Endocytosis is essential for cysteine deprivation-induced ferroptosis

Summary

Ferroptosis is a form of cell death caused by iron-dependent phospholipid peroxidation and subsequent membrane rupture. Autophagic degradation of the iron-storage protein ferritin promotes ferroptosis by increasing cytosolic bioactive iron, presumably explaining how lysosomal inhibitors suppress ferroptosis. Surprisingly, we found that lysosomal inhibitors suppress cysteine starvation-induced (CDI) ferroptosis even in autophagy-defective cells, and subsequently discovered that clathrin-mediated endocytosis (CME) of transferrin is essential for CDI ferroptosis. Blockage of lysosomal proteolytic activity failed to inhibit ferroptosis, whereas disrupting endosomal acidification and eliminating the endocytic protein AP2M1 both impeded ferroptosis. Conversely, repleting cellular iron with ferric ammonium citrate, but not transferrin, restored CDI ferroptosis in endocytosis-deficient cells. Unexpectedly, abolishing endosomal acidification, CME, and the associated increase of cellular labile iron could not prevent ferroptosis triggered by direct inhibition of the ferroptosis-suppressing enzyme glutathione peroxidase-4 (GPX4). Together, this study reveals the essential role of endocytosis specifically for CDI ferroptosis.

Graphical Abstract

Introduction

Ferroptosis is a cell death process driven by iron-dependent phospholipid (PL) peroxidation^1-3. Because PL peroxidation is a natural consequence of cellular metabolism, there are multiple surveillance mechanisms in the cell to attenuate PL peroxidation and thus prevent unwanted ferroptosis. Among these mechanisms, the most prevalent one is mediated by glutathione (GSH) peroxidase-4 (GPX4), an enzyme reducing PL peroxides into nonharmful PL alcohols. As such, direct inhibition of GPX4 or depletion of cysteine, an essential building block of GSH, can trigger ferroptosis^4,5. Iron is crucial for ferroptosis induction, as it can function both as a cofactor for enzymes generating PL peroxides and as a catalyst for non-enzymatic Fenton reaction to propagate PL peroxidation^1-3.

Previous research has established that autophagy, a stress-responsive survival pathway mediated by intracellular membrane reorganization and lysosomal degradation, can promote ferroptosis by degrading the iron-storage protein ferritin (aka ferritinophagy⁶), thereby increasing cellular labile iron availability^7,8. This process is mediated by the cargo receptor NCOA4, which directs ferritin to the lysosome for degradation^7,8. Consistently, lysosomal inhibitors, including V-ATPase inhibitors and chloroquine, which all function by disrupting lysosomal acidification⁹, can inhibit ferroptosis^7,8,10.

Surprisingly, we observed that even in autophagy-defective cells, these lysosomal inhibitors still possessed potent ferroptosis-inhibitory activity. Subsequently, through a series of pharmacological and genetic experiments targeting key components of clathrin-mediated endocytosis (CME), we revealed that these perturbations ablated cysteine deprivation-induced (CDI) ferroptosis through disrupting transferrin endocytosis and associated iron import. Importantly, CME is essential for CDI ferroptosis but not for GPX4 inhibition-induced ferroptosis, indicating that the basal level of cellular free iron is sufficient to mediate ferroptosis induced by GPX4 inhibition. Collectively, this study demonstrates that endocytosis, a cellular process essential for cell viability, also plays an indispensable role in promoting ferroptotic cell death triggered by cysteine deprivation.

Results

Lysosomal inhibitors can prevent ferroptosis independently of autophagy.

Lysosomal inhibitors, such as bafilomycin A1 (BafA1), concanamycin A (ConA), and chloroquine (CQ), have been shown to inhibit ferroptosis, and this effect has been attributed to their effect on the inhibition of autophagy^7,8,10. To further investigate this effect and the underlying mechanism, we tested these compounds on multiple cell culture models, including human fibrosarcoma HT1080 cells, human kidney cancer 786-O cells, and mouse embryonic fibroblasts (MEFs). As expected, treatment with 100 nM BafA1, 100 nM ConA, or 50 μM CQ prevented lipid peroxidation (Figures S1A-B) and protected the cells from ferroptosis induced by cystine deprivation, which was achieved by either directly removing cystine from the culture medium or by using imidazole ketone erastin (IKE), a chemical inhibitor of the system x_c⁻ cystine/glutamate antiporter¹¹ (Figures 1A-D, S1C-D). Surprisingly, although these lysosomal inhibitors almost completely ablated CDI ferroptosis, they failed to inhibit ferroptosis triggered by GPX4 knockout (KO) or RSL3, a pharmacological inhibitor of GPX4 (Figures 1E, S1E-G). This result indicates a context-dependent role of these lysosomal inhibitors, as well as lysosomal-autophagy activity, in ferroptosis regulation.

Figure 1. Lysosomal inhibitors can inhibit cysteine deprivation-induced ferroptosis in an autophagy-independent manner.

(A-D) Lysosomal inhibitors fully prevented cysteine deprivation-induced (CDI) ferroptosis in both HT1080 and MEF cells. (A-D) Time course of cell death in HT1080 (A-B) and MEF cells (C-D) treated with 100 nM bafilomycin A1 (BafA1), 100 nM concanamycin A (ConA), 50 μM chloroquine (CQ), or 2 μM liproxstatin-1 (Lip-1). Ferroptosis was induced by cystine starvation (A, C) or 500 nM IKE (B, D).

(E) Cell death analysis of GPX4 KO HT1080 cells treated with 50 μM deferoxamine (DFO), 2 μM Lip-1, 100 nM BafA1, 100 nM ConA, in the absence or presence of 2 μM Trolox as indicated.

(F-K) Lysosomal inhibitors blocked CDI ferroptosis in autophagy-deficient MEF cells. (F-G) Western blot analysis confirming knockout of ATG5 and ULK1 (F) or FIP200 (G) in MEF cells. (H, J) Time course of cell death in ATG5/ULK1/2-knockout (TKO) MEF cells (H) or FIP200 knockout (KO) MEF cells (J) treated with 100 nM BafA1, 100 nM ConA, 50 μM CQ, or 2 μM Lip-1. Ferroptosis was induced by cystine starvation. (I, K) Microscopy images showing cell death in TKO (I) or FIP200-KO (K) MEF cells treated as indicated for 18 hours. Upper panels: phase-contrast images; lower panels: SYTOX Green staining for dead cells (scale bar = 100 μm).

Data are presented as mean ± SD, n = 3 (A-D, H, J) or n = 5 (E) biologically independent samples. See also Figure S1.

If indeed these compounds inhibit CDI ferroptosis through autophagy inhibition, then one can make two predictions: (1) genetic ablation of autophagy should result in an equally potent, if not more, ferroptosis inhibition in comparison with these compounds; and (2) when autophagy is completely ablated genetically, these compounds should no longer exert inhibitory effect on ferroptosis. Surprisingly, all these lysosomal inhibitors exhibited a much stronger inhibitory effect on ferroptosis than that caused by genetic blockage of autophagy: while ATG5/ULK1/ULK2-triple knockout (TKO) MEFs and FIP200-KO MEFs showed delayed CDI ferroptosis in comparison with wildtype MEFs (Figures 1F-K), the compounds caused a much more pronounced inhibition than these genetic interventions (compare Figures 1A-D with 1H-K). More importantly, the lysosomal inhibitors further blocked ferroptosis in TKO and FIP200-KO MEFs towards a near complete inhibition (Figures 1H-K). Therefore, BafA1, ConA, and CQ can inhibit ferroptosis via both autophagy-dependent (ferritin degradation) and autophagy-independent mechanisms. Similarly, ferroptosis can still be triggered by RSL3 in TKO MEFs with BafA1, ConA, or CQ treatment (Figure S1H), further confirming the context-dependent role of both autophagy and these lysosome inhibitors in ferroptosis regulation.

Chaperone-mediated autophagy (CMA) was reported to be involved in ferroptosis regulation¹². To determine whether CMA contributes to the strong ferroptosis-inhibitory effect of the lysosome inhibitors, we generated HT1080 cells and MEFs lacking lysosome-associated membrane protein type 2a (LAMP2), an essential mediator of CMA, and found that BafA1, ConA, and CQ were still effective in inhibiting CDI ferroptosis (Figures S1I-K). Therefore, lysosomal inhibitors can block CDI ferroptosis through mechanisms independent of CMA.

Lysosomal proteolytic activity is not required for CDI Ferroptosis.

To determine whether inhibition of lysosomal proteolytic activity by these compounds leads to such an autophagy-independent blockage of CDI ferroptosis (e.g., by blocking the lysosomal degradation of a certain ferroptosis-suppressing protein), we examined the effect of a protease inhibitor cocktail. BafA1 but not the protease inhibitors prevented ferroptosis (Figures 2A, S2A-B), although BafA1 and the protease inhibitors similarly reduced lysosomal proteolytic activity as assessed by DQ-BSA assay¹³ (Figures 2B-C, S2C-D). This result indicates that lysosomal protease activity is not required for the autophagy-independent inhibition of CDI ferroptosis by the lysosomal inhibitors.

Figure 2. Endocytosis but not lysosomal protease activity is required for cysteine-deprivation-induced ferroptosis.

(A-C) BafA1 but not the protease inhibitors prevented ferroptosis. (A) Time course of cell death in HT1080 cells treated with protease inhibitor cocktail or 100 nM BafA1. Ferroptosis was induced by cystine starvation. (B) Fluorescence images of HT1080 cells treated with protease inhibitor cocktail or 100 nM BafA1. 100 μg/ml DQ Red BSA was added to the cells and incubated for 6 hours before imaging (scale bar = 100 μm). (C) Quantification of DQ-BSA fluorescence intensity of cells treated as described in (B). Statistical analysis was performed using 1-way ANOVA.

(D-K) Inhibition of endocytosis blocked ferroptosis induced by cystine starvation or IKE but not by RSL3. (D) Time course of cell death in HT1080 cells treated with 10 μM DN34-2, 100 nM BafA1, or 2 μM Lip-1. Ferroptosis was induced by cystine starvation. (E) Western blot analysis confirming the degradation of the dTAG-AP2M1 fusion protein within 2 hours after treatment with 100 nM dTAG^v-1. (F-H) Time course of cell death in dTAG-AP2M1 HT1080 cells treated with 100 nM dTAG^v-1, 100 nM BafA1, or 2 μM Lip-1. Ferroptosis was induced by cystine starvation (F), 500 nM IKE (G) or 100 nM RSL3 (H). (I) Microscopic images showing cell death in dTAG-AP2M1 HT1080 cells treated as indicated for 12 hours. Upper panels: phase-contrast images; lower panels: SYTOX Green staining for dead cells (scale bar = 100 μm). (J-K) Viability assay of dTAG-AP2M1 HT1080 cells. Cells were pretreated with or without 100 nM dTAG^v-1 for 6 hours, followed by indicated concentrations of IKE (J) or RSL3 (K) for an additional 24 hours.

Data are presented as mean ± SD, n = 3 (A, D, F-H, J-K) or n = 4 (C) biologically independent samples. See also Figure S2.

Endocytosis is required for cysteine deprivation-induced ferroptosis.

All lysosomal inhibitors used in this study function through disrupting lysosomal acidification, with BafA1 and ConA acting as V-ATPase inhibitors and CQ as a direct lysosomal pH neutralizer⁹. Importantly, in addition to lysosomes, various other endocytic vesicles also need to be acidified during endocytosis in a V-ATPase-dependent manner^14,15. As such, the function of these endosomes can also be ablated by lysosomal inhibitors. Therefore, an alternative mechanism could be that these compounds cause autophagy-independent ferroptotic blockage through inhibiting endocytosis.

To test this possibility, we employed chemical and genetic approaches targeting core components of the endocytic pathway. The core components of CME include the clathrin complex, the clathrin adaptor protein AP2 complex, and the GTPase dynamin^16-18. Pharmacologically, we used small molecule Dynole 34-2 (DN34-2), which blocks the GTPase activity of dynamin 1/2¹⁹. Treatment with 10 μM DN34-2 effectively blocked CDI ferroptosis, showing an inhibitory effect comparable to that of BafA1 in HT1080, MEF, and 786-O cells (Figures 2D, S2E-F). Genetically, we sought to target another key component of CME, AP2M1, which is crucial for clathrin adaptor complex assembly^16,20. As CME is essential for the cell, permanent elimination or knockdown of AP2M1 is not feasible. Therefore, we engineered HT1080 cells with a degradation tag (dTAG)^21,22 knocked in at the endogenous locus of the AP2M1 gene, enabling rapid degradation of the tagged endogenous AP2M1 protein upon the treatment of the cell with dTAG^v-1, a small molecule that can bring the tagged protein to the ubiquitin ligase VHL for ubiquitin-mediated proteasomal degradation²³. As shown in Figure 2E and S2G, treatment of the engineered cells with 100 nM of dTAG^v-1 led to complete degradation of AP2M1 within 2 hours without affecting the expression levels of known ferroptosis regulators, including GPX4, SLC7A11, and ACSL4. Importantly, dTAG^v-1 treatment nearly completely blocked ferroptosis induced by cystine starvation or IKE, but not RSL3, confirming that endocytosis is specifically required for CDI ferroptosis (Figures 2F-K).

Endocytosis mediates ferroptosis independently of proliferation regulation

Lysosomal acidity and iron homeostasis are known to influence cell proliferation²⁴, although the role of cell cycle arrest in ferroptosis regulation remains controversial^25,26. To determine whether lysosomal acidity and endocytosis regulate ferroptosis through their effects on cell proliferation, we first examined changes in cell proliferation under these conditions. Indeed, inhibition of endocytosis both genetically and chemically, as well as treatment with the lysosomal inhibitor BafA1, modestly reduced cell proliferation (Figures S2H-I). However, this reduction does not account for the observed suppression of ferroptosis. First, treatment with Abemaciclib, a cyclin-dependent kinase 4/6 (CDK4/6) inhibitor, similarly impeded cell proliferation but did not inhibit ferroptosis (Figures S2I-J), indicating that reduced proliferation alone is insufficient to confer resistance to ferroptosis. Further, dTAG^v1-induced degradation of AP2M1 did not result in a noticeable change in cell proliferation within 24 hours (Figure S2I), but within the same period, it almost completely inhibited CDI ferroptosis (Figures 2F-G).

Deficient endocytosis leads to cellular iron reduction.

The iron carrier protein transferrin (TF) has been reported to be essential for CDI ferroptosis²⁷. Through transferrin receptor (TFRC) and clathrin-mediated endocytosis, TF delivers iron into the cell and is recycled back to the plasma membrane after releasing the bound iron in acidified endosomes^28,29. Based on this, we hypothesize that blocking endocytosis will intervene TF-mediated iron import, leading to a reduction of cellular labile iron pool and consequently, ferroptosis inhibition.

To determine if lysosome inhibitors block ferroptosis by inducing iron deficiency, we treated HT1080 cells with BafA1 and stained them with FerroOrange, a fluorescent probe for ferrous iron. Consistent with a previous study²⁴, BafA1 significantly reduced intracellular ferrous iron levels (Figures 3A, S2K-L). Given that BafA1 prevents CDI ferroptosis through blocking endocytosis, we next investigated whether inhibiting endocytosis through other approaches could also decrease cellular ferrous iron. Indeed, inhibiting endocytosis by DN34-2 or by dTAG-mediated AP2M1 degradation also led to a significant reduction in intracellular ferrous iron levels (Figures 3B-C, S2K-L). Additionally, we observed a significant upregulation of TFRC and iron regulatory protein 2 (IRP2) in HT1080 cells treated with BafA1 or ConA, as well as in cells subjected to dTAG^v-1-mediated AP2M1 degradation, which could be suppressed by supplementation with ferric ammonium citrate (FAC), a form of non–transferrin-bound iron (NTBI) that bypasses transferrin and the endocytic system for cellular uptake³⁰ (Figures S2M-N). These changes are consistent with a cellular response to reduced intracellular iron levels.

Figure 3. Endocytosis inhibition and transferrin deprivation lead to iron deficiency and ferroptosis blockage.

(A-C) Lysosomal inhibitors and endocytosis inhibition both reduced intracellular ferrous iron levels. HT1080 cells were treated with DMSO,100 nM BafA1 (A) or 10 μM DN34-2 (B) for 24 hours. dTAG-AP2M1 HT1080 cells were treated with DMSO or 100 nM dTAG^v-1 (C) for 24 hours. After washing three times with PBS, the cells were incubated with FerroOrange for 30 minutes and then imaged by fluorescent microscopy. Right panels: Quantification of FerroOrange fluorescence intensity of cells treated as described above.

(D-F) TFRC knockdown inhibits ferroptosis induced by IKE but not RSL3. (D) Western blot analysis confirming knockdown of TFRC in HT1080 cells. (E) Cell death analysis of shNT and shTFRC HT1080 cells with (right) or without (left) 312.5 nM IKE for 24 hours. (F) Cell death analysis of shNT and shTFRC HT1080 cells with (right) or without (left) 62.5 nM RSL3 for 24 hours.

(G-H) Serum starvation inhibited iron accumulation and CDI ferroptosis, which was restored by transferrin supplementation. (G) Time course of cell death in HT1080 cells incubated in serum-free DMEM supplemented with 10% dialyzed FBS or 5 μg/mL recombinant human transferrin, in the absence or presence of 2 μM Lip-1 as indicated. Ferroptosis was induced by cystine starvation. (H) HT1080 cells were incubated in serum-free DMEM supplemented with 10% dialyzed FBS or 5 μg/mL recombinant human transferrin for 24 hours. After washing three times with PBS, the cells were incubated with FerroOrange and Hoechst 33342 for 30 minutes. Images were acquired with a Leica SP8 confocal laser scanning microscope (scale bar = 100 μm). Lower panel: Quantification of FerroOrange fluorescence intensity of cells treated as described above. Data are presented as mean ± SD, n = 3 (E-G) or n=4 biologically independent samples (A-C, H). Statistical analysis was performed using two-tailed t-test (A-C) or 1-way ANOVA (H). See also Figure S2.

Given that TFRC-mediated endocytosis of transferrin is a major pathway for cellular iron uptake, we knocked down TFRC in HT1080 cells (Figure 3D) and examined its effect on ferroptosis induced by different triggers. TFRC knockdown significantly inhibited ferroptosis induced by IKE (Figure 3E) or cystine starvation (Figure S2O), which is consistent with previous reports^27,31. Strikingly, TFRC knockdown had minimal impact on RSL3-induced ferroptosis (Figure 3F), which is in line with the differential effect of BafA1 or endocytosis on ferroptosis triggered by cysteine deprivation versus that by GPX4 inhibition.

Endocytosis mediates ferroptosis through transferrin internalization.

We next examined directly whether TF is responsible for the role of endocytosis in CDI ferroptosis. Consistent with the previous report²⁷, CDI ferroptosis was ablated by serum starvation, and the addition of 5 μg/mL recombinant human TF restored ferroptosis under this condition (Figures 3G, S2P-Q). Further, under serum-free condition, HT1080 and MEF cells exhibited a significant reduction in FerroOrange staining compared to the cells grown in media with 10% dialyzed FBS (diFBS), and supplementing with 5 μg/mL TF restored intracellular ferrous iron levels in the absence of serum (Figures 3H, S2R).

Subsequently, we assessed the role of endocytosis in TF internalization and the associated increase of cellular iron. Inhibition of endocytosis, either through DN34-2 treatment or AP2M1 degradation, reduced both cellular ferrous iron and transferrin uptake (Figures 4A-B, S3A; Video S1). Colocalization of TFRC and TF was arrested at the cell membrane by dTAG^v1 treatment, indicating impaired internalization (Figure S3B; Video S2). However, under the same endocytosis-inhibited conditions, exogenous iron supplementation with FAC restored both cellular ferrous iron levels and ferroptosis triggered by cystine starvation or IKE (Figures 4A-D, S3C). In contrast, the availability of TF in the cultural medium failed to restore iron accumulation in endocytosis-deficient cells (Figures 4A-B), and TF remained trapped at the cell membrane but was not internalized upon AP2M1 elimination (Figures 4B, S3A-B; Video S1-2). Importantly, FAC supplement bypassed the need for lysosomal/endosomal acidification in ferroptosis execution (Figures 4C-D, S3D-F) and restored CDI ferroptosis in TFRC-knockdown cells (Figure S2O). Notably, FAC supplement rescued ferroptosis but not cell proliferation in endocytosis-deficient cells (Figures S2K-L).

Figure 4. Endocytosis promotes cysteine deprivation-induced ferroptosis by mediating transferrin uptake.

(A-B) Inhibition of endocytosis reduced cellular ferrous iron and transferrin uptake. HT1080 cells were seeded on 24-well glass-bottom plates and grown with or without 10 μM DN34-2 or 75 μM ferric ammonium citrate (FAC) for 12 hours (A). dTAG-AP2M1 HT1080 cells were treated with DMSO or 100 nM dTAG^v-1 for 6 hours (B). Cells were washed twice with PBS, then incubated with FerroFarRed for 1 hour and transferrin-Alexa 488 (TF488) for 30 minutes. Images were acquired with a Leica SP8 confocal laser scanning microscope (scale bar = 100 μm).

(C-D) FAC supplementation restored ferroptosis triggered by cystine starvation or IKE in endocytosis-deficient cells. dTAG-AP2M1 HT1080 cells were seeded on a 24-well plate and grown for 12 hours. Cells were treated with DMSO (control) or 100 nM dTAG^v-1 for 6 hours. Then, cells were washed twice in PBS and treated with 75 μM FAC, with or without 2 μM Lip-1. Ferroptosis was induced by cystine starvation (C) or 500 nM IKE (D).

(E-F) Cystine starvation promoted transferrin internalization and increased intracellular ferrous iron accumulation. (E) HT1080 cells were treated as indicated for 5 hours. After washing three times with PBS, the cells were incubated with FerroOrange and Hoechst 33342 for 30 minutes. Images were acquired with a Leica SP8 confocal laser scanning microscope (scale bar = 100 μm). Statistical analysis was performed using 1-way ANOVA (E). (F) Representative still images from time-lapse imaging of HT1080 cells treated as indicated. Live cells were stained with Hoechst 33342 for 30 minutes and changed into media with 5% FBS and 2.5 μg/mL TF488 (scale bar = 20 μm). Lower panel: Quantification of intracellular TF488 fluorescence intensity per cell.

Data are presented as mean ± SD, n = 3 (C-D, F) or n = 6 (E) biologically independent samples. See also Figure S3 and Videos S1, S2, S3, and S4.

We next investigated whether cystine deprivation influences transferrin endocytosis and the subsequent iron accumulation. HT1080 cells under cystine starvation or IKE treatment showed increased FerroOrange staining and enhanced transferrin uptake rate, in comparison with cells in the presence of cyst(e)ine (Figures 4E-F, S3G; Video S3-4). However, treatment with DN34-2 disrupted this enhanced endocytic process, resulting in the accumulation of TFRC and transferrin at the cell surface (Figure S3G, third row; Video S4), further supporting the essential role of endocytic trafficking in iron uptake during CDI ferroptosis. As predicted, such cysteine deprivation-induced iron accumulation was not observed under serum-free conditions but was restored by adding TF to serum-free medium (Figure 4E).

Discussion

In this study, we provide compelling evidence demonstrating that the endocytic pathway is crucial for the execution of ferroptosis triggered by cysteine deprivation. Considering the essential role of endocytosis at both cellular and organismal levels, this finding is conceptually important as it links endocytosis with the execution of a specific cell death process. Mechanistically, we show that endocytosis enables CDI ferroptosis by facilitating transferrin trafficking and resulting in a consequent accumulation of intracellular labile iron.

This study clarifies a misconception in the fields of ferroptosis and autophagy. Previously, any cellular effect of lysosomal pH neutralizers (chloroquine and its derivatives) and V-ATPase inhibitors was commonly interpreted as the consequence of the inhibition of autophagy or lysosomal activity. For example, their capability to block ferroptosis was attributed to the inhibition of autophagy, or more specifically, ferritinophagy. However, this study showed that these compounds can ablate ferroptosis through preventing transferrin endocytosis, in an autophagy-independent manner. Furthermore, as various other endosomal vesicles in addition to lysosomes need to be acidified to function^14,15, the effect of these compounds can also be independent of lysosomes.

As iron is essential for the execution of ferroptosis, it is important to determine whether an increase in cellular labile iron is always associated with, and even indispensable for, ferroptosis. As demonstrated unambiguously in this study, although an increase of labile iron is crucial for ferroptosis triggered by cysteine deprivation, such an increase is not required for ferroptosis induced by direct inhibition of GPX4.

The differential requirement of iron quantity for ferroptosis induced by its distinctive triggers also raises a series of intriguing questions. For example, why is a higher level of iron specifically required for CDI ferroptosis? Does such a higher level of iron exert other ferroptotic function in addition to directly catalyzing and amplifying lipid peroxidation, and if so, how is this additional function only essential for CDI ferroptosis, but not for that triggered by GPX4 inhibition? Is iron required to be localized in a specific subcellular compartment (e.g., endosomes or lysosomes) for ferroptosis induction? Lastly, could the mechanism underlying CDI ferroptosis be fundamentally different from that triggered by GPX4 inhibition? This latter possibility challenges the dogmatic mechanism explaining how cysteine deprivation induces ferroptosis, i.e., cysteine deprivation leads to the depletion of glutathione, hence causing GPX4 inactivation and eventual ferroptosis^1-3. Accumulating observations as reported previously and described in this study suggest fundamental difference between cysteine deprivation and GPDX4 inhibition as ferroptosis triggers. These previous observations include prominent roles of mitochondria, autophagy, transferrin, and glutamine in CDI ferroptosis instead of that triggered by GPX4 inhibition^27,32, as well as a seemingly more significant role of ACSL4 in GPX4 inhibition-induced ferroptosis than CDI ferroptosis³³. Taken together, this study reveals the specific role and regulation of iron homeostasis in ferroptosis in a context-dependent manner.

Limitations of the Study

This study reveals that endocytosis is essential for cysteine deprivation-induced ferroptosis. As transferrin enters the cell and releases iron through the endosomal-lysosomal route, an obvious mechanistic question that is not addressed here is whether ferroptosis is triggered by iron within these specific organelles. Indeed, some recent studies, mainly through utilizing pharmacological tools, suggest that lysosomal iron may play a particularly important role in ferroptosis^34,35. However, it remains technically challenging to quantify iron in different subcellular compartments, and even more so to determine the function of iron within a specific subcellular compartment via definitive and causative experiments. The fact that cellular iron homeostasis is highly dynamic, i.e., change of iron contents within one compartment may alter that at other subcellular sites, further complicates the problem. Another limitation of the study is that although we showed here that cysteine deprivation enhanced the endocytosis of transferrin, the underlying mechanism has not been defined and warrants future investigation.

STAR Methods

Method Details

Cell culture

HT1080 fibroblast (ATCC) and Mouse Embryonic Fibroblast (MEFs) were cultured in high-glucose Dulbecco’s Modified Eagle’s Medium (DMEM) (MSKCC Media Preparation Core Facility) supplemented with 10% fetal bovine serum (Gibco), 2 mM L-glutamine and 100 U/mL Penicillin/Streptomycin (Gibco). 786-O cells were cultured in Roswell Park Memorial Institute medium (RPMI) (MSKCC Media Preparation Core Facility), supplemented with the same components. All cell lines were maintained at 37°C in a humidified incubator with 5% CO₂. Mycoplasma contamination was routinely checked using PCR screening. Cysteine-free media were supplemented with 10% dialyzed fetal bovine serum (Gemini BioProducts, Cat# 100-108-500), 2 mM L-glutamine, and 100 U/mL penicillin-streptomycin. For cystine starvation treatment, cells were first seeded in cystine-replete medium and allowed to adhere overnight. The next day, cells were gently washed with pre-warmed PBS and then switched to cystine-deficient medium (MSKCC Media Preparation Core Facility) to initiate starvation.

Generation of CRISPR knockout cells

GPX4-KO HT1080 cells, FIP200-KO MEFs, and ULK1/ULK2-double knockout (DKO) MEFs were previously established by our lab^32,36,37. ULK1/ULK2-DKO-Cas9 MEFs were generated by infecting ULK1/ULK2-DKO MEFs with the lentiCas9-Blast lentivirus (Addgene, Cat# 52962), followed by selection with 20 μg/ml blasticidin. Guide RNAs targeting mouse ATG5 were cloned into the gRNA_cloning vector (Addgene, Cat# 41824) at the AflII restriction site using the NEB Gibson Assembly Kit (New England Biolabs, Cat# E5510S). To create ATG5/ULK1/ULK2 triple knockout (TKO) MEFs, the ULK1/ULK2-DKO-Cas9 MEFs were transiently transfected with the 3 sgATG5 plasmids described above and selected with 400 μg/ml Neomycin. HT1080-LAMP2 knockout (KO) cells were generated by transient transfection with eSpCas9(1.1)-T2A-mCherry vectors (Addgene, Cat# 71814) encoding a guide RNA (gRNA) targeting human LAMP2 (Sigma, Cat# HSPD0000023589). Following transfection, mCherry-positive cells were sorted by fluorescence-activated cell sorting (FACS) to enrich successfully transfected cells.

Generation of locus-specific dTAG-AP2M1 knock-in (KI) HT1080 cells

To generate human dTAG-AP2M1 KI cells, HT1080 cells were transfected with vectors (eSpCas9(1.1)-T2A-mCHerry) encoding a single gRNA (1 μg) targeting the start codon of AP2M1, along with a donor plasmid carrying the eGFP-P2A-dTAG knock-in cassette flanked by AP2M1 homology arms (3 μg). Three days post-transfection, single cells expressing both eGFP and mCherry were sorted via FACS. Individual cells were plated in 96-well plates for further screening and validation.

Generation of shRNA expression cells

shTFRC cells were generated by infecting HT1080 cells with the indicated shTFRC lentiviruses (#1: TRCN0000057659; #2: TRCN0000057660; #3: TRCN0000057661), followed by selection with 2 μg/mL puromycin. shNT control cells were generated by infecting HT1080 cells with non-targeting shRNA lentivirus (Sigma, #SHC016).

Generation of stable expression cells

TFRC cDNA (GenScript, NM_003234.4) was subcloned into the pQCXIP retroviral vector with an N-terminal HaloTag. The HaloTag sequence was PCR-amplified from the PIEZO1-HaloTag construct (Addgene, Cat# 207834). To generate HaloTag-TFRC overexpression cells, the indicated cell lines were infected with retroviruses carrying the construct and selected with 2 μg/mL puromycin.

Measurement of lipid ROS

For lipid ROS imaging, cells were seeded in 24-well glass-bottom plates (Cellvis, Cat# P24-1.5H-N) at an appropriate cell density and cultured for 24 hours. After treatment, cells were washed twice with HBSS, incubated in HBSS containing 2 μM BODIPY 581/591 C11 (Invitrogen) for 20 minutes, and stained with Hoechst 33342 (0.1 μg/mL). Imaging was conducted using a Leica SP8 confocal microscope (MSKCC Molecular Cytology Core Facility) with a 63x oil immersion objective, capturing eight representative fields per condition under consistent settings.

Cell death assay

Cells were seeded in 24-well plates at an appropriate density and cultured under normal conditions for 24 hours. Cells were then stained with Hoechst 33342 (0.1 μg/mL) for 30 minutes, followed by a wash with 1x PBS. Afterward, cells were subjected to the indicated experimental treatments. Dead cells were labeled using 5 nM Sytox Green. Cells were incubated in a BioSpa system (BioTek Instruments) connected to a Cytation 5 imager and were captured every 2 hours using an automated protocol. Four images were taken per well using a 4x objective and stitched together using Gen 5 software. Cell death percentage was calculated by dividing the number of Sytox Green-positive cells by the total number of Hoechst 33342-positive cells.

Cell viability assay

Cells were seeded at the appropriate density and treated according to the specified conditions. Following treatment, cell viability was measured using the CellTiter-Glo Luminescent Cell Viability Assay (Promega, Cat# G7573), according to the manufacturer’s instructions. Relative viability was determined by normalizing ATP levels of treated samples to untreated controls.

Measurement of cellular labile ferrous iron (Fe²⁺)

Cells were seeded in 24-well glass-bottom plates at an appropriate cell density and cultured under normal conditions for 24 hours. After the indicated treatments, cells were washed three times with serum-free medium and incubated with 1 μM FerroOrange (Dojindo, Cat# F374-12) in DMEM for 30 minutes at 37°C and 5% CO₂. Cells were also stained with Hoechst 33342 (0.1 μg/mL). Imaging was performed using an ECLIPSE Ti2 or Leica SP8 confocal laser scanning microscope (MSKCC Molecular Cytology Core Facility) with a 63x oil immersion objective. Eight representative fields per condition were captured under identical settings.

Transferrin Uptake Assay

Cells were seeded in 24-well glass-bottom plates and cultured for 24 hours. After the indicated treatments, cells were washed three times with serum-free medium and incubated with 5 μg/mL transferrin-Alexa 488 (TF488) for 30 minutes. Cells were then washed twice with serum-free medium without phenol red and stained with FerroFarRed (Millipore Sigma, Cat# SCT037) according to the manufacturer’s instructions. Images were acquired using a Leica SP8 confocal microscope, capturing eight representative fields per condition under identical acquisition settings. For time-lapse imaging, cells were stained with Hoechst 33342 (0.1 μg/mL) and/or Janelia Fluor 646 HaloTag Ligand (200 nM) for 20 minutes, washed twice with HBSS, and subjected to treatments for 1 hour. Cells were then incubated in HBSS with 5% FBS containing 2.5 μg/mL TF488. Fluorescence and DIC images were captured every 5 minutes, and the images were analyzed using NIS Elements (Nikon) and ImageJ.

Analysis of lysosomal degradation capacity

Lysosomal degradation was assessed using DQ Red BSA (Fischer Scientific, Cat# D12051). Cells were seeded in 24-well glass-bottom plates and cultured for 24 hours. After treatment with DMSO (12 hours), protease inhibitor cocktail (12 hours), or 100 nM BafA1 (4 hours), the medium was supplemented with 0.3 mg/mL DQ Red BSA for 6 hours. Lysosomal degradation capacity was determined by the fluorescence signal generated from DQ Red BSA degradation. Imaging was performed using an ECLIPSE Ti2 microscope, capturing eight representative fields under identical settings for each condition.

Western Blot

Cells were lysed in RIPA buffer, and lysates were boiled and resolved via SDS-PAGE before being transferred to nitrocellulose membranes. Membranes were blocked with 5% skim milk for 1 hour at room temperature, followed by overnight incubation at 4°C with primary antibodies. After three washes with TBST, membranes were incubated with HRP-conjugated secondary antibodies (goat anti-mouse or donkey anti-rabbit, Invitrogen) for 1 hour at room temperature. After three washes with TBST, the membranes were developed using Clarity Western ECL Substrate (Bio-Rad) and imaged on an Amersham Imager 600 (GE Healthcare Life Sciences).

Quantification and Statistical Analysis

All statistical analyses were conducted using GraphPad Prism 9.0. Data are presented as mean ± SD from three or more independent experiments. For experiments involving two groups, Student’s t-test was performed; for experiments with more than two groups, one-way ANOVA (analysis of variance) followed by Tukey’s multiple comparisons test was performed. Significance is indicated as ns (not significant), *p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001.

Supplementary Material

1

Video S1. Time-lapse video of transferrin uptake in dTAG-AP2M1 HT1080 cells treated with (right) or without (left) 100 nM dTAG^v-1, related to Figures 4 and S3A.

2

Video S2. Time-lapse video of transferrin uptake in dTAG-AP2M1 HaloTag-TFRC HT1080 cells treated with (right) or without (left) 100 nM dTAG^v-1, related to Figures 4 and S3B.

3

Video S3. Time-lapse video showing transferrin uptake in HT1080 cells under three conditions: normal (control, left), ferroptosis induced by cystine starvation (middle), and ferroptosis induced by 500 nM IKE treatment (right), related to Figure 4F.

4

Video S4. Time-lapse video showing transferrin uptake in HaloTag-TFRC HT1080 cells under three conditions: control (left), ferroptosis induced by 500 nM IKE (middle), and co-treatment with 500 nM IKE and 10 μM DN34-2 (right), related to Figures 4 and S3G.

Key Resources Table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Anti-LAMP-2A	Thermo Fisher Scientific	Cat# 512200
Anti-AP2M1	Abcam	Cat# ab75995
Anti-ATG3	Cell Signaling Technology	Cat# 3415S
Anti-ATG7	Cell Signaling Technology	Cat# 2631S
Anti-IRP1	Cell Signaling Technology	Cat# 20272T
Anti-IRP2	Cell Signaling Technology	Cat# 37135T
Anti-TFRC	Abcam	Cat# ab214039
Anti-FTH1	Cell Signaling Technology	Cat# 3998S
Anti-GPX4	Abcam	Cat# ab116703
Anti-SLC7A11	Cell Signaling Technology	Cat# 12691
Anti-ACSL4	Invitrogen	Cat# PA5-27137
Anti-Vinculin	Sigma-Aldrich	Cat# V4505
Anti-GAPDH	Cell Signaling Technology	Cat# 2118
Anti-α-tubulin	Calbiochem	Cat# CP06
Bacterial and virus strains
DH5alpha	Thermo Fisher Scientific	Cat # 18265017
Chemicals, peptides, and recombinant proteins
Imidazole ketone erastin	MedChem Express	Cat# HY-114481
1S-3R RSL3	Cayman Chemicals	Cat# 19288-10
Liproxstatin-1	Cayman Chemicals	Cat# 17730
Bafilomycin A1	Cayman Chemicals	Cat# 11038-1
Concanamycin A	MedChem Express	Cat# HY-N1724
Chloroquine	Selleck	Cat# S6999
Deferoxamine Mesylate	Sigma-Aldrich	Cat# D9533
Trolox	Santa Cruz Biotechnology	Cat# sc-200810
Abemaciclib	MedChem Express	Cat# HY-16297A
Dynole 34-2 (DN34-2)	Thermo Fisher Scientific	Cat# 42- 221010
Ferric ammonium citrate (FAC)	Sigma-Aldrich	Cat# 5879
Human recombinant transferrin	Sigma-Aldrich	Cat# T3705
Transferrin From Human Serum, Alexa
Fluor™ 488 Conjugate (TF488)	Thermo Fisher Scientific	Cat# T13342
FerroOrange	Dojindo	Cat# F374-12
BioTracker Far Red Fe2+ dye	Millipore Sigma	Cat# SCT037
DQ Red BSA	Thermo Fisher Scientific	Cat# D12051
dTAGv1	Thermo Fisher Scientific	Cat# 69-145
Lipofectamine 3000	Thermo Fisher Scientific	Cat# L3000001
IQ™ SYBR® Green Supermix	Bio-Rad	Cat # 1708882
T4 DNA Ligase	NEB	Cat # M0202M
Phusion High Fidelity DNA Polymerase	Fisher Scientific	Cat # F530L
Janelia Fluor® 646 HaloTag® Ligand	Promega	Cat# GA1120
Experimental models: Cell lines
Human: HT1080	ATCC	N/A
Human: 786-O	ATCC	N/A
Mouse: MEF	ATCC	N/A
Mouse: MEF-ATG5/ULK1/ULK2TKO	This paper	N/A
Mouse: MEF-FIP200KO	Ganley, Ian G et al.36	N/A
Human: HT1080- LAMP2KO	This paper
Human: HT1080- GPX4KO	Gao et al.32	N/A
Recombinant DNA
Plasmid: gag/pol	Addgene	Cat# 14887
Plasmid: pCMV-VSV-G	Addgene	Cat# 8454
Plasmid: lentiCas9-Blast	Addgene	Cat# 52962
Plasmid: eSpCas9(1.1)	Addgene	Cat# 71814
Plasmid: TFRC cDNA ORF Clone	GenScript	OHu06795
Plasmid: shTFRC-1 (GCTGGTCAGTTCGTGATTAAA)	MSKCC gene editing core	TRCN0000057659
Plasmid: shTFRC-2 (CGTGAATTTAAACTCAGCAAA)	MSKCC gene editing core	TRCN0000057660
Plasmid: shTFRC-3 (GCCAGCTTTACTGGAGAACTT)	MSKCC gene editing core	TRCN0000057661
Plasmid: PIEZO1-HaloTag	Addgene	Cat# 207834
Plasmid: sgAP2M1 (GCATATACGATACAAGGCTG)	This paper	N/A
Plasmid: sgLAMP2 (TAGCAGTGCAGTTCGGACC)	This paper	N/A
Plasmid: sgATG5#1 (GATGAAAGGCCGCTCCGTCG)	This paper	N/A
Plasmid: sgATG5#2 (TATCCCCTTTAGAATATATC)	This paper	N/A
Plasmid: sgATG5#3 (TTCCATGAGTTTCCGGTTGA)	This paper	N/A
Plasmid: PQCXIP-HaloTag-TFRC	This paper	N/A
Software and algorithms
Illustrator 2022	Adobe	N/A
GraphPad Prism 9	GraphPad	https://www.graphpad.com/
ImageJ	NIH, USA	https://imagej.nih.gov/ij/
Deposited data
Western Blots raw data	This paper	Mendeley Data: https://data.mendeley.com/datasets/5s2np3c4y7/1

• Lysosomal inhibitors can suppress CDI ferroptosis independently of autophagy

• Endocytosis is required for CDI ferroptosis but not that induced by GPX4 depletion

• Endocytic defects lower cellular iron levels and block CDI ferroptosis

• Transferrin internalization by endocytosis drives CDI ferroptosis

Liu et al. demonstrate that lysosomal inhibitors can block CDI ferroptosis independently of autophagy and that clathrin-mediated endocytosis of transferrin is essential for cysteine deprivation-induced (CDI) ferroptosis by sustaining cellular iron. This study also reveals fundamental differences between CDI ferroptosis and that induced by GPX4 inhibition.