Influence of periaqueductal gray on other salience network nodes predicts social sensitivity

Abstract

The intrinsic connectivity of the salience network (SN) plays an important role in social behavior, however the directional influence that individual nodes have on each other has not yet been fully determined. In this study, we used spectral dynamic causal modeling to characterize the effective connectivity patterns in the SN for 44 healthy older adults and for 44 patients with behavioral variant frontotemporal dementia (bvFTD) who have focal SN dysfunction. We examined the relationship of SN effective connections with individuals' socioemotional sensitivity, using the revised self-monitoring scale, an informant-facing questionnaire that assesses sensitivity to expressive behavior. Overall, average SN effective connectivity for bvFTD patients differs from healthy older adults in cortical, hypothalamic, and thalamic nodes. For the majority of healthy individuals, strong periaqueductal gray (PAG) output to right cortical (p < .01) and thalamic nodes (p < .05), but not PAG output to other central pattern generators contributed to sensitivity to socioemotional cues. This effect did not exist for the majority of bvFTD patients; PAG output toward other SN nodes was weak, and this lack of output negatively influenced socioemotional sensitivity. Instead, input to the left vAI from other SN nodes supported patients' sensitivity to others' socioemotional behavior (p < .05), though less effectively. The key role of PAG output to cortical and thalamic nodes for socioemotional sensitivity suggests that its core functions, that is, generating autonomic changes in the body, and moreover representing the internal state of the body, is necessary for optimal social responsiveness, and its breakdown is central to bvFTD patients' social behavior deficits.

Keywords: behavioral variant frontotemporal dementia; dynamic causal modeling; periaqueductal gray; salience network; socioemotional sensitivity.

FIGURE 1

Matrices of mean effective connectivity of the ONC and bvFTD group and bvFTD compared with ONC, accompanied with schematic overviews of the effective connectivity results (thickness of lines equal the effective connectivity strength between node pairs). (a) At a Pp = 0.99, the ONC effective connectivity model shows reciprocal connections among the cortical vAI and ACC nodes. All subcortical bilateral counterparts are reciprocally connected as well. Self‐connections are all self‐inhibiting, except for the PAG. (b) At a Pp = 0.99, the bvFTD effective connectivity model shows reciprocal connections between the cortical SN nodes as well, as well as between the bilateral thalamic nodes. Self‐connections are self‐inhibiting in the bilateral hypothalamic nodes, the left thalamic node, the right vAI, and in the ACC, but are not significantly different in bilateral amygdala, right thalamus, and PAG. (c) The effective connectivity model of the bvFTD group compared with the effective connectivity model of the ONC group showed that at a Pp = .74 there was significantly weaker activation for the bvFTD group among (right) cortical nodes and subcortically between the bilateral thalamic nodes. Effective connectivity from the right to the left hypothalamus was significantly lower as well, and self‐connections were less self‐inhibiting in the right vAI, left hypothalamus and bilateral thalamic nodes. ACC, anterior cingulate cortex; amy, amygdala; bvFTD, behavioral variant frontotemporal dementia; hypo, hypothalamus; L, left; ONC, older normal controls; PAG, periaqueductal gray; Pp, posterior probability; R, right; thal, thalamus; vAI, ventral anterior insula

FIGURE 2

Schematic illustration of effective node connections that with a significance level of p < .05 or p < .01 predicted higher or lower RSMS EX scores, calculated separately for ONC and bvFTD patients (see Figure S2 for all effective node connections' confidence intervals in relation with the RSMS EX score). For the ONC group, PAG output to cortical and thalamic nodes positively predicted RSMS EX score, whereas input from the ACC, amygdala, and PAG to the left vAI positively predicted RSMS EX score for the bvFTD group. Mainly hypothalamic output had negative influence on the RSMS EX score for ONC, and bilateral thalamic connections for bvFTD patients. EX, expressive behavior; RSMS, revised self‐monitoring scale

FIGURE 3

Individual differences in strength of PAG output predict RSMS EX score. (a) Comparing the summary estimates for PAG output to cortical nodes with PAG output to subcortical nodes yielded three different cluster groups in ONCs, of which group 1 (blue) showed weakest PAG output strength toward cortical and subcortical SN nodes, group 2 (red) showed medium output strength, and group 3 (green) showed most output strength toward cortical and subcortical SN nodes. (b1) When comparing RSMS EX scores of ONCs that belong to cluster group 3 (high PAG output) to cluster group 2 (medium PAG output), cluster group 3 performed significantly better on the RSMS EX scale than group 2. (b2) When comparing RSMS EX scores of the whole sample (i.e., ONCs and bvFTD patients combined), members of cluster group 3 performed significantly higher on the RSMS EX scale than members of cluster group 1. This suggests that stronger PAG output (toward cortical and subcortical SN nodes) increased socioemotional sensitivity, both in disease and in normal brain function. (c) By taking the χ ² of cluster group membership per diagnostic group, bvFTD patients belonged significantly more to cluster group 1, and ONCs belonged significantly more to cluster group 3. This shows that compared with ONC, bvFTD patients belonged significantly more to the cluster group with weak PAG output and corresponding low RSMS EX scores. (d) Middle to high values on the silhouette plot shows that data can be safely divided into three clusters