β3-Adrenergic Receptors on White and Brown Adipocytes Mediate β3-Selective Agonist-induced Effects on Energy Expenditure, Insulin Secretion, and Food Intake

β3-Adrenergic receptors (β3-ARs) are expressed predominantly on white and brown adipocytes, and acute treatment of mice with CL 316,243, a potent and highly selective β3-AR agonist, produces a 2-fold increase in energy expenditure, a 50–100-fold increase in insulin levels, and a 40–50% reduction in food intake. Recently, we generated gene knockout mice lacking functional β3-ARs and demonstrated that each of these responses were mediated exclusively by β3-ARs. However, the tissue site responsible for producing these actions is unknown. In the present study, genetically engineered mice were created in which β3-ARs are expressed exclusively in white and brown adipocytes (WAT+BAT-mice), or in brown adipocytes only (BAT-mice). This was accomplished by injecting tissue-specific β3-AR transgenic constructs into mouse zygotes homozygous for the β3-AR knockout allele. Control, knockout, WAT+BAT, and BAT-mice were then treated acutely with CL, and the effects on various parameters were assessed. As previously observed, all effects of CL were completely absent in gene knockout mice lacking β3-ARs. The effects on O2 consumption, insulin secretion, and food intake were completely rescued with transgenic re-expression of β3-ARs in white and brown adipocytes (WAT+BAT-mice), demonstrating that each of these responses is mediated exclusively by β3-ARs in white and/or brown adipocytes, and that β3-ARs in other tissue sites were not required. Importantly, transgenic re-expression of β3-ARs in brown adipocytes only (BAT-mice) failed to rescue, in any way, CL-mediated effects on insulin levels and food intake and only minimally restored effects on oxygen consumption, indicating that any effect on insulin secretion and food intake, and a full stimulation of oxygen consumption required the presence of β3-ARs in white adipocytes. The mechanisms by which β3-AR agonist stimulation of white adipocytes produces these responses are unknown but may involve novel mediators not previously known to effect these processes.

onist stimulation of white adipocytes produces these responses are unknown but may involve novel mediators not previously known to effect these processes.
Obesity is a prevalent condition frequently associated with diabetes, hypertension, and cardiovascular disease. Because available treatments are minimally effective, substantial efforts have been directed toward the discovery of new, effective, anti-obesity drugs. The ␤3-adrenergic receptor (␤3-AR) 1 represents one of a number of potential anti-obesity drug targets for which selective agonists have been developed (1)(2)(3). The ␤3-AR is encoded by a distinct gene that is expressed predominantly in white and brown adipocytes (4 -7), important sites for energy storage and energy expenditure, respectively. Selective activation of ␤3-ARs leads to marked increases in triglyceride breakdown (lipolysis) and energy expenditure (1)(2)(3), and long term treatment of obese rodents with ␤3-selective agonists reduces fat stores and improves obesity-induced insulin resistance (1)(2)(3). Thus, ␤3-selective agonists are promising anti-obesity compounds.
Acute treatment of rodents with ␤3-selective agonists causes a number of diverse metabolic effects including a 2-fold increase in energy expenditure as measured by effects on oxygen consumption, a 10 -100-fold increase in insulin levels, and a 40 -50% reduction in food intake (8 -11). While the effects on lipolysis and energy expenditure are likely to be mediated by ␤3-ARs, uncertainty has existed regarding the identity of receptors mediating the acute effects on insulin levels and food intake. Recently, we generated gene knockout mice lacking functional ␤3-ARs (11). These animals have a slight increase in body fat; but otherwise, they appear to be normal, probably because of adaptations which compensate for the absence of ␤3-ARs, one example being the observed up-regulation of ␤1-AR gene expression in brown and white adipose tissue (11). Of significance, these mutant mice are completely resistant to the ability of CL 316,243 (1), a ␤3-selective agonist, to increase lipolysis, energy expenditure, insulin levels, and to reduce food intake (11). Thus, each of these effects is mediated exclusively by ␤3-ARs.
The relative role of ␤3-ARs in white versus brown adipocytes, as well as ␤3-ARs in other sites, in mediating each of these responses is also unresolved. For example, is the marked stimulatory effect of ␤3-selective agonists on energy expenditure mediated exclusively by stimulation of brown adipocytes, or must white adipocytes also be stimulated to supply free fatty acids as fuel for brown adipocytes? Furthermore, as ␤3-ARs have been reported to exist in the gastrointestinal tract, brain, heart, and prostate and since other sites of expression cannot be excluded (5,(12)(13)(14)(15), is it possible that receptors in nonadipocyte locations mediate some or all of the effects on energy expenditure, insulin secretion, and food intake? At present, the answers to questions such as these are unknown.
In general, it has been difficult to determine the relative role of various target tissues in mediating complex physiologic responses. Genetic engineering in mice, however, provides a means by which these issues can be addressed. In the present study, we have combined gene knockout and transgenic techniques to create mice in which functional ␤3-ARs are completely absent (knockout mice; see Ref. 11), or are expressed exclusively in selected tissues, namely white and brown adipose tissue (WATϩBAT-mice) or brown adipose tissue only (BAT-mice). To create WATϩBAT-mice and BAT-mice, transgenic constructs were generated in which murine ␤3-AR gene expression is driven by the tissue-specific promoter/enhancers, aP2 for white and brown adipose tissue expression (16) and UCP for brown adipose tissue expression (17,18). These transgenic constructs were then injected into fertilized mouse zygotes homozygous for the ␤3-AR gene knockout allele (11), thus creating mice in which functional ␤3-ARs are restricted to white and brown fat (WATϩBAT-mice), or brown fat only (BAT-mice). Control, knockout, WATϩBAT-, and BAT-mice were then used to investigate the relative role of ␤3-ARs in white versus brown adipose tissue, as well as ␤3-ARs in other sites, in mediating a number of responses to ␤3-selective agonists.
␤3-AR knockout mice were originally created on an inbred FVB background (11). The knockout allele and the WATϩBAT and BATtransgenes have been maintained on an inbred FVB background. Control animals were wild-type inbred FVB mice housed under conditions similar to the genetically modified animals. All animals were housed at 24°C, had free access to food (Purina Chow 5008) and water, and were handled in accordance with the principles and guidelines established by the National Institutes of Health.
RNA Analyses-Total RNA was isolated from white and brown adipose tissue, brain, colon, kidney, liver, and skeletal muscle of male mice, age 12 to 20 weeks old, using a Brinkmann homogenizer and RNA STAT-60 solution (Tel-Test "B", Inc., Friendswood, TX). ␤3-AR mRNA was detected using standard Northern blotting techniques and 20 g of total RNA. A ␤3-AR hybridization probe was generated by random priming from a 306-bp cDNA template corresponding to codons 120 -222. Note that this 306-bp cDNA region corresponds to the NheI to XhoI ␤3-AR fragment deleted during the creation of the ␤3-AR gene knockout allele. The gels were stained with ethidium bromide, and the abundance of 18 S and 28 S ribosomal bands were used to establish equal loading. To quantitate ␤3-AR mRNA expression, radioactive signals from all Northern blots were analyzed using a PhosphorImager (Molecular Dynamics, Image Quant software).
Assessment of ␤3-AR and Actin mRNA levels in Isolated Pancreatic Islets-Five hundred pancreatic islets were isolated from 10 normal mice as described previously (21). Total RNA was extracted using RNA STAT-60 solution and was then subjected to RT-PCR analysis using mouse ␤3-AR and actin primer sets (␤3-AR sense primer, CCTAGCT-GTCACCAACCCTTT; ␤3-AR antisense primer, GACGAAGAGCATCA-CAAGGAG; actin sense primer, GTGGGCCGCTCTAGGCACCA; actin antisense primer, CGGTTGGCCTTAGGGTTCAGGGGGG). RNA samples were treated with DNase prior to performing RT-PCR. The ␤3-AR primer set amplifies a 260-bp fragment, and the actin primer set amplifies a 245-bp fragment. First strand synthesis was performed using the SuperScript preamplification system procedure (Life Technologies, Inc., Gaithersburg, MD) and the antisense primers for ␤3-AR and actin listed above. PCR was then performed using TaKaRaEx (Otsu, Shiga, Japan) Taq polymerase. Manufacturers recommended buffer, nucleotide, and primer concentrations were employed. The samples were heated to 94 C for 1.5 min and then cycled for 30 cycles using the following protocol: 98°C for 20 s, 60°C for 1 min, and 68°C for 1.5 min. The PCR products were then analyzed using agarose gel electrophoresis and ethidium bromide staining.
Effect of CL on Oxygen Consumption-Oxygen consumption was measured in female, 16 -20-week old control, knockout, WATϩBAT-, and BAT-mice before and after treatment with CL at 1 mg/kg body weight (subcutaneously). Preliminary experiments with control mice indicated that this dose exerted maximum effects on oxygen consumption. Oxygen consumption was measured using computerized equipment that included a 1-liter chamber maintained at 28°C, an air flow of 500 ml/min (regulated with a mass flowmeter, Brooks Instrument Division, Emerson Electric), and an oxygen analyzer (Beckman Industrial Oxygen Analyzer model 755). Mice were awake and unrestrained for the study. The resting rate of oxygen consumption (basal) was assessed when the mice were curled up and still, usually 1-2 h after they had been placed in the chamber. The animals were then injected with CL and returned to the chamber. After injection of CL, a constant rate of oxygen uptake was assessed when mice were again resting.
Effect of CL on Adenylate Cyclase Activity in Membranes Isolated from Brown Adipose Tissue-Membranes were isolated from interscapular brown adipose tissue of control, knockout, and BAT-mice, and effects of CL on adenylate cyclase activity were assessed as described previously (11).
Effect of CL on Insulin, Glucose, and Free Fatty Acids (FFAs) Levels-CL (1 mg/kg body weight, intraperitoneal) or saline was injected into female, 8 -12-week old control, knockout, WATϩBAT-, and BATmice; 15 min later, the animals were quickly sacrificed using a small animal decapitator. Whole blood was collected and analyzed for blood glucose levels (One Touch Blood Glucose Meter, Lifescan, Inc., Milpitas, CA). Serum was then isolated and assayed for FFAs (NEFA C Kit, Wako Pure Chemical Industries, Ltd.) and insulin (Rat Insulin Kit, Linco Research Inc., St. Louis, MO). In a preliminary experiment, control mice were injected with CL, and blood samples were obtained before CL and 5, 10, 15, and 60 min after CL injection. Maximal increases in insulin were observed at the 15-min point.
Effect of CL on Insulin Secretion from Isolated Pancreatic Islets-Groups of 20 islets from control, knockout, and WATϩBAT-mice were pre-incubated for 30 min in 3 mM glucose as described previously (21). The media was removed and then replaced with 1 ml of the following test solutions: 3 mM glucose, 11 mM glucose, or 11 mM glucose plus 1 M CL. Of note, 11 mM glucose is a submaximal stimulatory concentration of glucose. Twenty min later, the incubation media was removed and assayed for insulin.
Effect of CL on Food Intake-The effect of a single injection of CL (1 mg/kg body weight, intraperitoneal) on food intake was assessed in male, 8 -12-week old control, knockout, WATϩBAT-, and BAT-mice. The mice were housed individually during the study. For three days prior to the day of CL treatment, the mice received daily injections of saline to acclimate them to handling. On the day of study, each experimental group was divided in half, and one half was treated with saline while the other half received CL. Food was weighed before and 24 h after injection, and the weight of food missing was assumed to represent g of food eaten. The cages were inspected for food spillage and none was noted.

RESULTS
Generation of Transgenic Mice-A large colony of homozygous ␤3-AR knockout mice was established (11) and used to generate homozygous ␤3-AR knockout zygotes. These zygotes were injected with the aP2-␤3-AR or UCP-␤3-AR transgenes ( Fig. 1) to create WATϩBAT-mice and BAT-mice, respectively. For the aP2-␤3-AR construct, nine founders were produced, six of which transmitted the transgene to offspring. Of these six transmitting WATϩBAT-lines, five lines expressed ␤3-AR mRNA in white and brown adipose tissue, and three of these expressing lines were selected for further analyses (WATϩBAT-1, WATϩBAT-2, and WATϩBAT-3). For the UCP-␤3-AR construct, six founders were produced, five of which transmitted the transgene to offspring. Of these five transmitting BAT-lines, four lines expressed ␤3-AR mRNA in brown adipose tissue, and two of these expressing lines were selected for further analyses (BAT-1 and BAT-2).
␤3-AR mRNA Expression in Control, Knockout, WATϩBAT, and BAT-Transgenic Mice-Northern blotting was used to determine the expression level of ␤3-AR mRNA. The probe used for these studies corresponded to the 306-bp ␤3-AR coding sequence segment, which was deleted during the creation of the knockout allele (Fig. 1). Consequently, no signal was detected in RNA samples isolated from homozygous ␤3-AR knockout mice (Fig. 2). As expected, WATϩBAT-lines expressed ␤3-AR mRNA in white and brown adipose tissue while BAT-lines expressed ␤3-AR mRNA in brown adipose tissue only. With respect to white adipose tissue, line WATϩBAT-1 expressed at a level that was 44% above control mice while lines WATϩBAT-2 and WATϩBAT-3 expressed at levels that were 14 and 47% below control mice, respectively. With respect to brown adipose tissue, lines WATϩBAT-2 and WATϩBAT-3 expressed at levels that were approximately 2.5-fold higher than control mice, lines WATϩBAT-1 and BAT-1 expressed at levels that were just slightly above controls, while line BAT-2 expressed at a level that was slightly below control mice.
Northern blot analysis of total RNA isolated from brain, colon, kidney, liver, and skeletal muscle demonstrated that ␤3-AR mRNA expression, with the possible exception noted below, was restricted to adipose tissue (Fig. 3). In the case of BAT-lines, no expression was detected outside of brown adipose tissue (no signal in white adipose tissue, brain, colon, kidney, liver, and skeletal muscle). In the case of WATϩBAT-lines, however, low-level signals were occasionally detected in skeletal muscle but not in brain, colon, kidney, or liver. It is difficult to establish whether this apparent expression in skeletal muscle was due to ectopic expression in myocytes or represents FIG. 1. The murine ␤3-AR gene, the targeted ␤3-AR gene, and the aP2-␤3-AR and UCP-␤3-AR transgenes. Shown is a partial restriction enzyme map of the murine ␤3-AR gene, the targeted ␤3-AR gene, and the two transgenes employed in the present study. Note that the targeted allele lacks 306 bp of ␤3-AR coding sequence, which is used in Northern blotting experiments to monitor ␤3-AR expression from the endogenous gene and the two transgenes. Large boxes refer to exons, the locations of which have been described previously (19,24). The translated segments are shown in black. ATG and TAA represent ␤3-AR start and stop codons, respectively; AATAAA represents the ␤3-AR transcription termination and polyadenylation signal. The long, narrow, lightly shaded boxes represent the aP2 and UCP promoter/ enhancer sequences. B, BamHI; Bg, BglII; E, EcoRI; H, HindIII; N, NheI; P, PstI; S, SalI; X, XhoI.
FIG. 2. Expression levels of ␤3-AR mRNA in brown and white adipose tissue. Total RNA was isolated from white and brown adipose tissue and was analyzed using Northern blotting techniques. The blots were probed using a 306-bp segment of ␤3-AR coding sequence deleted during the construction of the targeted allele. To quantitate ␤3-AR mRNA expression, radioactive signals from all Northern blots were submitted to PhosphorImager analysis (Molecular Dynamics, Image Quant software). Lanes that were not equally loaded as assessed by ethidium bromide fluorescence of 18 S and 28 S ribosomal bands were not analyzed. Expression in all groups was normalized to control mice. Results are expressed as mean Ϯ S.E. The N for each group is as follows. Left panel (white fat): control (CONT), n ϭ 12; knockout (KO), n ϭ 8; WATϩBAT-1, n ϭ 12; WATϩBAT-2, n ϭ 12; WATϩBAT-3, n ϭ 12; BAT-1, n ϭ 6; and BAT-2, n ϭ 6; right panel (brown fat): control, n ϭ 12; knockout, n ϭ 8; WATϩBAT-1, n ϭ 3; WATϩBAT-2, n ϭ 3; WATϩBAT-3, n ϭ 3; BAT-1, n ϭ 9; and BAT-2, n ϭ 9. expression in adipocytes that are resident within or surrounding this tissue. To address this issue, we carefully dissected away visible fat from skeletal muscle obtained from line WATϩBAT-1. Northern blot analysis of these RNA samples failed to detect any specific ␤3-AR mRNA signal, indicating that the weak signal observed in muscle samples from this line were likely due to the presence of adipocytes.
Effect of CL on Oxygen Consumption-Acute treatment with CL increased oxygen consumption in control mice by 89% (Fig.  4). As was previously observed (11), this response was completely absent in ␤3-AR knockout mice (increase of 0%). In WATϩBAT-lines, the effect of CL on oxygen consumption was restored, with lines WATϩBAT-1 (increase of 120%) and WATϩBAT-2 (increase of 103%) having responses that were equal to or greater than control mice and line WATϩBAT-3 (increase of 60%) having a response that was somewhat less than control mice. CL also increased oxygen consumption in BAT-1 (increase of 24%) and BAT-2 (increase of 17%) transgenic mice; however, the O 2 consumption response in these mice was notably less than that observed in control mice and in all three lines of WATϩBAT-transgenic mice.
Effect of CL on Adenylate Cyclase Activity in Membranes Total RNA was isolated from tissues shown above and was analyzed using Northern blotting techniques and the probe described in Fig. 2. F, skeletal muscle dissection ϭ RNA isolated from skeletal muscle following careful removal of surrounding white adipose tissue.
Isolated from Brown Adipose Tissue-Previously, it was assumed that the O 2 consumption response to CL was mediated exclusively by ␤3-ARs in brown adipocytes. Therefore, the markedly reduced O 2 consumption response of BAT-1 and BAT-2 mice was somewhat unexpected, especially since Northern blot expression data (Fig. 2) demonstrated that BAT-1 and BAT-2 mice have normal amounts of ␤3-AR mRNA in brown fat. However, it is formally possible that the mRNA transcribed from the UCP-␤3-AR transgene bears an unknown mutation created during transgene construction or that the transgenederived RNA transcript is translated at a lower efficiency due to differences in 5Ј-untranslated sequence. To rule out these important possibilities, we assayed for functional ␤3-ARs by assessing CL-mediated stimulation of adenylate cyclase activity in membranes isolated from interscapular brown adipose tissue of control, knockout, BAT-1, and BAT-2 mice. As previously reported (11), adenylate cyclase activity was not stimulated in membranes derived from knockout mice (Fig. 5). In contrast, adenylate cyclase activity was stimulated normally in membranes derived from BAT-1 and BAT-2 mice, demonstrating that BAT-mice express normal amounts of functional ␤3-ARs in brown adipose tissue and that the reduced O 2 consumption response in these mice must be due to the absence of ␤3-ARs in white adipose tissue.
Effect of CL on Insulin, Glucose, and FFAs Levels-Previous studies have demonstrated that acute treatment of rodents with ␤3-AR selective agonists increases serum insulin levels by 10 -100-fold (8,11). To establish the time course for this effect, as well as for the effects on FFA and glucose levels, an initial experiment was performed on control mice (Fig. 6). This study demonstrated that FFA levels increased by 4.4-fold, with a near maximal increase being observed as early as 5 min. Insulin increased by 50-fold with a near maximal increase being observed at the 10-min point. A substantially smaller increase in insulin was observed at the 5-min point. Glucose concentrations decreased by 58%, with a near maximal decrease being observed at the 10-min point. At the 60-min point, insulin levels fell slightly but were still markedly elevated (increased by 24-fold) despite the presence of hypoglycemia during the preceding 50 min. The large increase in insulin levels and the persistent elevation despite hypoglycemia indicate that the stimulus for insulin secretion was potent. In all further experiments, FFAs, insulin and glucose were assessed at the 15-min point.
As previously reported (11), the ability of CL to increase FFAs and insulin and to reduce glucose was absent in ␤3-AR knockout mice (Fig. 7). These responses were completely restored in WATϩBAT-1 mice (FFAs increased by 3.6-fold, insulin increased by 91-fold, and glucose decreased by 55%), were only partially restored in WATϩBAT-2 mice (FFAs increased by 2.7-fold, insulin increased by 13-fold, and glucose decreased by 49%), and were minimally restored in WATϩBAT-3 mice (FFAs increased by 1.6-fold, insulin increased by 3.7-fold, and glucose decreased by 13%). Of note, as previously shown in Fig.  2, left panel, ␤3-AR mRNA expression in white adipose tissue was reduced in WATϩBAT-2 and WATϩBAT-3 mice, with the lowest level of expression being observed in line 3. Thus, it seems likely that decreased ␤3-AR gene expression in white adipose tissue accounts for the reduced responsiveness of these two lines. In BAT transgenic mice (BAT-1 and BAT-2), no effect of CL was observed on FFAs, insulin, and glucose, supporting the idea that ␤3-ARs in white adipose tissue are required for these responses.
Assessment of ␤3-AR and Actin mRNA Levels in Isolated Pancreatic Islets-Total RNA was extracted from control mouse pancreatic islets and was then assessed for ␤3-AR and actin mRNA expression using RT-PCR (data not shown). Total RNA from white adipose tissue was used as a positive control. ␤3-AR mRNA was not detected in RNA samples isolated from pancreatic islets but was detected in RNA samples derived from white fat. In contrast, actin mRNA was detected in both islet and white adipose tissue RNA samples. Thus, mouse pancreatic islets appear not to express detectable levels of ␤3-AR mRNA.
Effect of CL on Insulin Secretion from Isolated Pancreatic Islets-To rule out any possibility that CL stimulated insulin secretion via a direct effect on pancreatic ␤-cells, islets were isolated from control, knockout, and WATϩBAT mice and then treated with CL in the presence of submaximal stimulatory concentrations of glucose (11 mM). The secretory response at 11 mM glucose is one third to half of that obtained upon addition of arginine or an increase in glucose to 20 mM (21), demonstrating the sensitivity of this system for further stimulation of insulin secretion. As shown in Fig. 8, CL had no effect on insulin secretion in any group. This is in agreement with a previous study of isolated islets (8). Thus, the stimulatory effect of CL on insulin secretion is not mediated via stimulation of ␤3-ARs within islets.
Effect of CL on Food Intake-Acute treatment of control mice with CL reduced food intake by 35-45% (Fig. 9, right and left panels). As was previously observed (11), this effect was completely absent in ␤3-AR knockout mice. Significantly, the CLmediated inhibition of food intake was observed in all WATϩBAT-transgenic mice (WATϩBAT-1, 89% inhibition; WATϩBAT-2, 94% inhibition; WATϩBAT-3, 63% inhibition). Of note, the degree of inhibition observed in each WATϩBAT transgenic line was greater than that observed in control mice. In contrast to WATϩBAT-transgenic mice, food intake was not inhibited in BAT-transgenic mice (BAT-1 or BAT-2 mice), consistent with the idea that ␤3-ARs in white adipose tissue are required for the effect on food intake. DISCUSSION ␤3-AR mRNA is expressed predominately in white and brown adipose tissue (4 -7), and ␤3-selective agonists are potential anti-obesity drugs (1)(2)(3). Chronic treatment of obese rodents with ␤3-selective agonists decreases fat stores and improves obesity-linked insulin resistance. Acute treatment of rodents causes a number of diverse metabolic effects including an increase in oxygen consumption and insulin levels, and a decrease in food intake. The role of ␤3-ARs, and ␤3-ARs in white versus brown adipocytes in particular, in mediating each of these responses has been unclear. This is especially true given (a) the diverse nature of these responses, (b) the lack of clear explanations as to how stimulated adipocyte ␤3-ARs might control insulin secretion and food intake, and (c) the recent observations suggesting the presence of ␤3-ARs in nonadipocyte sites (5,(12)(13)(14)(15). Recently, we generated gene knockout mice that lack functional ␤3-ARs (11). These animals are completely resistant to CL-mediated effects on oxygen consumption, insulin secretion, and food intake demonstrating that each of these responses are mediated exclusively by ␤3-ARs (11). To determine the role of white and brown adipocyte ␤3-ARs in mediating these responses, we have combined gene knockout and transgenic techniques to create mice in which functional ␤3-ARs are expressed selectively in white and brown adipocytes (WATϩBAT-mice) or in brown adipocytes only (BAT-mice). This was accomplished by injecting tissue-specific ␤3-AR transgenic constructs (aP2-␤3-AR for WATϩBAT-mice and UCP-␤3-AR for BAT-mice) into mouse zygotes homozygous for the ␤3-AR knockout allele (11). We then treated control, knockout, WATϩBAT, and BAT-mice with CL (1), the ␤3selective agonist, and determined effects on oxygen consumption, insulin secretion, and food intake.
The significant findings of this study are that transgenic re-expression of ␤3-ARs in white and brown adipose tissue (WATϩBAT-mice) completely rescued CL-mediated effects on oxygen consumption, insulin levels,and food intake. Thus, each of these responses are mediated exclusively by ␤3-ARs in white and/or brown adipocytes; receptors in non-adipocyte locations Mice were treated with CL, and 15 min later blood was obtained for determination of FFAs, insulin, and glucose. A small subset of animals from each genotype were given saline. In saline treated animals, FFAs ϭ 420 Ϯ 32 M, insulin ϭ 0.34 Ϯ 0.04 ng/ml, and glucose ϭ 147 Ϯ 12 mg/dl. Results are expressed as the mean Ϯ S.E., and the N for each CL treated group is as follows. Control (CONT), n ϭ 10; knockout (KO), n ϭ 10; WATϩBAT-1, n ϭ 6; WATϩBAT-2, n ϭ 14; WATϩBAT-3, n ϭ 10; BAT-1, n ϭ 3; and BAT-2, n ϭ 3.
are not required. Importantly, transgenic re-expression of ␤3-ARs in brown adipocytes only (BAT-mice) failed to rescue in any way CL-mediated effects on insulin levels and food intake and only minimally restored effects on oxygen consumption, indicating that any effect on insulin secretion and food intake and a full response on oxygen consumption requires the presence of ␤3-ARs in white adipocytes. Whether some or all of these responses require the simultaneous existence of ␤3-ARs in white and brown adipocytes or are partially or completely dependent upon ␤3-ARs in white adipocytes alone, is not resolved by the present study. In theory, this question could be addressed in the future by generating mice that express ␤3-ARs exclusively in white adipocytes.
It is noteworthy that ␤3-ARs must be present in both white and brown adipocytes in order for CL to maximally increase O 2 consumption (89% increase in control mice and 60 -120% increase in WATϩBAT-mice). Lack of ␤3-ARs in white adipocytes significantly limits the stimulatory effect of CL (17-24% increase in BAT-mice). The mechanism by which ␤3-ARs in white adipocytes permit maximal stimulation of energy expenditure is unknown but might possibly be related to their role in mobilizing FFAs from triglyceride stores in white adipocytes, which are then used as fuel for energy expenditure by brown adipocytes or possibly by some other tissue site. Alternatively, it is also possible that energy expenditure in white adipocytes becomes significant following stimulation by CL, an effect not previously appreciated. Such an effect could be mediated by uncoupling protein-2 (UCP2), a recently identified homologue of UCP1 that is abundantly expressed in white adipocytes (22). Prior studies have demonstrated that FFAs are potent activators of UCP1 (23), and based upon the marked amino acid similarity between UCP2 and UCP1, the same is expected to be true for UCP2. Thus, a CL-induced increase white adipocyte intracellular FFA should activate the uncoupling activity of UCP2, thereby increasing energy expenditure in white fat cells. Future studies will be needed to define the role of UCP2 in mediating ␤3-agonist-induced energy expenditure in white adipocytes. Nevertheless, the observation that white adipocyte ␤3-ARs are required for maximal stimulation of energy expenditure may have important implications for the treatment of human obesity. In humans, ␤3-ARs are abundant in brown adipocytes (12, 24 -26), however, unlike rodents, ␤3-ARs in human white adipocytes are either rare or absent (26 -28). Given this and the findings of the present study, it is reasonable to anticipate that the acute stimulatory effect of ␤3-agonists on energy expenditure in humans will be less than might have been expected. Whether this will translate into reduced anti-obesity effectiveness under chronic treatment situations is entirely unknown. It will be important to evaluate this possibility by comparing the long-term anti-obesity effectiveness of CL in mice with ␤3-ARs in white and brown fat (control mice and WATϩBAT-mice) with mice where ␤3-ARs are restricted to brown adipocytes only (BAT-mice).
The mechanism by which CL increases insulin secretion is not known. Nevertheless, the stimulus for insulin secretion is extremely potent as it causes a 50 -100-fold increase in insulin levels, which remain elevated (24-fold increase) despite the presence of hypoglycemia. This effect cannot be mediated by direct effects of CL on pancreatic ␤-cells since pancreatic islets appear not to express detectable levels of ␤3-AR mRNA and since insulin secretion is not stimulated following addition of CL to cultured pancreatic islets. Instead, the present study demonstrates that the effects of CL on insulin secretion are mediated by stimulation of ␤3-ARs on white adipocytes. Importantly, this observation vividly demonstrates that a signal or signals emanating from white adipocytes can, directly or indirectly, profoundly alter ␤-cell function. The identity of this signal is unknown but could be a protein, FFAs, or another lipid product released from fat. However, given that the response occurs within 10 min, and therefore probably does not involve transcription of new genes, and since adipocytes are not known to possess storage granules, the stimulus for insulin secretion may not be an adipocyte-derived protein. In contrast, FIG. 8. Effect of CL on insulin secretion from pancreatic islets isolated from control, knockout, and WAT؉BAT-mice. Groups of 20 islets from the indicated mice were preincubated for 30 min in 3 mM glucose-containing buffer. The media was then removed and replaced with 1 ml of buffer containing either 3 mM glucose, 11 mM glucose, or 11 mM glucose ϩ 1 M CL. Twenty min later, the buffer was removed and assayed for insulin. Results are expressed as the mean Ϯ S.E. of three determinations obtained from three separate animals.
FIG. 9. Effect of a single dose of CL (1 mg/kg) on food intake. Mice were housed individually and were injected daily with saline to acclimate the animals to handling. After 3 days of saline treatment, the groups were divided in half, and one half was treated with saline while the other half received CL. Food intake was determined during the 24 h following CL or saline treatment. Results are expressed as the mean Ϯ S.E., and the N for each group is as follows. Control it is possible that CL-induced increases in FFAs (3-4-fold) mediate the effect on insulin release. Previous studies have demonstrated that FFAs are secretagogues for insulin secretion (29,30), and the observed time course in the present study, i.e. increased FFA levels preceding increased insulin concentrations, is consistent with this possibility. However, the magnitude of the observed effect on insulin levels (50 -100-fold increase), and its persistence despite the presence of hypoglycemia, seems to implicate additional factors as well. This is especially true since FFAs do not stimulate insulin secretion at low glucose levels, and starvation-induced increases in FFA concentrations do not lead to increased insulin levels. Thus, it is plausible that some novel factor emanating from white adipocytes, possibly some lipid product other than FFAs, directly or indirectly stimulates insulin secretion. Given the extreme potency of this stimulus, it will be important to identify the mechanism by which this response occurs.
Similarly, the mechanism by which CL treatment acutely decreases food intake is unknown. From the present study, it is clear that this response is mediated by ␤3-ARs in adipose tissue and requires the presence of ␤3-ARs on white adipocytes. The nature of this signal is also unknown. Leptin is a fatderived protein that regulates appetite (31)(32)(33)(34); however, it cannot account for this effect since leptin levels decrease substantially following ␤3-agonist treatment (35), and this would be predicted to have the opposite effect on food intake. Insulin is another factor that has been shown to suppress food intake (36,37), and as discussed above, insulin levels rise substantially following treatment with CL. However, insulin may not be mediating the suppression of food intake since in WATϩBAT transgenic line 3, a minimal effect was observed on insulin levels but a substantial effect was observed on food intake. Heat has long been known to have effects on regulating appetite (38,39), and it has been postulated that heat generation by brown adipose tissue regulates food intake (40). Of note, heat production is predicted to be markedly increased in all lines that had decreased food intake (control mice and WATϩBAT mice, but not BAT mice). If heat is the signal mediating the suppression of food intake, then it is predicted that it requires the presence of ␤3-ARs in white as well as brown adipocytes. For this and other reasons, it would be of interest to create mice that express ␤3-ARs in white adipocytes only. Finally, other possible candidates for mediating the effects of CL on food intake are lipid fuels (FFAs, or ketone bodies) or possibly another fat-derived protein capable of regulating food intake. Future work will focus on identifying the mechanism responsible for this effect.
In summary, acute treatment with ␤3-selective agonists produces a number of diverse metabolic effects, including stimulation of O 2 consumption, insulin secretion, and inhibition of food intake. In the present study, we have used a transgenic and gene knockout approach to genetically identify the receptor and tissues responsible for mediating each of these important responses. From this and an earlier study (11), we can conclude that each of these effects are mediated exclusively by ␤3-ARs on adipocytes. In addition, ␤3-ARs in white adipocytes must be present to obtain any effect on insulin secretion and food intake and for a full effect on energy expenditure. Determining the mechanisms by which stimulation of adipocytes produces each of these responses will be the focus of future studies.