Danger and safety signals independently influence persistent pathological avoidance in anxiety-vulnerable Wistar Kyoto rats: A role for impaired configural learning in anxiety vulnerability

Early Safety Discrimination Under Uncertainty in Trait Anxiety: An Event-Related Potential Study

Detection of safety-threat signals during uncertainty is an important mechanism of developmental anxiety disorder (AD). Although extensive research has focused on the detection of uncertain threat signals in anxious individuals, relatively little attention has been given to the identification of safety signals during uncertainty, which is an important way to relieve anxiety in individuals with AD. To investigate this phenomenon, 16 subjects with high trait anxiety (HTA) and 16 with low trait anxiety (LTA) completed a modified cue-target task in certain and uncertain stimulus blocks. In the uncertain block, the cue was followed by a threat picture or safety picture in 20% of trials, respectively; in the certain block, the cue could be followed by a threat picture or a safety picture on 100% of trials. Behavioral responses and event-related potentials (ERPs) were recorded. The ERP results demonstrated that LTA participants exhibited larger P2 amplitudes in the detection of safety cues than of threat cues during the uncertain block, whereas HTA participants showed significant P2 amplitudes between the safety and threat cues during the certain block, impairing the detection of safety stimuli during uncertainty. However, all participants exhibited greater N2 amplitudes following threat cues in certainty or uncertainty conditions. These findings pertaining to the P2 amplitude suggested distinctive attentional biases between HTA and LTA individuals, whereas the N2 amplitude showed association learning in uncertain conditions, compensating for safety-threat detection in HTA individuals.

Dataset of active avoidance in Wistar-Kyoto and Sprague Dawley rats: Experimental data and reinforcement learning model code and output

Data were collected from 40 Wistar-Kyoto (WKY) and 40 Sprague Dawley (SD) rats during an active escape-avoidance experiment. Footshock could be avoided by pressing a lever during a danger period prior to onset of shock. If avoidance did not occur, a series of footshocks was administered, and the rat could press a lever to escape (terminate shocks). For each animal, data were simplified to the presence or absence of lever press and stimuli in each 12-second time frame. Using the pre-processed dataset, a reinforcement learning (RL) model, based on an actor-critic architecture, was utilized to estimate several different model parameters that best characterized each rat's behaviour during the experiment. Once individual model parameters were determined for all 80 rats, behavioural recovery simulations were run using the RL model with each animal's "best-fit" parameters; the simulated behaviour generated avoidance data (percent of trials avoided during a given experimental session) that could be compared across simulated rats, as is customarily done with empirical data. The datasets representing both the experimental data and the model-generated data can be interpreted in various ways to gain further insight into rat behaviour during avoidance and escape learning. Furthermore, the estimated parameters for each individual rat can be compared across groups. Thus, possible between-strain differences in model parameters can be detected, which might provide insights into strain differences in learning. The software implementing the RL model can also be applied to or serve as a template for other experiments involving acquisition learning. Reference for Co-Submission: K.M. Spiegler, J. Palmieri, K.C.H. Pang, C.E. Myers, A reinforcement-learning model of active avoidance behavior: Differences between Sprague-Dawley and Wistar-Kyoto rats. Behav. Brain Res. (2020 Jun 22[epub ahead of print])  doi: 10.1016/j.bbr.2020.112784.

A reinforcement-learning model of active avoidance behavior: Differences between Sprague Dawley and Wistar-Kyoto rats

Avoidance behavior is a typically adaptive response performed by an organism to avert harmful situations. Individuals differ remarkably in their tendency to acquire and perform new avoidance behaviors, as seen in anxiety disorders where avoidance becomes pervasive and inappropriate. In rodent models of avoidance, the inbred Wistar-Kyoto (WKY) rat demonstrates increased learning and expression of avoidance compared to the outbred Sprague Dawley (SD) rat. However, underlying mechanisms that contribute to these differences are unclear. Computational modeling techniques can help identify factors that may not be easily decipherable from behavioral data alone. Here, we utilize a reinforcement learning (RL) model approach to better understand strain differences in avoidance behavior. An actor-critic model, with separate learning rates for action selection (in the actor) and state evaluation (in the critic), was applied to individual data of avoidance acquisition from a large cohort of WKY and SD rats. Latent parameters were extracted, such as learning rate and subjective reinforcement value of foot shock, that were then compared across groups. The RL model was able to accurately represent WKY and SD avoidance behavior, demonstrating that the model could simulate individual performance. The model determined that the perceived negative value of foot shock was significantly higher in WKY than SD rats, whereas learning rate in the actor was lower in WKY than SD rats. These findings demonstrate the utility of computational modeling in identifying underlying processes that could promote strain differences in behavioral performance.