Jingfeng Zhou

The research interest of the lab has coalesced around a desire to understand the neural circuit mechanisms of animal behavior and cognition. We use in vivo electrophysiology as a primary tool, combined with complex behavioral tasks, optogenetics, and computational tools, to map neural dynamics embedded in the neural population activities—primarily in the prefrontal cortex and hippocampus—to the task or cognitive variables that underlie learning, memory, and decision-making. We also have a great interest in how these neural representations and their behavioral functions would be altered in neuropsychiatric disease conditions such as addiction, depression, and schizophrenia.

How do we understand the world around us?

Plenty of evidence has shown that our brain is very likely to construct a mental model of the surrounding environment. Such a model (i.e., cognitive map, a term coined by Edward Tolman in the 1940s) accounts for complex relationships between various parts of the environment including sensory stimuli, events, and consequences caused by actions. It is believed that we use these models to attend, perceive, predict, learn, remember, decide, and generate actions—spanning the triad of perception, action, and cognition. Thus, to answer the question above, it is necessary to figure out how such a model is instantiated through the firing activities of many single neurons, the building blocks of our brain.

About models, there is a famous quote often attributed to the British statistician George Box: “All models are wrong, but some are useful”. The quote was originally said about statistical models; however, it might also be true for mental models—we do not build a mental model that captures every detail of the environment like a camera, but instead, we build one that is most useful in maximizing utility in a complex world. That said, our cognitive maps are not a true reflection of reality but are tailored to meet the task demand at hand (i.e., a "bespoke" cognitive map), which is supported by our recent work (Zhou et al., Curr. Biol., 2019a,b) in the orbitofrontal cortex and hippocampus.

Another important feature about the cognitive map is that it should be dynamic or evolving in time. For example, right after a life experience, such as doing a unit-recording experiment in the lab, we would hold a good memory of the episode for a while, including when and where the event has happened. Over time, these details fade away, but more abstract and generalizable knowledge—how to record single-unit activities in behaving animals—is formed and maintained. Consistent with this idea, we have seen neural ensemble activities in the orbitofrontal cortex showing the conversion from detailed to more schematic task representations (Zhou et al., Nature, 2020).

The neural mechanisms of cognitive maps and schemas (i.e., generalized cognitive maps) are still largely unknown, although many interesting hypotheses have been proposed, such as that the mental models in the hippocampus are more episodic while those in the prefrontal cortex are more schematic. In testing these and other fascinating ideas, there would be a lot of exciting and important work that needs to be done. Moving forward, the lab will focus on the following questions.

· How are different types of mental models—from episodic to schematic—learned and organized in the brain?

· How do prior experience and knowledge guide new learning and behavior through neural representations of these models?

· How are these neural representations and their behavioral functions altered in neuropsychiatric disease conditions, such as addiction, depression, and schizophrenia?

Methodology

To answer these questions, we use an integrative approach derived from behavioral, cognitive, systems and computational neurosciences.

· Carefully designed complex behavioral tasks to isolate the task or cognitive variables of interest, independent of sensory and motor confounds.

· Large-scale in vivo electrophysiological recording and calcium imaging.

· State-of-the-art multivariate analyses including machine learning to read specific information from both single-unit and neural-ensemble activities.

· Neural perturbation methods such as optogenetics and chemogenetics for causal inference.

· Computational modeling.

Update to December 2021

^*Co-first author, ^#Co-corresponding author

1. Bruch, S^#., McClure, P., Zhou, J., Schoenbaum G. & Pereira F^#. (2021). Validating the representational space of deep reinforcement learning models of behavior with neural data, bioRxiv, 448556.

2. Zhou, J^#., Gardner, M. P., & Schoenbaum, G#. (2021). Is the core function of orbitofrontal cortex to signal values or make predictions? Current Opinion in Behavioral Sciences 41, 1-9.

3. Zhou, J^#., Zong, W., Jia, C., Gardner, M.P.H., Schoenbaum, G#. (2021). Prospective representations in rat orbitofrontal ensembles. Behavioral Neuroscience. Accepted.

4. Zhou, J^#., Jia, C., Montesinos-Cartagena, M., Gardner, M.P.H., Zong, W., Schoenbaum, G#. (2020). Evolving schema representations in orbitofrontal ensembles during learning. Nature. (2020).

5. Gardner, M^#., Sanchez, D., Conroy, J.S., Wikenheiser, A., Zhou, J., Schoenbaum, G#. (2020). Processing in lateral orbitofrontal cortex is required to estimate subjective preference during initial, but not established, economic choice. Neuron, 108, 1-12.

6. Gardner, M^#, Conroy, J., Sanchez, D., Zhou, J., Schoenbaum, G#. (2019). Real-time value integration during economic choice is regulated by orbitofrontal cortex. Current Biology, 29(24), 4315-4322.

7. Zhou, J^#., Montesinos-Cartagena, M., Wikenheiser, A., Gardner, M., Niv, Y., & Schoenbaum, G#. (2019). Complementary task structure representations in hippocampus and orbitofrontal cortex during an odor sequence task. Current Biology, 29(20), 3402-3409.

8. Zhou, J^#., Gardner, M., Stalnaker, T., Ramus, S., Wikenheiser, A., Niv, Y., & Schoenbaum, G#. (2019). Rat orbitofrontal ensemble activity contains multiplexed but dissociable representations of value and task structure in an odor sequence task. Current Biology, 29(6), 897-907. (F1000 recommended)

9. Wang, D., Li, Y., Feng, Q., Guo, Q., Zhou, J., & Luo, M. (2017). Learning shapes the aversion and reward responses of lateral habenula neurons. eLife, 6, e23045.

10. Zhang, J., Tan, L., Ren, Y., Liang, J., Lin, R., Feng, Q., Zhou, J., Hu, F., Ren, J., Wei, C., Yu, T., Zhuang, Y., Bettler, B., Wang, F., & Luo, M. (2016). Presynaptic excitation via GABA B receptors in habenula cholinergic neurons regulates fear memory expression. Cell, 166(3), 716-728.

11. Li, Y., Zhong, W., Wang, D., Feng, Q., Liu, Z., Zhou, J., Jia, C., Hu, F., Zeng, J., Guo, Q., Fu, L., & Luo, M. (2016). Serotonin neurons in the dorsal raphe nucleus encode reward signals. Nature Communications, 7.

12. Guo, Q*., Zhou, J*., Feng, Q., Lin, R., Luo Q., Zeng S., Luo, M., & Fu, L. (2015). Multi-channel fiber photometry for population neuronal activity recording. Biomedical Optics Express. 6(10), 3919-3931.

13. Luo M., Zhou J., Liu Z. (2015). Reward processing by the dorsal raphe nucleus: 5-HT and beyond. Learning and Memory, 22: 452-460.

14. Zhou, J., Jia, C., Feng, Q., Bao, J., & Luo, M. (2015). Prospective coding of dorsal raphe reward signals by the orbitofrontal cortex. Journal of Neuroscience, 35(6), 2717-2730.

15. Liu, Z*., Zhou, J*., Li, Y., Hu, F., Lu, Y., Ma, M., Feng, Q., Zhang, J., Wang, D., Zeng, J., Bao, J., Kim, J., Chen, Z., Mestikawy, S., & Luo, M. (2014). Dorsal raphe neurons signal reward through 5-HT and glutamate. Neuron, 81(6), 1360-1374. (F1000 recommended)

16. Zhan, C*^#., Zhou, J*., Feng, Q., Zhang, J. E., Lin, S., Bao, J., Wu, P., & Luo, M^#. (2013). Acute and long-term suppression of feeding behavior by POMC neurons in the brainstem and hypothalamus, respectively. Journal of Neuroscience, 33(8), 3624-3632. (F1000 recommended)