Ferroptosis is an iron-dependent form of regulated cell death induced by the lethal overload of lipid peroxides in cellular membranes. In recent years, modulating ferroptosis has gained attention as a potential therapeutic approach for tumor suppression. In the current study, retinol saturase (RETSAT) was identified as a significant ferroptosis mediator using a publicly accessible CRISPR/Cas9 screening dataset. RETSAT depletion protected tumor cells from lipid peroxidation and subsequent cell death triggered by various ferroptosis inducers. Furthermore, exogenous supplementation with retinoids, including retinol (the substrate of RETSAT) and its derivatives retinal and retinoic acid, also suppressed ferroptosis, whereas the product of RETSAT, 13, 14-dihydroretinol, failed to do so. As effective radical-trapping antioxidant, retinoids protected the lipid membrane from autoxidation and subsequent fragmentation, thus terminating the cascade of ferroptosis. Pseudotargeted lipidomic analysis identified an association between retinoid regulation of ferroptosis and lipid metabolism. Retinoic acid, but not 13, 14-dihydroretinoic acid, interacted with its nuclear receptor and activated transcription of stearoyl-CoA desaturase, which introduces the first double bond into saturated fatty acid and thus catalyzes the generation of monounsaturated fatty acid, a known ferroptosis suppressor. Therefore, RETSAT promotes ferroptosis by transforming retinol to 13, 14-dihydroretinol, thereby turning a strong anti-ferroptosis regulator into a relatively weak one.

Significance:

Retinoids have ferroptosis-protective properties and can be metabolized by RETSAT to promote ferroptosis, suggesting the possibility of targeting retinoid metabolism in cancer as a treatment strategy to trigger ferroptosis.

Ferroptosis is a form of non-apoptotic regulated cell death that results from excessive accumulation of phospholipid hydroperoxides in an iron-dependent manner (1). Its implication in various human pathologic conditions ranging from neurodegeneration to ischemia-reperfusion injury has been gradually uncovered (2). Interestingly, accumulating evidence shows that ferroptosis plays a key role in tumor development and treatment, and targeting ferroptosis might provide new therapeutic opportunities in treating cancers refractory to conventional therapies (3, 4).

Mechanistically, ferroptosis is induced by an imbalance between the production and elimination of reactive oxygen species (ROS), which leads to the peroxidation of polyunsaturated fatty acids (PUFA) in cell membranes, thus resulting in rapid membrane rupture and subsequent irreversible cell death (1). Because of enhanced metabolism, cancer cells usually experience extensive oxidative stress during tumorigenesis and in response to therapy but display low lipid peroxidation levels, suggesting the presence of a robust anti-ferroptosis system that neutralizes ROS damage (5). For instance, GPX4, a glutathione (GSH)-dependent selenoenzyme known as the most important ferroptosis suppressor, reduces toxic lipid hydroperoxides into nontoxic alcohol (3). Besides, ferroptosis suppressor protein 1 (FSP1) and dihydroorotate dehydrogenase (DHODH) confers protection against ferroptosis through catalyzing the regeneration of ubiquinol, the reduced form of CoQ10 and a radical-trapping antioxidant (RTA) that traps lipid peroxyl radicals and terminates the cascade of lipid peroxidation (6–8). Similarly, GTP cyclohydrolase 1 (GCH1) participates in synthesizing another RTA, BH4 (9). It has been established that pharmacologically targeting these proteins can trigger ferroptosis in a diverse panel of human cancer cell lines.

Moreover, cell sensitivity toward ferroptosis is determined by several metabolic factors such as ferrous ion availability and phospholipid metabolism (4). Specifically, PUFAs in glycerophospholipids are considered the substrates or “fuel” of ferroptosis, given their high susceptibility to redox attack of bis-allylic hydrogen atoms due to the presence of double bonds, whereas monounsaturated fatty acids (MUFA), such as palmitoleic acid and oleic acid, inhibit ferroptosis in a structure-specific manner (1, 10).

Although significant inroads have been achieved in better understanding the mechanisms underlying cell response to ferroptosis, it remains elusive whether it is possible to coordinate multiple downstream pathways, thus regulating ferroptosis from different dimensions. In the current study, using a publicly accessible CRISPR/Cas9 dataset, we found that retinoids, also known as vitamin A derivatives, confer dual protection against ferroptosis through their intrinsic radical-trapping properties and retinoic acid receptor (RAR)/SCD (stearoyl-CoA desaturase)-mediated MUFA generation. Furthermore, as the key components in retinoid metabolism, retinol saturase (RETSAT) and STRA6 play important roles in regulating ferroptosis. Thus, our data uncover retinoids as a novel anti-ferroptosis system that acts parallel to the canonical defense mechanisms and suggests an effective strategy of targeting RETSAT/STRA6 to modulate ferroptosis in cancer treatment.

Cell lines and culture condition

All cells were cultured in a 37°C incubator with humidified 5% CO2 atmosphere. The human non–small cell lung cancer cell lines A549, H1299, H23, H1975, H358, PC9, colorectal cancer cell line DLD1, SW480, HCT116, pancreatic cancer cell line MiaCaPa-2, fibrosarcoma cell line HT1080, and human embryonic kidney cell line HEK293T were cultured in high-glucose DMEM supplemented with 10% FBS (ScienCell) and 100 U/mL penicillin/streptomycin/amphotericin B (Sangon Biotech). All primary cell lines were purchased from the Chinese Academy of Science Cell Bank. Low passage cells (<30 passages) were used for all experiments. The cell lines were authenticated by short tandem repeat profiling in 2022 and were passaged every 3 to 5 days according to cells’ proliferating rates. To prevent Mycoplasma contamination, cells were tested every 2 months using a PCR-based method suggested by Uphoff and colleagues (11). Briefly, supernatant medium without any antibiotics was collected after 7 days cell culturing, from which the DNA was extracted and purified using silica-gel columns (TIANGEN). Then, PCR assays were performed with hot-start Taq DNA polymerase according to the protocol provided by manufacturer. PCR products were separated on a 1.3% agarose-TAE gel (Sangon Biotech) containing 0.3 μg/mL ethidium bromide and the results were visualized with a UV transilluminator. If the tested medium contained Mycoplasma, an additional band at 515 to 525 bp could be observed. Only the Mycoplasma-negative cells are kept for subsequent experiments.

Compounds

The following compounds were obtained from Topscience: RSL3 (T3646), ML210 (T8375), IKE (T5523), ferrostatin-1 (T6500), deferoxamine mesylate (T1637), Z-VAD(OMe)-FMK (T6013), necrosulfonamide (T7129), cisplatin (T1564), all-trans retinol (T1183), retinal (T5256), β,β-carotene (T1633), palmitoleic acid (T4872), and oleic acid (T2O2668). The following compounds were obtained from MedChemExpress: AM580 (HY-10475), Ch55 (HY-107397), AGN193109 (HY-U00449), and A939572 (HY-50709). Besides, all-trans retinoic acid (ST1627, Beyotime) and all-trans-13,14-dihydroretinol [13C(R)-enantiomer, SC-217607, Santa Cruz Biotechnology] were also purchased.

Cell viability assay

Cell viability was determined using Cell Counting Kit-8 (CCK8, Topscience), as described previously (12). Briefly, 5,000 cells (for cytotoxicity assay) or 1,000 cells (for proliferation assay) per well were seeded in 96-well plates and incubated for 24 hours. Subsequently, cells were treated as required and exposed to 10 μL of CCK8 reagent (100 μL medium per well) for 1 hour at 37°C, 5% CO2, in an incubator. The absorbance at a wavelength of 450 nm was measured using a Molecular Devices microplate reader.

qRT-PCR

RNA extraction and qRT-PCR were performed as described previously (12). Briefly, total RNA was extracted from cells using TRIzol reagent (TIANGEN), and cDNA was synthesized with a Hifair II 1st Strand cDNA Synthesis Kit (gDNA digester plus, Yensen Biotechnology). qPCR was performed with a Hifair III One Step qRT-PCR SYBR Green Kit (Yensen Biotechnology), and triplicate samples were run on an ABI QuantStudio 5 Real-Time PCR System (Thermo Fisher Scientific). The threshold cycle (Ct) values for each gene were normalized to GAPDH as the endogenous control, and the 2−ΔΔCt method was used for quantitative analysis. The primers used were synthesized by Sangon Biotech. GAPDH: F: AGAAGGCTGGGGCTCATTTG, R: AGGGGCCATCCACAGTCTTC; CHP1: F: AGTCAAATCACTCGCCTCTACA, R: TGATGGCAAGTTCTGGAATCCT; RETSAT: F: TACTTGGGACTATTCTCTGGCA, R: GCACTTGGTTGGCTGAAAAAG; GPAT4: F: GGTATCCGCAAACTCTACATGAA, R: CCACTTCGACGAATCTCTTTGA; SLC11A2: F: TGGAGATCATGGGGAGTCTG, R: AAGAAAACCTGGTCCGGTGAA; ABCA1: F: CCTGAAGCCAATCCTGAGAACACT, R: CAATACGAGACACAGCCTGGTAGA; RARA: F: AGAGCAGCAGTTCTGAAGAGATAGTG, R: CATAGTGGTAGCCTGAGGACTTGTC; SCD: F: TCTAGCTCCTATACCACCACCA, R: TCGTCTCCAACTTATCTCCTCC; SREBF1: F: ACAGTGACTTCCCTGGCCTAT, R: GCATGGACGGGTACATCTTCAA; ACACA: F: ATGTCTGGCTTGCACCTAGTA, R: CCCCAAAGCGAGTAACAAATTCT; FASN: F: AAGGACCTGTCTAGGTTTGATGC, R: TGGCTTCATAGGTGACTTCCA.

Immunoblotting

Western blotting was performed as described previously (12). Briefly, cells were lysed using RIPA buffer (Beyotime) with protease and phosphatase inhibitor cocktail (Beyotime), and the protein concentration was determined with a Bicinchoninic Acid Protein Assay Kit (Yensen Biotechnology). Next, proteins were boiled in 5× SDS-PAGE loading buffer (EpiZyme Biotech) for 10 minutes at 100°C. About 15–20 μg proteins were separated using SDS-PAGE (EpiZyme Biotech) and transferred onto polyvinylidene fluoride membranes (Merck-Millipore). The membranes were blocked with 5% nonfat milk and incubated overnight at 4°C with antibodies against RETSAT (1:1,000, HPA046513, ATLAS), GAPDH (1:2,000, AF0006, Beyotime), glutathione peroxidase-4 (GPX4; 1:1,000, DF6701, Affinity), acyl-coenzyme A synthetase long-chain family member 4 (ACSL4; 1:1,000, abs106075, Absin), SLC7A11 (1:1,000, DF12509, Affinity), carbohydrate-responsive element-binding protein (ChREBP; MLXIPL, 1:1,000, abs118005, Absin), lamin-B1 (1:1,000, AF1408, Beyotime), STRA6 (1:1,000, abs134246, Absin), p-AMPK (1:1,000, T172, Cell Signaling Technology), SCD (1:1,000, abs124771, Absin), CHP1 (1:1,000, abs152364, Absin), SLC11A2 (1:1,000, abs133855, Absin), GPAT4 (1:1,000, abs140719, Absin). After the membranes were washed with TBS-Tween solution, the secondary antibodies were added to the membranes at room temperature. Finally, the protein bands were visualized with a BeyoECL Plus Kit (Beyotime).

Determination of lipid peroxidation by BODIPY-C11

The determination of lipid peroxidation was performed as described previously (12). Briefly, cells were seeded on 12-well plates and incubated overnight. The next day, cells were treated as indicated, harvested by trypsinization and washed with PBS (Beyotime). Next, the cells were suspended in a fresh medium containing 4 μmol/L BODIPY 581/591 C11 dye (Thermo Fisher Scientific) at 37°C in a humidified 5% CO2 atmosphere. After 30 minutes of incubation, the cells were washed with PBS, and the lipid peroxidation levels were assessed using the flow cytometer Accuri C6 (BD Biosciences) and FACS Aria II (BD Biosciences) with a 488-nm laser. The results were analyzed in FlowJo software (TreeStar). A figure exemplifying the gating strategy is provided in Supplementary Fig. S1A.

Malondialdehyde assay

The measurement of cellular malondialdehyde (MDA) levels was performed using the MDA assay kit according to the manufacturer's protocol (Beyotime). Briefly, 1 × 106 cells were seeded in a 10-cm dish 24 hours before the experiment. The cells were treated as indicated, lysed using 200 μL lysis buffer (Beyotime), and collected using a cell scraper. Cell lysates were centrifuged for 10 minutes at 12,000 × g, and the supernatants were collected and aliquoted into two 100 μL replicates, with the remaining used for protein concentration measurement. Then, 200 μL thiobarbituric acid (0.37%) solution was added to each replicate and boiled at 100°C for 15 minutes. After being cooled down to room temperature, the samples were centrifuged for 10 minutes at 1,000× g, and the supernatants were pipetted into a microplate to quantify the absorbance at a wavelength of 532 nm. Results were normalized to the protein concentration.

Labile iron pool measurement

A total of 100,000 cells were seeded in 24-well plates and treated as indicated 24 hours later. The colorimetric ferrozine–based iron quantification assay was performed as described by the manufacturer (Applygen; ref. 13). Mechanistically, iron complexed to cellular proteins was made accessible to ferrozine by treatment of cell lysates with acidic KMnO4 solution. Then, the cellular iron pool was measured by adding the iron-detection reagent (ferrozine, neocuproine, ammonium acetate, and ascorbic acid) and reading the absorbance at 550 nm using a microplate reader). Results were normalized to the protein concentration.

Reduced GSH measurement

The reduced GSH level was determined as described previously (12). The treated cells were harvested by trypsinization, and 6 × 105 live cells from each sample were transferred to new tubes, washed in PBS, and centrifuged at 1,200 rpm at 4°C for 5 minutes. The cell pellets were resuspended in 60 μL protein removal solution, thoroughly mixed, and incubated at −196°C (liquid nitrogen) and 37°C sequentially twice for fast freezing and thawing, then incubated at 4°C for 5 minutes and centrifuged at 10,000 × g for 10 minutes. The supernatant was pipetted to determine the amount of GSH in the sample. This assay was conducted using the GSH and GSSG Assay Kit (Beyotime) according to the manufacturer's protocol.

Transmission electron microscopy

The samples for transmission electron microscopy were prepared as described previously (12). Briefly, treated cells cultured in 6 cm dishes were fixed with a solution containing 2.5% glutaraldehyde. After being washed in 0.1 mol/L phosphate buffer (pH 7.4) for three times, cells were postfixed with phosphate buffer containing 1% osmic acid, then washed in 0.1 mol/L phosphate buffer (pH 7.4) for three more times. After dehydration and embedding, samples were incubated in a 60°C drying oven for 48 hours. Ultrathin sections were prepared and stained with lead citrate and uranyl acetate. After drying overnight, the sections were examined with a Hitachi transmission electron microscope (Hitachi).

CRISPR/Cas9-mediated gene knockout

CRISPR/Cas9 technology was employed to knock out RETSAT and/or STRA6 in the cell lines. Single-guide RNAs (sgRNA) were cloned into the GV392 plasmids containing puromycin resistance gene and hSpCas9 gene. The following sgRNA sequences were used: sgRETSAT-#1, 5′- CCAGCTGCCCTTCAGTGATC-3′; sgRETSAT-#2, 5′-GGGGGCTGCTGTCATACCTT-3′; sgSTRA6-#1, 5′-TGGTACATCGATGAGCCCCA-3′; sgSTRA6-#2, 5′- CAGGCGTGGTACAGGCCGGG-3′; sgRNA-Control: 5′-CGCTTCCGCGGCCCGTTCAA-3′. Then, the edited vectors were packaged into lentiviruses in HEK293T cells. The design and construction of the vectors and lentiviruses were performed by Genechem Technology.

To establish cell lines with stable gene deletion or overexpression, the cells were seeded into 6-well plates with a concentration of 2 × 105 cells per well. After 24 hours, cells were infected with lentiviruses (multiplicity of infection = 10) and Hitrans G P as per the manufacturer's protocol (Genechem technology). At 72 hours after infection, cells were propagated in a selection medium containing 2.5 μmol/L puromycin (Beyotime) for 48 hours.

siRNA/short hairpin RNA–mediated gene knockdown

Small interfering RNAs (siRNA) and short hairpin RNAs (shRNA), as well as corresponding negative controls, used in the presented research, were purchased from RiboBio. The target sequences are listed below: siCHP1#1: GCATGATGGTCGGAGTAAA, siCHP1#2: TGCGAACTTTGGCTCATTT; siRETSAT#1: GAAGAAGGTTCTCAAACAA, siRETSAT#2: GCTGCTGACTCGTTTCTCT; siSLC11A2#1: GGAGGAATCTTGGTCCTTA, siSLC11A2#2: GTACCTGCATTCTGCCTTA; siGPAT4#1: GCACAACTGTGGTGGGATA; siGPAT4#2: AGACCATTATGGATGATGA; shRARA#1: GTGAGAAACGACCGAAACA, shRARA#2: ATTACTGACCTGCGAAGCA; shSCD#1: GGTACTACAAACCTGGCTTGC, shSCD#2: GCGATATGCTGTGGTGCTTAA. The cells were seeded into 6-well plates at 60%–80% confluence. The next day, the medium was replaced with fresh medium containing siRNAs and Lipo8000 (Beyotime) as the transfection reagent. The cells were harvested for subsequent analyses 48 hours after transfection. The protocol for lentiviral vector-mediated shRNA transfection is the same as that of the CRISPR/Cas9-mediated gene knockout described above. The knockdown efficiency at the level of protein expression was exhibited in Supplementary Fig. S1B.

Overexpression plasmid constructs

The GV341 plasmids containing the full-length RETSAT gene were packaged into lentivirus in HEK293T cells. To avoid being cleavaged by stably expressing Cas9-sgRNA in some cell lines, slient mutations were introduced into the sequences targeted by corresponding sgRNAs. The design and construction of the vector and lentivirus were performed by Genechem Technology.

Fluorescence-enabled inhibited autoxidation assay

The experiment was carried out as described by Shah and colleagues (14). For liposome preparation, egg phosphatidylcholine was dissolved in chloroform, treated by thin-film hydration (45°C, 100 rpm) on a rotary evaporator, and resuspended in 1×PBS followed by an ultrasonic dispersion to form lipid suspension with a final concentration of 20 mmol/L. Then, the lipid suspension was extruded through a polycarbonate membrane with pore sizes of 100 nm to prepare liposomes with good dispersity and diameters of 100 ± 0.13 nm. The size of the liposomes was measured on a Malvern Zeta-sizer, which showed a relatively uniform distribution of particles.

A solution containing liposomes (1 mmol/L), STY-BODIPY (1 mmol/L, Cayman Chemical) and PBS at pH 7.4 to a final volume of 196 μL was added to a black 96-well polypropylene plate (Beyotime). Next, 2 μL potential antioxidants was added at desired concentrations, including 2,2,5,7,8-pentamethyl-6-chromanol (PMC; T14008, Topscience), all-trans-retinol (ROL), 13, 14-DROL, RAL, RA, β, β-carotene, Ch55, AM580, and PBS as a negative control. The plate was incubated for 10 minutes at 37°C in the microplate reader, followed by a mixing protocol for 5 minutes. Then, 3 μL (E)-1,2-bis((2-methyldecan-2-yl)oxy)diazene (DTUN, 1 mmol/L in ethanol, Cayman Chemical) was added to the mix to initiate the autoxidation process, followed by another mixing protocol for 5 minutes. Next, the plate was incubated at 37°C for 10 minutes before data were acquired by excitation of the probes at 488 nm and emission was measured at 518 nm. The data were acquired every 2 minutes, and the whole measurement process lasted for 8 hours without interruption.

To determine the response factor of the presented experiment, a set of uninhibited autoxidation with varying concentrations of the STY-BODIPY (0.2, 0.4, 0.6, 0.8, 1 μmol/L) was included. As shown in Supplementary Fig. S4A, the obtained relative fluorescent unit values exhibited a good linear relationship with the concentrations of STY-BODIPY. The parameters, including the rate of initiation (Ri), inhibition rate constant (kinh), and stoichiometry (n) were calculated according to equations provided in Supplementary Fig. S4B as previously described by Shah and colleagues (14).

RNA sequencing and bioinformatic analysis

RNA sequencing (RNA-seq) and corresponding bioinformatic analyses were performed using the Illumina HiSeq platform (Illumina) as described previously (12, 15). Differentially expressed genes (DEG) were identified with the limma package, which implements an empirical Bayesian approach to estimate gene expression changes by using the moderated t test. |log FC| > 0.5 and a P value < 0.05 were considered as a cut-off criteria to screen for DEGs.

RAR-GFP reporter assay

Cells seeded in a 96-well dish were transfected with RAR-GFP reporter plasmid (Yeasen Biotechnology) with the assistance of lipo8000 (Beyotime) and cultured for 24 hours. The luciferase intensity was measured using a Luciferase Reporter Gene Assay Kit (Beyotime) according to the manufacturer's protocol.

Pseudotargeted lipidomic analysis

Sample preparation

A total of 1 × 106 A549 cells were seeded into a 10-cm dish. One day later, the cells were treated with vehicle (DMSO) and IKE (20 μmol/L) for 24 hours or with RSL3 (2 μmol/L) for 4 hours, respectively. After treatment, cells were washed with PBS, collected using a cell scraper, and centrifuged. The cell pellets were quenched with liquid nitrogen and then subjected to a modified Folch lipid extraction procedure as described previously (16). Briefly, the cells were resuspended in 600 μL methyl alcohol/water (1:1, v/v) containing isotope-labeled internal mix standards obtained from Avanti Polar Lipids and Sigma-Aldrich. Then, 600 μL chloroform was added to the tube, and the samples were subjected to ultrasonication for 3 minutes and ultrasonic extraction in an ice-water bath for 10 minutes, followed by incubation at 4°C for 30 minutes, and the chloroform layers were collected and lyophilized in a centrifugal vacuum evaporator. The residue was re-extracted using the same condition as above. Finally, samples were combined and reconstituted in isopropanol/methanol (1:1, v/v) and subjected to vortex for 30 seconds, ultrasonic extraction for 3 minutes, and centrifugation at 13,000 rpm, 4°C for 10 minutes. About 200 μL supernatant in each sample was collected for subsequent analyses. The quality-control sample was prepared by mixing equal volumes of aliquots of the supernatants from all samples.

LC/MS analysis

The LC/MS analysis was carried out by the Shanghai Lumin Biological Technology, LTD. The LC system includes an ExionLC System consisting of a binary high pressure mixing gradient pump with degasser, a thermostatic autosampler, and a column oven. The optimized conditions were as follows: temperatures of the autosampler: 55°C; sample injection volume: 5 μL; mobile phase A: acetonitrile/water (6:4, v/v) containing 0.1% formic acid and 10 mmol/L ammonium formate; mobile phase B: acetonitrile/methanol (1:9, v/v) containing 0.1% formic acid and 10 mmol/L ammonium formate; flow rate: 0.35 mL/minute. A 20-minute elution gradient with an UPLC HSS T3 (1.7 μm, 2.1 × 100 μm) column was performed as follows: during the first 1.5 minutes, eluent composition was set at 0% phase B, which was linearly changed to 55% phase B at 5 minutes, then 60% at 10 minutes, 70% at 13 minutes, and 90% at 15 minutes. The next 1-minute phase was then increased to 100% and kept for 2 minutes. Finally, the initial conditions were recovered and maintained for 2 minutes for column conditioning. The mass spectrometry (MS) system was QTRAP 6500+ (SCIEX) equipped with an IonDrive Turbo V source, and the assay was performed in the negative/positive-ion working mode with a time-scheduled multiple reaction monitoring (MRM) method. The source condition was as follows: curtain gas: 35 psi; medium: CAD; IS: −4.5 kV/+5.5 kV; Gas1: 40 psi; Gas2: 45 psi.

Data processing and analysis

The raw data files were processed and annotated using the MRMPROBS software designed by Tsugawa and colleagues (17). Principle component analysis (PCA) and differential lipid analysis were performed in R software (R Foundation for Statistical Computing).

Chromatin immunoprecipitation assay

The assay was conducted using the SimpleChIP Plus Enzymatic Chromatin IP Kit (Cell Signaling Technology) according to the manufacturer's protocol. Briefly, cells were fixed with formaldehyde to cross-link histone and non-histone proteins to DNA. Then, chromatin was digested with Micrococcal Nuclease into 150–900 bp DNA/protein fragments. IgG or antibody specific to RARα (1:50, #62294, Cell Signaling Technology) were added, the complex coprecipitated and was captured by Protein G magnetic beads. Next, the chromatin was eluted, and the cross-links were reversed. Finally, DNA was purified using spin columns and quantified by qRT-PCR. Primer 1: F: CAGAGGTTGCAGTGAAGCGAGAT, R: GCTCATCCAGTCATGCCTCAGAAG; primer 2: F: TCAACTGCCAGCTCCATCACTTAC, R: TCACCCAGCAGCAGGCGAAA; primer 3: F: GGTGGAAGAGAAGCTGAGAAGGAG, R: TCGTCCTGCCGTTGCCATTG.

Dual-luciferase reporter assay

SCD promoter region spanning from −2000 to +200 of the transcription start site and corresponding mutant sequences were cloned into PHY-811 vectors. The assay was conducted in the HEK293T cell line using a Luciferase Reporter Gene Assay Kit (Beyotime) as described previously (12).

Targeted lipidomic analysis

Sample preparation

After being collected using a cell scraper, cells were washed with PBS and centrifuged for 5 minutes at 1,000 × g. The cell pellets were homogenized in liquid N2, and the metabolites were extracted in 1 mL ice-cold methanol (70%, v/v), followed by sonication in an ultrasonic water bath for 30 minutes at a frequency of 40 kHz (25°C). Samples were centrifuged at 12,000 × g for 15 minutes at 25°C, and supernatants were filtered through a 0.2 μm membrane. Under vacuum conditions, the supernatant was evaporated to dryness, and the residue was dissolved in 1 mL of H2O/ methanol (1:1, v/v), followed by centrifugation (10 minutes at 12,000 × g), and 3 μL of the aqueous phase was retained for subsequent analysis.

MS conditions

The separation of the compounds was carried out on a Phenomenex Kinetex C18 operated at 40°C (1.8 μm, 100 × 3.0 mm). The mobile phase consisted of 0.1% formic acid in water (A) and acetonitrile (B), was delivered at a flow rate of 0.5 mL/minute under a gradient program. The gradient system was 0–0.5 minute, 10% B; 0.5–3.0 minutes, 10%–90% B; 3.0–5.0 minutes, 90%–90% B; 5.0–5.1 minutes, 90%–10% B. The diode-array detector was set to monitor at 254 nm, and the online UV spectra were recorded in the scanning range of 190–400 nm.

The mass spectra were acquired in negative and positive ESI mode using a TripleTOF 6500 system with a Duo Spray source (SCIEX). Optimized parameter for negative and positive mode was as follows: the ion spray voltage was set to 5,500 (positive ion mode) and −4,500 V (negative-ion mode); the Turbo V spray temperature, 600°C; nebulizer gas (Gas 1), 50 psi; heater gas (Gas 2), 60 psi; collision gas, medium; the curtain gas was kept at 30 psi; and declustering potential, 80 (positive ion mode) and −80 V (negative-ion mode). The collision energy was set at 35 (positive ion mode) and −35 V (negative-ion mode), and the collision energy spread was 15 V for MS-MS experiments. The data were analyzed by Peak View Software 2.2 (SCIEX).

Tumor xenograft experiment

All animal studies were conducted in compliance with the policies of the animal ethics committee of Zhongshan Hospital, Fudan University (Shanghai, P.R. China). Four-week-old male BALB/c nude mice were purchased from the Shanghai Jiesijie Laboratory Animal Company and maintained under pathogen-free conditions. A total of 2 × 106 Cas-NC or RETSAT-knockout (KO) A549 cells were resuspended in 100 μL cold PBS and subcutaneously injected into the right flank of each nude mouse. When the tumors reached 70–100 mm3, the mice were assigned randomly into six treatment groups (n = 8 for each group at the beginning). IKE (30 mg/kg), ROL (20 mg/kg), and RA (20 mg/kg) were dissolved in DMSO, diluted in corn oil (Beyotime), and intraperitoneally injected every 2 days. The drug administration was continued until the endpoint, as indicated in the corresponding figures. The tumor volume was measured using a caliper every 3 days until the endpoint and calculated according to the equation: v = length * width2 * 1/2. Mice that died before the endpoint were not included in the final analysis. The xenograft tumors were harvested for subsequent IHC at the end of the experiment.

IHC

Tissue specimens were obtained from the tumor xenograft experiment mentioned above and the patients diagnosed with lung adenocarcinoma (LUAD) who received surgery in the Department of Thoracic Surgery, Zhongshan Hospital, Fudan University (Shanghai, P.R. China). The use of patient samples in this study was approved by the Ethics Committee of Zhongshan Hospital, Fudan University, Shanghai, P.R. China (approval no: B2022-180R). All of the patients have provided written informed consent, and the study was conducted in accordance with the Declaration of Helsinki. The paraffin-embedded tissues were dewaxed, rehydrated, and stained using the IHC kit according to the manufacturer's protocol (SP-9002, ZSGB-BIO). The antibody used for IHC was anti-4-hydroxy-2-nonenal (4-HNE; 1:200, Abcam, ab46545), RETSAT (1:200, HPA046513, ATLAS), and STRA6 (1:100, abs134246, Absin).

Statistical analysis

All experiments were independently performed in at least triplicate. Unpaired Student t tests or one-way ANOVA were utilized to compare continuous variables between two groups or among multiple groups, respectively. The results are presented as means, and the error bars represent the SD unless stated otherwise. All statistical analyses were conducted in GraphPad Prism software (7.0) and R software. Bioinformatic analyses were conducted as described previously (12, 18). The P values were all two sided, and P values < 0.05 were considered statistically significant. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001; ns, not significant.

Code availability

The software and algorithms for data analyses used in this study are all well established from previous work. There is no unreported algorithm used in this article. The R codes for data processing are available from the corresponding author upon reasonable request.

Data availability

Raw sequencing data for CRISPR-Cas9 screens in UACC-257 melanoma cells was downloaded from the Gene Expression Omnibus (accession number GSE130982). Raw RNA-seq data were deposited in SRA database at PRJNA979805. Processed RNA-seq and pseudotargeted lipidomic data were deposited in figshare (DOI: 10.6084/m9.figshare.21026443). The Cancer Genome Atlas (TCGA; GDC TCGA LUAD) and DepMap data (Version: DepMap Public 22Q4) analyzed in this study were obtained at https://portal.gdc.cancer.gov/ and https://depmap.org/portal/. Other source data and supporting the findings of this study are available from the corresponding author on reasonable request.

CRISPR/Cas9 screening uncovers RETSAT as a significant ferroptosis mediator

To identify genes that mediate ferroptosis, we used the genome-wide synthetic lethal CRISPR/Cas9 screening dataset published by Zou and colleagues, in which five types of fatty acids (FA), including linoleic acid (LA; C18:2), arachidonic acid (AA; C20:4), α-linolenic acid (ALA; C18:3), docosapentaenoic acid (DPA; C22:5n-3), and docosahexaenoic acid (DHA; C22:6) were added to culture media to sensitize UACC-257 melanoma cell line to ML210-triggered ferroptosis (19). For all FAs except for LA, ACSL4, a known ferroptosis regulator involved in PUFAs esterification, emerged as one of the top hits with a P value < 0.05, together with five other candidates: POR, CHP1, RETSAT, GPAT4, and SLC11A2 (Fig. 1A and B).

Figure 1.

CRISPR/Cas9 screening links RETSAT to ferroptosis. A, Venn diagram showing the intersection of enriched ferroptosis-related oxidoreductases in cells treated with ML210 and different PUFAs. B, Partial volcano plots showing the top hits in ALA-, DPA-, DHA-, or AA-treated screening conditions in UACC-258-Cas9 cells. ACSL4, POR, CHP1, RETSAT, SLC11A2, and GPAT4 are highlighted. C, Cell viability in A549 or H1299 cells treated with RSL3 (A549, 4 μmol/L for 6 hours; H1299, 40 μmol/L for 24 hours) following siRNAs transfection for 48 hours, determined by CCK8. D, Cell viability in A549 cells treated with RSL3 for 6 hours following siRNAs transfection for 48 hours. E, Lipid peroxidation in A549 cells treated with 2 μmol/L RSL3 for 3 hours following siRNAs transfection for 48 hours, determined by BODIPY-C11 staining followed by FACS analysis. F, RETSAT protein levels in Cas-NC and RETSAT-KO (#1 and #2) cancer cell lines determined by Western blotting. G, Cell viability in Cas-NC and RETSAT-KO A549, HT1080, or H1299 cells treated with RSL3 for 6 hours (A549 and HT1080) or 24 hours (H1299). H, RETSAT protein levels in cancer cell lines with indicated genotypes. I, Cell viability in A549, HT1080, or H1299 cells with indicated genotypes treated with RSL3 for 6 hours (A549 and HT1080) or 24 hours (H1299). J, Cell viability in A549, HT1080, or H1299 cells with indicated genotypes treated with RSL3 (A549 and HT1080, 4 μmol/L for 6 hours; H1299, 40 μmol/L for 24 hours) combined with or without DFO (100 μmol/L), Fer-1 (10 μmol/L), z-VAD-FMK (10 μmol/L), or necrosulfonamide (0.5 μmol/L). K, Lipid peroxidation in Cas-NC and RETSAT-KO A549 cells treated with 2 μmol/L RSL3 for 3 hours. L, Transmission electron microscopy images of Cas-NC and RETSAT-KO A549 cells treated with 2 μmol/L RSL3 combined with or without 2 μmol/L ROL or RA for 5 hours. Scale bars, 4 μm. M, Protein levels of ChREBP, ACSL4, RETSAT, SLC7A11, and GPX4 in Cas-NC and RETSAT-KO A549 cells treated with DMSO or 0.5 μmol/L RSL3 for 48 hours. Data are presented as the mean ± SD; n = 3 independent repeats. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001.

Figure 1.

CRISPR/Cas9 screening links RETSAT to ferroptosis. A, Venn diagram showing the intersection of enriched ferroptosis-related oxidoreductases in cells treated with ML210 and different PUFAs. B, Partial volcano plots showing the top hits in ALA-, DPA-, DHA-, or AA-treated screening conditions in UACC-258-Cas9 cells. ACSL4, POR, CHP1, RETSAT, SLC11A2, and GPAT4 are highlighted. C, Cell viability in A549 or H1299 cells treated with RSL3 (A549, 4 μmol/L for 6 hours; H1299, 40 μmol/L for 24 hours) following siRNAs transfection for 48 hours, determined by CCK8. D, Cell viability in A549 cells treated with RSL3 for 6 hours following siRNAs transfection for 48 hours. E, Lipid peroxidation in A549 cells treated with 2 μmol/L RSL3 for 3 hours following siRNAs transfection for 48 hours, determined by BODIPY-C11 staining followed by FACS analysis. F, RETSAT protein levels in Cas-NC and RETSAT-KO (#1 and #2) cancer cell lines determined by Western blotting. G, Cell viability in Cas-NC and RETSAT-KO A549, HT1080, or H1299 cells treated with RSL3 for 6 hours (A549 and HT1080) or 24 hours (H1299). H, RETSAT protein levels in cancer cell lines with indicated genotypes. I, Cell viability in A549, HT1080, or H1299 cells with indicated genotypes treated with RSL3 for 6 hours (A549 and HT1080) or 24 hours (H1299). J, Cell viability in A549, HT1080, or H1299 cells with indicated genotypes treated with RSL3 (A549 and HT1080, 4 μmol/L for 6 hours; H1299, 40 μmol/L for 24 hours) combined with or without DFO (100 μmol/L), Fer-1 (10 μmol/L), z-VAD-FMK (10 μmol/L), or necrosulfonamide (0.5 μmol/L). K, Lipid peroxidation in Cas-NC and RETSAT-KO A549 cells treated with 2 μmol/L RSL3 for 3 hours. L, Transmission electron microscopy images of Cas-NC and RETSAT-KO A549 cells treated with 2 μmol/L RSL3 combined with or without 2 μmol/L ROL or RA for 5 hours. Scale bars, 4 μm. M, Protein levels of ChREBP, ACSL4, RETSAT, SLC7A11, and GPX4 in Cas-NC and RETSAT-KO A549 cells treated with DMSO or 0.5 μmol/L RSL3 for 48 hours. Data are presented as the mean ± SD; n = 3 independent repeats. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001.

Close modal

Zou and colleagues and Yan and colleagues previously established the key-mediating role of POR as an electron donor in ferroptosis (19, 20). In the current study, we knocked down the expression of the remaining four potential genes using siRNAs in two lung adenocarcinoma cell lines (Supplementary Fig. S2A) and found that only knockdown of RETSAT and GPAT4 significantly protected cells against ferroptosis induced by RSL3, IKE, and ML210 (Fig. 1C; Supplementary Fig. S2B), and the ferroptosis blockage effect of RETSAT knockdown was much stronger than GPAT4. We treated A549 cells with RSL3/IKE and H1299 cells with RSL3/ML210 because we previously confirmed that A549 and H1299 cells were virtually unaffected by ML210 and IKE, respectively. Then, we focused on RETSAT in subsequent analyses. We found that A549 and H1299 cells with RETSAT knockdown presented a marked resistance toward RSL3 at different doses (Fig. 1D; Supplementary Fig. S2C). Meanwhile, because ferroptosis is driven by phospholipid peroxidation, we stained cells with BODIPY-C11 and found that RETSAT-KD significantly blunted RSL3-induced lipid peroxidation (Fig. 1E; Supplementary Fig. S2D). Ferroptosis inducers (FIN) also induced RETSAT upregulation in a dose-dependent manner (Supplementary Fig. S2E and S2F). These results indicate that RETSAT is involved in the ferroptosis cascade.

RETSAT is a pro-ferroptosis gene that does not interfere with known ferroptosis regulators

To further validate the pro-ferroptosis role of RETSAT, we depleted its proteins using two sequence-independent sgRNAs in A549, H1299, and HT1080, a bona fide ferroptosis cell line model (Fig. 1F). Similar to siRNA assays, RETSAT-KO cells were more resistant to ferroptosis triggered by multiple FINs that act on different targets, including GPX4 inhibitors RSL3 or ML210 and the xCT inhibitor IKE (Fig. 1G; Supplementary Fig. S2G). We also tested cisplatin, a commonly used cytotoxic chemotherapy drug, but noticed that the protective effect of RETSAT-KO (Supplementary Fig. S2H) was much weaker in such a context. Sensitivity to FINs in RETSAT-KO cells was restored by its re-expression, thus excluding the off-target effects of CRISPR/Cas9-mediated RETSAT depletion (Fig. 1H and I; Supplementary Fig. S2I). The increased sensitivity to FINs caused by RETSAT-OE was rescued by the iron chelator deferoxamine (DFO) and RTA ferrostatin-1 (Fer-1), but not by inhibitors of apoptosis [z-VAD (OMe)-FMK] or necroptosis (necrosulfonamide), confirming the specific role of RETSAT in ferroptosis (Fig. 1J; Supplementary Fig. S2J). Moreover, RETSAT-KO prevented RSL3-induced lipid peroxidation measured by BODIPY-C11 and MDA (Fig. 1K; Supplementary Fig. S2K and S2L). Transmission electron microscopy revealed that A549 cells treated with RSL3 exhibited shrunken mitochondria with increased membrane density, a characteristic morphologic feature of ferroptosis, but this phenomenon was almost not visible in RETSAT-KO cells (Fig. 1L). Together, these findings indicate that RETSAT is a key pro-ferroptosis gene.

To characterize the alteration of gene expression profile after RETSAT-KO, we found that RETSAT-KO did not significantly influence the expression of known key ferroptosis-related factors like ACSL4, GPX4, and SLC7A11 in cells treated with DMSO, RSL3, and IKE using RNA-seq and Western blotting (Fig. 1M; Supplementary Fig. S3A and S3B). In addition, we confirmed that the GSH and labile iron levels were not impacted by RETSAT-KO (Supplementary Fig. S3C). These findings demonstrated that the pro-ferroptosis function of RETSAT was independent of canonical ferroptosis mechanisms. Of note, consistent with the high-throughput data analysis results in the Cancer Dependency Map (DepMap; ref. 21), we validated that RETSAT-KO did not significantly alter cellular viability or proliferation rate (Supplementary Fig. S3D), suggesting that it was not an important gene under standard culture conditions, and targeting RETSAT represents a pretty safe approach to inhibit ferroptosis in different disease contexts.

Retinoids prevent lipid peroxidation and subsequent ferroptosis

We next focused on elucidating the mechanism underlying the ferroptosis-promoting effect of RETSAT. The initial function of RETSAT is well established to catalyze saturation of ROL at the 13-14 double bond to generate all-trans-13, 14-dihydroretinol (13, 14-DROL). This process occurs via transferring a hydride ion from the reduced factor NAD(P)H or FADH2 together with a proton (22). Both the substrate (ROL) and the product (13, 14-DROL) can be stored as esters or further oxidized to corresponding aldehyde or acids (Fig. 2A; refs. 23, 24). Because ROL can only be synthesized by plants, mammals must intake ROL or β, β-carotene to meet normal growth and development demands. Also known as provitamin A carotenoid, β, β-carotene could be symmetrically cleaved into two retinal (RAL) molecules, and RAL converts mutually with ROL (Fig. 2A; ref. 25).

Figure 2.

Retinoids prevent ferroptosis as RTAs. A, A schematic model describing the metabolism of retinoids and the role of RETSAT and STRA6 in this process. B, Cell viability in A549, HT1080, or H1299 cells treated with RSL3 (A549 and HT1080, 4 μmol/L for 6 hours; H1299, 40 μmol/L for 24 hours) combined with or without retinoids as indicated. C, Lipid peroxidation in A549 cells treated with 2 μmol/L RSL3 combined with or without retinoids (2 μmol/L) for 3 hours. D, Protein levels of STRA6 in A549 cells with indicated genotypes. E, Cell viability in Cas-NC and STRA6-KO A549 cells treated with 4 μmol/L RSL3 combined with or without 2 μmol/L ROL for 6 hours. F, Lipid peroxidation in Cas-NC and STRA6-KO A549 cells with indicated genotypes treated with 2 μmol/L RSL3 combined with or without 2 μmol/L ROL for 3 hours. Data are presented as the mean ± SD; n = 3 independent repeats. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001.

Figure 2.

Retinoids prevent ferroptosis as RTAs. A, A schematic model describing the metabolism of retinoids and the role of RETSAT and STRA6 in this process. B, Cell viability in A549, HT1080, or H1299 cells treated with RSL3 (A549 and HT1080, 4 μmol/L for 6 hours; H1299, 40 μmol/L for 24 hours) combined with or without retinoids as indicated. C, Lipid peroxidation in A549 cells treated with 2 μmol/L RSL3 combined with or without retinoids (2 μmol/L) for 3 hours. D, Protein levels of STRA6 in A549 cells with indicated genotypes. E, Cell viability in Cas-NC and STRA6-KO A549 cells treated with 4 μmol/L RSL3 combined with or without 2 μmol/L ROL for 6 hours. F, Lipid peroxidation in Cas-NC and STRA6-KO A549 cells with indicated genotypes treated with 2 μmol/L RSL3 combined with or without 2 μmol/L ROL for 3 hours. Data are presented as the mean ± SD; n = 3 independent repeats. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001.

Close modal

Like GSH, vitamin C, vitamin E, vitamin K, BH4, or CoQ10, retinoids also possess antioxidant properties. The antioxidant activity of retinoids is conferred by the side chain of polyene unit, also known as a highly conjugated double bond system, that can quench ROS and stabilize peroxyl radicals (26, 27). Considering the mediating role of RETSAT in both ferroptosis and retinoid metabolism, we questioned whether retinoid could act as a membrane-permeant antioxidant that protects cells against ROS-induced lipid peroxidation and ferroptosis, as known ferroptosis “defenders” like GSH, BH4, or CoQ10 do (3, 6–9)? As expected, exogenous supplementation of ROL in the culture system and its downstream metabolites, including RAL and retinoic acid (RA), abrogated cell sensitivity to FINs in a dose-dependent manner (Fig. 2B; Supplementary Fig. S3E). Meanwhile, as indicated by BODIPY-C11 staining, ROL, RAL, and RA, but not β, β-carotene markedly eliminated lipid peroxidation caused by RSL3 treatment (Fig. 2C; Supplementary Fig. S3F). In addition, the morphologic alteration of mitochondria induced by RSL3 could be reversed by treatment with ROL or RA (Fig. 1L).

STRA6 protects against ferroptosis by mediating ROL's entry into cells

When it comes to the influx of ROL, STRA6, a membrane receptor, facilitates ROL uptake from extracellular ROL-binding protein (RBP4) to cytoplasm, especially in cells of blood–tissue barriers, such as Sertoli cells in the testis, choroid cells in the brain, and retina pigmented epithelium cells, while in other tissues the retinol also permeats the membrane in a nonspecific manner (26, 28). Therefore, to further address whether interfering ROL homeostasis indeed impacts cell sensitivity to ferroptosis, we generated STRA6-KO cell lines through CRISPR/Cas9 (Fig. 2D; Supplementary Fig. S3G) and found that the loss of STRA6 not only sensitized cells to FINs but also slightly impaired the anti-ferroptosis effect conferred by exogenously supplemented ROL, which is consistent with STRA6’s function (Fig. 2E and F; Supplementary Fig. S3H). Together, these findings firmly establish a yet-unrecognized role of retinoids in preventing lipid peroxidation and associated cell death caused by ferroptosis.

Retinoids are effective RTAs

On the basis of the observation that retinoids, including ROL, RAL, and RA, suppress ferroptosis and their widely-accepted antioxidant properties (26, 29), we sought to explore whether retinoids could act as a membrane-permeant antioxidant and thus inhibit lipid peroxidation occurred during ferroptosis. Therefore, we carried out a well-defined fluorescence-enabled inhibited autoxidation (FENIX) assay, which employs liposomes composed of egg phosphatidylcholine to simulate the cell membrane's lipid bilayer in vitro. A lipophilic radical initiator DTUN was used to generate lipid peroxyl radicals, while STY-BODIPY (an oxidizable probe that competes with phospholipids for propagating lipid peroxyl radicals and is converted to its oxidized products ox-STY-BODIPY with high fluorescence) was implemented to monitor the autoxidation reaction (Fig. 3A; Supplementary Fig. S4A–S4C; ref. 14). RTAs such as PMC applied as a positive control inhibited the autoxidation process and thus retarded STY-BODIPY oxidation. Therefore, the STY-BODIPY consumption rate reflected RTA's activity, as implied by the rate constant (kinh).

Figure 3.

RETSAT regulates ferroptosis through mediating retinoid metabolism. A, The general view of FENIX assay. DTUN was used to initiate the co-autoxidation of the polyunsaturated lipids of egg phosphatidylcholine liposomes and STY-BODIPY. The process was inhibited by RTAs and monitored by the fluorescence of oxidized STY-BODIPY. B, Representative autoxidation of STY-BODIPY (1 μmol/L)–embedded liposomes (1 mmol/L; extruded to about 100 nm particle size) suspended in PBS at 37°C, initiated by the addition of 3 μL DTUN (1 mmol/L in ethanol), and inhibited by different RTAs as indicated. C, Representative autoxidation of STY-BODIPY as in B, but with varying concentrations of ROL, 13, 14-DROL, RAL, RA, and β, β-carotene. D, Cell viability in Cas-NC and RETSAT-KO or OE A549, HT1080, or H1299 cells treated with RSL3 (A549 and HT1080, 4 μmol/L for 6 hours; H1299, 40 μmol/L for 24 hours) combined with or without ROL as indicated. Data are presented as the mean ± SD; n = 3 independent repeats. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001.

Figure 3.

RETSAT regulates ferroptosis through mediating retinoid metabolism. A, The general view of FENIX assay. DTUN was used to initiate the co-autoxidation of the polyunsaturated lipids of egg phosphatidylcholine liposomes and STY-BODIPY. The process was inhibited by RTAs and monitored by the fluorescence of oxidized STY-BODIPY. B, Representative autoxidation of STY-BODIPY (1 μmol/L)–embedded liposomes (1 mmol/L; extruded to about 100 nm particle size) suspended in PBS at 37°C, initiated by the addition of 3 μL DTUN (1 mmol/L in ethanol), and inhibited by different RTAs as indicated. C, Representative autoxidation of STY-BODIPY as in B, but with varying concentrations of ROL, 13, 14-DROL, RAL, RA, and β, β-carotene. D, Cell viability in Cas-NC and RETSAT-KO or OE A549, HT1080, or H1299 cells treated with RSL3 (A549 and HT1080, 4 μmol/L for 6 hours; H1299, 40 μmol/L for 24 hours) combined with or without ROL as indicated. Data are presented as the mean ± SD; n = 3 independent repeats. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001.

Close modal

As shown in Fig. 3B and C, supplementation with ROL, RAL, or RA, but not 13, 14-DROL or β, β-carotene, substantially delayed autoxidation in a dose-dependent manner, similar to PMC. Among these metabolites, RAL exhibited the strongest radical trapping activity. This finding might be due to its distinct functional groups besides its common conjugated double bond system because it has been proposed that aldehyde groups also, to some extent, possess radical-trapping activity (30), whereas carboxylic acid does not. These results reveal that retinoids can trap radicals in the lipid bilayer, thereby subverting lipid peroxidation and subsequent ferroptosis, which is exactly in agreement with our experiments in cell models.

RETSAT promotes ferroptosis by mediating ROL's reduction to 13, 14-DROL

It is worth noting that the product of RETSAT—13, 14-DROL—failed to exhibit significant anti-ferroptosis and radical-trapping activity similar to its precursor ROL in both cell experiments and FENIX assay (Figs. 2B, C, 3B and C; Supplementary Fig. S3E and S3F). In terms of molecular structure, the conjugated double bond system of 13, 14-DROL consists of three C = C double bonds in its side chain (exocyclic double bonds), whereas ROL, RAL, and RA, have four, which might account for the significantly impaired antioxidant activity of 13, 14-DROL compared with its precursor. Taking the enzymatic function of RETSAT into account, RETSAT transforms ROL (a “powerful” RTA) into 13, 14-DROL (a relatively weaker RTA), which is in agreement with the pro-ferroptosis role of RETSAT.

Consistently, our results suggested that supplementing Cas-NC, RETSAT-KO, and RETSAT-OE cells with ROL could reduce, and even almost completely abrogate their different sensitivity to FINs (Fig. 3D; Supplementary Fig. S4D). Ultimately, these data demonstrate that RETSAT mediates ferroptosis through enzymatically transforming ROL into 13, 14-DROL, thus impairing ROL's antioxidant activity.

RETSAT exerts its pro-ferroptosis function partly through the RAR pathway

We next investigated whether other mechanisms besides ROL's radical-trapping activity also account for retinoid's anti- and RETSAT's pro-ferroptosis function. As depicted in Fig. 2A, the final oxidized product of ROL is RA, while that of 13, 14-DROL is 13, 14-dihydroretinoic acid (13, 14-DRA). Moise and colleagues suggested that although 13, 14-DRA could also bind to and activate the RAR, the major molecular target of RA, its potency is substantially lower than that of RA. Accordingly, it can be inferred that by saturating ROL into 13, 14-DROL, RETSAT may inhibit RAR activity by “indirectly” transforming RA (a “powerful” RAR agonist) into 13, 14-DRA (a relatively weaker one). Consistently, our RNA-seq data indicated an extensive activation of the RAR signaling pathway in A549, HT1080, and H1299 cells harbouring RETSAT depletion, as revealed by the broad alteration of several well-known RA-inducible genes such as ABCA1 (Fig. 4A and B; Supplementary Figs. S4E and S5A; ref. 31). Considering that these genes are also regulated by other nuclear receptors like liver X receptor and retinoid X receptor, we further employed a RAR-GFP reporter system to specifically investigate the activating level of RAR after RETSAT-KO. As expected, the loss of RETSAT significantly activated the RAR signaling pathway (Fig. 4C; Supplementary Fig. S5B). Therefore, we assessed whether RA metabolism and the RAR pathway were involved in RETSAT-related ferroptosis sensitization.

Figure 4.

RETSAT promotes ferroptosis partly through the RAR pathway. A, Heat maps exhibiting the expression levels of a series of RAR pathway–related genes in Cas-NC and RETSAT-KO A549 cells treated with vehicle (DMSO), or 2 μmol/L RSL3, or 20 μmol/L IKE for 4 hours, determined by RNA-seq, n = 2 for each group. B, mRNA levels of ABCA1 in Cas-NC and RETSAT-KO A549 cells, determined by qPCR. C, Luciferase activity in A549 cells after transfection of RAR-GFP reporter plasmid for 24 hours. D, Cell viability in A549, HT1080, or H1299 cells treated with RSL3 (A549 and HT1080, 4 μmol/L for 6 hours; H1299, 40 μmol/L for 24 hours) following pretreatment with 5 μmol/L AM580, or 5 μmol/L Ch55, and/or 10 μmol/L AGN193109 for 36 hours. E, Cell viability in H23 cells treated with RSL3 (2 μmol/L for 6 hours) or IKE (8 μmol/L for 12 hours) following pretreatment with 10 μmol/L AGN193109 for 36 hours. F, Lipid peroxidation in A549 cells treated with 2 μmol/L RSL3 for 3 hours following pretreatment with 5 μmol/L Ch55 and/or 10 μmol/L AGN193109 for 36 hours. G and H, Cell viability in Cas-NC and RETSAT-KO A549 (G) or vehicle and RETSAT-OE A549 cells (H) treated with 4 μmol/L RSL3 for 6 hours following pretreatment with 2 μmol/L RA, 5 μmol/L Ch55, and/or 10 μmol/L AGN193109 for 36 hours. I and J, Lipid peroxidation in Cas-NC and RETSAT-KO A549 (I) or vehicle and RETSAT-OE A549 cells (J) treated with 2 μmol/L RSL3 for 3 hours following pretreatment with 10 μmol/L AGN193109 (I) or 5 μmol/L Ch55 (J) for 36 hours. Data are presented as the mean ± SD; n = 3 independent repeats except for A. **, P < 0.01; ***, P < 0.001; ****, P < 0.0001.

Figure 4.

RETSAT promotes ferroptosis partly through the RAR pathway. A, Heat maps exhibiting the expression levels of a series of RAR pathway–related genes in Cas-NC and RETSAT-KO A549 cells treated with vehicle (DMSO), or 2 μmol/L RSL3, or 20 μmol/L IKE for 4 hours, determined by RNA-seq, n = 2 for each group. B, mRNA levels of ABCA1 in Cas-NC and RETSAT-KO A549 cells, determined by qPCR. C, Luciferase activity in A549 cells after transfection of RAR-GFP reporter plasmid for 24 hours. D, Cell viability in A549, HT1080, or H1299 cells treated with RSL3 (A549 and HT1080, 4 μmol/L for 6 hours; H1299, 40 μmol/L for 24 hours) following pretreatment with 5 μmol/L AM580, or 5 μmol/L Ch55, and/or 10 μmol/L AGN193109 for 36 hours. E, Cell viability in H23 cells treated with RSL3 (2 μmol/L for 6 hours) or IKE (8 μmol/L for 12 hours) following pretreatment with 10 μmol/L AGN193109 for 36 hours. F, Lipid peroxidation in A549 cells treated with 2 μmol/L RSL3 for 3 hours following pretreatment with 5 μmol/L Ch55 and/or 10 μmol/L AGN193109 for 36 hours. G and H, Cell viability in Cas-NC and RETSAT-KO A549 (G) or vehicle and RETSAT-OE A549 cells (H) treated with 4 μmol/L RSL3 for 6 hours following pretreatment with 2 μmol/L RA, 5 μmol/L Ch55, and/or 10 μmol/L AGN193109 for 36 hours. I and J, Lipid peroxidation in Cas-NC and RETSAT-KO A549 (I) or vehicle and RETSAT-OE A549 cells (J) treated with 2 μmol/L RSL3 for 3 hours following pretreatment with 10 μmol/L AGN193109 (I) or 5 μmol/L Ch55 (J) for 36 hours. Data are presented as the mean ± SD; n = 3 independent repeats except for A. **, P < 0.01; ***, P < 0.001; ****, P < 0.0001.

Close modal

Although we have confirmed that RA protected cells from lipid peroxidation and ferroptosis (Fig. 2B and C; Supplementary Fig. S3E and S3F), this property is largely due to the radical-trapping activity of its conjugated double-bond system. To exclude this confounding effect and specifically focus on the RAR pathway, we selected two RAR agonists, AM580 and Ch55, and ruled out their potential radical-trapping ability through the FENIX assay (Fig. 3B). As exhibited in Fig. 4DF and Supplementary Fig. S5C–S5D, both compounds dramatically desensitized cells to lipid peroxidation and ferroptotic cell death induced by FINs, of which, Ch55 was more effective. Nevertheless, AGN193109, a RAR antagonist, failed to significantly sensitize the cells to ferroptosis, probably because basal RAR activation levels were low in these cell lines. To verify this hypothesis, we investigated the endogenous basal RAR signaling in a panel of cancer cell lines using ABCA1 as an indicator and found that when treated with AGN193109, H23, the non–small cell lung cancer cell line, exhibited the strongest reduction of ABCA1 expression (Supplementary Fig. S5E). This result was further confirmed by the RAR-GFP reporter assay, suggesting a high basal RAR signaling in this cell line (Supplementary Fig. S5F). As expected, in the absence of exogenous RAR agonist, RAR inhibition with AGN193109 in H23 significantly promoted ferroptosis (Fig. 4E). Whereas in A549, HT1080, and H1299, exogenous supplementation of RAR agonist was needed to boost RAR activity, thus revealing the difference in ferroptosis sensitivity between vehicle- and AGN193109-treated cells. Moreover, RAR inhibition markedly reversed the increased resistance of RETSAT-KO cells to FINs, while Ch55 treatment in RETSAT-OE cells generated the opposite effect (Fig. 4GJ; Supplementary Fig. S5G and S5H). Besides, we also generated RARA-KD A549, HT1080, and H23 with shRNA and obtained similar results with AGN193109 treatment, thus excluding the off-target effect of RAR inhibitor (Supplementary Fig. S5I–S5K).

Notably, AGN193109 or RARA-KD, and Ch55 could only “partly” rather than “completely” abrogate the effect of RETSAT alteration. Combined with our previous findings, we speculated that this interesting phenomenon was likely due to that RETSAT regulates ferroptosis through both abolishing retinoids’ radical-trapping property and inhibiting the RAR pathway. Collectively, these findings suggest that RAR activation desensitized cancer cells to ferroptotic cell death, which accounts to a certain extent for RETSAT's pro-ferroptosis role. This newly uncovered mechanism is independent of the radical-trapping properties of ROL and its derived metabolites.

Pseudotargeted lipidomics revealed that RETSAT depletion promotes MUFA generation

Considering that ferroptosis is fueled by lipid peroxidation and the RAR signaling pathway regulates lipid metabolism (32–34), we performed pseudotargeted lipidomic analysis to characterize lipid profile alterations in RETSAT-KO and corresponding parental A549 cells treated with vehicle (DMSO), IKE, and RSL3, respectively. Unsupervised PCA (visualized as PCA score plot) of extracted lipid features exhibited clear separation and tight clustering among the groups (Fig. 5A). Consistent with previous reports, the relative abundance of several lipid species, especially lysophospholipids, was markedly increased in response to IKE treatment (Fig. 5B; Supplementary Fig. S6A). This phenomenon is widely believed to be the consequence of the degradation by enzymatic cleavage of the oxidized PUFA tail upon FINs treatment (9, 35, 36). Unexpectedly, RSL3 treatment failed to yield such alterations, which might be due to the heterogeneity in concentration and treatment time. Compared with parental cells, we observed significant increases (FDR-corrected P < 0.05, |log2 (fold change)| > 0.5) of many fatty acids and phospholipids in A549 cells harboring RETSAT-KO. It is worth noting that among these lipid species, those containing MUFA (or acyl) chains, such as palmitoleic acid (C16:1) and oleic acid (C18:1), exhibited a much larger increase than PUFA. In contrast, saturated fatty acids (SFA; or acyl) were decreased in RETSAT-KO cells (Fig. 5B and C; Supplementary Fig. S6A and S6B). This observation was concordant with DMSO and FINs treatment findings, indicating that RETSAT-KO led to enhanced conversion from SFA to MUFA during glycerophospholipid synthesis.

Figure 5.

RETSAT-RAR pathway promotes ferroptosis through SCD-mediated MUFA generation. A–C, Pseudotargeted lipidomic data for Cas-NC and RETSAT-KO A549 cells treated with vehicle (DMSO), or 1 μmol/L RSL3, or 20 μmol/L IKE for 24 hours. A, PCA. B, Heat maps exhibiting the relative abundance of a series of SFA, MUFA, and PUFA or corresponding phospholipids in these cells. C, Volcano plots showing the difference in the abundance of fatty acids and phospholipids between Cas-NC and RETSAT-KO A549 cells treated as indicated. D, Cell viability in A549 cells treated with RSL3 (4 μmol/L for 6 hours) or IKE (40 μmol/L for 48 hours) following pretreatment with 10 μmol/L A939572 for 36 hours. E, Lipid peroxidation in A549 cells treated with 2 μmol/L RSL3 for 3 hours following pretreatment with 10 μmol/L A939572 for 36 hours. F, mRNA levels of a series of lipid metabolism–associated genes in Cas-NC and RETSAT-KO A549 cells. G, Protein levels of SCD in Cas-NC and RETSAT-KO A549 cells treated with 2 μmol/L RA for 48 hours. H, mRNA levels of SCD in A549 cells treated with 2 μmol/L RA, or 5 μmol/L Ch55, and/or 10 μmol/L AGN193109 for 12 hours. I, Protein levels of SCD in A549 cells treated with 2 μmol/L RA, or 5 μmol/L Ch55, and/or 10 μmol/L AGN193109 for 48 hours. J, Targeted lipidomics determined the abundance of several MUFAs in Cas-NC or RETSAT-KO A549 cells treated with 2 μmol/L RA or 5 μmol/L Ch55 for 24 hours. K, The relative positions and sequences of predicted BSs of RARα in the promoter region of SCD. L, The binding between RARα and the promoter region of SCD was quantified through a ChIP assay, followed by qPCR. M, Luciferase activity in Cas-NC and RETSAT-KO HEK293T cells treated with 2 μmol/L RA, or 5 μmol/L Ch55, or 10 μmol/L AGN193109 for 24 hours, following transfection of indicated plasmids. N, Cell viability in Cas-NC and RETSAT-KO A549 cells treated with 4 μmol/L RSL3 for 6 hours following pretreatment with 10 μmol/L A939572 for 36 hours. O, Lipid peroxidation in Cas-NC and RETSAT-KO A549 cells treated with 2 μmol/L RSL3 for 3 hours following pretreatment with 10 μmol/L A939572 for 36 hours. P, Cell viability in A549 cells treated with 4 μmol/L RSL3 for 6 hours following pretreatment with 10 μmol/L A939572 combined with or without 2 μmol/L RA or 5 μmol/L Ch55 for 36 hours. Q, Cell viability in vehicle and RETSAT-OE A549 cells treated with 4 μmol/L RSL3 for 6 hours following pretreatment with 20 μmol/L palmitoleic acid or oleic acid for 36 hours. R, Cell viability in A549 cells treated with 4 μmol/L RSL3 for 6 hours following pretreatment with 20 μmol/L palmitoleic acid or oleic acid combined with or without 10 μmol/L A939572 for 36 hours. Data are presented as the mean ± SD; n = 3 independent repeats. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001; ns, not significant.

Figure 5.

RETSAT-RAR pathway promotes ferroptosis through SCD-mediated MUFA generation. A–C, Pseudotargeted lipidomic data for Cas-NC and RETSAT-KO A549 cells treated with vehicle (DMSO), or 1 μmol/L RSL3, or 20 μmol/L IKE for 24 hours. A, PCA. B, Heat maps exhibiting the relative abundance of a series of SFA, MUFA, and PUFA or corresponding phospholipids in these cells. C, Volcano plots showing the difference in the abundance of fatty acids and phospholipids between Cas-NC and RETSAT-KO A549 cells treated as indicated. D, Cell viability in A549 cells treated with RSL3 (4 μmol/L for 6 hours) or IKE (40 μmol/L for 48 hours) following pretreatment with 10 μmol/L A939572 for 36 hours. E, Lipid peroxidation in A549 cells treated with 2 μmol/L RSL3 for 3 hours following pretreatment with 10 μmol/L A939572 for 36 hours. F, mRNA levels of a series of lipid metabolism–associated genes in Cas-NC and RETSAT-KO A549 cells. G, Protein levels of SCD in Cas-NC and RETSAT-KO A549 cells treated with 2 μmol/L RA for 48 hours. H, mRNA levels of SCD in A549 cells treated with 2 μmol/L RA, or 5 μmol/L Ch55, and/or 10 μmol/L AGN193109 for 12 hours. I, Protein levels of SCD in A549 cells treated with 2 μmol/L RA, or 5 μmol/L Ch55, and/or 10 μmol/L AGN193109 for 48 hours. J, Targeted lipidomics determined the abundance of several MUFAs in Cas-NC or RETSAT-KO A549 cells treated with 2 μmol/L RA or 5 μmol/L Ch55 for 24 hours. K, The relative positions and sequences of predicted BSs of RARα in the promoter region of SCD. L, The binding between RARα and the promoter region of SCD was quantified through a ChIP assay, followed by qPCR. M, Luciferase activity in Cas-NC and RETSAT-KO HEK293T cells treated with 2 μmol/L RA, or 5 μmol/L Ch55, or 10 μmol/L AGN193109 for 24 hours, following transfection of indicated plasmids. N, Cell viability in Cas-NC and RETSAT-KO A549 cells treated with 4 μmol/L RSL3 for 6 hours following pretreatment with 10 μmol/L A939572 for 36 hours. O, Lipid peroxidation in Cas-NC and RETSAT-KO A549 cells treated with 2 μmol/L RSL3 for 3 hours following pretreatment with 10 μmol/L A939572 for 36 hours. P, Cell viability in A549 cells treated with 4 μmol/L RSL3 for 6 hours following pretreatment with 10 μmol/L A939572 combined with or without 2 μmol/L RA or 5 μmol/L Ch55 for 36 hours. Q, Cell viability in vehicle and RETSAT-OE A549 cells treated with 4 μmol/L RSL3 for 6 hours following pretreatment with 20 μmol/L palmitoleic acid or oleic acid for 36 hours. R, Cell viability in A549 cells treated with 4 μmol/L RSL3 for 6 hours following pretreatment with 20 μmol/L palmitoleic acid or oleic acid combined with or without 10 μmol/L A939572 for 36 hours. Data are presented as the mean ± SD; n = 3 independent repeats. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001; ns, not significant.

Close modal

Currently, it is widely accepted that PUFAs are highly susceptible to ROS, thus leading to lipid peroxidation and “fueling” ferroptosis cascade (4). On the contrary, Magtanong and colleagues proposed that MUFAs could suppress this process by promoting the displacement of PUFAs from plasma membrane phospholipids (10). On the basis of this theory, our lipidomic study results were in agreement with the increased resistance toward ferroptosis in RETSAT-KO cell lines. Meanwhile, the impact of the RA-related pathway on lipid metabolism has been investigated by several researchers in the past few years (33, 34, 37). Together, these results indicate that the RETSAT-RAR pathway probably mediates ferroptosis by promoting MUFA generation.

The RETSAT-RAR pathway sensitizes cells to ferroptosis through SCD-mediated MUFA generation

We then explored how RAR activation caused by RETSAT-KO promoted MUFA generation and protected cells against ferroptosis. Heidenreich and colleagues suggested that RETSAT positively regulated both expression and nuclear accumulation of ChREBP, a central transcription factor involved in glucolipid metabolism, activating the transcription of its target genes, promoting de novo lipogenesis, and resulting in steatosis in mouse liver (38). However, our results showed that neither RETSAT depletion nor RA treatment significantly altered the expression or nuclear translocation of ChREBP in A549 and HT1080 cells (Supplementary Fig. S6C). Besides, Lee and colleagues demonstrated that energy stress–mediated AMPK activation inhibits ferroptosis, and the connection between AMPK and ferroptosis was established by AMPK-mediated phosphorylation of acetyl-CoA carboxylase and PUFA biosynthesis. Because it has been proposed by multiple researchers that RA activates the AMPK signaling pathway in different models (39, 40), we assessed whether this potential mechanism could be applied to our study, but found that RETSAT depletion did not significantly affect AMPK phosphorylation, irrespective of the use of RA supplementation (Supplementary Fig. S6D). Taken together, these findings suggest that it is less likely that the RETSAT-RAR pathway mediates ferroptosis through regulating ChREBP or AMPK, at least in the cell models we employed in this research.

Having excluded the involvements of ChREBP and AMPK in this process, all the evidence mentioned above pointed to SCD, the enzyme that catalyzes the conversion from SFA to MUFA by utilizing O2 and electrons from cytochrome b5 to introduce the first double bond into SFA (41). Our RNA-seq results suggested that SCD was upregulated in the RETSAT-KO group (Fig. 4A; Supplementary Fig. S4E), prompting us to validate the recently reported anti-ferroptosis function of SCD in A549, HT1080, and H1299 cells through SCD specific inhibitor A939572 (Fig. 5D and E; Supplementary Fig. S6E and S6F; ref. 42). Likewise, shRNA-mediated SCD knockdown generated a similar ferroptosis-sensitizing effect (Supplementary Fig. S6G–S6I). We further investigated whether SCD was a major downstream target of RETSAT and RAR that mediated sensitivity to ferroptosis induction. Consistent with RNA-seq, we found that loss of RETSAT significantly enhanced SCD expression levels (Fig. 5F and G; Supplementary Fig. S6J and S6K). Treatment with RAR agonists, including RA and Ch55, exerted similar effects, while RAR antagonist AGN193109 and RARA-KD reversed this alteration (Fig. 5H and I; Supplementary Fig. S6L–S6N). Few reports have indicated RA's impact on SCD expression and several other lipid metabolism–related genes such as SREBF1, FASN, and ACACA (33, 34, 43). However, RETSAT depletion did not significantly alter these potential targets in the tested cell lines in our study (Fig. 5F; Supplementary Fig. S6J). Furthermore, targeted lipidomics demonstrated that both RETSAT-KO and RAR agonist treatment substantially increased the levels of multiple palmitoleic acid–containing and oleic acid–containing phospholipids (Fig. 5J). These findings confirmed that RETSAT-KO and subsequent RAR activation induced SCD expression and subsequent MUFA generation.

To ascertain whether RAR directly regulated SCD, we turned to the ENCODE database and found the enrichment of RARα, the dominant subtype of RAR, at the promoter region of the SCD gene (Supplementary Fig. S6O). Further nucleotide sequence analysis through JASPAR identified a series of potential RARα binding sites (BS), also known as RA response elements, in this region (Fig. 5K). As expected, our chromatin immunoprecipitation (ChIP) assays with a RARα specific antibody revealed significant recruitment of RARα to the BS2 and BS3&4 (Fig. 5L). To further confirm the binding specificity, we carried out a luciferase reporter gene assay using an SCD promoter plasmid [wild-type (WT)] and introduced mutations (MUT) in the RARα BS. As exhibited in Fig. 5M, RETSAT depletion, as well as RAR agonist treatment, resulted in extensive transcriptional activation, while the RAR antagonist led to transcriptional inhibition, as reflected by the luciferase intensity in cells transfected with the WT but not the all-MUT plasmid. Mutations at the three BSs also abrogated these changes to varying degrees. These data confirmed that RAR directly activates SCD transcription by binding to its promoter.

Furthermore, pharmacologic inhibition of SCD by A939572 partly abrogated the inhibitory effect of RETSAT-KO and RA treatment on ferroptosis, while Ch55-induced ferroptosis resistance was completely blocked (Fig. 5NP; Supplementary Fig. S7A and S7B). Moreover, exogenous supplementation of palmitoleic acid (C16:1) or oleic acid (C18:1) partially restored the excessive sensitivity to RSL3 in cells harboring RETSAT-OE (Fig. 5Q; Supplementary Fig. S7C), whereas the rescue effects of the two MUFAs, especially the oleic acids, were more significant in cells treated with A939572 (Fig. 5R; Supplementary Fig. S7D). Taken together, these findings suggest that by indirectly transforming RA to 13, 14-DRA, RETSAT suppresses the activity of the RAR signaling pathway, thereby retarding the transcription of RAR's target SCD and resulting in decreased MUFA generation, which finally leads to ferroptosis sensitization. This mechanism and the mitigated radical trapping of retinoids, which are independent of each other, contribute to the pro-ferroptosis function of RETSAT together.

Retinoid metabolism alters tumor ferroptosis in vivo

To further elucidate the physiologic significance of retinoid metabolism in ferroptosis regulation, we investigated whether RETSAT depletion or exogenously supplemented ROL or RA impacted tumor growth in vivo. Tumor xenografts were generated with Cas-NC and RETSAT-KO A549 cell lines, and IKE was used as the FIN for its excellent stability in animal experiments. RETSAT depletion did not cause marked tumor growth inhibition or promotion, consistent with in vitro assays, indicating that RETSAT was quite a dispensable gene under normal culture conditions. IKE treatment resulted in a significant reduction in the growth of the Cas-NC tumors but not in the RETSAT-KO tumors. Meanwhile, treatment with ROL or RA largely restored the growth of Cas-NC tumors treated with IKE (Fig. 6A). Notably, both RETSAT depletion and retinoid supplementation mitigated the staining of 4-HNE, a lipid peroxidation marker, in resected tumor samples (Fig. 6B). In all animal studies, drug treatment did not significantly affect animal weight, suggesting that the treatment was well tolerated in vivo (Supplementary Fig. S7E). Together, our findings substantiate that enhanced retinoids level caused by RETSAT-KO or exogenous supplementation desensitized tumors to ferroptosis in vivo and raised the possibility of targeting retinoid metabolism in cancer treatment.

Figure 6.

Retinoid metabolism alters tumor ferroptosis in vivo and correlates with LUAD patients’ prognosis. A, Image of resected tumors from Cas-NC and RETSAT-KO A549 mice xenografts. Groups of mice were treated as indicated (n = 8 per group at the beginning, while mice dying before the endpoint were excluded). The growth of tumor volumes and the final weight of resected tumors are also shown (right). B, Representative IHC images of Cas-NC and RETSAT-KO A549 xenograft tumors, with the indicated treatments. Scale bars, 40 μm. C, KM curves showing the prognostic value of RETSAT and STRA6 in patients with LUAD retrieved from the KM plotter database. D, The mRNA levels of RETSAT and STRA6 in the tumor and normal tissue of patients with LUAD enrolled from TCGA database, determined by RNA-seq. E, KM curves showing the prognostic value of RETSAT and STRA6 in patients with LUAD, retrieved from our institute. F, The mRNA levels of RETSAT and STRA6 in the tumor and normal tissue of patients with LUAD enrolled from our institute, determined by RNA-seq. G, Representative IHC images of the tumor and adjacent normal tissue of patients with LUAD. Scale bars, 100 μm; n = 28. The staining results are summarized below. H, The protein levels of RETSAT and STRA6 in the paired tumor and adjacent normal tissue, determined by Western blotting (n = 6). I, A schematic model depicting how retinoids, and the key factors involved in their metabolism, including RETSAT and STRA6, regulate ferroptosis from two perspectives. See the main text for a detailed description. Data are presented as the mean ± SD. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001. (I, Created with BioRender.com.)

Figure 6.

Retinoid metabolism alters tumor ferroptosis in vivo and correlates with LUAD patients’ prognosis. A, Image of resected tumors from Cas-NC and RETSAT-KO A549 mice xenografts. Groups of mice were treated as indicated (n = 8 per group at the beginning, while mice dying before the endpoint were excluded). The growth of tumor volumes and the final weight of resected tumors are also shown (right). B, Representative IHC images of Cas-NC and RETSAT-KO A549 xenograft tumors, with the indicated treatments. Scale bars, 40 μm. C, KM curves showing the prognostic value of RETSAT and STRA6 in patients with LUAD retrieved from the KM plotter database. D, The mRNA levels of RETSAT and STRA6 in the tumor and normal tissue of patients with LUAD enrolled from TCGA database, determined by RNA-seq. E, KM curves showing the prognostic value of RETSAT and STRA6 in patients with LUAD, retrieved from our institute. F, The mRNA levels of RETSAT and STRA6 in the tumor and normal tissue of patients with LUAD enrolled from our institute, determined by RNA-seq. G, Representative IHC images of the tumor and adjacent normal tissue of patients with LUAD. Scale bars, 100 μm; n = 28. The staining results are summarized below. H, The protein levels of RETSAT and STRA6 in the paired tumor and adjacent normal tissue, determined by Western blotting (n = 6). I, A schematic model depicting how retinoids, and the key factors involved in their metabolism, including RETSAT and STRA6, regulate ferroptosis from two perspectives. See the main text for a detailed description. Data are presented as the mean ± SD. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001. (I, Created with BioRender.com.)

Close modal

Retinoid metabolism correlates with LUAD patient's prognosis

To investigate the clinical relevance of retinoid metabolism, we interrogated Kaplan–Meier (KM) plotter, an online database integrating the RNA expression and survival data derived from multiple sources (https://kmplot.com/analysis/; ref. 44), and found that lower RETSAT expression levels correlated with shorter overall survival in patients with LUAD while STRA6 exhibited opposite results (Fig. 6C). Meanwhile, the RNA-seq data analysis from TCGA database revealed significant downregulation of RETSAT and upregulation of STRA6 in LUAD tissue compared with normal tissue (Fig. 6D). We validated these findings using the transcriptomic data of 35 paired clinical human LUAD lesions and adjacent noncancerous tissue samples, as well as corresponding clinical follow-up data, in our institute (Fig. 6E and F; ref. 45). Next, IHC was also employed for analysis of the protein levels of RETSAT and STRA6 in a cohort of 28 clinicopathologically characterized LUAD cases. In agreement with the RNA-seq data, quantitative analysis of the IHC results demonstrated that RETSAT was abundantly expressed in normal lung tissues but weakly expressed in LUAD tissues, while STRA6 showed the opposite trend (Fig. 6G). Besides, similar results were observed in immunoblot analysis (Fig. 6H). These results indicated enhanced retinoid synthesis in cancer cells, which counteracted excessive ROS production during their growth, proliferation, and metabolic reprogramming, thus protecting them from ferroptosis and leading to increased malignancy. In addition, based on our previously published single-cell sequencing data of 8 patients with LUAD (12), after dimension reduction and cell-type annotation, we also noticed that the expression of RETSAT and STRA6 was mainly restricted to cancer cells rather than T or B cells (Supplementary Fig. S7F). Given that depletion of RETSAT and STRA6 did not significantly alter cell proliferation rate under normal culture conditions (Fig. 1P; Supplementary Fig. S7G), we propose that retinoid metabolism, especially the two key proteins, is a promising therapeutic target for activating ferroptosis in human cancer without causing severe toxicity.

In recent years, the involvement of ferroptosis in tumor development and treatment has attracted considerable attention (4). On the one hand, several cancer-associated signaling pathways have been shown to govern ferroptosis in cancer cells. For instance, p53, the famous cancer suppressor, promotes ferroptosis by repressing SLC7A11 expression (46). Inhibition of mTORC1, a critical factor regulating tumor metabolic reprogramming, sensitizes cancer cells to ferroptosis by decreasing GPX4 synthesis (47). On the other hand, the metabolic feature of cancer cells, their high load of ROS and their specific mutations render some of them intrinsically susceptible to ferroptosis, thereby exposing vulnerabilities that could be therapeutically targeted (4). Furthermore, ferroptosis could also be triggered by various cancer therapies, including radiotherapy, chemotherapy, and immunotherapy (12, 48, 49). Therefore, the modulation of ferroptosis is increasingly recognized as a potential avenue for developing therapeutics against malignancies.

There are several reports dating back 15 years of RETSAT's role in peroxide-mediated cell death, while its effects have not been discussed in much depth in relation to ferroptosis because the term was coined only 10 years ago (32, 50). In the current study, after discovering RETSAT as a pro-ferroptosis gene based on a CRISPR/Cas9 dataset, we provided compelling evidence of retinoids as a novel anti-ferroptosis system that exerts its function from two perspectives (Fig. 6I). On the one hand, several types of retinoids, such as ROL, RAL, and RA, possessing radical-trapping properties, exert an unexpected protective effect against lipid peroxidation and ferroptosis, whereas 13, 14-DROL and β, β-carotene failed to do so. On the other hand, RA, rather than 13, 14-DRA, contributes to ferroptosis resistance by interacting with nuclear RARs and activating the transcription of SCD, which introduces the first double bond into SFA, thus promoting MUFA generation. Notably, as a key regulator of retinoid metabolism, RETSAT transforms ROL to 13, 14-DROL. Our results demonstrated that loss of RETSAT leads to accumulation of ROL and its downstream metabolites RA, accounting for the ferroptosis-resistant phenotype of RETSAT-KO cancer cell lines. Meanwhile, depleting STRA6, the membrane receptor mediating cellular ROL influx, sensitized cells to FINs, further substantiating the role of retinoids in ferroptosis. Because previous researchers have investigated the role of retinoids in carcinogenesis and tumor development in multiple clinical trials, for example, serum β, β-carotene, and ROL levels were found to be inversely related to lung cancer risk, our findings on retinoids’ anti-ferroptosis property raised the question that whether should physicians monitor serum retinoids level and avoid excessive retinoids supplementation for patients diagnosed with lung cancer (51–55).

Initially, when we were exploring the mechanism by which RETSAT mediates ferroptosis, inspired by the role RETSAT plays in lipid metabolism and adipogenesis (38), and by its location in the endoplasmic reticulum (22, 56), we focused on the synthesis and desaturation of PUFA and MUFA and sought to exclude the involvement of ROL and 13, 14-DROL in this process, as previous researchers did (50). However, to our surprise, ROL protected cells against ferroptosis when exogenously supplemented to the medium. Therefore, to some extent, the anti-ferroptosis property of ROL was unearthed by accident. Mechanistically, ferroptosis is a radical chain reaction characterized by the continuous generation of lipid hydroperoxides initiated by the iron-dependent Fenton reaction. In addition to the canonical GPX4/GSH system, an increasing number of RTAs have emerged, including endogenous metabolites ubiquinol, BH4, and newly discovered squalene, 7-DHC, and vitamin K (3, 6–9, 57), as well as exogenous compounds, vitamin E, Fer-1, and Lip-1. They are able to trap chain-propagating lipid peroxyl radicals and prevent ferroptosis from entering the irreversible phase (32, 58). The discovery of these GPX4-independent ferroptosis defense mechanisms has aroused interest in other intracellular antioxidants. In addition to its effect on the maintenance of normal vision, ROL has been proposed as an antioxidant for several years (26, 59). However, this property has not received enough attention because most of ROL's predominant biological functions are widely thought to be attributed to its active metabolite, RA. At present, there are no convincing published studies on the role of ROL per se as a membrane-permeant antioxidant. To fill this knowledge gap, we estabolism a yet-unrecognized role of ROL and its metabolite in preventing lipid peroxidation through cellular experiments and the FENIX assay—an in vitro cell-free system. Unexpectedly, although β, β-carotene is the precursor of ROL and possesses a similar molecular structure, and Stockwell's lab showed that β-carotene (0.2 μmol/L) as an antioxidant could rescue erastin-induced cell death (BJ cells, fibroblast), although the name of “ferroptosis” had not been coined at that time (60), it failed to exhibit any anti-ferroptosis effect in our tests. An increasing body of evidence suggests that carotenoids are good singlet oxygen (1O2) quenchers and serve as antioxidants in various diseases (61, 62). This phenomenon might be attributed to several reasons. On the one hand, 1O2 is not the main source of ROS in the cascade of ferroptosis. More importantly, the ferroptosis inhibiting effect of a specific metabolite in different cell types is determined by numerous intrinsic and extrinsic elements such as the influx and efflux of antioxidants, environmental pH, oxygen concentration, and kinetics of the redox reaction (14). Similar to this phenomenon, it was also shown that RAL (5–20 μmol/L) induced ferroptosis in photoreceptor cells, further demonstrating that whether a metabolite is “pro-ferroptosis” or “anti-ferroptosis” is highly context dependent and concentration dependent (63).

The metabolism of ROL is quite complicated. After being absorbed in the intestine, dietary ROL is transported in plasma in a 1:1 complex with RBP4 (23), then shifted into cells by STRA6. As expected, we found that STRA6-KO led to ferroptosis sensitization. However, STRA6 knockout only slightly abrogated the anti-ferroptosis effect conferred by ROL supplementation or RETSAT-KO, this was due to only a fraction of exogenous ROL could bind to RBP4 in cultural media supplemented with FBS and was uptaken by STRA6, and ROL might also pass through the lipid bilayer in a STRA6-independent manner given its lipid solubility (28). Furthermore, after entry into cells, ROL's esterification, storage, isomerization, and oxidation are modulated by numerous enzymes (23, 24). Considering that different metabolites of ROL may have different radical-trapping activity, as exemplified by RAL and RA, further investigation is necessary to pinpoint the exact roles of these proteins in the regulation of ROL homeostasis and cell response to ferroptosis.

Overwhelming evidence substantiates the implication of RETSAT in hepatic glucose and lipid metabolism (22, 38, 50). Therefore, we performed pseudotargeted lipidomics, thereby excluding the involvement of ChREBP and AMPK pathways in RETSAT-mediated ferroptosis and finally fixing on SCD-induced MUFA generation. There has been growing appreciation that fatty acid metabolism profoundly influences ferroptosis. On the one hand, MUFAs compete with PUFAs to integrate into the plasma membrane, thus suppressing ferroptosis (10). Cancer cell membranes are characterized by an increased MUFA/PUFA ratio, resulting in reduced susceptibility to ferroptosis (64). On the other hand, the tumor microenvironment is enriched with fatty acids, which induces intratumoral CD8+ T-cell ferroptosis in a CD36-dependent manner (65). There has been an increasingly sophisticated understanding of the role the pathways regulating lipid synthesis, lipid storage, and β-oxidation plays in ferroptosis, such as AMPK (35) and mTOR (47). This evidence connects ferroptosis to the whole complicated metabolic network (not restricted to lipid), thus pointing out the future research direction in this area.

The anti-ferroptosis effect of SCD has been widely reported over the years (66). For the regulation of SCD expression, researchers have found that a variety of transcription factors, like SREBP and PPARα, can bind to its promoter (41, 67). In the current study, we confirmed the role of RAR in these candidates and bridged the relationship between SCD-mediated ferroptosis resistance and the RAR signaling pathway. As the active derivative of ROL, RA participates in various physiologic processes such as reproduction, differentiation, immunity, and metabolism via its nuclear receptors (59). Although mounting evidence suggests the involvement of RA in FA synthesis mediated by FASN, SREBP, and ACC (33, 34, 43), we failed to observe substantial alterations of these genes in the cell lines employed, suggesting that RA's function in lipid metabolism is a context-dependent process.

Our study linking retinoids to ferroptosis raises an intriguing question regarding whether retinoid regulation on ferroptosis plays a role in tumor biology and immunity. It is widely believed that ROL and RA are tumor suppressors because they are known to induce the differentiation of various types of tumor cells, especially acute promyelocytic leukemia (68). In contrast, RA possesses tumor-promoting properties. Haldar and colleagues proposed that tumor cells could evade immune responses by producing excessive RA into the tumor microenvironment, which directs intratumoral monocytes to differentiate into immunosuppressive macrophages rather than immunostimulatory dendritic cells. Therefore, blocking RA production enhances antitumor T-cell immunity (69). Importantly, ferroptosis also plays a significant role in tumor immunity (4, 70). On the one hand, CD8+ T-cell exerts its anticancer function partly through enhancing ferroptosis-specific lipid peroxidation in tumor cells (48). On the other hand, ferroptosis of intratumoral immune cells like CD8+ T and regulatory T cells also dampens their effector function. In light of the theories that ROL and RA inhibit cancer cell ferroptosis and impair the function of antitumor T cells, as suggested by our findings and Halder and colleagues’ study, it is tempting to consider ROL/RA blockage in tumor therapy, for instance, by activating RETSAT, which not only promotes cancer cell ferroptosis but also enhances antitumor immune responses.

Collectively, through its enzymatic function, RETSAT mitigates both retinoids’ radical trapping property and RA's transcriptional activating effect. On the basis of its ferroptosis-mediating role, we identified retinoids as a stand-alone parallel system, which cooperates with other regulators like GPX4, FSP1, and DHODH to suppress ferroptosis from two aspects. Future studies are warranted to further dissect the potential role of retinoid-induced ferroptosis inhibition in tumor development and treatment.

B. Gan reports personal fees from Guidepoint Global, Cambridge Solutions, and NGM Bio outside the submitted work. C. Zhan reports grants from Natural Science Foundation of Shanghai and Special Foundation for Supporting Biomedical Technology of Shanghai during the conduct of the study. No disclosures were reported by the other authors.

G. Bi: Conceptualization, resources, data curation, software, formal analysis, validation, investigation, visualization, methodology, writing–original draft. J. Liang: Conceptualization, data curation, software, formal analysis, writing–original draft. G. Shan: Conceptualization, validation, investigation, visualization, methodology, writing–original draft. Y. Bian: Investigation, visualization. Z. Chen: Visualization, methodology. Y. Huang: Software, validation. T. Lu: Formal analysis, validation. M. Li: Data curation, investigation. V. Besskaya: Software, formal analysis. M. Zhao: Formal analysis, validation. H. Fan: Supervision, writing–review and editing. Q. Wang: Conceptualization, supervision, project administration, writing–review and editing. B. Gan: Supervision, writing–review and editing. C. Zhan: Conceptualization, supervision, funding acquisition, project administration, writing–review and editing.

The authors thank Dr. Jing Wu for providing liposomes. The authors thank Home for Researchers (www.home-for-researchers.com) for editing the article. They thank OE Biotech Co., Ltd. (Shanghai, China) for their assistance on lipidomics, single-cell RNA-seq, and bioinformatic analysis. This research was supported by the Natural Science Foundation of Shanghai (no. 22ZR1411900 to C. Zhan) and the Special Foundation for Supporting Biomedical Technology of Shanghai (no. 22S11900300 to C Zhan).

The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.

Note: Supplementary data for this article are available at Cancer Research Online (http://cancerres.aacrjournals.org/).

1.
Stockwell
BR
,
Friedmann Angeli
JP
,
Bayir
H
,
Bush
AI
,
Conrad
M
,
Dixon
SJ
, et al
.
Ferroptosis: a regulated cell death nexus linking metabolism, redox biology, and disease
.
Cell
2017
;
171
:
273
85
.
2.
Jiang
X
,
Stockwell
BR
,
Conrad
M
.
Ferroptosis: mechanisms, biology and role in disease
.
Nat Rev Mol Cell Biol
2021
;
22
:
266
82
.
3.
Yang
WS
,
SriRamaratnam
R
,
Welsch
ME
,
Shimada
K
,
Skouta
R
,
Viswanathan
VS
, et al
.
Regulation of ferroptotic cancer cell death by GPX4
.
Cell
2014
;
156
:
317
31
.
4.
Lei
G
,
Zhuang
L
,
Gan
B
.
Targeting ferroptosis as a vulnerability in cancer
.
Nat Rev Cancer
2022
;
22
:
381
96
.
5.
Cerutti
PA
.
Prooxidant states and tumor promotion
.
Science
1985
;
227
:
375
81
.
6.
Doll
S
,
Freitas
FP
,
Shah
R
,
Aldrovandi
M
,
da Silva
MC
,
Ingold
I
, et al
.
FSP1 is a glutathione-independent ferroptosis suppressor
.
Nature
2019
;
575
:
693
8
.
7.
Bersuker
K
,
Hendricks
JM
,
Li
Z
,
Magtanong
L
,
Ford
B
,
Tang
PH
, et al
.
The CoQ oxidoreductase FSP1 acts parallel to GPX4 to inhibit ferroptosis
.
Nature
2019
;
575
:
688
92
.
8.
Mao
C
,
Liu
X
,
Zhang
Y
,
Lei
G
,
Yan
Y
,
Lee
H
, et al
.
DHODH-mediated ferroptosis defence is a targetable vulnerability in cancer
.
Nature
2021
;
593
:
586
90
.
9.
Kraft
VAN
,
Bezjian
CT
,
Pfeiffer
S
,
Ringelstetter
L
,
Müller
C
,
Zandkarimi
F
, et al
.
GTP cyclohydrolase 1/tetrahydrobiopterin counteract ferroptosis through lipid remodeling
.
ACS Cent Sci
2020
;
6
:
41
53
.
10.
Magtanong
L
,
Ko
P-J
,
To
M
,
Cao
JY
,
Forcina
GC
,
Tarangelo
A
, et al
.
Exogenous monounsaturated fatty acids promote a ferroptosis-resistant cell state
.
Cell Chem Biol
2019
;
26
:
420
32
.
11.
Uphoff
CC
,
Drexler
HG
.
Detection of mycoplasma contaminations
.
Methods Mol Biol
2005
;
290
:
13
23
.
12.
Bi
G
,
Liang
J
,
Zhao
M
,
Zhang
H
,
Jin
X
,
Lu
T
, et al
.
miR-6077 promotes cisplatin/pemetrexed resistance in lung adenocarcinoma via CDKN1A/cell cycle arrest and KEAP1/ferroptosis pathways
.
Mol Ther Nucleic Acids
2022
;
28
:
366
86
.
13.
Riemer
J
,
Hoepken
HH
,
Czerwinska
H
,
Robinson
SR
,
Dringen
R
.
Colorimetric ferrozine-based assay for the quantitation of iron in cultured cells
.
Anal Biochem
2004
;
331
:
370
5
.
14.
Shah
R
,
Farmer
LA
,
Zilka
O
,
Van Kessel
ATM
,
Pratt
DA
.
Beyond DPPH: use of fluorescence-enabled inhibited autoxidation to predict oxidative cell death rescue
.
Cell Chem Biol
2019
;
26
:
1594
607
.
15.
Bi
G
,
Bian
Y
,
Liang
J
,
Yin
J
,
Li
R
,
Zhao
M
, et al
.
Pan-cancer characterization of metabolism-related biomarkers identifies potential therapeutic targets
.
J Transl Med
2021
;
19
:
219
.
16.
Folch
J
,
Lees
M
Sloane Stanley
GH
.
A simple method for the isolation and purification of total lipides from animal tissues
.
J Biol Chem
1957
;
226
:
497
509
.
17.
Tsugawa
H
,
Arita
M
,
Kanazawa
M
,
Ogiwara
A
,
Bamba
T
,
Fukusaki
E
.
MRMPROBS: a data assessment and metabolite identification tool for large-scale multiple reaction monitoring based widely targeted metabolomics
.
Anal Chem
2013
;
85
:
5191
9
.
18.
Chen
Z
,
Huang
Y
,
Hu
Z
,
Zhao
M
,
Li
M
,
Bi
G
, et al
.
Landscape and dynamics of single tumor and immune cells in early and advanced-stage lung adenocarcinoma
.
Clin Transl Med
2021
;
11
:
e350
.
19.
Zou
Y
,
Li
H
,
Graham
ET
,
Deik
AA
,
Eaton
JK
,
Wang
W
, et al
.
Cytochrome P450 oxidoreductase contributes to phospholipid peroxidation in ferroptosis
.
Nat Chem Biol
2020
;
16
:
302
9
.
20.
Yan
B
,
Ai
Y
,
Sun
Q
,
Ma
Y
,
Cao
Y
,
Wang
J
, et al
.
Membrane damage during ferroptosis is caused by oxidation of phospholipids catalyzed by the oxidoreductases POR and CYB5R1
.
Mol Cell
2021
;
81
:
355
69
.
21.
Dempster
JM
,
Rossen
J
,
Kazachkova
M
,
Pan
J
,
Kugener
G
,
Root
DE
, et al
.
Extracting biological insights from the Project Achilles Genome-Scale CRISPR Screens in Cancer Cell Lines
.
bioRxiv
2019
:
720243
.
22.
Weber
P
,
Flores
RE
,
Kiefer
MF
,
Schupp
M
.
Retinol saturase: more than the name suggests
.
Trends Pharmacol Sci
2020
;
41
:
418
27
.
23.
Lidén
M
,
Eriksson
U
.
Understanding retinol metabolism: structure and function of retinol dehydrogenases
.
J Biol Chem
2006
;
281
:
13001
4
.
24.
D'Ambrosio
DN
,
Clugston
RD
,
Blaner
WS
.
Vitamin a metabolism: an update
.
Nutrients
2011
;
3
:
63
103
.
25.
von Lintig
J
.
Metabolism of carotenoids and retinoids related to vision
.
J Biol Chem
2012
;
287
:
1627
34
.
26.
Dao
DQ
,
Ngo
TC
,
Thong
NM
,
Nam
PC
.
Is vitamin A an antioxidant or a pro-oxidant?
J Phys Chem B
2017
;
121
:
9348
57
.
27.
Edge
R
,
Truscott
TG
.
Singlet oxygen and free radical reactions of retinoids and carotenoids-a review
.
Antioxidants
2018
;
7
:
5
.
28.
Kelly
M
,
Widjaja-Adhi
MAK
,
Palczewski
G
,
von Lintig
J
.
Transport of vitamin a across blood-tissue barriers is facilitated by STRA6
.
FASEB J
2016
;
30
:
2985
95
.
29.
Luo
J
,
le Cessie
S
,
van Heemst
D
,
Noordam
R
.
Diet-derived circulating antioxidants and risk of coronary heart disease: a mendelian randomization study
.
J Am Coll Cardiol
2021
;
77
:
45
54
.
30.
Maity
A
,
Hyun
S-M
,
Powers
DC
.
Oxidase catalysis via aerobically generated hypervalent iodine intermediates
.
Nat Chem
2018
;
10
:
200
4
.
31.
Zhao
J
,
Fu
Y
,
Liu
C-C
,
Shinohara
M
,
Nielsen
HM
,
Dong
Q
, et al
.
Retinoic acid isomers facilitate apolipoprotein E production and lipidation in astrocytes through the retinoid X receptor/retinoic acid receptor pathway
.
J Biol Chem
2014
;
289
:
11282
92
.
32.
Dixon
SJ
,
Lemberg
KM
,
Lamprecht
MR
,
Skouta
R
,
Zaitsev
EM
,
Gleason
CE
, et al
.
Ferroptosis: an iron-dependent form of nonapoptotic cell death
.
Cell
2012
;
149
:
1060
72
.
33.
Zolfaghari
R
,
Ross
AC
.
Recent advances in molecular cloning of fatty acid desaturase genes and the regulation of their expression by dietary vitamin a and retinoic acid
.
Prostaglandins Leukot Essent Fatty Acids
2003
;
68
:
171
9
.
34.
Liao
X-D
,
Zhou
C-H
,
Zhang
J
,
Shen
J-L
,
Wang
Y-J
,
Jin
Y-C
, et al
.
Effect of all-trans retinoic acid on casein and fatty acid synthesis in MAC-T cells
.
Asian-Australas J Anim Sci
2020
;
33
:
1012
22
.
35.
Lee
H
,
Zandkarimi
F
,
Zhang
Y
,
Meena
JK
,
Kim
J
,
Zhuang
L
, et al
.
Energy-stress-mediated AMPK activation inhibits ferroptosis
.
Nat Cell Biol
2020
;
22
:
225
34
.
36.
Yang
WS
,
Kim
KJ
,
Gaschler
MM
,
Patel
M
,
Shchepinov
MS
,
Stockwell
BR
.
Peroxidation of polyunsaturated fatty acids by lipoxygenases drives ferroptosis
.
Proc Natl Acad Sci U S A
2016
;
113
:
E4966
E75
.
37.
Krupková
M
,
Liška
F
,
Sedová
L
,
Křenová
D
,
Křen
V
,
Seda
O
.
Pharmacogenomic analysis of retinoic-acid induced dyslipidemia in congenic rat model
.
Lipids Health Dis
2014
;
13
:
172
.
38.
Heidenreich
S
,
Witte
N
,
Weber
P
,
Goehring
I
,
Tolkachov
A
,
von Loeffelholz
C
, et al
.
Retinol saturase coordinates liver metabolism by regulating ChREBP activity
.
Nat Commun
2017
;
8
:
384
.
39.
Sun
H-W
,
Chen
J
,
Wu
W-C
,
Yang
Y-Y
,
Xu
Y-T
,
Yu
X-J
, et al
.
Retinoic acid synthesis deficiency fosters the generation of polymorphonuclear myeloid-derived suppressor cells in colorectal cancer
.
Cancer Immunol Res
2021
;
9
:
20
33
.
40.
Zhang
J
,
Deng
B
,
Jiang
X
,
Cai
M
,
Liu
N
,
Zhang
S
, et al
.
All–retinoic acid suppresses neointimal hyperplasia and inhibits vascular smooth muscle cell proliferation and migration activation of AMPK signaling pathway
.
Front Pharmacol
2019
;
10
:
485
.
41.
Igal
RA
.
Stearoyl CoA desaturase-1: new insights into a central regulator of cancer metabolism
.
Biochim Biophys Acta
2016
;
1861
:
1865
80
.
42.
Tesfay
L
,
Paul
BT
,
Konstorum
A
,
Deng
Z
,
Cox
AO
,
Lee
J
, et al
.
Stearoyl-CoA desaturase 1 protects ovarian cancer cells from ferroptotic cell death
.
Cancer Res
2019
;
79
:
5355
66
.
43.
Repa
JJ
,
Liang
G
,
Ou
J
,
Bashmakov
Y
,
Lobaccaro
JM
,
Shimomura
I
, et al
.
Regulation of mouse sterol regulatory element-binding protein-1c gene (SREBP-1c) by oxysterol receptors, LXRalpha and LXRbeta
.
Genes Dev
2000
;
14
:
2819
30
.
44.
Győrffy
B
,
Surowiak
P
,
Budczies
J
,
Lánczky
A
.
Online survival analysis software to assess the prognostic value of biomarkers using transcriptomic data in non-small-cell lung cancer
.
PLoS One
2013
;
8
:
e82241
.
45.
Bi
G
,
Chen
Z
,
Yang
X
,
Liang
J
,
Hu
Z
,
Bian
Y
, et al
.
Identification and validation of tumor environment phenotypes in lung adenocarcinoma by integrative genome-scale analysis
.
Cancer Immunol Immunother
2020
;
69
:
1293
305
.
46.
Jiang
L
,
Kon
N
,
Li
T
,
Wang
S-J
,
Su
T
,
Hibshoosh
H
, et al
.
Ferroptosis as a p53-mediated activity during tumour suppression
.
Nature
2015
;
520
:
57
62
.
47.
Zhang
Y
,
Swanda
RV
,
Nie
L
,
Liu
X
,
Wang
C
,
Lee
H
, et al
.
mTORC1 couples cyst(e)ine availability with GPX4 protein synthesis and ferroptosis regulation
.
Nat Commun
2021
;
12
:
1589
.
48.
Wang
W
,
Green
M
,
Choi
JE
,
Gijón
M
,
Kennedy
PD
,
Johnson
JK
, et al
.
CD8 T cells regulate tumour ferroptosis during cancer immunotherapy
.
Nature
2019
;
569
:
270
4
.
49.
Lei
G
,
Zhang
Y
,
Koppula
P
,
Liu
X
,
Zhang
J
,
Lin
SH
, et al
.
The role of ferroptosis in ionizing radiation-induced cell death and tumor suppression
.
Cell Res
2020
;
30
:
146
62
.
50.
Pang
X-Y
,
Wang
S
,
Jurczak
MJ
,
Shulman
GI
,
Moise
AR
.
Retinol saturase modulates lipid metabolism and the production of reactive oxygen species
.
Arch Biochem Biophys
2017
;
633
:
93
102
.
51.
Hada
M
,
Mondul
AM
,
Weinstein
SJ
,
Albanes
D
.
Serum retinol and risk of overall and site-specific cancer in the ATBC study
.
Am J Epidemiol
2020
;
189
:
532
42
.
52.
Holick
CN
,
Michaud
DS
,
Stolzenberg-Solomon
R
,
Mayne
ST
,
Pietinen
P
,
Taylor
PR
, et al
.
Dietary carotenoids, serum beta-carotene, and retinol and risk of lung cancer in the alpha-tocopherol, beta-carotene cohort study
.
Am J Epidemiol
2002
;
156
:
536
47
.
53.
Yan
H
,
Jin
X
,
Yin
L
,
Zhu
C
,
Feng
G
.
Investigating causal associations of circulating micronutrients concentrations with the risk of lung cancer: a mendelian randomization study
.
Nutrients
2022
;
14
:
4569
.
54.
Abar
L
,
Vieira
AR
,
Aune
D
,
Stevens
C
,
Vingeliene
S
,
Navarro Rosenblatt
DA
, et al
.
Blood concentrations of carotenoids and retinol and lung cancer risk: an update of the WCRF-AICR systematic review of published prospective studies
.
Cancer Med
2016
;
5
:
2069
83
.
55.
Menkes
MS
,
Comstock
GW
,
Vuilleumier
JP
,
Helsing
KJ
,
Rider
AA
,
Brookmeyer
R
.
Serum beta-carotene, vitamins A and E, selenium, and the risk of lung cancer
.
N Engl J Med
1986
;
315
:
1250
4
.
56.
Moise
AR
,
Kuksa
V
,
Imanishi
Y
,
Palczewski
K
.
Identification of all-trans-retinol:all-trans-13,14-dihydroretinol saturase
.
J Biol Chem
2004
;
279
:
50230
42
.
57.
Mishima
E
,
Ito
J
,
Wu
Z
,
Nakamura
T
,
Wahida
A
,
Doll
S
, et al
.
A non-canonical vitamin K cycle is a potent ferroptosis suppressor
.
Nature
2022
;
608
:
778
83
.
58.
Zilka
O
,
Shah
R
,
Li
B
,
Friedmann Angeli
JP
,
Griesser
M
,
Conrad
M
, et al
.
On the mechanism of cytoprotection by ferrostatin-1 and liproxstatin-1 and the role of lipid peroxidation in ferroptotic cell death
.
ACS Cent Sci
2017
;
3
:
232
43
.
59.
Takahashi
N
,
Saito
D
,
Hasegawa
S
,
Yamasaki
M
,
Imai
M
.
Vitamin a in health care: suppression of growth and induction of differentiation in cancer cells by vitamin a and its derivatives and their mechanisms of action
.
Pharmacol Ther
2022
;
230
:
107942
.
60.
Yagoda
N
,
von Rechenberg
M
,
Zaganjor
E
,
Bauer
AJ
,
Yang
WS
,
Fridman
DJ
, et al
.
RAS-RAF-MEK-dependent oxidative cell death involving voltage-dependent anion channels
.
Nature
2007
;
447
:
864
8
.
61.
Green
AS
,
Fascetti
AJ
.
Meeting the vitamin a requirement: the efficacy and importance of -carotene in animal species
.
ScientificWorldJournal
2016
;
2016
:
7393620
.
62.
Kawata
A
,
Murakami
Y
,
Suzuki
S
,
Fujisawa
S
.
Anti-inflammatory activity of β-carotene, lycopene and Tri–butylborane, a scavenger of reactive oxygen species
.
In Vivo
2018
;
32
:
255
64
.
63.
Chen
C
,
Chen
J
,
Wang
Y
,
Liu
Z
,
Wu
Y
.
Ferroptosis drives photoreceptor degeneration in mice with defects in all-trans-retinal clearance
.
J Biol Chem
2021
;
296
:
100187
.
64.
Hoy
AJ
,
Nagarajan
SR
,
Butler
LM
.
Tumour fatty acid metabolism in the context of therapy resistance and obesity
.
Nat Rev Cancer
2021
;
21
:
753
66
.
65.
Ma
X
,
Xiao
L
,
Liu
L
,
Ye
L
,
Su
P
,
Bi
E
, et al
.
CD36-mediated ferroptosis dampens intratumoral CD8 T cell effector function and impairs their antitumor ability
.
Cell Metab
2021
;
33
:
1001
12
.
66.
Luis
G
,
Godfroid
A
,
Nishiumi
S
,
Cimino
J
,
Blacher
S
,
Maquoi
E
, et al
.
Tumor resistance to ferroptosis driven by stearoyl-CoA desaturase-1 (SCD1) in cancer cells and fatty acid binding protein-4 (FABP4) in tumor microenvironment promote tumor recurrence
.
Redox Biol
2021
;
43
:
102006
.
67.
Igal
RA
.
Stearoyl-CoA desaturase-1: a novel key player in the mechanisms of cell proliferation, programmed cell death and transformation to cancer
.
Carcinogenesis
2010
;
31
:
1509
15
.
68.
Jin
Y
,
Teh
SS
,
Lau
HLN
,
Xiao
J
,
Mah
SH
.
Retinoids as anti-cancer agents and their mechanisms of action
.
Am J Cancer Res
2022
;
12
:
938
60
.
69.
Devalaraja
S
,
To
TKJ
,
Folkert
IW
,
Natesan
R
,
Alam
MZ
,
Li
M
, et al
.
Tumor-derived retinoic acid regulates intratumoral monocyte differentiation to promote immune suppression
.
Cell
2020
;
180
:
1098
114
.
70.
Friedmann Angeli
JP
,
Krysko
DV
,
Conrad
M
.
Ferroptosis at the crossroads of cancer-acquired drug resistance and immune evasion
.
Nat Rev Cancer
2019
;
19
:
405
14
.