Disrupted effective connectivity between the amygdala and orbitofrontal cortex in social anxiety disorder during emotion discrimination revealed by dynamic causal modeling for FMRI

Abstract

Social anxiety disorder (SAD) is characterized by over-reactivity of fear-related circuits in social or performance situations and associated with marked social impairment. We used dynamic causal modeling (DCM), a method to evaluate effective connectivity, to test our hypothesis that SAD patients would exhibit dysfunctions in the amygdala-prefrontal emotion regulation network. Thirteen unmedicated SAD patients and 13 matched healthy controls performed a series of facial emotion and object discrimination tasks while undergoing fMRI. The emotion-processing network was identified by a task-related contrast and motivated the selection of the right amygdala, OFC, and DLPFC for DCM analysis. Bayesian model averaging for DCM revealed abnormal connectivity between the OFC and the amygdala in SAD patients. In healthy controls, this network represents a negative feedback loop. In patients, however, positive connectivity from OFC to amygdala was observed, indicating an excitatory connection. As we did not observe a group difference of the modulatory influence of the FACE condition on the OFC to amygdala connection, we assume a context-independent reduction of prefrontal control over amygdalar activation in SAD patients. Using DCM, it was possible to highlight not only the neuronal dysfunction of isolated brain regions, but also the dysbalance of a distributed functional network.

Keywords: DLPFC; amygdala; fMRI; orbitofrontal cortex; social anxiety disorder.

Figure 1.

Experimental paradigm. Facial emotion and object discrimination tasks were presented in alternating individual blocks for 20 s. Between task conditions, a white fixation cross was presented for 20 s to serve as a baseline condition. Each task block was repeated 5 times, yielding a total paradigm length of ∼7 min. The vertical lines depict the individual stimulus onsets in one randomly selected subject, the black time course presents the paradigm regressor convolved with SPM's canonical hemodynamic response function, which was used for the GLM and DCM analyses.

Figure 2.

Rendering of the statistical parametric map from the emotion discrimination tasks. This map is based on the group-level result from all subjects (n = 30), with FACE > OBJECT as contrast of interest, thresholded at P < 0.05 (FWE whole-brain corrected for multiple comparisons) with n = 75 voxels minimum cluster size. Right amygdala, OFC, and right DLPFC are highlighted and were used as volumes of interest for the subsequent DCM analysis.

Figure 3.

DCM networks for healthy controls (left) and SAD patients (right). BMA results reveal a group difference in model structure, connectivity parameter estimates, and modulatory effects of emotional faces (significant group differences highlighted in Fig. 4).

Figure 4.

Significant group differences between healthy controls (left) and SAD patients (right). Most importantly, the negative feedback loop between amygdala and OFC, found in HC, is dysfunctional in SAD (also see Fig. 5). SAD patients further exhibited decreased effective connectivity from amygdala and OFC to the DLPFC.

Figure 5.

Modulation of the OFC to amygdala connection by emotional faces. The modulatory influence of the emotion decimation task was not different between groups. However, in combination with the positive intrinsic connectivity, the negative modulatory influence was not sufficient to effectively downregulate amygdalar activation in SAD patients. Group BMA results and SD are shown, *P < 0.05 in 2-sided 2-sample t-test.