Efficiently allocating resources to meet many and varied needs is a hallmark of adaptive behavior. To maximize the return on these limited resources, one must estimate the expected value of an outcome before engaging in a particular action. Expected value can be signaled either explicitly through stimulus-reward associations or inferred from experience with the demands of a given environment. In rhesus monkeys, lesions of rhinal cortex (Rh) cause profound deficits in the ability to adjust behavior in response to reward-predicting visual cues. Recently, we found that lesions of orbitofrontal cortex (PFo) an area with strong reciprocal connections to Rh and functional disconnection of PFo and Rh also altered responses in an instrumental task in which visual cues predicted different levels of reward. Surprisingly, in these latter two groups, this impairment persists even in conditions in which reward predictions are generated by varying reward size across blocks of trials that lack visual cues. We wondered whether ablations of Rh would also alter behavior in a task in which rewards are predictable but expected value is not signaled by visual cues. To determine whether Rh is required primarily to form cue-reward associations or more generally for determining expected value, we tested monkeys with ablations of Rh (n=3) as well as a set of unoperated controls (n=3) on a simple lever pressing task. To earn a fluid reward, monkeys had only to press and release a touch bar. In this task, reward size (1, 2, 4 or 8 drops) varied across blocks (25 trials per block);no visual cues predicted reward nor instructed the animals to respond. Animals were free to press and release the lever at their own pace;releases were followed by visual feedback and reward delivery. Both the control and experimental group quickly learned that bar releases led to reinforcement. Control group performance was significantly affected by expected reward size. On average, controls adjusted their inter-response intervals (IRI) 1-2 trials after block transitions and had significantly shorter IRIs during large reward blocks. Surprisingly, the Rh group did not show a sensitivity to reward size in this task;averaging responses across all trials in a block revealed no effect of reward size on this groups performance, and examination of the within-block trial-by-trial dynamics did not reveal a tendency for the IRI to differ across reward sizes either early or late in a block. Our demonstration that lesions of rhinal cortex alter monkeys expected value estimates even when value is not signaled by sensory cues is unexpected given this structures long accepted role in multi-modal sensory processing and short term memory. Dopamine and noradrenaline, two important and closely related modulatory neurotransmitters, are critical for normal motivated behavior. Depletion of either of these transmitters severely curtails normal exploratory behavior. The neurons releasing these agents seem to be activated by salient stimuli, and the activation seems related to the values of stimuli used for predicting future behavior. However, dopamine and noradrenaline appear to be related to different functions, with dopamine being related to assessment of rewards and noradrenaline being related to arousal and/or attention suggesting that their roles in motivated behavior are different. The different roles may be related to the multiple influences of a reward. Rewards are defined by their appetitive and reinforcing influence on behavior. In addition, they are energizing, and enhance arousal and attention. These different roles are related but not completely overlapping. To improve our knowledge about the similarities and differences between LC and DA neurons, we have carried out a direct comparison of the similarities and differences in firing characteristics under a single set of arousing and rewarding. Specifically, we compare the activity of noradrenergic neurons from the locus coeruleus (LC) and dopaminergic neurons from the substancia nigra pars compacta (SNc) in different monkeys performing the same reward schedule task. Monkeys rapidly adjust their operant performance as a function of the progression through the schedules.. The operant performance across trials of the schedules reflects the monkeys motivation given the current trials contingency, so we use errors in the trials to quantify the operant performance and infer the predicted outcome value in a given trial. We also measure an appetitive Pavlovian response (lipping) and interpret this as reflecting the intrinsic value of salient task events such as cue onset or bar release. The results show that the times at which neuronal responses of SNc and LC neurons occur within this task are similar, in that they occur around the salient events that evoke lipping. However, the modulations of these responses as a function of motivational levels appear to reflect the predicted outcome value for the dopamine neurons and the the predicted cost to obtain the reward for the locus coeruleus neurons. Thus, although these two neuromodulatory systems respond to salient stimuli, their influences on target neurons will also be substantially different. One major obstacle for studying relationships between neuronal activity and behavior in nonhuman primates is the limited selectivity and/or temporal control of current methods such as ablation, pharmacology and electrical stimulation. Recently, new tools, based on molecular biology, have emerged - such as optogenetics, CNO activated DREADDs and ivermectin activated chloride channels. These tools allow, at least in principle, inducible and reversible silencing or activation of neuronal activity and can be genetically targeted into selected neuron populations with cellular resolution. However, because of the size and anatomy of the nonhuman primate brain, the biggest challenge for applying these methods in the monkey is how to target them into enough cells of a specific neuron population in order to achieve a behavioral effect. We have compared parameters for injection of adeno-associated (AAV) viruses and lentiviruses into cortical structures. From this work, we find that slow injection of AAV constructs provides good spread across cortical layers, whereas rapid injection of lentivirus constructs is needed to get a similar effect. We have shown that we can detect GFP co-expressing antisense RNA against the dopamine DRD2 receptor after injecting plasmid DNA into the rhinal cortex. Finally, we have also developed a lentivirus expressing GFP from a tyrosine hydroxylase promoter fragment targets dopaminergic and noradrenergic neurons in the substantia nigra and the locus coeruleus respectively. Our ability to target of such neuron populations genetically and verify targeting, if applied successfully, can allow controllable manipulation of neuronal activity while also monitoring correlated physiological and behavioral effects, therefore bridging the gap between cellular activity and behavior.
Showing the most recent 10 out of 18 publications
