Collectively, the extant data on the functions of vmPFC/OFC and VS in reward processing provides evidence for complementary roles that may be crucial for the control and execution of value-based decisions. A critical unresolved question, however, is how these two areas interact to mediate the observed functions. In this section, we outline research findings suggesting that vmPFC/OFC serves to modulate VS activity during reward-related behavior.
First, we note that anatomical and functional connectivity data from animals and humans are consistent with putative interactions between vmPFC/OFC and VS. Rodent studies have demonstrated direct glutamatergic projections from vmPFC/OFC to VS (Gabbott and others 2005
, Sesack and others 1989
, Voorn and others 2004
), while human fMRI studies indicate highly correlated activity between the vmPFC/OFC and VS at rest (Di Martino and others 2008
) and during tasks involving favorable outcomes or rewards (Cauda and others 2011
, Diekhof and others 2012
). While consistent with vmPFC/OFC modulation of VS activity, these circumstantial and correlational findings do not provide evidence of causality. Evidence corroborating the exact causal functional dynamics between the VS/NAc and vmPFC/OFC has only recently begun to emerge as a result of novel technological advances, which, at present, are only available for non-human animal studies.
Using computational modeling, Frank and Claus (2006)
provided a mechanistic account of fronto-striatal interactions derived from theories of reward and reinforcement learning. In this model, a striatal system receives dopaminergic inputs from the midbrain, which monitor the frequency of positive and negative decision outcomes via go and no-go (‘trial-and-error’) learning to refine motor actions. The OFC receives information about these positive and negative decision outcomes from the VS and constitutively updates information about the magnitude of an outcome to facilitate subsequent action selection on the basis of these updated value terms. Taking the OFC ‘off-line’ in this OFC-striatal neural network model results in deficits in decision-making, reversal learning, and devaluation, much like the effects seen in these tasks following lesions to the vmPFC/OFC across species. This model thus proposes that the expression of fast, flexible, and adaptive decision-making relies on intact interaction between the OFC and striatum. More specifically, it indicates a role for the OFC in the top-down biasing of striatal activity for action selection, wherein information about the magnitude or value of an outcome becomes integrated with information about simple frequencies of positive and negative outcomes to quickly and efficiently influence differentiation between ‘go’ and ‘no-go’ responses. This intriguing computational model preceded direct, in vivo
tests of VS-vmPFC/OFC interactions by several years, owed to the recent emergence of novel applications of multimodal techniques used to test the behavioral consequences of causal interactions between these two brain regions.
Novel combinations of techniques have only recently been used to relate the causal interactions between vmPFC/OFC and VS to animal behavior. Ghazizadeh and others (2012)
did this for the first time by combining electrophysiological recording of neurons in the NAcS with concurrent inactivation of the vmPFC in rodents during performance on a reward learning task. In the task, pressing an ‘active’ lever was required during the presentation of a tone to procure a sucrose reward and terminate the tone. Rats concurrently learned to inhibit responses to an ‘inactive’ lever which produced no reward if pressed during the presentation of a different tone. In contrast to control animals that were able to respond with greater frequency to the active lever while reducing lever-pressing during the neutral tone presentation, animals with inactivated vmPFCs lever-pressed during both tone presentations. This disinhibition following vmPFC inactivation was temporally specific to the presentation of the neutral tone, since responding was higher during the neutral tone presentation compared to other task epochs, such as the time delay between reward consumption and a new trial.
These compelling behavioral data call to question the exact influence that vmPFC can have on NAcS activity to affect appropriate time-sensitive responding. As expected, vmPFC inactivation results in direct modulation of NAcS neuronal activity and corresponds with response disinhibition during tone presentations (Ghazizadeh and others 2012
). vmPFC is specifically responsible for both a suppression of phasic firing that promotes behavioral cue responding and for providing excitatory input to increase the basal firing of NAcS neurons that inhibit responses generalized to reward-seeking (i.e., responses with no bearing on actual outcome). Putting this in the context of the computational model described above, vmPFC seems to control at least two distinct populations of neurons in the NAcS to mediate appropriate responding: (1) one population which facilitates actions (“go”) and (2) one population that inhibits responses (“no-go”). The dual nature of this modulation suggests a process of summation or integration of opposing signals for the ultimate expression of motoric output.
In a parallel effort to characterize the importance of the VS-vmPFC/OFC neural pathway on value-based decision-making, St. Onge and colleagues (2012)
performed concurrent inactivation of both brain regions and assessed the effects of this functional disconnection on rodent performance in a probabilistic discounting task. In this task, larger, uncertain rewards were pitted against smaller, sure rewards. Rats learned that pressing the lever that corresponded to large/risky outcomes was disadvantageous over time as the probability of obtaining a reward decreased over the course of the task. Although disconnection of the vmPFC and NAc did not impair the acquisition of probabilistic reward learning, the animals were less accurate and had slower response times and reduced locomotor activity. The authors suggest that these findings reflect impaired attention or vigilance.
This interpretation finds support in an earlier study showing that mPFC and NAc inactivation or disconnection resulted in attentional impairments in a five-choice serial reaction time task (Christakou and others 2004
). However, re-evaluating these findings in the context of the previously described working model of VS-vmPFC/OFC interactions provides an alternative interpretation of the data. If both VS and vmPFC/OFC are taken ‘off-line,’ there is no information from vmPFC/OFC about value to integrate with dopamine signals from the midbrain to influence and efficiently guide discrimination of ‘go’ or ‘no go’ responding. In effect, the most efficient (and perhaps most direct) pathway involved in guiding value-based decision-making has been “wiped out,” and it could be the case that other brain regions (i.e., amygdala, hippocampus, thalamus, dorsolateral PFC) are processing these reward-outcome associations, but to a much less efficient degree. This would explain the paradoxical nature of these findings, wherein learning is still intact but occurs at a much slower rate and is subject to error [or, in the context of Christakou and others, 2004
Despite these promising initial findings, substantially more work needs to be done to characterize the interactions within the VS-vmPFC/OFC circuit. Virtually no human studies have been done to understand the causal nature of these interactions and how these interactions relate to behavior. Future directions exploring the causal role of vmPFC/OFC activity on VS activity in humans could, like the previously described studies, combine multiple techniques. For example, in a recent study by Cohen and others (2012)
, NAc activity was recorded simultaneously with surface electroencephalogram (EEG) on patients with obsessive-compulsive disorder undergoing deep brain stimulation in the NAc during performance on a task of reward anticipation and motivation. Granger causality analyses showed that “top-down” or frontal cortical to NAc synchrony was stronger when rewards were being anticipated. As another example, a combined fMRI-lesion approach has been employed with amygdala-damaged patients (n
=2) to assess the causal role of the amygdala on vmPFC activity during engagement in a reversal learning task (Hampton and others 2007
). These studies presage future opportunities to examine the effects of manipulations of vmPFC/OFC function (e.g., through deep brain stimulation or neurological damage) for detectable differences in VS activity (compared to healthy controls) during their engagement in a reward processing task.
To this point, we have summarized evidence that adaptive value-based learning and decision-making depends critically on efficient integration of reward-related information within the VS-vmPFC/OFC pathway. A corollary of this supposition is that variation in the integrity of this circuit would be associated with variation in overt levels of social and affective functions related to reward processing. In the next section, we describe evidence that numerous psychiatric illnesses may involve deficiencies in VS-vmPFC/OFC circuit function.