Evidence for a release of brain-derived neurotrophic factor from the brain during exercise


Corresponding author P. Rasmussen: AN2041, Blegdamsvej 9, Copenhagen 2100, Denmark. Email: peter@prec.dk


Brain-derived neurotrophic factor (BDNF) has an important role in regulating maintenance, growth and survival of neurons. However, the main source of circulating BDNF in response to exercise is unknown. To identify whether the brain is a source of BDNF during exercise, eight volunteers rowed for 4 h while simultaneous blood samples were obtained from the radial artery and the internal jugular vein. To further identify putative cerebral region(s) responsible for BDNF release, mouse brains were dissected and analysed for BDNF mRNA expression following treadmill exercise. In humans, a BDNF release from the brain was observed at rest (P < 0.05), and increased two- to threefold during exercise (P < 0.05). Both at rest and during exercise, the brain contributed 70–80% of circulating BDNF, while that contribution decreased following 1 h of recovery. In mice, exercise induced a three- to fivefold increase in BDNF mRNA expression in the hippocampus and cortex, peaking 2 h after the termination of exercise. These results suggest that the brain is a major but not the sole contributor to circulating BDNF. Moreover, the importance of the cortex and hippocampus as a source for plasma BDNF becomes even more prominent in response to exercise.

Brain-derived neurotrophic factor (BDNF) is a key protein in regulating maintenance, growth and even survival of neurons (Mattson et al. 2004). Brain-derived neurotrophic factor also influences learning and memory (Tyler et al. 2002), and brain tissue from patients with Alzheimer's disease and clinical depression exhibit low expression of BDNF (Connor et al. 1997; Karege et al. 2002). Brain-derived neurotrophic factor has also been identified as a key component of the hypothalamic pathway that controls body weight and energy homeostasis (Wisse & Schwartz, 2003). Obese phenotypes are found in BDNF-heterozygous mice and are associated with hyperphagia, hyperleptinaemia, hyperinsulinaemia and hyperglycaemia (Lyons et al. 1999). In addition, BDNF reduces food intake and lowers blood glucose in diabetic mice (Nakagawa et al. 2000). In humans, similar symptoms are associated with the functional loss of one copy of the BDNF gene and with a mutation in the BDNF receptor Ntrk2 gene (Yeo et al. 2004; Gray et al. 2006).

Physically and socially more complex housing leads to increased neurogenesis, improved learning and less weight gain in rats (Young et al. 1999; Cao et al. 2004) associated with consistent up-regulation of BDNF expression, and a direct role for BDNF has recently been reported (Cao et al. 2009). A better understanding of therapeutic actions aimed at increasing BDNF levels, such as exercise (Neeper et al. 1995), is of clinical relevance. It is well known that BDNF synthesis is centrally mediated and activity dependent (Johnson & Mitchell, 2003) and that exercise enhances BDNF transcription in the brain (Oliff et al. 1998). In addition, exercise induces brain uptake of insulin-like growth factor 1, which is a prerequisite for the elevation in BDNF mRNA expression (Carro et al. 2000). However, the regions within the brain responsible for the production of BDNF are not known. Physical exercise increases circulating BDNF levels in healthy humans (Gold et al. 2003; Vega et al. 2006; Ferris et al. 2007) although the origin is unclear. We have recently identified BDNF as being a novel contraction-induced muscle cell-derived protein that activates and increases fat oxidation in skeletal muscle in an AMPK-dependent fashion (Matthews et al. 2009). Although BDNF was robustly up-regulated in contracting muscle fibres, muscles were not a source of circulating BDNF. Thus, the main source of circulating BDNF at rest and in response to exercise has not been defined.

Tang et al. (2008) proposed that platelets can explain the finding that that serum BDNF is increased in response to exercise. However, BDNF is also increased in plasma samples, suggesting that BDNF might originate from several other cell sources. We previously demonstrated cerebral output of BDNF in resting healthy humans (Krabbe et al. 2007). Alterations in plasma BDNF levels could therefore reflect variation in the release of BDNF from the brain (Lommatzsch et al. 2005). However, the contribution of the brain to the level of BDNF present in the internal jugular vein in humans during exercise is unknown. While the increase in serum BDNF depends on exercise intensity (Ferris et al. 2007), it is unclear whether exercise duration influences BDNF levels.

The aim of this study was to evaluate the contribution of the human brain to plasma BDNF at rest and during prolonged whole-body exercise through the measurement of arterial-to-internal jugular venous difference (a–v difference). To further identify the cerebral region(s) responsible for BDNF release and examine the time pattern of exercise-induced BDNF expression, mice brains were dissected and analysed for BDNF mRNA expression following treadmill exercise. We hypothesized that the release of BDNF from the human brain would progressively increase throughout the exercise.


Human study

Eight men aged 22–40 years participated in the study (height 1.88 ± 0.08 m, body mass 84 ± 9 kg and maximal oxygen uptake 4.8 ± 0.5 l min−1; means ±s.d.). The subjects provided written informed consent to the study as approved by the Ethics Committee of Copenhagen and Frederiksberg according to the principles established in the Declaration of Helsinki. Each participant visited the laboratory on two occasions. On day 1, the subjects performed incremental exercise to exhaustion on a wind-braked rowing ergometer (Concept II, Morrisville, VT, USA) to determine the workload associated with the lactate threshold, defined as the first inflection in the relation between blood lactate and exercise intensity. The subjects were instructed to eat a regular breakfast on the day of the main study. On day 2, following 30 min of supine rest, subjects performed a 4 h bout of ergometer rowing corresponding to a workload 10–15% below the lactate threshold (Hart et al. 2006) to evaluate the influence of exercise duration on BDNF concentration. We did not consider the exercise mode of importance because we did not have any a priori indications that exercise modality would impact BDNF response as long as the subjects are familiar with the activity. Subjects were encouraged to row as far as possible while maintaining a steady pace.

On the day of the main study, a catheter was placed with Seldinger technique, under local anaesthesia (2% lignocaine), retrograde in the right internal jugular vein (1.6 mm, 14 gauge; ES-04706, Arrow International, Reading, PA, USA), guided by an ultrasound image, and advanced to the bulb of the vein. Arterial blood was drawn from a catheter in the radial artery (1.1 mm, 20 gauge) of the non-dominant arm. Blood was sampled after 30 min of rest prior to the exercise bout and after 2 and 4 h of rowing as well as after 1 h of recovery in the supine position.

Blood was analysed for oxygen, carbon dioxide, glucose and lactate content (ABL 725, Radiometer, Copenhagen, Denmark), and heart rate (HR) was recorded on a wristband monitor (Polar Electro OY, Kempele, Finland). For determination of plasma BDNF, blood samples were drawn into glass tubes containing EDTA, which were immediately spun at 2600g for 15 min at 4°C. Plasma was isolated, respun at 10 000g for 10 min at 4°C, re-isolated for complete platelet removal, and stored at −80°C until analysed. We used an ELISA specific for BDNF (R&D Systems, Minneapolis, MN, USA) to measure plasma concentrations of BDNF. Samples were analysed in duplicate, and mean concentrations were calculated. Two measurements with arterial BDNF values >200 pg ml−1 at rest were excluded from the analysis because the values lay well outside the mean and 3 s.d. range compared with the remaining subjects of the group. The cerebral fractional release of BDNF (fBDNF) was calculated as v–a difference divided by venous concentration.

Animal study

Forty mice were divided into five groups of eight mice each. All animals were acclimated to the treadmill by 10 min of running on three separate days, with the last session being held 48 h before the experiment. One group of mice (‘Pre’) did not run acutely and served as control animals. The other mice exercised for 2 h on a treadmill (18 m min−1, 10% slope) until exhaustion and were euthanized by cervical dislocation either immediately after exercise (0 h), or after 2, 6 or 24 h of recovery. The brains were dissected immediately, and cerebellum, hippocampus and the remaining parts of the brain were separated and quickly frozen in liquid nitrogen for mRNA analysis. One mouse in the 24 h group had to be taken out of the experiment and, accordingly, seven mice were available for analysis in that group.

All mice were kept on a 12 h–12 h light–dark cycle and received standard rodent chow (Altromin no. 1324, Chr. Pedersen, Ringsted, Denmark). Experiments were approved by the Danish Animal Experimental Inspectorate and complied with the European convention for the protection of vertebrate animals used for experiments and other scientific purposes (council of Europe, no. 123, Strasbourg, France, 1985).

Isolation of RNA, RT and PCR

Isolation of RNA was performed on cerebellum, hippocampus and the remaining brain (mainly cortex) using a guanidinium thiocyanate–phenol–choloroform method modified from Chomczynski & Sacchi (2006) as described previously (Pilegaard et al. 2000). Reverse transcription (RT) was performed using the Superscript II RNase H-system (Invitrogen, Carlsbad, CA, USA) as previously described (Pilegaard et al. 2000). The amount of single-stranded DNA (ssDNA) was determined in the RT samples using the OliGreen reagent (Molecular Probes, Leiden, The Netherlands) as previously described (Lundby et al. 2005). The BDNF mRNA content was determined by fluorescence-based real-time PCR (ABI PRISM 7900 Sequence Detection System, Applied Biosystems, Foster City, CA, USA). Forward and reverse primers and TaqMan probe were designed from mouse specific sequence data (Ensembl, Sanger Institute, Hauxton, UK) using computer software (Primer Express, Applied Biosystems). The oligo sequences used to amplify a fragment of the BDNF mRNA were as follows: forward primer 5′-GGACAGCAAAGCCACAATGTTC-3′; reverse primer 5′-TCCGTGGACGTTTACTTCTTTCAT-3′; and TaqMan probe 5′-CGGTTGCATGAAGGCGGCG-3′. The probe was 5′6-carboxyfluorescein (FAM) and 3′6-carboxy-N,N,N′,N′-tetramethylrhodamine (TAMRA) labelled. Prior optimization was conducted to determine optimal primer concentrations, probe concentration and to verify the efficiency of the amplification. The PCR amplification was performed (in triplicates) in a total reaction volume of 10 μl and the Ct values were converted to a relative amount using the standard curve (Lundby et al. 2005). The BDNF mRNA content was normalized to the total ssDNA content in each sample, and this BDNF mRNA/ssDNA ratio is presented.

Statistical analysis

The effect of time was evaluated by using a one-way ANOVA with repeated measures in the human experiment and a one-way ANOVA in the mouse study (proc mixed, SAS 9.1, SAS Institute Inc., Cary, NC, USA). Following a significant F−test, Student–Neuman–Keuls post hoc test was used for multiple comparisons and statistical significance was accepted at P < 0.05. Values are expressed as means ±s.d., except in the figures, where error bars indicate s.e.m.


Human experiment

All volunteers completed 4 h ergometer rowing; however, one subject had to reduce pace after 3 h and 15 min of rowing to be able to complete the bout. Mean work rate and HR were 160 ± 38 W and 143 ± 8 beats min−1, respectively, and the subjects rowed 51.1 ± 8.3 km. No changes were observed in arterial haemoglobin oxygen saturation ( inline image; Table 1). Rowing exercise caused a marked impact on cerebral haemodynamics. Internal jugular venous haemoglobin oxygen saturation ( inline image) decreased after 4 h of exercise compared with rest, and the decrease became statistically significant (P < 0.05). Similarly, arterial carbon dioxide tension ( inline image) decreased throughout rowing to 4.5 ± 0.5 kPa after 4 h but the resting inline image (5.0 ± 0.3 kPa) was restored after 1 h of recovery. During the experiment, arterial glucose was relatively stable, ranging from 6.3 ± 0.9 mmol l−1 at rest to 5.3 ± 0.9 mmol l−1 after 1 h of recovery. No changes in arterial lactate concentration were observed.

Table 1.  Arterial ( inline image ) and jugular venous haemoglobin oxygen saturation ( inline image ) and arterial carbon dioxide tension ( inline image )
 RestRowingRecovery 1 h
2 h4 h
  1. Values are means ±s.d. for 8 subjects.

inline image (%)97.3 ± 0.696.5 ± 1.095.3 ± 4.096.1 ± 1.4 
inline image (%)70.9 ± 9.859.0 ± 7.262.2 ± 5.764.5 ± 8.7 
inline image (kPa) 5.0 ± 0.3 4.6 ± 0.2 4.5 ± 0.5 5.0 ± 0.28

Arterial BDNF increased during rowing (P < 0.05; Fig. 1). The internal jugular venous BDNF concentration increased from 442 ± 272 pg ml−1 at rest to 1172 ± 968 pg ml−1 after 4 h of exercise (P < 0.05) and returned to the resting level after 1 h of recovery (P < 0.05). At rest, BDNF was released from the brain, with the a–v difference being −347 ± 316 pg ml−1, and this release increased with exercise (−902 ± 876 pg ml−1, P < 0.05). After 1 h of recovery, the release of BDNF from the brain returned to resting levels. The fBDNF was 72 ± 32% at rest and 84 ± 8% during exercise, without reaching statistical significance. In the recovery period, the fBDNF decreased to 35 ± 44% (P < 0.05 versus rest).

Figure 1.

Arterial, jugular venous and venous–arterial difference across the brain for brain-derived neurotrophic factor (BDNF) before, during and 1 h after a 4 h rowing bout
Values are means ±s.e.m. for 8 subjects. *P < 0.05 different from rest and †P < 0.05 different from rest and exercise.

Mouse experiment

At rest, the level of BDNF mRNA was twofold higher (P < 0.05) in cortex than in the hippocampus and approximately threefold higher (P < 0.05) in hippocampus than in cerebellum. In response to 2 h of treadmill exercise, the expression of BDNF mRNA was increased in hippocampus and cortex, but not in cerebellum (Fig. 2). The BDNF mRNA expression peaked both in hippocampus and cortex at 2 h of recovery from exercise, with levels three- to fivefold higher (P < 0.05) than in control mice.

Figure 2.

Brain-derived neurotrophic factor in mouse brain cerebellum (open bars), hippocampus (hatched bars) and cortex (cross-hatched bars) before (Pre), immediately after (0 h) and 2, 6 and 24 h after a 2 h treadmill exercise bout
The BDNF mRNA content is normalized to the single-stranded DNA content in the samples. Each bar represents means of 8 animals (except the group at 24 h, in which 7 animals were available for analysis) and s.e.m. *P < 0.05 significantly different from ‘Pre’; †P < 0.05 significantly different from cerebellum.


The present results suggest that the brain has a significant BDNF production both at rest and during prolonged exercise, and that it may be a major source for increased plasma BDNF during exercise in healthy subjects. These observations are supported by BDNF mRNA present in all three examined brain parts in the mouse and an exercise-induced increase in BDNF mRNA expression in mouse hippocampus and cortex in response to a single exercise bout. The peak in mRNA expression extended into recovery, suggesting that exercise-induced BDNF gene regulation within the brain occurs into the recovery phase.

At rest, BDNF is released into the internal jugular vein, suggesting that the brain delivers BDNF to the circulation (Krabbe et al. 2007). The present findings confirm the presence of a cerebral output of BDNF at rest in trained subjects. Since BDNF can cross the blood–brain barrier in both directions (Poduslo & Curran, 1996; Pan et al. 1998), it is of interest to discern the contribution of the brain to the internal jugular venous concentration of BDNF at rest and during exercise. The finding that almost three-quarters of the BDNF present in the venous circulation originated from brain structures suggests that brain tissue is the main contributor to the circulating BDNF. However, other possible explanations for the release exist. Brain-derived neurotrophic factor is released from the cerebral vascular endothelium following hypoxic stress (Wang et al. 2006; Guo et al. 2008). While such stress may not be present at rest, exercise could result in cerebral hypoxic stress, since cerebral oxygen tension decreases during strenuous exercise (Nybo & Rasmussen, 2007). It remains to be determined from which cells the mRNA induction originates. Cerebral blood volume is reported to be less than 5% (Ito et al. 2005). Thus, the bulk of the cells in the central nervous system are not vascular endothelium cells. It is therefore most likely that the mRNA increase occurs in tissues outside the vascular endothelium. Brain-derived neurotrophic factor may also be released from activated platelets in the cerebral circulation as suggested in conditions such as sleep apnoea (Staats et al. 2005). The platelet content of BDNF represents the vast majority of BDNF in circulating blood, and the increase in a–v difference across the brain for BDNF during exercise could originate from activation of platelets transitioning through the cerebral vasculature. Conversely, the increase in brain BDNF mRNA expression is in accordance with findings by Neeper et al. (1995), and that BDNF found in the internal jugular vein originates from brain structures such as the hippocampus or the cortex. Contribution from other sources, however, cannot be excluded.

There was a dramatic reduction of BDNF concentration from the jugular vein to the radial artery. The fate, however, of the BDNF in the periphery remains unclear. Brain-derived neurotrophic factor enhances lipid oxidation in the muscles (Matthews et al. 2009) and it could thus be speculated that the muscles take BDNF up. However, clearance of BDNF by the liver as well as simple dilution cannot be ruled out.

In response to prolonged exercise, the contribution from the human brain to BDNF in the circulation was larger than at rest. The transient increase in BDNF plasma concentration in response to prolonged exercise confirms that a single exercise bout changes plasma BDNF concentrations (Gold et al. 2003; Vega et al. 2006). The concentration of BDNF in the internal jugular vein increased by two- to threefold after 4 h of exercise. The brain did not increase its BDNF release after 2 h but only after 4 h of exercise, suggesting that, at this specific exercise intensity, more than 2 h is needed to increase the cerebral BDNF release. This observation emphasizes the importance of the prolonged nature of this specific exercise stimulus. Accordingly, exercise duration seems to influence circulating BDNF in addition to exercise intensity (Ferris et al. 2007). We speculate that the volume of exercise could be the trigger to increase BDNF release or that BDNF is released as exercise becomes strenuous, and several of the subjects reported high perceived exertion towards the end of the rowing bout. The reduction in fBDNF during recovery might reflect that while the brain transiently increases the release of BDNF in response to exercise, plasma BDNF levels return to resting concentrations during recovery through a reduction of BDNF release by the brain. While hyperglycaemia attenuates BDNF release (Krabbe et al. 2007), the release of BDNF by the brain in the present study was not an effect of exercise-induced hypoglycaemia but exercise per se, since the release of BDNF returned to baseline after 1 h of recovery, when blood glucose was the lowest.

The levels of BDNF mRNA in the various brain parts and the transient up-regulation of BDNF mRNA expression in mouse hippocampus and cortex, but not cerebellum, in response to exercise emphasize the likely importance of specific parts of the brain as a source of BDNF at rest as well as during and after exercise. Surprisingly, although the cerebellum is involved in motor tasks, BDNF mRNA was not up-regulated in the cerebellum following exercise. Complex motor learning and moderate exercise produce different effects on the expression of BDNF (Klintsova et al. 2004). It may be that the synthesis of BDNF is part of the adaptation to a new stimulus, such as exercising to exhaustion, or coping with a new or stressful environment, such as the exercise laboratory. Since both running and rowing are relatively simple motor tasks once they have been learned, the cerebellum may not have been heavily activated. The cerebellum may thus not have been activated to the same level as the hippocampus and cortex and it may be that a threshold needs to be passed before an increase in BDNF mRNA is exhibited. Given the involvement of the hippocampus and cortex in memory and cognition, sensation of exertion may be part of the signal for the increase in BDNF mRNA rather than the motor task itself. We may hypothesize a link between the metabolic challenges imposed by strenuous exercise on the brain (Dalsgaard, 2006; Nybo & Rasmussen, 2007) and BDNF production. Unfortunately, we did not systematically evaluate perceived exertion. Further studies should evaluate whether a relation exists between perceived exertion and BDNF release from the brain. Moreover, the observation that the mRNA expression in the mouse brain peaked 2 h after the end of exercise, whereas the release of BDNF from the human brain peaked during the exercise period may indicate that exercise elicits multiple roles of BDNF exerted at different times, although species differences may exist. We suggest that the increase in circulating BDNF in humans engaged in prolonged, strenuous exercise originates from the brain. Our results from mice exercising to exhaustion support a cerebral contribution to circulating BDNF. We do, however, acknowledge that comparison of the exercise intensity and duration may be troublesome between humans and mice, and this may explain the different time courses of the two responses. We can, though, state that both protocols induced a significant BDNF response. The findings in mouse brain, however, suggest that a likely increased protective role of BDNF within the brain in response to exercise mainly occurs in the recovery phase or that repeated bouts of exercise may well be needed for detectable BDNF protein changes to be evident. Thus, training-induced elevated BDNF mRNA levels are suggested to derive from cumulative effects of transient increases after each exercise bout eventually leading to elevated protein levels. Still, it is likely that the increased BDNF levels detected in the internal jugular vein of the human subjects after exercise comes from a release of BDNF by the brain and not from other blood-borne sources. In this context, up-regulation of BDNF mRNA in the recovery period may serve as a super-compensatory adaptation to the increased demand for BDNF release by the brain during and after exercise.


Healthy trained men were investigated, and the generalization of the present results might be limited to this population. However, the aim of this study was to identify whether the brain is a source of BDNF during exercise in humans. Oestrogen is known to regulate the expression of BDNF (Sohrabji & Lewis, 2006), but plasma BDNF levels have been shown to be similar between sexes in healthy adults (Lommatzsch et al. 2005). However, women displayed significantly lower platelet BDNF levels than men in that study. Accordingly, a difference in BDNF release from the brain between sexes cannot be ruled out and limits the generalization of this study. Thus, to avoid potential sex hormonal effects on BDNF expression and/or release, we chose only to include male subjects in the present study.

There was also some interindividual variance in BDNF a–v difference. Two subgroups may be differentiated, maybe because the superior sagittal sinus drains into the right internal jugular vein in only approximately 50% of individuals (Lambert et al. 1991, 2000; Ferrier et al. 1993), and whether BDNF overflow is markedly different between cortical and subcortical brain regions is not known. Given that BDNF mRNA expression increased in cortex, it may be that for some subjects the a–v difference for BDNF could have been even larger than observed had the contralateral jugular vein been catheterized. Also, no significant elevation in arterial BDNF was present after 4 h of exercise. However, when the two subjects with outlying resting BDNF levels were excluded from analysis, the increase in arterial BDNF became significant. Since the exercise-induced increase in systemic BDNF depends on its resting levels and BDNF concentration is influenced by acute exercise (Ferris et al. 2007), it may be speculated that some of the variability in the resting BDNF levels may be due to subjects having performed some physical activity before reporting to the laboratory, e.g. commuting by cycling. We did not evaluate the general activity levels of the subjects. We included active rowers, and they must be assumed to be above-average physically active subjects. Accordingly, fitness level may have influenced our results, but the main goal was to verify whether the brain is a source of BDNF during exercise, rather than evaluating the impact of fitness level on the BDNF response to exercise.

We did not measure cerebral blood flow, and it could be speculated that the changes in the BDNF a–v difference were related to changes in cerebral blood flow. However, the changes in cerebral blood flow were moderate when gauged from inline image or inline image (Table 1) and are unlikely to account for the two- to fivefold increase in the BDNF a–v difference. Furthermore, when measuring a–v difference, it is not possible to discriminate whether the neurons, glial cells or the cerebrovascular endothelium produce BDNF. However, exercise is associated with an elevation of BDNF mRNA in the hippocampus and cerebral cortex in mice. Therefore, the brain is likely to be the main source of systemic BDNF during prolonged exercise.

Since we did not measure the contribution of platelets or other peripheral sources of BDNF, the importance of the brain as a source of BDNF might be overestimated relative to peripheral sources in the blood circulation draining the brain. The platelets were spun down and thus removed from the surfactant and thereby the analysis. The fBDNF does not take recirculation of BDNF into account, and the brain could thus contribute to a larger proportion of BDNF than illustrated by fBDNF.


In conclusion, the human and mouse data suggest that BDNF mRNA expression peaks in recovery after exercise, while an increased release occurs during exercise. We suggest that BDNF gene regulation during recovery from exercise serves as the basis for cumulative effects of repeated exercise bouts, eventually leading to detectable increases in BDNF protein content and concomitantly increased potential for BDNF release and neuroprotection in specific brain parts.



This work was supported by the Centre of Inflammation and Metabolism and the Copenhagen Muscle Research Centre. This study was further supported by the Danish Medical Research Council, the Commission of the European Communities (contract no. LSHM-CT-2004-005272 EXGENESIS) and the Lundbeck Foundation. Patrice Brassard is the recipient of a postdoctoral fellowship from the Fonds de la recherche en santé du Québec (FRSQ).