Regulation of social hierarchy learning by serotonin transporter availability

Abstract

Learning one's status in a group is a fundamental process in building social hierarchies. Although animal studies suggest that serotonin (5-HT) signaling modulates learning social hierarchies, direct evidence in humans is lacking. Here we determined the relationship between serotonin transporter (SERT) availability and brain systems engaged in learning social ranks combining computational approaches with simultaneous PET-fMRI acquisition in healthy males. We also investigated the link between SERT availability and brain activity in a non-social control condition involving learning the payoffs of slot machines. Learning social ranks was modulated by the dorsal raphe nucleus (DRN) 5-HT function. BOLD ventral striatal response, tracking the rank of opponents, decreased with DRN SERT levels. Moreover, this link was specific to the social learning task. These findings demonstrate that 5-HT plays an influence on the computations required to learn social ranks.

Fig. 1. Tasks and behavioral results.

A Social Dominance Hierarchy Learning task (left). Participants were led to believe they were competing against one of three real opponents. Unbeknownst to them, the probability they would win was predefined at P = 28%, 50%, and 72% for the superior, intermediate, and inferior opponents, respectively. In any one trial participants chose which one of two opponents they preferred to “compete” against. After competing in a perceptual decision-making task (circle with arrows), the outcome of the competition was delivered. For some trials (four per opponent), participants rated how confident they were of winning against the selected opponent (an example is shown in the bottom part of the panel). Note that the faces are derived from Todorov’s “25 Maximally Distinct Identities, Dominance” set. Middle: bar graphs represent the frequency with which they selected each opponent. Participants preferred to select the opponent against whom they had more chance of winning. Right: illustration of confidence rating through the social learning task. It is divided into the three reward probabilities (see Table S4 for more details on confidence rating). B Reinforcement Learning paradigm. Similar to the Social hierarchy learning task, participants chose which one of two slot machines among 3 they preferred to bet on (unknown winning probabilities: P = 28%, 50%, and 72% for the worst, intermediate and best chance to win, respectively). This was followed by an outcome phase in which they were informed if they had won or lost. In some trials, participants estimated how confident they were of winning on the selected slot machine (see bottom part of the panel). Middle: bar graphs represent the frequency that each slot machine was chosen. Right: illustration of the confidence rating through the non-social learning task. It is divided into the three reward probabilities. ***p < 0.001.

Fig. 2. Binding potential in the dorsal raphe nucleus modulates social learning and competitive behavior.

A Statistical map of the average BP_ND revealed a large distribution in the striatum and DRN. The white shape represents the DRN ROI. The bar graph represents the mean binding potential extracted in the DRN, ventral striatum left and right, amygdala left and right, and anterior PFC for illustration purposes. All extractions were performed using the ROI defined with the AAL3. B Left. Participants’ choice frequency during the social dominance learning task (dots) when facing the Inferior (green), Intermediate (orange), and Superior (red) opponents and model choice probability estimated by the RL algorithm. Right. Same illustration for the RL task. task. C Negative correlation between the BP_ND DRN, and participants’ learning rate in the social task. No correlation was observed between BP_ND, DRN and learning rate in the non-social task. D Competitive behavior related to the SERT level in the DRN. Competitive choices in the High and Low BP_ND groups. Individuals with lower BP_ND, in the DRN tended to increase their competitive choices, i.e. they chose to play against the stronger of the two opponents, in later trials. Interaction between trial bins and group (Low vs high BP_ND resulting from the median split) (F(_1,5) = 4.41, p = 0.037). Post-hoc tests conducted on the last bin revealed that the high BP_ND group made less competitive choices in the last bin of the task (Median rank = 18.87) compared to the low BP_ND group (Median rank = 12.13) (p = 0.036). The bar graphs show a between-groups difference in BP_ND level in the DRN based on a median split of individuals. Errors bars represent SEM. ***p < 0.001. BP_N non-displaceable binding potential, DRN dorsal raphe nucleus, RL reinforcement learning.

Fig. 3. Statistical maps of brain regions tracking the social dominance status of the opponent SDS(t) (bottom) and the social prediction error SPE(t) (top) at the outcome of the competition.

The graph on the left represents the evolution of the SDS(t) for a participant over the experiment. Positive encoding of the social prediction error SPE(t) was revealed in the bilateral VS, anterior mPFC, bilateral superior frontal gyrus, and posterior middle cingulate gyrus. Tracking SDS(t) engages the bilateral VS and the anterior mPFC. All statistical analyses were performed at a p < 0.05 cluster level corrected for Family Wise Error at the whole-brain level, with an initial cluster forming threshold of p < 0.001 uncorrected. VS ventral striatum, amPFC anterior medial prefrontal cortex.

Fig. 4. Negative correlation between SERT availability in the DRN and the BOLD response from the ventral striatum tracking opponents’ social dominance status at the outcome of the competition (in red).

Significant correlations between SERT availability in the DRN and Ventral Striatum BOLD response related to the tracking of the social dominance status SDS(t) during the outcomes of the competitive interaction. No significant correlation was observed between SERT availability in the DRN and BOLD response in the ventral striatum tracking the expected value of the slot machine Q(t) (in blue). Moreover, the direct comparison of the correlation coefficient revealed that the correlation coefficients is significantly lower in the social context, compare to the non-social one. VS ventral striatum, SERT serotonin reuptake transporter level, SDS social dominance status, DRN dorsal raphe nucleus.

Fig. 5. Brain regions tracking the expected value of the slot machine Q(t) and the prediction error PE(t) at the outcome.

The graph represents the evolution of the Q(t) for a participant over the experiment. Tracking Q(t) engages the anterior mPFC and tracking the prediction error PE(t) engages the right VS, the anterior medial prefrontal cortex, the superior frontal gyrus and the medial posterior cingulate gyrus. All statistical analyses were performed at a p < 0.05 cluster level corrected for Family Wise Error at the whole-brain level, with an initial cluster forming threshold of p < 0.001 uncorrected. VS ventral striatum, amPFC anterior medial prefrontal.