Normothermic human kidney preservation drives iron accumulation and ferroptosis

Abstract

Ex vivo normothermic machine perfusion has been proposed to protect deceased donor kidneys. However, its benefits remain ambiguous. We postulate that the use of red blood cells (RBCs) and associated secondary hemolysis may in fact cause renal injury, offsetting potential advantages. During 48-hour normothermic perfusion of seven human donor kidneys, we observed progressive hemolysis, leading to iron accumulation in perfusate and tissue. Untargeted lipidomic profiling revealed significant increases in oxidized phospholipid species in perfused kidneys, pointing towards iron-dependent cell death known as ferroptosis. Next, in twelve additional perfusions, we assessed strategies to mitigate hemolysis-driven injury. Dialysis-based free hemoglobin removal reduced lipid peroxidation, but a ferroptosis gene signature persisted. In contrast, cell-free perfusion at subnormothermia negated iron accumulation, the ferroptosis gene signature, phospholipid peroxidation, and acute kidney injury. Our findings highlight the pathological role of hemolysis and iron on the kidney, urging restraint in the clinical application of RBC-based kidney perfusion.

Fig. 1. Hemolysis induces iron accumulation during RBC-based human kidney perfusion.

A Schematic overview of the perfusion platform. Kidneys were perfused with an RBC-based perfusate using a pressure-controlled impeller pump set at 75 mmHg following oxygenation at normothermia. Urine was recirculated, and continuous hemofiltration allowed removal of small molecular weight waste products and substitution with fresh perfusate. B Renal blood flow during perfusion. C Perfusate free hemoglobin (fHb) during perfusion. D Association between perfusate fHb and urine fHb levels (Spearman’s rank correlation coefficient, two-sided. Shaded area indicates 95% CI). E Macroscopic appearance of a control kidney, Kidney1 with low fHb accumulation, and Kidney6 that failed after 34-h with significant fHb accumulation. F Immunohistochemistry for heme oxygenase-1 (HO1) in serial biopsies obtained during perfusion (n = 7). Representative images from Kidney3 are shown. Scale bar, 50 μm. G Experimental design comparing RBC-based perfusion with (n = 7) or without (n = 3) a deceased donor kidney present in the platform. H Perfusate fHb concentration during perfusion (Mean ± SD). I Perfusate iron concentration during perfusion (Mean ± SD. Two-way ANOVA). J Association between perfusate fHb and perfusate iron concentrations across both groups (Spearman’s rank correlation coefficient, two-sided. Shaded area indicates 95% CI). K Temporal changes in expression of genes involved in iron uptake, iron storage, iron export, iron scavenging, and redox defense during RBC-based human kidney perfusion. Log2 foldchange relative to T0 is indicated by color; significance (adjusted P ≤ 0.05) by dot size and border (two-sided Wald test with Benjamini-Hochberg (BH) correction). L Immunofluorescence for DMT1 (Fe-uptake), FTH1 (Fe-storage), and Ferroportin (Fe-export) in pre- and post-perfusion biopsies (n = 7). Representative images from Kidney3 are shown. Scale bar, 200 and 50 μm, respectively. M Enhanced Perl’s Prussian Blue (EPPB) staining for iron deposition in control (n = 4) and RBC-perfused kidneys (n = 7). Representative images and quantification (Mean ± SD. Unpaired two-sided t-test). Scale bar, 50 μm. A, G Illustrated by Manon Zuurmond (scientific illustrator). Source data are provided as a Source Data file.

Fig. 2. RBC-based human kidney perfusion induces ferroptosis.

A Experimental design comparing RBC-perfused kidneys (n = 7) to unperfused controls (n = 3) and RBC-only perfusions (n = 3). B Representation of 539 lipid species detected through untargeted LC-MS/MS lipidomics, grouped by lipid (sub-)class (also see Table S4). Heatmaps showing relative lipid abundance for individual lipid species (C) and grouped into 11 lipid (sub-)classes (D) in control and RBC-perfused kidneys (n = 3 and n = 7, respectively). Z-scores represent total area normalized abundance. TG Triradylglycerols, DG Diradylglycerols, FA Fatty acids, FE Fatty esters, PL Phospholipids, LPL Lyso-phospholipids, EtherPL Ether-linked phospholipids, oxPL Oxidized phospholipids, OtherPL Other phospholipids (BMP (4), CL (8), and MLCL (1)), SP Sphingolipids, ST Sterol lipids. E Volcano plot representing the 539 lipid species comparing RBC-perfused with control kidneys (unpaired two-sided t-test with Welch’s correction). OxPL species are highlighted in red. Significantly increased lipid species (Log2FC ≥ 1.5 and P-value < 0.05) are detailed in Table S5. Abundance of oxidized PCs/non-oxidized PCs (F) and oxidized phosphatidylinositols (PIs)/ non-oxidized PIs (G) (Mean ± SD; n = 3 and n = 7, respectively. Unpaired two-sided t-test). H Levels of arachidonic acid (AA)-derived lipid mediators in perfusate from RBC-perfused kidneys (n = 7) and RBC-only perfusions (n = 3) over time (Median with IQR). Box plot of gene set variation analysis (GSVA) enrichment score of WP_Ferroptosis (I, n = 64 genes) and FerrDb v2 driver, suppressor and marker genes ( J, n = 313 genes) (Mixed-effects ANOVA). Boxes represent median with 25th and 75th percentile and error bars minimum to maximum. K Volcano plots of differentially expressed genes over time during RBC-based kidney perfusion. Black highlights FerrDb v2 marker genes (n = 9). Red highlights validated FerrDb v2 driver or suppressor genes with a protein product (n = 312). L Dotplot illustrating temporal expression dynamics in FerrDb v2 driver, suppressor, and marker genes (|Log2FC| ≥ 2 and adjusted P ≤ 0.05 for at least one of the timepoints). Two-sided Wald test with BH correction (K, L). A Illustrated by Manon Zuurmond (scientific illustrator). Source data are provided as a Source Data file.

Fig. 3. Dialysis-based free hemoglobin removal reduces hemolysis-driven phospholipid peroxidation.

A Experimental design comparing RBC-based perfusion of deceased donor kidneys without (Group 1, n = 7) or with (Group 2, n = 4) dialysis-based free hemoglobin (fHb) removal. Perfusate fHb levels during perfusion (B), and perfusate (C) and effluent (D) fHb levels after 24 and 48 h of perfusion (Mean ± SD. Two-way ANOVA). E Total oxidized phosphatidylcholines (oxPC) in tissue as detected by targeted oxPC LC-MS/MS in control (n = 4), RBC-perfused (n = 7), and fHb removal (n = 3) groups (Mean ± SD. One-way ANOVA). F Heatmap of individual oxPC species. See Table S6 for lipid identities and their abbreviations. G Levels of arachidonic acid (AA)-derived lipid mediators in perfusate during RBC-based (n = 7) and fHb removal during RBC-based perfusion (n = 3–4) (Median with IQR). H, I Box plot of GSVA enrichment score of WP_Ferroptosis (I, n = 64 genes) and FerrDb v2 driver, suppressor and marker genes ( J, n = 313 genes) (One-way ANOVA). Boxes represent median with 25th and 75th percentile and error bars minimum to maximum. J Volcano plots of differentially expressed genes in kidneys after RBC-based perfusion or fHb removal during RBC-based perfusion, compared to control kidneys (left and middle panel). Correlation of Log2FC changes between RBC-perfused vs control kidneys and fHb removal vs control kidneys (right panel). Red highlights validated FerrDb v2 driver, suppressor, or marker genes with a protein product (n = 313). K Temporal expression dynamics in FerrDb v2 driver, suppressor, and marker genes for RBC-perfused and fHb removal groups relative to control kidneys (|Log2FC| ≥ 2 and adjusted P ≤ 0.05 for at least one of the comparisons). Two-sided Wald test with BH correction ( J, K). Association between perfusate fHb and lipid peroxidation markers after 48-h of perfusion: perfusate TBARS (MDA) (L) and tissue 4-hydroxynonenal (4HNE) (M) (Pearson’s correlation coefficient, two-sided. Shaded area indicates 95% CI). A Illustrated by Manon Zuurmond (scientific illustrator). Source data are provided as a Source Data file.

Fig. 4. Cell-free perfusion negates hemolysis-driven iron accumulation and phospholipid peroxidation.

A Experimental design comparing RBC-based perfusion (Group 1, n = 7) with cell-free perfusion (Group 3, n = 8). Perfusate iron levels during perfusion (B) and at 24 and 48 h (C) (Mean ± SD, two-way ANOVA). D Iron deposition visualized by Enhanced Perl’s Prussian Blue (EPPB) staining in control (n = 4), RBC-perfused (n = 7), and cell-free perfused (n = 5) kidneys. Representative images and quantification (Mean ± SD. One-way ANOVA). Scale bar, 200 μm. E Total oxidized phosphatidylcholines (oxPC) in tissue as detected by targeted oxPC LC-MS/MS in control (n = 4), RBC-perfused (n = 7), and cell-free perfused (n = 4) kidneys (Mean ± SD. One-way ANOVA). F Heatmap of individual oxPC species. See Table S6 for lipid identities and their abbreviations. G Perfusate TBARS (MDA) levels in RBC-perfused (n = 7) and cell-free perfused (n = 8) kidneys (Mean ± SD. Two-way ANOVA). H Levels of arachidonic acid (AA)-derived lipid mediators in perfusate during RBC-based (n = 7) and cell-free (n = 8) perfusion (Median with IQR). Box plot of GSVA enrichment score of GOBP_Iron_Ion_Transport (I, n = 58 genes, Kruskall-Wallis test), WP_Ferroptosis ( J, n = 64 genes, one-way ANOVA) and validated FerrDb v2 driver, suppressor and marker genes (K, n = 313 genes, one-way ANOVA). Boxes represent median with 25th and 75th percentile and error bars minimum to maximum. L Volcano plots of differentially expressed genes after RBC-based or cell-free perfusion, compared to control, and cell-free perfusion compared to RBC-based perfusion. Red highlights validated FerrDb v2 driver, suppressor, or marker genes with a protein product (n = 313). M Temporal expression dynamics in FerrDb v2 driver, suppressor, and marker genes after RBC-based or cell-free perfusion, compared to control kidneys, and cell-free perfusion compared to RBC-based perfusion (|Log2FC| ≥ 2 and padj ≤ 0.05 for at least one of the comparisons). Two-sided Wald test with BH correction (L, M). Perfusate LDH (N) and NGAL (O) concentration after 48 h of RBC-based (n = 7) and cell-free (n = 8) perfusion. (Mean ± SD. Mann-Whitney test, two-sided). P Association between perfusate NGAL and iron after 48 h of perfusion (Spearman’s rank correlation coefficient, two-sided. Shaded area indicates 95% CI). A Illustrated by Manon Zuurmond (scientific illustrator). Source data are provided as a Source Data file.