A pathway coordinated by DELE1 relays mitochondrial stress to the cytosol

Abstract

Mitochondrial fidelity is tightly linked to overall cellular homeostasis and is compromised in ageing and various pathologies^1-3. Mitochondrial malfunction needs to be relayed to the cytosol, where an integrated stress response is triggered by the phosphorylation of eukaryotic translation initiation factor 2α (eIF2α) in mammalian cells^4,5. eIF2α phosphorylation is mediated by the four eIF2α kinases GCN2, HRI, PERK and PKR, which are activated by diverse types of cellular stress⁶. However, the machinery that communicates mitochondrial perturbation to the cytosol to trigger the integrated stress response remains unknown^1,2,7. Here we combine genome engineering and haploid genetics to unbiasedly identify genes that affect the induction of C/EBP homologous protein (CHOP), a key factor in the integrated stress response. We show that the mitochondrial protease OMA1 and the poorly characterized protein DELE1, together with HRI, constitute the missing pathway that is triggered by mitochondrial stress. Mechanistically, stress-induced activation of OMA1 causes DELE1 to be cleaved into a short form that accumulates in the cytosol, where it binds to and activates HRI via its C-terminal portion. Obstruction of this pathway can be beneficial or adverse depending on the type of mitochondrial perturbation. In addition to the core pathway components, our comparative genetic screening strategy identifies a suite of additional regulators. Together, these findings could be used to inform future strategies to modulate the cellular response to mitochondrial dysfunction in the context of human disease.

Extended Data Fig. 1. CHOP and CHOP^Neon protein levels in the context of different types of pharmacological stimulation.

a, Wild-type HAP1 cells were treated as indicated for 9 h and analysed by immunoblotting. TM, tunicamycin. b, Schematic depicting the expected cellular activities of CCCP, tunicamycin and CDDO. ER, endoplasmic reticulum; LONP, LON protease 1; ΔΨ, mitochondrial membrane potential. c, CRISPR engineering of the DDIT3 locus, resulting in an endogenous in-frame fusion of CHOP with a triple Flag tag followed by mNeon as indicated. d, Pharmacological stimulation of wild-type HAP1 CHOP^Neon cells for 9 h (tunicamycin and CDDO) or 16 h (CCCP) leads to induction of the CHOP^Neonprotein, which was measured by flow cytometry (one representative experiment shown of three independent experiments). e, Schematic depicting the generation of CHOP^Neon cells and their interrogation in phenotypic genetic screening after exposure to cellular stress. Mutagenized cells were sorted on the basis of mNeon intensity (Neon^Lo and Neon^Hi populations) and gene-trap mutations in these populations were analysed by deep sequencing. f, Regulators of CHOP^Neon in the genome-wide screen using CDDO (n = 1.90 × 10⁷ interrogated single cells). Genes are coloured as in Fig. 1a, b, with dark grey denoting significant enrichment for mutations (two-sided Fisher’s exact test, FDR-corrected P value (P _adj) < 0.05). The two other eIF2α kinases GCN2 (encoded by EIF2AK4) and PKR (encoded by EIF2AK2) are also highlighted in black. g, HAP1 cells were exposed to sgRNAs that target one of the four eIF2α kinases, pharmacologically stimulated for 9 h as indicated and assayed for CHOP induction by immunoblotting (one representative experiment shown of two independent experiments). h–m, Common regulators of CHOP^Neon in the genome-wide screens using CCCP, tunicamycin or CDDO. Data are from Fig. 1a, b and panel f, with the statistical significance of mutation enrichment assessed and visualized in the same way. Genes that were identified as significant positive regulators (P _adj < 0.05) with a mutation ratio lower than 0.25 (h–j) or negative regulators with a mutation ratio higher than 4 (k–m) and which are shared across all three datasets are highlighted in red. WHSC2 is also known as NELFA.

Extended Data Fig. 2. Unique negative regulators of CHOP^Neon across genetic screens.

a–c, Data from Fig. 1a, b and Extended Data Fig. 1f, with statistical significance of mutation enrichment assessed and visualized in the same way. Genes that were identified as significant negative regulators (two-sided Fisher’s exact test, P _adj < 0.05) and which had a mutation ratio higher than 4 in the query dataset but not in either of the other two datasets are highlighted in red: CCCP versus tunicamycin and CDDO (a); tunicamycin versus CDDO and CCCP (b); CDDO versus CCCP and tunicamycin (c). ATP5B is also known as ATP5F1B; ATP5F1 as ATP5PB; ATP5O as ATP5PO; GNB2L1 as RACK1; C2or f29 as CNOT11; HIST1H4D as H4C4; and ICT1 as MRPL58.

Extended Data Fig. 3. Unique positive regulators of CHOP^Neon across genetic screens.

a–c, Data from Fig. 1a, b and Extended Data Fig. 1f, with statistical significance of mutation enrichment assessed and visualized in the same way. Genes that were identified as significant positive regulators (two-sided Fisher’s exact test, P _adj < 0.05) and which had a mutation ratio lower than 0.25 in the query dataset but not in either of the other two datasets are highlighted in red: CCCP versus tunicamycin and CDDO (a); tunicamycin versus CDDO and CCCP (b); CDDO versus CCCP and tunicamycin (c). ATPIF1 is also known as ATP5IF1; KIAA0141 is also known as DELE1; KIAA0664 is also known as CLUH; TH1L is also known as NELFCD; C11orf73 is also known as HIKESHI; C22orf29 is also known as RTL10; H2AFZ is also known as H2AZ1.

Extended Data Fig. 4. Overview of positive and negative regulators of CHOP^Neonafter treatment with different drugs.

a–c, Data from Fig. 1a, b and Extended Data Fig. 1f, with statistical significance of mutation enrichment assessed and visualized in the same way. For all datasets, significant positive regulators (two-sided Fisher’s exact test, P _adj < 0.05) with a mutation ratio lower than 0.25 (corresponding to an effect size of 4× or greater) are shown in blue, whereas significant negative regulators (two-sided Fisher’s exact test, P _adj < 0.05) with a mutation ratio higher than 4 are shown in yellow.

Extended Data Fig. 5. Intersection of the most pronounced positive regulators of CHOP^Neon across genetic screens and validation of DELE1 and HRI as mediators of mitochondrial stress.

a, Data from Fig. 1a, b and Extended Data Fig. 1f, processed in the same way, except that genes that were identified as highly significant positive regulators (two-sided Fisher’s exact test, P _adj < 10⁻⁵) with a mutation ratio lower than 0.25 in the CCCP dataset but not in either of the other two datasets are highlighted in red. b, Venn diagram showing the results of stringent filtering of positive CHOP^Neon regulators from a (two-sided Fisher’s exact test, P _adj < 10⁻⁵ and mutation ratio lower than 0.25) as intersections between all three of the genetic screens that were performed in this study. CCDC101 is also known as SGF29. c, HeLa cells, 293T cells, BJEH fibroblasts and N/TERT-1 keratinocytes were exposed to sgRNAs directed against the specified genes and pharmacologically stimulated for 9– 12 h before immunoblotting (one representative experiment shown of n = 2 (HeLa and 293T) or n = 3 (BJEH and N/TERT-1) independent experiments). sgCTR, non-targeting control sgRNA. d, Clonal HAP1 knockout and stably reconstituted cells were treated as indicated (CCCP, tunicamycin and oligomycin (OM), 9 h; CDDO and GTPP, 11 h) and analysed by immunoblotting. e, HAP1 CHOP^Neon cells of the indicated genotypes were treated for 9 h (CCCP, tunicamycin, CDDO) or 12 h (GTPP) and analysed by flow cytometry. Per genotype and treatment, the CHOP^Neonsignal was normalized to its DMSO control and statistical significance is indicated compared to identically treated wild-type cells (mean ± s.d. of n = 3 independent experiments; one-way ANOVA with Dunnett’s multiple comparisons correction). f, Wild-type, DELE1-deficient (clones #3 and #6) and stably reconstituted HAP1 CHOP^Neon cells were transduced with lentiviral constructs that contain a non-targeting control sgRNA or an sgRNA directed against LONP1. CHOP^Neon levels were analysed by flow cytometry. Per genotype, the CHOP^Neon fluorescence detected in sgLONP1-treated cells is shown relative to the CHOP^Neon fluorescence detected in the matching sgCTR-treated cells (n = 2 independent experiments). g, HeLa, 293T and HCT116 cells—either wild type or that were transiently exposed to an sgRNA directed against EIF2AK1 (HRI) or DELE1—were stimulated with CCCP for 1 h and the phosphorylation of eIF2α was analysed by immunoblotting (one representative experiment shown of two independent experiments).

Extended Data Fig. 6. DELE1 operates upstream of HRI, is cleaved and accumulates in the cytosol across different types of mitochondrial perturbation and different cellular systems.

a, b, HeLa wild-type, knockout and stably reconstituted cells were treated for 1 h (phosphorylated eIF2α) or 9 h (CHOP) with CCCP. Cells were analysed by immunoblotting (a) and the levels of phosphorylated eIF2α were quantified relative to ACTB (b; mean ± s.d. of n = 5 independent experiments; for each genotype, CCCP treatment was compared to DMSO using a paired two-tailed Student’s t-test). c, d, Subcellular localization of HA-tagged DELE1 or DELE1(ΔMTS) in HeLa cells analysed by confocal microscopy. TRAP1 staining identifies mitochondria and nuclei were visualized with DAPI. e, Top, clonal DELE1-knockout HAP1 CHOP^Neon cells were transiently transfected with full-length DELE1 or truncation mutants that lack the first 101 (DELE1(ΔMTS(N101))) or 115 (DELE1(ΔMTS(N115))) amino acids as indicated. Bottom, transfected cells were treated with CCCP or DMSO for 9 h and analysed by flow cytometry as in Fig. 2a (mean ± s.d. of n = 3 independent biological samples; two-way ANOVA with Tukey’s multiple comparisons correction; one representative experiment shown of three independent experiments). f, HeLa cells were transfected with DELE1-HA, treated with CCCP alone or CCCP and cycloheximide (CHX) as indicated and analysed for subcellular localization of DELE1. Mitochondria were stained with MitoTracker and nuclei were visualized using DAPI. g, CRISPR engineering of the DELE1 locus in 293T cells (introducing a C-terminal in-frame fusion with a triple HA tag), followed by treatment with CCCP alone or CCCP and cycloheximide and analysis by immunoblotting (one representative experiment shown of four independent experiments). h, Wild-type and HRI-knockout HAP1 cells were transfected with DELE1-HA and treated as indicated. The fate of the DELE1–HA protein was analysed by immunoblotting (n = 5 (HRI) or 2 (wild type) biologically independent clones). i, j, HCT116 (i) and HAP1 (j) cells were transfected with DELE1-HA, treated with the specified compounds and analysed for DELE1– HA cleavage and CHOP induction by immunoblotting (one representative experiment shown of two independent experiments). k, N/TERT-1 keratinocytes, hTERT⁺ (BJEH) fibroblasts and primary fibroblasts were transduced with DELE1–HA, stimulated for 4 h as indicated and assayed by immunoblotting. l, Wild-type 293T and HAP1 cells were transfected with DELE1-HA and treated with CCCP for 4 h. The localization of DELE1–HA was analysed as in Fig. 2d (one representative experiment shown of two independent experiments). m–o, Cells from k were stimulated as indicated for 4 h and the localization of DELE1–HA and MFN2 were analysed by confocal microscopy. TRAP1 staining (m), MitoTracker (n, o) and DAPI (o) were used to visualize cellular substructures (one representative experiment shown of two independent experiments). Scale bars, 10 μm (c, d, f, l–o).

Extended Data Fig. 7. Processing of full-length DELE1 into S-DELE1 is mediated by catalytically active OMA1.

a, Data from Fig. 1a, with statistical significance of mutation enrichment assessed and visualized in the same way (two-sided Fisher’s exact test, P _adj < 0.05). Known mitochondrial proteases^, are highlighted in red. XRCC6BP1 is also known as ATP23. b, Wild-type HAP1 cells and clonal OMA1 knockout cells were transfected with DELE1-HA, treated with CCCP and assayed by immunoblotting. c, HeLa cells of the stated genotypes were transfected with DELE1-HA and pre-treated with the indicated protease inhibitors for 5 h, followed by the addition of CCCP for 2 h. The fates of DELE1, OMA1 and OPA1 were monitored by immunoblotting. DCI, 3,4-dichloroisocoumarin; pepstat. A, pepstatin A; ophen, o-phenanthroline. d, Quantification of c. The fraction of S-DELE1 (versus L-DELE1) is indicated. Statistical significance was assessed using oneway ANOVA with Tukey’s multiple comparisons correction and represents the significance relative to the wild type plus CCCP condition (mean ± s.d. of n = 3 independent experiments). e, HeLa cells transiently exposed to an sgRNA directed against OMA1 were co-transfected with the indicated cDNAs and cleavage of DELE1–Flag–mNeon (FmN) after 2 h of CCCP treatment was monitored by immunoblotting (one representative experiment shown of two independent experiments). f, Workflow of the in vitro assay for DELE1 fate upon mitochondrial depolarization using mitochondria purified from wild-type or OMA1-deficient 293T cells that were transiently transfected with DELE1-HA. g, h, Purity analysis of isolated mitochondria using electron microscopy (g) and immunoblotting (h) (one representative experiment shown of two independent experiments for each). For immunoblotting, around 1% of the mitochondrial and 0.2% of the cytosolic fraction was used. The electron micrographs show a high proportion and purity of mitochondria, only slight contaminations of membrane fragments or other cell organelles and the complete absence of intact cells within the mitochondrial suspensions. Scale bars, 5 μm (left) or 1 μm (right). i–k, Isolated mitochondria from g, h were treated in vitro with FCCP and analysed for DELE1 processing. i, Mitochondria build up a membrane potential that remains stable for one hour (demonstrated by constant low fluorescence). FCCP (500 nM) was used to induce dissipation of the mitochondrial membrane potential, yielding an increase in fluorescence. j, After one hour of FCCP treatment, the mitochondria and supernatant were separated by centrifugation and the protein levels in both compartments were analysed by immunoblotting (top, long exposure; bottom, short exposure for DELE1–HA) k, The amount of S-DELE1 in the supernatants was quantified and compared to the wild type plus DMSO condition using two-way ANOVA with Tukey’s multiple comparisons correction. In i, k, data are mean ± s.d. of n = 5 independent experiments; j shows one representative experiment of n = 5 independent experiments, of which the purity analysis shown in g, h was performed for two independent experiments.

Extended Data Fig. 8. DELE1 cleavage and subcellular redistribution after CCCP treatment depend on OMA1 but not OPA1.

a–c, HeLa cells were treated with sgRNAs directed against OPA1, stimulated with CCCP and analysed as indicated together with wild-type HeLa cells by immunoblotting (a). OPA1 levels (b) and DELE1 cleavage (c) were quantified. Statistical significance was assessed using one-way ANOVA with Tukey’s multiple comparisons correction and represents the significance relative to the wild type (mean ± s.d. of n = 4 biologically independent samples). d, HAP1 cells were treated and analysed as in a (one representative experiment shown of two independent experiments). e, Data from Fig. 1a, with statistical significance of mutation enrichment assessed and visualized in the same way (two-sided Fisher’s exact test, P _adj < 0.05). OPA1 is highlighted in red. f, HeLa cells were treated with an sgRNA directed against OPA1, stimulated with CCCP and analysed for DELE1 localization and OPA1 status, alongside wild-type and OMA1-knockout cells. Arrowheads highlight OPA1-deficient cells. One representative experiment shown of two independent experiments. Scale bars, 10 μm.

Extended Data Fig. 9. Physical interaction between DELE1 and HRI.

a, 293T cells were co-transfected with StrepTagII-HA-HRI and the indicated Flag-mNeon-tagged cDNAs before stimulation with CCCP as indicated. Cells were then lysed and StrepTagII-containing protein complexes were immobilized on Strep-Tactin beads and analysed by immunoblotting along with the input lysate. DELE1^FL, full-length DELE1. b, A C-terminal in-frame fusion of a triple Flag tag was introduced into the endogenous EIF2AK1 (HRI) locus of 293T cells using CRISPR. c, Quantification of Fig. 3b. The bar chart shows the relative abundance of the slowly migrating species of HRI (phosphorylated HRI) compared to the faster migrating band. Data were normalized to the empty vector (black) and significance (compared to full-length DELE1 untreated, grey) was assessed using one-way ANOVA with Dunnett’s multiple comparisons correction (mean ± s.d. of n = 3 independent experiments). d, 293T cells were co-transfected with HRI-Flag and DELE1-HA and exposed to CCCP for the denoted amounts of time before immunoblotting. The shift in electrophoretic mobility of HRI upon activation is reversed by treatment of the lysates with alkaline phosphatase. e, 293T cells of the indicated genotypes were co-transfected with HRI-Flag and the specified HA-tagged constructs. After CCCP treatment, lysates were immobilized on anti-Flag beads and analysed by immunoblotting along with the input lysate (one representative experiment shown of four independent experiments). f, Quantification of e(mean ± s.d. of n = 4 independent experiments; two-way ANOVA with Tukey’s multiple comparisons correction). g, HeLa cells were transiently transfected with the indicated constructs, treated with CCCP or DMSO and analysed for subcellular localization of the transfected cDNAs by HA staining and confocal microscopy with cellular structures labelled as indicated. h, DELE1 mutants were transiently transfected into HeLa and 293T cells before exposure to CCCP for 2 h. The processing of DELE1–HA was monitored by immunoblotting. i, HeLa cells were transfected as indicated and analysed for subcellular localization as in g. j, 293T cells were co-transfected with HRI-Flag and the indicated StrepTagII-HA cDNAs. Cells were then lysed and StrepTagII-containing protein complexes were immobilized on Strep-Tactin beads and analysed by immunoblotting along with the input lysate. k, l, Clonal DELE1-knockout HAP1 cells were transiently transfected with the specified cDNAs together with mCherry and induction of CHOP^Neon was monitored by flow cytometry as in Fig. 2a. Per genotype, data were normalized to the empty vector and statistical significance (compared to DELE1(ΔMTS(N101))) was assessed using one-way ANOVA with Tukey’s multiple comparisons correction (mean ± s.d. of n = 4 independent experiments). m, Wild-type 293T cells were transiently transfected with HRI-Flag and the indicated HA-tagged cDNAs and processed as in e (one representative experiment shown of four independent experiments). n, DELE1-deficient HAP1 CHOP^Neon cells were transiently transfected as indicated and analysed by flow cytometry 32 h after transfection. Per genotype, data were normalized to the empty vector (black) and statistical significance (compared to DELE(ΔMTS)) was assessed using one-way ANOVA with Tukey’s multiple comparisons correction (mean ± s.d. of n = 3 independent experiments). o, Cell lysates from Fig. 3c were analysed by Coomassie staining before (lysate) and after exposure to GFP-TRAP beads (IP (GFP)). Scale bars, 10 μm (g, i).

Extended Data Fig. 10. Effects of haem and mitochondrial membrane potential on the OMA1–DELE1–HRI pathway.

a, Schematic depiction of haem metabolism. b, Data from Fig. 1a, with statistical significance of mutation enrichment assessed and visualized in the same way (two-sided Fisher’s exact test, P _adj < 0.05). Haem-related genes from a are highlighted in red. c, Immunoprecipitation experiment as in Extended Data Fig. 9a in the presence of haemin during cell culture (20 μM for 12 h before lysis) and immunoprecipitation (8 μM in lysis and wash buffers). HMOX1 is induced in response to increased haem levels (one representative experiment shown of two independent experiments). d, Model of the pathway identified here. Mitochondrial stress activates OMA1, which leads to cleavage of DELE1 and its subsequent cytosolic binding to and activation of HRI, triggering the ISR. e, Wild-type HAP1 and HeLa cells were treated with CCCP, and the mitochondrial membrane potential was visualized by TMRM using a live-cell-imaging system. H2B-GFP was transfected to stain nuclei. Scale bars, 100 μm. f, g, HAP1 CHOP^Neon cells were stimulated as indicated, and CHOP^Neon levels (g) and the mitochondrial membrane potential (f) were quantified by flow cytometry. Statistical significance is depicted relative to DMSO-treated control cells (mean ± s.d. of n = 3 independent experiments; one-way ANOVA with Tukey’s multiple comparisons correction). h, HeLa cells were transfected with the indicated sgRNAs and the mitochondrial membrane potential was quantified after 1 h of CCCP treatment by flow cytometry. The TMRM signal was normalized to the non-targeting control sgRNA and DMSO control, and for CCCP-treated samples statistical significance was assessed relative to the non-targeting control sgRNA and CCCP (mean ± s.d. of n = 3 independent experiments; two-way ANOVA with Dunnett’s multiple comparisons correction). i, Volcano plots of RNA-seq analysis of the indicated samples. For every gene, the fold change in expression after CCCP treatment (versus DMSO treatment; x axis) is plotted against the P value (two-tailed Wald test in DESeq2; y axis). Genes with a significant differential expression (P _adj < 10⁻¹² (Benjamini–Hochberg)) are depicted in dark grey. ATF4 and CHOP target genes^, (green) and heat-shock proteins (red) from Fig. 3e are highlighted (n = 4 biologically independent samples per group). FAM129A is also known as NIBAN1.

Extended Data Fig. 11. Cellular fitness and mitochondrial function in cells in which the DELE1–HRI–eIF2α axis is perturbed.

a, HAP1 cells were infected with lentivirus expressing TagRFP and the indicated sgRNAs. A 1:1 mixture of infected and uninfected cells was treated for 48 h with 5 μM CCCP or DMSO and analysed by flow cytometry. The abundance of RFP⁺ sgRNA-containing cells after CCCP treatment was compared with the respective DMSO-treated sample bearing the same sgRNA and normalized to a non-targeting control sgRNA. Statistical significance was assessed compared to a second non-targeting control sgRNA using one-way ANOVA with Dunnett’s multiple comparisons correction (mean ± s.d. of n = 5 independent experiments). b, c, HAP1 cells of the indicated genotypes were treated as indicated for 48 h and cell survival was visualized using crystal violet (one representative experiment shown of three independent experiments). d–j, Wild-type and DELE1-knockout HeLa cells stably expressing the indicated cDNAs were subjected to different mitochondrial stressors and the oxygen consumption rate (OCR) was monitored. Within the same clonal background, no significant difference between DELE1-deficient and DELE1-proficient cells could be observed in non-mitochondrial oxygen consumption (e), basal respiration (f), maximal respiration (g), proton leak (h), ATP production (i) and spare respiratory capacity (j). Mean ± s.d. of n = 4 independent culture wells from one representative experiment of three independent experiments is shown; statistical significance was assessed using one-way ANOVA with Tukey’s multiple comparisons correction. k, l, Wild-type, MFN2-knockout and stably reconstituted HAP1 cells were lentivirally transduced with a construct encoding TagRFP together with a non-targeting control sgRNA or an sgRNA directed against EIF2AK1 (HRI) (k) or DELE1 (l) and the fraction of RFP⁺ cells was monitored over time. The abundance of sgDELE1 or sgHRI-containing cells was normalized to non-targeting sgRNA and to time 0 (mean ± s.d. of n = 3 biologically independent, separately infected and cultured wells per sgRNA and genotype; one-way ANOVA with Dunnett’s multiple comparisons correction; one representative experiment shown of two independent experiments). m, SH-SY5Y cells were exposed to the indicated sgRNAs and treated as in Extended Data Fig. 5c (one representative experiment shown of four independent experiments). n, o, Processing and cellular localization of DELE1–HA in SH-SY5Y cells, as in Extended Data Fig. 6k, o. MFN2–Flag–mNeon served as a control (one representative experiment shown of two independent experiments). Scale bars, 10 μm.

Fig. 1. Genome-wide perturbation profiling reveals regulators of CHOP in the context of different triggers of the ISR.

a, b, Genetic screens showing positive and negative regulators of CHOP^Neon in cells that were treated with CCCP (a) or tunicamycin (b). Per gene (dots), the ratio of the mutation frequency in cells with high CHOP^Neon expression versus cells with low CHOP^Neon expression (y axis) is plotted against the combined number of unique mutations identified in both populations of cells (x axis) (n = 2.12 × 10⁷ (a) or 2.07 × 10⁷ (b) interrogated single cells). Genes enriched for mutations are shown in dark grey (two-sided Fisher’s exact test, false discovery rate (FDR)-corrected P value (P _adj) < 0.05). Red dots specify genes that are highly significant in the CCCP dataset but not in the tunicamycin or CDDO datasets. PERK is also known as EIF2AK3.

Fig. 2. DELE1 operates upstream of HRI and accumulates in the cytosol after OMA1-mediated cleavage.

a, b, HAP1 CHOP^Neon cells of the specified genotypes (ΔDELE1 or ΔHRI), transiently transfected with the indicated cDNAs or empty vector (EV), were treated for 9 h with CCCP or DMSO control and analysed by flow cytometry. The CHOP^Neon fluorescence intensity for each genotype was normalized to the empty vector plus DMSO condition, and compared with the empty vector for each treatment (mean ± s.d. of n = 3 independent biological samples; two-way ANOVA with Tukey’s multiple comparisons correction; one representative experiment shown of six independent experiments). c, Domain architecture of DELE1. d, Confocal microscopy of DELE1–Flag–mNeon and MFN2–Flag–mNeon expressed in HeLa cells. e, HeLa cells (top) or 293T cells (bottom) transfected with haemagglutinin (HA)-tagged DELE1 (DELE1-HA) were treated with CCCP for the indicated time periods before analysis of DELE1 cleavage by HA immunoblotting. L-DELE1 indicates the longer version of DELE1. f, Cells expressing endogenous HA-tagged DELE1 (DELE1^HA) were stimulated as indicated and analysed by immunoblotting. g, Wild-type (WT) and clonal OMA1-knockout (ΔOMA1) cells, reconstituted with the specified cDNAs, were transfected with DELE1-HA, treated with CCCP for the indicated periods and assayed by immunoblotting. h, Wild-type and clonal OMA1-knockout HeLa cells were transfected with the indicated cDNAs, treated with CCCP or DMSO control and assayed by confocal microscopy. DAPI (d) and wheat-germ agglutinin (WGA; h) staining identify the nuclei and cell membranes, respectively; TRAP1 staining identifies mitochondria. Scale bars, 10 μm (d, h).

Fig. 3. DELE1 binds to HRI in the cytosol, and a deficiency in this pathway mimics ISR inhibition and alters fitness in the context of mitochondrial stress.

a, Endogenous HRI was immobilized on beads and its ability to co-precipitate endogenous HA-tagged DELE1 (DELE1^HA) was analysed by immunoblotting. IP, immunoprecipitation; WB, western blot. b, Shift in electrophoretic mobility of endogenous Flag-tagged HRI (EIF2AK1^Flag) after the indicated treatments. c, EGFP or DELE1(TRP1–7)–EGFP were stringently purified from 293T cells that were treated with an sgRNA to deplete HRI and incubated with glutathione S-transferase (GST)-tagged HRI produced in E. coli. Interaction was assessed by Coomassie staining and immunoblotting. GST–HRI co-precipitation is accompanied by its depletion from the corresponding supernatant. d, Model of the pathway. e, Transcriptomic profiling of wild-type HAP1 cells (with or without ISRIB treatment) and knockout mutants after treatment with CCCP for 12 h (n = 4 independent biological samples per group). Differentially expressed genes after CCCP treatment (two-tailed Wald test in DESeq2, P _adj < 10⁻¹², |log_2-transformed fold change| > 1.5) were hierarchically clustered. ATF4 and CHOP target genes are enriched among genes that are strongly induced in cells proficient in DELE1–HRI–eIF2α signalling but not in cells deficient in this pathway (two-sided Fisher’s exact test, P _adj < 2 × 10⁻¹⁹, green box); heat-shock proteins are enriched in the cluster of genes induced in the opposite samples (P _adj < 3 × 10⁻⁷, red box). f, HAP1 CHOP^Neon cells of the specified genotypes (wild-type and DELE1-deficient clones #3 and #6) were treated as indicated for 9 h and analysed by flow cytometry. The CHOP^Neon fluorescence intensity for each genotype was normalized to its DMSO control (red dotted line; mean ± s.d. of n = 3 independent experiments; two-way ANOVA with Tukey’s multiple comparisons correction). g, HAP1 cells of the specified genotypes were treated as indicated for 48 h and surviving cells were stained with crystal violet (n = 2 independent experiments).