Radiotherapy and Immunotherapy Promote Tumoral Lipid Oxidation and Ferroptosis via Synergistic Repression of SLC7A11

Immune checkpoint blockade is a powerful oncologic treatment modality for a wide variety of human malignancies (1). Randomized clinical trials are assessing how best to interdigitate this treatment modality with traditional therapies including radiotherapy (2). Thus far, preclinical data have suggested that immune checkpoint blockade can synergize with radiotherapy through agonism of innate immune sensing pathways activated as a consequence of DNA damage (3–5).

It is well established that unrepaired DNA double-strand breaks dictate cellular survival following radiotherapy (6). These lesions promote mitotic catastrophe, an unregulated form of cell death (7). More recent examinations have implicated a number of other forms of regulated cell death, including apoptosis, necroptosis, and autophagy in the response to radiotherapy (8, 9). Radiotherapy stochastically induces oxidative damage in all cellular compartments, including the lipid membrane (10, 11). Toxic lipid peroxidation accumulation has only recently been implicated in directly causing a regulated form of cell death, termed ferroptosis (12). The sources of cell death–inducing reactive oxygen species (ROS) that predispose cells to undergo ferroptosis have not been defined. It has been suggested that radiation-induced lipid oxidation results in apoptosis (13), but the relationship between radiotherapy and ferroptosis remains undefined. Further, radiotherapy efficacy in vivo is dependent on the presence of CD8⁺ T cells (14, 15). We have recently discovered that CD8⁺ T cells modulate tumoral ferroptosis through IFNγ (16). It is unknown if ferroptosis underlies the interplay between the adaptive immune system and radiotherapy.

To address these questions, we explored the involvement of lipid peroxidation and ferroptosis in response to radiotherapy. We implicate ferroptosis as a previously unknown mechanism of tumor cell death following radiotherapy in vitro and in vivo. In addition, we have found that radiotherapy can sensitize tumor cells to ferroptosis agonists in vitro and in vivo, highlighting a novel strategy to radiosensitize tumors. Moreover, we have implicated ferroptosis as a direct link through which CD8⁺ T cells and radiotherapy cooperate to improve tumor control. Mechanistically, IFNγ derived from immunotherapy-activated CD8⁺ T cells and ATM activated following radiotherapy synergistically suppress SLC7A11, a unit of the glutamate–cystine antiporter xc⁻, resulting in enhanced tumor lipid oxidation and ferroptosis. This work is the first to nominate radiotherapy as a therapeutic ferroptosis inducer and suggest ferroptosis agonists as a novel class of sensitizers for radiotherapy and immunotherapy. Finally, our studies highlight lipid peroxidation as a previously unappreciated point of mechanistic interplay between radiotherapy and immunotherapy in cancer.

Ferroptosis is a recently discovered form of cell death that differs from apoptosis and results from iron-dependent lipid peroxide accumulation (12, 17). It is unknown whether lipid oxidation and ferroptosis are induced by radiotherapy. To test this in vivo, we inoculated ID8 ovarian cancer cells into the peritoneal cavity of C57BL/6 mice and treated the mice with a single dose of 8 Gray (Gy) ionizing radiation, liproxstatin-1, a ferroptosis antagonist, or both agents concurrently. ID8 tumor–bearing mice tolerated radiotherapy without changes in mouse weight (Supplementary Fig. S1A). Radiotherapy, but not liproxstatin-1, reduced tumor growth (Fig. 1A). Unexpectedly, liproxstatin-1 administration concurrent and adjuvant to radiotherapy diminished radiotherapy efficacy (Fig. 1A).

To extend this finding, we inoculated B16F10 melanoma cells into C57BL/6 mice and allowed the tumors to establish. Subsequently, B16F10-bearing mice were treated with a single fraction of 8 Gy. Again, radiotherapy reduced tumor growth (Supplementary Fig. S1B). Dual treatment with liproxstatin-1 and radiotherapy partially and significantly abrogated radiotherapy efficacy. Ferroptosis is accompanied by increased lipid peroxidation, which can be examined using the lipophilic redox-sensitive dye C11BODIPY (12). C11BODIPY quantification showed that radiotherapy increased lipid peroxidation in vivo, and concurrent liproxstatin-1 treatment abrogated radiotherapy-induced lipid peroxidation in B16F10 tumors (Fig. 1B; Supplementary Fig. S1C).

To orthogonally confirm that tumoral ferroptosis occurs following radiotherapy, we generated ferroptosis-resistant subclones of B16F10 through serial exposure to RSL-3, a ferroptosis agonist. We showed that RSL-3–resistant B16F10 cells were indeed resistant to RSL-3 and erastin, two ferroptosis agonists (Supplementary Fig. S1D). We then inoculated parental or RSL-3–resistant B16F10 cells into C57BL/6 mice and treated established tumors with radiotherapy. We observed that resistance to ferroptosis significantly reduced the efficiency of radiotherapy (Fig. 1C). Further, RSL-3–resistant tumors failed to manifest a robust increase in lipid peroxidation following radiotherapy as compared with parental tumors (Fig. 1D). We additionally generated ferroptosis-resistant ID8 subclones through serial exposure to erastin. We also showed that erastin-resistant ID8 cells were resistant to RSL-3 and erastin (Supplementary Fig. S1E). We inoculated parental and erastin-resistant ID8 cells into C57BL/6 mice and treated established ID8 tumors with radiotherapy. Again, we observed reduced radiotherapy efficacy in erastin-resistant tumors (Fig. 1E). Together, these data suggest that radiotherapy induces tumor ferroptosis and also that ferroptosis is critical for radiotherapy efficacy.

To understand whether radiotherapy directly induces tumor ferroptosis, we performed clonogenic survival assays following radiotherapy in HT1080 human fibrosarcoma cells in vitro. We observed that ferroptosis inhibitors liproxstatin-1 and ferrostatin-1, the lipophilic antioxidant trolox, and the iron chelator deferoxamine (DFO) increased clonogenic cell survival following radiotherapy (Fig. 1F). To determine whether radiotherapy directly affected lipid peroxidation, we conducted C11BODIPY staining in HT1080 cells in vitro. We observed that radiotherapy resulted in an increase in lipid ROS in HT1080 cells (Fig. 1G). We next examined clonogenic cell survival and lipid peroxidation in B16F10 and ID8 cells. Again, liproxstatin-1, as well as trolox and DFO, enhanced clonogenic survival in B16F10 (Supplementary Fig. S1F) and ID8 (Supplementary Fig. S1G) cells following radiotherapy. Radiotherapy induced an increased lipid peroxidation in B16F10 cells (Supplementary Fig. S1H) and ID8 cells (Supplementary Fig. S1I) in a dose-dependent manner. These results suggest that radiotherapy directly induces tumor ferroptosis.

Given that radiotherapy can induce tumor ferroptosis (Fig. 1A–G; Supplementary Fig. S1A–S1I), we wondered whether ferroptosis agonists could enhance radiotherapy efficacy and be exploited as novel radiosensitizers. To explore this possibility, we pretreated HT1080, B16F10, and ID8 cells with low doses of ferroptosis inducers (FIN) including sulfasalazine, RSL-3, erastin, and atorvastatin, and then irradiated these cells. We found that ferroptosis inducers reduced clonogenic cell survival as compared with radiotherapy alone in HT1080 cells in vitro (Fig. 1H). GPX4 utilizes glutathione to reduce oxidized lipid species and to limit ferroptosis. Glutathione levels are regulated by intracellular cystine concentrations. We depleted cystine and cysteine using cyst(e)inase, a recombinant human enzyme (18). Interestingly, cyst(e)inase strongly sensitized HT1080 cells to radiotherapy in vitro (Fig. 1I).

To support our finding in HT1080 cells, we treated B16F10 cells with sulfasalazine, RSL-3, and cyst(e)inase and observed decreased cell survival following radiotherapy as compared with radiotherapy alone (Supplementary Fig. S1J and S1K). Cell death quantification using propidium iodide (PI) staining confirmed that sulfasalazine and cyst(e)inase augmented radiation-induced cell death (Fig. 1J). Pretreatment with RSL-3, sulfasalazine, or cyst(e)inase also increased ID8 cell sensitivity to radiotherapy in vitro (Supplementary Fig. S1L and S1M). To examine the relevance of this finding in vivo, we established B16F10 tumors in C57BL/6 mice and treated mice with cyst(e)inase, radiotherapy, or the combination. Although cyst(e)inase had minimal efficacy as monotherapy, it enhanced radiotherapy efficiency (Fig. 1K). Moreover, increased tumor control was accompanied by enhanced lipid oxidation in the combination therapy group (Fig. 1L). Neither single treatment nor the combination treatment caused significant murine weight loss (Supplementary Fig. S1N).

To assess whether FINs can improve radiotherapy efficacy in vivo, we established B16F10 tumors in mice and treated tumors with radiotherapy, sulfasalazine, or both (dual treatment). We found that dual treatment enhanced tumor control (Supplementary Fig. S1O). Again, this improved tumoral control was accompanied by increased lipid oxidation in tumors treated with both agents (Supplementary Fig. S1P). These data suggest that ferroptosis can be targeted in vivo to enhance radiotherapy efficacy.

It has recently been shown that fatty-acid saturation modulates ferroptosis sensitivity. Specifically, ACSL4 promotes ferroptosis, whereas ACSL3 limits ferroptosis (19, 20). We generated ACSL4 knockout and ACSL3 knockout B16F10 cells (Supplementary Fig. S1Q and S1R). We observed that loss of the ferroptosis effector gene ACSL4 diminished radiotherapy efficacy in vivo (Fig. 1M). In contrast, deletion of the ferroptosis suppressor ACSL3 augmented radiotherapy efficacy in vivo (Fig. 1N).

To demonstrate that diminished cell survival following radiation was due to increased lipid peroxidation and ferroptosis, we quantified lipid peroxidation changes with C11BODIPY following FIN addition. We observed that in B16F10 cells, short exposures to low doses of sulfasalazine, RSL-3, and cyst(e)inase minimally altered lipid ROS. Combinations of FINs with radiation synergistically increased lipid ROS (Fig. 1O). FINs similarly augmented lipid ROS levels in concert with radiation in HT1080 cells (Supplementary Fig. S1S).

To generalize these findings to different radiation doses and fractionation schedules, we treated B16F10 tumors with 5 fractions of 3 Gy as well as a single fraction of 10 or 20 Gy. Quantification of lipid peroxidation changes with C11BODIPY showed that all radiotherapy doses induced increases in C11BODIPY, whereas higher single fractions of radiotherapy induced more lipid peroxidation (Fig. 1P). To confirm that ferroptosis resistance conferred radiotherapy resistance at ablative doses of radiotherapy, we treated wild-type (WT) and RSL-3–resistant B16F10 tumors with a single fraction of 20 Gy (21). We observed that even at higher doses of radiotherapy, ferroptosis-resistant tumors remained resistant to radiotherapy in vivo (Supplementary Fig. S1T).

Collectively, these data suggest that radiation induces tumor ferroptosis, and this ferroptosis can be pharmacologically augmented by ferroptosis agonists. Thus, targeting ferroptosis may be a novel therapeutic approach for radiation sensitization.

CD8⁺ T cells are required for efficacy of ablative doses of radiotherapy in vivo (21, 22). The cellular and molecular bases for the interaction between T cells and radiotherapy are not fully defined. We first examined the involvement of CD8⁺ T cells in radiation therapy efficacy in a B16F10 tumor–bearing mouse model at moderate clinically utilized radiation doses. We subcutaneously inoculated B16F10 cells into C57BL/6 mice and established tumors. Then, mice received a moderate dose of radiation (8 Gy) alone or with anti-CD8 mAb administration prior to radiotherapy. As expected, radiotherapy reduced tumor volume (Fig. 2A). However, CD8⁺ T-cell depletion with anti-CD8 mAb treatment largely abolished radiotherapy-induced tumor reduction (Fig. 2A). This radiation dose induced tumor-cell ferroptosis (Supplementary Fig. S1B and S1H). To understand if CD8⁺ T cells promoted ferroptosis in concert with radiotherapy, we treated B16F10 cells with naïve or activated T-cell supernatant. We found that exposure to a low dose of T-cell supernatant had minimal effect on B16F10 cell survival. Interestingly, T-cell supernatant synergized with radiotherapy to limit B16F10 clonogenic cell survival (Fig. 2B). Furthermore, T-cell supernatant and radiotherapy each independently enhanced lipid peroxidation in B16F10 cells, and their combination further augmented lipid peroxidation (Fig. 2C). Blockade of IFNγ signaling decreased T-cell supernatant-induced tumor STAT1 phosphorylation (Supplementary Fig. S2A) and diminished tumor lipid ROS production following combination treatment with T-cell supernatant and radiotherapy (Fig. 2C). These results suggest that CD8⁺ T cells may promote tumor ferroptosis and induce radiosensitization via IFNγ.

To assess a direct link between IFNγ signaling and radiotherapy efficacy, we treated ID8 cells with IFNγ and radiotherapy and examined clonogenic cell survival. We observed that IFNγ and radiotherapy synergistically reduced clonogenic survival (Fig. 2D). We also observed diminished clonogenic cell survival following IFNγ and radiation treatment in B16F10 cells (Supplementary Fig. S2B). IFNγ and radiotherapy synergistically increased lipid ROS levels in ID8 cells (Fig. 2E), B16F10 cells (Supplementary Fig. S2C), and HT1080 cells (Supplementary Fig. S2D). To confirm that IFNγ increased radiation-induced cell death, we quantified PI⁺ dying B16F10 cells following treatment with IFNγ, radiotherapy, or the combination. We observed a synergistic increase in cell death following combination treatment with radiotherapy and IFNγ (Fig. 2F). To implicate ferroptosis in cell death induced by the combination treatment of radiotherapy and IFNγ, we conducted clonogenic survival experiments in HT1080 cells with IFNγ, radiotherapy, and liproxstatin-1. We observed that liproxstatin-1 administration blocked the synergistic reduction of cell survival observed upon treatment with IFNγ and radiotherapy (Fig. 2G; Supplementary Fig. S2E). The data indicate that the combination of IFNγ and radiotherapy regulates tumor-cell ferroptosis in vitro. To examine whether IFNγ alters radiotherapy efficacy in vivo, we inoculated HT1080 cells into NOD/SCID IL2γc-deficient (NSG) mice and treated mice with recombinant human IFNγ, radiotherapy, or the combination. Although low doses of IFNγ and radiotherapy slightly diminished tumor growth, the combination potently promoted tumor regression (Fig. 2H). To assess whether this is associated with tumor ferroptosis, we performed IHC staining in tumor tissues and quantified 4-hydoxynoneal (4-HNE), a lipid peroxidation by-product (23). Indeed, combination treatment with IFNγ and radiotherapy significantly increased tumoral 4-HNE levels (Fig. 2I; Supplementary Fig. S2F). To functionally demonstrate that ferroptosis contributed to synergistic tumor regression in vivo, we inoculated HT1080 cells into NSG mice and treated mice with IFNγ plus radiotherapy, liproxstatin-1, or all three agents. As expected, IFNγ plus radiotherapy induced tumor regression. Importantly, this effect was abrogated by simultaneous treatment with liproxstatin-1 (Fig. 2J). Together, these data demonstrate that CD8⁺ T cells directly sensitize tumor cells to radiotherapy through IFNγ by increasing lipid peroxidation and ferroptosis.

We next sought to define the molecular mechanism by which IFNγ sensitizes to radiotherapy and promotes tumor ferroptosis. We first performed unbiased RNA sequencing in HT1080 cells following treatment with IFNγ, radiotherapy, or the combination. Consistent with published results, interferon response genes, including CXCL9 and IRF1, were upregulated strongly by IFNγ treatment (Fig. 3A; ref. 24). As expected, radiotherapy induced an ATM-dependent transcriptional program (25). Dual treatment with radiotherapy and IFNγ surprisingly downregulated multiple cystine and cysteine transport channels, including a component of the cystine–glutamate transporter system xc⁻, SLC7A11 (Fig. 3A), which is a critical regulator of ferroptosis (26). Other ferroptosis regulators including GPX4, ACSL4, LPCAT3, and lipoxygenases (Supplementary Fig. S3A and S3B) were not altered in the same manner by dual treatment with radiotherapy plus IFNγ treatment. We next validated this finding. Treatment with IFNγ and radiotherapy individually reduced SLC7A11 transcripts (Fig. 3B) and protein levels (Fig. 3C) in HT1080 cells. Dual treatment resulted in a dramatic reduction of SLC7A11 in HT1080 cells (Fig. 3B and C). We also performed IHC staining for SLC7A11 in HT1080 tumor tissues (Fig. 2H) and found that IFNγ and radiotherapy reduced SLC7A11 expression in vivo (Fig. 3D). To determine the functional relevance of SLC7A11 reduction, we examined radiolabeled cystine uptake in HT1080 cells. Although both radiation and IFNγ each reduced cystine uptake, dual treatments resulted in a maximal decrease in cystine uptake in HT1080 cells (Fig. 3E). Cystine is a precursor for glutathione production, which is a required cofactor for GPX4 activity (27). To confirm that diminished cystine import altered cellular antioxidant stores, we quantified glutathione level in HT1080 cells following treatment with radiation, IFNγ, and the combination. We observed that the combination treatment significantly decreased glutathione level (Fig. 3F). Thus, radiotherapy and IFNγ synergistically downregulated SLC7A11 expression in tumor cells, resulting in diminished cystine transport and depletion of antioxidant stores in tumor cells to induce ferroptosis.

Next, we explored how radiotherapy downregulates SLC7A11 in tumor cells. Cells respond to radiotherapy through defined signaling cascades, including ATM activation (6, 25, 28). ATM has recently been shown to promote ferroptosis (29). We wondered whether ATM activation is involved in radiation-induced ferroptosis by regulating SLC7A11. To test this possibility, we treated HT1080 cells with KU60019 (30), an ATM inhibitor. ATM inhibition abrogated radiotherapy-mediated downregulation of SLC7A11 transcription (Supplementary Fig. S3C) and protein levels (Fig. 3G). Indeed, treatment with KU60019 diminished radiotherapy-induced lipid peroxidation in HT1080 cells (Fig. 3H). Further, siRNA targeting ATM prevented downregulation of SLC7A11 transcription (Supplementary Fig. S3D) and protein (Fig. 3I) following radiotherapy. To confirm this mechanism in vivo, we treated WT and shATM HT1080 tumors with radiotherapy (Supplementary Fig. S3E). We observed that WT but not ATM-silenced tumors showed downregulation of SLC7A11 expression (Fig. 3J; Supplementary Fig. S3F). Further, WT tumors showed increased 4-HNE lipid peroxidation in vivo following radiotherapy as compared with ATM-silenced tumors (Supplementary Fig. S3G and S3H). Therefore, radiotherapy transcriptionally represses SLC7A11 expression via ATM to promote tumoral ferroptosis.

We have previously reported that IFNγ negatively regulates SLC7A11 expression in a JAK–STAT1-dependent manner (16, 31). To validate this finding, we silenced STAT1 with siRNAs (siSTAT1) in HT1080 cells and found that IFNγ downregulated SLC7A11 in a STAT1-dependent manner (Supplementary Fig. S3I; Fig. 3K). Interestingly, IFNγ efficiently and comparably reduced SLC7A11 expression in both ATM-proficient and ATM-deficient HT1080 cells (Supplementary Fig. S3J). Conversely, regardless of STAT1 expression, radiotherapy efficiently reduced SLC7A11 expression in HT1080 cells (Supplementary Fig. S3K). Thus, radiotherapy and IFNγ synergistically repress tumor SLC7A11 through ATM and STAT1 signaling, respectively. To confirm this mechanism in vivo, we treated WT and STAT1 knockout HT1080 tumors with IFNγ (Supplementary Fig. S3L). We observed that WT but not STAT1-deficient tumors showed SLC7A11 downregulation following IFNγ treatment (Fig. 3L; Supplementary Fig. S3M). We also observed that WT tumors showed increased 4-HNE lipid oxidation following IFNγ treatment as compared with STAT1 knockout tumors (Supplementary Fig. S3N and S3O).

To functionally connect SLC7A11 to tumor ferroptosis induced by radiotherapy, we used small hairpin RNAs against SLC7A11 (shSLC7A11) to knock down SLC7A11 in HT1080 cells (Supplementary Fig. S3P). We found that knockdown of SLC7A11 promoted radiotherapy-induced cell death in HT1080 cells (Fig. 3M) and B16F10 cells (Supplementary Fig. S3Q). This effect was rescued by liproxstatin-1 in HT1080 cells (Fig. 3M) and B16F10 cells (Supplementary Fig. S3Q). To determine the importance of SLC7A11 in radiotherapy in vivo, we used CRISPR/Cas9 to delete SLC7A11 in B16F10 cells. We confirmed deletion by quantifying SLC7A11 mRNA levels, and, consistent with loss of system xc⁻ function, knockout cells were unable to export glutamate (Supplementary Fig. S3R and S3S). In line with previous reports (12, 26), we found SLC7A11-deficient cells underwent spontaneous cell death, which can be rescued by liproxstatin-1 or 2-mercaptoethanol (Supplementary Fig. S3T). We inoculated SLC7A11-proficient and SLC7A11-deficient B16F10 tumors into C57BL/6 mice and treated mice with radiotherapy. Radiotherapy efficacy was enhanced in SLC7A11-deficient cells in vivo (Fig. 3N). Examination of lipid peroxidation showed that SLC7A11-deficient tumors had increased baseline levels of lipid ROS (Fig. 3O). Moreover, radiotherapy-induced alterations in lipid peroxidation were further increased by loss of SLC7A11 (Fig. 3O). To confirm the importance of SLC7A11 to survival following combination IFNγ and radiotherapy, we overexpressed SLC7A11 in HT1080 cells (Supplementary Fig. S3U). We observed that overexpression significantly increased clonogenic survival following treatment with radiotherapy and IFNγ (Fig. 3P). To further confirm the importance of SLC7A11 expression, we performed rescue experiments in which SLC7A11 was expressed in SLC7A11-deficient cells (Supplementary Fig. S3V). In addition, we observed that overexpression of SLC7A11 reversed the sensitivity of SLC7A11-deficient tumors to radiotherapy in vivo (Fig. 3Q). Collectively, these results suggest that radiotherapy-activated ATM and IFNγ-induced STAT1 signaling jointly target SLC7A11 to modulate cystine uptake and promote tumor lipid peroxidation and ferroptosis.

Immunotherapy augments radiotherapy efficacy in vivo, but the mechanisms through which immune checkpoint blockade enhances tumor-cell death in irradiated lesions are not well defined (4, 32, 33). Checkpoint therapy including anti–PD-L1 and anti-CTLA4 activates CD8⁺ T cells and potentiates the IFNγ signaling pathway to control tumor progression (1). We hypothesized that radiotherapy and immunotherapy would synergize via ferroptosis in vivo to improve the efficacy of both therapies. To test this hypothesis, we established B16F10 tumors in C57BL/6 mice and treated mice with anti-CTLA4 mAb, radiotherapy, or the combination. As previously reported, anti-CTLA4 monotherapy was minimally efficacious in B16F10 tumors (32). However, we found that it potentiated radiotherapy efficacy in irradiated tumors (Fig. 4A). Consistent with our observations (Fig. 1), radiotherapy increased lipid peroxidation. Interestingly, the combination treatment with anti-CTLA4 and radiotherapy further enhanced tumoral lipid peroxidation (Fig. 4B). The combination of anti-CTLA4 and radiotherapy enhanced CD8⁺ T-cell trafficking into tumors (Supplementary Fig. S4A and S4B). Radiotherapy increased tumor-infiltrating Ki67⁺CD8⁺ T cells, and this was augmented by anti-CTLA4 (Fig. 4C; Supplementary Fig. S4C). Furthermore, we detected enhanced CD8⁺ T-cell function following combined anti-CTLA4 and radiotherapy as marked by increased IFNγ and granzyme B production (Fig. 4D and E; Supplementary Fig. S4D and S4E). Similarly, we observed that combination therapy increased CD4⁺ T-cell number and IFNγ expression (Supplementary Fig. S4F and S4G). To determine whether tumor ferroptosis contributed to the synergy between anti-CTLA4 and radiotherapy in vivo, we inoculated B16F10 tumors in C57BL/6 mice and treated them with the combination of anti-CTLA4 plus radiotherapy, liproxstatin-1, or all three treatments. Interestingly, liproxstatin-1 diminished the combined therapeutic efficacy of anti-CTLA4 and radiotherapy as shown by tumor volume (Fig. 4F), as well as tumor lipid peroxidation (Supplementary Fig. S4H). To extend our findings to an additional model, we inoculated ID8 tumors into the peritoneal cavities of C57BL/6 mice, and treated mice with anti-CTLA4, radiotherapy, or the combination. Anti-CTLA4 and radiotherapy synergistically enhanced tumor control in vivo (Fig. 4G). Again, we observed increased lipid peroxidation following radiotherapy, and anti-CTLA4 mediated enhancement of lipid peroxidation following dual treatment with radiotherapy and anti-CTLA4 (Fig. 4H). To explore if other forms of immune checkpoint blockade also induced lipid peroxidation, we established B16F10 tumors in mice and treated the tumors with radiotherapy, anti–PD-L1, or both. Although anti–PD-L1 or radiotherapy alone controlled tumor growth, the dual treatment was more effective in reducing tumor growth than any single therapy (Fig. 4I). We also detected the highest levels of tumor lipid peroxidation following combination therapy (Fig. 4J).

SLC7A11-deficient cells were more sensitive to radiotherapy in vitro and in vivo because of their increased ferroptosis susceptibility (Fig. 3). We hypothesized that ferroptosis-sensitive tumors may benefit more from immune checkpoint blockade. To examine this, we established SLC7A11-proficient or SLC7A11-deficient B16F10 tumors in C57BL/6 mice and treated mice with anti–PD-L1. We detected higher tumoral lipid ROS in SLC7A11-deficient tumor cells than in WT tumor cells regardless of anti–PD-L1 therapy (Supplementary Fig. S4I). Furthermore, SLC7A11-deficient tumors were more sensitive to anti–PD-L1 therapy, as shown by increased tumor C11BODIPY fluorescence and reduced tumor volume (Supplementary Fig. S4J) compared with WT tumors.

To additionally support that ferroptosis is a mechanism of synergy between immunotherapy and radiotherapy, we inoculated parental or RSL-3 ferroptosis-resistant cells into mice and treated mice with anti–PD-L1 and radiotherapy. We observed that in contrast to the parental cells, RSL-3–resistant B16F10 cells were largely insensitive to combined treatment with anti–PD-L1 and radiotherapy (Fig. 4K). Moreover, this insensitivity was accompanied by diminished C11BODIPY in tumor cells (Fig. 4L). To define the mechanisms contributing to resistance to therapy, we established ovalbumin expressing parental or RSL-3 ferroptosis-resistant B16F10 tumors and treated the mice with OT-1 transgenic T cells. We observed that OT-1 therapy controlled WT but not RSL-3–resistant tumors. This suggests that ferroptosis-resistant tumors are resistant to T-cell effector function (Fig. 4M).

To examine whether SLC7A11 modulated the induction of durable immune responses, we established parental or SLC7A11-deficient B16F10 tumors and treated them with radiotherapy and anti–PD-L1 therapy. We observed that SLC7A11-deficient tumors were more sensitive to combination therapy, with complete long-term responses in 56% (14/25) of SLC7A11-deficient tumor-inoculated mice (Fig. 4N). To test whether these mice developed immunologic memory, WT B16F10 tumors were reinoculated in cured mice. We noted that tumors established in control but not previously cured mice, suggestive of the establishment of T-cell memory (Fig. 4O). Collectively, these data suggest that immunotherapy synergizes with radiotherapy to induce tumor ferroptosis and establish T-cell immunity.

A major challenge in oncology is how to rationally and effectively integrate PD-1/PD-L1 and CTLA4 immune checkpoint blockade with traditional oncologic treatment modalities, including radiotherapy (1). CD8⁺ T cells are required for the efficacy of both immunotherapy and radiotherapy. We have now demonstrated that effector T cells and radiotherapy interact through ferroptosis to promote tumor clearance. Radiotherapy classically promotes tumor-cell death primarily through the induction of double-stranded DNA breaks in tumor cells which, if unrepaired, lead to mitotic catastrophe and unregulated tumor cell death (7). We provide the first evidence that radiation also promotes lipid peroxidation, resulting in tumoral ferroptosis. Concurrent administration of genotoxic agents to augment DNA damage has been harnessed clinically to improve radiotherapy efficacy and improve the outcomes of patients with cancer. We have found that ferroptosis agonists can sensitize tumors to radiation both in vitro and in vivo, uncovering new therapeutic strategies to improve radiotherapy efficacy.

Ferroptosis is regulated by lipid metabolism, iron metabolism, and cysteine metabolism (34). ATM regulates cell survival following radiotherapy (25) and has also been shown to regulate ferroptosis (29). We have shown that radiotherapy activates ATM to suppress SLC7A11 expression, limit tumor cystine uptake, and diminish glutathione, and increases lipid oxidative damage to mediate tumor-cell ferroptosis. We have demonstrated that biochemical and genetic inhibition of ATM can prevent loss of SLC7A11 expression, reduce lipid oxidation, and rescue tumor-cell ferroptosis induced by radiotherapy. Similarly, ATM promotes DNA-damage repair and cell survival following radiation but is also required for radiation-induced ferroptosis. Along this line, p53 contributes to tumor-cell survival but also promotes tumor-cell death (35). Several pharmacologic inhibitors are being studied as radiosensitizers (36). Given that ATM signaling may play an unexpected role in promoting tumor-cell ferroptosis, additional study is required to understand the context-dependent consequence of ATM inhibition. Notably, we have focused on a role of SLC7A11 and ATM in radiotherapy-associated ferroptosis. We do not rule out that radiation may affect other ferroptosis-associated genes, which may collectively contribute to radiotherapy-mediated ferroptosis in different contexts.

CD8⁺ T-cell tumor infiltration and IFNγ signaling pathway activation are key features of effective immunotherapy in patients with cancer (1, 37, 38). We have recently shown that CD8⁺ T cells modulate tumoral ferroptosis (16). We now extend this work, showing that CD8⁺ T cells can alter the sensitivity of tumor cells to radiotherapy by promoting ferroptosis. CD8⁺ T cell–derived IFNγ and radiation coordinately inhibit tumor SLC7A11 expression to induce tumor-cell ferroptosis. This translates into potent tumor immunity, tumor ferroptosis, and tumor regression in several murine models treated with the combination of radiotherapy and PD-L1/PD-1 or CTLA4 blockade. Moreover, ferroptosis inhibition or resistance diminishes the synergy between immunotherapy and radiotherapy in vivo. The major immunosuppressive mechanisms in the tumor microenvironment include high PD-L1 expression (39, 40) and regulatory T-cell infiltration (41). Radiation can target PD-L1–expressing cells and regulatory T cells (15). More recently, cGAS–STING sensing of cytoplasmic DNA and micronuclei following radiation has emerged as an important connection between innate immunity and radiotherapy (3, 4, 32). Our work suggests that tumor ferroptosis is a novel intersection between radiotherapy and adaptive immunity.

In conclusion, this work highlights that ferroptosis may be an untapped therapeutic mechanism and focus for the development of effective pharmacologic and immunotherapeutic combinatorial approaches with radiotherapy for the treatment of cancer.

Ferrostatin-1 (347174-05-4), liproxstatin-1 (950455-15-9), trolox (53188-07-1), DFO (138-14-7), RSL-3 (1219810-16-8), erastin (571203-78-6), sulfasalazine (599-79-1), atorvastatin (599-79-1), and KU60019 (925701-46-8) were purchased from Cayman. The concentration used is shown in Supplementary Table. S1. C11BODIPY (D3861) was purchased from Invitrogen. Mouse recombinant IFNγ (485-MI-100), anti-mouse IFNγ block antibody (MAB485), and human recombinant IFNγ (285-IF-100) were purchased from R&D. Anti-mouse IFNGR1 antibody (16-1193-85) was purchased from Thermo Fisher. The following antibodies were used for immunoblots: Anti-human SLC7A11 antibody (12691), anti–β-actin (12262), anti-GAPDH (2118), anti-STAT1 (9172), anti–phospho-STAT1 (9167), anti-ATM (2873), and anti-ACSL4 (4047) were bought from Cell Signaling Technology. Anti-GPX4 (ab41787) was from Abcam. In addition, IgG (BE0087, BE0090), anti-mouse CTLA4 (BE0131), anti-mouse PD-L1 (BE0101), and anti-mouse CD8 (BE0117) were purchased from BioXcell. 4-HNE antibody (ab46545) was purchased from Abcam. Cyst(e)inase was generated as previously described (18).

All the cell lines used here, including B16F10, ID8, HT1080, and HEK-293T cell lines, were bought from the ATCC. Cells were regularly tested for Mycoplasma contamination using the Adenosine 5′-triphosphate (ATP) Bioluminescent Assay Kit (Sigma) every 2 weeks. Cells were thawed at early passage and cultured for up to 16 weeks in total.

To assess cell death, cells were treated as indicated, then collected and stained with PI. For in vitro C11BODIPY analysis, cells were treated with test compounds for the indicated times, then harvested and incubated with C11BODIPY (581/591) for 20 minutes at 37°C. For ex vivo C11BODIPY analysis, tumors were filtered and underwent gradient centrifugation, then were incubated with 2 μmol/L C11BODIPY and anti-CD45 antibody. C11BODIPY FITC channel median fluorescence intensity data were normalized to untreated controls. To quantify T-cell cytokine production, mononuclear cells were isolated, stained, and analyzed as previously described (40).

Radiotherapy was delivered using a Philips RT250 model orthovoltage unit (Kimtron Medical) or a Small Animal Radiation Research Platform (XStrahl Medical and Life Sciences) at a dose rate of approximately 2 Gy per minute as previously described (42). Experiments were performed in the Experimental Irradiation Shared Resource (University of Michigan). Focal irradiation was provided to flank tumors via lead shielding or collimation. Whole-abdominal radiotherapy was provided to intraperitoneal tumors with shield above the diaphragm and below the pelvis.

Clonogenic survival analysis was conducted as previously described. Normalized surviving fraction was represented as relative to compound alone as previously described (42).

Six- to 8-week-old NSG and WT C57BL/6 mice were obtained from the Jackson Laboratory. OT-I colony was maintained internally. All mice were maintained under specific pathogen–free housing with a maximum of 5 mice per cage. Mice were randomized to experimental groups when tumors reached 50 mm³ (HT1080, B16F10) or on day of first bioluminescent imaging (ID8). Tumor size was quantified, and anti–PD-L1 was provided as previously described (16). For IFN experiments, 65 μg of human IFNγ was given 24 hours prior to radiation (6–8 Gy, single fraction) as well as 2 days after radiotherapy. The primary endpoint of all experiments was tumor size. All the animal studies were conducted under the approval of the University of Michigan Committee on Use and Care of Animals (PRO00008278).

Single-cell suspension was prepared from spleens and lymph nodes from OT-I mice, then stimulated with 5 μg/mL OVA peptide (Sigma, S7951) in RPMI-1640 medium containing 10% FBS. After 3 days, CD8⁺ T cells were isolated via magnetically labeled antibodies (StemCell Technologies, 19853) and then cultured in RPMI-1640 medium containing IL2 (R&D Biosystems) and 2-mercaptoethanol (Gibco, 21985023). After 3 days, supernatant was harvested and filtered through 0.2-μm filter prior to subsequent use. For blocking, B16F10 cells were pretreated with 10 μg/mL anti-IFNGR as well as 10 ng/mL anti-IFNγ antibody.

Tumor tissues were harvested and fixed in 10% formalin and embedded in paraffin. Staining was performed on tumors treated with the indicated conditions using anti–4-HNE or anti-SLC7A11 antibody and counterstained using hematoxylin. The alkaline phosphatase optical density was quantified in ImageJ. Results were normalized to untreated controls.

Cell lines were treated as indicated, and total RNA was purified using an mRNA kit (Qiagen). Libraries were prepared as previously described, and the edgeR package was used to generate reads per kilobase of transcript per million mapped reads values (43). Differential expression analysis was performed using the limma-R package as previously described. Minmax data normalization was used for heat-map generation. Gene set enrichment analysis was performed to identify significantly enriched gene sets (FDR < 0.20). Gene Expression Omnibus accession number: GSE137946.

Total RNA was extracted using TRIzol and phenol-chloroform phase separation. Reverse transcription was performed using the RevertAid RT Reverse Transcription Kit (Invitrogen, K1691). SYBR Green (Invitrogen, 4368702) was used for quantitative PCR with a StepOne Real Time PCR system (Applied Biosystems). β-actin was used as an internal control. Results are represented as fold change from untreated controls. The sequences of primers for qPCR are as follows:

• Human SLC7A11 forward: 5′-TGCTGGGCTGATTTTATCTTCG-3′,

• Human SLC7A11 reverse: 5′-GAAAGGGCAACCATGAAGAGG-3′,

• Human β-actin forward: 5′-TGGTATCGTGGAAGGACTC-3′,

• Human β-actin reverse: 5′-AGTAGAGGCAGGGATGATG-3′,

• mouse ACSL3 forward: 5′-TCCTGGCTGCGATACACTTG-3′,

• mouse ACSL3 reverse: 5′- CCAAAGTCAAGGGCTCGGAT-3′,

• mouse β-actin forward: 5′-TGAGCTGCGTTTTACACCCT-3′,

• mouse β-actin reverse: 5′-TTTGGGGGATGTTTGCTCCA-3′

HEK-293T cells were transfected with a nontargeted retroviral vector, or a retroviral vector encoding shRNA targeting SLC7A11. Virus-containing supernatant was harvested, and HT1080 cells and B16F10 cells were transduced and selected prior to subsequent use.

Dharmacon shRNA sequence:

• Human shSLC7A11: ATAATAAAGAGATAATACG

• Mouse shSLC7A11: ATTAGCTGTATAACTCCAGGG

CRISPR/Cas9 constructs (Santa Cruz Biotechnology) were utilized to generate B16F10 subclones deficient in ACSL4 (sc-424503-NIC) and SLC7A11 (sc-424104-NIC). In addition, 2-mercaptoethanol (Gibco, 21985023) was added to support SLC7A11 knockout cell growth. lentiCRISPR v2 (Addgene, 52961)-based constructs were used for knockout of ACSL3 in B16F10 cells, and the designed gRNA sequences were GTAATGATATGCCGCAGACG and GGTGTG TACAATGACACCTT. HEK-293T cells were used to produce lentivirus. B16F10 cells were infected with scramble control or gene-targeted lentivirus and selected with 2 μg/mL puromycin prior to subsequent use.

To perform the ectopic SLC7A11 rescue assay, a cytomegalovirus promoter-driven mouse SLC7A11-IRES-GFP construct was generated by Gibson Assembly. This construct was transfected into SLC7A11-deficient B16F10 cells. Clones were selected with G418 and fluorescence-activated cell sorting.

HT1080 cells were transfected with control or siRNA targeting human ATM (4392420-S532284, Thermo Fisher) or STAT1 (SASI_Rn02_00204016, Sigma) using lipofectamine RNAiMAX (Invitrogen) prior to experimentation.

HT1080 cells were treated with either IFNγ (10 ng/mL), radiotherapy (10 Gy), or both for 24 hours. Medium was removed and changed to cystine-free medium. L-¹⁴C-Cystine (0.2 μCi/mL) was added and incubated for 15 to 45 minutes. Cells were washed and lysed. A Beckman liquid scintillation counter was used to quantify radioactive cystine uptake. Results were normalized to total protein as determined by Pierce BCA Protein Assay Kit (Thermo, 23225).

HT1080 cells were treated with either IFNγ (10 ng/mL), radiotherapy (16 Gy), or both for 24 hours. Glutathione was quantified using the GSH-Glo assay (Promega). Results were normalized to cell viability.

The Glutamate Colorimetric Assay Kit (BioVision, K629) was used for detection of extracellular glutamate released into the medium. Parental or SLC7A11 knockout B16F10 cells were seeded overnight in the presence of 2-mercaptoethanol to support knockout cell growth. Results normalized to cell viability.

Cells were collected and seeded into 96-well plates. To determine the effect of treatment on cell growth and viability, 10% volume of alamarBlue (Bio-Rad) was added directly into the medium and incubated for 4 hours. Absorbance at wavelengths of 570 and 600 nm was measured. Results were normalized to untreated controls and shown as relative cell viability (%).

Statistical analyses were performed with an unpaired Student t test for experiments with two groups (GraphPad Prism). One-way ANOVA was used to compare multiple experimental groups, and two-way ANOVA was used to compare continuous outcomes across multiple experimental groups. For all tests, P < 0.05 was considered significant. Sample size was not predetermined. Unless noted, samples were independent biological replicates.

E.M. Stone is a consultant for and has equity interest in Aeglea Biotherapeutics, and is an inventor on intellectual property related to this work owned by The University of Texas at Austin. G. Georgiou has ownership interest (including patents) in Aeglea Biotherapeutics. M.A. Morgan is an advisory board member for AstraZeneca, reports receiving a commercial research grant from AstraZeneca, and has received honoraria from the speakers’ bureau of AstraZeneca. W. Zou is a scientific board member at EMD, Cstone, Synlogic, and Henlius. No potential conflicts of interest were disclosed by the other authors.

Conception and design: X. Lang, M.D. Green, T.S. Lawrence, W. Zou

Development of methodology: X. Lang, M.D. Green, W. Wang, L. Jiang, Q. Zhang, I. Kryczek

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): X. Lang, M.D. Green, W. Wang, J. Yu, A. Dow, A.L. Saripalli, S. Wei, E.M. Stone, T.S. Lawrence

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): X. Lang, M.D. Green, W. Wang, J.E. Choi, G. Georgiou, M. Cieslik, M.A. Morgan, A.M. Chinnaiyan, T.S. Lawrence, W. Zou

Writing, review, and/or revision of the manuscript: X. Lang, M.D. Green, A.L. Saripalli, G. Georgiou, D.R. Wahl, M.A. Morgan, A.M. Chinnaiyan, T.S. Lawrence, W. Zou

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): M.D. Green, P. Liao, J. Zhou, Q. Zhang, S. Wei, W. Szeliga, L. Vatan, G. Georgiou, A.M. Chinnaiyan, W. Zou

Study supervision: M.D. Green, W. Zou

This work was supported in part by research grants from the NIH/NCI for W. Zou (CA217648, CA123088, CA099985, CA193136, and CA152470), E.M. Stone and G. Georgiou (CA189623), T.S. Lawrence (U01CA216440), and A.M. Chinnaiyan (1UM1HG006508); and the NIH through the University of Michigan Rogel Cancer Center Support Grant (P30CA46592). The authors would like to thank all members of the Zou lab for their insightful feedback, as well as acknowledge the support of the University of Michigan Experimental Irradiation Shared Resource, the Research Histology and Immunohistochemistry Core, and the Advanced Genomics Core.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.