Unveiling ac4C modification pattern: a prospective target for improving the response to immunotherapeutic strategies in melanoma

Abstract

Emerging evidence has confirmed the inextricable connection between N4-acetylcytidine (ac4C) mRNA modification and the clinical characteristics of malignancies. Nonetheless, it is uncertain whether and how ac4C mRNA modification patterns affect clinical outcomes in melanoma patients. This research integrated single-cell sequencing data and transcriptomics to pinpoint ac4C-related genes (acRG) linked to melanoma progression and evaluate their clinical implications. Cells with elevated acRG score were predominantly located within the melanocytes cluster. Intercellular communications between melanocytes and other cell subtypes were markedly strengthened in the acRG-high group. We developed and confirmed an excellent acRG-related signature (acRGS) utilizing a comprehensive set of 101 algorithm combinations derived from 10 machine learning algorithms. Hereby, the acRGS, including MYO10, ZNF667, MRAS, SCO2, MAPK10, PNMA6A, KPNA2, NT5DC2, BAIAP2L2 and NDST3, delineated ac4C-associated mRNA modification patterns in melanoma. The acRGS possesses distinctly superior performance to 120 previously reported signatures in melanoma and could predict the overall survival of melanoma patients across four external datasets. The substantial associations among immune checkpoint genes, immune cell infiltration, and tumor mutation burden with acRGS indicate that acRGS is helpful in identifying melanoma patients who are sensitive to immunotherapy. Besides, we confirmed that MYO10 was mainly overexpressed in melanoma tissues, and elevated MYO10 was positively correlated with malignant phenotypes and unfavorable prognosis in melanoma patients. Silencing MYO10 expression inhibited melanoma cell proliferation, migration and invasion in vitro as well as tumor growth in vivo. Taken together, the acRGS could function as a reliable and prospective tool to improve the clinical prognosis for melanoma individuals.

Keywords: Immunotherapy; MYO10; Machine learning; Melanoma; Tumor microenvironment; ac4C modification.

Fig. 1

The detailed research flowchart

Fig. 2

Cell type annotation based on distinct cell markers in melanoma samples. (A) A Uniform manifold approximation (UMAP) plot of acral melanoma and cutaneous melanoma is presented. (B) UMAP visualization of 23 clusters. (C) UMAP visualization of eight cell types, including 9,145 T cells, 6,567 NK cells, 37,879 melanocytes, 1,423 endothelial cells, 773 macrophages, 1,019 fibroblast cells, 950 smooth muscle cells and 1,426 plasma B cells. (D) The dot plot illustrates the average and percent expression levels of marker genes across various cell subtypes. (E) The line chart displays the distribution of each cell type throughout the samples. (F) GO and KEGG enrichment analysis of differentially expressed genes among various cell subtypes. (G-H) Cellchat analysis of all cell types. Both the numbers of interactions and their respective intensities were demonstrated

Fig. 3

Evaluation of acRG activity scores in different kinds of cells. (A) Bubble plots depict the activity score of acRG across eight cell types implementing AUCell, UCell, singscore, ssGSEA, AddModulescore, and Scoring (the sum of scores from other algorithms). (B) Violin plots depict the activity score of acRG across eight cell types implementing Scoring. (C) The bubble graphic demonstrates the cell cycle phases and expression of maker genes for eight cell types across two acRG groups. (D) The UMAP plot visualizes the acRG activity score in each cell type. (E) the bar graph exhibits cellular communication according to the number and strength of interactions. (F) Network diagrams demonstrate enhanced intercellular communication in the acRG-high group. (G) Comparison of signaling dynamics between acRG-high and acRG-low groups, proving more robust signaling in the acRG-high group. (H) Identification of the top 150 genes experiencing the strongest correlations with acRG score

Fig. 4

A prognostic acRGS was developed by applying machine learning algorithms. (A) Comparison of the number of signaling pathways between the acRG-high and acRG-low cohorts. (B) A total of 101 permutations of machine learning processes were employed to construct acRG-related signatures (acRGS), and the C-index for each model was determined across all validation datasets. The model established through the Lasso + StepCox(both) method is the most effective

Fig. 5

Assessing and contrasting the predictive value of acRGS in melanoma. (A–D) Kaplan–Meier curves indicated survival disparities between high- and low-acRGS groups across various cohorts, with the high-acRGS group exhibiting a considerably adverse prognosis. (E–H) Time-dependent ROC curves of the acRGS for predicting overall survival at 1, 3, and 5 years in the TCGA (E), GSE19234 (F), GSE53118 (G) and GSE54467 (H) datasets. (I–L) Principal component analysis (PCA) of melanoma patients according to the acRGS. (M–P) Comparison of the C-index between acRGS and 120 published signatures. acRGS indicated superior predictive performance across multiple datasets, achieving the highest C-index value

Fig. 6

The landscape of tumor microenvironment (TME) in different acRGS groups. (A) The heatmap explicitly illustrates the variations in immune cell infiltration within acRGS groups as assessed through seven predictive algorithms. The z-values represent the relative abundance of immune cell infiltration, clearly suggesting a heightened level of immunological infiltration in the low-acRGS group. (B) The heatmap reveals the varying degrees of connections between acRGS groups and immunomodulators. (C) The comparison of ICGs expression in the two acRGS groups indicates elevated levels of ICGs expression in the low-acRGS group. (D) The bubble plot shows the comparative expression levels of ICGs and acRGS modeling genes. (E) Correlations among acRGS score and stromal, immune, and ESTIMATE scores indicate a negative association with immune and stromal score while exhibiting a positive correlation with tumor purity. (F) The efficacy of acRGS in predicting the success of immunotherapy

Fig. 7

Genomic and transcriptomic characterization of acRGS in the TCGA cohort. (A) Analysis of chromosomal amplifications and deletions in high versus low acRGS groups utilizing GISTIC 2.0. (B) Differences in tumor mutation burden (TMB) and genomic alteration profiles between high-acRGS and low-acRGS groups. (C, D) There existed a notable disparity between the acRGS score and TMB. (E-F) Survival analysis across four subgroups delineated by acRGS and TMB revealed divergent prognostic outcomes. The prognosis was markedly poorer in the low-TMB group (E). The low-TMB + high-acRGS group exhibited the poorest prognosis, whereas the high-TMB + low-acRGS group demonstrated the most favorable prognosis (F)

Fig. 8

Biological characterization of acRGS in melanoma. (A) The correlation of acRGS with the steps of the cancer immunity cycle (right) and the expression of immune cells (left). (B) MsigDB-based GSVA analysis delineated the biological features of two acRGS groups. (C) t-SNE visualization of pathway activity disparity between acRGS groups, evaluated through GO and KEGG pathways

Fig. 9

Exploring the prognostic value of the critical gene MYO10 in melanoma. (A) The scatterplot illustrates the correlation between the acRGS score and MYO10 expression in melanoma (r = 0.52). (B) Assessment of MYO10 as a prognostic factor in melanoma. Patients exhibiting elevated MYO10 expression experienced negative clinical outcomes. (C) Univariate Cox regression analysis was conducted across multiple datasets. (D) The heatmap highlights the relationship between MYO10 expression and immune cell infiltration across various datasets. (E) Correlation between MYO10 expression levels and immunomodulators across several datasets. (F) Elevated MYO10 expression forecasts medication responsiveness in melanoma

Fig. 10

MYO10 downregulation inhibits melanoma progression both in vitro and in vivo. (A) The mRNA levels of MYO10 in ten pairs of melanoma and corresponding normal skin tissues are presented as (T/N). (B) QRT-PCR examined the efficacy of MYO10 inhibition. (C, D) CCK-8 assays demonstrated that MYO10 inhibition reduced the proliferative capacity of melanoma cells in vitro. (E–G) Colony formation assay exhibited a significant reduction in colonies number in the shMYO10 group. (H, I) Scratch-wound healing assay indicated significantly slower cell migration after MYO10 knockdown (scale bar, 200 μm). (J, K) Transwell assay revealed that downregulation of MYO10 expression impeded the migratory and invasive capacity of melanoma cells (scale bar, 200 μm). (L) Representative images of subcutaneous xenograft tumors (n = 5 per group). (M, N) The growth curves were established by measuring tumor volumes every five days, and the tumor weights were measured. Silencing of MYO10 expression markedly suppressed melanoma cell growth in nude mice. Note: ****P ≤ 0.0001.***P ≤ 0.001. **P ≤ 0.01