Autophagy Is Dispensable for Macrophage-Mediated Lipid Homeostasis in Adipose Tissue

Interest in adipose tissue (AT) macrophages (ATMs) developed about a decade ago when it was discovered that obesity increases ATM populations (1,2). Since then, much work (3–11) has focused on the inflammatory functions of ATMs and their contribution to obesity-induced pathology. More recently, studies have uncovered noninflammatory, adaptive, and homeostatic roles for ATMs, including their role in local lipid catabolism and buffering. With the onset of obesity, ATMs accumulate lipid and activate a catabolic program that is lysosome dependent (12–16). In mice and humans, deficiency of lysosomal acid lipase (Lipa), which encodes the lysosomal enzyme required for hydrolysis of triglycerides (TGs) and cholesterol esters, leads to AT atrophy and severe hepatic steatosis (17,18). The expression of Lipa specifically by macrophages prevents the complications of Lipa deficiency in mice, revealing a critical need for lipid metabolism by macrophages in AT homeostasis (17,19–21).

In most cells, including macrophages, neutral lipids accumulate in lipid droplets (LDs), specialized organelles that store and release lipid through classic neutral lipolysis (22). However, recent studies (23–25) revealed that lipids in LDs also undergo hydrolysis in lysosomes via autophagy, providing a pathway for lysosome lipid delivery in ATMs. Indeed, foam cells, the lipid-laden macrophages of atherosclerotic plaques, catabolize lipids in part through autophagy, and inhibition of autophagy leads to the accumulation of lipids (24). Autophagy is an evolutionarily conserved process by which single molecules, macromolecules, and organelles are delivered to lysosomes for degradation (26–29). The metabolic state of a cell or tissue profoundly regulates autophagy, with fasting or nutrient deprivation typically activating autophagy. The effects of overnutrition are more complex. In liver and foam cells, obesity impairs autophagy, whereas, conversely, when whole AT is analyzed, obesity has been found to activate autophagy (26–35). We and others had hypothesized, therefore, that autophagy delivers lipids to lysosomes in ATMs of obese animals and that impairment of autophagy would phenocopy the effects of Lipa deficiency (i.e., lipoatrophy and hepatic steatosis). Using both in vivo and in vitro systems, we found that excess lipids and obesity activate autophagy in ATMs, but, surprisingly, impairing autophagy genetically or pharmacologically does not alter the lipid content of ATMs in lean or obese animals, does not alter AT mass, and does not cause hepatic steatosis or impair systemic metabolism.

Male C57BL/6J and B6.V-Lep/obJ mice were obtained from The Jackson Laboratory (Bar Harbor, ME) at 10–12 weeks of age. To generate mice with myeloid-specific deletion of Atg7 (Mac^Atg7 KO mice), we crossed LysMCre B6.129P2-Lyz2^tm1(cre)Ifo/J (The Jackson Laboratory) and C57BL/6 Atg7^F/F mice (a gift from Masaaki Komatsu, Tokyo Metropolitan Institute of Medical Science, Tokyo, Japan), subsequently generating Mac^Atg7 KO (Atg7^F/F; LysMCre) and littermate controls (Atg7^F/F) (36,37). Mouse genotypes were determined using PCR amplification of DNA from tail lysate or blood immune cells (36). Mice were housed in a pathogen-free barrier facility, in ventilated cages with free access to autoclaved water, and were fed either a low-fat pellet diet (5% calories from fat) (PicoLab Rodent Diet 20; Purina Mills Inc) or high-fat diet (60% calories from fat) (D12492; Research Diets). Mice were maintained on a 12-h light/dark cycle and housed with three to five male mice per cage. Metabolic measurements were made beginning at 10 weeks of age until sacrifice at 19 weeks of age (lean mice) or 26 weeks of age (DIO mice). The Columbia University International Animal Care and Use Committee approved all procedures.

Following CO₂ asphyxiation and cervical dislocation, perigonadal AT (PGAT) was isolated using sterile techniques. PGAT was placed in FACS buffer (PBS; GIBCO), 0.2% BSA (Sigma-Aldrich), and 5 mmol/L EDTA (Sigma-Aldrich), and was minced into fine (<10 mg) pieces. Samples were centrifuged at 500g for 5 min. PGAT was then digested in DMEM (Invitrogen) with 10 mg/mL BSA and digestive enzymes (0.14 units/mL Liberase TM (Roche Applied Science) and 50 units/mL DNase I (Sigma-Aldrich) for 30 min at 37°C at 185 rpm on an orbital shaker. The digested material was passed through a sterile 250-µm nylon mesh (Sefar Filtration Inc.) and centrifuged for 5 min at 500g. The stromal vascular cells (SVCs) comprising the pellet were washed with FACS buffer and centrifuged at 500g for 5 min. For flow cytometry, SVCs were resuspended in erythrocyte lysis buffer (BD Biosciences). For immunofluorescence staining, SVCs were resuspended in culture medium (DMEM, 10% FBS [Invitrogen], and 1% penicillin-streptomycin [Invitrogen]) at 500,000 cells/mL, and were cultured overnight at 37°C, with 5% CO₂ in four-well chamber slides (LabTek). For Western blot analysis, cultured SVCs were treated with 20 μmol/L chloroquine (CQ; Sigma-Aldrich) or deionized H₂O for 16 h, collected using 0.02% EDTA (Sigma-Aldrich), and resuspended in Tissue Extraction Reagent I (Invitrogen) with a protease inhibitor cocktail (Sigma-Aldrich).

Bone marrow (BM)-derived ATMs (BM-ATMs) were differentiated as previously described (13). Briefly, BM cells were collected from femurs of 6- to 10-week-old control (Atg7^F/F), Mac^Atg7 KO (LysMCre⁺Atg7^F/F), control (Lipa^+/+), or Lipa knockout (KO; Lipa^−/−) mice by flushing with BM culture medium (minimum essential medium α [Invitrogen], 10% FBS, 1% nonessential amino acids [Invitrogen], 1% penicillin-streptomycin) (18). Cells were plated using 50–60 × 10⁶ cells/100 mL in a 175-mL tissue culture flask at 37°C in 5% CO₂. On the second day, nonadherent cells were collected, centrifuged at 500g for 5 min, and plated at 1.5 × 10⁶ cells/2 mL in BM culture medium supplemented with 30 ng/mL human macrophage colony stimulating factor (M-CSF; R&D Systems). After 3 d of culture with M-CSF, adherent cells were differentiated into either BM-macrophages (BM-Macs), BM-ATMs, or BM-foam cells. BM-Macs were generated by addition of fresh M-CSF supplemented BM and continued culture for five days. To generate BM-ATMs, on day three of culture 100 mg of PGAT was placed in porous tissue inserts (BD Biosciences) and added to adherent cells wells. To generate BM-foam cells, 150 μg/mL acetylated LDL (Biomedical Technologies, Inc.) was added to BM-Macs for 24 h (24,31). Lysosomal function was inhibited in cells by adding 20 μmol/L CQ for 16 h. On the eighth day, cells were collected using 0.02% EDTA (Sigma-Aldrich) for flow cytometry or protein isolation, or were used for immunofluorescence staining.

SVCs were resuspended in erythrocyte lysis buffer incubated at room temperature for 3 min and centrifuged at 500g for 5 min. Erythrocyte-free SVCs or collected BM cells were resuspended in FACS buffer at 7 × 10⁶ cells/mL with 2% Fc Block (BD Biosciences) for 30 min at 4°C. Fluorophore-conjugated antibodies were added for 30 min at 4°C with rotation, as follows: anti-CD45.2-Percp-Cy5.5 (1:100, v/v) immune cell marker (BD Biosciences), anti-F4/80-APC (1:20, v/v) macrophage marker (AbD Serotec), anti-CD11b (1:100, v/v) macrophage marker (Invitrogen), anti-CD11c-PE-TR (1:100, v/v), ATM marker (Invitrogen), and 4,4-difluoro-1,3,5,7,8-pentamethyl-4-bora-3a,4a-diaza-s-indacene (BODIPY 493/503)–FITC (1:300 v/v) lipid marker (Invitrogen). Cells were resuspended in FACS buffer with DAPI (Invitrogen) and analyzed using an LSRII Flow Cytometer (Becton Dickson). Data analysis was performed with FlowJo software.

Cultured cells were stained with 1:300 (v/v) BODIPY for neutral lipid and 5 nmol/L LysoTracker (Invitrogen) for lysosomes for 30 min, then were fixed with zinc formaldehyde (Z-Fix; Anatech Ltd.) for 10 min. For intracellular staining, cells were permeabilized using ice-cold methanol (Fisher Scientific) at −20°C for 10 min. Cells were blocked in blocking buffer (PBS, 10% goat serum; Sigma-Aldrich) for 1 h at room temperature. Primary antibodies were diluted in antibody dilution buffer (Cell Signaling Technology) containing PBS, 1% BSA (Sigma-Aldrich), and 0.3% Triton X-100 (Sigma-Aldrich), and were incubated overnight at 4°C. The following antibodies were used: anti-F4/80 (1 μg/mL), macrophage antibody (Invitrogen), anti-LC3 (1:250, v/v), an autophagosome marker (Cell Signaling Technology). On the following day, cells were incubated with secondary antibodies for 2 h at room temperature, as follows: 1:400 (v/v) Cy5 or Cy3 (Invitrogen). Samples were washed and incubated with 1:4,000 (v/v) DAPI for 10 min to visualize the nuclei. Glass coverslips were mounted using Fluoro-Gel Mounting Solution (Electron Microscopy Sciences). Slides were imaged using Nikon A1R MP Confocal Microscope. The mean fluorescence intensity (MFI) of BODIPY and LC3 was measured within F4/80⁺ cells. Background (average fluorescent intensity outside of cells) was subtracted from fluorescence signals, and the MFI per macrophage was measured. MFIs were calculated from individual images of all images from one well. Data were analyzed on Nikon NIS Elements imaging software.

Tissues were fixed for 48 h in Z-fix, washed with 70% ethanol for 24 h, then embedded in paraffin. The 5-µm sections were cut at 50-µm intervals and mounted on glass slides, deparaffinized in xylene, and counterstained with hematoxylin-eosin. To stain neutral lipids, SVCs were fixed with Z-fix for 10 min and stained with Oil Red O solution (Sigma-Aldrich) for 10 min, washed three times in deionized water, and counterstained with hematoxylin-eosin for nuclear staining.

SVCs and BM cells resuspended in tissue extraction reagent I (Invitrogen) with a protease inhibitor cocktail (Sigma-Aldrich) were homogenized and sonicated. Protein concentration was measured using Bio-Rad Protein Assay. Twenty to fifty micrograms of protein was denatured at 96°C for 5 min in 4× protein sample buffer (200 mmol/L Tris 6.8, 8% SDS, 0.4% bromophenol blue, 40% glycerol, and 5% β-mercaptoethanol) and loaded on 14% SDS-polyacrylamide gels (Invitrogen). Samples were transferred on nitrocellulose (Thermo Scientific) or polyvinylidene fluoride (Fisher Scientific) using either a semidry or wet transfer system (Bio-Rad). Membranes were counterstained with Ponceau stain, washed with Tris-buffered saline with Tween-20 (TBST), blocked with a 5% nonfat milk solution, and incubated with primary antibodies (anti-LC3 [Cell Signaling Technology], anti-Atg7 [Cell Signaling Technology], anti-Atgl [Cell Signaling Technology], and anti–β-actin [Sigma-Aldrich]) at 4°C overnight. Membranes were washed with TBST, incubated with secondary antibodies, and developed using an enhanced chemiluminescence detection system (GE Healthcare Life Sciences) or the fluorescence Odyssey CLx Imaging System (LI-COR). Band intensities were determined using image analysis (Quantity One; Bio-Rad) and were normalized to β-actin.

Serum metabolites and hormones were measured after a 6-h fast. Blood was collected for glucose and serum analysis through a submandibular bleed. Glucose was measured using a Freestyle Blood Glucometer (Abbott Laboratories). Serum measurements included insulin (Mouse Ultrasensitive Insulin ELISA; ALPCO), TGs (Infinity Triglycerides Reagent; Thermo Scientific), and nonesterified fatty acids (NEFAs; NEFA-HR; Wako Diagnostics). Body composition measurements were made with a minspec TD NMR analyzer (Bruker). Mice were fasted for 24 h before sacrifice. Tissues were stored in −80°C for later analysis, fixed for histological sectioning, or digested for SVC isolation.

Data are presented as the mean ± SEM. Differences were deemed significant if P values were <0.05 or corrected P values were <0.001. Sample comparison statistics were determined using a Student t test (two-tailed distribution, two-sample unequal variance) with a Bonferroni correction to analyze metabolic studies. All data were analyzed using Microsoft Excel and GraphPad Prism.

ATMs uptake and catabolize lipid in a manner that depends on Lipa and is strongly induced by obesity. The absence of Lipa in macrophages leads to the accumulation of lipid in ATMs, AT atrophy, and severe hepatic steatosis (12–16). We used flow cytometry, confocal microscopy, and immunohistochemistry to assess lipid content in ATMs from PGAT under various experimental conditions. Using BODIPY, a fluorescent dye that quantitatively stains neutral lipids, we found, consistent with previous reports, that the concentration of lipids in ATMs from leptin-deficient obese (Lep^ob/ob) mice is higher than in ATMs from lean (Lep^+/+) animals (Fig. 1A–D). Ex vivo, the catabolism of neutral lipid overtime is evident in ATMs from both lean and obese mice with decreases in lipid content between 4 and 48 h after isolation (Fig. 1E).

Using pharmacological inhibition of lysosomes, we previously demonstrated that lipid catabolism in ATMs is lysosome dependent (13). To determine whether LIPA, the only known lysosomal lipase, is required for lipid catabolism in ATMs, we differentiated BM stem cells from control (Lipa^+/+) and Lipa deficient (Lipa^−/−) mice in the presence of AT to generate BM-ATMs and Lipa KO BM-ATMs (18). As demonstrated previously, BM-ATMs mimic in vivo obese ATMs by expressing CD11c antigen, accumulating lipids, increasing lysosomal biogenesis, and forming multinucleated giant cells (13). Consistent with the catabolism of lipids in ATMs being Lipa dependent, the absence of Lipa increases the lipid content of BM-ATMs (Fig. 1F and G).

Autophagy is a process in which cellular content is delivered to lysosomes for degradation (26,27). LDs or portions of them can be incorporated into autophagosomes and conveyed to lysosomes through macroautophagy in a process dubbed lipophagy (25,27). Lipophagy occurs in hepatocytes, AgRP⁺ neurons, and foam cells, and is altered by obesity (23–25,30,32). Lipophagy requires the conjugation of microtubule-associated protein light chain 3 (LC3-I) with a phosphatidylethanolamine to yield LC3-II as part of the functional complex that forms the autophagosome membrane (38). LC3-II is a marker of autophagy and flux through autophagic pathways and can be assessed by measuring the accumulation of LC3-II when lysosome function is inhibited (i.e., the accumulation of LC3-II occurs because fusion of autophagosomes with lysosomes is prevented) (38). Similar to lipid-laden foam cells (Supplementary Fig. 1), autophagic flux is increased more than threefold in BM-ATMs relative to BM-Macs (Fig. 2A and B) (24). The treatment of ATMs with CQ confirms that LC3-II accumulation in these cells is not due to a defect in autophagy machinery per se (Fig. 2C and D).

These data suggest that, in vitro, AT induces autophagy in macrophages. To determine whether obesity similarly activated autophagy in primary ATMs in vivo, we isolated SVCs from fat of lean (Lep^+/+) and obese (Lep^ob/ob) animals. Among SVCs, we found that obesity increased the LC3-II content of cells (Fig. 3A and B). Treatment with CQ further increased LC3-II content in SVCs from obese mice compared with those from lean mice, which is consistent with an increase in autophagic flux in primary cells (Fig. 3A and B). Using immunofluorescence, we found that LC3 content in both ATMs and non-ATMs cells was similar, but in obese mice LC3 specifically increased in ATMs (Fig. 3C). Consistent with our in vitro data, autophagy is increased in ATMs during the development of obesity (Fig. 3D). Fasting increases lipid accumulation in ATMs and broadly activates autophagy in some key tissue. AT does not behave the same as most organs. Autophagy, as measured by LC3-II, was not increased in whole AT or SVCs, suggesting that, unlike most tissues, overnutrition, not fasting, specifically activates autophagy in AT and its constituents (Supplementary Fig. 2).

The concurrent increase in lysosomal-dependent lipid catabolism and autophagy in ATMs during the development of obesity suggested that lipid delivery to lysosomes occurs via lipophagy. To directly assess the role of autophagy in ATM lipid catabolism and AT homeostasis, we deleted Atg7, a gene required for autophagy, from myeloid cells. Using a Cre/loxP recombination system, we crossed mice carrying a floxed Atg7 allele (Atg7^F/F) with mice in which the Cre-recombinase was inserted in the Lysozyme 2 gene and expressed in myeloid cells, including macrophages (36,37). We confirmed the deletion of Atg7 from circulating myeloid cells, BM-Macs, whole AT, and SVCs (Supplementary Fig. 3). Although some studies (39,40) have implicated autophagy in M-CSF–dependent monocyte differentiation into macrophages, we found no difference in the efficiency of differentiation between control and Mac^Atg7 KO BM-Macs and BM-ATMs (Fig. 4A).

To determine whether lipid catabolism in ATMs is dependent upon autophagy, we studied BM-ATMs and primary ATMs from control and Mac^Atg7 KO mice. Quantifying lipid content using a fluorescent neutral lipid dye, we found no difference in lipid content between control and Mac^Atg7 KO BM-ATMs (Fig. 4B and C). In contrast and consistent with previous reports of foam cell lipid catabolism being dependent in part on autophagy, Mac^Atg7 KO BM-foam cells had twice the lipid content of control BM-foam cells (Fig. 4D and E). To confirm the unexpected finding that autophagy does not regulate lysosomal lipid metabolism in ATMs, we pharmacologically inhibited autophagy in BM-ATMs with 3-methyladenine (3-MA), and, consistent with our genetic findings, 3-MA did not alter lipid content in BM-ATMs (Fig. 4F and G). In contrast, lysosomal inhibition with CQ did increase lipid content in control and Mac^Atg7 KO BM-ATMs (Supplementary Fig. 4). The data demonstrate that lipid catabolism in ATMs is lysosomal dependent but not autophagy dependent in vitro.

To study autophagy-dependent metabolism of lipids in vivo, we isolated SVCs from lean and obese (high fat–fed) mice. If impairing autophagy reduces the delivery of lipid to lysosomes, then we predicted an increase in either the proportion of lipid-containing ATMs or an increase in the lipid content of each ATM. However, targeted deletion of Atg7 did not alter the distribution or size of ATM subpopulations in Mac^Atg7 KO mice, in either lean or obese animals (Fig. 5A). Neither CD11c⁻ATMs (FBs), a subpopulation of macrophages low in lipid content, nor CD11c⁺ ATMs (FBCs), a subpopulation with the highest lipid content, were affected by the deletion of Atg7 (Fig. 5B).

The deficiency of autophagy also did not affect the lipid content of individual ATMs. The FB and FBC ATMs from control and Mac^Atg7 KO mice contained equal amounts of neutral lipid in both lean and obese mice (Fig. 5C and D). This was confirmed histologically (Fig. 5E). The lack of effect on lipid content was not due to an induction of the primary cytosolic lipase adipose TG lipase (ATGL/PNPLA2) (Supplementary Fig. 5) (41).

Impairing the lysosomal catabolism of lipids in macrophages in addition to increasing the lipid content of ATMs also leads to AT atrophy and severe hepatic steatosis (18). However, consistent with normal ATM lipid metabolism and in stark contrast to myeloid Lipa deficiency, there was no effect of autophagy deficiency on body weight (data not shown), body composition, fat mass distribution, or organ size (i.e., no hepatosplenomegaly) (Fig. 6A–C). The analysis of AT depots confirmed that adipocyte size and numbers were not histologically different between control and Mac^Atg7 KO PGAT (Fig. 6D). Macrophage lysosome lipid catabolism is also required for normal systemic lipid homeostasis (19). However, in Mac^Atg7 KO mice the concentrations of serum TGs and NEFAs were not different than control mice (Fig. 6E and F). Fasting and fed glucose and insulin values were also similar between control and Mac^Atg7 KO mice (Fig. 6G and H).

Although autophagy-dependent catabolism of lipid is not required for ATM function and AT homeostasis in lean animals, obesity increases autophagy in ATMs. To test whether under the increased stress of obesity AT requires ATM autophagy for homeostasis, we placed control and Mac^Atg7 KO mice on a high-fat diet. Control and Mac^Atg7 KO mice were both comparably susceptible to diet-induced obesity (Fig. 7A–C). There were no differences in fat depot, liver, and spleen weights between control and Mac^Atg7 KO mice (Fig. 7D). Histologically, PGAT from Mac^Atg7 KO mice was indistinguishable from the fat of control animals (Fig. 7E). Circulating TG, NEFA, glucose, and insulin concentrations were not different between obese control and Mac^Atg7 KO mice either in the fasted or fed state (Fig. 7F–I).

With the onset of obesity, adipocytes become hypertrophic and store lipid with reduced efficiency (42). Simultaneously, AT is infiltrated by immune cells, with ATMs comprising the majority (1,2). ATMs exert inflammatory actions that can contribute to the adverse effects of obesity on local and systemic metabolism (3–11). However, ATMs also serve adaptive functions that include a recently identified pathway of lysosomal lipid catabolism (12–15). The importance of this pathway is revealed by the following two observations: adipose atrophy and severe hepatic steatosis development in mice and people deficient in the enzyme required for lysosome lipid hydrolysis (LIPA) and prevention of the phenotypes by macrophage-specific expression of LIPA (17,19–21). Efforts to understand the mechanisms by which macrophage-lipid catabolism contributes to AT homeostasis promise to provide insights into AT biology and systemic lipid metabolism, but require defining the cellular and tissue-specific processes involved in lipid delivery to lysosomes.

Recent studies (25) have found that the delivery of neutral lipids to lysosomes from lipophagy is critical for lipid homeostasis in hepatocytes. Ablation of Atg7 in hepatocytes increases hepatic TGs and cholesterol, and has been implicated in obesity-induced hepatic steatosis (25). In AT, autophagy has been implicated in adipocyte differentiation with the deletion of Atg7 in adipocytes, resulting in a reduction of white AT mass and a compensatory increase in brown fat and energy expenditure (43,44). A similar lean phenotype is observed in mice lacking Atg7 in AgRP⁺ neurons (23). Given these findings and that macrophage autophagy contributes to lipid catabolism in foam cells as well as cellular differentiation and macrophage apoptosis, inflammasome activation, and clearance of dead cells/pathogens, it seemed likely that lipid delivery to lysosomes in ATMs would occur via autophagy (24,29,31,35,40,45). Furthermore, defective autophagy has been implicated in the defective processing of excess lipid and debris by foam cells in atherosclerotic plaques (35,46).

Here in studying the delivery of lipid to lysosomes in macrophages, we expected that autophagy would provide an important transport mechanism. Instead, we found that, although obesity does activate autophagy in ATMs, autophagy is dispensable for lipid-dependent degradation of TGs by ATMs both in vitro and in vivo. Furthermore, unlike impairing lysosomal lipid metabolism, disabling autophagy in macrophages has no discernable effects on AT or whole-body metabolism.

These findings argue that a pathway distinct from classic autophagy delivers lipid to lysosomes in ATMs. Two possibilities come to mind. It is possible that ATMs possess a novel, nonclassic autophagy pathway that does not depend upon Atg7 and cannot be inhibited by 3-MA but still delivers lipid from LDs to lysosomes. There have been reports (47) of pathways that do not require specific ATGs, but even those reports suggest that chemical inhibitors are capable of blocking delivery to lysosomes. We have no evidence for such a pathway, which would require identifying molecular components of such a process and developing tools to inhibit their function.

An alternative possibility is that the lipid in ATMs is not in stored LDs, but is contained within a distinct set of lipid vesicles. The delivery of LDs or parts thereof to lysosomes depends on autophagy because their surface is delimited by a monolayer of amphipathic molecules, which by itself is not capable of direct fusion with lysosomes (22,48). Other organelles, including endosomes, possess bilayer membranes that permit fusion with lysosomes directly (49). Macrophages are professional phagocytes and deliver a diverse range of molecules, including lipids, to endocytic vesicles. Such vesicles can be rapidly be targeted to lysosomes. Therefore, it seems possible that lipids in ATMs are in fact contained within non-LD vesicles. However, if lipids in ATMs are contained within such vesicles, the question arises as to how and in what form the lipid is taken up by ATMs. Hepatic- and intestinal-derived lipoproteins circulate and are the primary source of lipid in foam cells; adipocytes are not known to release lipoproteins (48,49). Indeed, on the basis of the RNA sequencing and microarray expression studies (Gene Expression Omnibus database and unpublished data), we have not identified the canonical protein constituents of known lipoprotein particles (e.g., ApoB) in AT (data not shown). If an endocytic pathway does deliver lipid to lysosomes in ATMs, then there must exist a heretofore undefined mechanism of lipid release by adipocytes.

Our findings were unexpected and point to previously unappreciated complexity in the relationship between ATMs and adipocytes. These studies suggest the existence of an intratissue lipid cycle not accounted for by current models and point to further areas of study.

Acknowledgments. The authors thank Eleanor Ables (Department of Medicine, Columbia University) for technical support and Theresa Swayne (Herbert Irving Comprehensive Cancer Center, Columbia University) for assistance with confocal microscopy and quantification. The authors also thank Masaaki Komatsu (Tokyo Metropolitan Institute of Medical Science, Tokyo, Japan) for providing the Atg7^F/F mice. In addition, the authors thank Hong Du (Indiana University, Bloomington, IN) for providing the Lipa^−/− mice.

Funding. This research was supported by the National Institutes of Health, National Institute of Diabetes and Digestive and Kidney Diseases, including grants R01-DK-066525 and R01-DK-101942, and support from the Columbia University Diabetes Research Center (P30-DK-063608), the New York Obesity Nutrition Research Center (P30-DK-026687), the Institute of Human Nutrition (T32-DK-007647), and the Irving Institute for Clinical and Translational Research, National Center for Advancing Translational Sciences (UL1-TR-000040).

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

Author Contributions. A.G. designed the studies, performed all experiments other than the lysosomal acid lipase BM experiments, interpreted the results, and wrote the manuscript. X.X. performed lysosomal acid lipase BM experiments. A.W.F. designed the studies, interpreted the results, and wrote the manuscript. A.G. and A.W.F. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.