The default mode network and rumination in individuals at risk for depression

The study examines whether individuals at risk for depression, specifically females with high neuroticism but no current or lifetime mood or anxiety disorders, exhibit differential default mode network (DMN) activation linked to the processing of negative self-referential information and whether such activation is associated with rumination. The DMN is implicated in spontaneous self-referential thought and shows altered connectivity and activation in major depressive disorder (MDD), particularly in response to negative stimuli. Establishing whether similar DMN alterations exist prior to the onset of depression would illuminate neurocognitive vulnerabilities and mechanisms (like rumination) that may contribute to risk and inform prevention.

Prior work shows currently depressed individuals often exhibit increased resting-state DMN functional connectivity, though some studies report decreases. Task-based fMRI has revealed both increased and decreased activation across DMN regions in MDD when processing negative stimuli, suggesting altered self-referential network engagement in negative information processing. In at-risk populations, structural alterations have been reported, such as reduced medial prefrontal cortex (MPFC) thickness and its association with negative affect and polygenic risk for depression. Familial risk has been linked to increased resting-state DMN connectivity (e.g., left lateral parietal cortex and precuneus/posterior cingulate cortex), though many such samples include individuals with current or past psychopathology, limiting inferences about premorbid risk. Rumination, a key risk and maintenance factor for depression, has been associated with DMN connectivity patterns (e.g., PCC-subgenual cingulate, lateral temporal cortex-parahippocampal cortex) and relative DMN vs task-positive network activity, with depressive rumination positive and reflective rumination negative correlations to DMN dominance. MPFC-ACC connectivity has also been positively correlated with rumination. These literatures motivate examining DMN activation and its relation to rumination in individuals at risk but without clinical histories.

Participants: 1,126 individuals were pre-screened online. Eligible females were those scoring in the upper 80th percentile (at-risk, high neuroticism) or 40th–60th percentile (controls, average neuroticism) on the NEO-Five Factor Inventory (NEO-FFI) neuroticism scale; 115 attended a screening visit. Using the SCID-5-RV, 60 had no current or lifetime mood or anxiety disorders; 55 underwent MRI scanning. Final dataset: n = 25 at-risk females and n = 28 control females after exclusions (1 did not hear audio; 1 excessive motion). Groups did not differ in age or years of education. Self-identified race/ethnicity included African American (11.9%), Asian (35.6%), Caucasian (35.6%), Hispanic (5.1%), multiracial (11.9%). Measures: NEO-FFI (neuroticism α = 0.84) and Ruminative Responses Scale (RRS; 22 items, score range 22–88; α = 0.95). MRI acquisition: 3.0 Tesla Siemens MAGNETOM Prisma at Harvard Center for Brain Science; functional BOLD images with gradient-echo T2*-weighted sequence (TR/TE/flip angle = 650 ms/34.4 ms/54°), 2.3 × 2.3 × 2.3 mm resolution; T1-weighted high-resolution images (TR/TE/flip angle = 2200 ms/1.57 ms/7°), 1.2 × 1.2 × 1.2 mm resolution. Scanner software upgraded to VE11C after 28 subjects; version entered as regressor. Stimuli: Auditory blocks of four 30 s comments per block (critical and praise), adapted from maternal comments literature. Instruction: imagine comments spoken by someone very important to the participant. Order counterbalanced. Data preprocessing and analysis: SPM8 used for realignment (six-parameter rigid body), co-registration to T1, normalization to MNI space, and smoothing with 6 mm FWHM Gaussian kernel. Motion covariates (realignment parameters) included at first-level analysis; groups did not differ in motion (max absolute motion: at-risk M = 0.77, SD = 0.45; controls M = 0.81, SD = 0.52; t(51) = 0.05, P = 0.83). Conditions of interest were rest periods before/after criticism and before/after praise, focusing on post-comment rest as the likely period of processing and for comparability to DMN rest literature. Second-level flexible factorial GLM modeled Group (at-risk vs controls) × Condition (after/before criticism; after/before praise). Contrasts: (i) at-risk > controls for (after criticism > before criticism); (ii) at-risk > controls for (after praise > before praise). ROI analyses: A priori DMN ROIs defined via WFU PickAtlas: MPFC (BA 9,10,24,32), PCC/retrosplenial (BA 29,30/23,31), hippocampal formation (bilateral hippocampus/HF), lateral temporal cortex (BA 21), inferior parietal lobule (IPL; BA 39,40). Significance threshold P < 0.005 voxelwise; multiple comparisons controlled with AFNI 3DClustSim to achieve cluster-level α < 0.05. Cognitive measures analyses: Spherical ROIs (radius 5 mm) centered on peak voxels in MPFC and IPL showing significant criticism-specific activation. MarsBaR used to extract beta weights per condition. Two control participants with RRS scores > 58 (>2 SD above control mean) were removed from correlational analyses. Pearson correlations in SPSS v24 assessed associations between ROI beta weights (criticism-specific) and rumination.

- DMN hyperactivation specific to criticism: At-risk individuals showed greater activation than controls in bilateral MPFC and left IPL during rest after criticism relative to before criticism; no group differences after praise. • Left MPFC (BA 10): MNI = (−6, 60, 18), Z = 3.69, cluster size k = 81 voxels (3DClustSim-corrected). • Right MPFC (BA 9): MNI = (6, 58, 30), Z = 3.35, k = 61 voxels (3DClustSim-corrected). • Left IPL (BA 40): MNI = (−46, 36, 50), Z = 2.47, k = 212 voxels (3DClustSim-corrected). - Rumination association: In the at-risk group, criticism-specific left IPL activation correlated positively with rumination [r(25) = 0.479, P = 0.02]; in controls, the correlation was not significant [r(26) = −0.020, P = 0.93]. MPFC criticism-specific activation was not significantly correlated with rumination in either group (left MPFC: at-risk r = 0.245, P = 0.27; controls r = 0.291, P = 0.17; right MPFC: at-risk r = 0.121, P = 0.59; controls r = 0.337, P = 0.11). - No significant group differences after praise in MPFC or IPL. - Groups were comparable on motion metrics; imaging analyses accounted for motion and scanner software version.

Findings support the hypothesis that individuals at risk for depression exhibit differential DMN activation when processing negative self-referential information. Specifically, greater dorsal MPFC (DMPFC) and left IPL activation after criticism suggests enhanced self-referential evaluation and higher-order semantic/conceptual integration of negative comments in at-risk participants. The DMPFC (vs VMPFC) pattern aligns with roles in comparing incoming self-referential information to personal standards, indicating selective engagement for negative information. The IPL activation, part of a semantic network overlapping with the DMN, implies more extensive conceptual processing of critical content. The link between IPL activation and rumination in the at-risk group indicates that DMN nodes implicated in semantic/self-referential processing may contribute to ruminative tendencies before clinical onset. Absence of group differences in HF or PCC/retrosplenial regions may reflect stimulus characteristics (general imagined close person vs actual maternal comments), potentially engaging fewer autobiographical memory processes than studies using mothers' voices. Lack of LTC differences suggests that LTC abnormalities may be more specific to current depression rather than premorbid risk. Overall, results indicate that DMN hyperactivation to negative self-referential input, and its association with rumination, could represent a neurocognitive vulnerability pathway for depression.

Individuals at risk for depression (female high-neuroticism, no mood/anxiety history) exhibit increased DMN activation (bilateral MPFC, left IPL) after criticism but not praise, and IPL activation relates to rumination. This pattern indicates biased processing of negative self-referential information via DMN nodes, potentially reflecting a vulnerability for later depression. Future research should: - Use thought sampling to verify in-the-moment cognitions (rumination) during DMN hyperactivation. - Employ stimuli from personally relevant sources (e.g., mothers) in at-risk samples. - Conduct longitudinal studies tracking DMN activation and LTC changes across first depressive episode onset. - Explore neuromodulation (e.g., targeting MPFC or IPL) as preventative interventions for rumination and depression risk. - Screen comprehensively for multiple depression-specific risk factors to refine risk-related DMN signatures.

- Control group may have unassessed risk factors (e.g., family history of depression not assessed), complicating group comparisons. - Neuroticism and female sex are non-specific risk factors also relevant to anxiety; findings may generalize to anxiety risk as well as depression risk. - Rumination assessed via trait self-report (RRS); participants may not have been ruminating during measured DMN hyperactivation. In-the-moment thought sampling is needed. - Relatively small final sample size (n = 53) limits power and generalizability. - Stimulus design (imagined close person rather than actual maternal comments) may have reduced engagement of autobiographical memory networks, potentially contributing to null findings in HF/PCC.