Eed-flox mice were backcrossed to C57BL/6 for more than 7 generations (24). CD4-Cre transgenic mice were crossed with Eed-flox mice to generate Eed conditional knockout (cKO) mice (25). C57BL/6 (Ly5.2) mice were obtained from CLEA Japan and C57BL/6 background Ly5.1 mice were obtained from the Jackson Laboratory. Mice were bred and maintained under the specific pathogen-free conditions.

Single-cell suspensions from the spleen and bone marrow were resuspended in Gey’s or ACK solutions for red blood cell lysis. Cells were treated with Fc Block (2.4G2, BD Biosciences) or TruStain FcX (93, BioLegend) followed by staining with fluorochrome-conjugated antibodies or biotinylated antibodies. Anti-CD24 (M1/69), CD44 (IM7), TCRβ (H57-597), NK1.1 (PK136), CD4 (RM4-5), CD8 (53-6.7), PLZF (R17-809), RORγt (Q31-378), Ly49A (JR9-318), Ly49C and Ly49I (5E6), Ly49G2 (4D11), IL-4 (11B11), IL-17A (TC11-18H10), IFN-γ (XMG1.2) were purchased from BD Biosciences; Anti-T-bet (4B10) was purchased from BioLegend. The cells stained with biotinylated antibodies were detected by fluorochrome-conjugated streptavidin. Dead Cells were excluded using propidium iodide (Sigma-Aldrich). To detect iNKT cells, the cells were labeled with α-GalCer (Funakoshi)-loaded CD1d-tetramer PE (MBL) for 30 min at room temperature. For intracellular staining, the cells were fixed and permeabilized using a Foxp3 staining kit (Thermo Fisher Scientific). For cytokine staining, the cells were treated with 50 ng/ml of phorbol 12-myristate 13-acetate (PMA) and 1 μg/ml of ionomycin with BD GolgiStop containing monensin for 6 hours (h) followed by staining using Cytofix/Cytoperm Plus Fixation/Permeabilization Kit (BD Biosciences). For H3K27me3 and Eed staining, anti-trimethyl-Histone H3 (Lys27) polyclonal antibody (07-449, Upstate), anti-Eed (E4L5E) XP rabbit monoclonal antibody (85322, CST), and normal rabbit IgG (2729; CST) as isotype control were labeled using Zenon Alexa Fluor 647 rabbit IgG Labeling Kit (Z25308; Invitrogen) followed by intracellular staining procedure. Stained cells were analyzed using FACSCanto II (BD Biosciences) or CytoFLEX S (Beckman Coulter). Cell sorting was performed using FACSAria II (BD Biosciences). Data were analyzed on FlowJo software (BD Biosciences).

To isolate lymphocytes from the liver and lung, mice were anesthetized and perfused with PBS. For liver lymphocytes, the liver cells were filtered through 100-μm nylon mesh, washed with 2% FBS/PBS, and resuspended in 33.75% Percoll PLUS (GE Healthcare Life Sciences) followed by centrifugation at 2,300 rpm for 25 min at room temperature. For lung lymphocytes, the lung was cut into small pieces and digested with 1 mg/ml collagenase D (Roche) and 0.5 mg/ml DNase I (Roche) for 1 h at 37°C with shaking. The cells were filtered through 70-μm nylon mesh and then treated with Gey’s solution to remove red blood cells.

Fetal thymus organ culture (FTOC) was performed as described previously (26). For timed pregnancies, the day of the vaginal plug was designated E0.5. The thymic lobes were isolated from control or Eed cKO E15.5 fetuses, which were then transferred to standard FTOC conditions and analyzed for iNKT cell development on the indicated day.

Bone marrow (BM) cells (2x10⁶) obtained from control or Eed cKO mice (Ly5.2) were mixed with C57BL/6 (Ly5.1/5.2) BM cells at a 1:1 ratio and transferred into 9-Gy irradiated C57BL/6 (Ly5.1) recipient mice. Recipient mice were analyzed after 4 months (Mon).

APAP (00204-82, Nacalai Tesque) was dissolved at 15 mg/ml in PBS at 37°C. Seven- to eleven-week-old control and Eed cKO mice were fasted for 16 h before APAP treatment. For liver histology, female mice were administrated with 360 mg/kg APAP and sacrificed mice at 48 or 72 h after treatment. Hematoxylin and eosin (H&E) staining was performed with frozen liver tissues, firstly stained with Mayer’s Hematoxylin Solution (FUJIFILM) and 1% Eosin Y Solution (FUJIFILM), then followed by ethanol and xylene dehydration steps. The images were obtained from BZ-X800 microscope (Keyence). The percentage of necrosis area was measured by Keyence analysis software (Keyence). For the inflammation score (27), severity was categorized into four levels: 0, no inflammation; 1, moderate and inflammatory cells were scattered; 2, marked and inflammatory cells formed foci; 3, severe and cells were inflammatory diffuse. For iNKT transfer experiments, liver iNKT cells sorted from wild-type Ly5.1 mice were injected intravenously into control or Eed cKO mice at 3-5x10⁵ cells per mouse, then fasted mice were treated with APAP after 3 days. To analyze the survival, male mice were injected 375 mg/kg APAP by intraperitoneal and checked mice survival every 12 h.

CD4⁺CD8⁺TCRβ⁺CD69⁺ thymocytes (1x10⁵) were sorted from 3 pairs of each control or Eed cKO mice, then kept in RNA later solution (Thermo Fisher Scientific). Total RNA was isolated by using QIAzol and miRNeasy Micro kit (QIAGEN). cDNA synthesis (1x105) was used SMART-Seq v4 Ultra Low Input RNA Kit for Sequencing Kit (Takara). Libraries were prepared by Nextera XT v2 Kit (Illumina) and sequenced on an Illumina HiSeq 2500. Quality control and adapter trimming for raw reads were performed by fastp (v0.21.0). Trimmed read mapping and quantification were performed with RSEM (v1.3.1) using STAR (v2.7.10) as an aligner. Differential expressions were identified by using DEseq2 (v1.34.0). RNA-seq data were used for GSEA analysis with gene sets obtained from publicly available databases (28, 29).

H3K27me3 ChIP-seq dataset was obtained from GSE84238 (30). ChIP-seq raw reads were mapped onto reference mouse genome mm39 by using HISAT2 and peaks were called using MACS2. SparK and IGV were used to visualize bedGraph coverage tracks.

DP and iNKT cells sorted from control and Eed cKO mice were fixed with 1% formaldehyde for 10 min at room temperature and sonicated with Bioruptor, and then sonicated chromatins were immunoprecipitated by control rabbit IgG antibody (2729, CST) or anti-trimethylated H3K27 antibody (07-449, Upstate) coupled with SureBeads Protein A/G Magnet Beads (Bio-Rad) overnight. After washing and reverse crosslinking, DNA was purified using Phenol: Chloroform: Isoamyl alcohol (Nacalai Tesque). qPCR was performed using Thunderbird SYBR qPCR mix (TOYOBO) or TB Green Premix Ex Taq II (Takara). Primers are listed in Supplementary Table 1 .

All data analyses were performed with Prism 7 software (GraphPad). The results were presented as mean ± SD. p values were calculated by unpaired t-test, multiple t-tests with the Holm-Sidak method at α = 0.05, or one-way ANOVA. Statistical significance was determined with alpha<0.05 and presented as *, <0.05; **, <0.01; ***, <0.001; ****, <0.0001.

We examined H3K27me3 levels during iNKT cell development. H3K27me3 signals were detected in developing thymic iNKT cells from DP to the S3 stage, indicating that H3K27me3 histone modification occurs during the early development of iNKT cells. Notably, H3K27me3 levels significantly increased during the transition from DP/S0 to the S2 stage, suggesting that additional activation of the PRC2 complex leads to de novo deposition of H3K27me3 with stage progression ( Figure 1A ). We compared the mRNA expression of PRC2 core subunits among the T-cell subsets. Similar to the H3K27me3 levels, Eed mRNA expression levels were increased in thymic and splenic iNKT cells compared to thymic DP cells, whereas Ezh2 and Suz12 mRNAs were equal or downregulated ( Figure 1B ). These results prompted us to investigate the specific function of Eed in iNKT cell development.

We bred Eed-flox mice with CD4-Cre transgenic mice expressing Cre recombinase after the DP stage in the T-cell lineage (named Eed cKO mice). We first examined the deletion of Eed in different T-cell populations in Eed cKO mice. The expression of Eed protein remained intact in double-negative (DN) cells but was partially abrogated in DP cells and S0 iNKT cells, with nearly complete loss of Eed in CD4 or CD8 single-positive (SP) cells, S1/3 iNKT cells, and splenic T cells ( Figure 1C ; Supplementary Figure 1A ). Additionally, Eed mRNA was severely decreased in the S1 to S3 stages, but only partially decreased in the S0 stages of iNKT cells from Eed cKO mice ( Supplementary Figure 1B ). These results confirmed Eed protein expression in conventional T cells and iNKT cells, consistent with the mRNA expression profile, indicating that the deletion of Eed using CD4-Cre resulted in a delay beyond the thymic DP cell stage, similar to previous findings (31).

In Eed cKO mice, we observed a significant reduction in iNKT cells characterized by TCRβ⁺CD1d-tetramer (Tet)⁺ cells in various tissues, including the thymus, spleen, liver, and lung ( Figure 1D, E ). To further examine the early developmental stages of impaired thymic iNKT cells in Eed cKO mice, we utilized the surface markers CD24, CD44, and NK1.1, to distinguish the developmental stages of iNKT cells. Compared to control mice, Eed cKO mice showed a significant increase in the percentage of iNKT cells from the S0 to S2 stage, followed by a substantial decline at the S3 stage ( Figures 1F, G ). Furthermore, the absolute number of S3 iNKT cells in Eed cKO mice was severely reduced compared to that of control mice, indicating a critical role of Eed in the S2 to S3 transition during iNKT cell development, and the increased proportions of S0, S1, and S2 iNKT cells were mainly due to the reduction of the S3 iNKT cells. Conversely, the numbers of DN, DP, CD4 SP, CD8 SP, and Foxp3⁺ Tregs T-cell subsets were not affected by Eed deficiency in thymus ( Supplementary Figure 2 ). These results suggest that Eed plays an indispensable and specific role in the development of iNKT cells.

The differentiation of iNKT cells in mice begins after birth (32). To investigate potential age-related differences in iNKT cell development, we examined iNKT cell development in neonatal mice. TCRβ⁺CD1d-Tet⁺ iNKT cells were first detected at postnatal day 4 (P4), which was confirmed by the detection of NK1.1⁺ iNKT cells in P4 control mice ( Figure 2A ; Supplementary Figure 3 ). While the total numbers of thymocytes increased with postnatal days in both control and Eed cKO mice, a marked reduction in the iNKT cell fraction was observed in the Eed cKO thymus at P8 ( Figure 2B ). In control mice, the majority of iNKT cells were S2 cells; however, in Eed cKO mice, most iNKT cells were identified as S0 cells, with few S1, S2, and S3 cells ( Figures 2C, D ). We further utilized fetal thymic organ cultures to analyze iNKT cell development from embryonic progenitors. In FTOC, progenitors from the E15.5 fetal thymic lobe revealed a significant reduction in Eed-deficient iNKT cell numbers after day 11 of culture compared to that of controls, primarily due to a severe reduction in S2/3 cells, consistent with the neonatal results ( Figures 2E, F ).

CD1d and SLAM family receptors play critical roles in initial iNKT cell development (11, 33). To assess whether Eed affects the expression of these molecules, we measured the expression of CD1d, CD150 (Slamf1), and Ly108 (Slamf6) in DP thymocytes, and found no obvious differences between control and Eed cKO mice ( Supplementary Figure 4A ). In mice, over 80% of CD1d-restricted TCRαβ⁺ iNKT cells express Vα14-Jα18 chain combined with Vβ8, Vβ7, or Vβ2 chain (7). Therefore, we determined TCR Vβ usage on thymic iNKT cells in Eed cKO mice, finding no abnormalities in the distribution of TCRVβ2, Vβ5, Vβ7, Vβ8.1/8.2, and Vβ8.3 ( Supplementary Figures 4B, C ). Additionally, mRNA expression level of the Vα14-Jα18 chain was similar between control and Eed cKO DP cells ( Supplementary Figure 4D ). Thus, the abnormal development of iNKT cells in Eed cKO mice was not due to dysregulated expression of CD1d, SLAM family, Vα14-Jα18 chain, or Vβ usage on DP thymocytes.

Considering that the cellular interaction between DP thymocytes in the thymic microenvironment is crucial for the initial development of iNKT cells (4, 34), we aimed to discriminate the environmental influences caused by Eed deletion in other cell types from cell-intrinsic defects in iNKT cells. Therefore, we isolated BM cells from Ly5.2 control and Eed cKO mice and transferred them together with Ly5.1/5.2 wild-type BM cells into congenic Ly5.1 mice. The results showed a significant reduction in the percentage of iNKT cells developed from Eed cKO BM cells compared to that from control BM cells ( Figures 2G, H ). In summary, Eed is critically required for iNKT progenitors to differentiate into mature iNKT cells in a cell-intrinsic manner, particularly at the transition from the S0/S1 to S2 stage in embryos/neonates and from S2 to S3 in adults.

Shortly after the initial PLZF expression, iNKT cells transition to a progenitor state known as NKTp, which further differentiates into three mature NKT subsets exhibiting distinct effector functions and cytokine secretion (16). We examined the effect on the terminal maturation of iNKT subsets in Eed cKO mice based on the cytokine profiles of thymic iNKT cells, observing a significant reduction in both the percentage and cell number of IFN-γ-producing NKT1 cells and IL-17-producing NKT17 cells in Eed cKO mice compared to that of control mice, while those of IL-4-producing NKT2 cells were comparable with control mice ( Figures 3A, B ).

Different NKT subsets are characterized by specific transcription factors, including NKT1 (RORγt^loPLZF^loT-bet⁺), NKT2 (RORγt^loPLZF^hi), and NKT17 (RORγt^hiPLZF^int) (16). Further analysis based on transcription factor expression revealed a significant reduction in NKT1 and NKT17 cells in Eed cKO mice, whereas NKT2 cells remained comparable to those in the control mice ( Figures 3C, D ). Moreover, distinct iNKT subsets defined by cell-surface markers demonstrated that CD122⁺ NKT1 and CD138⁺ NKT17 cells were significantly decreased in Eed cKO mice, whereas the absolute number of CCR7^-PD-1⁺ NKT2 cells in Eed cKO mice was comparable to that in control mice ( Supplementary Figure 5 ). Since we detected Eed expression in NKT1, NKT2, and NKT17 cells, a different requirement of Eed in NKT subsets is not due to different expression levels of Eed ( Supplementary Figure 6 ). Overall, our findings suggest that Eed is indispensable for the development of both NKT1 and NKT17 cells, but not NKT2 cells.

To understand how Eed deficiency inhibits iNKT cell development at the transcriptional level, we performed RNA-sequencing (RNA-seq) analysis on CD4⁺CD8⁺TCRβ⁺CD69⁺ thymocytes from control and Eed cKO mice. We identified 83 upregulated and 125 downregulated genes (fold change >2 or <0.5, p <0.05) in Eed cKO mice compared to the control ( Figure 4A ). These differentially expressed genes (DEGs) were compared with previously reported iNKT cell stage-specific datasets (29), revealing that 11 of the 125 downregulated genes in Eed cKO thymocytes. St8sia6, S100a6, Csf1, Eed, Gpr68, Il18rap, Cxcr3, and Arsb were downregulated during the DP to S1 transition, whereas Il18rap, Cxcr3, Arsb, Klra9, Klrb1c, and Samd3 were downregulated during the S1 to S2 transition in an Eed-dependent manner, suggesting the importance of Eed in the developmental transition of iNKT cells ( Figure 4B ). Furthermore, S1 to S2, S2 to S3, and NKT1 transition datasets positively correlated with DEGs in control to Eed cKO DP cells ( Figure 4C ).

We then analyzed genes associated with natural cell function. Among the NK receptors (NKRs) normally expressed at the S2/S3 stage, the expression of Klra7, Klra9, Klrb1c, Klrd1, Klre1, Klri2, and Klrk1 were significantly downregulated in Eed cKO cells ( Figure 4D ). We confirmed reduced protein levels of NK1.1 (Klrb1c), Ly49C/Ly49I (Klra3/Klra7), and Ly49G2 (Klra9) in S2/3 (CD44⁺CD24^-) iNKT cells in Eed cKO mice ( Figure 4E ). These results indicate that Eed participates in the transcriptional regulation of various genes related to the effector function of iNKT cells.

We observed upregulation of Zbtb16, encoding the zinc-finger transcription factor PLZF, in Eed cKO DP cells ( Figure 4A ), whose expression is normally upregulated in S1/2 and then downregulated in S3 during iNKT cell differentiation (29). This suggests that the failure of iNKT cell development in Eed cKO mice may be associated with abnormal expression of PLZF due to a lack of H3K27me3-mediated inactivation of Zbtb16. Consequently, we first assessed the impact of Eed deletion on H3K27me3 levels in iNKT cells. Intracellular staining of H3K27me3 in thymocytes revealed a significant reduction of H3K27me3 levels in Eed cKO iNKT cells compared to that of controls at the S1, S2, and S3 stages by 3.0-, 11.1-, and 10.9-fold, respectively ( Figure 4F ). To further understand the relationship between H3K27me3 status and Zbtb16 transcription, we compared the peaks from RNA-seq and H3K27me3 Chromatin immunoprecipitation sequencing (ChIP-seq) obtained from iNKT cells around the Zbtb16 locus (30). H3K27me3 signals were detected at the transcriptional start and end regions of Zbtb16, corresponding to elevated RNA-seq peaks in Eed cKO cells ( Figure 4G ; Supplementary Table 2 ). Additionally, we detected H3K27me3 signals at the proximal promoter and 3′ untranslated region (3′ UTR) of Zbtb16 in iNKT cells, along with a known Eed target Hoxa11 ( Figure 4H ) (35). Notably, H3K27me3 enrichment was higher at the proximal promoter region of Zbtb16 in the DP stage, followed by a decline in the S1/S2 stage and subsequent recovery in the S3 stage, inversely correlating with the PLZF expression pattern ( Figure 4I ). As expected, PLZF protein expression significantly increased in Eed cKO iNKT cells at the S2 stage ( Figures 4J, K ). Thus, Eed likely negatively regulates PLZF expression through H3K27me3 at the proximal promoter and 3′ UTR region of Zbtb16 during iNKT cell development.

The proliferation and survival of iNKT cells significantly influence their development and population size (7). We evaluated these aspects in Eed cKO iNKT cells by assessing BrdU incorporation for proliferation and Annexin V staining for survival. While BrdU incorporation rates in Eed cKO iNKT cells were similar to or higher than those in control mice, a significant increase in the percentage of Annexin V⁺ iNKT cells was observed at the S2 stage in Eed cKO mice ( Figures 5A-D ). This indicates that impaired cell survival, particularly at the S2 developmental stage, contributes to developmental defects in Eed-deficient iNKT cells.

Tumor suppressor Cdkn2a, a target of PRC2 (36), encodes two proteins (p16Ink4a and p19Arf) that promote cell cycle arrest and apoptosis by regulating RB and p53 proteins, respectively (37). Additionally, Cdkn1a, encoding a p21 CDK inhibitor, is involved in the p53-RB signaling pathway that regulates the cell cycle (38). Both Cdkn2a and Cdkn1a are implicated in apoptosis induction in hematopoietic stem cells following Eed deficiency (36). We hypothesized that the abnormal expression of Cdkn2a and Cdkn1a contributes to the impaired survival and development of Eed-deficient iNKT cells. H3K27me3 ChIP-seq analysis using iNKT cells revealed H3K27me3 signals around the transcriptional promoter and 3′ UTR region of Cdkn2a and Cdkn1a in iNKT cells ( Figure 5E ). Moreover, proximal promoter regions of Cdkn2a and Cdkn1a were significantly enriched with H3K27me3 in control S1/2 cells in an Eed-dependent manner ( Figure 5F ). Furthermore, mRNA expression of p16Ink4a, p19Arf, and p21 was significantly increased in S2 cells from Eed cKO mice compared to that in the control ( Figure 5G ). These results strongly support that Cdkn2a and Cdkn1a are direct targets of Eed for H3K27me3, regulating cell death and promoting iNKT cell development.

As Eed deficiency leads to a reduction in thymic NKT1 cells, we examined NKT1 cells in peripheral tissues and observed a substantial proportion residing in the liver. However, the numbers of NKT1 cells in the spleen, liver, and lung of Eed cKO mice were significantly diminished ( Figure 6A ). CD1d- and Jα18-deficient mice lacking iNKT cells are highly susceptible to acetaminophen (APAP)-induced liver injury (AILI) (39, 40), a severe consequence of sudden hepatocyte injury, often induced by APAP overdose in developed countries (41), but the critical iNKT cell subset involved and the role of epigenetic modification remain unknown. To investigate the physiological significance of Eed-dependent iNKT cell development, we used an AILI mouse model. Compared with control mice, Eed cKO mice exhibited a lower survival rate after APAP challenge, with mortality being dose-dependent ( Supplementary Figure 7A ). Histological analysis of the livers collected 48h or 72h post-APAP treatment revealed significantly enlarged necrotic areas accompanied by severe inflammation in the liver of Eed cKO mice compared to controls ( Figures 6B, C ). In APAP-treated mice, iNKT cells were significantly reduced in Eed cKO mice, whereas the numbers of CD4⁺ and CD8⁺ T cells were comparable to those in control mice ( Supplementary Figures 7B, C ). To determine whether the defect induced in APAP-treated Eed cKO mice resulted from a lack of iNKT cells, we transferred iNKT cells from the wild-type liver (mostly NKT1 cells) into Eed cKO mice. The results showed that iNKT cell transfer significantly increased survival and reduced necrosis in Eed cKO mice 24 h after APAP treatment ( Figures 6D–F ). Collectively, Eed-dependent liver resident NKT cells, particularly NKT1 cells, play a critical role in protecting hepatocytes from APAP-mediated necrosis and liver inflammation.