Brown adipose tissue whitening leads to brown adipocyte death and adipose tissue inflammation[S]

In mammals, white adipose tissue (WAT) stores and releases lipids, whereas brown adipose tissue (BAT) oxidizes lipids to fuel thermogenesis. In obese individuals, WAT undergoes profound changes; it expands, becomes dysfunctional, and develops a low-grade inflammatory state. Importantly, BAT content and activity decline in obese subjects, mainly as a result of the conversion of brown adipocytes to white-like unilocular cells. Here, we show that BAT “whitening” is induced by multiple factors, including high ambient temperature, leptin receptor deficiency, β-adrenergic signaling impairment, and lipase deficiency, each of which is capable of inducing macrophage infiltration, brown adipocyte death, and crown-like structure (CLS) formation. Brown-to-white conversion and increased CLS formation were most marked in BAT from adipose triglyceride lipase (Atgl)-deficient mice, where, according to transmission electron microscopy, whitened brown adipocytes contained enlarged endoplasmic reticulum, cholesterol crystals, and some degenerating mitochondria, and were surrounded by an increased number of collagen fibrils. Gene expression analysis showed that BAT whitening in Atgl-deficient mice was associated to a strong inflammatory response and NLRP3 inflammasome activation. Altogether, the present findings suggest that converted enlarged brown adipocytes are highly prone to death, which, by promoting inflammation in whitened BAT, may contribute to the typical inflammatory state seen in obesity.

adipocytes. Importantly, macrophage infiltration of obese fat positively correlates with adipocyte size and CLS density (4)(5)(6)(7). In particular, the number of CLSs in obese fat is in direct proportion to the degree of adipocyte hypertrophy in both visceral and subcutaneous adipose tissue (7). However, their density is higher in visceral adipose tissue, where adipocytes are smaller, than in subcutaneous fat, where adipocytes are larger. This suggests that lipid overload results in adipocyte growth up to a critical size; growth beyond this size induces profound cell stress and leads to death. Thus, the critical size is smaller in visceral than in subcutaneous fat depots (6,7). Hypertrophic adipocytes from obese mice are prone to developing organelle dysfunction, such as endoplasmic reticulum (ER) stress (10), mitochondrial dysfunction (11), and activation of the NLRP3 inflammasome pathway, suggesting that hypertrophic stressed adipocytes undergo pyroptotic cell death and initiate an inflammatory response (12). Notably, organelle dysfunction and inflammasome activation are more marked in visceral than in subcutaneous adipocytes (12).
The mammalian adipose organ contains white and brown adipocytes (13). White adipocytes, which are found in white adipose tissue (WAT), store dietary energy as triglycerides in a single large lipid droplet (unilocular adipocytes); at times of high calorie requirements, triglycerides undergo hydrolyzation and the resulting fatty acids are secreted into the bloodstream and supplied to other tissues. Brown adipocytes are the main parenchymal cell type found in brown adipose tissue (BAT). They are characterized by multiple small triglyceride droplets (multilocular adipocytes) and contain numerous large mitochondria, where uncoupling protein 1 (UCP1) enables use of the energy derived from fatty acid oxidation for heat generation (nonshivering thermogenesis) (14). In normally fed mice maintained at normal ambient temperature, white and brown adipocytes are found together in several subcutaneous and visceral depots (13)(14)(15)(16). In contrast, in the adipose organ of obese animals and humans, typical brown adipocytes are barely detectable because most of them undergo a still poorly elucidated conversion to a "white-like" phenotype (13,17,18). Given the limited available data on the fate and possible proinflammatory role of white-like adipocytes derived from brown-to-white conversion in obese mice, the present study was performed to investigate the implications of brown-to-white-like conversion (in terms of the tendency of adipocytes to develop inflammation and to die) in the adipose depots of mice challenged with environmental, dietary, and genetic stimuli.

Animals and tissues
Adipose triglyceride lipase (Atgl)-knockout (ko) mice generated by targeted homologous recombination (19) were backcrossed at least seven times to a C57BL/6J genetic background. All experiments were carried out in male 8-to10-week-old Atgl-ko mice and corresponding wild-type littermates. The adipose-specific Atgl-ko (AAKO) mice (20) are described in the supplemental Materials and Methods. Db/+ and db/db female mice aged 5 weeks were purchased from Charles River (Lecco, Italy) and used for experimental procedures at 14 weeks of age (five animals per strain). C57BL/6J female mice (Harlan, Udine, Italy) aged 12 weeks were kept at 28°C (n = 5, warm-acclimated mice) or at 6°C (n = 5, cold-acclimated mice) for 10 days to reduce and respectively increase the noradrenergic inputs to the adipose organ. Adult mice lacking -adrenergic receptors (-less mice) were kindly provided by Dr. B. B. Lowell (Harvard Medical School, Boston, MA) (21). Animals were individually caged and maintained on a 12:12 h light/dark cycle with free access to standard pellet food and water. The animal experiments performed at the Department of Experimental and Clinical Medicine, Università Politecnica delle Marche, Ancona, Italy were in accordance with institutional and national guidelines and were approved by the institutional Animal Ethics Board of Università Politecnica delle Marche. Experiments carried out at the Institute of Molecular Biosciences, University of Graz, Graz, Austria were approved and performed according to the guidelines of the ethics committee of the University of Graz and the Austrian Federal Ministry for Science and Research.
For morphological studies, 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. BAT and WAT depots were dissected using a Zeiss OPI1 surgical microscope (Carl Zeiss, Oberkochen, Germany) and further fixed by immersion in 4% paraformaldehyde in PB overnight at 4°C. After a thorough rinse in PB, small fragments were collected for transmission electron microscopy (TEM; see below). The remaining tissue was dehydrated in ethanol, cleared in xylene, and embedded in paraffin. For molecular biology assays, animals were anesthetized with ISO-flo®/isoflurane (Abbott, Abbott Park, IL) and euthanized by cervical dislocation. BAT and WAT specimens were rapidly removed, snap-frozen in liquid nitrogen, and stored at 80°C until use.

Light microscopy and morphometry
Serial paraffin sections (3 m thick) were obtained from each adipose depot, placed on glass slides, and dried. Alternate sections were used for hematoxylin and eosin staining to assess morphology, and for immunohistochemical procedures to evaluate tissue protein expression. Adipocyte size was defined as mean adipocyte area (in square micrometers) using a drawing tablet and a morphometric program (Nikon LUCIA IMAGE, Laboratory Imaging, version 4.61; Praha, Czech Republic). Tissue sections were examined with a Nikon Eclipse E800 light microscope, and digital images were captured at 20× with a Nikon DXM 1200 camera (Nikon Instruments S.p.A, Calenzano, Italy). CLS density was determined by counting the total number of (MAC-2-positive) CLSs in each section compared with the total number of adipocytes and was expressed as CLS number/10,000 adipocytes. The CLS index was calculated by dividing CLS density by mean adipocyte size.

Immunohistochemistry
For immunohistochemistry, 3 m thick paraffin-embedded sections of the fat depots were dewaxed; they were then reacted with 0.3% H 2 O 2 (in methanol; 30 min) to block endogenous peroxidase, rinsed with PBS, and incubated in 3% normal serum blocking solution (in PBS; 30 min). Sections were then incubated overnight at 4°C with rat monoclonal anti-MAC-2 antibody (dilution 1:1,500; Cedarlane Laboratories, Burlington, Ontario, Canada), a marker of activated macrophages, or rabbit polyclonal anti-perilipin antibody (dilution 1:300; kindly provided by Dr. A. S. Greenberg, Boston, MA). After a thorough rinse in PBS, sections were incubated in 1:200 v/v horse anti-rat (MAC-2 schedule) or goat anti-rabbit (perilipin schedule) IgG biotinylated HRPconjugated secondary antibody solution (Vector Laboratories, Burlingame, CA) in PBS for 30 min. Histochemical reactions were performed using a Vectastain ABC kit (Vector Laboratories) and Sigma Fast 3,3′-diaminobenzidine (Sigma-Aldrich, Vienna, Austria) as the substrate. Sections were counterstained with hematoxylin, dehydrated in ethanol, and mounted in Eukitt® mounting medium (Sigma-Aldrich). Staining was never observed when the primary antibody was omitted. The procedures applied for immunofluorescence and confocal microscopy analysis are described in the supplemental Materials and Methods.

TEM
Small fragments of BAT and WAT from perfused mice were fixed in 2% glutaraldheyde-2% paraformaldehyde in PB for 4 h at room temperature, postfixed in 1% osmium tetroxide, dehydrated in a graded series of acetone, and embedded in an Epon-Araldite mixture. To determine the region of interest, semi-thin sections were cut and 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, The Netherlands).

RNA isolation and quantitative real-time PCR
Homogenization of frozen BAT and WAT for RNA isolation was performed using an Ultra-Turrax apparatus (IKA, Staufen, Germany) and TRIzol® reagent (Life Technologies, Invitrogen, Vienna, Austria) according to standard protocols. For gene expression analyses, RNA samples were treated with DNase I (Life Technologies, Invitrogen) and reverse transcribed into singlestranded cDNA using a high-capacity reverse transcription kit (Life Technologies, Applied Biosystems, Vienna, Austria). cDNA samples were amplified using Maxima SYBR Green/ROX Master Mix 2x (Fermentas Life Science, St. Leon-Roth, Germany) and primer pairs specific for the target genes (Life Technologies, Invitrogen). The primer sequences are available on request. Realtime PCR was performed on a C1000 thermocycler using the CFX96 real-time system (Bio-Rad Laboratories GmbH, Vienna, Austria). Relative target gene expression was normalized to the ribosomal gene, 36b4, and calculated using the Ct method (22). The procedures regarding RNA isolation and quantitative (q)PCR of AAKO mice are described in the supplemental Materials and Methods.

DNA isolation and quantitative mitochondrial DNA content analysis
Whole genomic DNA and mitochondrial (mt)DNA were extracted from adipose tissue depots using DNeasy blood and tissue kit (Qiagen Vertriebs GmbH, Vienna, Austria) according to the manufacturer's instructions. mtDNA copy number was determined with quantitative real-time PCR using SYBR green and specific primers for the mitochondrial-encoded cytochrome c oxidase 1 (Mt-Co1) and the single copy nuclear gene, NADH dehydrogenase (ubiquinone) flavoprotein 1 (Ndufv1), as control, as described previously (23). DNA samples were amplified using Maxima SYBR Green/ROX Master Mix 2x (Fermentas Life Science) and gene-specific primer pairs (Life Technologies, Invitrogen). PCR reactions were run on an ABI One Step PLUS system (Life Technologies, Applied Biosystems). Relative mtDNA content was determined by normalizing Mt-Co1 expression values to Ndufv1 and calculated using the Ct method as described recently (22).

Tissue homogenization and immunoblot analysis
Snap-frozen tissue was homogenized in ice-cold homogenization buffer (0.25 M sucrose, 1 mM EDTA, 1 mM DTT) containing protease inhibitors (20 g/ml leupeptin, 2 g/ml antipain, 1 g/ml pepstatin, pH 7.0) and a phosphatase inhibitor cocktail (Sigma-Aldrich) using a handheld dispenser (Ultra-Turrax). Samples were centrifuged for 30 min at 10,000 or 20,000 g at 4°C. For immunoblotting, specific antibodies against peIF2 and eIF2 were used and GAPDH served as a loading control (all from Cell Signaling Technology, Boston, MA). HRP-conjugated goat antirabbit antibody (Vector Laboratories) was used as a secondary antibody. Proteins were visualized using Amersham Hyperfilm ECL (GE Healthcare). Films were scanned (GS-8000; Bio-Rad) and signal density was determined with the Quantity One® program (all from Bio-Rad Laboratories). The target protein signal was normalized to GAPDH and the fold difference of wild-type to Atgl-ko samples was determined.

Statistical analysis
Data are expressed as mean ± SEM. Data were tested on normality prior to statistical analysis. Statistical significance was determined between two groups with unpaired two-tailed Student's t-test when normally distributed. Mann-Whitney test was performed when data did not show normality. Statistical significance of more than two groups was analyzed using one-way ANOVA. Group differences were considered significant for *P < 0.05, **P < 0.01, and ***P < 0.001. All statistical analyses were performed with Prism 6.0 (GraphPad Software Inc., La Jolla, CA).

Environmental and genetic conditions leading to brown-to-white adipocyte conversion and CLS formation in "whitened" BAT
All raw data of the mean adipocyte area and CLS density of the experimental mouse models examined in the present study are reported in supplemental Table S1.
Warm acclimation. Ambient temperature has a profound influence on body metabolism and energy expenditure, importantly affecting BAT morphology and thermogenesis (13,14,24). The impact of ambient temperature on BAT macrophage infiltration was assessed in interscapular BAT (iBAT) and mediastinal BAT (mBAT) from lean C57Bl/6j (B6) mice acclimated to cold (6°C) and warm (28°C) temperatures. In cold-acclimated mice, iBAT and mBAT were composed of typical multilocular adipocytes that exhibited a high degree of UCP-1 immunoreactivity (data not shown). MAC-2 immunohistochemistry demonstrated the absence of CLSs in iBAT from cold-acclimated mice (Fig. 1A, H) and their occasional presence in mBAT (Fig. 1C, H). Brown fat cells from lean mice kept at 28°C, a temperature close to thermoneutrality for rodents, acquired a white-like unilocular adipocyte phenotype. Notably, brown-to-white conversion did not involve significant changes in cell size either in iBAT or in mBAT (Fig. 1G). In contrast, brown-towhite conversion involved increased macrophage infiltration, which led to formation of several CLSs in iBAT (Fig.  1B, H); the density of MAC-2-positive CLSs in mBAT was also increased, even though the difference compared with cold-acclimated mice was not significant (Fig. 1D, H). To establish whether the increased CLS density seen in thermoneutral conditions was confined to BAT, inguinal WAT (iWAT) sections from cold-and warm-acclimated animals were examined. Very few CLSs were detected in iWAT from cold-acclimated mice (Fig. 1E, H), whereas in warm-acclimated animals, there was a significant increase in adipocyte area (Fig. 1G) that was, however, not matched by a significant increase in CLS density (Fig. 1F, H). These data suggest that brown adipocytes converting to a white-like unilocular phenotype at high ambient temperatures have a limited ability to store lipids and grow, and that they become more prone to death and to be cleared by CLSs compared with white adipocytes.
Leptin receptor deficiency. In obese mice, brown adipocytes display a white-like appearance similar to the phenotype found in warm-acclimated mice (13,24). We therefore investigated to determine whether iBAT from leptin receptor-deficient db/db mice showed signs of whitening and CLS formation like warm-acclimated mouse iBAT. Typical brown adipocytes were detected in db/+ mice ( Fig. 2A), whereas in db/db mice they showed a white-like appearance and a significant size increase (Fig. 2B, E). CLSs were never detected in iBAT from control animals ( Fig. 2A, E), whereas they were frequently observed in iBAT from db/db mice (Fig. 2B, E). Comparison of brown and white adipocyte area in iBAT, mBAT, and iWAT from cold-acclimated (6°C) and warm-acclimated (28°C) mice documenting a similar brown adipocyte size in both groups and a significant enlargement of subcutaneous inguinal white adipocytes from warm-acclimated mice (n = 5). H: Comparison of CLS density in cold-and warm-acclimated mice showing that warm acclimation significantly increases CLS density in iBAT, to some extent in mBAT, but not in iWAT (n = 5). Data are expressed as mean ± SEM. Statistical significance was determined using unpaired two-tailed Student's t-test for iBAT and the Mann-Whitney test for mBAT and iWAT *P < 0.05; n.d., not detectable.
The -less mice. BAT-to-WAT conversion in warm-acclimated and obese mice is held to be mediated by low activity of the sympathetic nervous system (13,14,24), whose tonic noradrenergic signaling is required for brown phenotype maintenance (25)(26)(27). To explore the whitening process in the absence of the adrenergic stimulus, iBAT morphology and cell composition were studied in -less mice (21). In these -less mice, BAT exhibited a marked white-like appearance and adipocyte hypertrophy (Fig. 2D, F) compared with BAT of control mice (Fig. 2C, F). In line with the abovereported findings, BAT from control mice did not contain CLSs (Fig. 2C, F), whereas whitened BAT from -less mice was characterized by a large number of CLSs (Fig. 2D, F).
Atgl deletion. ATGL is a key enzyme for lipolysis, the intracellular hydrolysis of triglycerides. Global Atgl ko induces mild obesity and adipose tissue expansion, even in mice fed a standard chow diet (19). ATGL activity is crucial for normal BAT morphology and function because its deletion leads to severe BAT hypertrophy and to the acquisition of a WAT-like phenotype (19,28). However, it is unclear whether BAT whitening in Atgl-ko mice also involves changes in macrophage infiltration and CLS formation. To gain insight into the question, adipocyte size and CLS density were assessed in iBAT, mBAT, and iWAT from wild-type and Atgl-ko mice kept in standard laboratory conditions. In wild-type mice, brown and white adipocyte appearance and size were normal in all three depots, whereas CLSs were rare in iBAT (Fig. 3A, H) and absent in mBAT (Fig. 3C, H) and iWAT (Fig. 3E, H). As expected (19,28), genetic Atgl ablation involved brown-to-white-like conversion in brown depots and an important size increase of both brown and white adipocytes (Fig. 3G). Interestingly, the size of transformed iBAT and mBAT brown adipocytes from these mice was not significantly different from the size of iWAT white adipocytes (Fig. 3G). In striking contrast, CLS density was very high in iBAT (Fig. 3B, H), high in mBAT (Fig. 3D, H), and low in iWAT (Fig. 3F, H). In this model of BAT whitening, enlarged white-like brown adipocytes therefore die in large numbers when they reach a size and lipid content that are still tolerated by white adipocytes. In epididymal WAT (eWAT) and mesenteric WAT visceral fat, Atgl deficiency involved an enlargement of white adipocytes associated with increased CLS density compared with control mice (supplemental Fig. S1).

Brown adipocyte death and CLS index
To estimate the number of healthy versus dysfunctional, or dying, whitened brown adipocytes in iBAT, we assessed the immunohistochemical expression of perilipin, a widely used marker of adipocyte viability (29)(30)(31). The brown adipocytes from control mice were strongly and consistently perilipin-immunoreactive (Fig. 4A). In contrast, perilipin expression was low, very low, or absent in a variable number of whitened brown adipocytes from the experimental groups. In particular, slightly positive or negative adipocytes were about 5% in warm-acclimated mice, about 5-10% in db/db mice, about 20-30% in -less mice, and about 30-40% in Atgl-ko mice (Fig. 4B). As also shown by double labeling experiments and confocal microscopy analysis (supplemental Fig. S2), specific perilipin staining was never detected in CLSs (Fig. 4C), which confirmed that they were formed of brown adipocyte debris, mainly lipids, surrounded by macrophages. Notably, perilipin staining was detected around small cytoplasmic lipid droplets in some endothelial cells of parenchymal capillaries from the whitened iBAT of Atgl-ko mice (Fig. 4D).
The above data suggest that lipid overload and/or functional inhibition induce brown adipocyte conversion to a white-like phenotype, which makes them highly prone to death and to be cleared by macrophages. Given that cell size is strongly involved in inducing such proneness to death (6,7,12), all mouse models were compared by their CLS index, i.e., the ratio between CLS density (CLSs/10,000 adipocytes) and the respective mean adipocyte area (in square micrometers). Interestingly, the CLS index was highest in Atgl-ko mouse iBAT (Fig. 5). ATGL deletion thus seems to be the most robust model to study the mechanisms and metabolic consequences of brown adipocyte whitening and CLS formation.

Ultrastructure of whitened BAT from Atgl-deficient mice
TEM examination of iBAT specimens from control mice revealed typical brown adipocytes filled with several small lipid droplets and containing numerous large mitochondria packed with laminar cristae (Fig. 6A). In Atgl-ko mice, the brown adipocytes were enlarged and their cytoplasm contained large and coalescing lipid droplets and numerous mitochondria (Fig. 6B). The majority of examined mitochondria still contained packed laminar cristae (Fig. 6D), but some mitochondria exhibited signs of degeneration characterized by the absence of cristae in restricted portions of the organelle (inset of Fig. 6B). Some whitened brown adipocytes exhibited dilated ER and cholesterol crystals (Fig. 6C). In the extracellular space, an increased number of collagen fibrils was found around the whitened brown adipocytes (Fig. 6C) and small lipid droplets were occasionally detected in the endothelial capillary wall (Fig. 6D), in line with perilipin immunostaining observed on light microscopy. All sections contained a large number of macrophages, some of which were clustered around adipocyte debris, sometimes forming syncytia (Fig. 6E). On the whole, the ultrastructure of the CLSs found in Atgl-ko mouse iBAT was very similar to that seen in subcutaneous and visceral WAT from obese animals (6, 7), suggesting that CLSs are a morphological "hallmark" of both white and brown adipocyte death.

Whitening of brown adipocytes in Atgl-deficient mice induces BAT inflammation and NLRP3 inflammasome activation
Increased CLS density is usually directly proportional to increased expression of inflammatory marker genes, as previously described (4-7). Therefore, we measured Comparison of CLS density in Atgl-ko and wild-type mice showing that Atgl deletion involves a significant increase in CLS density in all depots from the former mice (n = 3). Data are expressed as mean ± SEM. Statistical significance was determined using unpaired twotailed Student's t-test; *P < 0.05; **P < 0.01; ***P < 0.001. inflammatory and inflammasome marker expression in Atgl-ko mice. In line with the morphological data, the high CLS density found in BAT from Atgl-ko mice was associated with a robust inflammatory response in iBAT, which was reflected in an increased abundance of the mRNA of classic adipose tissue inflammation markers, such as F4/80, Tnf, and Mcp-1 (Fig. 7A). Interestingly, markers for M2 macrophage polarization (IL-10 and Mgl1), which indicate the presence of M2 macrophages, were also significantly upregulated in Atgl-ko mouse iBAT (Fig. 7A). Finally, induction of the expression of NLRP3 inflammasome markers, such as Casp-1 and IL-18, in iBAT of Atgl-ko mice (Fig. 7B) suggests that inflammasome activation might be the driver of brown fat inflammation and CLS formation in these animals, as previously hypothesized for obese WAT (12).

BAT inflammation in Atgl-deficient mice is independent of ER stress and mitochondrial dysfunction
ER stress, mitochondrial dysfunction, and the consequent presence of large amounts of reactive oxygen species in adipose tissue are held to be involved in adipocyte death and in the development of adipose tissue inflammation and dysfunction (10,11,(32)(33)(34). Therefore, established markers of ER stress and mitochondrial dysfunction were measured to assess whether either or both mechanisms might underpin inflammasome activation and inflammation in whitened Atgl-ko mouse iBAT. ER stress induction was analyzed by measuring the mRNA abundance of several members of the unfolded protein response, such as Grp78/Bip, Xbp-1 splicing, and eIF2 phosphorylation. However, there were no changes in Xbp-1 splicing (Fig.  8A), Grp78/Bip and Chop expression (Fig. 8B), or eIF2 phosphorylation (Fig. 8C). Similar results were obtained from the analyses of mitochondrial degeneration markers because neither cytochrome c release nor mtDNA content were altered in iBAT from Atgl-ko mice compared with controls (Fig. 8D, E). Finally, the expression of the oxidative stress marker, catalase, was not altered (Fig. 8F), whereas Gpx1 only moderately increased (Fig. 8G) in iBAT from Atgl-deficient mice.

iBAT from AAKO mice contains numerous CLSs and shows signs of inflammation
To exclude the possibility that the phenotype found in Atgl-deficient mouse BAT was due to the lack of ATGL in macrophages or in other cell types, the analysis was extended to AAKO mice, where adiponectin-driven Cre expression was used to selectively target Atgl deletion in white and brown adipocytes (20). As already reported (20), the iBAT of AAKO mice exhibited a marked white-like appearance. Both iBAT and eWAT from these mice exhibited a significant increase in adipocyte size and CLS density compared with wild-type controls (supplemental Fig. S3); notably, the increase in CLS density was especially marked in iBAT. The expression of some inflammatory markers, including F4/80, Chop/Ddit, GPX1, and Casp-1, was significantly increased in the iBAT of AAKO mice compared with control mice (supplemental Fig. S3). NLRP3 mRNA was not detected in the iBAT of wild-type control mice, but showed a low level of expression in AAKO mice. Interestingly, no signs of inflammasome activation were detected in the eWAT of AAKO mice (supplemental Fig. S3). Altogether, these data show that selective ATGL deletion in adipocytes is able to induce cell hypertrophy and death and to lead to macrophage infiltration and CLS formation that are particularly pronounced in BAT, where these processes may ultimately sustain tissue inflammation.

DISCUSSION
Obesity is a multifactorial disease resulting from a chronic imbalance between food intake and energy expenditure. An enduring and sustained positive energy balance leads to the progressive enlargement of subcutaneous and visceral WAT depots, due to an increase in white adipocyte number (hyperplasia) as well as size (hypertrophy). When the adipose organ is no longer capable of buffering the excess nutrients, ectopic lipid deposition occurs in liver, muscle, heart, and pancreas (35)(36)(37). Importantly, the enlarged obese WAT develops a low-grade inflammatory state (4,5). Histologically, obese WAT inflammation is marked by the presence of CLSs, where aggregates of macrophages surround dead adipocytes to clear the extracellular space of adipocyte debris, especially the large lipid droplets (4-7). Stressed and dysfunctional obese adipocytes and dead adipocytes trigger the production of proinflammatory adipokines and chemoattractants, which in turn instigate inflammatory cell infiltration, eventually leading to a chronic inflammatory state (8,35,38). The consequent upregulation of inflammatory cytokines and downregulation of anti-inflammatory adipokines are crucial to induce the major metabolic and cardiovascular consequences of obesity (35,38,39).
While numerous studies have explored the morphological and pathophysiological aspects of obese WAT, especially visceral WAT, whose involvement is linked to the more severe morbid consequences of obesity (40), obese BAT has not been extensively investigated.
In genetic or acquired obesity, BAT function declines and the tissue acquires a white-like appearance (whitening). Histologically, whitened brown adipocytes show a unilocular rearrangement of their lipid content but, as demonstrated by TEM, their cytoplasm still contains brownlike typical mitochondria filled with numerous packed cristae (13). By immunohistochemistry, whitened brown adipocytes retain a weak mitochondrial UCP1 immunoreactivity, but also begin to express leptin (18,41), the typical white adipocyte hormone, and reduce the expression of many classic brown genes (42). Even though de novo white adipocyte formation and/or infiltration of adjacent white adipocytes, which are usually located at the periphery of BAT lobules, could contribute to BAT whitening, these data suggest that in some conditions involving lipid overload and/or inhibited oxidation, brown adipocytes can directly convert to white-like adipose cells with few or no thermogenic properties (43,44).
Importantly, data from rodents and humans stress that the development of obesity and its complications largely depends on the balance between WAT, the main energy store  . The mRNA levels of the inflammation and inflammasome markers were measured by quantitative real-time PCR. Target gene abundance was normalized to 36b4 and expressed relative to wild-type levels of each marker. Data are expressed as mean ± SEM. Statistical significance between groups was calculated using unpaired two-tailed Student's t-test or Mann-Whitney test for Mgl1 mRNA abundance and is expressed as: *P < 0.05; **P < 0.01; ***P < 0.001. of the body, and BAT, whose task is to produce energy via nonshivering thermogenesis. For instance, in mice, BAT ablation leads to obesity (45), ectopic UCP1 expression in WAT results in resistance to obesity (46), and reduced BATdependent nonshivering thermogenesis impairs insulin sensitivity (47). In humans, a reduced brown adipose phenotype is associated with insulin resistance (48), and obese individuals show BAT atrophy associated with visceral fat accumulation, aging, and hyperglycemia (49), suggesting that defective BAT function may induce and/or exacerbate obesity and its complications.
However, the fate of whitened brown adipocytes and whether whitened BAT develops inflammation, possibly contributing to the systemic inflammatory state found in obesity, have not been explored in depth. The models of energy imbalance used in the present study suggest that whitening may be caused by leptin receptor deficiency, absence of -adrenergic signaling to brown adipocytes, or lipase deficiency. In each of these conditions, brown fat whitening results in macrophage infiltration, adipocyte death, and CLS formation. Notably, the most marked effect of brown-to-white conversion was seen in Atgl-deficient mice (19,28), which, even in normal temperature and dietary conditions, developed enlarged brown adipocytes and BAT inflammation with increased CLS formation and expression of inflammatory marker genes, such as Tnf and Mcp-1. These data are consistent with the findings of recent studies of whitened BAT from diet-induced obese mice, which have reported macrophage recruitment to BAT and increased expression of proinflammatory mediators (47,(50)(51)(52). Some ultrastructural aspects of the whitened brown adipocytes of Atgl-ko mice (such as cholesterol crystals) and the NLRP3 inflammasome activation (increased expression of Casp-1 and IL-18) suggest that brown adipocytes may undergo pyroptotic cell death, as described in hypertrophic and dysfunctional white adipocytes (12). ER stress and mitochondrial dysfunction have been implicated in causing adipocyte dysfunction and death and adipose tissue inflammation (10,11,33,34). Even though TEM data showed ER dilatation and mitochondrial degeneration in some whitened brown adipocytes, no clear evidence of ER stress or mitochondrial dysfunction in the iBAT of Atgl-ko mice was found on a molecular level. Therefore, in our study, few effects of BAT whitening on mitochondrial function occur in Atgl-ko mice, in accordance with the recent demonstration that iBAT retains fairly normal thermogenic activity in AAKO mice (53). However, a previous experiment had clearly identified mitochondrial dysfunction in the cardiac muscle of ATGL-ko mice (54). A close involvement of mitochondria in brown adipocyte whitening is also suggested by a recent study showing that loss of the scaffold protein, p62, in adipose tissue leads to BAT whitening, impaired nonshivering thermogenesis, and obesity that are linked to mitochondrial dysfunction (55). Impaired vascularization in vascular endothelium growth factor (VEGF)-ko mice has also been shown to be a primary factor leading to BAT whitening, which was characterized by lipid accumulation, mitochondrial , and eIF2a phosphorylation (C), shows no difference between wild-type and Atgl-ko mice, thus excluding ER stress in iBAT from Atgl-ko mice (n = 4, refed mice). Cytochrome c release in iBAT lysates (D) and qPCR analysis of iBAT mitochondrial (mt) DNA content (E): the relative abundance of the mtDNA-encoded gene Mt-Co1 was measured by qPCR, normalized to the nuclear single copy gene, Ndufv1, and expressed relative to wild-type levels (n = 4-6, refed mice), showing that mitochondrial integrity is not affected in Atgl-ko mouse iBAT. The mRNA abundance of catalase (F) and Gpx1 (G) in Atgl-ko mouse iBAT, which reflect oxidative stress, remained unchanged (n = 5-6, refed mice). Relative mRNA levels were measured by quantitative real-time PCR. Target gene abundance was normalized to 36b4 and expressed relative to wild-type levels of each marker. Data are expressed as mean ± SEM. Statistical significance between groups was calculated with unpaired two-tailed Student's t-test or Mann-Whitney test for Gpx1 mRNA abundance. dysfunction, and impaired noradrenergic signaling (56). In addition, it has been suggested that abnormal amounts of intracellular fatty acids, as could occur in ATGL-depleted adipocytes, can activate the NLRP3 inflammasome directly (57). Clearly, different molecular mechanisms can contribute to BAT whitening and obesity-related BAT dysfunction, and the primary cellular mechanism leading to whitening in Atgl-ko mice remains to be determined. However, our data suggest that enlargement or simply the acquirement of a spherical shape by whitened brown adipocytes that also show prominent organelle alterations by TEM may be sufficient to induce inflammasome activation in these cells, even in the absence of biochemically detectable signs of ER stress and mitochondrial dysfunction from the whole BAT depot.
The present data show that the histological appearance of brown and white adipocyte death is very similar at both light and electron microscopy. Indeed, brown and white stressed and dysfunctional adipocytes both appear to induce macrophage recruitment and to die, giving rise to CLSs. Thus, CLSs create the tissue microenvironment where adipocyte debris can adequately be resorbed. Importantly, they are also found in healthy brown and white fat, albeit in very small number [(6, 7) and present results]; therefore, CLSs are not necessarily a pathological feature, but one that may reflect normal adipocyte turnover. In this connection, numerous CLSs were unexpectedly detected in the whitened BAT of mice exposed to warm acclimation, which is not a pathological condition, but a condition that induces a significant size increase, further underlining the importance of normal morphology maintenance for adipocyte viability. This raises the interesting hypothesis of a "physiological" role of CLSs in normal fat turnover, and of the existence of "molecularly different" CLSs that in some instances are proinflammatory, while others are involved in normal homeostatic fat processes.
According to all the models of energy imbalance and obesity that were explored in this study, brown adipocytes show a substantial death rate. Thus, we can hypothesize that in lipid overload conditions brown and white adipocytes display a strikingly different behavior. White adipocytes, such as those from subcutaneous depots, have an outstanding ability to expand and reach a large critical death size, whereas whitened brown adipocytes have a more limited lipid storage capacity and a smaller critical death size, which results in a greater proneness to death. This mechanism may be of crucial importance for the pathophysiology of human obesity because, as suggested by NLRP3 inflammasome activation in whitened BAT of Atglko and AAKO mice, the death of brown adipocytes beyond a certain threshold may induce BAT inflammation; this process, combined with WAT inflammation, may compound the systemic inflammatory state typical of obesity. As documented by fluorodeoxyglucose positron emission tomography studies combined with computed tomography, metabolically active BAT is found in humans at visceral sites, mainly around the aorta and the proximal portion of its thoracic and abdominal branches (49,58,59). Chronic exposure to an obesogenic environment may induce whitening of these brown adipocytes which, given their high propensity to die, may thus provide a major contribution to obesity-related inflammation, offering a novel explanation for the strong role of visceral fat in determining the most adverse metabolic and cardiovascular outcomes of obesity.