Direct stimulation of anterior insula and ventromedial prefrontal cortex disrupts economic choices

The neural mechanisms underlying risky decision-making remain poorly understood, despite the widely accepted influence of the desirability of potential outcomes. Functional imaging and intracranial electrophysiology studies implicate the ventromedial prefrontal cortex (vmPFC) and anterior insular cortex (aINS) as crucial in shaping these decisions. Increased pre-stimulus activity in the vmPFC has been associated with increased risk-taking (overweighting potential gains), while increased aINS activity is linked to decreased risk-taking (overweighting potential losses). However, these studies are correlational, preventing conclusive establishment of a causal link. Further research highlights the involvement of both vmPFC and aINS in confidence judgments, suggesting a possible interplay between value and confidence neural codes. Previous causal studies have been limited due to challenges in precisely targeting these deep brain regions using non-invasive methods, inconsistent results from lesion studies, and a scarcity of intracranial electrical stimulation (IES) studies combining IES with cognitive paradigms to dissect decision-making processes. This study aimed to address these limitations by using IES in humans to investigate the causal role of vmPFC and aINS in risky decision-making and confidence judgments.

Existing literature suggests opposing roles for the vmPFC and aINS in economic decisions, mood, and learning. vmPFC activity is associated with increased risk-seeking, pleasantness, reward-based learning, and positive mood, while aINS activity is linked to risk aversion, unpleasantness, punishment avoidance, and negative mood. Studies using functional imaging and electroencephalography (EEG) have shown correlations between neural activity in these regions and decision-making, but these methods lack the causal power to definitively establish a direct link. Lesion studies have yielded inconsistent results, likely due to variations in lesion size and location. Previous IES studies have rarely focused on these regions or haven't combined IES with appropriate cognitive tasks. This study builds upon this existing research by using IES with high spatial resolution to target specific subregions within the vmPFC and aINS and investigate their causal impact on decision-making and confidence.

Fifteen patients with epilepsy implanted with intracerebral electrodes for pre-surgical evaluation participated. Intracranial electrical stimulation (IES) was applied to the aINS (38 sites in 13 participants) and vmPFC (16 sites in 9 participants). Participants performed an accept/reject task involving choices with varying gain and loss prospects and a subsequent challenge with varying difficulty. Before each challenge, participants provided confidence ratings. The IES was applied for 5 seconds, starting 1 second before the trial onset. The stimulation intensity was chosen based on clinical observations to be the highest intensity without inducing any clinical symptoms. A computational model based on expected utility theory was used to analyze the effects of IES on choice behavior, focusing on sensitivity to potential gains and losses. Anatomical landmarks were used to further define the dorsal and ventral subregions within the aINS and vmPFC. The location of stimulation sites was confirmed by anatomical labeling by co-registering pre-operative MRI with postoperative CT scan using IntrAnat Electrodes software. Data from 19 healthy participants performing the same task was used for comparative analysis.

IES effects on decision-making varied depending on the stimulation location within the aINS and vmPFC. Ventral aINS stimulation decreased risk-taking by increasing sensitivity to potential losses, whereas dorsal aINS and ventral vmPFC stimulation increased risk-taking by decreasing loss sensitivity. This dorso-ventral gradient was consistent across both statistical and anatomically defined subregions. The opponent effects of IES on choice in the dorsal versus ventral clusters were not influenced by the magnitude of gains or losses. K-means clustering and anatomical landmarks (Destrieux's atlas and the Julich Brain Atlas) were used to define the dorsal and ventral subregions, producing highly similar results across the two methods, demonstrating the robustness of the findings. Importantly, applying IES to these anatomically defined subregions of the alns differentially affected participants' choices in the same directions as in the previous analysis. aINS stimulation consistently decreased participants' confidence, regardless of stimulation location. vmPFC stimulation did not significantly affect confidence levels. Computational modeling revealed that IES modulated participants' sensitivity to potential losses but not gains, consistent with the observed effects on choice behavior.

The findings support the causal involvement of functionally distinct subregions within the aINS and vmPFC in decision-making and confidence. The opposing effects of stimulation on risk-taking in the dorsal and ventral aINS align with preclinical evidence demonstrating distinct areas governing approach and avoidance behaviors. The results also corroborate stimulation studies in non-human primates showing that ventral aINS stimulation reduces approach behaviors and is associated with negative feelings, while dorsal aINS stimulation is linked to positive behaviors or feelings. The consistent decrease in confidence with aINS stimulation provides causal evidence for its role in metacognitive judgments, aligning with previous correlational studies. The lack of a significant effect of vmPFC stimulation on confidence is surprising given previous research, and may reflect greater variability in metacognitive judgments or the temporal delay between stimulation and confidence ratings. The similar effects of IES on choice behavior and loss sensitivity in both vmPFC and dorsal aINS suggest that these effects could be mediated by distinct cognitive processes. For example, the vmPFC's involvement in self-referential episodic memory retrieval might explain the increase in risk-taking after vmPFC stimulation through reduced sensitivity to emotional markers of self-referential risk. Discrepancies between this study and previous invasive electrophysiology studies highlight the potential influence of statistical power and the complexity of investigating the effects of stimulation on functional connectivity.

This study provides causal evidence for the involvement of functionally dissociated subregions within the vmPFC and aINS in value-based decision-making and confidence. The findings demonstrate a dorso-ventral gradient of effects within these regions, with ventral aINS promoting risk aversion and dorsal aINS promoting risk-seeking. Future research should focus on further characterizing the functional roles of vmPFC subregions and investigate the network-level mechanisms underlying the effects of stimulation. These findings have implications for understanding neuropsychiatric disorders involving these brain regions and could guide the development of closed-loop brain stimulation therapies.

The study's reliance on patients with epilepsy could introduce biases, although participants' behavior mirrored that of healthy individuals and epileptic tissue can generate normal neuronal activity during cognitive tasks. The limited number of trials and low spatial sampling could have affected statistical power and the investigation of network-level mechanisms. The lack of sufficient data for the dorsal part of vmPFC also limited detailed investigation of this sub-region.