The authors used a within-subject blocking design to study the role of ventrolateral periaqueductal gray (v1PAG) opioid receptors in regulating prediction errors during Pavlovian fear conditioning. In Stage I, the authors trained rats to fear conditioned stimulus (CS) A by pairing it with shock. In Stage II, CSA and CSB were co-presented and followed with shock. Two novel stimuli, CSC and CSD, were also co-presented and followed with shock in Stage II. CSA blocked fear from accruing to CSB. Blocking was prevented by systemic pretreatment with naloxone. Blocking was also prevented in a dose-dependent and neuroanatomically specific fashion by vlPAG infusions of the micro-opioid receptor antagonist CTAP. These experiments show that v1PAG micro-opioid receptors contribute to Pavlovian fear learning by regulating predictive error.

Pavlovian fear conditioning depends on prediction error, or the discrepancy between actual and expected outcomes. We used immunohistochemistry, neuronal tract tracing, and reversible inactivation to study the role of prefrontal cortex and thalamocortical pathways in predictive fear learning. Unexpected, but not expected, conditioned stimulus (CS)-unconditioned stimulus (US) presentations caused increased c-Fos expression in the prefrontal cortex (PFC), midline thalamus, lateral amygdala, as well as retrograde labeled midline thalamic afferents to PFC. Reversible inactivation of dorsomedial PFC, but not infralimbic PFC, prevented the associative blocking of fear learning. These results suggest a role for dorsomedial PFC (dmPFC), and a thalamic → dmPFC pathway, in signaling whether or not aversive events are expected or unexpected and so whether they are to be learned about.

Pavlovian fear conditioning is not a unitary process. At the neurobiological level multiple brain regions and neurotransmitters contribute to fear learning. At the behavioral level many variables contribute to fear learning including the physical salience of the events being learned about, the direction and magnitude of predictive error, and the rate at which these are learned about. These experiments used a serial compound conditioning design to determine the roles of basolateral amygdala (BLA) NMDA receptors and ventrolateral midbrain periaqueductal gray (vlPAG) m-opioid receptors (MOR) in predictive fear learning. Rats received a three-stage design, which arranged for both positive and negative prediction errors producing bidirectional changes in fear learning within the same subjects during the test stage. Intra-BLA infusion of the NR2B receptor antagonist Ifenprodil prevented all learning. In contrast, intra-vlPAG infusion of the MOR antagonist CTAP enhanced learning in response to positive predictive error but impaired learning in response to negative predictive error-a pattern similar to Hebbian learning and an indication that fear learning had been divorced from predictive error. These findings identify complementary but dissociable roles for amygdala NMDA receptors and vlPAG MOR in temporal-difference predictive fear learning. (Merluzzi et al. 1991;Arntz et al. 1993;Kozak et al. 2007).A fundamental question concerns how fear learning is distributed among these structures. Of particular interest in the present experiments were the specific roles of vlPAG MOR and BLA NMDA receptors in Pavlovian association formation, given evidence that vlPAG contributes to predictive error during fear learning and the well-established role for BLA in fear learning. A difficulty in answering this question is that fear learning is not a unitary process. Many variables contribute to fear learning, including the physical salience of the events being learned about, the direction and magnitude of predictive error, and so forth. Answering these questions requires isolation of these different variables. The Temporal-Difference (TD) model allows one such approach based on real-time processing of predictive error (Sutton 1988;Sutton and Barto 1990). The TD model has been used to explain learning-related activity in primate midbrain dopamine neurons (e.g., Schultz et al. 1997) and human fMRI BOLD signals (e.g., Seymour et al. 2004), and recently has been employed to examine fear learning (see Cole and McNally 2007a).A unique assumption of the TD model is that earlier predictors of an outcome are more informative, and so are learned about at the expense of later predictors (Sutton 1988;Sutton and Barto 1990). Therefore, a well-trained CS may undergo a decrement in responding if it is subsequently arranged to precede the US more closely than a second, neutral stimulus, which in turn, should undergo an increment in responding (Sutton and Barto 1981;Kehoe et al. 1987;Jennings and Kirkpatrick 2006). A threestage blocking-unblocking des...

The orexin/hypocretin system is important for reward-seeking behaviors, however less is known about its function in non-homeostatic feeding. Environmental influences, particularly cues for food can stimulate feeding in the absence of hunger and lead to maladaptive overeating behavior. The key components of the neural network that mediates this cue-induced overeating in sated rats include lateral hypothalamus, amygdala, and medial prefrontal cortex (mPFC), yet the neuropharmacological mechanisms within this network remain unknown. The current study investigated a causal role for orexin in cue-driven feeding, and examined the neural substrates through which orexin mediates this effect. Systemic administration of the orexin-1 receptor (OX1R) antagonist SB-334867 had no effect on baseline eating, but significantly reduced cue-driven consumption in sated rats. Complementary neural analysis revealed that decreased cue-induced feeding under SB-334867 increased Fos expression in mPFC and paraventricular thalamus. These results demonstrate that OX1R signaling critically regulates cue-induced feeding, and suggest orexin is acting through prefrontal cortical and thalamic sites to drive eating in the absence of hunger. These findings inform our understanding of how food-associated cues override signals from the body to promote overeating, and indicate OX1R antagonism as a potential pharmacologic target for treatment of disordered eating in humans.

The amygdala, prefrontal cortex, striatum and other connected forebrain areas are important for reward-associated learning and subsequent behaviors. How these structurally and functionally dissociable regions are recruited during initial learning, however, is unclear. Recently, we showed amygdalar nuclei were differentially recruited across different stages of cue-food associations in a Pavlovian conditioning paradigm. Here, we systematically examined Fos induction in the forebrain, including areas associated with the amygdala, during early (day 1) and late (day 10) training sessions of cue-food conditioning. During training, rats in the conditioned group received tone-food pairings, while controls received presentations of the tone alone in the conditioning chamber followed by food delivery in their home cage. We found that a small subset of telencephalic and hypothalamic regions were differentially recruited during the early and late stages of training, suggesting evidence of learning induced plasticity. Initial tone-food pairings recruited solely the amygdala, while late tone-food pairings came to induce Fos in distinct areas within the medial and lateral prefrontal cortex, the dorsal striatum, and the hypothalamus (lateral hypothalamus and paraventricular nucleus). Furthermore, within the perifornical lateral hypothalamus, tone-food pairings selectively recruited neurons that produce the orexigenic neuropeptide orexin/hypocretin. These data show a functional map of the forebrain areas recruited by appetitive associative learning and dependent on experience. These selectively activated regions include interconnected prefrontal, striatal, and hypothalamic regions that form a discrete but distributed network that is well placed to simultaneously inform cortical (cognitive) processing and behavioral (motivational) control during cue-food learning.