Secretion of fatty acid binding protein aP2 from adipocytes through a nonclassical pathway in response to adipocyte lipase activity.

Adipocyte fatty acid binding protein 4, aP2, contributes to the pathogenesis of several common diseases including type 2 diabetes, atherosclerosis, fatty liver disease, asthma, and cancer. Although the biological functions of aP2 have classically been attributed to its intracellular action, recent studies demonstrated that aP2 acts as an adipokine to regulate systemic metabolism. However, the mechanism and regulation of aP2 secretion remain unknown. Here, we demonstrate a specific role for lipase activity in aP2 secretion from adipocytes in vitro and ex vivo. Our results show that chemical inhibition of lipase activity, genetic deficiency of adipose triglyceride lipase and, to a lesser extent, hormone-sensitive lipase blocked aP2 secretion from adipocytes. Increased lipolysis and lipid availability also contributed to aP2 release as determined in perilipin1-deficient adipose tissue explants ex vivo and upon treatment with lipids in vivo and in vitro. In addition, we identify a nonclassical route for aP2 secretion in exosome-like vesicles and show that aP2 is recruited to this pathway upon stimulation of lipolysis. Given the effect of circulating aP2 on glucose metabolism, these data support that targeting aP2 or the lipolysis-dependent secretory pathway may present novel mechanistic and translational opportunities in metabolic disease.


Animals
Plin1 Ϫ / Ϫ mice were obtained from the Jackson Laboratories (Bar Harbor, ME; stock number: 021887). Atgl fl ox/fl ox mice (C57/ BL6 background) were a kind gift from Dr. Erin Kershaw (University of Pittsburgh, Pittsburgh, PA), and Hsl fl ox/fl ox and Magl fl ox/fl ox mice (mixed background) were a kind gift from Dr. Rudolf Zechner (University of Graz, Graz, Austria). Adiponectin-Cre (Adipoq-Cre) mice (mixed background) were a kind gift from Dr. Evan Rosen (Beth Israel Deaconess Medical Center , Harvard Medical School) and were backcrossed onto the C57Bl/6J genetic background. In order to obtain adipose tissue-specifi c lipase-defi cient mice, homozygous fl oxed mice were crossed with Adipoq-Cre mice. Cre-positive progeny were then crossed with homozygous fl oxed mice to obtain fl ox/fl ox or fl ox/fl ox;Cre progeny for use in experiments. Adipose tissue explants were prepared from 10-to 16-week-old mice. Intralipid infusion experiments were performed using 24-week-old male mice. All mice were maintained on a 12 h light and dark cycle. Mice were maintained on regular chow diet (RD, PicoLab 5058 Lab Diet, 9% fat). The Harvard Medical Area Standing Committee on Animals approved all studies.

Intralipid infusion and quantifi cation of plasma aP2
Intralipid infusion was performed as previously described ( 33 ). Briefl y, WT male mice were fasted overnight before the experiments and were infused with Intralipid at 5 ml/kg/h (Baxter Healthcare Corporation) for 5 h. Blood was collected before and after Intralipid infusion. Plasma was separated by microcentrifugation of whole blood at 13,000 rpm for 30 min. Plasma aP2 was determined with an ELISA system specifi c to mouse aP2 (Biovendor Inc.).

Cell culture
3T3-L1 or in-house derived WT, aP2-deficient, or FABP ( aP2 Ϫ / Ϫ ; mal1 Ϫ / Ϫ )-defi cient preadipocytes were maintained in DMEM with 10% bovine calf serum. For differentiation, cells were seeded, and medium was changed after 2 days (day Ϫ 2). On day 0, adipocyte differentiation was induced in DMEM with 10% Cosmic Calf Serum (CCS) using 500 µM 3-isobutyl-1-methylxanthine (IBMX), 5 µg/ml insulin, 10 µM dexamethasone, and 10 µM Traditionally, the metabolic improvements observed upon aP2 defi ciency were attributed to changes in adipocytes or immune cells that in turn regulate secondary mediators to impact distant tissues and whole body homeostasis ( 10,11 ). However, we recently demonstrated that aP2 is secreted from adipocytes in a regulated manner and directly increases hepatic glucose output in mice ( 15 ). These studies also demonstrated that circulating aP2 levels are signifi cantly increased in dietary and genetic models of obesity, and that neutralizing circulating aP2 with a polyclonal antibody improved metabolic parameters and diabetes in mice ( 15 ). A growing body of association studies in humans also shows that circulating aP2 levels are positively correlated with obesity, metabolic pathologies, and cardiovascular disease (16)(17)(18)(19)(20)(21)(22)(23)(24)(25)(26). In addition, recent studies have revealed multiple other potential roles for extracellular aP2 ( 27,28 ). These observations strengthen the emerging view that aP2 is a hormone linking adipose tissue to systemic metabolism and suggest that targeting circulating aP2 may have potential therapeutic value for a wide array of complex diseases. Hence, understanding the mechanisms that underlie and regulate its secretion is of critical importance.
As is the case for other glucoregulatory hormones such as insulin and glucagon, aP2 secretion from adipose tissue is regulated by feeding and fasting, but it also occurs simultaneously with lipolysis ( 15 ). It is well established that lipolysis is increased in the context of long-term fasting, as well as obesity, insulin resistance, and diabetes ( 29,30 ). Lipolysis has mostly been explored in the context of signals that drive the response, rather than the hormonal signals that may be coupled to lipid breakdown and that integrate the status of adipose tissue depots with other metabolic responses of fasting or nutrient excess. We postulate that aP2 may be a key hormonal candidate coupled to lipolysis in adipocytes that supports systemic metabolic responses and adaptation to energy fl uctuations. However, neither the direct role of the lipolytic pathway in regulating aP2 nor the mode of secretion is currently known. Sequence analysis using SignalP 4.1 Server ( 31 ) reveals that aP2 lacks a signal peptide required for endoplasmic reticulum/Golgi-mediated classical secretion, and it is predicted by SecretomeP 2.0 Server ( 32 ) to be a nonclassically secreted protein. While this is consistent with the earlier observations that aP2 secretion is resistant to inhibitors of classical secretion, brefeldin A and monensin ( 15 ), the identity of the nonclassical route involved in aP2 secretion is unknown.
In this study, we utilized chemical and genetic approaches to demonstrate that aP2 secretion is highly dependent on the specifi c activity of lipases and is driven by elevated lipids. Furthermore, we found that lipolytic stimulation leads to recruitment of aP2 into multivesicular bodies (MVBs), a nonclassical secretory compartment, and externalization in exosome-like vesicles. These fi ndings support the concept that aP2 is an adipokine that responds to lipolytic activity in adipocytes to regulate metabolic status as controlled by liver and indicate a new pathway that is utilized by adipocytes to carry out, at least in part, their secretory and hormonal functions.

Electron microscopy
Adipocytes. 3T3-L1 cells were differentiated for 7 days and infected with lentivirus to express FLAG-tagged-GFP-aP2. On day 8, cells were kept in F12 medium for 1 h. They were then treated with isoproterenol (10 M) or water in F12 medium for 2 h. Cells were fi xed with formaldehyde (4% w/v in 200 mM HEPES pH 7.4 buffer) overnight at room temperature. Cells were washed two times using the same HEPES buffer and infi ltrated with a solution of 2.3 M sucrose for 48 h at 4°C. Subsequently, the adipocytes were washed two times in HEPES buffer containing 0.5% BSA and centrifuged to obtain a pellet. This was in turn embedded in 10% gelatin and cryosectioned ( 34 ). Sections were picked up using a 1:1 solution of 2% methyl cellulose and 2.3 M sucrose. Cells were then labeled with aP2 or GFP antibody followed by protein A gold (10 nm) for their fi nal visualization in transmission electron microscopy.
EVs. Isolated vesicles were adsorbed to nickel mesh grids that had been coated with a formvar fi lm and subjected to glowdischarge to increase hydrophilicity. After washing with water, the immunogold-labeled samples were silver-enhanced for 1 min using the HQ silver enhancement kit (Nanoprobes) and washed again with water, followed by positive staining with uranyl acetate. Samples were stabilized with a carbon coating prior to imaging with a JEM 1200EX electron microscope (JEOL USA) equipped with an AMT XR-60 digital camera (Advanced Microscopy Techniques Corp.).

Microsome isolation
Differentiated 3T3-L1 adipocytes were treated with IBMX to induce aP2 secretion as explained above, and cells were collected at the time points indicated. Microsomes were fractionated as described previously ( 35 ). Briefl y, cells were homogenized with a Tefl on pestle, nuclei were cleared at 800 g , and mitochondria were cleared at 6,000 g . The supernatant was ul tracentrifuged at 100,000 g for 1 h to pellet microsomes. The pellet was resuspended in lysis buffer and analyzed by Western blot.

EV isolation and continuous sucrose gradients
EV-depleted medium for use in conditioning was prepared by ultracentrifugation of DMEM with 20% CCS for 18 h at 28,500 rpm ( ‫ف‬ 100,000 g ) using SW32 rotor (Beckman Coulter). Media were then fi ltered (0.2 µm) and diluted to 10% serum with DMEM. Differentiated 3T3-L1 or in-house adipocytes were incubated for 24 h in this vesicle-depleted media for conditioning. Processing of the media from this point on was performed at 4°C. Conditioned media from differentiated adipocytes were collected and centrifuged at 200 g for 10 min, 500 g for 10 min twice, and then 2,000 g for 15 min to eliminate fl oating cells. Media were then fi ltered through 0.45 µm pore fi lters and ultracentrifuged at 9,000 rpm for 30 min with SW32 rotor to clear debris.
rosiglitazone. On day 4, medium was switched to DMEM with 10% CCS insulin and rosiglitazone. On day 6 and thereafter, medium was supplemented only with insulin. Lipolytic stimulation with or without inhibitors was done on days 8-12 of induction. For induction of aP2 secretion, adipocytes were treated with 1 mM IBMX or 20 µM forskolin (FSK) in DMEM with 10% CCS for 1 h, and medium was refreshed for another hour. Medium was collected and microcentrifuged at 5,000 rpm for 10 min. Supernatants were used for further analysis. For experiments involving inhibitors, differentiated adipocytes were pretreated with H89 (50 µM), Atglistatin (kind gift from Dr. Rudolf Zechner, University of Graz, Graz, Austria) (10 µM), 76-0079 (kind gift from Dr. Peter Kurtzhals and Dr. Christian Fledelius, Novo Nordisk, Denmark) (2.5 µM), CAY10499 (0.1 M), or DMSO for 2 h. Then cells were cotreated with IBMX and inhibitors as explained above. For lipid treatment experiments, stocks of free fatty acids prepared with fatty acid-free, low-endotoxin BSA were diluted in DMEM and 10% CCS. Adipocytes were incubated with vehicle or lipids (500 µM) for 24 h, and conditioned medium was processed as explained above. For extracellular vesicle (EV) isolation experiments after lipolytic stimulation, adipocytes were treated with indicated stimulants for 1 h, and medium was refreshed for another 3 h. Conditioned medium was processed further as explained below to isolate vesicles.

Adipose tissue explants and biochemical measurements
Atgl fl ox/fl ox or Hsl fl ox/fl ox mice were crossed with Adipoq-Cre mice to obtain adipose-specifi c lipase-defi cient models. Plin1 +/ Ϫ mice were intercrossed to obtain Plin1 Ϫ / Ϫ and WT littermates. Perigonadal adipose depots were removed for preparation of explants. Adipose tissue samples were washed in PBS and DMEM containing 10% CCS consecutively and minced into roughly 2 mm-size pieces with scissors. Explants were washed with DMEM with 10% CCS and incubated for 1 h in the same medium. In lipase-defi ciency models, lipolysis and aP2 secretion were induced with 1 mM IBMX, 20 µM FSK, 1 mM dibutyryl-cAMP (N 6 ,2 ′ -O -dibutyryladenosine 3 ′ ,5 ′ -cyclic monophosphate sodium salt), 10 µM isoproterenol, or 10 µM CL 316,243, and secreted aP2 levels in culture medium were evaluated using the protocol described above. In the perilipin1defi ciency model, explants were incubated for 16 h in DMEM with 10% CCS to assess the extent of aP2 secretion under basal conditions. Glycerol was measured using a commercial free glycerol determination system (Sigma-Aldrich), and NEFA levels were determined using HR Series NEFA-HR ( 2 ) (Wako Diagnostics) and normalized to total protein from cell or tissue lysates.

Vectors and expression
FLAG-tagged aP2 was cloned into pcDNA3 plasmid (Invitrogen). The lipid binding mutant (LBM) version of aP2 was produced using GeneTailor TM Site-directed mutagenesis system (Invitrogen). Adipocytes defi cient in endogenous aP2 ( aP2 were electroporated on day 10 of differentiation using Cell Line L Nucleofector system (Lonza) and AMAXA electroporator. Medium was changed the next day, and experiments were performed on day 13 of differentiation. FLAG-tagged green fl uorescent protein (GFP )-aP2 was produced by cloning aP2 into the lentiviral vector pRRLCMVGFP. 3T3-L1 adipocytes were infected with lentivirus for confocal and electron microscopy analysis.

Immunofl uorescence and confocal microscopy
3T3-L1 cells were differentiated for 7 days and infected with lentivirus to express FLAG-tagged-GFP-aP2. On day 8, cells were kept in F12 medium for 1 h. They were then treated with water or isoproterenol (10 M) in F12 medium for 1 h. Cells were fi xed with 4% formaldehyde for 15 min and stained with LipidTOX the role of lipases in aP2 secretion. We fi rst generated adipose tissue-specifi c ATGL-defi cient mice by crossing Atgl fl ox/fl ox mice with the Adipoq-Cre transgenic mice. After confi rmation of successful deletion of Atgl in the adipose tissue ( Fig. 2A ), we examined aP2 secretion. Perigonadal adipose explants from these mice exhibited diminished IBMX-and FSK-induced NEFA and glycerol release ( Fig.  2B, C ), validating the loss of ATGL activity. In control adipose tissue explants, a robust secretion of aP2 was detectable upon stimulation of lipolysis with multiple agents ( Fig. 2D-F ). Despite a similar intracellular aP2 protein level, aP2 secretion was undetectable from ATGL-defi cient explants upon stimulation of lipolysis with IBMX or FSK ( Fig. 2D ). We observed similar results with other stimulants of lipolysis such as dibutyryl-cAMP, CL 316,243, or Supernatant was collected and ultracentrifuged at 28,500 rpm with SW32 rotor for 90 min to pellet vesicles. The pellet was resuspended in PBS and repelleted. The pellet was resuspended in 2× SDS loading buffer for Western blot analysis or in PBS for sucrose gradient experiments. EVs resuspended in PBS were layered on top of a continuous sucrose gradient (0.25-2 M in 200 mM Tris) and ultracentrifuged at 28,500 rpm using SW41 rotor for 20 h at 4°C. Fractions (1 ml) were collected, and density was determined with a refractometer. Each fraction was then diluted in 10 ml PBS and ultracentrifuged at 28,500 rpm for 90 min using SW41 rotor. Pellets were resuspended in 2× SDS loading buffer for analysis by Western blotting.

Chemical and genetic manipulation of the lipolytic pathway inhibits aP2 secretion
During prolonged fasting, adipose tissue liberates its stored lipid depot via increasing lipolysis. This critical function is achieved by hormonal activation of the ␤ 3 -adrenergic receptor and a consequent increase in the secondary mediator cAMP. In turn, the cAMP-activated kinase, protein kinase A (PKA), indirectly activates ATGL and phosphorylates HSL, each of which liberate one molecule of free fatty acid from triacylglycerols (TAGs) and diacylglycerols (DAGs), consecutively. This cascade is finalized by the activity of monoacylglycerol lipase (MAGL), a constitutively active enzyme that hydrolyzes monoacylglycerols (MAGs) and liberates one free fatty acid and one glycerol molecule ( 36 ) ( Fig. 1A ).
We have previously shown that lipolytic stimuli that modulate this pathway at the level of the ␤ 3 -adrenergic receptor or those that increase cAMP induce aP2 secretion robustly from adipocytes in vitro, ex vivo, and in vivo ( 15 ). In order to test whether lipolysis and aP2 secretion occur independently, in parallel, or consecutively, and to characterize the molecular steps proximal to stimulated aP2 secretion, we fi rst evaluated aP2 secretion in the presence of chemical lipolytic inhibitors acting on different stages of the lipolytic pathway ( Fig. 1A ). We found that H89, an inhibitor of PKA, completely blocked IBMX-stimulated aP2 secretion from differentiated 3T3-L1 adipocytes along with lipolysis, as indicated by a decrease in glycerol release in inhibitor-treated cells ( Fig. 1B ). Having shown that blocking PKA activation upstream of the lipolytic pathway can diminish aP2 secretion, we next asked whether the downstream lipases themselves play a role in regulating this phenomenon. Indeed, treatment of cultured 3T3-L1 adipocytes with a synthetic ATGL inhibitor, Atglistatin ( 37 ); an HSL inhibitor, 76-0079 (NNC 0076-0000-0079, Novo Nordisk, Denmark) ( 38 ); and an HSL-MAGL dual inhibitor, CAY10499 ( 39 ), blocked or reduced IBMXinduced aP2 secretion and also inhibited lipolysis as indicated by glycerol release ( Fig. 1C-E ). These data suggest that aP2 release relies specifi cally on increased lipase activity in adipocytes.
To map this pathway in further detail and considering the potential nonspecifi c activities of chemical inhibitors, we used complementary genetic approaches to elucidate defi ciency, we observed a partial but signifi cant reduction in aP2 secretion from adipose tissue explants of Hsl fl ox/fl ox ; Adipoq-Cre mice upon stimulation with FSK, dibutyryl-cAMP, CL 316,243, and isoproterenol, but not with IBMX ( Fig. 3D-F ). In addition, we have not observed a signifi cant impact of MAGL-defi ciency on aP2 secretion from adipocytes in preliminary studies (data not shown). Collectively, our chemical and genetic data demonstrate that aP2 secretion is regulated directly by the activity of the adipocyte lipases ATGL and HSL, although in this experimental paradigm the impact of HSL defi ciency on aP2 secretion was smaller than that of ATGL defi ciency. Differential effects of ATGL and HSL defi ciency might suggest that there are lipase-specifi c properties regulating aP2 secretion. isoproterenol ( Fig. 2E, F ). Considering that ATGL is the primary enzyme for TAG hydrolysis in adipocytes ( 40 ), the dramatic effect we observed with ATGL defi ciency suggests that ATGL and products of TAG hydrolysis may be major regulators of aP2 secretion.
Next, we assessed aP2 secretion in adipose explants in which Hsl had been deleted ( Hsl fl ox/fl ox ;Adipoq-Cre). In this system, we also confi rmed the effective loss of HSL protein by Western blot analysis ( Fig. 3A ) and validated loss of stimulated lipolytic activity by decreased NEFA and glycerol release in HSL-defi cient samples ( Fig. 3B, C ). In these experiments, there was also a robust induction of aP2 secretion in control explants upon stimulation of lipolysis ( Fig. 3D-F ). Although not as dramatic as the effect of ATGL without an additional signal to induce lipolysis. We fi rst examined aP2 secretion in adipose tissue explants from perilipin1-defi cient mice, in which adipose triglycerides are continuously metabolized due to lack of the protective function of perilipin on lipid droplets (41)(42)(43)(44). In these explants, aP2 secretion was markedly increased compared with controls even in the absence of stimuli ( Fig. 4A ), accompanied by increased lipolysis, as measured by glycerol release ( Fig. 4B ).
These results raise the possibility that increasing lipid availability may increase aP2 secretion from adipocytes. To test this possibility in vitro, we treated cultured 3T3-L1 adipocytes with two saturated fatty acids, palmitic acid (C16:0)

Free fatty acid availability increases aP2 secretion
Numerous biochemical and structural studies have demonstrated that aP2 has a ligand binding domain formed by an antiparallel ␤ -barrel structure, and that this pocket can bind to several hydrophobic ligands, including free fatty acids ( 4 ). Because lipase activity is strongly tied to aP2 secretion and the end products of lipolysis can potentially bind to aP2, we next examined the potential role of lipid mobilization in regulating aP2 secretion from adipocytes and whether direct lipid binding to aP2 is a requirement for its secretion. To begin to address this question, we utilized conditions in which lipid breakdown or availability in adipocytes could be stimulated

A nonclassical vesicular pathway mediates aP2 secretion upon lipolytic stimulation
We have previously shown that aP2 release is resistant to inhibitors of classical secretion ( 15 ), consistent with its lack of a canonical signal peptide. To explore the nonclassical secretory pathway involved in aP2 secretion, we examined subcellular localization of aP2 upon lipolytic stimulation. First, to visualize localization in intact cells, we infected differentiated 3T3-L1 adipocytes with lentivirus to express GFP-fused aP2 and analyzed subcellular distribution by confocal microscopy with or without treatment with the ␤ -adrenergic receptor agonist isoproterenol. In resting cells, aP2 was diffusely distributed throughout the cytosol ( Fig. 5A ). However, upon induction of lipolysis, GFP signal became less diffuse and assumed a more peripheral and punctate pattern, suggesting that aP2 may be compartmentalized into structures that mediate its secretion. Nonclassical secretion can occur through several pathways, many of which include intracellular membranous structures and vesicles ( 46 ). Therefore, we asked whether aP2 localizes to a membranous compartment during lipolysis by conducting subcellular fractionation experiments. Differentiated 3T3-L1 adipocytes were treated with IBMX and harvested at several time points after induction. The cells were then fractionated to isolate total microsomes, which should contain a mixture of membranous and stearic acid (C18:0), and found that this resulted in signifi cantly increased aP2 secretion ( Fig. 4C ). Next, we administered a complex lipid mixture, Intralipid, into mice, and observed signifi cantly increased circulating aP2 levels ( Fig. 4D ). These data demonstrate that increasing lipids in vitro and in vivo was indeed suffi cient to increase aP2 release from adipocytes.
Finally, we utilized a lipid binding mutant (LBM) form of aP2 to further investigate whether direct binding to lipids was required for aP2 secretion. This mutant version of aP2 harbors two point mutations in the ligand binding pocket (R126L, Y128F) that disrupt hydrogen bonding with the carboxyl group of lipids ( 45 ). To prevent potential confounding effects of endogenous aP2, we reconstituted in vitro differentiated aP2-defi cient adipocytes with exogenous FLAG-tagged WT or LBM aP2 and compared their secretion in the presence of FSK ( Fig. 4E ). Interestingly, FSK stimulation induced the secretion of both WT and LBM aP2, suggesting that aP2 secretion is not entirely dependent on direct interactions between lipids and the aP2 ligand binding pocket. However, induction of LBM aP2 secretion tended to be lower, raising the possibility that this functionality contributes to stimulated aP2 secretion. Therefore, the induction of aP2 secretion may involve secondary effects of released free fatty acids on cellular processes as well as lipid binding activity for full capacity.   structures within the cytosol. We observed a time-dependent increase in aP2 localization to the total microsome fraction upon lipolytic stimulation ( Fig. 5B ), suggesting lipolysisinduced recruitment of aP2 to a membrane compartment.
To further defi ne the membrane compartment to which aP2 is localized within the adipocyte, we also analyzed intact 3T3-L1 adipocytes using electron microscopy to reach the necessary resolution. Immunogold labeling against GFP in adipocytes expressing GFP-fused aP2 protein revealed that aP2 was present in compartments resembling MVBs, while labeling was absent in cells that did not express the GFP-aP2 fusion construct ( Fig. 5C ).
MVBs are large cytosolic vesicles derived from late endosomes that invaginate to form smaller vesicles called intraluminal vesicles (ILVs) ( 47 ). Once MVBs fuse with the plasma membrane, ILVs are released into the extracellular space as exosomes. Exosomes, as well as other secretory vesicles, have been reported to have specifi c size and exosome-like vesicles from conditioned media of differentiated 3T3-L1 adipocytes using gradual ultracentrifugation with or without induction of lipolysis. We found that lipolysis induction with FSK or IBMX signifi cantly increased the level of vesicular aP2 ( Fig. 6C ). Taken together, our results demonstrated that lipolysis recruits aP2 to MVBs for nonclassical secretion partially via exosome-like vesicles. DISCUSSION aP2 (FABP4) has been implicated in the pathogenesis of many chronic metabolic diseases including type 2 diabetes, atherosclerosis, fatty liver disease, asthma, and cancer. Until recently, the biological functions of aP2 in the pathogenesis of chronic diseases have been attributed exclusively to its intracellular action. However, newer studies have demonstrated that aP2 is released from adipocytes; its levels are signifi cantly increased in experimental and human obesity and strongly correlate with cardiometabolic diseases in numerous human studies (15)(16)(17)(18)(19)(20)(21)(22)(23)(24)(25)(26). We have also shown that the secreted form of aP2 is a bona fi de adipokine that regulates systemic metabolism, and that antibody-mediated neutralization exhibits potent antidiabetic activity in preclinical models ( 15 ). Other groups have also demonstrated that the secreted form of aP2 plays other multiple potential roles ( 27,28 ). Despite this emerging paradigm of aP2 function as a hormone, mechanisms involved in aP2 secretion and the regulation of this process remained largely unknown.
Here, we describe a novel mechanism and route of aP2 hormone secretion from adipocytes through a nonclassical secretory pathway responsive to lipolytic activity and the marker profi les and fl oat to specifi c fractions in sucrose gradients ( 47,48 ). In order to further characterize the structures that contain aP2 and to test whether exosomes are associated with nonclassical aP2 secretion, we layered isolated EVs onto a continuous sucrose gradient and blotted for aP2 protein together with established exosomal markers. In these experiments, aP2-containing vesicles fl oated at a density of ‫ف‬ 1.20 g/ml and also contained established exosome markers CD63, TSG101 , and ALIX ( Fig. 5D ) ( 49 ). We also confi rmed in a separate sucrose gradient experiment that aP2-containing vesicles are positive for MFG-E8, another EV marker ( Fig. 5E ). These lines of evidence suggest that aP2 physically associates with secretory vesicles that have characteristics consistent with exosomes.
We then performed immunogold labeling against aP2 on EVs from the conditioned media of in-house-generated differentiated adipocytes by electron microscopy, using EVs isolated from FABP-defi cient ( aP2 Ϫ / Ϫ ; mal1 Ϫ / Ϫ ) cells as controls ( Fig. 5F ). In these experiments, aP2 was detected in ‫ف‬ 100 nm diameter vesicles isolated from WT but not FABPdefi cient adipocyte conditioned media. The size and homogeneity of the EVs containing aP2 were also consistent with the exosome-like properties of these vesicular structures.
To test whether lipolysis increases aP2 secretion through this vesicular pathway, we immunogold labeled aP2 in differentiated 3T3-L1 adipocytes treated with either vehicle or isoproterenol and quantifi ed immunogold labeling on MVBs. These experiments demonstrated that the presence of aP2 in these vesicular structures was signifi cantly increased upon lipolytic stimulation ( Fig. 6A ). Similar results were obtained by expressing GFP-fused aP2 in 3T3-L1 adipocytes and immunogold labeling against GFP ( Fig. 6B ). To confi rm that lipolysis increases aP2 secretion through this vesicular pathway biochemically, we also isolated Fig. 6. aP2 is recruited to the nonclassical exosome-like pathway upon lipolysis. A: 3T3-L1 adipocytes were treated with isoproterenol (10 M), fi xed, and immunogold labeled for aP2. Right panel: Gold particles per MVB were quantifi ed from at least 62 randomly selected MVBs in each treatment. aP2 labeling is indicated with arrows. B: 3T3-L1 adipocytes expressing FLAG-tagged-GFP-aP2 were treated with isoproterenol (10 M) or vehicle, immunogold labeled against GFP, and examined by electron microscopy. Lower magnifi cation of control is shown at top left panel. N, nucleus. GFP labeling is indicated with arrows. C: EVs were isolated from 3T3-L1 adipocyte CM treated with vehicle or lipolytic stimuli, and aP2 was measured by Western blot. MFG-E8 was used as vesicular marker. Western blots were quantifi ed (lower panel) using seven independent experiments. * P < 0.05, ** P < 0.01. Bars indicate SEM.
Our biochemical and microscopic analysis revealed that within adipocytes, lipolysis induces relocalization of aP2 to a nonclassical secretory pathway that involves the MVB compartment. Once secreted, the majority of aP2 is detectable in the free form in circulation, but a small fraction persists in vesicles (data not shown). The vesicles bearing aP2 resemble exosomes in terms of size and marker expression, in agreement with previous reports ( 57 ). Our analysis of the density of the aP2-containing particles showed that they are at the high end of the density range reported for exosomes (1.19-1.21 g/ml) ( 58 ). This may refl ect unique properties of adipocytes and the lipid composition that may infl uence the precise characteristics of these vesicles. Our earlier observations demonstrated that plasma aP2 levels increase ‫ف‬ 4-fold in mice with dietary or genetic obesity ( 15 ). However, in the circulation, the majority of aP2 is in the free form, and based on therapeutic effi cacy of an aP2 neutralizing antibody, free aP2 may be the biologically active form, at least in the contexts in which it has been studied so far. While the precise calculation of the ratio of the vesicular and free pools of aP2 is challenging, we predict that the vesicular fraction represents <5% of total aP2. At the moment, whether the free form is derived from breakdown of the vesicles exported through MVBs and/or is contributed through an independent secretory pathway remains unclear. It will be interesting to evaluate these possibilities as well as potential distinct functional properties of free and vesicle-bound aP2 in different target tissues in the future. For example, it was previously reported that adipose-derived exosome-like vesicles from ob/ob mice have proinfl ammatory effects on macrophages ( 59 ), suggesting that they may play a role in the metabolically driven infl ammatory aspects of diabetes.
The regulation of aP2 secretion by nutritional status and lipolysis, and its function in promoting hepatic glucose production and insulin resistance, fi ts well with our current understanding of blood glucose regulation during prolonged fasting, when the release of gluconeogenic substrates (i.e., glycerol) and signals from adipose tissue are synchronized. One condition in which specifi c organisms are challenged with prolonged fasting is hibernation. A recent study has demonstrated that during hibernation, grizzly bears exhibit reversible insulin resistance coupled with elevated lipolysis and markedly increased aP2 expression, and their glucose levels are well maintained at this stage ( 60 ). In the posthibernation period when bears are no longer reliant on fat stores, adipose aP2 levels are dramatically decreased and insulin sensitivity is increased. However, how circulating aP2 levels correlate with this remarkable regulation remains to be studied.
Finally, in obesity, adipose tissue becomes resistant to insulin-mediated suppression of lipolysis, which will drive aP2 release and contribute to increased liver glucose output, leading to hyperglycemia and diabetes. Hence, although aP2 carries the properties of a glucoregulatory hormone of fasting or catabolism, it is paradoxically elevated to signal a "pseudofasting" condition in obesity. Interestingly, the synthetic status of liver in obesity resembles that of fasting in a healthy animal ( 61 ). Lipase inhibitors availability of free fatty acids. We demonstrate that adipocyte lipase activities, specifi cally ATGL and, to a lesser extent, HSL, are involved in the regulated secretion of aP2, and genetic deletion or chemical inhibition of these enzymes dramatically diminishes aP2 secretion from adipocytes. It has been reported in various systems that cAMP signaling positively regulates classical and nonclassical hormone secretion (50)(51)(52)(53). However, induction of hormone secretion has not previously been associated with specifi c lipase activity or free fatty acid availability in adipocytes and has not been linked to aP2 secretion.
We have shown that increasing free fatty acid availability to mimic the state of increased lipolysis is suffi cient to elevate aP2 secretion in vitro and to increase circulating aP2 levels in vivo. It is possible that activation of lipases liberates fatty acids that may serve as ligands for aP2 and promote its secretion by this mechanism. While this possibility is consistent with stimulation of aP2 secretion in response to lipids, an experimental mutation that prevents lipid binding to aP2 was not suffi cient to completely block its secretion from cultured adipocytes. Hence, there are likely to be other cellular events that signal and support aP2 secretion. For example, it is possible that free fatty acids drive aP2 secretion indirectly by activating signaling cascades after they are further metabolized into alternative products or that they play a role as metabolites or signaling molecules for activation or synthesis of the machinery that supports aP2 secretion through vesicular pathways ( 54 ). In addition, lipids and lipidderived species such as phospholipids and ceramides may affect aP2 secretion via their incorporation into membranes. For instance, ceramide can be produced by condensation of palmitic acid with serine into sphingolipids; this particular lipid species is enriched in exosomal membranes and is potentially involved in budding of exosomes into MVBs ( 55 ).
Based on our studies in the genetic deletion models, we suggest that such lipid signals would be preferentially produced by ATGL, and less so by HSL activity, as deletion of ATGL essentially abolished aP2 secretion from adipocytes, whereas the impact of adipocyte-HSL defi ciency was only partial. This differential activity may guide the efforts to identify specifi c lipid molecules involved in regulating aP2 secretion in future studies. Recent studies on stereo/regioselectivity of ATGL and HSL revealed that ATGL prefers sn -2 and sn -1 positions of TAGs and generates sn -1,3 and sn -2,3 DAGs. HSL thereafter prefers sn -3 position on DAGs ( 56 ). The differential impact of ATGL and HSL defi ciency on aP2 secretion might suggest that fatty acids mostly found in the sn -2 or sn -1 but not sn -3 positions of TAGs may be important factors in aP2 secretion. Alternatively, if DAGs generated by ATGL activity as well as lipids in the sn -3 position liberated by HSL are important, the accumulation of DAGs in HSL defi ciency might undermine our ability to interpret the role of HSL activity and lipids in the sn -3 position on aP2 secretion. Another possible explanation is that the mild effect of HSL-defi ciency model refl ects the contribution of HSL activity to TAG hydrolysis, which is 10% of its activity on DAG hydrolysis ( 56 ).
have been suggested as therapeutic tools for inhibiting toxic effects of lipids resulting from increased lipolysis in patients with metabolic disease ( 37,62,63 ). Our data suggest that regulation of aP2 secretion or its neutralization may be appealing alternatives or additive therapeutic approaches that would circumvent limitation of inhibition of lipases and the potential for broader and undesired consequences.