A mitochondria-related gene-based signature predicts pancreatic ductal adenocarcinoma clinical outcome and revealed CAMK2A/THEM4 regulates progression phenotypes and mitophagy in vivo and in vitro.
Gang Nie, Zhongfei Zhu, Jing Huang, Xiaohan Shi, Xiaoyi Yin, Gang Li, Gang Jin
Abstract
Open AccessBACKGROUND: Pancreatic ductal adenocarcinoma (PDAC) is a highly lethal malignancy with limited available prognostic tools. There is a need to develop robust molecular signatures to improve risk stratification and identify potential therapeutic targets. METHODS: A 12-gene mitochondrial-related prognostic signature was developed using machine learning. The performance was compared with signatures developed using other algorithms and published signatures. The detailed synergistic role and molecular mechanism of signature gene CAMK2A and THEM4 was investigated. Additionally, an immunohistochemical (IHC) microarray and in vivo xenograft model was performed to evaluate the prognostic utility of candidate genes. RESULTS: The prognostic signature effectively stratified patients into high- and low-risk groups with significantly different survival outcomes (median OS: 15.4 vs. 24.1 months, p < 0.0001). It demonstrated robust predictive accuracy across datasets and outperformed 101 other machine learning-based signatures as well as previously reported models. High-risk samples showed enrichment in cancer-related pathways and unique adaptive drug profiles. Single-cell sequencing data indicated that the signature reflects immune cell infiltration and cell-cell interaction heterogeneity. Functional experiments revealed that CAMK2A suppresses PDAC progression by impairing mitochondrial function-reducing membrane potential, ATP production, and ROS levels-and facilitating THEM4 release from mitochondria. THEM4, which is downregulated in PDAC, inhibited tumor growth by suppressing AKT phosphorylation (p < 0.01). THEM4 knockdown accelerated in vivo tumor growth. Clinically, combining THEM4 and CAMK2A expression improved prognostic performance compared to either biomarker alone. CONCLUSIONS: This study establishes a mitochondria-driven prognostic model for PDAC and identifies the CAMK2A-THEM4-AKT axis as a novel therapeutic target. To our knowledge, this is the first pancreatic cancer signature that investigates the synergistic effect of candidate genes, offering both prognostic and mechanistic insights into PDAC progression.