Deletion of CD1d in Adipocytes Aggravates Adipose Tissue Inflammation and Insulin Resistance in Obesity

Obesity is a crucial risk factor for insulin resistance that leads to various metabolic disorders (1). In obese individuals, various immune cells respond to altered nutritional status and mediate chronic and low-grade inflammation in adipose tissues (1–5). Moreover, immune homeostasis is maintained in adipose tissues of lean subjects, whereas immune balance is disrupted in obese subjects (6–12).

Among various immune cells in adipose tissues, invariant natural killer T (iNKT) cells alleviate inflammation of adipose tissue in obese individuals (13–17). iNKT cells are the type I NKT cells that are characterized by the expression of semi-invariant CD1d-restricted T cell receptors (TCRs) (18). Also, iNKT cells recognize lipid antigens loaded on CD1d molecules (19). The CD1d molecule shows a homology with MHC class I (MHC-I) polypeptides and is expressed in several cell types including dendritic cells, macrophages, and B cells (18). Marine sponge–derived α-galactosylceramide (α-GC) is a potent CD1d-binding lipid antigen that activates iNKT cells (20). In addition, stimulated iNKT cells can secrete large quantities of Th1- and Th2-type cytokines (21). In iNKT cells, determination of cytokine characters into Th1 type or Th2 type is influenced by antigen-presenting cell (APC) types, environmental cytokine milieu, and lipid antigen species (18,21).

Although there is the controversy for the role of iNKT cells in obesity, we and other groups reported the protective roles of iNKT cells against obesity-induced adipose tissue inflammation (13–17,22–25). Compared with wild-type mice, iNKT cell–deficient animals such as Jα18-knockout (KO) or CD1d-KO mice exhibited higher proinflammatory responses and exacerbated insulin resistance upon high-fat diet (HFD) (13–17). Despite the establishment of these models and the reporting of their findings, the regulatory mechanisms for adipose iNKT cell activation have not been thoroughly understood.

Upon excess consumption of nutrients, lipid metabolites are dynamically regulated and processed in adipocytes (26). Interestingly, adipocytes highly express CD1d and potentially activate iNKT cells in vitro (13,17). In the current study, we investigated the in vivo roles of adipocyte CD1d in the regulation of adipose iNKT cells, adipose tissue inflammation, and insulin resistance in obesity. Accordingly, we generated adipocyte-specific CD1d-KO (CD1d^ADKO) mice and then analyzed the effects of adipocyte CD1d deletion on adipose tissue immune responses and metabolic alterations following HFD feeding. The present experiments suggest that adipocyte CD1d is a crucial activator of adipose iNKT cells and that adipose iNKT cell activation could alleviate adipose tissue inflammation and insulin resistance in obese subjects.

CD1d^ADKO mice were generated by crossing adiponectin cre mice (provided by Dr. Koh, KAIST, Daejeon, Korea) with CD1d^flox mice (C57BL/6-CD1d1^tm1.Aben/J; The Jackson Laboratory). C57BL6/J mice were purchased from Central Lab. Animal Inc. (Seoul, Korea). HFD feeding experiments were performed after feeding 8-week-old mice with a 60% HFD (Research Diets Inc., New Brunswick, NJ) for the indicated periods. Glucose tolerance tests (GTTs) were performed using intraperitoneal glucose injections (1 g/kg body weight) after overnight fasting. Insulin tolerance tests (ITTs) were performed using intraperitoneal injections of human insulin (0.75 units/kg body weight; Eli Lilly and Company, Indianapolis, IN). To test α-GC–induced activation of iNKT cells, α-GC (1 μg/mouse; Adipogen, San Diego, CA) was injected intraperitoneally. Prior to interleukin-4 (IL-4) supplementation experiments, IL-4 complex was prepared using a mixture of recombinant mouse IL-4 (BD Biosciences) and anti-mouse IL-4 antibody (BD Biosciences) as described previously (27,28). IL-4 complexes (2 μg IL-4 and 10 μg anti–IL-4 antibody/mouse) were injected intraperitoneally twice during the 8 days of HFD feeding. The Institute of Laboratory Animal Resources at Seoul National University reviewed the protocols and approved all animal experiments.

Quantitative RT-PCR was performed as described previously (17).

The protocol for cell preparation and antibody staining from adipose tissue and spleen was described previously (17). The gating strategies are presented in Supplementary Fig. 7. TCRβ (BD Biosciences) and phycoerythrin-conjugated PBS57-loaded CD1d-tetramer (National Institutes of Health Tetramer Core Facility) double-positive cells were gated as iNKT cells. For macrophage analyses, stromal vascular cells (SVCs) were stained with CD11b (BD Biosciences), F4/80 (eBioscience), CD11c (eBioscience), and CD206 (BioLegend, San Diego, CA) monoclonal antibodies (mAbs). Cells were analyzed using an FACSCanto II instrument (BD Biosciences). Apoptosis and proliferation of iNKT cells were assayed using Annexin V staining kits (BD Biosciences) and a Ki67 mAb (eBioscience). Intracellular cytokine staining was performed after 4-h treatments of SVCs with phorbol myristate acetate (50 ng/mL) and ionomycin (1 μg/mL) (PI), and Golgiplug (BD Biosciences) was added during the last 2 h. Analyses of CD1d expression on adipocytes were performed after incubating adipocytes with CD1d mAb (eBioscience) for 20 min at 4°C. For flow cytometry analysis with adipocytes, adipocytes were passed through the cell strainer (BD Falcon) to prevent cell blocking in instrument. Adipocyte fraction was gated by forward light scatter/side scatter difference comparing with the SVC fraction, as shown in Supplementary Fig. 7B. By using this method, CD1d deletion in adipocytes was nicely detected (Supplementary Fig. 1A).

To isolate macrophages and endothelial cells, epididymal adipose tissues (EATs) from normal chow diet (NCD)–fed mice (n = 5) were pooled, and SVC pellets were prepared. SVCs were then stained with mAbs against CD11b, CD11c, CD45, and CD31 (BD Biosciences), and macrophages (CD45⁺CD11b⁺) and endothelial cells (CD45⁻CD31⁺) were sorted using FACSAria (BD Biosciences). iNKT cells were sorted into TCRβ and PBS57-CD1d tetramer double-positive cells using FACSAria (BD Biosciences).

EATs (0.2 g) were chopped into small pieces. These fat tissues were incubated with DMEM/high glucose plus 10% FBS. After 48 h, we harvested the conditioned media (CM). Secreted levels of cytokines (IL-4, IL-6, and IL-1β) were measured by using a multiplex assay kit (eBioscience).

Bone marrow–derived macrophages (BMDMs) were differentiated as described previously (29). Briefly, bone marrow cells were isolated from femurs and tibias of wild-type mice. The cells were differentiated into BMDM in Iscove’s modified Dulbecco’s medium containing 15% of L-929 CM for 7 days. CM were mixed with same volume of BMDM differentiation medium and then treated with BMDM (M0) for 24 h.

Serum was prepared from mice that were fasted for 2 h. Serum insulin levels were measured using an Insulin ELISA kit (Morinaga Institute of Biological Science Inc., Yokohama-Shi, Japan). Serum free fatty acids (FFAs) were measured using an FFA assay kit (Half-Micro test; Roche, Mannheim, Germany).

EATs were harvested from 1 week of HFD-fed mice, and tissue lysates were prepared by using RIPA buffer. Anti-mouse arginase-1 (BD Biosciences) and anti-mouse GAPDH (Sigma-Aldrich) antibodies were used.

Data were presented as means ± SEM. Differences between two groups were identified using the Student t test in Excel (Microsoft). Differences among groups were evaluated using two-way ANOVA followed by post hoc Turkey tests. Repeated-measures ANOVA (RM-ANOVA) was used for GTT and ITT data. Values of P < 0.05 were considered significant.

To identify in vivo APCs for adipose iNKT cells, we determined CD1d expression in several adipose cell types. EATs were fractionated into adipocytes and SVCs, and SVCs were then sorted into M1 macrophages, M2 macrophages, and endothelial cells using FACS analysis. We confirmed cell sorting purity by examining the mRNA levels of cell-specific marker genes such as F4/80 (macrophage marker), CD11c (M1 marker), arginase-1 (M2 marker), and CD31 (endothelial marker) (Fig. 1A). CD1d mRNA was prominently expressed in adipocyte fractions compared with macrophages and endothelial cells (Fig. 1B). However, mRNA levels of other antigen-presenting molecules, such as MHC-I and -II, were not more abundant in adipocyte fractions than in macrophages or endothelial cells. In addition, we examined the levels of CD1d protein on the cell surface of adipocytes and adipose macrophages (CD11b⁺F4/80⁺). The level of CD1d protein on adipocytes was much greater than on macrophages (Fig. 1C) and downregulated after 16 weeks of HFD feeding (Fig. 1D). These data suggest that differentiated adipocytes could act as APCs for adipose iNKT cells and that reduced expression of CD1d might be associated with dysregulated adipose iNKT cells following diet-induced obesity (DIO).

Given that adipocyte is the highest CD1d-expressing cell type in adipose tissue and its lipid metabolism is dynamically regulated by nutritional status, we hypothesized the idea that adipocytes could indeed serve as an APC for adipose iNKT cells. To address this, we generated CD1d^ADKO mice by crossing adiponectin-cre mice with CD1d^flox mice. To investigate the alteration of phenotypes depending on CD1d deletion in adipocytes, the littermate CD1d^f/f mice were used as a control group. When we analyzed the levels of CD1d mRNA in several tissues, CD1d was selectively ablated in adipose tissues (Fig. 2A). The absence of CD1d protein in adipocytes was confirmed using whole-mount immunohistochemistry and flow cytometry (Fig. 2B and Supplementary Fig. 1A). Because macrophages are abundant in adipose tissues and well known as professional APCs (30), we carefully compared CD1d levels in adipose tissue macrophages from CD1d^f/f and CD1d^ADKO mice. Unlike adipocytes from CD1d^ADKO mice, the CD1d expression on macrophages was not abolished in CD1d^ADKO mice (Supplementary Fig. 1B), indicating that CD1d was selectively ablated in adipocytes but not macrophages.

Because CD1d is central to iNKT cell–mediated immune responses, we analyzed numbers and characters of iNKT cells in adipose tissues from CD1d^ADKO mice. The number of iNKT cells per fat mass and percentages of iNKT cells in SVCs and lymphocytes were significantly reduced in CD1d^ADKO mice (Fig. 2C–E). In contrast, numbers of splenic and hepatic iNKT cells did not differ between CD1d^ADKO and control mice (Fig. 2F and Supplementary Fig. 1C). In addition, we further investigated the cellular events leading to a decreased iNKT cell number in adipose tissue of CD1d^ADKO mice. The levels of apoptosis and proliferation in adipose iNKT cells were not greatly altered by CD1d deletion in adipocytes (Supplementary Fig. 1D and E), implying that reduced number of adipose iNKT cells in CD1d^ADKO mice might not be because of changes in apoptosis or proliferation. Taken together, our data suggest that adipocyte CD1d would have distinct roles in the maintenance of iNKT cell number in adipose tissue.

Candidate adipose APCs for iNKT cell activation include macrophages and dendritic cells. Thus, to examine in vivo roles of adipocyte CD1d in the regulation of iNKT cell activation in adipose tissues, we administrated α-GC into CD1d^f/f and CD1d^ADKO mice and assessed iNKT cell activation. Expression of the T cell activation marker CD69 in adipose tissues was upregulated on iNKT cells in α-GC–treated CD1d^f/f mice (Fig. 3A). On the contrary, it was abolished in adipose iNKT cells of CD1d^ADKO mice. In splenic iNKT cells, α-GC–induced CD69 expression on iNKT cells was not different between CD1d^f/f mice and CD1d^ADKO mice (Fig. 3B). Additionally, mRNA expression levels of the inflammatory molecules in adipose tissues were significantly stimulated by α-GC in CD1d^f/f mice, but not in CD1d^ADKO mice (Fig. 3C). In contrast, these differences were not observed in liver tissues from either CD1d^f/f or CD1d^ADKO mice (Fig. 3D), suggesting that adipocyte CD1d could regulate iNKT cell activation in adipose tissue.

To investigate the roles of adipocyte CD1d in DIO, we compared various metabolic parameters between CD1d^f/f and CD1d^ADKO mice following feeding on an NCD or HFD. In these experiments, body weight and masses of inguinal adipose tissues, EATs, and liver were not different between CD1d^f/f mice and CD1d^ADKO mice, regardless of NCD or HFD feeding (Fig. 4A and B). It is of interest to note that HFD-fed CD1d^ADKO mice showed higher fasting glucose levels than HFD-fed CD1d^f/f mice (Fig. 4C). However, the level of serum insulin was not altered by CD1d deletion in adipocytes (Fig. 4D). Accordingly, GTTs revealed no differences between NCD-fed CD1d^f/f mice and CD1d^ADKO mice (Fig. 4F). In contrast, glucose intolerance was exacerbated in HFD-fed CD1d^ADKO mice compared with HFD-fed CD1d^f/f mice. Moreover, ITTs revealed that HFD-fed CD1d^ADKO mice were less insulin sensitive than HFD-fed CD1d^f/f mice (Fig. 4G). In addition, serum FFA levels were higher in CD1d^ADKO mice after HFD feeding (Fig. 4E). Taken together, these data suggest that deletion of adipocyte CD1d would exacerbate insulin resistance in DIO.

In further experiments, we decided to identify factor(s) that mediate insulin resistance in HFD-fed CD1d^ADKO mice. Because CD1d was specifically deleted in adipocytes of CD1d^ADKO mice, we speculated that dysregulation of adipose iNKT cells may lead to insulin resistance in HFD-fed CD1d^ADKO mice. The numbers and percentages of adipose iNKT cells were decreased in CD1d^f/f mice upon HFD (Fig. 5A–C), whereas they were not greatly altered in CD1d^ADKO mice. Unlike adipose tissues, the percentage of iNKT cells in the spleen was not changed by HFD feeding (Fig. 5D). Although macrophages could also affect adipose iNKT cells, the level of CD1d expression was not affected by HFD or adipocyte CD1d deficiency (Supplementary Fig. 2A and B). Furthermore, the CD1d mRNA deletion was sustained in adipose tissues of CD1d^ADKO mice upon HFD feeding (Supplementary Fig. 2C and D). These data indicate that specific reduction of adipose iNKT cells by HFD might be sensitively regulated by adipocyte CD1d.

Murine iNKT cell population can be classified as CD4⁺ or CD4⁻ subsets (18,19). Moreover, it has been shown that human CD4⁺ iNKT cells are enriched for the expression of Th2-type cytokines including IL-4 compared with CD4⁻ iNKT cells, although this character of cytokine production of murine CD4⁺ iNKT cells has not been clearly described yet (21,31). To explore whether the property of adipose iNKT cells in CD1d^ADKO mice might be different from in CD1d^f/f mice, we further examined the percentage of CD4⁺ cells among adipose iNKT cells. CD4⁺ iNKT cells were increased in CD1d^f/f mice upon HFD (Fig. 5E). Moreover, in contrast with CD1d^f/f mice, CD4⁺ iNKT cell numbers were not increased in HFD-fed CD1d^ADKO mice. Next, we tested the differences in cytokine production between CD4⁺ and CD4⁻ iNKT cells following activation of SVCs from CD1d^f/f and CD1d^ADKO mice using PI to compare the potency of IL-4 production. IL-4 was found to be highly induced in PI-treated CD4⁺ iNKT cells but not in CD4⁻ iNKT cells (Fig. 5F), implying that reduced number of CD4⁺ iNKT cells might affect exacerbated adipose tissue inflammation in HFD-fed CD1d^ADKO mice.

Along with IL-4 production from CD4⁺ iNKT cells and altered percentage of CD4⁺ iNKT cells in HFD-fed CD1d^ADKO mice, IL-4 expression was tested in adipose iNKT cells. Upon HFD, the expression of IL-4 mRNA was increased in adipose iNKT cells (Supplementary Fig. 3A), implying that adipose iNKT cells may produce anti-inflammatory cytokines to confer resolving process against nutritional excess-induced adipose tissue inflammation. However, the iNKT cells from adipose tissues of HFD-fed CD1d^ADKO mice showed lower expression of IL-4, which was consistent with the reduced proportions of CD4⁺ cells among adipose iNKT cells (Fig. 5G). In a case of splenic iNKT cells, IL-4 expression was not different between CD1d^f/f and CD1d^ADKO mice. Moreover, the levels of IL-4 mRNA and protein secretion were greatly decreased in adipose tissue of HFD-fed CD1d^ADKO mice (Fig. 5H and I). These data imply that the deficiency of adipocyte CD1d might reduce HFD-induced IL-4 production from adipose iNKT cells, probably through dysregulation of CD4⁺ iNKT cells, eventually leading to boosted insulin resistance in DIO.

Because the accumulation of adipose tissue macrophages is a key feature of increased inflammatory response in obesity, we examined the number of adipose tissue macrophages in HFD-fed CD1d^ADKO mice to investigate the effects of dysregulated adipose iNKT cells in the regulation of adipose tissue inflammation. Among macrophage population (CD11b⁺F4/80⁺ cells), CD11c⁺ macrophages are known as M1-type macrophages that contribute proinflammatory responses in obese adipose tissue (9), whereas CD206⁺ macrophages represent M2-type macrophages that play anti-inflammatory roles (5,32). As it has been reported that CD11c⁺CD206⁺ macrophages show more proinflammatory phenotypes (33) and that anti-inflammatory macrophages were gated with CD206^high macrophages (14), we have gated CD11c⁺ macrophages and CD11c⁻CD206^high macrophages as M1 and M2, respectively. Whole-mount immunohistochemistry data showed increased numbers of CD11b⁺ and CD11c⁺ cells in CD1d^ADKO mice compared with CD1d^f/f mice upon HFD (Fig. 6A). Moreover, total numbers of macrophages (CD11b⁺F4/80⁺) and CD11c⁺ macrophages were significantly increased in adipose tissues of HFD-fed CD1d^ADKO mice relative to those in HFD-fed CD1d^f/f mice (Fig. 6B and C). Although the number of M2 (CD11c⁻CD206^high) macrophages was not different between CD1d^f/f and CD1d^ADKO mice upon HFD, ratios of M1/M2 macrophages, which were calculated by normalizing the numbers of CD11b⁺F4/80⁺CD11c⁺ cells to those of CD11b⁺F4/80⁺CD11c⁻CD206^high cells, were increased in adipose tissues of HFD-fed CD1d^ADKO mice (Fig. 6D and E).

Next, to determine the inflammatory status of adipose tissue, the levels of inflammatory cytokines were examined in adipose tissue from CD1d^f/f mice and CD1d^ADKO mice. As shown in Fig. 6F and G, when secreted cytokine levels were measured in CM from ex vivo–cultured EATs, the levels of IL-6 and IL-1β were significantly increased in CM from HFD-fed CD1d^ADKO mice compared with HFD-fed CD1d^f/f mice. These data suggest that CD1d deficiency in adipocytes would promote inflammatory responses by increased M1/M2 ratios in adipose tissues of HFD-fed mice.

Short-term (<1 week) HFD feeding rapidly enhances macrophage infiltration into adipose tissues (34,35) and activates adipose iNKT cells (17). Accordingly, we examined the regulatory roles of adipocyte CD1d in adipose tissue inflammation during the early stage of obesity in short-term (1 week) HFD-fed CD1d^ADKO mice. Similar to data from long-term (8 weeks) HFD-fed mice, short-term HFD-fed CD1d^ADKO mice showed decreased IL-4 expression in adipose iNKT cells compared with HFD-fed CD1d^f/f mice (Supplementary Fig. 3C). Moreover, body weights, tissue masses, and serum insulin level did not differ between CD1d^f/f and CD1d^ADKO mice (Supplementary Fig. 4A–C and E). In contrast, fasting blood glucose level was higher in HFD-fed CD1d^ADKO mice than CD1d^f/f mice (Supplementary Fig. 4D). Insulin resistance assessed by GTTs and ITTs was exacerbated in CD1d^ADKO mice following short-term HFD feeding (Fig. 7A and B). Compared with HFD-fed CD1d^f/f mice, the proportion of CD11c⁺ macrophages in SVCs of EAT was further enhanced in HFD-fed CD1d^ADKO mice (Fig. 7C and H), although total macrophage numbers were not different (Supplementary Fig. 4F). However, CD11c⁻CD206^high macrophage (M2) numbers and arginase-1 protein level were considerably diminished in HFD-fed CD1d^ADKO mice (Fig. 7D and Supplementary Fig. 4G). As shown in Fig. 7E, adipose tissues from CD1d^ADKO mice contained larger CD11c⁺ cell populations among adipose macrophages than those from CD1d^f/f mice after short-term HFD feeding. In contrast, percentage of CD11c⁻CD206^high cells among macrophages was decreased in short-term HFD-fed CD1d^ADKO mice (Fig. 7F). Thus, ratios of M1/M2 were greatly higher in CD1d^ADKO mice compared with CD1d^f/f mice during the early stage of obesity (Fig. 7G). In further studies, alterations in M1- and M2-type macrophages were determined according to M1 and M2 marker gene expression in adipose tissues from CD1d^f/f and CD1d^ADKO mice. In these experiments, M2-related genes were downregulated in EATs of HFD-fed CD1d^ADKO mice, whereas those of proinflammatory genes were upregulated (Fig. 7I).

To test whether some secretory factors might affect macrophage polarization in adipose tissues with or without adipocyte CD1d, CM from ex vivo cultured adipose tissues of short-term HFD-fed CD1d^f/f and CD1d^ADKO mice were harvested and treated to BMDMs. As shown in Fig. 7J, BMDMs treated with the CM from CD1d^ADKO mice suppressed expression levels of M2 marker genes compared with BMDMs treated with CM from CD1d^f/f mice. These results imply that the deficiency of adipocyte CD1d would increase M1/M2 ratio, which might result in the acceleration of adipose tissue inflammation and insulin resistance at the early stage of obesity.

Because adipose iNKT cells from HFD-fed CD1d^ADKO mice showed decreased IL-4 expression, we determined whether IL-4 supplementation might alleviate adipose tissue inflammation and insulin resistance in HFD-fed CD1d^ADKO mice. To address this, we administrated IL-4 into CD1d^f/f and CD1d^ADKO mice during 8 days of HFD feeding (Supplementary Fig. 5A). As an indicator of IL-4 responses, we measured mRNA level of the Ym1 gene, which was promoted by IL-4 (Supplementary Fig. 5B). The mRNA levels of other M2 marker genes such as arginase-1 and mgl-2 were not significantly affected by IL-4 injection (Supplementary Fig. 5C and D). Although IL-4 treatments had no significant effects on glucose tolerance in HFD-fed CD1d^f/f mice, improved glucose tolerance was observed in these HFD-fed CD1d^ADKO mice (Fig. 8A). In addition, the number of macrophages was reduced following IL-4 treatments of HFD-fed CD1d^ADKO mice (Fig. 8B and C), and reduced macrophage accumulations were observed in IL-4–treated HFD-fed CD1d^ADKO mice using immunohistochemistry (Fig. 8D). Furthermore, mRNA expression of inflammatory genes was markedly downregulated following IL-4 supplementation, but only in HFD-fed CD1d^ADKO mice (Fig. 8E). Collectively, these data suggest that dysregulated metabolic phenotypes and adipose tissue inflammation in HFD-fed CD1d^ADKO mice might be mediated, at least in part, by reduced IL-4 level.

In this study, we have demonstrated in vivo roles of adipocytes in the regulation of adipose iNKT cells by investigating the effects of adipocyte-specific CD1d deletion in CD1d^ADKO mice. In the current study with CD1d^ADKO mice, numbers of adipose iNKT cells decreased, and α-GC–induced stimulation of iNKT cell activity was abolished. In addition, HFD-fed CD1d^ADKO mice showed higher inflammatory responses including increased M1/M2 ratio and exacerbated insulin resistance. With respect to increased M1/M2 ratio, adipocyte CD1d deletion decreased IL-4 production in adipose iNKT cells. Nevertheless, IL-4 supplementation into HFD-fed CD1d^ADKO mice alleviated adipose tissue inflammation and insulin resistance. Hence, we suggest that adipocyte CD1d would stimulate anti-inflammatory cytokine production in adipose iNKT cells in vivo, leading to reduced proinflammatory responses and insulin resistance in DIO.

Because adipocyte-specific CD1d deletion reduced cytokine expression of adipose iNKT cells in DIO, it is plausible to speculate that adipocyte CD1d might be required for activation of adipose iNKT cells in the presence of specific lipid antigens or signals that are regulated by HFD. Although we were unable to specify the HFD-induced lipid antigen(s) for adipose iNKT cells because of technical limitations, it is well known that lipid metabolism is dynamically altered in adipocytes during the progression of DIO (26,34,36). Thus, adipose iNKT cells are likely activated by specific lipid metabolites via adipocyte CD1d, resulting in defensive immune responses against excessive inflammation following nutritional surplus.

Emerging evidence suggests that excessive energy intake such as HFD feeding would induce not only proinflammatory responses but also anti-inflammatory responses to maintain adipose tissue homeostasis in obesity (9,16,37). However, accumulated proinflammatory responses can overwhelm anti-inflammatory mechanisms in severe obesity, resulting in adipose tissue dysfunction. In the present experiments, both M1 and M2 macrophages accumulated in adipose tissues during short-term HFD feeding, whereas M2 macrophage accumulation was abrogated in HFD-fed CD1d^ADKO mice. These results suggest that anti-inflammatory responses, especially in the early stage of obesity, are downregulated in HFD-fed CD1d^ADKO mice compared with those in HFD-fed CD1d^f/f mice. In a previous study, high expression of IL-10 and arginase-1 in M2 macrophages limited arginine availability for nitric oxide synthesis in M1 macrophages and limited the ensuing proinflammatory responses (38). Because obesity-induced proinflammatory cytokines and chemokines stimulate M1 macrophage infiltration into adipose tissues via a positive-feedback loop, reduced M2 macrophage numbers may augment M1 macrophage accumulation because of diminished immune resolution in HFD-fed CD1d^ADKO mice (39).

IL-4 induces genes encoding M2 macrophage-signature proteins such as arginase-1 and Ym1 (5,6,15). Accordingly, the present data suggest that reduced IL-4 expression contributes to increased M1/M2 ratios in HFD-fed CD1d^ADKO mice. Moreover, IL-4 KO mice and iNKT cell–deficient mice were previously shown to have decreased arginase expression even after 4 days of HFD feeding, suggesting that iNKT cells potentiate M2 macrophage polarization by secreting IL-4 (16).

Although it has been reported that IL-4 improves insulin sensitivity in obese mice (28), we did not observe any significant effect of IL-4 administration in CD1d^f/f mice during short-term HFD feeding. The different results might be attributable to the short duration of IL-4 treatment, because Ricardo-Gonzalez et al. (28) observed IL-4 effects on insulin sensitivity after long-term (>8 weeks) administration of IL-4. Also, it is possible that the level of endogenous IL-4 would be sufficient to modulate insulin sensitivity in CD1d^f/f mice, leading to the blunted effect of IL-4 administration on the resolution of adipose tissue inflammation.

Recently, Satoh et al. (40) reported that HFD-fed adipocyte-specific CD1d KO mice are protective from DIO and exhibit reduced adipose tissue inflammation. These findings are somewhat consistent with their previous reports showing that adipose NKT cells might aggravate adipose tissue inflammation and insulin resistance in obesity (22,23). They have suggested that adipocyte CD1d deficiency would reduce interferon-γ (IFN-γ) production from NKT cells, which might mediate protective effects in obesity. In contrast, our present study reveals that decreased production of IL-4 from iNKT cells would mediate exacerbated adipose tissue inflammation in HFD-fed CD1d^ADKO mice. However, when we analyzed gene expression of adipose iNKT cells in wild-type mice, IL-4, but not IFN-γ, mRNA level was increased upon HFD (Supplementary Fig. 3A and B). It implies that adipose iNKT cells would contribute to IL-4–mediated anti-inflammatory roles rather than IFN-γ–mediated proinflammatory roles upon HFD. Also, other groups have shown the protective roles of adipose iNKT cells by performing elaborate experiments including iNKT cell adoptive transplantation and high-throughput microarray to characterize adipose iNKT cells (14,24). Currently, it is unclear how CD1d^ADKO mice could show different phenotypes upon HFD. However, this discrepancy might be resulted from several factors including the use of different types of control mice (adiponectin cre/CD1d^flox/+ mice vs. littermate CD1d^flox/flox mice), the ingredients of HFD (tallow and safflower oil of high oleic type vs. lard), and environments of animal facilities. As Th1/Th2 cytokine skewing of iNKT cells could be decided by the lipid species (18,41), the high oleic–type enriched diet could be one of the important causes showing different phenotypes in iNKT cell–deficient or adipocyte CD1d-deficient mice. These issues should be clarified in future studies.

In summary, adipocyte CD1d plays crucial roles in the regulation of adipose iNKT cell–mediated immune responses, which would subsequently induce anti-inflammatory response and help to alleviate proinflammatory response and insulin resistance in DIO (Supplementary Fig. 6). Our data propose one of the defense mechanisms against energy surplus, which is mediated by the interaction between adipocytes and iNKT cells in adipose tissue. Regarding the question of how adipose iNKT cells recognize energy surplus in obese adipose tissue, the identification of obesity-related lipid antigens that might be loaded on adipocyte CD1d would be a critical issue in future studies.

Acknowledgments. The authors thank the National Institutes of Health Tetramer Core Facility for generous provision of CD1d tetramers.

Funding. This work was supported by grants from the National Creative Research Initiative Program funded by the Ministry of Education, Science and Technology (2011-0018312). J.P. was supported by the Brain Korea 21 program.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. J.Y.H. and J.B.K. designed the project. J.Y.H. executed most of the experiments. J.P., J.I.K., and Y.K.L. helped to analyze animal experiments. J.Y.H., J.P., Y.J.P., and J.B.K. wrote the manuscript. J.B.K. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.