文献类型：期刊文献

英文题名：Machine learning integrated bulk and single cell transcriptomic analyses reveal oxidative stress associated mitochondrial polarization signatures in diabetic foot ulcers

作者：Guo, Zeao[1];Zeng, Zhaoyang[1,2];Ma, Xuepeng[1];Jing, Shuai-Shuai[1];Wu, Yu[1];Shan, Bin[1];Pan, Haibang[2];Shi, Yuhong[2];Liu, Ai[1];Jiang, Hugang[1];Li, Ning[2]

第一作者：郭志

通信作者：Zeng, ZY[1];Li, N[2];Zeng, ZY[2]

机构：[1]Gansu Univ Tradit Chinese Med, Sch Integrated Tradit Chinese & Western Med, Lanzhou, Peoples R China;[2]Gansu Univ Tradit Chinese Med, Gansu Univ Chinese Med, Affiliated Hosp, Lanzhou, Peoples R China

第一机构：甘肃中医药大学

通信机构：[1]corresponding author), Gansu Univ Tradit Chinese Med, Sch Integrated Tradit Chinese & Western Med, Lanzhou, Peoples R China;[2]corresponding author), Gansu Univ Tradit Chinese Med, Gansu Univ Chinese Med, Affiliated Hosp, Lanzhou, Peoples R China.|[10735b845793de6ae2b30]甘肃中医药大学第二附属医院;[10735]甘肃中医药大学;

年份：2026

卷号：14

外文期刊名：PEERJ

收录：;Scopus(收录号：2-s2.0-105036167236);WOS:【SCI-EXPANDED(收录号:WOS:001739457300001)】；

基金：The authors sincerely thank all members involved in the creation of this study and express gratitude to all developers of the tools used in this research.

语种：英文

外文关键词：Mitochondrial polarization; Diabetic foot ulcers; Oxidative stress; Single-cell transcriptomics; Machine learning; Macrophage polarization

摘要：Diabetic foot ulcers (DFU) are a major complication of diabetes, and its pathogenesis remains incompletely elucidated. Converging evidence indicates that oxidative stress and dysregulated mitochondrial polarization participate in DFU progression, nominating these processes as therapeutically actionable targets. This study integrates bulk and single-cell transcriptomic data with machine learning to reconstruct cross-scale, cell type-resolved molecular atlases and regulatory networks. Macrophages and fibroblasts emerged as communication hubs, dominating pathway enrichment and ligand-receptor programs such as macrophage migration inhibitory factor signaling pathway (MIF), ANNEXIN signaling pathway, and COMPLEMENT signaling pathway. Peptidylprolyl isomerase F (PPIF), which encodes cyclophilin D (CypD) and apolipoprotein E (APOE) were further prioritized as putative drivers within macrophages and fibroblasts, and a five-gene classifier was derived with robust performance (internal/external AUC = 0.833/0.933). Within DFU lesions, under the control of non-coding RNA circuitry, SOX5 may shape the inflammatory microenvironment, APOE may participate in lipid-metabolic remodeling, and PPIF (CypD) likely links reactive oxygen species (ROS) accumulation to a p53-dependent mitochondrial death pathway (necroptosis/apoptosis). Orthogonal validation showed significantly increased CypD in diabetic foot ulcer skin (DFUS) and diabetic foot ulcer tendon (DFUT) relative to diabetic foot skin (DFS) and DFT (Diabetic foot tendon), with up-regulated p53 and Cytc and down-regulated ApoE in DFUS; in primary foot-skin fibroblasts, a high-glucose plus tert-butyl hydroperoxide (HG+TBHP) model reproduced elevated ROS, loss of mitochondrial Delta psi m (mitochondrial membrane potential), growth restriction, and apoptosis, supporting a ROS-CypD/mPTP (mitochondrial permeability transition pore)-Delta psi m depolarization-p53/Cytc apoptosis axis. The delineated PPIF-centered regulatory network includes upstream transcription factors CEBPB/REL/SPI1 and a downstream ceRNA axis comprising miR-128-3p/miR-23a-3p-long non-coding RNA OIP5-AS1. Additionally, the significant role of polarization-specific reprogramming in regulating macrophage function highlights therapeutic strategies focused on metabolic reprogramming and inhibition of the PPIF/mPTP pathway. Collectively, a cell type-resolved molecular map of DFU is provided, healing-relevant cell populations and regulatory circuits are prioritized, and a translational, testable intervention framework is proposed.