Immune checkpoint inhibitor (ICI) therapy has transformed the treatment of advanced clear cell renal cell carcinoma (ccRCC), yet only a minority of patients achieve durable responses. Identifying reliable predictors of ICI response is a critical priority to guide patient selection and avoid exposing non-responders to treatment toxicity.
Exosomes are nano-sized extracellular vesicles secreted by virtually all cell types, including tumor cells. Tumor-derived exosomes carry a cargo of proteins, nucleic acids, and non-coding RNAs that can reprogram the immune microenvironment, suppress anti-tumor immunity, and facilitate immune evasion, making exosome biology highly relevant to immunotherapy response.
Long non-coding RNAs (lncRNAs) are increasingly recognized as functional components of exosome cargo, with the ability to modulate recipient cell gene expression and immune function. This study aimed to identify exosome-related lncRNAs (ERLRs) in ccRCC and develop a prognostic score linking their expression to overall survival and ICI response prediction.
Gene expression data from 539 ccRCC tissue samples in the TCGA-KIRC dataset were analyzed alongside a curated list of exosome-related genes derived from the ExoCarta database. lncRNAs correlated with these exosome-associated gene sets were identified as candidate ERLRs for further analysis.
LASSO (Least Absolute Shrinkage and Selection Operator) regression combined with Cox proportional hazards modeling was applied to reduce the candidate list to the minimum set of lncRNAs with independent prognostic value. LASSO penalizes model complexity, preventing overfitting and ensuring that only the most informative predictors are retained in the final signature.
The final ERLR score was constructed as a weighted linear combination of four lncRNA expression values with LASSO-derived coefficients. Patients were divided into high-risk and low-risk groups based on their ERLR score, and group differences in survival, immune infiltration, and genomic characteristics were systematically analyzed.
The four lncRNAs comprising the ERLR score are EMX2OS, AC026401.3, AC018690.1, and AL161935.1. Each was independently associated with overall survival and each carries a distinct regulatory role within the exosome-mediated communication network of the ccRCC tumor microenvironment.
EMX2OS is an antisense lncRNA to the EMX2 transcription factor gene, with documented roles in developmental signaling and emerging evidence for function in tumor suppression. Its dysregulation in high-risk ccRCC suggests loss of regulatory control over EMX2-dependent transcriptional programs.
The other three lncRNAs (AC026401.3, AC018690.1, AL161935.1) are less well-characterized but were consistently associated with immune-related gene expression programs and exosome pathway activity in the TCGA analysis. Their biological functions in ccRCC represent an important area for future mechanistic investigation.
High ERLR score patients had significantly worse overall survival compared to low-risk patients in Kaplan-Meier analysis, with the difference maintained across multiple subgroup analyses stratified by age, sex, and clinical stage. The ERLR score showed significant independent prognostic value in multivariate Cox regression controlling for established clinical factors.
High-risk patients also presented with more advanced pathological stage (T3-T4, higher nodal involvement) and higher histological grade, confirming that the ERLR score captures biologically aggressive disease characteristics rather than merely reflecting clinical stage.
Tumor mutational burden (TMB) was significantly elevated in high-risk ERLR patients, which is often associated with both worse prognosis and potential for ICI response in patients who can overcome immune suppression. This dual association adds nuance to the interpretation of high TMB in ccRCC.
Immune cell infiltration analysis revealed that high-risk ERLR patients harbored a more immunosuppressive tumor microenvironment, with higher abundance of regulatory T cells (Tregs), M2-polarized macrophages, and myeloid-derived suppressor cells. These populations are well-established mediators of T cell dysfunction that impair ICI therapeutic efficacy.
Despite the immunosuppressive baseline, ICI response prediction algorithms suggested that high-risk ERLR patients may show relative benefit from anti-PD1 and anti-CTLA4 therapy, likely because their elevated TMB and higher checkpoint expression levels create an ICI-susceptible immune state that can be unblocked by checkpoint inhibition.
These findings suggest that the ERLR score provides nuanced clinical guidance: high-risk patients should likely receive ICI-based regimens rather than targeted therapy alone, while low-risk patients may have adequate outcomes with targeted agents and could potentially be spared the toxicity of aggressive immunotherapy combinations.
ceRNA network analysis identified a specific regulatory axis involving EMX2OS, hsa-miR-31-5p, and TLN2 (talin 2). In this model, EMX2OS acts as a molecular sponge for hsa-miR-31-5p, preventing this miRNA from silencing its target TLN2. TLN2 encodes a cytoskeletal protein involved in cell adhesion, migration, and mechanical signal transduction.
Dysregulation of this axis in high-risk ccRCC could promote tumor cell invasion and metastasis by relieving miR-31-5p-mediated suppression of TLN2, allowing increased cytoskeletal remodeling and integrin-mediated signaling. This mechanism provides a concrete molecular pathway connecting EMX2OS dysregulation to aggressive tumor behavior.
The EMX2OS/hsa-miR-31-5p/TLN2 axis represents a potential therapeutic target. Restoring miR-31-5p activity or inhibiting TLN2 signaling could theoretically reverse the invasive phenotype in EMX2OS-high tumors, though these strategies require experimental validation in ccRCC preclinical models before advancing to clinical testing.