IL-33 Reverses an Obesity-Induced Deficit in Visceral Adipose Tissue ST2⁺ T Regulatory Cells and Ameliorates Adipose Tissue Inflammation and Insulin Resistance

Obesity-associated visceral adipose tissue (VAT) inflammation is thought to result in insulin resistance and contribute to the development of type 2 diabetes (1, 2). Diet-induced obesity (DIO) results in infiltration of many proinflammatory immune cells into the VAT, with strong evidence that accumulation of macrophages producing proinflammatory cytokines, such as TNF-α and IL-6, has a major causal role in driving inflammation (3, 4). In parallel to expansion of inflammatory macrophages, obese VAT has a reduced proportion and function of anti-inflammatory FOXP3⁺ regulatory T cells (Tregs), which normally comprise a significant proportion of CD4⁺ T cells in lean VAT (5–7). In addition, mice in which Tregs are ablated have worsened VAT inflammation and insulin resistance (6). These findings have led to the hypothesis that poor Treg function in obese VAT is a major factor contributing to chronic VAT inflammation. In support of this possibility, treatments that enhance Tregs, such as administration of anti-CD3 (5), or IL-2 anti–IL-2 complex (6), or adoptive transfer of Tregs themselves (8), can improve insulin sensitivity.

Tregs resident in lean VAT have a unique phenotype that is at least partly due to the fact they express PPARγ, a transcription factor associated with adipocyte differentiation and function, which cooperates with FOXP3 to drive a distinct VAT Treg gene signature (9). In addition, we previously showed that, in comparison with splenic Tregs, lean VAT-resident Tregs express high levels of IL-10 (7). This high expression of IL-10 by VAT Tregs is specifically altered in DIO: Tregs from obese VAT have a significant reduction in IL-10 production, with unchanged expression of CTLA-4, CD39, CD25, or LAP (7). Because IL-10 is necessary for Treg-mediated suppression of TNF-α production from macrophages (7), this obesity-induced change in VAT Treg function most likely contributes to inflammation and insulin resistance.

In this work, we further investigated the phenotype of Tregs in lean VAT and found that, in comparison with Tregs in the spleen, or conventional T cells (Tconv) in lean VAT or spleen, they also express proportionately higher levels of the ST2 chain of the IL-33R. The proportion of this pool of ST2⁺ Tregs is reduced in obese VAT and can be completely restored by treatment with IL-33. Treatment with IL-33 also reduced VAT inflammation and ameliorated insulin resistance. Hence, therapeutic maintenance of ST2⁺ Tregs in VAT by IL-33 may be a new approach to control obesity-associated inflammation and metabolic disorders.

Four-week-old male C57BL/6 FOXP3-eGFP or C57BL/6 FOXP3-RFP × IL10-GFP mice (bred in house) were fed a normal chow diet (NCD; Lab Diets catalog 5053) or 58 kcal% fat with sucrose diet (Research Diets catalog D12331) for additional 14–16 wk. Diets were not matched for micronutrients. Where indicated, mice were then injected i.p. for consecutive 10 d with 0.5 μg/day rIL-33 or PBS (eBioscience), while maintained on their respective diets, and were sacrificed 10 d after the start of treatment. To determine metabolic effects of IL-33, consecutive 10 d of injections were followed by one injection every 2–3 d of 0.5 μg IL-33 or PBS. Fasting blood glucose and plasma insulin levels were measured at 14 and 17 d after the start of treatment. All animal studies were approved by the Canadian Council on Animal Care.

Spleens and epididymal fat pads (VAT) were processed, as described (7). The stromal-vascular fraction (SVF) was obtained after centrifugation and passed through a cell strainer. Intestines were minced and digested with collagenase, and live cells were isolated using Percoll density centrifugation. Bone marrow–derived dendritic cells were differentiated with GM-CSF and 10 ng/ml IL-4 in RPMI 1640 and 10% FCS for 7 d, and then purified as CD11c⁺ cells (STEMCELL Technologies) and rested overnight in complete media, with or without 10 ng/ml IL-33. CD4⁺ T cells were magnetically sorted (STEMCELL Technologies) from spleens. A ratio of 10 T cells to 1 DC was cultured with 0.01–0.1 μg/ml anti-CD3 (2C11; BD Biosciences) in the presence or absence of 10 ng/ml IL-33 for 5 d.

All flow cytometry and ELISA Abs were commercially obtained from BD Biosciences or eBioscience. A portion of VAT was immediately immersed in lysis buffer and dissociated using gentleMACS (Miltenyi Biotec) for analysis of gene expression. Data were normalized to 18 s. The SVF from digested VAT was cultured overnight at 2 × 10⁶ cells/ml, and amounts of secreted IL-6 were measured in the culture media by cytometric bead array (BD Biosciences).

Mice were fasted for 4–6 h at days 14 and 17 after the start of treatment, respectively. Blood glucose was measured using a OneTouch Ultra 2 glucometer, and plasma insulin was measured by ELISA (ALPCO). Homeostatic model assessment of insulin resistance (HOMA-IR) was calculated as fasting blood glucose (mmol/l) × fasting insulin (mU/l)/22.5, as described (10).

Unpaired Student t tests were used to analyze significance. The p values are indicated as follows: *p < 0.05 and **p < 0.01.

We previously found that Tregs resident in VAT express high levels of IL-10, which is most likely involved in restraining the production of proinflammatory cytokines from macrophages (7). To further characterize the phenotype and function of VAT Tregs, we asked whether these cells also express ST2, one of the chains of the IL-33R, because expression of this protein has recently been reported on Tregs resident in other tissues (11, 12). Single-cell suspensions from the VAT of 20-wk-old mice maintained on NCD were analyzed by flow cytometry. We found that 54.5 ± 14.3% (n = 14) of VAT Tregs expressed ST2, compared with just 2.9 ± 0.9% (n = 14) of splenic Tregs (Fig. 1A). In addition, only 6.8 ± 4.2% (n = 14) of CD4⁺FOXP3⁻ Tconvs in the VAT expressed ST2. CD4⁺FOXP3⁺ Tregs in the VAT also had high expression of CD25, CTLA-4, and CD39 regardless of whether they express ST2 (Supplemental Fig. 1A). VAT Tconvs expressed significantly higher levels of CD25, CTLA-4, and CD39 compared with splenic Tconvs, albeit at lower levels than on Tregs, and interestingly, ST2⁺ Tconvs in the VAT expressed increased levels of CD25 compared with ST2⁻ Tconvs (Supplemental Fig. 1A). The proportion of Tregs expressing ST2 in lean VAT was significantly higher than in the intestinal lamina propria, in which an average of 17.7 ± 0.4% (n = 3) of Tregs expressed ST2 (Supplemental Fig. 1B).

ST2 expression has also been previously linked to Th2 cells (13, 14); thus, we measured expression of Th2-associated proteins CCR4 and GATA3 in VAT Tregs. We found that VAT Tregs expressed significantly higher levels of CCR4 and GATA3 than splenic Tregs (Fig. 1B), but did not express Th2-related cytokines such as IL-13 and IL-5 (data not shown). VAT Tconvs expressed slightly higher levels of GATA3, but not of CCR4, compared with splenic Tconvs (Fig. 1B). By comparison, VAT Tregs did not express increased levels of Th1 (TBET, CXCR3)- or Th17 (retinoic acid-related orphan receptor γt)-associated proteins (Supplemental Fig. 1C). However, VAT Tregs expressed significantly more Th17-associated CCR6 than splenic Tregs (Supplemental Fig. 1C), but so did VAT Tconvs, indicating that this change is not Treg specific.

To ask whether expression of Th2 markers is associated with expression of ST2, we examined expression of CCR4 and GATA3 in ST2⁺ versus ST2⁻ VAT Tregs. As shown in Fig. 1C, VAT-specific expression of GATA3 and CCR4 was restricted to ST2⁺ Tregs, indicating that ST2 appears to specifically mark Tregs with a Th2-like phenotype. Of note, ST2⁺ Tconvs also expressed significantly more CCR4 than ST2⁻ Tconvs (Fig. 1C). Because GATA3 enhances the expression of IL-10 in CD4⁺ T cells (15), and VAT Tregs express high levels of IL-10 (7), we used FOXP3-RFP × IL10-GFP mice to determine the correlation between ST2⁺ and IL-10. Indeed, the vast majority (90.8 ± 6.7%, n = 6) of IL-10–expressing Tregs in VAT also expressed ST2, and by corollary, ST2⁺ Tregs in the VAT expressed significantly more IL-10 than did ST2⁻ Tregs (Fig. 1D): 68.1 ± 13.1% of ST2⁺ Tregs expressed IL-10, compared with only 17.9 ± 9.1% of ST2⁻ Tregs.

To determine whether the Th2-like phenotype and IL-10 production by ST2⁺ Tregs were related to IL-33 signaling, we used bone marrow–derived dendritic cells and IL-33 to differentiate ST2⁺ Tregs in vitro from splenic CD4⁺ T cells, which had very low levels of ST2 expression prior to stimulation (16). This culture system with IL-33 resulted in the induction of ST2⁺ Tregs, which had increased expression of both CCR4 and IL-10 (Supplemental Fig. 2A, 2B).

Interestingly, ST2⁺ Tregs also had increased expression of neuropilin-1 (Nrp1) (Fig. 1E), which is known to promote Treg survival and function (17). In contrast, analysis of Helios revealed no significant differences in expression between ST2⁺ and ST2⁻ Tregs (Fig. 1F), suggesting that these cells are comprised of similar proportions of thymically and peripherally derived cells.

DIO results in a reduced proportion of VAT Tregs due to an expansion of proinflammatory Tconvs (5, 6). In addition, Tregs from obese VAT have diminished expression of IL-10, whereas other suppressive mechanisms, including CD39, CTLA-4, and LAP, remain unchanged compared with cells in lean VAT (7). To investigate the effect of DIO on ST2⁺ Tregs, littermates to the lean NCD mice shown in Fig. 1 were fed with a 58 kcal% fat with sucrose diet (high-fat/sucrose diet [HFD]). We found that obese VAT had a substantially reduced proportion of ST2⁺ Tregs (Fig. 2A): 54.5 ± 14.3% (n = 14) of VAT Tregs expressed ST2 in lean VAT, compared with just 16.3 ± 7.1% (n = 14) in obese VAT from HFD mice. There was a similar HFD-induced reduction in the minor population of VAT Tconvs that also expressed ST2.

In parallel, VAT Tregs from HFD mice also had reduced expression of multiple proteins that were associated with ST2 expression, including CCR4, GATA3, and Nrp1 (Fig. 2B–D). VAT Tregs in HFD mice had a 56.9 ± 19.7% (n = 12) reduction in CCR4, and a 43.6 ± 12.9% (n = 6) reduction in the expression of GATA3. VAT Tconvs had a similar percentage reduction of GATA3, but no reduction in the expression of CCR4. Interestingly, HFD induced a large 84 ± 11.6% (n = 7) reduction of Nrp1 in VAT Tconvs, compared with 53.4 ± 34.8% (n = 7) in VAT Tregs. The expression of Th1 (TBET, CXCR3)- or Th17 (retinoic acid-related orphan receptor γt, CCR6)-associated proteins on VAT Tregs was not affected by HFD (Supplemental Fig. 1C).

Within the populations of ST2⁺ Treg and ST2⁺ Tconvs in the VAT, DIO significantly reduced the expression of CCR4 (Fig. 2E) and Nrp1 (Fig. 2G), but not GATA3 (Fig. 2F). The expression of Nrp1 in ST2⁻ Tregs and ST2⁻ Tconvs was also downregulated in HFD mice (Fig. 2G), indicating that both ST2⁺ and ST2⁻ cells were affected by DIO. In addition, although the overall expression of CCR4, GATA3, and Nrp1 in total VAT Tregs and/or Tconvs was decreased in HFD mice, ST2⁺ Tregs and Tconvs retained higher expression of these proteins compared with ST2⁻ cells (Supplemental Fig. 3).

Injection of IL-33 is known to expand peripheral ST2⁺ Tregs in a model of cardiac transplantation (18), so we asked whether in vivo administration of IL-33 could reverse the diet-induced reduction of ST2⁺ Tregs in VAT. Mice were fed a NCD or HFD for 16 wk, then injected i.p. with IL-33, or PBS as a control, for consecutive 10 d. Notably, 10 d of IL-33 treatment did not change the body weights of NCD or HFD mice (Supplemental Fig. 4A, 4B). Mice fed an HFD had a significant reduction in the proportion of Tregs, and injection of IL-33 completely reversed this effect, resulting in restoration of the normal proportion of Tregs (Fig. 3A). IL-33 also significantly increased the absolute number of Tregs per gram of VAT. However, IL-33 did not increase the numbers of Tconvs, and, contrary to a previous report (18), IL-33 injections did not expand Tregs in the spleen.

IL-33 administration also increased the proportion of ST2-expressing Tregs in the VAT by 3-fold in HFD mice, resulting in restoration of the normal population balance of VAT ST2⁺ Tregs found in lean NCD control mice (Fig. 3B). The absolute number of ST2⁺ Tregs per gram of VAT increased by 8-fold, whereas the number of ST2⁺ Tconvs was not significantly affected. GATA3 expression and Ki67 staining were also increased in VAT Tregs of HFD mice receiving IL-33 treatments (Fig. 3C, 3D). Injection of IL-33 also upregulated ST2 expression on VAT Tconvs by 4-fold, and on splenic Tregs by 2-fold, although the overall proportion of ST2⁺ cells in these populations always remained significantly lower than that of VAT Tregs.

We next asked whether the IL-33–induced restoration of ST2⁺ Tregs in the VAT of HFD mice was accompanied by a reduction in inflammation. As expected, compared with NCD mice, HFD mice had an increase in F4/80⁺ macrophages infiltrating the VAT, and this was almost completely reversed by IL-33 administration (Fig. 4A). By comparison, IL-33 did not significantly alter the proportion of CD4⁺, CD8⁺, or CD11c⁺ cells in the VAT (Supplemental Fig. 4C). RT-PCR analysis of the VAT revealed that IL-33 treatment of HFD mice also reduced the expression of Tnfα mRNA in the VAT (Fig. 4B), but did not significantly affect levels of Ccl2 or Il10 mRNA (Supplemental Fig. 4D). We also isolated the SVF from the VAT, which contains all VAT-resident immune cells, and measured the amount of IL-6 released in the culture media. SVF cells isolated from the VAT of HFD mice that had been injected with IL-33 produced significantly less IL-6 than did SVF cells isolated from HFD mice receiving PBS (Fig. 4C). Indeed, IL-33 administration was able to reduce TNF-α and IL-6 in the VAT to the level of expression found in lean NCD control mice, as indicated by the dashed lines (Fig. 4B, 4C).

To determine whether IL-33–induced resolution of VAT inflammation could also lead to an improvement of metabolism, mice fed an HFD for 16 wk were injected with IL-33 for consecutive 10 d to expand VAT Tregs and reduce VAT inflammation, followed by one injection every 2–3 d. At days 14 and 17, mice were fasted for 4–6 h, and their fasting blood glucose and serum insulin were measured, and the HOMA-IR, a mathematical equation used to estimate insulin resistance using fasting blood glucose and insulin levels (19), was calculated. There was no significant difference in body weights between PBS- and IL-33–treated mice before and after treatment (Fig. 5A); however, paired analysis revealed IL-33 injections resulted in a reduction of body weight after 14 and 17 d of treatment, whereas PBS injections did not (Fig. 5B). IL-33 significantly reduced fasting glycemia and amounts of plasma insulin (Fig. 5C, 5D). Consequently, the HOMA-IR was reduced to normal levels in mice receiving IL-33 treatment (Fig. 5E), indicating a significant improvement in insulin resistance.

A deficit in FOXP3⁺ Tregs in the VAT is thought to be a key factor in the development of adipose inflammation, resulting in the accumulation of proinflammatory macrophages and cytokines that promote insulin resistance (1, 3, 6, 20–23). In this work, we show that the VAT contains a large proportion of ST2⁺ Tregs that have a phenotype that is distinct from ST2⁻ Tregs and consistent with cells likely to have superior stability and IL-10–dependent suppressive function. IL-33 can be used therapeutically to completely reverse the DIO-associated reduction in the proportion of ST2⁺ Tregs, resulting in reduced VAT inflammation and improved metabolic health of obese mice. Thus, manipulation of ST2⁺ Tregs may be a new approach to control obesity-associated inflammation and insulin resistance.

Previous studies have shown that the VAT of lean mice is populated with a high proportion of Tregs that are phenotypically unique compared with peripheral Tregs due to high expression of PPARγ (6, 9) and IL-10 (7). Our study adds to this knowledge by showing that the majority of Tregs in lean VAT express ST2, and that they can be therapeutically expanded with IL-33. Thus, VAT Tregs may be uniquely sensitive to IL-33, possibly requiring continual IL-33 stimulation for their survival, stability, and/or function (11, 16, 18). A previous study reported that Tregs in the colon also express ST2 (11). Interestingly, we found that the proportion of ST2⁺ Tregs was significantly higher in lean VAT compared with Tregs isolated from the whole intestinal lamina propria, suggesting that the IL-33:ST2 axis could be relatively more influential in VAT compared with the intestine.

ST2⁺ Tregs as well as ST2⁺ Tconvs in lean VAT expressed increased levels of the Th2-related proteins GATA3 and CCR4 (24, 25). Similar to VAT Tregs, ST2 expression on colonic Tregs is also associated with GATA3 expression (11). As GATA3 promotes Treg stability and function (26, 27), VAT-resident ST2⁺ Tregs may be more stable and potent than their ST2⁻ counterparts. In support of this hypothesis, ST2⁺ Tregs also have enhanced expression of IL-10 compared with ST2⁻ Tregs. Because IL-10 produced by Tregs inhibits TNF-α production by macrophages (7), and IL-10 is well known to limit obesity-associated inflammation (28, 29) and TNF-α–induced insulin resistance (4), ST2⁺IL-10⁺ Tregs are most likely key regulators of immune responses in the VAT. Evidence that ST2⁺ Tregs also express increased levels of Nrp1, which is critical for the maintenance and function of Tregs (17), further supports the hypothesis that ST2⁺ Tregs possess superior stability and function. Although ST2⁺ Tregs are not more suppressive than ST2⁻ Tregs in in vitro suppression assays (16, 18), evidence that ST2-deficient Tregs have attenuated suppressive function in vivo (11) supports the notion that this subset may have unique functional properties.

In obese VAT, not only is the frequency of Tregs reduced, but also the proportion of the remaining Tregs expressing ST2 is severely diminished. As a result, there was an overall reduction in the expression of CCR4, GATA3, and Nrp1 in VAT Tregs. It remains unclear whether this reduction is due to the loss of ST2⁺ Tregs, selective expansion of ST2⁻ Tregs, and/or an intrinsic change in the phenotype of VAT-resident ST2⁺ Tregs. Because, as discussed above, the expression of ST2, GATA3, and Nrp1 have all been associated with Treg expansion, stability, and function, the phenotype of Tregs in obese VAT suggests these cells may be less potent. This possibility is supported by our previous finding that Tregs from obese VAT have diminished expression of IL-10 (7). Because the HFD and NCD were not micronutrient matched, we cannot exclude the possibility that there may also be obesity-independent, but diet-related effects.

Therapeutic IL-33 administration corrected the obesity-associated deficit in VAT ST2⁺ Tregs, suggesting that IL-33 is sufficient to maintain a pool of functionally specialized VAT-resident ST2⁺ Tregs. IL-33 increased both the proportion and absolute numbers of ST2⁺ Tregs, indicating that IL-33 induced the active expansion of ST2⁺ Tregs. In contrast, although IL-33 treatment increased the proportion of ST2⁺ Tconvs, their absolute number did not significantly change. Evidence that IL-33 reprograms DCs to selectively expand ST2⁺ Tregs (16), and the paucity of ST2⁺ Tregs that could respond to IL-33 in HFD mice raises the possibility that, at least initially, the effect of IL-33 on the expansion of ST2⁺ Tregs may be indirect. Contrary to a previous report (18), we found that IL-33 injections did not expand Tregs in the spleen, possibly because the 20-wk-old mice used in these experiments had a higher age-related (30) proportion of Tregs in the spleen than the younger mice reported by Turnquist et al. (18).

IL-33 therapy also reversed VAT inflammation in obese mice, reducing expression of TNF-α and IL-6 to the levels found in the VAT of healthy lean mice. Because macrophages express almost all of the TNF-α, and a significant amount of IL-6, in the VAT (29), changes in these cytokines most likely originate from these cells. Although the reversal of VAT inflammation was correlated with the expansion of ST2⁺ Tregs in the VAT, IL-33 could also have many other effects in the adipose tissue, including the activation of eosinophils (31) and type 2 innate lymphoid cells (ILC2)-mediated beiging of VAT (32). IL-33 also affects macrophages; however, it is unclear whether IL-33 promotes the polarization of alternatively activated macrophages (33), or enhances inflammatory cytokine production (34), and studies of the effects of IL-33 specifically on VAT-resident macrophages remain to be done. Thus, the IL-33–mediated reduction of inflammation is most likely due to a combination of anti-inflammatory effects of this cytokine on multiple cell types.

Both TNF-α and IL-6 have previously been shown to induce insulin resistance in adipocytes (35, 36), leading us to the hypothesis that IL-33 may also improve the metabolic parameters of obese mice. One hallmark of obesity is an overproduction of insulin, which at high levels can inhibit IL-10–dependent Treg function (7). IL-33–induced reduction of hyperinsulinemia may thus dampen the inhibitory effect that insulin exerts on Tregs in obesity. Using HOMA-IR to estimate the severity of insulin resistance (19), we found that IL-33 decreased the severity of insulin resistance in obese mice. Notably, we observed the metabolic effects of IL-33 at day 14, but not at day 10 (data not shown). Because IL-33 had already reversed VAT inflammation at day 10, the delayed improvement in insulin sensitivity may thus be subsequent to the effect of reduced inflammation. Longitudinal analysis of individual mice revealed that IL-33 treatment reduced body weight at day 14, coincident with improved metabolic parameters. Although the difference between the body weights of PBS versus IL-33–treated mice was not significant, alleviation of insulin resistance may be partly related to the reduction of body weight.

Our data are consistent with a recent report that the majority of VAT-resident Tregs express ST2, and IL-33 treatment promoted the increase of VAT Tregs, reduced VAT inflammation, and improved glucose tolerance of HFD mice (37). In addition, we show that DIO diminishes ST2⁺ Tregs and the expression of ST2-associated proteins such as CCR4, GATA3, and Nrp1. IL-33–mediated correction of the Treg deficit and VAT inflammation lasted as long as 25 d after initial treatment (data not shown), suggesting a prolonged therapeutic effect. IL-33 has also been reported to reduce genetically driven insulin resistance in leptin-deficient (ob/ob) mice (38). However, because ob/ob mice have an abnormal T cell compartment (39) and leptin has direct effects on T cells, including Treg proliferation (40), it is difficult to interpret these findings as they relate to T cells.

In summary, we show that more than half of the Tregs in normal VAT have a unique Th2-like and ST2⁺ phenotype and express markers associated with potent and activated Tregs (i.e., IL-10, Nrp-1). DIO severely diminishes the proportion of VAT Tregs with this phenotype, and this effect can be completely reversed by treatment with IL-33. Treatment with IL-33 also reverses the infiltration of macrophages into the VAT and the accumulation of proinflammatory cytokines. Subsequent to the reduction in inflammation, IL-33 reduces body weight, hyperinsulinemia, and insulin resistance, resulting in an anti-obesity effect that targets multiple aspects of DIO. Thus, IL-33–induced therapeutic expansion of ST2⁺ Tregs may be a new approach to control obesity-associated inflammation and insulin resistance.

We thank Lixin Xu for flow cytometry support.

The authors have no financial conflicts of interest.