The role of mPFC and MTL neurons in human choice under goal-conflict

Resolving approach-avoidance conflicts relies on encoding motivation outcomes and learning from past experiences. Accumulating evidence points to the role of the Medial Temporal Lobe (MTL) and Medial Prefrontal Cortex (mPFC) in these processes, but their differential contributions have not been convincingly deciphered in humans. We detect 310 neurons from mPFC and MTL from patients with epilepsy undergoing intracranial recordings and participating in a goal-conflict task where rewards and punishments could be controlled or not. mPFC neurons are more selective to punishments than rewards when controlled. However, only MTL firing following punishment is linked to a lower probability for subsequent approach behavior. mPFC response to punishment precedes a similar MTL response and affects subsequent behavior via an interaction with MTL firing. We thus propose a model where approach-avoidance conflict resolution in humans depends on outcome value tagging in mPFC neurons influencing encoding of such value in MTL to affect subsequent choice.


Supplementary Data
Supplementary

Supplementary Note 1: Neural response to cue
The number of cue trials differentiate between conditions. To account for this, we randomly selected some of the control trials to match the amount of the uncontrol trials, resulting in an equal amount of trials for all conditions. While this accounts for the statistical biases it does not account for cognitive bias as participants comprehend uncontrol conditions as more rare and thus a saliency effect may be manifested in the results.
We found 23 of 79 (29%), 27 of 61 (44%), 20 of 63 (32%) and 40 of 107 (37%) neurons that significantly responded to at least one of the four outcome conditions in the Amygdala, Hippocampus, dmPFC, CC respectively. We interperate these results in the context of saliency. As uncontrol trials are more salient compared to control trials, they evoke a stronger neural response. Similarly, as control reward trials are more salient compared to control punishment trials, they also evoke a stronger neural response. Interestingly, both these effects are strongest and only significant in the amygdala, a major hub of the brain's salience network [Seeley et al., 2007].

Supplementary note 2: Effect of Seizure Onset Zone (SOZ)
Of the 14 patients, 5 has SOZ in the MTL and 1 had an SOZ in the mPFC. This included 29 neurons in the MTL (13 amygdala, 16 hippocampus) and no neurons in the mPFC that were within the SOZ. 31 amygdala neurons were responsive to at least one of the four outcome conditions and 5 of these were within the SOZ. 26 neurons were responsive to at least on of the four conditions and 6 of these were within the SOZ. Removing these neurons from the analysis did not effect results.

Outcome response probability
While mPFC showed a clear preference to control reward over control punishment (17 vs. 4 in the dmPFC and 20 vs. 10 in the CC), MTL neurons were unbiased (10 vs. 9 in the amygdala and 5 vs. 7 in the hippocampus).

Outcome response amplitude
Following the removal of SOZ neurons (3 neurons), ANOVA analysis of neurons with increased firing in one of the four conditions found similar results to those with these neurons: A repeated measures ANOVA with normalized FR increase following outcome (200-800msec) as the dependent variable and region-groups [MTL, mPFC], controllability (controlled/uncontrolled) and outcome-valence (reward/punishment) as the independent factors, revealed a greater response to controlled negative outcomes, specifically in the mPFC region group (

Effect of outcome response on subsequent behavior in HGC trials
after removing MTL neurons (2 responsive neurons from the left amygdala of patient 6) that were within the seizure onset zone this remained significant. We found that only MTL firing following punishment outcomes significantly correlated with behavior in subsequent HGC reward trials [beta=1.2, t=4.3, p<0.0001, FDR corrected], even after accounting for movement and time between punishment and subsequent HGC trials. Even after removing MTL neurons (2 responsive neurons from the left amygdala of patient 6) that were within the seizure onset zone this remained significant.
Comparing behavior to healthy controls 20 healthy participants (15 Females, age 32.9±3.7) performed the PRIMO task with a laptop in laboratory room as volunteers and did not receive money for performing the task. As expected, approach probability in this group was higher for the LGC