Transcriptome and single-cell transcriptomics reveal prognostic value and potential mechanism of anoikis in skin cutaneous melanoma

Abstract

Background: Skin cutaneous melanoma (SKCM) is a highly lethal cancer, ranking among the top four deadliest cancers. This underscores the urgent need for novel biomarkers for SKCM diagnosis and prognosis. Anoikis plays a vital role in cancer growth and metastasis, and this study aims to investigate its prognostic value and mechanism of action in SKCM.

Methods: Utilizing consensus clustering, the SKCM samples were categorized into two distinct clusters A and B based on anoikis-related genes (ANRGs), with the B group exhibiting lower disease-specific survival (DSS). Gene set enrichment between distinct clusters was examined using Gene Set Variation Analysis (GSVA) and the Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis.

Results: We created a predictive model based on three anoikis-related differently expressed genes (DEGs), specifically, FASLG, IGF1, and PIK3R2. Moreover, the mechanism of these prognostic genes within the model was investigated at the cellular level using the single-cell sequencing dataset GSE115978. This analysis revealed that the FASLG gene was highly expressed on cluster 1 of Exhausted CD8( +) T (Tex) cells.

Conclusions: In conclusion, we have established a novel classification system for SKCM based on anoikis, which carries substantial clinical implications for SKCM patients. Notably, the elevated expression of the FASLG gene on cluster 1 of Tex cells could significantly impact SKCM prognosis through anoikis, thus offering a promising target for the development of immunotherapy for SKCM.

Keywords: Anoikis; Single-cell data analysis; Skin cutaneous melanoma; TCGA.

Fig. 1

The Landscape of Anoikis-Related Genes in Skin Cutaneous Melanoma. A Heat map depicting the top 50 genes with the most significant differences. B The volcano map of the DEGs. C The forest plot shows the top 14 ANRGs (p < 0.01) through univariate Cox regression analysis. D The network diagram illustrates the correlations between the top 14 ANRGs. E Copy number variations (CNVs) of 14 ANRGs in TCGA-SKCM. F The Circos plot displays the chromosome region and alteration of ANRGs. G Protein–protein interaction network. H The GGI network was built using GeneMANIA

Fig. 2

Consensus Cluster Analysis. A A heat map representing the consensus clustering solution (k = 2) for 23 genes in SKCM data is presented. B The empirical cumulative distribution function (CDF) picture shows the constant distribution of the various K values. C Kaplan–Meier analysis results of two molecular subtypes. D–F PCA (D), tSNE (E), and UMAP (F) distinguish two subtypes based on the expression of ANRGs. G. Heat map of clinical information and gene expression profiles of the two molecular subtypes based on 14 ANRGs. H The patterns of immune infiltration in the two subtypes. I. Kaplan–Meier analysis results of 2 molecular subtypes based on 14 ANRGs. J. GSEA analysis identifies potential signaling pathways in cluster A

Fig. 3

Construction and Validation of Prognostic Model Related to Anoikis. A–C Kaplan–Meier analysis of DSS in melanoma patients in the training dataset. D–F The time-dependent ROC curves of DSS for 1 (D), 3 (E), and 5 (F) years. G A heat map showing the patient's predictive 3-gene signature in the TCGA database. H An alluvial diagram of subtype and living status. I A box plot of risk scores in clusters A and B. J The forest plot summarizes the multivariable Cox regression analyses of the clinical features and the risk score in SKCM patients. K A nomogram plot based on risk score and clinicopathological factors

Fig. 4

Correlations Between Anoikis and Prognosis in SKCM Patients. A–L The DSS Kaplan–Meier curve of statistically significant subgroups for age (> 65 years vs. = 65 years) (A and B), gender (female vs. male) (C and D), N (N0-1 vs. N2-3) (I and J), T (T0-2 vs. T3-4) (K and L) in SKCM patients, and age (> 65 years vs. = 65 years) (E and F), gender (female vs. male) (G and H) in GSE65904 dataset

Fig. 5

Drug Sensitivity Analysis. A–F The high-risk group had higher sensitivity to BI-2536 (A), ERK_2440 (B), ERK-6604 (C), Lapatinib (D), NVP-ADW742 (E), and SB505124 (F) than the low-risk group. G–L The sensitivity of the low-risk group to 5-Fluorouracil (G), AGI-5198 (H), AGI-6780 (I), Alisertib (J), BMS_345541 (K), and AMG-319 (L) was higher than that of the high-risk group

Fig. 6

Results of Anoikis at the Single-Cell Level. A. Comparison of immune cell components between the high-risk and low-risk groups.. B. Correlation between immune cells and three hub ANRGs. C. Single-cell sequencing data were reduced in dimension to 23 clusters. D. Annotated results of major cell types. E. Result of FASLG dimension reduction in UMAP. F. Violin plot of FASLG expression on major cell lineages. G. Cell communication between Tex-C1 cells and other clusters. H. Down-regulated HALLMARK gene sets in different clusters. I. Down-regulated KEGG gene sets in different clusters