Complementary functional gradient and SFC analyses reveal network abnormalities in adolescent depression subtypes

Adolescent major depressive disorder (MDD) emerges during a sensitive neurodevelopmental period and is frequently accompanied by non-suicidal self-injury (NSSI), a clinically important behavioral phenotype associated with affective dysregulation. Although prior resting-state fMRI studies have identified distributed functional abnormalities in adolescent MDD, it remains unclear whether MDD with NSSI is associated with alterations in macroscale functional organization, stepwise cross-network propagation, or both. We therefore combined functional gradient (FG) analysis and stepwise functional connectivity (SFC) to characterize group-related differences from the complementary perspectives of gradient-based functional organization and multi-step network propagation.

Using resting-state fMRI data from 135 adolescents classified as healthy controls (HC), adolescents with major depressive disorder and non-suicidal self-injury (MDD-NSSI), and adolescents with major depressive disorder without NSSI (MDD-noNSSI), we estimated functional gradients based on the Schaefer-400 parcellation. Between-group differences in Gradient 1 (G1) and Gradient 2 (G2) were assessed using pairwise Welch's t-tests, followed by global BH-FDR correction across all 2400 ROI-wise tests. Bonferroni correction was additionally used as a conservative sensitivity analysis. To avoid circular seed selection, weighted SFC was performed using an a priori sgACC/BA25 seed set rather than ROIs selected from the gradient analysis. Stepwise propagation maps were computed from step 1 to step 7. For ROI-wise SFC comparisons, pairwise Welch's t-tests were performed within each contrast and step, followed by exploratory within-network BH-FDR correction separately within each canonical functional network.

Functional gradient analysis revealed distinct gradient profiles across the two patient groups. After global BH-FDR correction across 2400 ROI-wise tests, G1 showed extensive differences in MDD-noNSSI versus HC and in MDD-NSSI versus MDD-noNSSI, whereas the G1 difference between MDD-NSSI and HC was relatively limited. By contrast, G2 showed extensive and robust differences in MDD-NSSI versus HC and in MDD-NSSI versus MDD-noNSSI, while MDD-noNSSI showed a weaker G2 difference relative to HC. The exploratory sgACC/BA25-seeded SFC analysis showed HC-relative propagation abnormalities mainly involving limbic temporal pole and orbitofrontal parcels. Compared with HC, MDD-NSSI showed broader exploratory within-network-corrected limbic involvement and additional somatomotor reductions, whereas MDD-noNSSI showed a more focal left temporal pole propagation pattern. No direct SFC difference between MDD-NSSI and MDD-noNSSI survived exploratory within-network BH-FDR correction.

FG findings provided direct evidence for different gradient-related profiles between adolescent MDD with and without NSSI, whereas sgACC/BA25-seeded SFC provided exploratory HC-relative propagation evidence. MDD-noNSSI was mainly associated with broad G1-related displacement, whereas MDD-NSSI was associated with G2-related redistribution and broader HC-relative limbic-cortical propagation abnormalities from the sgACC/BA25 seed.