Obese adipocytes show ultrastructural features of stressed cells and die of pyroptosis.

We previously suggested that, in obese animals and humans, white adipose tissue inflammation results from the death of hypertrophic adipocytes; these are then cleared by macrophages, giving rise to distinctive structures we denominated crown-like structures. Here we present evidence that subcutaneous and visceral hypertrophic adipocytes of leptin-deficient (ob/ob and db/db) obese mice exhibit ultrastructural abnormalities (including calcium accumulation and cholesterol crystals), many of which are more common in hyperglycemic db/db versus normoglycemic ob/ob mice and in visceral versus subcutaneous depots. Degenerating adipocytes whose intracellular content disperses in the extracellular space were also noted in obese mice; in addition, increased anti-reactive oxygen species enzyme expression in obese fat pads, documented by RT-PCR and immunohistochemistry, suggests that ultrastructural changes are accompanied by oxidative stress. RT-PCR showed NLRP3 inflammasome activation in the fat pads of both leptin-deficient and high-fat diet obese mice, in which formation of active caspase-1 was documented by immunohistochemistry in the cytoplasm of several hypertrophic adipocytes. Notably, caspase-1 was not detected in FAT-ATTAC transgenic mice, where adipocytes die of apoptosis. Thus, white adipocyte overexpansion induces a stress state that ultimately leads to death. NLRP3-dependent caspase-1 activation in hypertrophic adipocytes likely induces obese adipocyte death by pyroptosis, a proinflammatory programmed cell death.

FAT-ATTAC (fat apoptosis through targeted activation of caspase-8) mice were also used ( 19 ). Animals were individually caged and maintained on a 12 h/12 h light/dark cycle with free access to standard pellet food and water. Unless otherwise specifi ed, control mice were the heterozygous animals. Female Swiss CD-1 mice 3 weeks of age purchased from Charles River were caged and treated like the other mice for one week. After the acclimatization period, they were weighed and divided into two groups with similar mean body weight: one group (HFD mice) were fed a high-fat diet (Charles River; 50 kJ% from fat, 30 kJ% from carbohydrates, and 20 kJ% from proteins); the other group (control mice) were fed a low-fat diet (Charles River; 19 kJ% from fat, 50 kJ% from carbohydrates, and 31 kJ% from proteins). After 10 weeks under these feeding regimens, the mean body weight of HFD mice was 47 ± 3 g and that of control mice was 38 ± 1.7 g ( P < 0.05). All animal procedures were in accordance with Italian Institutional and National Guidelines and were approved by the Animal Ethics Committee of University of Ancona (Politecnica delle Marche). Mice were euthanized with an overdose of anesthetic (Avertin; Fluka Chemie, Buchs, Switzerland) and immediately perfused with 4% paraformaldehyde in 0.1 M phosphate buffer (PB), pH 7.4, for 5 min. Subcutaneous (inguinal) and visceral (mesenteric) WAT depots were dissected using a Zeiss OPI1 surgical microscope (Carl Zeiss, Oberkochen, Germany) and further fi xed by immersion in 4% paraformaldehyde in PB overnight at 4°C. After a thorough rinse in PB, specimens were dehydrated in ethanol, cleared in xylene, and embedded in paraffi n.

Light microscopy and morphometry
Serial paraffi n sections 3 m in thickness were obtained from each specimen. Some were stained with hematoxylin and eosin to assess morphology; the others were used for immunohistochemical and histochemical procedures. Adipocyte size was calculated as the mean adipocyte area of 300 random adipocytes (100 per section) from each depot of each mouse using a drawing tablet and the Nikon LUCIA IMAGE software (version 4.61; Laboratory Imaging, Praha, Czech Republic) of the morphometric program. Tissue sections were observed with a Nikon Eclipse E800 light microscope (Nikon Instruments, Firenze, Italy) using a 20× objective, and digital images were captured with a Nikon DXM 1220 camera.

Metabolic analysis
Plasma insulin was measured using a commercially available ELISA kit (Mouse Insulin ELISA, DRG Instruments, Marburg, Germany). Plasma glucose was measured using the glucose oxidase method (Glucose RTU, bioMerieux S.A., Marcy l'Etoile, France). The Homeostasis Model Assessment (HOMA) insulin resistance index was calculated using the following standard formula: HOMA = (FPG × FPI) / 22.5, where FPG and FPI are fasting plasma glucose (mMol) and fasting plasma insulin, respectively.

Transmission electron microscopy and morphometry
Small subcutaneous and visceral adipose tissue fragments were fi xed in 2% glutaraldheyde-2% paraformaldehyde in PB for 4 h at room temperature, postfi xed in 1% osmium tetroxide, and embedded in an Epon-Araldite mixture. Semithin sections (2 m) were stained with toluidine blue. Thin sections were obtained with an MT-X ultratome (RMC; Tucson, AZ), stained with lead citrate, and examined with a CM10 transmission electron microscope (Philips; Eindhoven, Netherlands). Morphometric evaluation of hypertrophic adipocytes was performed using image analysis software (Lucia IMAGE version 4.82, Laboratory characteristic chronic, low-grade infl ammation of obese fat importantly impairs peripheral insulin sensitivity by reducing adiponectin secretion and instigating the production of proinfl ammatory mediators, such as tumor necrosis factor (TNF) ␣ , inducible nitric oxide syntase , interleukin (IL)-6, and IL-1 ␤ ( 2, 3 ). The vast majority of macrophages infi ltrating obese adipose tissue are found at sites where adipocytes undergo necrotic-like death; there they form distinctive structures that we have called crown-like structures (CLS) ( 16 ). Ultrastructurally, CLSs are aggregates of activated macrophages, sometimes fused into syncytia (multinucleated giant cells), that surround individual dead adipocytes. Such structures, which are highly reminiscent of those seen in foreign body tissue reaction, suggest that the death of the obese adipocyte is a major event in obese adipose tissue and that it is followed by macrophage infi ltration whose function is to sequester and ingest adipocyte debris. Thus, we have advanced the hypothesis that the chronic, low-grade infl ammation seen in obese adipose tissue is essentially sustained by the need for clearing the extracellular space of adipocyte debris, particularly the large lipid droplet ( 16 ). Comparison of different visceral and subcutaneous fat depots from ob/ob and db/db obese mice indicates that the former depots are the main sites of adipocyte death and macrophage infi ltration ( 17,18 ), a fi nding that may explain the higher morbidity associated with visceral fat expansion.
In this study, we analyze by quantitative methods the fi ne structure of the hypertrophic adipocytes found in the subcutaneous and visceral adipose tissues of genetically obese ob/ob and db/db mice, which lack, respectively, leptin and the functional form of leptin receptor. Electron microscopic techniques document numerous organelle abnormalities in hypertrophic adipocytes that are consistent with severe cellular metabolic stress and often result in cell degeneration and subsequent macrophage clearing. The increased expression of anti-reactive oxygen species (ROS) enzymes, seen in the fat of obese mice compared with controls, confi rms the oxidative stress state of hypertrophic adipocytes. The abnormal traits found in the hypertrophic adipocytes are capable of activating infl ammasome pathways. Indeed, the nucleotide-binding oligomerization domain (NOD)-like receptor (NLR)-3 (NLRP3) infl ammasome is activated in the adipose depots of both genetically obese mice and in those where obesity is induced by a high-fat diet (HFD). Collectively, the strong caspase-1 expression detected by immunohistochemistry in hypertrophic adipocytes from obese mice and their degenerative features, disclosed by electron microscopy in massively obese leptindefi cient mice, lend support to the hypothesis that hypertrophic adipocytes die of pyroptosis.

Animals and tissues
Ob/+ , ob/ob , db/+ , db/db , and C57BL/6 wild-type (WT) female mice were purchased from Charles River (Lecco, Italy) at 5 weeks of age and used for experimental procedures at 14 weeks of age. for 1 min. Collagen was detected by picrosirius red staining ( 23 ). Adipose tissue sections were fi rst dewaxed and hydrated, then incubated in Sirius Red 0.1% in Sat'd Picric Acid for 30 min, rinsed in water for 10 min, washed for 2 min in 1 N hydrochloric acid, rinsed again in water, dehydrated, cleared, and mounted.

RNA isolation and analysis
Total RNA was extracted using TRIZOL reagent (Invitrogen, Milano, Italy), purifi ed, digested with RNase-free DNase, and concentrated using RNeasy Micro kit (Qiagen, Milano, Italy) according to the respective manufacturer's instructions.
For determination of mRNA levels, 1 µg of RNA was reversetranscribed with a high-capacity cDNA RT kit with RNase inhibitor (Applied BioSystems, Foster City, CA) in a total volume of 20 µl. Real time gene expression was analyzed in triplicate by using Taq Since _g1 gene expression assays, which do not exclude genomic DNA amplifi cation, were used for ASC, superoxide dismutase 1, TXNIP, and glutathione peroxidase 1, after RNA purifi cation, samples were subjected to DNase treatment with RQ1 RNase-Free DNase (Promega, Milano, Italy) followed by RNA cleanup procedure using RNeasy micro kit (Qiagen).
Reactions were carried out in an ABI 7300 system (Applied Biosystems) using 50 ng RNA in a fi nal reaction volume of 20 l and the following thermal cycle protocol: initial incubation at 95°C for 10 min, followed by 40 cycles of 95°C for 15 s and 60°C for 20 s. A control reaction where reverse transcriptase was omitted in the amplifi cation mixture was included for each sample to rule out genomic contamination. Relative mRNA expression was determined by the ⌬ ⌬ -Ct method using TBP levels as an endogenous control. Differences in starting total RNA and in cDNA synthesis effi ciency among samples were normalized using TBP expression; results were expressed as fold changes in relative gene expression compared with the control group.

Statistical analysis
Results are given as mean ± standard error of the mean (SEM). Differences between groups were evaluated by ANOVA (ANOVA) and by the unpaired t -test. Statistical analysis was performed with GraphPad InStat (version 3.00 for Windows) software (San Diego, CA). A level of P < 0.05 was considered signifi cant.

RESULTS
As expected, both ob/ob and db/db obese mice had higher body and fat depot weights than heterozygous and WT mice (data not shown). Numerous scattered MAC-2-positive CLSs were detected by light microscopy among white adipocytes in the subcutaneous and visceral fat depots of obese mice ( 16 ). Mean subcutaneous and visceral white adipocyte size was about 7 times greater in ob/ob mice and about 11 times greater in db/db mice compared with WT animals ( Table 1 ), in line with previous morphometric investigations ( 18 ). In addition, subcutaneous adipocytes were about 24% larger than visceral adipocytes in both obese strains ( Table 1 ). Imaging). Five animals per group were examined. Ten electron micrographs of each adipose depot of each mouse were taken at a magnifi cation of 4,900× to calculate peripheral cytoplasmic area; mitochondrial area; mitochondrial density; rough endoplasmic reticulum (RER) area; Golgi complex area; the percentage of hypertrophic adipocytes containing glycogen deposits, calcium deposits, and collagen fi brils; and the percentage of degenerating adipocytes.

High-resolution scanning electron microscopy
Tissue fragments were fi xed with 0.25% glutaraldehyde and 0.25% paraformaldehyde in PB for 15 min at room temperature. They were postfi xed in 1.25% potassium ferrocyanide and 1% osmium tetroxide for 2 h and then processed by the osmium tetroxide maceration method ( 20,21 ). After dehydration through a graded acetone series, they were critical-point dried (CPD 010 Balzers; Fachbericht, Liechtenstein) and gold-coated with an Emitech K550 sputter coater (Quorum Technology, Ashford, Kent, UK). Specimens were observed with a FESEM Zeiss Supra 40 VP.

Peroxidase immunohistochemistry
Immunohistochemistry was performed on 3 m-thick paraffi nembedded sections of fat depots. After dewaxing, antigen retrieval was achieved with a pressure cooker treatment (90°C for 20 min) by soaking sections in a sodium citrate buffer 0.01 M, pH 6.0. After a thorough rinse in phosphate buffered saline (PBS), sections were reacted with 0.3% H 2 O 2 (in PBS; 30 min) to block endogenous peroxidase, rinsed with PBS, and incubated in a 3% blocking solution (in PBS; 60 min). Then they were incubated with the primary antibodies (in PBS; overnight at 4°C). After a thorough rinse in PBS, sections were incubated in a 1:200 v/v biotinylated secondary antibody solution (in PBS; 30 min). Biotinylated HRPconjugated secondary antibodies (Vector Laboratories, Burlingame, CA) were goat anti-rabbit IgG (perilipin/ASC/SOD-1/ catalase) and horse anti-mouse IgG (MAC-2/caspase-1/GPX-1). Histochemical reactions were performed using Vectastain ABC kit (Vector Laboratories) and Sigma Fast 3,3 ′ -diaminobenzidine (Sigma-Aldrich) as the substrate. Sections were fi nally counterstained with hematoxylin, dehydrated, and mounted in Entellan. Staining was never observed when the primary antibody was omitted.

Histochemistry
Von Kossa staining was employed to detect calcium deposits according to Rungby et al. ( 22 ), with slight modifi cations. Paraffi nembedded sections were dewaxed, hydrated, and incubated with a solution of silver nitrate 5% in distilled water at room temperature for 1 h under a 100 W light bulb. They were rinsed with distilled water, incubated with sodium-carbonate-formaldehyde for 2 min, rinsed again, and fi nally incubated with Framer's reducer As regards glucose homeostasis ( Table 2 ), db/+ , ob/+ , and WT mice had comparable plasma glucose, insulin, and HOMA index values, whereas in ob/ob mice, plasma glucose levels were comparable to those found in controls and the insulin and HOMA index values were signifi cantly higher. The db/db mice exhibited the highest values for all these parameters. In synthesis, in ob/ob mice, hyperinsulinemia compensated for impaired insulin sensitivity, whereas db/db mice had both hyperinsulinemia and hyperglycemia.

Hypertrophic adipocytes of genetically obese mice show ultrastructural features of cellular stress
To document early signs of cellular stress in obese fat, we examined by electron microscopy the hypertrophic adipocytes found near CLSs. These cells appeared to lack direct contact with the macrophages infi ltrating adipose tissue and were always perilipin-immunoreactive, and hence, metabolically viable ( 19,24,25 ).
The cytoplasmic rim was thinner in hypertrophic ( Fig. 1A ) than in control adipocytes ( Fig. 1B ), even though the total cytoplasmic area was unchanged, and thinnest in visceral hypertrophic adipocytes ( Fig. 2A ). Whereas mitochondrial density was similar in adipocytes from both obese and control mice, mean mitochondrial area was signifi cantly smaller in the hypertrophic adipocytes of visceral depots ( Fig. 2B ). Notably, giant mitochondria twice the size of normal mitochondria and exhibiting abnormally arranged cristae were also occasionally found ( Fig. 1C, D ). In the vast majority of adipocytes of both obese strains, the RER was expanded and dilated ( Fig. 1E ), but the number of these cells was again higher in the visceral depots ( Fig. 2C ). Hypertrophic adipocytes were also characterized by a hypertrophic Golgi complex and numerous and clearly evident glycogen deposits ( Fig. 1F ). Interestingly, the percentage of hypertrophic adipocytes Plasma glucose was signifi cantly increased in db/db compared with ob/ob , wild-type (WT), and heterozygous mice. Insulin and HOMA index were signifi cantly higher in both ob/ ob and db/db mice compared with WT and heterozygous mice. Values are mean ± SEM. *P < 0.05, 5-7 animals per group. containing glycogen deposits was higher in visceral than in subcutaneous fat ( Fig. 2D ).
Electron dense material with the ultrastructural appearance of calcium deposits ( 26 ) was often found in The extracellular matrix (ECM) surrounding hypertrophic adipocytes differed markedly from the ECM around control adipocytes. A greater number of collagen fi brils were found in close contact with the external surface of the basal lamina of hypertrophic adipocytes ( Fig. 4A ). In line with this fi nding, increased picrosirius red staining was noted around hypertrophic adipocytes in paraffi n-embedded sections ( Fig. 4C ). In addition, three-dimensional evaluation performed with high-resolution scanning electron hypertrophic adipocytes; in many cases, this material appeared to accumulate very close to the lipid droplet ( Figs.  1D and 3A ). Light-microscopy examination of paraffi nembedded sections processed for calcium-specifi c von Kossa staining confi rmed the presence of abnormal calcium deposits in the hypertrophic adipocytes of obese mice ( Fig. 3B ). Finally, hypertrophic adipocytes also contained crystals with the ultrastructural appearance of cholesterol crystals ( 26,27 ) (see below). In subcutaneous fat all measures show the same tendency, but differences are not signifi cant except for cytoplasmic thickness in db/db mice. * P < 0.05; ** P < 0.01. expressed in the mesenteric depot of ob/ob mice ( Fig. 5D ), whereas control mouse adipocytes were negative or very weakly stained (inset of Fig. 5D ).

Hypertrophic adipocytes of genetically obese mice degenerate and die before CLS formation
Transmission electron microscopy (TEM) occasionally demonstrated adipocytes that contained disorganized lipid-like material and cholesterol crystals in the cytoplasm and that exhibited an obvious reduction of the organelle content ( Fig. 6A , B ). The observation that such degenerating adipocytes were closely adjacent to well-preserved hypertrophic adipocytes ( Fig. 6B ) ruled out that their appearance could be due to technical artifacts. Degenerating adipocytes were perilipin-negative by immunohistochemistry. A semiquantitative assessment performed by TEM disclosed that they were quite numerous (15.6% ± 4.4) in the visceral and subcutaneous fat depots of obese mice, rare (3.15% ± 1.4) in lean db/+ and ob/+ mice, and virtually absent in lean C57BL/6 WT mice. Degenerating adipocytes sometimes seemed to extrude lipid-like material into the interstitial space, as documented by both TEM and HR-SEM ( Fig. 7 ). Detection of free lipid droplets in the extracellular space and in neighboring macrophages suggested that the lipid-like material extruded from degenerating adipocytes was phagocytized by macrophages. Cholesterol crystals were also found in degenerating adipocytes and in surrounding macrophages ( Figs. 6C and 7 ).
In summary, the hypertrophic adipocytes of ob/ob and db/db mice exhibited a number of ultrastructural abnormalities (thinning of the cytoplasmic rim, abnormal mitochondrial size and shape, expansion and dilatation of RER and Golgi complex, abnormal storage of glycogen and calcium, cholesterol crystals, and increased amounts of collagen fi brils in the extracellular space) that are highly suggestive of a cellular stress condition. This was confi rmed by the increased expression of some anti-ROS enzymes. Many of these aspects were especially evident in hyperglycemic db/db mice compared with normoglycemic ob/ob animals and in visceral compared with subcutaneous adipose depots. Altogether, these morphological observations suggest a sequence of events taking place before CLS formation, as follows: 1 ) adipocyte hypertrophy, 2 ) degeneration of hypertrophic adipocytes with lipid droplet release into the extracellular space, 3 ) recruitment of activated macrophages, and 4 ) formation of multinucleated giant cells and fi nally of CLSs.

NLRP3 infl ammasome is activated in hypertrophic adipocytes of genetically obese mice
The NLRP3 infl ammasome is a cytoplasmic protein complex that comprises an intracellular sensor (the NODlike receptor), the precursor procaspase-1, and the adaptor ASC. The NLRP3 infl ammasome has recently been shown to be activated in obese fat, instigating local production of the proinfl ammatory IL-1 ␤ and IL-18 ( 28 ). In our experiments, RT-PCR analysis showed signifi cantly increased ASC ( Fig. 8A ), NLRP3 ( Fig. 8B ) and caspase-1 microscopy (HR-SEM) showed an intricate network of collagen fibrils on the surface of hypertrophic adipocytes ( Fig. 4D, E ).

Hypertrophic adipocytes of genetically obese mice undergo oxidative stress
Storage of excess energy in adipocytes results in mitochondrial stress and endoplasmic reticulum disruption, commonly giving rise to oxidative stress and subsequent ROS production ( 5, 6 ). A variety of mechanisms have been developed by cells to stem the harmful effect of ROS and free radical production, including production and/or activation of the enzymes SOD-1, catalase, and GPX-1 ( 5 ). To provide evidence of the stress state of hypertrophic adipocytes of genetically obese mice, the expression of these anti-ROS enzymes was sought in the adipose depots of these animals. RT-PCR analysis showed signifi cantly increased mRNA levels of catalase in the subcutaneous and visceral fat of ob/ob mice ( Fig. 5A ) and of GPX-1 in the subcutaneous and visceral adipose depots of ob/ob and db/db mice compared with lean mice ( Fig. 5B ). SOD-1 mRNA expression was increased in the subcutaneous depot of ob/ob and db/db mice and decreased in the visceral depot of db/db mice ( Fig. 5C ). All enzymes were detected in the cytoplasm of hypertrophic adipocytes by immunohistochemistry. According to RT-PCR data, catalase was strongly  mRNA levels ( Fig. 8C ) in the adipose depots of obese compared with lean mice. Notably, the mRNA levels of thioredoxin (TRX)-interacting protein (TXNIP), which drives NLRP3 infl ammasome activation following oxidative stress ( 29 ), were signifi cantly increased in the subcutaneous and visceral adipose depots of genetically obese mice ( Fig. 8D ).
As reported above, cholesterol crystals were detected in the cytoplasm of both hypertrophic and degenerating adipocytes. They were also found in macrophages that were closely apposed to hypertrophic-degenerating adipocytes, thus providing additional evidence that the latter cells   6. Transmission electron microscopic images of hypertrophic adipocytes in visceral fat of db/db mice. In (A), the cytoplasm of two closely apposed hypertrophic-degenerating adipocytes is disrupted, devoid of recognizable organelles, and contains vacuoles and a cholesterol crystal (arrowhead). In (B), a hypertrophic adipocyte (HA) with well-preserved organelles is found near a degenerating adipocyte (DA), whose cytoplasm lacks organelles, which are replaced by lipid-like material. In (C), several macrophages (Mac) are closely apposed to a HA containing a cholesterol crystal (arrowhead) in the degenerating cytoplasm; the macrophages contain lipid droplets (L) and numerous cholesterol crystals (arrows). m, mitochondria; lys, lysosome. Scale bar: 0.6 m for (A); 1 m for (B) and (C). larger in the HFD subcutaneous fat compared with the corresponding control depots. MAC-2-positive CLSs were not found in the fat of control mice but were frequently detected in HFD mice (visceral fat: 129 ± 30 CLSs/10,000 adipocytes; subcutaneous fat: 128 ± 112 CLSs/10,000 adipocytes). RT-PCR analysis showed a signifi cant increase of anti-ROS enzymes and TXNIP in the HFD fat (data not shown). A signifi cant increase in ASC ( Fig. 10A ), NLRP3 ( Fig. 10B ), and caspase-1 ( Fig. 10C ) mRNAs was found in the visceral WAT of HFD mice, whereas only NLRP3 and caspase-1 were increased in their subcutaneous depot. By immunohistochemistry, ASC ( Fig. 10D ), NLRP3 ( Fig. 10E ), and caspase-1 ( Fig. 10F ) protein were detected in CLSs and in the cytoplasm of several subcutaneous and visceral hypertrophic adipocytes.

DISCUSSION
In obese animals and humans, WAT expansion is associated with a local infi ltration of infl ammatory cells, mainly macrophages, that give rise to low-level, chronic infl ammation ( 14,15 ). Mounting evidence suggests that adipose tissue infl ammation strongly correlates with the appearance of peripheral insulin resistance and, consequently, to type-2 diabetes due to a reduction in adiponectin secretion and an increase in circulating infl ammatory cytokines ( 1-4 ). Thus, chronic adipose tissue infl ammation is a key element in the pathogenesis of metabolic syndrome; however, the factors that trigger and sustain the infl ammatory state remain elusive.
We previously showed that the vast majority (>90%) of activated macrophages infi ltrating obese adipose tissue are found around dead adipocyte remnants, giving rise to characteristic light microscopic fi gures that we called CLSs were indeed cleared by the macrophages infi ltrating obese adipose tissue. Size and cholesterol content correlate positively in adipocytes ( 30 ). Thus, the larger the fat cells, the more cholesterol they contain; under certain conditions cholesterol may precipitate, giving rise to crystals ( 31 ). Cholesterol crystals are able to activate the NLRP3 infl ammasome ( 31,32 ); their formation in hypertrophic adipocytes might thus be the critical event triggering NLRP3 infl ammasome activation. Immunohistochemical analysis showed caspase-1 ( Fig. 9A , B ), NLRP3 (not shown) and ASC ( Fig. 9C ) in the cytoplasm of many hypertrophic adipocytes and in CLS macrophages. Notably, caspase-1 and ASC immunoreactivity was mainly found in macrophages that were located in close apposition to adipocyte remnants, suggesting that staining might not only be due to direct protein production by macrophages but also to phagocytosis of caspase-1 and ASC-containing cytoplasmic debris from dead adipocytes. The absence of staining for these markers in macrophages not spatially close to CLSs is in line with this hypothesis ( Fig. 9B ). Notably, the infl ammasome proteins were found neither in the adipocytes of control mice nor in the adipocytes or CLSs of FAT-ATTAC mice ( Fig. 9E ), a transgenic model of apoptotic adipocyte death ( 19,33 ).

NLRP3 infl ammasome is activated in hypertrophic adipocytes of HFD mice
To exclude that the effects seen in ob/ob and db/db mice were peculiar to the leptin-defi cient state, the investigation was extended to HFD obese mice. Macroscopically, both their visceral and subcutaneous WAT depots appeared enlarged compared with those of controls. By light microscopy, the WAT contained hypertrophic adipocytes that were especially evident and numerous in the visceral depot; indeed, the mean adipocyte area was ‫ف‬ 2.7 times larger in the HFD mesenteric depot and ‫ف‬ 1.23 times ceral fat of obese animals ( 18 ) suggested to us that enlargement of white adipocytes beyond a certain depot-dependent, critical size may result in their death.
The present investigation has documented severe organelle alterations and signs of oxidative stress in the hypertrophic adipocytes of genetically obese mice. Adipocyte degeneration results in release of the cellular content into the extracellular space, and macrophage recruitment eventually leads to their removal and CLS formation. Most morphological and biochemical adipocyte changes were more marked in the hyperglycemic hyperinsulinemic db/db than the normoglycemic hyperinsulinemic ob/ob mice. Adipocytes are larger in db/db than ob/ob mice ( 18,35 ). A positive correlation between adipocyte size and fat infl ammation has been documented ( 15,18 ), as has a correlation between fat infl ammation and insulin resistance (1)(2)(3)(4). We noted that visceral fat infl ammation in db/db mice, measured as CLS density, was 3.4 times greater than that in ob/ob mice ( 18 ). Thus, the greater propensity of db/db mice to develop insulin resistance and type-2 diabetes can be explained by greater fat infl ammation.
Analysis of the time course of the histomorphological changes taking place in FAT-ATTAC mice, in which white ( 16 ). Ultrastructurally, CLSs correspond to remnants of hypertrophic adipocytes surrounded by lipid-resorbing macrophages. Based on such morphological fi ndings, we proposed that the death of hypertrophic adipocytes may be the trigger inducing macrophage recruitment and adipose tissue infl ammation in obese mice and humans. Notably, knockout of hormone-sensitive lipase (HSL), a major lipase in mature white adipocytes, results in increased lipid storage and white adipocyte hypertrophy. As a consequence, HSL-defi cient mice show adipocyte hypertrophy but not increased adipose mass or obesity ( 34 ). Also in such mice, we found typical CLSs with a frequency very similar to the one observed in obese mouse fat ( 16 ). This further observation confi rms that adipocyte death is a trigger for macrophage recruitment and suggests that the mechanisms causing adipocyte death are not linked to the obese condition per se but may correlate directly to adipocyte size. Indeed, a positive correlation between adipocyte size and degree of macrophage infi ltration, or CLS density, has been documented in the subcutaneous and visceral adipose depots of obese mice ( 14,15,18 ). Finally, the observation that visceral fat adipocytes are smaller than those of subcutaneous fat but that CLS density is greater in the vis- turn processes the cytosolic precursors of the proinfl ammatory cytokines IL-1 ␤ and IL-18, which are secreted as biologically active cytokines. In diet-induced obesity, in genetically obese mice ( ob/ob and db/db ), and in obese humans, the NLRP3-infl ammasome is activated in both visceral and subcutaneous fat and drives local production of IL-1 ␤ and IL-18 ( 28,37 ). In particular, IL-1 ␤ is elevated in obese individuals and rodents; it is also strongly implicated in impaired insulin secretion by pancreatic ␤ -cells ( 38 ) and in disruption of insulin signaling at cellular targets ( 39 ). Notably, blockade of IL-1 ␤ signaling by Anakinra, a specifi c recombinant human receptor antagonist, in patients with type-2 diabetes ( 40 ) and administration of anti-IL-1 ␤ antibodies to mice with diet-induced obesity ( 41 ) lead to a sustained reduction of systemic infl ammation and to signifi cantly improved glycemic control. That NLRP3-dependent caspase-1 activation and production of infl ammatory cytokines play a prominent role in metabolic syndrome has recently been confi rmed in mice lacking NLRP3, ASC, or caspase-1, which are resistant to the development of high-fat diet-induced obesity and insulin resistance ( 42 ). adipocyte apoptosis is induced through forced dimerization of a caspase-8 fusion protein ( 19,33 ), disclosed an analogous sequence of morphological events: apoptotic adipocytes fi rst exhibit organelle alterations; then they lose perilipin immunoreactivity, degenerate, and recruit neutrophils and lymphocytes; and fi nally, they are resorbed by macrophages in CLSs ( 36 ). Thus, the death of hypertrophic adipocytes appears to be a major event triggering macrophage recruitment and CLS formation in both genetic obesity and in the adipose tissue-specifi c apoptosis model.
NOD-like receptors (NLR) comprise a large family of intracellular proteins that are involved not only in the innate immune response to microbial pathogens (through the recognition of conserved pathogen-associated molecular patterns) but also in sterile infl ammation by sensing "danger signals" (i.e., endogenous molecules produced during cell and tissue damage). NLRP3 is currently the most fully characterized infl ammasome, consisting of the NLRP3 scaffold, the ASC (PYCARD) adaptor, and caspase-1. NLRP3 activation results in production of activated caspase-1, which in Fig. 9. NLRP3 activation in adipose tissue of genetically obese mice. Immunohistochemistry. Caspase-1 immunostaining is evident (brownish precipitate, arrows) in hypertrophic (or degenerating) adipocytes (A) and in a CLS (B) of the visceral fat of a db/db mice: here, staining is found in macrophages surrounding the large lipid droplet but not in those lying at a distance from it (arrowheads), which are not yet engaged in lipid clearing. In (C), ASC immunoreactivity is found in a CLS, but weak staining is also visible in the cytoplasmic rim of a hypertrophic adipocyte (arrow). In ( cholesterol crystals in the cytoplasm. Even though a causal link has not yet been demonstrated, such crystals may constitute a "danger signal" capable of activating the NLRP3 infl ammasome ( 31,32 ). It may be speculated that NLRP3 infl ammasome activation in hypertrophic adipocytes is triggered by the cholesterol crystals, which ultimately drive cells to pyroptosis. Importantly, in FAT-ATTAC mice, the apoptotic mode of cell death of hypertrophic adipocytes is not characterized by formation of cholesterol crystals ( 36 ); accordingly, the hypertrophic adipose tissue shows no signs of NLRP3 infl ammasome or caspase-1 activation (present results). Data from different experimental settings suggest that, in addition to cholesterol crystals, other factors may activate the NLRP3 infl ammasome in hypertrophic adipocytes, including extracellular ATP and high glucose concentrations, which may follow rupture of adipocyte cell membranes ( 29,47 ), and increased ROS and free radical production, as also suggested by the increased expression of anti-ROS enzymes detected in our tissue samples. The latter is consistent with the increased expression of TXNIP in the adipose tissue of obese mice ( 29 ). Recently, increased intracellular Ca 2+ also has been shown to activate the NLRP3 infl ammasome ( 48 ).
In conclusion, the evidence collected in the present work lends support to the hypothesis that, in obesity, hypertrophic adipocytes exhibit ultrastructural and biochemical alterations, including cholesterol crystals and accumulation of calcium and ROS, that might trigger the NLRP3 infl ammasome Caspases are cysteine proteases. They are tightly controlled by infl ammasome proteolytic activation, and they initiate or execute cellular programs, leading to infl ammation or cell death ( 43 ). Caspase-1 is a proinfl ammatory caspase whose activity can result in a highly infl ammatory form of cell death known as pyroptosis ( 44 ). At the cellular level, pyroptosis is characterized by plasma membrane rupture, water infl ux, cellular swelling, osmotic lysis, and release of proinfl ammatory cellular content. Although it is considered as a form of programmed cell death, pyroptosis is different from apoptosis, which is an immunologically silent programmed cell death. Although the mechanisms underpinning this process are not clear, the rate of pyroptosis appears to increase with increasing infl ammasome stimulation ( 45 ). We confi rmed NLRP3 infl ammasome activation and the presence of active caspase-1 in adipose tissue in genetically leptin-defi cient and HFD obese mice. ASC, NLRP3, and caspase-1 were detected by immunohistochemistry in the macrophages forming CLSs, as reported by other researchers ( 28 ), but strong immunoreactivity was also found in hypertrophic adipocytes. This fi nding is in line with in vitro observations showing that caspase-1 is upregulated during adipocyte differentiation and directs adipocytes toward a more insulin-resistant phenotype ( 37 ) and with in vivo data showing high levels of caspase-1 protein in the fat of ob/ob mice ( 46 ).
A distinctive ultrastructural feature of the hypertrophic and degenerating adipocytes observed in this study were , and caspase-1 (C). Mean ± SEM. * P < 0.05; ** P < 0.01; *** P < 0.001. Bottom panels. Immunohistochemical detection of NLRP3 markers in the visceral fat of HFD mice. In (D), specifi c ASC immunostaining can be seen as brownish precipitate in the cytoplasm of hypertrophic adipocytes (arrows). In (E), NLRP3 is found in some hypertrophic adipocytes (arrows) and in a CLS. In (F), a hypertrophic adipocyte exhibits strong cytoplasmic staining for caspase-1 (arrowheads). Scale bar: 20 m for (D) and (E); 7 m for (F). pathway with subsequent massive activation of caspase-1, which ultimately results in adipocyte death from pyroptosis. Pyroptosis, the infl ammatory mode of cell death, promotes recruitment of infl ammatory cells, including macrophages, whose prolonged presence and sustained activity in obese fat is needed to clear the extracellular space of lipid material, including cholesterol crystals, released by degenerating adipocytes. In the visceral fat of obese mice, this process can be accelerated by the smaller death size of hypertrophic adipocytes, thus offering a biological basis for the more severe clinical outcomes associated with visceral fat increase. Recently, it has been suggested that the mild infl ammation found in obesity may, to some extent, depend on the leakage of gut microbiota endotoxins into the circulation ( 49 ), resulting in nuclear factor (NF)-B activation in adipocytes and release of proinfl ammatory cytokines ( 50 ). Nonetheless, the distinctive histopathological fi ndings presented here and the sequence of cellular events they suggest point to a dysregulation of adipocyte storage capability and size as the primary triggers of adipose tissue infl ammation.