The discovery of drugs for obesity, the metabolic effects of leptin and variable receptor pharmacology: perspectives from β3-adrenoceptor agonists
- First Online:
- Cite this article as:
- Arch, J.R.S. Naunyn-Schmied Arch Pharmacol (2008) 378: 225. doi:10.1007/s00210-008-0271-1
- 393 Views
Although β3-adrenoceptor (β3AR) agonists have not become drugs for the treatment of obesity or diabetes, they offer perspectives on obesity drug discovery, the physiology of energy expenditure and receptor pharmacology. β3AR agonists, some of which also stimulate other βARs in humans, selectively stimulate fat oxidation in rodents and humans. This appears to be why they improve insulin sensitivity and reduce body fat whilst preserving lean body mass. Regulatory authorities ask that novel anti-obesity drugs improve insulin sensitivity and reduce mainly body fat. Drugs that act on different targets to stimulate fat oxidation may also offer these benefits. Stimulation of energy expenditure may be easy to detect only when the sympathetic nervous system is activated. Leptin resembles β3AR agonists in that it increases fat oxidation, energy expenditure and insulin sensitivity. This is partly because it raises sympathetic activity, but it may also promote fat oxidation by directly stimulating muscle leptin receptors. The β1AR and β2AR can, like the β3AR, display atypical pharmacologies. Moreover, the β3AR can display variable pharmacologies of its own, depending on the radioligand used in binding studies or the functional response measured. Studies on the β3AR demonstrate both the difficulties of predicting the in vivo effects of agonist drugs from in vitro data and that there may be opportunities for identifying drugs that act at a single receptor but have different profiles in vivo.
Keywordsβ3-adrenoceptor agonistAtypical β-adrenoceptorLeptinEnergy expenditureFat oxidationObesity drugInsulin sensitivityLigand-directed signalling
A brief selective history
Aspects of the history of the discovery of the β3-adrenoceptor (β3AR) and β3AR agonists have been described in previous reviews by the author of this article and his colleagues and collaborators (Arch 2000; Arch and Kaumann 1993). This brief history from the author’s perspective acknowledges the importance of the work of Professor Hans Zaagsma and sets the context for the rest of the article.
In Beecham Pharmaceuticals in the late 1970s, Mike Cawthorne had concluded that stimulation of energy expenditure might offer a better route to the discovery of novel anti-obesity drugs than inhibition of energy intake. His vision was supported by the work of Derek Miller. Miller’s group had demonstrated that ephedrine, a non-selective sympathomimetic agent, is thermogenic and slims genetically obese (Lepob/Lepob) mice (Massoudi et al. 1983; Massoudi and Miller 1977), a finding confirmed by workers at Eli Lilly (Yen et al. 1981) and Beecham (Arch et al. 1982). Other types of sympathomimetic drug also slimmed obese mice by raising energy expenditure (Arch 1981; Arch et al. 1987; Dulloo and Miller 1984; Dulloo and Miller 1987).
The problem was now to mimic the sympathetic nervous system without causing the hypertension, tachycardia, tremor or hypokalaemia that is mediated by β1AR and β2AR. In the absence of any better logic, the Beecham chemists led by AT Ainsworth and DG Smith decided to attach substituents that contained carboxylic acid moieties to the nitrogen of ethanolamine βAR agonists in the hope that, like fatty acids, they would form esters with glycerol and localise in adipose tissue. The first compounds synthesised were methyl esters rather than the target acids, but it was soon apparent that they had the desired selectivity in vivo. The two most studied compounds were BRL-26830A [(RR + SS)-(±)-methyl 4-[2-[(2-hydroxy-2-phenylethyl)amino]propyl]benzoate, (E)-2-butendioate (2:1) salt] and (from a different team of chemists) BRL-35135A [(RR + SS)-(±)-methyl 4-[2-[2-hydroxy-2-(3-chlorophenyl)ethyl-amino]-propyl]phenoxyacetate, hydrobromide salt]. Only the parent acids of these compounds (BRL-28410 [(RR + SS)-(±)-4-[2-(2-hydroxy-2-phenylethylamino)-propyl]benzoic acid)] and BRL-37344 [(RR + SS)-(±)-4-[2-[(2-(3-chlorophenyl)-2-hydroxyethyl)amino]propyl]phenoxyacetic acid]) were found in plasma after oral administration of the esters (Ida et al. 1996).
Studies on the acids using isolated rodent tissues showed that, rather than depositing as triglycerides in fat, they were selective stimulants of brown and white adipose tissue lipolysis compared to atrial rate or force, or uterine relaxation—a novel profile for βAR agonists. Moreover, their lipolytic effects were resistant to blockade by standard antagonists of β1AR and β2AR (Arch et al. 1984a; Wilson et al. 1984a). Zaagsma, Harms, de Vente and colleagues had previously shown that standard β1AR and β2AR antagonists displayed low potency as antagonists of rat white adipocytes lipolysis (de Vente et al. 1980; Harms et al. 1974). Moreover, the (+)-enantiomers of these antagonists were typically only tenfold less potent than the (−)-enantiomers as antagonists of lipolysis, compared to a 100-fold lower potency for antagonism of classical β1AR and β2AR-mediated responses (Harms 1976; Harms et al. 1977). Thus Zaagsma and his colleagues had already identified the β3AR. Others, in particular Furchgott (1972), had noted that the βAR in the gut is unusual, notably in its resistance to standard βAR antagonists (Arch and Kaumann 1993). However, it took the discovery of selective agonists, which relaxed the gut as well as stimulating lipolysis (Manara et al. 1995), and ultimately the cloning of the β3AR (Emorine et al. 1989) to convince the doubters. In view of the later discovery of atypical β1AR and β2AR, the doubters were right to doubt.
Since these early days, many pharmaceutical companies have attempted to develop β3AR agonists for the treatment of human obesity, but there is no report that a compound has progressed beyond phase II clinical trials. The first-generation compounds had poor selectivity as agonists of the human cloned β3AR, and second-generation compounds that were selective for the human β3AR mostly had poor oral bioavailability or pharmacokinetics, as illustrated by two of a series of papers from Merck (Biftu et al. 2000; Feng et al. 2000). There may be a more fundamental problem that the β3AR is less important as a regulator of energy balance in humans than in rodents; it is clearly less important for the regulation of lipolysis in human than in rodent white adipocytes (Barbe et al. 1996; Deng et al. 1997; Sennitt et al. 1998). There is, nevertheless, evidence that it plays a significant role in the regulation of energy balance in humans (Arch 2002). Moreover, recent evidence that there is a significant amount of brown adipose tissue, which expresses the β3AR, in many adult humans (Nedergaard et al. 2007) offers new hope. Treatment with β3AR agonists may increase the amount and activity of this brown adipose tissue (Holloway 1989; Kawashita et al. 2002). Meanwhile, there remains a long-standing interest in β3AR agonists for the treatment of depression (Stemmelin et al. 2008) and a more recent interest in the potential of β3AR agonists for the treatment of bladder disorders (Furuta et al. 2006; Michel and Vrydag 2006; Takasu et al. 2007). Relaxation of the uterus in the management of pre-term labour has also been suggested as a possible indication (Bardou et al. 2007).
Even though many companies have lost interest in the potential of β3AR agonists for the treatment of obesity and diabetes, they can offer useful perspectives on pharmacology, physiology and obesity drug discovery. Such perspectives are the focus of the rest of this article. Studies on the β3AR are hindered by the lack or limited availability of highly selective agonists (especially for the human receptor), antagonists and radioligands (Niclauss et al. 2006; Vrydag and Michel 2007). This does not invalidate the perspectives provided by the compounds, however. For example, in the section that follows, it is the linkage of fatty acid oxidation to other properties of the compounds that is the core of the argument, irrespective of whether they act exclusively via β3AR.
Insights for the discovery of anti-obesity drugs
Regulatory criteria and current therapies
Three drugs are marketed for the long-term treatment of obesity: orlistat, which inhibits pancreatic lipase and consequently fat digestion; sibutramine, which acts centrally to inhibit noradrenaline and serotonin reuptake; and rimonabant, which antagonises the cannabinoid receptor-1 in both the brain and peripheral tissues. Rimonabant is not approved in the USA owing to concerns about its depressant effects. The US Food and Drug Administration (FDA) draft guidance asks that anti-obesity drugs should produce weight loss of 5% compared to placebo treatment over 1year in phase III trials (http://www.fda.gov/cder/guidance/7544dft.pdf). The European Medicines Evaluation Agency (EMEA) draft guideline asks for a significant effect of the drug compared to placebo treatment plus weight loss of 10% from baseline, which includes the effects of diet and exercise recommended for all subjects (http://www.emea.europa.eu/pdfs/human/ewp/028196en.pdf). Since mean weight loss in the placebo group is often about 5% and the mean weight of the subjects is usually about 100 kg, this means that a drug should cause 5 kg more weight loss than the placebo treatment causes. Both authorities also give ‘softer’ measures of efficacy based on the proportions of drug- and placebo-treated patients who achieve these weight-loss criteria.
Orlistat, sibutramine and rimonabant fall short of the 5 kg target, though not by much in the case of sibutramine and rimonabant. Meta-analyses conclude that weight loss relative to placebo is 4.2 to 4.4 kg for sibutramine, 4.2 to 4.9 kg for rimonabant and 2.8 to 2.9 kg for orlistat (Christensen et al. 2007; Curioni and Andre 2006; Li et al. 2005; Rucker et al. 2007). As with earlier drugs, nearly all this weight loss is achieved in the first 6 months.
The EMEA guideline asks that the lost weight includes an “appropriate” loss of fat. The FDA guidance is that weight loss is caused “primarily by a reduction of fat”. It would be unreasonable to ask that all the weight lost in response to pharmacotherapy should be fat rather than lean tissue, especially in the case of drugs that reduce energy intake or absorption, because the weight lost in response to a low-calorie diet always includes lean tissue. The proportion is variable, but a systematic review reports median values of 14.0%, 23.4% and 22.5% for diets described as low in energy, very low, and very low plus exercise, respectively (Chaston et al. 2007). Orlistat and sibutramine are no better than diets in protecting lean body mass (Chaston et al. 2007), whilst data for rimonabant have not been published.
The EMEA and FDA also ask for reductions in visceral fat or waist-to-hip ratio. The rationale is that insulin resistance and other features of the metabolic syndrome are associated with excess visceral rather than subcutaneous fat. Subcutaneous fat may even be protective (Buemann et al. 2005; Livingston 2006). Orlistat, sibutramine and rimonabant all reduce visceral fat (Deprés et al. 2005; James et al. 2000; Kelley et al. 2004), but there is no evidence that they have less effect on subcutaneous fat, other than one unconfirmed report that sibutramine reduced visceral to subcutaneous fat ratio (P < 0.04; Faria et al. 2005; Van Gaal et al. 1998). A recent report suggests that diet also fails to distinguish between fat depots (Redman et al. 2007). By contrast, in the Diabetes Prevention Program, intensive lifestyle training, which included a significant exercise component, had more effect on central than on subcutaneous adiposity (Fujimoto et al. 2007). An earlier report also indicated that exercise reduces visceral fat selectively (Thomas et al. 2000).
What really matters is that anti-obesity drugs have metabolic benefits that are at least as great as those expected from the weight loss that they cause. Demonstrating that these benefits are associated with loss of visceral fat merely provides comfort that they probably have a physiological explanation, although there is no consensus as to why visceral fat is associated with impaired glucose homeostasis or dyslipidaemia (Jensen 2006; Ritchie and Connell 2007). Even the moderate weight loss elicited by orlistat is associated with significant metabolic benefits. In some studies, these were beyond those expected from the weight loss and perhaps were a consequence of reduced plasma non-esterified fatty acid concentrations (Kelley et al. 2004). Why this should be so is unclear, since orlistat inhibits fat digestion: one might expect a low-fat diet to have a similar effect. There is a report that sibutramine, which raises sympathetic activity (Hirsch et al. 2000), improves lipid profile beyond expectations from weight loss (James et al. 2000), but this appears to be all.
Rimonabant may be more interesting. Sanofi-aventis have claimed that in the SERENADE trial, which was conducted in drug-naïve patients, 55% of the reduction in haemoglobin A1c, an indicator of long-term blood glucose control, was not explained by weight loss. The evidence to support this claim has not been published at the time of writing. In the RIO-Lipids study, a similar percentage (57%) of the elevation of plasma adiponectin concentration by rimonabant was independent of weight loss (Deprés et al. 2005). Since adiponectin, which is released from adipose tissue, improves insulin sensitivity in skeletal muscle and liver (Kadowaki and Yamauchi 2005), this weight loss-independent effect on plasma rimonabant concentration could partly explain the weight loss-independent effect of rimonabant on blood glucose control. Studies on β3AR agonists that are described below illuminate an additional explanation. This is that rimonabant, partly through adiponectin but also directly, stimulates fat oxidation in adipose tissue, muscle and possibly liver (Lafontan et al. 2007). Rimonabant activates adenosine monophosphate (AMP)-activated protein kinase (AMPK) in human myotubes, suggesting a direct (adiponectin-independent) molecular link for these effects (Cavuoto et al. 2007).
β3AR agonists demonstrate benefits of stimulating fat oxidation
β1-selective and non-selective β1/2AR agonists stimulate fat oxidation in humans (Schiffelers et al. 2000; Schiffelers et al. 2001a; Schiffelers et al. 1999; Wheeldon et al. 1993). In some but not all of these studies, the respiratory exchange ratio decreased, indicating a selective stimulation of fat oxidation, and thermogenesis seemed to depend primarily on fat oxidation (Schiffelers et al. 1998; Schiffelers et al. 2001b). β2AR agonists also stimulate fat oxidation (Scheidegger et al. 1984), though in the presence of acipimox to inhibit lipolysis, the β2AR agonist salbutamol selectively stimulated carbohydrate oxidation (Hoeks et al. 2003).
Even though the β3AR is less important than the β2AR or β1AR in the regulation of white adipocyte lipolysis in humans (Barbe et al. 1996; Deng et al. 1997; Sennitt et al. 1998), the β3AR agonist CL-316243 (5-[(2R)-2-[[(2R)-2-(3-chlorophenyl)-2-hydroxyethyl]amino]propyl]-1,3-benzodioxole-2,2-dicarboxylic acid), which is an antagonist of the human β1AR and β2AR, increased fat oxidation and decreased carbohydrate oxidation in humans (Weyer et al. 1998). Presumably, the stimulation of fat oxidation did not depend on a strong direct stimulation of fat mobilisation. L-796568 ((R)-N-[4-[2-[[2-hydroxy-2-(3-pyridinyl)ethyl]amino]ethyl]phenyl]-4-[4-[4-(trifluoromethyl)phenyl]thiazol-2-yl]benzenesulfonamide), which is a more efficacious agonist of the human β3AR and a weak partial agonist of the human β1AR and β2AR, did stimulate lipolysis. It tended (P = 0.08) to lower the respiratory quotient in obese men (van Baak et al. 2002). The rodent β3AR agonist BRL-26830A, which elicits tremor via β2AR in humans and so cannot be totally β3AR-selective, stimulated fat but not carbohydrate oxidation in humans even when accompanied by a glucose load (Arch et al. 1989). BRL-26830A also stimulated fat oxidation only in rats (Wilson et al. 1986), and it and other β3AR agonists have little or no effect on energy expenditure in rats and mice if fat mobilisation is prevented (Gavrilova et al. 2000; Grujic et al. 1997; Wilson et al. 1986). So what might be the benefits of selectively stimulating fat oxidation?
Effects of β3-adrenoceptor agonists on physiologya
Effects of β3-adrenoceptor agonists
Demonstrated in rodents and humansb
Increased energy expenditure
Increased fat oxidation
Selective loss of fat; preservation of lean body mass
Insulin sensitivity improved beyond expectations from weight loss
Prolonged weight loss due to maintenance of lean body mass and avoidance of counter-regulatory mechanisms associated with reduced food intake
Selective loss of visceral adipose tissue: hypothesis based on report that Trp64Arg variant of β3AR selectively influences visceral fat loss (Tchernof et al. 2000)
There may be many benefits associated with maintaining lean body mass when losing body weight, not least for physical strength and cardiac function. Another benefit may be that weight loss will be maintained for longer than the 6 months that is typical for anorectic drugs. Basal metabolic rate is largely determined by lean body mass (Bitz et al. 2004; Garrow and Webster 1985; Nelson et al. 1992; Ravussin et al. 1986). Loss of lean tissue is probably one, if not the main reason why energy balance returns to equilibrium after 6 months’ pharmacotherapy with anorectic drugs (Flatt 2007). Maintenance of lean body mass may be one mechanism whereby physical activity helps in long-term weight maintenance (Stiegler and Cunliffe 2006; Wing and Phelan 2005). Although not specifically a benefit of stimulants of fat oxidation, thermogenic drugs may also cause prolonged weight loss because they do not provoke counter-regulatory mechanisms associated with reduced food intake—particularly those due to altered gut hormone release or hedonic deprivation. Prolonged weight loss may compensate for the possibility that the rate of weight loss in response to drugs that increase fat oxidation may be slow compared to that elicited by drugs that reduce energy intake or absorption. Indeed, it may be useful to add a thermogenic drug to an anorectic drug in order to maintain weight loss, just as physical exercise seems useful in maintaining weight loss (Wing and Phelan 2005).
There is no evidence that β3AR agonists selectively reduce visceral rather than subcutaneous adipose tissue, but this possibility is raised by reports that the Trp64Arg variant of the β3AR was associated with visceral obesity (Kim-Motoyama et al. 1997) and with a reduction in the loss of visceral adipose tissue of 43% in women undergoing a weight reduction programme. Loss of subcutaneous adipose tissue was reduced by 23%, but this was not statistically significant (Tchernof et al. 2000). It should be noted, however, that there is disagreement as to whether the Trp64Arg variant of the β3AR is hypofunctional in vitro or associated with impaired regulation of energy balance or glucose homeostasis (Kurokawa et al. 2001; Leineweber et al. 2004; Zhan and Ho 2005).
Irrespective of whether they reduce visceral fat, β3AR agonists, probably because they stimulate fat oxidation, elicit metabolic benefits, in particular insulin sensitisation, that are more than would be predicted from weight loss. Such weight-independent effects have been described in rodents, monkeys and humans (Arch 2002). Focussing on the human data, the weight-independent effects of BRL-35135 on glucose tolerance and insulin sensitivity (Mitchell et al. 1989; Smith et al. 1990) are relevant to the argument even though they may be due to stimulation of the β2AR as well as the β3AR because, like other βAR agonists, BRL-35135 probably stimulates fatty acid oxidation in humans. The weight-independent enhancement of insulin action by the more selective β3AR agonist CL-316243 was probably due to stimulation of β3AR only and was demonstrated to be associated with increased fat oxidation (Weyer et al. 1998). It used to be thought that stimulation of fat oxidation would exacerbate insulin resistance because fatty acid oxidation is increased in diabetes and inhibits glucose oxidation through mechanisms described by Randle and his co-workers (Bebernitz and Schuster 2002). Fat oxidation driven by unrestrained lipolysis is, however, a different scenario from fat oxidation drive by a more terminal catabolic mechanism. Moreover, there is increasing evidence that mitochondrial function and the capacity for fat oxidation are defective in both diabetes and obesity (Astrup et al. 1994; Blaak et al. 2001; Larson et al. 1995; Schiffelers et al. 2001a; Zurlo et al. 1990).
The following broad hypothesis may explain why insulin sensitisation is a consequence of increased fat oxidation. Stimulation of fat oxidation may reduce the concentration of various lipid metabolites (diacylglycerol and possibly ceramide or fatty acyl CoA) that affect insulin signalling at the level of insulin receptor substrates (Morino et al. 2006; Yu et al. 2002) or protein kinase B/Akt (Schmitz-Peiffer et al. 1999). Activation of isoforms of protein kinase C may be involved. Unfortunately, there have been few attempts to investigate the effects of β3AR agonists on the concentrations of fatty acid metabolites, though one study demonstrated a reduction in diacylglycerol concentration in skeletal muscle (Darimont et al. 2004). Clearly, changes in the concentrations of fatty acid metabolites may occur rapidly when fat oxidation is stimulated because the pool sizes are small, whereas it will take longer to drain the large store of triglyceride from white adipose tissue for oxidation elsewhere. Hence insulin sensitisation may occur long before weight loss is detectable. Negative energy balance due to an anorectic drug must also promote fat oxidation once glycogen reserves have been depleted, which may be one reason why it sensitises the body to insulin more than if the individual was ‘naturally’ leaner (Wing and Phelan 2005). However, the anorectic drug and counter-regulatory forces will affect fat metabolism in a way that is no different from when weight is lost by dieting. By contrast, stimulation of fat oxidation is opposed by a reduction in fatty acid supply as subjects get thinner, resulting in a dual influence on fatty acid metabolite levels.
Other targets associated with stimulation of fat oxidation
Other than β3AR agonists, what targets might provide drugs that selectively stimulate fat oxidation? One possible target is 11β-hydroxysteroid dehydrogenase type 1, the enzyme that converts cortisone (inactive) to cortisol (active) in humans and 11-dehydrocorticosterone (inactive) to corticosterone (active) in rodents (Wang 2006). Inhibitors of this enzyme improve insulin sensitivity and reduce body weight in obese mice partly by reducing food intake (Alberts et al. 2003; Hermanowski-Vosatka et al. 2005). One of them (BVT2733) prevented the fall in energy expenditure and lean body mass seen in pair-fed mice (Wang et al. 2006), similar to two features (thermogenesis and lean body mass preservation) of treatment with β3AR agonists. Deletion of the gene encoding 11β-hydroxysteroid dehydrogenase type 1 also affects energy expenditure, since it results in resistance to diet-induced obesity despite increased energy intake. Studies on this genetically modified mouse raise the possibility that inhibitors of 11β-hydroxysteroid dehydrogenase type 1 might selectively reduce visceral fat (Morton et al. 2004).
Acetyl CoA carboxylase (ACC) synthesises malonyl CoA. As well as being a precursor of fatty acids, malonyl CoA inhibits fat oxidation. ACC-2 rather than ACC-1 tends to predominate in tissues where the rate of fatty acid oxidation is high. ACC-2 knockout mice had elevated rates of fat oxidation and were protected from diet-induced obesity, hyperglycaemia and hyperinsulinaemia (Abu-Elheiga et al. 2001; Abu-Elheiga et al. 2003). Only body fat content was significantly reduced in the mutant mice, but there was a trend to a reduction in lean body mass. Whole body, muscle and adipocyte fat oxidation were increased (Abu-Elheiga et al. 2003; Oh et al. 2005). Some workers have found that carbohydrate oxidation as well as fat oxidation is increased. The diacylglycerol content of muscle and liver was decreased, and protein kinase C activity and PKB/Akt activity were increased (Choi et al. 2007a). Non-selective inhibition of ACC-1 and ACC-2 similarly increased fat oxidation in isolated muscle and in vivo, but surprisingly no increase in energy expenditure could be detected, contrasting with the thermogenic effect of the β3AR agonist CL-316243 in the same study (Harwood et al. 2003). Thus, ACC-2 inhibitors may produce similar benefits to β3AR agonists because both types of agent stimulate fatty acid oxidation, but ACC-2 inhibitors do not have a marked acute thermogenic effect. Activators of AMPK offer an indirect approach to ACC-2 inhibition (Cool et al. 2006), but AMPK also regulates enzymes in carbohydrate and protein metabolism and in other metabolic pathways.
Acyl-CoA: diacylglycerol acyltransferase (DGAT) is the last enzyme of the triacylglycerol synthetic pathway. Even though DGAT is an enzyme in an anabolic pathway, DGAT-1 knockout mice have increased energy expenditure. This is due in part to increased physical activity, but also to increased energy expenditure in brown adipose and other tissues. Increased fatty acid oxidation would explain why the concentration of diacylglycerol, the substrate of DGAT, tended to be low in white adipose tissue, skeletal muscle and liver of DGAT-1 knockout mice and not high as might be expected (Chen and Farese 2005; Matsuda and Tomoda 2007; Wang et al. 2007; Yu and Ginsberg 2004). When DGAT-1 knockout mice are given a high-fat diet, they are resistant to accretion of fat. In some experiments, lean body mass was actually increased in the mutant mice (Wang et al. 2007). Insulin sensitivity is less affected by the diet than in wild type mice. A recent publication describes a DGAT-1 inhibitor that reduced weight gain in diet-induced obese mice without reducing food intake or increasing locomotor activity (Zhao et al. 2008). However, suppression of DGAT-2, but surprisingly not DGAT-1, with antisense oligonucleotides improved hepatic insulin sensitivity and lowered hepatic diacylglycerol in rats with diet-induced non-alcoholic fatty liver disease. Expression of carnitine palmitoyl transferase 1 was increased, an indication of increased fatty acid oxidation (Choi et al. 2007b).
Peroxisome proliferator-activated receptor δ (PPARδ) is another target associated with fatty acid oxidation. Over-expression of PPARδ in muscle or white adipose tissue increased mitochondrial number and fatty acid oxidation and resulted in resistance to diet-induced obesity apparently because energy expenditure was raised (Wang et al. 2003). The PPARδ agonists GW-501516 [(2-methyl-4(((4-methyl-2-(4-trifluoromethylphenyl)-1,3-thiazol-5-yl)methyl)sulfanyl)phenoxy)acetic acid] and GW-0742 [2-[4-[[[2-[3-fluoro-4-(trifluoromethyl)phenyl]-4-methyl-5-thiazolyl]methyl]thio]-2-methylphenoxy]acetic acid] similarly increase fatty acid oxidation and reduce diet-induced obesity (Harrington et al. 2007; Tanaka et al. 2003). GW-0742 reduced fat but not lean mass. GW-501516 markedly improved glucose homeostasis despite having a modest effect on body weight. GW-501516 differs from β3AR agonists in not having such a marked thermogenic effect. It has been reported to increase oxygen consumption in mice fed on a high-fat diet (Tanaka et al. 2003), but oxygen consumption was expressed relative to body weight, which contributes less than lean tissue to energy expenditure (Arch et al. 2006). Energy expenditure per mouse appeared little changed.
It seems to be a common feature of the targets described above that their effects on energy expenditure are more subtle than those of β3AR agonists. β3AR agonists mobilise fat from white adipose tissue (Grujic et al. 1997) and uncouple oxidative phosphorylation in brown adipose tissue. They cause a large acute increase in energy expenditure, which is then opposed by limitations in fuel supply. By contrast, interventions in lipid metabolism may have a more sustained effect on energy expenditure. A simple calculation shows that a 5% increase in energy expenditure, which would be difficult to detect if steady over the day, would cause the loss of about 0.1g fat per day in mice—an acceptable effect for a candidate drug.
A final comment in this section is that, in view of their insulin-sensitising effects, it may be a good strategy to register drugs that stimulate fatty acid oxidation for diabetes before registering them for obesity. First, in the USA, payers are more likely to reimburse patients for diabetes than obesity therapy. Secondly, regulatory authorities encourage this strategy. Usually, they require less patient exposure over a shorter time for diabetes than obesity drugs. Furthermore, they state that when considering anti-obesity drugs for registration as diabetes drugs, they will only consider that component of their efficacy that is more than that expected from weight loss. Inexplicably, they do not ask this same question of metformin, pramlintide or exenatide, even though weight loss contributes to their anti-diabetic effects. Whilst regulatory authorities hold this perverse attitude, drugs that improve insulin resistance more than expected from weight loss should be developed first for diabetes, so that their full effect on HbA1c is taken into account in the registration process and weight control can be claimed as an added benefit.
Insights into the mechanism of action of leptin
Although the following comments relate specifically to leptin, there may be similar lessons for the understanding of other hormones and drugs that increase energy expenditure.
Leptin reduces body weight in rodents both by reducing food intake and by increasing energy expenditure (or preventing the decrease that usually accompanies reduced food intake; Halaas et al. 1997; Hwa et al. 1997; Levin et al. 1996). Like β3AR agonists, leptin stimulates fat oxidation and, at least when compared with pair-fed animal, it preserves lean tissue (Alonso and Maren 1955; Breslow et al. 1999; Hwa et al. 1997; Rafael and Herling 2000; Steinberg et al. 2002). Some reduction in lean body mass is to be expected if leptin reduces food intake, but this may be difficult to detect in experiments in which the anorectic effect of leptin wanes as animals lose body fat (Chen and Heiman 2000; Rafael and Herling 2000). Leptin is less effective in lean mice than in Lepob/Lepob mice, which lack leptin (Pelleymounter et al. 1995). This is in part because leptin receptors and downstream signalling mechanisms are upregulated in Lepob/Lepob mice, but it is interesting that β3AR agonists are also less effective in lean than in Lepob/Lepob mice and that the availability of fat for oxidation may be one cause of this difference (Arch et al. 1984b; Wilson et al. 1986).
Leptin also resembles β3AR agonists in that it improves insulin sensitivity in rodents. Both leptin and β3AR agonists must be given for some days for this effect to develop (Widdowson et al. 1998) In most (Grasso et al. 2001; Hidaka et al. 2002; Lin et al. 2002; Shi et al. 1998; Shimomura et al. 1999; Sivitz et al. 1997; Yaspelkis et al. 1999) but not all (Rouru et al. 1999) studies, the effect of leptin was greater than could be explained by any effect that it had on food intake. It was not, however, demonstrated that the improvement in insulin sensitivity was greater than would be achieved by diet restriction sufficient to cause the same weight loss as caused by leptin. This would require a more severe diet restriction than pair feeding.
There is little evidence that leptin raises energy expenditure in humans, but then it has also proved difficult to demonstrate that it reduces food intake in either obese or non-obese humans (Heymsfield et al. 1999). As in rodents, those subjects that are most susceptible to the anorectic effect of leptin are those very rare individuals that lack their own leptin. There was no major effect of leptin on basal metabolic rate or free-living energy expenditure in leptin-deficient subjects, but it is notable that basal metabolic rate did not fall as weight was lost in contrast to weight loss achieved by other means (Farooqi and O’Rahilly 2004). It is also important to recognise that Lepob/Lepob mice have a normal or higher energy expenditure than wild-type mice when they are studied at thermoneutrality and that it is their ability to raise their energy expenditure at lower temperatures that is compromised (Rafael and Herling 2000; Thurlby and Trayhurn 1979; Wilson et al. 1984b). Moreover, in normal mice, leptin did not elicit a general increase in energy expenditure but prevented it from falling in animals that were below thermoneutrality and had a restricted energy intake (Dow 1997). Human subjects are usually studied fed, clothed and in a warm environment—conditions under which the thermogenic effect of leptin may be difficult to detect.
There is some evidence that leptin increases insulin sensitivity in humans. High plasma leptin concentrations usually correlate with an adverse metabolic profile (as they do in diet-induced obese rodents) and an increased risk of type 2 diabetes and cardiovascular disease. However, when the influence of obesity as a confounder is taken into account, high leptin levels are associated with a reduced risk of these diseases (Arch 2007; Ruige et al. 2006; Schmidt et al. 2006). In other words, for a given level of obesity, a high plasma leptin concentration seems to be of benefit to metabolism.
Mechanism of the metabolic effects of leptin
To what extent are those effects of leptin that are independent of reduced food intake mediated centrally, possibly by the sympathetic nervous system, and to what extent by stimulation of peripheral leptin receptors, in particular skeletal muscle receptors?
Intracerebroventricular administration of leptin increases energy expenditure (Hwa et al. 1996; Mistry et al. 1997) and improves insulin sensitivity (da Silva et al. 2006; Lin et al. 2002) consistent with an involvement of the sympathetic nervous system. Furthermore, leptin increases noradrenaline turnover in brown and white adipose tissue (Collins et al. 1996). In fact, leptin seems to activate of all branches of the sympathetic nervous system in rats (Dunbar et al. 1997). However, it enhanced hypothermia-stimulated sympathetic outflow in brown adipose tissue and not in the kidney (Hausberg et al. 2002). This is consistent with leptin being required for thermoregulatory thermogenesis, as mentioned above.
But none of this proves that the centrally-mediated effect of leptin on energy expenditure is entirely due to activation of the sympathetic nervous system. An alternative mechanism may be needed to account for a report that denervation of interscapular brown adipose tissue did not prevent centrally administered leptin from increasing GDP binding to rat brown fat mitochondria (a measure of brown fat activation) in fasted rats (Surmely et al. 1998). The authors suggested that leptin prevented a fall in thyroid hormone levels. Could release of adrenaline in response to leptin be an alternative explanation? Adrenaline release cannot, however, explain a report that combined blockage of α1, β1 and β2 adrenoceptors, or of the β3AR, did not prevent centrally administered leptin from lowering blood glucose in insulin-deficient rats (da Silva et al. 2006).
On the other hand, another intervention, chemical sympathectomy, strongly implicated a role for the sympathetic nervous system in mediating weight loss in response to leptin in rats (Dobbins et al. 2003). Moreover, studies on the effect of leptin in mice that lack all three βARs suggested that the sympathetic nervous system plays a role in leptin-mediated thermogenesis, though the results were not clear-cut. The thermogenic effect of leptin was less than 50% of that in wild-type mice, but some effect seemed to remain, even though it was not statistically significant. In contrast to the sympathectomy study, the weight-reducing effect of leptin was not affected by the absence of βARs (Asensio et al. 2008).
There is little work in humans that supports a role for the sympathetic nervous system in mediating leptin-stimulated thermogenesis, if indeed leptin does stimulate thermogenesis in humans. However, there is a report that plasma leptin concentrations correlated strongly with resting metabolic rate in normal subjects but not in spinal-injury patients in whom the activity of the sympathetic nervous system was impaired (Jeon et al. 2003).
In contrast to this evidence for an involvement of sympathetic activation in leptin-mediated thermogenesis are reports that leptin stimulates oxygen consumption by isolated mouse muscles. Thermogenesis appears to be due to increased lipogenesis and fat oxidation and involve phosphatidylinositol-3-kinase and AMPK signalling (Solinas et al. 2004).
It may be that those metabolic effects of leptin that are independent of its effect on food intake are mediated by both central leptin receptors that raise sympathetic activity and peripheral leptin receptors. Thus, there is evidence that both mechanisms contribute to activation of AMPK by leptin in mouse skeletal muscle. Leptin elicited a biphasic activation of AMPK. One peak of AMPK activation, 15 min after administration of leptin, was not affected by surgical or pharmacological (αAR) blockade of the sympathetic nervous system, whereas a second peak after 6 h was abolished (Minokoshi et al. 2002). Although the increase in AMPK activity after 6 h appeared to be mediated by an αAR rather than a βAR, it was presumably a consequence of the release of catecholamines that would activate all adrenoceptors, and so energy expenditure would be raised. Therefore, given that β3AR agonists elicit clear increases in energy expenditure, whereas interventions in fat metabolism, such as inhibition of ACC-2 or activation of PPARδ, elicit more subtle changes, this report seems consistent with evidence that the effect of leptin on energy expenditure is not immediate but takes at least an hour to develop even when it is given centrally (Mistry et al. 1997): leptin seems to activate the sympathetic nervous system slowly, and it is only when the sympathetic nervous system is activated that an effect on energy expenditure is easily detected. In mice that lack all three βARs, although stimulation of energy expenditure by leptin was reduced, stimulation of fat oxidation was actually much greater in the mutant than in wild-type mice (Asensio et al. 2008). It would not be surprising if fat oxidation was increased within 15 min of administration of leptin and was linked to the early activation of AMPK in muscle.
Finally, it may be noted that Biovitrum has an interest in putative leptin mimetics, exemplified by BVT.3531. This compound reduced food intake and body weight when infused centrally in rats over 7 days (Mirshamsi et al. 2007). If it is truly a leptin mimetic, one would predict that it should activate AMPK in skeletal muscle after 15 min as leptin does (Minokoshi et al. 2002), but only after some hours will sympathetic activity and energy expenditure be increased. The latter effects will be most obvious in fasted animals held at temperatures well below thermoneutrality.
Insights for pharmacology
The pharmacology that led to the identification of the β3AR was quite distinct from the classical pharmacology of β1ARs and β2ARs. To recapitulate, the β3AR has low affinity for standard β1ARs and β2AR antagonists and distinguishes poorly between the more and less active enantiomers. In addition, agonists were discovered that were highly selective stimulants of the β3AR. This situation was confused, however, by the discovery that various β1ARs and β2AR antagonists, notably CGP12177, were, at high concentrations, agonists of the rodent and human β3AR (Arch 2000; Granneman et al. 1991; Nahmias et al. 1991). CGP12177 and other β-blockers had also been shown by Kaumann and co-workers to stimulate cardiac β-adrenoceptors and had been termed ‘non-conventional partial agonists’. Cardiac responses to non-conventional partial agonists, like responses to β3AR agonists, are resistant to most β-blockers. To add to the confusion, Kaumann (1989) had coined the term ‘third cardiac β-adrenoceptor’, though this was subsequently revised to ‘putative β4-adrenoceptor’ (Kaumann 1997). A receptor with similar pharmacology in rodent adipocytes was also called the ‘β4AR’ (Galitzky et al. 1997).
Despite their similarities, the pharmacologies of the β3AR and the putative β4AR are clearly different. The most obvious difference is that various selective agonists of the β3AR have neither agonist nor antagonist effects in cardiac preparations, or if they do so, these are mediated by β1AR or β2AR (Kaumann and Molenaar 1996; Malinowska and Schlicker 1996; Malinowska and Schlicker 1997; Sennitt et al. 1998). Moreover, the ‘β4AR’ was present in atria and brown adipocytes from β3AR knockout mice (Cohen et al. 2000; Kaumann et al. 1998).
The ‘β4AR’ is not a distinct gene product like the β3AR, however. Rather, it is a site or conformation of the β1AR that has low affinity for βAR ligands. Thus ‘β4AR’ pharmacology is absent in tissues from β1AR knockout mice (Kaumann et al. 2001; Konkar et al. 2000). The pharmacology of this secondary or low affinity or CGP12177 site of the β1AR has been studied in cardiac preparations and in cell lines expressing the human β1AR by, among others, Kaumann, Molenaar and co-workers (Kaumann et al. 2007) and Baker and co-workers (Baker 2005b). These workers have shown that agonist effects of β-blockers are not confined to the secondary β1AR site but can also occur at the primary/high affinity/catecholamine site (Baker et al. 2003a). It is possible that the different sites can utilise different signalling pathways (Baker and Hill 2007) and even that the low-affinity site can adopt distinct conformations that signal either by cyclic AMP or p38 mitogen-activated protein kinase (MAPK; Galandrin and Bouvier 2006).
The β2AR also seems to have more than one binding site or conformation. There is strong evidence that these site or conformations can use different signalling pathways (Baker et al. 2003b; Baker and Hill 2007; Heubach et al. 2004; Seifert et al. 1999). Studies using the cloned human β2AR suggest, however, that the low-affinity state of the β2AR is induced over time by high efficacy agonists, which cause receptor phosphorylation (Baker et al. 2003c), whereas the low affinity state of the β1AR is not induced in this way (Baker 2005a).
My purpose is not to review the pharmacology of atypical β1AR or β2AR in any further detail but merely to point out that the discovery of the β3AR prepared minds for their discovery. Comparison of site-directed mutagenesis studies between the βARs might help in the understanding of the structural basis of their pharmacologies. Mutation of Asp138 in the β1AR to Glu has demonstrated that this amino acid is required for the high-affinity but not the low-affinity binding site (Joseph et al. 2004a). The equivalent amino acid in the human β3AR (Asp 117) is important for ligand binding and signal transduction (Gros et al. 1998), so there is no simple equivalence between the β3AR site and the low-affinity β1AR site.
Variable β3AR pharmacology
The existence of dual (or multiple) β1AR and β2AR pharmacology has cautioned pharmacologists against assuming that receptor pharmacology is as simple as written in textbooks and opened drug discovery possibilities that might otherwise have been assumed to be impossible. β3AR pharmacology is also variable. Baker (Baker 2005a) has reported that the affinities of antagonists vary 4- to 50-fold depending on the agonist used to stimulate cyclic AMP formation by human cloned β3AR. Although this is less than the 30- to 1,000-fold variability found in similar studies using β1AR (Baker 2005b; Joseph et al. 2004b), it suggests that the human β3AR has at least two binding sites or conformations.
Baker’s evidence for the existence of at least two binding sites or conformations is based on studies in which the only response measured was cyclic AMP formation. There is also evidence for a conformation of the β3AR that links to a different signalling pathway. Stimulation of the mouse cloned β3AR expressed at low levels or of 3T3-F442A adipocytes increased the rate of extracellular acidification by a cyclic AMP-independent mechanism, possibly involving MAPK and extracellular signal-regulated kinase 1/2 (Erk1/2). Some ligands, notably SR59230A, increased the extracellular acidification rate and activation of the kinases but antagonised cyclic AMP formation. The efficacy of SR59230A was greater than that of CL-316243 for stimulation of the extracellular acidification rate and the kinases but less than that of CL-316243 for stimulation of cyclic AMP formation (Hutchinson et al. 2005; Sato et al. 2007). A study from the same laboratory (of Summers) using the human cloned β3AR found that zinterol had similar potency for stimulation of cyclic AMP accumulation and extracellular acidification rate, whereas CL316243, salbutamol and clenbuterol were as much as 100-fold more potent as stimulants of extracellular acidification rate (Hutchinson et al. 2006). These results suggest not only that there is a conformation of the β3AR associated with activation MAPK and Erk1/2 but not cyclic AMP formation but also that some ligands direct signalling selectively down this pathway. Similar evidence had been presented previously for ligand-directed activation of either Erk1/2 activation or cyclic AMP formation by the β3AR (Gerhardt et al. 1999). This concept is important to pharmacologists, owing to the opportunities that it presents for discovering drugs with different properties that act at the same receptor (Urban et al. 2007).
Ligand-directed signalling means that the relative affinities of ligands may vary according to which site mediates the functional response measured. It also suggests that the relative affinities of ligands in binding studies may vary according to the site to which the radioligand binds. Iodination of cyanopindolol has little influence on its affinity for the β3AR: the Kd value of [125I]-cyanopindolol was 2.5 nM, and the Ki value of cyanopindolol was 8.2nM when competing with [125I]-cyanopindolol for the human β3AR (Wilson et al. 1996). However, the Ki value of cyanopindolol when it was competing with the radiolabelled active enantiomer of BRL-37344 was far higher than the Kd value of [125I]-cyanopindolol. BRL-37344, by contrast, had similar potency relative to isoprenaline whether competing with [125I]-cyanopindolol or the radiolabelled active enantiomer of BRL-37344. The nature of the labelled ligand altered the relative potencies of BRL-37344 and [125I]-cyanopindolol/cyanopindolol by 400-fold (Muzzin et al. 1994; Muzzin et al. 1991). In another study, cyanopindolol was 370-fold more potent than isoprenaline when displacing [125I]-cyanopindolol from membranes that expressed human β3AR, whereas isoprenaline was 230-fold more potent than cyanopindolol as a stimulant of cyclic AMP accumulation in whole cells (Arch 2002; Wilson et al. 1996). Even more intriguing, isoprenaline was only 1.5-fold more potent than cyanopindolol when the response measured was cyclic AMP production by membranes from the cells used for measuring cyclic AMP accumulation. This paradox has never been resolved. Drug-discovery scientists now recognise that screens for agonists should ideally be functional screens and that the choice of functional response measured may be important in predicting the properties of a drug. Those who sought β3AR agonists in the past discovered these issues for themselves.
This indeed is the message of this review: the study of the β3AR and β3AR agonists has given us insights and perspectives. They may yet give us a drug.
The author thanks his many colleagues and collaborators in his work on β3-adrenoceptor agonists, notably Mike Cawthorne, Shelagh Wilson, John Clapham and Alberto Kaumann. He thanks Kenneth Langlands, Rob Ward, Don Smyth, Matthew Coghlan and Julie Cakebread for assistance in the preparation of this manuscript, and Frederic Preitner and Cedric Asensio for helpful discussions.