PEBP1 Wardens Ferroptosis by Enabling Lipoxygenase Generation of Lipid Death Signals

Abstract

Ferroptosis is a form of programmed cell death that is pathogenic to several acute and chronic diseases and executed via oxygenation of polyunsaturated phosphatidylethanolamines (PE) by 15-lipoxygenases (15-LO) that normally use free polyunsaturated fatty acids as substrates. Mechanisms of the altered 15-LO substrate specificity are enigmatic. We sought a common ferroptosis regulator for 15LO. We discovered that PEBP1, a scaffold protein inhibitor of protein kinase cascades, complexes with two 15LO isoforms, 15LO1 and 15LO2, and changes their substrate competence to generate hydroperoxy-PE. Inadequate reduction of hydroperoxy-PE due to insufficiency or dysfunction of a selenoperoxidase, GPX4, leads to ferroptosis. We demonstrated the importance of PEBP1-dependent regulatory mechanisms of ferroptotic death in airway epithelial cells in asthma, kidney epithelial cells in renal failure, and cortical and hippocampal neurons in brain trauma. As master regulators of ferroptotic cell death with profound implications for human disease, PEBP1/15LO complexes represent a new target for drug discovery.

Keywords: GPX4; PEBP1/15LO complex; acute kidney injury; asthma; brain trauma; cell death; ferroptosis; ferrostatin-1; phosphatidylethanolamine oxidation; redox phospholipidomics.

Figure 1. PEBP1 forms complexes with 15LO1 and 15LO2

(A) 1) Far Western blotting gels showing binding of PEBP1 with 15LO. Ponceau S staining of proteins (left panel). Immuno-detection of PEBP1 bound to 15LO using anti-PEBP1 antibody (middle and right panels). 2) Quantitation of 15LO1 bound to wt PEBP1 and P112E PEBP1 with anti-15LO1 antibody (means±SD, *p < 0.05 vs. wt PEBP1, N=7/group. (B) SDS-PAGE gels illustrating interactions of 15LO1 with wt PEBP1 and P112E PEBP1 revealed by cross-linking using glutaric dialdehyde (GDA). Proteins were detected by silver-staining or by anti–PEBP1 antibodies. C) Computational modeling of human PEBP1/15LO1 complex. 15LO1 (gray) and PEBP1 (red) bound to the lipid bilayer are displayed, with the PL head phosphorus atoms in tan spheres, and 15LO1 R402, crucial for PEBP1/15LO1 interactions, in pink spheres. Inset: Coordination of Fe²⁺ at 15LO1 catalytic byH360, H365, H540 and H544. (D) Close-up view of PEBP1/15LO1 interface. The interface between PEBP1 (yellow) and 15LO1 (gray) complex closely neighbors the AA binding site located near the catalytic site (highlighted in ellipse). K156, F166, D173, R402 in 15LO1, and P112, D144 and Y181 in PEBP1 make interfacial contacts that stabilize the PEBP1/15LO1 complex. (E) Coarse-grained molecular dynamics (CGMD) simulations confirm the inability of the P112E to bind 15LO1. Four independent runs are displayed, two for wt PEBP1 (CGMD1 and 2), and two for the mutant (CGMD3 and 4). The ordinates denote the closest distance between the two proteins, originally separated by 20 Å. In CGMD1 and 2, stable bound poses are reached ~80–100 ns; the minimal distance attained and maintained during the remaining 150 ns. For the mutant, there is either no formation of complex (3), or transient binding (4) indicating weaker (if any) binding of P112E to 15LO1. (F) Binding of P112E to 15LO1 in silico shows the weaker affinity and distinctive binding. Left: the optimal binding pose for mutant P112E is shifted away from the putative allosteric binding site near R402. Right: interfacial interactions. Wt PEBP1 exhibits several close contacts (see also Figure S5), which are broken due to the repulsion between mutated P112 (E112) and E175 on 15LO1. P112E and 15LO1 residues are colored as blue and green, respectively. E112 on PEBP1 and E175 on 15LO1 undergo side chain rotations in the mutant complex that weaken the intermolecular interactions. (G) Binding of wt PEBP1 and P112E to Raf-1 kinase whose binding domain (KBD) (PDB id: 3c4c) (Tsai et al., 2008) was docked against human PEBP1 (left) and P112E mutant (right). Top docking poses of PEBP1 were in the vicinity of KBD N-terminus. Both wt (left) and P112E (right) utilize the region G143-R146 on PEBP1 for interacting with Raf-1 KBD (Deiss et al., 2012). (H) CGMD of Raf-1 binding to wt PEBP1 (left) and P112E (right). In both cases, the complex was formed (distance less than 0.5 nm, indicated by dashed redline) at approximately 175ns (left) and 200ns (right).

Figure 2. PEBP1 co-localizes with15LO1 and 15LO2 in cells

(A) 3D volume views showing the immuno-localization of PEBP1 (red) with 15LO1 (green) in HAECs (nuclei blue) in untreated (upper panel) and IL13 treated cells (lower panel). (B) Objects identified as containing both PEBP1 and 15LO1 (yellow) were identified by 3D object-based segmentation and co-localization analysis. (C) The number of co-localized objects for 15LO1 and PEBP1 expressed as a percentage of the total number of puncta in HAECs following 5 days of culture without or with IL13 (means±SD, *p < 0.05 vs. untreated. (D) Fluorescence resonance energy transfer (FRET) based analysis showing the close physical proximity (within 10 Á) of 15LO1 with PEBP1 in HAEC (confirmed by proximity ligation assay). Upper panels: the donor (cy3) emissions are colored green and the acceptor (cy5, FRET) colored red. Lower panel: FRET ratio (donor/acceptor) which has been pseudo-colored (range 0–5, blue-red). FRET was confirmed by acceptor photobleaching (right hand panels) with an increase in the ratio following photobleaching of cy5 within the circled regions. (E) Co-localization of PEBP1 (red) with 15LO2 (green) in HK2 cells exposed to LPS. The yellow shows the spots positive for both PEBP1 and 15LO2. In this example, there were 241 spots positive for both 15LO2 and PEBP1 in untreated cells, and 753 positive for both proteins in LPS treated cells. (F) Number of co-localized objects positive for both 15LO2 and PEBP1 in untreated and LPS treated HK2 cells, expressed as a percentage of the total number of puncta (means±SD, *p< 0.05 vs. untreated, N=5/group). (G) Co-localization of 15LO2 and PEBP1 in HT22 cells. Left panel: a 3D volume rendering of immunostaining for 15LO2 (green), and PEBP1 (red) with nuclei in blue. (H) Percentage of objects identified as positive for 15LO2 alone, PEBP1 alone or both 15LO2 and PEBP1 together (means±SD, N=3/group). See also Figure S2.

Figure 3. 15LO1 catalyzes PEBP1-dependent production of PEox in IL13 stimulated HAECs

(A) Normal phase LC/MS/MS chromatogram (black) and mass spectra of PLs in HAECs. BMP-bis-monoacylglycero-phosphate; PG-phosphatidylglycerol; PE-phosphatidylethanolamine; PS-phosphatidylserine; CL-cardiolipin; PC-phosphatidylcholine; PI-phosphatidylinositol. (B) Contents of PLox in HAECs treated with IL13 and AA (means±SD. *p<0.05 vs. control, N=3/group). (C) Volcano plots of IL13 induced changes of PEox in wt (left plot) and PEBP1 KD HAECs (right plot) (log₂ (fold-change) vs. significance (log₁₀ p-value). Cells were exposed to IL13 in the presence of AA. (D) Quantitation of PEBP1 KD by siRNA in HAECs (means±SD, *p<0.05 scrambled vs. siPEBP1, N=3/group). Insert: A typical Western blot for PEBP1. (E) 15LO1 KD changes HETE-PE and HpETE-PE in HAECs exposed to IL13 and AA, (N=4, different human donors). See also Figure S3 and S4.

Figure 4. RSL3-induced ferroptosis depends on endogenous or exogenously added PEBP1

(A) Quantitation of PEBP1 (normalized to actin) in overexpressing (pCMV6-PEBP1) and empty vector (pCMV6) cells. Note, PEBP1-myc-FLAG and constitutive PEBP1 migrated differently due to the differences in molecular masses. Insert: Western blots show overexpressed PEBP1 (upper band, PEBP1-myc-FLAG) in transfcted HK2 cells. (B) HK2 cells with elevated PEBP1 are more sensitive to ferroptosis. Insert: Ferrostatin-1-inhibitable cell death (means±SD, *p<0.05 vs. pCMV6, **p< 0.05 vs. pCMV6/RSL3, ***p< 0.05 vs pCMV6-PEBP1/RSL3, ^#p<0.05 vs. pCMV6/RSL3, N=3/group). Insert: quantitation of ferroptosis as differences in LDH in ±FER-treated cells. (C) PEBP1 KD suppresses RSL3-induced ferroptosis in HAEC cells (means±SD, *p<0.05 RSL3 (wt) vs. DMSO (wt) **p<0.05 PEBP1 KD vs. RSL3 (wt), N=3/group). (D) PEBP1 KD suppresses RSL3-induced ferroptosis in HT22 cells (means±SD, *p<0.05 vs. scrambled siRNA control, **p<0.05 vs. scrambled siRNA treated cells+RSL3, N=3/group). (E) Composite MS³ spectrum of HpETE-PE with m/z 798.47 from supernatants after incubations of MLE cells exposed to AA-PE/15LO2/PEBP1. Fragments of HpETE-PE are shown in red. Insert: structural formula of OOH-AA-PE and fragments formed during MS²/MS³ analysis (red). (F) PEBP1 enhances RSL3-induced ferroptosis in MLE cells incubated with exogenously added AA-PE/15LO2 (means±SD, *p<0.05 vs. control, ^#p< 0.05 vs. RSL3. N=3/group). See also Figure S5.

Figure 5. Locostatin enhances RSL3-induced ferroptosis

(A) HAECs (means±SD, *p<0.05 vs. no RSL3/no locostatin, **p<0.05 vs. locostatin, N=3/group) (B) PHKCs (means±SD, *p<0.05 vs. no RSL3/no locostatin/no FER, **p<0.05 vs. RSL3, ^#p<0.05 vs. RSL3, ^##p<0.05 vs. RSL3/locostatin, N=3/group). (C) Ferrostatin-1 (but not zVAD-fmk or Necrostatin-1s) inhibits RSL3+locostatin induced ferroptosis in HT22 cells (means±SD, *p<0.05 vs. no RSL3/no locostatin/no FER, **p<0.05 vs. RSL3/locostatin, N=3/group). (D) Ferrostatin-1 (FER) suppresses RSL3-induced PEox in PHKCs in the presence of locostatin. Volcano plot of PEox changes (log₂ (fold-change) vs. significance (−log₁₀ (p-value)), N=3/group. (E) Locostatin enhances RSL3-induced ferroptosis in HK2 cells (means±SD, *p<0.05 vs. control (no RSL3, no FER, no locostatin), #p<0.05 vs. RSL3, N=3/group). (F) Necrostatin-1s and zVAD-fmk do not suppress RSL3/locostatin induced death in HK2 cells (means±SD, *p<0.05 vs RSL3 only; #p<0.05 vs. RSL3 plus locostatin, N=3/group).

Figure 6. Binding of free AA by PEBP1 facilitates 15LO-catalyzed oxygenation of AA-PE

PEBP1 stimulates accumulation of OOH-AA-PE (but not free AA-OOH) catalyzed by: (A) porcine 15LO1 and (B) human recombinant 15LO2 (means±SD, *p<0.05 vs. PE, ** p<0.05 vs. AA, N=4/group). (C) A typical dot-blot illustrates binding of PEBP1 with free AA (3 separate experiments). (D) AA causes concentration-dependent decrease of monomeric PEBP1 and increase of its oligomers (means±SD, *p<0.05 vs. PEBP1 without AA, N=4/group). Insert: a typical Blue native PAGE gel (silver-staining), (E) Electrospray ionization mass spectrometry (ESI-MS) demonstrates AA binding by PEBP1. ESI-MS spectra before and after incubation with AA (left panels). PEBP1 mass was 21391 Da. AA + PEBP1 yielded 3 additional species with masses 22359, 23331 and 24303 Da corresponding to PEBP1 plus 3, 6 and 9 AA molecules (with 1, 2 and 3 acetic acid molecules), respectively. AA binding to the PEBP1 mutant Y176X (right panels). MS spectra of Y176X before and after incubation with AA. Mutant mass was 19850 Da. AA + Y176X mutant yielded 2 additional species, albeit at much lower levels, with masses 20818 and 21727 corresponding to Y176X plus 3 AA and 6AA ligands, respectively (plus 1 acetic acid molecule). Shown are typical spectra from at least four independent experiments. (F) Computational docking shows multiple binding sites for AA on PEBP1. Up to nine AA molecules bind to PEBP1 with energies ranging from −6.2 to −4.0 kcal/mol. See also Figure S6 and Tables S1, S2 and S3.

Figure 7. PEBP1/15LO complexes and PEox are detectable in vivo in pro-inflammatory disease conditions

(A) Increased co-localization of PEBP1/15LO1 correlates with asthmatic T2 inflammation in freshly brushed HAECs. 1) Fresh HAEC cytospin from a patient with asthma. Top panel: a volume view of the immunolocalization of 15LO1 (green), PEBP1 (red), and nuclei (blue). Lower panel: the complexes containing both PEBP1 and 15LO1 (yellow). 2) The number of objects identified as positive for both 15LO1 and PEBP1 correlates strongly with fraction of exhaled nitric oxide (FeNO) in human subjects (N=6, p<0.001) (B) Increased contents of PEox in urine cell pellets from patients with acute kidney injury (AKI). Volcano plot of changes in PEox (log₂ (fold-change) vs. significance (−log₁₀ (p-value)) in urine pellet samples from patients with AKI (N=5/group). (C) Scatter plots of PEox species from the plot in B (as indicated) in urine cell pellets from patients with AKI. (D) Changes in ferroptotic protein expression in rat brain cortex after CCI. Note increased 15LO2 levels (*p<0.05; N=3–4/group), and decreased GPX4 levels (*p<0.05; N =3–4/group) at 4 h after CCI in ipsilateral cortex of injured vs. naïve rats. The amount of PEBP1 remained unchanged. (E) Changes in GPX4 activity in rat brain cortex at 4 h after CCI (means±SD, *p<0.05 vs. naïve rats, N=3/group). (F) Co-localization of PEBP1 and 15LO2 in brain tissue. Stitched image showing high resolution large area confocal scanning of 3x5 image fields. Left panels: the overlayed emissions for the immunolocalization of PEBP1 (red), 15LO2 (green) and nuclei (blue). Right panels: co-localization analysis for 15LO2 and PEBP1, with the number of spots having both proteins appearing yellow. Scale bar is 200 microns. (G) Number of co-localized 15LO2 and PEBP1 in brain tissue 4 h after CCI (means±SD, *p<0.001 vs. sham, N=5/group). (H) Volcano plot demonstrates changes in the content of PEox at 1h post CCI (N=4/group). (I) Identification of pro-ferroptotic PEox in rat brain cortex after CCI using high resolution Orbitrap Fusion^™ Lumos^™ Tribrid^™ Mass Spectrometer. 1) Full mass spectrum of PE (rat brain after CCI). Insert: Spectrum in the range of m/z from 782.1 to 782.9. Molecular ion with m/z 782.5350 (PEox) is shown in red. 2)-3) MS² spectra of precursor ions with m/z 750.5451 and 766.5411 containing AA and corresponding to PE-O-18:1/20:4 and PE-18:0/20:4, respectively. 4) MS² spectra of PEox with m/z 782.5350. Note 2 species with 1 and 2 oxygens formed after oxidation of PE(18:1/20:4) and PE(18:0/20:4) respectively. 5) and 6) Fragmentation patterns of ions with m/z 319 and m/z 317 (335-H₂O) generated by MS³ analysis of PEox with m/z 782.5350. The fragment with m/z 113 is diagnostic of the OH- and OOH-groups at 15^th carbon of AA. See also Figure S7 and S8.