Deficiency of TNFα Converting Enzyme (TACE/ADAM17) Causes a Lean, Hypermetabolic Phenotype in Mice

Abstract

Energy homeostasis involves central nervous system integration of afferent inputs that coordinately regulate food intake and energy expenditure. Here, we report that adult homozygous TNFα converting enzyme (TACE)-deficient mice exhibit one of the most dramatic examples of hypermetabolism yet reported in a rodent system. Because this effect is not matched by increased food intake, mice lacking TACE exhibit a lean phenotype. In the hypothalamus of these mice, neurons in the arcuate nucleus exhibit intact responses to reduced fat mass and low circulating leptin levels, suggesting that defects in other components of the energy homeostasis system explain the phenotype of Tace^ΔZn/ΔZn mice. Elevated levels of uncoupling protein-1 in brown adipose tissue from Tace^ΔZn/ΔZn mice when compared with weight-matched controls suggest that deficient TACE activity is linked to increased sympathetic outflow. These findings collectively identify a novel and potentially important role for TACE in energy homeostasis.

TNFα CONVERTING ENZYME (TACE/ADAM17) is a prototypic member of the disintegrin-metalloprotease (ADAM) family that is involved in proteolytic ectodomain shedding (1,2,3). TACE was originally identified as the metalloprotease responsible for ectodomain cleavage of the membrane-bound TNFα precursor to generate the soluble cytokine, TNFα (4,5). Since then, TACE has been shown to possess broad sheddase activity that is required for the efficient ectodomain cleavage of a variety of type I and II transmembrane proteins including growth factors, cytokines, cytokine receptors, and cell adhesion molecules in vitro (1,2,3). However, despite the identification of this large number of diverse TACE substrates by in vitro analysis, the physiological importance of these shedding events in vivo, for the most part, has not been determined.

Recently several insights into TACE function in vivo have been obtained through analysis of mice lacking functional TACE (Tace^ΔZn/ΔZn). Tace^ΔZn/ΔZn mice have a targeted deletion of exon 11 that encodes the catalytic active site of the TACE metalloprotease domain, resulting in a lack of enzymatic activity (6). These animals display substantial perinatal lethality with several phenotypic defects characteristic of epidermal growth factor (EGF) receptor (EGFR)-deficient (Egfr^−/−) mice (7,8,9). These include the open-eye phenotype and altered eyelid, hair and whisker development of TGFα-deficient mice (10,11), the aberrant heart valve development characteristic of heparin binding EGF-like growth factor (HB-EGF)-null (12,13), and defects in mammary morphogenesis observed in amphiregulin-deficient mice (14). Based on these observations, TACE is hypothesized to be an essential sheddase for the activation of at least three EGFR ligands during development. More recently the in vivo analysis of radiation chimeric mice reconstituted with Tace^ΔZn/ΔZn hematopoietic cells has demonstrated an important role for TACE in shedding of several leukocyte substrates including TNFα, TNF receptor (TNFR)-1, TNFR2, and l-selectin (15,16).

Modulation of TACE expression and activity can also alter substrate shedding and therefore affect downstream signaling and cellular responses (17,18,19). For example, when heterozygous Tace^+/ΔZn mice, which are viable and fertile (6), are made homozygous for an impaired EGFR allele (wa-2), more animals are born with the open eye phenotype, suggesting that in the haplo-insufficient state, TACE activity is limiting for effective EGFR signaling in vivo (19).

Recently, we characterized a population of homozygous Tace^ΔZn/ΔZn null mice that survive to adulthood and demonstrated that non-cell autonomous TACE expression was required for T cell development and peripheral B cell maturation in vivo (20). The impaired B cell follicle organization and germinal center formation in secondary lymphoid organs observed in Tace^ΔZn/ΔZn mice displayed features that overlap with those found in TNFα-deficient mice, which suggests a physiological role for TACE in activating TNFα signaling (20,21,22,23). Another feature of the adult Tace^ΔZn/ΔZn mice was a dramatic reduction in body weight. However, TNFα signaling is unlikely to play a major role in this phenotype because neither TNFα- nor TNFR-deficient mice show significant changes in body weight when fed a standard chow diet (21,22,23).

The generation of adult Tace^ΔZn/ΔZn mice provides a unique opportunity to further investigate the role of TACE signaling in adult physiology. The aim of the present study was to characterize in detail the reduced body weight phenotype of these mice. We found that adult Tace^ΔZn/ΔZn null mice have a lean, profoundly hypermetabolic phenotype. A central feature of this phenotype is a dramatic increase of metabolic rate that is not due to increases of physical activity, body temperature, or thyroid function. Adult Tace^ΔZn/ΔZn mice also display appropriate responses of arcuate nucleus (ARC) neurons to reduced fat mass and leptin levels but, despite this, fail to mount the expected decrease in energy expenditure or increased food consumption normally observed in this setting. Comparison with wild-type littermates that were calorically restricted to achieve a fat mass similar to that of Tace^ΔZn/ΔZn mice revealed disturbances of several downstream signaling events within both the central nervous system (CNS) and periphery of Tace^ΔZn/ΔZn mice. Collectively, these data suggest a previously unrecognized role for TACE in the control of energy homeostasis.

Materials and Methods

Mice

All experiments were conducted using 8- to 16-wk-old male Tace^ΔZn/ΔZn mice and control age-matched wild-type littermates unless otherwise stated. Mice were maintained in a temperature-controlled room with a 12-h light, 12-h dark cycle and were provided with ad libitum access to standard laboratory chow and water. All procedures were approved by the Animal Care and Use Committees (University of Washington and Pacific Northwest Research Institute) in accordance with National Institutes of Health Guidelines for the Care and Use of Animals.

Body weight and length measurements

Cumulative body weight was measured once per week in group-housed mice from 4 to 12 wk of age. Snout-anus length was measured on 13-wk-old anesthetized animals.

Adipocyte differentiation

Mouse embryonic fibroblasts (MEFs) were derived from 14.5-d-old wild-type, Tace^ΔZn/+ and Tace^ΔZn/ΔZn embryos. Early passage cells (passage 3 or earlier) were used for differentiation of primary cells. Adipocyte differentiation was initiated 2 d after cells reached postconfluence by the addition of differentiation medium [10% fetal bovine serum-DMEM containing 5 μg/ml insulin, 1 μm dexamethasone, 0.5 mm 3-isobutyl-1-methylxanthine, and 10 μm troglitazone] (24). After 2 d, differentiation mixture was removed and culture was continued in 10% fetal bovine serum-DMEM containing insulin and troglitazone. Fresh media was added every 2–3 d. At d 7–10, cells were with fixed 4% paraformaldehyde and stained with Oil Red O. For quantization, Oil red O staining was eluted from cells with 0.5 ml 60% isopropanol and absorbance read at 540 nm. All experiments were performed in triplicate.

Body composition analysis

In vivo body composition analysis of lean mass, fat mass, and water content from conscious, immobilized mice was performed by quantitative magnetic resonance (QMR) (EchoMRI whole-body composition analyzer; Echo Medical Systems, Houston, TX) (25,26).

Indirect calorimetry

Mice were individually housed and acclimated to the calorimeter cages for 1 d before 1–3 d of data collection of gas exchanges and food intake. Indirect calorimetry was performed with a computer-controlled open circuit calorimetry system (Oxymax; Columbus Instruments Co., Columbus, OH) comprised of four respiratory chambers equipped with a stainless steel elevated wire floor, water bottle, and food tray connected to a balance. Oxygen consumption and CO₂ production were measured for each mouse at 6-min intervals, and outdoor air reference values were determined after every 10 measurements. Instrument settings were: gas flow rate = 0.5 liters/min, settle time = 240 sec, measure time = 60 sec. Gas sensors were calibrated daily with primary gas standards containing known concentrations of O₂, CO₂, and N₂ (Tech Air). A mass flow meter was used to measure and control air flow. Oxygen was measured by an electrochemical sensor using a limited diffusion metal air battery. CO₂ was measured with a spectrophotometric sensor. Respiratory quotient (RER) was calculated as the ratio of CO₂ production (liters) over O₂ consumption (liters). Energy expenditure was calculated by the equation: energy expenditure = (3.815 + 1.232 × VCO₂/VO₂) × O₂ consumption (rate of oxygen uptake; VO₂)_. For thermoneutral conditions, calorimetry measurements were performed as described above except that the ambient temperature of the facility housing the calorimetry system was raised to 31 C.

Locomotor activity and feeding behavior

Locomotor activity, feeding, and drinking behavior were monitored continuously during all indirect calorimetry experiments. Locomotor activity was evaluated using an Opto-Varimetrix-3 sensor system (Columbus Instruments). Consecutive adjacent infrared beam breaks were scored as an ambulatory count. Cumulative ambulatory activity counts were recorded every hour for 24 h. The feeding behavior (i.e. frequency/timing/duration and the amount of food/water consumed) was quantified using feed-scale (mass) measurements. In separate experiments, daily food intake was monitored by weighing food hoppers.

Caloric-restriction and fasting studies

Wild-type age- and sex-matched littermates were calorically restricted to a similar relative fat mass as Tace^ΔZn/ΔZn null mice. This was achieved by providing mice with only 70% of their normal food intake for the light and dark cycles at 0900 and 1700 h, respectively. Body composition was monitored daily by QMR. Upon reaching the appropriate reduced fat mass, mice were examined by indirect calorimetry and for locomotor activity, food intake, and water consumption. Caloric restriction of mice was continued during indirect calorimetry experiments. In separate experiments, wild-type age- and sex-matched littermates were fasted for 24–48 h. Ad libitum wild-type littermates were used as controls for both experimental conditions.

Plasma measurements

Leptin and corticosterone plasma levels were determined using specific RIAs that were performed by the Vanderbilt Mouse Metabolic Phenotyping Center. Free T₄ plasma levels were determined by competitive enzyme immunoassay (Leinco Technologies, St. Louis, MO).

Tissue isolation

Mice were anesthetized and blood collected by cardiac puncture. All tissues were rapidly dissected and frozen for subsequent biochemical analysis or mRNA determination as previously described (27). For the hypothalamus, a rectangular region of mediobasal hypothalamus (defined caudally by the mamillary bodies; rostrally by the optic chiasm; laterally by the optic tract; and superiorly by the apex of the hypothalamic third ventricle) was isolated (27).

mRNA analysis

RNA from tissue was isolated and underwent RT-PCR quantification as previously described (27). Total RNA was extracted from tissue using RNAzol B according to the manufacturers’ instructions (Tel-Test, Inc., Friendswood, TX). RNA was calculated by spectrophotometry at 260 nm, and 1 μg RNA was reverse transcribed with 10 U avian myeloblastosis virus reverse transcriptase (Promega Corp., Madison, WI). PCR was performed on a LightCycler (Roche Molecular Biochemicals, Indianapolis, IN) using a 50-ng sample of cDNA template added to the commercially available LightCycler PCR master mix (FastStart DNA Master SYBR Green I; Roche Molecular Biochemicals). Primers were designed to span an exon/intron boundary and optimized for mRNA encoding Npy, Agrp, Pomc, Mch, Crh, Trh, and Gapdh. Primer sequences can be obtained on request. Expression levels of individual hypothalamic mRNAs were normalized to glyceraldehyde-3-phosphate dehydrogenase mRNA content and nontemplate controls were incorporated into each PCR run.

Determination of cellularity

Four random digital images were captured from individual hematoxylin and eosin (H&E) stained sections of white adipose tissue (WAT) and brown adipose tissue (BAT) obtained from four to five different female mice of each genotype. The number nuclei were determined in three fields for each fat pad using the National Institutes of Health ImageJ software (Bethesda, MD). Data are presented as mean ± sem.

Western blotting

BAT or gonadal WAT was dissected and homogenized. Protein content was determined and Western blots using 20 μg of total protein were performed using goat antimouse uncoupling protein (UCP)-1 antibody (1:1000) and goat anti-actin antibody (1:500) (Santa Cruz Biotechnology, Santa Cruz, CA) as previously described (28).

Histology and immunohistochemistry

All tissues were immersion fixed with 4% paraformaldehyde unless otherwise stated. Tissues were processed, paraffin embedded, and tissue sections (8 μm) prepared by routine procedures. H&E staining were performed by standard histological methods.

Statistical analysis

All results are expressed as mean ± sem. A two-sample unpaired t test was used for two-group comparisons. A one-way ANOVA with a Newman-Keuls ad-hoc test was used to compare means between multiple groups. In all instances, P < 0.05 was considered significant.

Results

Generation of Tace^ΔZn/ΔZn mice that survive to adulthood

In the current study, all analyses were performed using Tace^ΔZn/ΔZn mice and littermate controls on a mixed (C57BL/6 × 129) background (20) (supplemental methods, published as supplemental data on The Endocrine Society’s Journals Online web site at http://endo.endojournals.org). Embryos derived from heterozygous Tace^ΔZn/+ matings were examined at embryonic days (E) 12.5, 14.5, and 17.5. Consistent with the findings of Peschon et al. (6), the three expected genotypes were found at the normal Mendelian ratio until E17.5, when a reduction in the appearance of Tace^ΔZn/ΔZn embryos was observed (supplemental Table 1). Similarly, a large proportion of Tace^ΔZn/ΔZn pups died in the first 24 h after birth and during the first weeks of postnatal life. However, in contrast to the original report by Peschon et al. (6), in which a few Tace^ΔZn/ΔZn pups survived to weaning, we found that a significant number (∼25%) of Tace^ΔZn/ΔZn pups survived to adulthood (>8 wk of age) (supplemental Table 2). The explanation for the approximate 2- to 3-fold increase in survivability of Tace^ΔZn/ΔZn pups is not known but is likely associated with subtle differences in strain background that can influence the phenotype of EGFR-deficient mice (29) and by special attention paid to animal husbandry in the current study (Materials and Methods).

We previously demonstrated that the adult Tace^ΔZn/ΔZn mice have defective shedding of several well-established TACE substrates including TNFR1, TNFR2, and l-selectin, confirming that these mice lack TACE proteolytic activity (20). As expected, adult Tace^ΔZn/ΔZn mice display many characteristics previously described in mice with TACE deficiency including open eyelids at birth, stunted and curly vibrissae, perturbed hair coat, and reduced body weight (6,12,14,30,31,32) (supplemental Fig. 1). Specifically, whereas body weight was normal at birth, both female and male Tace^ΔZn/ΔZn mice exhibited significantly reduced body weight by 4 wk of age (Table 1). By 8 wk of age, the magnitude of this weight difference declined in females but not males. Body length was also reduced in female and male Tace^ΔZn/ΔZn mice by 19.7 ± 2.5 and 19.3 ± 2.6%, respectively, at 8 wk of age. By contrast, female and male heterozygous Tace^ΔZn/+ littermates had body lengths and body weights similar to that of wild-type (WT) controls at all ages examined (Table 1).

Table 1.

Phenotypic analysis of Tace^ΔZn/ΔZn, Tace^+/ΔZn, and WT littermates

	WT	Tace+/ΔZn	TaceΔZn/ΔZn	P	n
Body length (cm)
≥8 wk male	9.94 ± 0.11	9.98 ± 0.24	7.98 ± 0.25	<0.001	≥11
≥8 wk female	9.24 ± 0.10	8.99 ± 0.20	7.46 ± 0.24	<0.001	≥9
Body weight (g)
P1	1.50 ± 0.04	1.56 ± 0.03	1.35 ± 0.09a	<0.05	≥8
4 wk male	17.38 ± 0.45	18.67 ± 0.48	9.49 ± 0.91	<0.001	≥10
4 wk female	14.26 ± 0.57	14.79 ± 0.38	9.85 ± 1.00	<0.001	≥6
8 wk male	24.44 ± 0.74	24.98 ± 0.78	18.37 ± 1.17	<0.001	≥8
8 wk female	20.40 ± 0.65	21.53 ± 0.76	18.52 ± 0.28a	<0.05	≥9

^aTace^ΔZn/ΔZn mice were significantly different only from Tace^+/ΔZn mice. 

Tace^ΔZn/ΔZn mice have reduced fat mass despite normal food intake

To determine whether reductions of body fat mass, lean mass, or both contribute to the reduced body weight of Tace^ΔZn/ΔZn mice, QMR measurements of body composition were conducted. At age 12–14 wk, male Tace^ΔZn/ΔZn mice were characterized by decreases of both total lean and fat mass compared with controls (Fig. 1, A, C, and D), with the reduction of fat mass but not lean mass, remaining significant when normalized to total body weight (30.6% reduction, n = 10–14, P < 0.05) (Fig. 1, E and F; supplemental Table 3). As predicted by the reduced fat mass, plasma leptin levels were also decreased in Tace^ΔZn/ΔZn mice compared with WT controls (Fig. 1B). To investigate whether the reduction of plasma leptin levels and other responses are appropriate for the decrease of body fat mass, a subset of Tace^ΔZn/ΔZn mice were compared with WT littermates that were calorically restricted (WT-CR) to achieve a comparable decrease of body fat content. As expected, WT-CR mice exhibited reductions of total body weight, lean mass, and fat mass comparable with that of Tace^ΔZn/ΔZn mice, yet plasma leptin levels remained lower in TACE-deficient mice than in the WT-CR group (Fig. 1B). Whereas normal animals exhibit hyperphagia in response to reduced levels of body fat mass and plasma leptin (33,34), food intake of Tace^ΔZn/ΔZn mice was comparable with that of WT mice fed ad libitum (Fig. 1G).

Figure 1.

Tace^ΔZn/ΔZn mice have reduced fat mass despite normal food intake. QMR measurements were performed on 12- to 14-wk-old male Tace^ΔZn/ΔZn, WT, and WT-CR mice. A, Body weight (***, WT vs. Tace^ΔZn/ΔZn, P < 0.001; **, WT vs. WT-CR; #, Tace^ΔZn/ΔZn vs. WT-CR, P < 0.01). B, Leptin plasma levels (*, WT vs. Tace^ΔZn/ΔZn, P < 0.05). C, Fat mass (***, WT vs. Tace^ΔZn/ΔZn, P < 0.001; *, WT vs. WT-CR, P < 0.05). D, Lean mass (***, WT vs. Tace^ΔZn/ΔZn, P < 0.001; *, WT vs. WT-CR, P < 0.05). E, Fat mass as a percentage of total body weight (*, WT vs. Tace^ΔZn/ΔZn, P < 0.05). F, Lean mass as a percentage of total body weight. G, Food intake was measured as described in experimental procedures. In panels, data are presented as mean ± sem. For A, C, D, E, and F, n = 10–14; B, n = 4–10; G, n = 4–5.

Reduced WAT content in Tace^ΔZn/ΔZn mice is not due to defective adipocyte differentiation

TACE proteolytic activity has been linked to the regulation of several signaling pathways involved in adipocyte differentiation including TNFα and preadipocyte factor 1 (PREF-1) signaling (35,36). To determine whether the capacity for adipocyte differentiation is reduced in Tace^ΔZn/ΔZn mice, MEFs were isolated from E14.5 wild-type and Tace^ΔZn/ΔZn embryos and subsequently examined for their ability to differentiate in vitro. Low-passage, confluent cultures of both WT control and Tace^ΔZn/ΔZn MEFS exhibited similar patterns (Fig. 2, A and B) and levels (Fig. 2C) of triglyceride accumulation, as revealed using Oil Red O staining 8 d after onset of differentiation. In cell culture, therefore, TACE activity does not appear to be essential for adipocyte differentiation.

Figure 2.

MEFs from Tace^ΔZn/ΔZn mice are not defective in adipocyte differentiation. Primary MEFs derived from wild-type (A) and Tace^ΔZn/ΔZn (B) 14.5-d-old embryos were treated with differentiation mixture (DM) as described under experimental procedures. At d 8 after induction, cells were stained for lipid droplets with Oil Red O. C, Quantification of neutral lipid (Oil Red O) accumulation during adipocyte differentiation from wild-type and Tace^ΔZn/ΔZn MEFs treated at d 8 after induction with or without DM. Data are presented as mean ± sem from triplicate cultures. Scale, 40 μm.

To further investigate adipocyte maturation in Tace^ΔZn/ΔZn mice, histological analysis of WAT was performed. In agreement with QMR fat mass analysis, gross inspection of fat pads from Tace^ΔZn/ΔZn mice demonstrated a marked reduction in size of all WAT depots. H&E sections of gonadal WAT taken from age-matched female WT and Tace^ΔZn/ΔZn mice revealed fat cells from TACE-deficient mice that were smaller in size and had a multilocular appearance with increased eosin staining (Fig. 3, A and B). Consistent with these observations, the number of fat cells per unit area was increased in Tace^ΔZn/ΔZn mice compared with WT littermates (Fig. 3C). These phenotypic changes are inconsistent with failure of adipocyte differentiation and are reminiscent of those observed in WAT that has been genetically or pharmacologically altered to increase its metabolic activity (37,38,39,40). To investigate this hypothesis further, we measured WAT expression of the mitochondrial UCP-1, which is normally found only in BAT (41,42). UCP-1 protein was detected at variable levels in two of five WAT samples taken from Tace^ΔZn/ΔZn mice but was not detected in any WAT sampled from control littermates (data not shown). Thus, the appearance of BAT characteristics in WAT from Tace^ΔZn/ΔZn mice is a finding associated with increased whole-body metabolic rate in several other mutant mouse models (28,43,44).

Figure 3.

WAT from Tace^ΔZn/ΔZn mice exhibits a multilocular appearance and increased cellularity. Paraffin-embedded sections of gonadal WAT from wild-type (A) and Tace^ΔZn/ΔZn (B) littermates stained with H&E. WAT from Tace^ΔZn/ΔZn mice showed a multilocular appearance with increased eosin staining. C, Quantitation of the number of adipocytes and nuclei within WAT H&E sections. At least three fields were evaluated per fat pad. Data are presented as mean ± sem (n = 5). Differences between WT and Tace^ΔZn/ΔZn mice were evaluated with Student’s t test. **, P < 0.001. Scale, 10 μm.

Tace^ΔZn/ΔZn mice have perturbed energy homeostasis

The finding of reduced fat mass despite normal daily food intake in Tace^ΔZn/ΔZn mice raises the possibility that energy expenditure is increased in these animals. To test this hypothesis, indirect calorimetry measurements of whole-animal energy expenditure were performed based on VO₂ normalized to lean body mass (Fig. 4, A and B). As expected, WT-CR to achieve fat mass comparable with Tace^ΔZn/ΔZn mice exhibited a 34 ± 4% reduction in mean dark cycle VO₂ compared with ad libitum-fed WT controls, although this difference was not observed during the light cycle (when metabolic rate of both groups is relatively reduced). This reduction in dark-cycle energy expenditure is a homeostatic response that conserves energy stores in the face of weight loss due to energy restriction. By contrast, the energy expenditure of Tace^ΔZn/ΔZn mice was dramatically increased in both dark (by 1.8 ± 0.1-fold, P < 0.001, n = 4) and light cycles (by 2.42 ± 0.1-fold, P < 0.001, n = 4) compared with WT controls fed ad libitum (Fig. 4, A and B). A similar outcome was observed when energy expenditure was expressed either as total VO₂ or as VO₂ normalized to total body weight (supplemental Table 3).

Figure 4.

Tace^ΔZn/ΔZn mice have dramatically increased energy expenditure. A, VO₂ in WT, Tace^ΔZn/ΔZn, and WT-CR mice. Arrows indicate when WT-CR mice were given access to food as described in experimental procedures. B, Mean VO² consumption during dark and light cycles (***, WT vs. Tace^ΔZn/ΔZn; ###, Tace^ΔZn/ΔZn vs. WT-CR; **, WT vs. WT-CR, P < 0.001). C, Mean RER during dark and light cycles (*, WT vs. WT-CR and Tace^ΔZn/ΔZn vs. WT-CR, P < 0.05). D, Mean ambulatory activity during dark and light cycles (**, WT vs. Tace^ΔZn/ΔZn and WT vs. WT-CR, P < 0.01). E, Mean VO₂ consumption at thermoneutrality in the dark cycle (***, WT vs. Tace^ΔZn/ΔZn, P < 0.001). F, Mean VO₂ consumption at thermoneutrality in the light cycle (***, WT vs. Tace^ΔZn/ΔZn, P < 0.001). All values are given as mean ± sem (n = 4–10/group).

This pattern of markedly increased energy expenditure in Tace^ΔZn/ΔZn mice is therefore opposite to how normal animals adapt to reduced body fat stores. In addition, whereas mean dark cycle respiratory quotient was reduced in WT-CR compared with WT controls, consistent with increased use of fat as an energy source (Fig. 4C), the mean dark and light RER values in Tace^ΔZn/ΔZn mice were similar to those of WT mice. Therefore, these animals do not display the adaptive increase of fat, relative to carbohydrate, oxidation characteristic of animals in which body fat stores are reduced by energy restriction. Unlike these metabolic responses, Tace^ΔZn/ΔZn mice exhibited a decrease in nocturnal ambulatory activity (compared with WT controls fed ad libitum) that resembled the response of WT-CR mice (Fig. 2D). Thus, increased energy expenditure in mice lacking TACE occurs despite reduced physical activity.

An alternative explanation for the increased energy expenditure of Tace^ΔZn/ΔZn mice is that their hair defect (supplemental Fig. 1) causes loss of body heat and that a compensatory increase of thermogenesis is required to maintain core body temperature. To test this possibility, indirect calorimetry measurements of WT control and Tace^ΔZn/ΔZn mice were performed at temperatures that approach thermoneutrality (Fig. 4, E and F). As expected, mean dark and light cycle oxygen consumption decreased in a temperature-dependent manner in both WT controls and Tace^ΔZn/ΔZn mice, but mean VO₂ values of Tace^ΔZn/ΔZn mice remained markedly increased during both dark and light cycles compared with WT littermate controls at ambient temperatures in which loss of body heat to the environment is minimal. Therefore, the hypermetabolic phenotype of Tace^ΔZn/ΔZn mice is not a compensatory response elicited by excessive loss of core body heat. Because, in addition, circulating concentrations of thyroid hormone were not elevated in these animals (Table 2), these data collectively exclude increased levels of physical activity, heat loss, or thyroid hormone in the pathogenesis of hypermetabolism in TACE-deficient mice.

Table 2.

Neuropeptide mRNA expression profiles in the hypothalamus and circulating levels of centrally regulated hormones

	Wild type	TaceΔZn/ΔZn	WT-CR	P	n
Mch mRNA	100.0 ± 9.7	67.5 ± 10.5	87.8 ± 6.0	NS	≥5 (5–7)
Trh mRNA	100.0 ± 22.9	154.6 ± 26.6	117.9 ± 23.9	NS	≥5 (5–7)
Free T4 (μ g/ml)	0.79 ± 0.03	0.50 ± 0.11	1.15 ± 0.29	NS	≥3 (3–4)
Crh mRNA	100.0 ± 9.0	46.9 ± 8.8	152.7 ± 32.5	<0.05a	≥5 (5–7)
				<0.001b
Corticosterone (ng/ml)	308 ± 35	285 ± 34	701 ± 35	<0.001c	≥8 (8–10)

^aP values were significantly different between Tace^ΔZn/ΔZn and WT. 

^bP values were significantly different between Tace^ΔZn/ΔZn and WT-CR. 

^cP values were significantly different between Tace^ΔZn/ΔZn and WT-CR and WT and WT-CR. 

UCP-1 mRNA and protein levels in BAT from Tace^ΔZn/ΔZn mice

Prominent among the remaining mechanisms that might explain increased energy expenditure in Tace^ΔZn/ΔZn mice is an increase of sympathetic nervous system (SNS) outflow to thermogenic tissues such as BAT. In this setting, BAT hyperplasia, combined with increased expression of UCP-1, causes energy dissipation as heat and increases whole-body oxygen consumption. To investigate this hypothesis, BAT weight and histology were analyzed (Fig. 5, A and B). Although the weight of BAT from Tace^ΔZn/ΔZn mice as a percentage of total body weight did not differ significantly from controls (WT, 0.93 ±.05 g vs. WT-CR, 0.94 ± 0.04 vs. Tace^ΔZn/ΔZn, 1.06 ± 0.05 g, n = 5–15), cellularity of BAT from Tace^ΔZn/ΔZn mice was increased (Fig. 5, A–C). Furthermore, BAT Ucp-1 mRNA levels were reduced in WT-CR mice, as expected for animals with depleted body fat reserves, whereas Ucp-1 mRNA levels in Tace^ΔZn/ΔZn mice did not differ significantly from WT controls (supplemental Fig. 2A). However, at the protein level, a substantial increase in the expression of UCP-1 was detected in BAT taken from Tace^ΔZn/ΔZn mice as determined by Western blotting (Fig. 5D), and this was confirmed by UCP-1 immunostaining in tissue sections from BAT (supplemental Fig. 2, B–E). This pattern of increased cellularity and UCP-1 expression in BAT from mice lacking TACE is consistent with a role for increased SNS outflow to BAT in the lean, hypermetabolic phenotype of these animals.

Figure 5.

BAT from Tace^ΔZn/ΔZn mice exhibits increased cellularity and UCP-1 protein levels. Paraffin-embedded sections of BAT from WT (A) and Tace^ΔZn/ΔZn (B) littermates stained with H&E. BAT from Tace^ΔZn/ΔZn mice showed increased eosin staining. C, Calculation of the number of adipocyte nuclei within BAT H&E sections. At least three fields were evaluated per fat pad. Data are presented as mean ± sem (n = 4–5). Differences between WT and Tace^ΔZn/ΔZn mice were evaluated with Student’s t test (**, P < 0.05). Scale, 10 μm. D, A representative example of Western blot analysis for UCP-1 expression in protein extracts of BAT obtained from Tace^ΔZn/ΔZn and sex-matched WT control mice. β-Actin was used a protein loading control.

Effect of TACE deficiency on hypothalamic neuropeptide gene expression

Because the hypothalamic ARC plays a key role in energy homeostasis, we sought to determine whether TACE deficiency affects neuropeptide gene expression in this brain region. We found (using real-time PCR) increased hypothalamic Npy and Agrp and reduced Pomc mRNA levels in both Tace^ΔZn/ΔZn and WT-CR mice compared with ad libitum-fed WT controls (Fig. 6). This finding suggests that ARC neurons sense and respond appropriately to deficient energy stores and reduced leptin levels in mice with TACE deficiency and therefore that defects in other components of the energy homeostasis system explain the phenotype of Tace^ΔZn/ΔZn mice.

Figure 6.

Tace^ΔZn/ΔZn mice display appropriate neuropeptide expression within the ARC of the hypothalamus. Neuropeptide gene expression in the ARC from WT, Tace^ΔZn/ΔZn and WT-CR littermates was examined by real-time PCR. A, NPY mRNA levels (**, WT vs. Tace^ΔZn/ΔZn, P < 0.001; *, WT vs. WT-CR; #, Tace^ΔZn/ΔZn vs. WT-CR, P < 0.05). B, Agouti-related peptide (AgRP) mRNA levels (**, WT vs. Tace^ΔZn/ΔZn, P < 0.001; *, WT vs. WT-CR; #, Tace^ΔZn/ΔZn vs. WT-CR, P < 0.05). C, Proopiomelanocortin (POMC) mRNA levels (**, WT vs. Tace^ΔZn/ΔZn and WT vs. WT-CR, P < 0.01). All values are given as mean ± sem (n = 5–7/group).

This finding prompted us to measure expression of hypothalamic neuropeptides involved in energy homeostasis that are expressed outside the ARC as well as circulating hormones regulated by these peptides (Table 2). Although melanocyte concentrating hormone (Mch) mRNA levels tended to be decreased in Tace^ΔZn/ΔZn mice compared with either WT group, this effect failed to reach statistical significance. Similarly whereas hypothalamic thyroid-releasing hormone (Trh) mRNA levels tended to be increased in Tace^ΔZn/ΔZn mice, this difference also did not achieve statistical significance. As expected for mice with depleted energy reserves, both CRH (Crh) mRNA levels and circulating corticosterone hormone levels were increased in the WT-CR compared with WT animals. In contrast, Crh mRNA levels were decreased in Tace^ΔZn/ΔZn mice, despite circulating levels of corticosterone levels that were similar to those of WT mice fed ad libitum and were far lower than those of WT-CR mice (Table 2). Unlike the ARC, therefore, the expression of CRH mRNA in the hypothalamic paraventricular nucleus of mice lacking TACE appears to differ from that of WT-CR mice.

To exclude the possibility that alterations in neuronal development contributed to these differences in neuropeptide gene expression and neuroendocrine function, we performed morphological analysis using Nissl staining of cell bodies and fibers within the hypothalamus of TACE-deficient mice and WT controls. This analysis failed to reveal any gross morphological difference of hypothalamic structure between WT and Tace^ΔZn/ΔZn mice (supplemental Fig. 3, A and B), nor did we detect differences with respect to either numbers of hypothalamic neuropeptide Y (NPY)-positive cell bodies and fibers within the ARC (supplemental Fig. 3, C and D) or paraventricular nucleus (supplemental Fig. 3, E–H), or synaptophysin staining in the ARC (supplemental Fig. 3, I and J). Thus, the dramatic effect of TACE deficiency on energy homeostasis is unlikely to arise from defective neuronal development.

Discussion

Through the process of energy homeostasis, changes of fat mass (induced, for example, by energy restriction) elicit compensatory adjustments of feeding behavior and energy metabolism that favor the return of energy stores to their baseline level. In the current work, we describe fundamental defects of energy homeostasis induced by TACE deficiency that result in a lean, profoundly hypermetabolic phenotype. The reduced fat mass of Tace^ΔZn/ΔZn mice was accompanied by decreases of circulating leptin levels and changes of ARC neuropeptide gene expression (increased Npy and Agrp mRNA levels and decreased Pomc levels) expected for animals with depleted body fat stores, yet these mice exhibited neither the increase of food intake nor the decrease of energy expenditure that serve to restore depleted fat mass in normal animals. Instead, Tace^ΔZn/ΔZn mice displayed markedly increased energy expenditure and consumed food in an amount that, although comparable with WT controls, was lower than expected for energy-depleted animals. Together, these results suggest that TACE deficiency causes a hypermetabolic phenotype in which the capacity to transduce key initial components of the response to depleted fuel stores (involving leptin and ARC neurons) into appropriate behavioral and metabolic outputs is impaired, resulting in excessive leanness (Table 3).

Table 3.

Summary of homeostatic responses to reduced body fat mass in TACE-deficient mice

	Expected response to reduced leptin and fat mass	Observed response of TACE-deficient mice
	↓ Leptin	Yes
	↓ Ambulatory activity	Yes
	↑ Npy mRNA	Yes
ARC responses	↑ Agrp mRNA	Yes
	↓ Pomc mRNA	Yes
	↑ Crh mRNA	No
	↑ Corticosterone	No
	↓ UCP-1 in BAT	No
	↑ Food intake	No
	↓ VO2	No
	↓ RER	No

Previous studies characterizing Tace^ΔZn/ΔZn null mice were performed on several different strain backgrounds (6,12,14,19,30,31,32,45,46,48). In the original report, Tace^ΔZn/ΔZn mice on a mixed (C57BL/6 × 129) background exhibited perinatal lethality due, in part, to defective EGFR receptor signaling that is likely associated with defects in cardiac development observed in both HB-EGF- and EGFR-deficient mice (6,12,19,48). By contrast, mice lacking functional TNFα signaling are viable and fertile indicating that reduced TNFα signaling is unlikely to be the primary factor causing postnatal lethality observed in Tace^ΔZn/ΔZn mice (21,22,23). Whether the defective shedding of other TACE substrates is involved in the postnatal lethality of Tace^ΔZn/ΔZn mice has not been determined.

Several studies have demonstrated that some EGFR-deficient mice could survive to weaning (9,29), as do a subset of Tace^ΔZn/ΔZn mice, although neither the percentage of surviving pups nor whether these animals could survive beyond weaning was reported (6). Compared with previous studies, we obtained an estimated 2- to 3-fold increase in the number of Tace^ΔZn/ΔZn pups surviving to and beyond weaning. Interestingly, the number of surviving Tace^ΔZn/ΔZn null mice observed in our study resembles that of mice with a compound heterozygous Egfr^wa-2/wa-5 mutation. The latter mice express hypomorphic (wa-2) and antimorphic (wa-5) Egfr alleles and, like TACE-deficient mice, have markedly reduced EGFR activity (49). Thus, a comparable reduction, but not the complete absence, of EGFR signaling may explain the limited survival of both mouse models into adulthood (6). Whereas differences in strain background can influence the survival of EGFR-deficient mice (29) (our unpublished observation), both HB-EGF- and EGFR-deficient mice on congenic C57BL/6 backgrounds display a range of survivability, indicating that other unknown factors influence the survival of these mice (29,50) (our unpublished observation). Taken together, these findings suggest that the increase in the percentage of surviving Tace^ΔZn/ΔZn mice in the current study reflects differences in strain background combined with factors such as improved husbandry.

One plausible mechanism to explain hypermetabolism of TACE-deficient mice involves increased heat loss arising from a combination of reduced body size (which increases the surface area to mass ratio) (51,52) and altered hair phenotype that in turn triggers a compensatory increase of thermogenesis to maintain core body temperature. Interestingly, in diacylglycerol acyltransferase 1-deficient mice, which also display a hair/coat defect and increased energy expenditure, neither the enhanced oxygen consumption nor the core body temperature of these mice was altered by changes in ambient temperature, indicating that the hair defect did not cause body heat loss (53). Whereas the ambient temperature needed to achieve thermoneutral conditions for Tace^ΔZn/ΔZn mice has not been defined, we found that indirect calorimetry measurements performed at temperatures approaching thermoneutrality showed a persistently elevated rate of oxygen consumption in TACE-deficient mice. Combined with our finding that Tace^ΔZn/ΔZn null mice maintain core body temperature normally compared with their WT controls (our unpublished observation), these data suggest the hypermetabolic phenotype of Tace^ΔZn/ΔZn mice is not due to either a defect in thermoregulation or a compensatory response elicited by loss of core body heat. Similarly, our findings also exclude hyperthyroidism and increased physical activity as factors contributing to the hypermetabolic phenotype of mice lacking TACE.

Several members of the matrix metalloproteinase (MMP)/ADAM/tissue inhibitor of matrix metalloproteinase (TIMP) axis are expressed during adipocyte differentiation and are implicated in this process (36,54,55,56,57). For example, genetic deficiency or overexpression of ADAM12 can alter adipocyte differentiation (58,59,60), and membrane type 1-matrix metalloproteinase acts as a dominant adipogenic factor during WAT development (61). Our current studies, however, suggest that WAT differentiation proceeds normally in the absence of TACE, because MEFs from Tace^ΔZn/ΔZn mice readily differentiate into adipocytes in cell culture. Relevant to this finding are recent studies demonstrating an important role for TACE in the shedding of PREF-1, in which the cleaved form of PREF-1 is a negative regulator of adipogenesis (62). In TACE-deficient mice, the lack of PREF-1 shedding would be predicted to cause a loss of PREF-1 signaling and therefore would create permissive conditions for adipogenesis, a result that would be in agreement with the demonstrated capacity of Tace^ΔZn/ΔZn MEFs to differentiate into adipocytes. It is noteworthy that, despite this evidence of normal adipocyte differentiation ex vivo, WAT taken from adult animals lacking TACE displays features characteristic of BAT. Specifically, gonadal fat pads from Tace^ΔZn/ΔZn mice were characterized by increased cellularity and adipocytes with a multilocular appearance that have variable UCP-1 expression. Whether alterations in WAT phenotype arise from direct or indirect consequences of TACE deficiency, and the extent to which the hypermetabolic phenotype of Tace^ΔZn/ΔZn mice arises from increased metabolic activity of WAT, are important questions for future study.

Developmental defects observed in TACE-deficient mice have been linked to impaired EGFR and/or TNFα receptor signaling (6) and raise the possibility that the energy homeostasis phenotype of Tace^ΔZn/ΔZn mice is linked to one or both of these signaling pathways. Although decreased TNFα shedding is associated with protection against diet-induced insulin resistance and diabetes (45,46) and TNFα-deficient mice have a mild reduction of body weight and fat mass due, in part, to improved insulin sensitivity (63,64,65), these effects were not associated with changes of food intake or energy expenditure (63,64). Interestingly, Tace^ΔZn/ΔZn mice do display lower blood glucose levels and improved glucose tolerance and are more insulin sensitive (supplemental Table 4 and supplemental Fig. 4), which is consistent with a loss of TNFα signaling contributing to changes in insulin sensitivity. However, we believe that the primary cause of increase insulin sensitivity of Tace^ΔZn/ΔZn mice is their reduced body fat and/or hypermetabolic phenotype. Indeed, at least some phenotypic features of Tace^ΔZn/ΔZn mice (including perinatal lethality) are observed in the complete absence of TNFR1 and TNFR2 receptors (6,66,67,68,69,70), suggesting that reduced TNFα/TNFR signaling alone is unlikely to explain the energy homeostasis phenotype of TACE-deficient mice.

Developmental defects of the heart cause heart enlargement and impaired cardiac function in both TACE- and EGFR-deficient mice that could potentially contribute to the hypermetabolic phenotype in adult Tace^ΔZn/ΔZn mice. However, HB-EGF-deficient mice, which display the same cardiac defects as the Tace^ΔZn/ΔZn mice (12), show only a small reduction in body weight (our unpublished observation). In addition, Tace^ΔZn/ΔZn mice show a preferential reduction in total fat mass whereas a distinguishing feature of weight loss induced by heart failure is a predominant reduction in lean mass (71). Whereas we cannot rule out the possibility that the cardiac defects observed in Tace^ΔZn/ΔZn mice contribute to increased energy expenditure, we believe that they are unlikely to be its principal underlying cause.

In the developing and adult CNS, protein family receptor signaling functions in astrocyte development, cell survival (72), and neuronal precursor migration (73,74). However, the complexity of the ErbB ligand/receptor interactions and the shortened life span of ErbB receptor-deficient mice has thus far precluded detailed analysis of energy homeostasis in adult animals. Mice with neuron-specific ablation of ErbB4 (a known TACE substrate) (75) have no overt energy homeostasis defect (76) and the energy homeostasis of phenotype of neuron-specific ErbB2-deficient mice has not been described (77). Transgenic mice ubiquitously overexpressing TGFα ligand for EGFRs have reduced body weight and fat mass, but energy expenditure was not perturbed in these animals (10). Similarly, intracerebroventricular injection of recombinant TGFα or neuregulin into the third ventricle of hamsters reversibly inhibited food intake (78). Whereas mice lacking individual EGFR ligands do not recapitulate the lean, hypermetabolic phenotype of Tace^ΔZn/ΔZn mice, therefore, the lack of phenotype in these mice probably reflects the overlapping and/or redundant functions of different EGFR ligands (10,11,12,13,79,80). By contrast, EGFR-deficient mice show significantly reduced body weights (7,8,9,29), and the lean phenotype of adult Tace^ΔZn/ΔZn mice closely resembles that of adult compound heterozygous Egfr^wa2/wa5 mice (49) (Threadgill, D., personal communication).

Whereas the above observations raise the possibility that reduced EGFR signaling contributes to the hypermetabolic phenotype of mice lacking TACE, EGFR deficiency also has adverse consequences for brain development that were not observed in Tace^ΔZn/ΔZn mice. In situ hybridization experiments have shown Tace mRNA to be widely expressed throughout the CNS including the hypothalamus, in which it is expressed in astrocytes, tanycytes, and neurons (81,82,83,84). Although ErbB receptors are implicated in CNS development (7,8,9,72,73,74), morphological analysis reported here failed to identify differences in fiber or cell body staining within the adult hypothalamus of Tace^ΔZn/ΔZn mice, nor were gross architectural changes observed in the brain of these animals during postnatal development (postnatal d 1, 6, and 14) (Hevner, R., personal communication). Thus, Tace^ΔZn/ΔZn mice do not overtly display the developmental defects seen in these ErbB-deficient mice, despite the close association between functional ErbB ligand signaling and TACE activity in other developmental settings (6,12,13,14,67). The extent to which defective ErbB signaling within the CNS contributes to the energy homeostasis phenotype of TACE-deficient mice is therefore an important question that awaits further study.

In rodents, UCP-1 expression in BAT is dynamically regulated by changes in metabolic state such as exposure to cold or changes in food intake. These stimuli differentially control SNS activity that is responsible for the activation of adrenergic receptors and downstream expression of UCP-1 in brown adipocytes (43,85). In Tace^ΔZn/ΔZn mice, the increased cellularity and UCP-1 expression in BAT is consistent with the possibility that perturbations in SNS outflow to BAT may contribute to the lean, hypermetabolic phenotype of these animals. In a similar manner, stimuli such as exposure to cold or β-adrenergic agonists can also induce WAT to BAT differentiation (28,43). Recently phosphatidylinositol 3-kinase activation in primary leptin-responsive neurons of the CNS was shown to increase SNS activity of WAT that lead to WAT to BAT remodeling, which caused the increased energy expenditure and leanness in these mice (44). Interestingly, the WAT taken from Tace^ΔZn/ΔZn mice also has a multilocular appearance with increased cellularity and aberrant UCP-1 expression. Although it is difficult to assign tissue-specific defects in the regulation of energy homeostasis in global knockout mice, these findings raise the possibility that perturbations in SNS outflow to both WAT and BAT may contribute to the lean, hypermetabolic phenotype of Tace^ΔZn/ΔZn mice.

The hypermetabolic phenotype of adult TACE-deficient mice suggests a novel, unexpected and potentially important role for TACE in energy homeostasis. Indeed, the increased energy expenditure of Tace^ΔZn/ΔZn mice is among the most dramatic yet reported in a rodent system. Given the variety of known TACE substrates, additional insight into mechanisms underlying the effect of TACE deficiency on energy balance will benefit from studies in which metalloproteinase activity is selectively inhibited in distinct hypothalamic nuclei or specific neuronal subsets. Gene therapy approaches (such as brain region specific adenoviral gene delivery) to increase or block TACE activity, and cell type-specific conditional deletion (47) or rescue of TACE activity, should also help to clarify how TACE activity participates in energy homeostasis and whether inhibition of TACE signaling has potential in the future of obesity treatment.