Biology subjects

Kelsch, W.

Publications and source records attributed to Kelsch, W..

6 recordsLinked to original sources

bioRxiv2026

Error-driven representation learning in the mesolimbic system

In reinforcement learning, an agent learns to map representations of the environment state to predictions of future reward. Most prior work in neuroscience has assumed a fixed representation and studied how reward prediction errors (thought to be conveyed by phasic dopamine signals) are used to update the mapping from representations to predictions. However, work in machine learning has demonstrated that much more powerful predictive systems can be learned by using the errors to update the representations themselves. We study whether the brain does something similar by leveraging simultaneous recordings of striatal projection neurons in the olfactory tubercle (putatively representing state features) and dopamine neurons in the ventral tegmental area. We show that trial-by-trial changes in striatal activity are more consistent with dopamine-driven representation learning than a variety of alternative updating schemes. This result suggests a convergence of representation learning principles in biological and artificial systems.

neuroscience↗

bioRxiv2026

Sniffing Shapes Dopamine Signals for Reward Prediction

Adaptive behaviors depend on predicting outcomes from sensory evidence. Dopamine neurons in the ventral tegmental area (VTA) broadcast reward-prediction signals that guide learning. Yet the principles functionally coordinating information flow from input regions to VTA are incompletely understood. In the olfactory system, the sniff cycle structures sampling and odor encoding. We therefore asked whether this rhythm also entrains the ventral striatum to VTA communication and if so, how this shapes the implementation of predictive coding in dopamine neurons. We recorded identified dopamine neurons throughout olfactory conditioning and found that their firing shifted systematically to the post-inspiratory phase of the sniff cycle with learning. This temporal realignment predicted a neurons engagement in value encoding along the optimism-pessimism-spectrum of distributional reinforcement learning. This is associated with an enhanced phase-gated communication channel from the striatal olfactory tubercle to dopamine neurons, the strength of which predicts task performance. Thus, the sniffing rhythm provides a scaffold for information flow, revealing a phase-gating mechanism for the integration of outcome predicting sensory evidence to dopamine neurons during reinforcement learning. Significance StatementPredicting the future from sensory cues is central to adaptive behaviors. We show that sniffing temporally organizes the flow of information between the olfactory tubercle of ventral striatum and midbrain dopamine neurons during odor-guided learning in mice. Phase-specific coupling in the respiratory cycle determines when sensory information reaches dopamine neurons and how predictive signals are encoded. These findings link active-sensing rhythms to predictive reinforcement signals. This temporal scaffold yields a gradient of "optimistic" to "pessimistic" predictions consistent with distributional reinforcement-learning theories. These insights contribute to a better understanding of the large-scale computations across multiple brain regions to predict future outcomes.

neuroscience↗

bioRxiv2025

Individual differences drive social hierarchies in mouse societies

Social hierarchies structure groups and confer advantages on high-ranking individuals. In mice, individual position in hierarchies may emerge situationally from current group compositions, or, alternatively, may remain largely stable across groups as an internalized feature. Dominance and subordination are expressed in behaviors like tube competitions or agonistic chasing. The interaction of these behaviors in the shaping of social position in larger male mouse groups remains largely unknown. To address these questions, we developed the NoSeMaze, a semi-naturalistic, open-source, modular platform that enables automated long-term tracking of unperturbed groups. Across more than 4,000 mouse-days, hierarchies derived from incidental competitions in the integrated tube tests were non-despotic, transitive, and stable even when group compositions changed. This stability supports an internalized component of competition-based social rank. Chasing was also stable across contexts. Notably, chasing was concentrated among high-ranking individuals, consistent with ongoing negotiation of social rank among individuals at the upper end of the hierarchy. The link between chasing and social rank strengthened in groups with less well-defined rank structure, where mice rely more on aggressive signaling to assert their position. Chasing and social rank were associated with certain dimensions of simultaneously measured physical and cognitive features. In summary, high-dimensional tracking with the NoSeMaze reveals that social position in mice is multifaceted and shaped by stable dimensions of individual behavior that persist across changing social contexts. The approach thus enables longitudinal modeling of individuality and social position as key resilience factors.

neuroscience↗

bioRxiv2025

Stable clique membership in mouse societies requires oxytocin-enabled social sensory states

The ability to form stable de novo relationships in complex environments is essential for social functioning and is impaired in severe psychiatric disorders including autism. Yet, the neurobiological basis and cognitive processes enabling the formation of stable bonds in larger groups remain poorly understood, thereby limiting our ability to develop effective therapies. Here, we establish a semi-naturalistic model of clique formation in mouse societies, where individuals are tracked longitudinally from massive video data. Small, stable rich-clubs develop within these mouse social networks. Consistent with human rich-clubs, these cohesive cliques tended to have high social rank and exerted influence on non-members. Interestingly, neither prior rich-club-membership in a different group nor kinship facilitated entry into rich-clubs. Mimicking sparse population genetics, we probed the open question whether a subtle neuro-cognitive phenotype, namely impaired induction of social sensory processing states by cortical oxytocin signaling, disrupts higher-order social bonding in these complex social environments. Despite preserved social motivation, mice with alterations in this oxytocin subsystem failed to join rich-clubs. They approached group members less consistently, and connections from others towards them fluctuated more as well. This reciprocal disorganization highlights how interactional dynamics within social networks can amplify individual-level deficits, consistent with models of emergent properties of social behavior. These findings underscore the role of oxytocin in tuning sensory systems into a social processing state. Its dysfunction affects an individuals ability to establish stable relationships in complex social networks, with profound implications for social functioning deficits in psychiatric disorders.

neuroscience↗

bioRxiv2021

Hierarchical cross-scale analysis identifies parallel ventral striatal networks coding for dynamic and stabilized olfactory reward predictions

The unbiased identification of brain circuits responsible for behavior and their local cellular computations is a challenge for neuroscience. We establish here a hierarchical cross-scale approach from behavioral modeling and fMRI in task-performing mice to cellular network dynamics to identify how reward predictions are represented in the forebrain upon olfactory conditioning. fMRI identified functional segregation in reward prediction and error computations among olfactory cortices and subcortical circuits. Among them, the olfactory tubercle contributed both to dynamic reward predictions and prediction error. In this region, cellular recordings revealed two parallel neuronal populations for prediction coding. One population produced stabilized predictions as distributed stimulus-bound transient network activity; the other evolved during anticipatory waiting and fully reflected predicted value in single-units, dynamically integrating the recent cue-specific history of uncertain outcomes. Thus, the cross-scale approach revealed regional functional differentiation among the distributed forebrain circuits with a limbic hotspot for multiple non-redundant reward prediction coding.

neuroscience↗

bioRxiv2019

Phasic dopamine enhances the distinct decoding and perceived salience of stimuli

Subjects learn to assign value to stimuli that predict outcomes. Novelty, rewards or punishment evoke reinforcing phasic dopamine release from midbrain neurons to ventral striatum that mediates expected value and salience of stimuli in humans and animals. It is however not clear whether phasic dopamine release is sufficient to form distinct engrams that encode salient stimuli within these circuits. We addressed this question in awake mice. Evoked phasic dopamine induced plasticity selectively to the population encoding of coincidently presented stimuli and increased their distinctness from other stimuli. Phasic dopamine thereby enhanced the decoding of previously paired stimuli and increased their perceived salience. This dopamine-induced plasticity mimicked population coding dynamics of conditioned stimuli during reinforcement learning. These findings provide a network coding mechanism of how dopaminergic learning signals promote value assignment to stimulus representations.

neuroscience↗