T-Cell Costimulation Protects Obesity-Induced Adipose Inflammation and Insulin Resistance

Inflammation is widely believed to play a key pathogenic role in the development of obesity-induced insulin resistance (IR) and type 2 diabetes (1–3). While innate immune activation typified by infiltrating macrophages are widely believed to represent important mediators of obesity-related complications, the role of adaptive immune responses is less well characterized. Recent findings suggest that T-cells may play an important role in this process (4–6). Activated CD8⁺ effector T-cells have recently been shown to promote adipose inflammation by enhancing macrophage recruitment and activation (5,6). The precise mechanisms by which T-cells are activated in obesity remains poorly characterized.

Classically, T-cell activation requires two different types of signals, both of which are delivered by antigen-presenting cells (APCs) (7). The first signal derives from the interaction between T-cell receptor and peptide major histocompatibility complex, whereas the second signal (also called costimulatory signal) is provided by the binding of CD28 to the costimulatory molecules such as CD80 (B7-1) and CD86 (B7-2) (8–11). As the most classical costimulation, B7 is involved in various inflammatory diseases (11–17). Nevertheless, whether B7 molecules play a role in regulation of adipose inflammation in IR is completely unclear. Prevention of B7-mediated costimulation has been suggested to result in impaired T-cell activation, which may be protective in a number of experimental contexts such as experimental allergic encephalomyelitis (11,14), allograft transplantation (12), arthritis (13,16), hypertension (17), and type 1 diabetes (15). In this investigation, we elucidated a homeostatic role of B7-mediated costimulation in diet-induced obesity using CD80/CD86 double knockout (B7 KO) mice and investigated the relevance of this process in humans with obesity and IR.

All procedures of this study were approved by the Committees on Use and Care of Animals. B7 KO (B6.129S4-Cd80^tm1Shr Cd86^tm2Shr/J), Foxp3-GFP knock-in (B6.Cg-Foxp3^tm2(EGFP)Tch/J), and C57BL/6 mice were purchased from The Jackson Laboratory (Bar Harbor, ME). Eight-week-old wild-type (WT) and B7 KO mice were randomized to a normal diet (ND) or a high-fat (HF; 42% calories from fat, Harlan TD.88137) diet for 12 weeks. At sacrifice, mice were fasted overnight and serum was collected for insulin (Crystal Chem, Inc., Downers Grove, IL) and leptin (R&D Systems, Minneapolis, MN) ELISA. Homeostatic model assessment of IR (HOMA-IR) was calculated as follows: HOMA-IR = fasting serum glucose (mg/dL) × fasting plasma insulin (μU/mL) / 405.

We enrolled prospective obese subjects (BMI >30) as part of research protocol investigating the role of visceral adipose inflammation in obesity. Greater omental adipose tissue was obtained during endoscopic gastric bypass surgery. As controls, human visceral adipose tissue (VAT) was procured during endoscopic repair of hernias from lean controls (BMI <30). Human subcutaneous adipose tissue (SAT) was obtained from patients undergoing lipoaspiration/liposuction as part of a separate institutional review board–approved protocol. The stromal vascular fraction (SVF) from VAT/SAT was isolated by digesting it with 1 mg/mL collagenase type 2 from Clostridium histolyticum (Sigma-Aldrich, St. Louis, MO) as described previously (18). All procedures of this study were approved by the Office of Responsible Research Practices, Human Institutional Review Board of The Ohio State University under university protocol 2008H0177. Human informed consent was obtained in writing, and a copy was inserted in the patients’ medical records.

Magnetic resonance imaging (MRI) was performed to assess degree of adiposity. After 12 weeks of HF diet, MRI was performed on a 9.4 T Bruker BioSpin system equipped with ParaVision 4.0 software as previously described (19). A spin echo sequence (repetition time 920 ms, echo time 12 ms, in-plane resolution 256 × 256 μm; 2, 4 averages) was used to acquire 50 transversal, 1-mm thick slices that covered from the top of the kidneys to the hind legs. Fat quantification analysis was performed using OsiriX software (OsiriX Foundation, Geneva, Switzerland). Thresholding technique was applied to all images to separate fat and water signal.

Bone marrow–derived dendritic cells (BMDCs) were generated as previously described (20). For the induction of bone marrow–derived macrophages (BMMs), bone marrow cells were cultured in RPMI-1640 suspended with 10% FBS and 10 ng/ml recombinant mouse M-CSF (R&D Systems). Media were replaced every 2 days. Adherent cells (BMMs) were collected for experiment at day 7.

Intraperitoneal glucose tolerance tests (IPGTTs) and insulin tolerance tests (ITTs) were performed following 12 weeks of diet feeding as detailed previously with minor modifications (21). For IPGTT, mice were weighed and baseline blood glucose was determined after overnight fasting (16–18 h) with free access to water. Animals were then intraperitoneally injected with 2 g/kg d-glucose. Blood glucose was detected 30, 60, 90, and 120 min after the glucose injection by using a Contour blood glucose meter (Bayer, Pittsburgh, PA). For ITT, body weight and baseline blood glucose were measured after 4 h of fasting. Mice were then intraperitoneally injected with 0.75 units/kg insulin, and blood glucose was detected at 30, 60, 90, and 120 min.

All antibodies used in flow cytometry were purchased from BioLegend (San Diego, CA) or BD (San Jose, CA). Cells were stained with indicated antibodies as described previously (18) and then analyzed on a BD LSR II cytometer. For intracellular staining of Foxp3, cells were incubated with antibodies against surface maker at 4°C for 15 min. After a wash with PBS, cells were then fixed and permeabilized using a Foxp3 staining kit (BioLegend) as instructed by the manufacturer.

Total RNA were isolated from epididymal fat using Trizol Reagent (Life Technologies, Grand Island, NY). cDNA were synthesized from mRNA using a high-capacity cDNA reverse transcriptase kit (Life Technologies) according to the manufacturer’s instruction. Real-time PCR was performed using a LightCycler 480 SYBR Green I Master kit (Roche, Indianapolis, IN). 18s rRNA was used as internal control. Primers are listed in Supplementary Table 1.

Serum insulin level and leptin level were measured using an ultra-sensitive mouse insulin ELISA kit (Crystal Chem, Inc.) and a quantikine mouse leptin immunoassay kit (R&D Systems) as instructed by the manufacturer.

Total proteins were prepared from adipose tissue using M-PER mammalian protein extraction reagent (Thermo Scientific, Waltham, MA) supplemented with protease inhibitor cocktail (Sigma-Aldrich). Western blot analysis was carried out as reported by probing the blots with an indicated primary Ab (Santa Cruz, CA) followed by a horseradish-peroxidase-conjugated secondary antibody (Santa Cruz, CA). The reactive bands were visualized using an ECL Plus Western Blotting kit (Pierce, Rockford, IL).

Regulatory T-cells (Tregs) were isolated from the lymph nodes and spleen of Foxp3-GFP knock-in mice by flow sorting. CD4⁺ GFP⁺ natural Tregs and CD4⁺ GFP^- naïve T-cells (Tn) were cocultured with BMMs at the concentration of 1 × 10⁶/mL for 5 days. Cells were then harvested for the detection of macrophage polarization and Treg proliferation. For cell proliferation assay, Tregs were labeled with CellTrace violet stain (CellTrace Violet Cell Proliferation Kit, Life Technologies) before coculture according to the manufacturer’s instruction. Fluorescence intensity was analyzed on an LSR II flow cytometer, and proliferation was evaluated by the percentage of divided cells (with lower fluorescence intensity). For the analysis of Treg development, thymocytes were isolated from Foxp3-GFP knock-in mice and cocultured with BMMs or BMDCs derived from WT or B7 KO mice. Treg percentage in thymocytes was evaluated after 5 days of coculture by flow cytometry.

For the adoptive transfer study, natural Treg and Foxp3^- Tn were prepared from Foxp3-GFP knock-in mice by flow sorting. Treg or Tn (0.5 × 10⁶/mouse) were then suspended in 200 µL 1 × PBS and adoptive transferred into B7 KO mice by intravenous injection. The day after first injection, mice were fed an HF diet for 12 weeks. The adoptive transfer was performed every 2 weeks until sacrifice.

All data are presented as mean ± SEM. A P value of <0.05 was considered statistically significant. GraphPad Prism 4 was used for statistical analysis using Student t test or one-way ANOVA and Bonferroni post hoc test where appropriate.

To investigate a potential role for costimulatory molecules CD80 (B7-1) and CD86 (B7-2) in obesity-induced IR, VAT and SAT isolated from lean and obese humans were used for flow cytometric detection of CD80 and CD86. Supplementary Table 2 provides baseline metabolic data of patients recruited to the study. Expression of CD80 and CD86 was lower in monocyte/macrophage lineage cells (CD45⁺CD14⁺) and mature macrophages (defined as CD45⁺CD11b⁺) in VAT from obese patients compared with that of lean (BMI <30) controls (Fig. 1A–D). Similarly, CD80 and CD86 expression was lower in CD45⁺CD14⁺ and CD45⁺CD11b⁺ cells from SAT of obese subjects (BMI ≥ 30) compared with lean controls (BMI <30) (Fig. 1E–H). CD80 and CD86 expression on adipose tissue macrophage (ATM) in VAT was negatively correlated with HOMA-IR, suggesting a potential role of CD80 and CD86 in the pathogenesis of obesity-induced IR (Fig. 1H and I). Consistent with human data, expression of CD80 and CD86 on ATM of mice fed an HF diet was lower compared with ND-fed mice (data not shown).

To investigate the involvement of costimulatory molecules CD80 and CD86 in obesity and IR, WT and B7 KO (CD80 and CD86 double knockout) mice were fed an HF diet or ND for a duration of 12 weeks. There were no significant differences in body weight between WT and B7 KO fed ND. In response to HF diet, both WT and B7 KO mice significantly gained weight compared with ND-fed animals, but the degree of body weight gain was identical in B7 KO and WT animals (48.9 ± 1.2 vs. 49.0 ± 1.3 g; P > 0.05) (Fig. 2A). B7 KO mice had significantly lower epididymal fat after 12 weeks of HF diet feeding (1.31 ± 0.04 vs. 0.92 ± 0.06 g for WT vs. B7 KO; P = 0.005) (Fig. 2B). MRI analysis of VAT content demonstrated similar findings. Abdominal fat analysis by MRI (visceral and subcutaneous) demonstrated a reduction in both compartments in B7 KO mice (Fig. 2C). Liver weight was higher in B7 KO mice on HF diet (3.92 ± 0.14 vs. 4.38 ± 0.08 g for WT vs. B7 KO; P = 0.02) (Fig. 2D). Hepatic triglyceride and cholesterol was significantly increased in response to HF diet but without significant differences between WT and B7 KO mice (Fig. 2E).

Figure 2F and G depicts insulin and glucose homeostasis measures as evidenced by IPGTT and ITT. B7 KO mice exhibited significantly worse fasting glucose levels even under ND conditions. Fasting blood glucose levels significantly increased after 12 weeks of HF diet in both WT and B7 KO mice. Overnight fasting blood glucose levels in B7 KO mice fed an HF diet were higher than WT HF-fed animals (Fig. 2H) (ND 78.4 ± 1.5 vs. 94 ± 7.7 mg/dL, n = 10, P = 0.041; HF 107.6 ± 4.7 vs. 127 ± 7.9 mg/dL, n = 10, P = 0.042). Fasting insulin levels were significantly higher in B7 KO mice on both ND and HF diet compared with WT mice on corresponding diets (Fig. 2I) (ND 0.46 ± 0.13 vs. 0.96 ± 0.11 ng/mL, P = 0.024; HF 1.27 ± 0.10 vs. 1.93 ± 0.04 ng/mL, P = 0.027). HOMA-IR values were significantly higher in the B7 KO mice on both ND and HF diet (Fig. 2J). There was an increase in serum leptin level upon HF diet feeding. However, no significant difference in leptin levels was observed between WT and B7 KO mice (Fig. 2K).

We next evaluated the contribution of B7 to inflammation in circulation and adipose tissue in diet-induced obesity. We first confirmed the absence of CD80 and CD86 expression in splenic and peritoneal macrophages in B7 KO mice (data not shown). As depicted in Fig. 3A and B, comparable amounts of macrophages (CD11b⁺ F4/80⁺) and total T-cells (CD45⁺CD3⁺) were detected in the spleen of WT and B7 KO mice. However, B7 KO mice had reduced CD4⁺ T-cells and increased CD8⁺ T-cells in the spleen (Fig. 3C). Further analysis demonstrated that B7 KO mice had decreased populations of activated CD4⁺ T-cells (less expression of T-cell activation marker CD44 and CD69) in the spleen compared with WT mice (Fig. 3D–G). However, no significant differences were found in splenic CD8⁺ T-cells (Fig. 3D–G), suggesting a reduced circulating CD4 T-cell activation in B7 KO mice. In contrast, innate adipose inflammation was enhanced in B7 KO mice. Increased CD11b⁺F4/80⁺ macrophages and expression of tumor necrosis factor-α (TNFα) were observed in VAT of B7 KO mice (Fig. 3H–J). There was an approximately twofold increase of TNFα in the VAT of WT mice fed an HF diet compared with ND, with a further increase seen in B7-deficient mice fed an HF diet (Fig. 3H and J). Monocyte chemotactic protein-1 (MCP-1) level in mice fed an HF diet increased 1.9-fold compared with that of ND mice. However, no significance was observed between WT and B7 KO mice (Fig. 3I and J). No significant difference in expression of interleukin-10, an anti-inflammatory cytokine, was observed (Fig. 3K). The percentage of CD45⁺CD11b⁺ macrophages increased in VAT of HF WT mice compared with that of ND WT mice (increased from 27 to 36.9%) (Fig. 3J). Compared with WT mice, B7 KO mice had a higher level of macrophage infiltration under both ND and HF diet conditions (Fig. 3L). Consistent with these findings, there was an enhanced hepatic inflammation in B7 KO mice, as evidenced by increased TNFα expression and immune cell infiltration (data not shown).

Prior studies have highlighted the importance of Tregs in obesity and IR, with studies suggesting reduced local differentiation of Tregs as a critical determinant of heightened innate immune responses (4–6). Since CD28 and CTLA-4 have been implicated in Treg development, we reasoned that one mechanism for heightened innate immune activation in adipose tissue may involve Treg pathways. As shown in Fig. 4A, CD3⁺ T-cell infiltration in VAT increased in response to HF diet. However, CD3⁺ T-cells decreased in B7 KO mice in both ND and HF groups. CD4 significantly decreased in the visceral fat of B7 KO mice. CD8 also slightly decreased in B7 KO mice (data not shown). We next quantified Foxp3, a Treg-restricted transcription factor, in VAT of both WT and B7 KO mice in response to HF diet. Consistent with prior studies, Foxp3 expression decreased in the VAT of HF-fed WT mice compared with ND-fed WT mice (Fig. 4B). Foxp3 gene expression was, however, markedly decreased in the VAT of B7 KO mice compared with that of WT mice under both ND and HF diet conditions (Fig. 4B). Flow cytometry confirmed the decrease of Treg in VAT of B7 KO mice (Fig. 4C). Treg numbers were also decreased in the spleen, blood, and lymph nodes of B7 KO mice (Fig. 4D), suggesting a systemic reduction of Tregs in B7 KO mice. To further confirm the relation between ATM B7 expression and Treg in human, human VAT samples were analyzed for the cytometric detection of Treg (CD4⁺Foxp3⁺CD25⁺) and B7-expressing ATMs (CD80⁺CD86⁺CD11b⁺). As shown in Fig. 4E, there was a positive correlation between B7 expression on ATM and Treg percentage in VAT (P = 0.01).

We next investigated the requirement of B7-mediated costimulation in Treg expansion. Treg cells were flow sorted from Foxp3-GFP knock-in mice and fluorescently labeled using a CellTrace violet cell proliferation kit. Labeled Treg cells were then cocultured with BMMs from WT or B7 KO mice for 5 days, and proliferation of Treg was evaluated by the intensity of the fluorescence (Fig. 5A). As demonstrated in Fig. 5B and C, loss of B7-mediated costimulation significantly impaired the ability of macrophages to stimulate Treg proliferation (5.4 ± 0.3 vs. 2.3 ± 0.4%; P < 0.05). To investigate the suppressive function of Treg cocultured with B7^−/− macrophages, WT splenic T-cells were cocultured with Treg cells harvested from the coculture system depicted in Fig. 5A. After 5 days of coculture, T-cells were collected for flow cytometric detection of T-cell activation marker CD44 and CD69. As shown in Fig. 5D and E, Treg cells cocultured with B7^−/− macrophages were less effective in suppressing T-cell activation.

Professional APCs, including dendritic cells and macrophages, are required for the development of Treg in thymus and peripheral tissues (22–25). To investigate the role of B7 molecules in Treg development, thymocytes from Foxp3-GFP knock-in mice were cocultured with WT or B7^−/− BMMs. After 5 days of coculture, CD4⁺ cells were gated for the detection of Treg (GFP⁺) population (Fig. 6A). Smaller amounts of Treg cells were present in thymocytes cocultured with B7^−/− BMMs (19.6 ± 2.2 vs. 12.2 ± 1.1% for WT vs. B7 KO; P < 0.05) (Fig. 6B and C). Similarly, B7^−/− BMDCs have a reduced ability to drive Treg development (Fig. 6D and E). It has been shown that Foxp3^- T-cells can be converted into Foxp3⁺ Treg cells (26). To investigate whether B7 molecules play a role in the conversion of Foxp3^- T-cells to Treg, Foxp3^- T-cells were sorted from Foxp3-GFP knock-in mice and cocultured with WT or B7^−/− BMDCs. As a result, B7^−/− BMDCs were less potent in converting Foxp3^- T-cells to Foxp3⁺ Treg cells (11.1 ± 0.3 vs. 5.9 ± 1.2% for WT vs. B7 KO; P < 0.05) (Fig. 6F and G).

To confirm the involvement of Treg in preventing adipose inflammation, human VAT samples were used for flow cytometric detection of Treg (CD4⁺Foxp3⁺CD25⁺) and CD11c⁺CD11b⁺ cells. As depicted in Fig. 7A, there was a negative association between Treg and CD11c⁺CD11b⁺ cells in VAT (R² = 0.29; P = 0.048). To further investigate whether reduced Treg in B7 KO mice is responsible for heightened adipose innate immune inflammation and IR, the effect of Treg on macrophage activation was examined. As depicted in Fig. 7B and C, Treg reduced CD11c⁺CD11b⁺ cells (72.3 ± 1.9 vs. 51.2 ± 0.8% for WT vs. B7 KO; P < 0.05), a subpopulation that plays an important role in adipose inflammation and IR (27). Consistently, higher percentages of CD11c⁺CD11b⁺ cells were observed in B7 KO mice on both ND and HF diet (Fig. 7D).

To further confirm whether enhanced IR in B7 KO mice is dependent on the decrease of Tregs, CD4⁺GFP⁺ Treg or CD4⁺GFP^- Tn cells from Foxp3-GFP knock-in mice were adoptively transferred into B7 KO mice (once every other week) and fed an HF diet. After 12 weeks of HF diet feeding, transferred Treg (CD4⁺GFP⁺ cells) presented in both circulation (data not shown) and VAT of mice injected with Treg (Fig. 8A). No significant changes in body weight and visceral fat weight were observed (data not shown). There was an ∼2.1-fold increase of Foxp3 expression in the VAT of mice transferred with Treg (Fig. 8B). Macrophage infiltration and its classical activation were reduced by exogenous Treg transfer (Fig. 8C and D). Adoptive transfer of Treg also restored insulin sensitivity in B7 KO mice, as evidenced by improved IPGTT and ITT (Fig. 8E and F).

In this article, we demonstrate a previously unrecognized homeostatic function of costimulation in preserving Treg numbers and modulating innate immune responses. The main novel findings of our work are as follows: 1) expression of B7-1 and B7-2 (CD80 and CD86) on APCs in both VAT and SAT is reduced in diet-induced obesity in mice and human obesity; 2) loss of B7-mediated costimulation exacerbates diet-induced IR despite lower visceral adiposity in both ND and HF diet contexts; 3) contrary to the expected reduction in innate immune activation, genetic deletion of B7 promoted adipose inflammation; 4) B7-mediated costimulation is required for the development and expansion of Treg; and 5) adoptive transfer of Tregs restores insulin sensitivity in response to genetic B7 deficiency.

Recent studies have suggested a critical role of T-cells in mediating obesity-induced adipose inflammation (4–6). Inflamed adipocytes and activated T-cells in the VAT are widely believed to secrete a number of chemokines to recruit macrophages to adipose tissue. However, the underlying mechanisms for impaired T-cell homeostasis in obesity is poorly understood.

B7/CD28-mediated costimulatory signal is required for activation and effector function of T-cells (7). Typically, engagement of the T-cell receptor in the absence of costimulation leads to suboptimal T-cell stimulation, anergy, and apoptosis. CD28 is constitutively expressed by naïve T-cells and binds to the ligands CD80 (B7-1) and CD86 (B7-2), expressed on the surface of professional APCs. Whether B7 molecules are involved in obesity and IR is completely unclear. The contribution of costimulation has been widely studied in the context of autoimmune disease where, in general, it has been shown to play a facilitatory role in promoting inflammation. Approaches to decrease costimulation such as genetic or immune disruption of CD28, CTLA-4, or CD80 and/or CD86 have been shown to attenuate transplant rejection, autoimmune arthritis, and experimental allergic encephalitis. The use of CTLA-4 Ig, which reduces costimulation by binding to CD80 and CD86, has been shown to be protective in humans with type 1 diabetes by slowing the rate of β-cell loss when administered at an early stage of the disease. Unexpectedly, in this study, we observed an enhanced adipose inflammation and IR in mice deficient for CD80/CD86 despite decreased T-cell activation in circulation.

To investigate the role of costimulatory molecules CD80 and CD86 in obesity, human VAT and SAT derived from lean as well as obese/insulin-resistant humans were used for the detection of CD80 and CD86 on APCs. The obese adipose samples were obtained from obese patients undergoing endoscopic gastric bypass surgery. Most of the obese patients were on antidiabetic medication due to morbid obesity and the blood glucose and HbA_1c level was controlled in a relatively normal range. Despite relatively normal metabolic parameters in those obese patients, a correlation between B7 expression and degree of IR was observed. Further studies are required to confirm the implication of B7 in IR in humans. Despite higher indices of innate and adaptive immune inflammation, macrophages expressing CD80⁺CD86⁺ were decreased in both VAT and SAT compartments of obese patients, as well as in the VAT of HF-fed mice. A trend toward a decrease in CD80 and CD86 single positive cells was also noted, although no statistical significance was observed due to the relatively limited sample size. Further study is required to address the significance of CD80 and CD86 single positive cells. CD80/CD86 may be expressed in alternate cell types such as adipocytes and endothelial cells. The expression of CD80/CD86 on endothelial cells is typically low under unstimulated conditions (28) and is insufficient to activate T-cells (29). Since adipocytes constitute dominant cells in adipose, we investigated the expression in these cells and did not find a difference. Therefore, the expression of CD80/CD86 on macrophages, the dominant APCs in adipose tissue, may play an essential role in regulating costimulatory signaling of adipose tissue. Expression of CD80 and CD86 on macrophages in human VAT but not SAT was negatively correlated with HOMA-IR, suggesting a potential role of costimulation in IR. Consistent with our expectation that CD80/CD86 deletion would reduce T-cell activation, we observed reduced T-cell numbers and T-cell activation in CD80/CD86 B7 KO mice (7,10). Due to defective B7-mediated costimulation, T-cell numbers reduced in both spleen and adipose tissue. However, in contrast to our expectation, multiple measures of glucose/insulin homeostasis were impaired by deficiency of B7, despite lack of change in overall body weight and lower visceral adiposity. Since innate immune activation and inflammation are major drivers of IR and metabolic derangement (30–33), we observed an increased level of TNFα and MCP-1, as well as macrophage infiltration in VAT of both obese mice and obese humans.

Prior studies have implicated CD28 and CTLA-4 in Treg development (34–36). Blocking the B7/CD28 axis by neutralizing antibodies or CTLA-4 Ig reduced Treg in C57BL/6 and NOD mice (34,35). Using a CD80/CD86 B7 KO model, we provide direct evidence supporting the essential role of the B7/CD28 axis in Treg development and proliferation in this study. Tregs in spleen, blood, lymph nodes, as well as VAT were dramatically reduced in B7 KO mice. More importantly, a positive correlation between B7 expression on ATM and adipose-tissue-resident Treg was noted in human VAT samples. During the development of obesity, it has been suggested that there is a shift to proinflammatory M1-type macrophages, which contributes to IR. Tregs play an important role in maintaining immune balance by interacting with inflammatory cells, including T-cells and APCs (37). By coculturing Treg or Tn with macrophages, we confirmed that Treg suppresses CD11c⁺ macrophage, a proinflammatory subpopulation of macrophage contributing to obesity-induced complications, including IR. In human VAT samples, we also observed a negative association between Treg and CD11c⁺CD11b⁺ cells. Consistent with this, the CD11c⁺CD11b⁺ population significantly decreased in VAT of B7 KO mice. Furthermore, adoptive transfer of Treg to B7 KO mice abrogated the increase of both macrophage infiltration and macrophage classical activation in VAT.

We next demonstrated that B7-deficient APCs displayed a reduced ability to promote Treg proliferation, differentiation from thymocytes, and conversion from Foxp3^- T-cells. Moreover, Treg cocultured with B7^−/− APCs has decreased suppressive functions, indicating that B7-mediated costimulation is essential for the differentiation, expansion, as well as function of Treg. Furthermore, Treg transfer restored insulin sensitivity and reduced macrophage-mediated adipose inflammation in B7 KO mice. It has been reported that Treg downregulates CD80 and CD86 on APCs through CTLA-4-mediated transendocytosis (38). On the basis of our observations, we propose a bidirectional control loop between Treg and B7 molecules on APCs in healthy individuals with prevention of excessive T-cell activation (Fig. 8G). In obese patients, it is likely that this feedback control loop between Treg and B7 molecules is impaired, leading to a progressive decrease of B7 and Treg, which enhances adipose inflammation (Fig. 8H). Our results do not provide insights on the primacy of each pathway or whether a decrease in CD80/CD86 occurs earlier in obesity than perturbations in Treg numbers. However, our results in ND-fed animals suggest that disruption of CD80/CD86 may alter Treg numbers/function even in the absence of HF diet.

Understanding the mechanisms underlying obesity-induced inflammation and IR is of critical importance in designing novel therapeutic approaches for the treatment of type 2 diabetes. In the current study, we demonstrate for the first time that CD80 and CD86 on ATM are essential in sustaining immune balance in adipose tissue. Expression of CD80 and CD86 decreased during obesity and may represent an important mechanism of impaired Treg homeostasis in obese patients. Our results may have implications for therapies that target costimulation for the treatment of type 2 diabetes and suggest that their application may need to consider potential adverse consequences on Tregs.

Acknowledgments. The authors thank Randy Seeley, University of Cincinnati College of Medicine, for valuable suggestions on the manuscript.

Funding. This work was supported by R01-ES-015146, R01-ES-017290, R21-HL-108467, and R21-DK-088522. J.Z. was supported by RC063384OSU, 13POST17210033, and 81101553. J.A.D. was supported by a National Research Service Award (F32-DK-083903).

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

Author Contributions. J.Z. researched data and wrote the manuscript. X.R. and Z.B. researched data. A.T., V.N., J.H., D.M., B.N., J.R., and J.A.D. contributed to clinical tissue collection. Q.S. and A.R.S. contributed to discussion. S.R. reviewed and edited the manuscript. J.Z. 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.