Hippocampal contributions to dopamine prediction error signals

• Applicant: Kauê Machado Costa, Ph.D.
• Subject Area: Cognitive, Systems and Behavioural Neurobiology ; Molecular Biology and Physiology of Neurons and Glial Cells
• Term: from 2018 to 2023
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 419309165

• Final Report Year: 2023

Final Report Abstract

I received a DFG fellowship from 2019 to 2022 to work as a postodoctoral fellow at the laboratory of Geoffrey Schoenbaum, at the National Institute on Drug Abuse. During this time I performed a set of experiments combining classical learning theory and state-of-the-art techniques for recording and manipulating neural activity in freely-moving rats that aimed to better understand the neural mechanisms of learning, specifically differentiating learning that is based on reward (value-based) and neutral sensory (value-less) prediction errors. The main findings of my DFG-sponsored work were: 1) The discovery that the orbitofrontal cortex is necessary for the value-less latent learning that underlies latent inhibition, or the process by which animals learn to ignore irrelevant stimuli. This finding supports the hypothesis that the orbitofrontal cortex is involved in the latent learning of task states, which helps to better understand the role of this critical brain region. 2) The discovery that the orbitofrontal cortex is critical for learning the specific sensory identities involved in associative learning, but not for the general process of building associative models, or cognitive maps, of a task. This work strongly supports the so called “cartographer” hypothesis of orbitofrontal cortex function, which stipulates that this area is specifically involved in updating cognitive maps. Additionally we had the surprising finding that this update function is highly selective for sensory-specific properties of associative elements. 3) The discovery that dopamine signaling in specific striatal subregions (the nucleus accumbens and dorsomedial striatum) reflects value-less sensory prediction errors in addition to reward prediction errors and novelty. This upturns common assumptions about the dopamine system, which is typically thought to only be involved in reward-based learning. 4) The discovery that dopamine signals in the nucleus accumbens reflect inferred prediction errors in a manner that is dependent on orbitofrontal cortex activity. This was a surprising finding that opens new avenues of investigation. Taken together, this set of findings revealed that the orbitofrontal cortex is important for sensory-specific learning, especially when this learning is not directly related to reward or necessary to execute a task, and that dopamine signals, commonly thought to be related only to learning based on rewards, also supports learning about value-less sensory stimuli withing a similar general computational framework (prediction error coding).

Publications

• The orbitofrontal cortex is necessary for learning to ignore. Current Biology, 31(12), 2652-2657.e3.
  Costa, Kauê Machado; Sengupta, Ayesha & Schoenbaum, Geoffrey
  
• The role of the lateral orbitofrontal cortex in creating cognitive maps. Nature Neuroscience, 26(1), 107-115.
  Costa, Kauê Machado; Scholz, Robert; Lloyd, Kevin; Moreno-Castilla, Perla; Gardner, Matthew P. H.; Dayan, Peter & Schoenbaum, Geoffrey
  
• Striatal dopamine release reflects domain general prediction errors
  Costa K.M.; Raheja N.; Mirani J.; Sercander C. & Schoenbaum G.