Social conditioning of neutral faces in families genetically enriched for social anxiety disorder

Abstract

Background: Social anxiety disorder (SAD) runs in families, but insight in the genetic susceptibility to SAD is still scarce. The Leiden Family Lab study on SAD (LFLSAD) [1] has a multiplex, multigenerational design, especially suitable to investigate endophenotypes of social anxiety [2]. Here, we used data from the LFLSAD to examine whether the behavioral and neurobiological responses related to processing neutral faces, as well as to learning the social value of these faces, are potential candidate endophenotypes. We investigated two endophenotype criteria: the ‘co-segregation of the endophenotype with social anxiety (SA) within families’, as well as the ‘heritability of the endophenotype’.

Methods: Participants (n=108; age-range 9.0–61.5 y) originated from eight families genetically enriched for SAD. Experienced clinicians determined the presence of (sub)clinical SAD and other DSM-IV diagnoses, while all participants completed age-matched questionnaires on the level of SA and depression. Functional (f)MRI data were acquired during the Social Conditioning Paradigm (SCP) [3], in which three neutral faces are repeatedly presented and consistently paired with self-relevant sentences containing either positive endorsements, negative comments or neutral statements. This enabled participants to learn the social value of the faces. At several timepoints during the SCP, participants rated the faces on likeability; we tested whether the initial likeability ratings, as well as the difference in these ratings related to learning the social value of the faces (∆Likeability_SCP), were behavioral candidate endophenotypes.

Next, imaging analyses were performed using voxelwise regression models, with SA-level as predictor and activation related to viewing the faces (contrast: ‘all faces>baseline’) as well as activation related to learning the social value of the faces (contrast: ‘first half of the SCP>last half of the SCP’, for all faces as well as separate for the three conditions) as candidate endophenotypes. These analyses were corrected for age, gender, level of depressive symptoms and family structure (cf [4]) and performed within masks of the amygdala. Results: Participants learned the social value of the faces (repeated measures ANOVA with condition and time as within-subject factors: main effect of condition: F(1.9,199.4)=18.9, p=0.009; interaction time*condition: F(3.4,362.8)=37.2, p<0.001). Association analyses showed that the initial response to the faces was not significantly related to SA-level within the families, but there was a significant negative effect of SA-level on ∆Likeability_SCP (β±SE:0.08±0.03, p=0.003), suggesting that the conditioning sentences had a stronger negative effect on the likeability ratings in family-members with higher SA-levels. This behavioral effect was, however, not heritable. fMRI analyses revealed that SA-level was positively related to activation related to viewing the neutral faces in both the left (36 voxels, p=0.03) and right amygdala (164 voxels, p=0.02). Furthermore, SA-level was negatively related to activation associated with learning the negative value of the faces in the right amygdala (138 voxels, p=0.02). Within these amygdala clusters, several voxels displayed at least moderate (h2>0.2) heritability. Conclusions: These findings provide initial evidence for amygdala activation in response to neutral faces, as well as during learning the social value of these faces, as candidate endophenotypes of SA.