PRESENTED BY
Jeff Markowitz, Ph.D.
Postdoctoral Fellow
Harvard Medical School

ABSTRACT
With remarkable ease, our brains are able to process rewards, goals, external cues, and our internal state, and select a single action to execute among myriad alternatives. Decades of research both in the clinic and in animal models point to the basal ganglia (BG) as a critical neural circuit for this process. The BG are thought to implement a simple push-pull model, wherein two classes of principal cells (direct and indirect pathway neurons) initiate or suppress actions. While the push-pull model is appealing in its simplicity, it may be a consequence of how behavior is typically studied in the lab. To address this, we used custom machine learning techniques that enable us to comprehensively map spontaneous 3D sub-second behaviors exhibited by a mouse in the open field. At the same time, we monitored activity in direct and indirect pathway neurons. Rather than a simple push-pull dynamic, we found that these cells encode the timing of transitions between behaviors, reflect the content of individual movements and their sequence context, and directly control the statistics of how movements are concatenated into sequences.
 
The BG also interface with the brain’s reward systems via the neuromodulator dopamine. Dopamine is a known potent reinforcer, however, relatively little is known about how dopamine influences ongoing spontaneous behavior. To address this, instead of training a mouse to perform a stereotyped, learned behavior, we developed a new closed-loop system that can automatically recognize behaviors as they are performed by a freely moving mouse. With this system, we optogenetically triggered dopamine release as mice performed specific actions spontaneously in an open field. Since the mouse is freely moving, we could assess how dopamine reward impacts the full landscape of the mouse’s behavior. Surprisingly, we found that dopamine plays a complex role in ongoing actions, controlling the probability with which behaviors are expressed, their time-scale, how they are executed, and how they are sequenced over time.

BIOGRAPHY
As both a graduate student and a postdoc, I have focused on the question of how neural activity is orchestrated to produce natural behaviors. To that end, my training has spanned theoretical neuroscience, engineering, machine learning, and experimental approaches in multiple model organisms. As a graduate student with Timothy Gardner, I focused on how the nucleus HVC controlled vocalizations in the zebra finch. These experiments exploited a custom, minimally-invasive electrode that I helped to develop when I first entered the lab. I also developed statistical methods to capture the long-range dependencies found in canary song. This led to the surprising finding that, depending on the note being sung, the decision about what to sing next could depend on what the canary sang up to 10 seconds into the past. As a postdoctoral fellow jointly advised by Bernardo Sabatini and Sandeep Datta, I have combined 3D depth imaging, time-series modeling, and neural recording and perturbation to understand the role of the basal ganglia in action selection. In my own lab, I will use advances in this technology to understand the link between brain and behavior through delivering targeted perturbations to the animal’s neural or behavioral state as they are freely moving.