The CSF and arterial to internal jugular venous hormonal differences during exercise in humans


Corresponding author M. K. Dalsgaard: Department of Anaesthesia, Rigshospitalet 2041, Blegdamsvej 9, DK-2100 Copenhagen Ø, Denmark. Email:


Strenuous exercise increases the cerebral uptake of carbohydrate out of proportion to that of oxygen, but it is unknown whether such enhanced carbohydrate uptake is influenced by the marked endocrine response to exercise. During exhaustive exercise this study evaluated the a–v differences across the brain (a–v diff) of hormones that could influence its carbohydrate uptake (n= 9). In addition, neuroendocrine activity and a potential uptake of hormones via the cerebrospinal fluid (CSF) were assessed by lumbar puncture postexercise and at rest (n= 6). Exercise increased the arterial concentration of noradrenaline and adrenaline, but there was no cerebral uptake. However, following exercise CSF noradrenaline was 1.4 (0.73–5.5) nmol l−1, and higher than at rest, 0.3 (0.19–1.84) nmol l−1 (P < 0.05), whereas adrenaline could not be detected. Exercise increased both the arterial concentration of NH4+ and its a–v diff, which increased from 1 (–12 to 5) to 17 (5–41) μmol l−1 (P < 0.05), while the CSF NH4+ was reduced to 7 (0–10) versus 11 (7–16) μmol l−1 (P < 0.05). There was no release from, or accumulation in the brain of interleukin (IL)-6, tumour necrosis factor (TNF-α), heatshock protein (HSP72), insulin, or insulin-like growth factor (IGF)-I. The findings indicate that for maximal exercise, the concentration of noradrenaline is increased within the brain, whereas blood borne hormones and cytokines are seemingly unimportant. The results support the notion that the exercise-induced changes in brain metabolism are controlled by factors intrinsic to the brain.

Activation of the human brain by exercise provokes an increase in its uptake of carbohydrate that is out of proportion to that of oxygen (Dalsgaard et al. 2002). Intense neuronal activity may be fuelled by lactate from glycolysis (Sappey-Marinier et al. 1992) and/or glycogenolysis in neighbouring astrocytes (Brown et al. 2003). Moreover, with exhaustive exercise a marked lactate uptake by the brain is likely to be metabolized (Dalsgaard et al. 2004) and this aggravates the reduction in the cerebral metabolic ratio (O2/[glucose +½ lactate]). Such enhanced glucose and lactate uptake by the brain during exercise takes place while plasma insulin is expected to be low (Aarnio et al. 2001). However, the maximal carbohydrate uptake is reached during the first five minutes of the recovery (Dalsgaard et al. 2002) which may coincide with a rebound increase in plasma insulin (Aarnio et al. 2001). Furthermore, exercise elevates the plasma concentrations of other hormones such as noradrenaline, adrenaline, cortisol, and insulin-like growth factor (IGF)-I that may influence glucose and/or glycogen metabolism (Galbo et al. 1987; Schwarz et al. 1996; Del Corral et al. 1998).

This study evaluated if such endocrine responses could explain the cerebral carbohydrate uptake during and after intense exercise by determination of their arterial to internal jugular venous differences (a–v diff). It was furthermore assessed whether the selected hormones became detectable in the CSF. Noradrenaline and vasoactive intestinal peptide (VIP) are important regulators of glycogen metabolism within the brain (Magistretti, 1988; Cambray-Deakin et al. 1988) and neuroendocrine activity may be too discrete to be expressed as the balance over the brain. Also, insulin, cortisol and in particular IGF-I, may enter the brain via the CSF (Reinhardt & Bondy, 1994; Pulford & Ishii, 2001).

Exercise induces changes not only in hormones but also in substrates and cytokines. Release of ammonia (NH3) from intensely contracting skeletal muscles may raise the ammonium (NH4+) concentration in plasma (van Hall et al. 1995). Plasma NH4+ was assessed as it stimulates glycogen metabolism in astrocytes (Tsacopoulos et al. 1997) and cerebral uptake could interfere with oxidative metabolism. Finally, we considered the concentrations of the cytokines interleukin (IL)-6 and tumour necrosis factor (TNF-α) as well as heat shock protein (HSP72) for their potential association with glucose metabolism. The brain releases IL-6 after prolonged exercise, and in particular when exercise is repeated (Nybo et al. 2002), which may act as a signal demonstrating low energy reserves in the brain. A similar effect is demonstrated with glycogen depletion in skeletal muscle which increases IL-6 release (Steensberg et al. 2001) and gene transcription (Keller et al. 2001). Equally, HSP72 messenger RNA is up-regulated by glycogen depletion (Febbraio et al. 2002).


Two female and seven male healthy subjects gave their written informed consent to participate in the study [median age 26 years (range 23–28); height 182 (168–192) cm; weight, 80 (66–94) kg] as approved by the Ethics Committee of Copenhagen (KF 01–034/02).

Work was made as strenuous as possible by combined arm and leg exercise in order to provide for a marked endocrine response. During the current experiment with these subjects, exercise also induces a large surplus uptake of carbohydrate by the brain (Dalsgaard et al. 2004). A minimum of three days prior to the main study, the subjects had their work capacities determined for the arms and for the legs by incremental exercise, which were found to be 159 (88–191) W and 320 (206–353) W, respectively. The participants reported to the laboratory following an overnight fast, although drinking water was permitted. Subjects were allowed to recover after the invasive procedures, and then exercise commenced (combined arm and leg cycle ergometer at 60 r.p.m.). The work rate increased from a warm up load at ∼30% of the individual work capacity by ∼10% every second minute until exhaustion, which occurred on average after 12 min (range 8–16).

The a–v diff over the brain was determined from a catheter (14 gauge; 2.2 mm) in the right internal jugular vein with the tip advanced retrogradely to the superior bulb and another catheter (20 gauge; 1.1 mm) in the radial artery of the non-dominant arm. Pair wise blood samples were drawn at rest, at the point of exhaustion and after 30 min of recovery. Twenty-five millilitres of blood was drawn into anaerobic test tubes prepared with EGTA (for noradrenaline and adrenaline), EDTA (for IL-6, TNF-α, and NH4+), a combination of EDTA and Trasylol® (Bayer, Leverkusen, Germany; for cortisol, insulin, and IGF-I), and a pro-coagulant (for HSP72). Within half an hour of collection blood was spun at 3500 r.p.m. for 10 min at 4°C. Plasma (serum for HSP72) was divided into fractions, frozen in dry ice, and stored at −40°C, except for the plasma used for the determination of NH4+, which had boric acid and l-serine added.

The CSF was drained through a lumbar puncture as soon as possible after termination of exercise, usually within 2–3 min. The puncture was preceded by local anaesthesia of the skin (2% Lidocain) and a 25 gauge pencil-point cannula (Braun, Melsungen, Germany) was advanced between the third and fourth lumbar vertebrae. ‘Spinal headache’ was prevented by the administration of 1 l of isotonic saline intravenously during the first hour of the recovery. Also, to avoid more than one lumbar puncture for each subject, resting values were obtained from an additional three male and three female subjects [24 (22–29) years, 180 (174–184) cm, 75 (65–84) kg]. Five millilitres of CSF was divided into several tubes without additive, frozen and stored for later analysis.

Cortisol was measured in unextracted plasma by double antibody radioimmunoassay (RIA) using a kit (Diagnostic Products Corporation, Los Angeles, USA). The least detectable quantity was 6.1 nmol l−1 and the intra- and interassay coefficients of variation were 4.5% and 5.7%, respectively. Insulin was measured by enzyme-linked immunosorbent assay (ELISA) technique (DAKO, Glostrup, Denmark). The IGF-I was determined by RIA with intra- and interassay variations less than 6% and 9%, respectively. High-sensitivity ELISA kits were used to measure IL-6 and TNF-α (R & D Systems Europe, Oxon, UK), and HSP72 (StressGen Biotechnologies Corp., Victoria, BC, Canada). The VIP was measured in CSF only (Fahrenkrug et al. 1977). Evaluation of the cerebral metabolism of O2, glucose, and lactate has been reported (Dalsgaard et al. 2004).

Intravascular mean arterial pressure (MAP) was obtained by a Bentley transducer (Uden, Holland) connected to a patient monitor (Dialogue 2000, Danica Electronic, Copenhagen, Denmark). Subjects rated their perceived exertion (RPE) on a scale from ‘6’ to ‘20’, representing rest and the hardest exercise imaginable, respectively (Borg, 1970).

Values are presented as median and range unless otherwise stated. Changes with time were detected by Friedman's test and when found significant, deviating results were located using Wilcoxon's signed test by rank. The Mann–Whitney test was applied for comparison of the CSF variables at rest and after exercise. Linear regression determined a relationship between the arterial and the a–v diff for NH4+, using and the least squares principle. A P value of < 0.05 was considered significant.


The cardiovascular and cerebral metabolic response to exercise has been published (Dalsgaard et al. 2004). With exercise MAP increased from 91 (81–100) to 109 (98–131) mmHg (P < 0.05) and at exhaustion all subjects expressed a RPE of 20. The a–v diff for glucose increased from 0.6 ± 0.0 mmol l−1 to a maximum of 1.3 ± 0.3 mmol l−1 in the immediate recovery, and the a–v diff for lactate rose even faster (from zero) to a peak value of 1.3 ± 0.2 mmol l−1 at exhaustion. Consequently, the cerebral metabolic ratio (O2/[glucose +½ lactate]) decreased to ∼2.8 from the mean resting value close to 6, without accumulation of lactate in the CSF or in the brain as determined by magnetic resonance spectroscopy (Dalsgaard et al. 2004).

Exercise increased the arterial concentrations of noradrenaline and adrenaline markedly, but the a–v diff did not change significantly (Fig. 1A; Table 1). Nevertheless, the CSF concentration of noradrenaline was higher after exercise than at rest (P < 0.05; Fig. 1B). Adrenaline was not detectable in CSF and the concentration of VIP did not increase with exercise. Exercise increased the arterial concentration of NH4+ and its a–v diff (R= 0.71, P < 0.0001; Fig. 2). Conversely, CSF NH4+ was slightly lower after exercise compared to the resting value (P < 0.05). Exercise did not change the concentration of cortisol in blood or in the CSF. The arterial concentration of insulin decreased with exercise but increased after 30 min of recovery to a concentration above that obtained at rest (P < 0.05). The a–v diff for insulin did not convey an uptake at any time point and insulin could not be detected in CSF. Exercise raised the arterial concentration of IGF-I, IL-6 and TNF-α, but the a–v diff and the concentration in CSF did not change significantly.

Figure 1.

A, arterial concentrations of noradrenaline and adrenaline at rest, during and after exhaustive whole body exercise (n= 9). B, CSF noradrenaline concentration immediately after exercise from the same subjects as in A and at rest (n= 6); also given is the median. Values are median with the 25th and the 75th percentiles. * Different compared to rest P < 0.05.

Table 1.  Concentrations of hormones and other variables in arterial blood, cerebrospinal fluid and as a–v differences over the brain in response to exercise



a–v diff
Peak exercise


a–v diff
(2–5 min)
Recovery (30 min)


a–v diff
  1. Values are presented as median with range. CSF, n= 7; blood, n= 8; n= 5; n= 6; * Different compared to rest P < 0.05; IGF, insulin-like growth factor; TNF, tumour necrosis factor; HSP72, heat shock protein; NH4+, ammonium.

Noradrenaline (nmol l−1)0.3 (0.19–1.84)0.63 (0.36–1.3)–0.11 (–0.5–0.31)58 (19–119) *0.19 (–7.1–5)1.4 (0.73–5.5)*1.4 (0.75–3.0)0.12 (–0.89–0.44)
Adrenaline (nmol l−1)0 (0–0.22)0.25 (0.03–0.38)0.01 (–0.37–0.2)15 (3.2–69) *0.23 (–2.0–57.1)00.38 (0.1–3.0)0.04 (–0.16–2.7)
Cortisol (nmol l−1)19.9 (12.3–35.4)554 (417–1047)29 (–145–86)534 (422–828)§31 (–133–146)§23 (19–28)677 (342–1132)§–15 (–116–117)§
Insulin (nmol l−1)37 (14–57)–2.4 (–3.5–0.8)26 (16–34)*–0.1 (–1–8.1)88 (41–371)*0.15 (–4–16)
IGF-I (ng ml−1)< 21199 (148–256)–7 (–25–19)233 (166–340)§*– 2 (–42–56)§< 21204 (164–257)§1.5 (–35–15)§
IL-6 (pg ml−1)1.8 (0–2.5)0.89 (0–2.2)0.11 (–0.41–0.48)1.8 (0.84–3.2)*– 0.03 (–0.58–2.1)1.5 (0–2.4)1.5 (0.06–2.89)–0.09 (–2.7–0.03)
TNF-α (pg ml−1)00.83 (0.29–2.28)0.05 (–0.76–0.23)0.93 (0.65–3.21)*0.09 (–0.23–0.43)00.60 (0.19–2.61)–0.08 (–0.49–0.25)
HSP72 (pg ml−1)0.04 (0–0.86)0 (0–1.22)0 (–0.38–0.21)0 (0–1.06)0 (–0.22–0.48)0 (0–0.14)0 (0–0.60)0 (–0.15–0.18)
NH4+ (μmol l−1)11 (7–16)23 (15–30)1 (–12–5)67 (49–142)*17 (5–41)*7 (0–10)*26 (14–44)3 (–14–7)
VIP (pmol l−1)11 (0–47)33 (15–41)
Figure 2.

A, arterial concentration and a–v difference over the brain for NH4+ at rest (n= 9), during and after exhaustive exercise (n= 8). B, CSF NH4+ concentration immediately after exercise from the same subjects as in A and at rest (n= 6); also given is the median. C, the arterial–venous differences over the brain for NH4+ as a function of the arterial concentration at rest (○), during exhaustive exercise (•), and in the recovery (▴); the best fit is y= 0.30 x–5.77, R= 0.71, P < 0.0001. Values are presented as median with the 25th and the 75th percentiles. * Different compared to rest P < 0.05.


The human brain takes up carbohydrate out of proportion to O2 during and following intense exercise (Dalsgaard et al. 2002) in a manner similar to that seen in conditions of increased brain activity (Fox et al. 1988). Glucose transporters in the brain are mainly of the insulin-insensitive subtype 1 and 3, although the insulin-sensitive type 4 has been identified (Vannucci et al. 1998). Insulin enters the brain after peripheral administration (Reinhardt & Bondy, 1994) where it promotes glucose transport, particularly in cortical areas (Bingham et al. 2002). Also, insulin stimulates incorporation of glucose into glycogen (Dringen & Hamprecht, 1992; Choi et al. 2003). In the present study, as expected, the arterial concentration of insulin decreased with exercise (Guezennec et al. 1982) and was at its lowest when the cerebral carbohydrate uptake was at its highest. In the late recovery insulin had increased to above pre-exercise concentration, but the a–v diff did not change. It therefore appears that the cerebral metabolic ratio recovers largely without the influence of insulin. On the other hand, replenishment of cerebral glycogen may take longer than the time it took to deplete (Dienel et al. 2002). Hence, insulin could become important during such prolonged resynthesis of glycogen, especially when in these fasting subjects plasma glucose and insulin increase upon food intake after the study.

Compared to insulin, the structurally and functionally homologous IGF-I traverses more easily from the blood into the brain tissue (Reinhardt & Bondy, 1994). In addition, IGF-I more potently induces glucose metabolism and promotes transport and consumption of glucose in the brain (Cheng et al. 2000) as well as restoration of astrocyte glycogen (Dringen & Hamprecht, 1992). In this study, the 10–15% increase in IGF-I from baseline with exercise is in line with previous findings (Schwarz et al. 1996), and may enter the brain from blood or via the CSF (Carro et al. 2000; Pulford & Ishii, 2001). However, there was no cerebral uptake of IGF-I and it could not be detected in CSF. Yet, an increase in CSF IGF-I from a concentration around 3 ng ml−1 at rest (Haselbacher & Humbel, 1982) could occur but remain below our detection limit (∼21 ng ml−1) during exercise or at a later recovery stage (Pulford & Ishii, 2001).

Noradrenaline and VIP mediate not only the breakdown of glycogen in the brain (Cambray-Deakin et al. 1988) but also the subsequent resynthesis (Sorg & Magistretti, 1992; Allaman et al. 2000). For noradrenaline such an effect presumably occurs mainly via β-adrenergic receptors compared to α1-receptors, as illustrated when blockade by propranolol prevents glycogen repletion in cell cultures (Cambray-Deakin et al. 1988; Allaman et al. 2000) and the surplus carbohydrate uptake by the brain in physically active rats (Schmalbruch et al. 2002). Superimposing arm exercise on leg exercise effectively raised the arterial concentration of noradrenaline to higher values than reported with only leg exercise (Pott et al. 1996), most probably through a combined effect of a large contracting muscle mass (Savard et al. 1987) and an intense mental effort, that through sympathetic activation provokes noradrenaline spill-over from organs such as skin, kidneys and adrenal medulla (Goldstein et al. 1987a). A substantial increase in arterial noradrenaline may influence its concentration in CSF (Goldstein et al. 1987b), but the a–v diff remained stable. Therefore, the ∼4 fold increase in CSF noradrenaline compared to baseline is likely to result from neuronal activity (perhaps originating in the nucleus locus coeruleus), and a similar increase with exercise is noted in the dog (Radosevich et al. 1989). It is noteworthy that this increase in brain turnover of noradrenaline coincides with a low cerebral metabolic ratio. Furthermore, spillover from VIP-containing neurones confined to the cerebral cortex may be the source of the VIP in CSF, which was unaffected by exercise.

The increase in the arterial concentration of adrenaline exceeded that obtained during maximal leg exercise (Galbo et al. 1987; Pott et al. 1996). Adrenaline may play a role in neurotransmission within the rodent brain during motor activity (Stone et al. 2003). However, cerebral metabolism in the rat is unaffected by infusion of adrenaline (De Bruin et al. 1990), and in the present study there was no cerebral uptake and adrenaline was not detectable in CSF. Even though cortisol is reported to increase in blood and CSF in proportion with exercise intensity and mental effort (Radosevich et al. 1989), all measured variables for cortisol remained stable.

A cerebral uptake of NH4+ could challenge the cerebral metabolic ratio as NH4+ stimulates glycolysis in glia (Tsacopoulos et al. 1997) and may interfere with neurotransmitter (glutamate and γ-amino butyric acid) metabolism (Guezennec et al. 1998). In the present study, the two-fold increase in the arterial concentration of NH4+ was paralleled by an increase in the a–v diff, in agreement with NH3, which is in fast equilibrium with NH4+, traverses rapidly from blood to brain (Felipo & Butterworth, 2002). Despite the cerebral NH4+ uptake, the CSF concentration of NH4+ was slightly lower after exercise as compared with rest. This suggests that the NH4+ taken up by the brain is metabolized, and in support of this notion, the intra cerebral pools of glutamate and alanine increase with brain activity in the rat (Dienel et al. 2002). Integrating the NH4+ a–v diff curve during exercise and in the recovery, and assuming a cerebral blood flow of 0.7 l min−1 as previously described (Dalsgaard et al. 2004), the amount of NH4+ taking up, and seemingly metabolized, is estimated to 0.2 mmol (even when taking into account the reduction in CSF NH4+). Such uptake of NH4+ can, however, explain only ∼3% of the surplus uptake of carbohydrate by the brain that is not oxidized in these subjects (Dalsgaard et al. 2004).

In addition, it is unlikely that a cerebral net uptake of NH4+ would lead to a low CSF NH4+. Thus, at least during maximal exercise, the observation of a low CSF NH4+ contradicts the hypothesis that a cerebral uptake of NH4+ is important for central fatigue (Guezennec et al. 1998).

Following prolonged exercise to exhaustion the brain releases IL-6 and this is particularly pronounced if exercise is repeated (Nybo et al. 2002). Efflux of IL-6 from the brain may express low energy reserves as exemplified by depletion of muscle glycogen (Keller et al. 2001). It is suggested that IL-6 could be implicated in central fatigue. Likewise, shortness of energy supply, as a low glycogen level in skeletal muscle, appears to up-regulate the HSP72 messenger RNA (Febbraio et al. 2002). In the present study, exercise raised the arterial concentration of IL-6, but there was neither a net release nor an uptake by the brain and the CSF concentration remained unchanged. The IL-6 release from the brain was recently demonstrated to diminish during prolonged exercise with hypoglycaemia compared to prolonged exercise when glucose was supplemented (Nybo et al. 2003). Thus, factors other than glucose metabolism in the brain appear to cause IL-6 release, although we cannot rule out that the time of the lumbar puncture, i.e. ∼15 min after exercise start, was too early for IL-6 production to become manifest in the CSF.

The TNF-α was included in the evaluation of brain glucose metabolism for its potential interaction with IL-6 (Sawada et al. 1992). However, there was no cerebral uptake or release or change in the CSF concentration. Equally, following fatiguing exercise in the rat there is no change in plasma TNF-α or in the expression of messenger RNA in the brain (Colbert et al. 2001). The TNF-α seems to increase primarily with inflammation (Sawada et al. 1992; Colbert et al. 2001) and in contrast to the neuroprotective effects also exhibited by IL-6, TNF-α may be mainly pro-inflammatory (Chang et al. 2001).

We conclude that during exercise the reduction of the cerebral metabolic ratio takes place while noradrenergic activity within the brain is high, and while NH4+ is taken up and seemingly utilized by the brain. In contrast, systemically provided noradrenaline, adrenaline, and cortisol appear not to be important for brain metabolism, and the cerebral metabolic ratio seems to recover without an uptake of insulin or IGF-I. Furthermore, for IL-6, TNF-α, and HSP72 there seem to be no uptake or release from the brain during a maximal bout of exercise and these cytokines did not build up or become depleted from CSF. The findings suggest that during maximal exercise the reduction and subsequent recovery of the cerebral metabolic ratio is governed by factors intrinsic to the brain.



The study was supported by The Danish National Research Foundation, Grant 504–4, and the Danish Medical Research Council, Grant 52-00-0098. Mads K. Dalsgaard received founding from the University of Copenhagen Medical Faculty Foundation and from the Medical Society of Copenhagen. The authors express their gratitude to J. Westergaard (Braun, Germany) for providing the spinal needles and Peter Nissen, Elsa Larsen and Nine Scherling for technical assistance.