HMOX1 controls a heme-ferritin switch that protects cells from ferroptosis

Abstract

Modulating the intracellular labile iron pool (LIP) has emerged as a promising strategy to induce ferroptosis in cancer cells, offering a way to overcome resistance to apoptosis-based therapies. One of the main contributors to LIP is heme catabolism mediated by heme oxygenase-1 (HMOX1), which promotes ferroptosis sensitivity by releasing free iron. Beyond its role as an iron donor, heme can influence diverse proteins and signaling pathways that drive tumor progression, but how heme regulates ferroptosis remains poorly understood. Here, we uncover a paradoxical, protective function of heme in the absence of HMOX1 activity. When HMOX1 is inactive, heme becomes stabilized, leading to ferritin up-regulation, suppression of ferroptosis, and rescue of cell death induced by both pharmacological and genetic inhibition of GPX4. Our findings reveal an unrecognized heme-HMOX1-ferritin axis that controls ferroptosis sensitivity. Targeting this pathway may offer a new therapeutic strategy to modulate ferroptosis in cancer.

INTRODUCTION

Ferroptosis is a distinct, iron-dependent form of cell death characterized by the uncontrolled accumulation of lipid hydroperoxides within cellular membranes (1). Unlike apoptosis or necrosis, ferroptosis is triggered when a lipid repair enzyme that detoxifies phospholipid hydroperoxides, glutathione peroxidase 4 (GPX4), is inhibited or inactivated. GPX4 normally detoxifies phospholipid hydroperoxides by reducing them to nontoxic lipid alcohols. In its absence, lipid hydroperoxides accumulate, destabilize membranes, and trigger cell death (2–4).

The central driver of this process is the Fenton reaction, in which ferrous iron Fe(II) reacts with hydrogen peroxide (H₂O₂) to generate highly reactive oxygen radicals, including hydroxyl radicals (HO•). These radicals abstract a bis-allylic hydrogen atom from polyunsaturated fatty acids within phospholipids, generating carbon-centered radicals. The latter react with molecular oxygen (O₂) to form lipid peroxyl radicals (HOO•), which perpetuate chain reactions by extracting hydrogen atoms from nearby acyl chains (5). Under physiological conditions, GPX4 stops this cascade, thereby maintaining membrane integrity. However, when GPX4 is inactivated, uncontrolled iron-driven lipid peroxidation disrupts the cell membrane, leading to cell death (2, 4, 6).

Because cancer cells maintain higher intracellular iron levels than nonmalignant cells, they are believed to be especially susceptible to ferroptosis (7, 8). Cancer cells tightly regulate iron homeostasis to ensure that enough iron is available for essential cellular processes, including enzyme function, heme synthesis, and iron-sulfur cluster assembly. This balance also minimizes the risk of iron-mediated oxidative damage (9–12), and if it is disrupted, the intracellular labile iron pool (LIP) increases, sensitizing cells to ferroptosis (13). In this context, manipulating iron metabolism has emerged as a potential strategy to treat drug-resistant tumors. One major contributor to LIP is the degradation of free heme, an iron-containing protoporphyrin complex. This process is catalyzed by heme oxygenase-1 (HMOX1), releasing carbon monoxide, biliverdin, and ferrous iron. The release of iron through this pathway directly contributes to ferroptosis sensitivity (14–17).

However, a great deal of evidence has shown that the role of heme in cancer biology and ferroptosis extends beyond serving as an iron donor. Heme regulates transcription, translation, drug metabolism, and signal transduction pathways that affect tumor growth and proliferation (18–21). Despite this, the biological relevance of heme in cancer and ferroptosis has often been attributed exclusively to its iron content, overlooking potential regulatory functions mediated by the intact heme molecule.

In the present study, we explored the regulatory role of heme in ferroptosis in cancer cells and uncovered an unexpected cytoprotective function. We found that in the absence of HMOX1, heme suppressed ferroptosis by up-regulating ferritin expression. The effect required the loss of the catalytic activity of HMOX1, rather than its nuclear localization, and occurred independently of the BACH1/NRF2 pathway, a key regulator of antioxidant defense. In addition, we found that stabilization of heme preserved cell viability even in cells without GPX4, a condition that normally produces a lethal phenotype with increased susceptibility to ferroptosis. This indicates that heme stabilization mediated by loss of HMOX1 strongly suppresses ferroptosis. Mechanistically, we observed that stabilization of heme, combined with an iron sequestration response caused by inactivation of iron regulatory protein 2 (IRP2), a key promoter of iron uptake under low-iron conditions, leads to increased expression of the ferritin heavy and light chains. Collectively, these findings uncover a previously unrecognized ferroptosis-suppressive role of heme in the absence of HMOX1, mediated in part by ferritin induction.

RESULTS

Hemin protects HMOX1-knockout cells from ferroptosis

By degrading heme and releasing ferrous iron, HMOX1 contributes to the LIP and thereby increases ferroptosis sensitivity (Fig. 1A). To further investigate the link between heme accumulation and ferroptosis sensitivity, we used CRISPR-Cas9 gene editing to knock out HMOX1 (HMOX1^KO) in A375, A549, and HT1080 cells. We confirmed successful HMOX1 knockout by immunoblotting. HMOX1 expression in these cells was induced by adding hemin, the oxidized and more stable form of heme, which is widely used experimentally to mimic free heme exposure and induce HMOX1 expression. As expected, immunoblotting revealed strong HMOX1 expression in control cells sgNT (single-guide Non-Targeting) but not in HMOX1^KO cells (Fig. 1B and fig. S1A). Uncontrolled heme exposure can cause cell damage as it exerts pro-oxidant effects (22) by catalyzing reactive oxygen species (ROS) formation and promoting lipid peroxidation. Consistent with this, we observed that HMOX1 deficiency strongly sensitized the cells to high doses of hemin (0 to 128 μM) (Fig. 1C and fig. S1B). A nontoxic concentration of hemin (10 μM), combined with GPX4 inhibitor ML210, a ferroptosis inducer, enhanced ferroptosis sensitivity in control cells, which is fully rescued by the antiferroptotic agent liproxstatin-1 (Lip-1; fig. S1C). Notably, however, the same treatment rendered HMOX1^KO cells markedly more resistant to ferroptosis, producing the opposite effect (Fig. 1D and fig. S1, D and E). This unexpected result reveals that the absence of HMOX1 in the presence of hemin paradoxically protects against ferroptosis.

Fig. 1. Hemin protects HMOX1-knockout cells from ferroptosis.

(A) Schematic view of the known role of hemin in ferroptosis. Heme synthesized in mitochondria through sequential enzymatic steps is exported to the cytosol, where it can be degraded by HMOX1 into biliverdin, carbon monoxide (CO), and ferrous iron (Fe²⁺). The released iron contributes to the LIP, promoting lipid peroxidation and ferroptotic cell death. Under ferroptosis-inducing conditions (e.g., GPX4 inhibition by RSL3 or ML210), the accumulation of lipid hydroperoxides (PLOOH) leads to membrane rupture. Created in BioRender. C. Rocha (2026) https://BioRender.com/osoaf9w. (B) Immunoblot analysis of HMOX1 and vinculin in A375 sgNT and HMOX1^KO cells. (C) Dose-dependent cytotoxicity in A375 sgNT and HMOX1^KO cells treated with increased concentration of hemin (0 to 128 μM) for 72 hours. (D) Dose-dependent cytotoxicity in A375 sgNT and HMOX1^KO cells pretreated with hemin for 24 hours (10 μM) followed by ML210 treatment for 72 hours. (E) Immunoblot analysis of HMOX1 and vinculin in A375 HMOX1^KO cells reconstituted with either an empty vector control or a vector encoding HMOX1. (F) Dose-dependent cytotoxicity of 10 μM hemin (24 hours) and ML210 (72 hours) treatment in A375 sgNT and A375 HMOX1^KO cells reconstituted with either an empty vector control or a vector encoding HMOX1. (G) Crystal violet staining of A375 HMOX1^KO cells reconstituted with either an empty vector control or a vector encoding HMOX1 after 10 μM hemin and 0.1 μM ML210 cotreatment. (H) Dose-dependent cytotoxicity in A375 HMOX1^KO cells reconstituted with either an empty vector control or a vector encoding HMOX1, after hemin pretreatment (10 μM) followed by ML210, erastin, or H2O2 (72 hours). Cell viability was assessed using Alamar Blue. Data represent mean ± SD of triplicates from one representative experiment of three performed. IC₅₀, median inhibitory concentration.

To confirm that the protective effect resulted from HMOX1 deficiency, A375 HMOX1^KO cells were transduced with lentiviral vectors encoding either wild-type HMOX1 (to restore HMOX1 expression) or an empty control vector (mock) (Fig. 1E). The complemented cells were subsequently treated with increasing concentrations of hemin. As expected, reintroduction of HMOX1 rendered the A375 HMOX1^KO cells (HMOX1-reconstituted cells) less sensitive to high doses of hemin (fig. S1F). These cells also regained sensitivity to cotreatment with 10 μM hemin and ML210, as assessed by Alamar Blue and crystal violet cell viability staining. In contrast, HMOX1^KO cells transduced with an empty lentiviral vector maintained their high resistance to ferroptosis after 10 μM hemin and ML210 cotreatment (Fig. 1, F and G). Notably, hemin protected HMOX1^KO cells transduced with an empty lentiviral vector against GPX4 inhibitor, but it did not confer protection against erastin, hydrogen peroxide, or other cytotoxic agents tested with various mechanisms of action (Fig. 1H and fig. S1G).

Nontoxic hemin supplementation protects HMOX1/GPX4 double knockout cells from ferroptosis

Since accumulation of lipid peroxides is a hallmark of ferroptosis, we monitored lipid peroxidation using the BODIPY-C11 probe. The BODIPY-C11 probe is a fluorescent dye that shifts from red to green emission upon oxidation, allowing it to serve as a reporter of lipid peroxidation in membranes. We observed that HMOX1 deficiency attenuated both lipid peroxidation and ROS triggered by combined hemin and ML210 treatment (fig. S2, A and B). In addition, because lipid metabolism is a key feature in ferroptosis, we explored the lipidomic profile of cells treated with hemin. This analysis showed no substantial difference in the levels of polyunsaturated species across various phospholipid classes (fig. S2, C to F).

Because both pharmacological inhibition (ML210) and genetic knockout of GPX4 trigger ferroptosis, we tested whether hemin could protect cells in each context, thereby excluding off-target effects or residual GPX4 activity. To do this, we knocked out GPX4 in A375 wild-type cells, generating GPX4^KO, and in A375 HMOX1^KO cells, generating GPX4^KO/HMOX1^KO. We confirmed that these cells lack GPX4 without altering the antiferroptotic protein FSP1 (ferroptosis suppressor protein 1) levels (Fig. 2A). As a positive control, we treated cells with Lip-1 and hemin. We found that hemin treatment increased cell viability by ∼70% in GPX4^KO/HMOX1^KO cells (Fig. 2B). This finding was consistent with the protection of GPX4^KO/HMOX1^KO hemin-treated cells observed by microscopy (fig. S2G). To confirm that the protective effect required the HMOX1 absence, we reconstituted HMOX1 expression in GPX4^KO/HMOX1^KO cells using a doxycycline-inducible vector, which restored HMOX1 after doxycycline treatment and abolished hemin-mediated protection (Fig. 2, C and D).

Fig. 2. Nontoxic hemin supplementation protects HMOX1/GPX4 double knockout cells from ferroptosis.

(A) Immunoblot analysis of GPX4, HMOX1, FSP1, and vinculin in A375 GPX4^KO and GPX4^KO/HMOX1^KO cells. (B) Cell viability of A375 GPX4^KO and GPX4^KO/HMOX1^KO after treatment with hemin (10 μM, 72 hours) or Lip-1 (500 nM, 72 hours). (C) Immunoblot analysis of GPX4, HMOX1, and vinculin in A375 GPX4^KO/HMOX1^KO cells reconstituted with either the pSB empty vector or the pSB vector encoding HMOX1 treated with doxycycline (Dox) (500 ng/ml). (D) Cell viability of A375 GPX4^KO/HMOX1^KO cells reconstituted with either the pSB empty vector or the pSB vector encoding HMOX1 after treatment with hemin (10 μM, 72 hours) or Lip-1 (500 nM, 72 hours). (E) Chemical structures of representative iron-containing heme and metal-substituted or metal-free porphyrins. Structures were redrawn on the basis of entries from PubChem. (F) Immunoblot analysis of HMOX1 in A375 GPX4^KO and GPX4^KO/HMOX1^KO cells treated with hemin, ZnPP, CoPP, or protoporphyrin IX (10 μM, 24 hours). (G) Cell viability of A375 GPX4^KO and GPX4^KO/HMOX1^KO cells after treatment with hemin, ZnPP, CoPP, or protoporphyrin IX (10 μM, 24 hours) or Lip-1 (500 nM, 72 hours). (H) Crystal violet staining of A375 GPX4^KO and GPX4^KO/HMOX1^KO cells after treatment with hemin, ZnPP, CoPP, protoporphyrin IX (10 μM, 24 hours), or Lip-1 (500 nM, 72 hours). (I) Immunoblot analysis of HMOX1 and vinculin expression in A375 sgNT and HMOX1^KO cells. (J) Dose-dependent cytotoxicity of 10 μM ZnPP, CoPP, or protoporphyrin IX in A375 sgNT and HMOX1^KO cells pretreated with the analogs at 24 hours followed by ML210 treatment at 72 hours. Cell viability was measured using Alamar Blue. Dose-response curve data represent mean ± SD of triplicates from one representative experiment of two. Bar graph data represent three independent experiments. Each dot represents a replicate. ****P < 0.0001.

To determine whether this effect was specifically triggered by hemin (protoporphyrin ring + iron) rather than the protoporphyrin ring alone, we treated HMOX1^KO and GPX4^KO/HMOX1^KO cells with protoporphyrin analogs ZnPP, CoPP, and Proto-IX (Fig. 2E). All analogs induced HMOX1 expression but did not affect cell viability after pharmacological inhibition of GPX4 with ML210 (GPX4i) or in GPX4^KO cells (Fig. 2, F to J). Collectively, these findings demonstrate that hemin exerts a context-dependent and paradoxical function in ferroptosis: In the presence of HMOX1, it enhances GPX4 inhibition or GPX4^KO-induced cell death, whereas in its absence, it unexpectedly protects cells by increasing their resistance to ferroptosis.

Catalytic inactivation of HMOX1, not nuclear localization, underlies heme-mediated ferroptosis resistance

Previous studies have shown that HMOX1 is anchored to the endoplasmic reticulum (ER), where it catalyzes heme degradation (23). However, under stress, HMOX1 can be cleaved from the ER and translocated to the nucleus to act in an enzymatic-independent manner (24). Given the dual and context-dependent role of heme in ferroptosis, acting to either increase or decrease cellular sensitivity, we investigated whether this role was dependent on the catalytic activity or cellular localization of HMOX1. To do this, we used a truncated form of the protein shown to localize to the nucleus. Specifically, we reconstituted HMOX1^KO cells with HMOX1 variant truncated at serine-275, referred to as HMOX1^trunc (24, 25). The truncated form lost the typical ER localization and also localized to the nucleus, as confirmed by immunofluorescence (Fig. 3A). These cells exhibited increased resistance to hemin toxicity, confirming successful reintroduction of the protein (Fig. 3B).

Fig. 3. Catalytic inactivation of HMOX1, not nuclear localization, underlies heme-mediated ferroptosis resistance.

(A) Immunofluorescence analysis of A375 HMOX1^KO cells reconstituted with either an empty vector control or a vector encoding HMOX1, HMOX1-trunc, or HMOX1-truncH25A tagged with an alfa-tag. Scale bars, 10 μm. DAPI, 4′,6-diamidino-2-phenylindole. (B) Dose-dependent cytotoxicity of increased concentration of hemin (0 to 128 μM, 72 hours) in A375 HMOX1^KO cells reconstituted with either an empty vector control or a vector encoding either HMOX1 or HMOX1-trunc. (C) Dose-dependent cytotoxicity RSL3 (72 hours) in A375 HMOX1^KO cells reconstituted with either an empty vector control or a vector encoding either HMOX1 or HMOX1-trunc pretreated for 24 hours with 10 μM hemin (bottom graph) or vehicle [dimethyl sulfoxide (DMSO), top graph]. (D) Immunoblot analysis of HMOX1 and vinculin in A375 HMOX1^KO cells reconstituted with either an empty vector control or a vector encoding HMOX1 or HMOX1-H25A. (E) Dose-dependent cytotoxicity of increased concentration of hemin (0 to 128 μM, 72 hours) in A375 HMOX1KO cells reconstituted with either an empty vector control or a vector encoding either HMOX1, HMOX1-H25A, or HMOX1-truncH25A. (F) Dose-dependent cytotoxicity of RSL3 (72 hours) in A375 HMOX1^KO cells reconstituted with either an empty vector control or a vector encoding HMOX1, HMOX1-H25A, or HMOX1-truncH25A pretreated for 24 hours with 10 μM hemin (right graph) or vehicle (DMSO, left graph). Cell viability was assessed using Alamar Blue. Data represent mean ± SD of triplicates from a representative experiment, repeated three times independently. Dose-response curves were fitted by nonlinear regression. IC₅₀ values are shown with 95% confidence intervals derived from the regression model.

By inhibiting GPX4 using the ferroptosis inducer RSL3 (RAS-selective lethal 3), we found that HMOX1^KO cells, which exhibit ferroptosis resistance when treated with hemin, were resensitized to GPX4 inhibition under hemin treatment when reconstituted with HMOX1^trunc (Fig. 3C). This indicated that although HMOX1 can translocate to the nucleus, its concurrent cytosolic presence ensures that nuclear localization does not contribute to the heme-mediated ferroptosis resistance. We therefore tested whether loss of HMOX1 catalytic activity, independent of nuclear localization, was required for resistance. To do this, we reconstituted HMOX1^KO cells with a catalytically inactive mutant, HMOX1^H25A (25), either wild type or truncated (HMOX1^trunc-H25A) (Fig. 3D). These cells were more sensitive to toxic doses of hemin, confirming loss of enzymatic activity (Fig. 3E). In addition, both HMOX1^H25A- and HMOX1^trunc-H25A-reconstituted cells retained their resistance to ferroptosis when treated with nontoxic hemin treatment (Fig. 3F). These findings suggest that the absence of HMOX1 catalytic activity likely leads to heme accumulation, which is required to protect cells from ferroptosis.

Heme stabilization drives ferroptosis resistance independently of the BACH1-NRF2 pathway

Heme promotes proteasomal degradation of the transcriptional repressor BACH1, thereby activating the transcriptional factor NRF2 and inducing a network of antioxidant genes, including HMOX1, that can protect cells from ferroptosis (26). Thus, we hypothesized that in the absence of HMOX1, intracellular heme would be expected to accumulate, thereby promoting more efficient BACH1 degradation and NRF2 activation and leading to ferroptosis resistance (27). As expected, hemin supplementation reduced BACH1 protein levels (fig. S3A). To directly assess the role of BACH1 in this context, we generated BACH1-deficient cell lines (fig. S3B). As shown in Fig. 4A, BACH1 knockout alone was not sufficient to protect cells from ferroptosis. Increased resistance to GPX4 inhibition was observed only in HMOX1^KO cells treated with hemin (Fig. 4A). Next, to test whether NRF2 contributes to ferroptosis protection, we disrupted HMOX1 in NRF2-deficient cells (NRF2^KO) (fig. S3C) and reexpressed wild-type NRF2 across all cell lines. We validated successful generation of these cell lines by elevated expression of NRF2 target genes, such as xCT, FSP1, and NQO1 (fig. S3D). As shown in Fig. 4B, hemin treatment in HMOX1^KO cells is sufficient to confer protection from ferroptosis even in an NRF2^KO background, thereby suggesting that this protective mechanism operates independently of NRF2. Together, these data indicate that heme-mediated ferroptosis resistance in HMOX1^KO cells is BACH1 and NRF2 independent.

Fig. 4. Heme stabilization drives resistance independently of the BACH1-NRF2 pathway.

(A) Dose-dependent cytotoxicity of RSL3 treatment (72 hours) in A375 sgNT, HMOX1^KO, BACH1^KO, and HMOX1^KO/BACH1^KO cells pretreated for 24 hours with 10 μM hemin (right graph) or vehicle (DMSO, left graph). (B) Cell viability of A375 NRF2^KO and NRF2^KO/HMOX1^KO cells reconstituted with NRF2 after hemin pretreatment (10 μM, 24 hours), followed by RSL3 (72 hours). Cell viability was assessed using Alamar Blue. Data represent mean ± SD of triplicates from three independent experiments. **P < 0.01 and ****P < 0.0001. Each dot represents a replicate. (C) Immunoblot analysis of NRF2, TXNRD1, ACSL4, FSP1, GPX4, IRP2, TfR1, FTH1, FTL, HMOX1, and vinculin in A375 HMOX1^KO cells reconstituted with an empty vector control or a vector encoding HMOX1. (D) Dose-dependent cytotoxicity of RSL3 treatment (72 hours) in A375 sgNT, HMOX1^KO, IRP2^KO, and HMOX1^KO/IRP2^KO cells pretreated for 24 hours with 10 μM hemin (bottom graph) or vehicle (DMSO, top graph). (E) Immunoblot analysis of FTH1, FTL, HMOX1, and vinculin in A375 HMOX1^KO and HMOX1^KO/IRP2^KO reconstituted with an empty vector control or a vector encoding IRP2. (F) Dose-dependent cytotoxicity of 10 μM hemin pretreatment (24 hours) and RSL3 treatment (72 hours) in A375 HMOX1^KO and HMOX1^KO/IRP2^KO reconstituted with an empty vector control or a vector encoding IRP2. Cell viability was measured using Alamar Blue. Data represent mean ± SD of triplicates from one representative of three independent experiments. Dose-response curves were fitted by nonlinear regression. IC₅₀ values are shown with 95% confidence intervals derived from the regression model. WT, wild type.

The transcription factor NRF2 induces the expression of ferritin, the major intracellular iron storage complex (28). Ferritin is composed of heavy [ferritin heavy chain 1 (FTH1)] and light [ferritin light chain (FTL)] chain subunits. By sequestering labile iron, ferritin prevents iron-dependent lipid peroxidation and thereby protects cells from ferroptosis (29). Unexpectedly, we observed that FTH1 was strongly induced by hemin even in the absence of NRF2 (fig. S3, D and E). We further examined the expression of a broad range of ferroptosis-related genes after hemin treatment and found that only FTH1 and FTL were differentially induced (Fig. 4C). In contrast to NRF2, which promotes ferritin expression, the iron regulatory protein IRP2 posttranscriptionally represses FTH1 and FTL, thereby limiting ferritin translation and increasing ferroptosis sensitivity (30, 31). Although IRP2 levels were not reduced in HMOX1^KO cells after hemin treatment, its activity appeared diminished, as its targets FTH1 and FTL were induced while TfR1 was suppressed (Fig. 4C). To investigate whether IRP2 contributes to ferritin regulation in the absence of HMOX1, we generated IRP2 knockout and IRP2/HMOX1 double knockout cells. We observed that loss of IRP2 increased expression of FTHL and FTL, and this effect was enhanced by the HMOX1 absence (fig. S3, F and G). Consistently, these cells were more resistant to ferroptosis (Fig. 4D). Notably, hemin treatment still induced ferritin expression and conferred ferroptosis resistance even in IRP2^KO and IRP2^KO/HMOX1^KO cells (Fig. 4D and fig. S3G). Using IRP2-reconstituted cells (IRP2 add-back), we confirmed that loss of IRP2 contributes to ferritin up-regulation in the absence of HMOX1, as IRP2 reconstitution in IRP2-HMOX1 double knockout cells resulted in lower ferritin levels (Fig. 4E). Furthermore, in IRP2-HMOX1 double knockout cells reconstituted with IRP2, hemin does not conferred protection (Fig. 4F), indicating that decreased ferritin levels may attenuate heme-mediated resistance. Notably, lower ferritin levels increased the sensitivity of these cells to 10 μM hemin, accounting for the baseline differences between the cell lines. Collectively, these findings suggest that heme, likely stabilized in the absence of HMOX1, suppresses ferroptosis independently of BACH1 and NRF2, while IRP2 inactivation partially contributes to ferritin up-regulation and resistance to ferroptosis.

Heme stabilization prevents ferroptosis by regulating ferritin levels

Building on our previous observations, we hypothesized that ferritin may contribute to hemin-mediated ferroptosis resistance. RNA sequencing (RNA-seq) analysis comparing hemin-treated HMOX1^KO (mock) and HMOX1^KO-HMOX1 (add-back) revealed a limited set of significantly up-regulated transcripts in HMOX1^KO cells. Among the top hits, SRXN1, an antioxidant enzyme involved in peroxiredoxin repair and redox homeostasis, was up-regulated, whereas GLDC, a mitochondrial enzyme linked to glycine metabolism and glutathione synthesis, was down-regulated. Notably, FTH1 and FTL emerged as the most strongly induced mRNAs (log₂ fold change, false discovery rate < 0.05) (Fig. 5A). Consistently, heatmap analysis across biological replicates confirmed the robust up-regulation of ferritin genes (fig. S4A). To further validate this, we examined the levels of FTH1 and FTL in HMOX1^KO cells transduced with a vector encoding either a control cassette (mock), wild-type HMOX1, or a catalytically inactive mutant, HMOX1^H25A. As shown in Fig. 5B, hemin increased FTH1 and FTL levels in HMOX1^KO and HMOX1^H25A-reconstituted HMOX1KO cells, but not in HMOX1^WT-reconstituted HMOX1^KO cells. Notably, hemin was the only protoporphyrin able to induce FTH1 and FTL in HMOX1^KO cells. Because FTH1 contains the ferroxidase site required for iron oxidation, which limits the LIP that drives lipid peroxidation during ferroptosis, we generated FTH1 knockout (FTH1^KO) and FTH1/HMOX1 double knockout (FTH1^KO/HMOX1^KO) cells (Fig. 5C). Notably, hemin failed to confer ferroptosis protection in the FTH1^KO/HMOX1^KO cells (Fig. 5D). In addition, FTH1^KO cells were sensitive to hemin, with the effect most pronounced in double knockout cells, indicating a strong dependence on ferritin under these conditions (fig. S4, B and C). Next, we reintroduced FTH1 in FTH1^KO and FTH1^KO/HMOX1^KO cells (Fig. 5E), which rescued heme-mediated ferroptosis protection (Fig. 5F). These findings indicate that heme, likely stabilized and accumulated in HMOX1-deficient cells, mediates ferroptosis suppression partly by up-regulating ferritin.

Fig. 5. Heme stabilization prevents ferroptosis by regulating ferritin levels.

(A) Volcano plot showing differential gene expression between A375 HMOX1^KO cells reconstituted with an empty vector control or a vector encoding HMOX1, both treated with hemin. Each dot represents one transcript; the x axis indicates log₂ fold change (FC) and the y axis −log10 adjusted P value (false discovery rate). The samples were classified as follows: down (downregulated, blue), non sig (nonsignificant, grey), or up (upregulated, red). (B) Immunoblot analysis of FTH1, FTL, and vinculin in A375 HMOX1^KO cells reconstituted with an empty vector control or a vector encoding HMOX1 or HMOX1^H25A. Immunoblot analysis of FTH1, FTL, and vinculin in A375 HMOX1^KO cells reconstituted with an empty vector control or a vector encoding HMOX1 after treatment with hemin, ZnPP, CoPP, or protoporphyrin IX (10 μM, 24 hours). (C) Immunoblot analysis of FTH1, HMOX1, and vinculin in A375 sgNT, HMOX1^KO, FTH1^KO, and HMOX1^KO/FTH1^KO cells. (D) Dose-dependent cytotoxicity of RSL3 treatment (72 hours) in A375 sgNT, HMOX1^KO, FTH1^KO, and HMOX1^KO/FTH1^KO cells pretreated for 24 hours with 10 μM hemin (right graph) or vehicle (DMSO, left graph). Cell viability was measured using Alamar Blue. Data represent mean ± SD of triplicates from one representative of three independent experiments. (E) Immunoblot analysis of FTH1, HMOX1, and vinculin in A375 FTH1^KO and HMOX1^KO/FTH1^KO cells reconstituted with either an empty vector control or a vector encoding FTH1. (F) Dose-dependent cytotoxicity of 10 μM hemin pretreatment (12 hours) and RSL3 treatment (72 hours) in A375 FTH1^KO and HMOX1^KO/FTH1^KO cells reconstituted with either an empty vector control or a vector encoding FTH1. Cell viability was measured using Alamar Blue. Data represent mean ± SD of triplicates from one representative of four independent experiments.

DISCUSSION

Beyond its role as an iron donor, the intact heme molecule exerts regulatory effects on signaling and drug metabolism, but these functions in cancer biology and ferroptosis remain unclear (32). Intracellular heme accumulation induces HMOX1 activity, leading to the generation of ferrous iron, carbon monoxide, and biliverdin. Carbon monoxide and biliverdin exert anti-inflammatory and antioxidant effects, respectively. Reduced biliverdin levels were associated with increased sensitivity to ferroptosis (33). Conversely, excess heme expands the pool of chelatable, redox-active iron, thereby promoting ferroptosis (34). Despite these observations, how heme signaling intersects with ferroptosis, particularly in cancer, remains poorly understood. Here, we uncover a previously unknown role of heme in ferroptosis, showing that it functions as an HMOX1-dependent switch that determines whether cells undergo or resist ferroptosis. This dual effect depends on the catalytic activity of HMOX1: In its absence, heme stabilizes and elevates ferritin levels, conferring marked protection against GPX4 inhibition.

Heme is synthesized endogenously through a multistep process involving both mitochondrial and cytosolic compartments (34). While non-erythroid cells normally maintain low intracellular free heme levels (∼300 nM), cancer cells frequently exhibit enhanced heme biosynthesis and uptake to sustain their elevated metabolic and proliferative demands (35–38). Increased heme availability has been linked to intensified mitochondrial respiration, oxidative phosphorylation, and tumorigenic potential, particularly in highly metabolic cancers such as lung and pancreatic cancer (37, 38). Emerging evidence shows that cancer cells present higher activity of heme-containing enzymes and greater regulation of heme import and export channels (37, 39). These observations suggest that tumors actively reprogram heme metabolism to support malignant proliferation and survival.

While increased heme synthesis and import contribute to its intracellular accumulation, its steady-state concentration is also regulated by HMOX1-mediated degradation (38, 40). Previous studies have shown that increased heme degradation via HMOX1 can suppress tumor growth through ferroptosis (15, 41–43). In this study, we confirmed this observation by showing that increased HMOX1 levels induced by hemin promote ferroptosis sensitivity in different cancer cell types. However, we also found that in the absence of HMOX1, heme exerts strong antiferroptotic activity. Specifically, loss of HMOX1 enzymatic activity leads to intracellular heme stabilization, which inhibits ferroptotic cell death. The protective effects associated with hemin accumulation in HMOX1-deficient cells are more apparent under direct GPX4 suppression than following erastin treatment, likely reflecting distinct ferroptotic dependencies imposed by each trigger, consistent with previous studies (1, 44, 45), although the basis for these differences remains under active investigation. While erastin induces broader redox and metabolic perturbations, including mitochondrial ROS generation, direct GPX4 inhibition more specifically disrupts lipid peroxide detoxification (1). These broader redox alterations induced by erastin may mask or bypass the heme-mediated protective mechanism.

Notably, we identified ferritin as a central mediator of heme-mediated ferroptosis resistance in HMOX1-deficient cells. In the canonical setting, hemin-induced heme degradation promotes iron release, leading to FBXL5-mediated degradation of IRP2 and consequent ferritin up-regulation (46–48). While in HMOX1-deficient cells, ferritin is up-regulated despite limited intracellular iron and stable IRP2 protein levels. However, the reversal of ferritin induction upon IRP2 overexpression further indicates that IRP2 is functionally relevant in this context. Together, our findings support a previously unknown mechanism of ferritin regulation by hemin stabilization. Furthermore, FTH1 knockout cells were hypersensitive to ferroptosis upon hemin treatment, whereas reintroduction of FTH1 fully restored resistance. These findings are consistent with a role for ferritin in buffering intracellular iron and protecting against ferroptotic cell death (10, 49–51).

Hemin is therapeutically administered for acute porphyrias, highlighting the biological significance of exogenous heme in disease (52). Our study expands this perspective by showing that exogenously acquired heme, when stabilized by loss of HMOX1 activity, promotes ferroptosis resistance across multiple tumor types, including melanoma, lung carcinoma, and fibrosarcoma. Although our findings were established using engineered models of HMOX1 deficiency, similar conditions may arise in physiopathological contexts in vivo. Tumors frequently exhibit profound metabolic rewiring associated with elevated heme synthesis, increased uptake of extracellular heme, and enhanced erythrocyte turnover or intratumoral hemorrhage, all of which may increase intracellular heme burden (38, 39). Similar conditions may also occur in pathological settings characterized by excessive heme release, including ischemia-reperfusion injury, tissue damage, inflammation, hemorrhagic disorders, and hemolysis (53). For instance, during hemolysis, extracellular heme concentrations are transiently elevated, releasing ∼2 to 5 μM of bioavailable heme into the microenvironment, comparable to doses of hemin used in this study (54). Moreover, under excessive heme conditions, HMOX1 activity may become limiting or insufficient relative to heme load, leading to intracellular heme accumulation. In addition, HMOX1 expression and activity are highly heterogeneous across tumor types and microenvironmental conditions, suggesting that localized states of elevated heme combined with reduced degradative capacity may occur in subsets of tumors (55, 56). Our findings, therefore, raise the possibility that heme-mediated ferroptosis resistance may represent a previously unrecognized adaptive mechanism operating in tumors and other pathological contexts associated with altered heme homeostasis.

MATERIALS AND METHODS

Chemicals and reagents

Hemin (catalog no. H9039), Lip-1 (SML1414), ML210 (SML0521), RSL3 (SML2234), ZnPP (691550-M), CoPP (C1900), and Proto-IX (P8293) were obtained from Sigma-Aldrich. BODIPY-C11 (catalog no. D3861) was purchased from Thermo Fisher Scientific. KI696 was purchased from MedChem Express (catalog no. HY-101140). All drugs were prepared as stock solutions in dimethyl sulfoxide (DMSO) under sterile conditions. Stock solutions were stored at −20°C.

Cell culture conditions

The cell lines A375 (human melanoma), A549 (human lung carcinoma), and HT1080 (human fibrosarcoma) were purchased from American Type Culture Collection. All cell lines were authenticated by Short Tandem Repeats (STR) profiling and verified to be negative for mycoplasma contamination. They were cultured in Dulbecco’s modified Eagle’s medium (Gibco), supplemented with 10% fetal bovine serum (Gibco) and penicillin/streptomycin (100 U/ml; Gibco), at 37°C in a humidified 5% CO₂ atmosphere.

Cell viability assays

For the Alamar Blue assay, 3000 cells were seeded on 96-well plates and pretreated with hemin, ZnPP, CoPP, or Proto-IX for 24 hours (unless stated otherwise). Next, RSL3 or ML210 was added and incubated with the cells for 72 hours. Cell viability was assessed using Alamar Blue. Alamar Blue solution was prepared by dissolving 1 g of resazurin sodium salt in 200 ml of sterile phosphate-buffered saline (PBS) and filtered through a 0.22-μm filter. Stock solutions were stored at 4°C. The working solution was made freshly by adding 100 μl of the stock solution to 10 ml of growth media. After 1 to 2 hours of incubation time, viability was estimated by measuring the fluorescence using a 540/35 excitation filter and a 590/20 emission on a Spark microplate reader (Tecan, Zürich, Switzerland).

For the crystal violet assay, 5000 cells were seeded per well in a 12-well plate. Hemin, ZnPP, CoPP, or Proto-IX were added 6 hours and RSL3 or ML210 24 hours after seeding. The plates were cultured for 10 days under respective conditions. The colonies were stained with 0.01% (w/v) crystal violet-PBS solution.

Generation of knockout cell lines

Single sgRNA guides were chosen using the Vienna Bioactivity CRISPR (VBC) score (https://vbc-score.org/) (57). Guide sequences were cloned into the pLenti-CRISPR-V2 vector by annealing oligonucleotides (Eurofins Genomics) with specific overhangs complementary to BsmBI-digested sites (Addgene #98293 and #98290). Plasmid DNA was confirmed by Sanger sequencing. Cells were transduced with lentiviruses expressing the indicated constructs and subjected to selection for 7 days. Cells were used as pools, or single clones were generated when necessary. Loss of protein expression in cells was validated by Western blot analysis. The following single-guide RNA (sgRNA) primer sets were used: hHMOX1_sgRNA3, TACCTGGGTGACCTGTCTGG; hGPX4_sgRNA1, GAAGCGCTACGGACCCATGG; hBACH1_sgRNA1, GTACCTGCGGAGTTTACACTG; hNRF2_sgRNA1, GGATCTGCCAACTACTCCC; hIRP2_sgRNA2, AAAATCTTGAAGAAGAACAC; hFTH1_sgRNA1, GCACCATGGACAGGTAAACGT.

Lentiviral particles were produced with a third-generation packaging system and X-tremeGENE HP DNA Transfection Reagent (Sigma-Aldrich, 6366546001). Human embryonic kidney (HEK) 293T cells were transfected with the lentiviral transfer plasmids, two packaging plasmids (pMDLg/pRRE, Addgene #12251 and pRSV-Rev, Addgene #12253), and the envelope plasmid (pCMV-VSV-G, Addgene #8454). Viral particles present in the cell culture supernatant were harvested 48 and 72 hours after transfection and used to transduce the target cell line by directly incubating cells with HEK293T supernatants filtered through a 0.44-μm membrane. All experimental procedures for lentivirus production and transduction were performed in a biosafety level 2 laboratory.

Lentivirus production for gene overexpression

Codon-optimized HMOX1 and IRP2 were synthesized by Integrated DNA Technologies (IDT) as gBlocks and subcloned into the p442-IRES vector. The following constructs were generated: p442-IRES-Mock, p442-IRES-HMOX1-Alfatag, p442-IRES-HMOX1(trunc)-Alfatag, p442-IRES-HMOX1(trunc-H25A)-Alfatag, p442-IRES-HMOX1(H25A)-Alfatag, and p442-IRES-IRP2-Flag. HMOX1 overexpression constructs were additionally generated using the transposase vector pSB-ET-iE, as previously described (58). pSB-based expression vectors were cotransfected with the Sleeping Beauty transposase vector pCMV-(CAT)T7- SB100, generating pSB-HMOX1. The cDNA of NFE2L2 was subcloned from pSB-NFE2L2 into p442-IRES, generating the p442-IRES-NFE2L2 plasmid. Similarly, the cDNA of FTH1 was subcloned into p442-IRES and fused to FKBP12^F36V. All plasmids were verified by Sanger sequencing and subsequently used for lentiviral transduction.

GPX4^KO cell experiments

A375 cells deficient in GPX4 or harboring double knockouts were cultured in medium supplemented with 0.5 μM Lip-1 to inhibit ferroptosis. One day after plating, the medium was removed, and cells were washed once with 200 μl of 1× PBS. Following the wash, cells were incubated with medium containing the indicated treatments of interest with or without Lip-1 supplementation. Cell death was determined 72 hours later.

Western blot

Cell lysates were prepared using radioimmunoprecipitation assay buffer supplemented with a protease inhibitor cocktail (Roche). Proteins were separated by SDS–polyacrylamide gel electrophoresis and electrotransferred onto polyvinylidene difluoride membranes. Membranes were subsequently blocked with 5% nonfat milk in tris-buffered saline containing 0.1% Tween 20 (TBS-T) for 1 hour at room temperature (RT). Primary antibodies were diluted in 5% bovine serum albumin (BSA) in TBS-T solution, and membranes were incubated with the primary antibody solution at 4°C overnight. After washing with TBS-T, membranes were incubated at RT for 2 hours with horseradish peroxidase–conjugated secondary antibody (1:3000, anti-rabbit #7074, anti-mouse #7076, or anti-rat) diluted in 5% BSA in TBS-T. Protein bands were visualized using the Immobilon Western Chemiluminescent horseradish peroxidase substrate (Sigma-Aldrich) and detected by enhanced chemiluminescence. Images were acquired using the Amersham ImageQuant 800 detection system (Cytiva). The following antibodies were used against vinculin: 1:1000, sc-73614, Santa Cruz Biotechnology; HMOX1: 1:1000, D60G11, Cell Signaling; GPX4: 1:1000, ab125066, Abcam; FSP1: 1:10, antibody raised against recombinant human FSP1 protein, clone 6D8-11, developed in M. Conrad (Munich, Germany); xCT: 1:10, antibody raised against recombinant human xCT protein, developed in M. Conrad (Munich, Germany); IRP2: 1:1000, D6E6W, Cell Signaling; FTH1: 1:1000, ab65080, Abcam; FTL: 1:1000, ab69090, Abcam; TfR1: 1:1000, ab84036, Abcam; NRF2: 1:1000, ab62352, Abcam; TXNRD1: 1:1000, D1T3D, Cell Signaling; β-actin: 1:5000, A5441, Sigma-Aldrich; BACH1: 1:1000, sc-271211, Santa Cruz; ACSL4: 1:1000, ab205197, Abcam; FluoTag-X2 anti-ALFA-tag (sequence SRLEEELRRRLTE): 1:500 for immunofluorescence, 1:1000 for Western blot, N1502-At488-L, NanoTag Biotechnologies.

Assessment of lipid peroxidation and ROS

A total of 150,000 cells per well were seeded in six-well plates 1 day before the experiment. The next day, cells were treated for 3 hours and then incubated with C11-BODIPY (581/591) (1 μM) or CellROX Green (5 μM) for 30 min at 37°C. Subsequently, cells were trypsinized, resuspended in 500 μl of fresh PBS, and analyzed by flow cytometry using a 488-nm excitation laser to record fluorescence (FACS Canto II, BD Biosciences) in phycoerythrin and fluorescein isothiocyanate channels. The samples were analyzed using FlowJo software (BD Biosciences). Cells were first gated on the basis of forward and side scatter (FSC-A versus SSC-A) to exclude debris. Viable cells were selected on the basis of scatter properties. The C11-BODIPY and CellROX Green fluorescence intensity (FITC channel, 488-nm excitation/510- to 550-nm emission) was quantified within the viable population. Unstained cells were used to define background fluorescence. C11-BODIPY was gating for oxidized versus reduced populations, which was established using untreated controls.

Immunofluorescence

Cells expressing the ALFA-tagged constructs were seeded on glass coverslips in six-well plates. The next day, cells were washed with PBS and fixed in 4% paraformaldehyde (Sigma-Aldrich) for 30 min, then quenched in PBS containing 0.1 M glycine for 10 min, and permeabilized and blocked for 15 min in PBS containing 10% normal goat serum and 0.1% Triton X-100. Coverslips were then incubated for 1 hour with FluoTag-X2 anti-ALFA conjugated to Atto488 (diluted 1:500 in 3% normal goat serum and 0.1% Triton X-100). DNA was counterstained by incubating coverslips with 4′,6-diamidino-2-phenylindole (Thermo Fischer Scientific, 1:10,000 in PBS) for 10 min, followed by two PBS washes for 10 min each. Coverslips were subsequently mounted onto microscope slides using DABCO (Sigma-Aldrich). Confocal fluorescence images were obtained using a Leica SP8 microscope with a HCX PL APO CS 63× 1.4 numerical aperture oil objective.

Lipidomic analysis

Water-soluble metabolites were quantified by liquid chromatography–mass spectrometry. Briefly, 1 × 10⁶ cells were harvested, washed with PBS, and snap-frozen in liquid nitrogen. Metabolites were extracted with 500 μl of ice-cold methanol/water (80:20, v/v) supplemented with 0.01 μM lamivudine and 1 μM each of D₂-glucose, D₄-succinate, D₅-glycine, and ¹⁵N-glutamate (Sigma-Aldrich). After centrifugation, supernatants were evaporated and reconstituted in 150 μl of 5 mM NH₄OAc in acetonitrile/water (50:50, v/v), and 20-μl aliquots were injected for analysis. Metabolites were separated on an XBridge Premier BEH Amide UPLC column (2.5 μm, 100 mm by 2.1 mm; Waters) using a DIONEX Ultimate 3000 UHPLC (Ultra-High-Performance Liquid Chromatography) system. Solvent A was 5 mM NH₄OAc in CH₃CN/H₂O (40:60, v/v), and solvent B was 5 mM NH₄OAc in CH₃CN/H₂O (95:5, v/v). Separation was performed at a flow rate of 200 μl/min and at 45°C using a linear gradient from 100% B to 10% B over 23 min, followed by 16 min in 10% B and a 2-min return to 100% B. Columns were equilibrated with 100% B for 7 min before each injection. High-resolution mass spectrometry was performed on a Q Exactive mass spectrometer equipped with a HESI (Heated Electrospray Ionization) probe, operated in alternating positive and negative full mass spectrometer mode (69 to 1000 mass/charge ratio, 70,000 resolution; Thermo Fisher Scientific) using standard electrospray ionization parameters. Data were analyzed using TraceFinder v5.1, integrating peaks corresponding to the calculated monoisotopic masses of the metabolites [±H⁺, ±3 mMU (milli-Mass Unit)]. All analyses were conducted in three independent biological replicates. Processed lipidomic datasets are publicly available in Mendeley Data (10.17632/43dz5xhgxt.2).

RNA-seq analysis

Total RNA was extracted using the Monarch Total RNA Miniprep Kit according to the manufacturer’s instructions. RNA-seq data were generated from A375 HMOX1 knockout (mock) and A375 HMOX1 add-back cells, treated or untreated with hemin, with three biological replicates per condition. Sequencing and primary data processing, including normalization of gene expression matrices, were performed by the Core Unit of Systems Medicine at the University of Würzburg. Differential expression analysis was performed using the Limma package implemented in R/Bioconductor, which allows statistical analysis of prenormalized expression data using moderated linear models with empirical Bayes correction. Genes with an adjusted P value < 0.05 were considered significantly differentially expressed unless otherwise stated. Processed RNA-seq datasets are publicly available in Mendeley Data (10.17632/43dz5xhgxt.2).

Quantification and statistical analysis

Details of the statistical analyses for each experiment are provided in the corresponding figure legends, with P values indicated in the figures. Unless stated otherwise, data are shown as mean ± SD from three independent biological replicates conducted on separate days. Statistical significance was assessed using one-way or two-way analysis of variance (ANOVA) followed by Dunnett’s or Tukey’s multiple comparisons test (ns indicates not statistically significant; *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001), performed in GraphPad Prism v10 (GraphPad Software). The outcomes of these analyses are reported in the figures. No statistical methods were used to predetermine sample size, and no data were excluded from the analyses. Experiments were not randomized, and investigators were not blinded to allocation during experimentation or outcome assessment.

Supplementary Materials

This PDF file includes:

Figs. S1 to S4