Dissociable Fronto-Operculum-Insula Control Signals for Anticipation and Detection of Inhibitory Sensory Cue

Abstract

The ability to anticipate and detect behaviorally salient stimuli is important for virtually all adaptive behaviors, including inhibitory control that requires the withholding of prepotent responses when instructed by external cues. Although right fronto-operculum-insula (FOI), encompassing the anterior insular cortex (rAI) and inferior frontal cortex (rIFC), involvement in inhibitory control is well established, little is known about signaling mechanisms underlying their differential roles in detection and anticipation of salient inhibitory cues. Here we use 2 independent functional magnetic resonance imaging data sets to investigate dynamic causal interactions of the rAI and rIFC, with sensory cortex during detection and anticipation of inhibitory cues. Across 2 different experiments involving auditory and visual inhibitory cues, we demonstrate that primary sensory cortex has a stronger causal influence on rAI than on rIFC, suggesting a greater role for the rAI in detection of salient inhibitory cues. Crucially, a Bayesian prediction model of subjective trial-by-trial changes in inhibitory cue anticipation revealed that the strength of causal influences from rIFC to rAI increased significantly on trials in which participants had higher anticipation of inhibitory cues. Together, these results demonstrate the dissociable bottom-up and top-down roles of distinct FOI regions in detection and anticipation of behaviorally salient cues across multiple sensory modalities.

Keywords: dynamic causal modeling; fMRI; human; prefrontal cortex; response inhibition.

Figure 1.

Task design and brain activations. (a) SST paradigms SST1 and SST2. The design of the 2 tasks is similar except that an auditory stop cue was used in SST1, while a visual stop cue was used in SST2. (b) Increased activation of FOI, sensory and other cortical regions during successful stop trials (SuccStop) compared with Go trials during SST1 and SST2. (c) Regions of interest (ROIs), based on SST1 and SST2 task activations, which were used in the DCM analysis.

Figure 2.

Dynamic causal interactions of the FOI during detection of inhibitory cues. (a, b) In SST1, the strength of causal interactions from the auditory cortex was significantly greater to the AI, compared with the IFC. (c, d) In SST2, the strength of causal interactions from the visual cortex was significantly greater to the AI, compared with the IFC. *P < 0.05; **P < 0.01. SST1, Stop-signal task 1; SST2, Stop-signal task 2; r, right. Line and arrow width is set as exponential power of estimated weight.

Figure 3.

Bayesian prediction of behavior and brain activation associated with anticipation of inhibitory cues. (a) Participants showed faster RTs on Go and unsuccessful stop trials (UnsuccStop) with low, compared with high, probability of expecting an inhibitory cue (low-p_stop vs. high-p_stop; “left panel”). Participants had higher stop-signal expectation (“middle panels"), and had shorter SSD times on successful, compared with unsuccessful stop trials (SuccStop vs. UnsuccStop; “right panel”). (b) Bilateral activation of AI and IFC associated with successfully stopping on trials with low versus high probability of expecting an inhibitory cue (“left panel”), as well as unsuccessful, compared with successful stop trials (UnsuccStop vs. SuccStop, “right panel”). (c) Activation levels in unbiased AI and IFC ROIs from a previous study of right FOI parcellation (Cai et al., 2014). **P < 0.01; ***P < 0.001.

Figure 4.

Dynamic causal interactions of the FOI associated with anticipation of inhibitory cues. (a) Causal interactions associated with successfully stopping on trials with low and high probability of anticipating inhibitory cues (SuccStop-low-p_stop, SuccStop-high-p_stop). (b) The strength of casual interactions from the IFC to AI was significantly greater during trials with high, compared with low, probability of anticipating inhibitory cues (low-p_stop vs high-p_stop). *P < 0.05; **P < 0.01. Line and arrow width is set as exponential power of estimated weight.