Comprehensive bioinformatics analysis of EXOSC family genes in lung adenocarcinoma Authors: Ning Li, Xiaolei Wang, Tingting Chu, Yongxin Pan, Xuezhong Zhang Published: 2026 ABSTRACT The RNA exosome complex (EXOSC1-10) orchestrates 3’-5’ RNA processing and decay, yet its family-wide landscape and clinical relevance in lung adenocarcinoma (LUAD) remain incompletely defined. Here, we conducted an integrative multi-omics analysis of EXOSC family members in LUAD using public transcriptomic and proteomic resources, external validation cohorts, and single-cell data. Across TCGA (59 normal vs. 515 tumors), all EXOSC genes were significantly upregulated in tumors, with concordant increases for most proteins in CPTAC and immunohistochemistry evidence from the Human Protein Atlas. Survival analyses identified EXOSC2, EXOSC3 and EXOSC5 as prognostically informative, with elevated EXOSC2/EXOSC5 also associated with inferior disease-specific survival. Receiver operating characteristic analyses indicated moderate-to-high diagnostic performance for several EXOSC genes, supported by qRT-PCR validation in LUAD cell lines and confirmation in GSE31210. Clinicopathological correlations linked EXOSC1-5 and EXOSC8-10 to advanced stage and metastatic features, whereas EXOSC6 and EXOSC7 showed comparatively limited or context-dependent associations, suggesting functional heterogeneity within the family. Genomic profiling revealed recurrent alterations in 11.83% of patients, with EXOSC4 exhibiting the highest alteration frequency (predominantly amplifications), and network analyses identified core interacting partners centered on RNA surveillance machinery. EXOSC expression was broadly correlated with RNA modification regulators (m1A/m5C/m6A) and displayed heterogeneous relationships with immune checkpoint genes and immune infiltration, while higher EXOSC expression consistently associated with lower ESTIMATE-derived microenvironment scores and an immune profile characterized by Th2/T helper enrichment with reduced cytotoxic and antigen-presenting populations. Gene set enrichment analyses implicated cell-cycle regulation, senescence, chromatin-related programs and extracellular matrix pathways, and EXOSC-high tumors were associated with increased tumor mutational burden and homologous recombination deficiency, together with enrichment of DNA damage response pathways. Single-cell analysis (GSE146100) localized highest EXOSC expression to proliferating monocytes/macrophages and suggested TGF-β-linked communication with endothelial and T-cell compartments. Exploratory translational analyses further indicated an association between EXOSC9 and predicted cisplatin sensitivity. Collectively, these findings position the EXOSC family as a clinically and biologically relevant axis in LUAD, linking RNA surveillance dysregulation to immune contexture and genomic instability, and highlight subunit-specific heterogeneity that warrants mechanistic validation.Supplementary InformationThe online version contains supplementary material available at 10.1007/s12672-026-05258-9. Introduction Lung adenocarcinoma (LUAD), the most common subtype of non-small cell lung cancer (NSCLC), remains a leading cause of cancer-related mortality owing to late diagnosis, high recurrence rates, and therapeutic resistance [1, 2]. Despite advances in targeted therapies and immunotherapy, reliable prognostic biomarkers and effective tools for therapeutic stratification are still lacking [1, 3, 4]. The exosome complex (EXOSC) family (EXOSC1-10) constitutes a highly conserved ribonuclease complex responsible for 3′-5′ RNA degradation, including ribosomal RNA (rRNA), messenger RNA (mRNA), and various non-coding RNAs [5–8]. Dysregulation of EXOSC subunits can impair RNA surveillance, compromise genome integrity, and disrupt cell cycle regulation, thereby contributing to tumorigenesis and cancer progression [5, 6]. Increasing evidence indicates that abnormal expression of specific EXOSC genes is associated with unfavorable prognosis and resistance to therapy in several malignancies, such as head and neck squamous cell carcinoma (HNSC) and hepatocellular carcinoma [5, 6]. Moreover, recent studies have reported that EXOSC3 is functionally linked to progression of LUAD [9, 10], while EXOSC5 has been identified as a potential prognostic biomarker indicative of poor survival in LUAD [11]. Nevertheless, the biological impact of the EXOSC family appears heterogeneous: some subunits, such as EXOSC6 and EXOSC7, show largely neutral or context-dependent associations across cancer cohorts, implying that EXOSC functions may vary with tumor type, molecular background, and microenvironmental state rather than conforming to a uniform family-wide effect [5, 12]. This functional heterogeneity underscores the need for systematic, integrative interrogation of the entire EXOSC family in LUAD. However, existing findings are fragmented and gene-specific; a comprehensive evaluation of all ten EXOSC family members in LUAD, integrating expression patterns, prognostic implications, genomic alterations, immune features, and RNA-modification associations-has not yet been performed. Moreover, no study has explored EXOSC-centered co-expression networks or developed prognostic signatures derived from EXOSC-associated pathways in LUAD, leaving key mechanistic and clinical questions unresolved. Here, we conducted an integrative multi-omics and multi-cohort analysis to delineate the landscape of EXOSC family dysregulation in LUAD. By combining transcriptomic and proteomic profiling, survival and diagnostic assessment, genomic alteration analysis, pathway enrichment, immune infiltration and immune checkpoint correlations, co-expression network interrogation, and construction of an prognostic model [13–18], we aimed to clarify the biological and clinical relevance of EXOSC family members. We further incorporated single-cell RNA-sequencing to localize EXOSC expression within the tumor microenvironment and to infer cell-cell communication features, and explored translational hypotheses by relating EXOSC activity to genomic instability metrics and predicted drug sensitivity patterns. Together, these analyses sought to establish a coherent framework linking EXOSC-associated RNA surveillance programs to LUAD progression and potential therapeutic vulnerability. Methods Gene and protein expression analysis Differential expression of EXOSC genes between LUAD and normal lung tissues was assessed using UALCAN (http://ualcan.path.uab.edu/index.html), based on TCGA transcriptomic data [19]. Protein expression levels were obtained from the CPTAC LUAD dataset via UALCAN [20]. Statistical significance of differences was evaluated using the unpaired Wilcoxon rank-sum test and the Wilcoxon signed-rank test. Statistical significance was defined as follows: *P < 0.05, **P < 0.01, and ***P < 0.001. Survival analysis The prognostic relevance of EXOSC genes was evaluated using TCGA-LUAD clinical data. Patients were dichotomized into high- and low-expression groups according to median gene expression. The statistical analyses were conducted in R (version 4.2.1). Kaplan-Meier curves for overall survival (OS) and disease-specific survival (DSS) were generated using the survival (v3.5-7) and survminer (v0.4.9) packages. Hazard ratios (HRs) and 95% confidence intervals (CIs) were calculated via Cox proportional hazards models, with significance assessed by log-rank tests. Statistical significance was defined as follows: *P < 0.05, **P < 0.01, and ***P < 0.001. Diagnostic performance analysis Diagnostic utility was analyzed using receiver operating characteristic (ROC) curves generated with pROC (v1.18.4). AUC values were computed to assess discriminatory performance. Statistical significance was defined as follows: *P < 0.05, **P < 0.01, and ***P < 0.001. Human protein atlas The Human Protein Atlas (HPA, https://www.proteinatlas.org/) is an open-access database that provides detailed information on the localization and expression of proteins and genes across cell lines, normal tissues, and pathological specimens [21]. Immunohistochemical (IHC) images of EXOSC family genes in normal lung and LUAD tissues were retrieved from the HPA for comparative analysis. Cell culture The human bronchial epithelial cell line BEAS-2B and the lung adenocarcinoma cell lines A549 and H1299 were purchased from the Cell Bank of the Shanghai Institute of Life Sciences (Shanghai, China). Cells were maintained in Dulbecco’s Modified Eagle’s Medium (DMEM) containing 10% fetal bovine serum (FBS) and 1% penicillin-streptomycin solution (final concentrations: 100 U/mL penicillin and 100 µg/mL streptomycin). Cultures were incubated at 37 °C in a humidified environment with 5% CO₂. Cells in the exponential (log) growth phase were harvested and used for the subsequent analyses. Quantitative real-time polymerase chain reaction (qRT-PCR) Total RNA from cultured cells was isolated using TRIzol reagent (Takara, Shiga, Japan). Reverse transcription was performed with the First Strand cDNA Synthesis Kit (Yeasen, Shanghai, China) to generate complementary DNA (cDNA). qRT-PCR was performed on an ABI 7900HT system using standard cycling conditions: 50 °C for 5 min, 90 °C for 10 min, followed by 45 cycles of 95 °C for 15 s and 60 °C for 1 min. Relative expression levels were calculated using the 2−ΔΔCT method, and β-Actin was used as an endogenous control. Melting-curve analysis confirmed amplification specificity. Primer sequences are provided in Supplementary Table 1. Clinicopathological correlation analysis Associations between EXOSC expression and clinical parameters (T/N/M stage, overall stage, laterality, age, sex, smoking status, tumor site) were examined using TCGA data. Statistical comparisons used the Wilcoxon rank-sum test, Kruskal-Wallis test, or chi-square test as appropriate. Statistical significance was defined as follows: *P < 0.05, **P < 0.01, and ***P < 0.001. Genetic alteration analysis Genetic alterations of EXOSC family genes were examined using cBioPortal (https://www.cbioportal.org) based on the TCGA-LUAD PanCancer Atlas dataset [22]. Data from 507 patients with complete mutation and CNV profiles were included. Alteration types (missense mutations, structural variants, amplifications, deep deletions) were visualized using cBioPortal oncoprints. Correlation analysis of gene expression Pairwise correlations among EXOSC genes were computed using Pearson correlation (psych v2.3.9) and visualized as heatmaps with pheatmap (v1.0.12). Statistical significance was defined as follows: *P < 0.05, **P < 0.01, and ***P < 0.001. Protein-protein interaction (PPI) network construction Potential functional interactors of EXOSC family genes were identified using STRING (https://string-db.org, version 12.0) [23] and GeneMANIA (http://genemania.org) [24]. STRING networks were constructed for Homo sapiens with sources including text mining, experiments, databases, co-expression, neighborhood, gene fusion, and co-occurrence, applying a minimum confidence score of 0.7. GeneMANIA was used to validate and expand the PPI network using co-expression, physical interactions, co-localization, and pathway associations. Intersection and correlation analysis of shared interactors Shared interactors predicted by both STRING and GeneMANIA were identified using VennDiagram (v1.7.3). Pairwise correlations among common interactors (C1D, DIS3, DIS3L, HBS1L, MPHOSPH6, MTREX) were visualized using ggcorrplot (v0.1.4). Statistical significance was defined as follows: *P < 0.05, **P < 0.01, and ***P < 0.001. Correlation between EXOSC expression and RNA modification regulators Correlations between EXOSC family genes and RNA-modification regulators, including N1-methyladenosine (m1A), N5-methylcytosine (m5C), and N6-methyladenosine (m6A), were assessed using curated gene sets from recent literature and MSigDB. Pearson correlation analysis (|R| > 0.30, P < 0.05) was performed in R (v4.2.2), and results were visualized as heatmaps using pheatmap via ggplot2[3.4.4]. Statistical significance was defined as follows: *P < 0.05, **P < 0.01, and ***P < 0.001. Correlation analysis with immune checkpoint genes Relationships between EXOSC family genes expressions and immune checkpoint molecules, including inhibitory receptors (PDCD1, CTLA4, TIGIT, LAG3, BTLA, CD27, VSIR), ligands (CD274, TGFB1, TGFBR1), and other regulatory genes (ADORA2A, CD160), were analyzed using TCGA-LUAD expression data and Pearson correlations. Results were visualized with ggcorrplot via ggplot2[3.4.4]. Statistical significance was defined as follows: *P < 0.05, **P < 0.01, and ***P < 0.001. Immune cell infiltration analysis Single-sample gene set enrichment analysis (ssGSEA) quantified infiltration of 24 immune cell types, including activated dendritic cells (aDC), B lymphocytes (B cells), CD8⁺ T lymphocytes (CD8 T cells), Cytotoxic lymphocytes (Cytotoxic cells), Dendritic cells (DC), Eosinophilic granulocytes (Eosinophils), immature dendritic cells (Idc), Macrophages, Mast cells, Neutrophilic granulocytes (Neutrophils), CD56 bright natural killer cells (NK CD56bright cells), CD56 dim natural killer cells (NK CD56dim cells), Natural killer cells (NK cells), Plasmacytoid dendritic cells (pDC), T lymphocytes (T cells), T helper lymphocytes (T helper cells), Central memory T cells (Tcm), Effector memory T cells (Tem), T follicular helper cells (TFH), γδ T cells (Tgd), Type 1 T helper cells (Th1 cells), Type 17 T helper cells (Th17 cells), Type 2 T helper cells (Th2 cells), Regulatory T cells (Treg) [25]. Marker genes were sourced from immunogenomic studies, and ssGSEA was performed using GSVA (v1.46.0) [26]. Correlations with EXOSC expression were computed using Pearson correlation. Differences between high- and low-expression groups (median split) were evaluated by Wilcoxon rank-sum tests, visualized with ggplot2 (v3.4.2). Statistical significance was defined as follows: *P < 0.05, **P < 0.01, and ***P < 0.001. To investigate the relationship between the expression of EXOSC family genes and tumor microenvironment characteristics, we calculated three comprehensive scores for each sample based on gene expression data using the R package “ESTIMATE” (version 1.0.13): the ImmuneScore, StromalScore, and ESTIMATEScore. Subsequently, Pearson correlation coefficients (R) and their corresponding two-tailed p-values were computed between the expression levels of EXOSC family genes (log₂(TPM + 1) transformed values) and each of the above scores using the cor.test function. The correlation results were visualized using a lollipop plot, generated with the ggplot2 package (version 3.4.0), in which the length of each lollipop represents the absolute value of the correlation coefficient and the direction indicates positive or negative correlation. Statistical significance was defined as follows: *P < 0.05, **P < 0.01, and ***P < 0.001. Gene set enrichment analysis (GSEA) GSEA was performed using clusterProfiler (v4.4.4) to identify biological pathways associated with each EXOSC gene [27]. Patients were stratified into high- and low-expression groups by median expression. MSigDB hallmark (h.all.v7.5.1) and canonical pathway (c2.cp.v7.5.1) gene sets were used. Enrichment was significant with NES > 1, nominal P < 0.05, and FDR < 0.25. Top pathways were selected for interpretation. Statistical significance was defined as follows: *P < 0.05, **P < 0.01, and ***P < 0.001. Co-expression network construction and functional enrichment analysis Co-expressed genes were identified by Pearson correlation analysis (|R| > 0.15, P < 0.001) using TCGA-LUAD transcriptomic data. Top 10 positively and negatively correlated genes for each EXOSC member were visualized with corrplot and ggplot2. Shared co-expressed genes were identified with VennDiagram. Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses were conducted using clusterProfiler, with P < 0.05 and FDR < 0.05. Results were categorized as biological process, cellular component, or molecular function, and visualized with bubble and bar plots. Construction of prognostic model and nomogram Univariate Cox regression identified prognostic co-expressed genes, followed by LASSO Cox regression (glmnet) to reduce dimensionality. The optimal penalty parameter (λ) was determined by 10-fold cross-validation. A multigene risk score was calculated as a weighted sum of expression values by regression coefficients. Patients were stratified into high- and low-risk groups by the median score. Kaplan-Meier analysis and log-rank tests compared survival outcomes. Time-dependent ROC analysis (timeROC) assessed predictive performance at 1-, 3-, and 5-year intervals. Multivariate Cox regression, including clinical covariates (age, sex, TNM stage, smoking), evaluated independent prognostic significance. A nomogram integrating risk score and clinical features was generated using rms, and calibration curves assessed prediction accuracy. The relationship between EXOSC family genes levels and tumor mutational burden (TMB) and homologous recombination deficiency (HRD) in LUAD Tumor mutation burden (TMB) data for LUAD-TCGA samples were obtained from the Genomic Data Commons (GDC) portal and processed using MuTect2 software. TMB scores were calculated using the maftools R package (version 2.8.05). Homologous recombination deficiency (HRD) scores for LUAD were retrieved from The Immune Landscape of Cancer study. The associations between the expression levels of EXOSC family genes and TMB or HRD across LUAD-TCGA were explored using R software. Statistical significance was defined as follows: *P < 0.05, **P < 0.01, and ***P < 0.001. Single-cell analysis of the EXOSC family genes in LUAD Using the Tumor Immune Single-cell Hub 2 (TISCH2) platform and the CellChat toolkit [28], we profiled the expression of exosome component (EXOSC) family genes within the TME of lung adenocarcinoma using a publicly available single-cell RNA-sequencing dataset (GSE146100). Following a standardized workflow including quality control, batch-effect correction, and unsupervised clustering, we identified 11 major cell types. We then quantified the distribution of EXOSC gene expression across cellular subsets and systematically inferred intercellular communication networks involving EXOSC-high proliferating monocytes/macrophages and other TME populations, with particular emphasis on the potential mediating role of TGF-β signaling. Relationship of EXOSC genes with PARP1 and drug sensitivity in LUAD To interrogate molecular vulnerabilities associated with EXOSC high-expression tumors from a synthetic-lethality perspective, we leveraged DepMap (https://depmap.org/repurposing) LUAD cell line data to assess whether EXOSC family gene essentiality co-varies with DNA repair capacity. Specifically, gene dependency scores (CRISPR-derived) for EXOSC members were correlated with PARP1 expression across LUAD cell lines using correlation testing, and significance was evaluated with two-sided P values. In parallel, to estimate potential therapeutic sensitivities in patients, transcriptomic profiles from TCGA-LUAD were analyzed with the OncoPredict framework to impute drug response, generating predicted IC50 values for chemotherapy (cisplatin) and targeted therapy (osimertinib). Associations between EXOSC expression and predicted IC50 values were then quantified by correlation analysis to identify EXOSC-linked drug sensitivity patterns, highlighting candidate combination strategies for EXOSC-high tumors (e.g., PARP1-related vulnerabilities and cisplatin sensitivity signals, including the EXOSC9–cisplatin association). Statistical significance was defined as follows: *P < 0.05, **P < 0.01, and ***P < 0.001. Results Expression, survival and ROC analyses of EXOSC family genes in LUAD Analysis of TCGA data using the UALCAN platform (59 normal vs. 515 tumor samples) revealed distinct expression patterns of EXOSC family members in LUAD. As shown in Fig. 1A, all EXOSC genes (EXOSC1-10) were significantly upregulated in tumor tissues compared with normal counterparts (P < 0.001). Consistently, CPTAC data demonstrated markedly elevated protein levels of EXOSC1-8 and EXOSC10 in primary LUAD tissues relative to normal controls (P < 0.001), whereas EXOSC9 showed an increasing trend without statistical significance (Fig. 1B). In addition, IHC results from HPA confirmed the protein expression profiles of EXOSC genes in LUAD. Specifically, EXOSC1, EXOSC2, EXOSC3, EXOSC4, EXOSC5, EXOSC7, EXOSC8, EXOSC9, and EXOSC10 exhibited higher protein expression in LUAD tissues compared with normal tissues (Fig. 1C). Fig. 1Differential expression of EXOSC family members in LUAD at mRNA and protein levels (*p < 0.05, **p < 0.01, ***p < 0.001). A UALCAN analysis of TCGA-LUAD data (59 normal vs. 515 tumor samples) revealed significantly higher mRNA expression of all EXOSC genes (EXOSC1-10) in tumor tissues compared with normal lung. B Protein expression levels of EXOSC family members in LUAD and normal tissues derived from CPTAC datasets. C Immunohistochemistry (IHC) images from the Human Protein Atlas (HPA) showing protein expression of EXOSC family members in LUAD and normal lung tissues Differential expression of EXOSC family members in LUAD at mRNA and protein levels (*p < 0.05, **p < 0.01, ***p < 0.001). A UALCAN analysis of TCGA-LUAD data (59 normal vs. 515 tumor samples) revealed significantly higher mRNA expression of all EXOSC genes (EXOSC1-10) in tumor tissues compared with normal lung. B Protein expression levels of EXOSC family members in LUAD and normal tissues derived from CPTAC datasets. C Immunohistochemistry (IHC) images from the Human Protein Atlas (HPA) showing protein expression of EXOSC family members in LUAD and normal lung tissues Kaplan-Meier analyses based on TCGA data were conducted to assess the prognostic value of EXOSC family members in LUAD, including OS and DSS (sample sizes in Supplementary Table 2). As shown in Fig. 2A, EXOSC2, EXOSC3, and EXOSC5 were significantly associated with OS (P < 0.05), whereas other family members showed no notable correlation. Similarly, Fig. 2B demonstrates that elevated expression of EXOSC2 and EXOSC5 predicted inferior DSS, while no significant associations were observed for the remaining EXOSC genes (P > 0.05). To evaluate diagnostic performance, ROC analyses were performed (sample sizes in Supplementary Table 3). As presented in Fig. 2C, EXOSC genes exhibited moderate-to-high discriminatory accuracy for LUAD, with AUC values of 0.889 (EXOSC1), 0.820 (EXOSC2), 0.718 (EXOSC3), 0.869 (EXOSC4), 0.896 (EXOSC5), 0.573 (EXOSC6), 0.528 (EXOSC7), 0.649 (EXOSC8), 0.556 (EXOSC9), and 0.658 (EXOSC10). Validation using qRT-PCR further supported these findings. As shown in Fig. 2D, EXOSC family genes were markedly upregulated in the LUAD cell line A549 relative to the normal bronchial epithelial cell line BEAS-2B (all P < 0.05). Consistently, Fig. 2E shows that an additional LUAD cell line (H1299) displayed similar overexpression patterns (all P < 0.05). We further examined EXOSC expression using the GSE31210 dataset. As shown in Fig. 2F, EXOSC1-8 and EXOSC10 were significantly upregulated in LUAD tissues compared with normal lung samples (all P < 0.05), whereas EXOSC9 showed no significant difference. Fig. 2Prognostic, diagnostic, and expression profiles of EXOSC family genes in LUAD (*p < 0.05, **p < 0.01, ***p < 0.001). A Kaplan-Meier analysis of overall survival (OS) in LUAD patients from TCGA, showing that high expression of EXOSC2, EXOSC3, and EXOSC5 is significantly associated with poor OS. B Kaplan-Meier analysis of disease-specific survival (DSS), indicating that elevated EXOSC2 and EXOSC5 expression predicts worse DSS, while other family members show no significant correlation. C Receiver operating characteristic (ROC) curves for EXOSC family genes in LUAD, demonstrating moderate-to-high diagnostic accuracy with area under the curve (AUC) values ranging from 0.528 to 0.896. D, E qRT-PCR validation of EXOSC expression in LUAD cell lines (A549 and H1299) relative to normal bronchial epithelial cells (BEAS-2B), showing significant upregulation of all examined EXOSC genes. F Analysis of the GSE31210 dataset showing significant overexpression of EXOSC1–8 and EXOSC10 in LUAD tissues compared with normal lung tissues, whereas EXOSC9 shows no significant difference Prognostic, diagnostic, and expression profiles of EXOSC family genes in LUAD (*p < 0.05, **p < 0.01, ***p < 0.001). A Kaplan-Meier analysis of overall survival (OS) in LUAD patients from TCGA, showing that high expression of EXOSC2, EXOSC3, and EXOSC5 is significantly associated with poor OS. B Kaplan-Meier analysis of disease-specific survival (DSS), indicating that elevated EXOSC2 and EXOSC5 expression predicts worse DSS, while other family members show no significant correlation. C Receiver operating characteristic (ROC) curves for EXOSC family genes in LUAD, demonstrating moderate-to-high diagnostic accuracy with area under the curve (AUC) values ranging from 0.528 to 0.896. D, E qRT-PCR validation of EXOSC expression in LUAD cell lines (A549 and H1299) relative to normal bronchial epithelial cells (BEAS-2B), showing significant upregulation of all examined EXOSC genes. F Analysis of the GSE31210 dataset showing significant overexpression of EXOSC1–8 and EXOSC10 in LUAD tissues compared with normal lung tissues, whereas EXOSC9 shows no significant difference Correlation between expressions of EXOSC family and clinicopathological parameters of LUAD Overall, EXOSC family gene expression in LUAD showed distinct clinicopathological associations (Supplementary Table 4). EXOSC1-5 and EXOSC8-10 were primarily associated with advanced tumor stage, metastasis, and smoking history, whereas EXOSC6 and EXOSC7 exhibited more limited or context-dependent relevance. Specifically, EXOSC1 expression correlated with pathologic T stage (P = 0.018), overall stage (P = 0.025), and right-sided tumors (P = 0.013), with higher levels observed in T2-T4 and stage III-IV tumors. EXOSC2 was linked to advanced T stage (P = 0.004), nodal metastasis (P = 0.015), distant metastasis (P = 0.037), overall stage (P = 0.003), and right-sided tumors (P = 0.001), particularly enriched in T4, N2-N3, M1, and stage III-IV cases. EXOSC3 was associated with distant metastasis (P = 0.032), advanced stage (P = 0.023), smoking status (P = 0.015), and tumor laterality (P = 0.014). EXOSC4 correlated with T stage (P = 0.029) and tumor laterality (P = 0.003), while EXOSC5 was linked to T stage (P = 0.028), nodal involvement (P = 0.014), and heavy smoking (≥ 40 pack-years, P = 0.019). EXOSC6 showed higher expression in peripheral tumors (P = 0.035), whereas EXOSC7 was associated with T stage (P = 0.0496) and gender (P = 0.042). EXOSC8 displayed strong associations with distant metastasis (P = 0.006), advanced stage (P = 0.005), male gender (P = 0.042), and younger age (P = 0.028). EXOSC9 was correlated with distant metastasis (P = 0.020), and EXOSC10 was linked to distant metastasis (P = 0.032). Mutation and correlation analysis of EXOSC family genes in LUAD Epigenetic and genetic alterations play critical roles in tumor initiation and progression. Using the cBioPortal platform, we analyzed genetic alterations of EXOSC family genes in LUAD, including mutations and copy number variations (CNVs). As shown in Fig. 3A, EXOSC4 exhibited the highest alteration frequency (6%), predominantly amplifications, suggesting it may be the most frequently altered member in LUAD. EXOSC10 showed the second highest frequency (1.8%), comprising amplifications, deep deletions, and missense mutations. Other genes displayed relatively low alteration frequencies, including EXOSC5 (1%), EXOSC2 (0.8%), EXOSC8 (0.8%), EXOSC3 (0.6%), EXOSC6 (0.6%), EXOSC7 (0.6%), and EXOSC9 (0.6%), whereas EXOSC1 exhibited the lowest frequency (0.4%). Overall, alterations in EXOSC family genes were detected in 11.83% of LUAD patients (507 cases), including mutations, structural variants, amplifications, and deep deletions (Fig. 3B). These alterations may contribute to tumor development, as amplifications can drive oncogene overexpression, while deep deletions may result in loss of tumor suppressor function. Pearson correlation analysis was performed to further explore the relationships among EXOSC family members. As shown in Fig. 3C, most EXOSC genes exhibited significant positive correlations (P < 0.05). Notably, no significant correlations were observed between EXOSC4 and EXOSC6, EXOSC4 and EXOSC9, or EXOSC5 and EXOSC6, suggesting potential heterogeneity in their regulatory mechanisms. Fig. 3Clinicopathological relevance, genetic alterations, and correlation analysis of EXOSC family genes in LUAD (*p < 0.05, **p < 0.01, ***p < 0.001). A, B Genetic alteration landscape of EXOSC family genes in LUAD analyzed via the cBioPortal platform. C Pearson correlation heatmap showing expression relationships among EXOSC family members in LUAD Clinicopathological relevance, genetic alterations, and correlation analysis of EXOSC family genes in LUAD (*p < 0.05, **p < 0.01, ***p < 0.001). A, B Genetic alteration landscape of EXOSC family genes in LUAD analyzed via the cBioPortal platform. C Pearson correlation heatmap showing expression relationships among EXOSC family members in LUAD Protein-protein interaction (PPI) network of EXOSC family genes in LUAD To systematically identify potential functional partners of the EXOSC family in humans, we constructed protein-protein interaction (PPI) networks using STRING and GeneMANIA. As shown in Fig. 4A, the STRING database predicted 50 EXOSC-binding proteins, including MPHOSPH6, DIS3, C1D, and DIS3L. GeneMANIA analysis further identified 20 significant interacting proteins, such as PNPT1, XRN1, KHSRP, ZFP36L1, ZFP36, SUPT6H, MTREX, PARN, AURKAIP1, TNPO1, DCP1A, DCP2, EXD3, EXD1, and DDX47 (Fig. 4B). Among these, MPHOSPH6 and DIS3 showed the highest interaction scores, indicating the central role in the EXOSC-associated network. Intersection analysis identified C1D, DIS3, DIS3L, HBS1L, MPHOSPH6, and MTREX as shared partners across the two prediction platforms (Fig. 4C). To further explore the relationships among these six genes, pearson correlation analysis was performed. As shown in Fig. 4D, significant positive correlations were observed among them (P < 0.001), highlighting their potential cooperative roles in LUAD. Fig. 4Protein–protein interaction (PPI) network and functional enrichment of EXOSC family genes in LUAD (*p < 0.05, **p < 0.01, ***p < 0.001). A PPI network constructed using the STRING database, revealing 50 predicted EXOSC-interacting proteins, including MPHOSPH6, DIS3, C1D, and DIS3L. B GeneMANIA network analysis identifying 20 additional functional partners such as PNPT1, XRN1, KHSRP, ZFP36L1, ZFP36, and MTREX. C Intersection analysis between STRING and GeneMANIA predicted networks highlighting six shared interactors: C1D, DIS3, DIS3L, HBS1L, MPHOSPH6, and MTREX. D Pearson correlation analysis of these six hub genes showing significant positive correlations, indicating their potential cooperative involvement in LUAD progression Protein–protein interaction (PPI) network and functional enrichment of EXOSC family genes in LUAD (*p < 0.05, **p < 0.01, ***p < 0.001). A PPI network constructed using the STRING database, revealing 50 predicted EXOSC-interacting proteins, including MPHOSPH6, DIS3, C1D, and DIS3L. B GeneMANIA network analysis identifying 20 additional functional partners such as PNPT1, XRN1, KHSRP, ZFP36L1, ZFP36, and MTREX. C Intersection analysis between STRING and GeneMANIA predicted networks highlighting six shared interactors: C1D, DIS3, DIS3L, HBS1L, MPHOSPH6, and MTREX. D Pearson correlation analysis of these six hub genes showing significant positive correlations, indicating their potential cooperative involvement in LUAD progression Association between EXOSC family genes expression and RNA modification in LUAD RNA modifications are critical regulators of tumor progression, including angiogenesis. In LUAD, EXOSC family genes show broad and significant correlations with m1A, m5C, and m6A modification pathways (Fig. 5, sample sizes in Supplementary Table 5). As shown in Fig. 5A, most EXOSC genes display positive associations with m1A regulators such as ALKBH1, TRMT61A, TRMT10C, TRMT6, YTHDF1, and ALKBH3, whereas EXOSC4 is negatively correlated with YTHDC1. For m5C-related genes (Fig. 5B), EXOSC members are positively associated with NSUN5, NSUN2, DNMT3A, and DNMT3B, while EXOSC1, EXOSC4, and EXOSC5 show negative correlations with TET2, and EXOSC4 is also inversely related to NSUN3 and TRDMT1. Regarding m6A pathways (Fig. 5C), most EXOSC genes exhibit strong positive correlations with key regulators, but EXOSC4 is negatively correlated with FMR1, FTO, METTL14, YTHDC1, YTHDC2, and ZC3H13, and EXOSC5 with FTO, METTL4, YTHDC2, and ZC3H13. Collectively, these results suggest that RNA modifications may modulate EXOSC gene expression, revealing a potential mechanistic relationship between EXOSC family members and RNA modification pathways in LUAD. Fig. 5Association between EXOSC family gene expression and RNA modification in LUAD (*p < 0.05, **p < 0.01, ***p < 0.001). Correlation analyses for RNA modification patterns: m1A (A), m5C (B), and m6A (C) Association between EXOSC family gene expression and RNA modification in LUAD (*p < 0.05, **p < 0.01, ***p < 0.001). Correlation analyses for RNA modification patterns: m1A (A), m5C (B), and m6A (C) Association between EXOSC family gene expression and immune checkpoint genes in LUAD To explore the potential involvement of the EXOSC family in immune regulation of LUAD, we analyzed the correlations between EXOSC family genes expressions and immune checkpoint molecules (Fig. 6, sample sizes in Supplementary Table 6). Distinct patterns were observed across family members. EXOSC1, EXOSC4, and EXOSC5 showed broad negative correlations with multiple immune checkpoints, including PDCD1, CTLA4, CD274, TIGIT, and VSIR. In contrast, EXOSC3 and EXOSC9 were positively associated with several checkpoints, such as CD274, CTLA4, TIGIT, and TGFBR1, suggesting a potential role in immune activation. EXOSC2 displayed both positive (e.g., CD274, LAG3, TGFBR1) and negative (BTLA, CD27) associations, while EXOSC6, EXOSC7, and EXOSC8 were largely negatively correlated with inhibitory checkpoints, though limited positive associations were noted (e.g., EXOSC6 with ADORA2A and TGFBR1, EXOSC8 with CD160). EXOSC10 was positively correlated with several checkpoints, including CD274, LAG3, TGFBR1, and TGFB1. Overall, these findings indicate that different EXOSC family members may exert heterogeneous effects on immune checkpoint regulation, with some members potentially facilitating immune evasion, whereas others may contribute to immune activation in LUAD. Fig. 6Correlation between EXOSC family gene expression and immune checkpoint genes in LUAD (*p < 0.05, **p < 0.01, ***p < 0.001) Correlation between EXOSC family gene expression and immune checkpoint genes in LUAD (*p < 0.05, **p < 0.01, ***p < 0.001) Correlation between EXOSC family genes expressions and immune cell infiltration in LUAD As shown in Fig. 7 (sample sizes in Supplementary Table 7)., ssGSEA analysis revealed distinct associations between EXOSC family members and the relative abundance of 24 immune cell types in LUAD. EXOSC1 was positively correlated with Th2 cells and Tgd, but negatively associated with multiple immune cells including B cells, macrophages, neutrophils, T cells, NK cells, dendritic cells, and Tregs (Fig. 7A). EXOSC2 showed a positive relationship with Th2 and T helper cells, while exhibiting negative correlations with B cells, NK cells, CD8 + T cells, and other innate and adaptive immune subsets (Fig. 7B). Similarly, EXOSC3 expression was positively linked to Th2 cells, T helper cells, Tgd, and Tcm, but negatively correlated with NK cells, macrophages, neutrophils, cytotoxic cells, and B cells (Fig. 7C). EXOSC4 expression correlated positively with NK CD56bright and Th17 cells, but negatively with macrophages, neutrophils, T cells, B cells, dendritic cells, and cytotoxic cells (Fig. 7D). EXOSC5 was positively related to Th2 cells, while negatively correlated with T cells, B cells, dendritic cells, macrophages, and neutrophils (Fig. 7E). EXOSC6 showed a positive association with Tcm but negative correlations with cytotoxic, CD8 + T, NK, macrophages, neutrophils, and regulatory T cells (Fig. 7F). EXOSC7 expression exhibited broad negative correlations, particularly with B cells, T cells, neutrophils, macrophages, and NK CD56dim cells (Fig. 7G). EXOSC8 was positively related to Th2 and T helper cells but negatively associated with a wide range of immune cells, including NK cells, dendritic cells, macrophages, and CD8 + T cells (Fig. 7H). Both EXOSC9 (Fig. 7I) and EXOSC10 (Fig. 7J) were positively associated with Th2, T helper cells, and Tcm, but negatively correlated with B cells, CD8 + T cells, dendritic cells, macrophages, neutrophils, and cytotoxic cells. Collectively, these findings highlight that EXOSC family genes exhibit diverse and complex immune correlations, suggesting potential roles in shaping the tumor immune microenvironment of LUAD. Fig. 7Correlation between EXOSC family gene expression and immune cell infiltration in LUAD (*p < 0.05, **p < 0.01, ***p < 0.001). Lollipop plots showing correlations between EXOSC1-10 (A-J) expression and the infiltration of 24 immune cell types in LUAD Correlation between EXOSC family gene expression and immune cell infiltration in LUAD (*p < 0.05, **p < 0.01, ***p < 0.001). Lollipop plots showing correlations between EXOSC1-10 (A-J) expression and the infiltration of 24 immune cell types in LUAD To explore potential associations between the expression of EXOSC family genes and features of the tumor immune microenvironment, we computed ImmuneScore, StromalScore, and ESTIMATEScore for each sample using the ESTIMATE algorithm. (Fig. 8A-J). The results revealed statistically significant negative correlations between the expression of all EXOSC family genes and each of the three microenvironment scores. These findings suggest that elevated expression of EXOSC family genes may be associated with reduced infiltration of certain immune cells in the tumor microenvironment, indicating a potential role for these genes in modulating the tumor immune landscape. Further functional experiments and validation in clinical cohorts are warranted to confirm these observations. Fig. 8Correlation between EXOSC family gene expression and tumor microenvironment scores in LUAD (*p < 0.05, **p < 0.01, ***p < 0.001). ESTIMATE-derived ImmuneScore, StromalScore, and ESTIMATEScore show significant negative correlations with all EXOSC family genes (A-J) in LUAD Correlation between EXOSC family gene expression and tumor microenvironment scores in LUAD (*p < 0.05, **p < 0.01, ***p < 0.001). ESTIMATE-derived ImmuneScore, StromalScore, and ESTIMATEScore show significant negative correlations with all EXOSC family genes (A-J) in LUAD To evaluate the relationship between EXOSC family genes and immune infiltration, we performed a Wilcoxon rank-sum test to compare immune cell enrichment between high- and low-expression groups (Fig. 9A-J, sample sizes in Supplementary Table 8). Overall, high expression of EXOSC1, EXOSC2, EXOSC3, EXOSC5, EXOSC8, EXOSC9, and EXOSC10 was consistently associated with increased Th2 or T helper cell enrichment, but reduced infiltration of a wide range of immune subsets, including B cells, macrophages, dendritic cells (DCs and pDCs), NK cells, CD8 + T cells, cytotoxic cells, neutrophils, Th1 cells, and Tregs. Specifically, EXOSC3 and EXOSC9 additionally correlated with elevated Tcm levels, whereas EXOSC8 and EXOSC9 were enriched in T helper cells. Distinct patterns were also observed. EXOSC4 high expression was positively associated with NK CD56bright cells, Tgd, and Th2 cells, while showing reduced enrichment of most innate and adaptive immune cells. EXOSC6 was uniquely linked to increased Tcm infiltration but reduced CD8 + T cells, NK cells, and Th1/Th2 cells. In contrast, EXOSC7 high expression primarily exhibited negative correlations, especially with B cells, neutrophils, T cells, Th1 cells, and Tregs, without significant enrichment of other immune subsets. Collectively, these findings indicate that elevated EXOSC family expression in LUAD is generally characterized by enrichment of Th2/T helper cell subsets alongside broad suppression of cytotoxic and antigen-presenting immune populations, suggesting a potential role of EXOSC genes in shaping an immunosuppressive tumor microenvironment. Fig. 9Comparison of immune cell infiltration between high- and low-expression groups of EXOSC family genes (A-J) in LUAD (*p < 0.05, **p < 0.01, ***p < 0.001) Comparison of immune cell infiltration between high- and low-expression groups of EXOSC family genes (A-J) in LUAD (*p < 0.05, **p < 0.01, ***p < 0.001) GSEA of EXOSC family genes in LUAD To investigate the potential functional roles of the EXOSC family in LUAD, we performed GSEA by stratifying patients into high- and low-expression groups according to the median expression of each gene (Fig. 10). Distinct enrichment patterns were observed across family members. Specifically, EXOSC1 was enriched in chromatin-modifying enzymes, cellular senescence, and Histone acetyltransferases acetylate histones (Fig. 10A). EXOSC2 was associated with cell cycle, mitotic processes, and Wnt signaling (Fig. 10B). EXOSC3 correlated with HDAC-mediated histone deacetylation, cellular senescence, and histone acetylation (Fig. 10C). EXOSC4 and EXOSC5 were both enriched in cellular senescence, histone acetylation, and programmed cell death-related pathways (Figs. 10D-E). EXOSC6 was linked to matrisome-associated processes, innate immune responses, and post-translational protein modification (Fig. 10F). EXOSC7 showed enrichment in cellular senescence, histone acetylation, and cell cycle (Fig. 10G). EXOSC8 was associated with histone methylation, histone demethylation (HDMs), and Notch signaling (Fig. 10H). EXOSC9 correlated with extracellular matrix (ECM)-related processes, NRF2 signaling, and matrisome pathways (Fig. 10I). Finally, EXOSC10 was enriched in histone acetylation, cellular senescence, and DNA replication (Fig. 10J). Significance was determined using thresholds of FDR < 0.25, P < 0.05, and NES > 1. Pathways with the highest NES values were prioritized for interpretation, highlighting the biological implications of EXOSC gene expression in LUAD. Fig. 10Gene set enrichment analysis (GSEA) of EXOSC family genes (A-J) in LUAD. Pathway enrichment significance was defined by an FDR < 0.25, P-value < 0.05, and NES > 1 Gene set enrichment analysis (GSEA) of EXOSC family genes (A-J) in LUAD. Pathway enrichment significance was defined by an FDR < 0.25, P-value < 0.05, and NES > 1 Genes co-expressed with EXOSC family genes in LUAD and enrichment analysis of EXOSC family co-expression genes in LUAD Transcriptome analysis of TCGA data revealed extensive gene co-expression networks associated with the EXOSC family. Specifically, the numbers of positively correlated genes for EXOSC1-10 were 6,478, 10,081, 9,771, 4,792, 5,707, 10,732, 9,307, 8,356, 13,029, and 12,956, respectively. Conversely, the numbers of negatively correlated genes were 4,118, 2,367, 2,034, 8,085, 5,988, 731, 973, 2,284, 711, and 463, respectively (sample sizes in Supplementary Table 9). The top 10 positively and negatively correlated genes for each EXOSC family member in LUAD are depicted in Fig. 11A-J. Additionally, A Venn diagram identified 291 overlapping genes co-expressed with all EXOSC members (|R| > 0.15, P < 0.001; Fig. 12A). To clarify their potential biological roles, GO and KEGG enrichment analyses were performed. GO analysis revealed that co-expressed genes were mainly enriched in ribonucleoprotein complex biogenesis, ncRNA and rRNA processing, and ribosome biogenesis (Biological Process, BP); mitochondrial matrix, preribosome, spliceosomal complex, and small-subunit processome (Cellular Component, CC); as well as RNA catalytic activity, RNA helicase activity, and snoRNA binding (Molecular Function, MF). KEGG analysis indicated that these genes were primarily involved in the spliceosome, ribosome biogenesis, nucleocytoplasmic transport, mRNA surveillance, and aminoacyl-tRNA biosynthesis pathways (Fig. 12B). Fig. 11Top co-expressed genes of EXOSC family members in LUAD (*p < 0.05, **p < 0.01, ***p < 0.001). Correlation networks for EXOSC1-10 (A-J) Top co-expressed genes of EXOSC family members in LUAD (*p < 0.05, **p < 0.01, ***p < 0.001). Correlation networks for EXOSC1-10 (A-J) Fig. 12Shared co-expressed genes and functional enrichment of EXOSC family co-expression networks in LUAD. A Venn diagram showing 291 overlapping genes co-expressed with all EXOSC family members (|R| > 0.15, P < 0.001). B Gene Ontology (GO), and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis of the 291 shared co-expressed genes. GO enrichment analysis: Biological Process (BP), Cellular Component (CC), and Molecular Function (MF) Shared co-expressed genes and functional enrichment of EXOSC family co-expression networks in LUAD. A Venn diagram showing 291 overlapping genes co-expressed with all EXOSC family members (|R| > 0.15, P < 0.001). B Gene Ontology (GO), and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis of the 291 shared co-expressed genes. GO enrichment analysis: Biological Process (BP), Cellular Component (CC), and Molecular Function (MF) Development of a prognostic model and nomogram based on EXOSC family co-expression genes in LUAD To identify prognostic biomarkers among EXOSC family co-expressed genes in LUAD, we first performed univariate Cox regression analysis, which revealed 72 genes significantly associated with overall survival (Fig. 13A-B). Subsequently, LASSO Cox regression refined this set to nine key prognostic genes: AMPD1, ANP32B, CHCHD4, DDX56, GTF3C6, PAICS, PTGES3, RCCD1, and TRIM28 (Fig. 13C). Based on these genes, we established a prognostic risk model with the following formula: Risk score = (AMPD1 × -0.02578) + (ANP32B × 0.0940) + (CHCHD4 × 0.0237) + (DDX56 × 0.1927) + (GTF3C6 × 0.1927) + (PAICS × 0.0066) + (PTGES3 × 0.1103) + (RCCD1 × 0.1168) + (TRIM28 × 0.0741). We applied this model to the TCGA LUAD cohort (n = 539) and stratified patients into high-risk (n = 270) and low-risk (n = 269) groups according to the median risk score. The prognostic utility of the model was supported by three observations. First, survival analysis showed that high-risk patients had significantly poorer overall survival and higher mortality than low-risk patients (Fig. 13D). Second, Kaplan-Meier curves confirmed a marked survival difference between the two groups (P < 0.001; Fig. 13E). Third, time-dependent ROC analysis demonstrated moderate predictive performance, with AUCs of 0.675, 0.639, and 0.616 for 1-, 3-, and 5-year survival, respectively (Fig. 13F). We further validated the nine-gene risk model using the GSE31210 dataset. The model demonstrated strong diagnostic capability, with an AUC of 0.939 (Fig. 13G). In addition, Kaplan-Meier analysis revealed a significant survival difference between the high- and low-risk groups (P < 0.001; Fig. 13H). Time-dependent ROC curves indicated moderate prognostic performance, yielding AUCs of 0.730, 0.674, and 0.764 for 1-, 3-, and 5-year survival (Fig. 13I), respectively. To facilitate clinical translation, we constructed a nomogram integrating the LASSO-derived risk score with clinical variables (Fig. 14). Importantly, multivariate Cox analysis indicated that the risk score was one of the strongest independent predictors of prognosis, highlighting its potential application in individualized risk stratification and clinical decision-making. Fig. 13Development and validation of a nine-gene prognostic model based on EXOSC family co-expressed genes in LUAD (*p < 0.05, **p < 0.01, ***p < 0.001)A-B Univariate Cox regression identified 72 EXOSC co-expressed genes significantly associated with overall survival. C LASSO Cox regression refined these to nine key prognostic genes (AMPD1, ANP32B, CHCHD4, DDX56, GTF3C6, PAICS, PTGES3, RCCD1, TRIM28). D Distribution of risk scores and survival status of LUAD patients in high- and low-risk groups (n = 270 and 269, respectively). High-risk patients exhibited higher mortality. Integrated visualization of risk scores, survival outcomes, and clinical variables. The risk score panel shows individual scores sorted by magnitude. Survival times are plotted on the y-axis in the time status panel, with colors representing different outcomes. Risk groups are indicated by distinct colors, and the heatmap displays both numerical and categorical variables, with a unified color scale applied to numerical data. (E) Kaplan–Meier survival curves showing significantly poorer overall survival in the high-risk group compared with the low-risk group (P < 0.001). (F) Time-dependent ROC analysis of the prognostic model with AUCs of 0.675, 0.639, and 0.616 for 1-, 3-, and 5-year survival, respectively, indicating moderate predictive performance. (G-I) Validation in the GSE31210 dataset confirmed the model’s strong diagnostic capability (AUC 0.939), significant survival differences between risk groups, and moderate prognostic performance for 1-, 3-, and 5-year survival (AUC 0.730, 0.674, 0.764) Development and validation of a nine-gene prognostic model based on EXOSC family co-expressed genes in LUAD (*p < 0.05, **p < 0.01, ***p < 0.001)A-B Univariate Cox regression identified 72 EXOSC co-expressed genes significantly associated with overall survival. C LASSO Cox regression refined these to nine key prognostic genes (AMPD1, ANP32B, CHCHD4, DDX56, GTF3C6, PAICS, PTGES3, RCCD1, TRIM28). D Distribution of risk scores and survival status of LUAD patients in high- and low-risk groups (n = 270 and 269, respectively). High-risk patients exhibited higher mortality. Integrated visualization of risk scores, survival outcomes, and clinical variables. The risk score panel shows individual scores sorted by magnitude. Survival times are plotted on the y-axis in the time status panel, with colors representing different outcomes. Risk groups are indicated by distinct colors, and the heatmap displays both numerical and categorical variables, with a unified color scale applied to numerical data. (E) Kaplan–Meier survival curves showing significantly poorer overall survival in the high-risk group compared with the low-risk group (P < 0.001). (F) Time-dependent ROC analysis of the prognostic model with AUCs of 0.675, 0.639, and 0.616 for 1-, 3-, and 5-year survival, respectively, indicating moderate predictive performance. (G-I) Validation in the GSE31210 dataset confirmed the model’s strong diagnostic capability (AUC 0.939), significant survival differences between risk groups, and moderate prognostic performance for 1-, 3-, and 5-year survival (AUC 0.730, 0.674, 0.764) Fig. 14Nomogram integrating the nine-gene risk score and clinical variables for LUAD prognosis Nomogram integrating the nine-gene risk score and clinical variables for LUAD prognosis The relationship between EXOSC family genes levles and TMB and HRD in LUAD We further investigated the associations between EXOSC family gene expression and genomic instability–related indicators, including tumor mutational burden (TMB) and homologous recombination deficiency (HRD), in lung adenocarcinoma (LUAD). As shown in Fig. 15A, the expression levels of EXOSC1, EXOSC2, EXOSC3, EXOSC5, EXOSC8, EXOSC9, and EXOSC10 were positively correlated with TMB in LUAD. Consistently, Fig. 15B demonstrates that the expression of EXOSC1, EXOSC2, EXOSC3, EXOSC5, EXOSC8, EXOSC9, and EXOSC10 was also positively correlated with HRD, indicating a potential association between elevated EXOSC expression and increased genomic instability in LUAD. Fig. 15Association of EXOSC family gene expression with genomic instability in LUAD (*p < 0.05, **p < 0.01, ***p < 0.001). A Correlation analysis between EXOSC family gene expression and tumor mutational burden (TMB) in TCGA-LUAD. B Correlation analysis between EXOSC family gene expression and homologous recombination deficiency (HRD) scores in TCGA-LUAD Association of EXOSC family gene expression with genomic instability in LUAD (*p < 0.05, **p < 0.01, ***p < 0.001). A Correlation analysis between EXOSC family gene expression and tumor mutational burden (TMB) in TCGA-LUAD. B Correlation analysis between EXOSC family gene expression and homologous recombination deficiency (HRD) scores in TCGA-LUAD In addition, GSEA was performed to determine whether high expression of EXOSC family genes was associated with activation of DNA damage response-related pathways (Supplementary Fig. 1). Elevated EXOSC1 expression was significantly enriched in pathways related to mitotic cell cycle, cell cycle regulation, telomere stress–induced senescence, cell cycle checkpoints, and the G2/M DNA damage checkpoint. High EXOSC2 expression was predominantly enriched in mitotic cell cycle, cell cycle checkpoints, cell cycle, and telomere stress-induced senescence pathways. Similarly, increased EXOSC3 expression was associated with enrichment of mitotic cell cycle, telomere stress-induced senescence, the G2/M DNA damage checkpoint, cell cycle checkpoints, and cell cycle pathways. High EXOSC4 expression showed enrichment in telomere stress-induced senescence, mitotic cell cycle, cell cycle, the G2/M DNA damage checkpoint, and cell cycle checkpoints, whereas elevated EXOSC5 expression was enriched in telomere stress-induced senescence, mitotic cell cycle, cell cycle, cell cycle checkpoints, and the G2/M DNA damage checkpoint. In addition, high EXOSC7 expression was associated with cellular senescence, cell cycle, mitotic cell cycle, telomere stress-induced senescence, cell cycle checkpoints, and the G2/M DNA damage checkpoint. High EXOSC8 expression was enriched in cellular responses to stimuli and mitotic cell cycle, while elevated EXOSC10 expression was linked to enrichment of cellular senescence, diseases of programmed cell death, cell cycle checkpoints, the G2/M DNA damage checkpoint, mitotic cell cycle, and cell cycle pathways. Exploring EXOSC family genes expressions across various cells of the tumor microenvironment (TME) through single-cell data analysis To comprehensively characterize the TME associated with EXOSC family genes, we analyzed single-cell RNA sequencing (scRNA-seq) data from lung adenocarcinoma (LUAD). Analysis of the GSE146100 dataset identified 18 distinct cell clusters, which were further classified into 11 major cell types (Fig. 16A). We found that EXOSC family genes were most highly expressed in proliferating monocytes/macrophages, followed by ten other cell types: conventional CD4⁺ T cells, regulatory T cells, CD8⁺ T cells, natural killer cells, B cells, dendritic cells, mast cells, endothelial cells, fibroblasts, and epithelial cells (Fig. 16B). Cell-cell interaction analysis indicated that proliferating monocytes/macrophages cells primarily interacted with endothelial cells, dendritic cells, conventional CD4⁺ T cells, and CD8⁺ T cells. Furthermore, pathway analysis revealed that key signaling cascades such as TGF‑β was involved in the communication network between EXOSC‑high tumor cells and other constituents of the tumor microenvironment. (Fig. 16C). Fig. 16Single-cell landscape of EXOSC family gene expression and intercellular communication in LUAD. A UMAP visualization of 18 cell clusters from the GSE146100 scRNA-seq dataset, annotated into 11 major cell types. B Dot/violin plot summarizing EXOSC family gene expression across TME cell populations, with highest expression observed in proliferating monocytes/macrophages, followed by conventional CD4⁺ T cells, regulatory T cells, CD8⁺ T cells, natural killer cells, B cells, dendritic cells, mast cells, endothelial cells, fibroblasts, and epithelial cells. C Cell–cell interaction analysis highlighting major communication partners of proliferating monocytes/macrophages (including endothelial cells, dendritic cells, conventional CD4⁺ T cells and CD8⁺ T cells) and indicating involvement of TGF-β signaling within the inferred communication network Single-cell landscape of EXOSC family gene expression and intercellular communication in LUAD. A UMAP visualization of 18 cell clusters from the GSE146100 scRNA-seq dataset, annotated into 11 major cell types. B Dot/violin plot summarizing EXOSC family gene expression across TME cell populations, with highest expression observed in proliferating monocytes/macrophages, followed by conventional CD4⁺ T cells, regulatory T cells, CD8⁺ T cells, natural killer cells, B cells, dendritic cells, mast cells, endothelial cells, fibroblasts, and epithelial cells. C Cell–cell interaction analysis highlighting major communication partners of proliferating monocytes/macrophages (including endothelial cells, dendritic cells, conventional CD4⁺ T cells and CD8⁺ T cells) and indicating involvement of TGF-β signaling within the inferred communication network Relationship of EXOSC genes with PARP1 expression and drug sensitivity in LUAD cell models Using the DepMap database, we evaluated the association between dependency scores of EXOSC family genes and expression of the DNA repair gene PARP1 in LUAD cell lines (Fig. 17A). The dependency scores of EXOSC1, EXOSC2, EXOSC3, EXOSC6, EXOSC7, EXOSC8, EXOSC9, and EXOSC10 showed showed negative correlation trends with PARP1 expression, whereas EXOSC4 and EXOSC5 exhibited a positive correlation trend; however, none of these associations reached statistical significance (all P > 0.05). In addition, drug response prediction was conducted based on TCGA transcriptomic data using the OncoPredict algorithm to assess the relationship between EXOSC expression and predicted sensitivity to chemotherapy (cisplatin) and targeted therapy (osimertinib), with sensitivity represented by estimated IC50 values. In LUAD, expression levels of EXOSC2, EXOSC5, EXOSC6, EXOSC7, EXOSC8, EXOSC9, and EXOSC10 showed a negative correlation trend with cisplatin IC50, among which EXOSC9 demonstrated a statistically significant association (P < 0.05), whereas EXOSC1 and EXOSC4 exhibited a positive correlation trend with cisplatin IC50 (all others P > 0.05, Fig. 17B). By contrast, predicted sensitivity to osimertinib showed negative correlation trends with EXOSC1, EXOSC5, and EXOSC9, and positive correlation trends with EXOSC2, EXOSC6, EXOSC7, EXOSC8, and EXOSC10, although none of these associations reached statistical significance (all P > 0.05, Fig. 17C). Fig. 17Associations of EXOSC genes with PARP1 and predicted drug sensitivity in LUAD models. A DepMap-based correlation analysis between EXOSC gene dependency scores and PARP1 expression across LUAD cell lines, showing negative correlation trends for EXOSC1/2/3/6/7/8/9/10 and positive trends for EXOSC4/5, without statistical significance (all P > 0.05). B OncoPredict analysis in TCGA-LUAD showing correlations between EXOSC expression and predicted cisplatin IC50, with negative trends for EXOSC2/5/6/7/8/9/10 (EXOSC9 significant, P < 0.05) and positive trends for EXOSC1 and EXOSC4. C Correlation of EXOSC expression with predicted osimertinib IC50, showing non-significant negative trends for EXOSC1/5/9 and positive trends for EXOSC2/6/7/8/10 (all P > 0.05) Associations of EXOSC genes with PARP1 and predicted drug sensitivity in LUAD models. A DepMap-based correlation analysis between EXOSC gene dependency scores and PARP1 expression across LUAD cell lines, showing negative correlation trends for EXOSC1/2/3/6/7/8/9/10 and positive trends for EXOSC4/5, without statistical significance (all P > 0.05). B OncoPredict analysis in TCGA-LUAD showing correlations between EXOSC expression and predicted cisplatin IC50, with negative trends for EXOSC2/5/6/7/8/9/10 (EXOSC9 significant, P < 0.05) and positive trends for EXOSC1 and EXOSC4. C Correlation of EXOSC expression with predicted osimertinib IC50, showing non-significant negative trends for EXOSC1/5/9 and positive trends for EXOSC2/6/7/8/10 (all P > 0.05) Discussion LUAD continues to pose major clinical challenges due to its genomic complexity, marked heterogeneity, and limited therapeutic responsiveness [1, 2]. Despite recent advances in targeted therapy and immunotherapy, reliable biomarkers for risk stratification and mechanistically informed therapeutic targets remain insufficient. In this study, we integrated multi-omics datasets to characterize the landscape of EXOSC family genes in LUAD. By integrating transcriptomic, proteomic, genomic, immune, epigenetic, and co-expression data, we generated a comprehensive view of EXOSC biology in LUAD. Importantly, all molecular and clinical relationships described in this study should be interpreted as associative and correlative rather than causal, unless otherwise demonstrated by future functional experiments. The RNA exosome is a highly conserved 3’-5’ ribonuclease complex essential for RNA surveillance, rRNA/ncRNA processing, and degradation of aberrant or misprocessed transcripts [5, 6]. We observed consistent overexpression of multiple EXOSC members (EXOSC1-10) at both mRNA and protein levels across TCGA, CPTAC, and HPA datasets, suggesting that elevated EXOSC expression commonly coexists with LUAD tumor states. These observations are aligned with previous studies showing that EXOSC components can accompany proliferative and transcriptionally active tumor phenotypes [10, 11]. Our results extend this observation to a family-wide evaluation in LUAD. Among all EXOSC members, EXOSC2-3 and EXOSC5 exhibited a significant association with shorter OS, suggesting they may reflect a prognostically relevant vulnerability. In addition, several members (EXOSC1, EXOSC4, EXOSC5) showed high diagnostic accuracy, indicating potential value in distinguishing LUAD from normal tissues. While EXOSC6, EXOSC7, and EXOSC9, with AUC values ranging from approximately 0.53 to 0.58, are therefore classified as having “limited diagnostic utility.” Clinicopathological analyses revealed that higher EXOSC expression was frequently associated with advanced T, N, and M classifications, as well as smoking history. EXOSC2, for example, showed enrichment in T4, N2-N3, M1, and late-stage tumors. These patterns suggest that elevated EXOSC levels coexist with clinically aggressive LUAD phenotypes. EXOSC6 and EXOSC7 demonstrated context-dependent associations, supporting functional heterogeneity among exosome subunits. Epigenetic modifications are recognized as key contributors to the initiation and progression of malignancies [29]. EXOSC4 exhibited the highest alteration frequency (predominantly amplifications), followed by EXOSC10. Amplifications have been associated in other malignancies with increased ribonuclease complex abundance and enhanced RNA-degradation capacity [30]. Although causality cannot be inferred from our correlative data, the coexistence of EXOSC amplifications with elevated transcript levels and co-expression patterns suggests a possible genomic contribution to elevated EXOSC activity in LUAD. The strong positive correlations among most EXOSC members may reflect shared transcriptional control or convergent epigenetic activation. Discrete exceptions highlight potential internal divergence within the complex, which aligns with reports that different exosome modules (cap components vs. core barrel) may be regulated by distinct upstream factors. EXOSC4 is a structural component of the RNA exosome complex and plays a critical role in mRNA degradation, non-coding RNA processing, and RNA quality control [31]. In the present study, integrative bioinformatics analyses revealed that EXOSC4 is amplified and overexpressed in LUAD, suggesting that tumor cells may develop a dependency on its elevated expression. Such amplification may enhance RNA metabolic capacity and sustain the high transcriptional burden characteristic of rapidly proliferating tumor cells, reflecting a potential “transcriptional” or “RNA processing” addiction. This provides a theoretical basis for synthetic lethality-based or dependency-oriented therapeutic strategies. However, as a non-enzymatic structural protein essential for normal cellular function, EXOSC4 lacks well-defined small-molecule binding pockets, and its direct inhibition may carry substantial toxicity risks [31]. Therefore, it does not represent a conventional druggable target. Determining whether EXOSC4 amplification confers cancer cell-specific dependency is crucial for evaluating its therapeutic potential. If amplification enhances RNA exosome activity or sustains oncogenic transcriptional programs, tumor cells may exhibit dose-dependent reliance on EXOSC4 expression. Functional interrogation using CRISPR-mediated perturbation, combined with RNA sequencing to compare amplified and non-amplified cells, would help determine whether a phenomenon analogous to “oncogene addiction” exists. From a therapeutic perspective, a more feasible approach may involve exploiting synthetic lethality by targeting cooperative pathways involved in RNA processing and transcription, such as RNA polymerase II, CDK7/CDK9, or spliceosome-related regulators, thereby exacerbating transcriptional stress and inducing tumor cell collapse. Although RNA interference or antisense oligonucleotide based strategies are theoretically possible, their application would require clear evidence that tumor cells are substantially more dependent on EXOSC4 than normal cells. Furthermore, if EXOSC4 amplification is associated with immune suppression or increased replication stress, combination strategies incorporating immune checkpoint blockade or ATR/CHK1 inhibitors may offer additional therapeutic opportunities. Nevertheless, these therapeutic considerations remain hypothesis-generating, because the current analyses do not establish that EXOSC4 directly drives these phenotypes. Protein-protein interaction (PPI) network analysis identified canonical RNA-processing partners (DIS3, DIS3L, C1D, MPHOSPH6), and enrichment analyses consistently linked EXOSC-associated proteins to RNA degradation, exoribonuclease activity, rRNA processing, and ribosome biogenesis. These data imply EXOSC upregulation in LUAD coexists with broad activation of ribosome production and RNA-processing pathways, a hallmark of proliferative tumors with high biosynthetic demands. Furthermore, co-expression analysis revealed a shared set of 291 overlapping genes involved in ribonucleoprotein assembly, ncRNA/rRNA maturation, and mRNA surveillance. These results support a model in which elevated EXOSC expression aligns with a global high-flux RNA metabolic state, possibly reflecting increased transcriptional burden or enhanced turnover of unstable transcripts in LUAD tumors [5]. However, the observed links between EXOSC expression and these functional pathways should be interpreted strictly as correlative co-enrichment patterns rather than evidence of direct pathway regulation. Immune infiltration is a critical determinant of LUAD progression and treatment response [32]. In our study, EXOSC family genes showed consistent associations with distinct immune signatures across ssGSEA profiling, high-low expression subgroup analyses, and ESTIMATE-based purity correction. Although these results are correlative, they outline immune features that frequently accompany elevated EXOSC expression in LUAD. Across multiple EXOSC members, including EXOSC1, EXOSC2, EXOSC3, EXOSC5, EXOSC8, EXOSC9, and EXOSC10, higher expression was generally associated with increased enrichment of Th2 and T helper cell subsets, coupled with reduced signatures of cytotoxic and antigen-presenting populations, such as CD8⁺ T cells, NK cells, dendritic cells, macrophages, and cytotoxic lymphocytes. This overall pattern resembles immunosuppression-compatible immune states reported in recent LUAD immunogenomic studies [33, 34], though mechanistic involvement cannot be inferred from transcriptomic associations alone. Gene-specific distinctions were observed as well: EXOSC4 showed positive associations with NK CD56bright and Th17 cells, whereas EXOSC7 exhibited predominantly negative correlations across most immune subsets. The observed trends were further supported by purity-corrected analyses. All EXOSC members showed significant negative correlations with ImmuneScore, StromalScore, and ESTIMATEScore, suggesting that elevated EXOSC expression tends to align with reduced immune and stromal infiltration at the bulk-tumor level. These findings indicate that the patterns identified are unlikely to be solely driven by sample purity or proliferative transcriptional backgrounds, although additional refined deconvolution analyses could strengthen this interpretation. Associations with immune-checkpoint molecules provide additional context. EXOSC3, EXOSC9, and EXOSC10 showed positive correlations with PD-L1 and CTLA4, consistent with LUAD subsets characterized by diminished cytotoxic activity and enhanced regulatory signaling [35]. Nevertheless, these associations with immune infiltration and immune checkpoint genes do not imply direct regulatory effects or causal effects, and may instead reflect shared transcriptional programs. These observations generate testable hypotheses regarding potential interactions between RNA-metabolic states and immune contexture in LUAD, warranting further mechanistic studies. Based on bioinformatic analyses suggesting that members of the EXOSC family (e.g., EXOSC4/5) may be associated with immune suppression in LUAD, a multi-level experimental framework is required to systematically determine whether this association is genuine, causal, and mechanistically defined. At the in vitro molecular level, LUAD cell lines can be engineered using CRISPR-Cas9 knockout, shRNA-mediated knockdown, and overexpression systems, preferably comparing amplified and non-amplified models. RNA sequencing should then be performed to assess changes in immune-related genes, including immune checkpoint molecules (PD-L1), chemokines (CCL2, CXCL9, CXCL10), antigen presentation genes, and signaling pathways such as IFN-γ and JAK/STAT, in order to determine whether EXOSC directly regulates immune-associated transcriptional programs. RNA stability assays using actinomycin D treatment can further clarify whether EXOSC exerts its effects through modulation of mRNA half-life. These findings should be validated at the protein level by Western blotting, flow cytometry, and ELISA. At the functional level, tumor-immune cell co-culture assays can be performed to evaluate whether EXOSC depletion enhances CD8⁺ T cell proliferation, IFN-γ secretion, and cytotoxic activity, or alters macrophage polarization (M1/M2 balance), thereby establishing its role in regulating immune cell function. For causal validation in vivo, immunocompetent syngeneic mouse models should be employed to assess the impact of EXOSC manipulation on tumor growth, survival, and the tumor immune microenvironment, including infiltration of CD8⁺ T cells, Tregs, MDSCs, and macrophages. The potential synergistic effects with anti-PD-L1 therapy should also be evaluated. Finally, clinical relevance can be examined through tissue microarray-based immunohistochemistry, single-cell RNA sequencing, and spatial transcriptomics to determine whether high EXOSC expression correlates with an immunosuppressive tumor microenvironment or spatial “immune-cold” regions. If these findings collectively support a coherent mechanistic axis-EXOSC upregulation leading to altered immune-related RNA regulation, modulation of chemokine and PD-L1 expression, reduced CD8⁺ T cell infiltration, enhanced tumor growth, and altered immunotherapy responsiveness-this would establish EXOSC as a functional regulator of tumor immune suppression and provide a foundation for further investigation into pathways such as STAT3/NF-κB signaling, STING activation, or dsRNA-mediated innate immune sensing. Members of the EXOSC family constitute core components of the RNA exosome complex, which mediates 3′-5′ RNA processing and degradation and maintains RNA quality control [5, 6]. Therefore, their impact on immune cell infiltration in LUAD is more likely attributable to post-transcriptional regulation of RNA homeostasis rather than direct RNA chemical modification. Mechanistically, altered EXOSC activity may influence the stability of transcripts encoding chemokines and cytokines, such as CXCL9, CXCL10, CCL2, and CXCL8, thereby modulating the recruitment of T cells, monocytes/macrophages, and neutrophils to the tumor microenvironment [5]. In addition, impaired RNA surveillance may lead to the accumulation of aberrant or endogenous double-stranded RNA, triggering RIG-I/MDA5-MAVS signaling and type Ι interferon responses, which can further reshape immune activation and infiltration [36]. Dysregulation of the nuclear RNA exosome may also promote R-loop accumulation, replication stress, and DNA damage, potentially activating the cGAS-STING pathway and downstream inflammatory programs [37]. Moreover, altered RNA turnover may influence cellular stress responses, including ER stress and the unfolded protein response, thereby indirectly affecting immune-related gene expression and tumor immunogenicity [38]. Although EXOSC proteins are not canonical RNA modification enzymes, they may indirectly interact with epitranscriptomic regulation by modulating the expression of m6A machinery components or altering the abundance and isoform distribution of modified transcripts. Collectively, these considerations provide biologically plausible explanations for the observed correlations, but they remain speculative mechanistic models that have not yet been validated experimentally in LUAD. Accumulating evidence suggests that RNA modifications play a vital role in the development and progression of human cancers and may represent promising targets for tumor therapy [39]. GSEA highlighted several impactful biological programs associated with EXOSC expression: Chromatin remodeling and histone modifications, suggesting that EXOSC expression may coincide with broader epigenomic reprogramming that supports oncogenic transcription [40]– [42]. Cell cycle and mitotic regulation, particularly for EXOSC2, aligning with proliferative phenotypes in LUAD [43, 44]. Wnt, Notch signaling, observed for EXOSC2 and EXOSC8, respectively, which may reflect lineage specification or stem-like features [45]. ECM remodeling and NRF2-stress pathways, linked to EXOSC9 and EXOSC6, compatible with invasive and metabolic adaptation phenotypes [46]. These enrichments provide multiple mechanistic hypotheses-though all remain correlative-that EXOSC dysregulation may interact with LUAD biology through RNA metabolism-linked coordination of transcriptional and post-transcriptional programs. We observed notable correlations between EXOSC expression and m1A, m5C, and m6A RNA modification regulators. Methylation writers (METTL3, NSUN2, TRMT6/61) showed positive correlations, whereas demethylases (FTO, ALKBH5) exhibited negative associations for several EXOSC members. Individual EXOSC members exhibited distinct molecular signatures. EXOSC4, which demonstrated the highest mutation rate and multiple negative correlations with regulators of m¹A, m⁵C, and m⁶A pathways (such as FTO, METTL14, and YTHDC1), was associated with transcriptional programs related to chromatin remodeling, nucleolar activity, and RNA-processing intensity. Recent studies suggest that RNA-processing factors often coincide with epigenetically remodeled tumor states [47, 48], implying that elevated EXOSC4 expression may characterize LUAD subsets with high transcriptional throughput; however, functional involvement requires experimental validation. In contrast, EXOSC9 was the only family member consistently associated with disease-specific survival, and it showed stronger associations with immune-related signatures, including lower infiltration of cytotoxic immune cells and higher expression of immune checkpoints. These findings resemble transcriptional patterns compatible with immunosuppression reported in recent pan-cancer datasets. However, the relationships between EXOSC genes and RNA modification regulators are likewise correlative rather than causal, and do not establish direct biochemical interactions or regulatory hierarchies between EXOSC activity and RNA modification machinery. The broad correlations between EXOSC genes and m¹A, m⁵C, and m⁶A regulators further link the exosome to epitranscriptomic programs. Emerging evidence demonstrates that RNA methylation can reshape RNA structure, modulate reader-protein binding, or alter the accessibility of transcripts to ribonucleases [49, 50]. Consistent with this, most EXOSC members showed positive correlations with m¹A writers/readers (TRMT6, TRMT61A, and YTHDF1), m⁵C regulators (NSUN2, DNMT3A/3B), and m⁶A machinery (METTL3/4). Meanwhile, EXOSC4 and EXOSC5 demonstrated unique negative correlations with demethylases such as FTO and ALKBH5. Although these findings do not establish directionality, they raise mechanistically plausible hypotheses. One possibility is that certain RNA-modification states may influence substrate recognition during RNA surveillance [51]. For example, m⁶A-dependent structural switches are known to expose or occlude ribonuclease-accessible regions, whereas m⁵C and m¹A can modulate RNA stability or reader interactions [52]. Alternatively, co-activation of EXOSC and RNA-modification pathways may reflect broader transcriptomic remodeling programs in rapidly proliferating LUAD cells. These interpretations remain speculative and require biochemical testing. Functional enrichment analyses further showed that EXOSC-high tumors were enriched for pathways associated with chromatin organization, ribosome biogenesis, DNA damage response, oxidative stress, and cell-cycle progression. These transcriptional programs parallel accumulating evidence linking RNA-decay dysregulation to R-loop accumulation, chromatin-state transitions, and replication stress in cancer. Nonetheless, the present data cannot establish whether EXOSC expression contributes to these programs or simply coexists with them within highly proliferative LUAD subtypes. Taken together, the findings support a conceptual “EXOSC-RNA metabolism-immune/epigenetic axis” in LUAD, in which: (i) elevated EXOSC expression aligns with enhanced ribosomal and RNA-processing signatures; (ii) EXOSC members correlate with distinct m¹A, m⁵C, and m⁶A modification profiles; and (iii) multiple EXOSC genes-particularly EXOSC9-track with immune features compatible with reduced cytotoxic activity and increased immune-checkpoint expression. While causality cannot be established, this axis provides a structured framework for interpreting the coordinated molecular patterns associated with EXOSC dysregulation. This study provides an expanded biological context for understanding the multifaceted roles of EXOSC family members in LUAD and generates testable hypotheses regarding how RNA degradation, RNA modifications, and immune regulation may intersect. Future work incorporating perturbation experiments, RNA-decay kinetics, epitranscriptomic profiling, and immune-functional assays will be essential to dissect potential mechanistic interactions within this axis and to determine whether specific EXOSC members may serve as biomarkers or therapeutic targets. The nine-gene signature shows only moderate discriminatory performance and should currently be regarded as a hypothesis-generating tool rather than an immediately applicable clinical assay. Some signature components may function as surrogate markers rather than direct mediators of EXOSC biology. Some signature genes (e.g., TRIM28, DDX56, RCCD1) have mechanistic relevance to chromatin regulation or nucleolar biology, consistent with EXOSC-associated pathways [53, 54]. However, the model’s discriminatory power is modest, and several components may function as surrogate markers rather than direct mediators of EXOSC biology. Accordingly, the signature should be regarded as hypothesis-generating, requiring multi-cohort validation and mechanistic grounding before clinical translation. Translating a nine-gene prognostic model from statistical significance into a clinically applicable tool generally requires four sequential stages: technical validation, clinical validation, assessment of clinical utility, and standardized implementation. First, the clinical positioning of the model must be clearly defined, such as postoperative recurrence risk stratification in early-stage LUAD, survival prediction in advanced disease, evaluation of immunotherapy efficacy, or guidance for adjuvant treatment decisions. The model should be considered as a complementary risk stratification instrument to the TNM staging system rather than a replacement for established staging criteria. Second, a standardized detection platform must be established to convert the transcriptome-based algorithm into a clinically feasible assay. Platforms such as qRT-PCR or NanoString using formalin-fixed paraffin-embedded specimens are practical options, provided that stringent protocols are implemented for RNA quality control, reference gene normalization, and correction of inter-batch variability. Prospective validation is the cornerstone for incorporation into clinical guidelines. Independent multicenter cohorts encompassing diverse populations and treatment backgrounds should be recruited, with a recommended sample size of 300–500 patients to ensure adequate event numbers. Risk score cutoffs, primary endpoints (e.g., overall survival or disease-free survival), and predefined subgroup analyses must be prospectively registered to minimize overfitting. Model performance evaluation should extend beyond hazard ratios and include discrimination metrics (C-index, time-dependent AUC, and ROC curves), calibration analysis, and decision curve analysis. Importantly, the incremental predictive value of “TNM plus the nine-gene model” should be directly compared with TNM staging alone. Stratified analyses across disease stages, mutational subgroups, and immunotherapy-treated populations are essential, as is assessment of cross-platform reproducibility. Crucially, the model must demonstrate an ability to influence clinical decision-making. For example, if significant differences in recurrence rates are observed between high- and low-risk groups, subsequent studies should determine whether risk-adapted therapeutic strategies improve outcomes. An optimal translational framework includes both prospective observational validation and higher-level risk-stratified interventional trials to assess whether model-guided management enhances survival. Common causes of failure include lack of external validation, batch effects, unstable cutoffs, and failure to outperform existing staging systems, underscoring the need for rigorous statistical procedures such as cross-validation, bootstrap resampling, adherence to TRIPOD reporting standards, and net reclassification improvement analysis. If developed as an in vitro diagnostic assay, additional analytical and clinical performance validation, quality system compliance, and regulatory approval are required. Ultimately, only by completing the full translational pathway-from prediction to stratification, treatment modification, and survival improvement-can the nine-gene model achieve genuine clinical utility. We also explored the potential associations between EXOSC family gene expression and multiple dimensions of genomic instability, tumor microenvironment heterogeneity, and therapeutic response in LUAD. Correlation analyses indicated that several EXOSC family members exhibited positive associations with TMB and HRD, suggesting that higher EXOSC expression levels tend to co-occur with increased mutational load and homologous recombination-related genomic alterations. It should be emphasized that these findings reflect statistical correlations rather than causal relationships; they do not demonstrate that EXOSC expression induces genomic instability, nor that genomic instability necessarily drives EXOSC upregulation. Consistent with these observations, GSEA results showed that high expression of multiple EXOSC genes was enriched in pathways involved in cell cycle regulation, mitotic progression, telomere stress-induced senescence, and DNA damage checkpoints. These pathway-level associations support the notion that EXOSC genes may be preferentially expressed in tumors with active cell cycle dynamics and DNA damage response signaling, although functional validation is required to clarify their mechanistic roles. Notably, the associations were not uniform across the EXOSC family. EXOSC6 and EXOSC7 displayed relatively limited or context-dependent correlations with TMB, HRD, and drug sensitivity, in contrast to other family members. This heterogeneity suggests functional diversification within the EXOSC family, rather than a fully homogeneous role of the RNA exosome complex in LUAD. Such subunit-specific or context-dependent functions are consistent with recent studies reporting differential roles of RNA exosome components in cancer biology, stress responses, and genome maintenance [55, 56]. Gene set enrichment analysis further showed that high expression of multiple EXOSC genes was enriched in pathways related to cell cycle regulation, mitotic progression, telomere stress-induced senescence, and DNA damage checkpoints. These findings are in line with emerging evidence that RNA metabolism-related factors are closely associated with cell cycle dynamics and genome stability [57]. However, these pathway-level enrichments primarily reflect co-activation patterns and do not establish direct regulatory roles of EXOSC genes in DNA damage response processes. Single-cell transcriptomic analysis further revealed that EXOSC family genes were not restricted to malignant epithelial cells but were also prominently expressed in proliferating monocytes/macrophages and various immune and stromal cell populations within the TME. This broad cellular distribution suggests that EXOSC expression may reflect both tumor-intrinsic proliferative states and the activity of specific immune or stromal compartments. Moreover, inferred cell-cell communication analyses highlighted signaling pathway TGF-β in EXOSC-high contexts, implying potential links between EXOSC expression and inflammatory or growth factor-related signaling networks in the TME. These results again indicate associations rather than direct regulatory effects and should be interpreted with caution. These findings are consistent with recent reports highlighting interactions between RNA regulatory machinery and tumor-associated inflammatory signaling [58], although they remain correlative in nature. And then, analyses of LUAD cell line data showed no statistically significant correlations between EXOSC dependency scores and PARP1 expression, and only limited associations between EXOSC expression and predicted drug sensitivity. While certain EXOSC genes displayed trends toward increased cisplatin sensitivity, with EXOSC9 reaching statistical significance, most relationships with cisplatin or osimertinib IC50 values were modest and non-significant. Thus, the observed links between EXOSC expression and therapeutic response should also be interpreted as correlative rather than predictive or causal, and they are currently insufficient to support the use of EXOSC genes as validated biomarkers of drug sensitivity or DNA repair dependency. Together, these findings suggest that EXOSC family genes may be linked to genomic instability and cell cycle-related features in LUAD, but their utility as predictive markers for DNA repair dependency or drug response remains uncertain. Further experimental and clinical studies are needed to determine whether these observed correlations translate into functional relevance or therapeutic implications. Recent studies have increasingly highlighted the value of integrative computational approaches in cancer and disease research. Representative examples include network toxicology and molecular docking analyses of chemical exposure-related effects [59, 60], immune feature-based prognostic modeling and deep learning-based prediction of immunotherapy response [55], multimodal artificial intelligence frameworks for prognostic assessment in specific tumor types [61], and broad narrative reviews of multi-component therapeutic strategies [62]. The present study was not designed as a single-gene, single-pathway, or single-task predictive analysis; rather, it constitutes a family-wide, disease-focused integrative investigation of EXOSC1-10 in LUAD. By integrating transcriptomic and proteomic analyses, external cohort validation, clinicopathological correlations, genomic alteration profiling, analyses of immune infiltration and immune checkpoint associations, correlations with RNA modification regulators, co-expression network construction, prognostic signature development, single-cell localization, inference of cell-cell communication, and exploratory assessments of genomic instability and drug sensitivity, we established a broader analytical framework. In this regard, our study extends beyond previous work primarily focused on exposure-related target prediction, immune biomarker modeling, or tumor-specific multimodal risk prediction, and offers a systematic, family-level perspective on EXOSC dysregulation and its potential biological and clinical relevance in LUAD. The future studies should prioritize functional perturbation experiments, such as CRISPR/Cas9-mediated knockout, RNA interference, rescue assays, RNA-decay kinetics, epitranscriptomic profiling, immune co-culture systems, and spatial/single-cell validation, to determine whether key EXOSC members directly regulate LUAD progression and immune remodeling. Limitations This study has several limitations. First, although we integrated transcriptomic, proteomic, genomic, immune-profiling, and RNA-modification datasets, all associations identified in this study are correlative rather than causal. In particular, the relationships between EXOSC expression and immune infiltration, immune checkpoint genes, RNA modification regulators, functional pathway enrichment, genomic instability features (including TMB and HRD), and therapeutic response measures should not be interpreted as evidence of direct biological regulation or causal dependence. These findings cannot establish mechanistic relationships between EXOSC dysregulation and immune remodeling, RNA-metabolism alterations, pathway activation, genome maintenance defects, or drug sensitivity in LUAD. Second, while we constructed and externally validated an EXOSC-associated prognostic signature, the model was developed using retrospective public datasets, and additional prospective cohorts are required to confirm its clinical applicability. Third, variability in sequencing platforms, sample processing, and annotation across TCGA, CPTAC, HPA, and GEO may introduce technical bias despite normalization and harmonization procedures. Fourth, our analyses do not replace experimental validation; the biological mechanisms by which specific EXOSC members influence RNA degradation, immune regulation, genomic stability, or LUAD progression warrant further investigation through functional and mechanistic studies. Additionally, TMB, HRD, and drug sensitivity estimates are indirect measures that may be influenced by tumor heterogeneity, model assumptions, and methodological constraints. The limited availability of single-cell cohorts, together with heterogeneity in patient origin and treatment background, may influence the accuracy of immune cell proportion estimates and inferred cell-cell communication patterns. Moreover, as the single-cell data were derived from a specific cohort, they may not fully reflect the biological diversity and complexity of the LUAD tumor microenvironment. Further functional and clinical studies are warranted to clarify the context-dependent roles of individual EXOSC family members and to determine whether the observed correlative associations have genuine biological or therapeutic relevance. Conclusion The study presents an integrated characterization of the EXOSC family in lung adenocarcinoma (LUAD) using bulk multi-omics, external validation, and single-cell analyses. Overall, EXOSC1-10 were broadly upregulated in LUAD, with several members, particularly EXOSC2, EXOSC3, and EXOSC5, showing prognostic and diagnostic relevance. Functional analyses linked elevated EXOSC expression to RNA processing, chromatin regulation, cell-cycle progression, senescence-associated pathways, and genomic instability, while a nine-gene prognostic signature demonstrated independent predictive value. Immunologically, high EXOSC expression was associated with a microenvironment consistent with immune suppression, characterized by lower stromal and immune scores, enrichment of Th2/T helper populations, and reduced cytotoxic and antigen-presenting immune components. Single-cell data further localized EXOSC activity to proliferative monocyte/macrophage populations and suggested TGF-β-related intercellular communication. Although exploratory analyses indicated potential associations with chemotherapy sensitivity, the findings also highlighted substantial functional heterogeneity among EXOSC subunits, especially for EXOSC6 and EXOSC7. Taken together, these results suggest that dysregulated EXOSC activity may contribute to LUAD progression through coordinated effects on RNA metabolism, tumor proliferation, immune remodeling, and genome maintenance, while supporting the need for further mechanistic and clinical validation of specific EXOSC members as biomarkers and therapeutic targets. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1. Supplementary Material 1. Supplementary Material 2. Supplementary Material 2. Supplementary Material 3. Supplementary Material 3. Supplementary Material 4. Supplementary Material 4. Supplementary Material 5. Supplementary Material 5. Supplementary Material 6. Supplementary Material 6. Supplementary Material 7. Supplementary Material 7. Supplementary Material 8. Supplementary Material 8. Supplementary Material 9. Supplementary Material 9. Supplementary Material 10. Supplementary Material 10.