Gpx4 regulates iNKT cell homeostasis and function by preventing lipid peroxidation and ferroptosis

Abstract

Invariant natural killer T (iNKT) cells are a group of innate like T cells that plays important roles in immune homeostasis and activation. We found that iNKT cells, compared to CD4⁺ T cells, have significantly higher levels of lipid peroxidation in both mice and humans. Proteomic analysis also demonstrated that iNKT cells express higher levels of phospholipid hydroperoxidase Glutathione peroxidase 4 (Gpx4), a major antioxidant enzyme that reduces lipid peroxidation and prevents ferroptosis. T cell specific deletion of Gpx4 reduces iNKT cell population, most prominently the IFNγ producing NKT1 subset. RNAseq analysis revealed that IFNγ signaling, cell cycle regulation, as well as mitochondrial function are perturbed by Gpx4 deletion in iNKT cells. Consistently, we detected impaired cytokine production, elevated cell proliferation and cell death, and accumulation of lipid peroxides and mitochondrial ROS in Gpx4 KO iNKT cells. Ferroptosis inhibitors, iron chelators, vitamin E and vitamin K2 can prevent ferroptosis induced by Gpx4 deficiency in iNKT cells and ameliorate the impaired function of iNKT cells due to Gpx4 inhibition. Lastly, vitamin E rescues iNKT cell population in Gpx4 KO mice. Altogether, our findings reveal the critical role of Gpx4 in regulating iNKT cell homeostasis and function, through controlling lipid peroxidation and ferroptosis.

Introduction

Invariant natural killer T (iNKT) cells are a unique group of T cells that share properties with innate immune cells. They reside in the tissues ready to respond to stimulation within hours by secreting effector cytokines or expressing cytotoxic molecules, thus playing important roles in the immune system. Distinct from conventional T cells, iNKT cells express semi-invariant T cell receptors, restricted by CD1d, recognize lipid antigens, and gain functional maturity during thymic development independent of inflammation^1–4. It has also been shown that iNKT cells are tumor infiltrating and constitute a promising platform for expressing Chimeric Antigen Receptors (CARs) to treat human malignancy^{5, 6}. Understanding the regulation of iNKT cell development, function, and homeostasis will provide new insights for therapeutics.

iNKT cells are modulated by various cell death regulators. These include Fas-FasL dependent apoptosis⁷, pyroptosis induced through the TNF superfamily receptor OX40⁸, necroptosis regulators Ripk1 and Ripk3^{9, 10}. Further, autophagy is essential for iNKT cell development as we and others have shown^11–14. In addition, iNKT cells in the peripheral organs express the purinergic P2X7 receptor (P2RX7) and the activating ectoenzyme ARTC2.2^15–17. ADP-ribosylation of P2RX7 by ARTC2.2 induces irreversible pore formation and cell death¹⁸. Several studies have shown that this pathway modulates iNKT cell survival in vitro, ex vivo and in vivo^15–17. Furthermore, a recent study showed that iNKT cells are susceptible to oxidative stress and express higher levels of reactive oxygen species (ROS) as compared to conventional CD4⁺ or CD8⁺ T cells¹⁹. Interestingly, deletion of the gene encoding Keap1, a negative regulator of Nrf2 which is a master transcription factor of the antioxidant response, impairs iNKT cell development, indicating ROS might be critical for iNKT cell survival and homeostasis²⁰. However, the specific reactive oxygen species that are elevated in iNKT cells remain unknown.

Excessive lipid hydroperoxides formed by iron dependent lipid peroxidation can trigger a new form of regulated cell death called ferroptosis, which is involved in tissue damage and carcinogenesis²¹. The selenoenzyme Glutathione peroxidase 4 (Gpx4) reduces lipid hydroperoxides into non-toxic lipid alcohols at the expense of reduced glutathione thus critically protecting cells from ferroptosis. The roles of Gpx4 and ferroptosis in regulating the survival and function of CD8⁺ T cells, Treg and Tfh cells have been reported, while the naïve and effector populations of conventional CD4⁺ T cells in unchallenged mice are relatively normal in the absence of Gpx4^22–24. How ferroptosis regulates iNKT cells has not been resolved to date.

Here we report that iNKT cells show elevated lipid peroxidation as compared to CD4⁺ T cells in both mice and humans. Gpx4 protein expression is also significantly higher in iNKT cells. Deletion of Gpx4 impairs iNKT cell homeostasis and function, primarily NKT1, which can be rescued by blocking lipid peroxidation. Our results reveal an essential role for Gpx4-ferroptosis axis in iNKT cells.

Materials and Methods

Mice

All mice experiments were approved by Institutional Animal Care and Use Committee at the Oklahoma Medical Research Foundation. Mice used in this study were 6–12 weeks old. The B6.Cg-Tg(Cd4-cre)1Cwi/BfluJ (Cd4-cre) (022071), C57BL/6J (B6) (000664) and B6.Cg-Tg(LCKprBCL2L1)12Sjk/J (Bcl-XL transgenic (Tg)) (013738) mice were purchased from the Jackson Laboratory. Gpx4-flox/flox (C57BL/6 background) strain was obtained from H. Van. Remmen. T cell-specific deletion of Gpx4 was generated by crossing the Gpx4-flox/flox mice with Cd4-cre. The anti-ARTC2.2 (s+16a) nanobody (50 μg) was injected 30 min prior to organ harvesting throughout this study, unless indicated otherwise.

Human PBMCs

Studies on PBMCs were performed in accordance with the Declaration of Helsinki and approved by the Institutional Review Board at the Oklahoma Medical Research Foundation. Adult volunteers, both males and females, self-enrolled and provided written informed consent. PBMCs were isolated by density gradient centrifugation using Histopaque-1077 (Millipore-Sigma), washed with RPMI-1640 basal media, and processed for flow cytometry as described below.

Antibodies and Reagents

The antibodies conjugated with various fluorophores are from Biolegend: (anti-mouse)- B220 (RA3-6B2), TCRβ (H57-597), CD3ε (145-C11), CD8a (53-6.7), CD4 (GK1.5), ICOS (C398.4A), IFNγ (XMG1.2), IL-4 (11B11), (anti-human)-CD19 (HIB19), CD14 (HCD14), CD3ε (OKT3), CD4 (SK3), CD8a (SK1). The antibodies against transcription factors Tbet (O4-46), PLZF (R17-809), and RORγt (Q31-378) are from BD. The biotinylated anti-mouse TER-119 (TER-119), CD8a (53-6.7), CD19 (6D5), CD62L (MEL-14), CD11b (M1/70), CD11c (N418), F4/80 (BM8), and Ly-6G/Ly-6C (RB6-8C5) antibodies, as well as the anti-ARTC2.2 (s+16a) nanobody are from Biolegend. Human and mouse CD1d/PBS-57 tetramers are from the NIH Tetramer Core Facility.

Cytofix/Cytoperm Fixation/Permeabilization Kit and 7-aminoactinomycin D (7-AAD) are from BD Biosciences. The eBioscience^™ Foxp3 / Transcription Factor Fixation/Permeabilization Concentrate and Diluent as well as the permeabilization buffer, and the live/death fixable blue dead cell stain kit are from ThermoFisher. KRN700/αGC is from Avanti Polar Lipids.

Enrichment and isolation of T cells for proteomic and transcriptomic analysis

iNKT cells were negatively enriched from splenocytes using biotinylated anti- mouse CD8a, CD19, Ter119, CD11c, CD11b, F4/80, Ly-6G/Ly-6C, CD62L antibodies and EasySep^™ Mouse Streptavidin RapidSpheres^™ Isolation Kit (STEMCELL). iNKT cells were sorted as 7-AAD⁻, TCRβ^int, CD1d-tet⁺ cells, NKT1 cells as 7-AAD⁻, TCRβ^int, CD1d-tet⁺, ICOS⁻, CD4^+/− cells and naïve CD4⁺ T were sorted as 7-AAD⁻, CD1d-tet⁻, TCRβ⁺ CD4⁺, CD25⁻, CD62L⁺ cells.

Proteomics

Freshly sorted naïve CD4⁺ T and NKT1 were washed PBS once and lysed with protein lysis buffer (150 mM sodium chloride, 1% NP-40, 0.5% sodium deoxycholate, 0.1% sodium dodecyl sulfate, 50 mM Tris, pH7.4) with 1 × Halt Protease and Phosphatase Inhibitor cocktail (Thermo Scientific). Proteomic profiling of the protein lysates was performed by the IDeA National Resource for Quantitative Proteomics. The heatmap of the differentially expressed proteins of naïve CD4⁺ T and NKT1 was generated using the free web server tool, Heatmapper (http://www.heatmapper.ca/expression). The row Z-score, a scaling method for visualization in heatmap, was shown. The proteomics data is in the process of being deposited at Zenodo with pending accession number.

Flow cytometry

Single cell suspensions were prepared from thymus, spleen, or liver. Before antibody staining, cells were pre-incubated with Fc block (BD Biosciences). Staining for transcription factors was performed using reagents and protocol from the eBioscience^™ Foxp3 TF staining kit (ThermoFisher). Staining for intracellular cytokines was performed using reagents and protocol from Cytofix/Cytoperm Fixation/Permeabilization Kit (BD). Flow cytometry data were acquired on Cytek Aurora (Cytek Biosciences) and analyzed with FlowJo V10.8.2 (TreeStar). The dead cells (7-AAD⁺ or live/death fixable blue dead cell stain⁺) and cell doublets were excluded from the analysis.

Lipid ROS and cytosolic general ROS detection

For lipid ROS detection, the splenocytes or PBMCs were stained with 200 nM C11-BODIPY 581/591 (Invitrogen, USA) in PBS for 30 min at 37°C. The oxidized form was detected using the excitation 488nm (Blue laser) and emission 510 nm (Channel B2 using 5 laser Aurora). For cytosolic ROS detection, the splenocytes were stained with 1 μM CM-H₂DCFDA (Invitrogen) in PBS for 20 min at 37°C.

Mitochondrial mass and mitochondrial ROS detection

To assess the mitochondrial mass, the cells were stained with 20 nM MitoTracker green FM (Invitrogen) in T cell culture medium (below) for 15 min at 37°C. To examine the mitochondrial ROS, the cells were incubated with 5 μM MitoSOX (Invitrogen) for 15 min or 20 nM MitoTracker orange CM-H₂TMRos (Invitrogen) for 20 min in T cell culture medium at 37°C.

Cell culture and in vitro iNKT activation assays

For the in vitro culture, the splenocytes from B6 were treated with or without ferroptosis inhibitors liproxstatin-1 (Lip1) (Tocris) (100 nM, final), deferoxamine (DFO) (Tocris) (400 μM), MK4/Vit K2 (Sigma-Aldrich) (1 μM), α-tocopherol/Vit E (Cayman Chemical) (100 μM), Z-VAD-FMK (RayBiotech) (100 μM), necrostatin-1 (Nec-1) (Tocris) (10 μM), MCC950 (Selleck Chemical) (10 μM), and VX-765 (Sigma-Aldrich) (5 μM). RSL3 (Tocris) was used at 5 μM, and αGC/KRN7000 (Avanti) at 250 ng/mL. For cytokine detection, Brefeldin A (BFA) at (2.5ug/ml) was included. The cells were cultured in T cell culture medium (DMEM medium (Gibco) supplemented with 1% penicillin/streptomycin (Gibco), 1 mM sodium pyruvate (Gibco), 1% non-essential amino acids (Gibco), 10 mM HEPES (Gibco), 2-mercaptoethanol (Gibco), and 10% FBS (R&D) for 4 h at 37°C. The cell viability and lipid ROS were measured by live/dead blue stain and C11-BODIPY, respectively. The cytokine production were measured as mentioned above.

RNA sequencing (RNA-seq)

RNA was isolated from sorted T cells using NucleoSpin RNA XS Micro kit (Macherey-Nagel). The library was prepared using a QIAseq UPXome RNA Lib Kit (QIAGEN). The pooled libraries were sequenced with 150bp paired-end configuration on NovaSeq X Plus (Illumina). The RNA-seq data were analysed using CLC genomic workbench software (QIAGEN) and IPA. The RNA-seq data is in the process of being deposited at the Gene Expression Omnibus (National Center for Biotechnology Information) with pending accession number.

In vivo vitamin E supplementation

For in vivo postnatal treatment, the littermate pups were fed with 50 mg/kg of αToc every other day starting from P6 to P35. For vitamin E supplementary diet, the littermate mice were fed with vitamin E high (556 IU/kg, #5053 customized) (LabDiet) or chow (99 IU/kg, #5053) diet (LabDiet) for 3 consecutive weeks starting from the age of 3 week.

Statistics

All data are presented as mean ± SD and at least three independent experiments were performed. Statistical analysis was performed using GraphPad Prism 9 software. The difference between two groups were analyzed by the unpaired or paired Student’s t test.

Results

Distinct expression of proteins regulating the redox balance and cell death pathways in iNKT cells

To find new regulators for iNKT cells, we compared the proteomes of freshly isolated NKT1 cells (the major subset of iNKT cells in C57BL/6 mice) and naïve CD4⁺ T cells from spleen. Over 4,500 proteins were identified and quantified in an unbiased manner. Differentially expressed (DE) proteins (adj p<0.05) were analyzed by QIAGEN Ingenuity Pathway Analysis (IPA) to determine the significantly affected pathways that are predicted to be activated or inhibited (Fig. 1a). TCR signaling^{25, 26}, Natural Killer Cell signaling²⁷, NF-κB activation²⁸ and Nrf2 mediated responses²⁹ are key regulators for iNKT cell development and function and showed positive Z score in iNKT cells indicating potential activation (Fig. 1a, Supplemental Fig. 1a, b). A recent study showed that compared to CD4⁺ T cells, iNKT cells had reduced elevation in oxygen consumption rate (OCR) after palmitate-BSA treatment and reduced sensitivity to etomoxir, an inhibitor of carnitine palmitoyltransferease-1 (CPT-1)³⁰. Consistent with this, we observed that mitochondrial fatty acid beta-oxidation pathway in iNKT cells compared to naïve CD4⁺ T cells (Fig. 1a) showed negative Z score indicating potential inhibition. Similarly, we detected a potential reduction in PTEN signaling which parallels transcriptomic changes in iNKT cells compared to CD4⁺ or CD8⁺ thymocytes³¹ (Fig. 1a, Supplemental Fig. 1c).

Figure 1. iNKT cells have distinct redox regulation and higher lipid peroxidation.

(a) Ingenuity Pathway Analysis (IPA) of the canonicals pathways enriched in the differentially expressed proteins between NKT1 and naïve CD4⁺ T cells. (b) Heatmap of the differentially expressed proteins associated with oxidative stress and ferroptosis pathways (columns represent biological replicates of each cell type). (c) Histogram overlays and mean fluorescence intensities (MFI) of oxidized C11-BODIPY staining signal reporting lipid ROS in CD4⁺ T and iNKT from the spleens of C57BL/6J mice or the PBMC from healthy donors, considering the difference in cell size as detailed in Supplemental Fig. 2. The data on mice shown in c is representative of four independent experiments, while human PBMC data depict representative histograms and individual responses from 18 donors. **p < 0.01, ****p < 0.0001, paired Student’s t test.

Consistent with a critical role for redox regulation in iNKT cells, two important superoxide dismutases, Sod1 and Sod2, were expressed at higher levels in iNKT cells as compared to CD4⁺ T (Fig. 1b). In contrast, catalase (Cat), the main enzyme breaking down hydrogen peroxides, was downregulated in iNKT cells. Thioredoxin (Trx) and glutathione (GSH) systems are the major thiol dependent antioxidant mechanisms in cells³². Thioredoxin reductase 1 (TxnRd1) and N-ribosyldihydronicotinamide: quinone dehydrogenase 2 (Nqo2), a member of the thioredoxin family of enzymes, were also expressed at higher levels in iNKT cells (Fig. 1b). Many glutathione related enzymes similarly had higher expression in iNKT cells, including glutamate-cysteine ligase catalytic subunit (Gclc), glutaredoxin (Glrx), glutathione-disulfide reductase (Gsr), glutathione S-transferase omega 1 and pi1 (Gsto1, Gstp1), while some had lower expression in iNKT cells, including glutamate-cysteine ligase modifier subunit (Gclm) and glutathione S transferase kappa 1 (Gstk1).

Interestingly, immunogenic cell death signaling pathway was predicted to be activated in iNKT cells (Fig. 1a). Proteins involved in apoptosis (Annexin A1/Anxa1, Rock1, Bax, Casp3, Pycard; Supplemental Fig. 1d) and autophagy (Atg5, Becn1; Supplemental Fig. 1d) were mostly expressed at higher levels in iNKT cells. Although these two programmed cell deaths (PCD) pathways are generally considered non-immunogenic, they share complex crosstalk with multiple immunogenic PCD pathways such as pyroptosis (Casp-1, P2rx7, Gasdermin D/Gsdmd; Supplemental Fig. 1d) and necroptosis (Ripk3; Supplemental Fig. 1d)³³. Heat shock proteins (HSPs; Hsp90b1, Hspa5, Hspa1a, Hspa1b, Hspa1l, Hspa9; Supplemental Fig. 1d), molecular chaperones important to maintain cellular homeostasis, senescence and PCD^{34, 35}, were generally upregulated in iNKT cells compared to naïve CD4⁺ T cells.

iNKT cells show higher lipid peroxidation.

Ferroptosis, a type of programmed cell death dependent on iron and lipid peroxides²¹, was predicted to be upregulated in iNKT cells based on proteome (Fig. 1a). Apart from higher expression of Gpx4 (Fig. 1b), proteins involved in iron metabolism (Ferritin heavy chain/Fth1, Cysteine desulfurase/Nfs1; Fig. 1b) and fatty acid synthesis (Acetyl-CoA carboxylase α/Acaca; Fig. 1b) were also differentially expressed in iNKT cells as compared to naïve CD4⁺ T cells, suggesting ferroptosis regulation is critical for iNKT cells. Furthermore, using the lipid peroxidation sensor BODIPY581/591 undecanoic acid (C11-BODIPY) we found high accumulation of lipid ROS in iNKT cells compared to CD4⁺ T cells. Similar trends were observed with lymphocytes from either mouse spleen or human peripheral blood mononuclear cells (PBMCs) (Fig. 1c, gating strategy shown in Supplemental Fig. 2a), after accounting for comparable sizes between cells (Supplemental Fig. 2b). Our data indicate that dysregulation of lipid peroxidation and ferroptosis could have detrimental effects on iNKT cells.

Gpx4 is required for peripheral iNKT cell homeostasis.

It has been shown that Gpx4 is important for the homeostasis of CD8⁺ T cells, but not CD4⁺ T cells in vivo²², although Treg suppressive function²³ and follicular helper T cell differentiation after immunization²⁴ are impaired in Gpx4 knockout cells. Since iNKT cells express higher levels of Gpx4 than naïve CD4⁺ T cells (Fig. 1b), we hypothesized that Gpx4 is important for iNKT cell development and/or maintenance in vivo. We therefore established a T cell specific Gpx4 knockout strain by crossing Gpx4-flox/flox with CD4cre (Gpx4-cKO hereafter). Consistent with previous reports that thymocytes were intact in the absence of Gpx4, iNKT cells developed normally (Supplemental Fig. 2c). Interestingly, iNKT cell populations in the spleen and liver were significantly impaired in Gpx4-cKO as compared to WT littermate controls (Supplemental Fig. 2c). In line with this, we found that lipid ROS level was higher in splenic iNKT cells than thymic iNKT cells (Supplemental Fig. 2d).

Studies showed that the extracellular NAD⁺ and ATP released during organ processing can trigger the activation of P2X7R through ADP ribosylation by ecto-ADP-ribosyltransferase, ARTC2.2, and cell death^{15, 17, 36}. Blocking ARTC2.2 with nanobodies improved the recovery and IFNγ production of iNKT cells ex vivo¹⁵. To rule out the possibility that the defect of Gpx4-cKO iNKT cells was due to experimental artefacts related to ARTC2.2/P2RX7 mediated cell death, we injected anti-ARTC2.2 nanobodies prior to organ harvesting and confirmed that Gpx4 deficiency impairs the homeostasis of iNKT in spleen and liver, but not in thymus and lymph nodes (Fig. 2a and data not shown). All results shown hereafter were generated with anti-ARTC2.2 nanobodies injection prior to organ harvesting to ensure quality and rigour.

Figure 2. Gpx4 is required for iNKT homeostasis.

(a) Left panel: representative flow cytometry plots of iNKT cells from thymus, spleen, and liver of WT and littermate Gpx4-cKO mice. Percentages show iNKT cells among live, CD19⁻CD8a⁻ splenocytes; Right panel: The percentage among total live cells and absolute cell numbers of iNKT cells. (b) Left panel: representative flow cytometry plots of NKT1 (Tbet⁺ RORγt⁻), NKT2 (Tbet⁻ RORγt⁻), and NKT17 (Tbet⁻ RORγt⁺) cells among total iNKT cells in WT and littermate Gpx4-cKO mice. Right panel: The percentages among total iNKT cells and absolute cell numbers of all subsets. (c) Total iNKT cell percentages among total live cells and cell numbers in Bcl-XL-Tg⁺; Gpx4-flox/flox; CD4cre⁻ and littermate Bcl-XL-Tg⁺; Gpx4-flox/flox; CD4cre⁺ mice. (d) iNKT cell subsets among total iNKT cells and subset cell number in Bcl-XL-Tg⁺; Gpx4-flox/flox; CD4cre⁻, and littermate Bcl-XL-Tg⁺; Gpx4-flox/flox; CD4cre⁺ mice. Live/death blue stain was used for cell viability. The data shown in a and b are from one representative experiment out of three independent experiments, and the data shown in c and d are pooled from three independent experiments. ns: non-significant, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001, unpaired Student’s t test.

iNKT cells differentiate into functional subsets NKT1, NKT2 and NKT17 at steady state defined by the expression of transcription factors Tbet, PLZF and RORγt, respectively. The splenic iNKT cell subsets in the Gpx4-cKO mice were skewed towards NKT2 and NKT17, while NKT1 cells were impaired (Fig. 2b), suggesting iNKT cell subsets have differential dependence on Gpx4. BCL-XL transgene did not rescue the defect of iNKT cell differentiation in Gpx4-cKO mice (Fig. 2c, d), suggesting non-apoptotic cell death is responsible for the observed phenotype.

Iron and lipid peroxides promote ferroptosis in Gpx4 deficient iNKT cells.

In line with the reduced presence in vivo, Gpx4-cKO iNKT cells showed significantly lower survival compared to littermate WT iNKT cells after 4 h culture ex vivo (Fig. 3a). Inhibitors of apoptosis (z-VAD-fmk), necroptosis (Necrostatin-1/Nec-1), NLRP3 inflammasome (MCC950), and pyroptosis (VX765) had minimal effect on viability deficit (Fig. 3b). On the contrary, ferroptosis inhibitor (Liproxstatin-1) and the extracellular iron chelator deferoxamine (DFO) profoundly improved cell survival of Gpx4-cKO iNKT cells, suggesting ferroptosis is responsible for the impaired homeostatic survival of Gpx4-cKO iNKT cells in vivo.

Figure 3. Lipid peroxidation promotes ferroptosis in Gpx4 deficient iNKT cells.

(a) Cell viability of WT and Gpx4-cKO iNKT cells from littermate mice measured immediately ex vivo (0 h) or after 4 h culture at 37°C. (b) Normalized cell viability of WT and Gpx4-cKO iNKT cells from littermate mice cultured for 4 h in the presence of various compounds. (c) Cytosolic ROS level assessed by CM-H₂DCFA and (d) lipid ROS by C11-BODIPY in iNKT cells. (c, d) Intensities normalized to MFI signals in WT cells. (e) WT iNKT cells cultured for 4 h as in (b) with or without RSL3 (10 μM), by itself or in combination with other compounds as indicated, before lipid ROS was assessed by C11-BODIPY staining. (f) Normalized cell viability of WT and Gpx4-cKO iNKT cells from littermate mice cultured for 4 h with or without MK-4 or αToc. For cell viability, live/death blue was used. The data shown is from one representative experiment out of three independent experiments. ns: non-significant, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001, unpaired Student’s t test.

Interestingly, both cytosolic ROS (indicated by CM-H2DCFDA) and lipid peroxides (C11-BODIPY) showed significant increases in Gpx4-cKO iNKT cells ex vivo (Fig. 3c). Furthermore, after normalization to ROS signals in WT cells run in parallel, more drastic increases of ROS were observed in Gpx4 deficient NKT1 cells, as compared to NKT2 and NKT17 (Fig. 3c, d), indicating that Gpx4 plays a critical role in NKT1 cells to control oxidative stress. In the absence of Gpx4, increased lipid ROS can trigger ferroptosis (Fig. 3b) which could account for the significantly reduced NKT1 cell numbers in vivo (Fig. 2b). We therefore treated WT iNKT cells with (1S,3R)-RSL3 (RSL3), a potent inhibitor of Gpx4, to examine the immediate effect of Gpx4 inhibition. RSL3 potently increased lipid peroxides (Fig. 3e) in iNKT cells, assessed by C11-BODIPY, which could be prevented by ferroptosis inhibitor/Liproxstatin-1, DFO and a lipid soluble antioxidant vitamin E (α-tocopherol, αToc), but not other PCD inhibitors (Fig. 3e). A recent study showed that Menatetrenone (MK-4), a vitamin K2 derivative, efficiently suppressed lipid peroxidation dependent on AIFM2/FSP1^{37, 38}. We similarly found that MK-4 reduced lipid peroxide accumulation in RSL3 treated iNKT cells (Fig. 3e). αToc and MK-4 also significantly improved the survival of Gpx4-cKO iNKT cells (Fig. 3f). Altogether, these data suggest that lipid peroxides and iron promote ferroptosis of Gpx4-cKO iNKT cells.

Gpx4 deficient iNKT cells are functional impaired.

We next stimulated iNKT cells with α-Galactosyl Ceramide (αGC), the agonist antigen for iNKT cells. Cytokine production induced by αGC, however, was impaired in live Gpx4-cKO iNKT cells (Fig. 4a). To rule out the possibility that reduced NKT1 subset in Gpx4-cKO is responsible for the impaired IFNγ production, we stained for intracellular cytokines together with transcription factor Tbet. Comparing Tbet⁺ (NKT1) subset in WT and Gpx4-cKO iNKT cells, αGC induced cytokine was reduced in Gpx4-cKO cells (Supplemental Fig. 3a), suggesting that Gpx4 is required for iNKT cell function on a per cell basis. RSL-3 treatment in WT iNKT cells for 4 h minimally affected cell survival (Fig. 4b) and led to functional impairment as seen in Gpx4-cKO cells (Fig. 4c). This defect was more pronounced for IFNγ⁺IL4⁺ (Fig. 4c) and IFNγ⁺ (Supplemental Fig. 3b) population, and less for IL4⁺ population (Supplemental Fig. 3b). Inhibitors for iron dependent lipid peroxidation and ferroptosis significantly enhanced cytokine production from RSL-3 treated cells (Fig. 4c, Supplemental Fig. 3b). Altogether, these data indicate Gpx4 deficiency leads to dysregulated lipid peroxidation that impairs TCR induced cytokine production from iNKT cells.

Figure 4. iNKT cell function is regulated by Gpx4.

(a) Splenocytes from WT and littermate Gpx4-cKO mice were cultured in the presence of αGC (250 ng/ml) and BFA (2.5 μg/ml) for 4 h at 37°C and intracellular cytokines in live iNKT cells assessed by flow cytometry. (b, c) WT splenocytes were cultured for 4 h with αGC (250 ng/ml) and BFA (2.5 ug/ml) in the absence or presence of RSL3 (10 μM), by itself or in combination with other compounds as indicated. Cell viability (b) and relative cytokine productions (c) from total live INKT cells relative to αGC stimulated control were shown. For cell viability, live/death blue was used. The data shown in a are pooled from four independent experiments, and the data shown in b and c are from one representative experiment out of three independent experiments. ns: non-significant, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001, unpaired Student’s t test.

Gpx4 deficiency impairs cell proliferation and mitochondrial fitness in iNKT cells.

To identify potential mechanisms, we carried out RNAseq to compare the transcriptomes of Gpx4-cKO and WT iNKT cells from littermate control mice. At steady state, Gpx4 deficient iNKT cells showed 86 differentially expressed genes (DEGs; absolute fold change [FC]>2 and false discovery rate [FDR]<0.1). Consistent with higher cell death (Fig. 3a) and higher ROS (Fig. 3c, d) in Gpx4-cKO iNKT cells, Ingenuity Pathway Analysis (IPA) showed that Ferroptosis Signaling Pathway, Immunogenic Cell Death Signaling Pathway, and Oxidative Stress Induced Senescence were significantly enriched as potential upregulated canonical pathways (Fig. 5a). Interferon gamma signaling was among the potential downregulated pathways (Fig. 5a), reflective of the impaired TCR activation induced cytokine production in these cells (Fig. 4). Interestingly, Cell Cycle Checkpoints was prominently enriched as a potential upregulated pathway, suggesting cell cycle progression could be affected by Gpx4 deficiency. Using Ki-67 expression as readout, we observed organotypic changes in iNKT cell proliferation. While Gpx4 deficiency did not change Ki-67 expression in thymic iNKT cells, splenic NKT1 subset and liver iNKT cells showed elevated Ki-67 level (Fig. 5b), suggesting that Gpx4 regulates iNKT cell homeostasis through modulating cell proliferation. The impaired cell proliferation was not influenced by BCL-XL (Fig. 5c), indicating apoptosis is not involved.

Figure 5. Gpx4 deficiency leads to abnormal proliferation and accumulation of mitochondrial ROS in iNKT.

(a) Ingenuity Pathway Analysis (IPA) of canonical pathways enriched in the differentially expressed genes comparing Gpx4-cKO to WT iNKT cells. (b) Proliferation of iNKT subsets from thymus, spleen, and liver was assessed by Ki-67 expression. Measurement of mitochondrial mass using MitoTracker Green (MTG) (c), mitochondrial ROS using MitoSOX (d) and MitoTracker orange CM-H₂TMRos (MTO) (e) in WT and Gpx4-cKO iNKT cells from littermate mice, as well as normalization by MFI (MTG) were shown (d, e). Live/death blue was used for viability. The data shown in b and c are pooled from four and three independent experiments, respectively. The data shown in d, e and f are from one representative experiment out of three independent experiments. ns: non-significant, *p < 0.05, ***p < 0.001, ****p < 0.0001, unpaired Student’s t test.

Gpx4 deficiency has been shown to elevate mitochondrial ROS and reduce expression of mitochondrial genes in Treg cells²³. Although Mitochondrial Dysfunction did not show significant enrichment in Gpx4-cKO iNKT cells in our transcriptomic analysis, Oxidative Phosphorylation was among the potential downregulated canonical pathways predicted by IPA in these cells (Fig. 5a). Therefore, we examined mitochondria fitness and oxidative stress associated with Gpx4-deficiency. Overall, we detected comparable mitochondrial mass between WT and Gpx4-cKO iNKT cells (Fig. 5d), as assessed using MitoTracker Green (MTG). In contrast, loss of Gpx4 led to significantly higher levels of mitochondrial superoxide in iNKT cells, measured by MitoSox staining, which was retained even after normalization to mitochondrial mass/MTG (Fig. 5e). To confirm our findings, we next used MitoTracker Orange CM-H2TMRos, which is a reduced, nonfluorescent rosamine-based MitoTracker dye that only fluoresces upon oxidation³⁹. Like MitoSox, Gpx4-cKO iNKT cells showed higher staining with MitoTracker Orange (MTO) with or without normalization to mitochondrial mass/MTG (Fig. 5f). These data suggest that homeostatic iNKT cells require Gpx4 to maintain low levels of lipid peroxides important for mitochondrial fitness.

Vitamin E restores iNKT cell homeostasis in Gpx4-cKO mice.

Given that lipid soluble antioxidant vitamin E (αToc) rescued the survival of Gpx4-cKO iNKT cells in vitro (Fig. 3f), we examined the effects of αToc in vivo. iNKT cells are detected in vivo early postnatally. We administered αToc (50mg/kg diluted in corn oil) or corn oil (equivalent volume) via oral gavage in WT and Gpx4-cKO mice every other day from postnatal day 6 to day 35 (Fig. 6a). αToc treatment restored iNKT cell percentage as well as cell number in Gpx4-cKO mice to the WT level (Fig. 6b). Furthermore, the defects in iNKT cell subset distribution as well as the cell proliferation were also fully rescued by αToc (Fig. 6c, d). These data suggest that Gpx4 controls lipid ROS level to prevent ferroptosis in iNKT cells in vivo, and lipid antioxidants, such as vitamin E, can potentially offset Gpx4 deficiency.

Figure 6. Vitamin E rescues iNKT homeostasis in Gpx4-cKO mice.

(a) Experimental scheme to study the effects of αToc/Vit E supplementation from perinatal day 6 in WT and littermate Gpx4-cKO mice. (b) Left panel: Representative flow cytometry plots of splenic iNKT cells from WT and littermate Gpx4-cKO mice, treated with vehicle (Corn oil) or αToc (50 mg/kg). Right panel: The percentages among total live cells and absolute cell numbers of iNKT. (c) Left panel: Representative flow cytometry plots of NKT1, NKT2, and NKT17 cells among total iNKT cells in WT and littermate Gpx4-cKO mice. Right panel: The absolute cell numbers of iNKT subsets. (d) Proliferation of splenic iNKT subsets in different experimental groups as assessed by Ki-67 expression. Live/death blue stain was used for viability. The data shown is each from one representative experiment out of three independent experiments. ns: non-significant, **p < 0.01, ***p < 0.001, ****p < 0.0001, unpaired Student’s t test.

To test if Gpx4 is required beyond perinatal developmental stage, we next fed Gpx4-cKO mice and WT littermate controls with high vitamin E diet (556 IU/KG) or control chow (99IU/KG) for 3 weeks after weaning at day 21. Consistent with αToc treatment, the VitE high diet fully rescued iNKT cell percentage, cell number and subset distribution in Gpx4-cKO mice to WT level (Supplemental Fig. 4a, b). Our data suggests that lipid ROS are generated in the peripheral iNKT cells, and Gpx4 is required constitutively to prevent the accumulation of lipid ROS that may lead to ferroptosis in these cells.

Discussion

iNKT cells undergo unique developmental and functional maturation as compared to conventional T cells before they become memory like T cells and form the first line of defense in the peripheral tissue together with other innate immune cells. Metabolic processes have emerged as key regulators for iNKT cells. Upregulation of Nrf2 mediated transcriptional programs in Keap1 knockout mice specifically impaired iNKT cell development suggesting that redox balance is critical for these cells²⁰. Our proteomic analysis also uncovered NRF2-mediated pathway and Oxidative Stress Induced Senescence as positively enriched in iNKT cells compared to naïve CD4⁺ T cells (Fig. 1a). Cellular ROS in T cells can serve important functions when induced at physiological level⁴⁰, while release of high concentration of ROS is a known cytotoxic mechanism employed by innate immune cells during infection⁴¹. It has been shown that iNKT cells exhibited high levels of ROS in spleen and liver, which was proposed to be mainly produced by NADPH oxidases¹⁹. Mitochondria are a critical source of cellular ROS and we have shown that iNKT cells have a higher spare respiratory capacity (SRC) than conventional CD4⁺ or CD8⁺ T cells in the thymus⁴². Deficiency of Rieske iron-sulfur protein (RISP), an essential subunit of the mitochondrial complex III drastically impaired iNKT cells development, but not conventional T cells³⁰. Deficiency of Opa1, a dynamin related GTPase that controls mitochondrial membrane fusion, cristae formation, and energetics, also resulted in severely diminished iNKT cells⁴³. Inhibition of mitochondrial oxidative phosphorylation by oligomycin A induced cell death and functional defects in splenic iNKT cells⁴⁴, and treatment with a mitochondria targeted antioxidant SkQ1 reduced NKT cell percentage⁴⁵. We have detected higher protein level of mitochondria specific superoxide dismutase SOD2 in iNKT cells (Fig. 1b). Whether mitochondria derived ROS contributes to the overall essential role mitochondria play during iNKT cell development, homeostasis and function remains to be defined.

The ferroptosis pathway is enriched in splenic iNKT cells based on proteomic analysis (Fig. 1a). Ferroptosis is a type of programmed cell death dependent on iron and accumulation of lipid peroxides²¹. In a recent study, thymic iNKT cells showed a higher labile iron pool (LIP) than other thymocytes based on calcein-AM dye staining⁴⁶. In the absence of the E3 ubiquitin ligase cullin 3 (Cul3), thymic iNKT cells displayed increased LIP and impaired development, however, lipid peroxidation was unchanged, suggesting that iron metabolism rather than ferroptosis promoted the iNKT cell deficiency during thymic development in Cul3 KO mice⁴⁶. In the present study, knocking out Gpx4, the primary protective enzyme against ferroptosis, did not affect thymic iNKT cells (Fig. 2). We found lipid peroxides were lower in thymic iNKT cells than in splenic iNKT cells (Supplemental Fig. 2d). This suggests that lipid peroxidation and ferroptosis are not constitutively ongoing in these cells. In contrast, iNKT cells from mouse peripheral organs such as the spleen display higher lipid peroxidation than CD4 T cells, as reported by C11-Bodipy (Fig. 1c), as well as higher Gpx4 protein (Fig. 1b). This indicates that elevated lipid ROS are actively maintained at homeostasis, and Gpx4 and other redox regulators are required to control their cytotoxic effects in iNKT cells in these organs. Indeed, Gpx4 deficiency resulted in drastic increases in ROS in splenic iNKT cells as measured by CM-H2DCFDA and C11-Bodipy (Fig. 3c, d), cell death (Fig. 3a) and impaired iNKT cell responses to TCR stimulation (Fig. 4), which can only be rescued by inhibitors for ferroptosis (Lip1), iron chelators (DFO), vitamin K2 (MK-4) and vitamin E (αToc), but not by inhibitors for other forms of PCD (Figs. 3, 4). These data indicate that Gpx4 is essential for iNKT cell homeostasis through reducing lipid ROS and preventing ferroptosis, resembling its role in CD8⁺ T at steady state and follicular T cells induced after vaccination^{22, 24}. It is reported that Gpx4 is dispensable for CD4⁺ T cells at homeostasis^{22, 24}, although it is necessary for CD4⁺ T cell survival and sterile expansion in the mixed bone marrow chimera environment²².

Interestingly, Gpx4 deletion specifically reduced NKT1 cells but not NKT2 and NKT17 cells in the spleen (Fig. 2). Likewise, the increase of cellular ROS in the absence of Gpx4 is more drastic in NKT1 cells (Fig. 3c, d), and dysregulated cell proliferation was observed in this subset (Fig. 5b). The function of NKT1 was also impaired by Gpx4 deficiency (Supplemental Fig. 3a). Given that lipid ROS level is comparable among the subsets in WT, these data suggest that Gpx4 plays a nonredundant role in NKT1 cells at homeostasis to control the level of lipid ROS, while in NKT2 and NKT17 cells, either the production of lipid peroxides is lower or alternative antioxidant mechanisms act in parallel with Gpx4. Recently, Fsp1/AIFM2 was shown to generate lipophilic antioxidants by reducing coenzyme Q10 instead of glutathione to suppress the production of lipid peroxides^{37, 47}, and could provide an alternative to Gpx4. CD8⁺ T cells also express Tbet and are more dependent on Gpx4 than CD4⁺ T cells for survival²². Interestingly, Gata3 and RORγt expression, but not Tbet expression in Gpx4-cKO T cells were elevated compared to WT control²⁴, suggesting similar differential requirements of Gpx4 might be among Th1, Th2 and Th17 cells. Whether this analog persists among functional subsets of other innate like T cells such as MAIT cells or γδT cells, and the corresponding mechanisms remain to be discovered. Further, we focused our analysis on lymphoid organs and liver. It is important for future studies to determine if the differential requirements for Gpx4 in iNKT cell subsets persists in other organs such as the lung and intestine.

Like Treg cells²³, iNKT cells accumulated mitochondrial ROS in the absence of Gpx4 while maintaining the same mitochondrial mass (Fig. 3d–f). Three isoforms of GPX4 localize in the cytoplasm, mitochondria, or nucleus^48–50. Gpx4 gene deletion potentially leads to the loss of all three, and the mitochondrial isoform could be responsible for maintaining mitochondrial fitness. To establish the contribution of subcellular ROS to iNKT cell ferroptosis, future studies need to examine whether any GPX4 isoform can rectify iNKT cells deficient in Gpx4. GPX4 is a selenoprotein that requires selenium for its enzymatic function. Selenium supplementation was shown to upregulate GPX4, increased the number of follicular helper T cells, boosted humoral responses in a mouse model, and enhanced human Tfh function and vaccine response²⁴. A similar strategy could aid iNKT cells as well, and we show here that, alternatively, antioxidant vitamins can counteract the ROS imbalance due to Gpx4 deficiency. Given the promising potential of using iNKT cells as CAR carrier for cancer immunotherapy, it will be interesting to determine if GPX4 activity is fine tuned to maintain the optimal level of lipid peroxides (Fig. 1c) for iNKT cell maintenance and function.

Summary.

• iNKT cells show higher levels of lipid peroxidation than CD4⁺ T cells.

• Gpx4 is required for iNKT cell homeostasis and function.

• Vitamin E, a lipid soluble antioxidant, can rescue iNKT cells in Gpx4 knockout mice.