The diagnosis of major depressive disorder (MDD) urgently requires objective biomarkers for clinical translation. In this study, we established a multicenter cohort to date (N = 1,816; comprising 910 MDD patients and 906 healthy control subjects), using 23,608 standardized speech samples. Based on 6,373 acoustic-prosodic features, we develop a deep learning framework that employs a self-supervised architecture to leverage speech biomarkers. We perform a systematic comparative analysis among pretrained foundation models, including WavLM and HuBERT, and traditional acoustic features extracted from openSMILE. Model performance is evaluated using the area under the receiver operating characteristic curve (AUC), sensitivity, and specificity. Our framework achieves an AUC of 0.932 in internal validation (n = 333), significantly outperforming conventional methods, while maintaining efficacy in external validation (n = 160, AUC = 0.879). Self-supervised representations demonstrate robust diagnostic accuracy compared to other models. Leveraging the large speech biomarker dataset, our findings provide a rapid, cost-effective, and non-invasive approach for assisted depression assessment.

BACKGROUND:Non-suicidal self-injury (NSSI) imposes a significant burden on adolescents, making it essential to understand its underlying functions to develop effective interventions. This study aims to investigate the NSSI functional network and clarify the "psychosocial factors-personality-behavior" chain to inform the development of interventions for NSSI. METHODS:A total of 265 psychiatric inpatients aged 16-25 years with NSSI participated in this cross-sectional study. Network analysis was used to examine relationships among NSSI functions, and mediation analysis explored the pathways from psychosocial factors and personality to these functions. RESULTS:Five stable NSSI functional communities were identified, including internal emotion regulation (IER), psychological pain avoidance (PPA), social influence (SI), suicide resistance (SR), and sensation seeking (SS). The core function of IER exhibited the highest centrality. Psychosocial factors, including health adjustment (βIER = 0.450, βPPA = 0.303, βSI = 0.322, βSS = 0.130), interpersonal stress (βIER = 0.228, βPPA = 0.196, βSI = 0.301), neuroticism (βIER = 0.205, βPPA = 0.165), childhood emotional mistreatment (βIER = 0.204, βPPA = 0.195), openness to experience (βSI = 0.146), conscientiousness (βSR = 0.066), and extraversion (βSR = -0.079) were significantly associated with various NSSI functions. Neuroticism mediated most of the pathways connecting psychosocial factors to NSSI functions such as IER, PPA, and SI. CONCLUSIONS:Internal emotion regulation was the central driver among five functions of NSSI. Neuroticism played a key mediating role in the link between psychosocial factors and NSSI functions. Tailored interventions targeting emotion regulation are essential.

Increasing evidence supports associations between cognitive function and autoimmune disorders, yet the underlying genetic mechanisms remain unclear. Using large-scale genome-wide association statistics for seven cognitive traits and fifteen autoimmune disorders, together with two independent cohorts comprising 522 healthy individuals and 80 patients with schizophrenia, we performed multi-level pleiotropic analyses. Genetic correlation analyses identified nine significant cognition-immune linkage pairs, whereas Mendelian randomization (MR) analyses suggested that these associations were largely driven by shared genetic architecture rather than causality. Using PLACO, colocalization analysis, and MAGMA analyses, we identified 46 pleiotropic loci and 169 pleiotropic genes across the linkage pairs. Polygenic risk scores derived from pleiotropic variants were associated with cognition in healthy adults and showed nominal associations in schizophrenia. Enrichment analyses linked these genes to cognition-immune-related tissues, pathways, and biological processes, while multi-trait colocalization highlighted CD33 as a potential key mediator. Finally, summary-based MR analyses of both pleiotropic genes and anti-inflammatory drug target genes highlighted AMT, CRAT, ERAP2, ERBB3, GNL3, IRF3, MST1R, RPS26, SH2B1, SULT1A1, SULT1A2, TMEM258, CYP2D6 and MAPK3 as promising therapeutic targets for both cognitive function and autoimmune disorders. This study delineates the shared genetic architecture underlying the cognition-immune nexus and identifies novel candidate targets, highlighting the value of integrative genetic approaches for advancing diagnosis and treatment.

While urbanicity increases the risk of mental health issues, its effects on brain networks are heterogeneous and underexplored in relation to different exposome factors. Using a coordinate network mapping strategy termed exposure network mapping (ENM) across eight datasets, this study first consolidated heterogeneous findings of urbanicity to a significant, replicable network involving the middle frontal gyrus, orbital gyrus, and anterior cingulate gyrus. Afterwards, among the other factors examined (air pollution, noise, income, stress, green space), only stress converged into a distinct common network, highlighting the orbital gyrus, caudate, anterior/middle cingulate gyrus, hippocampus, and middle frontal gyrus. This ENM-stress map exhibited the highest correlation with both the ENM-urbanicity map (r = 0.77) and a transdiagnostic map (r = 0.72). In addition, sleep-related coordinates also formed a consistent network, involving the middle cingulate gyrus, orbital gyrus, caudate, and putamen, which correlated strongly with urbanicity (r = 0.75), stress (r = 0.80), and the transdiagnostic pattern (r = 0.55). Collectively, this study highlights the potential risks of urbanicity and stress, as well as the protective role of sleep on brain networks, which may offer new insights for preventing mental health issues in urban environments.

Schizophrenia and immune-mediated diseases are globally prevalent and highly heritable conditions that frequently co-occur, posing major public health burdens. However, their shared genetic architecture remains poorly understood. We applied the bivariate causal mixture model (MiXeR) to investigate the polygenic overlap between schizophrenia and eight common immune-mediated diseases, using genome-wide association study summary statistics comprising 2,489 to 67,323 cases and 9,066 to 497,622 controls. Shared loci were identified through conditional/conjunctional false discovery rate (cond/conjFDR), local genetic correlation (LAVA), and colocalization analyses. Subsequently, gene mapping, functional annotation, expression-trait association, and drug-gene interaction analyses were performed to explore shared genes and enriched pathways, and genetic risk scores (GRS) from the UK Biobank were used to validate the findings. MiXeR estimated substantial polygenic overlap between schizophrenia and immune-mediated diseases, and conjFDR identified 133 shared loci, with eight prioritized through local genetic correlation and colocalization signals. These eight loci were mapped to 85 protein-coding genes enriched in pathways essential for B cell function. Among them, S-PrediXcan analyses identified 14 genes whose expression in brain tissues or blood was associated with both diseases. These genes also interact with immunomodulatory or antihypertensive drugs. Additionally, 11 of the 14 genes were linked to innate immunity and/or cognitive traits. Using UK Biobank data, we further confirmed that overall, shared gene, and B cell activation and receptor signaling pathway–specific genetic risk for schizophrenia is associated with immune-mediated disease susceptibility. These findings underscore the shared genetic architecture of schizophrenia and immune-mediated diseases, advancing insights at the interface of psychiatric genetics and immunology.

Bipolar disorder (BD) manifests both genetic predispositions and brain imaging abnormalities. Genetic analyses provide a powerful approach to disentangling the associations between BD and brain imaging-derived phenotype (IDP). However, in East Asian (EAS) populations, the genetic basis of these neuroimaging changes and their clinical relevance remains underexplored due to limited integrative evidence. We investigated the genetic architecture and clinical relevance of BD and brain IDPs in East Asian populations using individual-level data from the Chinese Bipolar Disorder (CN-BD) cohort (N = 3,920), summary statistics of BD from the Psychiatric Genomics Consortium (PGC) (N = 16,915) and of IDPs from the CHIMGEN Consortium (N = 7,058), and genotype and clinical data from the Chinese Longitudinal and Systematic Study of Bipolar Disorder cohort (CLASS-BD). We employed genetic correlation (GNOVA), multi-trait meta-analysis (MTAG, CPASSOC), and ancestry-specific fine-mapping (MESuSiE) to identify shared genetic loci. Functional annotation, protein association (SMR, BLISS), and spatiotemporal expression trajectory analyses were conducted. Bidirectional Mendelian Randomization (MR) assessed causal relationships, and polygenic risk score (PRS)-PheWAS explored phenome-level associations with clinical traits. Our multi-trait meta-analysis identified 11 pleiotropic single nucleotide variants (SNVs) implicated in BD-IDP pairs, showing convergent enrichment in lipid metabolism and neurotransmitter biosynthesis pathways. Fine-mapping revealed an EAS-specific causal signal (rs7941324 near RPS27P20), which demonstrated high posterior inclusion probability in EAS but not European ancestries, and colocalized with BD and specific brain alterations. MR revealed 27 IDP-to-BD and 35 BD-to-IDP causal relationships, distinguishing structural changes of the parietal and temporal gyri as markers preceding BD onset and the right precuneus area as a state marker that followed BD onset. Clinically, BD-IDP polygenic risk scores were significantly associated with depressive symptoms, anxiety, and suicide risk. Mediation analysis further indicated that depressive and anxiety symptoms statistically mediated the relationship between BD-IDP PRS and suicide risk. Moreover, PRS-PheWAS identified 135 significant associations with metabolic phenotypes, highlighting a psycho-metabolic nexus in BD. Our findings reveal associations between polygenicity and distinct BD brain patterns and clinical profiles. We identified ancestry-specific causal variants and bidirectional BD-IDP relationships. PRS associations with depressive symptoms, anxiety, suicide risk, and metabolic phenotypes highlight precision medicine potential for BD.

Schizophrenia is frequently comorbid with dyslipidemia and hyperglycemia. However, whether metabolic-modifying agents aggravate schizophrenia progression remains unclear. We perform a drug-target genetic association study in two independent Han Chinese schizophrenia cohorts (N = 2,111/292 for discovery/validation). Leveraging metabolic genome-wide association studies, we generate genetic risk scores (GRSs) for lipid-modifying and hypoglycemic targets. Those with higher APOC3 (inhibited by volanesorsen/olezarsen) GRS exhibit attenuated triglycerides and improvement in negative symptoms assessed by Positive and Negative Syndrome Scale (PANSS) (β = 1.23, 95% confidence interval [CI]: 0.30-2.16). Higher GCK (activated by dorzagliatin) GRS is associated with decreased glucose and less improvement across PANSS total (β = -1.70, 95% CI: -2.91-0.50), positive, negative, general subscales. Causal associations of GCK are replicated in independent validation. The effects of APOC3 and GCK on negative symptom recovery are robust in hyperlipidemic/diabetic subgroups. Genetically proxied proteomics analysis provides further functional validation for the identified target-outcome associations. Our findings suggest volanesorsen/olezarsen as potential adjunctive candidates; dorzagliatin warrants prudence in schizophrenia with metabolic disturbance.

Contextual fear generalization is a hallmark of psychiatric disorders including post-traumatic stress disorder (PTSD), generalized anxiety disorder and panic disorder. Among these disorders, PTSD is particularly associated with contextual fear generalization. However, the mechanisms by which fear responses extend to similar contexts remain unclear. Here we tested whether corticotropin-releasing hormone (CRH) neurons in the dorsal bed nucleus of the stria terminalis (dBNST) contribute to contextual fear generalization. Using a contextual fear generalization paradigm with variable shock timing, we observed that intense foot shock conditioning enhanced freezing responses in both the training context and the generalization context in adult male mice, accompanied by elevated dBNST activity. Fiber photometry showed that dBNST CRH neurons were strongly engaged by foot shock and exhibited clear activity changes around freezing onset during both training and generalization tests. Functionally, chemogenetic inhibition of dBNST CRH neurons reduced freezing responses in the generalization test, whereas activation increased generalized freezing and promoted contextual fear generalization. Mechanistically, local pharmacological blockade of CRH receptor 1 (CRHR1) within the dBNST reduced freezing responses in the generalization test, indicating that CRHR1-dependent signaling within the dBNST contributes to fear generalization. Together, these findings identify dBNST CRH neurons and local CRHR1 signaling as key components underlying contextual fear generalization.

Antipsychotics-induced metabolic syndrome (APs-induced MetS) is a common side-effect of antipsychotics, significantly increasing the risk of cardiovascular diseases and mortality. However, the genetic risk factors underlying APs-induced MetS remain poorly understood. Thus, we conducted a sex-stratified genome-wide association study (GWAS) in 3067 patients from Schizophrenia In Non-Occidental participants (SINO) trial, and significant results were validated in an independent cohort (all samples = 200) and proteomic data. Post-GWAS analyses were used to further explore the genetic mechanisms involved in APs-induced MetS. Multi-omics prediction incorporating both polygenic risk and proteomic markers was conducted. After quality control, 1956 patients (965 males, 991 females) were included. We identified significant genetic variants (rs73762168; P = 1.77 × 10-8) on chromosome 6q21, associated with three highly linked genes, NR2E1, SNX3 and AFG1L/LACE1, which were correlated with APs-induced MetS in male patients. Top SNP genotype was validated in independent cohort, showing associations with increased weight and waist circumference. Enrichment analyses across genetic and proteomic data consistently highlighted the PPAR signaling pathway involved in oxidative stress and fatty acid metabolism as a key contributor to APs-induced MetS development. Proteomic analyses confirmed baseline SNX3 protein levels associated with weight gain (P = 0.03) and increased waist circumference (P = 8.87 × 10-3) following six-week antipsychotic treatment. The multi-omics prediction (R2 = 0.18) yielded better prediction of APs-induced metabolic side effects than using either marker alone(R2 = 0.13 or 0.07). This study provides novel genetic insights into the development of APs-induced MetS, particularly in males. The identified genetic variants and pathways offer potential targets for early risk prediction and personalized treatment strategies.

BACKGROUND:Treatment response to antipsychotic drugs in schizophrenia (SCZ) is highly variable, necessitating predictive biomarkers for personalized treatment. While neuroimaging-based predictive models (NPMs) offer promise, their reliance on correlation-based methods renders them vulnerable to confounders. A general framework is required to integrate NPMs with biologically causal information for generalizable and replicable predictions. METHODS:We propose causality-informed neuroimaging prediction (CINP), a framework that incorporates Mendelian randomization-derived causal effects as biological priors into model inference. This framework aims to transcend purely correlational approaches by establishing biologically constrained neuroimaging predictions. To test feasibility, we collected multimodal magnetic resonance imaging data from patients with SCZ across 2 independent longitudinal datasets (Peking University Sixth Hospital: n = 37; Zhumadian Psychiatric Hospital: n = 58). Antipsychotic responses were quantified using the percentage reduction in Positive and Negative Syndrome Scale (PANSS) scores from baseline to follow-up. CINP models were trained to predict individualized PANSS score reduction using neuroimaging data. Model accuracy was evaluated via internal cross-validation and generalizability by intersite cross-validation. RESULTS:CINP achieved an average threefold improvement over conventional NPMs in predicting antipsychotic treatment response. Among various neuroimaging modality configurations, the white matter tract-based CINP yielded the highest prediction accuracy (mean r = 0.652) and cross-site generalizability (mean r = 0.551). Furthermore, predictive features' weights were highly consistent across datasets (similarity = 0.397 to 0.510), indicating replicable patterns. CONCLUSIONS:Beyond feasibility and validity in predicting antipsychotic response, CINP provides a flexible computational framework for incorporating diverse causal constraints, thereby enabling robust neuroimaging-based predictions in multiple clinical contexts.

Gene-environment interaction (G×E) analyses play a crucial role in advancing genetic discovery, addressing missing heritability, and facilitating precision medicine. However, existing G×E methods are mostly designed for cross-sectional data, limiting the utility of longitudinal data. Here we propose SAGELD, a scalable and accurate genome-wide G×E method for longitudinal traits that controls for sample relatedness in large-scale datasets. SAGELD uses matrix projection to construct test statistics and the SPAGRM framework to efficiently control for sample relatedness, achieving 10- to 10,000-fold speedups over existing methods while maintaining greater power than cross-sectional analyses. We evaluated SAGELD through extensive simulations and UK Biobank analyses. Using age and body mass index as environmental exposures, we identified 74 loci with genetic × age interactions and 5 loci with genetic × adiposity interactions in the pooled analysis of longitudinal primary care data and cross-sectional assessment data. These results highlight the advantages of leveraging longitudinal data in G×E analyses.

Epidemiological and clinical observations linking schizophrenia (SCZ) to increased dementia risk, together with the occurrence of psychosis in Alzheimer's disease and related dementias (ADRD), suggest that shared genetic liabilities may contribute to their co-occurrence. Leveraging large-scale genome-wide association study summary statistics for SCZ (53,386 cases and 77,258 controls) and ADRD (111,326 cases and 677,663 controls), we systematically investigated their shared genetic architecture and potential biological mechanisms. We identified three significant local genetic correlations (P < 2.0 × 10⁻⁵) and cross-trait polygenic enrichment between SCZ and ADRD, with 39 genomic loci jointly associated at conjunctional false discovery rate (conjFDR) < 0.05. Fifteen high-confidence genes (CNIH4, CD302, PCGF3, TFR2, EPHX2, SNX32, EFEMP2, CTSW, ASPHD1, TAOK2, INO80E, DOC2A, MAPK3, KANSL1, and XPNPEP3) were consistently prioritized across positional, expression quantitative trait locus, and chromatin-interaction mapping. Tissue- and cell-type enrichment analyses highlighted cerebellar tissues and ependymal-cell-related signals, while pathway analyses implicated synaptic signaling, axonal growth, and presynaptic structural organization. At the locus level, colocalization and transcriptome-wide association analyses converged on 16p11.2, prioritizing INO80E, YPEL3, SLX1B, and TMEM219. Developmental trajectory modeling further revealed region- and stage-specific expression divergence of prioritized 16p11.2 genes, with prominent differences spanning childhood and adulthood. Brain-wide association analysis linked the 16p11.2 lead variant rs9932702 to cortical gray-white contrast (β = -0.062, P = 7.5 × 10⁻¹⁵), a neuroimaging phenotype related to gray-white boundary microstructure and myelination. Finally, bidirectional Mendelian randomization supported a modest directional association between genetic liability to SCZ and increased ADRD risk, but not the reverse direction. Collectively, these findings provide convergent genetic, regulatory, transcriptomic, developmental, and imaging evidence for partial shared liability between SCZ and ADRD, highlighting 16p11.2 and biological processes related to neurodevelopment, synaptic and axonal organization, myelination-related microstructure, and later-life brain vulnerability.

Introduction The incidence of depression among children and adolescents has been increasing in recent years, posing significant challenges to public health and clinical care. A variety of treatments, including pharmacotherapy, psychotherapy and physical interventions, are widely used in clinical practice. However, a comprehensive synthesis of the evidence on the efficacy and acceptability of all these treatment modalities is currently lacking. This study aims to use network meta-analysis (NMA) to compare the efficacy and acceptability of all available treatments for depression in children and adolescents, offering valuable insights to inform clinical decision-making and guide future research in this critical area.Methods and analysis We will include randomised controlled trials evaluating active interventions for depressive disorders in children and adolescents. Seven electronic databases (PubMed, Embase, the Cochrane Library, Web of Science, PsycINFO, Scopus and ClinicalTrials.gov) were searched from inception to 2 July 2024 and updated on 2 November 2025. Two of four investigators will independently screen studies, extract data from eligible articles and assess the risk of bias using the Cochrane Risk of Bias 2.0 tool. The primary outcome will be the change in depressive symptoms. Secondary outcomes will include acceptability (all-cause discontinuation), response rate, remission rate and overall functioning. Pairwise and Bayesian NMA will be conducted. Small-study effects and publication bias will be assessed. The certainty of the evidence will be evaluated according to the Confidence in Network Meta-Analysis approach.Ethics and dissemination As this review involves secondary analysis of previously published studies, ethical approval is not required. The findings will be disseminated through publication in peer-reviewed journals.PROSPERO registration number PROSPERO-ID CRD42024557384.

Depression is a major public health concern, yet many patients fail to achieve remission with standard pharmacotherapy or psychotherapy, and often experience relapse and persistent cognitive impairments. Enhancing treatment outcome is therefore an urgent clinical need and aerobic exercise may have great potential. However, its efficacy in depression treatment and improving cognition, as well as the underlying mechanisms remains unclear. This device-monitored aerobic exercise for depression and cognition improvement (DAECD) trail will enroll 80 patients with depression, randomly assigned (1:1) to an aerobic exercise group or a waiting control group. The intervention group will receive a device-measured aerobic exercise prescription for eight weeks, consisting of moderate-intensity sessions lasting 30 to 50 min, three times per week, alongside their routine treatment. The control group will continue with routine treatment and will initiate the same exercise intervention after the 8-week follow-up period. No additional pharmacological treatment will be initiated or adjusted during the intervention period as part of the study protocol. For those not currently on medication, the intervention will consist of aerobic exercise only. Outcomes will be assessed at baseline, mid-intervention (week 4), and post-intervention (week 8) to evaluate the efficacy of aerobic exercise. The primary outcome is the change in Hamilton Depression Rating Scale score from baseline. Secondary outcomes include cognitive function, depressive and anxiety symptoms, rumination, sleep quality, psychological resilience, safety, physical fitness, exercise capacity, and biological markers. All analyses will follow the intention-to-treat principle. A mixed-effects model will be used for continuous outcomes, and modified Poisson regression will be applied for binary outcomes. Neuroimaging, laboratory and multi-omics tests will be conducted at both baseline and post-intervention, to explore the potential mechanisms through which aerobic exercise exerts its antidepressant effects. This study aims to evaluate the efficacy of aerobic exercise in improving depression and cognition, elucidate its underlying mechanisms, and identify patient subgroups most likely to benefit. The findings may contribute to improved treatment outcomes for depression and inform the development of individualized clinical exercise prescriptions. ClincalTrials.gov NCT06594588, registered 5 Sep 2024.

Although copy number variants (CNVs) represent well-established genetic contributors to schizophrenia (SCZ), their role in bipolar disorder (BD), especially within non-European ancestries, has been inadequately explored. We evaluated the genome-wide load of rare CNVs, encompassing deletions and duplications, in a Han Chinese sample of 3915 BD cases and 7820 ethnically matched controls. We observed a marked overrepresentation of rare deletions in BD patients relative to controls, with affected genes showing enrichment in neural signaling and dosage-dependent networks, indicating that haploinsufficiency in neurodevelopmental loci could underlie a central etiological pathway in BD. Among the 12 previously reported CNV loci from European cohorts, only deletions at 3q29 and 15q11.2 exhibited robust associations with BD susceptibility in Han Chinese individuals. Through genome-wide, gene-centric CNV association testing, we uncovered novel BD-linked loci, including deletions spanning GLIS2 and PAM16 at 16p13.3, GRID2IP at 7p22.1, and CFLAR at 2q33.1, alongside a duplication affecting ZNF878 and ZNF844 at 19p13.2. These disrupted genes are chiefly implicated in neuronal maturation, synaptic modulation, and mitochondrial dynamics. This work delivers the most thorough delineation of BD-associated CNVs in Han Chinese to date, underscoring the imperative for ancestry-inclusive research to comprehensively unravel psychiatric genomics and unveiling fresh mechanistic perspectives on BD etiology.

Psychiatric and cardiovascular diseases (CVDs) are frequently comorbid and are interconnected through the brain-heart axis. However, the underlying shared genetic etiology remains unknown in East Asians. To address this critical gap, we conducted a genome-wide pairwise trait pleiotropy study by leveraging genome-wide association studies of three major psychiatric disorders (schizophrenia [SCZ], bipolar disorder [BIP], major depressive disorder [MDD]) and ten cardiovascular traits (including eight CVDs) in East Asians. We identified genetic overlaps across seven disease pairs, such as SCZ with coronary artery disease. Through this pairwise approach, six of a total of 18 pleiotropic loci demonstrated tissue-specific expression in brain and cardiovascular systems. In the cross-ancestry replication, nine of the pleiotropic loci were validated. Among the novel pleiotropic genes, TPCN1, CACNA2D2, CACNA1D, and ATP2B1 are involved in voltage-dependent calcium channel activity, regulation of calcium influx, enriched in calcium-related pathway. We validated association with calcium signal pathway in an independent cohort. Calcium pathway-specific polygenic risk score for SCZ was associated with prolonged corrected QT (QTc) interval, which remained robust among individuals free from QTc-affecting drugs. Given that calcium-channel blockers are commonly prescribed for heart and blood vessel conditions, we performed drug target analysis by integrating gene expression profiles from the brain and cardiovascular tissues. Our findings implicated that calcium-channel blockers and peripheral vasodilators elevated SCZ risk, diuretics reduced the risks of SCZ, BIP, and MDD. Our study reveals extensive shared genetic architectures underlying psychiatric and CVDs, which warrant prudence in the use of calcium channel blockers among patients with concurrent psychiatric and CVDs.

BACKGROUND:Nonsuicidal self-injury (NSSI) in adolescents with depressive disorder is characterized by recurrent impulsivity and is a predictor of future suicidal behavior, necessitating effective and feasible interventions. Mindfulness-based cognitive therapy for adolescents (MBCT-A), a first-line treatment for depression, was evaluated in this trial for its efficacy in reducing NSSI in this population. METHODS:In this randomized, single-blind, controlled trial, 94 hospitalized adolescents diagnosed with depressive disorder and engaging in NSSI were allocated to either MBCT-A plus treatment-as-usual (TAU) or TAU alone. The primary outcome was the severity of NSSI, measured by the Adolescent Self-Harm Scale (ASHS) at baseline, post treatment (4 weeks), and at 1-month follow-up. Secondary outcomes encompassed symptoms of anxiety, depression, emotion regulation, risk behaviors, life satisfaction, resilience, rumination, coping styles, and mindfulness. Both intention-to-treat (ITT) and per-protocol (PP) analyses were conducted. RESULTS:Both ITT and PP analyses revealed significant between-group differences in NSSI severity at 4 weeks. Compared to the control group, the MBCT-A group showed significant improvements in mindfulness awareness, nonreactivity, emotional awareness, depressive symptoms, emotion regulation, and rumination (all P < 0.05). At the 1-month follow-up, the intervention group maintained a significant reduction in NSSI severity (interaction effect: F = 9.454, P < 0.001). CONCLUSIONS:MBCT-A is effective in reducing NSSI behavior and improving mindfulness, emotion regulation, and depressive symptoms among adolescents with depressive disorder and NSSI. These findings support the clinical application of MBCT-A for this high-risk population.

Non-suicidal self-injury (NSSI) is a common behavior among adolescents, particularly within psychiatric populations. While neurobiological and psychosocial risk factors have been extensively studied, the mechanisms underlying NSSI’s heterogeneity remain unclear. This study investigated 304 hospitalized adolescents/young adults (16–25 years) with NSSI and comorbid psychiatric diagnoses (major depressive disorder [MDD], bipolar disorder [BD], eating disorders [ED]) using psychological assessments and resting-state fMRI data from 163 participants. Orthogonal projection non-negative matrix factorization of Ottawa Self-Injury Inventory responses identified two latent factors: self-related factor and social-related factor. The self-related factor correlated with amygdala-centered cortico-limbic emotional regulation networks and predominated in affective disorders (MDD/BD), while the social-related factor linked to frontoparietal cognitive control and frontotemporal social cognition networks, particularly in ED. Fuzzy C-means clustering revealed three NSSI functional subtypes, independent of diagnostic categories: self-subtype primarily driven by self-related functions, social-subtype influenced by both self-related and social-related functions with greater exposure to psychosocial risks, and non-specific subtype characterized by mixed motivations. No subtype was exclusively driven by social-related functions. The “self-social” dual-dimensional framework with distinct neural mechanisms demonstrated subtype-specific profiles in functional connectivity, psychosocial risk exposure, and clinical features. Self-related mechanisms primarily engaged emotional regulation circuits, whereas social-related mechanisms emphasize the role of psychosocial risk factors and cognitive-emotional circuits. These findings provide neural evidence for the functional heterogeneity of NSSI and highlight the need for personalized interventions. Treatments targeting emotion regulation may benefit all subtypes, individuals with prominent social-related motivations may additionally require interventions aimed at improving interpersonal functioning.