The ATP-P2X₇ Signaling Axis Is Dispensable for Obesity-Associated Inflammasome Activation in Adipose Tissue

Wild-type (WT) C57/B6 (CD45.2⁺; 000664), B6.129P2-P2rx7^tm1Gab/J (P2X₇^−/−; 005576), B6.129P2-P2ry2^tm1Bhk/J (P2Y₂^−/−; 009132), B6.SJL-Ptprc^a Pepc^b/BoyJ (CD45.1⁺; 002014), and B6.V-Lep^ob/J (ob/ob; 000632) mice were purchased from The Jackson Laboratory and bred in our facility. The P2X₇^−/− and P2Y₂^−/− mice had been back-crossed to the B6 background for 7 and 12 generations, respectively. WAT tissue lysates from 16-week HFD caspase-1^−/− mice were provided by Dr. Rinke Stienstra (Wageningen University). Mice were housed in our pathogen-free facility with 13% low-fat diet (LFD), with 13% fat, 67% carbohydrate, and 20% protein from Harlan Teklad (2914), and at 6 weeks of age were placed on 60% HFD, with 60% fat, 20% carbohydrate, and 20% protein from Research Diets Inc. (D12492). All animal procedures were approved by the Cornell Institutional Animal Care and Use Committee.

Six-week-old male mice were fed with either 13% LFD or 60% HFD for the indicated period of time from 1 to 29 weeks. For GTT, mice were fasted for 16–18 h followed by an injection of glucose (Sigma) at 1 g/kg body weight (BW). For ITT, mice were fasted for 5 h followed by an injection of insulin (Sigma) at 36 µg/kg BW. Blood glucose was monitored using One-Touch Ultra (LifeScan) or AlphaTRAK Glucometer (Abbott). Fasting insulin levels were measured after a 4-h fast. Mice were killed by cervical dislocation, and tissues were either fixed for histology or snap frozen in liquid nitrogen for quantitative PCR (Q-PCR) and Western blot analyses. Frozen tissues were stored at −80°C.

Eight-week-old WT or P2X₇^−/− mice were injected intraperitoneally with LPS at 2 ng/g BW (L4130; Sigma) 1.5 h prior to the ATP challenge intraperitoneally (50 mM ATP, adjusted to pH 7 with NaOH; Sigma) at 10 µL/g BW. One hour later, mice were killed and blood serum was collected.

Bone marrow was obtained by flushing the femurs from sex-matched donor mice with PBS. Eight-week-old recipient male mice were lethally irradiated (10 Gy) and transplanted with 5 million bone marrow cells via intravenous injection. One week prior to and 2 weeks after bone marrow transplantation (BMT), 2 mg/mL neomycin (Cellgro) was added to drinking water. After 4 weeks recovery, recipients were placed on 60% HFD for 16 weeks prior to metabolic phenotyping.

Epididymal fat pads were collected and minced with razor blades. After two washes with KRBH buffer (10 mmol/L HEPES, pH 7.4, 15 mmol/L NaHCO₃, 120 mmol/L NaCl, 4 mmol/L KH₂PO₄, 1 mmol/L MgSO₄, 1 mmol/L CaCl₂, and 2 mmol/L sodium pyruvate) supplemented with 200 nM adenosine, 5 mM glucose, and 2% BSA, fat pads were digested with 1.5 mg/mL of type II collagenase (for adipocytes, C6885; Sigma) in KRBH buffer for 45 min with gentle shaking at 37°C. Cell suspension was filtered through a 300-μm nylon filter (Small Parts, Inc.) and then spun at 2,000 rpm for 5 min to separate floating adipocytes from the pellets termed stromal vascular cells (SVCs). Erythrocytes in the SVCs were lysed using the cell lysis buffer (BD Biosciences) prior to collection. Floating primary adipocytes were washed with KRBH buffer and spun twice more to remove the contaminating immune cells. Both SVCs and primary adipocytes were snap frozen for RNA extraction.

Mice were injected intraperitoneally with 2 mL 4% thioglycollate (211716; BD Biosciences). Four days later, mice were killed and macrophages were harvested by flushing the peritoneum with cold PBS twice and plated at 2 × 10⁶ cells/mL in Dulbecco’s modified Eagle’s medium (Cellgro) supplemented with 10% FBS (Sigma) and 1% penicillin/streptomycin (Cellgro). Peritoneal macrophages were treated with 0.1 µg/mL LPS (L4391; Sigma) for 5 h followed by the addition of 0.5 or 5 mmol/L ATP (Sigma) for 30 min followed by protein extraction.

Adipose tissues were fixed in 4% formaldehyde, embedded in paraffin, and sectioned by the Cornell Histology Core Facility. Pictures were taken using the Axiovert 200M microscope (Zeiss).

Blood was collected in animals upon 4 h fasting during the day. Circulating insulin levels were measured using the kit from Millipore. IL-1β level in circulation and adipose tissue was measured using the ELISA kit from eBioscience per supplier’s protocol. For tissue ELISA, a piece of WAT was lysed in 1 mL Tris-based lysis buffer by homogenization and IL-1β level was normalized to total WAT weight.

RNA extraction from cells and tissues and Q-PCR were carried out as previously described (32) using Trizol plus QIAeasy kit (Qiagen) for adipocytes and adipose tissues with DNase digestion (Roche). Q-PCR data collected on the Roche LightCycler 480 were normalized to ribosomal l32 gene in the corresponding sample. For RT-PCR, the equivalent amount of RNA was reverse transcribed and amplified by 30 PCR cycles (95°C for 15 s, 55–58°C for 15 s, and 72°C for 20 s). The PCR products were separated on a 1.5% agarose gel and visualized using Bio-Rad ChemiDoc XRS+ system after exposure. Supplementary Table 1 lists primer sequences used in this study.

Tissues or cells were lysed in Tris-based lysis buffer containing 1% Triton X-100. Total lysates (10–20 µg) were used in a mini–SDS-PAGE as previously described (32). Antibodies specific for p-S473 serine/threonine protein kinase AKT (9271S; Cell Signaling), AKT (sc-1618; Santa Cruz), inhibitor of nuclear factor-κB (IκB) (9242; Cell Signaling), caspase-1 p10 (sc-514; Santa Cruz), caspase-3 (9665; Cell Signaling), and arginase-1 (Arg1) (sc-18351; Santa Cruz) were used at 1:1,000–2,000, and for the loading control, heat shock protein 90 (HSP90) (sc-7947; Santa Cruz) was used at 1:6,000. Caspase-1 p20 antibody clone 4B4 (a gift from Dr. Vishva Dixit, Genentech, Inc., San Francisco, CA) was used at 1:400. The secondary antibody goat anti-rabbit IgG horseradish peroxidase (HRP) and goat anti-mouse IgG HRP (1:10,000) were from Bio-Rad, and donkey anti-goat IgG HRP (1:10,000) were from Jackson ImmunoResearch. Quantification of band density was carried out using the ImageLab software of the Bio-Rad ChemiDoc XRS+ system after exposure.

Microarray analyses of WAT were performed as previously described (33) with four groups (n = 3–4 mice each): age- and sex-matched WT and P2X₇^−/− mice under either LFD or 12 weeks of HFD. Genes were filtered according to expression value >20 in greater than two arrays with a minimum of 10 probe sets per gene. This resulted in 16,001 genes that were used to generate the scatter plot. Thereafter, genes were selected that were differentially expressed between HFD and LFD using q < 0.001 and fold change >1.5, resulting in 3,097 genes that were used to generate the heat map. Data have been deposited into the GEO datasets (GSE36033).

Results are expressed as mean ± SEM. Comparisons between groups were made using either unpaired, two-tailed Student t test of the EXCEL software for two-group comparisons or the two-way ANOVA test with the Bonferroni post-tests of the PRISM software for multigroup comparisons. P < 0.05 was considered statistically significant.

HFD feeding with 60% calories from fat progressively increased BW, gonadal WAT weight, fasting blood glucose, and insulin levels (Fig. 1A). Upon 8 weeks of feeding of 60% HFD, mice developed mild hyperglycemia and hyperinsulinemia. RT-PCR analysis revealed that components of NLRP3 inflammasome such as Nlrp3, Casp1, Asc, Il1b, and Il18 were highly expressed in mouse epididymal adipose tissues, whereas other types of inflammasome sensors such as Nlrp1, Nlrc4, and Aim2 were expressed at much lower levels (Fig. 1B). Furthermore, most NLRP3 inflammasome components including Nlrp3, Casp1, Asc, and Il18 were expressed at comparable levels in both adipocytes and SVCs in WAT, whereas Il1b was predominantly present in the SVC fraction of WAT (Fig. 1C).

Intriguingly, most NLRP3 inflammasome components, including Nlrp3, Asc, and Casp1, were highly responsive to HFD feeding and elevated starting at 6 weeks HFD (Fig. 1C), as were other types of inflammasome sensors such as Nlrp1 and Nlrc4 (Supplementary Fig. 1). Interestingly, the Il1b mRNA level was not changed whereas Il18 was decreased with HFD (Fig. 1C). In support, at the protein levels, procaspase-1 p45 was increased fivefold after 8 weeks HFD (Fig. 1D and E). The cleaved fragment of caspase-1 p35, a marker for caspase-1 activation (13,16,34,35), exhibited a similar expression pattern (Fig. 1D and E). The specificity of the caspase-1 antibody was confirmed using adipose tissue from caspase-1^−/− animals (Fig. 1F). Consistently, the IL-1β level in epididymal adipose tissue was significantly increased after 12 weeks, but not 6 weeks, HFD (Fig. 1G). Intriguingly, at 8 weeks, not 6 weeks, HFD, cell death in WAT was significantly increased, as measured by levels of cleaved caspase-3 (Fig. 1H and I), as well as the crown-like structures (Supplementary Fig. 2). Thus, we conclude that inflammasome activation in adipose tissue coincides with increased cell death and the onset of hyperglycemia and hyperinsulinemia.

Early studies have implicated a significant role of ATP-P2X₇ signaling in inflammasome activation in macrophages and dendritic cells and in the clearance of dead cells (18,21,26). However, whether P2X₇ plays a role in adipose tissue remains unclear. RT-PCR analyses revealed that among all P2X family members P2X_1–7, P2X₇ was expressed at relatively high levels in both adipocytes and SVCs of mouse epididymal adipose tissue (Fig. 2A). Both mRNA and protein levels of P2X₇ were significantly increased in the WAT of 13-week-old ob/ob mice relative to age-matched WT lean mice (Fig. 2B and C). Similar observation was obtained in the liver but not muscle or pancreas (Fig. 2D).

We next asked when the induction of P2X₇ occurred in adipose tissue using HFD-feeding mouse models. Intriguingly, induction of P2X₇ protein also occurred at 8 weeks HFD feeding (Fig. 2E and F), a pattern strikingly similar to that of inflammasome activation. Preceding the induction of its protein, P2X₇ mRNA level was elevated starting at 6 weeks HFD (Fig. 2G). Thus, P2X₇ protein is induced in adipose tissue after 8 weeks feeding of 60% HFD, coinciding with the activation of inflammasome and the onset of hyperglycemia and hyperinsulinemia.

We next addressed whether the ATP-P2X₇ signaling axis plays a role in metabolic regulation in HFD-induced obesity by placing P2X₇^−/− mice on 60% HFD feeding for 12 weeks. Obese P2X₇^−/− mice exhibited similar body and gonadal fat weights relative to the WT cohort (Fig. 3A and B). There were no gross morphological differences in WAT, liver, and pancreas tissues between knockout (KO) and WT cohorts (Fig. 3C and Supplementary Fig. 3). Systemic glucose homeostasis and insulin sensitivity were not affected by P2X₇ deficiency as shown by GTT, fasting insulin levels, and AKT phosphorylation, respectively (Fig. 3D–F).

Given the known role of P2X₇ in ATP-induced inflammasome activation, we next addressed whether the ATP-P2X₇ signaling axis is required for inflammasome activation in adipose tissue upon 12 weeks HFD. Unexpectedly, levels of caspase-1 and its cleavage product p35 were not affected in P2X₇^−/− WAT (Fig. 4A and B), nor was the IL-1β level in WAT (Fig. 4C) or mRNA levels of key inflammasome components (Fig. 4D), suggesting that P2X₇ may be not involved in inflammasome activation in adipose tissue during diet-induced obesity. In direct contrast, P2X₇-deficient peritoneal macrophages and mice were not able to respond to LPS plus ATP both in vitro (Fig. 4E) and in vivo (Fig. 4F), further supporting the notion that P2X₇ is the only purinergic receptor that activates inflammasome in response to ATP stimulation. Consistently, protein levels of inflammatory markers such as the inhibitor of nuclear factor-κB (IκB) and Arg1 (36,37) in WAT were not affected by P2X₇ deficiency (Fig. 4A). In support, the expression of several key pro- and anti-inflammatory genes was not affected by P2X₇ deficiency (Fig. 4G). This was further confirmed using an unbiased cDNA microarray analysis of WAT from WT or P2X₇^−/− mice under LFD and 12 weeks HFD where the HFD effect was almost identical between two groups. As shown in both scatter plots and heat map, the loss of P2X₇ had no significant effect on the expression of genes that were regulated by HFD (Fig. 4H and Supplementary Fig. 4A), including many inflammatory genes (not shown). Taken together, these data suggest that the ATP-P2X₇ signaling axis is dispensable for obesity-associated inflammasome activation in WAT.

In addition to P2X₇, several other P2 receptors, including P2X_1–6 and P2Y_{1–2, 4, 11, 12}, are capable of sensing extracellular ATP but are not associated with inflammasome activation (38). Loss of P2X₇ did not lead to compensatory increases of other P2 receptors in adipose tissue (Supplementary Fig. 4B). Among the P2 family members, P2Y₂, a G protein–coupled receptor that has been implicated in phagocyte migration in response to extracellular nucleotides released by apoptotic cells (26), is highly expressed in adipose tissue (Supplementary Fig. 4C). To further exclude a possible role of P2Y₂ in vivo, we characterized the metabolic phenotypes and inflammatory status in WAT of P2Y₂^−/− mice after 15 weeks HFD (Supplementary Fig. 5). We failed to observe any alternations in metabolic phenotypes (Supplementary Fig. 5A–E), inflammatory responses (Supplementary Fig. 5F and G), or inflammasome activation (Supplementary Fig. 5G and H) in adipose tissue compared with age-matched WT controls. These results excluded the involvement of another ATP receptor, P2Y₂, in inflammasome activation in obese WAT.

To exclude a possible role of P2X₇ in somatic cells, such as pancreatic β-cells (39) and adipocytes (40), that may affect the outcome of inflammasome activation in adipose tissue, we next specifically ablated P2X₇ in hematopoietic cells using BMT. Preliminary BMT experiments using congenic CD45.1⁺ and CD45.2⁺ mice showed efficient immune cell reconstitution in SVCs of WAT (Supplementary Fig. 6). Next, lethally irradiated WT mice were reconstituted with bone marrow from P2X₇^−/− or WT mice. Metabolic phenotypes and inflammasome status were examined after 16 weeks HFD after 4 weeks recovery (Fig. 5A). The loss of P2X₇ in hematopoietic cells did not affect BW gain, fat weight, and adipose tissue morphology (Fig. 5B–D). Liver and pancreas tissue morphologies were also comparable between experimental and control cohorts (Supplementary Fig. 7). Systemic glucose tolerance and fasting insulin levels were not affected by hematopoietic-specific P2X₇ ablation (Fig. 5E and F). Moreover, mice with P2X₇ deficiency in hematopoietic cells exhibited no alterations in inflammasome activity as shown by the protein levels of caspase-1 p45 and p35 (Fig. 6A and B) as well as mRNA levels of three key inflammasome components (Fig. 6C). Additionally, the overall inflammation level in WAT was not altered, as shown by protein levels of IκB (Fig. 6A) and mRNA levels of various inflammatory genes (Fig. 6D). Thus, these data suggest that hematopoietic P2X₇ is not required for inflammasome activation in obese adipose tissue.

Previous studies have shown that inflammasome is activated in obese adipose tissue either after 36 weeks 60% HFD or 12 weeks 45% HFD feeding (13,15), and it remains unclear how long it takes for HFD feeding to activate inflammasome in adipose tissue. Here, we report a dramatic activation of inflammasome in adipose tissue starting at 8 weeks feeding of 60% HFD, coinciding with the onset of hyperglycemia and hyperinsulinemia and with cell death. This is consistent with the general timeline of inflammation during diet-induced obesity, which was reported to be up-regulated at 9 weeks in adipose tissue based on the infiltration of macrophages with 60% HFD and the global induction of inflammatory genes in microarray analysis with 45% HFD (28,41). In parallel with the onset of inflammation, cell death in adipose tissue was elevated at about 4–8 weeks 60% HFD and peaked at about 12 weeks HFD (Fig. 1H and I and Supplementary Fig. 2) (28). These tight correlations among HFD feeding, cell death, and inflammatory response strongly suggest that dying cells in obese adipose tissue may release cell death–mediated “danger signals” to activate inflammasome in situ. However, our data demonstrate that extracellular ATP-P2X₇ signaling is not involved in inflammasome activation in obese adipose tissue in vivo. This observation is in line with a recent study showing that adipocyte cell death is not sufficient to drive inflammation (30).

We observed that P2X₇, an inflammasome-activating ATP receptor in several disease models (17–21,23,26), was highly induced at 8 weeks HFD, coinciding with the inflammasome activation in obese adipose tissue. In line with our observation, a recent study showed that P2X₇ expression is increased in adipose tissue of human patients with metabolic syndrome (42). However, our data demonstrated that, although required for LPS-ATP–induced inflammasome activation both in vitro and in vivo (Fig. 4E and F), P2X₇ is dispensable for inflammasome activation in obese adipose tissue in vivo. One possible explanation is that the extent of cell death in adipose tissue during obesity may be much lower than the systems used in previous studies, such as drug-induced tumor cell death, host-verses-graft disease, and sterile inflammasome (12–15). Additionally, ATP may be quickly hydrolyzed in extracellular space (22,43). Therefore, as P2X₇ receptor requires a high ATP level to be activated (∼5 mM) (Fig. 4E) (43), extracellular ATP in obese adipose tissue may not reach the activation threshold for P2X₇ and thus fail to activate inflammasome in vivo. This possibility is supported by a study showing that CD39^−/− mice, which are incapable of hydrolyzing extracellular ATP, exhibit exacerbated insulin resistance (44). Although the inflammatory status was not examined in the CD39^−/− mice, this report suggests that extracellular ATP may have a metabolic effect when it reaches a certain level.

Though our data suggest that extracellular ATP may not be the “danger signal” for inflammasome activation in WAT during obesity, it does not exclude the potential importance of cell stress and death in obese adipose tissue during the pathogenesis of obesity. Lipid spillover during adipocyte expansion and death has been linked to macrophage infiltration and obesity-associated inflammation (45). Indeed, several studies using lipodystrophy mouse models suggest that adipocyte cell death is important for macrophage infiltration into adipose tissue (46,47). Furthermore, both palmitate and ceramides, but not unsaturated oleate, have been reported to activate NLRP3 inflammasome in bone marrow–derived macrophages in vitro (14,15). As saturated fatty acid has been demonstrated to induce ceramide biosynthesis in vivo (48), it remains to be tested whether palmitate and ceramides activate inflammasome in adipose tissue via the same mechanism. Thus, the function of lipids as inflammasome-activating signals in obese adipose tissue and the underlying molecular mechanism are worth further studies.

Alternatively, inflammasome in obese adipose tissue may be activated by mechanisms unrelated to in situ cell death. A high glucose level has been suggested to induce inflammasome activation. In isolated pancreatic islets, a high glucose level was shown to induce thioredoxin interacting protein (TXNIP) expression and then activate inflammasome, which was not mediated by ATP-P2X₇ signaling (12). In another independent study, similar induction of TXNIP and IL-1β by glucose was observed in adipose tissue explants and primary adipocytes (16). These reports indicate that hyperglycemia during obesity may lead to inflammasome activation. Interestingly, we observed that the onset of both hyperglycemia and inflammasome activation in adipose tissue occurred at 8 weeks HFD, whereas 6 weeks HFD feeding was not sufficient. Further studies are needed to show the potential role of high circulating glucose levels in inflammasome activation in adipose tissue in vivo.

To date, four classes of inflammasomes have been shown to activate caspase-1 in vivo, including NLRP3, NLRP1, NLRC4, and AIM2 inflammasome (10). Different classes of inflammasomes may respond to specific environmental cues. For example, the NLRC4 inflammasome is activated by gram-negative bacteria, the AIM2 inflammasome senses cytosolic double-stranded DNA, and the NLRP3 inflammasome responds to a number of pathogen-associated or endogenous danger–associated activators (10). Previous studies on inflammasomes in adipose tissue focused on NLRP3 inflammasome (13,15). However, the nature of inflammasomes in adipose tissue remains unclear and their cell-type specificity has not been examined. Here, we determined the expression of the four classes of inflammasomes in epididymal adipose tissue and liver by assessing the relative transcript levels of their sensors. Besides NLRP3, our data show that both NLRP1 and NLRC4 were also expressed in adipose tissue and were highly induced after 8 weeks HFD. In contrast, AIM2 expression was very low in adipose tissue. Thus, the role of non-NLRP3 inflammasomes in adipose inflammation in obesity requires further investigation.

Inflammasome activation generates two active caspase-1 subunits, p20 and p10, with enzymatic activities (10). In addition, as first reported by Tschopp and colleagues (34) in a cell-free system, the caspase-1 intermediate product p35 may also serve as an indicator for the activation of caspase-1. Recently, several studies by different groups have used p35 as a marker of inflammasome activation in tissues (13,16,35). Indeed, our data show that the p35 band is caspase-1 specific as it disappears in the caspase-1–deficient adipose tissue (Fig. 1F), and more importantly, that the appearance of caspase-1 p35 subunit in WAT under various physiological settings is in line with that of caspase-1 p20 (Fig. 1D and F, Fig. 4A, and Fig. 6A vs. Supplementary Fig. 8). Thus, the caspase-1 p35 fragment, albeit possibly not the active form, may be a bona fide indicator of inflammasome status in WAT.

P2X₇ was previously reported to regulate pancreatic β-cell functions by modulating insulin secretion and β-cell mass (39). Upon feeding with high-fat/high-sucrose Surwit diet (58% fat with majority of saturated fat, 26% carbohydrate, and 16% protein, no fiber) for 12 weeks, P2X₇^−/− mice developed hyperglycemia, severe glucose intolerance, decreased β-cell mass, and impaired β-cell function (39). Such β-cell–associated diabetic phenotype, however, was not observed in our study (Fig. 3 and Supplementary Figs. 3 and 7). The discrepancies may be due to specific diets used in the studies. In addition, the ATP-P2X₇ signaling axis has been reported to induce lipolysis and inhibit insulin signaling in adipocytes in vitro by upregulating cyclic AMP levels or inhibiting phosphorylation on insulin receptor substrate and AKT in rat primary adipocytes (40,49,50). In contrast, others have reported that insulin signaling was downregulated by ATP-P2X₇ in astrocytes (51,52). Here in our study, animals with both global and hematopoietic-specific deletion of P2X₇ did not exhibit altered systemic insulin sensitivity and adiposity, suggesting a dispensable role of ATP-P2X₇ signaling in regulating adipocyte function and insulin sensitivity in vivo.

In conclusion, our study showed that inflammasome is activated concurrently with the development of hyperglycemia and hyperinsulinemia. Unexpectedly, the ATP-P2X₇ signaling axis seems not to play a critical role in this process. Further studies are required to identify and characterize the endogenous signal(s) that is responsible for inflammasome activation in adipose tissue during obesity.

S.S. is supported by the International Student Research Fellowship from the Howard Hughes Medical Institute. This study was also funded by the Netherlands Nutrigenomics Centre (to S.K.), Cornell startup package, American Diabetes Association (7-08-JF-47 and 1-12-CD-04), and National Institute of Diabetes and Digestive and Kidney Diseases (R01-DK-082582 and R01-DK-082582-S1 to L.Q.).

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

S.S. conceived the framework of the study, designed the experiments, performed most in vivo and in vitro studies, analyzed data, and wrote, commented on, and approved the manuscript. S.X. performed the flow cytometric analysis of BMT, analyzed data, and commented on and approved the manuscript. Y.J. helped with animal dissections and Q-PCR analysis, analyzed data, and commented on and approved the manuscript. S.K. performed the microarray analysis, analyzed data, and commented on and approved the manuscript. L.Q. conceived the framework of the study, designed the experiments, analyzed data, and edited, commented on, and approved the manuscript. L.Q. 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.

Parts of this study have been submitted in abstract form for presentation at the 72nd Scientific Sessions of the American Diabetes Association, Philadelphia, Pennsylvania, 8–12 June 2012.

The authors thank Dr. Rinke Stienstra for helpful suggestions and sharing Casp1^−/− WAT, Dr. Joseph Wakshlag (Cornell University) and Dr. Vishva Dixit for caspase-3 and caspase-1 p20 antibodies, Dr. Qiaoming Long (Cornell University) for use of the microscope, Cindy Wang (Cornell University) for technical assistance, and other members of the Qi Laboratory at Cornell University for critical discussions/suggestions.