Am J Physiol Regul Integr Comp Physiol. 2011 May; 300(5): R1053–R1069.
Published online 2011 Feb 23. doi: 10.1152/ajpregu.00776.2010
Emily E. Noble, Charles J. Billington, Catherine M. Kotz, and ChuanFeng Wang
Brain-derived neurotrophic factor (BDNF) mediates energy metabolism and feeding behavior. As a neurotrophin, BDNF promotes neuronal differentiation, survival during early development, adult neurogenesis, and neural plasticity; thus, there is the potential that BDNF could modify circuits important to eating behavior and energy expenditure. The possibility that “faulty” circuits could be remodeled by BDNF is an exciting concept for new therapies for obesity and eating disorders. In the hypothalamus, BDNF and its receptor, tropomyosin-related kinase B (TrkB), are extensively expressed in areas associated with feeding and metabolism. Hypothalamic BDNF and TrkB appear to inhibit food intake and increase energy expenditure, leading to negative energy balance. In the hippocampus, the involvement of BDNF in neural plasticity and neurogenesis is important to learning and memory, but less is known about how BDNF participates in energy homeostasis. We review current research about BDNF in specific brain locations related to energy balance, environmental, and behavioral influences on BDNF expression and the possibility that BDNF may influence energy homeostasis via its role in neurogenesis and neural plasticity.
Keywords: food intake, body weight, ventromedial hypothalamus, paraventricular nucleus, brain-derived neurotrophic factor
brain-derived neurotrophic factor (BDNF) is a member of the neurotrophin family of growth factors (151), along with nerve growth factor (152), and neurotrophin (NT) 3 (67, 163), NT 4/5 (28), and NT 6 (88). Neurotrophins are synthesized as 32–35-kDa pro-isoforms, which are later cleaved to mature forms that dimerize after translation and then act as receptor ligands (136). Whereas the precursor forms of other neurotrophins are constitutively secreted, the 32-kDa pro-BDNF is packaged into vesicles of a regulated pathway and is secreted in an activity-dependent manner (87). Pro-BDNF may be secreted as is (48), cleaved by the extracellular protease plasmin (202), or interact with the pan-neurotrophin receptor p75NTR and other receptors that cause an independent biological effect (244). Alternatively, pro-BDNF is processed to the mature form intracellularly by furin or proconvertases, where it forms C-terminal dimers (212, 226).
Mature BDNF is considered the biologically active form, which has a high affinity for the tropomyosin-related kinase B (TrkB) receptor (130). Both BDNF and TrkB are present in presynaptic axon terminals and postsynaptic dendritic compartments of neurons, and they are capable of bidirectional release and activity [for review, see Tyler et al. (259)]. Typical of the neurotrophic factors, BDNF stimulates the development and differentiation of new neurons (3, 131) and promotes long-term potentiation (LTP) (139, 140, 205), and neuron survival (97, 105, 116). BDNF is abundantly expressed throughout the developing and mature CNS and in many peripheral tissues, including muscle, liver, and adipose (42, 159, 182, 261). Regional differences between BDNF mRNA levels and protein concentrations in the CNS are often reported (7, 51, 192, 193), which may be related to regulatory mechanisms, mRNA decay (164), or BDNF anterograde transport (7).
BDNF is synthesized in several areas of the hypothalamus, including the paraventricular nucleus (PVN), the ventromedial hypothalamic nucleus (VMN), the dorsomedial hypothalamic nucleus (DMN), and the lateral hypothalamic area (LH) (51). Additionally, BDNF-immunoreactive fibers have been identified in the arcuate nucleus (Arc); however, the Arc does not appear to be a site of BDNF synthesis (51). BDNF is also widely expressed throughout the hippocampus; amygdala; select areas of the thalamus, including the ventral tegmental area (VTA); and in areas of the hindbrain, including the dorsal vagal complex (DVC) (17, 51). In recent years, much attention has been given to BDNF for its role in energy homeostasis. While examining BDNF and neuronal plasticity in vivo, Lapchak et al. (145) noted that chronic intraventricular administration of BDNF prevented weight gain. It has since been observed in many studies that central administration of BDNF induces appetite suppression and weight loss (207, 267, 269, 270), increases locomotor activity (186), and resting metabolic rate (266, 268). An obese phenotype is also observed in BDNF-conditional knockout mice, where BDNF is deleted after birth and the knockout is restricted to the brain (216). In addition to central effects, BDNF exerts peripheral actions that affect glucose metabolism (287, 289, 290), energy expenditure (288, 290), and food intake (287). Both central and peripherally administered BDNF lowers blood glucose and increases energy expenditure in animal models of type 2 diabetes (188). The combined effects of central and peripheral BDNF are apparent from instances where BDNF is globally reduced, as is the case in rodents and human subjects with haploinsufficiency for the gene, which results in obesity and hyperphagia (91, 99, 123). The neurotrophin receptor TrkB is a receptor tyrosine kinase, which upon activation, results in receptor dimerization, followed by receptor transphosphorylation and the initiation of intracellular signaling cascades. A human mutation affecting the ability of the TrkB receptor to autophosphorylate is associated with obesity and hyperphagia (291). The TrkB receptor exists in full length and two truncated forms, but only the full-length receptor contains intracellular tyrosine kinase activity (4). The two truncated forms are generated by alternative splicing of the full-length receptor, and they are capable of inhibiting activity of BDNF by forming heterodimers with full-length TrkB (65), or by binding and internalizing BDNF (98). The role of BDNF in obesity appears to, at least in part, involve signaling through the full-length TrkB receptor, as TrkB hypomorphs, expressing ¼ of the full-length TrkB receptor of wild-type mice, are obese (283, 284). In the absence of BDNF the antiobesity effect of TrkB is still possible. This is evidenced by the use of two different BDNF agonists, the TrkB ligand NT4 and a TrkB-specific antibody that acts as a receptor agonist, which when administered to the hypothalamus of mice, caused reductions in food intake and resistance to diet-induced obesity (DIO), as well as polygenic and leptin receptor deficiency-associated obesity (255). It is important to note that BDNF appears to be the main natural ligand for TrkB, as rodents heterozygous for BDNF, but not NT4, display the obese phenotype (123). Thus, both BDNF and TrkB are necessary for BDNF-mediated effects on energy balance (255).
The human BDNF gene contains a total of 10 exons coding for the 5′ untranslated region and are alternatively spliced to a common 3′ coding exon, resulting in 34 possible mRNA transcripts (211). In rodents BDNF contains 9 exons, encoding for 24 different mRNA transcripts, each of them ultimately translating into an identical mature BDNF [reviewed by Cunha (54)]. The expression of BDNF transcripts is tissue-specific, that is, it is differentially expressed throughout different brain sites and peripheral tissues (29, 247). Moreover, environmental cues, such as stress, can alter transcript expression (80, 167, 248), which is associated with alterations in pro-BDNF/total BDNF ratio (248). The exact role of each of these individual transcripts is still largely unknown. In the VMN of rodents, transcripts encoding for BDNF exons I, II, and IV are expressed, whereas exon III is not (254). Transcription of exons I and IV in this region is regulated by steroidogenic factor 1 (SF-1), which, when reduced (as in the case of SF-1 heterozygotes), impairs hypothalamic function and results in hyperphagia and weight gain (254). Different genetic variants impact the activity of BDNF by affecting biosynthesis and/or post-translational processing of the pro-BDNF precursor (183). One variant, in particular, the single nucleotide polymorphism (SNP), which results in a substitution of a valine for methionine residue at position 66 (Val66Met), has been identified as having a strong correlation with eating disorders (ED), specifically anorexia nervosa (AN) of the restricting type, and is associated with low body mass index (BMI) (214). This polymorphism alters the intracellular packaging of pro-BDNF and may affect activity-dependent secretion of the mature BDNF peptide (48, 64). According to a recent meta-analysis, the Val66Met polymorphism is associated with a 33% increased risk for ED (90); however, this analysis did not subdivide ED by category. Recent evidence is conflicting regarding this gene variant, as some have found no association between ED and Val66Met (11), or AN (56), and one study reported a link between Val66Met and obesity in females (24). The conflicting evidence regarding ED and the BDNF SNP Val66Met reflects both the complexity of eating disorders and the range of factors affecting feeding behavior.
In a recent genome-wide association study, BDNF was one of 18 gene loci, where having a certain SNP variant was associated with higher BMI (235). Another study examined 41 different SNPs near the BDNF locus in 87 adults with sudden death and made comparisons between BMI and BDNF mRNA in the VMN of the cadavers. In the subjects with extreme obesity (BMI ≥ 40 kg/m2), BDNF expression was reduced twofold compared with overweight and obese individuals. There was an association between homozygosity for the minor C allele at rs12291063, reduced VMN BDNF expression, and high BMI, which suggests that having the SNP at rs12291063 may be a risk factor for obesity (201). In mice, heterozygosity for Bdnf decreases hypothalamic expression and results in hyperphagia and obesity (123). WAGR (Wilms’ tumor, aniridia, genitourinary anomalies, and mental retardation) syndrome is a rare disorder characterized by heterozygous gene deletions in at least two genes located near BDNF in the 11p13 region, and sometimes accompanies the heterozygous deletion of BDNF. In a study of individuals with WAGR syndrome, 100% of those heterozygous for BDNF deletion were obese by the age of 10, in contrast with 20% of those without BDNF deletion (99).
BDNF and the Central Regulation of Energy Metabolism
BDNF was first observed to affect energy metabolism with intracerebroventricular administration (207). In recent years, studies have identified additional sites of BDNF action regulating energy balance, including the DVC, hypothalamic PVN and VMN, the VTA, amygdala, and possibly the hippocampus (17, 18, 32, 52, 57, 58, 266–270, 272). In regulating energy balance, BDNF interacts with several other neuropeptides, including melanocortin (18, 39, 196, 255, 284), leptin (17, 18, 40, 137, 268, 269, 272), corticotrophin-releasing hormone (CRH) (35, 39, 40, 251, 271), and thyrotropin releasing hormone (TRH) (35, 233, 260).
In 1995, Pelleymounter et al. (207) discovered that intracerebroventricular administration of BDNF decreased energy intake and body weight of rats, which was associated with a dose-dependent increase in serotonin turnover. A pair-fed group had comparable weight loss, but the recovery weight gain in these rats was much faster than the BDNF-infused group (207), suggesting BDNF promotes lasting metabolic changes. Additional evidence for a central role of BDNF in the regulation of energy balance is apparent, as animals with reduced Bdnf expression, either due to a conditional homozygous knockout in the brain or due to heterozygous gene expression (Bdnf +/−), develop hyperphagia, obesity, and resistance to insulin and leptin (123, 216). Administration of BDNF intracerebroventricularly reverses the hyperphagic and obese phenotype of Bdnf +/−mutant mice (123). A single intracerebroventricular injection of BDNF is sufficient to improve insulin receptor signaling in the liver of streptozotocin-induced diabetic mice, whereas no direct effect of BDNF on cultured hepatocytes has been observed (256). This indicates that independently of anorectic effects, centrally administered BDNF may affect glucose metabolism. Nonomura et al. (197) observed that intracerebroventricular BDNF dose-dependently lowers blood glucose and increases pancreatic insulin content in leptin receptor-deficient db/db mice and does so independently of food intake. These improvements were also associated with increased norepinephrine turnover and uncoupling protein-1 (UCP-1) expression in brown adipose tissue (BAT) (197). Taken together, it appears the central BDNF enhances energy expenditure via activation of the sympathetic nervous system and improves blood glucose in obese, diabetic rodents. It should be noted that BDNF does not affect blood glucose in normoglycemic rats (207).
The adult rat hypothalamus contains high levels of BDNF (119, 193, 241), and most hypothalamic neurons express the TrkB receptor (43, 168, 176). Overexpression of the Bdnf gene in the hypothalamus is associated with increased heat production, respiratory exchange ratio, and resting metabolism, and increased hypothalamic expression of TrkB, insulin receptor, CRH, and TRH (39, 40). Increased Bdnf expression is also associated with sharp decreases in leptin and insulin concentrations, and increases in the adipose tissue-secreted hormone adiponectin (39), which is associated with increased fatty acid oxidation, enhanced glucose metabolism, and weight loss (79). We have found that BDNF reduces food intake and influences energy expenditure when injected into certain specific hypothalamic sites, but not others. For example, although the LH expresses BDNF and its receptor, specific site injection of BDNF did not significantly reduce feeding and body weight (267). The hypothalamic PVN is responsive to physiological stimuli, is involved in stress responses (21, 86), and contains high levels of BDNF and TrkB mRNA (241). We found that injections of BDNF in the PVN increases energy expenditure, mainly by increasing resting metabolic rate, and increasing thermogenic capacity, as indicated by elevation of UCP-1 expression in BAT (266). We have also found that a single injection of BDNF reduces food intake and body weight (267) for up to 48 h after injection, suggesting a prolonged and potent effect. Animals who were made obese with a high-fat diet (HFD) and subsequently given PVN injections of BDNF on alternate days over an extended period, had significant reductions in energy intake, body weight, and body fat (including visceral fat), compared with artificial cerebrospinal fluid-injected controls. BDNF also normalized HFD-induced hyperglycemia, hyperlipidemia, hyperinsulinemia, and hyperleptinemia, suggesting chronic BDNF in the PVN improves metabolic syndrome and associated resistance to insulin and leptin (270). Furthermore, the animals more susceptible to DIO were also more sensitive to PVN injections of BDNF (270).
Stress paradigms increase expression of BDNF mRNA in the PVN (213), which is associated with decreased inhibitory synaptic input leading to activation of PVN neurons (264, 265). The mechanism for BDNF-associated removal of inhibition was elucidated by Hewitt and Bains (103), who observed that through TrkB receptor activation on postsynaptic neurons, BDNF reduces the surface expression of inhibitory GABAA receptor clusters in the PVN. Thus, it is likely that BDNF increases the firing rate of PVN neurons involved in the stress response. CRH neurons in the PVN regulate locomotor activity and body temperature (158, 217), and, thus, the removal of inhibitory GABAA receptors on these neurons would likely account for the physiological and behavioral effects observed when BDNF is injected directly in this area. Naert et al. (186) observed that BDNF infused continuously into the lateral ventricle of rats causes a significant increase in paraventricular CRH and AVP mRNA, which is associated with increased locomotor activity, body temperature, and reduced body weights (186). BDNF overexpression in the hypothalamus also resulted in increased CRH expression (39), and hypothalamic increases in BDNF and CRH were associated with β-adrenergic receptor activation (40). Again, these data support the notion that the sympathetic nervous system is important in BDNF-associated metabolic changes. Toriya et al. (251) report that PVN-injected BDNF-induced reduction in feeding and body weight is mediated via CRH-R2, and accordingly, the effects of BDNF were blocked using a CRH antagonist. Similarly, we have observed that the effect of BDNF on feeding and body weight gain due to BDNF injections in the VMN or PVN was attenuated by pretreatment with a CRH antagonist (271). In the PVN, mRNA for BDNF and CRH are colocalized (186), and there is also colocalization for TrkB receptor and CRH (251), suggesting potential signaling between the two neuropeptides in the regulation of energy metabolism. (Fig. 1/See link)
The PVN is a site for TRH synthesis and secretion, and TRH plays an important role in the control of energy homeostasis (149) through the TRH-TSH-tyrosine hydroxylase (TH) cascade. Triiodothyronine (T3) activates orexigenic neurons of the VMN and stimulates food intake independently of energy expenditure (138). TRH neurons express BDNF in response to immobilization stress (233). Interestingly, BDNF and T3 have opposing effects on the expression of several obesity-related genes in the hypothalamus (35). BDNF increases, while T3 decreases expression of BDNF, leptin receptor, proopiomelanocortin (POMC), TRH, and agouti-related protein (AgRP) (35). Through TrkB receptor signaling, BDNF increases the expression of the TRH precursor pre-pro-TRH mRNA in PVN neurons (260). Thus, in the PVN, BDNF may, in part, contribute to negative energy balance via increasing TRH, or, conversely, TRH may affect energy balance by increasing BDNF.
POMC and AgRP [neuropeptide Y (NPY/AgRP)] neurons are two distinct populations of neurons that project from the hypothalamic Arc to the PVN and release alpha-melanocyte stimulating hormone (α-MSH) and AgRP, respectively (71, 72), which interact with the melanocortin receptor (MC3/4R) in the PVN to elicit opposing actions on food intake. The MC4R is a membrane-bound α-MSH receptor (199), highly expressed in both the hypothalamus and brain stem (281), and is important for maintaining a lean phenotype. Animals heterozygous for, or lacking the MC4R gene become obese, (108), and MC4R agonist infusion reduces food intake and lowers body weight in HFD-fed rats (225). Delivery of BDNF gene into the hypothalamus increases the expression of MC4R in animals fed with regular chow or a HFD (39), suggesting BDNF affects melanocortin signaling (39). Notably, activation of MC4R leads to acute elevations of hypothalamic BDNF, which is critical for melanocortinergic effects on appetite and body temperature (196). BDNF is highly expressed in the VMN (262), which is an important area for regulating energy metabolism (220). MC4R controls BDNF expression in the VMN, and infusion of BDNF in the brain of MC4R-deficient mice attenuates hyperphagia and excessive weight gain induced by a moderate fat diet (25.1% calories from fat) (284). In addition, the phenotype of the TrkB homomorph, a mutant with reduced BDNF/TrkB signaling, is similar to the MC4R-null mutant mouse (284). The effects of BDNF in MC4R signaling are dependent on TrkB activation, as TrkB ligands reduce food intake and body weight downstream of MC4R in the hypothalamus of mice (255). Together, these data suggest that BDNF is a downstream effector of MC4R activation and that BDNF-TrkB signaling is an essential part of the mechanism for the anorectic and obesity-resistant effects of MC4R agonists in the VMN. (Fig. 1/See link)
The VMN is involved in the control of autonomic responses that contribute to the prevention of obesity [reviewed by King (126)]. Neurons of the VMN project to many areas associated with feeding behavior, including the amygdala (37, 221), Arc (237), LH (224), PVN (157), the DMN (162, 245), as well as areas relating to rewarding aspects of feeding behavior, including the VTA (221), the NA, and the nucleus of the solitary tract (37). The VMN receives inputs from the Arc (15, 101), the LH (70, 221, 245), and the amygdala (161, 169). The VMN is involved in promoting satiety, as lesions in this area are associated with hyperphagic behavior (127). SF-1 is a nuclear hormone receptor important to the developmental structure of the VMN (109, 231). SF-1 is coexpressed with BDNF in the VMN and is involved in BDNF synthesis. Reduced levels of SF-1, as seen in heterozygous animals are associated with reduced BDNF, increased weight gain, hyperphagia, and lower daytime metabolic rate (254).
We found that a single injection of BDNF directly into the VMN, at doses not causing taste aversion, significantly decreases normal feeding and deprivation- and NPY-induced feeding for up to 48 h (269). No effects on feeding behavior were observed during the initial 4-h post injection, indicating that the feeding effects of injected BDNF might be indirect, or might take place as a result of retrograde (184, 185) or anterograde (7, 51, 234) transfer to a different brain location (269). The possibility that BDNF may act locally to reduce food intake by altering synaptic strength or receptor expression in the VMN has not been adequately investigated, but it is worth considering. In contrast to the delayed anorectic action, BDNF in the VMN immediately increased energy expenditure by elevating resting metabolic rate and physical activity (268). Chronic BDNF in the VMN also reduces HFD-induced obesity by reducing energy intake and/or increasing energy expenditure based on the phenotype of the animals on a HFD (272). Unlike in the PVN, BDNF in the VMN significantly increases physical activity in animals on regular chow (268) or HFD, suggesting that elevated physical activity-induced energy expenditure contributes to increased thermogenesis induced by BDNF in the VMN. BDNF in the VMN also decreases respiratory exchange ratio in animals on regular chow (268) and a HFD (272), indicating that BDNF stimulates fat metabolism, which partially explains the preferential loss of fat (vs. lean) mass after BDNF treatment. Deletion of BDNF in the VMN causes hyperphagia and obesity in mice (262), further confirming the importance of BDNF as a contributor to the maintenance of a lean phenotype. BDNF expression is responsive to dietary cues, as transcriptional levels of BDNF in the VMN are reduced by fasting (284) and elevated by glucose (262). It is possible that HFD-induced obesity results from lack of responsiveness to these dietary cues. In support of this, Yu et al. (292) observed decreased BDNF mRNA in the VMN of diet-induced obese mice, compared with mice resistant to HFD-induced obesity (292).
The adipokine leptin is produced and released from adipose tissue. Leptin signaling in the CNS inhibits food intake and increases energy expenditure, and in so doing, counters the accumulation of adiposity. Of note, leptin increases BDNF in certain brain areas, including the DVC (17), VMN, and DMN (137), with reductions in food intake (17, 18, 137, 268, 269) (Fig. 1/See link). However, the antiobesity antidiabetic effects of BDNF on energy metabolism occur downstream of leptin signaling, as the effects are observed in leptin receptor deficient db/db (256–258), Kkay (an animal model of metabolic syndrome) (187), and diet-induced obese (DIO) mice (187, 255). Since leptin receptor signaling is important for leptin-dependent BDNF up-regulation (137), the obese phenotype might partially be related to the inability of leptin to increase hypothalamic BDNF expression. While leptin increases hypothalamic BDNF, BDNF decreases leptin production in adipocytes, an effect that involves sympathoneural β-adrenergic signaling and the hypothalamic-pituitary-adrenal axis (HPA) (40). HFD induces leptin resistance, characterized as a reduced anorectic response to leptin, as well as hyperleptinemia. We found that VMN BDNF significantly attenuated hyperleptinemia compared with that prior to BDNF intervention, or to vehicle-treated control animals (272) on a HFD. Additional studies are needed to explore whether the observed attenuated hyperleptinemia after BDNF is associated with or is an indication of improved leptin sensitivity.
Both BDNF and the TrkB receptor are highly expressed in the DVC of the hindbrain (51). The DVC is located in the caudal brain stem, an autonomic integrator of food intake control (17) and is involved in integrating satiety signals emanating from peripheral fat stores (148). BDNF acts as an anorexigenic factor in the DVC; Bariohay et al. (17) reported that BDNF infusion in the DVC induced anorexia and weight loss. Noteworthy, the efficacy of BDNF as an anorectic agent in the DVC decreased over a 14-day infusion period, indicating some compensation or desensitization occurs. The protein content of BDNF in the DVC decreases after 48 h of food deprivation and increases upon refeeding. Furthermore, the anorexigenic hormones leptin and CCK injected peripherally increase BDNF content in the DVC (17). The DVC contains the neural network responsible for the central pattern generator of swallowing (111) and BDNF-TrkB signaling inhibits the swallowing reflex via modulation of GABAergic signaling (19). Similarly, increased stimulation of the superior laryngeal nerve decreases BDNF in the DVC, indicating positive feedback allowing the swallowing reflex to continue with the presence of a food stimulus (19, 148). BDNF is a downstream effector of the MC4R signaling pathway in the DVC and is necessary for the anorexigenic effect of MC4R activation (18). The orexigenic effect of an MC4R antagonist is abolished with coadministration of BDNF, and pharmacological blockade of the TrkB receptor attenuates the anorexigenic effect of an MC4R agonist (18). Taken together, BDNF in the DVC appears to be responsive to hormonal satiety signals, as well as physical signals of the presence of a food stimulus, and coordinates swallowing. (Fig. 1/See link)
MC4R is expressed in the amygdala (181) an area involved in regulating macronutrient selection (128) and some reward aspects of feeding behavior (75, 122). In the amygdala, injection of MC4R agonist causes a dose-dependent reduction in food intake, which is greater in animals fed a HFD (32). Surprisingly, while injections of the orexigenic hormone AgRP in the amygdala increases food intake, it is also associated with elevated amygdala BDNF mRNA (32). This effect is unexpected and warrants further investigation, as compelling evidence suggests that BDNF is a downstream effector of anorexigenic melanocortinergic signaling (32).
BDNF and TrkB are expressed in the mesolimbic dopamine system, which is associated with hedonic reward (198, 229). The consumption of palatable high-fat foods alters the expression of BDNF and TrkB receptor in the VTA, but not the nucleus accumbens (NAc) of wild-type mice (52). BDNF is not highly expressed in the NAc. Most of the BDNF found in the NAc is produced in the VTA and anterogradely transported from neurons that originate there (51, 198). The neurons of the NAc release dopamine in response to palatable foods (22). Site-specific viral depletion of BDNF in the VTA causes excessive intake of a palatable HFD, but not standard chow, whereas reduced BDNF in the VMN results in indiscriminate hyperphagia of either HFD or chow (52, 262). Peripheral administration of a D1 receptor agonist normalizes the caloric intake of palatable HFD in BDNF mutant mice (52), indicating that BDNF synthesis in the VTA is possibly involved in dopamine secretion from neurons of the NAc and thus BDNF may play a role in hedonic reward. (Fig. 1/See link)
The hippocampus, which has long been associated with learning and memory (110, 236), has been implicated in having a potential involvement in energy balance (58). Part of what makes this an attractive hypothesis is that areas of the hippocampus, in particular, field CA1 neurons, project to the LH, Arc, PVN, DMN, and VMN, all important hypothalamic areas involved in feeding behavior (44). Several multisynaptic pathways have been identified that connect brain stem feeding control areas to the hippocampus (96, 180). Amnesic patients with hippocampus damage showed reduced sensitivity to interoceptive signals of hunger and satiety (102, 218). Compared with intact controls, rodents with selective lesions of the hippocampus exhibit increased appetitive responding for food (49, 59, 223). In a recent study by Davidson et al. (57), lesioning of the complete hippocampus resulted in increased food intake, body weight gain, appetitive behavior, and metabolic activity. When lesioning was restricted to the ventral pole, which projects to the lateral hypothalamus (44), animals had increased food intake and body weight (57). Functional magnetic resonance imaging (fMRI) identified the hippocampus and prefrontal cortex as the sites of greatest activation in obese people (273). DelParigi et al. (60) also noticed a decreased hippocampal blood flow in obese and formerly obese people after they consumed a liquid meal to satiation. These findings suggest that the hippocampus plays an important role in the regulation of energy metabolism. BDNF and TrkB are highly expressed in the hippocampus. Many studies have reported that exercise increases hippocampal BDNF expression (47, 95, 227) and that these increases are associated with enhanced cognition (26, 143, 195). Exactly what role, if any, hippocampal BDNF plays in energy metabolism is still unclear. Some evidence suggests that hippocampal BDNF might be related to factors affecting the memory of food, and, therefore, motivation to eat (84). A/J mice, who behaviorally model activity-induced anorexia and exhibit reduced food anticipatory activity, have significantly lower BDNF expression in the hippocampus during feeding times, compared with mice that have normal food anticipatory activity (84). Dietary restriction has no effect on hippocampal BDNF in A/J mice, whereas in mice without activity-induced anorexia, dietary restriction increases hippocampal BDNF (62, 84, 150), as well as increasing the full-length TrkB receptor (150). This is likely not directly due to altered glucose levels, as no changes in BDNF expression were observed in the hippocampus with intracerebroventricular glucose administration (262). A HFD decreases hippocampal BDNF (178, 204, 278–280), as does a HF-high sugar diet (118, 177, 239); however, the type of sugar matters as the combination of a HFD and dextrose decreased BDNF, whereas HFD and sucrose did not. In this study, rats gained similar amounts of weight, indicating the differences in BDNF expression were not directly related to body weight (118). Furthermore, a HFD does not always decrease hippocampal BDNF. In an animal model of early life trauma, in which rats were separated from dams for about 2 wk after birth, a HFD was associated with increased hippocampal BDNF (166). Yu et al. (292) observed that in animals prone to DIO, hippocampal BDNF was reduced in response to a HFD, whereas in obesity-resistant animals (DRO), or in pair-fed DIO animals, it was not. The authors speculate that the decreases in hippocampal BDNF signaling may correspond to weakened inhibitory control of HF food intake and promote obesity (292). Davidson et al. (58) proposed a “vicious circle” model: an unhealthy diet (such as HFD) reduces hippocampal BDNF, which causes hippocampal dysfunction (such as hypermnesia), which may result in impaired feeding behavior and overeating in an “obesigenic” environment, which further reduces hippocampal BDNF and damages hippocampal function, and continues the circle (58).
Peripheral Actions of BDNF
In addition to the brain, BDNF is expressed in many tissues important to the regulation of energy homeostasis, namely adipose tissue, skeletal and smooth muscle, and liver (42, 159, 182, 261). It is, therefore, important to consider that the effect of BDNF in these peripheral tissues might also contribute to the overall maintenance of energy balance.
BDNF mRNA and protein expression are increased in exercising skeletal muscle, an effect associated with the phosphorylation of AMPK and acetyl-coA carboxylase β (ACCβ), as well as increases in fatty acid oxidation (170). AMPK “senses” a high-energy state in muscle, and when activated, it phosphorylates the mitochondrial ACCβ, thereby inhibiting increases in malonyl-CoA levels (an action that ultimately leads to increased mitochondrial fatty acid transport and oxidation) (41, 172, 173). BDNF elevates malonyl-CoA independently of AMPK; however, the effect of BDNF on muscle cell fatty acid oxidation is AMPK dependent (170). Skeletal muscle has a high nutritive demand, and poor intramuscular fatty acid metabolism may contribute to obesity (106) and insulin resistance (107, 222). The finding that BDNF increases ACCβ phosphorylation and fatty acid oxidation may be one of the ways in which peripheral BDNF increases insulin sensitivity and weight loss. Muscle-derived BDNF is not released into circulation (170), and although exercise is known to increase circulating levels of BDNF (74), the source of this increase is unlikely from muscle cells (170).
In the liver, BDNF contributes to the development of hyperglycemia, hyperinsulinemia, elevated serum cholesterol, and triglycerides associated with eating a HFD (242). When fed a HFD, liver-specific BDNF-knockout mice have elevated levels of peroxisome proliferator-activated receptor alpha (PPARα) and fibroblast growth factor 21 (Fgf21) compared with wild-type mice (242). Fgf21 is a downstream target of PPARα, which is important for hepatic lipid oxidation, and insulin sensitivity (14, 125, 285). Fgf21 dose dependently reduces body weight and adiposity and reduces the expression of a variety of genes involved in fatty acid and triglyceride synthesis (285). The protection against HFD-induced hyperglycemia and hyperinsulinemia observed in liver-specific knockouts is inconsistent with previous research, which reported improvements in liver histology after subcutaneous BDNF treatment (288). Thus, the improved liver histology viewed in these cases is likely secondary to other peripheral effects of BDNF.
Mature BDNF (∼13.6 kDa), which is less than half the size of pro-BDNF (∼32 kDa), does not cross the blood-brain barrier (203), and therefore, studies reporting physiological effects of subcutaneous injections of BDNF may be reflective of peripheral actions. Subcutaneous BDNF enhances glucose utilization in muscle and BAT of db/db mice, but not normoglycemic animals (290). Additionally, BDNF restores levels of insulin-secreting granules in beta cells and maintains their histologic cellular organization in db/db mice, even though there is no TrkB receptor in pancreatic islets (286). Subcutaneous injections of BDNF reduced body weight, fat pad weight, and liver weight in db/db mice compared with pair-fed Troglitazone-treated db/db mice (288). A single injection of BDNF has a prolonged hypoglycemic effect, which is not attributable to reductions in food intake alone (289). Taken together, this indicates that BDNF may act peripherally to normalize blood glucose in hyperglycemic rodents, or in those no longer sensitive to leptin, possibly by enhancing muscle utilization or via the stimulation of BAT.
While both central and peripheral activation of the TrkB receptor reduces food intake and obesity in rodents, this effect is not conserved across all species. In monkeys, the response of peripheral TrkB activation is orexigenic and obesity promoting, while the central activation parallels the anorectic effects observed in mice (156). The involvement of BDNF in energy homeostasis in humans is difficult to study, and studies are usually limited to correlations between serum or plasma BDNF and body weight or adiposity. Serum BDNF levels are lower in human type-2 diabetic patients (82), AN (191, 219), and in extremely overweight children (66), but not in bulimia nervosa (219), recovered AN (191), or healthy controls (191, 219). Thus, the relationship between serum BDNF and BMI is not clear. In one study, there was a significant positive relationship between the two (219), and in another, serum BDNF negatively correlated with both BMI and body fat (66). It is likely that serum BDNF reflects the amount stored in platelets, which is released during the clotting process (81, 174). This level is not acutely altered by food intake (66). Plasma BDNF, on the other hand, likely has many sources (34, 85, 124, 189). Plasma BDNF is higher in obese women, but these levels are significantly dropped after bariatric surgery (175). Conversely, Mercader et al. (174) observed no correlation between plasma BDNF and BMI in clinical subgroups of eating disorder patients. Krabbe et al. (141) report that plasma levels of BDNF are decreased in human type 2 diabetics, independent of obesity. By sampling from the internal jugular vein and comparing plasma BDNF from arterial and venous samples, Krabbe et al. (141) demonstrated that cerebral output of BDNF is reflected in the circulation. They observed that plasma BDNF levels are directly, inversely related to fasting plasma glucose levels and that BDNF output from the brain is inhibited when blood glucose levels are elevated.
BDNF and Neuronal Plasticity
Neuronal plasticity is defined as an experience-dependent change in synaptic strength (31). A well-studied example of this is long-term potentiation (LTP), which has particularly been associated with neurons of the hippocampus. LTP is the activity-dependent strengthening of a synapse, which is typically induced by high-frequency stimulation of excitatory input. Many studies have identified the importance of BDNF in the development of LTP (76, 117, 139, 205, 283, 293) and also in the development from LTP to long-term memories (6, 249, 250). BDNF is secreted in response to a high-frequency stimulation and is dependent on Ca2+ influx through voltage-gated Ca2+ channels or NMDA receptors (2, 16, 100). BDNF can bind TrkB receptors on either side of the synapse (61). In the hippocampus, it functions to facilitate the presynaptic release of excitatory neurotransmitters (33), as well as postsynaptic AMPA receptor insertion (154, 155) and dendritic spine maintenance (45, 55, 240). In contrast to LTP promotion by mature BDNF, recent evidence indicates a potential role for pro-BDNF in facilitating long-term depression (LTD) through activation of p75NTR receptor (69, 160, 277).
Factors related to energy balance have been described to affect hippocampal LTP. When fed a diet high in both fat and sucrose (HFS), rats were less capable in a spatial learning capacity task than rats fed standard chow (278). There was no sign of neuronal degeneration in the HFS fed rats. Additionally, rats with the lowest hippocampal BDNF, who also had lower levels of CREB and the vesicle-associated synapsin I (114), had the lowest learning capacity (278). A HFD also has a negative impact on hippocampal LTP and plasticity (210), and some studies suggest that this impairment is related to BDNF. A diet high in saturated fat and refined sugars significantly reduced hippocampal BDNF compared with rats fed standard chow, which was accompanied by impaired LTP and poor performance on the Morris water maze, a spatial learning task (177). A diet high in saturated fat reduced BDNF in both the ventral hippocampus and medial prefrontal cortex with associated impairment in reversal learning (118). Taken together, these data suggest that a HFD may affect synaptic plasticity in rat hippocampal neurons. Conversely, caloric restriction (77) and exercise (238) enhance LTP and are associated with increased BDNF (150, 194). Hippocampal BDNF levels are increased with running (53), and running protects against stress-related downregulation of BDNF (1).
Although the role of BDNF in neuronal plasticity has been well studied in the hippocampus, the possibility that BDNF contributes to plasticity of hypothalamic neurons related to energy balance is less well studied. The molecular mechanisms by which BDNF acts in the hypothalamus to affect energy homeostasis have not been characterized. However, an example of BDNF-dependent hypothalamic plasticity has been recently described in thermal sensation and temperature control development (120). Several plasticity-related genes (including BDNF) were reported to be differentially expressed in high and low body weight chickens during development, suggesting different inherent capacities of these animals to adapt appetite circuitry (115). We have recently observed reductions in feeding behavior in rats after BDNF administration into the VMN and PVN. We observed this feeding inhibition between 4 and 24 and 24 and 48 h postinjection in both sites; however, we did not observe an effect of BDNF during the first 4 h (267, 269). The possibility that BDNF acts to suppress food intake via a plasticity-related mechanism in the hypothalamus is thus worth investigating. Neuronal activity in the PVN is modulated by excitatory glutamatergic signaling, as well as other excitatory and inhibitory neurotransmitters. Recently, NMDA receptor-dependent plasticity in the PVN has been described to reduce excitatory glutamatergic signaling in spontaneously hypertensive rats (153), and the NMDA receptor-dependent plasticity was induced environmentally by chronic intermittent hypoxia (50). We have observed that a single injection of BDNF in the PVN dose dependently suppresses feeding for up to 48 h (267). The duration of the effect of a single injection of BDNF into the PVN, as well as characterization of plasticity mechanisms in the PVN, makes neural plasticity seem a candidate mechanism for the observed anorectic response.
BDNF and Neurogenesis
The adult central nervous system contains neuronal stem cells capable of generating new neurons (8), and several progenitor cells in certain brain regions have been identified (89, 243). Neurogenesis has been well studied in the subventricular zone of the lateral ventricles, and the subgranular zone of the hippocampal formation (83). In the hippocampus, Nakatomi et al. (190) observed the regeneration of CA1 pyramidal neurons of the adult rodent brain following ischemic degeneration, which was facilitated by the intracerebroventricular infusion of two different growth factors (FGF-2 and EGF). Novel to this study was the observation that growth factors signal the recruitment of progenitors from areas near the hippocampus to facilitate neurogenesis in areas where no progenitors are available. Thus, extensive neurogenesis is possible in brain sites not containing stem cells (190). Mitosis of progenitor cells lasts about 24 h for the rat (36), and 14 h in the mouse (165), after which it takes 3–4 wk for new neurons to mature and fully integrate into the circuitry (263). Thus, any effect of BDNF on neurogenesis would likely be observable more over the long term.
Energy balance has been described to affect BDNF and neurogenesis in the hippocampus. Dietary restriction (DR) increases neurogenesis in the adult mouse hippocampus, an effect associated with elevated BDNF expression, yet this effect is absent in BDNF heterozygous mice (150). However, DR normalizes BDNF in the hippocampus, striatum, and cerebral cortex of Huntington mutant mice, in whom BDNF expression is decreased, such that it is equivalent to the expression in wild-type ad libitum-fed mice, and reverses obesity, abnormal locomotor activity, and hyperphagia (62). Conversely, a diet high in saturated fat decreases BDNF and compromises cognitive performance (177). In mice susceptible to DIO, a HFD decreases BDNF and TrkB mRNA in the hippocampus compared with mice resistant to DIO (292). Exposure to enriched environment and exercise elevates BDNF in the hippocampus and improves learning (200). Conversely, it has been observed in rats that the consumption of a HFD, particularly one high in saturated fat, decreases BDNF and adult hippocampal neurogenesis (204). Park et al. (204) observed that the prevention of neurogenesis was associated with increased levels of malondialdehyde (MDA) in the hippocampus, which is an indicator of lipid peroxidation, and they found a direct effect of MDA administration on inhibiting neurogenesis. A HFD fed to dams is associated with obesity and hyperlipidemia in offspring. These offspring have impaired hippocampal neurogenesis, which parallels with the degree of neuronal impairment observed when treating cells in vitro with MDA (253). Tozuka et al. (252) observed that a maternal HF diet caused reductions in hippocampal BDNF and impaired dendritic arborization of hippocampal neurons in pups.
Neurogenesis in the hypothalamus is less well characterized. The effect of a maternal HFD on hypothalamic neurogenesis in pups has recently been investigated, where HFD increased neurogenesis of neurons expressing orexigenic peptides galanin, enkephalin, and dynorphin in the PVN and orexin and melanin-concentrating hormone in the perifornical lateral hypothalamus (46). These hypothalamic changes were associated with increased body weight, leptin, insulin, dietary fat preference, triglycerides, and galanin expression in the PVN (46). Thus, it appears that energy balance affects neurogenesis; however, a compelling question that remains to be addressed is whether neurogenesis also affects energy balance.
Neuronal stem cells have been identified in the hypothalamus, and hypothalamic neurogenesis has been described to occur at a low rate (134). Ciliary neurotrophic factor (CNTF) is similar to BDNF in that it promotes neuronal survival (146, 228) and the maintenance of neuronal stem cells (230), and it activates signaling cascades in the hypothalamus involved in feeding and energy homeostasis (30, 144). Kokoeva et al. (135) observed an increase in neurogenesis in several hypothalamic feeding centers, particularly the median eminence-Arc, in CNTF-treated mice fed a HFD compared with non-CNTF-treated controls. Furthermore, the CNTF-treated mice were resistant to weight gain on the HFD, which persisted for more than a week after the cessation of CNTF treatment. The antimitotic agent cytosine-B-d-arabinofuranoside prevented the inhibition of weight gain after the cessation of CNTF treatment, but not during, indicating neurogenesis is partially responsible for the sustained antiobesigenic effect of CNTF (135). Additionally, the new neurons expressed proteins involved in energy balance such as POMC, neuropeptide Y, and phosphorylated STAT3, an indicator of leptin signaling (135). That CNTF is capable of inducing hypothalamic neurogenesis and that this neurogenesis has an antiobesigenic effect is a demonstration that hypothalamic neurogenesis can, in fact, play an important role in regulating energy metabolism.
Using BrdU, Pencea et al. (208) were the first to observe that BDNF is capable of inducing hypothalamic neurogenesis after continuous intracerebroventricular administration of BDNF for 12 days. Neurogenesis triggered by BDNF correlates with levels of TrkB expression, but the TrkB receptor is not directly integrated into new neurons. Thus, hypothalamic regions that contained high levels of the TrkB receptor (e.g., the PVN) had greater amounts of BrdU cells than areas with less TrkB expression, even if those areas were located closer to the site of BDNF infusion (208). Recently, Kumar et al. (142) used BrdU to measure neurogenesis in response to DR during a model of cytotoxic injury. Alternate-day DR was associated with increased BDNF and neurogenesis in several brain regions, including the median eminence-Arc of the hypothalamus (142). The type of new neurons generated in the Arc with dietary restriction remains to be elucidated, as well as precisely how they might affect energy balance.
Neuroprotection and Survival
During development, neurotrophic factors are critical to survival because they inhibit apoptosis of developing neurons (282). BDNF is neuroprotective in the hippocampus, particularly against ischemic damage (23, 133, 147). Neuronal apoptosis, or programmed cell death, is characterized by the activation of caspases, specifically caspase 9, which acts upstream of caspase 3 (246). BDNF reduces glutamate-induced apoptotic cell death upstream of the activation of caspase-3-like enzymes and increases expression of the antiapoptotic protein B-cell lymphoma 2 (Bcl-2) (5). Additional studies have shown BDNF induces increases in Bcl-2 (12, 129, 209). It is important to note, however, that BDNF is not always protective. To some cultured neurons of the hippocampus and cerebrocortex BDNF is toxic (78). While activation of TrkB receptors increases LTP and neuronal survival, activation of p75NTR can lead to apoptosis (20) and LTD (277).
Oxidative stress (112) often leads to a loss of cell function, apoptosis, or necrosis [reviewed by Azad (13)]. BDNF has been implicated in being protective against oxidative stress by preventing the accumulation of peroxides and increasing antioxidant enzymes in hippocampal neurons (171). An abundant oxidative stress marker is 4-hydroxynonenal (HNE), which is generated through peroxidation of omega 6-polyunsaturated fatty acids (25, 68). HNE is highly diffusible and may contribute to oxidative stress far from a site of injury (68). Local application of BDNF on the dorsal hemisected spinal cord in rats resulted in reduced lipid peroxidation, as shown by decreased HNE-immunoreactive staining (113). This effect was observed within 48 h of BDNF application and was associated with a reduction in activated microglial cells, which may have contributed to decreased oxidative stress (113).
In brain regions exhibiting neuronal loss, as is associated with Huntington’s disease (73) and Alzheimer’s disease (104), BDNF levels are low, which contributes to the neuronal degeneration associated with these diseases (104). Tg2576 mice are a widely used model of Alzheimer’s disease. They develop amyloid plaques and have declining cognitive function at 6 mo old (274). Kohjima et al. (132) observed Tg2576 mice fed a HFD developed obesity and insulin resistance due to hyperphagia, and that the abnormal feeding behavior was associated with increased amyloid plaque formation and decreased hypothalamic BDNF. Oxidative stress is present early in the pathogenesis of Alzheimer’s disease (121, 215). Vitamin E is a known antioxidant, which can prevent the development of oxidative stress. Vitamin E supplementation ameliorates HFD-induced reductions in BDNF, suggesting that BDNF levels may be altered in response to oxidative stress (279). Several studies have identified that there is a relationship between a HFD and impairments in cognitive performance, particularly in rats fed a diet high in saturated fats (92, 93, 275, 276). In addition to affecting cognitive function, a HFD induces apoptosis of hypothalamic neurons, such as POMC neurons (179). Thus, the neuroprotective effect of BDNF may have a role in the central regulation of energy metabolism; however, further studies are needed to define the role of BDNF in hypothalamic neuroprotection and how it relates to energy balance.
Factors Affecting Expression of Hypothalamic BDNF
Hypothalamic BDNF and TrkB content are affected by age. In rats raised in standard laboratory conditions, the mature form of hypothalamic BDNF peaks at about 1 wk postnatally in rats, remains elevated for the first month of life, and declines with age (232). Additionally, declining levels of TrkB receptor begin in rats at around 2 mo, with extreme reductions observed by 22 mo (232). It is possible that environmental factors could prevent age-related decline in BDNF. Recently, Cao et al. (40) reported BDNF expression is increased in mice housed in an enriched environment. The environmental enrichment included a large open space, toys, and a running wheel. Earlier studies have shown that in this type of complex environment, animals exhibit improved learning and memory and increased neurogenesis (38, 63). Additionally, mice living in an enriched environment remain leaner than mice in standard housing, an effect associated with consistently elevated BDNF expression in the Arc (40). Elevated BDNF in the hypothalamus associated with environmental enrichment decreased leptin levels, which inhibited cancer tumor growth in several models of cancer. The metabolic profiles of mice in an enriched environment were mimicked when BDNF was overexpressed in the hypothalamus of animals living in standard housing using a recombinant adeno-associated virus vector (40).
Four weeks of running increased slightly, but not significantly, expression of BDNF in the Arc, whereas an enriched environment significantly increased Arc BDNF expression after only 2 wk. While both the running mice and the enriched environment mice had similar decreases in body weight, the groups differed in metabolic gene expression, and only the enriched environment group had a corresponding decrease in leptin and tumor growth (40). The study by Cao et al. (40) opens the door to many questions about the relationship between environment and metabolism, more specifically, how does environmental enrichment impact food intake and energy expenditure to contribute to obesity resistance? What constitutes an enriched environment? In another model of environmental enrichment, Angelucci et al. (9) report that music increased BDNF in the mouse hypothalamus.
Acute immobilization stress causes rapid increases in hypothalamic BDNF (186, 213). These increases are accompanied by decreased body weight and increased locomotor activity, as well as activation of the HPA (186). Conversely, neonatal stress induced by separating rat pups from their dams for 180 min per day is associated with elevations in hippocampal BDNF protein expression, which is likely exerting a protective effect on existing neurons there (94).
Kohjima et al. (132) report that 16 wk of a HFD, which was sufficient to induce insulin resistance, obesity, and amyloid plaques in Tg2567 mice, decreased hypothalamic BDNF and led to elevated feeding behavior. In the VMN specifically, BDNF, but not TrkB mRNA, is lower in DIO mice compared with DRO, suggesting that the DIO phenotype may, in part, be mediated through low BDNF expression in this important feeding center (292). No differences in BDNF expression were observed between the two phenotypes in the Arc, the DMN or the PVN (292). However, Sprague-Dawley rats fed either a high-energy diet (HE) or high energy plus Ensure liquid diet (HE + E) expressed increased TrkB receptor, but not BDNF, in the VMN (10). Both groups decreased food intake and body weight when switched from their respective diets to standard chow, and in both cases, this was accompanied by decreased BDNF. The HE, but not the HE+E group also had decreased TrkB expression upon switching to standard chow (10). Differences in TrkB mRNA are difficult to interpret, because not all forms of TrkB are active and some might serve to inhibit TrkB signaling. We have recently demonstrated that chronic administration of BDNF into the PVN reduced HFD-induced obesity and that animals with greater body fat were more responsive to the effect of added BDNF (270). This is in line with recent evidence that suggests that animals resistant to HFD-induced obesity maintain higher basal levels of BDNF and TrkB in the hypothalamic VMN compared with animals susceptible to DIO (292). Moreover, VMN BDNF levels were further reduced in DIO mice on a HF diet, and they were negatively correlated with plasma glucose and positively correlated with plasma adiponectin (292). Adiponectin is an adipokine associated with increased insulin sensitivity and reduced appetite (27). This suggests that BDNF and TrkB expression in the hypothalamus might play a role in determining susceptibility to obesity (292). The antiobesity effect of BDNF in the PVN was related to a significant reduction in energy intake (270), and previously, we have observed that BDNF in the PVN increases energy expenditure and resting metabolic rate (266).