Single-nucleus transcriptomic analysis identifies a putative Purkinje fiber-associated remodeling program after myocardial infarction

Abstract

Purkinje fiber (PF) remodeling may contribute to post-infarction ventricular arrhythmia, but rare conduction-associated cardiomyocyte states in injured human myocardium remain incompletely characterized. We use TsPF to denote transitional PFs, a study-specific transitional Purkinje fiber subpopulation resolved by cardiomyocyte re-clustering and characterized by an intermediate transcriptional phenotype. In this study, transitional refers to the observed expression phenotype; temporal progression, lineage conversion, and anatomical Purkinje localization require separate evidence. We performed a secondary analysis of the human myocardial infarction atlas reported by Kuppe et al. The analyzed discovery matrix was derived from 29 single-nucleus RNA-sequencing libraries representing 28 tissue specimens from 20 individuals (4 donor controls and 16 individuals with myocardial infarction). After quality control and Seurat/Harmony integration, ventricular cardiomyocytes were re-clustered. PF/TsPF annotation, MiloR, high-dimensional weighted gene co-expression network analysis (hdWGCNA), scMetabolism, PF-anchored pseudotime, and CellChat were evaluated with an explicit separation between abundance testing, descriptive molecular patterns, and model-derived hypotheses. The quality-controlled matrix contained 172,522 nuclei and 28,122 genes. All MiloR neighborhoods had a spatial false discovery rate (SpatialFDR) of at least 0.05. A secondary tissue-sample-level TsPF: mature-PF ratio was lower in myocardial infarction than control in the displayed comparison (unadjusted Wilcoxon P  = 0.0345); several tissues came from the same donors, and mature-PF denotes the PF-labeled denominator. Using β = 5, hdWGCNA resolved nine colored modules; the selected blue module contained 1,050 genes, including NT5C, ITPA, GUK1, and ADSS1. TsPF-labeled nuclei showed higher nucleotide-metabolism scores (average log2 fold change = 0.9982; Bonferroni-type adjusted P  = 4.63 × 10 − 27; cell-level comparison). CellChat network summaries showed 1,191 versus 805 inferred interactions and aggregate strength 0.095 versus 0.027 in control versus myocardial infarction; these values are descriptive model outputs rather than donor-level differential tests. This secondary analysis prioritizes a putative PF-associated transcriptional remodeling program for follow-up. The abundance result remained negative after SpatialFDR correction, while the co-expression, metabolic, ordering, and communication analyses provide candidates for hypothesis-driven work. Establishing a distinct TsPF state and its functional relevance will require independent donor-aware, spatial/histological, molecular, and electrophysiological validation.