How NINJ1 mediates plasma membrane rupture and why NINJ2 cannot

Abstract

Ninjurin-1 (NINJ1) is an active executioner of plasma membrane rupture (PMR), a process previously thought to be a passive osmotic lysis event in lytic cell death. Ninjurin-2 (NINJ2) is a close paralog of NINJ1 but cannot mediate PMR. Using cryogenic electron microscopy (cryo-EM), we show that NINJ1 and NINJ2 both assemble into linear filaments that are hydrophobic on one side but hydrophilic on the other. This structural feature and other evidence point to a PMR mechanism by which NINJ1 filaments wrap around and solubilize membrane fragments and, less frequently, form pores in the plasma membrane. In contrast to the straight NINJ1 filament, the NINJ2 filament is curved toward the intracellular space, preventing its circularization or even assembly on a relatively flat membrane to mediate PMR. Mutagenesis studies further demonstrate that the NINJ2 filament curvature is induced by strong association with lipids, particularly a cholesterol molecule, at the cytoplasmic leaflet of the lipid bilayer.

Keywords: NINJ1; NINJ2; inflammatory cell death; lytic cell death; membrane solubilization; necroptosis; ninjurin-1; ninjurin-2; plasma membrane rupture; pyroptosis.

Figure 1. Cryo-EM structures of the human NINJ2 and NINJ1 oligomers.

(A) Representative cryo-EM 2D class averages of the NINJ2 oligomer. (B) Low-pass filtered cryo-EM density map of the NINJ2 oligomer. Note the curved shape of the filament. (C) High-resolution cryo-EM density map of the NINJ2 oligomer shown at different views. Three subunits in each of the two layers of the filament are highlighted with different color. Lipids trapped in between the two layers are colored as yellow. Dotted line in the middle panel marks the clipping plane for the view shown in the right panel. (D) Architecture of the NINJ2 oligomer shown with atomic models. Lipids are not modeled and shown as cryo-EM densities. N and C denote the amino- and carboxyl-terminus of the amino acid chain, respectively. The two arrows denote the anti-parallel relationship of the two layers in the filament. (E) Representative cryo-EM 2D class averages of the NINJ1 oligomer. (F) Low-pass filtered cryo-EM density map of the NINJ1 oligomer. Note the straightness of the filament compared to the NINJ2 oligomer shown in (B). (G, H) High-resolution cryo-EM density map (G) and atomic models (H) of the NINJ1 oligomer shown at different views. All the coloring and labeling follow the same scheme as explained in (C) and (D).

Figure 2. Atomic interactions responsible for the assembly of NINJ1 or NINJ2 filament.

(A) Architecture of the NINJ1 filament shown with three adjacent subunits. The middle subunit is color-coded for four named regions, with their protein sequence boundaries shown in the bar above. AH, amphipathic helix; TM, transmembrane helix. (B) Hydrophobic interactions between AH2 and the TM helices. For clarity, the hydrophobic sidechains on the TM helices are not shown. (C) Polar interactions between AH2 and the TM helices. The green dashed lines represent possible hydrogen bonds. The red dot represents a modeled water molecule that mediates some of the hydrogen bond interactions among the protein side chains. (D) Hydrophobic interactions between AH1 and the TM helices. (E) Polar interactions between AH1 and the TM helices. The green dashed lines represent possible hydrogen bonds, and the red dashed line represents salt bridge interaction. Note that the interactions in (D) and (E) are mainly between AH1 and the TM helices of a neighboring subunit, resulting in crosslink of the two adjacent subunits. (F) Surface rendering of the two sides of the NINJ1 filament showing their distinctive difference in hydrophobicity. A hydrophobicity scale bar is placed at the bottom. The middle insert that helps to locate the two surfaces is the same view as the right panel in Figure1H. (G-L) Architecture and atomic interactions in the NINJ2 filament. Each panel is presented in the same way as that for NINJ1 at corresponding position on the left. The middle insert in panel L is the same view as the right panel in Figure1D.

Figure 3. How protein-lipid interactions in the NINJ1 and NINJ2 filaments affect their curvature.

(A) Open-up view of the NINJ1 filament and the associated lipids. The double-layered cryo-EM density map of NINJ1 filament is cut open along the middle line, and one-half is shown here. The NINJ1 densities are rendered as white, semi-transparent surface, with atomic models of three adjacent subunits fitted in and differentially colored. Lipid densities are colored as yellow. (B) NINJ1 protein-lipid interactions at the exoplasmic leaflet of the lipid bilayer. This is a zoom-in view of the orange box in (A) but from a different angle, as denoted by the eye symbol. Some lipid-interacting sidechains are highlighted. (C) NINJ1 protein-lipid interactions at the cytoplasmic leaflet of the lipid bilayer. This is a zoom-in view of the blue box in (A) but with only atomic models to highlight some side chains. (D) Open-up view of the NINJ2 filament and the associated lipids. Same as in (A), except that some obvious cholesterol densities are highlighted with brown color. The aliphatic-chain lipids in the cytoplasmic leaflet are arbitrarily modeled as phosphatidylcholine (PC). (E) NINJ2 protein-lipid interactions at the exoplasmic leaflet of the lipid bilayer. This is a zoom-in view of the orange box in (D) from the extracellular space. Note the differences compared with (B). (F) NINJ2 protein-lipid interactions at the cytoplasmic leaflet of the lipid bilayer. This is a zoom-in view of the blue box in (D) and is to be compared with (C). (G) Superposition of the NINJ1 and NINJ2 models to show their nearly identical structures at the subunit level. (H) Measurements of the inter-subunit distances at three different positions in the NINJ1 (top) or NINJ2 (bottom) filaments. Pairs of amino acid alpha carbons used for the distance measurement are marked. (I) A schematic explanation of how strongly associated lipids at the exoplasmic or cytoplasmic leaflet affects the filament curvature of NINJ1 (top) or NINJ2 (bottom). NINJ1 or NINJ2 subunits are depicted as inverted trapezoids to reflect their molecular shapes as shown in (G). Strongly associated lipids are depicted as yellow ellipses that “glue” the adjacent protein subunits. (J) Virtually assembled long filaments of NINJ1 and NINJ2. Many copies of the cryo-EM density maps of NINJ1 or NINJ2 were fitted head-to-tail to virtually extend the filaments. The NINJ1 filament is nearly straight, while the NINJ2 filament goes around a cylinder with a diameter measured to be about 45nm (bottom right panel).

Figure 4. Mutagenesis studies on the effect of lipid association in NINJ1 or NINJ2 filament assembly and PMR capability.

(A) A table summarizing NINJ1 and NINJ2 mutants with their lipid-interacting residues swapped. (B) The killing scores of NINJ1 or NINJ2 wild type (WT) and mutants. Killing scores are measurements of cytotoxicity, which is correlated with NINJ1 or NINJ2’s capability of mediating PMR, as defined in reference . Data are means (bars) ±standard error of mean (SEM, error bar) of three individual replicates (diamonds). The killing score of NINJ1-WT was defined as 1 in each measurement. (C) Negative-staining EM micrographs of NINJ1 or NINJ2 WT and mutants. See main text for detailed explanation.

Figure 5

A schematic model for NINJ1 activation and plasma membrane rupture.