Main Menu (Mobile)- Block

Main Menu - Block


custom_misc-custom_misc_header_title_slogan | block

Dudman Lab / Research

Our lab studies a critical nexus in the mammalian brain where sensory information and motor planning come together to subserve volition - the basal ganglia.
janelia7_blocks-janelia7_secondary_menu | block
node_title | node_title
node_body | node_body

Functional architecture of basal ganglia circuits

While much progress was made over the last several decades describing the anatomy of the basal ganglia a number of pressing questions remain. We approach these questions from a normative perspective in the sense that we first ask what features the circuit ought to have given the computations it has been proposed to perform. 

We next ask whether the circuit does, indeed, have those properties and how those properties are implemented by the biophysics and connectivity of constituent neurons. Finally, our long term goal is always to evaluate whether the key architectural features of the circuit that we identify are indeed necessary or sufficient for behavior.

A representative project is our recent study: The inhibitory microcircuit of the substantia nigra provides feedback gain control for basal ganglia output . As illustrated in this project we use patch clamp electrophsiology, extracellular recording, two photon microscopy, and functional circuit mapping to study the functional architecture of basal ganglia circuits. In collaboration with other labs at Janelia Research Campus we are developing molecular, genetic, optical, and chemical tools to characterize the integrative properties of neuronal circuits and precisely perturb their function during behavior.

Adapting behavior to the expected timing of reward

To make optimal decisions in the presence of uncertainty requires the inference of probabilistic models. For example, financial decision theory posits that selection of optimal portfolios requires information about both the mean and variance of expected returns. In the timing literature it is often suggested that rodents learn the delay until reward delivery with a fixed relative uncertainty. However, in a natural environment the timing of events may have an arbitrary uncertainty. Thus, we asked whether mice could infer probabilistic models of timing in highly dynamic environments. Recently, we developed a behavioral paradigm to address these questions. As we argue in our recent paper Mice infer probabilistic models for timing, mice can accumulate reward timing information over many tens of trials to infer accurate probabilistic models and make optimal decisions.

What computations the brain uses to infer probabilistic models remains a fundamental, but unanswered, question. Our lab is currently using dynamic tasks like the one above to first describe the neural correlates of inference. We use extracellular recording and two-photon calcium imaging to monitor that activity of many tens of neurons simultaneously during behavior. By developing tasks in which we have a good handle on the psychophysics of the mouse's behavior we disambiguate the multiple underlying neural signals related to task performance. Once we have described the underlying neural correlates, we will use the tools for cell-type specific perturbation available in the mouse to shed light on the computations performed and their circuit mechanisms.

Dynamics of the basal ganglia during behavior

If we are to understand how an organism can adapt their behavior to be in the right place at the right time to collect rewards, then we need to understand how voluntary behaviors are selected and specified. We focus on the basal ganglia which are critical for voluntary, reward-seeking behaviors. It is clear that perturbations of basal ganglia signalling, either induced experimentally in model organisms or as a consequence of disease, produce profound deficits in the vigor, timing, and selection of appropriate reward seeking actions. However, a number of fundamental questions remain unanswered. For example, it is unclear to what degree an appropriate action, to be taken at the correct time, is selected in the sense of opening a gate or whether the properties of the action are specified. Currently, we are exploring this question by asking what is represented about an action in the dynamics of the activity of populations of basal ganglia neurons.

A representative project is our recent study: Neural signals of extinction in the midbrain. We found that GABAergic neurons in the ventral midbrain, including the substantia nigra pars reticulata, exhibited short latency responses to conditioned stimuli (in this case pure tones). At the output nucleus of the dorsal bagal ganglia, the substantia nigra, we have recently observed that a conditioned stimulus (CS) recruits a precise sequence of activations in distinct cell-types. The first two populations to respond are GABAergic neurons (confirmed by in vivo using optogenetic tagging) followed by the phasic response of dopamine (DA) neurons as illustrated in the figure at right. The recruitment of this sequence of cell types is contingent upon learned associations between CS and a subsequent reward delivery. What remains unclear is how this sequential recruitment of circuits in the ventral midbrain is related to features of the learned response.

Tools and technologies for neuroscience

We are actively pursuing this question using a combination of extracellular recording and new analysis methods developed in collaboration with other groups at Janelia. In these projects we combine optical recording with extracellular and intracellular electrophysiological recording in behaving mice. We use both mice trained to behave under a microscope or freely moving mice with custom electronics. In collaboration with several groups at Janelia we are drawing upon the expertise of the Instrument Design and Fabrication group to design and build custom electronics, electrodes and recording systems. We continuously make some of these developments available in the resources section of the website where there is software available for download.