David Tsao, Heather Koenig Thomsen, Joyce Chou, Jennifer Stratton, Michael Hagen, Carole Loo, Carlos Garcia, David L. Sloane, Arnon Rosenthal and John C. Lin

Mutations in the tyrosine kinase receptor trkB or in one of its natural ligands, brain-derived neurotrophic factor (BDNF), lead to severe hyperphagia and obesity in rodents and/or humans. Here, we show that peripheral administration of neurotrophin-4 (NT4), the second natural ligand for trkB, suppresses appetite and body weight in a dose-dependent manner in several murine models of obesity. NT4 treatment increased lipolysis, reduced body fat content and leptin, and elicited long-lasting amelioration of hypertriglyceridemia and hyperglycemia. After treatment termination, body weight gradually recovered to control levels in obese mice with functional leptin receptor. A single intrahypothalamic application of minute amounts of NT4 or an agonist trkB antibody also reduced food intake and body weight in mice. Taken together with the genetic evidence, our findings support the concept that trkB signaling, which originates in the hypothalamus, directly modulates appetite, metabolism, and taste preference downstream of the leptin and melanocortin 4 receptor. The trkB agonists mediate anorexic and weight-reducing effects independent of stress induction, visceral discomfort, or pain sensitization and thus emerge as a potential therapeutic for metabolic disorders.

THE LAST DECADE has witnessed a global rise of obesity prevalence along with its complications such as type 2 diabetes. Advances in our understanding of the molecular pathways and neuroendocrine mechanisms of energy homeostasis, such as leptin and melanocortin systems (1, 2), may help to avert or alleviate the obesity epidemic.

Brain-derived neurotrophic factor (BDNF) and its receptor trkB were originally studied for their function in sensory neuron development (3), but are also implicated in the neuroendocrine control of mammalian feeding behavior and energy homeostasis (4). Mice (5) or humans (6) with a hypomorphic or loss-of-function allele of the tyrosine kinase receptor trkB display excessive appetite, reduced energy expenditure, and morbid obesity. The expression of BDNF in the ventromedial hypothalamus (VMH), an important component of the central nervous system (CNS) energy homeostasis circuitry, is reduced after food deprivation (5, 7), and mice with genetically reduced BDNF expression (7, 8, 9) or with postnatal forebrain excitatory neuron-selective BDNF deficiency (10) also develop severe obesity. TrkB is thought to act downstream of the melanocortin 4 receptor (Mc4r), a well established hypothalamic signaling system that mediates control of appetite in the arcuate nucleus by hormones such as leptin and insulin (4, 11). Thus, the Mc4r-deficient Ay lethal yellow mice, which are hyperphagic and obese, display reduced expression of BDNF in the VMH, and their obese phenotype is rescued by intracerebroventricular injections of BDNF (5).

The CNS feeding circuits receive hormonal signals such as leptin that reflect long-term energy status as well as vagally transmitted neural signals that sense short-term accumulation and movement of food in the gastrointestinal tract. Interestingly, mice deficient in neurotrophin-4 (NT4), the second trkB ligand, have 55% loss of nodose ganglion neurons and 80–90% of vagal intraganglionic mechanoreceptors in the small intestine. These mice suffer alterations in short-term meal patterns but display normal daily food intake and body weight because of compensatory changes in other meal parameters (12). One explanation for the lack of overt hyperphagia and obesity phenotype in the NT4-deficient mice is that BDNF, but not NT4, is expressed in the CNS regions critical for long-term feeding behavior and body weight homeostasis, such as the VMH. Alternatively, the phenotypic difference between BDNF−/− and NT4−/− mice regarding feeding and obesity may result from quantitative or qualitative differences in their ability to activate the common trkB signals. Support for the second hypothesis is provided by the finding that the trkBshc/trkBshc mutant mice exhibited a nearly complete loss of the NT4-dependent sensory neurons but only a modest loss of the BDNF-dependent neurons (13).

The mechanism by which trkB regulates feeding behavior and weight is not well understood. In addition, it is not known whether the central anorexigenic trkB system can regulate obesity of different causes, including monogenic leptin receptor deficiency, polygenic obesity, and high-fat diet-induced obesity (DIO). Leptin, for example, is effective in treating obesity due to leptin deficiency but not in high-fat DIO. Moreover, because BDNF and NT4 bind and signal also through the low-affinity neurotrophin receptor p75NTR, it is not known whether trkB activation alone is sufficient to mediate antiobesity effects. In addition, because trkB is implicated in a nociceptive pathway (14, 15), and severe pain can lead to anorexia and weight loss, it is important to establish whether the metabolic and nociceptive functions of trkB are dissociable. Finally, it would be important to determine whether the trkB system plays any role in taste preference, an important aspect of feeding behavior.

To expand our insight into the role of trkB agonists in energy homeostasis, we conducted a series of gain-of-function experiments using NT4 and a trkB-specific agonist antibody as pharmacological tools. We found that exogenously applied trkB agonists effectively reduce food intake and body weight and modify taste preference to high-calorie liquid food. TrkB agonists did not induce starvation-related stress response, LiCl-like visceral discomfort, or nerve growth factor (NGF)-like pain sensitization and were effective in several mouse models, including DIO, polygenic obesity, and db/db mice. Our pharmacological studies, together with the previous genetic evidence, strengthen the notion that trkB signaling is a key mediator of energy homeostasis that acts downstream of both the melanocortin and the leptin cascades. Given the prevalence of human obesity caused by Mc4r dysfunction and widespread leptin resistance in the human population, trkB modulators could emerge as attractive therapeutics for metabolic disorders.

Proteins

Recombinant human NT4 protein was purified from an Escherichia coli culture engineered to overexpress NT4, using a modification of published procedures (16, 17). Briefly, E. coli cell paste was suspended at a concentration of 100 mg/ml in Tris/EDTA buffer (20 mM Tris; 5 mM EDTA, pH 8.0) and disrupted by homogenization (three passes at 660 bar). NT4, in insoluble inclusion bodies, was then harvested by centrifugation. This insoluble fraction was then solubilized in 6 M urea, 25 mM dithiothreitol (DTT). Polyethyleneimine was added to 0.2%, and the sample was clarified by centrifugation. This clarified supernatant was applied to a DEAE-Sepharose FF column (GE Healthcare, Piscataway, NJ), which had been equilibrated with 20 mM Tris, 6 M urea, and 10 mM DTT (pH 8.0). The pH of the nonbound flow-through fraction was adjusted to 5.0 with glacial acetic acid, and this fraction was applied to an SP-Sepharose FF column (GE Healthcare) that had been equilibrated with 20 mM sodium acetate, 6 M urea (pH 5.0). NT4 was eluted from this column by washing the column in 20 mM sodium acetate, 6 M urea, 0.5 M NaCl (pH 5.0). The NT4 in the eluate was greater than 95% pure. The protein concentration of the eluate was adjusted to 10 mg/ml and buffer exchanged to 0.2 M Tris, 4 M GuHCl, 5 mM DTT (pH 8.3) by ultrafiltration /diafiltration on cellulose acetate membranes (Millipore, Billerica, MA). The NT4 was then refolded by first adding oxidized glutathione to 20 mM, then adding 19 vol 100 mM Tris, 20 mM glycine, 1 M GuHCl, 15% PEG 300 (pH 8.3), and then adding cysteine to 3 mM. This solution was then saturated with nitrogen and was stirred at 4 C for 16–18 h. The progress of refolding was monitored by reverse-phase HPLC. When refolding was determined to be complete, the protein was concentrated approximately 10-fold and then buffer exchanged to 10 mM sodium acetate (pH 4.0). NT4 was diluted to 0.5 mg/ml, and 0.750 M NaCl was added. This solution was then applied to a phenyl-650 M hydrophobic interaction chromatography column (Tosoh, Tokyo, Japan). The resin was equilibrated with 10 mM sodium acetate (pH 4.0) and 2.5 M NaCl. The protein was washed with 10 mM sodium acetate (pH 4.0) and 0.5 M NaCl. NT4 was eluted with 10 mM sodium acetate (pH 4.0). Protein was concentrated to approximately 3 mg/ml and tested for purity, stability, and endotoxin levels. LAL assay (Charles River-Endosafe-PTS, Wilmington, MA) resulted in less than 3.5 EU/mg after hydrophobic interaction chromatography purification. Before use in the animal studies, each lot of purified NT4 protein was shown to bind and activate trkB receptor in vitro (18) as well as supporting survival of embryonic nodose neurons (data not shown).

TrkB agonist monoclonal antibody (mAb)

TrkB antibodies were generated by immunizing BALB/c mice with the extracellular domain of recombinant trkB protein. Hybridomas were screened for trkB binding and agonist activity (experimental details will be described elsewhere). The ascites obtained from a single BALB/C mouse was purified batchwise with protein A resin. Approximately 1 ml resin equilibrated with PBS was incubated with ascites fluid overnight at 4 C with gentle agitation. The resin was spun down and washed twice with PBS. Antibody was eluted from the resin using 5 vol 50 mM sodium citrate-phosphate buffer (pH 3). Eluates were immediately neutralized with 1 M HEPES buffer (pH 7) and then dialyzed into PBS, concentrated, and sterile filtered. The antibody was shown to be a specific TrkB agonist by a cell-based receptor tyrosine kinase activation assay as shown in supplemental Fig. S4 (published as supplemental data on The Endocrine Society’s Journals Online web site at http://endo.endojournals.org) (18).