Integrated re-analysis of bulk and single-cell transcriptomic data reveals limited ferroptosis and neuroinflammation signatures after sevoflurane exposure in the aged rat hippocampus and human prefrontal cortex cells

Abstract

Experimental studies have implicated ferroptosis in sevoflurane-associated neural injury, but the consistency of this signal across transcriptomic systems, species, and neural cell types remains uncertain. We therefore reassessed whether publicly available transcriptomic datasets support a coherent ferroptosis and neuroinflammation signature after sevoflurane exposure. We re-analysed two public transcriptomic datasets: GSE141242, comprising hippocampal tissue from aged rats exposed to 2.5% sevoflurane for 4 h (n = 3) or oxygen control (n = 3), and GSE196239, comprising human embryonic prefrontal cortex mixed-cell cultures exposed to sevoflurane, propofol, recovery conditions, or untreated control (one library per condition). After quality control, 67,272 single cells were retained. Bulk differential expression was assessed with an empirical-Bayes linear model, whereas the unreplicated single-cell conditions were summarized descriptively using key-gene expression, predefined gene-set scores, and observed cell-type composition. In addition, qRT-PCR was performed in C6 glioma cells and HA astrocytes assigned to Control, 3.3% sevoflurane for 6 h, H 2 O 2 (200 μM, 24 h), or Sevo + H 2 O 2 groups to compare ferroptosis-related gene regulation across the four conditions. In the aged-rat hippocampus, no probe set reached FDR < 0.05 (minimum FDR = 0.071), and the two groups did not separate clearly in principal-component space. In the human prefrontal cortex cultures, selected ferroptosis-related genes showed only small descriptive differences between the single control and sevoflurane libraries, and predefined gene-set summaries were likewise similar. Observed cell-type proportions differed by less than two percentage points, and no broad cell type showed a coherent sevoflurane-associated ferroptosis or inflammation pattern. The datasets were compared qualitatively rather than by formal cross-species correlation because of their marked biological and replication differences. In the four-group qRT-PCR experiment, sevoflurane and H 2 O 2 each reduced GPX4 and SLC7A11 and increased ACSL4 and PTGS2 in C6 and HA cells; the combined group showed the largest changes, and all treatment groups differed from Control (Dunnett-adjusted P < 0.05). Because the single-cell dataset contains only one library per condition, all between-condition single-cell comparisons are descriptive. The available public transcriptomic data do not provide robust evidence for a coordinated ferroptosis or neuroinflammation program attributable to sevoflurane across the examined systems. The four-group experiment showed that sevoflurane and H 2 O 2 each regulated ferroptosis-related genes, with the largest changes in the combined group, providing complementary evidence of pathway perturbation under both exposures. Although the single-agent groups clarify the direction of each exposure-associated change, qRT-PCR alone cannot establish ferroptotic cell death or a pharmacologic interaction. Replicated studies with larger biological sample sizes and orthogonal ferroptosis assays are still required to establish whether ferroptosis is a reproducible component of sevoflurane-associated neurotoxicity.