Orbitofrontal cortex inactivation impairs between- but not within-session Pavlovian extinction: An associative analysis

Abstract

The orbitofrontal cortex (OFC) is argued to be the neural locus of Pavlovian outcome expectancies. Reinforcement learning theories argue that extinction learning in Pavlovian procedures is caused by the discrepancy between the expected value of the outcome (US) that is elicited by a predictive stimulus (CS), and the lack of experienced US. If the OFC represents Pavlovian outcome expectancies that are necessary for extinction learning, then disrupting OFC function prior to extinction training should impair extinction learning. This was tested. In experiment 1, Long Evans rats received infusions of saline or muscimol targeting the lateral OFC prior to three appetitive Pavlovian extinction sessions. Muscimol infused into the OFC disrupted between-session but not within-session extinction behaviour. This finding was not due to muscimol infusions disrupting the memory consolidation process per se as there was no effect of muscimol infusion when administered immediately post session (experiment 2). These findings support a role for the OFC in representing outcome expectancies that are necessary for learning. A number of ways in which disrupting outcome expectancy information might block learning will be discussed in the context of traditional associative learning theories and the associative structures they depend on.

Introduction

Extinction learning procedures form the basis of a number of animal models of psychopathology and behavioural therapy (Huston et al., 2013, Kaplan et al., 2011, Laborda et al., 2012). Similarly, learning and performance in these animal models and in extinction depend on limbic, hippocampal and prefrontal cortical structures (Delamater, 2004, Herry et al., 2010, Sah and Westbrook, 2008). In simple Pavlovian extinction procedures a cue (CS) that reliably predicts an outcome (US; e.g. the delivery of food) is presented in the absence of the US. As extinction proceeds, conditioned responses (CRs) to the CS decline and new extinction learning occurs (Bouton, 1993, Delamater, 2004, Rescorla, 2004). It is therefore important to specify the exact role that these neural regions might play in supporting the associative structures underlying Pavlovian extinction.

Reinforcement learning theories argue that extinction learning in Pavlovian procedures is caused by the discrepancy between the expected value of the US that is elicited by the CS, and the lack of experienced US (Rescorla and Wagner, 1972, Sutton and Barto, 1998). This discrepancy between expected and actual outcomes generates a prediction error signal that drives learning. The rodent orbitofrontal cortex (OFC) has been proposed as a neural site for representing US expectancy information (Delamater, 2007, Ostlund and Balleine, 2007, Schoenbaum et al., 2011). For example, in an odour discrimination task single unit neural firing increases in the period after CS presentation in anticipation of the US (Roesch, Calu, Burke, & Schoenbaum, 2007). It is therefore likely that OFC activity is required for prediction error learning in extinction. This is in line with other evidence that OFC lesions and inactivation impair reversal learning, a procedure that involves extinction learning (Boulougouris et al., 2007, Butter, 1969, Rudebeck and Murray, 2008, Schoenbaum et al., 2003). The non-human primate OFC is also implicated in extinction (Izquierdo and Murray, 2005, Reekie et al., 2008).

Recently, Burke, Takahashi, Correll, Brown, and Schoenbaum (2009) found that inactivation of rodent OFC did not impair extinction learning in a Pavlovian learning task. This finding is at odds with the hypothesised role of the OFC in representing Pavlovian US expectancy. However, the rats used in this experiment had previously been exposed to extinction of the cues that were used before inactivation (Takahashi et al., 2009). Furthermore, while deficits in reversal learning is the hallmark impairment associated with OFC (Boulougouris et al., 2007), this impairment is abolished when animals are exposed to reversal learning prior to OFC damage (Boulougouris & Robbins, 2009). Therefore, Burke et al.’s (2009) failure to find an effect of OFC inactivation in extinction may well have been due to the animals having pre-exposure to extinction. The re-expression of extinction may not require prediction error learning (Hart, Harris, & Westbrook, 2010) and hence may be independent of outcome expectancy and the OFC.

The first experiment sought to establish the role of the OFC in Pavlovian extinction. Pavlovian extinction learning has been shown to be dependent on prediction error signal (Leung and Westbrook, 2008, Steinberg et al., 2013). Prediction error signalling in dopamine neurons has been found to be dependent on outcome expectancy information from the OFC (Takahashi et al., 2011). Together, these findings suggest that the OFC should be necessary for prediction error learning in Pavlovian extinction.

We conducted two experiments to directly assess the role of the rodent OFC in simple Pavlovian extinction learning. Both experiments used an appetitive Pavlovian procedure in which animals initially acquired an auditory click CS – food pellet US association. Cannulae targeting OFC were then implanted, and OFC function was inactivated (via micro-infusions of the GABA_A agonist muscimol) prior to three sessions of extinction training (experiment 1). Experiment 2 replicated the procedures in experiment 1 except that post-session OFC inactivation was used to determine if the results of experiment 1 were due to disruption of memory consolidation processes.