HomeResearchComputational Principles Underlying Neural Circuit Dynamics and Behavior

Our Scientists

Computational Principles Underlying Neural Circuit Dynamics and Behavior

Research Summary

Anthony Leonardo is interested in general principles underlying neural information processing. His lab investigates the computations neural circuits implement, how robust they are, and how different circuits are linked into systems to produce behaviors.

How do behaviors emerge from computations distributed across many neural circuits? Behaviors, such as the singing of a bird or the flight of a bee, unfold before us smoothly and reveal no sign that they are generated by a vast network of interconnected neurons whirring away in the background. It is this enigmatic connection between circuit dynamics, computation, and behavior that drives us forward. We study how neural circuits work, what they do, and how they are linked together to produce computations and behaviors.

Our research is currently focused on the neural basis of prey capture, in both salamanders and dragonflies. Prey capture is sophisticated enough that understanding it will be a significant step beyond the more autonomous behaviors that have been well characterized. Yet for all its behavioral richness, prey capture is not overwhelmingly complex. The basic loop from sensation to motor output is only a few processing stages; these computations begin in early visual circuits in which we already have some degree of strong computation footing. Ultimately we hope to characterize the entire sequence of circuit dynamics underlying prey capture, from the eyes to the muscles, thereby forming a closed-loop system. Comparisons of the two systems will allow us to address whether particular neural circuits are so adept at producing certain computations that nature has used the same circuits again and again, such that there is a toolbox of fundamental neural circuit primitives from which all behaviors are constructed.

Our work takes a systems-level approach, combining behavior, neural recordings, and theory. First and foremost, we make behavioral measurements of the precision and robustness of prey capture in a variety of experimental conditions. Second, we use both electrophysiological and optical tools to measure the responses of large numbers of neurons. Finally, we build models of neural circuits based on control and dynamical systems theory. These models quantitatively describe how neural circuits embody particular computations and when these computations will succeed and fail. The models also make predictions about the behavior and the nature of downstream circuits, and are the engine that drives much of our experimental work.

Precision and Robustness of Salamander Prey Capture
Salamanders are predatory amphibians; they ballistically launch their tongues to catch rapidly moving insects. Ethological studies with freely moving salamanders and flies allow us to measure salamander prey capture precision and robustness in a wide variety of conditions. These studies define the basic operating regime of the behavior and the conditions over which the neural circuits underlying the behavior must function. Correspondingly, these studies allow us to test quantitative predictions from retinal circuit models regarding how prey capture precision and robustness should change as a function of prey size, speed, and background lighting. This is one way in which we attempt to link macroscopic behavioral events back to microscopic components of neural circuit models.

Circuit Dynamics and Computation in the Retina During Prey Capture
The first processing stage in prey capture is the retina—an ideal place in which to investigate neural circuit dynamics because its anatomy is well defined and amenable to electrophysiological measurement and pharmacological manipulation. We use multielectrode array and patch recordings to measure the electrical activity of retinal neurons during visual stimulation by small moving targets. In the future, we hope to develop optical tools for measuring the responses of hundreds of retinal neurons simultaneously, in collaboration with Luke Lavis and Eric Betzig (both HHMI, Janelia Farm Research Campus).

Tectal Recordings
There is little hope in understanding computation in the retina by studying only the retina. Ultimately it is necessary to record downstream, in the optic tectum, to directly measure how these neurons interpret and modify the retinal output. As our understanding of the retina's role in prey capture grows more sophisticated, we are gearing up to begin recording from single tectal neurons. These studies will focus on quantifying tectal neuron responses in the context of the retinal circuits that provide them with input. This will enable us to develop circuit models for the tectum, which will in turn allow us to begin unraveling the precise computational role of the tectum in prey capture.

Precision and Robustness of Dragonfly Prey Capture
Dragonflies are arguably the most sophisticated hunting and flying machines ever produced by evolution. By making rapid flights that predict and intercept the prey insect's trajectory, dragonflies capture nearly all of the insects they chase. We have constructed a large indoor flight arena where dragonflies can fly in controlled experimental studies. Analogous to our work in salamanders, the flight arena will be used to measure the operating regime of dragonfly prey capture, the conditions under which the neural circuits must function, and as a vehicle to test quantitative predictions from neural circuit models regarding how prey capture precision and robustness should change in different environmental conditions.

Circuit Dynamics of Target-Selective Neurons in the Dragonfly Nerve Cord
The final premotor computations in dragonfly prey capture are implemented by a set of target-selective descending neurons (TSDNs) in the ventral nerve cord. This set is composed of 16 identified neurons that map visual space and are sensitive to the motion of small moving targets, yet also modulate wing muscle activity. TSDNs were first described by Rob Olberg (Union College), who hypothesized that their function is to alter the dragonfly's flight course to intersect that of its prey. We are developing circuit models, in combination with electrophysiological recordings of TSDN activity, that formally describe the time dependent responses of these neurons to moving targets. These circuit models will provide a compact description of the computations implemented by the TSDN population and should provide insight into how TSDN outputs are used to drive turning muscles during pursuit flights.

Wireless Recordings in Flying Dragonflies
To fully understand behaviors requires measuring the responses of neurons while the animal is moving. In the case of the dragonfly, this requires recording from dragonfly neurons and muscles during flight. In part, our choice of the dragonfly as a system in which to study neural circuits arises from the fact that it is one of the few insects large enough to carry substantial weight. In collaboration with Reid Harrison (University of Utah), we have begun the development of a wireless multichannel neural amplifier that dragonflies can carry during prey capture. In collaboration with Rob Olberg, we are developing custom electrodes to record from TSDNs in intact behaving dragonflies. These tools will be used, in the large flight arena, to record from target-selective neurons and muscles while the dragonfly is catching prey. In combination with high-speed video and accelerometers, this will provide us with a complete measurement of all the control variables used in the end stages of dragonfly prey capture: responses of visual premotor neurons and the muscles they drive, positional changes in the configurations of the wings, the ensuing turning forces on the body, and manner in which these change the dragonfly's position relative to the prey it is chasing.

This research was supported in part by the Helen Hay Whitney Foundation, the Grass Foundation, and a Career Award from the Burroughs Wellcome Fund.

As of July 06, 2010

Scientist Profile

Janelia Group Leader
Janelia Research Campus
Neuroscience