brain-derived neurotrophic factor
granular layer of dentate gyrus
mitogen-activated protein kinase phosphatase-1
neurological severity scores
object recognition test
Exercise benefits the brain in many ways, e.g. promoting neuron repair and inhibiting neuroinflammation. However, current clinical practices often advise patients suffering head injury to rest during the post-traumatic period.
This study used a mouse model to investigate whether and how exercise retarded the brain structural and functional losses induced by a head impact.
An early moderate-exercise protocol (starting 2 days postimpact and lasting for 7 or 14 days) reversed the impact-induced rapid loss of recognition memory and prevented most of the delayed neuronal loss and neuroinflammation.
However, the same exercise protocol started 9 days postimpact was unable to restore deficits in the recognition memory, even though it still retarded the late-phase neuroinflammation.
These beneficial effects of exercise were probably mediated by the timely recovery of neurotrophic factors (brain-derived neurotrophic factor and mitogen-activated protein kinase phosphatase-1) in the injured brain.
Abstract Closed-head injury (CHI) usually involves both physical damage of neurons and neuroinflammation. Although exercise promotes neuronal repair and suppresses neuroinflammation, CHI patients currently often remain resting during the post-traumatic period. This study aimed to investigate whether and how postinjury exercise benefited the brain structure and function in mice after CHI. Closed-head injury immediately caused an elevated neurological severity score, with rapid loss of object recognition memory, followed by progressive location-dependent brain damage (neuronal loss and activation of microglia in the cortex and hippocampus). An early exercise protocol at moderate intensity (starting 2 days postimpact and lasting for 7 or 14 days) effectively restored the object recognition memory and prevented the progressive neuronal loss and activation of microglia. However, if the exercise started 9 days postimpact, it was unable to recover recognition memory deficits. In parallel, early exercise intervention drastically promoted neurite regeneration, while late exercise intervention was much less effective. We also tested the possible involvement of brain-derived neurotrophic factor (BDNF) and mitogen-activated protein kinase phosphatase-1 (MKP-1) in the exercise-induced beneficial effects. Exercise gradually restored the impact-abolished hippocampal expression of BDNF and MPK-1, while oral administration of triptolide (a synthesis inhibitor of MKP-1 and an antagonist of nuclear factor-κB) before each bout of exercise blocked the restorative effects of exercise on MKP-1 and recognition memory, as well as the exercise-induced retardation of neuronal loss. Although triptolide treatment alone inhibited activation of microglia and maintained neuronal numbers, it did not recover the injury-hampered recognition memory. Overall, moderate exercise shortly after CHI reversed the deficits in recognition memory and prevented the progression of brain injury.
Traumatic brain injury is one of the most common causes of death and disability in young people (Ghajar, 2000). Clinical brain concussion and oedema can be mimicked with a closed-head injury (CHI) mouse model, which produces a focal brain injury without skull trephination by a freely falling weight (Shapira et al. 1988; Tang et al. 1997a, b; Flierl et al. 2009). The CHI brain injuries are hallmarked by a bilateral neuronal loss in the cerebral cortex and the hippocampus (Tang et al. 1997b; Lammie et al. 1999; Grossman et al. 2003) and a delayed neuroinflammation (Israelsson et al. 2009). Moreover, learning and memory deficits persist after CHI treatment, similar to those observed in human postconcussive syndrome (Chen et al. 1996; Tang et al. 1997a; Zohar et al. 2003).
Although exercise exerts many beneficial effects on the brain, including the promotion of neuronal repair (Vaynman & Gomez-Pinilla, 2005) and inhibition of neuroinflammation (Leem et al. 2011), the effects of exercise in animal studies of traumatic brain injury have been controversial (McCrea et al. 2009). Moreover, most postconcussion patients are advised not to exercise because of concerns regarding exacerbation of symptoms (Leddy et al. 2007). We have shown that exercise improves learning and memory (Huang et al. 2006; Liu et al. 2009) and enhances neuroprotective effects against brain degeneration (Wu et al. 2007, 2011). Both these beneficial effects of exercise are largely due to the upregulation of brain-derived neurotrophic factor (BDNF) in the brain. Interestingly, BDNF is likely to be involved in CHI, because CHI reduces its production in the hippocampus (Yaka et al. 2007). Moreover, administration of stem cells or stromal cells to traumatically injured animals during the acute phase enhances the BDNF expression, decreases the trauma-induced neuronal apoptosis, and improves the neurological severity scores (NSS; Mahmood et al. 2004; Kim et al. 2010b). However, exactly how BDNF benefits CHI is unclear at present. A recent study has shown that BDNF facilitates axonal branching and remodelling via upregulation of mitogen-activated protein kinase phosphatase-1 (MKP-1) to destabilize the axonal microtubules (Jeanneteau et al. 2010). Mitogen-activated protein kinase phosphatase-1, a key negative immune regulator that prevents excessive inflammation (Eljaschewitsch et al. 2006; Lang et al. 2006) is increased in peripheral leucocytes by exercise training (Chen et al. 2010). It is therefore plausible to assume that certain exercise regimes may restore the CHI-diminished BDNF to promote neuronal repair and to suppress neuroinflammation via the BDNF–MKP-1 pathway.
This study aimed to investigate whether and how exercise prevented the progression of CHI in mice. We also tested the possible involvement of BDNF and MKP-1 in the beneficial effects of exercise on CHI. Our results demonstrated that an early postinjury moderate-exercise protocol effectively restored the CHI-impaired object recognition memory, and prevented the progressive neuronal loss and activation of microglia. Furthermore, these beneficial effects were probably mediated by the exercise-restored BDNF and MKP-1 in the brain after CHI.
Ethical approval, animals and experimental groups
This study was conducted in conformity with the principles of UK regulations on research involving experimental animals. The animal research protocol was approved by the review committee in National Cheng Kung University (IACUC approved number 98145). Seven-week-old male ICR mice (body weight 34–36 g) were purchased from the National Cheng Kung University Animal Centre and housed, four per cage, in an environmentally controlled room, with a temperature of 25 ± 1°C and 12 h–12 h light–dark cycle, with chow and water available ad libitum. Initially, all mice were familiarized with the treadmill for 1 week (details provided in the ‘Exercise protocol’ section below). To investigate the effects of CHI and exercise, mice were randomly assigned to a sham control group or they received a CHI. The ‘CHI mice’ were assigned to early postinjury exercise, later postinjury exercise or sedentary groups. Some mice were further assigned into vehicle control and triptolide-treated groups to examine the possible role of MKP-1 in the effects of exercise on CHI. Oral intubation for administration of triptolide (Sigma-Aldrich; 32 μg kg−1 day−1, in 0.001% dimethyl sulfoxide; Lin et al. 2007) or vehicle into the stomach was performed 20 min before daily treadmill running. Mice were killed by CO2 inhalation and brain tissues collected for immunostaining and single-neuron labelling.
Closed-head injury model
Our CHI method was modified from that of Flierl et al. (2009). The ICR strain of mice was used because they weighed more than 32 g at the age of 7 weeks. The CHI experiments were performed under full surgical anaesthesia induced and maintained by inhalation of isoflurane. Briefly, mice were anaesthetized by inhalation of isoflurane for about 3 min, with a minimal alveolar concentration of isoflurane of 1.0–1.5%. The anaesthetized mice showed no flexion or withdrawl in response to toe pinch and regained consciousness soon after exposure to fresh air.
A mid-line longitudinal incision was performed under general anaesthesia to expose the mouse skull. A metal rod weighing 311 g was dropped from a height of 2.5 cm onto the left anterior hemisphere 1 mm lateral to the sagittal suture and the lambdoid suture. The rod was retracted immediately to prevent unwarranted secondary trauma. Metallic closure clips were used to seal the skin wound rapidly. The whole procedure was completed in 1 min. Impacted animals usually became conscious in less than 2 min after the incision closure. Animals resumed food and water intake within 2 h after the impact; they demonstrated normal seeking behaviour and startle reflex. Two days after the impact, they showed no impairment of walking and were capable of running at moderate intensity without any difficulty (details provided in the ‘Exercise protocol’ section below). Sham-operated control mice were subjected to anaesthesia and skin incision and suturing only. This CHI protocol provided a degree of head injury with distinct motor and memory deficits that could be quantified in the different experimental protocols.
Neurological severity score measurements
The NSS consisted of 10 clinical parameters that assessed motor function, alertness and physiological behaviour (Chen et al. 1996; Flierl et al. 2009). The specific measurements were the presence of mono- or hemiparesis, inability to walk along three beams (32 cm in length and with widths of 3, 2 and 1 cm), inability to balance on a beam (0.7 cm wide) or on a cylindrical rod (0.5 cm diameter), failure to exit a 30-cm-diameter circular enclosure within 3 min, inability to walk straight, loss of startle behaviour due to an acoustic burst and loss of seeking behaviour. One point was scored for the inability to perform a particular task. This parameter was evaluated 1 h after CHI, with repeated testing once a day for the next 16 consecutive days.
Object recognition test (ORT)
This test is based on the tendency of mice to explore a novel object rather than a familiar one (Ennaceur & Delacour, 1988). This test was performed after CHI as described previously (Biegon et al. 2004). The ORT was performed only once for each animal at different time points following CHI. After being habituated to a polycarbonate box (47.5 cm × 25.8 cm × 21 cm) for 1 h per day for three consecutive days, a mouse was put back into the same box containing two identical objects (250 ml glass beakers, 6.5 cm diameter and 9 cm high, placed upside down) separately positioned 5 cm away from a wall. The cumulative time spent by the mouse in exploring each of the objects was recorded manually during a 5 min period. Four hours later, the mouse was reintroduced into the box, where one of the two objects was replaced by a new one (an iron rectangular cuboid, 6.5 cm × 6.5 cm × 7.5 cm). The time spent in exploring each object during a 5 min period was recorded. The baseline trial performance was indicated as the ratio of the time spent in exploring one of two identical objects over the total exploring time. Memory test trial performance was expressed as the percentage of time exploring the new object, i.e. the ratio of the time spent on exploring the novel object over the total exploring time.
In order to become familiar with treadmill running at the beginning of the experiment, mice ran on a motor-driven levelled treadmill (Model T408E; Diagnostic & Research Instruments Co., Taoyuan, Taiwan), starting at a very low speed and gradually reaching 9 m min−1 for 10 min each day for 5 days, prior to the CHI procedure. Based on our previous studies, the following moderate-exercise training protocol (reaching about 60% of maximal oxygen consumption) was used because it has many beneficial effects on brain functions (Huang et al. 2006; Wu et al. 2007, 2011; Liu et al. 2008, 2009). Mice in the early exercise groups (starting 2 days after CHI as our standard exercise regime) were placed on the treadmill with an initial speed of 9 m min−1 that was progressively increased to 13.5 m min−1 for 1 h per day for 14 days. To rule out the possible daily handling effects, sedentary control mice were placed on the treadmill for 1 h per day without receiving any exercise for 14 days. One day after the last exercise session, mice were killed by CO2 inhalation and then perfused with cold PBS. The brain was quickly removed and fixed with 4% paraformaldehyde in 0.1 m phosphate buffer for 48 h at 4°C.
Immunostaining of neurons, microglia, BDNF and MKP-1
Fixed brain tissues were immersed in 30% sucrose solution and embedded in cryo-embedding medium. Each brain specimen was cut into an average of 50 coronal sections with a thickness of 20 μm, between −1.46 and −2.46 mm from bregma (Paxinos & Franklin, 2004). Frozen tissue sections were immersed in 3% H2O2 to block the endogenous peroxidase. After overnight incubation at 4°C with either anti-NeuN antibody (1:500 dilution; Chemicom, Temecula, CA, USA) or anti-Iba1 antibody (1:1500 dilution; Wako Chemical, Osaka, Japan), tissue sections were incubated with secondary antibodies, either biotinylated horse antimouse IgG or goat anti-rabbit IgG antibodies (Vector Laboratories, Burlingame, CA, USA). The avidin–biotin peroxidase (ABC; Vector Laboratories) and substrate of 3,3′-diaminobenzidine tetrahydrochloride (DAB, Sigma-Aldrich) were used for colour development. Either primary or secondary antibody was omitted for the detection of non-specific binding.
In order to quantify labelled cells, a stereology protocol was used by an investigator who was blinded to the experimental conditions (Wu et al. 2011). The number of neurons and microglia was counted in every sixth section (120 μm intervals). Using a ×10 objective (Plan-NEOFLUAR; Zeiss, Oberkochen, Germany), NeuN+ cells larger than 7 μm were counted as neurons, while Iba1+ cells larger than 4 μm were counted as microglia. The regions of primary interest included the cortex and the hippocampal CA1 region that were directly under the impact site and their counter-regions in the right hemisphere. In the hippocampal CA1 area, only NeuN+ cells in the pyramidal cell layer were counted. The cell density in each defined region (0.5 mm × 1 mm) was represented by the average of counted cells in eight sections.
To visualize BDNF and MKP-1, tissue sections containing the hippocampal region were treated with a blocking buffer (Dako, Carpinteria, CA, USA), and subsequently incubated overnight with either anti-BDNF antibody (1:200 dilution, N20; Santa Cruz Biotechnology, Santa Cruz, CA, USA) or anti-MKP-1 antibody (1:100 dilution; C19; Santa Cruz) at 4°C. The immunofluorescence staining was achieved by incubation with Alexa Fluor 488-conjugated goat anti-rabbit IgG (1:200 dilution; Invitrogen, Carlsbad, CA, USA) for 2 h. The cell nucleus was counterstained with 4′,6-diamidino-2-phenylindole.
Single-cell labelling of pyramidal neurons in the hippocampal CA1 region
Single-neuron labelling with the fluorescent dye Lucifer yellow was carried out according to methods described in detail previously (Pace et al. 2002). Briefly, anaesthetized mice were perfused from the left ventricle with 20 ml of artificial cerebrospinal fluid (mm: 117 NaCl, 4.7 KCl, 2.5 CaCl2, 11 glucose, 1.2 MgCl2, 25 NaHCO3 and 1.2 NaH2PO4, pH 7.4) followed by 20 ml of 4% paraformaldehyde and 10 ml of 4% sucrose, both in 0.1 m phosphate buffer. Brain specimens were rapidly removed and placed in 4% paraformaldehyde overnight at 4°C. Coronal sections (200 μm thick) passing through the impact site were viewed under differential interference contrast optics to identify neuron soma in the left hippocampal CA1 region. The pyramidal neuron soma was impaled with a glass electrode filled with saturated Lucifer yellow. The fluorescent dye was delivered iontophoretically over a 15 min period to fill the entire neuron. A stack of fluorescent images (175 μm × 175 μm) covering the lateral dendrites of a labelled pyramidal cell was taken from a high-sensitivity CCD camera (CoolSNAP fx; Photometrics, Tuson, AZ, USA) attached to a fluorescence microscope (E600FN; Nikon, Tokyo, Japan) with a ×40 objective lens (NA = 0.8, water immersion). These images were compiled together using ImageJ software (National Institutes of Health, Bethesda, MD, USA), and fluorescent dendrites in a compiled image were traced using the NeuronJ plug-in.
Data are expressed as means ± SEM. Commercial statistical software was used to determine statistical significance (GraphPad, San Diego, CA, USA). The NSS results were analysed by repeated-measures ANOVA. Results of ORT and numbers of NeuN+ and Iba1+ cells were analysed by one-way ANOVA, followed by Newman–Keuls post hoc multiple comparisons analysis. A value of P < 0.05 was considered to be a statistically significant difference. Two-way ANOVA was used to analyse the effects of exercise and triptolide on CHI mice using Newman–Keuls post hoc multiple comparisons analysis.
Effects of CHI on the behavioural tests
After receiving a CHI impact at the left cerebral hemisphere, the mouse was subjected to two different behavioural tests, i.e. NSS and ORT. The NSS tests were performed repeatedly to evaluate the recovery of brain injury in the same animal, while ORT was performed only once for each animal to avoid possible confounding factors. The CHI mice showed a 4.2 ± 0.3 NSS 1 h after the impact, which rapidly recovered to a 2.6 ± 0.1 NSS at 2 days and remained the same over the next 2 weeks (Fig. 1A). The sham-operated mice received an NSS of 0 for successfully passing all NSS test items. The traumatized mice demonstrated normal seeking behaviour and startle reflex and showed no walking impairments 2 days after the impact. However, at 2 days they were typically still unable to balance on the cylindrical rod (5 mm in diameter) or the rectangular beam (7 mm in width).
There was no significant difference in the total time spent exploring the two identical objects or the exploration of any one of the objects during the baseline assessment between the CHI group and sham control animals, indicating that the injury did not affect the motor ability for exploration. In the baseline test, all groups spent approximately 50% of the time exploring the two identical objects (Fig. 1B). However, while the control group spent much of their exploring time (71.5 ± 3.2%) with the new object, the CHI groups spent equal exploring time on both novel and familiar objects, even at 2 weeks after the impact, indicating a persistent loss of recognition memory after CHI. Overall, CHI immediately caused a high NSS that partly recovered later; however, it persistently impaired the recognition memory from the beginning.
Effects of CHI on neuronal loss and activation of microglia in the cortex and the hippocampus
Although CHI might cause widespread brain injuries, we focused primarily on both ipsilateral (left) and contralateral (right) sides of the cortex and the hippocampus. As expected, the NeuN+ cells (neurons) underwent spatial and temporal losses in both hemispheres (Fig. 2; left cortex, P < 0.0001; right cortex, P < 0.0001; left CA1, P < 0.0001; and right CA1, P < 0.0001). The first phase of neuronal loss happened in less than 2 days on both sides of the cortex and the hippocampus. In addition, there was a very pronounced second phase of neuronal loss that happened directly underneath the impact site (the left cortex and the left dorsal hippocampus), i.e. the local brain tissue was virtually non-existent when examined 2 weeks after the injury. In contrast, minimal neuronal losses happened in the deep brain areas of the hippocampus, i.e. the hippocampal granular layer of dentate gyrus (GrDG) and the dentate gyrus (DG) regions (Supplementary Fig. S1).
Microglia are the major immune cells in the brain, and their number and activation indicate neuroinflammation. The activated microglia are characterized by increased soma size, irregular shape and intensified Iba1 staining. While most microglia remained resting in the sham-operated and 2 days post-CHI groups, they became activated 9 or 16 days after CHI (Fig. 3A). Quantitative results also demonstrated that the numbers of microglia increased progressively from day 9 to 16 after CHI in both hemispheres (Fig. 3B and C; left cortex, P < 0.001; right cortex, P < 0.0001; left CA1, P < 0.0001; and right CA1, P < 0.05). The tissues directly underneath the impact point disappeared 16 days after CHI, which resulted in a reductio in the microglia count (Fig. 3B). Overall, CHI induced a broad spectrum of brain pathophysiological changes, including rapid and progressive neuronal loss and delayed activation of microglia, especially near the impact site. However, these pathophysiological changes did not correspond temporally to the post-CHI performance of NSS or ORT.
Effects of postinjury exercise on the progression of CHI
As previously mentioned (see Effects of CHI on the behavioural tests), the mice showed partial recovery of the NSS and a normal capacity for movement 2 days after CHI. We chose this time point to start treadmill exercise at moderate intensity (the common exercise protocol usually applied to train healthy mice). Our standard exercise protocol did not further reduce the NSS (Fig. 4A). For the ORT, although neither CHI nor exercise affected the baseline trial performance, exercise almost completely reversed the CHI-induced memory impairment on day 9 (P < 0.001) and day 16 (P < 0.01); the mice could distinguish a novel object from an object to which they were exposed 4 h earlier (Fig 4B). Exercise did not affect the memory of mice in the sham-operated group. However, if the exercise intervention was started later, i.e. started 9 days post-CHI, there was no restoration of the ability to recognize a previously seen object (P > 0.05). Therefore, moderate treadmill exercise started shortly after CHI effectively restored the previously lost recognition memory.
Although our standard exercise protocol did not affect the number of neurons in sham-operated mice, it effectively prevented the progressive loss of neurons in the cortex and the hippocampus (Fig. 5). This neuroprotective effect was very remarkable, especially in the left cortex and hippocampus (locations directly underneath the impact spot); most of the neurons alive on day 2 after CHI survived at least until the end of the experiment, i.e. day 16 after CHI. Two-way ANOVA comparing the sham-operated and day 16 CHI groups revealed that both CHI (left cortex, P < 0.001; right cortex, P < 0.001; left CA1, P < 0.001; and right CA1, P < 0.001) and exercise (left cortex, P < 0.001; right cortex, P < 0.001; left CA1, P < 0.001; and right CA1, P < 0.001) significantly affected neuronal density, and there were interactions between these two factors (left cortex, P < 0.001; right cortex, P < 0.001; left CA1, P < 0.001; and right CA1, P < 0.01). Post hoc analyses indicated that the effect of exercise on the number of neurons was derived from the CHI group, but not the sham group. If the exercise intervention began 9 days post-CHI, it was not as effective in retarding the progressive loss of neurons (P > 0.05). As expected, exercise training did not affect the deeper brain regions, which showed minimal neuronal losses following CHI (data not shown).
We then investigated whether postinjury exercise was capable of suppressing the CHI-induced neuroinflammation. Delayed increases of Iba1+ cells (activated microglia) were obvious in the cortex and dorsal hippocampus of mice following the CHI, while these increases were almost absent in mice that started exercise 2 days following CHI (Fig. 6). For example, on day 16 both CHI (left cortex, P > 0.05; right cortex, P < 0.001; left CA1, P < 0.001; and right CA1, P < 0.001) and exercise (left cortex, P < 0.01; right cortex, P < 0.001; left CA1, P < 0.001; and right CA1, P < 0.001) affected the density of microglia, with significant interactions between these two factors except in the right CA1 (left cortex, P < 0.001; right cortex, P < 0.05; left CA1, P < 0.001; and right CA1, P > 0.05). Post hoc analyses indicated the effect of exercise on the number of Iba1+ cells was derived from the CHI group, but not the sham-operated group. If the exercise intervention was initiated late following the CHI, this was somewhat effective in retarding neuroinflammation (left cortex, P < 0.001; right cortex, P < 0.05; left CA1, P < 0.01; and right CA1, P > 0.05).
Finally, we used a single-cell labelling technique to visualize directly the CHI-induced physical damage and the progressive recovery/loss of pyramidal neurons located in the left side of the hippocampal CA1 area. Closed-head injury not only caused progressive neuron loss but also drastically changed the morphology of surviving neurons in a time-dependent way, i.e. a few neurites remained on day 2, some neurites regenerated on day 9, and yet almost all pyramidal neurons in this region disappeared on day 16 (Fig. 7A–D). The early exercise intervention boosted the neurite regeneration (Fig. 7E), while the delayed exercise intervention mainly prevented the ultimate loss of pyramidal neurons (Fig. 7F). Overall, postinjury exercise at moderate intensity, when commenced at an early time, not only prevented the CHI-induced neuron loss and delayed neuroinflammation but also restored the CHI-impaired recognition memory. However, a similar exercise intervention was less effective in retarding neuronal loss and neuroinflammation when it was initiated later. Moreover, when initiation of the exercise intervention was delayed, it was totally ineffective in reversing the deficits of object recognition memory.
Involvement of BDNF and MKP-1 in the beneficial effects of exercise on CHI
CHI involves both acute phase and progressive phase injuries of the brain, and we found that early postinjury exercise benefited both. We further investigated the possible involvement of BDNF and MKP-1 in the beneficial effects of exercise on CHI. Immunofluorescence staining revealed that CHI abolished the expression of BDNF and MKP-1 in the hippocampus, whereas early exercise intervention for 2 weeks reinstated the CHI-induced decrease in the levels of these proteins (Fig. 8A). This early exercise protocol did not increase hippocampal BDNF levels in sham-operated mice.
In order to examine whether MKP-1 was needed for the exercise benefits, triptolide (an MKP-1 synthesis inhibitor) was administered orally before each bout of exercise. The MKP-1 immunostaining results showed that the exercise-induced restoration of MKP-1 expression in CHI mice was blocked by administration of triptolide (Fig. 8B). Moreover, triptolide also completely blocked the restorative effects of exercise on the recognition memory deficits in CHI mice (Fig. 9; day 9 post-CHI, P < 0.001; and day 16 post-CHI, P < 0.001). Triptolide alone did not affect the recognition memory in either sham-operated (P > 0.05) or CHI mice (P > 0.05).
In contrast, triptolide as an anti-inflammatory agent (nuclear factor-κB inhibitor) might affect neuronal survival and activation of microglia. As shown in Fig. 10, triptolide administration alone prevented the CHI-induced neuron loss in the left cortex (P < 0.01) and left hippocampal CA1 area (P < 0.01). Similar but less dramatic responses were also noticed in the right hemisphere (Supplementary Fig. S2). However, triptolide hampered the exercise-induced neuronal protection in CHI mice (Fig. 10; left cortex, P < 0.001; and left CA1, P < 0.001).
We then examined the effect of triptolide on CHI-induced activation of microglia in exercised and sedentary groups. Our results showed that CHI-induced activation of microglia was significantly blocked by triptolide alone in the right hemisphere with intact cortex (Supplementary Fig. S3); however, the prevention of increased numbers of activated microglia in the CHI mice by exercise was not further reduced by co-administration of triptolide (Fig. 11 and Supplementary Fig. S3). Therefore, although triptolide alone effectively retarded the CHI-induced neuronal loss and neuroinflammation, it blocked many of the beneficial effects of exercise on CHI, especially the exercise-induced reversal of memory deficits.
Closed-head injury induced pronounced brain damage, including immediately elevated NSS, rapid loss of recognition memory, progressive neuronal loss and activation of microglia. An early and suitable postinjury exercise programme in mice mitigated most of the debilitating consequences of CHI. Our early exercise regime reversed the CHI-induced rapid loss of recognition memory and blocked most of the progressive loss of neurons and the increase in activated microglia.
In our view, depending on the location relative to the impact site, the injured neurons may be divided into three general categories according to the severity of injury, i.e. severely, moderately and minimally injured neurons. Although severely injured neurons died either immediately or soon afterwards, minimally injured neurons not only survived the initial impact but also reconnected some damaged neural circuits, thus explaining the rapid reduction of NSS score in the first couple of days. This viewpoint is consistent with the current consensus that CHI is characterized by an immediate high rise of NSS, which partly recovers in the first couple of days, with thin-beam walking and balance remaining deficient (Shapira et al. 1988, 1993; Chen et al. 1996). However, the moderately injured neurons might either recover or expire depending on whether they were provided in time with favourable conditions for repair. Our early exercise intervention was effective in recovering the lost object recognition memory because it helped the repair of moderately injured neurons at the right time. Additionally, this treatment prevented further progression of brain tissue damage. In contrast, if a similar exercise intervention was started on day 9 after the impact, many moderately injured neurons were probably too weak to recover or at least incapable of re-establishing their original connections. As a consequence, even though this delayed exercise regime impeded the late-phase neuronal loss, it was ineffective in recovering the impact-hampered object recognition memory.
Although neuronal loss is a good indicator of function in neurodegenerative diseases, this parameter in a traumatic brain injury model apparently does not faithfully reflect the time-dependent loss of brain functions, such as the performance of NSS and ORT. In principle, a functional deficit should be reflected by the loss of neural circuits involved in that particular function. Thus, severely hampered performance in a behavioural task may be accompanied by severe damage of neural circuits but not a drastic reduction of neuronal number. In the acute phase of CHI, many neurons with intact somata may be totally dysfunctional owing to instant depolarization (Yang et al. 1993; Reeves et al. 2000) and diffuse axonal injury (Feliciano et al. 2008). This would explain the rapid rise of NSS and the rapid loss of ORT. The spontaneous recovery of NSS over a couple of days could be due to repolariztion of certain impact-depolarized neurons. However, the functional recovery of learning and memory apparently need the reconstruction of an intact neural network in the hippocampus, which may take several days even in favourable conditions, such as with the help of early exercise intervention.
Despite the fact that exercise exerts many beneficial effects on the brain (Vaynman & Gomez-Pinilla, 2005; Leem et al. 2011), whether postconcussion patients should take some forms of exercise remains unsettled (McCrea et al. 2009). Postinjury exercise in brain traumatized rats has raised concerns owing to the possibility of competition for metabolites and energy supply between damaged brain tissues requiring repair and the activity of neural tissues required for performing the exercise (Yoshino et al. 1991; Ginsberg et al. 1997; Moore et al. 2000). Additionally, although postimpact exercise improves performance of the Morris water maze and upregulates hippocampal BDNF, these beneficial effects are disrupted when exercise is administered too early, i.e. between days 0 and 6 after the traumatic brain injury (Griesbach et al. 2004, 2007, 2009). In contrast, CHI mice in this study were mostly asymptomatic 2 days after the impact, i.e. they showed a low NSS score by demonstrating normal seeking, startle reflex and walking behaviours. Even though the CHI damage of the brain was evident at this time point (indicated by deficient recognition memory and loss of BDNF/MKP-1 in brain areas close to the impact site), the neuronal loss was relatively mild and the neuroinflammation was still dormant. In comparison, the exercise intervention was not as effective if it started too late, e.g. on day 9 postimpact in our CHI model. Therefore, 2 days postimpact was the right time to initiate the exercise with moderate intensity (to avoid possible systemic energy depletion and overwhelming disturbances in cardiovascular parameters) in our animal CHI model. The apparent discrepancy between early studies and the present one may be due to differences in the animals used (∼300 g rats vs. ∼35 g mice), the impact induction/severity (lateral fluid percussion of ∼250 kPa vs. CHI with initial NSS of ∼4.5), exercise starting date (day 0 vs. day 2) and exercise protocol/intensity (voluntary wheel running vs. moderate treadmill running).
Traumatic brain injury-induced damage usually occurs in two phases (Tang et al. 1997a, b; Israelsson et al. 2009). The acute phase involves axonal breaks, with minimal neuron death in the damaged area. This disrupts the neuronal circuitry in this region. The late phase is characterized by cumulative neuronal loss associated with regional neuroinflammation. In our CHI model, the benefits of postinjury exercise mostly happened during the acute phase of the injury. The exercise prevented, or at least significantly mitigated, the progression of the acute-phase injury, allowing neuronal circuits to reconnect. Thus, the animals had recognition memory impairments at day 2 postimpact, but following a week of exercise the recognition memory ability had returned to normal. The exercise-induced recovery of brain BDNF and MKP-1 in post-CHI animals (Fig. 8) could explain the rapid network repair, because BDNF facilitates axon branching and remodelling via upregulation of MKP-1 to destabilize the axonal microtubules (Jeanneteau et al. 2010). This is the most probable explanation for the inhibition by triptolide of the benefits of exercise on recognition memory deficits (Fig. 9), because triptolide prevents the synthesis of MKP-1. Apparently, the newly repaired axon branches were able to find the right track back to the original target cell and successfully reconstruct a functional neural circuit. Brain-derived neurotrophic factor plays a major role in the beneficial effects of exercise in rats following traumatic brain injury. Voluntary exercise is effective in recovering the hippocampal BDNF and neuroplasticity-related pathways, and this is abrogated by pre-administration of TrkB-IgG to block the BDNF receptor (Griesbach et al. 2004, 2009). The benefits of exercise on the recovery from brain injury depend on the timing of when the exercise is implemented. Severely injured animals may need to rest for longer before they can benefit from postinjury exercise, because the time window for the exercise-induced increase in BDNF and neuroplasticity-related pathways is dependent on the severity of the injury (Griesbach et al. 2007).
It is interesting that the exercise-induced elevation in BDNF was more pronounced in CHI animals with low basal BDNF levels than in normal animals with relatively high basal BDNF levels. Similar phenomena have been observed in other disease or injury animal models. That is, while exercise interventions have minimal effects in control animals, they significantly increase BDNF levels in the brain of aged rats, kainic acid-treated rats and lipopolysaccharide-treated mice (Gobbo & O’Mara, 2005; O’Callaghan et al. 2009; Wu et al. 2011). Unlike the hippocampal TrkB level, the hippocampal BDNF level in normal animals increases only transiently after treadmill running and voluntary wheel running; it returns to the original value 1 day after the termination of exercise (Huang et al. 2006; Liu et al. 2009). Given that exercised animals in the present study were killed 1 day after the last run in order to avoid the acute effects of exercise, the exercise-induced BDNF elevation mainly happened in CHI mice.
Additionally, the beneficial effects of postinjury exercise in our CHI model included the prevention of neuroinflammation and neuronal loss in the late phase. In our opinion, these effects were likely to be secondary to the primary effects of exercise on the acute-phase injury, because an exercise-accelerated repair of the neuronal circuit would lead to a reduction of the neuronal loss and the later neuroinflammation. Nevertheless, the postinjury exercise might reduce apoptosis and neuroinflammation to ameliorate the late-phase injury as well. Treadmill exercise in a rat model of traumatic brain injury alleviated the impairment of a step-down avoidance task, decreased the hippocampal expression of pro-apoptotic molecules (caspase-3 and Bax), and increased the expression of the anti-apoptotic protein Bcl-2 (Kim et al. 2010a). However, the benefits become minimal if exercise is commenced after the beginning of neuroinflammation and scar formation (Myer et al. 2006; Seo et al. 2010).
Triptolide is a reagent that inhibits both the biosynthesis of MKP-1 and the activation of nuclear factor-κB (Qiu et al. 1999; Chen et al. 2002). In principle, this drug has dual effects in modulating inflammation, i.e. it is pro-inflammatory by inhibiting the negative regulator MKP-1 and anti-inflammatory by inhibiting the pro-inflammatory factor nuclear factor-κB. Results from our triptolide experiments indicated that triptolide treatment alone reduced both neuronal loss and activation of microglia in the CHI mice, indicating a predominantly anti-inflammatory role in our experimental conditions. Consistently, triptolide promotes repair of traumatic spinal cord injuries by inhibiting astrogliosis and inflammation (Su et al. 2010). In the present study, however, triptolide administered after day 2 postimpact was unable to recover the CHI-induced loss of recognition memory. Moreover, co-administration of triptolide with the exercise intervention blocked the beneficial effects of exercise on the object recognition memory, indicating that it had detrimental effects on axonal repair due to an early inhibition of the BDNF–MKP-1 pathway. Given that neuroinflammation happened in the late phase of our CHI model, it would be interesting to investigate whether a late administration of triptolide would further improve the beneficial effects of early postinjury exercise.
Our study fits well with the ABCDE (Awakening and Breathing Co-ordination, Delirium monitoring and Exercise/Early mobility) model used in the intensive care setting for prevention of brain dysfunction (Vasilevskis et al. 2010). We concur with the general recommendation in clinical practice that concussion patients should rest until they are asymptomatic (Leddy et al. 2007), but once they reach this stage we believe that a proper exercise intervention programme should be initiated. Such interventions could provide significant benefits and dramatically reduce patients’ recovery period, as seen in our model. However, one must be very cautious in extrapolating the basic science data to clinical applications, because the final outcome may be critically dependent on the severity of the impact, the time of commencing the exercise intervention, the intensity and duration of the exercise regime, and the parallel drug treatments. In particular, it is essential to establish an applicable method for evaluating the rehabilitative exercise window time in humans with traumatic brain injury, e.g. measuring the Glasgow Coma Scale, the metabolic changes and the serum BDNF levels.
The major experimental design was by M.-F.C., Y.-M.K. and C.J.J.; the experiments were carried out by M.-F.C. and T.-Y.H.; the behaviour tests were formulated by L.Y.; and the manuscript drafts were written by M.-F.C., H.-i.C. and C.J.J. All authors approved the final version of the manuscript.
We thank Mr Ju-San Chen in the Department of Mechanical Engineering for constructing the closed-head injury set-up and Dr Paul R. Saunders for critically reading the manuscript. This study was supported by grants from the National Science Council, Taiwan (NSC 98-2320-B-006-019-MY3 and NSC 98-2320-B-006-028-MY3).