Comprehensive Analysis of Expression Regulation for RNA m6A Regulators With Clinical Significance in Human Cancers

Abstract

Background: N6-methyladenosine (m6A), the most abundant chemical modification on eukaryotic messenger RNA (mRNA), is modulated by three class of regulators namely "writers," "erasers," and "readers." Increasing studies have shown that aberrant expression of m6A regulators plays broad roles in tumorigenesis and progression. However, it is largely unknown regarding the expression regulation for RNA m6A regulators in human cancers.

Results: Here we characterized the expression profiles of RNA m6A regulators in 13 cancer types with The Cancer Genome Atlas (TCGA) data. We showed that METTL14, FTO, and ALKBH5 were down-regulated in most cancers, whereas YTHDF1 and IGF2BP3 were up-regulated in 12 cancer types except for thyroid carcinoma (THCA). Survival analysis further revealed that low expression of several m6A regulators displayed longer overall survival times. Then, we analyzed microRNA (miRNA)-regulated and DNA methylation-regulated expression changes of m6A regulators in pan-cancer. In total, we identified 158 miRNAs and 58 DNA methylation probes (DMPs) involved in expression regulation for RNA m6A regulators. Furthermore, we assessed the survival significance of those regulatory pairs. Among them, 10 miRNAs and 7 DMPs may promote cancer initiation and progression; conversely, 3 miRNA/mRNA pairs in kidney renal clear cell carcinoma (KIRC) may exert tumor-suppressor function. These findings are indicative of their potential prognostic values. Finally, we validated two of those miRNA/mRNA pairs (hsa-miR-1307-3p/METTL14 and hsa-miR-204-5p/IGF2BP3) that could serve a critical role for potential clinical application in KIRC patients.

Conclusions: Our findings highlighted the importance of upstream regulation (miRNA and DNA methylation) governing m6A regulators' expression in pan-cancer. As a result, we identified several informative regulatory pairs for prognostic stratification. Thus, our study provides new insights into molecular mechanisms of m6A modification in human cancers.

Keywords: DNA methylation; N6-methyladenosine; The Cancer Genome Atlas; microRNA; prognosis.

Figure 1

Pan-cancer expression alterations and prognostic values of m6A regulators. (A) RNA m6A modification is regulated by RNA m6A regulators, including “writers”-methyltransferase, “erasers”-demethylase, and “readers”-RNA m6A binding proteins. “Writers” consist of core components METTL3, METTL14, WTAP and other factors (KIAA1429, ZFP217, RBM15, RBM15B, and CBLL1). FTO and ALKBH5 are two “erasers.” “Readers” include HNRNPC, HNRNPA2B1, YTHDF1, YTHDF2, YTHDF3, YTHDC1, YTHDC2, IGF2BP1, IGF2BP2, IGF2BP3, and EIF3A. (B) Expression profiles of RNA m6A regulators in 13 cancer types. Up represents higher expression and down represents lower expression. The circle size represents the statistical significance after controlling FDR. (C) Representative examples of expression patterns of m6A regulators across four cancer stages. *P < 0.05, **P < 0.01, and ***P < 0.001. (D) Overview of prognostic effects of m6A regulators. High represents the patients with better prognosis when gene expression level is high, and low represents the patients with better prognosis when gene expression level is low.

Figure 2

The regulatory network and enriched pathways of miRNA-m6A regulators. (A) The regulatory network of miRNAs and m6A regulators in pan-cancer. In the pie chart, different colors represent different cancers, and size reflects the number of regulatory pairs. The circle represents miRNAs. The m6A regulators’ names were labeled. (B) The HNRNPA2B1 associated regulatory pairs in the pan-cancer network. The line width represents the number of cancers with this regulatory pair. (C) Statistics of hub miRNAs in 12 cancer types. When the connection of miRNA node in the network is greater than or equal to 2, the node is defined as hub miRNA. The top bar out of chart represents the number of hub miRNAs for each cancer and the right bar indicates the number of cancers for each miRNA. The redder the color, the more the connections. (D) Statistics of hub genes in 12 cancer types. When the connection of gene node is greater than or equal to 4, the node is defined as hub gene. The top bar out of chart is the number of hub regulators for each cancer. The right bar presented the number of cancers for each regulator. (E) Disease enrichment analysis of miRNAs. (F) Pathway enrichment analysis of miRNAs.

Figure 3

Summary of regulatory relationships between miRNAs and m6A regulators that potentially function as tumor-promoting (A) and tumor-antagonizing regulatory pairs (B). Lines of the same color represent the same type of cancer.

Figure 4

Construction of DMP-mRNA regulatory network. (A) Boxplot of Spearman’s correlation between DNA methylation data and mRNA-seq data across 11 cancer types. *P < 0.05, ***P < 0.001, and ****P < 0.0001. (B) The number of differentially methylated probes in different cancer types. (C) The regulatory network of DNA methylation probes and m6A regulators in pan-cancer. In the pie chart, different colors represent different cancers, and size reflects the number of regulatory pairs. The circle represents DMPs. (D) Statistics of the number of DMPs regulating m6A regulators in the pan-cancer regulatory network.

Figure 5

Summary of regulatory relationships between DNA methylation probes and m6A regulators that potentially affect patients prognosis in cancers.

Figure 6

Identification and analysis of key regulatory pairs in KIRC. (A) The process of building the signature using LASSO regression algorithm. (B) Biological process enrichment analysis of differentially expressed genes between high-risk and low-risk groups. (C) KEGG pathway enrichment analysis of differentially expressed genes in high-risk and low-risk groups. (D) Functional similarity analysis of gene sets in terms of biological processes. The order of enrichment items in the upper part of triangle is consistent with that of the left side. (E) Comparison of infiltration scores of 28 immune cell types between high-risk and low-risk groups. **P < 0.01, ***P < 0.001, and ****P < 0.0001.