Abstract
Self-reproach affects decision-making, mediates the impact of social distress on mental health, and is predictive of more severe depressive symptoms in patients with major depressive disorder (MDD) and subthreshold depression (SthD). Resilience and executive functioning promote stress coping. We examined the shared neural correlates of self-reproach, response failure, spatial planning, and resilience in the white matter connectome of patients with SthD. Whole-brain tractograms of the streamlines that connect 68 cortical and 14 subcortical regions of gray matter were estimated for 25 SthD patients, 29 MDD patients, and 23 healthy controls (HC) from their diffusion-weighted brain magnetic resonance images using a probabilistic algorithm, iFOD2. In mat1 (connection density), edge weights were defined as streamline counts connecting two different nodes scaled by the inverse of the summated two nodal volumes. In mat2 (streamline counts), unscaled streamline counts were assigned as edge weights. In mat3 [mean fractional anisotropy (FA)], the top N percent-ranked edges in terms of the streamline counts survived in the network sparsity (K) of N/100 and were recoded with edge weight values of the mean FA values of streamlines connecting two nodes. Using the graph theory approach, five global graph metrics were calculated to determine the optimal network sparsity range within which network connectedness, small-worldness, and modular organization were satisfied in mat1/mat2/mat3. With a network sparsity of K = 0.11–0.24, a total of 4 local graph metrics of the eigenvector (EC), degree (deg), betweenness (BC), and edge betweenness (EBC) centralities were derived and averaged over K = 0.11–0.24 per individual (and mat1/2/3 separately) to define group-level hubs (ranked in the top 12% of the total 82 nodes, in terms of 2–3 centralities out of the BC/deg/EC) and group-level principal edges (ranked in the top 12% for EBC and had 1–2 group-level hubs) for the 3 matrices separately. In the between-group comparisons of nodal centralities, significant differences (all P values < 0.05) observed in 2–3 centralities of 1–3 (hubs) or 2–3 (nonhubs) matrices [mat1/mat2/mat3] were regarded as meaningful at the study level. Between-group comparisons of EBC aimed to identify the shortcut edges (connecting group-level hubs with these nodes with between-group differences) among the group-level principal edges having either ‘2 group-level hubs’ or ‘1 group-level hub and 1 node with between-group differences’; significant differences in EBC (all P values < 0.05) in 2–3 matrices were considered meaningful at the study level. Furthermore, possible associations of centralities with self-reproach [7th item of the Patient Health Questionnaire-9 (PHQ-9); PHQ97], resilience [Resilience Appraisal Scale (RAS) total score], failure of response generation [Rapid Visual Processing (RVP) total miss; RVPTM], and spatial planning-problem solving [One Touch Stockings of Cambridge (OTS) mean choices to correct 5 moves; OTSMCC5] were examined using linear regression models [centralities ~ 1 + age + sex + education year + PHQ-9 total score (without the 7th item) + Generalized Anxiety Disorder-7 (GAD-7) total score + PHQ97 + RAS + RVPTM + OTSMCC5]. In nodes, significant t-statistics (all Ps < 0.05) for each coefficient in 2–3 centralities [BC/deg/EC] in 1–3 (hubs) or 2–3 (nonhubs) matrices [mat1/mat2/mat3] were regarded as the study-level significance of associations. Shortcut edges connecting these nodes with associations with PHQ97/RAS/RVPTM/OTSMCC5 to other group-level hubs were searched only among the group-level principal edges comprising either ‘2 group-level hubs’ or ‘group-level hub and associated node’; edges with significant t-statistics [P < 0.017 (= 0.05/3 matrices)] for each coefficient in the linear regression model of EBC in 1–3 matrices were defined as those with study-level significance. The line of best fit of PHQ97/RAS/RVPTM/OTSMCC5 with significant centrality values in each node/edge, as determined by multivariate linear regression models, is reflected in the scatterplot. The results first revealed no significant differences in the 5 global graph metrics (averaged in K = 0.11–0.24) among the HCs, SthD patients, and MDD patients (analyses of variance; all P values > 0.05). Second, between-group comparisons of EBC for group-level principal edges demonstrated lower participation of edges connecting the bilateral thalamus with the contralateral caudate, putamen, and insula in the SthD patients than in the MDD patients in the shortcuts connecting different nodes. Third, higher centralities in the right pallidum and right inferior parietal lobule were associated with lower self-reproach. Higher centralities in the right temporal pole in addition to higher EBC values (more participation as shortcut components in the structural connections between nodes) at edges connecting the right thalamus with the left putamen and insula were related to higher resilience. Higher centralities in the left medial orbitofrontal gyrus and right superior frontal gyrus were associated with less failure to generate a response (measured using the RVPTM). Higher centralities in the right middle temporal and pericalcarine gyri were related to better spatial planning-problem solving (reflected in the OTSMCC5). Fourth, the degree of participation of edges (EBC) connecting the thalamus and insula between these regions and with the striatum, pallidum, and left prefrontal regions (superior frontal, medial orbitofrontal, and pars opercularis) as shortcut components connecting different regions was shared in the associated features of self-reproach, executive functions, and resilience. Future longitudinal studies to examine the treatment efficacy of neuromodulation targeting the superior frontal and medial orbitofrontal cortices, insula, and white matter tracts connecting the thalamus with the insula and left prefrontal cortex to improve self-reproach/resilience/executive function are needed.