Cell type specific transcriptomic differences in depression show similar patterns between males and females but implicate distinct cell types and genes

Major depressive disorder (MDD) affects 200–300 million people worldwide and is a leading cause of disability. Epidemiological and clinical data show pronounced sex differences: women have roughly double the prevalence, more comorbid anxiety and recurrent episodes, whereas men have higher comorbid substance use and suicide mortality. Prior genetic and transcriptomic studies indicate heritability and brain transcriptomic alterations in MDD, but the molecular etiology remains incompletely understood. Recent human bulk-tissue transcriptomic studies and animal models suggest largely sex-specific molecular signatures in MDD with minimal overlap in DEGs across sexes. Given the higher prevalence in women and evidence for sex-specific mechanisms, the study aimed to dissect sex- and cell type-specific transcriptomic alterations in the dorsolateral prefrontal cortex (dIPFC) using single-nucleus RNA sequencing (snRNA-seq), testing whether global depression-associated gene expression patterns within cell types are shared between sexes while identifying sex-specific cell types and genes driving MDD.

Previous bulk transcriptomic studies in human MDD reported distinct male versus female differential expression patterns with limited overlap (e.g., Labonté et al., 2017; Seney et al., 2018). Reviews posit reduced microglial activation and increased synaptic connectivity in females with MDD, with opposite trends in males. snRNA-seq has provided cellular resolution in neuropsychiatric disorders and prior work in male MDD PFC implicated deep layer excitatory neurons and oligodendrocyte precursor cells (OPCs) (Nagy et al., 2020). Sex differences in immune responses and microglial biology are well documented in humans and rodents, including hormonal and sex-chromosome influences on microglial function and brain physiology. Collectively, this literature motivated a sex-stratified, cell type-resolved investigation to clarify concordant and discordant transcriptional changes across sexes in MDD.

Design and cohorts: snRNA-seq was performed on postmortem human dIPFC (Brodmann area 9). Female cohort: 20 MDD cases and 18 neurotypical controls obtained from Douglas Bell-Canada Brain Bank and University of Miami Brain Endowment Bank. Male cohort: previously published dataset (GSE144136) reprocessed with a unified pipeline. Total: 71 subjects (37 cases, 34 controls) and 160,711 high-quality nuclei (51% female; 58% MDD). Nuclei isolation and library prep: Nuclei extracted from frozen dIPFC gray/white matter sections; 10x Genomics Chromium Single Cell 3' v2/v3 chemistries used (loading targeted ~3000 nuclei/sample). Sequencing primarily on Illumina NovaSeq 6000 (female cohort; two samples on BGI DNB-seq). Alignment: Cell Ranger v5.0.1 against GRCh38, including introns. Preprocessing and integration: Quality filters (mitochondrial % <10; cohort-specific UMI/gene thresholds) yielded 79,058 male and 81,653 female nuclei. SCTransform per sample; PCA (100 PCs) with Harmony batch/chemistry/sample correction; UMAP visualization. Clustering and annotation: Seurat clustering optimized via scclusteval (Jaccard-based stability), selecting 70 Harmony PCs, k=30, resolution=0.7, yielding 41 clusters. Marker-based annotation identified 7 broad cell types: excitatory neurons (ExN), inhibitory neurons (InN), oligodendrocytes (Oli), astrocytes (Ast), OPCs, endothelial (Endo), and microglia (Mic); plus one mixed cluster. Subtypes included 20 excitatory and 10 inhibitory clusters (e.g., ExN10_L46, InN1_PV, InN9_PV). Oligodendrocyte lineage (OPCs/Oli) was examined with slingshot pseudotime. Cross-dataset validation: MetaNeighbor matched clusters to published human brain datasets (Allen Institute, STAB), confirming correspondence. Spatial label transfer from spatial transcriptomics supported layer annotations. Cell-type proportion analysis: Per-subject nucleus proportions compared (cases vs controls) with Wilcoxon tests, bootstrapped p-values, and subsampling robustness checks. Differential expression (DE): Pseudobulk per subject per cluster/broad type using muscat and edgeR (glmQLFit/QLF). Inclusion: ≥10 cells (broad) or ≥5 cells (cluster) per subject; covariates age, pH, PMI, batch. Significance: FDR<0.05 (Benjamini-Hochberg), |log2FC|>log2(1.1), expression in ≥3 samples. Outliers flagged (not removed). Female microglia sub-clustered to exclude OL-like contaminants; DE results robust to this filtering. Between-sex comparisons: Rank-Rank Hypergeometric Overlap (RRHO2) using signed statistics (logFC × -log10 p). Permutation analyses (100 label permutations within batch) assessed chance concordance and unique DEG counts. Meta-analysis: Fisher’s method (metaRNASeq) combined male and female uncorrected p-values per gene within each cell type/cluster; FDR<0.05; discordant-direction genes removed. Functional interpretation: Pre-ranked GSEA (FGSEA) on Reactome pathways for female Mic1, InN1_PV, InN9_PV using same ranking metric as RRHO. PsyGeNET text-mined gene–disease associations assessed enrichment and evidence indices. Protein–protein interactions (STRING v11.5, high confidence >0.7) for female Mic1 and PV DEGs; networks visualized in Cytoscape. Cell–cell communication (CellChat) on female Mic1 and PV nuclei estimated ligand–receptor interactions and pathway-level signaling differences (nboot=1000). WGCNA on pseudobulk expression (female Mic1 and combined PV clusters) identified gene modules associated with case status; module–DEG overlap and Reactome enrichment tested (clusterProfiler).

• Dataset and cell composition: 160,711 high-quality nuclei from 71 donors (37 MDD, 34 controls); 7 broad cell types and 41 clusters identified.
• Proportion shifts in MDD: Decreased astrocyte and OPC proportions in cases (Ast FDR=3.46×10^-6; OPC FDR=5.32×10^-6), with increased excitatory neuron proportion (FDR=0.0477). At cluster level, reductions in Ast1 (FDR=0.00188), Ast2 (FDR=0.00291), OPC1 (FDR=0.009799), and OPC2 (FDR=0.0168). Patterns similar across sexes.
• Between-sex global concordance vs DEG overlap: Few significant DEGs overlapped between sexes, yet RRHO2 revealed moderate-to-strong concordance of threshold-free MDD-associated patterns within most broad cell types (notably Ast, ExN, InN up/down direction consistency; Mic upregulated in MDD), with discordance particularly in oligodendrocyte lineage clusters (Oli2, Oli3, OPC1). Permutations showed real male–female correlations exceeded permuted in ~91% (broad) and ~90% (concordant clusters) of comparisons.
• Male-specific strongest DEGs: Broad types: astrocytes (90/151 DEGs, 60% of male broad DEGs) and OPCs (54/151, 36%). Clusters: ExN10_L46 (deep layer excitatory; 238/447, 53%) and Ast1 (98/447, 22%). Most DEGs downregulated in cases (broad: 110/151, 73%; cluster: 358/447, 80%) and largely cell-type specific (broad: 96% unique; cluster: 89% unique).
• Female-specific strongest DEGs: Broad types: microglia only (74/85 DEGs, 87%). Clusters: Mic1 had 68/180 (38%) of female cluster-level DEGs, 53/68 (78%) overlapping with broad-level microglial DEGs. 47/68 (69%) microglial DEGs confirmed as transcribed and translated in microglia using TRAP data from an LPS mouse model. Additional female DEGs concentrated in inhibitory neuron clusters: InN1_PV, InN9_PV, InN2_SST, InN8_ADARB2. Most DEGs upregulated in cases (broad: 70/85, 82%; cluster: 140/180, 78%) and cell-type specific (broad: 99% unique; cluster: 92% unique).
• Meta-analysis (combined sexes): Broad types: upregulated microglial DEGs (172; more than female alone) and downregulated astrocytic DEGs (53; fewer than male alone). OPC DEGs decreased (22 vs 54 male), suggesting sex discordance; oligodendrocyte DEGs increased (21 vs 7 male). Retention: 54% of male astrocyte DEGs and 76% of female microglial DEGs recovered. Clusters: Mic1 upregulated DEGs (128; > female alone) and ExN10_L46 downregulated DEGs (254; > male alone) were top findings.
• Functional pathways in females: GSEA in microglia showed negative enrichment of inflammatory pathways (Interferon-γ, IL4/IL13, IL10, TNFR2 non-canonical NF-κB) and positive enrichment of neuronal system pathways (voltage-gated K+ channels, class C/3 metabotropic glutamate receptors, neurexins/neuroligins). PV interneurons (InN1_PV, InN9_PV) showed negative enrichment of HSF1-related pathways (HSF1 activation/transactivation), and positive enrichment of cellular response to external stimuli and RNA metabolism; InN1_PV also enriched for innate/adaptive immune and estrogen receptor (ESR) signaling.
• Gene–disease associations: Female cluster-level DEGs enriched for depressive disorders in PsyGeNET (hypergeometric p=0.0378) and alcohol use disorders (p=0.0141); depressive disorders had the most associations (>60).
• Microglia–PV interactions: STRING PPI linked microglial and PV DEGs with shared directionality; notable pairs: ROBO2 (up in microglia) and its ligand SLIT3 (up in InN9_PV); ADAMTSL1 (up in microglia) and THSD4/ADAMTSL6 (up in InN1_PV), suggesting ECM- and guidance-related crosstalk affecting perineuronal nets (PNNs). CellChat indicated increased overall ligand–receptor communication within/between microglia and PV interneurons in cases, with decreased GAS pathway and increased SPP1 pathway. Specifically, probable increase in SPP1–integrin signaling and decrease in GAS6–MERTK signaling between these cell types in cases.
• WGCNA support: Microglia: 8/44 modules associated with case status; MEturquoise positively correlated with MDD (r=0.627, p=7.26×10^-10), overlapping upregulated microglial DEGs and enriched for ion channels, neurotransmitter receptors, neuronal system pathways. PV interneurons: 16/55 modules associated; downregulated DEGs overlapped MEturquoise negatively associated with MDD (r=-0.582, p=0.00016), with overlap of 30 Reactome pathways identified by GSEA (including HSF1 and ESR signaling). Upregulated PV DEGs overlapped modules MEred (r=0.568, p=0.0002) and MEgreenyellow (r=0.426, p=0.0085).

This study demonstrates that while the specific DEGs and implicated cell types differ between sexes in MDD, the global, threshold-free patterns of depression-associated gene expression changes within most cell types are largely concordant across sexes. In males, deep-layer excitatory neurons, astrocytes, and OPCs are most dysregulated; in females, microglia and PV interneurons show the strongest alterations. Female microglia display downregulation of both pro- and anti-inflammatory pathways alongside upregulation of neuronal system-related genes, aligning with theories of reduced microglial activation and increased synaptic connectivity in female MDD. PV interneurons in females exhibit signatures of cellular stress (HSF1 pathway downregulation) and estrogen receptor signaling changes. Protein interaction and ligand–receptor analyses suggest altered microglia–PV crosstalk, potentially mediated by ECM and guidance molecules (e.g., SLIT3–ROBO2, ADAMTS-like proteins) and shifts in SPP1–integrin and GAS6–MERTK pathways, which may influence PNN stability and neuronal function. These findings refine the understanding of sex-specific cellular contributions to MDD, highlight shared overarching transcriptional patterns, and suggest that therapeutic strategies might need to be tailored by sex and cell type, particularly targeting microglial and PV interneuron pathways in females and excitatory neuron–glia interactions in males.

Using the largest snRNA-seq dataset of human dIPFC in MDD to date, the study provides a sex- and cell type-resolved map of transcriptomic alterations. Key contributions include: (1) identification of predominant dysregulation in deep-layer excitatory neurons, astrocytes, and OPCs in males versus microglia and PV interneurons in females; (2) evidence that global, threshold-free expression change patterns within cell types are broadly similar between sexes, despite divergent significant DEGs; (3) meta-analytic consolidation highlighting Mic1 (upregulated) and ExN10_L46 (downregulated) as principal clusters across sexes; and (4) mechanistic hypotheses of altered microglia–PV communication in females via ECM/guidance and ligand–receptor signaling. Future research should incorporate larger and more diverse cohorts, unified prospective sample processing across sexes, spatial transcriptomics and proteomic validation to resolve cellular proximity and PNN involvement, and functional studies to test microglia–PV signaling (e.g., SPP1–integrin, GAS6–MERTK) and estrogen-related pathways as potential therapeutic targets.

• Sex datasets analyzed separately: inability to model sex×disease interactions directly; differences in library chemistry, tissue collection, or nuclei isolation across cohorts could contribute to observed sex differences, though a unified preprocessing and joint clustering were applied.
• DEG robustness: Permutations indicated female cluster-level DEGs were less robust than male cluster-level and broad-level results; microglial findings were stronger and corroborated by overlap with broad-level DEGs and WGCNA modules.
• Sample size and generalizability: Despite >160,000 nuclei, subject numbers (71) limit power for gene-level detection and may affect generalizability; further validation in larger, diverse populations is needed.
• snRNA-seq constraints: Cytoplasmic transcripts are underrepresented, potentially limiting detection of microglial activation states; no distinct disease-associated microglial subpopulation was identified.
• Inference limitations: STRING and CellChat analyses are hypothesis-generating; lack of spatial and protein-level validation prevents conclusions about microglia–PV proximity, PNN presence, or protein expression changes.
• Technical clusters: A few clusters likely reflect technical effects (e.g., ExN17, ExN5, Mix) but contributed minimally to DE results.