Transcriptional dissection of symptomatic profiles across the brain of men and women with depression

MDD is a heterogeneous syndrome defined by depressed mood and anhedonia, accompanied variably by cognitive impairment, anxiety, changes in weight/appetite, fatigue, psychomotor changes, sleep abnormalities, low self-esteem, indecision, and suicidality. Diagnosis relies on subjective clinical criteria and symptom counts rather than intensity, despite likely distinct biological substrates for different symptoms. Neuroimaging studies link specific symptom dimensions (e.g., rumination, negative bias, threat dysregulation, anxiety, cognitive control deficits, anhedonia) to functional changes in large-scale networks and regions, but the underlying molecular mechanisms are unclear. Prior transcriptomic work has shown sex-specific and region-specific differential expression and network alterations in MDD and stress models, and distinctions between trait vs state depression. However, it remains unknown whether specific transcriptional signatures map onto particular symptom domains in males and females. The study aims to associate brain-region transcriptional networks with symptom profiles in MDD, testing sex specificity and identifying hub genes and pathways that may underlie symptom expression.

Functional imaging has associated depressive symptom dimensions with activity/connectivity of specific networks: default mode and negative affect networks with rumination and negative bias; salience/attention/cognitive control networks with anxiety and cognitive dyscontrol; positive affect networks with anhedonia. Transcriptomic studies in postmortem MDD brains and stress models reveal gene expression and co-expression alterations, including sex-specific network changes affecting intracellular cascades, and signatures distinguishing trait vs state depression. Network-based, systems biology approaches have successfully characterized molecular architectures in complex neuropsychiatric disorders (Alzheimer's disease, autism, bipolar disorder, schizophrenia, MDD), supporting the use of co-expression networks to parse disease heterogeneity. The present work extends these concepts to link transcriptional modules to specific symptom domains by sex.

Human postmortem brain study integrating RNA sequencing and systems-level network analyses. Cohort: 89 individuals across six regions (aINS, OFC/BA11, vmPFC/cg25, dIPFC/BA8/9, NAc, vSub): 25 male MDD, 25 female MDD, 17 male controls, 22 female controls. Psychological autopsies (SCID-based DSM-IV) provided symptom presence (depressed mood, anhedonia, appetite/weight change, insomnia/hypersomnia, psychomotor agitation/retardation, fatigue, low self-esteem, difficulty in concentration/indecision, suicidal thoughts); since some symptoms were universally present (depressed mood, anhedonia, fatigue, recurrent suicidal thoughts), analyses focused on the remaining variable symptoms. Tissue processing: Dissections performed at 4 °C; RNA extracted (RNeasy/Trizol, DNase I), RIN assessed. Library prep: ScriptSeq Complete Gold (rRNA depletion), spike-in controls; Illumina HiSeq2500, 50 bp paired-end, ~50M reads/sample, multiplexed. Data processing: Quality control (FASTQ/FASTX), alignment with TopHat to GENCODE 2019 (GRCh38.p12), counts via HTSeq. Low-expression filtering (<5 reads in ≥20% samples per group). Normalization with voom (limma). Batch/unwanted variation estimated with RUVseq using spike-ins; top factor included as covariate. PCA identified region-specific clinical/technical covariates (e.g., age, PMI, pH, cohort, drug abuse, smoking, RIN, early life adversity), which were adjusted in downstream analyses. Differential expression: limma GLM per region with sex and phenotype as factors, adjusting for covariates; DEGs defined at nominal P ≤ 0.05 (multiple testing not applied for DEG calling). Transcriptional overlap: RRHO compared ranked signed gene lists between sexes and to prior cohort; BY correction, heat-mapped -log10 adjusted P-values. Gene ontology: g:Profiler with FDR < 0.05. Co-expression networks: WGCNA constructed sex-specific networks for each region and combined-sex networks per region; outliers removed based on standardized connectivity (<-3.5). Networks built using biweight midcorrelation, unsigned adjacency with soft-thresholding, topological overlap, dynamic tree cut; modules named by color; hub genes defined as top 5% intramodular connectivity. Module enrichment for DEGs via GeneOverlap (Fisher’s exact test, padj < 0.05). Module differential connectivity (MDC): ratio of within-module pairwise connectivity between conditions (MDD vs CTRL; male vs female); MDC > 1 indicates gain of connectivity (GOC), <1 loss (LOC); significance via FDR permutation. Module preservation across sexes assessed with WGCNA statistics; Zsummary >10 considered preserved. Symptom associations: For top 200 DEGs per region, hierarchical clustering assessed gene–symptom patterns. Module–symptom associations computed via point-biserial correlations between module eigengenes (first PC) and dichotomous symptom presence in males and females separately; P-values adjusted via permutation (1000) and BH. Gene-level contribution assessed by correlating gene significance (GS) with module membership (KME). Functional interpretations from GO enrichment. Reproducibility assessed by overlap with previous datasets.

- Differential expression by sex and region: Large numbers of DEGs in both sexes across all six regions with limited overlap between males and females (3.2%–35.9% overlap depending on region). Lowest overlap in NAc (3.2%) and vSub (4%), higher in cortex (dIPFC 35.9%, vmPFC 11.9%). RRHO showed strong concordant overlap in cortical regions (OFC, vmPFC, dIPFC; max P ≈ 1.0E-250), but little overlap in limbic NAc and vSub. Some genes displayed opposite regulation between sexes (e.g., ZNF729, RXFP3, OR52A5, EIF4EBP2, CARTPT). - Functional convergence despite sex differences: GO terms differed by sex per region, but shared functional domains were noted (e.g., GABAergic synaptic function in aINS; neuropeptide signaling in vmPFC; AMPA receptor activity and neurotransmitter receptor complex in vSub). - Network architecture and preservation: WGCNA identified 20–109 modules per region (50–7662 genes each), covering broad functional domains. A substantial fraction of male modules were preserved in females (Zsummary >10): NAc 76.8%, aINS 69.7%, vmPFC 66.3%, dIPFC 55.0%, OFC 52.2%, vSub 34.6%. - Module differential connectivity (MDC): Many sex-unique modules showed significant MDC in MDD vs CTRL. Females: vSub 67.3%, NAc 57.9%, OFC 53.3%; no unique aINS modules showed MDC. Males: vSub 41.4%, NAc 36.8%, vmPFC 27.9%. Preserved modules rarely showed MDC vs CTRL (<20% in most regions; exceptions: male NAc 31.7%, female vmPFC 21%). However, preserved modules frequently differed between male vs female MDD (e.g., dIPFC 72.7%, vSub 67.6%, NAc 42.9%). Enriched functions among MDC modules included synaptic function, mitochondrial function, intracellular signaling, and nuclear gene regulation. - Symptom–module associations are sex- and region-specific: Hierarchical clustering revealed clear gene expression patterns aligned with symptom presence in females (e.g., insomnia/hypersomnia across regions; psychomotor changes in aINS/vmPFC/dIPFC; appetite/weight changes in vmPFC/vSub), but not in males. Point-biserial correlations of module eigengenes identified numerous sex-specific module–symptom associations: in males, appetite/weight (OFC 26.9%), psychomotor agitation/retardation (NAc 31.0%), low self-esteem (dIPFC 26.1%), indecision/concentration (OFC 11.5%); in females, appetite/weight (dIPFC 13.0%, aINS 12.7%), psychomotor agitation/retardation (dIPFC 30.4%), low self-esteem (aINS 12.7%), indecision/concentration (vmPFC 46.0%). Insomnia/hypersomnia associated with the largest proportion of modules in both sexes. - Specific high-impact modules: • Male OFC Ivory (synaptic signaling, GABAergic neurotransmission): Strongly associated with appetite/weight change (r=0.76, Padj<0.001), insomnia/hypersomnia (r=0.68, Padj<0.005), and indecision/concentration (r=0.73, Padj<0.005). Enriched for GABAergic genes (Padj<5e-8). Hubs GAD1, GAD2, NXPH1; enriched for symptom-associated genes (FET Padj: appetite/weight <1e-18, insomnia/hypersomnia <5e-11, indecision/concentration <5e-17). • Male NAc Darkviolet: Associated with psychomotor agitation/retardation (r=−0.69, Padj<0.0001). Male dIPFC Darkred: Associated with low self-esteem (r=−0.83, Padj<0.0001). • Female aINS Darkorange (synapse, cell junction): Uniquely associated with appetite/weight change; enriched for downregulated DEGs (Padj<3e-10) and symptom-associated genes (39% of module; FET Padj<5e-26; module-level Padj<1e-57). Hubs include CLSTN1, CLSTN3, PIK4KA (downregulated). • Female dIPFC Saddlebrown (cellular protein catalytic processes): Strongly associated with psychomotor agitation/retardation (r=0.891, Padj<0.0001); enriched for upregulated DEGs (Padj<5e-37); 34.1% of genes associated with the symptom (Padj<1e-17). Hubs: SELENOT, ACTR3, CHMP2B, SGPP1, TM9SF3. • Additional female modules: Skyblue (OFC) with insomnia/hypersomnia (r=−0.747, Padj<0.0001), Purple (aINS) with low self-esteem (r=−0.827, Padj<0.0001), Palevioletred3 (vmPFC) with indecision/concentration (r=−0.60, Padj<0.0001). - Limited cross-sex commonality with opposite directions: Identified 22 modules associated with the same symptom in both sexes; 77% showed opposite correlation directions and largely distinct driving genes. Example: OFC Orange (synaptic transmission) correlated positively with indecision/concentration in males (r=0.484, Padj<0.05) but negatively in females (r=−0.622, Padj<0.05); only 6 of 97 symptom-associated genes were shared; hub PPP3R1 common, male-specific hubs included GABRB3, PRKCE; female-specific hubs included SYN1, ATP9A, SNAP91, RAB6B. NAc Lightcyan1 (mitochondrial function) correlated with insomnia/hypersomnia negatively in males (r=−0.409, Padj=0.078) and positively in females (r=0.875, Padj<0.05); enriched for upregulated genes in male MDD (Padj<0.01). Hubs predominantly female-associated for the symptom (GPX4, PSMB5, PSMB6, PRDX5, ASNA1, EIF4H), with UROD shared. - Reproducibility: Strong overlap of current transcriptional profiles with previous cohorts at DEG and RRHO levels, supporting robustness of findings.

The study demonstrates that distinct MDD symptom domains map onto sex-specific transcriptional networks across multiple brain regions. Although many network structures are preserved across sexes, their modular connectivity changes and symptom associations diverge markedly between males and females. Key implicated pathways include GABAergic and glutamatergic neurotransmission, mitochondrial metabolism, intracellular signaling, and synaptic processes. Findings align with neurocircuit models linking region/network dysfunction to symptom dimensions and provide a molecular substrate for these associations. Notably, even when modules are associated with the same symptom in both sexes, the direction of association and the specific driving genes frequently differ, suggesting sex-dependent mechanisms within shared functional pathways. These transcriptional reorganizations may alter regional activity and broader network connectivity that underlie symptom expression, consistent with evidence from stress models. The work supports a precision psychiatry framework wherein symptom-targeted, sex-informed molecular signatures guide mechanistic hypotheses and potential interventions.

This study dissects MDD heterogeneity by linking symptom domains to sex-specific, regionally distributed transcriptional networks. It identifies conserved network structures with divergent symptom associations by sex and highlights hub genes and pathways (e.g., GABAergic signaling in male OFC; synaptic/catalytic modules in female aINS and dIPFC) that may underlie particular symptoms. The approach validates the feasibility of systems biology methods to map clinical features onto molecular architectures and encourages larger, longitudinal, symptom-resolved cohorts, ideally integrating peripheral transcriptomics and GWAS, to stratify patients and tailor diagnosis and treatment toward symptom dimensions rather than broad syndromes.

- Power constraints: Despite a large multi-region dataset, sample size limited the ability to analyze symptom co-occurrence or clusters at the network level. - Postmortem design: Precludes assessment of symptom intensity, recurrence, temporal dynamics, and fine-grained phenotypes (e.g., distinguishing insomnia vs hypersomnia; appetite gain vs loss; psychomotor agitation vs retardation), which may have distinct substrates. - DEG multiple testing: Differential expression used nominal P-value thresholds without multiple-testing correction, which may include false positives; downstream network analyses were multiple-testing corrected. - Functional validation: Predicted roles of modules and hubs were not experimentally validated in this study. - Batch and cohort effects: Although modeled and adjusted (RUVseq, covariates), residual confounding cannot be fully excluded.