Brain-derived neurotrophic factor mediates an excitoprotective effect of dietary restriction in mice


Address correspondence and reprint requests to Mark P. Mattson, Laboratory of Neurosciences, National Institute on Aging Gerontology Research Center, 5600 Nathan Shock Drive, Baltimore, MD 21224, USA. E-mail:


Dietary restriction (DR; reduced calorie intake) increases the lifespan of rodents and increases their resistance to cancer, diabetes and other age-related diseases. DR also exerts beneficial effects on the brain including enhanced learning and memory and increased resistance of neurons to excitotoxic, oxidative and metabolic insults. The mechanisms underlying the effects of DR on neuronal plasticity and survival are unknown. In the present study we show that levels of brain-derived neurotrophic factor (BDNF) are significantly increased in the hippocampus, cerebral cortex and striatum of mice maintained on an alternate day feeding DR regimen compared to animals fed ad libitum. Damage to hippocampal neurons induced by the excitotoxin kainic acid was significantly reduced in mice maintained on DR, and this neuroprotective effect was attenuated by intraventricular administration of a BDNF-blocking antibody. Our findings show that simply reducing food intake results in increased levels of BDNF in brain cells, and suggest that the resulting activation of BDNF signaling pathways plays a key role in the neuroprotective effect of DR. These results bolster accumulating evidence that DR may be an effective approach for increasing the resistance of the brain to damage and enhancing brain neuronal plasticity.

Abbreviations used

ad libitum


brain-derived neurotrophic factor


dietary restriction


enzyme-linked immunosorbent assay


kainic acid


nerve growth factor


Tris-buffered saline with Tween-20

Reducing the calorie intake of rats and mice, while maintaining nutrition, can greatly increase their lifespans (Weindruch et al. 1986; Sohal and Weindruch 1996), and this also appears to be the case in monkeys (Lane et al. 1999) and humans (Roth et al. 1999). Several such dietary restriction (DR) feeding regimens can extend lifespan, with the two most commonly used protocols being intermittent (every-other-day) feeding and paired feeding that employs food pellets containing 30–40% less calories than pellets in the control diet (Goodrick et al. 1983a; Talan and Ingram 1985; Sohal et al. 1994). DR can suppress the development of age-related cancers, immune and neuroendocrine alterations, and motor dysfunction (Kritchevsky and Klurfeld 1986; Ingram et al. 1987; Wachsman 1996). DR has also been shown to exert beneficial effects on the brain including improved performance on learning and memory tasks (Idrobo et al. 1987; Ingram et al. 1987; Stewart et al. 1989); suppression of age-related increases in levels of glial fibrillary acidic protein (Finch and Morgan 1997); and increased resistance of neurons to age-related and disease-specific insults (Bruce-Keller et al. 1999; Duan and Mattson 1999; Yu and Mattson 1999). Moreover, epidemiological data suggest that individuals with a low calorie intake are at reduced risk for Parkinson's and Alzheimer's diseases (Logroscino et al. 1996; Mayeux et al. 1999). Available data suggest two possible mechanisms that might explain the beneficial effects of DR on aging and disease. One possibility is that, because less glucose is available to mitochondria during DR, fewer oxyradicals are produced over time and therefore less free radical-mediated damage to cells occurs (Dubey et al. 1996; Sohal and Weindruch 1996). The second possibility is that the mild metabolic stress associated with DR induces cells to produce ‘stress proteins’, such as heat-shock protein 70, that increase cellular resistance to disease processes (Aly et al. 1994; Ehrenfried et al. 1996; Heydari et al. 1996; Lee et al. 1999; Yu and Mattson 1999).

Neurotrophic factors include a diverse array of growth factors and cytokines that promote survival and plasticity (neurite outgrowth, synaptogenesis and changes in synaptic efficacy) of neurons (see Shen et al. 1997; Jankowsky and Patterson 1999; McKay et al. 1999 for review). Several such neurotrophic factors, including nerve growth factor (NGF), brain-derived neurotrophic factor (BDNF) and basic fibroblast growth factor, can protect neurons against death in experimental animal and cell culture models of both acute (stroke, trauma and seizures) and chronic (Alzheimer's and Parkinson's diseases) neurodegenerative conditions (Cheng and Mattson 1994; Tsukahara et al. 1995; Ren and Finklestein 1997; Guo et al. 1999; Larsson et al. 1999; Endres et al. 2000). Various types of injury and cellular stress can induce the expression of neurotrophic factors in the brain (see Mattson and Furukawa 1996 for review). BDNF production by brain cells is increased by severe cellular stresses that can cause neuronal degeneration including cerebral ischemia, hypoglycemia, epileptic seizures and trauma (Ballarin et al. 1991; Lindvall et al. 1992; Hicks et al. 1997; Lee et al. 1997; Rudge et al. 1998). Neurotrophic factor production is also induced by lower levels of cellular stress that occur under physiological conditions. For example, levels of BDNF are increased in neurons in response to depolarization and during learning and memory (Patterson et al. 1992; Ghosh et al. 1994; Johnston et al. 1999; Mu et al. 1999). We now report results of studies which reveal that levels of BDNF are increased in the hippocampus of mice maintained on a DR feeding regimen, and that such increased neurotrophic factor signaling plays an essential role in the neuroprotective effect of DR.

Materials and methods

Animals and diets

Sixty-four adult (2-month-old) male C57BL/6 mice obtained from the National Cancer Institute were maintained under temperature- and light-controlled conditions (20–23°C, 12-h light/12-h dark cycle). For each experiment, mice were divided into two groups (6–8 mice/group), an ad libitum (AL) group which had continual access to food, and a DR group which was provided food on alternate days. Mice were maintained on the diets for 3 months, at which time they were subjected to experimental treatments and/or sacrificed without further treatment (for analyzes of BDNF mRNA and protein levels). Previous studies have shown that rats and mice maintained on such an alternate day feeding schedule will consume 30–40% less calories over time compared to animals fed AL (Goodrick et al. 1983a; Talan and Ingram 1985). Food was withdrawn from mice in both groups 6 h prior to experimental treatment or euthanasia.

Experimental treatments and quantification of neuron survival

BDNF antibody (1 µg) and saline (vehicle) were infused into the right lateral ventricle in a volume of 5 µL administered during a 10-min period at 2 h prior to kainic acid (KA) administration. The BDNF antibody was a rabbit polyclonal antibody purchased from Santa Cruz Biotechnology (Santa Cruz, CA, USA). KA was administered via stereotaxic injection into the dorsal aspect of the right hippocampus of mice using methods described previously (Bruce et al. 1996; Furukawa et al. 1997). Briefly, mice were anesthetized and KA (0.2 µg in a volume of 0.5 µL) was injected unilaterally into the dorsal hippocampus (D/V, − 1.8; M/L, + 2.4; A/p, − 2.0). Mice were sacrificed 24 h after KA administration and perfused transcardially with saline followed by cold phosphate-buffered 4% paraformaldehyde. Coronal brain sections were cut at 30 µm on a freezing microtome and stained with cresyl violet. Nissl-positive surviving neurons were counted in three 40 × fields in regions CA1, CA3 and hilus of both the ipsilateral and contralateral hippocampus in six adjacent sections located between 0.2 and 0.6 mm rostral to the injection site on the ipsilateral side and at an equivalent rostro-caudal window on the contralateral side. The mean number of cells/field was determined such that the value obtained for each hippocampus represents an average number of neurons counted/40 × field. Neurons were scored as undamaged if they were Nissl-positive with a round to oval shaped cell body that exhibited no evidence of cell shrinkage. All slides were coded and the analyzes were done by an investigator (W.D.) blinded as to dietary history and experimental treatment of the mice from which the sections were taken. Comparisons of numbers of undamaged neurons in hippocampal regions among treatment groups were made using anova followed by Scheffe tests for pairwise comparisons.

ELISA analysis of BDNF and NGF protein levels

Hippocampal, cortical and striatal tissues were homogenized in lysis buffer (137 mm NaCl, 20 mm Tris, 1% NP-40 detergent, 10% glycerol, 1 mm phenylmethylsulfonylflouride, 10 µg/mL aprotinin, 1 µg/mL leupeptin, and 0.5 mm sodium orthovanadate; pH 7.2) at 4°C. Homogenates were centrifuged at 2000 g for 20 min (4°C), and supernatants were used for ELISA analysis. BDNF and NGF protein levels were quantified using commerically available kits (Promega) according to the manufacturer's protocol. Briefly, samples were processed by acidification and subsequent neutralization. Ninety-six-well plates were coated with either mouse monoclonal BDNF or rabbit polyclonal NGF antibodies, incubated in the presence of ‘block and sample’ buffer and washed in TBST (Tris-buffered saline with Tween-20 ). Samples (300 µg protein for the BDNF ELISA and 500 µg protein for the NGF ELISA) were added to triplicate wells in each plate, and serial dilutions of BDNF or NGF standard (0–500 pg/mL) were added to duplicate wells in each plate in order to generate a standard curve. Plates were incubated for 2 h, washed five times in TBST, and incubated for 2 h in a solution containing either rabbit polyclonal BDNF or mouse monoclonal NGF antibodies. Wells were washed five times with TBST, and a hydrogen peroxide solution was added together with a peroxidase substrate, and plates were incubated for 10 min. The intra- and interassay variabilities for the BDNF ELISA were 0.55 pg and 1.24 pg, respectively, at a sample protein concentration of 300 µg. The intra- and interassay variabilities for the NGF ELISA were 14.0 pg and 17.9 pg, respectively, at a sample protein concentration of 500 µg. Reactions were stopped by adding 100 µL 1 m phosphoric acid, and absorbance was measured at 450 nm using a plate reader. The concentrations of BDNF and NGF protein in each sample were determined in triplicate, and the average of the three values was used as the value for that mouse for the statistical analysis. Values were expressed as pg BDNF or NGF/mg protein.


Dietary restriction increases BDNF production in several brain regions of mice

Groups of young adult male mice were maintained either on an alternate day DR feeding schedule or were fed AL for 3 months at which time the average body weight of the DR groups was significantly reduced to 84% of the body weight of the AL-fed control group (Fig. 1a). Levels of BDNF protein were quantified by ELISA analysis in tissue homogenates from hippocampus, striatum and cerebral cortex from eight DR and eight AL-fed mice. Levels of BDNF were significantly increased by approximately 50% in hippocampus, 60% in striatum and 40% in cerebral cortex of mice maintained on DR compared to AL-fed control mice (Fig. 1b). In contrast to the increased levels of BDNF in brain tissues from mice maintained on DR, levels of NGF were not different in hippocampus, striatum or cerebral cortex of DR and AL-fed mice (Fig. 1c).

Figure 1.

Levels of BDNF protein are increased in several brain regions of mice maintained on a DR regimen. (a) Body weights of mice after 3 months of maintenance on either an AL or DR diets. Values are the mean and sd (n = 8 mice/group). (b) and (c) Levels of BDNF (b) and NGF (c) protein in tissue homogenates of hippocampus, striatum and cerebral cortex from AL-fed control mice (AL) and mice that had been maintained on DR for 3 months were quantified by ELISA analysis. Values are the mean and SD of determinations made in samples from eight mice. *p < 0.05, **p < 0.01 compared to corresponding AL value (anova with Scheffe posthoc tests).

Dietary restriction increases resistance of hippocampal neurons to excitotoxic injury: involvement of BDNF

It was recently reported that rats maintained on an alternate day feeding DR regimen exhibit increased resistance of hippocampal neurons to seizure-induced injury (Bruce-Keller et al. 1999). In light of previous studies documenting neuroprotective effects of BDNF in cell culture and in vivo, we therefore tested the hypothesis that increased levels of BDNF in mice maintained on DR might contribute to the neuroprotective effect of DR. A BDNF-blocking antibody or saline (control) was infused into the lateral ventricles of DR and AL-fed mice 2 h prior to unilateral intrahippocampal injection of KA. In preliminary experiments we showed that BDNF-protected cultured hippocampal neurons against glutamate toxicity and that the BDNF antibody completely abolished the neuroprotective effect of BDNF (data not shown). Mice were killed 24 h following KA administration, and coronal brain sections were stained with cresyl violet. Neurons that had not died during the 24-h period were scored as either undamaged or damaged according to the criteria described in the Methods section. Extensive damage to CA3, CA1 and hilar neurons occurred in the ipsilateral (KA-injected) hippocampus of AD-fed control mice such that only 10–25% of the neurons remained and were undamaged (Figs 2 and 3). Mice maintained on DR exhibited highly significant, two- to three-fold increases in numbers of undamaged neurons present in each region of ipsilateral hippocampus (Figs 2 and 3). In AL-fed mice the BDNF-blocking antibody had no significant effect on KA-induced neuronal damage in any of the three regions of hippocampus. In contrast, KA-induced neuronal loss was significantly enhanced in regions CA3 and hilus of BDNF antibody-treated DR mice (Figs 2 and 3). The BDNF antibody did not effect the vulnerability of CA1 neurons to KA-induced damage in DR mice. Mice maintained on DR exhibited a significant increase in the survival of CA3 and CA1 neurons in the contralateral hippocampus. The BDNF antibody significantly decreased neuronal survival in region CA3, but not in region CA1, of DR mice. There was no significant loss of hilar neurons in the contralateral hippocampus of KA-treated AL-fed or DR mice (Fig. 3).

Figure 2.

Micrographs show Nissl-positive cells in region CA3 of hippocampi from representative AL and DR mice in the indicated experimental groups. After 3 months on either AL or DR feeding regimens, mice were administered either saline or BDNF antibody via intraventricular infusion beginning 2 h prior to intrahippocampal KA administration. Mice were sacrificed 24 h after KA administration, and coronal brain sections were stained with cresyl violet. Note that KA-induced damage to CA3 neurons is decreased in DR mice compared to AL mice, and that the BDNF antibody attenuated the neuroprotective effect of DR.

Figure 3.

BDNF-blocking antibody attenuates the neuroprotective effect of DR. Numbers of surviving neurons in regions CA3 (a), CA1 (b) and hilus (c) of the ipsilateral (KA-injected) hippocampus and in regions CA3 (d), CA1 (e) and hilus (f) of the contralateral (uninjected) hippocampus were quantified in brain sections from AL and DR mice subjected to the indicated treatments. Values are the mean and SD of measurements made in six mice/group. *p < 0.01 compared to corresponding value for AL-fed mice. ##p < 0.01 compared to value for DR mice administered KA alone. anova with Scheffe posthoc tests.


The increase in BDNF levels induced by DR was associated with increased resistance of hippocampal neurons to excitotoxic injury, and the neuroprotective effect of DR was significantly reduced in mice administered a BDNF-blocking antibody demonstrating a requirement for BDNF, and presumably signaling via its high affinity receptor trkB (Black 1999), in the neuroprotective action of DR. NGF levels were not different in hippocampus, striatum and cerebral cortex of DR and AL-fed mice, indicating that the effect of DR on BDNF levels is selective, and also suggesting that a change in NGF signaling does not mediate the neuroprotective effect of DR. Recent studies have provided evidence that endogneous BDNF can increase resistance of neurons to excitotoxic and ischemic injury (Larsson et al. 1999; Endres et al. 2000). On the other hand, infusion of BDNF immediately prior to and after KA administration although reducing damage to CA1 neurons, exacerbated loss of CA3 neurons (Rudge et al. 1998). The endangering effect of BDNF on CA3 neurons in the latter study might be caused by an enhancement of calcium influx by BDNF resulting in excitotoxic necrosis of CA3 neurons, whereas CA1 neurons may die mainly by apoptosis in this model. In support of the latter possibility, it was recently shown that BDNF protects cultured hippocampal neurons against glutamate-induced apoptosis, but exacerbates glutamate-induced necrosis (Glazner and Mattson 2000). In the present study we observed no effect of BDNF antibody on vulnerability of hippocampal neurons to KA-induced damage in AL-fed mice suggesting that basal levels of BDNF may not be sufficient to protect the neurons against excitotoxic damage. Our findings therefore suggest that the tonic elevation of BDNF levels induced by DR may be particularly effective in increasing resistance of hippocampal neurons to excitotoxic injury, possibly as the result of effects on gene expression (Mattson et al. 1995).

Whereas DR protected each of the three populations of hippocampal neurons (CA3, CA1 and hilar neurons) against KA-induced injury, the BDNF-blocking antibody significantly reduced the neuroprotective effect of DR in regions CA3 and hilus, but not in CA1. There are several possible explanations for the latter result. One possibility is that there is differential expression of the BDNF receptor trkB among CA3, CA1 and hilar neurons. Studies in which the cellular localization of trkB mRNA and protein were examined suggest that trkB levels may indeed be higher in CA3 and hilar neurons than in CA1 neurons (Tokuyama et al. 1998). Ultrastructural immunolocalization analyzes have shown that trkB may be differentially localized in membranes of axonal and dendritic terminals of pyramidal neurons and granule cells (Drake et al. 1999). It will therefore be of considerable interest to determine whether DR affects trkB expression and/or localization. It should also be considered that DR may cause a differential increase in BDNF levels among populations of hippocampal neurons, because BDNF levels are higher in CA3 neurons than in CA1 neurons under normal feeding conditions (Tokuyama et al. 1998). Finally, there may be neuroprotective signaling mechanisms activated by DR that do not involve BDNF, and such mechanisms may promote survival of CA1 neurons under conditions where BDNF signaling is blocked. Indeed, recent findings suggest that levels of the neuroprotective proteins heat-shock protein-70 and glucose-regulated protein-78 are increased in neurons in several different brain regions of rats and mice maintained on a DR feeding regimen (Duan and Mattson 1999; Lee et al. 1999; Yu and Mattson 1999).

Although the present study did not establish the mechanism whereby DR induces BDNF production, a likely possibility is that the mild metabolic stress associated with DR upregulates expression of BDNF. DR is known to increase production of stress proteins in neurons of rats and mice (Duan and Mattson 1999; Lee et al. 1999) consistent with the presence of cellular stress. Many previous studies have shown that cellular stress, including metabolic stress, can induce BDNF production in neurons in the hippocampus and other brain regions of rodents (Ballarin et al. 1991; Lindvall et al. 1992). BDNF levels were increased to similar amounts in cerebral cortex, striatum and hippocampus indicating that this effect of DR on BDNF levels is widespread, consistent with a cellular response to a general change in brain metabolism that might be expected to result from a chronic reduction in food intake. It is conceivable that the increase in BDNF levels in the brains of DR mice might be a consequence of increased motor activity, because increased motor activity can increase BDNF production in the brains of rats (Neeper et al. 1996). However, this may not be the case because an intermittent feeding DR regimen similar to that employed in the present study does not increase, but rather decreases, motor activity in young adult rats (Goodrick et al. 1983b). Our data suggest that the increase in BDNF levels induced by DR plays an important role in protecting hippocampal neurons against excitotoxic injury. Although we did not determine whether the increases in BDNF levels in cerebral cortex and striatum also mediate the previously demonstrated neuroprotective effects of DR on neurons in those brain regions (Bruce-Keller et al. 1999; Duan and Mattson 1999; Yu and Mattson 1999), it has been shown that BDNF can promote survival of cortical (Cheng and Mattson 1994) and striatal (Bemelmans et al. 1999) neurons. The mechanism whereby BDNF protects neurons against excitotoxic injury in vivo is not known, but cell culture studies have shown that BDNF can suppress oxyradical production and stabilize cellular calcium homeostasis in neurons (Cheng and Mattson 1994; Mattson et al. 1995). The signal transduction pathways that mediate the neuroprotective action of BDNF may involve mitogen-activated protein kinase and phosphatidyl inositol-3 kinase (Hetman et al. 1999) which may induce expression of antiapoptotic proteins such as bcl-2 (Sanna et al. 1998).

Our discovery that diet can affect levels of a neurotrophic factor in the brain has profound implications for dietary modification of the many properties of neurons regulated by neurotrophic factors including synaptic plasticity and cell survival. BDNF production is stimulated in hippocampus during conditions associated with enhanced learning and memory including ‘enriched environments’ (Young et al. 1999), physical exercise (Russo-Neustadt et al. 1999) and stimulation of hippocampal circuits at frequencies that induce long-term potentiation of synaptic transmission (Bramham et al. 1992). An important role for BDNF in regulating synaptic plasticity is suggested by studies showing that exogenous BDNF can enhance synaptic transmission in the hippocampus (Levine et al. 1995), and that BDNF knockout mice (Korte et al. 1998) and animals administered BDNF antibodies (Johnston et al. 1999) exhibit deficits in learning and memory. In light of our data and previous studies showing that rats and mice maintained on DR exhibit enhanced performance on learning and memory tasks (Idrobo et al. 1987; Ingram et al. 1987; Stewart et al. 1989), it therefore seems likely that increased BDNF signaling may play a role in the synaptic plasticity-enhancing effect of DR. Finally, because many different neurodegenerative disorders involve increased oxidative and metabolic stress, and overactivation of glutamate receptors (Beal 1998; Mattson et al. 1999; Mattson 2000) our findings suggest that, by increasing neurotrophic factor production, DR may protect the brain against such neurodegeneration.