Introduction
Endothelial cell ion channels play critical roles in vascular mechanosensing, volume regulation, and the control of membrane potential and vascular tone. These functions are disrupted in heart failure, contributing to impaired cardiac perfusion and disease progression. Single-cell transcriptomics offers an unparalleled opportunity to characterise these changes at cell-type resolution, but individual studies are underpowered to identify consistent transcriptional signatures. Here we apply KOSMIC – an open-source platform we developed for standardised processing and patient-level meta-analysis of single-cell transcriptomic data – to identify robust ion channel transcriptional changes in cardiac endothelial cells across independent heart failure cohorts..

Methods
We systematically searched the Gene Expression Omnibus to identify single-cell transcriptomic datasets from failing and non-failing human hearts. Five single-nucleus RNA-sequencing studies met inclusion criteria, providing 241,043 endothelial cells from 156 donors (85 failing, 71 non-failing) after standardised quality control, clustering, and cell-type annotation using KOSMIC. Patient-level differential expression analysis of ion channel expression was performed on endothelial cells. To ensure reliable quantification, meta-analysis was restricted to ion channel genes detected in ≥5% of cells in all five cohorts, reducing a curated reference set of 303 ion channel genes to 46 high-confidence targets, which were then pooled using Restricted Maximum Likelihood (REML) random-effects models.

Results
Meta-analysis identified a consistent and selective pattern of ion channel dysregulation in endothelial cells of the failing heart. Of 46 high-confidence ion channel genes taken forward for analysis, 16 were significantly differentially expressed after correction for multiple testing (FDR<0.05). Importantly, 9 of these (56%), including the volume-regulated anion channel subunit LRRC8B and the polycystin PKD2, would have been missed by the analysis of any single study and were only detectable by pooling evidence across cohorts. At the family level, 8 of 35 ion channel families were significantly dysregulated (FDR<0.05). Chloride intracellular channels were the most robustly downregulated family (logFC −0.36, FDR=2×10⁻¹⁷; I²=0%), driven by CLIC5 (logFC −0.50, FDR=4×10⁻¹⁵, 5/5 studies concordant) and CLIC1/3/4. Aquaporins were also markedly reduced (logFC −0.54, FDR=5×10⁻¹²; I²=0%), with AQP3 showing the most consistent and pronounced downregulation of any individual gene (logFC −2.45, FDR=5×10⁻²⁸). In contrast, mechanosensitive channels were consistently upregulated: PIEZO channels (logFC +0.20, FDR=5×10⁻³, 5/5 studies directionally concordant) and calcium-activated chloride channels (Anoctamins, logFC +0.14, FDR=0.023), driven by ANO10 and ANO8.

Conclusions
Meta-analysis of single-nucleus RNA-sequencing data reveals a consistent and selective pattern of ion channel dysregulation in heart failure endothelium: mechanosensing machinery is upregulated while volume regulation, chloride transport, and KATP-mediated metabolic sensing are suppressed. These findings identify endothelial ion channel remodelling as a consistent feature of heart failure and highlight specific channel families as potential therapeutic targets in cardiovascular disease.