Identification of Molecular Subtypes Based on Tertiary Lymphoid Structure in Skin Cutaneous Melanoma

Abstract

Background: The tertiary lymphoid structures (TLSs) play a crucial role in the prognosis and response to skin cutaneous melanoma (SKCM) immunotherapy. Nevertheless, the molecular mechanisms underlying this connection remain unclear. Therefore, this study aimed to investigate the molecular subtypes of SKCM based on TLSs and to develop a scoring model using advanced bioinformatic approaches. Methods: The Cancer Genome Atlas (TCGA) data were retrieved from University Of Cingifornia Sisha Cruz Xena (UCSC-Xena), and the SKCM series matrix file GSE19234 was obtained from Gene Expression Omnibus (GEO). Using R 4.1.1 and the Sva package, the sample data were integrated to eliminate batch effects. Univariate Cox regression and Kaplan-Meier analysis were employed for the identification of prognosis-related tertiary lymphoid structure-related genes (TRGs). Based on TRGs expression, SKCM patients were categorized into different subtypes. Furthermore, unsupervised learning-cluster analysis was used to evaluate their clinical features, prognosis, and gene expression. Similarly, pathway enrichment and immune infiltration differences were assessed using Molecular Signatures Database (Msigdb), gene set variation analysis (GSVA), and single-sample gene set enrichment analysis (ssGSEA). Subsequently, subtypes were verified using limma, Gene Ontology (GO), and Kyoto Encyclopedia of Genes and Genomes (KEGG) analyses. Additionally, a scoring model was developed to analyze immune infiltration and the relationship between clinical features and drug sensitivity in SKCM. Results: Analysis of TCGA-SKCM revealed that over 80% of the SKCM patients exhibited TRG mutations, with missense mutation being predominant. Moreover, the Kaplan-Meier (K-M) survival analysis revealed that all TRGs acted as protective factors in SKCM, leading to prolonged survival among patients with elevated TRG expression. TRG expression clustered SKCM patients into three subtypes: Cluster A, Cluster B, and Cluster C. The K-M survival analysis revealed that patients in Cluster B exhibited the best prognosis and longest survival time as well as the highest TRG expression followed by Cluster A. Using 23 prognostic-related differentially expressed genes (DEGs), a scoring model was developed indicating worse prognosis and lower immune infiltration with higher scores. Lower scores correlated with longer survival, lower recurrence/metastasis, higher immune checkpoint expression, and better immunotherapy outcomes. Higher scores correlated with lower chemokine expression and reduced sensitivity to immune checkpoint inhibitors. Conclusions: We established a new classification of SKCM based on TLSs. The constructed scoring model could evaluate the immune infiltration, prognosis, and response of immunotherapy in SKCM, thereby assisting clinical treatment.