Liver X receptors regulate natural killer T cell population and antitumor activity in the liver of mice

The nuclear receptors liver X receptor α (LXRα) and LXRβ are lipid sensors that regulate lipid metabolism and immunity. Natural killer T (NKT) cells, a T cell subset expressing surface markers of both natural killer cells and T lymphocytes and involved in antitumor immunity, are another abundant immune cell type in the liver. The potential function of the metabolic regulators LXRα/β in hepatic NKT cells remains unknown. In this study, we examined the role of LXRα and LXRβ in NKT cells using mice deficient for LXRα and/or LXRβ, and found that hepatic invariant NKT (iNKT) cells are drastically decreased in LXRα/β-KO mice. Cytokine production stimulated by the iNKT cell activator α-galactosylceramide was impaired in LXRα/β-KO hepatic mononuclear cells and in LXRα/β-KO mice. iNKT cell-mediated antitumor effect was also disturbed in LXRα/β-KO mice. LXRα/β-KO mice transplanted with wild-type bone marrow showed decreased iNKT cells in the liver and spleen. The thymus of LXRα/β-KO mice showed a decreased population of iNKT cells. In conclusion, LXRα and LXRβ are essential for NKT cell-mediated immunity, such as cytokine production and hepatic antitumor activity, and are involved in NKT cell development in immune tissues, such as the thymus.

The liver is home to a variety of immune cells, including resident Kupffer cells, bone marrow-derived macrophages (BMDMs), natural killer (NK) cells, natural killer T (NKT) cells, T lymphocytes and B lymphocytes 1 . An understanding of the roles of hepatic immune cells is important to develop therapeutic and preventative strategies for liver diseases, such as nonalcoholic steatohepatitis and hepatocellular carcinoma 2 . NKT cells are a subset of T lymphocytes expressing surface markers common to NK cells and T lymphocytes, and recognize lipid antigens, such as α-galactosylceramide (α-GalCer), which are presented with the major histocompatibility complex class I-like molecule CD1d on antigen-presenting cells 3 . Invariant NKT (iNKT) cells, which express a semi-invariant T cell receptor (TCR) containing Vα14-Jα18 and Vβ11 in human and Vα14-Jα18 and Vβ8.2, Vβ7 or Vβ2 in mice, are enriched in mouse liver (20-30% of hepatic lymphocytes) and possess a strong activity for the production of the Th1 cytokine interferon-γ (IFN-γ) and the Th2 cytokine interleukin-4 (IL-4) 4 . Hepatic iNKT cells are involved in inflammation, antitumor immunity, repair and regeneration in mice [5][6][7][8] . Impaired iNKT cell function is associated with several human cancers, including hepatocellular carcinoma 9 , and NKT cell therapy is a promising approach for antitumor immunotherapy, such as chimeric antigen receptor NKT cell therapy against neuroblastoma 10,11 . Although iNKT cells play an important role in hepatic immunity, the function of metabolism regulators in iNKT cells remains largely unknown.
The nuclear receptors liver X receptor α (LXRα) and LXRβ are transcription factors that are activated by oxysterols and regulate expression of genes involved in lipid metabolism 12 . LXRα is abundantly expressed in liver, intestine, adipose tissue, kidney and macrophages, while LXRβ is ubiquitously present in the body. LXRs also regulate immune responses in immune cells, such as macrophages, through transrepression of other transcription factors, including nuclear factor κB and activation factor-1, and induction of genes involved in lipid metabolism,

Results
Decreased population of iNKT cells in the liver and spleen of LXRα/β-KO mice. We previously reported that the total number of MNCs and a population of F4/80 lo CD11b + BMDMs are increased in the liver of LXRα/β-KO mice 17 . We examined the effect of deletion of LXRα and LXRβ on the population of other types of hepatic immune cells. The number of hepatic NKT cells (βTCR + NK1.1 + ) was decreased in LXRα-KO mice and LXRβ-KO mice compared to wild-type (WT) mice, and further lowered in LXRα/β-KO mice (Fig. 1a, b). In contrast, the number of NK cells (βTCR − NK1.1 + ) was not affected by deletion of LXRα or LXRβ. Interestingly, iNKT cells (βTCR + CD1d-tetramer + ) were drastically decreased in the liver of LXRα/β-KO mice, while the cell number was decreased to half and one-third of WT mice in LXRα-KO mice and LXRβ-KO mice, respectively (Fig. 1c, d). The percentage of NKT cells positive for CD69, an activation marker in NK cells and NKT cells 19 , was decreased in the liver of LXRα/β-KO mice, while CD69 + cells were not altered in NK cells in all groups of mice ( Fig. 1e, f). Since NKT cells are matured in thymus and peripheral organs such as liver 3,20 , we examined the surface expression of CD44 and NK1.1 in βTCR + CD1d-tetramer + cells from the liver. iNKT cells in stage 1 (CD44 − NK1.1 − ) and stage 2 (CD44 + NK1.1 − ) were comparable between WT and LXRα/β-KO mice, whereas stage 3 (CD44 + NK1.1 + ) cells were significantly decreased in LXRα/β-KO mice. (Fig. 1g).Thus, NKT cells, particularly iNKT cells, are decreased in the liver of LXRα-KO mice and LXRβ-KO mice, with more severe deficiency in the liver of LXRα/β-KO mice.
Next, we examined NKT cell abundance in the spleen of LXR-deficient mice. Similar to the findings in the liver, splenic NKT cells were decreased in LXRα/β-KO mice, but not in LXRα-KO mice or LXRβ-KO mice (Supplementary Fig. 1a, b). iNKT cells were significantly decreased in the spleen of LXRα/β-KO mice ( Supplementary  Fig. 1c, d). The number of NK cells was not significantly altered in LXRα-KO, LXRβ-KO or LXRα/β-KO mice compared to WT mice ( Supplementary Fig. 1b). There was no difference in the percentage of CD69 + NKT cells or NK cells in all groups of mice ( Supplementary Fig. 1e, f). iNKT cells in stage 1 were increased and those in stage 3 were decreased in LXRα/β-KO mice compared to WT mice ( Supplementary Fig. 1g). Therefore, LXRα/β deletion decreases NKT cells, particularly iNKT cells, in the liver and spleen.
LXRα/β-KO hepatic iNKT cells are incompetent in cytokine production. To examine the effect of LXR deletion on NKT cell cytokine production in vitro, we isolated hepatic MNCs from WT and LXRα/β-KO mice, cultured with the iNKT cell-specific activator α-GalCer, and analyzed for expression of cytokine genes. In WT cells, α-GalCer treatment effectively induced mRNA expression of Il4 (the gene encoding IL-4), Ifng (the gene encoding IFN-γ) and Tnf (the gene encoding tumor necrosis factor-α (TNF-α)) ( Fig. 2a). In contrast, expression of these genes was not altered after α-GalCer treatment in LXRα/β-KO cells, and Ifng mRNA levels in untreated LXRα/β-KO cells were higher than in WT cells. Expression of Nr1h3 (the gene encoding LXRα) was unaffected and that of Nr1h2 (the gene encoding LXRβ) was slightly increased by α-GalCer treatment in WT cells. Interestingly, secreted IFN-γ protein levels were much higher in untreated LXRα/β-KO hepatic MNCs than in WT cells (Fig. 2b), consistent with Ifng mRNA expression (Fig. 2a). α-GalCer stimulation increased protein levels of IL-4, IFN-γ and TNF-α in conditioned media of WT cells but not in LXRα/β-KO cells (Fig. 2b). Impaired Il4 and Tnf mRNA induction was also observed in splenocytes from LXRα/β-KO mice, while WT cells increased expression of these genes efficiently after α-GalCer treatment (Fig. 2c). Ifng mRNA expression was not changed both in WT and LXRα/β-KO splenocytes at 6 h after α-GalCer stimulation. These findings indicate that iNKT cell responses are impaired in hepatic MNCs and splenocytes derived from LXRα/β-KO mice.
We next examined iNKT cell responses in vivo by injecting α-GalCer intravenously to WT, LXRα-KO, LXRβ-KO and LXRα/β-KO mice and measuring plasma cytokine levels. Injection of α-GalCer to WT-mice increased plasma IL-4 and IFN-γ, reaching peak levels at 3 and 24 h, respectively (Fig. 3a), as reported previously 18 . Plasma IL-4 levels at the 3-h time point were lower in LXRβ-KO mice and comparable to control levels in LXRα/β-KO mice (Fig. 3a). Plasma IFN-γ levels at the 24-h time point were also lower in LXRα/β-KO mice and LXRβ-KO mice, and did not increase in LXRα/β-KO mice. Consistent with in vitro studies (Fig. 2a, b), the basal level of plasma IFN-γ at the 0-h time point in LXRα/β-KO mice was higher than that in WT mice. Con-A, a lectin derived from jack beans, also induces IL-4 production by iNKT cells via a mechanism distinct from α-GalCer 21,22 . Intravenous Con-A treatment effectively increased plasma IL-4 levels in WT mice, although to a lesser extent than α-GalCer treatment (Fig. 3b), as reported previously 23 . The induction of IL-4 in LXRα/β-KO mice was much weaker than that in WT mice (Fig. 3b). These findings indicate that cytokine production by iNKT cell stimulation is impaired in LXRα/β-KO mice.   in the liver, are involved in antitumor immunity 6 . EL-4 cells are mouse T cell lymphoma-derived cells which express CD1d and when injected into mice are directly killed by a mechanism requiring NKT cells 6 . Survival of WT mice and LXRα/β-KO mice was assessed after EL-4 cell injection. Mice were injected with EL-4 cells intravenously and then treated with α-GalCer for iNKT cell activation. Five of 12 WT mice (33%) without α-GalCer treatment survived at 70 days after EL-4 cell injection (Fig. 4a). iNKT cell stimulation by α-GalCer increased the survival rate of mice to 93% (12 of 13 mice), consistent with our previous report 7 . On the other hand, only 2 of 10 LXRα/β-KO mice (20%) without α-GalCer treatment survived, and α-GalCer treatment did not increase the survival rate significantly (4 of 10 mice (40%); P > 0.05). Next, we analyzed metastatic tumors in the liver of α-GalCer-untreated mice 28 days after injection of EL-4 cells. All mice were alive at the 28-day time point (Fig. 4a). We observed several hepatic tumors in both WT and LXRα/β-KO mice (Fig. 4b).
Although there was no statistically significant difference in survival between WT and LXRα/β-KO mice (33% vs 20%) in the absence of α-GalCer treatment when monitored for 70 days (Fig. 4a), the tumor number in the liver of LXRα/β-KO mice was much higher than that in WT mice at 28-day time point (Fig. 4b). Liver histology showed more tumor cell infiltrate around the portal vein in the liver of LXRα/β-KO mice compared to WT mice (Fig. 4c). Liver weight, plasma aspartate aminotransferase and alanine aminotransferase levels were increased in EL-4 cell-injected LXRα/β-KO mice (Fig. 4d), indicating that liver metastasis is associated with hepatocyte damage. Next, we examined NKT cell-mediated antitumor effect in co-culture experiments. We attempted to isolate hepatic iNKT cells from WT and LXRα/β-KO mice but could not obtain enough number of cells, because iNKT cells were severely decreased in LXRα/β-KO mice (Fig. 1d). Then, we isolated hepatic MNCs containing iNKT cells from WT and LXRα/β-KO mice, cultured these cells with calcein-labeled tumor cells, and evaluated tumor cell killing by measuring fluorescence intensity in cultured media. We compared the antitumor effect against NK cell-resistant EL-4 cells and NK cell-sensitive Yac-1 cells 24,25 . The antitumor effect of hepatic MNCs against EL-4 cells was lower in LXRα/β-KO cells when compared to WT cells (Fig. 4e), consistent with the in vivo results (Fig. 4a-c) Antitumor effect against Yac-1 cells was not changed or slightly increased in LXRα/β-KO cells. These findings indicate that NKT cell-mediated antitumor effect is decreased in hepatic MNCs from LXRα/β-KO mice.

Expression of the MHC class I-like molecule CD1d and the chemokine C-X-C motif ligand (CXCL) 16 is not decreased in the liver of LXRα/β-KO mice. The MHC class I-like molecule CD1d
presents a glycolipid antigen to the invariant TCR of iNKT cells and is essential for iNKT cell development and activation 4 . CD1d expression was unaffected in total hepatic MNCs, neutrophils (F4/80 − CD11b + ) or BMDMs (F4/80 lo CD11b + ) when comparing WT and LXRα/β-KO mice, while it was slightly decreased in resident Kupffer cells (F4/80 hi CD11b − ) from LXRα/β-KO mice (Fig. 5a). Splenic total MNCs, neutrophils and BMDMs had similar CD1d expression in WT and LXRα/β-KO mice. While splenocytes from LXRα/β-KO mice had slightly lower Cd1d1 expression compared to WT splenocytes, Cd1d1 mRNA levels did not differ in hepatic MNCs or whole liver between WT and LXRα/β-KO mice (Fig. 5b), suggesting that Cd1d1 expression is not decreased in a major population of hepatic MNCs and parenchymal cells, 70-80% of which are hepatocytes. We isolated F4/80 + Kupffer cells/macrophages, CD146 + liver sinusoidal endothelial cells and CD11c + dendritic cells from hepatic MNCs. Cd1d1 mRNA expression tended to decrease in F4/80 + cells and CD146 + cells and to increase in CD11c + cells in the liver of LXRα/β-KO mice, although these differences were not statistically significant (Fig. 5c). The decrease of iNKT cells in LXRα/β-KO mice does not appear to be due to deficient CD1d presentation by other immune cells such as dendritic cells. Since CXCL16 in liver sinusoidal epithelial cells and Kupffer cells/macrophages is necessary for the accumulation of iNKT cells, which express the C-X-C motif chemokine receptor 6, and NKT-mediated antitumor immunity in the liver [26][27][28] . Cxcl16 mRNA expression was not decreased in hepatic MNCs from LXRα/β-KO mice (Fig. 5d), and also not in F4/80 + Kupffer cells/macrophages, CD146 + liver sinusoidal endothelial cells or CD11c + dendritic cells from the liver of LXRα/β-KO mice (Fig. 5e). Thus, decreased abundance of iNKT cells in LXRα/β-KO mice are not associated with CXCL16 expression.

Decreased population of iNKT cells in the thymus of LXRα/β-KO mice.
We further investigated the effect of LXR deficiency on iNKT cell population in young mice. The number of iNKT cells increases in the liver during development from youth to middle age in mice 29 . We analyzed a population of iNKT cells in the liver and spleen in 4-week-old young mice, and found that iNKT cells were decreased in the liver of LXRα/β-KO mice compared to WT mice (Fig. 6a). A similar tendency was observed in the spleen although it was not statistically significant. Next, we performed bone marrow transplantation experiments utilizing bone marrow cells derived from CD45.1 WT mice to irradiated CD45.2 WT or LXRα/β-KO recipient mice, and found that iNKT cell development from WT bone marrow cells was impaired in both liver and spleen of LXRα/β-KO mice (Fig. 6b). The populations of NK cells and conventional T cells were not affected LXR deficiency (Supplementary Fig. 2). Thus, iNKT cell development may be impaired in the absence of LXRα/β function at a stage following bone marrow exit.
Lymphoid precursors leave the bone marrow and develop into iNKT cells in the thymus 3 . We examined the effect of LXR deficiency on thymic iNKT cell development. The number of iNKT cells and level of Cd1d1 mRNA expression were decreased in the thymus of LXRα/β-KO mice (Fig. 7a, b). The transcription factor promyelocytic leukemia zinc finger (PLZF) is essential for the development of NKT cell lineage at the early developmental stage 30 . The expression of Zbtb16 (the gene encoding PLZF) was not altered in LXRα/β-KO mice (Fig. 7b). In thymocytes, Nr1h2 (the gene encoding LXRβ) was highly expressed compared to Nr1h3 (the gene encoding LXRα) (Fig. 7c). mRNA levels of Nr1h3 and Nr1h2 in sorted thymic iNKT cells were similar to those in thymic epithelial cells (TECs), CD4 + T cells and CD8 + T cells in the thymus (Fig. 7d)   www.nature.com/scientificreports/ Compared to WT mice, stage 1, stage 2 and stage 3 cells were significantly decreased in LXRα/β-KO mice (Fig. 7e). Stage 0 cells tended to be lower in LXRα/β-KO mice, but this difference was not statistically significant. Therefore, iNKT cell development is impaired in the thymus of LXRα/β mice.

Discussion
In this study, we report that LXRα and LXRβ are essential for iNKT cell-mediated immunity in mice. While MNCs and F4/80 lo CD11b + cells are increased in the liver of LXRα/β-KO mice 17 , iNKT cells were nearly absent in the liver and spleen of LXRα/β-KO mice ( Fig. 1 and Supplementary Fig. 1). We previously reported that the percentage of NKT cells in hepatic MNCs is not different in LXRα-KO mice and WT mice fed a standard diet 18 .
In the current study, we calculated cell numbers per liver weight and observed that hepatic NKT cells and iNKT cells are decreased in LXRα-KO mice. LXRβ-KO mice also had a decreased population of iNKT cells in the liver, and LXRα/β-KO mice showed more severe phenotypes (Fig. 1). Consistent with the decreased number of iNKT cells, cytokine production was impaired in LXRα/β-KO hepatic MNCs and splenocytes after stimulation with the NKT cell-specific activator α-GalCer and also in LXRα/β-KO mice treated with α-GalCer or Con-A (Figs. 2  and 3). The percentage of iNKT cells in hepatic MNCs is decreased in LXRα-KO mice fed high-fat and highcholesterol diet and the response to α-GalCer is impaired in these mice 18 . Con-A-induced hepatitis is attenuated in LXRα-KO mice and is exacerbated in constitutively active LXRα knock-in mice 33 . These findings indicate that LXRα and also LXRβ are involved in hepatic iNKT cell-mediated immune responses.
Cytokine production (such as IL-4, IFN-γ and TNF-α) induced by the iNKT cell activator α-GalCer was impaired in LXRα/β-KO hepatic MNCs and in LXRα/β-KO mice (Figs. 2 and 3), supporting the importance of LXR in iNKT immunity. Interestingly, hepatic Ifng mRNA expression, IFN-γ protein levels in conditioned media of hepatic MNCs and plasma IFN-γ levels were higher in LXRα/β-KO mice than WT mice in the absence of α-GalCer stimulation (Figs. 2 and 3). It has been reported that IFN-γ + CD4 + T cells and IFN-γ + CD8 + T cells are increased in LXRβ-KO mice 34 and that LXR ligand activation decreases IFN-γ expression in CD4 + T cells 35 , suggesting that CD4 + and/or CD8 + T cells are involved in the increased expression of IFN-γ in the liver of LXRα/β-KO mice. On the other hand, Ifng mRNA expression in splenocytes was not affected by LXR deficiency (Fig. 2). Treatment with α-GalCer did not change Ifng expression at 6 h in both WT and LXRα/β-KO splenocytes. Basal and α-GalCer-induced Ifng expression may be influenced by tissue-selective factors, such as metabolic environment and distinct immune cell composition.
Importantly, LXRα/β-KO mice were incompetent in mounting antitumor immunity against EL-4 tumor cells in in vivo and in vitro experiments (Fig. 4). LXR antitumor effect has also been reported in mice and humans via a mechanism in which induction of the LXR target apolipoprotein E depletes myeloid-derived suppressor cells and activates cytotoxic T cells 36 . In contrast, LXRα activation inhibits dendritic cell-dependent antitumor responses 37 , although LXRα and LXRβ are necessary for dendritic cell chemotaxis 38,39 . Thus, LXRs function in antitumor immunity in an immune cell type-dependent manner. In contrast to the effect against NK cell-resistant EL-4 cells, antitumor effect against NK cell-sensitive Yac-1 cells was not repressed in LXRα/β-KO cells (Fig. 4). We examined the LXR antitumor effect in mice and hepatic MNCs, which contain iNKT cells and other immune cells. Depletion of Kupffer cells decreases liver metastases in EL-4-injected mice 40 . iNKT cells are essential for antitumor immunity against EL-4 cells 41 . These findings suggest that repressed antitumor immunity in the liver of LXRα/β-KO mice is due to impaired iNKT cell-mediated immunity, including a decreased population of these cells. However, we could not compare immune functions of hepatic iNKT cells isolated from WT and LXRα/β-KO mice because iNKT cells were nearly absent in the liver of LXRα/β-KO mice (Fig. 1). In addition, the involvement of other immune cells and parenchymal cells in LXR-mediated antitumor immunity cannot be ruled out. Further studies, such as analysis of mice with iNKT cell-specific deletion of LXR and evaluation of iNKT cells derived from pluripotent stem cells with LXR deletion, will help elucidate the role of LXR in hepatic immunity directly or indirectly regulated by iNKT cells. Recently, several clinical trials utilizing the antitumor activity of iNKT cells have been reported for the treatment of cancer, including as melanoma 42 , non-small cell lung cancer 43,44 , head and neck cancer 45 , and neuroblastoma 11 . Thus, LXR may be used as a therapeutic target in iNKT cell antitumor immunotherapy.
iNKT cells were decreased in the liver and spleen of both young and adult LXRα/β-KO mice (Figs. 1, 6, 7 and Supplementary Fig. 1) and also in those tissues of LXRα/β-KO mice transplanted with WT bone marrow (Fig. 6). Analysis of developmental stages of thymic iNKT cells showed a decrease in stage 1, stage 2 and stage 3 iNKT cells in LXRα/β-KO mice (Fig. 7). In addition, stage 3 iNKT cells were decreased in the liver and spleen of LXRα/β-KO mice ( Fig. 1 and Supplementary Fig. 1). These finding suggest that LXR deficiency disturbs iNKT cell development or distribution in both thymus and peripheral organs and leads to a marked decrease in iNKT cells in the liver. PLZF is necessary for development of the NKT cell lineage 30 , and its expression is induced by the transcription factor early growth response 2 at the developmental stage 0 46 . The expression of Zbtb16 (the gene coding PLZF) was not altered in the thymus of LXRα/β-KO mice (Fig. 7). These findings indicate that LXRs are involved in NKT cell development at the pathway downstream of or independent of PLZF signaling. A recent report by Chan et al. shows that LXRα and LXRβ are involved in thymic T cell development through mechanisms of action on multiple cell types, including TECs and thymocytes 47 . Developmental stages of iNKT cells are partly overlapped with those of conventional T cells but are distinct such as in NK1.1 expression 3 . LXRdependent mechanisms for conventional T cell development may also be involved in iNKT cell development in the thymus. iNKT cells were decreased in the liver and spleen of LXRα/β-KO mice transplanted with WT bone marrow compared to WT mice transplanted with WT bone marrow (Fig. 6), but the number of conventional T cells in these tissues was not different (Supplementary Fig. 2). These findings suggest that LXR in non-bone marrow-derived cells plays a role in development of iNKT cells. Generation of CD45.1 LXRα/β-KO mice and bone marrow transplantation experiments utilizing these mice will be helpful for evaluating the intrinsic role of www.nature.com/scientificreports/ LXR in bone marrow-derived cells. iNKT cell development during animal development should also be further investigated in mice with cell-specific LXR deletion. The population of hepatic iNKT cells in adult LXRα/β-KO mice was markedly decreased compared to young LXRα/β-KO mice (Figs. 1 and 6). LXRα/β-KO mice have increased F4/80 lo CD11b + Kupffer cells/macrophages in the liver and also increased cholesterol content in hepatic MNCs 17 . High-fat and high-cholesterol diet feeding increased hepatic MNCs and F4/80 + CD68 + CD11b + Kupffer cells/macrophages in the liver of LXRα-KO mice 18 . Interestingly, these mice demonstrate decreased iNKT cell numbers and cytokine production. Activation of Kupffer cells/macrophages decreases the number of iNKT cells by inducing cell death in mouse models of hepatic steatosis and nonalcoholic steatohepatitis 48,49 . Deletion of LXRα/β in macrophages induces lipid accumulation in the thymus but does not interfere with T cell production 47 . NKT cells develop in the thymus and also in other organs, such as the liver 20 . Increased and activated Kupffer cells/macrophages might be involved in iNKT suppression in the liver of LXRα/β-KO mice. The immune cell communication between Kupffer cells/macrophages and iNKT cells should be further investigated.
In conclusion, we have found that LXRs play an essential role in iNKT cell-mediated immunity, including hepatic antitumor activity. Our findings provide a mechanistic insight into LXR-targeted immunotherapy.

Materials and methods
Compounds. α-GalCer was purchased from Funakoshi Co. (Tokyo, Japan) and Con-A was from Vector Laboratories (Burlingame, CA).  50,51 , and were backcrossed with C57BL/6 J mice for at least ten generations. Mice were maintained under controlled temperature (23 ± 1 °C) and humidity (45-55%) with a 12-h light, 12-h dark cycles, and with free access to water and chow (CE-2; Nihon CLEA). Young male mice (4 weeks old) and adult male mice (12-16 weeks old) were used for experiments. To activate NKT cells in vivo, mice were injected intravenously with α-GalCer (0.1 mg/kg of body weight) or Con-A (12.5 mg/kg of body weight) as reported previously 18 . Tissue samples were collected after euthanization with CO 2 inhalation. All animal experiments were performed according to the protocols, which adhered to the Nihon University Rules concerning Animal Care and Use, approved by Nihon University Animal Care and Use Committee (AP14M066, AP15M027, AP17M014, AP17M015, AP17M073, AP18MED007-1), and conformed to the ARRIVE guidelines.
Isolation of hepatic MNCs, splenocytes and thymocytes. MNCs were isolated with or without collagenase digestion and separated from parenchymal cells with Percoll solution (Sigma-Aldrich, St. Louis, MO) as reported previously 17,18 . For detection of lymphocytes including iNKT cells, MNCs were isolated without collagenase digestion. For detection of Kupffer cells and myeloid cells, MNCs were isolated with the 0.5 mg/mL collagenase digestion (Wako Pure Chemical Industries, Osaka, Japan) for 25 min at 37 °C. F4/80 + cells, CD146 + cells and CD11c + cells were isolated from the liver with collagenase perfusion using QuadroMACS sorting system (Miltenyi Biotec, Bergisch Gladbach, Germany) with Anti-F4/80 MicroBeads, CD146 (LSEC) MicroBeads and CD11c MicroBeads UltraPure (Miltenyi Biotec).
Splenocytes or thymocytes were isolated after filtration with 40 μm nylon mesh and removal of red blood cells. For NKT cell activation in vitro, cells were isolated with collagenase digestion and treated α-GalCer for 6 h for gene expression analysis, or 16 h for enzyme-linked immunosorbent assay.