N6-Methyladenosine regulator RBM15B acts as an independent prognostic biomarker and its clinical significance in uveal melanoma.

Uveal melanoma (UM) is the most frequent intraocular malignant tumor in adults. N6-Methyladenosine (m6A) methylation is recognized as the most critical epigenetic change and is implicated in the development of many malignancies. However, its prognostic value in UM is poorly understood. RNA-seq and clinical data from The Cancer Genome Atlas (TCGA) help us better understand the relationship between m6A regulators and UM patients. Herein, four UM groups established by consensus clustering were shown to have different immune cell infiltrations and prognostic survival. Five m6A regulators, including RBM15B, IGF2BP1, IGF2BP2, YTHDF3, and YTHDF1, were associated with the prognosis of UM patients. Intriguingly, RBM15B was confirmed to be the only independent prognostic factor for UM and it was significantly correlated with clinicopathologic characteristics of UM. Notably, RBM15B expression was significantly negatively correlated with immune checkpoints. Furthermore, LINC00665/hsa-let-7b-5p/RBM15B axis and LINC00638/hsa-miR-103a-3p/RBM15B axis were found to be potential prognostic biomarkers in UM. In a nutshell, this work, through bioinformatics analysis, systematically described the gene signatures and prognostic values of m6A regulators. RBM15B is an independent protective prognostic factor, which may help us better understand the crosstalk within UM.

Introduction

Uveal melanoma (UM) is the most frequent intraocular malignant tumor in adults, developing from melanocytes located in the choroid (90%), ciliary body (6%), or iris (4%) (1). UM is more common between the ages of 50 and 70, and it is quite uncommon among children (2). Although UM accounts for approximately 5% of all primary melanoma patients, nearly 50% of UM becomes metastatic and transfers to the liver, reducing the quality of life and adding a significant cost to people and society (3, 4). The emergence and progression of uveal malignant melanoma is a complex and multifactorial process. Several therapeutic approaches have been tried in UM, including enucleation, brachytherapy, stereotactic radiotherapy, and proton therapy; however, few hopeful outcomes have been reported (5, 6). Therefore, it is imperative to identify reliable biomarkers for prognostic prediction and targeted treatment.

m6A has recently garnered a lot of attention (7, 8). Discovered in 1974, m6A is described as the methylation of the nitrogen atom (N) at the sixth position of adenine (9). m6A is the most prevalent transcriptional modification of eukaryotic mRNA. This methylation modification of m6A is reversible, with the involvement of methyltransferases (writers), demethylases (erasers), and methyl-binding proteins (readers) (10, 11). Dysregulation of m6A modification has also been linked to tumorigenesis, prognosis, and treatment of various cancers (12–14). However, just a few research focused on its influence on UM (15–18). He and colleagues, for example, revealed that Beta-Secretase 2 (BACE2) presented an increased level of m6A RNA methylation, which led to the upregulation of BACE2 mRNA (16). A previous study revealed an association between ferroptosis-related lncRNAs and uveal melanoma and further identified a five genes novel signature which has effects on prognosis for UM patients (19). The m6A regulator METTL3 was markedly increased in UM cells and proved to be an important oncogenic factor in UM progression (18). However, the mechanism of the additional m6A regulators in uveal melanoma tumorigenesis warrants more investigation.

RNA Binding Motif Protein 15B (RBM15B), which regulates the alternative mRNA splicing and functions as an mRNA export factor, is essential for m6A methylation (20). RBM15B is a functional competitor of the serine-arginine protein, inhibiting the activity of the CDK11(p110)-cyclin L2α complex and acting as a new CDK11(p110) binding partner (21). The physiological and pathological role of RBM15B in uveal melanoma has been unknown so far.

As a result, we performed a systematic bioinformatics analysis to reveal the effect of m6A regulators on UM patients. For the first time, we demonstrated that RBM15B is an independent prognostic factor for UM. Furthermore, we built the LINC00665/hsa-let-7b-5p/RBM15B axis and LINC00638/hsa-miR-103a-3p/RBM15B axis to illustrate the function of RBM15B. These data indicate that RBM15B is a potential diagnostic and prognostic target for UM.

Materials and methods

Data acquisition and identification of m6A-related regulators

The TCGA database (https://portal.gdc.cancer.gov/) was used to gather all normalized RNA-seq and clinical data from UM patients ( ). A total of 20 m6A-related regulators identified in the literature were included in the study, including seven writers (RBM15B, VIRMA, RBM15, METTL3, ZC3H13, WTAP, METTL14), two erasers (FTO, ALKBH5), and eleven readers (IGF2BP2, HNRNPA2B1, IGF2BP1, YTHDF3, IGF2BP3, HNRNPC, RBMX, YTHDC2, YTHDF1, YTHDC1, and YTHDF2) ( ) (22).

Genetic alteration and consensus clustering of m6A regulators

The cBioPortal (http://www.cbioportal.org/) database was used to determine the genetic changes in m6A regulators and their relationships to survival prognosis. The R software package “pheatmap” was used to show the 20 m6A regulators correlation map. Also, the geneMANIA online tool (https://genemania.org/) was used to examine the candidate proteins that were most connected with the 20 m6A regulators (23). The R package “consensus cluster plus” was used to divide the UM patients into four clusters to further understand the various etiology and clinical prognostic features of m6A. The maximum number of clusters was six, and in this procedure, 80% of the samples were drawn 100 times. The delta area curve of consensus clustering was then used to determine the relative change in area under the cumulative distribution function (CDF) curve. The R packages “ggplot2” and “pheatmap” were used to compare the expression distribution of m6A regulators across four groups.

Differences in immune cell infiltration across four clusters

QUANTISEQ algorithm is based on a novel signature matrix and a constrained least square regression, which is specifically designed for RNA-seq analysis. It also performes well with deconvolution in different cancer types. In our study, it analyzed 10 immune cell types and uncharacterized cells. Meanwhile, MCPCOUNTER algorithm, another scoring method based on a stringent and robust set of marker genes, was also used in our study to quantify 8 immune cells, fibroblasts, and endothelial cells. The packages “survival” and “survminer” were used to construct the Kaplan-Meier ′ s survival curve of UM patients (24).

Univariate and multivariate cox regression analyses

The “survival analysis” module from GEPIA2 (gene expression profiling interactive analysis 2) (http://gepia2.cancer-pku.cn) was used to perform survival analysis based on RBM15B expression (25), and the group cutoff was 50%. Furthermore, univariate and multivariate cox regression analyses were performed to investigate the potential correlation between 20 m6A regulators and OS of UM patients using the “survival” package (26). Furthermore, after demonstrating that RBM15B was the sole independent prognostic factor for OS, the R software packages “ggrisk”, “survival”, “survminer”, and “timeROC” were used to explore the survival value of RBM15B in UM patients.

Enrichment and immune infiltration analyses

The low and high-RBM15B expression data were used in Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) studies to investigate signaling pathways activated in UM. The criteria were set as follows: |logFC|>2, p<0.01. The R packages “ggplot2” and “clusterprofiler” were used throughout this process (27). In addition, the TIMER2.0 (http://timer.cistrome.org/) online database was used to explore the potential correlation between the amount of immune cell infiltration and RBM15B expression based on the R packages “ggplot2” and “pheatmap” (24). Furthermore, the expression of eight immune checkpoints their correlation with RBM15B was also explored. The eight immune checkpoints include programmed death-ligand 1 (CD274), cytotoxic t-lymphocyte-associated protein 4 (CTLA4), hepatitis a virus cellular receptor 2 (HAVCR2), lymphocyte activating 3 (LAG3), programmed cell death 1 (PDCD1), programmed cell death 1 ligand 2 (PDCD1LG2), T cell immunoreceptor with Ig and ITIM domains (TIGIT), and sialic acid binding Ig like lectin 15 (SIGLEC15). The spearman correlation of tumor mutation burden (TMB) and RBM15B gene expression was also investigated (28, 29).

Construction of ceRNA networks

Clinical data from the TCGA database were downloaded from the Genomic Data Commons (GDC) portal. ceRNA networks were established as follows: miRNAs and lncRNAs were predicted based on interactions in the Encyclopedia of RNA Interactomes (ENCORI); all predicted miRNAs and lncRNAs exhibited significant prognostic survival and correlations with RBM15B based on the OncomiR database.

All analyses were performed with the R software (version 4.0.3) and online databases ( ).

Results

Different expressions of m6A regulators revealed by consensus clustering

The analysis of genetic changes revealed the majority of m6A regulators had no genetic variants. As such, no significant correlation was found between m6A regulator mutations and overall survival (OS) or disease-free survival (DFS) ( ). According to the heatmap of the spearman correlation analysis, the 20 m6A regulators revealed complex relationships across writers, readers, and erasers. RBM15B was negatively correlated with most of the regulators, including VIRMA ( , R = -0.440, p < 0.001). In addition, the GeneMANIA data indicated that the most associated molecules of m6A regulator were METTL4, ALKBH7, CBLL1, and RIDA in protein-protein interaction networks ( ). Consensus clustering analysis was performed in 80 UM patients based on the TCGA to better understand the association between the 20 m6A regulators and clinical prognosis features. The results showed that k=4 was the best value for stable clustering when k ranged between 1 and 6 ( ). The clinical characteristics of the four groups are illustrated in . The expression levels of these regulators were compared across the four groups; the findings revealed that all 20 m6A regulators were expressed differently. (all p<0.05) ( ).

Figure 1

Correlations and consensus clustering analysis of 20 m6A regulators. (A) Heatmap of the correlation between the 20 m6A regulators. Red color represents positive correlation and blue represents negative correlation. (B) Twenty most associated molecules associated with 20 m6A regulators. (C) TCGA uveal melanoma cohort was classified into four clusters. (D) Cumulative distribution function (CDF) curve and (E) Delta area curve of consensus clustering. (F) Comparison of gene expression levels of 20 m6A regulators among four groups. *p<0.05, **p<0.01, ***p<0.001.

Immune cell infiltration contributes to different prognostic survivals

There were significant disparities in immune cell infiltration among the four subgroups. Two algorithms (QUANTISEQ and MCPCOUNTER) were used to examine the variations in immune cells across the four groups. The MCPCOUNTER algorithm revealed a statistically significant difference among all immune cells except B cells ( ). At the same time, the QUANTISEQ algorithm demonstrated a statistically significant difference across all the immune cells except monocytes and regulatory T cells ( ). The analysis of the survival status of the four subgroups revealed that group 4 had a worse OS and PFS than the other 3 groups ( ).

Figure 2

Differences of immune cells using QUANTISEQ and MCPCOUNTER algorithms, and overall survival and disease-free survival analysis in four groups. (A) MCPCOUNTER algorithms showed the differences of immune cells in different groups of samples. (B) QUANTISEQ algorithms showed the differences of immune cells in different groups of samples. (C) Kaplan-Meier OS survival analysis of the four different subtypes. (D) Kaplan-Meier PFS survival analysis of the four different subtypes. *p<0.05, **p<0.01, ***p<0.001.

Key m6A regulators based on GEPIA and various cox regressions

Analysis of the GEPIA2 database revealed that five genes (RBM15B, IGF2BP1, IGF2BP2, YTHDF3 and YTHDF1) significantly influenced the prognosis of UM patients ( ). Moreover, univariate and multivariate cox regression model were used to identify genes of prognostic significance. Univariate cox regression showed that RBM15B (p<0.001, HR=0.031), IGF2BP1 (p=0.039, HR=2.456), IGF2BP2 (p<0.001, HR=0.110), IGF2BP3 (p=0.007, HR=3.658), YTHDF3 (p=0.027, HR=2.864), and YTHDF1 (p=0.021, HR=3.225) were correlated with UM prognosis ( ). Multivariate cox regression demonstrated that RBM15B was the sole independent prognostic factor for OS (p=0.006, HR=0.053) ( , ).

Figure 3

Identification of RBM15B as an independent prognostic biomarker and its correlation with clinical characteristics. Overall survivals of UM patients were calculated by Kaplan-Meier curves based on different expression of (A) RBM15B, (B) IGF2BP1, (C) IGF2BP2, (D) YTHDF3 and (E) YTHDF1. (F) Univariate cox regression analysis and (G) Multivariate cox regression analysis of m6A regulators. RBM15B expression was significantly correlated with (H) OS, (I) DSS, (J) PFI, (K) Age and (L) Cancer stage. *p<0.05, ***p<0.001. OS, overall survival. DSS, disease specific survival. PFI, progression free interval.

Correlation of RBM15B expression with clinicopathologic features

RBM15B was discovered in the nucleoplasm and its RNA expression exhibited a low tissue specificity; however, it was not translated into proteins in certain tissues ( ). RBM15B had no genetic variants, according to genomic analysis. As such, no significant correlation was revealed between RBM15B and OS abnormalities and disease-specific survival (DSS) or progression-free survival (PFS) ( ). We separated UM patients into distinct subgroups to reveal the correlations between RBM15B and clinicopathologic features. The results demonstrated that increased levels of RBM15B expression were significantly correlated with improved OS, DSS, and progression-free interval (PFI) ( ). Moreover, significantly different RBM15B expression was observed in subgroups of patients’ age and cancer stages based on the UALCAN database ( ) (30). Logistic regression analysis revealed a significant correlation of RBM15B expression with clinical stage and histological type (p<0.05) ( ).

Predictive power of RBM15B expression and GO and KEGG enrichment analyses in UM

The scatter plot, heat map of gene expression, and receiver operating characteristic curves were used to evaluate the prognostic and diagnostic significance of RBM15B expression. The dotted line separated patients into low-risk and high-risk groups. According to the scatter plot and gene expression heatmap, UM patients with a higher RBM15B expression had a better OS than those with a low level of RBM15B expression (p<0.001) ( ). The time-dependent ROC analysis demonstrated that the area under the curve (AUC) for different survival years was 0.808, 0.791, and 0.767, respectively ( ). In addition, GO analysis revealed that RBM15B expression was primarily associated with immune-related gene terms: BP terms, including immune response-activating cell surface receptor signaling pathway, T cell activation, regulation of lymphocyte activation, humoral immune response, and lymphocyte differentiation ( ); CC terms, including plasma membrane receptor complex, external side of the plasma membrane, T cell receptor complex, immunoglobulin complex circulating, and host cell cytoplasm ( ); MF terms, including antigen-binding, cytokine activity, immunoglobulin receptor binding, MHC protein binding, and coreceptor activity ( ). KEGG pathway analysis revealed that RBM15B was significantly associated with natural killer cell-mediated cytotoxicity, T cell receptor signaling pathways, and primary immunodeficiency pathways ( ).

Figure 4

Prognostic value and GO and KEGG analysis of RBM15B. (A) Patients were divided into low and high groups based on RBM15B expression. (B) Time-dependent receiver operating characteristic analysis of RBM15B. Bubble charts showed the top 5 elements of (C) Biological process; (D) Cellular component; (E) Molecular function; and (F) Kyoto encyclopedia of genes and genomes analysis. GO, gene oncology, KEGG, kyoto encyclopedia of genes and genomes.

Correlations between RBM15B expression and immune checkpoints

The tumor immune dysfunction and exclusion (TIDE) algorithm was used to predict the potential immune checkpoint blockade (ICB) response (31). This demonstrated the significant difference in immune response scores between high and low RBM15B expression groups ( ). Then, using the R package “GSVA”, we analyzed the enrichment score distribution of immune cells between low and high RBM15B expression groups ( ); the results revealed significant differences in most of immune cells, except for CD 8+ T cell, mast cells, natural killer (NK) cells, plasmacytoid dendritic cell (pDC), and Central Memory T cells (32, 33). However, no significant relationships between RBM15B expression and tumor mutation burden (TMB) score or microsatellite instability (MSI) scores were observed ( ).

Figure 5

Function and pathway enrichment analysis of RBM15B and CD8(+) immune infiltrations in uveal melanoma cases. (A) distribution of immune response scores between high and low RBM15B expression groups in uveal melanoma. (B) Immune cell infiltration between low and high expression of RBM15B groups. (C) Expression distribution of immune checkpoints gene between low and high expression of RBM15B groups. Spearman correlation of RBM15B with expression of (D) CD274; (E) CTLA4; (F) HAVCR2; (G) LAG3 in UM. (H) Heatmap of immune checkpoints gene expression. The different colors represent the trend of gene expression in different samples. Spearman correlation of RBM15B with expression of (I) PDCD1; (J) PDCD1LG2; (K) TIGIT; (L) SIGLEC15 in UM. *p<0.05, **p<0.01, ***p<0.001. ns, no significance.

Additionally, the expression of eight immune checkpoints (including CD274, CTLA4, HAVCR2, LAG3, PDCD1, PDCD1LG2, TIGIT, and SIGLEC15) was compared between patients with low and high RBM15B expression. The results showed that patients with low RBM15B expression expressed more immune checkpoints (except for CD274 and PDCD1LG2) ( ). Furthermore, RBM15B was significantly negatively correlated with the expression of most of immune checkpoints (except for CD274 and PDCD1LG2) ( ).

Construction of networks of LncRNA-miRNA-mRNA

ENCORI and OncomiR databases were used to filter out 51 and 199 candidate miRNAs, respectively, to further investigate the potential functions of RBM15B. These candidate target miRNAs were then intersected to yield seven miRNAs ( ), of which hsa-let-7b-5p, hsa-miR-199a-5p, hsa-miR-10b-5p, hsa-miR-346, hsa-miR-532-5p, and hsa-miR-103a-3p were identified as potential RBM15B targets; this was based on their prognostic values and correlations with RBM15B. All these six miRNAs were strongly linked with RBM15B expression. Also, increased expression of these six miRNAs had an adverse prognostic value for UM patients ( ). The ENCORI database was used to identify potential lncRNAs, and the results revealed LINC00665 and LINC00638 as the potential target based on their expression levels and associations with hsa-let-7b-5p, hsa-miR-103a-3p, and RBM15B ( ). In this view, the LINC00665/hsa-let-7b-5p/RBM15B axis and LINC00638/hsa-miR-103a-3p/RBM15B axis were constructed, which are potential prognostic biomarkers in UM.

Figure 6

competing endogenous RNA network construction. (A) 7 potential target miRNAs were obtained from ENCORI and OncomiR databases. (B-H) Kaplan-Meier plots revealed the overall survival of 7 potential target miRNAs, including (B) hsa-let-7b-5p, (C) hsa-miR-199a-5p, (D) hsa-miR-10b-5p, (E) hsa-miR-346, (F) hsa-miR-532-5p, (G) hsa-miR-103a-3p and (H) hsa-miR-17-5p. (I–N) among 7 potential target miRNAs, 6 miRNAs, (I) hsa-let-7b-5p, (J) hsa-miR-199a-5p, (K) hsa-miR-10b-5p, (L) hsa-miR-346, (M) hsa-miR-532-5p, and (N) hsa-miR-103a-3p significantly correlated the expression of RBM15B. Kaplan-Meier plots revealed the overall survival of 2 potential target lncRNAs, (O) LINC00665 and (P) LINC00638, predicted by ENCORI. 2 potential target lncRNAs, (Q) LINC00665 and (R) LINC00638, significantly correlated the expression of RBM15B.

Discussion

Uveal melanoma is a cancerous condition that poses a threat to people’s health, and UM is different from cutaneous melanoma in terms of genetic background and clinical behavior (34). Therefore, finding effective treatment approaches for this cancer is critical. Subgroup classification is crucial in uveal melanoma because it dictates the specificity of treatment and prognosis (35, 36). The present investigation using m6A-related regulators classified TCGA patients into 4 groups and narrowed them down to the relevant subgroups for different analysis. The analysis revealed cross-talk among 20 m6A regulators in UM patients, suggesting that m6A may play important roles in the course and prognosis of UM cases. Consensus clustering analysis suggested that different expression levels of 20 m6A regulators resulted in different survival prognoses among four UM groups, indicating that the 20 m6A regulators have a potentially significant impact on the prognosis of UM patients. Moreover, two algorithms (QUANTISEQ and MCPCOUNTER) revealed multiple statistically significant changes in immune cell populations across the four UM groups. Intriguingly, CD8(+) T cells, myeloid dendritic cells, and macrophages were the most statistically significant infiltrating cells. we found that different m6A regulator expressions in these four subgroups contributed to different prognosis, but these four subgroups also had different immune cell infiltrations. Thus, we had the hypothesis that m6A regulators may correlate with immune cells to influence patients’ prognosis.

However, it was unclear which of the m6A regulators might impact the survival prognosis of UM patients. RBM15B, IGF2BP2, IGF2BP1, YTHDF3, and YTHDF1 with significantly results were identified as candidate genes in the GEPIA2 database. Meanwhile, univariate and multivariate cox regression analyses revealed that 6 regulators (including RBM15B, IGF2BP1, IGF2BP2, IGF2BP3, YTHDF3, and YTHDF1) were associated with UM prognosis, and RBM15B was the only independent prognostic factor for OS. These results were quite encouraging; therefore, we focused on the molecular characterization and possible clinical applications of the RBM15B gene.

RBM15B is one of the most essential N6-Methyladenosine methyltransferases. Emerging evidence implicates RBM15B plays an important part in carcinoma growth and metastasis in several cancers (37, 38). Zhang and colleagues, for example, demonstrated that RBM15B was highly expressed in ovarian cancer and increased expression of RBM15B correlated with worse PFS and ovarian cancer cell metastasis (37). In addition, another study showed that RBM15B was highly expressed in hepatocellular carcinoma, and it promoted hepatocellular carcinoma cell growth, invasion and metastasis in vivo and in vitro, thus resulting in a poor prognosis (38). In our study, Logistic regression analysis revealed a significant correlation of RBM15B expression with clinical stage and histological type. Moreover, stage 4 of uveal melanoma had a markedly lower expression level compared with stage 3, indicating that RBM15B inhibit tumor growth and progression. Our present investigation also revealed that increased RBM15B levels were strongly linked with improved OS, DSS, and PFI in UM patients. Therefore, we confirmed that RBM15B inhibit UM growth and progression. We also speculate that RBM15B inhibits the metastasis of UM patients.

Furthermore, multivariate Cox regression revealed that RBM15B was the only independent prognostic factor for UM patients. But what exactly is the mechanism? Interestingly, GO and KEGG enrichment analysis confirmed that RBM15B expression was primarily associated with immune-related terms (including T cell activation, T cell receptor complex, and humoral immune response) and pathways (including T cell receptor signaling pathway), implying that RBM15B may influence the survival prognosis of UM patients by regulating specific immune-related pathways. In this view, we looked at the relationships between RBM15B expression and several immune checkpoint molecules. Immune checkpoints are expressed on a wide range of immune cells, which suppresses the immune function. When immune cells fail to mount an effective anti-tumor immune response, tumor immune escape causes tumor progression and distant metastasis. Our study revealed that RBM15B expression was significantly negatively correlated with 6 immune checkpoints (including CTLA4, HAVCR2, LAG3, PDCD1, TIGIT, and SIGLEC15). When RBM15B is highly expressed, 6 immune checkpoints are low expressed. Thus, we believe that there is a reciprocal association between RBM15B and 6 immune checkpoints expressions. RBM15B may affect the expressions of 6 immune checkpoints, or 6 immune checkpoints may affect the expressions of RBM15B, or both RBM15B and 6 immune checkpoints are regulated by a third unknown factor in organism. The question which is the right one needs verification in our next work. Briefly speaking, RBM15B expression was negatively correlated with 6 immune checkpoints, thus influencing the prognosis of UM patients.

The ceRNA regulation of RBM15B, including long noncoding RNAs (lncRNAs) and microRNAs, was also investigated in this work. It has been reported mRNAs and lncRNAs “talk” to each other using microRNA response elements (MREs) as letters of a new language. MicroRNAs can decrease the stability of target RNAs and prohibit their translation. Seven predicted programs (including microT, miRanda, miRmap, PITA, RNA22, PicTar, TargetScan) were used to predict miRNA-RBM15B interactions to identify the upstream miRNAs of RBM15B. Meanwhile, the OncomiR database was used to extract a list of miRNAs that are strongly linked to UM patient survival. The results showed 7 miRNAs that functioned as oncogenic miRNAs and showed markedly negative correlations with RBM15B. has-let-7b-5p, for example, exhibited a negative correlation with RBM15B (r=-0.309, p=5.26e-03) and had an unfavorable overall survival in UM (p=0.0064). Similarly, according to the ceRNA theory, we discovered 2 potential lncRNAs, LINC00665 and LINC00638 (39). Survival analysis and correlation analysis demonstrated that LINC00665/hsa-let-7b-5p/RBM15B and LINC00638/hsa-miR-103a-3p/RBM15B axes are prognostic biomarkers in UM.

Some limitations should be noted when drawing conclusions from our study. Firstly, since our results are generated from bioinformatics analysis, one severe limitation is that no validation experiments on human samples were performed. Secondly, the data used in this study was downloaded from TCGA and other online databases, more clinical data and experiments are needed to further confirm the prognostic value of RBM15B in UM.

In conclusion, this study investigated 20 m6A-related regulators, performed m6A regulator consensus clustering, and found that RBM15B was the only independent prognostic factor for UM. Two upstream ceRNA regulation mechanisms of RBM15B were also identified in UM. In addition, it also showed that RBM15B positively influenced the survival prognosis of UM patients by decreasing the expression of immune checkpoints ( ). However, further in-depth molecular research and large clinical trials would be necessary for the future to verify these findings.

Figure 7

The model of regulatory mechanism of RBM15B in carcinogenesis of UM.

Data availability statement

The original contributions presented in the study are included in the article/ . Further inquiries can be directed to the corresponding author.

Ethics statement

This study was exempted from approval by the institutional ethics committee of Shanghai Tenth People’s Hospital affiliated to Tongji University School of Medicine, China, and the need for informed consent was also waived for its data were obtained from publicly available databases that have approvals.

Author contributions

TW: Writing- Original draft preparation, Conceptualization, Methodology, Data curation, Validation, Formal analysis. JB: Visualization, Investigation, Formal analysis, Resources. YZ and YX: Conceptualization, Methodology. Software, Validation. QP: Supervision, funding acquisition, Project administration. All authors contributed to the article and approved the submitted version.

Fundings

This work was supported by the National Natural Science Foundation of China [grant number 81470029]; the Shanghai Municipal Health Bureau [grant number ZY (2018-2020)-ZWB-1001-CPJS10]; and the Three-Year Action Plan for Promoting Clinical Skills and Clinical Innovation in Municipal Hospitals [grant number SHDC2020CR5014].

Acknowledgments

The authors thank the all the contributors of the TCGA (https://tcga-data.nci.nih.gov/) database and the other databases and concerned authors for sharing their data on open access. The authors also thank Figdraw (www.Figdraw.com) website.

Conflict of interest

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.