The prognostic values of m6A RNA methylation regulators in uveal melanoma

Uveal melanoma is the most common type of malignant tumor of the adult eye, with an overall mortality rate of 50% [1, 2]. The prognosis for patients of UM remains poor, though advances in diagnosis and treatment have been reported [3, 4]. Therefore, it is important to explore the molecular mechanism underlying the survival events of UM and identify new prognostic factors and therapeutic targets.

It is well known that both DNA and histone proteins control gene expression by dynamic and reversible chemical modifications. RNA modification, like DNA and protein modification, is dynamically regulated by methyl-transferases [5]. The most prevalent RNA methylation is N6-methyladenosine (m6A), which exists in about 25% of transcripts at the genome-wide level and was firstly discovered in the 1970s. m6A RNA methylation regulators modify translocation, stability, RNA splicing and translation [6]. m6A is dynamically regulated by the ‘writers’ (RNA methyltransferases), such as METTL14, METTL3 and WTAP, is removed by ‘erasers’ (the demethylases), such as ALKBH5 and FTO, and ‘readers’ (the binding proteins), such as YTHDF2 and YTHDF1 [7]. RNA methyltransferases, the demethylases, and the binding proteins are often upregulated in a variety of human cancer types, increasing the expression of Oncogenes and Oncoproteins, augmenting the proliferation, progression, and metastasis of cancer cells [8].

m6A modification not only plays a vital role in the pathogenesis of a variety of human disease including obesity, neuronal disorders and immunological disease, but also has been shown to contribute to tumor initiation and promote progression of cancer and recurrence [9]. In addition, growing evidence suggests that gene mutation and abnormal expression of m6A regulators are intimately associated with the malignant progression of various cancers [10]. Although it is recognized that RNA methylation plays a critical role in different types of cancers, little is known about the relationship between m6A-related genes and UM.

Hence, in this study, we systematically evaluated the expression of m6A regulators in 80 UM samples from The Cancer Genome Atlas (TCGA) dataset as well as the association between the genetic alterations and clinical characteristics and validation in 28 UM samples from Gene Expression Omnibus (GEO) dataset. We found that the expression of m6A regulators plays critical roles in the malignant process of UM, and we identified three m6A regulators as potential biomarkers through their prognostic signatures.

The growing genome-wide studies demonstrated that most of the human genome is transcribed, which exists a complex network consist of large and small RNA molecules in human cells. However, only 1 to 2% of the transcripts have the capacity for protein translation [11‐13]. In fact, post-transcriptional regulation at the RNA level through cis-and trans-mechanisms is essential to control the gene expression procedures that determine cellular function and cell fate [14]. To date, more than 150 chemical modifications have been described for RNA. Among them, m6A is the most prevalent posttranscriptional modification of eukaryotic mRNAs and long noncoding RNAs. Recent studies have indicated that m6A regulators have been shown to play important regulatory roles in diverse biological processes in human cancer [15]. However, despite of the increasing evidence for their implication in cancers, the potential role of m6A regulators in UM prognosis is little known about.

In this study, we demonstrated that the expression of m6A regulators is also intimately related to the prognosis and malignancy of UM. Based on the expression of m6A regulators, we identified two UM subgroups, namely C1/2 molecular subgroup, by applying consensus clustered method. The C1/2 molecular subgroup not only affected the clinical and prognosis features but also closely associated with biological signals and malignant hallmarks of UM. Survival analysis showed that C1 subgroup have worse overall survival than C2 subgroup. In addition, C1 subgroup have higher percentage of subtype 4 which have been proven the worst outcome subgroup in previous UM TCGA study [16]. GSEA analysis showed that positive regulation of T cell mediated pathways and malignant hallmarks such as mTORC1 signaling, oxidative phosphorylation, interferon-a response and apoptosis signaling positively activate in C1 subgroup. In fact, T cells like active CD 4 + and CD 8 + cells have antitumor immunity and therapy functions [17]. As to C1 molecular subtype, lots of malignant hallmark of pathways were enriched. Thus, it is reasonable to believe that clustered molecular subtypes C1/2 are closely correlated to the malignancy and prognosis of UM. Moreover, extensive researches also suggest that UM with monosomy 3 is associated with a dramatically poor prognosis, which consistent with our research [18‐20]. The subgroups analysis of Chromosome.3.status showed that the percentage of Monosomy 3 in C1 molecular subgroup is much higher than this in C2 molecular subtype (Table 1). The different analysis of m6A regulators between C1/2 molecular subtype showed that “eraser” like ALKBH5, “writer” like METTL3, METTL14, and WTAP and “readers”, like YTHDF1 and YTHDF2 have significant differences. (Fig. 1e, f). The different expression of these m6A regulators may eventually lead to the various of survival outcomes [21]. Among the m6A regulators, previous studies indicated that the eraser ALKBH5 can induce breast cancer stem cell and glioblastoma stem-like cell proliferation and tumor initiation, [22] the writers METTL3 and METTL14 were reported to enhance glioblastoma growth and suppress Liver cancer metastasis, [23‐25] the reader YTHDF1 and YTHDF2 induce cancer cell proliferation in colon cancer and lung oncogenic effects [26, 27]. These findings manifested that high or low expression of specific m6A regulators are related to misregulation of RNAs in tumors, and the same m6A regulator may take different functions in various tumors [28, 29].

By analyzing the mutation annotation files of the TCGA-UVM cohort, we identified 5 highly variant mutated genes and SF3B1 is the most significantly influence the expression of m6A regulators. SF3B1 (splicing factor 3subunit B1) mutations can be generally found in 10 to 21% of cases of UM. Previous researches have shown that SF3B1 mutations in UM patients are associated with favorable prognosis [30]. Survival analysis also indicated that SF3B1-mutated UM had a better survival than the SF3B1 wild-type. In our research, the results showed that the mutation of SF3B1 will generally significantly down-regulated the expression of m6A regulators, including “eraser” such as ALKBH5 and FTO; “writer” such as WTAP and KIAA1429; “reader” such as YTHDF1, YTHDF2 and YTHDC2 (Fig. 2b). Therefore, it easily envisaged that the mutant of SF3B1 may lead to down-regulate the expression of “eraser” such as ALKBH5 and FTO and finally result in a better survival in UM.

What’s more, we also distinguished a prognostic risk signature with three identified m6A regulators (ALKBH5, YTHDF1 and KIAA1429), which divided the overall survival of UM into high- and low-risk subgroups. Kaplan-Meier analyses indicated that high-risk subgroups with a poor survival. Stratified analysis of clinical characteristics between low- and high- risk groups also revealed that lots of risk factors like mortality rate, subtype 4 and monosomy 3 are take higher percentage in high-risk group (Table 2). Furthermore, UM patients in high risk group had higher immune infiltration than low risk group. The risk sores of monosomy 3, SF3B1-wildtype, and subtype 4 were respectively higher than disomy 3, SF3B1-mutated and subtype 1 in UM, which was consistent with previous researches. Notably, compared with the 5 years AUC values of previous prognostic markers (stage, subtype and chromosome 3 status), our signature can achieve similar accuracy value. Besides, only the risk score had significant associations with OS no matter in univariate or multivariate regression analysis. In sum, the signature we constructed might be regarded as a new promising biomarker which supply more simple and accurate clinical applications. For example, in human breast cancer cells, knockdown ALKBH5 contributed to significantly decrease the number of cancer stem cells and the opportunity of tumorigenesis. In addition, the high expression of ALKBH5 in glioblastoma can lead to stem-like cell proliferation and tumorigenesis [31].

In summary, we firstly comprehensively evaluated the expression, potential function, and prognostic value of m6A regulatory genes in UM from TCGA dataset and have validated in GEO dataset, which should be helpful for UM early diagnosis and might be regarded as a new promising biomarker for UM prognosis and treatment.