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

Abstract

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.

Fig. 1. The paradigm.

The goal of the game was to earn virtual money by catching shekel signs and avoiding balls. A small avatar on a skateboard was located at the bottom of the screen and subjects had to move the avatar right and left using right and left arrow keys, in order to catch the money and avoid the balls falling from the top of the screen. There were two ways to gain or lose money—a “controlled” condition, where players actively approached green money signs (marked here as dollar signs) and avoided red balls, and an “uncontrolled” condition, where although cues appeared on the top of the screen (reward—cyan dollar sign, punishment—magenta ball), they always hit the avatar with no relation to the players’ action (they chase the avatar during their fall). Each money catch resulted in a five-point gain and each ball hit resulted in a loss of five points, regardless of controllability (the outcome was shown on the screen after each trial). All four outcome event types occurred roughly at the same frequency, adaptive to the player’s behavior. Each money trial was separated by a jittered interstimulus interval (ISI), which varied randomly between 550 and 2050 ms.

Fig. 2. Example neural responses to the different outcome conditions.

a Sagittal slices show the location of active electrode contacts in mPFC and MTL areas (yellow markers, registered to an MNI atlas) that refer to the raster plots and peri-stimulus time histograms (PSTH), per condition (see legend for color codes). b, c PSTH from two different amygdala neurons showing significant increase in firing following reward outcome in both controlled and uncontrolled conditions. d–f PSTH from three different mPFC neurons showing significant increase in firing following punishment outcome in controlled but not in uncontrolled condition. Time 0 on the x-axis represents the timing of outcome (coin or ball hit the avatar). FR firing rate.

Fig. 3. Neurons’ response probability in different regions and outcome conditions.

Percent of neurons per region presenting a significant change in firing rate (FR) between 200 and 800 ms in response to: a Controlled (black) or Uncontrolled (gray) outcomes (across valence); N = 79, 61, 63, and 107 independently sampled neurons for the amygdala, hippocampus, dmPFC, and CC, respectively. A two-sided McNemar’s exact test found effects at p = 0.088, p = 0.011, and p = 0.049 for the amygdala, dmPFC, and CC, respectively, FDR corrected. Asterisks represent significant at p < 0.05. b Reward (black) or Punishment (gray) outcomes (across controllability); N is similar to (a). c Controlled rewards or punishments (N = 93 independenty sampled neurons from four or two brain regions, p = 0.065 and p = 0.014 using χ² test, respectively). Diamond denotes significant valence preference between MTL and mPFC at p < 0.05. d Uncontrolled rewards or punishments (N = 39 independenty sampled neurons from four or two brain regions, p = 0.97 and p = 0.99 using χ² test, respectively). Amy amygdala, Hip hippocampus, dmPFC dorsomedial prefrontal cortex, CC cingulate cortex.

Fig. 4. Selectivity of neural time-course responses per outcome types.

Average normalized FR for positively responsive neurons shown for (a) each of the four recording regions N = 12, 9, 11, and 22 for Amygdala, Hippocampus, dmPFC, and CC, respectively. b Combined regions in MTL and mPFC groups, N = 24 MTL and 34 mPFC neurons. Shaded area corresponds to standard error of mean (SEM). Source data are provided as a Source Data file. c Responsivity profile projected on two sagittal atlas slices for mPFC (upper panel) and MTL (lower panel) region groups. Coloring is according to the averaged normalized FR change for each condition (coloring key is presented in lower square). Circle size corresponds to the number of neurons from each contact groups, from 1 (smallest) to 5 (largest) (see key). Time 0 on the x-axis represents the timing of outcome (coin or ball hit the avatar). Amy amygdala, Hip hippocampus, dmPFC dorsomedial prefrontal cortex, CC cingulate cortex, MTL medial temporal lobe, mPFC medial prefrontal cortex, STDs standard deviations, Rew reward, Pun punishment, Uncon uncontrol, Con control.

Fig. 5. Neural firing change and their effect on subsequent behavioral choice.

a Time courses for mean normalized FR change in mPFC (brown trace) and MTL (blue trace) sites for each outcome type. Asterisks mark for which 200 ms window the FR is significantly above baseline (two-sided signed rank test, p < 0.05 FDR corrected, Control Reward N = 14 and 9 for MTL and mPFC, respectively). Control Punishment N = 12 and 29 for MTL and mPFC, respectively. Uncontrol Reward N = 8 and 8 for MTL and mPFC, respectively. Uncontrol Punishment N = 8 and 2 for MTL and mPFC, respectively. Note that for both controlled outcomes mPFC neurons fire slightly before MTL neurons (left) but not for uncontrolled outcomes (right). Time 0 on the x-axis represents the timing of outcome (coin or ball hit the avatar). Shaded area corresponds to SEM. Source data are provided as a Source Data file. b, c The effect of neural responses following controlled outcomes on subsequent behavioral choice under HGC condition. b The mean probability for approaching a coin, subsequent to trials where a neuron fired 200–800 ms following a controlled outcome (black bars) vs. trials where a neuron did not fire (white bars), shown for MTL (b, top) and mPFC (b, bottom) neuron groups per outcome type. Note that only MTL neuron showed a consistent pattern of subsequent behavior of less approach after punishment outcome (b, top right). c Approach probability change following neural firing to controlled rewards (green markers) and controlled punishments (red markers) in MTL (c, top) and mPFC (c, bottom). c (top) Asterisk denotes a significant two-sided Mann–Whitney test, p = 0.01, FDR corrected, N = 21 neurons (10 punishment, 11 reward). STDs standard deviations, Delta Prob difference in probability.