Fat apoptosis through targeted activation of caspase 8: a new mouse model of inducible and reversible lipoatrophy
We describe the generation and characterization of the first inducible ‘fatless’ model system, the FAT-ATTAC mouse (fat apoptosis through targeted activation of caspase 8). This transgenic mouse develops identically to wild-type littermates. Apoptosis of adipocytes can be induced at any developmental stage by administration of a FK1012 analog leading to the dimerization of a membrane-bound, adipocyte-specific caspase 8–FKBP fusion protein. Within 2 weeks of dimerizer administration, FAT-ATTAC mice show near-knockout levels of circulating adipokines and markedly reduced levels of adipose tissue. FAT-ATTAC mice are glucose intolerant, have diminished basal and endotoxin-stimulated systemic inflammation, are less responsive to glucose-stimulated insulin secretion and show increased food intake independent of the effects of leptin. Most importantly, we show that functional adipocytes can be recovered upon cessation of treatment, allowing the study of adipogenesis in vivo, as well as a detailed examination of the importance of the adipocyte in the regulation of multiple physiological functions and pathological states.
Current murine models of lipodystrophy or lipoatrophy have contrib- uted substantially to our understanding of adipocyte physiology1–4, as well as to our understanding of diseases associated with excessive adipose mass, such as atherosclerosis, diabetes and hypertension. One consistent disadvantage to current lipoatrophic mouse models is the lack of inducibility. In addition, loss of adipose tissue in the current mouse models is irreversible such that the constitutive loss of adipocytes leads to lipotoxicity5–8, which makes the distinction of the acute loss of adipokine-induced effects versus the prolonged pathologic changes associated with lipoatrophy difficult. Here, we describe the generation and characterization of the FAT-ATTAC mouse, a transgenic mouse in which we can induce the ablation of both white and brown adipocytes within 1–2 weeks of treatment at any stage of development and without substantial secondary effects in other tissues.
RESULTS
Generation and characterization of an inducible fatless mouse The p20 and p10 catalytic domains of human caspase 8 were fused to serial FKBPv (Phe36Val mutant FKBP) domains that can be tethered using regulated homodimerization with introduction of a FK1012 ana- log, AP20187. AP20187 binds with 1,000-fold greater (subnanomolar) affinity to FKBPv than to endogenous FKBP. To effectively increase local concentrations of the fusion protein, a myristoylation site was included to provide membrane attachment for the caspase 8–FKBP (Fig. 1a). The adipocyte-specific Fabp4 promoter was used to drive expression of fusion protein in stably transfected 3T3-L1 preadipocytes. Treatment with AP20187 results in widespread apoptosis with loss of adipocyte membrane integrity and lipid contents (Fig. 1b). Similar to previous studies with caspase 8 in cardiac myocytes9 and caspase 3 in hepato- cytes10, we generated transgenic mice that carry this transgene. Upon northern analysis, only one line of six that had germline transmission showed marked expression at the mRNA level (data not shown) and this line was used for all subsequent experiments. Transgene expression was limited to all white adipose depots examined, as well as interscapu- lar brown adipose tissue, but was completely absent in all other tis- sues (Fig. 1c), which was further verified by RT-PCR (data not shown). Furthermore, western blot analysis using FLAG-specific antibodies indi- cated that fusion protein expression was only detectable in white and brown adipose tissue (data not shown). Based on the strategy used, we refer to this mouse as the FAT-ATTAC mouse.
To determine whether treatment with AP20187 dimerizer was equally effective in promoting adipocyte apoptosis in vivo, we analyzed inguinal and epididymal adipose depots from treated FAT-ATTAC mice. Short- term treatment with dimerizer (7 d) is sufficient to cause disruption of local adipose tissue structures with associated infiltration of macro- phages and fibrosis (Fig. 2a); the number of cells expressing the mac- rophage-specific F4/80 antigen is markedly increased in adipose from FAT-ATTAC mice in close proximity to adipocytes undergoing apoptosis (Fig. 2a). Longer courses of dimerizer treatment (14–28 d) result in progressive disruption of adipose architecture with concomitant apoptosis of lipid-laden adipocytes. TUNEL-positive nuclei were present in both white and brown adipose tissue, whereas no substantial TUNEL staining could be observed in nonadipose tissues or in adipose tissue from control mice (Fig. 2a). TUNEL-positive cells were also positive for activated caspase 3, the downstream effector of the caspase cascade (data not shown). Unusual granuloma-like structures were abundant in FAT-ATTAC mice treated for 28 d. Though the exact nature of these structures is unclear, cells surrounding the central lumen express neither perilipin nor F4/80, suggesting that they are neither adipocytes nor mac- rophages and may represent reactive fibrosis to widespread adipocyte apoptosis (Supplementary Fig. 1 online). Adipose tissue is morphologi- cally normal in vehicle-treated transgenic mice and dimerizer-treated wild-type littermates and no detectable histological changes were pres- ent in any other tissue, confirming the adipocyte-specific expression of the transgene. Furthermore, there was no evidence of lipid accumulation in the liver (Supplementary Fig. 2 online), suggesting that widespread adipocyte apoptosis does not result in the accumulation of triglycer- ide stores in peripheral tissues without persistently elevated serum tri- glyceride levels. Most importantly, 6 weeks after cessation of dimerizer treatment, adipose tissue from FAT-ATTAC mice appeared histologically identical to untreated mice, indicating that in the absence of continued dimerizer treatment, endogenous preadipocytes can successfully dif- ferentiate, enabling adipose tissue regeneration.
The absence of adipocytes observed histologically is mirrored by marked reductions (>95%) of the adipocyte-specific secretory proteins adiponectin and resistin by day 7 of treatment (Fig. 2b), and serum concentrations remain at near-knockout levels throughout the course of dimerizer treatment. Despite changes in circulating adipokines, fatless FAT-ATTAC mice in the genetic background of FVB mice showed no significant differences in fasting serum insulin, glucose, triglyceride or free fatty acid levels throughout the course of dimerizer treatment (data not shown). Upon cessation of treatment, adiponectin and resistin lev- els rebounded, indicating that adipocyte progenitor cells remain intact and the effects of the widespread adipocyte apoptosis are reversible. Progressive weight gain confirmed the global, robust return of adipose stores (Fig. 2c). Six weeks after removal from dimerizer treatment, FAT- ATTAC mice were heavier and showed increased body fat relative to pretreatment state as assessed by magnetic resonance imaging (MRI; Fig. 2d). Although metabolically identical to wild-type littermates before dimerizer treatment (Fig. 2e), fatless FAT-ATTAC mice were markedly glucose intolerant (Fig. 2e), consistent with decreased circulating adi- ponectin levels. Glucose intolerance persisted 6 weeks after cessation of dimerizer treatment (Fig. 2e), an observation that may be attributed to the fact that adiponectin levels only recover to approximately 50% of baseline levels. Circulating leptin levels are substantially decreased in fatless mice as well (Fig. 2f), leading to marked hyperphagia (50% increased food intake).
Figure 1 Adipocyte-specific expression of the caspase 8–FKBP transgene in vitro and in vivo. (a) Linearized expression construct introduced into blastocysts for transgenic expression of myristoylated caspase 8–FKBP fusion protein (top), and mechanism of activation through forced dimerization of adjacent FKBP molecules by dimerizer (bottom). (b) 3T3-L1 preadipocytes stably transfected with expression construct above were differentiated into adipocytes, then exposed to dimerizer (right) or vehicle (4% ethanol; left) treatment for 36 h. Note complete loss of lipid contents and adipocyte morphology in dimerizer-treated cells. (c) Tissue northern analysis on various tissues from 6-week-old male FAT-ATTAC mice. We hybridized total RNA with probes to human caspase 8 (top) or mouse -actin (bottom) for normalization purposes. Samples were unequally loaded to show absolute absence of transgene in nonadipose tissues. Caspase 8–FKBP transgene expression is limited to both brown and white adipose tissue (WAT).
FAT-ATTAC ob/ob mice
To determine whether the metabolic phenotype and associated changes in feeding behavior of the FAT-ATTAC mouse were attributable solely to the absence of leptin, an adipocyte-derived satiety signal, we generated FAT-ATTAC ob/ob mice and compared them to leptin-deficient ob/ob littermates. We observed similar weight curves for FAT-ATTAC ob/ob mice and ob/ob littermates throughout development (data not shown). After 10 weeks of dimerizer treatment, FAT-ATTAC ob/ob mice weighed nearly 30 g less than treated ob/ob littermates (Fig. 3a). Upon dissection, these FAT-ATTAC ob/ob mice showed near-complete absence of white adipose tissue (Fig. 3b). This finding was confirmed by MRI (data not shown). Histologically, there was an absence of lipid-laden adipocytes in white adipose depots (data not shown), similar to fatless FAT-ATTAC mice in the FVB background. Circulating levels of adipocyte-specific secretory proteins adiponectin and resistin were markedly reduced, whereas MCP-1 levels were elevated nearly fourfold (Supplementary Fig. 3 online).
Fatless FAT-ATTAC ob/ob mice on dimerizer showed progressive increases in daily food intake (Fig. 3c), resulting in a 2.5-fold increase of caloric intake per gram body weight in the fatless ob/ob mice. There was no evidence of malabsorption of nutrients. Decreased body weight in the presence of excess caloric intake is indicative of elevated meta- bolic rate. To indirectly assess metabolic rate, we surgically implanted temperature-sensitive transmitters into the interscapular region of anesthetized mice and monitored core body temperature in conscious, unrestrained animals in either the fasted or fed state. Fatless FAT-ATTAC ob/ob mice showed a significantly higher core body temperature than ob/ob littermates (Fig. 3d), indicative of elevated metabolic rate.
Despite the massive weight loss and increased metabolic rate of these animals, complete loss of adipose function in the FAT-ATTAC ob/ob mice seemed to exacerbate the metabolic disor- ders inherent to this mouse model. Fatless FAT- ATTAC ob/ob mice showed further elevations over already high levels of fasting and fed plasma glucose (60% and 25% increase, respec- tively; Fig. 3e) and fasting and fed triglycerides (100% and 300% increase, respectively; Fig. 3f) relative to ob/ob littermates, whereas there was no significant dif- ference in free fatty acids (Fig. 3g). In addition, histology of ob/ob versus fatless FAT-ATTAC ob/ob livers showed increased steatosis in the FAT-ATTAC ob/ob mouse (Fig. 3h). Furthermore, we showed that thiazolidinedione treatment is ineffective in ameliorating the diabetic phenotype of ob/ob mice in the absence of functional adipose tissue (Supplementary Fig. 4 online). Together with the data generated from the fatless FAT-ATTAC mice in the FVB background, this suggests that the severity of the metabolic complications associated with fat loss is increased with the quantity of lipid that is displaced and that mere deletion of adipose tissue is not the simple solution to type 2 diabetes.
Diminished glucose-stimulated insulin secretion
Decreased serum insulin levels (75% and 66% decrease in fasting and fed insulin levels, respectively; Fig. 4a) persisted in the face of elevated plasma glucose in FAT-ATTAC ob/ob mice throughout the course of dimerizer treatment, despite increased insulin secretion in ob/ob littermates, which parallels progressive metabolic dysfunc- tion in this mouse model (Fig. 4b). But 3 months after cessation of dimerizer treatment, serum insulin levels in FAT-ATTAC ob/ob mice returned to the severe hyperinsulinemic levels observed in control mice. An arginine challenge triggered similar levels of insulin secre- tion, and thus, a normal response to a glucose-independent challenge during the fatless state (Fig. 4c). To test whether the decrease in serum insulin resulted from a defect in glucose-stimulated insulin secretion (GSIS), we compared the effect of a glucose gavage on serum insu- lin levels in euglycemic FAT-ATTAC ob/+ mice. Although the mice showed the expected increase in insulin levels in response to the glu- cose challenge before initiation of fat loss, once FAT-ATTAC ob/+ mice became fatless, they did not increase serum insulin in response to an acute elevation of plasma glucose (Fig. 4d). There was no difference in pancreatic islet area or morphology in fatless FAT-ATTAC ob/+ mice compared to controls (Fig. 4e,f). Similar findings were evident in the FAT-ATTAC ob/ob mice (data not shown). Together, these data suggest that the diminished GSIS in the fatless FAT-ATTAC mouse may result in part from the absence of an adipocyte-specific factor necessary for maximal GSIS. Furthermore, 3 adrenergic-stimulated insulin secretion did not occur in the fatless FAT-ATTAC mouse (Fig. 4g), consistent with the adipocyte-specific expression of mouse 3 adrenergic recep- tors11. Upon cessation of dimerizer treatment, FAT-ATTAC mice showed a partial recovery of 3 adrenergic-stimulated insulin secre- tion, indicating reconstitution of adipose tis- sue function (Fig. 4h). Similar to decreased serum adiponectin levels after adipose tissue reconstitution, the incomplete recovery of 3-stimulated insulin secretion may be the result of either a quantitative or qualitative difference in adipose depots. Future stud- ies are necessary to determine whether an increased dimerizer-free interval would lead to full recovery of adi- pocyte function. Further, although the mechanism mediating the effect of adipocyte 3 stimulation on insulin secretion is unknown, and the possibility of intrinsic beta-cell dysfunction unaccompanied by obvious disruption of islet architecture or response to arginine challenge cannot be formally excluded, the effect lends additional support to a role for adi- pocytes in the regulation of insulin secretion and beta-cell function.
Relative contributions of adipose to the inflammatory state Recent work from our group and others has shown that adipose tis- sue can mediate a considerable proportion of the systemic response to various inflammatory stimuli12–14. To determine the contribution of adipocytes to circulating inflammatory cytokines and acute-phase reactants, we challenged fatless FAT-ATTAC mice and wild-type lit- termates with a sublethal dose (250 ng/g body weight) of lipopolysac- charide (LPS). In fatless mice, there is a significant reduction in the magnitude of the acute-phase response, as shown by a relative reduction in circulating serum amyloid A3 (Fig. 5a) and IL-6 (Fig. 5b) after LPS challenge. Furthermore, despite increased macrophage staining in white adipose depots in fatless FAT-ATTAC mice, we detected decreased circu- lating IL-6 in the basal (nonstimulated) state in the FAT-ATTAC mouse (15 pg/ml versus 55 pg/ml; Fig. 5c). The decreased basal inflammatory tone and reduced response to endotoxin may therefore reflect the loss of the adipocyte as a source of inflammatory cytokines. Alternatively, this may represent the loss of local adipocyte factors that stimulate adipose-resident macrophages, which are abundant in FAT-ATTAC mice (Supplementary Fig. 1 online), to produce inflammatory cytokines in response to acute or chronic inflammatory challenges15.
DISCUSSION
The ability to induce widespread and reversible loss of adipocytes at any time during development represents a major technologi- cal advance. The initial characterization of the FAT-ATTAC mouse presented here shows that this mouse is unequivocally fatless. We present data detailing the necessity of adipose tissue for maxi- mal GSIS, evidence for requirements of adipocyte-derived factors other than leptin in the maintenance of food intake and energy expenditure, as well as the effects of the adipocyte on the systemic inflammatory state. These findings were obtained in a model that does not possess the added secondary complications of long-term insulin resistance present in other lipoatrophic mouse models. The reversibility of our lipoatrophic mouse model provides a unique tool to study de novo adipocyte differentiation in vivo, as well as the in vivo characterization of many inhibitors initially identified in 3T3-L1 adipocyte differentiation in cell culture. The FAT-ATTAC mouse model may also be useful to address the contribution of local adipocyte-derived factors to morbidity and mortality in pathologic states, and will surely highlight the importance of the adipocyte in the regulation of multiple physiological functions.
METHODS
Materials and animals. C57BL/6J ob/+ animals were supplied by Jackson Laboratories. We maintained mice on a 12-h light-dark cycle and standard chow diet. We administered AP21087 (Ariad Pharmaceuticals) or vehicle (4% ethanol, 10% PEG-400, 1.75% Tween-20 in water) every 3 d by intraperitoneal injection at a dose of 0.2 g/g body weight unless otherwise indicated. Animals were between 6 and 8 weeks of age upon the initiation of dimerizer treatment.
All animal experimental protocols were approved by the Institute for Animal Studies of the Albert Einstein College of Medicine.
Cell culture. Subconfluent 3T3-L1 mouse fibroblasts were cotransfected with the FAT-ATTAC transgenic targeting vector and pCB7 to allow selection for hygromycin resistance. We manually selected hygromycin-resistant stable clones, propagated them and differentiated them as previously described16. Cell lines expressing caspase 8–FKBP–FLAG were identified by immunoblotting using M2 FLAG-specific antibody (Sigma-Aldrich); these cells were allowed to fully dif- ferentiate into adipocytes, then treated for 2 d with either vehicle (0.5% ethanol) or AP20187 (100 nM).
Generation of FAT-ATTAC transgenic mice. Activated human caspase 8 (p20- p10 transgene corresponding to Ser217–Asp479) was amplified, 3 FLAG-tagged and subcloned in-frame as a SpeI-BamHI fragment into vector pC4M-Fv2E (Ariad Pharmaceuticals) to generate a cDNA encoding a myristoylated caspase 8–FKBP–FKBP fusion protein. A further PCR amplification step of the entire cDNA including downstream rabbit -globin intron, polyadenylation sequence and 3 untranslated region for transgene stability with flanking NotI sites allowed subcloning of the entire 2.8 kb sequence into a pBluescriptKS+-based vector containing the minimal 5.4 kb of the Fabp4 promoter required for adipocyte- specific expression (provided by B. Spiegelman, Dana Farber Cancer Institute). pBluescriptKS+ sequence was excised with a Sac2/SalI digest and the remaining 8.2 kb linearized plasmid injected into pronuclei of fertilized mouse eggs derived from FVB animals. Transgenic progeny were identified by Southern analysis using an 800 bp fragment of human caspase 8 as a probe; subsequent generations were screened by PCR of tail-derived genomic DNA using human caspase 8–specific primers (forward: 5´-GAAAGTGCCCAAACTTCACAGCATTAGG-3´; reverse: 5´-CTTGTCATCCTTGTTGCTTACTTCATAG-3´), giving rise to a 600 bp prod- uct. Transgene expression was verified by northern blot and RT-PCR analysis of total RNA isolated from white adipose and other tissues using Trizol (Life Technologies). Western blot analysis of 20 g of total protein from tissue lysates in PBS + 2 mM phenylmethylsulfonylfluoride and protease-inhibitor cocktail (Boehringer-Mannheim) was performed as described previously17.
Assays. We measured serum values for glucose using FastBlue B glucose assay (Sigma-Aldrich), triglycerides using Infinity Triglycerides kit (ThermoDMA), free fatty acids using NEFA-C kit (Wako Chemicals), insulin using rat-sensitive insulin RIA kit (LINCO Research), and adiponectin using mouse adiponectin RIA kit (LINCO Research). Other adipokines were measured using LINCOPLEX reagents. All metabolic parameters measured in the fed state were in animals that had food withdrawn at 4 PM, replaced at 7 PM and assayed at 11 PM; for fasted-state measurements, we assayed animals at 10 AM after a 16-h fast. We performed oral glucose-tolerance tests in animals that did not have access to food for 2 h before administration of 2.5 mg/g body weight glucose load by oral gavage, and during the course of the study. We performed arginine-tolerance tests by intraperitoneal administration of a single 1 mg arginine/g body weight dose in animals fasted for 16 h. The effect of 3 agonists on insulin secretion was tested by intraperitoneal injection of CL316423 (1 g/g body weight) into fed animals. We performed LPS challenge by administration of 250 ng/g body weight of E. coli-derived LPS (Sigma-Aldrich) by intraperitoneal injection; we analyzed blood obtained through tail vein for serum IL-6 by Quantikine IL-6 ELISA (R&D Systems) and serum amyloid A3 (ref. 12) by quantitative western blotting using an 125I-derivatized secondary goat rabbit-specific antibody (Amersham) and ana- lyzed it with a Phosphoimager (Molecular Dynamics). For thiazolid inedione efficacy studies, we administered daily doses to FAT-ATTAC ob/ob animals and ob/ob littermates for 10 d by oral gavage of 10 mg/kg/d rosiglitazone in 0.25% carboxymethylcellulose.
Immunohistochemistry. Freshly isolated tissues were fixed with phosphate- buffered formalin and then embedded in paraffin wax. We incubated sections (5 m) with affinity-purified rabbit antibody to perilipin (Affinity BioReagents) or rat antibody to F4/80. After washing in PBS, we incubated slides with biotinylated goat antibody to rabbit IgG or goat antibody to rat IgG at 5 g/ml (Vector Laboratories). Slides were developed using a peroxidase detection kit (Vector Laboratories) and counterstained with hematoxylin (Sigma-Aldrich). We performed TUNEL staining and cleaved caspase 3 immunohistochemical detec- tion on paraffin-embedded sections as previously described18. We performed oil red O staining on 10 m sections of liver frozen in optimal tissue cutting support medium (Tissue-Tek). Hematoxylin was used as a counterstain. Pancreatic islet area was quantified using US National Institutes of Health Image J Software. We sampled at least 80 islets per group.
Magnetic resonance imaging. All images were obtained using a 9.4 Tesla imaging system. We anesthetized mice with isoflurane (1.5% in O2) delivered using a nose cone. To quantitatively assess whole body fat and water, we subjected each mouse to a four-scan pulse-acquire sequence in a 40 mm 1 H coil, and obtained and analyzed spectra, including the water and fat peaks. For imaging, several datasets of eight slices of 2 mm thickness spanning the whole body were acquired using a routine spin echo pulse sequence (18 ms echo time, 400 ms repetition time and two-signal averages).