Identification of prognostic m⁶A modification patterns and score system in melanoma patients

Abstract

N6-methyladenosine (m⁶A) is the most common modification on RNAs and LncRNAs. It plays an important role in cancer stem cell differentiation, T cell differentiation, and immune homeostasis. In this study, we explored the potential roles of m⁶A modification of RNA in melanoma and investigated the immune cell infiltration in tumor microenvironment in diverse m⁶Aclusters and different m⁶Ascore groups. A consensus clustering algorithm determined m⁶A modification patterns based on 14 m⁶A regulators, and further explored the biological functions and the connection with TME. An m⁶A-related gene signature (m⁶Ascore) was constructed based on m⁶A-related genes using principal component analysis. Three m⁶A modification patterns were identified based on 14 m⁶A regulators, named as m⁶Aclusters A-C. The prognosis of m⁶Acluster A was more favorable than m⁶Aclusters B and C, and it was more closely associated with immune regulation. To quantify the m⁶A modification patterns of individual tumor, an m⁶Ascore was constructed, and patients were classified into high and low m⁶Ascore groups. The low m⁶Ascore group, which had a favorable prognosis, was more relevant to immunology. The expression of PD-L1 was higher and the immunophenoscore (IPS) revealed stronger response to immunotherapy in the low m⁶Ascore group. This study identified 3 m⁶A modification patterns with different immune characteristics and constructed an m⁶Ascore system to predict prognosis and immunogenicity of patients, which is conducive to clinical prognosis judgment and individual treatment.

1. Introduction

Melanoma is a highly malignant cancer originating from melanocytes.^[1] Globally, there are about 232,100 new cases of cutaneous melanoma, and almost 55,500 deaths annually account for 0.7% of all cancer deaths.^[2] Most patients newly diagnosed at an early stage are suitable for surgical resection and have a favorable prognosis. The prognosis of advanced patients with metastatic disease is unsatisfactory under the conventional treatment, including surgical resection, chemotherapy, and targeted therapies.^[3] Previous studies demonstrated that melanoma is one of the most immunogenic tumors and has the greatest potential to respond to immunotherapy.^[4] Immune checkpoint inhibitors, including antiPD-1 and antiCTLA4 antibodies, have overturned the traditional treatment, and nearly 50% of patients are likely to experience tumor regression and long-lasting disease control, compared to <10% previously.^[5] To better improve the clinical application of immunotherapy, we need to understand more about the relationship between the individual tumor microenvironment (TME) and immunogenicity.

N6-methyladenosine (m⁶A) is the most common type of RNA methylation modifications. At present, m⁶A modification is widely detected in mRNA, lncRNA, and miRNA.^[6,7] Previous studies indicated that m⁶A modification is involved in a variety of biological processes, such as obesity, developmental defects, immunoregulation, and carcinogenesis.^[8,9] In immune regulation, m⁶A modification participates in immune recognition, innate immune response to viral infection, activation of adaptive immune response, and immune cell differentiation.^[10] Moreover, m⁶A plays a key role in TME, including the development of immune cells and stromal cells. Simultaneously, m⁶A is also regulated by various factors in TME, such as hypoxia and cellular stress.^[11] Therefore, m⁶A and TME mutually interact in tumor occurrence and progression.^[12] The dynamic function of m⁶A modification is mainly regulated by the modification proteins: methyltransferases (“writers”), demethylases (“erasers”), and binding proteins (“readers”).^[13] These regulators play indispensable roles in m⁶A modification, including RNA metabolism, processing, export, stability, and translation.^[13,14]

Recently, numerous studies reported that most of these m⁶A regulators are frequently overexpressed in various cancer tissues, and some regulators promote tumor progression through demethylation.^[15–17] For example, methyltransferase-like 3 (METTL3) was upregulated in human melanoma tissues, and overexpression of METTL3 increased m⁶A activity, migration, and invasion of melanoma cells.^[18] Another study reported that fat mass and obesity-associated protein (FTO) could reduce the response to antiPD-1 inhibitor immunotherapy in melanoma patients.^[19] Here, we further explored m⁶A regulators to help identify novel therapeutic biomarkers and targets to develop effective treatment strategies for melanoma.

2. Materials and methods

2.1. Datasets of skin cutaneous melanoma

We downloaded the gene expression data and corresponding clinical annotation from publicly available datasets of the Cancer Genome Atlas (TCGA), Genomic Data Commons (https://portal.gdc.cancer.gov/) and Gene-Expression Omnibus database (GSE65904). Samples without survival information were eliminated from further analysis. Finally, a total of 668 patients were enrolled in our research, including 458 patients in TCGA and 210 patients in gene-expression omnibus, and each sample corresponded to 1 patient. Data collection was performed in August 2021.

2.2. Consensus cluster of 14 m⁶A regulators

We sorted out 19 m⁶A regulators for our study according to the previous research.^[16,20–22] These 19 m⁶A regulators included 7 writers (METTL3, METTL14, METTL16, WTAP, ZC3H13, RBM15, RBM15B), 10 readers (YTHDC1, YTHDC2, YTHDF1, YTHDF2, YTHDF3, HNRNPC, LRPPRC, HNRNPA2B1, RBMX, ELAVL1), and 2 erasers (FTO, ALKBH5). The RCircos R package was used to show the copy number variation (CNV) location of 19 m⁶A regulators on 23 pairs of chromosomes.^[23] Then Kaplan–Meier analysis evaluated the relationship between the prognosis of melanoma and 19 m⁶A regulators, which found 14 m⁶A regulators were associated with the prognosis of melanoma. Based on the expression of 14 m⁶A regulators, we performed unsupervised clustering analysis to identify distinct m⁶A modification patterns. The above steps were run by the Consensus Cluster Plus package and were repeated 500 times at least to ensure the stability of classification. The number of clusters was determined by the dispersion, cophenetic, and silhouette coefficients.

2.3. Gene set variation analysis (GSVA)

GSVA enrichment analysis with the “GSVA” R package was performed to identify the difference on biological processes and pathways among the m⁶A modification patterns. The gene sets of “c2.cp.kegg.v7.4.symbols” downloaded from MSigDB database was used for running GSVA analysis. Gene ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses for m⁶A phenotype-related genes were performed in clusterProfiler R package with the cutoff value of false discovery rate (FDR) < 0.05.

2.4. ssGSEA on immune cell infiltration estimation

Single sample gene set enrichment analysis (ssGSEA) was carried out to quantify the relative abundance of each immune cell infiltration in TME of 3 modification patterns and m⁶Ascore groups. The enrichment scores were calculated by ssGSEA analysis and represented the relative abundance of each immune cell in TME in each sample.

2.5. Identification of DEGs between m⁶A distinct patterns

We identified m⁶A-related differentially expressed genes (DEGs) among the 3 m⁶A modification patterns with “limma” R package. An adjusted P value <.001 (P < .001) was set for the significance filtering criterion of DEGs.

2.6. Generation of m⁶Ascore

We constructed a scoring system to quantify the m⁶A modification patterns of individual tumor – an m⁶A gene signature, which was called the m⁶Ascore. The steps for establishing m⁶A gene signature were as follows:

We extracted the overlapping DEGs identified from 3 m⁶Aclusters. A univariate cox regression model was used to identify genes associated with the prognosis of melanoma among overlapping DEGs. The genes associated with significant prognosis were screened out for further analysis. Then based on the expression data of prognostic genes, patients were classified into several groups through unsupervised clustering. The consensus clustering algorithm determined the number (n = 3) and the stability of gene clusters. According to the expression of prognostic genes, we constructed m⁶A gene signature using principal component analysis; principal component 1 and 2 were both selected as signature scores. We then figured out a formula according to previous research: m⁶Ascore= ∑(PC1i + PC2i),^[24,25] where “ i ” represents the expression of significant m⁶A phenotype-related genes. Based on the median score, we divided the patients into high and low m⁶Ascore groups.

2.7. Correlation between m⁶Ascore and immunity

To identify the association between the m⁶Ascore and TME, we performed a correlation analysis between m⁶Ascore and 23 types of immune cell infiltration. The immunophenoscore (IPS), a quantification of determinants of tumor immunogenicity, is a superior predictor of response to the immune checkpoint blockade in melanoma.^[26] The scoring scheme consists of 4 categories of immune-related genes: MHC molecules, immunomodulators, effector cells, and suppressor cells. The weighted average Z-score is calculated from the average z-score of the samples of the 4 categories within the category, and the sum of the weighted average Z-score is IPS.

2.8. Statistical analysis

The Kaplan–Meier (KM) method was applied to draw the survival curves for the prognostic analysis, and log-rank tests were applied to ascertain significance of differences. A univariate Cox regression model was utilized to calculate the hazard ratios for m⁶A regulators and m⁶A phenotype-related genes. The multivariate Cox regression model was used to determine the independent prognostic factors. The mutation prospect of high and low m⁶Ascore in patients was presented by the maftools package with waterfall function. The statistical analyses in this study were performed by R software (R version 4.1.0). All 2-sided P values <.05 (P < .05) were considered statistically significant.

3. Results

3.1. Characteristics of melanoma patients

The TCGA cohort with 470 samples and the GSE65904 cohort with 214 samples were downloaded for our study. Samples without survival information were eliminated from further analysis. In total, 668 patients with melanoma were enrolled in our study, including 458 patients from TCGA and 210 patients from the GSE65904 dataset, with an average age of 59.5 years (15–91 years). The clinicopathological characteristics of these patients are shown in Table 1 and 2. The design process of our study is shown in the flow chart (Fig. 1).

Table 1.

Characteristics of 458 patients with melanoma in TCGA.

Characteristics	N(%)
Age
≤65	296 (64.6)
>65	162 (35.4)
Gender
Male	284 (62.0)
Female	174 (38.0)
Stage
I/II NOS	10 (2.2)
Stage 0	6 (1.3)
Stage I	76 (16.6)
Stage II	139 (30.3)
Stage III	169 (36.9)
Stage IV	22 (4.8)
Unknown	36 (7.9)
T
T0	23 (5.0)
T1	41 (9.0)
T2	76 (16.6)
T3	89 (19.4)
T4	151 (33.0)
Tis	7 (1.5)
TX	44 (9.6)
Unknown	27 (5.9)
M
M0	409 (89.3)
M1	23 (5.0)
Unknown	26 (5.7)
N
N0	228 (49.8)
N1	73 (15.9)
N2	49 (10.7)
N3	54 (11.8)
NX	35 (7.6)
Unknown	19 (4.1)

TCGA = the cancer genome atlas.

Table 2.

Characteristics of 210 patients with melanoma in GSE65904.

Characteristics	N(%)
Age
≤65	116 (55.2)
>65	93 (44.3)
Unknown	1 (0.5)
Gender
Male	124 (59.0)
Female	86 (41.0)
Stage
Stage I–II	49 (23.3)
Stage III–IV	154 (73.3)
Unknown	7 (3.3)

Figure 1.

Workflow of this study.

3.2. Landscape of genetic variation of m⁶A regulators in melanoma

In this study, a total of 19 acknowledged m⁶A regulators including 7 writers, 10 readers and 2 erasers from the previous literature review^[16,20–22] were enrolled for analyses. We first made a summary of the somatic mutations and copy number variations of 19 m⁶A regulators in melanoma. Mutations occurred in 92 of the 467 samples from the TCGA melanoma mutation database. The highest mutation frequency was 3%, including YTHDC1, ZC3H13, LRPPRC, YTHDC2 and YTHDF1; whereas FTO, HNRNPC, and ALKBH5 did not show any mutations in the melanoma samples (Fig. 2A). Frequency analysis of 19 m⁶A regulators presented a prevalent alteration in CNV. More than half of the alteration in m⁶A regulators showed deletion in copy number; especially RBM15, WTAP, FTO and ZC3H13 had a widespread CNV deletion, while YTHDF1 and YTHDF3 displayed prevalent CNV amplification (Fig. 2B). Location of the CNV alterations of m⁶A regulators on chromosomes is shown in Figure 2C. The KM curve revealed that 14 of the 19 m⁶A regulators were closely associated with prognosis of patients with melanoma (P < .05, Fig. 2D). Overexpression of METTL3, METTL14, RBMX, WTAP, YTHDC2, and YTHDF2 showed positive correlation with survival. In contrast, overexpression of ALKBH5, ELAVL1, HNRNPA2B1, LRPPRC, RBM15B, YTHDF1, YTHDF3, and ZC3H13 showed negative correlation with survival. These 14 regulators were selected for further study.

Figure 2.

Genetic variation of m⁶A regulators in melanoma and their effect on prognosis. (A) Summary of the somatic mutations and copy number variations of 19 m⁶A regulators in 467 melanoma patients (TCGA-SKCM mutation data). (B) The copy number variation (CNV) frequency of m⁶A regulators in the TCGA cohort. Green dots: deletion frequency. Red dots: amplification frequency. (C) The location of CNV alterations of m⁶A regulators on 23 chromosomes from TCGA cohort. (D) The KM-curve of 14 m⁶A regulators with significant prognosis in 668 melanoma patients (TCGA and GSE65904 cohort). KM = Kaplan-Meier, m6A = N6-methyladenosine, TCGA = the cancer genome atlas.

3.3. Generation of m⁶A modification patterns

A comprehensive landscape of the interaction and prognostic significance of 14 m⁶A regulators in melanoma patients is illustrated in the m⁶A regulator network (Fig. 3A). Based on the expression of 14 m⁶A regulators, patients were classified into 3 modification patterns by unsupervised clustering (Supplemental Digital Content (Figure S1, http://links.lww.com/MD/M303), A-D), including 298 cases in cluster A, 295 cases in cluster B, and 75 cases in cluster C (Fig. 3B). The principal component analysis diagram displayed significant differences in the expression profile of m⁶A regulators within the 3 modification patterns (Fig. 3C). Compared with cluster B and C, cluster A had a significant survival advantage (Fig. 3D). Additionally, GSVA showed that immune-related pathways such as cytokine-cytokine receptor interaction, autoimmune thyroid disease, graft-versus-host disease, asthma, and systemic lupus erythematosus were mainly enriched in m⁶Acluster A (Fig. 4A and C). While m⁶Acluster B was associated with pathways of ubiquitin mediated proteolysis, cell cycle and RNA degradation (Fig. 4A and B). And m⁶Acluster C was related to phenylalanine, drugs, retinol, tyrosine and some other metabolic pathways (Fig. 4B and C). Subsequently, the analysis of TILs (tumor infiltrating lymphocytes) showed that cluster-A had abundant immune cell infiltration, including B, CD4 T, dendritic, gamma delta T, and natural killer cells (Fig. 4D).

Figure 3.

Three m⁶A methylation modification patterns based on 14 m⁶A regulators. (A) The interaction between 14 m⁶A regulators in melanoma. The calculated values of log-rank test were P < .0001, P < .001, P < .01, P < .05, and P < 1, respectively. Green dots: favorable factors. Purple dots: risk factors. (B) Unsupervised clustering of 14 m⁶A regulators in the melanoma cohorts, named as m⁶Aclusters (A–C) (TCGA and GSE65904). (C) Principal component analysis performed on the transcriptome profiles of 3 m⁶A modification patterns. (D) Survival analyses for the 3 m⁶A modification patterns based on 668 patients with melanoma, including 298 samples in m⁶Acluster-A, 295 samples in m⁶Acluster-B, and 75 samples in m⁶Acluster-C (P = .004). m⁶A = N6-methyladenosine, TCGA = the cancer genome atlas.

Figure 4.

Characteristics of 3 m⁶A modification patterns. (A–C) Biological functions and pathways in 3 m⁶A modification patterns shown by GSVA enrichment analysis. These biological processes were visualized by heat map, with red representing activated pathways and blue representing inhibited pathways. The melanoma cohorts (TCGA and GSE65904) were used as the patients’ annotations. (A) m⁶Acluster A vs m⁶Acluster B; (B) m⁶Acluster B vs m⁶Acluster C; (C) m⁶Acluster A vs m⁶Acluster C. (D) The abundance of immune infiltrating cells in these 3 m⁶A modification patterns. Statistical P value was indicated by the asterisks (*P < .05, **P < .01, ***P < .001). m6A = N6-methyladenosine, TCGA = the cancer genome atlas.

3.4. Analysis of DEGs between different m⁶A modification patterns

To elucidate further the underlying biological behavior of each m⁶A modification pattern, we identified 4430 DEGs based on the m⁶A regulator expression (Fig. 5A). GO enrichment analysis illustrated that these DEGs were related to RNA splicing, RNA localization, and transcription coregulator activity (Fig. 5B). KEGG enrichment analysis displayed that these genes were associated with viral infection, mRNA surveillance pathway, the cell cycle, and chronic myeloid leukemia (Fig. 5C). We then screened out 845 genes associated with the prognosis of melanoma among overlapping DEGs for further analysis. Based on these overlapping prognostic genes, patients were classified into 3 m⁶A modification genomic phenotypes by unsupervised clustering (Supplemental Digital Content (Figure S1, http://links.lww.com/MD/M303), E-H), including 279 cases in gene cluster A, 317 cases in gene cluster B and 72 cases in gene cluster C (Fig. 5D). The heatmap displayed the characteristics of each patient in the m⁶A gene cluster (Fig. 5D). The expression of m⁶A regulators was significantly different in the 3 m⁶A gene clusters (Fig. 5E). The prognostic analysis of the 3 gene subtypes showed that m⁶A gene cluster A had a better survival advantage than m⁶A gene clusters B and C. (Fig. 5F).

Figure 5.

m⁶A modification pattern-related DEGs in melanoma. (A) A total of 4430 m⁶A modification pattern-related DEGs are shown by the Venn diagram. (B-C) GO and KEGG enrichment analysis show the functional annotation of m⁶A-related genes. (D) Unsupervised clustering of overlapping m⁶A-related genes classify patients into different genomic subtypes, named as m⁶A gene clusters (A-C) (E) The expression of 14 m⁶A regulators in 3 gene clusters. Statistical P value was indicated by the asterisks (*P < .05, **P < .01, ***P < .001). (F) Survival analyses for the 3 m⁶A modification genomic phenotypes based on 668 patients with melanoma, including 279 samples in gene cluster-A, 317 samples in gene cluster-B, and 72 samples in gene cluster-C (P < .001). DEGs = differentially expressed genes, GO = gene ontology, KEGG = Kyoto encyclopedia of genes and genomes, m6A = N6-methyladenosine.

3.5. Generation of m⁶Ascore

The above analyses indicated that m⁶A modification could play crucial part in immune regulation and prognosis of melanoma, but these results are based only on the patient population. To predict accurately the m⁶A modification pattern of individual patient with melanoma, we needed to consider individual heterogeneity and complexity. Therefore, based on the aforementioned prognostic m⁶A pattern-related DEGs, a scoring system termed the m⁶Ascore was constructed. The attribute changes of individual patient were visualized by an alluvial diagram (Fig. 6A). Compared with the high m⁶Ascore group, patients with a low m⁶Ascore were significantly associated with better survival (Fig. 6B). We further investigated the infiltration of immune cells in the 2 groups. The differential analysis of immune cell infiltration confirmed that the low m⁶Ascore group had abundant immune cell infiltration compared to the high m⁶Ascore group (Fig. 6C). Most of these 23 types of infiltrated immune cells significantly correlated with the m⁶Ascore, including B cells, CD8 T cells, natural killer cells, eosinophils, MDSCs (myeloid-derived suppressor cells), macrophages, and regulatory T cells (Fig. 6D).

Figure 6.

Construction of the m⁶Ascore. (A) Alluvial diagram shows the changes of m⁶Acluster, gene cluster, m⁶Ascore, and survival status. (B) Survival analyses for the low (383 samples) and high (285 samples) m⁶Ascore groups based on 668 patients with melanoma (P = .005). (C) The abundance of immune infiltrating cells in low and high m⁶Ascore groups. Statistical P value is indicated by the asterisks (*P < .05, **P < .01, ***P < .001). (D) Correlations between m⁶Ascore and 23 types of known immune infiltrating cell cells. Positive correlation is shown in red and negative correlation is shown in blue. m6A = N6-methyladenosine.

3.6. Immunotherapeutic characteristics of high- and low-m⁶Ascore groups

To explore the role of the m⁶Ascore in clinical applications, Kaplan–Meier analysis was performed to investigate the prognostic value of the m⁶Ascore in patients at different stages. Patients with a low m⁶Ascore significantly associated with favorable prognosis in both stage I to II and stage III to IV (Fig. 7A and B). Patients with high tumor mutation burden (TMB) are reported more likely to benefit from immune checkpoint inhibitors (ICIs) therapy.^[27] The distribution of somatic mutations illustrated that TMB was more extensive in the low m⁶Ascore group (94.76% vs 88.11%, Fig. 7C–D). Furthermore, survival analysis confirmed that the prognosis of patients with higher TMB was favorable, and the high-TMB + low-m⁶Ascore group was superior to others for survival (Fig. 7E–F). The expression of PD-L1 in the low m⁶Ascore group was higher than that of high m⁶Ascore group (Fig. 7G). We then investigated the association between the IPS and the low/high m⁶Ascore group. Based on the IPS, we observed the difference of immunotherapy scores between high and low m⁶Ascore groups. We found that in the CTLA4_negative + PD-1_positive and CTLA4_positive + PD-1_positive types, the IPS of the low m⁶Ascore group was significantly higher than that of high m⁶Ascore group (Fig. 7I and K). In the CTLA4_negative + PD-1_negative and CTLA4_positive + PD-1_negative types, the median IPS for these 2 groups was nearly matching (Fig. 7H and J).

Figure 7.

Immunotherapeutic characteristics of the m⁶Ascore. (A-B) Survival analyses for patients with low and high m⁶Ascore in different tumor stages (A) patients with I to II, P = .044; (B) patients with III-IV, P = .021). (C-D) The waterfall plot of tumor somatic mutation in patients with high m⁶Ascore (C) and low m⁶Ascore (D). (E) K-M curve of patients with high or low TMB (P < .001). (F) K-M curve of patients with high (or low) TMB and high (or low) m⁶Ascore (P < .001). (G) The expression of PD-L1 in low and high m⁶Ascore groups (P = .0011). (H-K) Immunophenoscore (IPS) comparison between low and high m⁶Ascore groups in melanoma patients in the CTLA4 negative/positive or PD-1 negative/positive types. CTLA4_positive represented antiCTLA4 therapy; PD1_positive represented antiPD-1/PD-L1 therapy. m6A = N6-methyladenosine.

4. Discussion

The TME plays an important role in the occurrence and development of tumorigenesis, and epigenetic modification is a crucial mechanism in tumorigenesis.^[12] The m⁶A modification, a common epigenetic modification in RNA methylation, plays a pivotal role in TME and m⁶A regulators are key characters in the biological function and processing of m⁶A methylation modification.^[28–31]

Here, we screened out 19 acknowledged m⁶A regulators and identified 14 regulators associated with prognosis of melanoma. We considered that the genetic variation of m⁶A regulators could affect their expression in melanoma, and suspected that disturbance in the expression of m⁶A regulators might function in the occurrence and progression of melanoma.

Based on 14 m⁶A regulators, 3 m⁶A methylation modification patterns were identified that correlated with different immune infiltration characteristics. The m⁶Acluster A was characterized by immune-related pathways, summarized as an immune-inflamed phenotype. We speculate that m⁶Acluster A is mainly related to immune activation. While m⁶Acluster B was not connected with immunity, the enriched pathways are often dysregulated in cancer. The m⁶Acluster C was characterized by noninflammatory pathways such as metabolism and neuroendocrine features, corresponding to an immune-desert phenotype. We further proved that the m⁶Acluster-A had more abundant immune cell infiltration than the other 2 clusters and correlated with a favorable prognosis. These 3 distinct m⁶A modification patterns, presenting significant differences in biological function, signaling pathways, and prognosis, also differed in immune cell infiltration and immunogenicity. In summary, we view the m⁶Acluster A as an inflamed tumor (hot tumor), while cluster B and C are noninflamed tumors (cold tumor). Generally, inflamed tumors are more sensitive to immune checkpoint inhibitors, and switching the tumor phenotype from cold to hot could enlarge the application of immunotherapies.^[32,33] Our results showed that these 3 m⁶A modification patterns were linked to distinct immune characteristics, and patients in m⁶Acluster-A with high immunogenicity might benefit more from immunotherapies.

Further, DEGs identified from the 3 patterns were enriched in m⁶A modification and immune-related pathways. Based on DEGs with prognostic value for melanoma, 3 genomic subtypes were identified, consistent with the results of m⁶A modification clustering and were termed as gene clusters A-C. These new genomic subtypes were likely connected to different immune characteristics in melanoma because the expression of the 14 m⁶A regulators in the 3 gene clusters were different, as well as the mRNA transcriptome in the 3 m⁶A modifications. We considered these DEGs as m⁶A signature genes. Furthermore, because of individual tumor heterogeneity and complexity, we constructed a score system to quantify different m⁶A modification patterns of individual tumor, named the m⁶Ascore. As a result, the m⁶Acluster A characterized as an immune-inflamed phenotype exhibited a lower m⁶Ascore, and the other clusters considered as a noninflamed phenotype showed a higher m⁶Ascore. This score system classified patients into 2 groups, and the prognosis of low m6Asocre group was favorable. Moreover, the same results presented in patients with different stage tumors, and patients with alive status showed a lower m⁶Ascore. These observations imply that the m⁶Ascore is a new marker that can predict the prognosis of melanoma.

Melanoma is associated with a tremendous number of somatic genetic alterations.^[34,35] Theoretically, the more mutations, the tumor is more likely to generate neoantigens and to have a better response to ICIs.^[27] The distribution of somatic mutations illustrated that TMB was more extensive in the low m⁶Ascore group. We deduced that the TMB of the low m⁶Ascore group was higher than that of the high m⁶Ascore group. Studies have shown that high TMB was more sensitive to immunotherapy, and the prognosis of patients with high TMB was favorable after receiving ICIs.^[36,37] The K-M curve in our study also showed that patients with high TMB had a better survival prognosis. Moreover, the combined m⁶Ascore and TMB survival analysis showed that the prognosis of high-TMB + low-m⁶Ascore group was more favorable than other groups. Based on these results, we conclude that patients with a low m⁶Ascore and high TMB would be better responders to immunotherapies; the survival prediction was more favorable as well.

Melanoma is an immunogenic tumor and is highly responsive to immunotherapy. However, the clinical application of PD-1/PD-L1 is limited because of individual heterogeneity and high cost of medical treatment. Thus, predictors of efficacy to PD-1/PD-L1 are vital for selection of subjects for immunotherapy, including PD-L1 expression, TMB, TILs, mismatch-repair (MMR) deficiency, and microsatellite instability (MSI).^[38] The expression of PD-L1 in the low m⁶Ascore group was higher than that of high score group. Further analyses revealed that the m⁶Ascore system showed a negative correlation with numerous immune cells, and the low m⁶Ascore group presented with an abundant enrichment in TILs, which implied that low m⁶Ascore group had a strong immune infiltration. Consideration of expression of PD-L1 and TILs in low-m⁶Ascore group, we speculated that m⁶Ascore could predict the efficacy of PD-1/PD-L1 inhibitors, and patients with melanoma in low m⁶Ascore group were probably the optimal candidates for immunotherapy.

The immunophenotypic score (IPS) is based on the expression of important components of tumor immunity, including MHC molecules, immunomodulatory molecules, effector cells and suppressor cells. We found that in the Cytotoxic T lymphocyte-associated protein-4 (CTLA4)_negative + PD-1_positive and CTLA4_positive + PD-1_positive types, the low m⁶Ascore group exhibited significant higher IPS. CTLA-4 and PD-1 are immune checkpoint inhibitors that have been identified as antibody immunotherapy targets for the treatment of melanoma, and the combination of CTLA4 and PD-1 was shown to significantly improve the survival of patients with metastatic melanoma.^[39–42] The analysis of combinations of antiPD-1 and antiCTLA-4 also revealed that the low m⁶Ascore group showed a higher positive response to antiPD-1 therapy and combination therapy of antiCTLA4 and antiPD-1. Our findings predicted that the m⁶Ascore correlates with the response to immunotherapy, and the low m⁶Ascore is more suitable for immunotherapy. Moreover, patients in the low m⁶Ascore group showed a higher positive response to antiPD-1 therapy and combination therapy of antiCTLA4 and antiPD-1.

Taken together, in this study, we determined 3 m⁶A modification patterns with different immune characteristics in melanoma patients. By considering the heterogeneity and complexity of individual tumor, we constructed a score system (m⁶Ascore) to quantify the m⁶A methylation modification, which could be an independent prognostic biomarker to predict patients’ survival. The m⁶Ascore can also be considered a novel biomarker that might be used an effective predictive strategy for immunotherapy.

This study also has some limitations. The m⁶A regulators involved in our study were the recognized from the previous literature, however, we might ignore some unknown m⁶A regulators. A set of recently discovered regulators must be integrated into the system to enhance the precision of the m⁶A patterns. Meanwhile, the interaction and cooperation of m⁶A regulators remain unknown, and the inside mechanism of m⁶A modification in regulating the immune infiltration of melanoma needs to be further explored. Additional research is needed to evaluate the effectiveness of m⁶Ascore in forecasting the reaction to ICIs in more extensive clinical trials, and to investigate the connections among these m⁶A regulators in immune response.

5. Conclusions

In conclusion, 3 m⁶A modification patterns with different immune characteristics were identified in melanoma and a score system (m⁶Ascore) was constructed to quantify the m⁶A methylation modification patterns, which correlate with clinical prognosis and individual treatment.

Author contributions

Data curation: Si Ouyang.

Formal analysis: Feixiang Wang, Peijie Chen, Kaixin Xiong, Yao Wang.

Funding acquisition: Yao Wang.

Investigation: Peijie Chen.

Methodology: Feixiang Wang, Peijie Chen, Kaixin Xiong, Zichuan Liu.

Writing – review & editing: Yao Wang.

Supplementary Material