Reactive Oxygen Species Regulate the Inflammatory Function of NKT Cells through Promyelocytic Leukemia Zinc Finger

Invariant NKT cells express highly restricted TCR repertoire and share characteristics of both T cells and NK cells (1). NKT cells arise from a common precursor of CD4⁺CD8⁺ double-positive thymocytes that have undergone TCR gene rearrangement and expression. In the thymus, they undergo progressive maturation after positive selection. Using the expression pattern of CD24, NK1.1, and CD44, stage 0 (CD24⁺NK1.1⁻CD44⁻), stage 1 (CD24⁻NK1.1⁻CD44⁻), stage 2 (CD24⁻NK1.1⁻CD44⁺), and stage 3 (CD24⁻NK1.1⁺CD44⁺) NKT cells have been identified (2). Additionally, NKT cell functional subsets, NKT1, NKT2, and NKT17, can be defined based on their distinct expression of transcription factors T-bet, GATA3, and retinoic acid–related orphan receptor (ROR)γt and promyelocytic leukemia zinc finger (PLZF) (3). Thymus-derived NKT cells undergo further differentiation and functional specialization in the periphery and produce a broad range of cytokines to exhibit both proinflammatory or immunoregulatory characteristics (4).

NKT cells express PLZF that is a master transcription factor in determining the innate T cell fate (5, 6). PLZF, encoded by Zbtb16, plays an important role in the acquisition of effector functions such as cytokine production and migration properties in NKT cells during the developmental process (5, 6). It is critical for NKT cell development and function. Ectopic expression of PLZF in conventional T cells leads to phenotype and function similar to NKT cells (5–7). Additionally, PLZF works as a negative regulator of cell cycle progression ultimately leading to growth suppression (8). Enforced expression of PLZF in myeloid cell lines resulted in inhibition of proliferation and differentiation (9). A recent study showed that PLZF promotes hepatic gluconeogenesis (10), suggesting a plausible role of PLZF in cell metabolism.

Unlike conventional T cells that undergo Ag-dependent differentiation to become effector cells in the periphery, each effector NKT cell type is generated as they develop in the thymus (11). The signaling pathways that control cellular metabolism have been shown to have a crucial role in dictating the outcome of T cell activation and effector function (12). Distinct T cell subsets adopt specific metabolic programs to support their needs. Studies have suggested that changes in the metabolic profile of T cells are responsible for defining the effector functions and T cell subsets (13, 14). It remains to be investigated whether NKT cell functions are also controlled by cell metabolism during their development in the thymus.

NKT cells play a unique and important role in many immune diseases, including metabolic diseases such as autoimmune hepatitis (15–17) and obesity (18–20). It seems that the effector function of NKT cells is different depending on the tissue type. Hepatic NKT cells are known to induce inflammation, whereas NKT cells in adipose tissues that might have developed differently from hepatic NKT cells in the thymus possess a regulatory function (21). Currently, the role of cell metabolism in controlling the NKT cell functions is not well understood although studies have shown that disruption of Vps34 (a class III PI3K) (22), metabolic regulator Fnip1 (23), or the autophagy pathway (24, 25) causes a defect in NKT cell development and function. NKT cells are also sensitive to changes in signaling pathways mediated by CD28 (26), ICOS (27), PI3K (28), PDK1, Wnt/β-Catenin (29), and mTOR (30, 31), all of which regulate metabolic pathways.

Reactive oxygen species (ROS) are byproducts of aerobic metabolism that include superoxide anion, H₂O₂, and hydroxyl radicals (32). ROS mediate redox biology and oxidative stress, and they contribute to both physiological and pathological conditions (32). ROS act as essential second messengers in innate and adaptive immune cells (33). However, increased levels of ROS within immune cells can result in hyperactivation of inflammatory responses, resulting in tissue damage and pathology (32, 34). In the immune cells, ROS are known for killing the pathogens in the phagosomes through oxidative burst mediated by NADPH oxidase (NOX) (35). In T cells, ROS levels increase upon activation, and the major initial source of ROS required for T cell activation is mitochondria (36). In the absence of ROS produced by mitochondria, Ag-specific responses of T cells are compromised, whereas the cells proliferate under a lymphopenic environment (36). NOX can be invoked in response to mitochondria-produced ROS to further sustain ROS levels so as to maintain T cell activation (37).

In the present study, we investigated the role of ROS in NKT cell functions. We found that ROS levels in NKT cells are dramatically higher compared with CD4 and CD8 T cells in the spleen and liver but not in the thymus and adipose tissues. As a consequence, NKT cells are susceptible to oxidative stress. Importantly, the inflammatory function of NKT cells correlates with the levels of ROS. Our study also revealed that ROS levels and ROS-mediated changes of NKT cells are controlled by PLZF. Collectively, our study reveals that PLZF governs homeostasis and inflammatory function in NKT cells by controlling the ROS levels.

PLZF transgenic mice driven by the cd4 promoter (6) and the lck promoter (7), PLZF-deficient mice (38), and Vα14 TCR transgenic mice (39) have been previously described. All mice were bred and maintained under specific pathogen-free conditions at the University of Michigan animal facility and used at 8–12 wk of age. All animal experiments were performed under protocols approved by the University of Michigan Institutional Animal Care and Use Committee.

Whole spleen and liver cells were mechanically disrupted onto a 100-μm cell strainer to collect single-cell suspensions. Homogenized spleen cells were subjected to ACK lysis to remove RBCs, washed, and then resuspended in PBS supplemented with 2% FBS (FACS buffer). Homogenized liver cells were resuspended in 40% isotonic Percoll solution, loaded on the top of a 70% Percoll solution (GE Healthcare), and then centrifuged at 970 × g at room temperatures with no brakes for 30 min. Nonparenchymal cells were collected at the interface of the two Percoll layers, washed, and resuspended in FACS buffer. To sort NKT and CD4 T cells, spleen cells were incubated with synthetic peptide PBS57-loaded CD1d tetramers, anti-TCR-β Ab, and anti-CD4 Ab for 30 min on ice. NKT and CD4 T cells were then sorted using FACSAria III (BD Biosciences). To sort ROS-low and ROS-high NKT cells, splenocytes were stained with 2′,7′-dichlorofluorescin diacetate (DCFDA) for 30 min prior to surface staining. Cells were sorted using FACSAria III (BD Biosciences) to isolated ROS-high and ROS-low cells.

The following Abs were used: anti-mouse TCR-β (H57-597) allophycocyanin or Pacific Blue, PBS57-loaded CD1d tetramer allophycocyanin or Pacific Blue, anti-mouse CD4 (GK1.5) PerCP-Cy5.5, anti-mouse CD8 (53-6.7) AmCyan, anti-mouse NK1.1 (PK-136) PE-Cy7, anti-mouse CD44 (IM7) PerCp-Cy5.5, anti-mouse CD69 (H1-2F3) PE-Cy7, anti-mouse CD62L (MEL-14) eVolve605, anti-mouse IFN-γ (XMG1.2) FITC or allophycocyanin, anti-mouse IL-4 (11B11) PE-Cy7, anti-mouse IL-17 (TC11-18H10) allophycocyanin–eFluor 780, anti–T-bet (eBio4B10) FITC, anti-RORγt (AFKJS-9) Pacific Blue, anti-GATA3 (L50-823) allophycocyanin, and anti-PLZF (Mags-21F7) PE (all from eBioscience). Dead cells were excluded by staining with 1 μg/ml propidium iodide (Sigma-Aldrich). To measure intracellular cytokines, cells were stimulated for 5 h with PMA (50 ng/ml; Sigma-Aldrich) and ionomycin (1.5 μM; Sigma-Aldrich) in the presence of Monensin (3 μM; Sigma-Aldrich), permeabilized using Cytofix/Cytoperm Plus (BD Biosciences), and then stained with the appropriate Abs. Transcription factor staining to identify committed cells was performed using the Foxp3/transcription factor staining kit (eBioscience) and intranuclear staining for T-bet, RORγt, GATA3, and PLZF. Data were acquired on a FACSCanto II (BD) and analyzed using FlowJo (software version 9.9; Tree Star).

To measure total ROS, cells (1 × 10⁶/ml FACS tube) were incubated in RPMI 1640 media containing 10% FBS with 1 μM DCFDA (Invitrogen) for 30 min at 37°C. Cells were washed with FACS buffer and stained with relevant surface Abs for flow cytometry. Dead cells were eliminated by staining with 1 μg/ml propidium iodide (Sigma-Aldrich). To measure ROS produced by mitochondria, 2.5 μM MitoSox (Invitrogen) was used. Data were acquired on a FACSCanto II (BD Biosciences) and analyzed using FlowJo (software version 9.9; Tree Star). For sorting of ROS-high and ROS-low NKT cell populations, splenocytes were stained with DCFDA as described above followed by surface staining to identify NKT cells. The two populations were gated based on histogram showing two separate peaks of DCFDA staining and sorted on FACSAria II (BD Biosciences).

One to two × 10⁶ total splenocytes or lymphocytes from liver were treated with 0, 10, 30, or 150 μM H₂O₂ in RPMI 1640 media containing 10% FBS for 30 min at 37°C. The cells were washed with FACS buffer. For detecting ROS levels, the cells were stained with DCFDA followed by relevant surface Abs. For detecting apoptosis, the cells were stained with surface Abs followed by annexin V as per the manufacturer’s instructions (BD Biosciences).

Sorted NKT cells were resuspended at 3 × 10⁵ cells per 96-well flat tissue culture plate in RPMI 1640 medium supplemented with 10% FBS with soluble α-galactosylceramide (α-GalCer; 100 ng/ml) for 3 d at 37°C with 5% CO_2. Sorted CD4 T cells (3 × 10⁵ cells per well) were stimulated with plate-bound anti-CD3 Ab (5 μg/ml) with soluble anti-CD28 Ab (2 μg/ml). To measure intracellular cytokines expression, the cells were restimulated with PMA (50 ng/ml; Sigma-Aldrich) and ionomycin (1.5 μM; Sigma-Aldrich) for 5 h followed by intracellular cytokine staining.

Total RNA was isolated from sorted splenic NKT and CD4 T cells from C57BL/6 mice using RNeasy Plus mini kit (Qiagen) following the manufacturer’s instructions. Then, 50 ng of total RNA was used to do single-strand cDNA synthesis with the RT2 first strand kit (Qiagen). Predesigned primer sets for each gene, that is, Nox1, Nox2, Nox3, Nox4, dual oxidase (Duox)1, and Duox2, were purchased from IDT. Either β-actin or GAPDH were used as internal controls. The inverse log of the ΔΔCT was then calculated.

Data for all experiments were analyzed with Prism software (Prism version 6; GraphPad Software, San Diego, CA). An unpaired Student t test was used for comparison of experimental groups. Correlation coefficient (r²) was calculated by fitting the data to a linear regression curve. The following p values were considered statistically significant: *p <0.05, **p <0.01, ***p <0.001, ****p <0.0001.

To examine whether NKT cells show different metabolism, we measured ROS levels as a parameter of cell metabolism using DCFDA, a fluorogenic dye that is commonly used to detect hydroxyl, peroxyl, and other ROS activity within the live cells (40). ROS levels in NKT cells were similar to CD4 or CD8 single-positive cells in the thymus but greatly increased in the spleen and liver with a mixture of ROS-high and ROS-low cells (Fig. 1A). We further characterized the NKT cell population to get a better understanding of the ROS levels and the NKT cell phenotype using CD44 and CD62L that are markers of NKT cell maturity and activation (6). We compared the levels of ROS of CD44-high and CD44-low cells as well as CD62L-high and CD62L-low NKT cells. The results showed that CD44-high or CD62L-low splenic NKT cells had high levels of ROS, and NK1.1 expression did not correlate with ROS levels (Fig. 1B). The high levels of ROS in NKT cells did not associate with the effector phenotype or being innate type cells because neither CD44⁺ effector CD4 T cells nor NK cells showed high ROS from the spleen and liver (Fig. 1C). Thus, highly elevated ROS is a unique property of NKT cells.

Next, we investigated the source of ROS in NKT cells. Because ROS are mainly produced by NOX and mitochondria (41), we first measured ROS produced by mitochondria (mtROS) using MitoSox (25). Unlike high ROS measured by DCFDA, a small fraction of NKT cells produced mtROS, and both the frequency of mtROS-producing NKT cells and the amount of mtROS measured by mean fluorescence intensity were lower than that of CD4 T cells (Fig. 1D), indicating that ROS in NKT cells are not produced by mitochondria in the steady-state. To determine whether ROS were produced by NOX, we compared the gene expression of members of the NOX family—Nox1, Nox2, Nox3, Nox4, Duox1, and Duox2—in NKT and CD4 T cells using quantitative PCR. CD4 T cells predominantly expressed Nox2, whereas NKT cells expressed both Nox1 and Nox2 (Fig. 1E). Nox3, Nox4, and DUOX mRNA levels were below detection or very low (data not shown). Together, ROS in NKT cells are likely generated by NOX1 and NOX2.

Because excessive ROS can cause stress and damage the cells, we asked whether NKT cells are susceptible to oxidative stress. To test this, we treated the cells with H₂O₂. An appropriate concentration of H₂O₂ provides a permissive oxidative environment for cellular signaling that is ideal to maintain homeostasis and to adapt to the stress. However, H₂O₂ levels above the optimal range cause oxidative damage and aberrant cell signaling, resulting in pathologies (34). We subjected total splenocytes to a range of H₂O₂ concentrations (10–150 μM) and measured apoptosis to compare responses. NKT cells showed increased percentages of apoptotic cell death at 30 μM, whereas a higher dose of H₂O₂ was required to induce cell death of CD4 T cells (Fig. 2A). The data were consistent with the reduced NKT cell frequency and lower ROS in remaining live cells (Fig. 2B). If apoptosis is caused by high ROS, cell death can be prevented by reducing ROS. This was tested by treating cells with antioxidants N-acetylcysteine (NAC) or diphenylene-iodonium (DPI) prior to addition of H₂O₂. NAC is an effective free radical scavenger, whereas DPI inhibits the activity of NOXs and DUOXs that generate superoxide anion and H₂O₂, respectively (42, 43). Both NAC and DPI pretreatment reduced NKT cell death caused by H₂O₂ (Fig. 2C). However, MitoQ, a mitochondria-targeted antioxidant that inhibits mitochondrial ROS (44), failed to protect the oxidative damage caused by H₂O₂ (Fig. 2C), indicating that mtROS do not induce oxidative stress in NKT cells. Next, we asked whether high levels of ROS in NKT cells would render sensitive responses to other cellular stress conditions. To test this, we treated NKT and CD4 T cells with different concentrations of thapsigargin that inhibits endoplasmic reticulum (ER) Ca₂⁺ ATPase to induce stress (45). In contrast to H₂O₂ treatment, NKT cells were highly tolerant to thapsigargin, whereas CD4 T cells showed cell death at 100 nM concentration (Fig. 2D). Therefore, high ROS in NKT cells do not cause sensitive responses to ER stress.

As effector cells, NKT cells can produce cytokines upon short stimulation (46). Because effector functions of T cells are regulated by ROS (36, 47), we asked whether ROS also control the function of NKT cells. To test this, we reduced ROS by pretreating freshly isolated lymphocytes with antioxidant for 30 min followed by stimulation with PMA and ionomycin for 5 h. The ROS level was decreased upon pretreatment with 20 μM DPI in NKT cells but not in CD4 T cells from the spleens (Fig. 3A). Hepatic NKT and CD4 T cells showed the same pattern (Fig. 3B). When cytokine expression was examined, DPI-treated NKT cells from spleen and liver had more IL-4⁺ cells but fewer IFN-γ⁺ and IL-17⁺ cells than without DPI treatment (Fig. 3C, upper panel; Fig. 3D, upper panel). CD4 T cells did not show a measurable change with DPI treatment except an increase in IFN-γ⁺ cells (Fig. 3C, lower panel; Fig. 3D, lower panel). Because it is possible that DPI treatment might have caused a selective loss or gain of an effector cell type, we sorted ROS-high and ROS-low NKT cells from spleen (Supplemental Fig. 1A) and examined NKT cell effector subsets, that is, NKT1, NKT2, and NKT17 cells, using PLZF and RORγt. The results showed that ROS-high NKT cells had more NKT1 and NKT17 cells but fewer NKT2 than did ROS-low cells (Fig. 3E, left panel), which correlated with the cytokine data.

Having observed the correlation between ROS levels and NKT effector types, we asked whether NKT cells in BALB/c mice have more ROS-low cells than in C57BL/6 mice because BALB/c mice are known to have more NKT2 cells (3). To test this, we examined NKT cells from BALB/c mice. Similar to C57BL/6 mice, ROS levels in BALB/c NKT cells were elevated compared with CD4 T cells (Supplemental Fig. 1B). In line with our expectation, the percentage of ROS-low NKT cells was higher in BALB/c than in C57BL/6 mice (Fig 3F). Furthermore, NKT2 cells were highly enriched in the ROS-low populations of BALB/c mice, whereas ROS-high NKT cells were comprised mainly of NKT1 and NKT17 cells (Fig. 3E, right panel). Our finding that BALB/c mice have higher percentages of NKT2 cells in ROS-low populations explains the abundance of the NKT2 phenotype in these mice. Taken together, both the cytokine data and the functional subset analyses suggest that the amount of ROS correlates with the NKT cell function.

It is reported that CD4 T cells increase mtROS production upon activation (36). Therefore, we asked whether NKT cells also produce mtROS upon activation, further increasing total ROS. To answer this, NKT cells from Vα14 transgenic (Vα14^Tg) mice were used to obtain a sufficient number of NKT cells for the assays. Vα14^Tg NKT cells had ROS levels comparable to NKT cells from C57BL/6 mice (Supplemental Fig. 1C). Splenic Vα14^Tg NKT and CD4 T cells from C57BL/6 mice were sorted and stimulated with α-GalCer or anti-CD3 plus anti-CD28, respectively. We then compared the amounts of total ROS and mtROS before and after 3 d of stimulation. As expected, both total and mtROS increased upon activation at day 3 in CD4 T cells. In contrast, in NKT cells, total ROS levels were reduced despite a comparable increase of mtROS to CD4 T cells (Fig. 4A). An increase of CD69 expression indicates that both cell populations were activated (Fig. 4A). NKT cells showed similar response when stimulated with α-GalCer or with anti-CD3 and anti-CD28 Abs (data not shown). Moreover, we found that CD62L-low and CD62L-high NKT cells exhibit decreased and increased ROS, respectively, after stimulation (Fig. 4B, Supplemental Fig. 1D). CD62L-high but not CD62L-low CD4 T cells also showed an increase in ROS levels. We showed that low ROS by DPI pretreatment changed cytokine expression (Fig. 3C). Therefore, we wanted to test whether cytokine expression in activated NKT cells would be similar to that of DPI-pretreated NKT cells due to low ROS. As shown in Fig. 4C, activated NKT cells showed less IFN-γ⁺ and IL-17⁺ but more IL-4⁺ cells than before activation. Activated CD4 T cells also showed a similar response to those treated with DPI (Fig. 4D).

It is known that NKT cells but not CD4 T cells express PLZF, so we looked into a possible role of PLZF in increased ROS levels of NKT cells. To answer this, we examined γδ T cells that are known to express PLZF (48, 49). We found that splenic and hepatic γδ T cells had low and high ROS, respectively (Fig. 5A). To further investigate the role of PLZF in ROS, we compared the amounts of ROS with PLZF expression levels using CD4, NKT, and γδ T cells from C57BL/6 mice. As shown in Fig. 5B, we observed a positive correlation between PLZF and ROS levels. Cells expressing higher levels of PLZF such as NKT cells had more ROS than did those with less or no PLZF, strongly suggesting that the level of ROS is controlled by PLZF. To further examine the role of PLZF in ROS levels in mice without a genetic modification, we examined NKT cells in visceral adipose tissues (VAT). A study reported that these NKT cells do not express PLZF and are different from those in other tissues in their development, phenotype, and function (23, 50). We found that NKT cells from VAT (VAT-NKT) had lower levels of ROS than did those from spleen or liver (Fig. 5C). As reported, we observed that VAT-NKT cells had lower PLZF expression compared with splenic or hepatic NKT cells (Supplemental Fig. 1E). Lastly, we compared PLZF expression in sorted ROS-high and ROS-low NKT cells from C57BL/6 mice that were used in Fig. 3E. ROS-high NKT cells expressed higher PLZF than did ROS-low cells (Fig. 5D), further indicating a role of PLZF in the regulation of ROS levels.

If PLZF controls the level of ROS, it is expected that NKT cells with reduced PLZF should have lower ROS than do NKT cells with normal amounts of ROS. To test this, we examined NKT cells from PLZF haplodeficient (PLZF^+/−) mice, which express less PLZF. Total deficiency of PLZF is not informative because PLZF is required for NKT cell development (5, 6). The ROS levels were greatly decreased in NKT but not CD4 T cells from PLZF^+/− mice compared with the wild-type (WT) mice (Fig. 5E). Consequently, PLZF^+/− NKT cells were more tolerant to oxidative stress than WT NKT cells (Fig. 5F). Next, we asked whether PLZF expression is sufficient to elevate ROS levels. To answer this, we examined CD4 T cells from PLZF transgenic (PLZF^Tg) mice under the premise that expression of PLZF in CD4 T cells that normally do not express PLZF would increase ROS levels and exhibit a sensitive response to oxidative stress similar to NKT cells. Indeed, ROS levels were greatly increased in CD4 T cells from PLZF^Tg mice compared with WT CD4 T cells, from both spleen and liver (Fig. 6A) and became sensitive to a low concentration of H₂O₂ treatment (Fig. 6B). NKT cells were comparable between WT and PLZF^Tg mice (Fig. 6A). The induction of ROS in PLZF^Tg CD4 T cells was not due to an artifact associated with the line of transgenic mouse because NKT cells from independently generated PLZF^Tg mice (cd4cre-PLZF^Tg) showed the same response (Fig. 6C). Similar to activated NKT cells, activated of PLZF^Tg CD4 T cells also showed reduction of ROS (Fig. 6D, 6E) and decreased IFN-γ– and IL-17–expressing cells but not IL-4⁺ cells (Fig. 6F). Collectively, PLZF regulates ROS levels, which in turn controls inflammatory function of NKT cells.

During thymic development, NKT and CD4 T cells showed little difference in ROS levels but, in the periphery, ROS were greatly elevated in NKT cells whereas CD4 T cells maintained the similar levels. It is not clear what causes the increase of ROS in the peripheral NKT cells, but it does not seem to be due to TCR-mediated signaling. It is reported that NKT cells receive very weak, if any, TCR stimulation in the steady-state measured by Nur77 reporter expression (50). Additionally, we showed that α-GalCer stimulation reduces ROS. Therefore, it is unlikely that NKT cells are continuously undergoing TCR-mediated activation leading to the increase of ROS. Although TCR stimulation of T cells induces ROS production in the mitochondria (36), our data showed that only a small fraction of freshly isolated NKT cells produced mtROS, indicating that ROS were from a different source. The primary source of ROS in NKT cells seems to be NADPH oxidases, and NKT cells may have a higher capacity to produce ROS using the NOX system. NK cells, a relative of NKT cells, do not have high ROS, which argues against the fact that the innate nature of NKT cells is responsible for elevated ROS. Rather, it is a unique feature of NKT cells.

Although NKT cells are effector cells, the high level of ROS is not simply due to the effector phenotype of the cells because CD44-high CD4 T effector cells did not have as high ROS as did NKT cells. Instead, it seems that ROS levels increase as NKT cells mature. It is not clear why NKT cells make more ROS in the steady-state because the consequence of high ROS is likely detrimental to NKT cell survival. A slight increase of ROS in a local environment may result in a selective loss of NKT cells but sparing CD4 and CD8 T cells to prevent nonspecific NKT cell–driven immune responses as bystanders. To prevent spontaneous cell death in vivo, it is possible that NKT cells have a built-in mechanism to cope with high ROS. The response of NKT cells to ER stress was not altered, implying that high ROS do not make NKT cells sensitive to other cellular stress. It is also interesting that NKT cells are more tolerant to ER stress than CD4 T cells, suggesting that intracellular environments of the two cell types are different. Upon activation via TCR signaling, in contrast to CD4 T cells, NKT cells downregulate total amounts of ROS, although mtROS levels are upregulated in NKT cells, suggesting different uses of ROS during activation. Perhaps ROS produced by the NOX system in the steady-state of NKT cells is reduced and switched to mitochondria production of ROS. Although underlying mechanisms for this switch need to be investigated, proper control of ROS in NKT cells is important for their survival.

Reducing ROS by either TCR-mediated activation or antioxidant treatment decreased NKT cells expressing IFN-γ or IL-17 but increased IL-4⁺ cells. The role of ROS in NKT cell function is further strengthened by the results that more inflammatory NKT1 and NKT17 cells were found among ROS-high cells, whereas NKT2 cells are more abundant in ROS-low cells. It is possible that the expression of T-bet, GATA3, and RORγt is regulated by ROS-mediated signaling. Alternatively, the functional difference may reflect the heterogeneity of glycolytic potential of NKT cells. It is known that both Th1 and Th17 are highly glycolytic, which induces ROS generation (51). Th17 cells had higher expression levels of glycolytic genes and levels of glycolytic intermediates. Th1 cells were also dependent on glycolysis, but appeared to regulate this pathway posttranscriptionally without accumulation of glycolytic intermediates (52). NKT cells show differential glucose uptake during maturation and effector cell generation in the thymus, but ROS levels of thymic NKT cells are similar at different stages. Although we cannot rule out the possibility that glycolysis contributes to the effector cell differentiation in the thymus, it seems that the maintenance of each effector NKT subset in the periphery requires different cell metabolism, which is reflected by the ROS levels.

The present study provided much evidence demonstrating PLZF as a regulator of ROS and that the level of ROS correlates with the inflammatory and regulatory function of NKT cells. PLZF might act by regulating the expression of genes that are involved in antioxidant pathways or by directly regulating the metabolic pathways. PLZF has been shown to regulate glucose homeostasis via gluconeogenesis and glucose output in the liver (10). Additionally, PLZF affects lipid metabolism by exerting anti-adipogenic activity (53). Note that although PLZF is a critical regulator of ROS in peripheral NKT cells, we did not observe a difference of ROS in thymic NKT and CD4 T cells, and PLZF^Tg CD4 T cells did not increase ROS levels in the thymus. Therefore, PLZF does not appear to control ROS levels in thymocytes. NKT cells reside in the liver and VAT, two organs that play a critical role in the development of meta-inflammation. Interestingly, NKT cells in VAT do not express PLZF (54). We showed in this study that NKT cells in the liver and VAT are ROS-high and ROS-low, respectively, which correlates with the inflammatory and regulatory function of NKT cells and the expression of PLZF.

In-depth studies of mechanisms for PLZF-mediated ROS regulation and NKT cell metabolism are warranted to have a better understanding of NKT cell–mediated inflammatory and regulatory immune responses under a different local environment in vivo. However, these studies require an appropriate tool that allows deleting PLZF after NKT cell development in a tissue type–specific manner. Additionally, it is challenging to explore the role of ROS in NKT function in vivo. Use of pro-oxidants is toxic and administration of antioxidants would affect multiple cell types, complicating the interpretation of the results. Note that although clinical studies showed the beneficial effects of antioxidants to prevent and treat liver disease, including chronic hepatitis C virus (55), alcoholic hepatitis or cirrhosis (56), and nonalcoholic fatty liver disease (57), the therapeutic effect of antioxidants on NKT cells has not been demonstrated. Without long-term longitudinal studies, it is difficult to assess the clinical effect of antioxidant.

In summary, we propose two roles of ROS in the regulation of NKT cell–mediated immune responses. First, under conditions leading to oxidative stress without Ag exposure, NKT cells likely undergo apoptotic cell death. It is possible that ROS-low NKT2 cells survive better than NKT1 or NKT17, skewing the distribution of functional subsets toward NKT2, which would prevent an unwanted inflammatory immune response. Second, when NKT cells get activated upon encountering Ags, NKT cells reduce ROS levels that then rescue them from cell death regardless of effector types. This will initiate a proper immune response. Note, however, that the ROS level is not a sole determinant of NKT cell–mediated immunity. Nevertheless, our study revealed, to our knowledge for the first time, that ROS regulates homeostasis and effector function of NKT cells, both of which are regulated by PLZF.

We thank Dr. Phil King (University of Michigan) for critical reading of the manuscript and insightful comments. The National Institutes of Health Tetramer Facility provided CD1d tetramers.

The authors have no financial conflicts of interest.