|Home | About | Journals | Submit | Contact Us | Français|
Adipose-resident invariant natural killer T (iNKT) cells are key players in metabolic regulation. iNKT cells are innate lipid sensors, and their activation, using their prototypic ligand α-galactosylceramide (αGalCer), induces weight loss and restores glycemic control in obesity. Here, iNKT activation induced fibroblast growth factor 21 (FGF21) production and thermogenic browning of white fat. Complete metabolic analysis revealed that iNKT cell activation induced increased body temperature, V02, VC02, and fatty acid oxidation, without affecting food intake or activity. FGF21 induction played a major role in iNKT cell-induced weight loss, as FGF21 null mice lost significantly less weight after αGalCer treatment. The glucagon-like peptide 1 (GLP-1) receptor agonist, liraglutide, also activated iNKT cells in humans and mice. In iNKT-deficient mice, liraglutide promoted satiety but failed to induce FGF21, resulting in less weight loss. These findings reveal an iNKT cell-FGF21 axis that defines a new immune-mediated pathway that could be targeted for glycemic control and weight regulation.
Obesity threatens to shorten the human lifespan by 5–20 years, the biggest burden being obesity-induced diseases (Wang et al., 2011). Obesity is far more complex and touches many more aspects of biology than was previously appreciated. Currently, therapy for obesity is limited by an incomplete understanding of how body weight is controlled. A promising potential approach for treating obesity is activation of brown adipose tissue (BAT). In contrast to energy-storing white adipose tissue (WAT), BAT contains many thermogenic mitochondria that express uncoupling protein 1 (UCP-1), which dissipates chemical energy into heat (Enerbäck et al., 1997). Recently, it was shown that multi-locular adipocytes expressing UCP-1 can also be induced in WAT, known as beige or brite cells (Cousin et al., 1992; Enerbäck, 2009). Browning of WAT uses large amounts of energy through induction of β-oxidation, resulting in increased metabolic rate and weight loss. Chronic cold exposure and β-adrenergic stimulation are physiological inducers of browning in WAT (Cousin et al., 1992; Himms-Hagen et al., 1994). More recently, the hormone fibroblast growth factor 21 (FGF21), produced in liver, WAT, and BAT, was shown to improve metabolic disease and induce weight loss in humans and mice (Gaich et al., 2013; Hanssen et al., 2015; Kharitonenkov and Adams, 2013; Kharitonenkov et al., 2005; Samms et al., 2015). Recently, a synthetic FGF21 variant, LY2405319, was shown to reduce low-density lipoprotein (LDL) cholesterol and triglycerides, increase adiponectin levels, improve fasting insulin, and induce weight loss in obese patients with type 2 diabetes (Gaich et al., 2013). FGF21 administration has been associated with increased UCP-1 levels, which are required for FGF21-induced thermogenesis; however, FGF21 reduces food intake, contributing to weight loss, independently of UCP-1 (Samms et al., 2015). These studies position FGF21 as a promising drug target for the treatment of metabolic disorders.
Recent studies have highlighted a role for the innate immune system in activation of BAT and induction of browning in WAT. Alternatively activated macrophages can produce catecholamines in response to cold exposure, which activates BAT and induces lipolysis in WAT (Nguyen et al., 2011). Eosinophils, which secrete IL-4 and IL-13, sustain alternative macrophage activation and catecholamine production in cold settings (Qiu et al., 2014). Independent of cold exposure, cytokines including IL-33 are critical for maintaining innate lymphoid cells (ILCs) in adipose tissue, which are key players in regulating energy expenditure (Molofsky et al., 2013). ILCs can also induce browning by producing methionine-enkephalin, which upregulates UCP1+ beige adipocytes in WAT (Brestoff et al., 2015). Thus, recent studies have solidified the role of the adipose innate immune system in the regulation of metabolism and body weight.
Invariant natural killer T (iNKT) cells are one such innate immune cell type with an important role in weight and glycemic control. iNKT cells are activated by lipid antigens presented by CD1d molecules (Brigl and Brenner, 2004). We and others have shown that iNKT cells are enriched in human and murine adipose tissue (Lynch et al., 2009, 2012), and that iNKT cells can regulate body weight and restore metabolic homeostasis in obesity (Hams et al., 2013; Huh et al., 2013; Ji et al., 2012; Lynch et al., 2012; Schipper et al., 2012). Recently, we found that adipose iNKT cells are a unique regulatory subset of iNKT cells with a distinct gene expression profile and anti-inflammatory functions (Lynch et al., 2015). Adipose iNKT cells are reduced in obesity, but activating iNKT cells with their prototypical lipid ligand α-galactosylceramide (αGalCer) causes their expansion, which induces potent weight loss; however, the weight loss mechanism is not understood. Here, we report that iNKT cell-induced weight loss occurs through browning of WAT and thermogenesis, without loss of appetite. Furthermore, FGF21 induction is a key player in this weight loss pathway. This iNKT-FGF21 pathway plays a physiological role in a subset of the actions of glucagon-like peptide 1 (GLP-1), a gut hormone that controls glycemia and satiety (Turton et al., 1996), leading to weight loss (Baggio and Drucker, 2007). In rodents, GLP-1 also regulates body temperature (O’Shea et al., 1996) and activates pathways associated with increased numbers of thermogenic beige and brown adipocytes (Lockie et al., 2012). Here, we show that murine iNKT cells contribute to the weight loss effects observed with administration of GLP-1 receptor (GLP-1R) agonists.
To investigate the mechanism of iNKT cell-induced weight loss, we specifically activated iNKT cells in obese mice by using their lipid antigen ligand αGalCer. A single intraperitoneal (i.p.) injection of αGalCer potently induced weight loss in obese mice (Figures 1A and and1B).1B). The effects of αGalCer were dose dependent; 1 μg αGalCer induced an average of 2.5 g of body weight in obese mice on high-fat diet (HFD) for 6–12 weeks (n = 15) (Figures 1A and and1B;1B; p ≤ 0.0001), while mice fed an HFD of longer duration (16–20 weeks) received 2 μg αGalCer and lost significantly more body weight (3.7 g lost with αGalCer, n = 10; versus 0.3 g increase with vehicle control, n = 10). We have previously shown that αGalCer reduces the percentage of fat mass, but not lean mass, in mice as measured by dual X-ray absorptiometry (DEXA) (Lynch et al., 2012). The reduction in total body fat after αGalCer prompted us to look at other fat depots in addition to epididymal fat pads. We found that iNKT cells are enriched in all adipose depots examined compared to spleen, except omentum, which contained fewer iNKT cells (Figure 1C). αGalCer activated iNKT cells in all adipose depots (as measured by iNKT cell expansion >3-fold and CD69 upregulation; data not shown) and induced a significant decrease in fat pad mass in the epididymal, subcutaneous, and perirenal adipose depots (Figure 1D). Thus, specific αGalCer activation of iNKT cells induces potent body weight loss associated with a reduction in the fat mass of several fat depots.
To understand how iNKT cell activation induces a rapid reduction of ~10% of body weight in 4 days, we measured whole-body metabolism using Complete Lab Animal Monitoring System (CLAMS). Obese mice (n = 6) were placed in the CLAMS and served as their own controls: they were first acclimatized and then received vehicle control (1 μg), metabolic readouts were measured for 70 hr, and then they received 1 μg αGalCer treatment and remained in the chambers for 70 hr. As before, mice lost significant weight after αGalCer treatment, but not vehicle control. Unlike control-treated mice, a single αGalCer injection induced an increase in VO2 and VCO2 (Figures 1F and and1G).1G). There was a slight, but not significant, reduction in locomotor activity after αGalCer treatment, and despite weight loss, there was no reduction in daily food intake (Figures S1A and S1B, available online). αGalCer treatment induced a robust and consistent increase in thermogenesis (Figure 1H). To confirm this increase in thermogenesis, we performed a separate experiment where obese mice received vehicle (n = 5) or αGalCer (n = 5), and body temperature was visualized using a forward-looking infrared (FLIR) camera, showing that αGalCer treatment induced a 1° rise in body temperature (Figure 1I). Most strikingly, αGalCer induced a significant decrease in the respiratory exchange ratio (RER) (Figure 1J), which is a measurement of substrate utilization. Decreased RER shows preferential burning of fat, rather than carbohydrates, as the main source of energy, indicating induction of β-oxidation. After adjusting for food intake and body weight of each mouse, energy expenditure calculations showed that one αGalCer treatment induced significantly higher energy expenditure in mice in each 24 hr period following treatment (Figure 1K). In addition to the weight loss, we confirmed our previous finding (Lynch et al., 2012) that αGalCer activation of iNKT cells in obese mice also improves fasting glucose levels (Figure 1L). This striking change in metabolic rate and energy homeostasis would explain the potent effects on weight loss seen with αGalCer treatment.
The induction of β-oxidation, thermogenesis, and reduced WAT suggested that activation of browning of WAT may have occurred. Consistent with this possibility, we found significant induction of UCP1+ cells in inguinal WAT within 24 hr after αGalCer (Figure 2A). Cold temperatures and β-adrenergic stimulators are physiological activators of browning in WAT. As mice treated with αGalCer were housed at room temperature where full BAT activation is not required, we investigated iNKT-dependent mechanisms leading to browning of WAT. FGF21 has been shown to improve metabolic disorders and induce weight loss in humans and mice (Kharitonenkov and Adams, 2013). FGF21 has been implicated in inducing a thermogenic program in WAT through activation of PGC1α (Fisher et al., 2012), as well as by induction of adiponectin, leading to increased energy expenditure (Holland et al., 2013). Administration of αGalCer strongly induced FGF21 transcripts in inguinal adipose tissue within 3 hr of administration, to a similar extent as the β3-adrenoreceptor agonist CL316,243 (Figure 2B). At 3 hr after αGalCer, FGF21 was not activated in BAT; however, by 24 hr after αGalCer treatment, FGF21 transcripts were also significantly induced in BAT and further increased in inguinal adipose tissue (Figures 2B and and2C).2C). In contrast, FGF21 expression was not induced in WAT or BAT of CD1d−/− (iNKT deficient) mice after αGalCer, confirming an iNKT-dependent effect (Figure 2C). FGF21 protein secretion was also significantly increased from explants of inguinal WAT and BAT after αGalCer treatment (Figure 2D). Furthermore, serum FGF21 levels were significantly elevated after αGalCer treatment, at even higher levels than those detected after administration of CL316,243 (Figure 2E). Surprisingly, despite the liver being another key source of iNKT cells and FGF21, αGalCer treatment did not increase FGF21 levels in liver (Figure S1C). αGalCer also induced adiponectin in inguinal adipose tissue of wild-type (WT), but not in CD1d−/−, mice (Figure 2F). Although αGalCer is a specific activator of iNKT cells and is not a β-adrenoreceptor agonist, we assessed if iNKT cells could induce FGF21 without αGalCer activation. To examine this, we adoptively transferred iNKT cells into obese CD1d−/− mice (in vivo), and in another experiment, we co-cultured iNKT cells with obese adipose tissue (in vitro). In both types of experiments, iNKT cells increased levels of FGF21 protein, and adiponectin in adipose tissue (Figure 2G). Thus, our data show that iNKT activation or adoptive transfer induces FGF21, thermogenesis, and weight loss, implicating FGF21 as an important mediator of this pathway. To examine if FGF21 was required for iNKT-mediated weight loss, we investigated FGF21 null mice. FGF21−/− mice were fed an HFD for 12 weeks and treated with αGalCer or vehicle control. In the absence of FGF21, αGalCer-induced weight loss was significantly attenuated, compared to WT controls (Figure 2H). There was no statistical difference in the mean weight of the epididymal fat pads between groups (Figure 2I). Thus, iNKT cell activation or adoptive transfer induces FGF21, adiponectin, β-oxidation, browning of WAT, and increased energy expenditure leading to weight loss, which is partly dependent on induction of FGF21.
Discovery of an iNKT-FGF21 thermogenic pathway led us to question if this pathway was physiologically relevant in other weight loss settings. As we have previously shown that liraglutide and native GLP-1 can activate human iNKT cells (Hogan et al., 2011), and GLP-1 receptor signaling can regulate lymphocyte maintenance and proliferation (Hadjiyanni et al., 2010), we examined whether GLP-1R signaling-induced weight loss was dependent in part on iNKT cell activation. Obese mice fed an HFD for 8 weeks were treated with the GLP-1R agonist liraglutide or PBS control by i.p. injection daily for 5 days. As expected, liraglutide treatment lowered fasting glucose, improved glucose handling (Figure S2A), reduced body weight and epididymal fat pad weight (Figures S2B and S2C), and decreased adipocyte size (Figure S2D). Consistent with our previous finding in humans, liraglutide increased iNKT cell frequency in blood and adipose tissue (Figures 3A and and3B).3B). Adipose iNKT cells were activated by liraglutide treatment, evidenced by an increased expression of the activation marker CD69 in vivo (Figure 3C). Liraglutide also led to proliferation of adipose iNKT cells (Figure S2E) and increased their IL-10 production in vivo (Figure 3D), a hallmark of adipose iNKT cell regulatory function (Lynch et al., 2015). Furthermore, when tested in vitro, liraglutide directly induced proliferation in murine iNKT cell primary lines (Figure S2F) and induced their IL-10 production (Figures S2G and S2H). These results show that the GLP-1 analog liraglutide can activate iNKT cells in vivo and directly in vitro.
As liraglutide activates iNKT cells, and both liraglutide (Lovshin and Drucker, 2009) and iNKT cell activation can promote weight loss (Lynch et al., 2012), we determined whether GLP-1-iNKT cell interactions were relevant to the metabolic and weight loss effects of liraglutide GLP-1-based therapy. We fed WT and CD1d−/− mice (lacking iNKT cells) an HFD for 8 weeks and treated them with liraglutide. The absence of iNKT cells did not impact the glycemic effects of liraglutide (Figures 3E and and3F),3F), for liraglutide induces similar reductions in fasting glucose and glycemic excursion in both WT and CD1d−/− mice. However, the presence of iNKT cells was required for the full weight loss effects after several days of liraglutide administration; CD1d−/− mice lost approximately one-third less overall weight than WT mice after liraglutide treatment (Figure 3G). Although liraglutide produced a significant reduction in food intake in both WT and CD1d−/− (Figure S2I), CD1d−/− mice regained their weight and exhibited no weight loss between days 2 and 6 following liraglutide treatment, unlike WT mice, which continued to lose weight (Figures 3H and and3I).3I). Similar findings (less weight loss with liraglutide) were seen in Ja18−/− mice on HFD, a second murine model that lacks iNKT cells due to deletion of the invariant Ja18 T cell receptor chain (Figure S3). Ja18−/− mice and WT mice lost similar weight in the first 24 hr, when food intake is severely reduced, but following this initial 24 hr period, WT continued to lose weight while Ja18−/− gained weight (Figures S3B–S3D). To rule out a central defect in GLP-1R signaling as a mechanism for resistance to liraglutide-induced weight loss, we found that GLP1r expression in the hypothalamus was similar in WT and iNKT-deficient mice (CD1d−/−) (Figures S3E and S3F). CLAMS studies revealed that CD1d−/− mice had a lower metabolic rate than WT mice (Figures S4A and S4B), consistent with the increase in adiposity previously described (Lynch et al., 2012). Liraglutide induced a slight decrease in overall VO2 and VCO2 in both WT and CD1d−/− mice (Figures S4A and S4B), in agreement with reports that liraglutide-induced weight loss mainly reflects reduced appetite and energy intake (Harder et al., 2004; O’Shea et al., 1996; van Can et al., 2014). Nevertheless, central GLP-1R signaling can also induce thermogenesis through activation of BAT, which may account for a small proportion of liraglutide weight loss effects (Lockie et al., 2012). Thus, we investigated if i.p. liraglutide treatment induced browning of white inguinal fat. Liraglutide induced a robust thermogenic gene expression program in WT mice, including upregulation of UCP-1, PGC1α, and Cidea (Figure 3J). Importantly, this response was blunted or lost in CD1d−/− mice (Figure 3J). Liraglutide administered intraperitoneally did not induce a significant change in thermogenic genes in BAT, although a trend was seen in WT, but not CD1d−/−, mice (Figure S4C). Together, our findings show (1) liraglutide activates iNKT cells, (2) iNKT cells are required for the thermogenic program in WAT induced by liraglutide, and (3) iNKT activation induces FGF21. Previous reports have linked GLP-1 analogs with increased FGF21 expression (Nonogaki et al., 2014; Yang et al., 2012). Thus, we examined whether iNKT cells were required for the induction of FGF21 by liraglutide. Liraglutide robustly induced FGF21 expression in WT inguinal adipose tissue; however, FGF21 was not induced in mice lacking iNKT cells (Figure 3K). Liraglutide also induced expression of FGF21 receptors FGFR and β-klotho in adipose tissue of WT, but not in CD1d−/−, mice (Figure 3L). Adiponectin, a downstream regulator of FGF21 signaling, was also robustly induced in adipose tissue by liraglutide in WT, but not iNKT, cell-deficient mice (Figure 3M). Thus, although the primary mechanism of action for the weight loss effects of liraglutide is a reduction of energy intake, liraglutide also induces adiponectin, FGF21, and an adipose thermogenic program that requires a functional iNKT cell system.
Previously, we have shown that liraglutide can activate human iNKT cells in vitro (Hogan et al., 2011) and can increase levels of circulating adiponectin in obese individuals with type 2 diabetes (Hogan et al., 2014). To ascertain the existence of a GLP-1-iNKT cell-FGF21 axis in humans, we studied a cohort of nine newly diagnosed obese type 2 diabetes patients before and after 8 weeks of liraglutide therapy. HbA1c and body weight were reduced following liraglutide administration (Figures 4A and and4B).4B). This was paired with a significant expansion in peripheral iNKT cell levels (Figure 4C). Liraglutide also significantly elevated circulating FGF21 levels (Figure 4D), which strongly correlated with the extent of weight loss: two individuals without weight loss did not display increases in FGF21, and the individuals who lost the most weight had the largest increases in FGF21 (Figure 4E). These data demonstrate that liraglutide expands iNKT cell number and increases FGF21 levels in obese humans, consistent with our findings in high-fat-fed mice.
Our data provide a mechanism whereby iNKT cells, by adoptive transfer or activation with their specific ligand αGalCer, induce rapid and robust weight loss. Our study shows that an innate T cell population, resident in adipose tissue, drives thermogenesis and weight loss. Thus, iNKT cells join the list of other immune cells that have been reported to control weight, namely ILC2s (Brestoff et al., 2015; Lee et al., 2015) and macrophages (Nguyen et al., 2011). iNKT cell actions are distinct from those previously published. ILC2 cell production of methionine-enkephalin peptides drove UCP1 expression and thermogenesis in WAT (Brestoff et al., 2015), while macrophages can produce catecholamines to sustain thermogenesis in cold temperatures (Nguyen et al., 2011). Here we show that iNKT activation drives production of FGF21 by adipocytes in WAT and a robust and sustained reduction in RER, indicating β-oxidation, coupled with UCP-1 expression, leading to thermogenic weight loss without affecting appetite. We also show that this pathway plays a role in another setting: liraglutide-induced weight loss. Although this pathway is not the main mechanism of liraglutide action, which is primarily satiety, our results highlight a role for activation of this iNKT-FGF21 pathway in the maximal weight loss effects of liraglutide in mice. Of additional relevance, the actions of liraglutide to induce FGF-21 and adiponectin also required a functional iNKT cell axis. Unlike other pathways of cytokine-driven activation of immune cells, iNKT cells can be specifically targeted in the clinic by αGalCer and related lipid ligand administration. αGalCer has been given to patients in several different clinical trials for cancer and has been proven safe and capable of activating human iNKT cells in vivo, with minimal side effects. This study suggests that targeting iNKT cells with specific ligands could represent a new therapeutic approach for subjects with metabolic disorders, such as nonalcoholic fatty liver disease or insulin resistance, that are sensitive to the actions of metabolic mediators such as adiponectin and FGF-21.
Details of patient collection are in the Supplemental Experimental Procedures. Blood was collected from ten type 2 diabetes patients before and after 8 weeks of GLP-1 analog therapy (0.6 mg once daily for 2 weeks, then 1.2 mg once daily by i.p. injection), and metabolic status, iNKT, and FGF21 levels were measured. Informed consent was obtained from all participants, and approval to conduct this study was obtained from the St. Vincent’s Healthcare Group Ethics and Medical Research Committee.
Details of mouse strains are in the Supplemental Experimental Procedures. C57BL/6 mice were purchased from Jackson Laboratory. CD1d−/− and Ja18−/− were provided by Mark Exley (Harvard). FGF21 null mice were kindly provided by Eli Lily (Badman et al., 2009). All animal work was approved by and in compliance with the Institutional Animal Care and Use Committee guidelines of the Dana Farber Cancer Institute and Harvard Medical School.
Details of cell isolation and flow cytometry can be found in the Supplemental Experimental Procedures. iNKT cells were measured using αGalCer analog PBS-57-loaded or empty CD1d tetramers provided by the NIH tetramer facility. For in vivo treatment, mice were injected intraperitoneally with a single injection of αGalCer or vehicle in 150 μL volume. For liraglutide treatment, mice received daily injection of GLP-1 analog (50 μg/kg intraperitoneally) for 5 days. For CLAMS analysis, mice were singly housed and acclimated, then administered vehicle control, monitored for 70 hr, given 1 μg αGalCer in 150 μL intraperitoneally, and monitored for a further 70 hr.
The changes of outcomes between baseline and the follow-up were evaluated using paired t tests. The difference between treatment groups was tested using unpaired two-sample t tests with Welsh’s correction for unequal variances and one-way ANOVA followed by Tukey’s post hoc test. p ≤ 0.05 was considered significant.
The authors would like to thank Eli Lily for the use of FGF21−/− mice and the NIH Tetramer Core for mouse CD1d-PBS57 tetramers. L.L. was supported by an American Diabetes JF Development Award (116JDF061), a BWH Evergreen Innovation Grant, a BADERC grant, and an ERC Stg grant (679173); M.B. is supported by research grants from the NIH (AI063428 and AI028973); and A.H. and D.O.S. are supported by the Health Research Board Ireland. M.B. is supported by NIHR01 111144. D.J.D. is supported by the Canada Research Chairs program, CIHR grant 123391, and a BBDC-Novo Nordisk Chair in diabetes research.
AUTHOR CONTRIBUTIONSL.L. designed and performed experiments and analyzed data; A.E.H. and D.O.S. designed and performed the human work. D.D., C.L., A.G., M.C., D.S., and K.L. performed experiments; A.B., M.B., D.E.C., E.M.-F., and D.J.D. contributed to the design of experiments and provided materials and tools; B.L. performed statistical analysis; and L.L., D.J.D., and M.B. wrote the paper.