Neural Mechanisms of Motivation and Reward

National Institute Of Mental Health

Investigators

Linked publications, trials & patents

Abstract

Previous experiments from our group and others have shown that medial temporal lobe (MTL), orbital prefrontal cortex (OFC), and ventral striatum (VS) play important roles in the integration of visual and reward value information. Damage to any of those structures has an adverse effect on the ability to judge relative reward values, that is, after bilateral damage to any of these structures monkeys have difficulty in recognizing which rewards are biggest (best) and which are worth less. In our effort to learn how the signals in these interconnected brain regions interact to give rise to reward expectancy, we are using tools from molecular genetics. We are using nonreplicating viral vectors that carry genes into neurons. We have recently injected the OFC with a lentivirus virus expressing anterogradely transported green fluorescent protein (GFP) and visualized strong projections to both VS and the MTL, identifying a feedforward connection. To identify the feedback connections from OFC to medial temporal lobe, we injected a retrogradely transported Lentivirus (Lenti-Fug-E-syn::GFP) into the OFC projection site of the feedforward expression. In a second monkey, we injected a retrogradely transported Adeno-Associated Virus (AAV-retro-hSyn-GCaMP). Antibody staining for the GFP reporter expression, followed by immunohistochemistry as well as confocal microscopy, revealed common regions of afferent VS and OFC projections from the basolateral amygdala (BLA) and perirhinal cortex (PR). It also showed that area TE, in the anterior temporal lobe, projects strongly to the OFC while there were few cells projecting to the VS. Conversely, there is a dense entorhinal cortex projection to the VS but not to the OFC. We are now using a two-component retrograde virus system expressing (chemogenetic) receptors that allow chemically activated reversible silencing of targeted neurons to examine the functional properties of the projections reported above. Because of our previous work showing the importance of the ventral striatum in reward expectation, and because of the strong projections from both MTL and OFC seen in the viral experiments described above, we recorded neuronal responses from two major classes of VS neurons: TANs (Tonically Active Neurons) and PANs (Phasically Active Neurons) in VS of two monkeys. During the recording the monkeys performed a task in which 9 combinations of reward were offered by combining one reward size (2, 4 or 6 drops of water) with one delay for reward delivery (1, 4 or 7s). A visual cue presented throughout the trial indicated the combination being offered. The monkeys were sensitive to the information about reward size and delay carried by the visual cues: always accepting large rewards with short delays, and mostly refusing small rewards with long delays. The behavior was well-fit by acceptance rate=R/ekD , where R is the reward size in drops, D is the delay in seconds, and k is a free fitting parameter. Our analysis of the neural activity focused on the period immediately following the appearance of the visual cue but before any action were required, approximately the first 750ms after the cue was on. The response of both groups of neurons were analyzed for sensitivity to reward size and delay. When the number of neuronal firings (spike count) was used to measure the responses 31/50 TANs were modulated by the information available in the cue. From visual inspection of the data (rasters) it appeared that the responses of some of the TANs changed their patterns over time, so we used a method that would be sensitive to patterns over time, principal component analysis. This approach showed that the number of neurons sensitive to the reward-delay combination increased from 31 to 44, indicating that many neurons carry information in a temporal code, and that such information can be missed when examining only simple spike counts. This method showed that representing the pattern of the response in time was almost always a better way to measure the responses of TANs. Of the 101 PANS, 68 showed significant spike count modulation in the ANOVA. There were 4 more using PC1 as dependent variable. Thus, it appears that the TANs use a temporal code to convey reward information and the PANs do not. We believe this is an important finding. Ever since Adrian recorded from single neurons, there has been speculation that spike timing or patterns of spikes might carry behaviorally relevant information. Examples where it is clear that timing of spikes carries information are in the sonar system of bats and in the sensory field oscillations of electric fish. In these examples, the neural responses reflect the transformation of information about the outside world to a sensory neural representation. There has been less success in identifying whether the timing within the spike train carries a substantial amount of information for cognitive assessments of events to guide behavior. Here in the TANs, for the first time that we know of, a pure temporal code is used to convey cognitive information. The finding that the TANs, which are interneurons, are neurons with a temporal code is unexpected. Is temporal coding more likely to be a property of interneurons? This finding leaves us with a puzzle. The TANs are thought to influences the response of the PANs. Why do we not see a reflection of the temporal coding the responses of the PANs? We are looking into this issue.