NINJ1 regulates plasma membrane fragility under mechanical strain

Abstract

The integrity of the plasma membrane is vital for nearly all aspects of cell functioning¹. Mechanical forces can cause plasma membrane damage², but it is unclear whether there are large molecules that regulate the integrity of the plasma membrane under mechanical strain. Here we constructed a 384-well cellular-stretch system that delivers precise, reproducible strain to cultured cells. Using the system, we screened 10,843 small interfering RNAs (siRNAs) targeting 2,726 multipass transmembrane proteins for strain-induced membrane permeability changes. The screen identified NINJ1-a protein that was recently proposed to regulate pyroptosis and other lytic cell death^3,4-as the top hit. We demonstrate that NINJ1 is a critical regulator of mechanical-strain-induced plasma membrane rupture (PMR), without the need for stimulating any cell death programs. NINJ1 levels on the plasma membrane are inversely correlated with the amount of force required to rupture the membrane. In the pyroptosis context, NINJ1 on its own is not sufficient to fully rupture the membrane, and additional mechanical force is required for full PMR. Our study establishes that NINJ1 functions as a bona fide determinant of membrane biomechanical properties. Our study also suggests that PMR across tissues of distinct mechanical microenvironments is subjected to fine-tuning by differences in NINJ1 expression and external forces.

Fig. 1. High-throughput genetic screen identifies NINJ1 as a regulator of PMR induced by mechanical strain.

a, Schematic of the high-throughput stretch system. A vacuum applied to sealed wells deforms the PDMS membrane, stretching the adherent cells on top. b, The two operating modes of the system: single-well imaging through a microscope and full-plate imaging through a plate reader. c, Images of HeLa-YFP cells immediately before and 120 s after strain application. Scale bar, 150 µm. d, YFP-intensity traces before and after the strain application. Each trace represents data from a single well; 10 wells were analysed and the red trace is the average. e,f, Trypan Blue and DRAQ7/Hoechst staining of the cells after strain application. Scale bars, 100 µm. g, The workflow of the siRNA screen for regulators of strain-induced membrane damage. h, Overview of the primary screen hitpicking, covering 2,726 genes encoding multipass transmembrane proteins. Each dot represents one of 10,843 siRNAs; the red dots indicate primary hits (z score > 1.5). i, Final validation of the top 20 candidate and 4 control genes showed that NINJ1 was the sole hit. j, YFP quenching in HeLa-YFP cells transfected with pooled siRNAs against NINJ1 or scrambled control. n = 4 trials per group. Statistical analysis was performed using a two-sided unpaired Student’s t-test versus the scrambled control. k, Western blot analysis of endogenous NINJ1 in HeLa cells 48 h after siRNA transfection. l, Trypan Blue staining of the control and NINJ1-knockdown cells after mechanical strain. Scale bar, 100 µm. m,n, LDH release (m) and the percentage of DRAQ7⁺ cells (n) in the control and NINJ1-knockdown groups, with or without 50% strain for 5 s. n = 4–8 trials per group. Statistical analysis was performed using using two-way analysis of variance (ANOVA) with Bonferroni correction. Unless otherwise indicated, data are mean ± s.e.m. **P < 0.01 versus scrambled. The diagrams in a, b and g were created using BioRender. Source data

Fig. 2. NINJ1 renders the plasma membrane susceptible to rupture under mechanical strain.

a, LDH release in HEK293T cells transfected with vector or CMV-driven NINJ1-IRES-mCherry was measured 24 h after transfection. n = 3 trials per group. Statistical analysis was performed using two-sided an unpaired Student’s t-test. b, Western blot analysis of NINJ1 in non-transfected, vector-transfected and CMV-NINJ1-transfected HEK293T cells. c, Western blot analysis of NINJ1 in HEK293T cells stably expressing TRE3G-NINJ1-IRES-mCherry induced with 0–300 ng ml⁻¹ doxycycline (Dox) for 24 h; parental cells were used as controls. d, Confocal images of TRE3G-NINJ1-mCherry stable cells with or without 100 ng ml⁻¹ doxycycline induction for 24 h. The arrowheads indicate fluorescent puncta. Scale bar, 10 µm. e,f, LDH release (e; n = 3–5 wells per group) and the DRAQ7⁺ cell percentage (f; n = 3–4 trials; 300–500 cells per group) 30 min after strain application in parental or NINJ1-inducible HEK293T cells pretreated with doxycycline. The duration of strain application was 5 s. g, Western blot analysis of endogenous NINJ1 in HeLa and ZOS cells. h,i, LDH release (h; n = 8–12 wells per group) and DRAQ7⁺ percentage (i; n = 3–6 trials; 300–500 cells per group) in HeLa and ZOS cells 30 min after strain. Statistical analysis was performed using two-way ANOVA with Bonferroni correction. j–l, Relative NINJ1 mRNA levels (j; n = 3 per group), LDH release (k; 50% strain; n = 6 trials per group) and the DRAQ7⁺ cell percentage (l; 50% strain; n = 3 trials; 300–400 cells per group) across human osteosarcoma cell lines. Statistical analysis was performed using one-way ANOVA with Bonferroni correction versus HeLa. m–o, Single-cell clones of 143B cells (n = 18) were analysed for NINJ1 mRNA using qPCR after treatment with 55% strain. m, Schematic of the experiment. LDH release (n) and DRAQ7⁺ (o) percentages were measured and correlated with NINJ1 expression (300–400 cells per clone). Unless otherwise indicated, data are mean ± s.e.m. The diagram in m was created using BioRender. Source data

Fig. 3. NINJ1 levels on the plasma membrane are inversely correlated with rupture tension.

a, Schematic of the micropipette aspiration assay. HeLa cells expressing NINJ1–mCherry fusion were treated with 2 mM NEM to generate GPMVs, which were aspirated through a micropipette to determine the lysis tension. Components of the diagram were created by Z. Zhuoyi Xu and using BioRender. b, Representative images of GPMVs with high or low NINJ1–mCherry fluorescence levels collected during the aspiration protocol. GPI–eGFP was used as a membrane marker. Scale bar, 10 µm. c, Traces showing the volume changes in NINJ1^high and NINJ1^low GPMVs during the pressure ramping. d, Lysis tension of the GPMVs as a function of membrane NINJ1–mCherry fluorescence. GPMVs were generated from HeLa cells overexpressing WT NINJ1–mCherry (n = 30 vesicles), K45Q mutant NINJ1-mCherry (n = 8 vesicles) or GPI–eGFP alone (n = 10 vesicles) across three independent experiments. Unless otherwise indicated, data are mean ± s.e.m. Error bars that were too small, or out of the logarithmic scale, are not displayed. Source data

Fig. 4. NINJ1 promotes PMR under mechanical strain during lytic cell death.

a, Representative images of THP-1 cells treated with 5 μg ml⁻¹ nigericin (Nig). Plasma membranes were labelled with CellBrite (red); DRAQ7 (yellow) marks nuclei of permeabilized cells. b, Time-course of LDH release, the percentage of DRAQ7⁺ cells and the percentage of ballooned cells. c, Images of parental and NINJ1-KO THP-1 cells 6 h after nigericin treatment. Permeabilized nuclei (DRAQ7⁺Hoechst⁺) appeared white; intact nuclei (Hoechst⁺ only) appeared blue. d, Time-course of LDH release, DRAQ7⁺ cells and ballooned cells in parental and NINJ1-KO cells. e, The workflow of the experiment testing the effect of flow on PMR. The diagram was created using BioRender. f, Shear rates in various human vascular beds as previously reported^,. g,h, Images (g) and quantification (h) of pyroptotic THP-1 cells after flow stimulation for 30 min. 5 μg ml⁻¹ nigericin was added to cells 6 h before. The arrows in the left and right panels indicate fully ruptured cells (DRAQ7⁺ nuclei lacking surrounding CellBrite staining). i, Images of parental and NINJ1-KO THP-1 cells treated with 5 ng ml⁻¹ nigericin for 6 h, then with flow at 2,073 s⁻¹ for 30 min. Static cultures were used as controls. j, LDH release and the full PMR rate in parental and NINJ1-KO cells after nigericin and flow treatment. k, LDH release and the full PMR rate in parental and NINJ1-KO cells after intracellular LPS and flow treatment. l,m, Quantification of dsDNA released from pyroptotic THP-1 cells treated with nigericin (l) and LPS (m) and flow for 30 min. Data are mean ± s.e.m. from three independent trials per group. Image quantification was done with 250–300 cells per trial. Statistical analysis was performed using two-sided unpaired Student’s t-tests versus the control (b) and two-way ANOVA followed by Bonferroni correction (d and j–m) versus the parental + nigericin or parental + LPS group. For a, c, g and i, scale bars, 25 μm. *P < 0.05. Source data

Extended Data Fig. 1. Characterization of the high-throughput cellular stretch system and the cellular response to mechanical strain.

a, Image of the HT cellular stretch system. b, Finite element analysis of the PDMS membrane under different vacuum levels, showing the strain pattern across the whole well. c, Illustration of the particle imagery assay used to empirically estimate the strain values with the formula shown. d is the distance of the bead from the centre of the well at the resting state, and d’ is the distance of the same bead from the centre in the stretched state. d, Bead images before stretch (grey) and at −30 kPa (magenta) were overlayed. e, Estimated strain values at different vacuum levels from the bead imagery analysis (n = 10 wells from 3 independent experiments). f, YFP intensity traces of HeLa cells subjected to 5 s stretch at a series of strain levels. Each trace is from 150 ~ 200 cells, n = 3 ~ 4 trials per group. g, The inhibition curve of the DCPIB, a non-specific chloride channel inhibitor, on the quenching of YFP induced by 40% strain. Each data point is from ~200 cells of each condition, n = 5 trials per group. h, Inhibition curve of the DCPIB on the quenching induced by 50% strain for 5 s. Each data point is from ~220 cells of each condition, n = 5 trials per group. Half-inhibition concentrations (IC₅₀) are indicated. i, YFP quenching over the course of 60 min. Fluorescence baseline was recorded for 10 min, then the following stimulations were applied: 5 µM RSL3, TSZ, and 500 ng ml⁻¹ LLO. Mechanical stimulation was applied for 5 s at 50% and 55% strain. Control was 0.5% DMSO. Each data point is from ~150 cells of each condition, n = 3 ~ 4 trials per group. j, YFP quenching over the course of 60 min under 20 ng ml⁻¹ nigercin or mechanical strain at 50% or 55% for 5 s. Each data point is from ~150 cells of each condition, n = 3 ~ 4 trials per group. Unless otherwise noted, all data are presented as mean ± s.e.m. Source data

Extended Data Fig. 2. Assessment of the role of PIEZO1 and PIEZO2 in strain-induced plasma membrane rupture.

a, LDH release of PIEZO1, PIEZO2 knock-down and control HeLa cells. n = 4–8 trials per group. b, Percentage of DRAQ7⁺ cells of PIEZO1, PIEZO2 knock-down and control HeLa cells. n = 4–8 trials per group from ~150 cells of each group. 2-way ANOVA followed by Bonferroni corrections. n.s., not significant versus Scrambled. c, d, LDH release and the percentage of DRAQ7⁺ cells from the HEK-293T transiently-transfected with human PIEZO1 and PIEZO2. Vector-transfected cells, as well as TRE3G-NINJ1 HEK-293T stable cells under 100 ng ml⁻¹ Dox induction were used as controls. Assay was conducted 24 h after transfection. n = 3 trials for each group, 2-way ANOVA followed by Bonferroni corrections. ** p <0.01 versus Vector at the same strain level. e, The lysis tension of PIEZO1 and NINJ1-expressing vesicles. Owing to the drastically larger size of PIEZO1 (38 transmembrane segments, 287 kDa, compared to 3 transmembrane and 16 kDa for NINJ1), protein levels were derived from fluorescence intensity of mCherry fused to C-terminal of PIEZO1, corrected for number of transmembrane segments to reflect the footprint, and plotted as size-corrected molecular density on the membrane in arbitrary units (A.U.) for a fair comparison with NINJ1. 30 vesicles harbouring WT NINJ1 and 13 vesicles with PIEZO1 were measured. 10 additional vesicles that only overexpress GPI-eGFP were included as the zero intensity points for both NINJ1 and PIEZO1. Unless otherwise noted, all data are presented as mean ± s.e.m. Source data

Extended Data Fig. 3. Validation of NINJ1-deficient animals, primary cells and THP-1 cells.

a, Mortality rate of the Ninj1^−/− mice within 2 months from birth. b, Relative Ninj1 mRNA levels from the spleen of WT and Ninj1^−/− mice. n = 4 animals per genotype. Two-sided unpaired Student’s t-test. * p < 0.05 versus WT. c, d, The LDH release and the percentage of DRAQ7⁺ primary BMDMs from Ninj1^−/− mice and WT littermates after the application of 5 s stretch at 45% strain. Each data point is from 100 ~ 200 cells in one of the stretched wells. For each genotype, primary BMDMs were isolated and pooled from 4 animals. 2-way ANOVA followed by Bonferroni corrections. * p < 0.05, ** p < 0.01 vs WT. Unless otherwise noted, all data are presented as mean ± s.e.m. e, Sanger sequencing results of the NINJ1 KO THP-1 cells and the parental cells. The KO cells have 38 bp missing starting from base pair 202, causing a frameshift. f, Western blot of endogenous NINJ1 in the NINJ1 KO and the parental cells. GAPDH was used as the control for protein loading. Experiments have been repeated at least 3 times independently with similar results. Source data

Extended Data Fig. 4. Assessment of ectopic and endogenous NINJ1 expression in various cell lines.

a, Relative NINJ1 mRNA level of vector- and CMV-NINJ1-IRES-mCherry-transfected HEK-293T cells 24 h post-transfection. n = 3 for each group. Two-sided unpaired Student’s t-test, ** p < 0.01 vs Vector. b, Representative image of HEK cells overexpressing NINJ1-mCherry fusion under the CMV promoter 24 h post-transfection. Fluorescent puncta were visible (arrowheads) and plasma membrane ballooning were evident in some cells (arrows). NINJ1 was visualized with direct mCherry fluorescence. Nuclei were live stained with Hoechst. Scale bar, 25 µm. c, Relative NINJ1 mRNA levels of HEK-293T parental cells and the cells stably expressing NINJ1 driven by TRE3G promoter. mRNAs were measured 24 h after induction by Doxycycline (Dox) of various doses. n = 4 trials per group. One-way ANOVA followed by Bonferroni corrections. ** p <0.01 versus Parental. d, Western blot of endogenous NINJ1 from HeLa and ZOS cells. 3 replicates were done for each cell line. e, Relative NINJ1 mRNA level of HeLa and ZOS cells. 3 trials were conducted for each group. Two-sided unpaired Student’s t-test, ** p <0.01 versus HeLa. Unless otherwise noted, all data are presented as mean ± s.e.m. Source data

Extended Data Fig. 5. Assessment of cytoskeleton organization and lipid composition in TRE3G-NINJ1-IRES-mCherry stable HEK-293T cells.

a-b, Cells were treated Dox at various doses. Where applicable, mCherry fluorescence intensity were used to assess the relative expression levels of NINJ1. Quantification data are from cells from 3 separate fields for each group using one-way ANOVA followed by Bonferroni corrections. a, Representative images and quantitation of immunofluorescence staining of F-actin in cells express different levels of NINJ1. b, Representative images and quantitation of β-tubulin staining. Average intensity, tubule lengths and tubule count per cells were quantified. c-d, TRE3G-NINJ1-IRES-mCherry stable HEK-293T cells were transfected with Dox-inducible fluorescent probes of phosphatidic acid (PA) and phosphatidylserine (PS), then incubated with 300 ng ml⁻¹ Dox to induce NINJ1 and lipid probe expression. Quantification data are from cells from 3 separate fields for each group and analysed by one-way ANOVA followed by Bonferroni corrections. c, Representative images and quantitation of average intensity of PA-GFP. d, Representative images and average intensity of PS-GFP. e, TRE3G-NINJ1-IRES-mCherry stable HEK-293T cells were incubated with Dox at various doses to induce NINJ1 expression, then stained for Annexin V and quantified for Annexin V intensity (n = 6 ~ 15 wells from 3 independent experiments, one-way ANOVA followed by Bonferroni corrections). f, Validation of Annexin V antibodies was carried out by staining cells treated with nigericin to induce cell death (n = 10 ~ 11 wells from 3 independent experiments, two-sided unpaired Student’s t-test, ** p <0.01 versus control). Unless otherwise noted, all data are presented as mean ± s.e.m. All scale bars are 25 µm. Source data

Extended Data Fig. 6. Evaluation of NINJ1 K45Q mutant’s effect on plasma membrane rupture under mechanical strain.

a, The relative mRNA levels of NINJ1 WT and NINJ1 K45Q in overexpressed in HeLa cells. 3 replicates were done for each group. n = 3 trials per group. One-way ANOVA followed by Bonferroni corrections. ** p <0.01 versus vector. b, c, LDH release and the percentage of DRAQ7⁺ HeLa cells overexpressing vector control, NINJ1 WT and NINJ1 K45Q, after mechanical stimulation at various strain levels. n = 3 ~ 6 trials each group. 2-way ANOVA followed by Bonferroni corrections. * p < 0.05, ** p <0.01 versus vector at the same strain level. All data are presented as mean ± s.e.m. Source data

Extended Data Fig. 7. The pyroptotic THP-1 cells maintain the ballooned morphology in the absence of mechanical stimulation.

a, Western blot of endogenous NINJ1 of THP-1 cells on a native BN-PAGE gel with or without incubation with 5 µg ml⁻¹ Nigericin for 90 min. b, THP-1 cells were incubated with 5 µg ml⁻¹ Nigericin and stained for plasma membrane (CellBrite, red), nuclei of cells with compromised membrane (DRAQ7, yellow) and all nuclei (Hoechst, blue) 48 h later. Cells with permeabilized plasma membrane showed DRAQ7/Hoechst double staining, represented in white. Scale bar, 50 µm. These experiments have been repeated at least 3 times independently with similar results.

Extended Data Fig. 8. NINJ1 facilitates full PMR under mechanical strain in intracellular LPS induced cell death.

a, NINJ1 KO and parental THP-1 cells were electroporated with LPS and stained with CellBrite (red), DRAQ7 (yellow) and Hoechst (blue) at 1 h, 8 h, 16 h and 24 h after LPS treatment. Almost all cells showed DRAQ7 staining, indicating large breaches on the plasma membrane, yet still displayed intact ballooned shape without full rupture. b, Quantification of LDH release, percentage of DRAQ7⁺ cells and percentage of ballooned cells at various time points after electroporation of LPS. n = 3 for each group. c, NINJ1 KO and parental THP-1 cells were electroporated with LPS 1 h before being subjected to flow treatment at shear rates ranging from 829 s⁻¹ to 2073 s⁻¹. A significantly lower percentage of NINJ1 KO cells show DRAQ7⁺/Hoechst⁺ bare nuclei, indicative of full PMR, after being subjected to flow. In all assays, data are from 3 trials. To quantify DRAQ7 staining and ballooned cell percentage, 250 ~ 300 cells examined for each trial. Data were analysed using 2-way ANOVA followed by Bonferroni corrections and presented as mean ± s.e.m. Scale bars, 25 µm. Source data

Extended Data Fig. 9. NINJ1 does not affect full PMR under mechanical strain in LLO induced cell death.

a, NINJ1 KO and parental THP-1 cells were treated with 500 ng ml⁻¹ LLO and stained with CellBrite (red), DRAQ7 (yellow) and Hoechst (blue) at 1 h, 2 h, 8 h, 12 h and 24 h later. b, Quantification of LDH release, percentage of DRAQ7⁺ cells and DRAQ7⁺ cells with bare nuclei at various time points after LLO treatment. n = 3 for each group. Data were analysed using 2-way ANOVA followed by Bonferroni corrections. ** p < 0.01 versus parental+LLO. c, NINJ1 KO and parental THP-1 cells were treated with LLO for 1 h before flow stimulation at the shear rate 829 s⁻¹. Cells were stained after 30 min as previously described. d, Cells were treated with 500 ng ml⁻¹ LLO for 1 h, then subjected to flow at 829 s⁻¹ in shear rate. Measurement of the LDH release, percentage of fully ruptured parental and KO cells were carried after 30 min. Data are from 3 trials. To quantify DRAQ7 staining and ballooned cell percentage, 250 ~ 300 cells examined for each trial, and analysed using two-sided unpaired Student’s t-test, n.s., not significant versus parental. All data are presented as mean ± s.e.m. Scale bars, 25 µm. Source data

Extended Data Fig. 10. NINJ1 facilitates full PMR under osmotic stress in Nigericin-induced pyroptosis.

a, Representative images of NINJ1 KO and parental THP-1 cells subjected to a series of hypotonic (215, 240 and 260 mOsm) and isotonic (310 mOsm) stimulation. Cells were stained with CellBrite (red), DRAQ7 (yellow) and Hoechst (blue). b, Pyroptosis was induced by 20 µg ml⁻¹ nigericin in NINJ1 KO and parental THP-1 cells in hypotonic and isotonic buffers for 2 h. A substantial percentage of cells showed DRAQ7 and Hoechst double staining, indicating large gaps or openings on the plasma membrane, yet most cells still displayed intact ballooned shape without full rupture. Scale bar, 25 µm. c-e, Quantification of LDH release, the percentage of DRAQ7⁺ cells and the DRAQ7⁺ bare nuclei in hypotonic and isotonic conditions. n = 3 trials per group. Data were analysed using 2-way ANOVA followed by Bonferroni corrections and presented as mean ± s.e.m. * p < 0.05, ** p < 0.01 vs parental. Source data