Developing and Testing a Neural Model of Conditioned Fear Generalization

Abstract

Fear overgeneralization is a promising pathogenic mechanism of anxiety-related disorders. Despite the ubiquity of fear generalization in human experience and its potential pathogenic contribution to clinical anxiety, neural investigations of human generalization have only recently begun. The present work provides the first meta-analysis of this growing literature to delineate brain substrates of conditioned fear-generalization and formulate a working neural model. Included studies (K = 6, N = 176) reported whole-brain fMRI results and applied generalizationgradient methodology to identify brain activations that gradually strengthen (positive generalization) or weaken (negative generalization) as presented stimuli increase in CS+ resemblance. Positive generalization was instantiated in cingulo-opercular, frontoparietal, striatal-thalamic, and midbrain regions (locus coeruleus, periaqueductal grey, ventral tegmental area), while negative generalization was implemented in default-mode network nodes (ventromedial prefrontal cortex, hippocampus, middle temporal gyrus, angular gyrus) and amygdala. Findings are integrated within an updated neural account of generalization centering on the hippocampus. This model posits that the hippocampus matches presented cues against the stored representation of the CS+. Cues with sufficient CS+ overlap trigger pattern completion and activation of positive generalization/fear excitatory circuits; cues with insufficient CS+ overlap trigger pattern separation and activation of negative generalization/fear inhibitory circuits. According to this model, overgeneralization occurs when the hippocampus fails to distinctly encode benign stimuli with low similarity to previously encountered threat cues, triggering excessive retrieval of stored threat representations. To causally test this model, we applied hippocampal network targeted transcranial magnetic stimulation (HNT-TMS), a protocol that has been shown to modulate hippocampal activation, connectivity, and hippocampal-iii dependent neutral memory encoding. Consistent with our neural model and previous HNT-TMS studies, we hypothesized that HNT-TMS would strengthen encoding of perceptually similar stimuli and thereby reduce retrieval errors (i.e., sharpen discrimination) in participants with posttraumatic stress symptoms (PTSS). We predicted that this effect would emerge for fear/fear-like stimuli as measured by the Farmer Task and neutral stimuli as measured by the Mnemonic Similarity Task. Continuous theta-burst stimulation (cTBS) was delivered to individualized left posterior-parietal targets derived via precision functional mapping, seed-based connectivity, and electric-field modeling methods. A vertex control target was also stimulated in a within-subject design (final N = 25). Multilevel models did not reveal significant effects of HNT-TMS on fear or neutral stimulus discrimination. However, HNT-TMS strengthened fear discrimination in participants with lower fear sensitization, indexed by responsivity to a control stimulus perceptually unrelated to the CS+. Sensitization reflects indiscriminate fear responding unrelated to CS+ similarity and is not expected to engage the hippocampal CS+ matching function. Our findings indicate that HNT-TMS may selectively sharpen fear discrimination when the hippocampal CS+ matching function is strongly engaged. These findings provide early causal support for our hippocampal centered model of fear generalization and motivate additional investigation in future studies that examine the capacity of HNT stimulation to produce stronger effects in samples stratified prior to stimulation by putative schematic matching engagement.