Neural Mechanisms of Motivation and Reward

National Institute Of Mental Health

Investigators

Linked publications & trials

Abstract

Primates, including old-world monkeys, learn stimulus-reward associations seemingly without effort. We previously showed that rhinal cortex has a causal role in associating value with visual stimuli, with monkeys without a rhinal cortex being less sensitive to previously learned stimulus-reward associations and learning new associations between visual cues and reward schedule states slowly, if at all. Rostromedial caudate (rmCD) also plays a causal role in stimulus-reward associations. When rmCD is inactivated chemogenetically, monkeys show a deficit in evaluating the size of the upcoming reward signaled by visual cues. There is a known connection between deep layers of the rhinal cortex projecting to rmCD. Here, we interrupted this connection by combining a unilateral rhinal cortex lesion with contralateral expression of an hM4Di DREADD via a retrograde lentivirus (FugE) injected into rmCD. The monkey was trained to perform a task with four visual cues, which signaled the upcoming reward of one, two, four or eight drops of water, respectively. A new set of cues was used each day. Without chemogenetic inactivation, the monkey learned quickly to accept only the two larger sizes of reward and refuse the smaller rewards. When all neurons projecting to rmCD were silenced by systemically injected DCZ, the monkey needed significantly more trials to learn the stimulus-reward mapping. The monkeys decision time also increased. To manipulate projection neurons from rhinal cortex to rmCD selectively, on some days, microinjections (3ul x 4 sites, 100nM) of DCZ were made into perirhinal cortex. The microinjections led to a small but significant deficit in the stimulus-reward mapping. Thus, our results show that projection neurons from rhinal cortex to rmCD are causally involved in the learning of visual stimulus-reward associations. The orbital frontal cortex (OFC) and rostromedial caudate (rmCD) are brain regions considered important for reward value processing and decision-making. The OFC is thought to be involved mainly in reward value assessment, whereas the rmCD is thought to play a role in reward expectation and action selection. To identify neurons that project from other areas to both of these regions, we injected retrograde viruses expressing red or green fluorescent reporters into the OFC and rmCD, respectively. After injection of the viral vectors, histological analysis showed a population of neurons in many regions co-expressing both fluorescent reporters, indicating that these neurons co-project to the OFC and rmCD. These neurons were located within higher order visual areas (V4 and TE) but were also identified in brain areas involved with processing spatial and emotional salience, such as the amygdala, rhinal cortex, and insula. Perhaps surprisingly, in the hemisphere contralateral to the injection site, these same regions had a larger than expected percentage of co-projecting neurons, that is, neurons that expressed both the red and green reporters, relative to neurons projecting to either OFC or rmCD alone. For example, in the amygdala of the ipsilateral hemisphere, approximately 20% of the neurons projecting to OFC also projected to rmCD, but from the contralateral amygdala nearly all OFC-projecting neurons also projected to rmCD. We are further characterizing the population of co-projecting neurons, which presumably carries identical signals to two brain regions thought to underlie reward evaluation and may play a role in the normalization of sensory information. Signals for reward value are distributed widely throughout the brain and are thought to be important for evaluating sensory stimuli when faced with behavioral choices. One mechanism for normalizing the value of sensory information is divisive normalization. In the early visual system normalization of luminance is achieved through feedback inhibition of multifurcating complex cells onto simple cells. At the other end of the decision-making hierarchy, there is evidence for divisive normalization, that is, a signal to which all other signals are seen as relative rather than absolute, in human and old-world monkey choice behaviors. Here we present evidence for the existence of a widespread neuron population that could be part of a substrate for normalizing the value of sensory signals from both hemispheres by co-projecting to orbital prefrontal cortex (OFC) and the rostromedial caudate (rmCD), key structures for evaluating and updating reward signals. We identified this co-projecting neuron population in two monkeys by unilaterally injecting a retrograde lentivirus expressing a fluorescent protein with one color into OFC and a retrograde lentivirus expressing another color into rmCD. Clusters of individual neurons that expressed both colors, and thus project to both regions, were found throughout the temporal lobe and other cortical regions responsible for processing visual and emotionally relevant information, including insular cortex and the dorsomedial thalamus. The highest percentages of co-projecting neurons, relative to numbers of neurons only projecting to one of the two regions, were found in areas of the basal and medial amygdala rich in dopamine innervation, the parahippocampus (TF) and the temporal parietal occipital area (TPO). In the same regions in the contralateral hemisphere, co-projecting neurons were found in smaller numbers but with a much higher percentage of co-projecting neurons compared to populations only projecting to the OFC or rmCD individually. Any signal responsible for normalization of value-reward association would need to be collected from sensory processing regions of both hemispheres, making this neuron population a prime candidate for sending identical value information to OFC and rmCD, which could be used to feed back a signal for adjusting gains in sensory processing areas.