Single cell and bulk transcriptomics reveal metabolic reprogramming biomarker signatures during extracorporeal membrane oxygenation for cardiogenic shock.

Cardiogenic shock (CS) patients receiving extracorporeal membrane oxygenation (ECMO) exhibit profound immune and metabolic disturbances, which may influence early clinical outcomes. However, the transcriptomic features linking metabolic reprogramming, immune-cell alterations, and early outcomes after ECMO initiation remain incompletely understood. The GSE182600 dataset from the public database was analyzed to identify differentially expressed genes (DEGs) between ECMO-treated CS samples with successful and failed outcomes. Multiple machine learning algorithms and the Shapley Additive exPlanations (SHAP) framework were applied to identify biomarkers associated with metabolic reprogramming. Functional enrichment analysis, immune infiltration analysis, drug prediction analysis, and molecular docking analysis were conducted to explore regulatory mechanisms and potential compound targets. Single-cell RNA sequencing (scRNA-seq) analysis was further conducted to determine key cell types and expression dynamics of the biomarkers. Clinical validation was achieved through quantitative reverse transcription polymerase chain reaction (RT-qPCR) assays on patient-derived specimens. PLIN2, TKTL1, NR1H4, GPX3, and NNMT were identified and selected as potential biomarkers. These biomarkers were primarily involved in metabolic, inflammatory, and immune processes. Immunological analysis revealed higher infiltration levels of CD8⁺ naive T cells in the failure group, by contrast, osteoblasts and plasma cells exhibited higher infiltration levels in the success group. Doxorubicin hydrochloride showed strong binding affinity to GPX3 and NR1H4, which may link the compound's cardiotoxicity to cardiac injury. Additionally, scRNA-seq analysis indicated monocytes as the predominant cell type, with dynamic expression of NNMT, TKTL1, PLIN2, and GPX3 along monocyte differentiation trajectories. In the RT-qPCR assay, the expression levels of TKTL1, NR1H4, and NNMT were significantly reduced in the success group compared with the failure group (P < 0.05), whereas GPX3 and PLIN2 showed consistent downward trends but did not reach statistical significance. PLIN2, TKTL1, NR1H4, GPX3, and NNMT may reflect immunometabolic features associated with early outcomes after ECMO initiation in patients with CS, with RT-qPCR providing preliminary support for the differential expression of TKTL1, NR1H4, and NNMT.