a, Gating strategy to identify human ILCs. Stromal vascular fraction (SVF) cells from human abdominal subcutaneous white adipose tissues (WAT) were isolated and subjected to flow cytometric analyses. First plot pre-gated on singlets. Lineage cocktail 1 (Lin1): CD3, CD5, TCRαβ. Lineage cocktail 2 (Lin2): CD19, CD56, CD11c, CD16. ILCs are identified as Lin-negative cells that are CD25⁺ CD127⁺. Plots shown are from an obese donor.b–e, SVF cells from murine epididymal (E)-WAT were isolated and subjected to flow cytometric analyses. ILCs were defined as live CD45⁺ Lin⁻ CD25⁺ CD127⁺ cells. The lineage (Lin) cocktail included CD3, CD5, CD19, NK1.1, CD11c, CD11b and FcεRIα. Comparison of Lin⁻ CD25⁺ CD127⁺ cells in E-WAT ofb,Id2^+/+ versusId2^−/− bone marrow chimaeras,c,Tcf7^+/+ versusTcf7^−/− mice andd,Rag2^−/− versusRag2^−/− γ_c^−/− mice.n = 3–8 mice per group from 2 independent experiments.e, E-WAT SVF cells from C57BL/6 mice were treated with PMA (100 ng ml⁻¹) and ionomycin (1 μg ml⁻¹) in the presence of Brefeldin A (10 μg ml⁻¹) for 4 h and stained for ILCs. Live CD45⁺ Lin⁻ CD25⁺ CD127⁺ cells were pre-gated, and IL-5 and IL-13 protein levels were assessed. Plot shown is representative ofn = 12 mice from 3 independent experiments.f, Human WAT ILC2 frequencies were compared in the 7 youngest donors (36.0 ± 3.5 years old) versus the 7 oldest donors (55.9 ± 1.9 years old).g, Human WAT ILC2 frequencies in female non-obese donors with body mass index (BMI) < 30.0 kg m⁻² versus female obese donors with BMI ≥ 30.0 kg m⁻². Student’st-test. **P < 0.01, ***P < 0.001. Data are shown as mean ± standard error and are representative of 2 independent experiments. Sample sizes are biological replicates.

Il33^+/+ (n = 6) orIl33^−/− (n = 5) mice were fed a low-fat diet (10% kcal fat) for 12 weeks starting at 7 weeks of age.a, Representative plots and frequencies of live CD45⁺ Lin⁻ CD25⁺ IL-33R⁺ ILC2s in epididymal (E)-WAT (data are fromFig. 2a) and iWAT. Plots pre-gated on CD45⁺ Lin⁻ cells that lack CD3, CD5, CD19, NK1.1, CD11c, CD11b and FcεRIα.b, Frequencies of IL-5⁺ IL-13⁻ and IL-5⁺ IL-13⁺ ILC2s in E-WAT and iWAT of wild-type and IL-33-deficient mice. E-WAT stromal vascular fraction cells were treated with PMA (100 ng ml⁻¹) and ionomycin (1 μg ml⁻¹) in the presence of brefeldin A (10 μg ml⁻¹) for 4 h before staining for ILC2s and intracellular cytokines. Pre-gated on CD45⁺ Lin⁻ CD25⁺ IL33R⁺ ILC2s.c, Inguinal white adipose tissue (iWAT) sections were haematoxylin and eosin stained and imaged at ×40 magnification. Adipocyte area was calculated from 25–40 adipocytes total from 2–3 images per mouse.d, 16-h fasting blood glucose concentrations.e, 16-h fasting serum insulin concentrations.f, Homeostatic model assessment of insulin resistance (HOMA-IR) index values.g, Glucose tolerance test (GTT) with 2 g per kg body weight glucose following a 16-h fast.h, Insulin tolerance test (ITT) with 0.5 U per kg body weight insulin following a 5-h fast. For panelsa–f, groups were compared using Student’st-test, *P < 0.05, ***P < 0.001. For panelsg–h, a two-way ANOVA with repeated measures was performed followed by Tukey post-hoc test. *P < 0.05, **P < 0.01, ***P < 0.001. Data shown are from a single cohort and are representative of 2 independent experiment. Sample sizes are biological replicates.

Male C57BL/6 mice were placed on a control diet (CD) or HFD (60% kcal fat) at age 8 weeks. On the first day of feeding, CD mice were treated with PBS and HFD mice were treated with PBS or recombinant murine (rm)IL-33 (12.5 μg per kg body weight) once every 4 days by intraperitoneal injection for 4 weeks.a, E-WAT ILC2 numbers per gram of adipose,b, body weight,c, relative E-WAT weight andd, relative iWAT weight at week 4.e, 16-h fasting blood glucose concentrations andf, glucose tolerance testing during week 3.g, Frequencies and representative plots of E-WAT T_regs defined as live CD45⁺ CD3⁺ CD4⁺ Foxp3⁺ cells. Plots are gated on live CD45⁺ CD3⁺ CD4⁺ cells, and numbers are the percentage of CD4⁺ T cells that are Foxp3⁺ T_regs.h, Numbers of T_reg cells per gram of adipose. All panels includen = 10 mice per group from 2 independent cohorts, except panel A which includesn = 16 CD PBS andn = 18 HFD PBS from 4 independent cohorts.a–e, One-way ANOVA with Tukey post-hoc test, *P < 0.05, **P < 0.01, ***P < 0.001.f, Two-way ANOVA with repeated measures, ***P < 0.001 comparing CD PBS versus HFD PBS,∧∧∧P < 0.001 comparing HFD PBS versus HFD IL-33. Data are shown as mean ± standard error. Sample sizes are biological replicates.

a, Male C57BL/6 mice were treated with PBS or recombinant murine (rm)IL-33 (12.5 μg per kg body weight) daily for 7 days (PBSn = 10, rmIL-33,n = 12). Over a 24 h period between days 6 and 7, food intake and ambulatory activity were measured over 15-min intervals. The average difference in food intake or ambulatory activity between PBS- and rmIL-33-treated mice was calculated for each 15-min interval, and the differences in food intake and ambulatory activity were related by linear regression. Solid line, best-fit line. Dashed curves, upper and lower 95% confidence intervals around the best-fit line. Data are shown as mean differences for each interval and are representative of 2 independent experiments.b–d, Male C57BL/6 mice were treated with PBS or recombinant murine (rm)IL-33 (12.5 μg per kg body weight) once and monitored for the first 3 h post-treatment using CLAMS cages (n = 4 per group).b, Energy expenditure,c, food intake andd, ambulatory activity (beam breaks) were measured over of the first 3 h post-treatment. Student’st-test. ***P < 0.001. Data are shown as mean ± standard error and are representative of 1 independent experiment. Sample sizes are biological replicates.

C57BL/6 male mice (10 weeks old) were treated with PBS (n = 8) or IL-33 (12.5 μg per kg body weight,n = 8) daily by intraperitoneal injection for 7 days.a, Representative plots and frequencies of Lin⁻ CD25⁺ IL-33R⁺ ILC2s in interscapular BAT. Gated on live CD45⁺ Lin⁻ cells.b, Numbers of ILC2s per gram of BAT.c,Ucp1 expression in BAT by real-time PCR.d, UCP1 immunohistochemistry of BAT at ×10 magnification. Scale bars, 100 μm.e, ×40 magnification ofd. Scale bars, 100 μm. Student’st-test, *P < 0.05, ***P < 0.001. Data are shown as mean ± standard error and are representative of 2 independent experiments. Sample sizes are biological replicates.

a, Experimental design for panelsa,b. Live CD45⁺ Lin⁻ CD25⁺ IL-33R⁺ ILC2s were sort-purified from E-WAT of CD45.1⁺ mice treated with 12.5 μg per kg body weight recombinant murine (rm)IL-33 daily for 7 days by intraperitoneal injection. PBS (n = 8) or ILC2s (1 × 10⁵ total,n = 8) were transferred to CD45.2⁺ recipient mice daily for 4 days by subcutaneous injection near iWAT (5 × 10⁴ ILC2s, split evenly bilaterally) and intraperitoneal injection (5 × 10⁴ ILC2s). Tissues were harvested on day 5 for analyses.b, Donor and recipient ILC2s in E-WAT, brown adipose tissue (BAT), mesenteric lymph nodes (mLN) and lung. iWAT ILC2 plots from this experiment are shown in mainFig. 3g. Pre-gated on Live CD45⁺ Lin⁻ CD25⁺ IL-33R⁺ ILC2s. Donor ILC2s are defined as CD45.1⁺ CD45.2⁻, whereas recipient ILC2s are defined as CD45.1⁻ CD45.2⁺. Representative plots shown. Frequencies represent percent of ILC2s that are recipient or donor cells. Student’st-test, ***P < 0.001.c, Experimental design for panelsc–e. Sort-purified CD45.1⁺ ILC2s (×10⁵) from E-WAT of IL-33-treated mice (as described above) were transferred intoRag2^−/− γc^−/− recipients by a single intraperitoneal injection. ILC2-sufficientRag2^−/− mice, ILC2-deficientRag2^−/− γc^−/− mice and ILC2-reconstitutedRag2^−/− γc^−/− mice were treated with PBS or rmIL-33 (12.5 μg per kg body weight) by intraperitoneal injection daily for 7 days. There weren = 4 mice per group. This experimental design corresponds to mainFig. 3l-m.d, Representative plots of live CD45.1⁺ Lin⁻ CD25⁺ IL33R⁺ ILC2s in E-WAT. Blue, recipient cells. Red, donor cells. Lineage cocktail includes CD3, CD5, CD19, NK1.1, CD11c, CD11b and FcεRIα.e, iWAT UCP1 IHC. Scale bars, 100 μm. ANOVA with Tukey post-hoc test, ***P < 0.001. Data are shown as mean ± standard error and are representative of 2 independent experiments. Sample sizes are biological replicates.

a–f, Wild-type (Balb/c),DblGata1 mice that lack eosinophils orIl4ra^−/− mice that have dysregulated alternatively activated macrophages (AAMacs) (both mutant strains on a Balb/c background) were treated with PBS or recombinant murine (rm)IL-33 (12.5 μg per kg body weight) daily by intraperitoneal injection for 7 days.a, iWAT ILC2 numbers per gram of adipose andb, iWAT UCP1 immunohistochemistry (IHC) in Balb/c mice (PBS,n = 4; rmIL-33,n = 3).c, iWAT ILC2 numbers per gram of adipose andd, iWAT UCP1 IHC inDblGata1 mice (PBS,n = 5; rmIL-33,n = 6).e, iWAT ILC2 numbers per gram of adipose andf, iWAT UCP1 IHC inIl4ra^−/− mice (PBS,n = 4; rmIL-33,n = 6).g,h, Live CD45⁺ Lin⁻ CD25⁺ IL-33R⁺ ILC2s were sort-purified from E-WAT of C57BL/6 mice treated with rmIL-33 (12.5 μg per kg body weight) daily for 5–7 days by intraperitoneal injection toRag1^−/− mice on a C57BL/6 background. ILC2s (1 × 10⁵ total) were transferred to recipient mice daily for 4 days by subcutaneous injection (PBS,n = 8; ILC2,n = 8).g, iWAT ILC2 numbers per gram of adipose andh, iWAT UCP1 IHC.i–l, Live CD45⁺ Lin⁻ CD25⁺ IL-33R⁺ ILC2s were sort-purified from E-WAT of Balb/c mice treated with rmIL-33 (12.5 μg per kg body weight) daily for 5–7 days by intraperitoneal injection. ILC2s (1 × 10⁵ total) were transferred to recipient mice daily for 4 days by subcutaneous injection.i, iWAT ILC2 numbers per gram of adipose andj, iWAT UCP1 IHC inDblGata1 recipients (PBS,n = 6; ILC2,n = 6).k, iWAT ILC2 numbers per gram of adipose andl, iWAT UCP1 IHC inIl4ra^−/− recipients (PBS,n = 3; ILC2,n = 4). Scale bars, 100 μm. Student’st-test, *P < 0.05. Data are shown as mean ± standard error and are representative of 2 independent experiments. Sample sizes are biological replicates.

Interleukin (IL)-33 acts on group 2 innate lymphoid cells (ILC2s) to upregulate production of the effector molecules IL-5, IL-13 and enkephalin peptides. ILC2-derived IL-5 promotes eosinophil homeostasis in WAT, and eosinophils in turn produce IL-4 to sustain alternatively activated macrophages (AAMacs) in WAT. ILC2-derived IL-13 can also promote AAMac responses. In the setting of chronic exposure to a cold environment, eosinophil-derived IL-4 stimulates AAMacs to produce catecholamines such as noradrenaline, which acts directly on beige adipocytes to upregulate uncoupling protein 1 (UCP1) expression and promote mitochondrial biogenesis. Although it remains unknown whether ILC2-derived IL-5 and IL-13 contribute to cold-stress-induced beiging, ILC2-derived enkephalin peptides can act directly on beige adipocytes to upregulate UCP1 and promote beiging. This results in increased energy expenditure and decreased adiposity that may counteract weight gain. In the setting of obesity, IL-33 expression in WAT is increased; however, WAT ILC2s are paradoxically decreased in both mice and humans, suggesting that the IL-33/ILC2 axis is dysregulated in obesity. This may impede the ability of ILC2s to contribute to the function of beige fat, resulting in a vicious cycle that promotes weight gain.