Our research complies with all relevant ethical regulations and guidelines. All zebrafish procedures were performed in accordance with the UK Animals (Scientific Procedures) Act with appropriate Home Office Project and Personal animal licences and with local Ethics Committee approval. The studies were performed in accordance with PREPARE and ARRIVE guidelines. Ethical approval was obtained from the University of Cambridge Animal Welfare and Ethical Review Body (AWERB) in accordance with the UK Animals (Scientific Procedures) Act under project licence P173861E7—Protocol 7 for Fig. 2a (survival assay) and Protocol 9 for all other zebrafish data.

ReagentsCell lines

HeLa cells (CCL-2; American Type Culture Collection, CVCL-0030) were cultured in DMEM medium (Merck, D6546) supplemented with 10% fetal bovine serum (Merck, F7524), 2 mM l-glutamine (Merck, G7513), and 100 U ml−1 penicillin and 0.1 mg ml−1 streptomycin (Merck, P0781). SH-SY5Y (ECACC, 85120602) cells were cultured in DMEM-F12 medium (Merck, D6421) supplemented with MEM non-essential amino acids (Thermo Fisher, 11140050), 10% fetal bovine serum, 2 mM l-glutamine and 100 U ml−1 penicillin and 0.1 mg ml−1 streptomycin. The HeLa CRISPR–Cas9 ATG2A/B-double-KO cell line (a gift from N. Mizushima35) was cultured in the same medium as the HeLa cells. The ATG16L1-KO HeLa cells and its wild-type control cell line were made in house. The Beclin 1-KO cell line and its wild-type control were a gift from W. Wei (Peking University, Beijing). All cell lines were maintained at 37 °C and 5% CO2, and were regularly tested for mycoplasma.

Antibodies

The following primary antibodies were used for western blots (working concentration, 1:1,000): rabbit anti-WDR45 (WIPI4; 19194-1-AP) from Proteintech; mouse anti-α-tubulin (9026) and rabbit anti-actin (A2066) from Sigma Aldrich; rabbit anti-ATG2A (MBL Life Science, PD041); rabbit anti-GFP (ab6556), mouse anti-NDUFA9 (ab14713) and rabbit anti-Sec23a (ab137583), rabbit anti-TOMM20 (ab186735), mouse anti-GM130 (ab52649), mouse anti-LAMP1 (ab24170), rabbit anti-KDEL (ab2898) and mouse anti-β III tubulin [2G10] (ab78078) from Abcam; rabbit anti-LC3 (Proteintech, 14600-1-AP); mouse anti-TOMM40 (Santa Cruz Biotechnology, sc-365467); rabbit anti-TMEM41b (Novus Biologicals, NBP1-81552); rabbit anti-calreticulin (12238), rabbit anti-cleaved caspase 3 (9661), rabbit anti-ATG16L1 (8089), rabbit anti-ATG7 (2631) and rabbit anti-MAP2 antibody (8707S) from Cell Signaling Technologies; and rabbit anti-PISD (Atlas Antibodies, HPA031091).

The following primary antibodies were used for immunofluorescence (working concentration, 1:100): mouse anti-TOMM20 (Santa Cruz, sc-17764), mouse anti-calnexin (Abcam, ab112995), rabbit anti-hNIX (Cell Signaling Technologies, 12396), mouse monoclonal antibody to FLAG M2 (Sigma, F1804-200UG), mouse anti-ORP8 (Santa Cruz, sc-134409), rabbit anti-ORP5 (Atlas Antibodies, HPA038712-100UL), rabbit anti-IP3 receptor 1 (D53A5) (Cell Signaling Technologies, 8568) and mouse monoclonal [20B12AF2] anti-VDAC1/Porin + VDAC3 (Abcam, ab14734).

Drugs and probes Drugs

The following drugs were used: Z-VAD-fmk (Promega, G7231), necrostatin-1 (Sigma, N9037), liproxstatin-1 (Sigma, SML1414), Fer-1 (Sigma, SML0583), N-acetyl-l-cysteine (Sigma, A7250), MitoTEMPO (Sigma, SML0737), deferoxamine mesylate salt (Merck, D9533), staurosporine (Sigma, S5921), erastin (Sigma, E7781), 1S,3R-RSL3 (Sigma, SML2234), iFSP1 (Sigma, SML2749), arachidonic acid (Sigma, 10931), FIN56 (Sigma, SML1740), sorafenib (Santa Cruz, sc-220125), inPISD inhibitor (or STK770988; Vitasmlab.biz), mitoquinol (Cayman Chemicals, 89950), CCCP (Sigma, C2759) and ULK1 inhibitor SBI-0206965 (Sigma, SML1540-5MG).

Probes

The following probes were used: BODIPY FL C12 (Invitrogen, D3822), BODIPY581/591C11 (Invitrogen, D3861), MitoPerOx (Abcam, ab146820) and nonyl acridine orange (Thermo Fisher Scientific, A1372).

Constructs and siRNA oligonucleotides

We used empty pEGFP (Clontech), pEGFP-C1-hATG2A (Addgene, 36456), GFP-WIPI4 (ref. 36) and Lact-C2-GFP (Addgene, 22852) plasmids. PLKO.1 transfer plasmids expressing shRNAs targeting human WIPI4 (RHS4533-EG11152) and mouse WIPI4 (RMM4534-EG54636) were purchased from the Horizon TRC Lentiviral shRNA library. Human ATG2A was cloned onto an EF1A lentiviral vector adapted from pMK1253 by removing the Tau.K18 (P301L/V337M) insert. The primers used are listed in Supplementary Table 2. Mitochondrial-only PISD-myc-DDK was made from PISD-myc-DDK isoform d (Origene, RC222868) by mutating three leucine residues, required for lipid droplet localization, to positively charged residues (I80K, I89K, I90R).

CRISPR–Cas9 WIPI4-KO constructs were generated as previously described37. Briefly, pairs of guide RNAs (sgRNAs) nicking WDR45 were designed (Supplementary Table 4) using the CRISPR design tool available at http://www.genome-engineering.org/crispr/. Guide RNAs with overhangs were annealed and cloned into a BbsI-digested pSpCas9n(BB)-2A-GFP (PX458) vector (Addgene, 48138). HeLa cells were transfected with pairs of sgRNAs and cultured for two days before use for cell death assays. Only transient KO of WDR45 was performed in this study and a pair of sgRNAs targeting exon 3 of the WDR45 gene was used.

The scramble siRNA (ON-TARGETplus non-targeting control; D-001810), and siRNAs targeting human WIPI4 (J-019758-10 for general usage and J-019758-12 as oligonucleotide 2) and ORP5 (J-009274-10), ORP8 (J-009508-06) were predesigned. Mitochondrial PISD-specific siRNA oligonucleotides (Oligo 1, 5′-GAGAAGCUGGGAUUGGAGAUU-3′; and Oligo 2, 5′-GCACCGAUGUCUGGGAUUACU-3′) were designed against the coding region specific to mitochondrial isoforms of PISD (transcript variants NM_001326411–NM_001326421)23. The siRNAs targeting human NIX (5′-CCAUAGCUCUCAGUCAGAAUU-3′), ATG2A/B, WIPI2, ATG16L1 and ATG7/10 were reported previously6,38. The oligonucleotides were obtained from Dharmacon-Thermo Scientific.

TechniquesTransfection

For the overexpression experiments, 1 µg complimentary DNA construct per well of a six-well plate was transfected with TransIT-2020 (Mirus, MIR5400). For the knockdown experiments, cells were transfected with either a single or double round of 20–100 nM siRNAs using Lipofectamine RNAiMAX (Invitrogen, 13778). In the single transfections, the cells were incubated with siRNA for 72 h or DNA constructs for (the last) 24 h. For the double transfections, the cells were transfected with 50 or 100 nM siRNA, followed by another transfection with 50 nM siRNA after 24–48 h.

Mutagenesis

GFP–WIPI4 was used as a template to mutate amino acids of the ATG2A binding site (LOOP3)9 to alanine residues; the siRNA (J-019578-12) recognises the sequence without altered amino acid residues. The pEGFP-C1-hATG2A (Addgene, 36456) plasmid was used as a template to generate the mitochondria-localization-deficient mutant (GFP–ATG2A-ΔMLD) according to previously published work18 by truncation, WIPI4 interaction deficient (GFP–ATG2A-YFS) mutant by mutating the three relevant amino acids to alanine residues9 and the lipid transfer-deficient mutant (GFP–ATG2A-LTD) according to published work18 by mutating the nine indicated amino acid residues to aspartates. Mutagenesis was performed using a QuickChange multi site-directed mutagenesis kit (Agilent Technologies, 200515) or Q5 site-directed mutagenesis kit (NEB, E0554) and the primers used are listed in Supplementary Table 2.

Virus packaging for overexpression and knocking down genes in neurons

WIPI4-targeting shRNA or GFP–hATG2A lentivirus were packaged using the third-generation packaging system. HEK293T cells seeded in a 10 cm dish coated with poly-d-lysine hydrobromide (Merck, P6407) were transfected with 3 µg transfer plasmids, 2.5 µg psPax2 and 1.6 µg PMD2.G following the manufacturer’s datasheet for Lipofectamine LTX transfection reagent with plus reagent (Thermo Fisher, 15338100). Growth medium (4.5 ml) without antibiotics was added to the top of the transfection mixture and changed to viral harvest medium (30% FBS + 10 U potassium penicillin and 100 μg streptomycin sulfate per 1 ml growth medium) the next day. Virus was harvested at 40 and 64 h post transfection. The collection was filtered using low-protein-binding Durapore (0.45 μm) filter (Merck, SLHVR33RS) and then centrifuged at 114,000g using SW40Ti rotor for 90 min at 4 °C. After disposing the supernatant, the virus pellet was resuspended in 150 µl medium and the aliquots were stored at −80 °C.

Culture and virus infection of neurons

The maintenance and differentiation of i3 neurons were conducted following published protocols39. Wild-type iPSC cells were seeded in plates coated with Matrigel (Corning, 354277) containing induction medium supplemented with 2.5 µM of the Rho-associated protein kinase (ROCK) inhibitor Y-27632 (Tocris Bioscience, 1254). The medium was replaced with fresh induction medium containing doxycycline, but without Y-27632, for two consecutive days. The desired number of Day 3 neurons were seeded in cortical culture medium39 onto plates coated with poly-l-ornithine (Sigma, P3655) diluted in borate buffer (Thermo Fisher, 28341). Half of the medium was replaced every 3–5 days (96-well plate; 1 × 104 cells per well). The isolation and culture of mouse primary cortical neurons was described previously7. Lentivirus was added to the i3 neurons 15 days after induction and to mouse primary neurons 5 days after in vitro culture. The medium was aspirated and replaced with fresh culture medium the day after infection. Experiments were performed 24–72 h after infection.

Cell cytotoxicity assay

Cell cytotoxicity was measured using an LDH Assay Kit (Abcam, ab65393) according to the manufacturer’s instructions. Briefly, cells were plated on a six-well plate. After the drug treatments, knockdown and/or overexpression, the medium with which cells were incubated for 24 h was used for the assay. The medium was exposed to LDH reaction mix for 30 min and absorbance was measured at both 450 and 650 nm (reference wavelength). Cell cytotoxicity was calculated according to the equation: (test sample absorbance − low control absorbance) ÷ (high control absorbance − low control absorbance). The high/positive control used was cells lysed with lysis buffer and the low/negative control was untreated cells.

Cell viability assay

Cell viability was assessed using AquaBluer (MultiTarget Pharmaceuticals LLC, 6015) according to the manufacturer’s instructions. Briefly, cells were cultured in a 96-well plate for 24 h before the assay following the indicated treatments. The cells were washed twice with PBS and then incubated for 4 h in medium containing AquaBluer. The fluorescence intensities were measured at an excitation wavelength of 540 nm and emission of 590 nm using a TECAN plate reader. Cell viability was calculated based on the formula: (average fluorescence value of the treated cells) ÷ (average fluorescence value of the untreated cells). Results were presented as the percentage of viable cells with respect to the control group.

IncuCyte assay

The IncuCyte S3 incubated live imaging system was also used for monitoring cell death. At least 30 images were taken for each sample and the images were analysed using the IncuCyte 2020 software. For cells that did not express GFP-tagged protein, the CellTox green cytotoxicity assay (Promega, G8743) was used to stain dead cells and the ratio of green area to total area (phase) was used as the measurement of cell death. For cells that expressed GFP-tagged proteins, IncuCyte Cytotox red dye was used for counting dead cells (Sartorius, 4632) and the number of dead cells (counted as red dots) per well was used as the measurement of cell death. In Extended Data Fig. 5c, the number of dead cells was normalized to the number of transfected cells (count of GFP-positive objects) to calibrate to the expression levels of ATG2A WT and ATG2A-mLIR.

Western blot analysis

Cells were washed and lysed with RIPA buffer (50 mM Tris–HCl pH 7.4, 150 mM NaCl, 1% Triton X-100, 0.5% sodium deoxycholatemonohydrate and 0.1% SDS supplemented with protease (cOmplete, mini, EDTA-free Protease I; Merck) and phosphatase (Sigma Aldrich, P5726 and P0044) inhibitors cocktails). The lysates were incubated on ice for 15 min and centrifuged at 16,100g and 4 °C for 10 min to remove debris. The protein concentration was determined using a BCA protein assay kit (Thermo Fisher Scientific, 23225). Blots were probed overnight with the indicated primary antibodies and then with DyLight Fluors-conjugated (Invitrogen) secondary antibodies for 1 h before detection on an infra-red imaging system (LICOR Odyssey system). Densitometric analysis of the immunoblots was performed using IMAGE STUDIO Lite software.

Immunoprecipitation assay

Cells plated in 100 mm dishes were washed twice with PBS and lysed with cold lysis buffer (20 mM Tris–HCl pH 7.4, 100 mM NaCl, 0.5 mM EDTA, 0.5% NP40, and protease and phosphatase inhibitors cocktails). The lysates were incubated on ice for 15 min and isolated by centrifugation at 16,100g and 4 °C for 10 min to remove debris. The supernatants were moved to new tubes. A fraction (5%) of the sample was stored to be used as the input control. The remaining lysate was incubated overnight with primary antibodies or a control IgG antibody (Cell Signaling, 2729S) at 4 °C with gentle agitation. Subsequently, the lysate was mixed with Dynabeads protein A (Life Technologies) and incubated at 4 °C for 2 h. The beads were washed five times with lysis buffer and the immunoprecipitated proteins were eluted and denatured by boiling with 2×Laemmli buffer containing β-mercaptoethanol for 10 min at 100 °C. Proteins were detected by SDS–PAGE. EGFP-tagged proteins were pulled down using GFP-TRAP beads (ChromoTek) according to the manufacturer’s protocol.

Membrane pull down with ATG2A

We identified the lipids associated with endogenous ATG2A that were oxidized using a combination of a pull down and lipid peroxidation assay using BODIPY581/591C11. Cells were cultured in a 100 mm dish and stained with 2.5 µM BODIPY581/591C11 diluted in Hanks’ Balanced Salt Solution medium (Invitrogen) for 15 min. The cells were washed twice with PBS, collected by gently scraping in 1 ml cold PBS and centrifuged at 1,000g and 4 °C for 5 min. The pellets were resuspended in isotonic buffer (10 mM Tris–HCl pH 7.4, 250 mM sucrose and 1 mM EDTA supplemented with protease and phosphatase inhibitors cocktails) without detergents and gently homogenized by passing them 12 times through a 26-gauge needle on ice. The lysates were again centrifuged at 1,000g and 4 °C for 5 min and subsequently incubated overnight with 5 µl anti-ATG2A in 500 µl isotonic buffer at 4 °C with gentle agitation. The lysates were mixed with Dynabeads protein A (Thermo Fisher, 10001D) and incubated at 4 °C for 2 h. The beads were washed five times with PBS, transferred into a chambered coverglass (Invitrogen, 155382) and imaged using a LSM880 Zeiss confocal microscope. Volocity software (PerkinElmer) was used for quantification and analysis of the images. In parallel, proteins on the beads were eluted with 2×Laemmli buffer for 10 min at 100 °C and subjected to western blot analysis to detect proteins immunoprecipitated with the ATG2A.

MDA measurement

MDA concentrations were assessed using a Lipid peroxidation assay kit (Abcam, ab118970) according to the manufacturer’s instructions. One confluent 60–100 mm dish of cells or 30 zebrafish at 5–10 d.p.f. were required for each sample. Cells or fish tissue were lysed with MDA lysis buffer supplemented with butylated hydroxytoluene (provided in the kit) stop solution. The cell lysates were gently homogenized by passing them 12 times through a 27-gauge needle on ice. Before fish homogenization in MDA lysis buffer (provided in the kit), 5 d.p.f. larvae were deyolked by pipetting up and down in calcium-free Ringer’s solution (5 M NaCl, 1 M KCl and 1 M NaHCO3 in double-distilled water) to eliminate lipids in the yolk sac. The deyolked fish were collected after gentle centrifugation and supernatant removal. Deyolking is not required at 10 d.p.f. when lipids from the yolk are largely depleted. Homogenization of fish larvae in MDA lysis buffer supplemented with butylated hydroxytoluene was performed by sonication in the water-bath of a diagenode BIORUPTOR (three cycles of 20 s each). The cell or fish homogenates were centrifuged at 13,000g for 10 min at room temperature, and the supernatants were collected for the MDA assay. A TECAN Spark multimode microplate reader was used for measurements at the excitation and emission wavelengths of 532 and 553 nm, respectively. The values were referenced against a standard curve and calculated as the MDA concentration (nmol µg−1 protein).

Cell fractionation and mitochondria purification

Cells were cultured in a 100 mm dish for cell fractionation and two 140 mm dishes for mitochondrial purification. The cells were washed twice with PBS, centrifuged at 500g for 5 min and homogenized by passing them 12 times through a 26-gauge needle or 20 times through a Dounce homogenizer on ice in isotonic buffer (10 mM Tris–HCl pH 7.4, 250 mM sucrose and 1 mM EDTA containing protease and phosphatase inhibitor cocktails) for cell fractionation or (10 mM HEPES pH 7.4, 70 mM sucrose, 210 mM mannitol and 1 mM EDTA containing protease and phosphatase inhibitor cocktails) for mitochondrial purification. The cell homogenates were centrifuged at 1,500g and 4 °C for 10 min to remove unbroken cells and nuclei. The post-nuclear supernatant was subjected to two sequential centrifugations at 10,500g and 4 °C for 10 min with one wash of the pellets with isotonic buffer in between. The pellets consist of purified mitochondria and the supernatant is the post-mitochondrial fraction.

Phospholipid measurement

Intracellular or mitochondrial PE, PC and PS were assessed using PE, PC and PS assay kits (Abcam, ab241005; Sigma-Aldrich, MAK049; and Abcam, ab273295, respectively) according to the manufacturer’s instructions. Briefly, for the PE and PC assays, cells from a 100 mm dish were used for phospholipids in total cell lysate and two 140 mm dishes were used for mitochondrial phospholipids. The cells were washed twice with PBS and collected into a tube for total cell lysates, whereas mitochondria were purified following the method described above. The pellets were resuspended in a 5% (vol/vol) solution of peroxide-free Triton X-100 in water and the samples were calibrated according to the protein concentration. The samples were heated to 80 °C and cooled to room temperature; this cycle was repeated another two times to fully extract lipids. The lysates were centrifuged at 10,000g and 4 °C for 10 min to remove insoluble structures. For the PS assay, lipids were extracted using a Lipid extraction kit (chloroform free; Abcam, ab211044) according to the manufacturer’s instructions following mitochondrial purification. The lipid film was resuspended in 100 µl of 1% Triton X-100. The fluorescence intensity of lipids stained with the corresponding dyes was measured using a TECAN instrument at 538 nm excitation and 587 nm emission. The levels of PE, PC and PS were calculated according to the equations in the protocols.

Measurement of unsaturated fatty acids

The levels of unsaturated fatty acids were measured using a Lipid assay kit (unsaturated fatty acids; Abcam, ab242305). The same amount of HeLa cells was used for mitochondria extraction and lipid extraction as described earlier. Each lipid sample was resuspended in 60 µl DMSO. We incubated 5 µl of each resuspended sample (recalibrated according to the protein concentration) or standards with 150 μl of 18 M sulfuric acid in a 1.5 ml Eppendorf tube at 90 °C for 10 min and followed the manufacturer’s instructions for the subsequent steps.

Immunofluorescence microscopy of cultured cells

HeLa cells cultured on coverslips were fixed in 4% paraformaldehyde for 5 min. Fluorescence was detected using a LSM880 Zeiss confocal microscope with a 63× oil-immersion lens. Images were acquired using the ZEN Black 2.6 Carl Zeiss Microscopy software. Volocity software (PerkinElmer) was used for co-localization analysis (Pearson’s correlation coefficient). Image J (Fiji) was used to quantify the intensity and area of the fluorescent signals. The cells in Extended Data Fig. 5e were stained live with MitoTracker deep red FM (Thermo Scientific, M22426) for 20 min before fixation, and anti-ATG2A (MBL, PD041) after fixation, and mounted with ProLong gold antifade mountant (Thermo Fisher, P10144). Confocal images in other figures were imaged with cells mounted with Invitrogen ProLong gold antifade mountant with DAPI DNA stain (Thermo Fisher, P36941). The time period of RFP–LC3 and GFP–ATG2A co-localization was quantified from movies imaged live using a LSM780 Zeiss confocal microscope. HeLa cells stably expressing RFP–LC3 (made in house) were transfected with GFP–ATG2A construct and cultured for 24 h before live imaging. The time interval between frames was 3.87 s. The movies started when a GFP and RFP double-positive event was identified and ended when GFP–ATG2 left RFP–LC3. There were a maximum of 36 frames in the movies, making the maximum observed time period 140 s. There might be events where GFP–ATG2 and RFP–LC3 co-localize for longer than 140 s but the events we observed were generally shorter than that.

Super-resolution microscopy

HeLa cells seeded on high-precision size 1.5 coverslips (Carl Zeiss Ltd) were transfected with BFP–Sec61β and pSpCas9(BB)-2A-GFP-scr/pSpCas9(BB)-2A-GFP-Exon3 sgRNA construct and cultured for 36 h. The cells were then stained with antibodies to ATG2A (MBL, PD041) and TOM20-F10 (Insight Biotech, sc-17764), and mounted with Invitrogen ProLong gold antifade mountant (Thermo Fisher, P10144). Super-resolution structured illumination microscopy was performed using an Elyra PS1 instrument (Carl Zeiss Ltd). The samples were examined on the microscope using a ×63 1.4 numerical aperture plan-APO Carl Zeiss objective lens and Immersol 518 F (23 °C) immersion oil. Image acquisition was carried out using the ZEN 2012 Elyra edition software in which datasets were collected with five grating phases, five rotations and sufficient z positions spaced 110 nm apart to form a 2 mm-deep volume of raw super-resolution structured illumination microscopy data. Optimal gratings were selected for each wavelength used.

Thresholding was based on the underlying raw data and determined algorithmically using the Imaris software as part of the modelling process. The data were captured with identical acquisition settings in all cases and therefore the range of image intensities was directly comparable between all images.

Proximity ligation assay

Rabbit anti-IP3R3 (ITPR3) (Chemicon), mouse anti-VDAC1 (Abcam), PLA probe anti-rabbit PLUS (Sigma, DUO92002–100RXN), anti-mouse MINUS (Sigma, DUO92004-100RXN) and Duolink detection fluorophore red (Sigma, DUO92008; excitation, 594 nm and emission, 624 nm) were used. The experiments were conducted as described in the figure legend. In principle, if the two proteins of interest were located ≤40 nm apart, the connector oligonucleotides hybridized with the PLA probes and after ligation, the signal was amplified by rolling circle amplification.

Flow cytometry

SH-SY5Y cells were stained for 30 min in growth medium supplemented with 5 μM BODIPY581/591C11. After staining, the medium was collected; the cells were washed with PBS and lifted with trypsin. The cells were collected in PBS and combined with the previous PBS and medium. Next, the cells were sorted on a Becton Dickinson LSR Fortessa flow cytometer. The oxidized and total signals were acquired simultaneously using 488 and 568 nm lasers, and detected with 530/30 and 590/30 filters, respectively. For the measurement of mitochondria mass, HeLa cells were stained with 2.5 μM nonylacridine orange (Thermo Fisher Scientific, A1372; excitation, 490 nm and emission, 540 nm) at 37 °C for 30 min for labelling mitochondria, and sorted using an Attune NxT analyser. The signals were acquired with a 488 nm laser and detected with a BL1 530/30 nm detector. Median fluorescence intensity analysis of labelled mitochondria was performed by gating on single cells. The data were processed using the FlowJo v10.8 Software.

Preparation of liposomes

A 3 mM phospholipid working mixture was prepared using 18:1-12:0 NBD PS (Avanti Polar Lipids, 810195C) and 18:0-16:0 PC (Avanti Polar Lipids, 850465) dissolved in import buffer (300 mM sucrose, 10 mM Tris–HCl pH 7.5, 150 mM KCl and 1 mM dithiothreitol) at a ratio of 75:25% (PC:NBD-PS). The mixture was left at room temperature for 1 h for the liposomes to form. Unilamellar liposomes were formed using an Avanti extruder (30 passes).

Measurement of de novo mitochondrial PE synthesis

HeLa cells cultured in live-imaging dishes (VWR International, 734-2905) were stained with MitoTracker Red CMXRos (Thermo Fisher, M7512) diluted 1:2,000 in growth medium for 10 min in the incubator to segment the mitochondrial area for future quantification. MAS Buffer was prepared (220 mM mannitol, 70 mM sucrose, 10 mM KH2PO4, 5 mM MgCl2 and 2 mM HEPES). The cells were first washed with PBS (containing MgCl2 and CaCl2) and then with MAS Buffer pre-warmed to 37 °C. The NBD-PS/PC liposomes in warm MAS buffer containing 5 µl ml−1 Duramycin–Cy5 conjugate (Molecular Targeting Technologies, D-1002) were loaded for 50 min and live imaged within 5 min (Time 0) using a LSM880 Zeiss confocal microscope with a ×63 oil-immersion lens. After imaging at Time 0, the samples were washed three times with PBS buffer and fresh warm MAS buffer was added. After 40 min in the incubator, the cells were stained again with Duramycin–Cy5 for 50 min and imaged (Time 1.5 h). For the samples with Seahorse treatment, the cells were incubated with MAS buffer (with 1 mM EGTA) containing 2 nM Seahorse XF-PMP (Agilent, 102504-100) for 15 min before imaging as described above. Image J was used to quantify the grey value of the Cy5 signal in the region of interest, which was double-positive for NBD signal and MitoTracker Red. The grey value of Cy5 was then normalized to the grey value of NBD in the corresponding region of interest to calculate the Cy5 intensity per arbitrary unit of NBD-PS.

Zebrafish maintenance and crosses

Adult fish that were between 6 and 18 months old were bred to generate embryos and larvae for the experiments described hereafter. For Fig. 2a (survival assay, Protocol 7), 0–7-week-old zebrafish were used. Zebrafish larvae were used for different experiments at the following ages: 0–10 d.p.f. for Figs. 2b,c, 3c,d and Extended Data Fig. 1m; 0–5 d.p.f. for Fig. 2d; 5 d.p.f. for Extended Data Fig. 1l, and 0–2 d.p.f. for Fig. 6i and Extended Data Fig. 8b.

The zebrafish were maintained and cultured under standard conditions under a 14-h light and 10-h dark cycle. Embryos were collected from natural spawnings, staged according to established criteria40 and reared in embryo medium (5 mM NaCl, 0.17 mM KCl, 0.33 mM CaCl2, 0.33 mM Mg2SO4 and 5 mM HEPES pH 7.2) at 28.5 °C in the dark. Embryos from crosses of Tupfel Longfin wild-type zebrafish were used for the CRISPR–Cas9 experiments. The rho:EGFP line [Tg2(rho:EGFP)cu3] expresses EGFP in the rods of the fish retina41 and was used to investigate rod-cell death after genetic and pharmacological manipulation with ferroptosis modulators. Offspring from outcrosses of the rho:EGFP line (heterozygous) to AB wild-type fish were treated with 0.003% phenylthiourea from 24 h.p.f. to inhibit pigmentation and to allow visualization of the fluorescent photoreceptors and the selection of EGFP-positive fish at 3 d.p.f. Fish collected for experimental analyses were culled by an overdose of 1 mg ml−1 3-amino benzoic acid ethyl ester (MS222) before sample processing.

CRISPR–Cas9 micro-injections

Four CRISPR sgRNAs per gene designed by Dharmacon (GE Healthcare Dharmacon, Inc.) were used together to target the zebrafish wdr45 or atg7 genes (sequences in Supplementary Table 3). To maximize the knockdown efficiency, 100 ng of each sgRNA was mixed with 4 µl TRACR RNA (Dharmacon Edit-R CRISPR–Cas9 synthetic tracrRNA; Dharmacon, U002005) and 1.6 μl nuclease-free water42. The mixtures were incubated at room temperature for 10 min before the addition of 2.64 μl of 2 M KCl and stored at −80 °C as 1.5 μl aliquots. On the day of injections, 5 μg Cas9 nuclease protein NLS (Horizon Discovery, CAS12206) was added to each thawed aliquot and incubated for 5 min at 37 °C, after which 0.2 μl phenol red was added to the samples to visualize the injected droplet.

Embryos were collected immediately after spawning and injected at the one-cell stage. the volumes injected were calibrated to a final amount of 0.32 ng CRIPSPR sgRNAs targeting atg7 and/or 0.64/0.96 ng for wdr45. For the double-knockdown experiments, CRISPR injections were performed sequentially—wdr45-targeting CRISPR–Cas9 solution was injected first, followed by atg7-CRISPR–Cas9 solution. Uninjected siblings were kept from each clutch in every experiment for comparison.

Endogenous wdr45 expression analysis

To determine the efficiency of the wdr45 knockdown by CRISPR injection, we compared the expression levels of endogenous wdr45 of CRISPR-injected fish with their uninjected siblings at 5 d.p.f. Pools of ten fish from seven independent clutches were collected at 5 d.p.f. and stored at −80 °C. Messenger RNA was extracted using an RNeasy-plus mini kit (Qiagen) and a QIAshredder kit column (Qiagen) according to the manufacturer’s instructions. A total of 1 μg RNA from each sample was then used to generate cDNA using a High-capacity reverse transcription kit (Applied Biosciences, 4368814) following the standard protocol. A primer pair targeting the last exons of the wdr45 gene (exons 10 and 11) was used for quantitative PCR (primer sequences in Supplementary Table 2). Quantitative PCR was performed in triplicate and the relative gene expression level was calculated after normalization to rbs11 internal controls using the 2−ΔΔCT method with logarithmic transformation for statistical analysis.

Drug treatment of zebrafish

GFP-positive offspring from crosses of the rho:EFGP line, previously treated with phenylthiourea from 1 d.p.f. and screened for EGFP expression at 3 d.p.f. as described earlier, were treated with Fer-1 at a concentration of 10 µM from 5 to 10 d.p.f. The drug was replenished daily until collection of fish for processing.

Cell death in rho:EGFP zebrafish—fixation, embedding and cryosectioning

Quantification of rod photoreceptor degeneration was analysed as previously described41. EGFP-positive offspring from rho:EGFP crosses to wild-type fish were culled at 10 d.p.f. by the addition of an excess of MS222 and fixed overnight in 4% paraformaldehyde in PBS at 4 °C. After washes in PBS, the zebrafish were equilibrated overnight in 30% sucrose in PBS at 4 °C, embedded in Scigen Tissue Plus optimal cutting temperature compound and frozen on dry ice before being stored at −20 °C. Transverse cryosections (10 μm) were cut through the central retina using a LEICA CM3050 S cryostat and collected on Thermo Scientific Menzel Gläser, SuperFrost Plus slides. Imaging and analysis of images were performed as described41.

Survival study

Wild-type TL fish injected with 0.96 ng wdr45-targeting CRISPR–Cas9 were raised in parallel to their uninjected siblings up to the age of 7 weeks to evaluate changes in survival. After injection, the fish were maintained in the dark in an incubator at 28.5 °C until 5 d.p.f. and then transferred to the aquarium facility (60 fish per group). The numbers in each group were counted on a weekly basis. Survival was analysed using the log-rank (Mantel–Cox) method to compare the two groups.

ATG2A construct micro-injections and abnormality quantification

Embryos from crosses of wild-type TL fish were injected with pEGFP, pEGFP-C1-hATG2A, pEGFP-C1-hATG2A-YFS, pEGFP-C1-hATG2A-ΔMLD or pEGFP-C1-hATG2A-LTD as described earlier. The DNA constructs were diluted to 100 or 200 ng μl−1 in Danieau’s solution (17.4 mM NaCl, 210 μM KCl, 120 μM MgSO4, 180 μM Ca(NO3)2 and 1.5 mM HEPES pH 7.6) and injected in different volumes to result in final DNA amounts ranging from 60 to 450 pg, into embryos at the one-cell stage. EGFP expression was visualized after 24 h using a LEICA M205 FA fluorescence microscope and abnormal fish were quantified in each group at 24 and 48 h.p.f. Any fish with aberrant morphology and developmental defects compared with their uninjected siblings were scored as abnormal. The number of fish in each group depended on the clutch size, with a minimum of 12 fish per condition.

Live imaging of zebrafish embryos

Live imaging of embryos injected with ATG2 constructs was performed using a LEICA M205 FA microscope equipped with a LEICA DFC7000T camera and the Leica Application Suite X (LAS X) software. Bright-field and EGFP-fluorescence images were taken using a LEICA 10450028 lens and the EGFP filter of an X-Cite 200DC fluorescence illuminator lamp. Larvae were anaesthetized using MS222 at 48 h.p.f. before imaging. Anaesthetic was not required at 24 h.p.f. as motility was limited by the presence of the chorion.

Statistics and reproducibilityGeneral information on statistical methods

Tissue culture data are presented as the (normalized) mean ± s.d., except for Extended Data Fig. 4e where data are the median ± 95% confidence interval. Tissue culture data were analysed using a two-tailed Student’s t-test, except for Extended Data Fig. 5d where a χ2 test was used. All in vivo experiments and tests were randomly assigned but no randomization was performed for the cell culture experiments. For the western blots, protein levels were normalized to the indicated internal control proteins.

General information on reproducibility

For the western blotting, immunomicroscopy, flowcytometry, lipid assays, cytotoxicity and viability assays, quantification and statistics were derived from n = 3 independent experiments, unless otherwise specified in the legends. Sample sizes were chosen on the basis of extensive experience with the assays we have performed. No data were excluded from the analyses.

Zebrafish data are presented as the mean ± s.e.m. or s.d., as indicated on each graph, and were analysed using a two-tailed Student’s t-test or log-rank (Mantel–Cox) test.

All statistical analyses were performed using GraphPad Prism 9 or Microsoft Excel; *P < 0.05, **P < 0.01, ***P < 0.001 and NS, not significant.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.