Circuit Mechanisms of Learning
- Research Goal and Significance
One central goal of neuroscience is to elucidate how learning is encoded and executed by the nervous system. To fully understand complex learned behaviors, we need to uncover the underlying neural circuits, define the properties of each functional node to address how a given circuit processes sensory inputs into motor outputs, illuminate the consequence of learning at each tier of the circuit, and characterize how these changes lead to learned behavior. We also need to understand how neuromodulators, such as amines and growth factors, set the global state of the neural circuits to modulate behavior. Our long-term goal is to achieve an integrative understanding of learning by addressing all levels of the underlying neural circuit from sensory perception to motor execution. Eventually, we will be able to test our understanding by manipulating learned behavior with genetic or optical tools, “writing in” experience-dependent changes to predictably modulate behavior.
- System and Approach
To tackle these questions, we study learning in C. elegans. A major advantage of this system is that the complete anatomical connectome that describes the connectivity of all 302 neurons in its nervous system. Moreover, this well-defined nervous system can be dissected with powerful molecular, cellular and genetic tools. The functional properties of each neuron in sensorimotor transformation and learning – properties that are largely stereotyped among individuals - can also be characterized with optical neurophysiology in live animals as they perform complex, experience-dependent behaviors.
We focus on olfactory learning. Because olfaction plays an essential role in locating food sources, olfactory learning is similarly displayed across animal species. Thus, the underlying mechanisms are likely to be conserved. Olfaction is critical for C. elegans to navigate the environment, and the well-characterized olfactory neurons use conserved signaling pathways to process sensory inputs. We integrate quantitative behavioral analysis, molecular cellular genetics, optical imaging and ontogenetics in live animals that perform defined behavioral tasks to pursue a full scale model of learning that will ultimately incorporate all of the regulatory and modulatory elements in a small nervous system.
You can navigate to the “Movies” to see some cool displays of our methods.
We seek the operational rules by which neural circuits for experience-dependent olfactory behavior utilize molecular and cellular mechanisms to shape neuronal properties and generate behavioral changes.
In the past several years, we have established an olfactory learning paradigm whereby the nematode learns to avoid the smell of pathogenic bacteria after ingestion, analogous to Garcia effect found in many animals, including humans [Zhang et al., Nature 438:179]. We have developed an automated assay to quantify this aversive olfactory learning and functionally mapped a neuronal network that encodes both the naive and learned olfactory preferences [Luo et al., J Neurophysiology 99:2617; Ha et al., Neuron 68:1173]. We have also systematically characterized the property of the learning network [Ha et al., Neuron 68:1173; Hendricks et al., Nature 487:99; Shen et al., eLife e14197; Qin et al, J Neuroscience 33:925; Liu et al, Neuron 97:390]. Particularly, we have identified, for the first time, the compartmentalized axonal activity in a C. elegansneuron that plays a key role in the circuit of the olfactory learning [Hendricks et al., Nature 487:99; Liu et al, Neuron 97:390]. This compartmentalized activity represents an internal feedback signal from motor program that resembles corollary discharge, a major means by which the nervous system monitors self-generated motion. Corollary discharge has been implicated in the learning ability of more complex brains and pathology of neurological disorders, such as schizophrenia. Identifying a simple form of corollary discharge in a genetically tractable model organism will allow us to study the mechanistic underpinnings of this important neural signaling and its role in complex brain function. We have also characterized the spatial and temporal activity of neuromodulators, such as insulin-like peptides and a TGF-beta growth factor, in regulating learning [Zhang and Zhang, PNAS 109:17081; Chen et al., Neuron 77:572; Fernandes de Abreu DA et al, PLoS Genetics 10(3):e1004225; Wu et al., Neuron in press]. Our findings have elucidated remarkably conserved features in the functional structure, computational property and modulatory signaling of neural circuits that underlie experience-dependent behavioral plasticity.
Our results provided critical foundations to further characterize the circuit mechanisms of learning. Our current research addresses the following questions.
Characterize functional consequence of learning in the underlying neural circuit To understand how the network encodes learning, we will characterize the mechanisms underlying the neuronal correlates of learning, i.e. the experience-dependent changes in the learning circuit that correlate with learning. We will analyze the neurotransmission and genetic underpinnings of the learning correlates. We will also define how these learning correlates interact at circuit level.
Characterize the causal link between learning correlates and learned behavior To understand how learning is encoded by the experience-dependent changes in the underlying circuit, we will characterize the causal role of the identified learning correlates in generating learning. We will manipulate the properties of the key learning neurons to eliminate the learning correlates in trained animals and “build” the learning correlates into the circuit of naive animals with optogenetics and other conditional regulation of neuronal activities, and examine the resulting effects on learning.
Characterize an ILP-network that regulates learning in the large context of animal physiology We showed that the insulin-like peptides (ILPs) regulate learning by modulating the activity of learning circuit. ILPs also respond to various environmental conditions. Thus, we hypothesize that multiple ILPs regulate learning in a context-dependent manner. We address this hypothesis by characterizing the underlying genetic framework and cellular circuit.