Spontaneous animal behaviour is built from action modules that are concatenated by the brain into sequences1,2. However, the neural mechanisms that guide the composition of naturalistic, self-motivated behaviour remain unknown. Here we show that dopamine systematically fluctuates in the dorsolateral striatum (DLS) as mice spontaneously express sub-second behavioural modules, despite the absence of task structure, sensory cues or exogenous reward. Photometric recordings and calibrated closed-loop optogenetic manipulations during open field behaviour demonstrate that DLS dopamine fluctuations increase sequence variation over seconds, reinforce the use of associated behavioural modules over minutes, and modulate the vigour with which modules are expressed, without directly influencing movement initiation or moment-to-moment kinematics. Although the reinforcing effects of optogenetic DLS dopamine manipulations vary across behavioural modules and individual mice, these differences are well predicted by observed variation in the relationships between endogenous dopamine and module use. Consistent with the possibility that DLS dopamine fluctuations act as a teaching signal, mice build sequences during exploration as if to maximize dopamine. Together, these findings suggest a model in which the same circuits and computations that govern action choices in structured tasks have a key role in sculpting the content of unconstrained, high-dimensional, spontaneous behaviour.

Keypoint tracking algorithms have revolutionized the analysis of animal behavior, enabling investigators to flexibly quantify behavioral dynamics from conventional video recordings obtained in a wide variety of settings. However, it remains unclear how to parse continuous keypoint data into the modules out of which behavior is organized. This challenge is particularly acute because keypoint data is susceptible to high frequency jitter that clustering algorithms can mistake for transitions between behavioral modules. Here we present keypoint-MoSeq, a machine learning-based platform for identifying behavioral modules ("syllables") from keypoint data without human supervision. Keypoint-MoSeq uses a generative model to distinguish keypoint noise from behavior, enabling it to effectively identify syllables whose boundaries correspond to natural sub-second discontinuities inherent to mouse behavior. Keypoint-MoSeq outperforms commonly-used alternative clustering methods at identifying these transitions, at capturing correlations between neural activity and behavior, and at classifying either solitary or social behaviors in accordance with human annotations. Keypoint-MoSeq therefore renders behavioral syllables and grammar accessible to the many researchers who use standard video to capture animal behavior.

Rationale The basolateral amygdala (BLA) and medial geniculate nucleus of the thalamus (MGN) have both been shown to be necessary for the formation of associative learning. While the role that the BLA plays in this process has long been emphasized, the MGN has been less well-studied and surrounded by debate regarding whether the relay of sensory information is active or passive. Objectives We seek to understand the role the MGN has within the thalamoamgydala circuit in the formation of associative learning. Methods Here, we use optogenetics and in vivo electrophysiological recordings to dissect the MGN-BLA circuit and explore the specific subpopulations for evidence of learning and synthesis of information that could impact downstream BLA encoding. We employ various machine learning techniques to investigate function within neural subpopulations. We introduce a novel method to investigate tonic changes across trial-by-trial structure, which offers an alternative approach to traditional trial-averaging techniques. Results We find that the MGN appears to encode arousal but not valence, unlike the BLA which encodes for both. We find that the MGN and the BLA appear to react differently to expected and unexpected outcomes; the BLA biased responses toward reward prediction error and the MGN focused on anticipated punishment. We uncover evidence of tonic changes by visualizing changes across trials during inter-trial intervals (baseline epochs) for a subset of cells. Conclusion We conclude that the MGN-BLA projector population acts as both filter and transferer of information by relaying information about the salience of cues to the amygdala, but these signals are not valence-specified.

RationaleThe basolateral amygdala (BLA) and medial geniculate nucleus of the thalamus (MGN) have both been shown to be necessary for the formation of associative learning. While the role that the BLA plays in this process has long been emphasized, the MGN has been less well-studied and surrounded by debate regarding whether the relay of sensory information is active or passive.ObjectivesWe seek to understand the role the MGN has within the thalamoamgydala circuit in the formation of associative learning.MethodsHere, we use optogenetics to dissect the MGN-BLA circuit and explore the specific subpopulations for evidence of learning and synthesis of information that could impact downstream BLA encoding. We employ various machine learning techniques to investigate function within neural subpopulations. We introduce a novel method to investigate tonic changes across trial-by-trial structure, which offers an alternative approach to traditional trial-averaging techniques.ResultsWe find that the MGN appears to encode arousal but not valence, unlike the BLA which encodes for both. We find that the MGN and the BLA appear to react differently to expected and unexpected outcomes; the BLA biased responses toward reward prediction error and the MGN focused on anticipated punishment. We uncover evidence of tonic changes by visualizing changes across trials during inter-trial intervals (baseline epochs) for a subset of cells.ConclusionWe conclude that the MGN-BLA projector population acts as both filter and transferer of information by relaying information about the salience of cues to the amygdala, but these signals are not valence-specified.PrefaceIn tribute to Nadia Chaudhri, and her discoveries regarding how contexts can modulate the representation of cues in the amygdala (Sciascia et al., 2015), and other mesolimbic circuit components (Chaudhri et al., 2009, 2008; Valyear et al., 2020), here we explore what information processing precedes entry into the basolateral amygdala (BLA) from the thalamus and develop visualization approaches for exploring across-trial temporal dynamics in addition to within-trial population activity.