Immune classification and identification of prognostic genes for uveal melanoma based on six immune cell signatures

Abstract

Cutaneous melanoma could be treated by immunotherapy, which only has limited efficacy on uveal melanoma (UM). UM immunotyping for predicting immunotherapeutic responses and guiding immunotherapy should be better understood. This study identified molecular subtypes and key genetic markers associated with immunotherapy through immunosignature analysis. We screened a 6-immune cell signature simultaneously correlated with UM prognosis. Three immune subtypes (IS) were determined based on the 6-immune cell signature. Overall survival (OS) of IS3 was the longest. Significant differences of linear discriminant analysis (LDA) score were detected among the three IS types. IS3 with the highest LDA score showed a low immunosuppression. IS1 with the lowest LDA score was more immunosuppressive. LDA score was significantly negatively correlated with most immune checkpoint-related genes, and could reflect UM patients' response to anti-PD1 immunotherapy. Weighted correlation network analysis (WGCNA) identified that salmon, purple, yellow modules were related to IS and screened 6 prognostic genes. Patients with high-expressed NME1 and TMEM255A developed poor prognosis, while those with high-expressed BEX5 and ROPN1 had better prognosis. There was no notable difference in OS between patients with high-expressed LRRN1 and ST13 and those with low-expressed LRRN1 and ST13. NME1, TMEM255A, Bex5 and ROPN1 showed potential prognostic significance in UM.

Figure 1

Consensus clustering identified three ISs of UM. (A) Venn diagram of intersection of immune cell signature significantly related to OS in patients with UM from TCGA and GSE22138 cohort. (B) CDF curves of TCGA cohort samples. (C) CDF Delta area curve of TCGA cohort samples (Delta area curve of consensus clustering, indicating the relative change in area under the CDF curve for each category number k compared with k − 1. The horizontal axis represents the category number k and the vertical axis represents the relative change in area under CDF curve). (D) The sample clustering heat map at k = 3. (E,F) Kaplan–Meier survival curves of patients with different ISs in the GSE22138 cohort.

Figure 2

Genetic mutations in different ISs. (A) TMB differences between the three ISs. (B) Difference analysis was used to compare the number of gene mutations in the three IS samples. (C) Mutation characteristics of the top5 significantly mutated genes in each IS samples. *P < 0.05, **P < 0.01. NS: not significant.

Figure 3

Associations between ISs and classical markers and immune checkpoints of chemotherapy-induced immune response. (A) Difference in expression of classic markers of chemotherapy-induced immune response in the three IS samples. (B) Differences in IFN-γ scores among the three ISs. (C) CYT score differences between the three ISs samples. (D) Differences in angiogenesis score among the three ISs. (E) The expression of 47 immune checkpoint-related genes differed in the three ISs. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001; Ns: not significant.

Figure 4

Association between IS and tumor immune microenvironment. (A) Heatmap of the degree of immune infiltration of 22 kinds of immune cells in different IS samples. (B) Box plots showing the relationships between IS and immune cell infiltrations. (C) Enrichment of three ISs in 10 oncogenic pathways. (D) Differences in immune scores between the three IS samples. (E) Comparison of the distribution of immune subtypes in different ISs. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001; Ns: not significant.

Figure 5

The LDA score can be used to measure the different immune characteristics of patients. (A) Relationship between the first 2 features of LDA score and ISs. (B) LDA score difference of different IS samples in TCGA data set. (C) LDA score differences of three ISs in GSE22138 data set. (D) ROC curve of LDA score in TCGA data set. (E) ROC curve of LDA score in GSE22138 data set. ***P < 0.001; ****P < 0.0001.

Figure 6

Relationship between LDA score and immunotherapy. (A) Relationship between LDA score and immune checkpoint. (B) Correlation between LDA score and PDCD1. (C) Correlation between LDA score and CD274 expression. (D) Correlation between LDA score and CTLA4 expression. (E) The LDA score of patients who responded to PD-1-blocking immunotherapy was significantly higher than that of patients did not respond to immunotherapy. *P < 0.05; **P < 0.01; ***P < 0.001.

Figure 7

Construction of co-expression network and identification of IS-related modules. (A) Clustering tree of UM samples in TCGA. (B,C) Analysis of the scale-free fit index and mean connectivity for various soft-thresholding powers (β). (D) 17 modules were identified by unsupervised clustering in WGCNA. (E) Transcript number statistics of each module. (F) Correlation between 17 modules and LDA score. (G) Correlation analysis between each module and clinical information. (H) Scatter plots of module eigengenes in the salmon, purple and yellow modules. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001.

Figure 8

Identification of prognosis- related genes in the module. (A) Univariate Cox analysis of the correlation between modules related to LDA score and OS. (B) Potential genetic markers associated with LDA score. (C) The correlation between 13 genes and UM prognosis was analyzed by univariate Cox analysis. (D–G) Kaplan–Meier curves of NME1, TMEM255A, BEX5, and ROPN1 expression in UM patients.