Selective disruption of lipid peroxide homeostasis in intratumoral regulatory T cells by targeting FSP1 enhances cancer immunity

Abstract

A burgeoning approach to treating cancer is the pharmacological induction of ferroptotic cell death of tumor cells. However, the impact of disrupting antiferroptotic pathways in the broader tumor microenvironment (TME), such as in immune cells, is still undefined and may complicate treatments. Here, we show that ferroptosis suppressor protein 1 (FSP1/Aifm2) is critically required for regulatory T cell (T_reg cell) resistance to ferroptosis and their immunosuppressive function within the TME. Compared to other canonical ferroptosis regulators such as GPX4 and NRF2, only FSP1 was induced upon T cell activation. Deletion of Aifm2 in all T cells, or T_reg cells specifically, enhanced tumor control by selectively disrupting T_reg cell immunosuppression within tumors without inciting autoimmune pathology in mice. As opposed to deletion of Gpx4 in all T cells, T cell deletion of Aifm2 did not impair antigen-specific CD8⁺ T cell responses. These results reveal an opportunity for targeting a regulator of ferroptosis that can not only directly target cancer cells but also simultaneously enhance anticancer immune responses without inciting autoimmunity.

Deletion of Aifm2 disrupts T_reg cell immunosuppression within tumors without inciting autoimmune pathology in mice.

INTRODUCTION

Ferroptosis, an iron-dependent form of regulated cell death driven by unchecked lipid peroxide accumulation, has emerged as a promising approach for treating therapy-resistant cancers (1–3). Unlike apoptosis, ferroptosis can release danger-associated molecular patterns that activate immune responses (4, 5). While glutathione peroxidase 4 (GPX4) is the primary defense against ferroptosis, cancer cells often express alternative protective ferroptotic regulators, such as ferroptosis suppressor protein 1 (FSP1) and GTP cyclohydrolase 1 (GCH1), to resist GPX4 inhibition (6–11).

Regulatory T cells (T_reg cells), a suppressive subset of CD4⁺ T cells, pose a major barrier to effective cancer immunotherapy by inhibiting antitumor responses through direct cell contact and the release of immunosuppressive molecules (12–14). T_reg cells are uniquely adapted to the nutrient-deprived and hypoxic tumor microenvironment (TME) by preferentially using oxidative phosphorylation, importing lipids via CD36, and using lactate and fatty acids as alternative carbon sources for their metabolism (15–18). Moreover, T_reg cells up-regulate antioxidant pathways—including glutathione synthesis and thioredoxin—to mitigate oxidative stress and preserve their function in the TME (19–21). Although T_reg cells depend on GPX4 to prevent ferroptosis, inhibiting GPX4 is not a viable strategy for cancer immunotherapy as its inhibition also impairs effector CD8⁺ and CD4⁺ T cell survival (22, 23).

In this study, we establish that FSP1/Aifm2 is a critical effector protein for T_reg cell resistance to ferroptosis induction in vitro and T_reg cell function in vivo in tumors. We found that T_reg cells use FSP1 to prevent lipid peroxide accumulation despite having higher levels of intracellular reactive oxygen species (ROS). Aifm2 deficiency was sufficient to increase the sensitivity of T_reg cells to ferroptosis induction with GPX4 inhibition in vitro. Aifm2 deficiency in T_reg cells decreased the expression of proteins essential for T_reg effector function but did not affect effector T cell immune homeostasis or peripheral tolerance in young or aged mice. Total ablation of Aifm2 in all T cells did not disrupt antitumor T cell responses but did diminish the immunosuppressive function of T_reg cells. These findings are important for the translation of pharmacological FSP1 inhibitors for cancer as they show that FSP1 inhibition not only can directly sensitize cancer cells to ferroptosis (24, 25), but it may also enhance immune responses to cancer by selectively blocking T_reg cell function in the TME without generating autoimmune toxicity.

RESULTS

T_reg cells resist lipid peroxide accumulation and ferroptosis

To understand the regulation of ROS across different T cell populations, we fluorescence-activated cell sorting (FACS)–purified naive T_reg cells and conventional CD4⁺ T cells (Tconv) for activation in vitro and measured their total ROS and lipid peroxide levels using fluorescent reagents detected by flow cytometry (Fig. 1A). In accordance with previous studies, we observed that T_reg cells had increased (~2-fold) intracellular ROS levels compared to Tconv cells (Fig. 1B) (18, 26). Next, we directly measured lipid peroxidation by measuring the amount of oxidized C11-BODIPY and observed that T_reg cells had decreased amounts of lipid peroxides compared to Tconv cells (Fig. 1C). While RSL3 increased oxidized C11-BODIPY in both T_reg cells and Tconv cells, Tconv cells increased CD11-BODIPY^OX significantly more than T_reg cells (Fig. 1D). Only Ferrostatin-1 (Fer-1), a radical-trapping antioxidant, but not the pan-caspase inhibitor Z-VAD-FMK or RIP1 kinase inhibitor Nectrostatin-1 (Nec-1) treatments, reversed CD11-BODIPY^OX elevation, as expected (Fig. 1D). Improved control of lipid oxidation ultimately made T_reg cells more resistant to ferroptotic cell death upon inhibition of GPX4 across a series of concentrations of RSL3 (Fig. 1E). Decreased cell viability was confirmed to be due to ferroptosis, as only treatment with Fer-1, but not other programmed cell death inhibitors, was able to rescue cell viability (Fig. 1F). The differences in sensitivity to ferroptosis were not due to the expression of the essential T_reg cell transcription factor Foxp3, as in vitro induced T_reg cells (iT_reg cells) were as sensitive to GPX4 inhibition as Tconv cells (Fig. 1E) (19). To confirm these findings without activating or culturing T cells long term in vitro, we isolated T cells from spleens of mice and treated them with RSL3 alone or RSL3 and Fer-1 for 4 hours ex vivo. This revealed that while CD8⁺ and CD4⁺ Tconv cells were highly sensitive to GPX4 inhibition, T_reg cells remained highly resistant to ferroptosis ex vivo (Fig. 1G). Therefore, despite having higher amounts of ROS, T_reg cells reduce their accumulation of lipid peroxides, which likely increases their resistance to ferroptosis.

Fig. 1. T_reg cells naturally resist lipid peroxide accumulation and ferroptosis.

(A) Schematic for in vitro T cell assays. (B) Representative flow plots (left) and quantification (right) of bulk ROS by CellROX staining in T_reg cells versus Tconv cells after activation and culture for 4 days in vitro. Data pooled from five independent experiments. (C) Representative flow plots for detection of oxidized C11-BODIPY in T_reg cells versus Tconv cells that were activated for 4 days in vitro. (D) Quantification of oxidized C11-BODIPY in T_reg cells versus CD4⁺ Tconv cells in vitro +/− 1 μM RSL3 in combination with 2 μM Fer-1, 100 μM Z-VAD-FMK, or 10 μM Nec-1. Data pooled from three experiments. (E) Cell viability of T_reg cells (nT_reg cells), CD4⁺ Tconv cells, and induced T_reg cells (Foxp3⁺ iT_reg cells) treated with the indicated concentrations of RSL3 for 4 hours. Representative data from two independent experiments. (F) Cell viability of T_reg cells versus CD4⁺ Tconv cells treated in vitro +/− 1 μM RSL3 in combination with 2 μM Fer-1, 100 μM Z-VAD-FMK, or 10 μM Nec-1. Data pooled from two experiments. (G) Frequencies of viable CD8⁺ T (left), CD4⁺ Tconv (middle), and T_reg cells (right) from spleen plotted as the percent of CD45⁺ cells +/− 1 μM RSL3 and 2 μM Fer-1. Data from n = 7 mice. For all plots, *P < 0.05, **P < 0.01, and ***P < 0.001 by two-way analysis of variance (ANOVA) (D to F) or one-way ANOVA (G) or Student’s t test (B), mean ± SD [(B) and (D) to (F)] or mean ± SEM (G). ns, not significant. D0, day 0; D0-D4, days 0 to 4. APC, allophycocyanin. gMFI, geometric mean fluorescent intensity.

FSP1 expression in T_reg cells confers resistance to ferroptosis

T cell activation increases levels of ROS, which can be tuned by the expression of antioxidant genes to regulate cytosolic signaling pathways (Fig. 2A) (22, 27–32). Analysis of gene expression by targeted quantitative polymerase chain reaction (qPCR) or bulk RNA sequencing of FACS-purified T_reg cells and Tconv cells that were naive (CD44^loCD62L^hi) or in vivo activated (CD44^hiCD62L^lo) revealed that Aifm2 and Gch1 exhibited increased expression in activated T_reg cells and Tconv cells (Fig. 2B and fig. S1, A and B) or upon in vitro T cell activation, but not other antiferroptotic proteins, e.g., Gpx4, Nfe2l2, Keap1, Acsl3, Acsl4, or Lpcat3. Next, a direct comparison of in vitro–activated T_reg cells versus Tconv cells (analyzed 4 days after anti-CD3/anti-CD28 of naive cells) revealed a significant induction of Aifm2 and Gch1 expression in T_reg cells compared to Tconvs, but not Keap1, Nfe2l2, Gpx4, Acsl3, Acsl4, or Lpcat3 (fig. S1C). These results suggested that Aifm2/FSP1 and Gch1 are more highly expressed in T_reg cells, and this may promote their enhanced resistance to ferroptosis.

Fig. 2. FSP1 drives T_reg cell resistance to ferroptosis.

(A) Schematic of ferroptotic regulatory pathways and key protein mediators. (B) Fold change in expression of indicated antiferroptotic genes from sorted naive (CD44^loCD62L^hi) versus in vivo activated (CD44^hiCD62L^lo) T_reg cells by reverse transcription (RT)–qPCR (n = 3 mice). (C) Fold change in gene expression of indicated antiferroptotic genes comparing activated CD4⁺ Tconv cells to nT_reg cells and iT_reg cells by RT-qPCR (n = 5 per group from two pooled experiments). (D) Treatment of wild-type versus Aifm2-deficient T_reg cells with indicated concentrations of RSL3 for 4 hours. Error bars are mean ± SD from three technical replicates of cells from a single mouse. Data shown are from one of two experimental repeats. (E) Aifm2 expression from bulk RNA sequencing of human peripheral blood mononuclear cell (PBMC) populations from Database of Immune Cell eQTLs, expression, epigenomics. (F) Differential expression analysis of Aifm2 expression across different PBMC populations from data in (E). For all plots, *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001 by one-way ANOVA (C) or two-way ANOVA [(B) and (D)], mean ± SD [(B) to (D)]. T_FH, T follicular helper; T_H2, T helper 2; NK, natural killer; TPM, transcripts per million; eQTL, expression quantitative trait loci.

Differentiation of Tconv cells into iT_reg cells by induction of Foxp3 did not confer resistance to ferroptosis (Fig. 1E). We hypothesized that Foxp3 expression alone in iT_reg cells may not induce Aifm2 or Gch1, thereby leaving iT_reg cells susceptible to ferroptosis induction. Therefore, we performed qPCR analysis for the expression of key ferroptosis regulators in iT_reg cells. This revealed that iT_reg cells expressed antioxidant genes similarly to Tconv cells and did not exhibit elevated expression of Aifm2 or Gch1 (Fig. 2C). In addition, we observed that natural T_reg cells (nT_reg cells) had reduced levels of Gpx4, Nfe2l2, and Lpcat3 in comparison to Tconv cells, suggesting a different antioxidant program in each T cell subset (Fig. 2C). These results demonstrate that although nT_reg cells and iT_reg cells both express Foxp3, their expression of ferroptosis regulators is distinct.

While GCH1 was shown to be crucial for T cell proliferative capacity, the role of FSP1 in T_reg cell biology is unexplored (31). To determine whether increased FSP1 expression in T_reg cells contributes to their enhanced resistance to ferroptosis compared to Tconv and iT_reg cells, we crossed mice carrying LoxP-flanked Aifm2^fl alleles with Foxp3-GFP-hCre to specifically delete Aifm2 in T_reg cells (called Treg.Aifm2^Δ/Δ mice) (33). Treatment of Aifm2-deficient T_reg cells with RSL3 to induce ferroptosis revealed that Aifm2-deficient T_reg cells were significantly more sensitive to RSL3-mediated cell death compared to wild-type T_reg cells (Fig. 2D). Thus, FSP1 promotes nT_reg cell resistance to ferroptosis in vitro.

Last, to investigate whether FSP1 may function similarly in human T_reg cells, we used comparative gene expression analysis from human peripheral blood mononuclear cells (PBMCs) and confirmed that activated human T_reg cells (memory T_reg cell) also exhibit the highest expression of AIFM2 compared both to naive T_reg cells and all other activated T cell populations (Fig. 2, F to G). These results strongly suggest that increased FSP1/Aifm2 expression in mouse and human T_reg cells protects them from ferroptosis induction in vitro.

T_reg cell–specific FSP1 deletion does not affect peripheral tolerance

Disruption of prosurvival genes in T_reg cells often leads to increased susceptibility to autoimmunity due to a loss of T_reg cell immunosuppression. To directly test whether T_reg cell–specific deletion of Aifm2 disrupted T_reg cell function in vivo, we analyzed lymphoid and nonlymphoid tissues from wild-type versus Treg.Aifm2^Δ/Δ mice in early (16 weeks) and late (28 weeks) life. There were no changes to the frequency of CD4⁺ Tconv, CD8⁺ T, or T_reg cells across lymphoid organs (lymph node and spleen) at any age (Fig. 3A and fig. S2, A and B). In the organ tissues examined by flow cytometry (colon, skin, and lung), while CD4⁺ Tconv and CD8⁺ T cell frequencies did not change in Treg.Aifm2^Δ/Δ mice, we did note a modest reduction in T_reg cell frequency in the lung and colon of 16-week-old Treg.Aifm2^Δ/Δ mice (Fig. 3A and fig. S2C). However, there were no changes to the activation state of CD4⁺ Tconv, CD8⁺ T, or T_reg cells in any of these tissues examined (Fig. 3B and fig. S2D). Examination of organ tissues (lung, liver, and kidney) did not reveal any signs of increased immune infiltration or autoimmunity by histological analyses (Fig. 3C). These data suggest that T_reg cell–specific Aifm2 deficiency does not increase the prevalence of autoimmunity in mice.

Fig. 3. Deletion of Aifm2 in T_reg cells does not disrupt peripheral tolerance in mice.

(A) Frequency of total CD8⁺ T cells (left), CD4⁺ Tconv cells (middle), and T_reg cells (right) as a percentage of total CD45⁺ immune cells in lymphoid tissues and organ tissues in 16-week-old mice. (B) Frequencies of activated (CD44^hiCD62^lo) CD8⁺ T cells (left), CD4⁺ Tconv cells (middle), and T_reg cells (right) in lymph nodes (LN) and spleen of 16-week-old mice. (C) Representative hematoxylin and eosin histology images of liver, lung, and kidney (left) and cumulative pathology scoring (right) for 28-week-old Treg.Aifm2^+/+ and Treg.Aifm2^Δ/Δ mice. Data from n = 4 for Treg.Aifm2^+/+ and n = 5 for Treg.Aifm2^Δ/Δ mice.¹Liver, kidney, and lung: Based on immune cell infiltrates (−, no immune infiltration; +, mild/sporadic immune infiltration, ++, high immune infiltration). All representative images are in dimensions of 1000 by 1000 pixels. Scale bars, 50 μm. (D) Schematic of pathway for conversion of adenosine triphosphate (ATP) to immunosuppressive adenosine mediated by the ectonucleotidases CD39 and CD73. (E) Representative histograms and gMFI of CD73 and CD39 in wild-type versus Aifm2-deficient T_reg cells across lymphoid tissues and nonlymphoid tissues in 16-week-old mice. Data from n = 5 for Treg.Aifm2^+/+ and n = 6 for Treg.Aifm2^Δ/Δ mice for 16-week-old mice. For all plots, *P < 0.05, **P < 0.01, and ***P < 0.001 by two-way ANOVA [(A) and (E)] or Student’s t test (B), mean ± SEM. PE, phycoerythrin; PE-Cy7, phycoerythrin-cyanine 7.

Phenotyping of the T_reg cells from Treg.Aifm2^Δ/Δ mice revealed no significant changes in the expression of canonical lineage defining T_reg cell markers (i.e., CD25, Foxp3, and NRP1) in either lymphoid or organ tissues at 16 weeks (fig. S2E). However, there was a significant decrease in CD73 expression on colonic and lung T_reg cells (Fig. 3, D and E). CD73, in conjunction with CD39, acts as an ectonucleoside that breaks down adenosine triphosphate to generate adenosine, a critical immunosuppressive molecule generated by T_reg cells. Both CD73 and CD39 expression were significantly decreased in the spleens of 28-week-old mice (fig. S2F). To further test whether these changes in key effector molecules had an impact on the fitness of T_reg cells, we analyzed tissues from female mosaic mice that were generated to harbor both wild-type and Aifm2-deficient T_reg cells (32, 34, 35). In this female mosaic model, 50% of T_reg cells express the diphtheria toxin receptor (DTR) driven by Foxp3 (Foxp3^DTR-GFP), and 50% express Foxp3-YFP-Cre combined with Aifm2^+/+ or Aifm2^fl/fl mice, generating a 50:50 mix of Aifm2 wild-type T_reg cells (GFP⁺RFP^neg.) with either Aifm2^+/+ or Aifm2^Δ/Δ T_reg cells (GFP⁺RFP⁺; fig. S3A). We observed no difference in the fitness of Aifm2^+/+ or Aifm2^Δ/Δ T_reg cells compared to Aifm2^WT T_reg cells in lymphoid or nonlymphoid organ tissues at homeostasis (fig. S3B). These results indicate that Aifm2 deficiency in T_reg cells does not markedly affect the frequency, phenotype, or fitness of T_reg cells in lymphoid or nonlymphoid tissues. Thus, Aifm2 deficiency in T_reg cells does not lead to a systemic breakdown in immunotolerance or organ-specific autoimmunity.

FSP1 deficiency specifically in T_reg cells impairs immunosuppression in response to cancer

Since the pharmacological inhibition of FSP1 in cancer cells is moving to clinical trials, we next wanted to assess tumor T_reg cell responses to the inhibition of FSP1. First, we tested whether different subsets of T cells from tumors exhibited different levels of resistance to ferroptosis induction, similar to experiments in Fig. 1G. Here, we grew MC38 colon carcinoma tumors subcutaneously for 14 days in wild-type mice, then harvested tumors, made single-cell suspensions, and treated all cells for 4 hours with or without 1 μM RSL3 to induce ferroptosis. Notably, T_reg cells again exhibited strong resistance to the induction of ferroptosis with GPX4 inhibition as compared to Tconv and CD8⁺ T cells (Fig. 4A). To test whether T_reg cell–specific FSP1 activity promoted tumor T_reg cell resistance to ferroptosis and supported tumor growth, we compared MC38 tumor growth in wild-type versus Treg.Aifm2^Δ/Δ mice (Fig. 4B) (35). Tumor growth was significantly reduced in Treg.Aifm2^Δ/Δ mice (Fig. 4C). While Treg.Aifm2^Δ/Δ mice did not exhibit changes to T_reg cell frequencies or tumor-specific CD8⁺ T cells in tumors (Fig. 4D and fig. S4A), Treg.Aifm2^Δ/Δ mice did exhibit increased CD8⁺ T cell to T_reg cell ratios in tumors, indicative of a selective impact on T_reg cells compared to effector T cells in tumors with Aifm2 deficiency in T_reg cells (Fig. 4D). However, an examination of the viability of Aifm2^Δ/Δ compared to wild-type T_reg cells in tumors only revealed a small increase in the death of Aifm2^Δ/Δ T_reg cells, suggesting that uncontrolled lipid oxidation in T_reg cells may not be inducing more ferroptosis of T_reg cells but also may affect T_reg cell function in tumors (fig. S4, B and C).

Fig. 4. Deletion of Aifm2 in T_reg cells improves cancer immunity.

(A) Schematic for treating single-cell suspensions from MC38 tumors +/− 1 μM RSL3 ex vivo (left). Frequencies of viable CD8⁺ (left), CD4⁺ Tconv (middle), and T_reg cells (right) from tumor single-cell suspensions as a percentage of CD45⁺ cells +/− 1 μM RSL3 for 4 hours. Data from n = 6 mice. (B) Experimental schematic for measuring tumor growth in wild-type versus Aifm2-deficient T_reg cell mice. (C) Growth of MC38 tumors in Treg.Aifm2^+/+ (n = 5) versus Treg.Aifm2^Δ/Δ (n = 6) mice. (D) Frequency of Foxp3⁺ T_reg cells as a percentage of total CD45⁺ immune cells (left) and CD8:T_reg cell ratio (right) in MC38 tumors at day 29. (E) Experimental schematic for measuring tumor growth in wild-type versus Aifm2-deficient T_reg cell mice treated +/− 200 μg anti-CD8b.2 intraperitoneally every 7 days beginning with tumor inoculation. (F) Growth of MC38 tumors in Treg.Aifm2^+/+ + isotype (n = 5), Treg.Aifm2^+/+ + anti-CD8b.2 (n = 5), Treg.Aifm2^Δ/Δ + isotype (n = 5), and Treg.Aifm2^Δ/Δ + anti-CD8b.2 (n = 5). (G) Experimental schematic for measuring tumor growth in wild-type versus Aifm2-deficient T_reg cell mice treated +/− 10 mg/kg liproxstatin (injected intraperitoneally daily). (H) Growth of MC38 tumors in Treg.Aifm2^+/+ + vehicle (n = 3), Treg.Aifm2^+/+ + Lip-1 (n = 4), Treg.Aifm2^Δ/Δ + vehicle (n = 5), and Treg.Aifm2^Δ/Δ + Lip-1 (n = 5). For all plots, *P < 0.05, **P < 0.01, and ***P < 0.001 by Student’s t test [(A) and (D)] or two-way ANOVA [(C), (F), and (H)], mean ± SEM.

In addition, we did not observe altered expression of the exhaustion markers PD-1 or LAG-3 in CD8⁺ T cells in Treg.Aifm2^Δ/Δ mice (fig. S4D). To validate that CD8⁺ T cells were necessary for improving the antitumor response when T_reg cells were deficient for FSP1/Aifm2, we depleted CD8⁺ T cells upon MC38 tumor inoculation and measured tumor growth (Fig. 4E). Depletion of CD8⁺ T cells fully restored tumor growth in Treg.Aifm2^Δ/Δ mice to a degree comparable to tumor growth in Treg.Aifm2^+/+ mice depleted of CD8⁺ T cells (Fig. 4F). To demonstrate that the reduction in tumor growth was due to the accumulation of lipid peroxides in Aifm2-deficient T_reg cells, we treated wild-type and Treg.Aifm2^Δ/Δ tumor–bearing mice with the radical trapping antioxidant liproxstatin-1 (Lip-1), a ferroptosis inhibitor that is functional in vivo in mice (Fig. 4G) (36). Lip-1 treatment restored tumor growth in Treg.Aifm2^Δ/Δ mice, while Lip-1 treatment had no effect on tumor growth in wild-type mice (Fig. 4H and fig. S4, E and F), suggesting that the antioxidant activity of Lip-1 acted on Aifm2-deficient T_reg cells to restore their activity and promote tumor growth. These results indicate that FSP1 activity in T_reg cells is required to control their accumulation of lipid peroxides within the TME, and disruption of FSP1 activity can prevent immunosuppression by T_reg cells, likely in a manner independent of ferroptosis induction in vivo, to promote better tumor control in a CD8⁺ T cell–dependent manner.

FSP1 deficiency in all T cells enhances T cell responses to infection and cancer

T cell deficiency for Gpx4- or Gch-1 was recently shown to negatively affect T cell responses in several disease settings (23, 31, 37). If Aifm2 deficiency similarly disrupts effector T cell responses, then this would diminish its applicability in the setting of cancer, where drugs targeting FSP1 would block its activity in both T_reg cells and effector T cells. To test whether Aifm2 deficiency negatively affects effector T cell responses, we generated Aifm2-deficient T cell mice (CD4-Cre;Aifm2^fl/fl, called T.Aifm2^Δ/Δ) and infected these mice with the bacteria Listeria monocytogenes, an intracellular pathogen that stimulates a robust CD8⁺ T cell response (fig. S5A) (38, 39). T.Aifm2^Δ/Δ mice had no impairment in bacterial control, as there was no difference in bacterial colony-forming units (CFUs) recovered from spleens or livers of wild-type versus T.Aifm2^Δ/Δ mice 6 days postinfection (fig. S5B). Furthermore, the frequency of Listeria-specific CD8⁺ T cells was not reduced, nor were the frequencies of bulk CD8⁺ T or CD4⁺ Tconv cells as a percentage of all CD45⁺ cells, indicative of normal antigen-specific T cell expansion in response to Listeria (fig. S5, C and D). However, the frequency of T_reg cells was decreased in T.Aifm2^Δ/Δ compared to wild-type mice (fig. S5E). These results revealed that only Aifm2^Δ/Δ T_reg cells are impaired during an immune response to bacterial infection, while Aifm2^Δ/Δ CD8⁺ and CD4⁺ effector T cells respond normally.

Next, to test the impact of deleting Aifm2 in all T cells in the setting of cancer, as would happen with a pharmacological FSP1 inhibitor, we implanted MC38 tumors into wild-type versus T.Aifm2^Δ/Δ mice and measured tumor growth (Fig. 5A). Tumor growth was significantly decreased in mice harboring Aifm2-deficient T cells (Fig. 5B). While there was no change in the frequency or viability of T_reg cells in MC38 tumors, there was an increase in the CD8:T_reg cell ratio in tumors from T.Aifm2^Δ/Δ compared to wild-type mice, just as we observed in tumors with T_reg cell–specific Aifm2 deficiency (Fig. 5, C and D, and fig. S6, A and B) (40, 41). Furthermore, analysis of MC38 tumors at day 16 did not show differences in the total number of T_reg cells or CD8⁺ T cells, further suggesting that ferroptotic cell death of T_reg cells does not explain the enhanced tumor control with Aifm2 deficiency in T cells (fig. S6C). Assessment of CD8⁺ T cell exhaustion in tumors by staining for both programmed cell death protein 1 (PD-1) and lymphocyte-activation gene 3 (LAG-3) proteins showed no changes in the bulk CD8⁺ T cell populations (Fig. 5E). However, tumor-specific CD8⁺ T cells, while unchanged in total frequencies, exhibited decreased frequencies of PD-1⁺Lag3⁺ MuLV/H2K^b+ CD8⁺ T cells (Fig. 5F) (42). These results contrast starkly with the consequences of T cell deficiencies in other antioxidant genes, i.e., Gpx4 or Gch1, where deficient T cells completely lose their capacity to clonally expand and function properly to control infections or cancer (23, 31). Instead, deficiency for FSP1/Aifm2 in all T cells completely preserves CD8⁺ and CD4⁺ Tconv cell function while potently disrupting the function of T_reg cells.

Fig. 5. Total T cell deficiency for Aifm2 preserves CD8⁺ T cell function and improves tumor immunity.

(A) Experimental schematic for measuring tumor growth in wild type versus mice with all T cells deficient for Aifm2. (B) Growth of MC38 tumors in wild-type versus Aifm2-deficient T cell mice. Data shown from n = 5 to 6 mice per group from one of two repeated experiments. (C) Frequency of Foxp3⁺ T_reg cells as a percentage of total CD45⁺ immune cells in MC38 tumors at day 29 from (B). (D) CD8:T_reg cell ratio in MC38 tumors at day 29 from (B). (E) Representative flow plot for PD-1 and LAG3 staining in bulk CD8⁺ T cells (left) and quantification of the frequency of PD-1⁺LAG3⁺ bulk CD8⁺ T cells in MC38 tumors (right). (F) Frequency of tumor-specific (MuLV/H2K^b+) CD8⁺ T cells as a percentage of all CD8⁺ T cells (left) and frequency of PD-1⁺Lag3⁺ of MuLV/H2K^b+ CD8⁺ T cells in MC38 tumors (right). For all plots, *P < 0.05, **P < 0.01, and ***P < 0.001 by two-way ANOVA (B) or Student’s t test [(C) to (F)], mean ± SEM. PERCP, peridinin-chlorophyll-protein.

DISCUSSION

In this study, we demonstrate that FSP1 (Aifm2) plays a unique role in sustaining the function of T_reg cells as compared to effector CD8⁺ and CD4⁺ T cells. Unlike other ferroptosis resistance factors such as GPX4, which are essential for effector T cell expansion, differentiation, function, and survival in response to infections or cancer, FSP1 was uniquely required for T_reg cell function. As FSP1 inhibitors are in development to target cancer cells directly to induce ferroptosis (24, 25), our finding of an additional benefit of blocking the activity of FSP1 in the TME in T_reg cells is important and timely.

We show that FSP1 expression is induced with T cell receptor (TCR) activation, distinguishing it from other ferroptosis regulators like GPX4, GCH1, or NRF2 that are constitutively expressed or posttranscriptionally regulated (23, 43–46). The increase in FSP1 protein with TCR engagement may be important for counteracting increased levels of mitochondrial-derived ROS that occur during T cell activation (47, 48) . Proper control of ROS-mediated lipid peroxidation may also be important for tuning TCR signaling in T cells upon activation (23, 30, 46). Our data indicate that FSP1 is critical to support T_reg cell survival in vitro and T_reg cell suppressive function in vivo by limiting the accumulation of lipid peroxides, which would otherwise accumulate in T_reg cells due to their higher levels of ROS generation compared to Tconv cells.

The differences in requirements for the activity of FSP1 in T_reg cells versus effector T cells were particularly prominent in the setting of cancer, where intratumoral T cells are likely receiving chronic TCR stimulation and have heightened sensitivity to increased mitochondrial-derived ROS (49–51). We hypothesize that increased FSP1 expression in T_reg cells evolved to prevent sublethal oxidative damage due to high ROS levels that would otherwise disrupt membrane-bound organelles, such as the mitochondria, which are essential for preserving T_reg cell function (26). This is an important contrast to GPX4-targeted approaches, which are being explored clinically, because GPX4 inhibition has been shown to have a negative impact on all populations of T cells. Thus, GPX4 inhibition carries a narrow therapeutic window due to its essential role in all cell types and the potential for significant toxicity with its inhibition (23, 37, 52, 53). Notably, Aifm2 deficiency in T_reg cells did not lead to any signs of overt autoimmune phenotypes or toxicities, thus widening the potential therapeutic window for pharmacologically targeting FSP1. Targeting FSP1 versus other antioxidant pathways is significant, as balanced regulation of antioxidant pathways in T_reg cells is essential for the maintenance of their suppressive function. For example, T_reg cell–specific deletion of Keap1, the critical upstream regulator of the antioxidant master transcription factor NRF2, leads to excessive activation of antioxidant pathways, which disrupts the maintenance of T_reg cell lineage identity, leading to spontaneous autoimmunity (53). In addition, while we hypothesize that the genetic ablation of FSP1 in all T cells resulted in better tumor control due to a T_reg cell–instrinsic role of FSP1 in supporting T_reg cell immunosuppression, it is also possible that the loss of FSP1 activity in effector T cells could have affected their activity intrinsically in beneficial ways, such as reducing the expression of canonical exhaustion markers (e.g., PD-1 and LAG3) on tumor-specific CD8⁺ T cells (54). Overall, our results support the critical importance of uncovering antioxidant pathways that can be targeted to block intratumoral T_reg cell function but without compromising T_reg cell–mediated peripheral tolerance or diminishing beneficial T cell immunity.

Notably, the induction of Foxp3 in iT_reg cells was insufficient to rescue these cells from undergoing ferroptosis with RSL3-mediated inhibition of GPX4 in vitro. Thus, Foxp3 expression alone does not directly control resistance to ferroptosis in T_reg cells. It is established that thymically derived nT_reg cells and iT_reg cells differ in their epigenetic regulation (54, 55). Most notably, TCR signaling during thymic T_reg cell development is important in setting the epigenetic state of nT_reg cells, which is unique to nT_reg cells compared to peripherally induced T_reg cells (pT_reg/iT_reg cells) (54–56). The heightened TCR signaling during T_reg cell development in the thymus may serve as an important signal to establish increased FSP1 expression in nT_reg cells to help them regulate their inherently increased lipid peroxidation (56–60).

The implications of our findings are important. By specifically targeting FSP1, it may be possible to alleviate the immunosuppressive effects of T_reg cells within the TME, thereby enhancing the efficacy of cancer immunotherapies. FSP1 inhibition, which targets T_reg cells, may also synergize with immune checkpoint blockade (e.g., anti–PD-1 therapies) and cellular immunotherapies such as chimeric antigen receptor T cells by preventing T cell exhaustion and preserving cytotoxic function. PD-1 blockade has also been shown to induce interferon-γ (IFN-γ) production by CD8⁺ T cells, which induced ferroptosis in cancer cells directly by affecting the lipid composition through Acsl4 expression (61). Thus, it may be possible that with FSP1 inhibition, IFN-γ produced by CD8⁺ T cells may also act on T_reg cells and induce their ferroptosis or decrease their immunosuppressive activity. We hypothesize that targeting FSP1 may be a particularly effective approach due to its potential to have dual effects targeting both tumor cells directly through ferroptosis induction while simultaneously reducing intratumoral T_reg cell function.

Our study also underscores the importance of metabolic regulation in T_reg cell biology. T_reg cells exhibit a unique metabolic profile characterized by a reliance on oxidative phosphorylation and fatty acid utilization, which supports their survival in the nutrient-deprived, hypoxic environment of most tumors (15–17). Despite high ROS levels due to these forms of metabolism, T_reg cells maintain low lipid peroxide accumulation, in part by maintaining higher expression of FSP1. Disrupting this balance through FSP1 inhibition leads to functional defects in T_reg cells without appearing to cause their complete elimination in vivo, which may be preferable in preserving immune homeostasis systemically, which again, T_reg cell–specific Aifm2-deficient mice showed no signs of autoimmunity or uncontrolled T cell responses.

In conclusion, our results position FSP1 as a T_reg cell–specific ferroptosis regulator and a promising therapeutic target for enhancing cancer immunity. By blocking FSP1 activity, it may be possible to reduce T_reg cell–mediated immunosuppression in tumors, preserve effector T cell responses, and ultimately improve the outcomes of cancer immunotherapy. Future studies should focus on translating these findings into pharmacological interventions, as well as exploring the potential of FSP1-targeting strategies in combination with existing immunotherapies to achieve synergistic effects.

MATERIALS AND METHODS

Mice

CD4-Cre mice were obtained from the Jackson Laboratory (JAX:022071) and bred in house. Foxp3-GFP-hCre mice were obtained from the Jackson Laboratory (JAX:023161) and bred in house. Aifm2^fl mice were generated as described (33). For tumor studies, syngeneic C57BL/6J mice were inoculated with 5.0 × 10⁵ MC38 in phosphate-buffered saline (PBS) subcutaneously. Tumor measurements were performed blindly across the entire experiment by a single operator measuring three dimensions of the tumor with calipers three times per week. For in vivo liproxstatin treatments (Selleck Chemicals), mice were injected intraperitoneally daily with Lip-1 (10 mg/kg) dissolved in dimethyl sulfoxide and diluted in polyethylene glycol, molecular weight 300, Tween 80, and dH₂O. For CD8⁺ T cell depletion, mice were injected intraperitoneally every 7 days with anti-CD8b.2 (2 mg/ml; Leinco, catalog no. C2832, clone 53-5.8) diluted in PBS at the start of tumor inoculations. All the experiments were conducted according to the Institutional Animal Care and Use Committee guidelines of the University of California, Berkeley under project license AUP-2017-05-9915-2.

Cell lines

MC38 cell lines were provided by J. Bluestone’s laboratory (62). All cell lines were maintained in Dulbecco’s modified Eagle’s medium (DMEM; Gibco) supplemented with 10% fetal bovine serum (FBS), sodium pyruvate (Gibco), 10 mM Hepes (Gibco), and penicillin-streptomycin (Gibco). Tumor cells were grown at 37°C with 5% CO₂.

L. monocytogenes strains

All strains of L. monocytogenes were derived from the wild-type 10403S strain. All strains were cultured in filter-sterilized nutrient-rich Brain Heart Infusion (BHI) media (BD Biosciences) containing streptomycin (200 μg/ml; Sigma-Aldrich).

Intravenous and Intratumoral Listeria infection

Overnight cultures were grown in BHI + streptomycin (200 μg/ml) at 30°C. The following day, bacteria were grown to logarithmic phase by diluting the overnight culture in fresh BHI + streptomycin (200 μg/ml) and culturing at 37°C shaking. Log-phase bacteria were washed and frozen in 9% glycerol/PBS. For intravenous infections, frozen stocks were diluted in PBS to infect via the tail vein with 5 × 10³ CFU log-phase bacteria. The mice were euthanized 6 days after intravenous injections, and organs were collected for CFU or flow analysis.

Primary T_reg cell and T cell culture

Spleens and lymph nodes were collected from Foxp3-GFP-hCre;R26^LSL-RFP; Aifm2^+/+, Foxp3-GFP-hCre;R26^LSL-RFP; Aifm2^Δ/Δ, CD4-Cre;Aifm2^+/+ or CD4-Cre;Aifm2^Δ/Δ. Single-cell suspensions were generated and enriched for CD4⁺ T cells by negative selection using an EasySep magnetic bead kit (STEMCELL Technologies) and stained with anti-mouse CD4 (RM4-5, BioLegend), anti-CD62L (MEL1-14, BioLegend), anti-CD8 (53-6.7, BioLegend), anti-CD25 (PC61, BioLegend), and anti-CD357 (GITR) (DTA-1, BioLegend). Naïve T_reg cells (CD4⁺Foxp3GFP⁺RFP⁺CD62L⁺ or CD4⁺CD25⁺GITR⁺CD62L⁺) and naive CD4⁺ T cells (CD4⁺CD62L⁺Foxp3GFP⁻RFP⁻ or CD4⁺CD62L⁺GITR⁻CD25⁻) were sorted using an Aria Fusion sorter (BD Biosciences) with a 70-μm nozzle. Cells were activated with anti-CD3– and anti-CD28–coated beads (Dynabeads Mouse T-Activator CD3/CD28, Invitrogen) at a ratio of 1:1 for Tconv or 1:3 ratio for T_reg cells (cell:bead). Cells were kept at a concentration of 10⁶ cells/ml in DMEM medium supplemented with 10% FBS, 1% nonessential amino acids, 1 mM sodium pyruvate, 2 mM l-glutamine, 10 μM Hepes, and 55 μM β-mercaptoethanol, and cells cultured more than 24 hours were supplemented with recombinant human interleukin-2 [IL-2; 200 to 2000 IU/ml; National Institutes of Health (NIH)].

RT-qPCR analysis

Naïve T_reg cells (CD4⁺Foxp3GFP⁺RFP⁺CD62L⁺) and naive CD4⁺ T cells (CD4⁺CD62L⁺Foxp3GFP⁻RFP⁻) were sorted as described above and cultured for 4 days in DMEM medium supplemented with 10% FBS, 1% nonessential amino acids, 1 mM sodium pyruvate, 2 mM l-glutamine, 10 μM Hepes, and 55 μM β-mercaptoethanol with IL-2 (2000 U/ml) and CD3/CD28 Dynabeads. Upon collection, total RNA was isolated using the RNeasy Mini kit (QIAGEN). cDNA was subsequently synthesized using the iScript cDNA Synthesis kit (Bio-Rad). Reverse transcription (RT)–qPCR was performed using Fast SYBR Green Master Mix (Applied Biosystems) and a QuantStudio 6 Pro Real-Time PCR machine (Applied Biosystems). Primer sequences used in this study for specific genes are in Table 1. Relative gene expression was determined by the 2-ΔΔCt method.

Table 1. Primer sequences for RT-qPCR analysis of ferroptotic regulatory genes. For, forward primer strand; Rev, reverse primer strand.

Primers	Sequences
Aifm2 For	TTGGTGACTGTGCCGATAC
Aifm2 Rev	GATCTGACCCACGCCATCAT
Nfe2l2 For	CTGAACTCCTGGACGGGACTA
Nfe2l2 Rev	CGGTGGGTCTCCGTAAATGG
Gch1 For	TCCATTTGTAGGAAGGGTCCA
Gch1 Rev	GCAATCTGTTTGGTGAGGCG
Keap1 For	TGCCCCTGTGGTCAAAGTG
Keap1 Rev	GGTTCGGTTACCGTCCTGC
Acsl4 For	CTCACCATTATATTGCTGCCTGT
Acsl4 Rev	TCTCTTTGCCATAGCGTTTTTCT
Acsl3 For	GGGACTACAATACCGGCAGA
Acsl3 Rev	ATAGCCACCTTCCTCCCAGT
Lpcat3 For	GGCCTCTCAATTGCTTATTTCA
Lpcat3 Rev	AGCACGACACATAGCAAGGA

Tissue collection and preparation for flow cytometry

Flow cytometry was performed on an BD LSR Fortessa X20 (BD Biosciences), Cytek Aurora (Cytek Biosciences), or LSRFortessa (BD Biosciences), and datasets were analyzed using FlowJo software (Tree Star). Single-cell suspensions were prepared in ice-cold FACS buffer [PBS with 2 mM EDTA and 1% bovine serum albumin (BSA)] and subjected to red blood cell lysis using ACK buffer [150 mM NH₄Cl, 10 mM KHCO₃, and 0.1 mM Na₂EDTA (pH 7.3)]. Dead cells were stained with a Live/Dead Fixable Blue or Aqua Dead Cell Stain kit (Molecular Probes) in PBS at 4°C. Cell surface antigens were stained at 4°C using a mixture of fluorophore-conjugated antibodies. Surface marker stains for murine samples were carried out with anti-mouse CD3 (17A2, BioLegend), anti-mouse CD4 (RM4-5, BioLegend), anti-mouse CD8a (53-6.7, BioLegend), anti-mouse, CD44 (IM7, BioLegend), anti-mouse CD45 (30-F11, BioLegend), anti-H-2Kb MuLV p15E Tetramer-KSPWFTTL (Medical & Biological Laboratories), and anti-H-2Kb SIINFEKL Tetramer (NIH Tetramer Core) in PBS, 0.5% bovine serum albumin. Cells were fixed using the eBioscience Foxp3/Transcription Factor staining buffer set (eBioscience), before intracellular staining. Intracellular staining was performed using anti-mouse Foxp3 (FJK-16S, eBioscience) at 4°C, according to the manufacturer’s instructions. Cells were resuspended in PBS and filtered through a 70-μm nylon mesh before data acquisition. Datasets were analyzed using FlowJo software (Tree Star).

Statistical methods

P values were obtained from unpaired two-tailed Student’s t tests for all statistical comparisons between two groups, and data were displayed as mean ±SEMs or mean ±SD. For multiple comparisons, one-way analysis of variance (ANOVA) or two-way ANOVA was used. For tumor growth curves, two-way ANOVA was used with Sidak’s multiple comparisons test performed at each time point or by multiple regression analysis. P values are denoted in figures by *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001.

Supplementary Materials

This PDF file includes:

Figs. S1 to S6