Fatty acid-binding protein 5 is a functional biomarker and indicator of ferroptosis in cerebral hypoxia

Abstract

The progression of human degenerative and hypoxic/ischemic diseases is accompanied by widespread cell death. One death process linking iron-catalyzed reactive species with lipid peroxidation is ferroptosis, which shows hallmarks of both programmed and necrotic death in vitro. While evidence of ferroptosis in neurodegenerative disease is indicated by iron accumulation and involvement of lipids, a stable marker for ferroptosis has not been identified. Its prevalence is thus undetermined in human pathophysiology, impeding recognition of disease areas and clinical investigations with candidate drugs. Here, we identified ferroptosis marker antigens by analyzing surface protein dynamics and discovered a single protein, Fatty Acid-Binding Protein 5 (FABP5), which was stabilized at the cell surface and specifically elevated in ferroptotic cell death. Ectopic expression and lipidomics assays demonstrated that FABP5 drives redistribution of redox-sensitive lipids and ferroptosis sensitivity in a positive-feedback loop, indicating a role as a functional biomarker. Notably, immunodetection of FABP5 in mouse stroke penumbra and in hypoxic postmortem patients was distinctly associated with hypoxically damaged neurons. Retrospective cell death characterized here by the novel ferroptosis biomarker FABP5 thus provides first evidence for a long-hypothesized intrinsic ferroptosis in hypoxia and inaugurates a means for pathological detection of ferroptosis in tissue.

Fig. 1. Identification of ferroptosis-specific surface proteins.

A Experimental strategy for identifying candidate biomarkers in ferroptosis-sensitive HT-1080 cells by surface biotinylation and protein mass spectrometry. B Lipid peroxidation induced by RSL3 (300 nM) treatment for 3 h in HT-1080 cells. Treated and untreated cells measured by BODIPY 581/591 C11 stain (BODIPY-C11). A typical flow histogram of three independent experiments is depicted. C Flow cytometry of untreated cells compared to short term (3 h) RSL3 (250 nM) treated HT-1080 cells. Selected populations designated by indicated gates and ratios in Table 1 (U1, untreated; ES, end-stage ferroptosis; P1, unshifted RSL3-treated; P2, FSC-shifted RSL3-treated). D Volcano plot of biotinylated candidate proteins (P2) identified by mass spectrometry with ≥2 peptides in relation to control population (U1). Selected significant proteins (Students t-test, above dotted line, p < 0.05) were investigated further.

Fig. 2. FABP5 is concentrated at cell surface during ferroptosis.

A Immunohistochemistry staining in normal human cerebral cortex sections showing endogenous expression using antibodies against CALML5, FABP5, CTSV, LGALS7, S100A14 as well as TfR1 identified previously in [28]). The data shown here are derived from the Human Protein Atlas [68]. B Normalized fluorescence intensity heat map of candidate biomarker proteins following treatment with equivalent death-inducing concentrations of different pharmacological agents in HT-1080 cells. C Validation of FABP5 antibody specificity. FABP5 overexpressing HT-1080 cells (FABP5 OE) with C-terminal FLAG tag compared to control (Ctrl) by fluorescence staining and Western blot. Results observed with two antibodies (see Supplemental Reagents Table) were indistinguishable. Right panel, FABP5 expression level was detected by whole membrane Western blot in given populations. D Timecourse of lipid peroxidation in HT-1080 as indicated by flow cytometry analysis of Bodipy C-11 fluorescence and FABP5 cell surface staining in unpermeabilized cells and confocal images of the same cells.

Fig. 3. FABP5 is specifically upregulated during ferroptosis.

A Timecourse in hours (h) of RSL3 (200 nM) or staurosporine (50 nM) induced changes in FABP5 as detected by confocal microscopy in FABP5 OE cells. B High-content analysis of FABP5 in HT-1080 cells is shown as normalized mean fluorescent intensity ± SD of n = 4 replicate samples representative of at least three independent repetitions of the experiment in A treated with RSL3 (200 nM) or staurosporine (50 nM). Western blots show FABP5 protein detection at the same time points following treatment. C Relative FABP5 expression by quantitative PCR with RSL3 (200 nM) time course. Values are shown as mean ± SD of n = 3 technical replicates related to untreated (0 h). Statistics were calculated using two-way ANOVA against respective control conditions.

Fig. 4. FABP5 biomarker validation in RSL3-treated and GPX4 knockout cells.

A FABP5 fluorescence intensity changes after RSL3 or GPX4 knockout (KO) in cell lines of different etiologies. Calu-1, HCC827, HT-1080 and U-138MG cells after treatment with RSL3 (200 nM, 5 h), and SH-SY5Y and human fetal fibroblast (hFF) cells with RSL3 (200 nM) + ammonium ferric citrate (FAC, 250 μM) for 3 h. 2 × 10³ cells were assayed per condition. RSL3 final intensity is shown as mean ± SD of n = 3 independent experiments with three replicate wells each. Fold change at 96 h post infection with GPX4 KO or control (empty vector) lentivirus is shown as mean ± SD of at least three independent experiments with 2 × 10³ cells assayed per condition in each of three replicate wells. SH-SY5Y and hFF are viable following GPX4 KO-induced lentivirus. B Live cell brightfield and fluorescence images of HT-1080 viability (indicated by DAPI penetration) were taken hours post infection (p.i.) with GPX4 KO and control virus at 72 and 96 h. Increased loss of viability was observed at 96 h, while early detection of FABP5 antigen at 72 h is shown in C when only a fraction of cells have died. C Western analysis of respective proteins in HT-1080 cells 72 h following GPX4 ablation. Statistics were calculated using two-way ANOVA (*p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001; ns not significant).

Fig. 5. Elevated FABP5 expression induces PUFA-containing lipids and ferroptosis sensitivity.

A Viability of FABP5 OE compared to control HT-1080 cells treated with ferroptosis inducer IKE (1.25 µM, 16 h, two-way ANOVA (*p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001; ns not significant)). B Viability of FABP5 knockdown (KD) compared to control HT-1080 cells treated with ferroptosis inducer IKE (1.25 µM, 16 h, two-way ANOVA (*p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001; ns not significant)). C, D Total lipid abundance and distribution by peak values as revealed by mass spectrometry for FABP5 OE cells compared to control cells. No significant differences were observed in total lipid content. Class terms are derived from LIPIDMAPS [69]. E Lipid heatmap annotations showing z-scores of all significantly misregulated phospholipid species in FABP5 OE and control cell samples compared with controls. Nearly all misregulated lipids are classified as PUFA-containing glycerophospholipids (boxed entries) with the number of double bonds indicated after the colon. DG diacylglycerol, LPA lysophosphatidic acid, LPE 1-acyl-sn-glycero-3-phosphoethanolamine, LPS 1-acyl-sn-glycero-3-phosphoserine, PC phosphatidylcholine, PE phosphatidylethanolamine, PE phosphatidylinositol, PS phosphatidylserine, O ether lipids (Welch t-test, n = 3, ns not significant). F Cumulative changes in each saturation class of phospholipids up- or downregulated in FABP5 OE cells. Lipids downregulated represent a smaller fraction of the total lipid pool, as represented by the proportion of the ‘down’ pie chart. Volcano plots demonstrate fold changes of individual lipids according to saturation, the dotted line indicates significant lipids shown in (E). MUFA monounsaturated, SFA saturated, PUFA polyunsaturated.

Fig. 6. FABP5 demarcates dying neurons in cortical and hippocampal sections with hypoxic damage.

A 15 h post-transient middle cerebral artery occlusion in mouse show evidence of ferroptosis by FABP5 staining. Left image: hematoxylin/eosin stain at the level of basal ganglia. The area of the lesion is bordered by stippled line. The red box marks an area of the penumbra, the black box an area of the corresponding healthy contralateral side. The images on the right show examples of the immunohistochemical detection of FABP5 in proximal areas. Arrows in the penumbra indicate almost exclusively neurons with partly shrunken nuclei and cytoplasmic FABP5 expression, while contralateral side neurons are healthy and unstained (open arrows). Scale bars correspond to 800 µm (left) and 20 µm (right images), respectively. B Human cerebellar cortex of a control case (C04) and a case with hypoxic damage (H04). (Left) Hematoxylin/eosin stains. In contrast to intact Purkinje cells (PCs; open arrows), hypoxically damaged PCs (solid arrows) are shrunken with eosinophilic cytoplasm, condensed nuclei, and an undefined nucleolus. (Right) Immunohistochemistry for FABP5 in neighboring sections, only hypoxically damaged PCs express FABP5 (solid arrows) in contrast to intact PCs (open arrows). Scale bar corresponds to 20 µm. C Human hippocampus of a control case (C01) and a case with hypoxic damage (H01). (Left) Hematoxylin/eosin stains. Hypoxically damaged shrunken pyramidal cells with condensed nuclei and eosinophilic cytoplasm (arrows). (Right) FABP5 immunohistochemistry in adjacent sections. In hypoxic damage almost all pyramidal cells strongly express FABP5 (arrows) in contrast to the control case without FABP5 expression. The weak brown color in the control case corresponds to lipofuscin. Scale bar corresponds to 20 µm.