Comprehensive Analysis of m6<scp>A</scp>‐Related Programmed Cell Death Genes Unveils a Novel Prognostic Model for Lung Adenocarcinoma

Xiao Zhang, Yan‐Pei Cao, Jiatao Liu, Wei Wang, Qiuyue Yan, Zhibo WangJournal of Cellular and Molecular Medicine 2025 · DOI

TL;DR

This study introduces a novel 13-gene prognostic signature (mPCDS) for lung adenocarcinoma (LUAD) by integrating m6A RNA modification and programmed cell death (PCD) pathways. The mPCDS demonstrates superior predictive power for patient survival and treatment response compared to existing clinical markers, offering a new avenue for personalized LUAD management.

What Problem Does This Paper Address?

Lung adenocarcinoma (LUAD) remains a significant health challenge with a poor 5-year survival rate, largely due to the lack of effective prognostic biomarkers for risk stratification and personalized treatment. While advances in targeted therapies and immunotherapies have shown promise, accurately predicting patient outcomes and treatment response is still difficult. This paper addresses this gap by investigating the complex interplay between N6-methyladenosine (m6A) RNA modification and programmed cell death (PCD) – two critical cellular processes implicated in cancer progression – to develop a more robust prognostic model for LUAD.

Methods

The researchers employed a multi-pronged approach:

  • Data Integration: They combined bulk RNA sequencing data from The Cancer Genome Atlas (TCGA) with single-cell RNA sequencing (scRNA-seq) data from the Gene Expression Omnibus (GEO) to gain a comprehensive understanding of LUAD at different resolutions.
  • Gene Identification: They systematically identified m6A regulatory genes and programmed cell death (PCD) genes, then performed correlation analyses to pinpoint m6A-related PCD genes.
  • Machine Learning for Model Development: A key innovation was the use of machine learning algorithms to build a prognostic signature. They tested 101 different combinations of algorithms, including CoxBoost and SuperPC, to select the optimal model. This involved a two-step process: CoxBoost for initial gene selection and dimensionality reduction, followed by SuperPC for final model construction, aiming to balance predictive accuracy with model complexity and reduce overfitting.
  • Validation: The developed model (mPCDS) was rigorously validated across multiple independent LUAD datasets.
  • Functional Analysis: Beyond prognosis, they explored the mPCDS's association with genomic profiles, tumor microenvironment (TME), immune cell infiltration, immunotherapy response, and drug sensitivity.
  • Experimental Validation: The role of a key identified gene, RCN1, was experimentally validated using Western blot, immunohistochemistry, siRNA knockdown, and cell proliferation/migration assays.

Datasets

  • Training Data: TCGA (The Cancer Genome Atlas) dataset for bulk RNA sequencing and clinical data.
  • Validation Datasets: Multiple GEO (Gene Expression Omnibus) datasets, including GSE171145 (single-cell RNA sequencing), GSE11969, GSE13213, GSE2693924, GSE29016, GSE31210, and GSE72094.
  • Immunotherapy Response Data: Data from the OAK trial and GSE91061 (melanoma patients treated with anti-CTLA4 and anti-PD1 therapies).
  • Drug Sensitivity Data: CTRP2.0 and PRISM databases.
  • Clinical Samples: Six pairs of LUAD tumor and adjacent non-tumor tissue samples from the First Affiliated Hospital of Nanjing Medical University for experimental validation.

The paper does not introduce new benchmark datasets but leverages existing public repositories.

Key Insights and Ideas

  • The m6A-PCD Axis is Crucial in LUAD: The study strongly suggests that the interplay between m6A RNA modification and programmed cell death is a significant driver of LUAD progression and prognosis, a link that was previously underexplored.
  • Single-Cell Data Illuminates Cellular Heterogeneity: Integrating scRNA-seq data revealed that m6A-related PCD activity varies significantly across different cell types within LUAD tumors, with myeloid and circulating cells showing higher activity. This highlights the importance of considering cellular heterogeneity in understanding cancer biology.
  • Machine Learning for Robust Biomarker Discovery: The sophisticated machine learning approach, combining CoxBoost and SuperPC, effectively identified a concise yet powerful 13-gene signature (mPCDS) that outperforms traditional clinical factors and existing prognostic models.
  • mPCDS as a Multifaceted Prognostic Tool: The mPCDS not only predicts survival but also offers insights into the tumor microenvironment, immune response, and potential drug sensitivities, making it a comprehensive tool for personalized medicine.
  • RCN1 as a Key Oncogene: The experimental validation of RCN1 as a driver of LUAD proliferation, migration, and invasion positions it as a promising therapeutic target.

Key Results

  • Identification of 43 Prognostic Genes: The study identified 43 m6A-related PCD genes with prognostic value.
  • Development of the 13-Gene mPCDS: A 13-gene risk signature (mPCDS) was developed using machine learning, demonstrating superior predictive performance.
  • Superior Prognostic Accuracy: The mPCDS consistently achieved higher C-index scores across multiple validation datasets compared to established clinical factors (age, gender, stage, etc.) and existing LUAD prognostic signatures.
  • TME and Immunotherapy Insights: The mPCDS was found to be significantly correlated with immune cell infiltration and predicted differential responses to immunotherapy, with lower mPCDS scores generally associated with better outcomes.
  • Drug Sensitivity Prediction: The mPCDS identified potential therapeutic drugs (e.g., paclitaxel, gemcitabine) that may be more effective for patients with high mPCDS scores.
  • RCN1 Validation: Experimental data confirmed that RCN1 is overexpressed in LUAD, promotes cell proliferation, migration, and invasion, and is associated with poorer prognosis.

Limitations and Future Work

  • Sample Size and Diversity: The authors acknowledge that the sample size of LUAD cases and the depth of follow-up clinical data are currently inadequate for truly reliable internal validation. They emphasize the need for additional, larger, and more diverse cohort studies to further assess the model's generalizability.
  • Mechanistic Understanding: The precise molecular mechanisms and biological functions of the 13 m6A-related PCD genes in LUAD are not yet fully clarified. Future research is needed to elucidate their roles in disease progression, immune evasion, and other oncogenic processes.
  • Experimental Validation of Drug Candidates: While potential therapeutic drugs were identified computationally, their efficacy requires experimental validation.
  • Clinical Translation: The authors suggest that future work should focus on developing user-friendly tools for clinicians to calculate mPCDS scores and conducting rigorous validation through multicenter randomized controlled trials to facilitate clinical application.
  • Pan-Cancer Applicability: While the mPCDS showed promise in a pan-cancer context, further investigation into its specific roles and mechanisms in various cancer types is warranted.

Summary updated

Abstract

Lung adenocarcinoma (LUAD) involves complex dysregulated cellular processes, including programmed cell death (PCD), influenced by N6-methyladenosine (m6A) RNA modification. This study integrates bulk RNA and single-cell sequencing data to identify 43 prognostically valuable m6A-related PCD genes, forming the basis of a 13-gene risk model (m6A-related PCD signature [mPCDS]) developed using machine-learning algorithms, including CoxBoost and SuperPC. The mPCDS demonstrated significant predictive performance across multiple validation datasets. In addition to its prognostic accuracy, mPCDS revealed distinct genomic profiles, pathway activations, associations with the tumour microenvironment and potential for predicting drug sensitivity. Experimental validation identified RCN1 as a potential oncogene driving LUAD progression and a promising therapeutic target. The mPCDS offers a new approach for LUAD risk stratification and personalised treatment strategies.
BibTeX
@article{zhang2025comprehensive,
  author = {Xiao Zhang and Yan‐Pei Cao and Jiatao Liu and Wei Wang and Qiuyue Yan and Zhibo Wang},
  title = {{Comprehensive Analysis of m6<scp>A</scp>‐Related Programmed Cell Death Genes Unveils a Novel Prognostic Model for Lung Adenocarcinoma}},
  journal = {Journal of Cellular and Molecular Medicine},
  year = {2025},
  doi = {10.1111/jcmm.70255},
}
Figure