Pan-Cancer Analysis of a Depression-Related Gene Signature Identifies an Immunosuppressive Tumor Microenvironment and Links to Immunotherapy Response in Breast Cancer Authors: Guanling Chen, Baoliang Yin, Muhammad Khan, Yipeng Yang, Qian Sun, Tingfeng Zhang, Hong Hu, Jie Lin Published: 2026 ABSTRACT AbstractBackgroundDepression has been increasingly associated with cancer progression and patient outcomes, yet its molecular and immunological correlates within tumors remain incompletely understood. This study aimed to investigate depression-associated gene signature across cancers and evaluate their relationship with tumor biology, immune microenvironment, and clinical outcomes, with a focus on breast cancer.MethodsA depression-related gene signature comprising 382 genes was obtained from the Molecular Signatures Database and used to compute a DScore via single-sample gene set enrichment analysis (ssGSEA) across over 10,000 tumor samples from 33 cancer types in The Cancer Genome Atlas (TCGA). Tumor microenvironment features were assessed using ESTIMATE and CIBERSORT, and cell-type–specific patterns were evaluated using single-cell RNA sequencing. CACNA1H was validated using qPCR and IHC in breast cancer samples.ResultsDScore exhibited tissue-specific patterns across cancers and was elevated in several tumor types, including gastrointestinal, brain, renal, and endocrine malignancies. Higher DScore was associated with older age, advanced stage, and poorer survival across multiple malignancies, including breast, ovarian, uterine, and lung cancers, whereas inverse associations with survival were observed in selected brain, kidney, and endocrine cancers. Higher DScore was further associated with genomic instability, an immunosuppressive tumor microenvironment (TME), and poorer survival across multiple cancers, particularly in breast and ovarian cancers. In breast cancer, this phenotype was characterized by monocytic-lineage enrichment, increased M2 macrophages, and reduced CD8⁺ T-cell infiltration. Consistent with these findings, higher DScore was associated with reduced benefit from immune checkpoint therapy in breast and bladder cancers, with non-responders exhibiting higher DScore predominantly enriched in monocytic-lineage cells. CACNA1H was identified as a depression-related gene correlated with poor prognosis in breast cancer and was validated in independent cohorts.ConclusionDepression-associated gene expression patterns are linked to tumor biology and immune features across cancers, with particularly strong relevance in breast cancer. These findings highlight the need for further research into how psychological, molecular, and immune factors interact in cancer progression and treatment response.Keywords: depression, breast cancer, cancer risk, computational biology, immunotherapy, pan-cancer Introduction Depression imposes a substantial healthcare and socioeconomic burden, impacting approximately 300 million individuals globally.1,2 Under stress conditions, depression is triggered by alterations in the hypothalamic-pituitary-adrenal (HPA) axis and the immune system, resulting in an increase in the production of proinflammatory cytokines (such as TNFα and IL-6) and glucocorticoids.3,4 These factors also induce the conversion of tryptophan into kynurenine, leading to reduced serotonin production and an elevation in the formation of neurotoxic substances like glutamate.4–6 The persistent and chronic release of these elements plays a significant role in the manifestation of behavioral changes associated with depression, spanning from fatigue to thoughts and acts of self-harm. Depression is frequently linked to cancer and is a prevalent diagnosis among individuals affected by this disease.7–9 Clinical investigations in cancer patients have predominantly concentrated on understanding the relationship between depression and cancer-related morbidity and mortality.10,11 Notably, a positive correlation between depression and lung and liver cancers has been observed.11,12 Moreover, depression has been associated with decreased survival rates in cancer patients.13–15 Conversely, cancer patients are more susceptible to developing depression, which can complicate the underlying pathophysiology of cancer and potentially influence the response to cancer treatments.7–9,16,17 Additionally, cancer patients with depression may exhibit reduced adherence to cancer therapies, further impacting their overall treatment outcomes.18 The immune system serves as a common factor in both cancer and depression. Depression can trigger inflammatory responses, and inflammation has been associated with cancer development.4,19 Oxidative stress and the production of reactive oxygen species (ROS), which are involved in the pathophysiology of both depression and cancer, contribute to this link.20,21 Alterations in growth factors, such as brain-derived neurotrophic factor (BNDF), fibroblast growth factor-2 (FGF-2), and insulin-like growth factor-1 (IGF-1), which are critical for cancer progression, have also been observed in individuals with depression.22–25 Recent advancements have highlighted the impact of inflammation on the tumor microenvironment (TME), consisting of tumor cells, stromal cells, and various immune cells, in cancer development.19 The interplay between inflammation, immune response, and the TME provides further insights into the complex relationship between cancer and depression. While prior studies have largely focused on clinical associations between depression and cancer, emerging transcriptomic approaches provide an opportunity to investigate depression-associated molecular patterns within tumors. In this study, we utilized a depression-related gene signature to derive a Depression Score (DScore), which reflects transcriptional enrichment of depression-related genes rather than a direct measure of clinical depression. This distinction is important, as transcriptomic signatures may capture biological processes related to stress and immune regulation but do not directly represent patient-level psychological states. Through a pan-cancer analysis, we systematically evaluated the relationship between DScore and tumor biology, immune microenvironment features, and clinical outcomes. Breast cancer was selected for in-depth analysis based on its strong and consistent associations with immune remodeling observed in the pan-cancer analysis, as well as its clinical relevance in immunotherapy research. In this context, we further investigated CACNA1H, a candidate gene identified through transcriptomic analyses, and explored its potential role as a biomarker in breast cancer. Methods & Materials TCGA Pan-Cancer & GTEx Data Processing Transcriptome profiling data for over 10,000 samples encompassing 33 cancer types were retrieved from The Cancer Genome Atlas (TCGA) database (https://portal.gdc.cancer.gov/). These cancer types include adrenocortical carcinoma (ACC), bladder urothelial carcinoma (BLCA), breast invasive carcinoma (BRCA), cervical and endocervical cancers (CESC), cholangiocarcinoma (CHOL), colon adenocarcinoma (COAD), Lymphoid neoplasm diffuse large B-cell lymphoma (DLBC), esophageal carcinoma (ESCA), glioblastoma multiforme (GBM), head and neck squamous cell carcinoma (HNSC), kidney chromophobe (KICH), kidney renal clear cell carcinoma (KIRC), kidney renal papillary cell carcinoma (KIRP), acute myeloid leukemia (LAML), Lower grade glioma (LGG), liver hepatocellular carcinoma (LIHC), lung adenocarcinoma (LUAD), lung squamous cell carcinoma (LUSC), pancreatic adenocarcinoma (PAAD), mesothelioma (MESO), ovarian serous cystadenocarcinoma (OV), pheochromocytoma and paraganglioma (PCPG), prostate adenocarcinoma (PRAD), rectum adenocarcinoma (READ), sarcoma (SARC), skin cutaneous melanoma (SKCM), stomach adenocarcinoma (STAD), testicular germ cell tumors (TGCT), thyroid carcinoma (THCA), thymoma (THYM), uterine corpus endometrial carcinoma (UCEC), uterine carcinosarcoma (UCS), and uveal melanoma (UVM). To analyze gene expression profiles of normal tissues, we utilized The Genotype-Tissue Expression (GTEx) dataset (V8.0) (https://commonfund.nih.gov/GTEx/) comprising 15,253 samples obtained from 54 tissues from 838 healthy individuals. External MDD Dataset Processing and Analysis Transcriptomic data from whole blood samples of MDD patients (n = 128) and healthy controls (n = 64) were obtained from GEO dataset GSE98793. The dataset was generated using the GPL570 platform (Affymetrix Human Genome U133 Plus 2.0 Array). Probe identifiers were mapped to gene symbols using the GPL570 annotation file, and when multiple probes mapped to the same gene, their expression values were averaged. The expression matrix was log2-transformed when required and normalized using the limma package. DScore was then calculated using ssGSEA implemented in the GSVA package based on the depression-related gene set. The resulting ssGSEA scores were normalized and used to compare depression-associated transcriptional activity between MDD patients and healthy controls. Statistical comparisons between groups were performed using the Wilcoxon rank-sum test, and p-values were displayed on the plots. Construction of the Depression Score Model The depression-related gene (DRG) signature (HP_DEPRESSION.v2022.1.Hs.gmt), comprising 382 genes and curated by the Human Phenotype Ontology (HPO) group, was obtained from the Molecular Signatures Database (MSigDB) (Table S1). Gene expression data from TCGA and GTEx datasets were processed in a consistent manner, including mapping to gene symbols, removal of duplicate genes, and exclusion of negative expression values. For TCGA data, expression values were log2-transformed [log2(x + 1)] prior to analysis. For each sample, a Depression Score (DScore) was calculated using single-sample gene set enrichment analysis (ssGSEA) implemented in the GSVA R package.26 The resulting enrichment score represents the relative enrichment of the depression-associated gene set within each sample. Genes not present in a given dataset were excluded from the analysis. Circadian rhythm and melatonergic gene-set enrichment scores were calculated using ssGSEA in the GSVA package, following the same procedure used for DScore calculation. Associations between DScore and circadian rhythm or melatonergic scores were evaluated using Pearson correlation analysis across TCGA cancer types and GTEx tissue samples. DRGs Differential Expression Across Cancers Differential expression analysis between tumor and normal samples was performed across TCGA cancer types. For each cancer type, samples were classified as tumor or normal based on TCGA barcode annotations, and only cancer types with available normal samples were included in the analysis. Gene expression data were preprocessed by removing negative values and imputing missing values using k-nearest neighbors. Differential expression analysis was conducted using the limma package. Genes with an absolute log2 fold change ≥ log2(1.5) and a false discovery rate (FDR) < 0.05 were considered differentially expressed. The numbers of upregulated and downregulated genes were summarized for each cancer type. Assessment of Clinical Features The prognostic potential of the DScore among cancers was assessed using the R package “survival”. Patients were stratified into high and low groups based on the median DScore within each cancer type. Data on phenotype-related characteristics, including immune subtypes and molecular subtypes of TCGA samples, were obtained from the UCSC Xena database (https://xenabrowser.net/). The association between DScore and various clinical features, such as body mass index (BMI) and cigarette exposure (pack years), was examined using Pearson correlation. Differences in DScore across clinical characteristics, including age, gender, race, tumor stage, immune and molecular subtypes, therapy response rates, and treatment outcomes, were evaluated using Wilcoxon rank-sum or Kruskal–Wallis tests, as appropriate. Survival outcomes were evaluated using Kaplan–Meier analysis with Log rank tests, and hazard ratios were estimated using univariate Cox proportional hazards regression. Single Cell Data Analysis To investigate the expression of DRGs at the single-cell level, we utilized The Tumor Immune Single-cell Hub (TISCH) (http://tisch.comp-genomics.org/), which provides curated tumor microenvironment (TME) annotations.27 TISCH employs the MAESTRO pipeline for preprocessing; therefore, additional quality control steps were not required.28 We downloaded the single-cell expression matrix and metadata files from the TISCH database for three breast cancer datasets (BRCA_GSE148673–10,359 cells; GSE1147727–19,676 cells; and BRCA_GSE136206_mouse_aPD1aCTLA4–27,532 cells), one colorectal cancer dataset (GSE146771–10,468 cells), and one bladder cancer (BLCA) dataset (GSE145282-aPDL1–14,462 cells). The data were processed using the Seurat R package. Normalization was performed using the NormalizeData function with the LogNormalize method and a scale factor of 10,000. Highly variable genes were identified using FindVariableFeatures (method = “vst”). Data scaling and dimensionality reduction were conducted using ScaleData, RunPCA, and RunTSNE functions, and cell cluster identities and lineage annotations were retained from the TISCH metadata. To quantify depression-associated transcriptional activity in breast cancer, a DScore was defined based on a curated depression-related gene signature. The signature was intersected with genes expressed in each dataset, and for each cell, DScore was calculated as the mean expression of the matched genes and stored as cell-level metadata within the Seurat object. DScore distributions across major cell compartments (immune, stromal, and malignant lineages) were visualized using violin plots with overlaid boxplots, and statistical differences between groups were assessed using Wilcoxon rank-sum tests implemented in the ggpubr package. Human DRGs were mapped to mouse orthologs through Ensembl BioMart prior to analysis of mouse single-cell transcriptomes. Additional datasets, including HNSC (GSE139324; 130,721 cells) and KIRC (GSE139555; 49,907 cells), were analyzed directly using the TISCH online platform under the Gene module. Gene Set Enrichment Analysis To investigate pathways associated with depression-related transcriptional activity, gene set enrichment analysis (GSEA) was performed following the methodology described by Subramanian et al (2005)29 For each tumor type, samples were stratified into high- and low-DScore groups corresponding to the top 30% and bottom 30% of DScore values. Differential expression analysis between these groups was conducted using the limma package, and genes were ranked based on log2 fold change. GSEA was subsequently performed using the clusterProfiler R package with Hallmark gene sets obtained from the Molecular Signatures Database (MSigDB), as well as curated immune-related gene sets representing immune pathways and functional cell-associated signatures. Enrichment significance was assessed using permutation-based testing, and multiple testing correction was applied using the Benjamini–Hochberg false discovery rate (FDR) method. All pathways were evaluated, and adjusted p-values (FDR) were used for interpretation of significance. Additionally, complementary functional enrichment analysis of the 382 DRGs was performed using the Metascape database (https://metascape.org/).30 Single Nucleotide Variation Analysis The “maftools” R package was used to construct the mutation frequency and oncoplot of 382 DRGs in 33 types of cancers.31 The mutation frequency was calculated by dividing the number of mutated samples by the total number of cancer samples. Additionally, the gene set SNV was calculated to assess the overall SNV status of the input DRGs for each sample. Tumor samples were categorized into two groups based on the presence of deleterious mutations: Mutant group (at least one gene in the input gene set was mutated) and WT group. The survival time and status within these groups were analyzed using the survival package in R. Cox proportional-hazards model and Log rank tests were conducted to examine the survival differences between the WT and Mutant groups. Somatic Copy Number Variation Analysis To conduct copy number analysis, GISTIC_2.0, available on GenePattern (https://cloud.genepattern.org), was utilized to identify genomes with significant amplifications or deletions.32 The burden of copy number changes, encompassing both losses and gains, was determined by calculating the total number of genes exhibiting copy number alterations at the focal and arm levels. Furthermore, the gene set CNV was computed, providing an indication of the integrated copy number variation status of the input DRGs for each sample. The samples were categorized into three groups: WT (wild-type), Amp. (amplified), and Dele. (deleted). A sample was classified as “Amplified” or “Deleted” if it consistently displayed amplification or deletion of at least one DRG. Samples with inconsistent gene status, where one gene was amplified while another was deleted, were excluded from the subsequent analyses. To evaluate survival differences between the groups, the R package survival was employed to fit survival time and status within each group. Cox proportional-hazards models and Log rank tests were performed to assess the survival disparities between the groups. External Validation To validate the prognostic significance of CACNA1H in independent cohorts, we employed the Kaplan-Meier Plotter database’s BRCA domain (https://kmplot.com/). Overall survival was analyzed using data from 1879 breast cancer patients derived from 50 GEO datasets. Profiling the Tumor Microenvironment Landscape The ESTIMATE algorithm was applied to infer tumor microenvironment composition by calculating immune and stromal scores, representing the levels of immune cell infiltration and stromal content, respectively.33 Gene expression data were filtered to retain tumor samples based on TCGA barcode annotations, and genes with zero mean expression were excluded prior to analysis. The relative abundance of 22 immune cell types was estimated using the CIBERSORT algorithm with the LM22 reference signature.34 Expression data were preprocessed and transformed using the voom method, followed by analysis with 1000 permutations and quantile normalization enabled. Only samples with CIBERSORT P values < 0.05 were retained for downstream analyses. To comprehensively evaluate tumor microenvironment and genomic features associated with DScore, Spearman correlation analysis was performed within each cancer type to assess relationships between DScore and multiple variables, including immune score, stromal score, immune cell fractions, stemness indices (RNAss and DNAss), and genomic instability indicators such as homologous recombination deficiency (HRD) and HRD-associated loss of heterozygosity (HRD-LOH). Stemness scores and genomic instability metrics were obtained from the UCSC Xena database (https://xenabrowser.net/). Correlation coefficients and corresponding P values were calculated, and results were visualized using bubble plots, with color representing the correlation coefficient and point size indicating −log10(P value). Independent Prognostic Evaluation of DScore Overall survival (OS) was analyzed using Cox proportional hazards regression. The continuous DScore was dichotomized into high- and low-DScore groups using the optimal cutoff determined by the surv_cutpoint function in the survminer R package. Univariate Cox regression analysis was first performed to evaluate the association of age, race, molecular subtype, tumor stage, and DScore group with OS. Variables were subsequently included in multivariate Cox regression analysis to assess their independent prognostic value. Hazard ratios (HRs) and 95% confidence intervals (CIs) were calculated, and a two-sided p < 0.05 was considered statistically significant. Immunotherapy Response Immunotherapy response datasets included the IMvigor210 dataset (Urothelial carcinoma; n=348; ICI=anti-PD-L1), GSE176307 dataset (metastatic urothelial carcinoma; n=88; anti-PD-1/PD-L1), GSE91061 dataset (melanoma; n=51), GSE78220 dataset (metastatic melanoma; n=27; ICI=anti-PD-1), and E-MTAB-3218 dataset (ccRCC; n=58) obtained from ArrayExpress. Cell Culture and Experimental Conditions Human breast cancer cell lines MCF-7 and BT549, along with the non-malignant breast epithelial cell line MCF-10A, were obtained from the Cell Bank of the Chinese Academy of Sciences (Shanghai, China). MCF-7 and BT549 cells were cultured in Dulbecco’s Modified Eagle’s Medium (DMEM) supplemented with 10% fetal bovine serum (FBS), 100 U/mL penicillin, and 100 µg/mL streptomycin. Cultures were maintained at 37 °C in a humidified incubator with 5% CO2. The cell lines were routinely validated by monitoring morphological features and screened for mycoplasma contamination to ensure experimental integrity. Quantitative Real-Time PCR (qPCR) Total RNA was extracted using TRIzol reagent (Takara, Otsu, Japan) according to the manufacturer’s instructions. The RNA was reverse-transcribed into complementary DNA (cDNA), which served as the template for subsequent qPCR assays. Amplification reactions were performed using SYBR Green PCR Master Mix (Takara, Otsu, Japan). Expression levels were normalized to GAPDH as an internal control, and relative quantification was determined using the 2−ΔΔCt method. Primer sequences used for amplification are listed in Table 1. Table 1Primers Used in This StudyGenesPrimersCACNA1H-ForwardATGCCGACTGCCACATAGAGCACNA1H -ReverseGTTGTACGGGTTCCGGATGTGAPDH-ForwardCCAGCAAGAGCACAAGAGGAGAPDH-ReverseGGGGAGATTCAGTGTGGTGG Primers Used in This Study Clinical Specimens Tumor and matched adjacent normal tissue samples were obtained from 12 patients with breast cancer treated between January 2024 and November 2025 at the Division of Breast Surgery, Department of General Surgery, Shenzhen People’s Hospital (The First Affiliated Hospital of Southern University of Science and Technology; The Second Clinical Medical College of Jinan University), Shenzhen 518020, China. All participants provided written informed consent, and the study protocol was approved by the Institutional Review Board and Ethics Committee of the same institution (approval number: LL-KY-2023144-01). Immunohistochemical Analysis Formalin-fixed, paraffin-embedded (FFPE) breast tumor tissues and matched adjacent normal specimens were sectioned at a thickness of 4 μm. Sections were deparaffinized in xylene and rehydrated through a graded ethanol series, followed by antigen retrieval in citrate buffer (pH 6.0) using microwave heating. Endogenous peroxidase activity was quenched with 0.3% hydrogen peroxide, and nonspecific binding was blocked by incubation with 5% bovine serum albumin (BSA) in phosphate-buffered saline (PBS). Slides were incubated overnight at 4 °C with a primary antibody against CACNA1H (mouse monoclonal, Invitrogen, #MA5-27656; dilution 1:1000). The following day, sections were treated with a biotinylated secondary antibody at room temperature, and visualization was achieved using 3,3′-diaminobenzidine (DAB) as the chromogen. Hematoxylin was applied as the counterstain. Staining was evaluated semi-quantitatively by two independent observers. Staining intensity was scored on a 4-point scale (0 = none, 1 = weak, 2 = moderate, 3 = strong), and the percentage of positively stained cells was scored on a 5-point scale (0, <5%; 1, 5–25%; 2, 26–50%; 3, 51–75%; 4, >75%). The final IHC score for each specimen was obtained by multiplying the intensity and percentage scores. In cases with heterogeneous staining, distinct regions were scored separately and averaged to obtain the final value. Statistical Analysis To compare the two or more groups, we employed statistical tests including Student’s t-test/Wilcoxon test and ANOVA/Kruskal–Wallis test. Correlation analysis was performed using Spearman correlation test and Pearson correlation test as appropriate. For assessing prognostic significance, Kaplan-Meier survival curves with Log rank tests were utilized. Survival risk was estimated using the Cox proportional-hazards model, and the Log rank test was used to determine statistical differences. A significance level of p < 0.05 was considered statistically significant. All statistical analyses were conducted using R v4.0.3 software (http://www.r-project.org). Results Pan-Cancer Distribution of a Depression-Related Gene Signature The association between depression and cancer has been a topic of ongoing discussion for several years. In order to explore this connection, we assessed the enrichment of a depression-related gene signature (DScore) (HP_DEPRESSION.v2022.1.Hs.gmt), which consists of 382 genes obtained from The Molecular Signatures Database (MSigDB), across 33 cancer types in The Cancer Genome Atlas (TCGA) (Table S1 and Figure 1A). The analysis of these genes using the Metascape database revealed their robust involvement in behavioral and neurological processes, as depicted in Figure S1A. Probing the depression signature in detail by breaking it down into smaller protein-protein interactive networks, known as MCODE (the Molecular Complex Detection algorithm), revealed the involvement of six major interactive networks (Figure S1B).35 These networks encompassed critical pathways, namely: 1) G protein coupled receptor signaling; 2) Mismatch repair; 3) PD-1 signaling; 4) regulation of protein kinase activity; 5) ATP synthesis and electron transport chain; and 6) neuron projection maintenance and organization. These components represent essential factors that have been previously implicated in the pathophysiology of depression, including neurotransmitter dysregulation, genetic variations, stress response abnormalities, inflammatory processes, and neuroplasticity changes.36–39 This breakdown provides valuable insights into the underlying molecular mechanisms and biological processes associated with depression. Upon projecting the DScore onto a dataset containing more than 9000 normal tissue samples extracted from the Genotype-Tissue Expression (GTEx) database, we observed a higher DScore in the pituitary gland, brain, nerves, and adrenal gland compared to other organs (Figure 1B). This finding highlights the significance of DScore in relation to depression, as these organs are primarily involved or affected in depressive conditions.4 Further validation of this observation was conducted by analyzing cell- and tissue-specificity using the PaGenBase database, which confirmed the cortex as the main tissue associated with the enrichment of DScore (Figure 1C).40 These results highlight the specific involvement of the cortex and provide additional evidence supporting the relationship between the DScore and depression. The relationship between depression and circadian rhythm as well as melatonin activity has been established.41–43 To investigate this association, we examined data from normal tissue expression in the GTEx database and cancer samples from 33 types of cancer in TCGA database. Our findings revealed a positive correlation between DScore and circadian rhythm-related genes in both GTEx and TCGA datasets, suggesting an association between depression-related transcriptional activity and circadian regulation (Figure 1D and E). Conversely, a negative correlation was observed between melatonergic genes and DScore in the majority of normal and cancer samples, suggesting an inverse relationship between melatonergic pathways and depression-related transcriptional activity43,44 (Figure 1F and G). These findings further solidify the notion that the DScore capture biological processes relevant to depression. Furthermore, a significant increase in DScore was observed in blood samples collected from patients with major depressive disorder (MDD) when compared to samples from healthy individuals45 (Figure 1H). Additionally, enrichment analysis using DisGeNET indicated the association of DRGs with the common adverse symptoms associated with depression, and are also commonly encountered by cancer patients during their illness and treatment (Figure 1I).46 Taken together, these findings support the use of DScore as a transcriptional proxy of depression-related activity. Assessing the reliability of DScore as an indicator of depression-related transcriptional activity. (A) Schematic overview of the analytical workflow, including (i) calculation of the Depression Score (DScore) from gene signature enrichment, (ii) analysis of its associations with genomic and functional features across cancers, and (iii) evaluation of its clinical and immunotherapy relevance. (B) Bar plot illustrates the DScore across 31 types of normal tissues. (C) Summary of gene set enrichment analysis (GSEA) in PaGenBase depicting the enrichment of DRGs in various tissues and cells. (D-G) Correlation between DScore and the enrichment of circadian rhythm (D and E) and melatonergic (F and G) gene sets in normal tissues (GTEx; D, (F) and cancers (TCGA; E, G). Pearson correlation; color indicates statistical significance at P < 0.05. (H) Comparison of DScore, as determined by the gene set enrichment analysis of DRGs, between patients with major depressive disorder (MDD) and healthy controls. Wilcoxon rank-sum test; P < 0.05 was considered significant. (I) Summary of GSEA in DisGeNET depicting the enrichment of DRGs in various human disease/symptoms. By utilizing gene set enrichment analysis on a vast collection of over 10,000 samples across 33 different cancer types from TCGA database, we investigated the distribution of DScore across cancers, as illustrated in Figure 2A and B. Our findings revealed a distinct tissue-specific pattern in DScores (Figure 2B and C). Notably, cancers originating from the brain (GBM, LGG), digestive system organs (gastrointestinal tract [GI], hepatobiliary system), kidneys and secretory glands (ACC, PCPG) exhibited a higher DScores. Conversely, lung cancers (LUAD, LUSC), soft tissue cancers (SKCM, SARC, MESO), female reproductive organ (FRO) cancers (UCEC, CESC), and hematopoietic and immune cancers (LAML, THYM, DLBC) displayed lower DScores. When comparing DScore in cancer tissues with adjacent normal tissues, we observed elevated DScore in gastrointestinal cancers (COAD, HNSC, LIHC, STAD), lung cancers (LUAD, LUSC), THCA and BRCA, while BLCA, GBM, and kidney cancers (KIRC, KIRP) showed lower DScore in cancer samples compared to their normal tissue counterparts (Figure 2D). Figure 2E provides a visual representation of the distinct patterns of differentially expressed DRGs in GI cancers and male/female reproductive organ (M/FRO) cancers. In GI cancers, there is a noticeable increase in the number of upregulated DRGs, whereas M/FRO cancers exhibit a relatively balanced distribution of both up- and down-regulated DRGs. Interestingly, the analysis presented in Figure 2F reveals a higher number of shared differentially expressed genes among reproductive organ cancers, whereas Figure 2G demonstrates a lower number of shared DRGs in GI cancers. These findings suggest that the patterns associated with DScore vary depending on the specific origin of the cancer. Collectively, these results highlight tissue-specific variation in DScore across cancers and suggest heterogeneous patterns of depression-associated transcriptional activity. Comprehensive quantification of DScore in cancers and tissues. (A) Circos diagram illustrating the distribution of DScore across various cancer types. Tissue and cancer types, along with participant counts and corresponding DScore, are depicted from the inner circle to the outer circle. (B) The box chart presents the DScore in different cancer types, arranged in descending order from high to low DScore. (C) The box chart displays the DScore in various tissue-specific cancer types, arranged in descending order from high to low DScore. (D) Comparison of DScore between cancer samples (Orange) and adjacent normal tissues (purple). Wilcoxon rank-sum test; *P < 0.05; **P < 0.01; ***P < 0.001. (E) Bar plot illustrating the number of up- and down-regulated DRGs in various cancers. (F) Upset plot represents the number of intersecting sets of differentially expressed DRGs in cancers derived from reproductive organs. (G) Upset plot represents the number of intersecting sets of differentially expressed DRGs in cancers originating from organs comprising the gastrointestinal system. Association Between DScore and Genomic Alterations Across Cancers Emerging evidence suggests a noteworthy association between depression and genomic alterations.37,47,48 The exploration of mutations, copy number variations, and genomic instability in the context of depression holds promise for advancing our knowledge of the biological underpinnings of depression in cancer. To assess this relationship, we examined the influence of DScore on cancer mutations and genomic instability. Frequent mutations were observed in individual DRGs across various cancer samples (Figure 3A). Among the top five cancers with the highest mutation frequency in DRGs were UCEC, SKCM, COAD, STAD, and LUAD (Figure 3A). These alterations were detected in 56.83% of the cancer samples, representing 4480 out of 7883 samples (Figure S2A). Missense mutations were the predominant single nucleotide variants (SNVs), with C>T mutations being the most common type (Figure S2B). The top five mutated DRGs, namely PIK3CA (18%), USH2A (14%), ADGRV1 (11%), PCDH15 (11%), and KRAS (10%), are well-established cancer-specific oncogenes (Figure S2A). Our results also demonstrated the cancer-specific nature of these mutations. For example, PIK3CA was found to be frequently mutated in over eight different cancer types, USH2A in four cancers (LUAD, LUSC, SKCM, COAD), KRAS in three cancers (LUAD, COAD, PAAD), ADGRV1 in SKCM, and ARTX in LGG (Figure S2A). To assess the potential association between DScore and cancer-specific oncogenes, we examined the mutation frequency of BRCA driver oncogenes in depression-stratified BRCA subgroups. We found that PIK3CA, TP53, and TTN mutations were more common in the BRCA low DScore subgroup (Figure 3B and C). A similar negative association was also observed between DScore and total mutation burden (TMB) in COAD and LGG, while positive correlations were seen in CHOL, SKCM, TGCT, and UCEC (Figure 3D). To assess the association between DScore and cancer stemness scores and genomic alterations, we examined their correlation. We found a predominantly positive association between Dscore and DNAss, while a contrasting pattern was observed with RNAss (Figure 3D). Furthermore, a consistent positive correlation pattern emerged between Dscore and genomic instability markers, including microsatellite instability (MSI), homologous recombination deficiency (HRD), and HRD with loss of heterozygosity (HRD-LOH), in GI cancers (COAD, ESCA, HNSC), lung cancers (LUAD, LUSC, MESO), and cancers of the female reproductive organs (UCEC and UCS). The pattern of this correlation for BRCA is highlighted in Figure 3E. The relationship between DScore and genomic variations at the pan-cancer level. (A) Histogram showing frequency of single nucleotide variations in each cancer type for top 30 DRGs. (B) The Spearman correlation between genomic alterations/instability markers and the DScore. Blue & red represent negative and positive correlation respectively. (C) The heatmap displays the top mutation events for each TCGA BRCA patient categorized into high- and (D) low-DScore subgroups. (E) Circular chart illustrates the correlation between genomic alterations/instability markers and the DScore in TCGA BRCA dataset. (F) The Spearman correlation between somatic copy number variations (SCNV), comprising focal- and arm-level amplifications and deletions, and the DScore. Green & red represent negative and positive correlation respectively. (G) Focal-level copy number alteration patterns between high- and low-DScore subgroups in TCGA BRCA, summarized by cytoband and visualized using LOESS smoothing. (H) Arm-level amplification and deletion frequencies between high- and low-DScore subgroups in TCGA BRCA, with statistical significance assessed using Fisher’s exact test; *P < 0.05. Next, we examined the influence of DScore on the frequencies of somatic copy number variations (SCNVs). SCNVs were assessed in terms of both focal and arm-level burden, encompassing gains and losses, and exhibited a similar pattern to that observed for mutations and genomic alterations (Figure 3F). Notably, BRCA, KIRC, and LGG showed higher focal and arm-level burden in the low DScore subgroups, indicating increased SCNV frequencies in these cancers. Conversely, GI cancers and UCEC displayed the opposite trend, with higher CNV frequencies in the high DScore subgroups. In case of BRCA, we can precisely observe the greater amplification and deletion frequency at various focal and arm levels in low DScore subgroup (Figure 3G and H). When examining the amplification and deletion rates of individual DRGs, a negative correlation was observed between the average amplification rate and DScore at pan-cancer level (Figure S2C). Conversely, a positive trend was apparent in the case of deletions, indicating a higher occurrence of deletions among these genes as a collective group in individuals with high DScore (Figure S2D). Furthermore, we conducted an assessment of how genomic variations may affect cancer prognosis. It was observed that gene set SNVs had a significant impact on prognosis in various cancers, such as BRCA, KIRC, KIRP, PCPG, SKCM, and UCS (Figure S2E–L). Patients with mutant gene set SNVs in BRCA, KIRC, PCPG, and UCS exhibited the poorest prognosis (Figure S2E–I). Additionally, the influence of gene set copy number variations (CNVs) on prognosis was evident in BLCA, KIRP, and THYM (Figure S2L). Collectively, these results highlight the association between depression-related transcriptional activity and genomic alterations in cancer. DScore Shapes Inflammatory and Tumor Microenvironmental Features Across Cancers, Marked by Immunosuppressive Microenvironment in Breast Cancer The TME has long been recognized as a critical factor influencing cancer progression and response to treatment. The TME consists of a complex interplay between tumor cells, immune cells, and stromal cells, with their relative proportions varying across different cancer types.33 In this study, we employed the ESTIMATE (Estimation of STromal and Immune cells in MAlignant Tumor tissues using Expression data) algorithm to assess the influence of DScore on TME characteristics. DScore exhibited a positive correlation with Tumor parity, which represents the tumor component of the TME, and demonstrated a negative association with the proportions of immune cells (immune score), stromal cells (stromal score), and the non-tumor components of the TME (ESTIMATE score) across various cancer types, except for GI cancers, LAML, OV, UCS, and UVM (Figure 4A). Moreover, we confirmed these findings by examining single-cell datasets of BRCA and CRC, which revealed contrast TME characteristics associated with DScore. As illustrated in Figure 4B–F, the DScore exhibited significantly higher enrichment in malignant cells compared to immune and stromal cells in breast cancer. In contrast, Figure 4G–K depicted a contrasting pattern in colorectal cancer, wherein malignant cells displayed the lowest enrichment, while stromal and immune cells were more enriched with DRGs. These findings were consistent with the enrichment analysis of inflammatory and immune pathways observed in various cancer types (Figure 4L and M). In GI cancer patients with a high DScore, a notable enrichment of several inflammatory and immune pathways was observed through gene set enrichment analysis. These patients exhibited a pronounced level of inflammation, increased expression of human leukocyte antigens (HLA) and chemokine receptors (CCR), which coincided with a high infiltration of lymphocytes (TIL) and the presence of checkpoint molecules (Check-point) (Figure 4L). These findings suggest that DScore may have potential relevance for immunotherapy response in specific cancer types. Furthermore, our evaluation of hallmark cancer-promoting inflammatory pathways further supported the observed pattern of enrichment, as illustrated in Figure 4M. Figure 4N demonstrates the enrichment of cancer-related pathways, including the P53 pathway, apoptosis, hypoxia, reactive oxygen species (ROS), epithelial mesenchymal transition (EMT), WNT/β-catenin signaling, TGF-β signaling, and angiogenesis, primarily in cancer subgroups with high DScore. These pathways have also been associated with an immunosuppressive TME.49–51 In conclusion, DScore is associated with distinct tumor microenvironment characteristics in cancer, including altered immune and stromal cell proportions and enrichment of pathways linked to immunosuppressive features, particularly in breast cancer. DScore Correlates with Tumor Microenvironment Composition and Immune–Inflammatory Signaling. (A) Heatmap summarizes Spearman correlation between DScore and ESTIMATE-stratified components of TME features including non-tumor components (stromal and immune scores) and tumor component (tumor parity). The heatmap indicates positive correlations in red and negative correlations in blue. (B) t-SNE plots showing major cell types, (C) TME lineages, and (D) DScore distribution in the single-cell breast cancer dataset (GSE148673), respectively. (E) Violin plot depicting enrichment of DScore-associated DRGs across major cell lineages. Wilcoxon rank-sum test; ****P < 0.0001. (F) Bar plot illustrating the proportions of single cells with high and low DScore in each lineage. (G) t-SNE plots showing major cell types, (H) TME lineages, and (I) DScore distribution in the single-cell colorectal cancer dataset (GSE146771), respectively. (J) Violin plot depicting enrichment of DScore-associated DRGs across major cell lineages. Wilcoxon rank-sum test; ****P < 0.0001. (K) Bar plot illustrating the proportions of single cells with high and low DScore in each lineage. (L–N) Gene set enrichment analysis (GSEA) comparing high- and low-DScore tumor subgroups using curated immune-related gene sets (L), Hallmark inflammatory and immune pathways (M), and Hallmark oncogenic and immunosuppressive pathways (N). Genes were ranked by log2 fold change derived from limma differential expression analysis, and enrichment significance was assessed using permutation-based testing with Benjamini–Hochberg false discovery rate (FDR) correction. NES indicates normalized enrichment score. Monocytic-Lineage Enrichment Characterizes DScore-Associated Immunosuppressive Features Across Cancers, with Macrophage Dominance and CD8⁺ T Cell Depletion in Breast Cancer We further explored the details of DScore relationship with the cancer TME features in terms of infiltration of immune cells. Dissecting immune composition of the TME demonstrated a positive correlation, that was consistent across tumors with few exceptions, between DScore and infiltration of monocytes and macrophages (M0 and M2) (Figure 5A). DScore showed a positive correlation with M0 and M2 (anti-inflammatory phenotype), and rather negative association with M1 (pro-inflammatory phenotype), indicating the presence of an anti-inflammatory immunosuppressive TME phenotype (Figure 5B). Simultaneously, a negative trend in the infiltration of CD8+ T cells was also evident, which supports the description of cold TME (Figure 5A and C). In Figure 5D, we can clearly observe the infiltration patterns of various types of immune cells deconvoluted by TIMER and CIBERSORT algorithms for TCGA BRCA categorized into high- and low-DScore subgroups. The TIMER algorithm indicates the low CD8+ T cell and high macrophage infiltration in the DScore-high subgroup. While CIBERSORT algorithm further breaks down the infiltration of specific macrophage phenotypes with high M0/M2 and low M1 enrichment in high-DScore subgroup. Moreover, association of DScore with TCGA immune subtypes revealed a high DScore in C4 and C5 which are characterized by a lower infiltration of lymphocytes (Figure 5E). Moreover, 20 types of cancers had significant difference in the DScore of their immune subtypes (Figure S3A). We further validated these observations in a single cell dataset of BRCA (BRCA_GSE114727) which was exclusively comprised of immune cell subsets (Figure 5F and G). The enrichment of DScore in this dataset showed that DRGs are predominantly expressed in monocytes and macrophages as compared to T cell subsets (Figure 5H). Violin plot of single cell expression pattern also showed a higher expression in monocytic-lineage cells as shown in Figure 5I. The DScore of Mon/Macro cells was significantly higher than the CD8 T cells (Figure 5J). Moreover, there were significantly more M2 phenotype macrophages with high DScore than the M1 phenotype, complementing our bulk RNA-sequenced analysis in breast cancer (Figure 5K). A similar pattern of DScore dynamics could be appreciated for KIRC (Figure S3B–D) and HNSC single cell dataset (Figure S3E–G). In summary, these findings highlight the enrichment of M2 macrophages and reduced T cell infiltration in association with DScore in cancer. DScore, Inflammation, and Characteristics of the Tumor Microenvironment. (A) Heatmap summarizes Spearman correlation between pyroptosis level and infiltration of 22 types of immune cells estimated by CIBERSORT algorithm. Red & blue indicate positive and negative correlation. (B and C) Pearson’s correlation between DScore and the CIBERSORT-estimated fractions of M2 macrophages (B) and CD8⁺ T cells (C) across TCGA cancers; color indicates statistical significance at P < 0.05. (D) Heatmap illustrates the infiltration pattern of immune cells, assessed by TIMER and CIBERSORT algorithms, and distribution of immune subtypes in DScore-stratified TCGA BRCA subgroups. (E) Boxplot depicts the DScore of TCGA immune subtypes. (F) t-SNE plots showing major cell types, (G) TME lineages, and (H) DScore distribution in the single-cell breast cancer dataset (GSE114727), respectively. (I) Violin plot showing enrichment of DScore-associated DRGs across major cell types. (J and K) Violin plots comparing DScore enrichment between CD8⁺ T cells and monocyte/macrophage cells (J), and across CD8⁺ T-cell subtypes (Tcm and Tem) and macrophage subtypes (M1 and M2) (K). Wilcoxon rank-sum test; ***P < 0.001; ****P < 0.0001; ns, not significant. DScore Correlates with Clinical and Molecular Features and is Associated with Poor Outcomes in Breast Cancer Depression has been linked to various patient characteristics, including age, gender, and socioeconomic status.52,53 In our study, we observed noteworthy variations in DScore across different age groups in seven types of cancer (ESCA, KICH, LAML, LIHC, PAAD, SARC, and THYM) (Figure 6A). Specifically, patients aged over 60 exhibited higher DScore in ESCA, LAML, LIHC, and THYM. Gender associations were also evident in select cancers, with depression appearing to impact kidney cancers (KIRC, KIRP) in women and gastrointestinal (GI) cancers (HNSC, READ) in men. Racial disparities were noted in DScore among seven cancers (BRCA, HNSC, KIRC, LIHC, OV, READ, STAD) (Figure 6A and B). In GI cancers (LIHC, STAD, HNSC) and ovarian cancer (OV), Black individuals had elevated DScore followed by White and Asian individuals. Moreover, Asian patients with KIRC and White patients with BRCA demonstrated the highest DScore compared to other racial groups. Cigarette smoking and body mass index (BMI) showed a positive trend with DScore in GI cancers (Figure 6B). DScore also exhibited an association with cancer stage in six cancers (COAD, KIRC, KIRP, PAAD, TGCT, and UVM), with a negative trend observed in the enrichment of DRGs from stage I to stage IV in KIRC, KIRP, and PAAD, while the opposite trend was evident in COAD, TGCT, and UVM (Figure S4A). Significance was also reached in BRCA when the stage categories were restricted to early-stage (T1/2; I/II) and late-stage (T3/4; III/IV) (Figure 6D and E). Notably, significant differences were observed among cancer subtypes, with aggressive and invasive molecular subtypes such as mesenchymal (GBM, HNSC) and basal (BRCA, HNSC, LUSC) associated with lower DScore (Figure 6F–H and Figure S4B–E). Similarly, high microsatellite instability (MSI-H) exhibited lower DScore compared to microsatellite stable (MSS) and low microsatellite instability (MSI-L) (Figure 6H and S4E). The dynamics of MSI instability; however, varied among cancers as illustrated in Figure S4E. These findings highlight associations between DScore and clinical and molecular features, underscoring their potential relevance in cancer assessment and management. Clinical relevance of DScore. (A) Association of DScore and patient’s characteristics. Wilcoxon rank-sum test; P value are shown as *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001. (B) Correlation between DScore and cigarette smoking (pack years exposure) and body mass index (BMI). (C) Heatmap illustrating association between DScore and clinicopathological features of the breast cancer patients (TCGA BRCA). Chi-square test; P value are shown as *P < 0.05; ***P < 0.001. (D) Boxplot illustrates the comparison of DScore between breast cancer stages categorized at TNM (T1-4) and (E) AJCC staging (I–IV). (F) Boxplot shows the comparison of DScore among breast cancer, (G) glioblastoma subtypes. (H) Boxplot illustrates the association of DScore and microsatellite instability (MSI) subtypes across cancers. (I) Heatmap depicts the survival significance of DScore across TCGA cancers assessed with cox-regression and Kaplan-Meier survival analysis. NS = not significant (J) Kaplan-Meier curves depicting overall survival (OS), (K) progression-free survival (PFS), and (L) disease-specific survival (DSS) difference between low DScore and high DScore BRCA and (M–O) KIRC subgroups. (P–Q) Forest plots showing univariate Cox regression analysis (P) and multivariate Cox regression analysis (Q). Hazard ratios (HRs) with 95% confidence intervals and corresponding P values are presented. Furthermore, survival analysis was performed to seek the prognostic relevance of DScore in cancer patients. Risky trend of DScore was evident in GI (HNSC, COAD, ESCA, READ, and STAD), female reproductive organ (BRCA, OV, UCEC, and UCS) and lung cancers (LUAD, LUSC) (Figure 6I). Notably, statistically significant results were found in BRCA, OV, UCEC, and UVM, indicating a strong prognostic association of DScore in these cases (Figure 6I–L and Figure S4F-H). Conversely, a protective role of DScore was demonstrated in brain cancers (GBM, LGG), kidney cancers (KICH, KIRC, KIRP), secretory tissue cancers (ACC, PAAD, PCPG), and hepatobiliary cancers (LIHC, CHOL) (Figure 6I and M–O). The significant protective trend was observed in ACC, LGG, KIRC, and PAAD (Figure 6M–O and Figure S4I-M). To further assess the prognostic value of DScore in breast cancer, patients within the BRCA cohort were stratified into high- and low-risk groups based on the optimal cutoff derived from the DScore. Univariate Cox regression analysis identified age, tumor stage, and DScore-based risk group as factors significantly associated with overall survival (Figure 6P). Subsequent multivariate Cox regression analysis demonstrated that age (HR = 1.040, 95% CI: 1.024–1.056, p < 0.001), tumor stage (HR = 2.382, 95% CI: 1.824–3.112, p < 0.001), and DScore-based risk group (HR = 1.686, 95% CI: 1.050–2.707, p = 0.031) remained significantly associated with overall survival (Figure 6Q). Importantly, patients in the high DScore group exhibited significantly poorer survival outcomes compared to those in the low DScore group, indicating that DScore serves as an independent prognostic factor in BRCA beyond established clinical variables. DScore and Therapeutic Sensitivity Across Cancers, with Links to Immune Response and Immunotherapy Outcomes in Breast and Bladder Cancer Conventional cancer treatments consist of relatively aggressive interventions, such as radiotherapy and chemotherapy, which induce direct DNA damage and cellular death.54,55 Consequently, the association between DScore and treatment outcomes for these therapies may appear limited. As anticipated, there was no disparity in the number of patients with tumors or those who were tumor-free after cancer therapy in the majority of cancers (n=29) based on their DScore (Figure S5A). In KIRC and PCPG, DScore was associated with tumor-free status while the opposite was true for ESCA and SKCM. However, if we disregard the heterogeneity caused by variations in tumor types, it was evident that patients with tumors had significantly higher DScore compared to those without tumors (Figure 7A). Likewise, there were no significant variations in the response to therapy at the individual cancer level, except for PCPG (Figure S5B–C). However, when considered collectively, a lack of response was found to be associated with the DScore (Figure 7B). Furthermore, cancer patients exhibiting high DScore demonstrated a higher incidence of tumor recurrence rather than the occurrence of primary new tumors, with a greater tendency towards local recurrence as opposed to distant metastasis (Figure 7C). BLCA, ESCA, and HNSC individually exhibited a higher DScore linked to distant metastasis compared to local recurrence and the development of new primary tumors (Figure S5D). These findings highlight the potential relevance of DScore in cancer prognosis and therapeutic response. Association of DScore with Cancer Outcomes and Immunotherapy Response. (A) Comparison of DScore between cancer patients with and without tumor; (B) response to conventional anti-cancer therapy; and (C) disease progression or recurrence. Wilcoxon test; P value less than 0.05 was considered significant. (D) Kaplan-Meier curves depicting overall survival difference between low DScore and high DScore TCGA BLCA subgroups, (E) IMvigor210 urothelial carcinoma subgroups, and (F) top 30% IMvigor210 urothelial carcinoma subgroups. Log rank test; P value less than 0.05 was considered significant. (G) t-SNE plots showing major cell types, (H) minor cell types, and (I) DScore distribution in single-cell bladder cancer dataset (GSE145281), respectively. (J) Violin plot showing enrichment of DScore-associated DRGs across minor cell types. (K and L) Violin plots showing DScore enrichment between responders and non-responders to immunotherapy (K) and across major cell types (L). Differences were assessed using the Wilcoxon rank-sum test; ***P < 0.001; ****P < 0.0001; ns, not significant. (M) t-SNE plots showing major cell types, and (N) DScore distribution in single-cell breast cancer mouse model dataset (GSE136206), respectively. (O–Q) Violin plots showing DScore enrichment between CD8⁺ T cells and monocyte/macrophage cells (O), between 7-day treated and untreated samples (P), and across major cell types under treated and untreated conditions (Q). Differences were assessed using the Wilcoxon rank-sum test; P value are shown as **P < 0.01; ***P < 0.001; ****P < 0.0001; ns not significant. Theoretically, molecular targeted therapy, such as monoclonal antibodies targeting immune checkpoints, may be more sensitive to variations in DScore due to its association with immune system. To assess this, we categorized the available immunotherapy cohorts into low and high depression subgroups based on the enrichment score of DRGs. In big IMvigore210 urothelial cancer dataset, DScore was associated with differences in immunotherapy benefit in these patients (Figure 7D–F). As opposed to the protective role of DScore in TCGA BLCA dataset, immunotherapy benefited the patients with low DScore in IMvigor210 UC cohort (Figure 7D–F). The difference in overall survival reached significance when the top 30% were included in the analysis (Figure 7F). No difference in response rate was seen (Figure S5E–F). Consistent with the findings from TME analysis using DScore, the excluded immunophenotype exhibited higher DScore compared to the desert and inflamed immunophenotype (Figure S5G). Among a limited group of melanoma patients, a noteworthy positive correlation was observed between the DScore and the response to immunotherapy, although this relationship was not evident in patients with metastatic melanoma (Figure S5H–M). No difference in survival benefit or response to immunotherapy was noticed in the clear cell Renal Cell Carcinoma (ccRCC) dataset (Figure S5N–P). Unlike IMvigor210, these datasets had a limited number of participants, potentially affecting the significance of DScore. We further validated these results in a single cell dataset of BLCA patients (GSE145281) treated with anti-PD-L1 immunotherapy (Figure 7G and H). In line with our previous observations, DScore was higher in monocytic-lineage cells compared to T (CD4/CD8+ T cells) and NK cells (Figure 7I and J). The non-responders had a high enrichment of DRGs compared to responders (Figure 7K). Moreover, significant differences were observed in the enrichment of DRGs in CD8+ T cells, monocytes/macrophages, and NK cells between responders and non-responders, suggesting the importance of DScore-associated phenotypes of these cells in determining the immunotherapy response (Figure 7L). Extending these observations to a BRCA single-cell RNA-seq dataset, we detected similar patterns: M2 macrophages showed higher DScore compared to CD8+ T cells (Figure 7M–O). Moreover, samples treated with immunotherapy exhibited lower DScore values than untreated samples (Figure 7P). Accordingly, both M2 macrophages and CD8+ T cells displayed higher DScore in the untreated group relative to the immunotherapy-treated group (Figure 7Q). Together, these results suggest that DScore-associated immune phenotypes, particularly M2 macrophages and CD8⁺ T cells, are linked to variation in immunotherapy response. CACNA1H as a Candidate DScore-Associated Biomarker in Breast Cancer Differential gene expression analysis between TCGA BRCA tumor and normal samples identified 1483 DEGs (Figure 8A). Among these, 17 genes overlapped with the DRGs (DRGs, n=381) (Figure 8B). A heatmap illustrates the expression profiles of these 17 genes in TCGA BRCA samples (Figure 8C). Survival analysis revealed that CACNA1H was the only gene significantly associated with poor prognosis in breast cancer patients (Figure 8D). Kaplan–Meier survival curves further demonstrated distinct survival differences between BRCA patients with high versus low CACNA1H expression (Figure 8E). The prognostic relevance of CACNA1H overexpression was validated in an independent cohort of 1879 breast cancer patients using the Kaplan–Meier Plotter (kmplot.com) (Figure 8F). Expression analysis confirmed significantly higher CACNA1H expression in TCGA BRCA tumor samples compared to normal tissues, with paired-sample analysis yielding consistent results (Figure 8G and H). Additionally, mRNA expression profiling in breast cancer cell lines demonstrated elevated CACNA1H levels in cancer cells relative to normal breast epithelial cells (Figure 8I). Representative IHC images of CACNA1H expression in breast cancer tissues (n=12) versus normal breast tissues, along with a bar plot quantifying these differences, further confirmed its significant overexpression (Figure 8J and K). CACNA1H as a Depression-Related Biomarker in Breast Cancer. (A) Differential gene expression analysis between TCGA BRCA tumor and normal samples identified 1483 DEGs. (B) Venn diagram showing 17 overlapping genes between DEGs and depression-related genes (DRGs, n=381). (C) Heatmap displaying expression profiles of the 17 overlapping genes in TCGA BRCA samples. (D and E) Survival and Kaplan–Meier analyses showing that CACNA1H is the only DRG significantly associated with poor prognosis, with higher expression linked to worse survival outcomes. (F) Validation of CACNA1H prognostic relevance in an independent breast cancer cohort (n=1879) using the Kaplan–Meier Plotter. (G and H) Expression analyses showing significantly elevated CACNA1H mRNA levels in TCGA BRCA tumor tissues compared with normal samples, confirmed by paired-sample comparison. Wilcoxon rank-sum test; ***P < 0.001 (I) CACNA1H expression in breast cancer cell lines versus normal breast epithelial cells. The data represent the mean ± SEM (standard error of the mean) of n = 3 independent experiments (biological replicates) for each condition. Statistical significance was determined using a two-tailed unpaired t-test; ***P < 0.001; ****P < 0.0001. (J and K) Representative IHC images and quantitative analysis showing overexpression of CACNA1H protein in breast cancer tissues compared to normal tissues. Paired t test; *P < 0.05. Discussion The existing clinical and preclinical evidence provides strong support for the correlation between depression and cancer. However, a comprehensive assessment of this association at pan-cancer level is still lacking. To address this gap, we utilized a computational approach to quantify depression-associated transcriptional activity and explore its associations with cancer genomic features and TME features. Furthermore, we investigated its significance for clinical outcomes, and immunotherapy response in various types of cancer. This comprehensive analysis sheds light on the complex association between depression-related transcriptional patterns and cancer, providing valuable insights for future research and clinical considerations. The focus of clinical research in cancer patients has primarily been on assessing the risk of depression and its negative impact on patient outcomes.7–18 Factors such as increased depression in cancer patients have been linked to treatment side effects and prognosis.13–18 However, there has been limited investigation into depression as a potential etiological factor for cancer development, and the existing findings are inconclusive.56–62 Our study suggests potential associations between depression-related transcriptional patterns and cancer. This is supported by the differential enrichment of DScore in normal and cancer tissues, as well as the association between DScore and genomic alterations/instability, and remodeling of the TME. Distinct enrichment patterns of DRGs were observed across cancer types, reflecting tissue-specific expression and systematic variation in depression-associated transcriptional activity. For example, GI cancers, lung cancers, and cancers originating from female reproductive organs showed high enrichment of DRGs in cancer tissues compared to normal. On the other hand, a suppressive pattern was evident in brain and urinary tract cancers. The emerging evidence linking depression and genomic alterations presents a compelling association, shedding light on the intricate biological underpinnings of depression in the context of cancer. Studies have reported a strong impact of depression on genomic instability and an association with genetic variations in psychiatric conditions.37,47,48 Considering that cancer is commonly regarded as a disease of the genome at cellular level, it is plausible to suggest that genomic instability may represent a shared feature associated with both depression-related transcriptional activity and cancer.63 Our study demonstrates a robust correlation between DScore and mutation frequency, somatic copy number variations, and markers of genomic instability (MSI, HRD, HRD-LOH). Moreover, DScore also correlated with the mutation rates of cancer-specific oncogenes. The associations were positive in cancers with upregulated expression of DRGs, as mentioned above, and negative in brain cancers and kidney cancers with few exceptions. Remarkably, these patterns of enrichment and genomic instability corresponded to the prognostic relevance of DScore in these cancer types. Overall, our findings contribute to the growing body of evidence linking depression-associated transcriptional activity with cancer biology. Further research is needed to better understand the underlying mechanisms and their potential relevance for cancer progression and therapeutic response. In addition, breast cancer emerged as a representative model of DScore-associated genomic and microenvironmental remodeling. DScore-associated genomic instability corresponded with distinct molecular features in breast cancer, including elevated CACNA1H expression—a depression-associated candidate biomarker validated across transcriptomic datasets, paired samples, and immunohistochemical analyses. Together, these findings underscore the possible association between depression-related transcriptional patterns and tumor progression in breast cancer. The immune system’s role in cancer development and therapy is crucial, and recent advancements in understanding the TME have highlighted its significance.19 Depression has been shown to disrupt immune functions, establishing a relevant link between depression and cancer development.4,64 Previous studies have primarily focused on the activation of proinflammatory cytokines and their impact on diseases associated with depression.4,65 In our study, we observed this pattern in gastrointestinal (GI) cancers and certain female reproductive organ cancers. DScore exhibited a positive correlation with immune and stromal components, and patients with higher DScore showed enrichment of inflammatory and immune pathways. These findings suggest that immune and stromal cells are associated with DScore-related patterns in the tumor microenvironment. Conversely, we also discovered associations between DRGs and cancer cells in various cancer types, including brain, breast, lung, kidney (KIRC), and bladder (BLCA) cancers. In these cases, cancer cells showed higher DScore enrichment. The differences in enrichment targets between these two groups of cancers underscore the diverse biological context of depression-related gene expression in cancer. These variations likely contribute to the differences observed in enriched pathways, the TME landscape, and prognosis which would require further investigations. On the other hand, a limited number of studies have also emphasized the immunosuppressive role of depression. Oxidative stress and inflammation are closely interconnected, with oxidative stress arising as a result of the release of proinflammatory cytokines during depressive conditions.20,65,66 Additionally, oxidative stress can either initiate the development of neoplasia or be triggered by the presence of a growing tumor.20 Oxidative stress disrupts the equilibrium of cytokines and contributes to the formation of an immunosuppressive tumor microenvironment (TME), promoting cancer growth and metastasis.67 Our results also indicated the prevalence of immunosuppressive TME as demonstrated by the enrichment of immunosuppressive hallmark pathways including hypoxia, ROS, EMT, WNT/β catenin signaling, TGF-β signaling, and angiogenesis in high DScore patients across various cancers.49–51 Both the direct impact of oxidative stress caused by tumor cells and the oxidative stress mediators within the TME significantly suppress the function of effector T cells and induce T cell death.68 Our results support this observation, as we found a strong negative correlation between DScore and CD8+ T cells. Furthermore, enhanced infiltration of CD4 memory T cells, which indicates an immune response to various stimuli like infection and stress-inducing agents, was observed.69 In addition, lower infiltration of NK cells, as previously reported in individuals with depression, was also evident.70 The most noteworthy finding was the significant presence of M2 macrophages in the TME and the enrichment of DRGs in these cells. M2 polarization, which is associated with an anti-inflammatory state and immunosuppression, is closely linked to pathways involved in immunosuppression such as EMT, WNT/β-catenin signaling, TGF-β signaling, and angiogenesis, which were enriched in patients with high DScore.71–73 In breast cancer, these immunosuppressive features were particularly evident, where monocytic-lineage enrichment and M2 macrophage dominance defined the DScore-associated TME. Single-cell analyses further confirmed macrophages as the principal DScore-enriched population, accompanied by reduced infiltration of CD8⁺ T cells and NK cells. Moreover, the same DScore-associated TME features were observed in immunotherapy response cohorts of breast and bladder cancers, where non-responders exhibited higher depression-linked transcriptional activity. These observations suggest that DScore-associated immune phenotypes may be linked to variation in antitumor immunity and immunotherapy response. An additional factor to consider regarding the presence of M2 macrophages and the prevalence of an immunosuppressive TME is their involvement in wound healing.74,75 M2 polarization can be initiated after an initial inflammatory phase, accompanied by the subsequent emergence of oxidative stress, aiming to mitigate inflammation and facilitate tissue healing and fibrosis.75 While M2 macrophages have diverse functions, their relevance in the context of depression in cancer may lie in their ability to release neurotransmitters and hormones, such as catecholamines, substance P, adiponectin, and growth factors.76–79 Alternatively, catecholamines can enhance the arginase activity of M2 macrophages during stress response, and arginase-1 expressing M2 macrophages have been shown to suppress chronic inflammation and fibrosis by inhibiting antigen-specific T cell responses.80–82 In conclusion, M2 macrophages appears to be the crucial link between depression and cancer which needs further investigations. Further research is needed to fully understand the connection between depression, M2 macrophages, and cancer. Thus far, we have explored the possibility that depression-related transcriptional patterns are associated with cancer development by inducing genomic alterations/instability and/or TME remodeling. A reverse phenomenon is likely to exist, where the shared underlying mechanisms between cancer and depression may contribute to the activation or expression of DRGs. As a result, DRG enrichment may reflect biological processes associated with behavioral and psychological symptoms commonly associated with depression. This observation could also supplement the traditional perception that depressive individuals are more likely to develop cancer, instead it could be the presence of depressive symptoms in individuals with undiagnosed cancer. Furthermore, there is clinical evidence supporting this hypothesis. For instance, studies have shown that healthy individuals exhibited typical depressive symptoms upon administration of endotoxin, leading to the release of cytokines.83 Similarly, a subset of hepatitis C patients developed major depression following interferon treatment.84 In conclusion, this perspective challenges the traditional view that depression solely increases the risk of cancer development. Instead, it suggests a more complex interplay where cancer development itself may contribute to the enrichment of DRGs and the manifestation of depressive symptoms. Further research is needed to unravel the intricate relationships between cancer and depression, shedding light on the bidirectional influences and potential implications for both physical and mental health. Our identification of CACNA1H as a DScore-related prognostic biomarker in breast cancer adds a novel dimension to the literature, bridging neuropsychiatric and oncologic domains. Although previous work on calcium channel genes (for example, the L-type channel subunit CACNA1C) has largely been in the context of mood disorders and neurodevelopmental illness, the T-type channel gene CACNA1H has been less well characterized in these settings. In neural circuits, a recent study showed that CACNA1H upregulation in ventral hippocampal COCH-neurons, via MTF1 activation, drives burst firing and social-stress-induced anxiety-like behavior.85 In oncology, a review of voltage-gated calcium channels noted that CACNA1H is typically down-regulated in several cancers including breast and bladder, suggesting a tumor-suppressor role.86 By contrast, our data demonstrate overexpression of CACNA1H in breast tumor tissue and breast cancer cell lines, and a strong association with worse survival. This suggests a context-specific switch in channel function, rather than simply loss of expression, and supports the idea that CACNA1H may represent a potential convergence point between cancer progression and depression-related biology, potentially via calcium signaling, altered electrical excitability or calcium-driven downstream pathways such as autophagy or ER-stress. Further mechanistic studies are warranted to disentangle whether CACNA1H may be involved in tumor cell behavior, neural-tumor interactions, or systemic neuroendocrine links between depression and tumor microenvironment. Our study has several limitations. In this study, we evaluated the depression via gene set enrichment analysis of a set of 382 genes acquired from The Molecular Signatures Database (MSigDB). However, the extent to which the enrichment of these genes accurately reflects depression-associated transcriptional activity is yet to be determined. Nevertheless, we made concerted efforts to establish the association between these genes and depression through comprehensive bioinformatics analyses involving diverse populations, cancer patients, and individuals with major depressive disorders. Moreover, it is important to note that our findings are derived from RNA-level quantifications of DRGs, whereas the process of depression primarily involves proteins. Therefore, there may be significant differences in protein-level dynamics that need to be considered. Additionally, the lack of available phenotype data related to depression prevented us from confirming the actual level of depression in these cancer patients. Conclusions In conclusion, this comprehensive computational analysis highlights associations between depression-associated transcriptional activity and cancer-related processes. Our findings indicate that these transcriptional patterns are linked to genomic alterations, tumor microenvironment features, and immune-related characteristics across cancers. These results emphasize the value of integrating molecular and clinical perspectives in cancer research. Future studies incorporating protein-level validation and detailed phenotypic data are needed to further clarify the biological basis of these associations and assess their potential clinical relevance. Funding Statement Shenzhen Key Medical Discipline Construction Fund (No.SZXK015);Guangdong Provincial and National Key Clinical Specialty Construction Project and National Key Clinical Specialty Construction Project; Sanming Project of Medicine in Shenzhen (No.SZSM202411026); Heyuan Social Development Science and Technology Plan Project (No.250612161605480); Wu Jieping Medical Foundation (No.320.6750.2023-18-90). Data Sharing Statement The datasets supporting the conclusions of this study are available from The Cancer Genome Atlas (TCGA) at https://portal.gdc.cancer.gov/. Further information and requests for resources should be directed to the Lead Contact, Jie Lin (linjie@gzhmu.edu.cn). Ethics Approval and Consent to Participate Studies involving human participants were reviewed and approved by the Institutional Review Board and Ethics Committee of Shenzhen People’s Hospital (The First Affiliated Hospital of Southern University of Science and Technology; The Second Clinical Medical College of Jinan University), Shenzhen, China (approval number: LL-KY-2023144-01). Written informed consent was obtained from all participants prior to enrollment. All procedures were conducted in accordance with the ethical principles of the Declaration of Helsinki. Author Contributions All authors made a significant contribution to the work reported, whether that is in the conception, study design, execution, acquisition of data, analysis and interpretation, or in all these areas; took part in drafting, revising or critically reviewing the article; gave final approval of the version to be published; have agreed on the journal to which the article has been submitted; and agree to be accountable for all aspects of the work. Disclosure The authors report no conflicts of interest in this work. References Associated Data Data Availability Statement The datasets supporting the conclusions of this study are available from The Cancer Genome Atlas (TCGA) at https://portal.gdc.cancer.gov/. Further information and requests for resources should be directed to the Lead Contact, Jie Lin (linjie@gzhmu.edu.cn).