuits at cellular resolution. Through collaborations with research groups in Korea working in biomaterials chemistry, nanoscience, microfluidics, and organoids, our lab has contributed to the development of technologies including ZOOM and OMNIA (ACS Applied Materials & Interfaces 2021, Bioconjugate Chemistry 2020, ACS Nano 2021, Acta Biomaterialia 2023, Science Advances 2022; related review: Molecules and Cells 2016; related book chapter: Seo et al. 2019).

• Park et al. Scalable and Isotropic Expansion of Tissues with Simply Tunable Expansion Ratio. Advanced Science (2019). Read the paper

Explore additional papers beyond the selected original research and reviews in the lists below. These also include collaborations and Sung-Yon Kim’s work before establishing the lab at Seoul National University.

PubMed | Google Scholar | ORCID