Identification and validation of a CCL18-related signature for prediction of overall survival in patients with uveal melanoma

Uveal melanoma (UM), the most frequent primary intraocular tumor in adults, has poor prognosis. High C-C motif chemokine ligand 18 (CCL18) has been detected in various tumors and is closely correlated with patients' clinicopathological characteristics. However, the essential role of CCL18 in UM remains unclear. Therefore, this study aimed to explore the prognostic value of CCL18 in UM. Uveal melanoma cells (M17) were transfected with pcDNA3.1-CCL18 si-RNA using Lipofectamine™ 2000. Cell growth and invasion abilities were measured through Cell Counting Kit-8 assay and invasion assay. RNA expression data and clinical and histopathological details were downloaded from the UM in The Cancer Genome Atlas (TCGA-UM) and GSE22138 datasets, which were defined as the training and validation cohorts, respectively. Univariate and multivariate Cox regression analyses were performed to identify significant prognostic biomarkers. The coefficients of these significant biomarkers generated by multivariate Cox proportional hazard regression analysis were used to establish a risk score formula. Functional enrichment analyses were also carried out. We found that downregulated CCL18 inhibits M17 cell growth and invasion in vitro. CCL18 may affect UM progression by altering C-C motif receptor 8 related pathways. Higher CCL18 expression was associated with worse clinical outcomes and tumor-specific death in the TCGA-UM dataset. Based on the coefficients obtained from the Cox proportional hazard regression analysis, a CCL18-related prognostic signature formula was constructed as follows: risk score = 0.05590 × age +2.43437 × chromosome 3 status +0.39496 × ExpressionCCL18. Notably, in this formula, the normal chromosome 3 was coded as 0, whereas the chromosome 3 loss was coded as 1. Each patient was assigned to either low-risk or high-risk groups using the median cut-off in the training cohort. High-risk patients survived for a shorter time than low-risk patients. The time-dependent and multivariate receiver operating characteristic curves showed promising diagnostic efficacy. Multivariate Cox regression analysis demonstrated the potential of this CCL18-related signature as an independent prognostic indicator. These results were validated using the GSE22138 dataset. In addition, in both TCGA-UM and GSE22138 datasets, stratification of clinical correlations and survival analyses based on this signature indicated the involvement of clinical progression and survival outcome in UM. In the high-risk group, Gene Ontology analyses mainly indicated the enrichment of immune response pathways, such as the T cell activation, response to interferon-gamma, antigen processing and presentation, interferon-gamma-mediated signaling pathway, MHC protein complex, MHC class II protein complex, antigen binding, and cytokine binding. Meanwhile, Kyoto Encyclopedia of Genes and Genomes analyses showed enrichments of pathways in cancer, cell adhesion, cytokine-cytokine receptor interaction, chemokine signaling pathway, Th1 and Th2 cell differentiation, and chemokine signaling pathway. Moreover, single-sample gene set enrichment analysis demonstrated the enrichment of almost all immune cells and immune functions in the high-risk group. In summary, a new prognostic CCL18-related signature was successfully established using the TCGA-UM dataset and validated using the GSE22138 dataset with meaningful predictive and diagnostic efficacies. This signature could serve as an independent and promising prognostic biomarker for patients with UM.