Seventy-nine male Wistar rats (Japan SLC, Tokyo, Japan) weighing 220–260 g (approx. 8 weeks of age) at the time of surgery were used in this study. The rats were randomly assigned to the ischaemia-non-exercised control (IC; n = total 37; n = 6 at 1-, 3-, 5-, 7-, 14-, and 28-day time points; one rat expired within 24 h of surgery), sham-exercise (SE; n = 5) and ischaemia-exercise (IE; n = total 37; n = 6 at 1-, 3-, 5-, 7-, 14-, and 28-day time points; one rat expired within 24 h of surgery) groups before surgery. The SE rats underwent sham occlusion, the IC and IE rats underwent left MCA occlusion, and the SE and IE rats ran on a treadmill. Two rats died within 24 h after insult because of severe cerebral infarction, possibly because of atypical MCA variations. One of the rats that died belonged to the IE group, and the other was from the IC group. These rats were excluded from the study. No rat died between 24 h after the stroke and end of the experiments as a result of treadmill exercise or disease. Seventy-seven rats completed the experiment. The animals were socially housed (2–3 rats per cage) under a 12 h reverse light/dark cycle in clear Plexiglas cages with water and food available ad libitum. The experimental protocols accorded with established guidelines determined by the Animal Care Committee of the Ethics Board of the Institute of Laboratory Animal Sciences of Kagoshima University.

The animals were anaesthetized via the injection of 4% chloral hydrate (10 mL kg⁻¹, intraperitoneally) and were given further doses as necessary to ensure adequate anaesthesia during surgery. Rectal temperature was monitored throughout the surgical procedures and was maintained at 37 °C using a heating blanket controlled by an electronic temperature controller (KN-474; NATUME, Tokyo, Japan). Stroke was induced by 90 min left MCA occlusion using an intraluminal filament (^{Longa et al. 1989}). Briefly, a midline incision was made, and the left common carotid artery (CCA), the external carotid artery (ECA) and the internal carotid artery (ICA) were exposed. The distal ECA branch was coagulated completely. The CCA and ECA were then tied with a white thread. A monofilament (a 4-0 nylon suture with a blunted tip of 0.2–0.3 mm in diameter and 16 mm in length) was inserted into the left CCA via an arteriotomy. It was then passed up the lumen of the ICA into the intracranial circulation and lodged into the narrow proximal anterior cerebral artery, blocking the MCA at its origin. After 90 min of MCA occlusion, reperfusion was established by withdrawal of the filament. Regional cerebral blood flow was measured by a laser Doppler flowmeter (ALF-21; Advance Co, Tokyo, Japan) before and during MCAO. A probe in the shape of a flat rectangular sheet (7.5, 3.5 and 1.0 mm in length, width and thickness respectively) was placed over the ischaemic side in the natural pocket formed between the temporal muscle and the lateral aspect of the skull. All animals exhibited significantly reduced blood flow (a >20% reduction compared with the pre-ischaemic baseline) during MCA occlusion, and none were excluded from the study because of lack of reduction of blood flow. For sham occlusion, the left CCA was exposed following general anaesthesia similar to the procedure for stroke induction. The ICA was separated at the junction of the left ICA and the left ECA but was not occluded. Body temperature was maintained as described above.

The rats in both the SE and IE groups underwent treadmill running, which began 24 h after the surgery. The rats in the IC group were allowed to move freely in their cages, but no additional treadmill running was employed. The rats in the IE group underwent treadmill running for 20 min a day every day for a maximum of 28 days. The rats in the SE group ran on a treadmill for 28 consecutive days after surgery. A motorized treadmill (Rat runner, RR-1200; AKK, Shimane, Japan) that forced the rats to run via an electric stimulation system installed on the rear floor was used. All animals ran at a speed of 3 m min⁻¹ for the first 3 days after surgery, 8 m min⁻¹ from 4 to 6 days after surgery and then 13 m min⁻¹ on subsequent days with an inclination of 0°. Two animals were housed per cage, and all animals were allowed to recover in their cages in a warm environment with food and water supplied ad libitum. The exercises were performed at normal room temperature, in a normal light setting and with normal room noise. To monitor the stress induced in the rats by treadmill running, body weight was measured periodically (^{Ding et al. 2004b}). None of the animals were excluded from the study because of an unwillingness to run.

Animals from the SE, IE and IC groups were examined with a motor test paradigm (limb placing: motor behaviour test) and for neurological deficits before and 1, 3, 5, 7, 14 and 28 days after surgery. All rats were tested three times on each trial day.

In the motor behaviour test, which was modified from the scoring system of ^{Fenny et al. (1982)}, the locomotion of the rats was evaluated pre-operatively and post-operatively using a beam-walking task with an elevated narrow beam (150 cm long × 2.5 cm wide). The worst score (‘0’) was given if the rat was unable to traverse the beam and could neither place the affected limbs on the horizontal surface nor maintain balance. A score of ‘1’ was given if the rat was unable to traverse the beam or to place the affected limbs on the horizontal surface of the beam but was able to maintain balance. A score of ‘2’ was given if the rat was unable to traverse the beam but placed the affected limbs on the horizontal surface of the beam and maintained balance. A score of ‘3’ was given if the rat used the affected limbs in less than half of its steps along the beam. A score of ‘4’ was given if the rat traversed the beam and used the affected limbs to aid more than 50% of its steps along the beam. A score of ‘5’ was given if the rat traversed the beam normally with no more than two foot slips. The day prior to surgery, the animals in all groups underwent the motor behaviour test to ensure that their performance score was 5.

A neurological grading system with a 5-point scale (0–4) described by ^{Menzies et al. (1992)} was used: 0 = no apparent deficits; 1 = right forelimb flexion; 2 = decreased grip of the right forelimb while tail pulled; 3 = spontaneous movement in all directions while right circling only if pulled by tail; 4 = spontaneous right circling.

Of the two observers rating motor behaviour and performing the neurological assessments, one was not informed which rats belonged to the SE, IC, and IE groups, and his scores were mainly employed for the subsequent analyses.

The rats in the IC and IE groups were randomly killed at 1, 3, 5, 7, 14 and 28 days (n = 6 at each time point) after the induction of ischaemia. The rats in the SE group were killed at 28 days after surgery.

All rats were deeply anaesthetized via an injection of 4% chloral hydrate (10 mL kg⁻¹, intraperitoneally) before being killed by decapitation. The brain was carefully removed and cut into six 2-mm thick coronal sections from its frontal tip using a brain slicer. The fresh brain slices were then immersed in a 2% solution of 2,3,5-triphenyltetrazolium chloride (TTC) in physiological saline at 37 °C for 5 min and were then fixed in 4% paraformaldehyde in 0.1 M phosphate buffer (pH 7.4) at 4 °C overnight, dehydrated, embedded in paraffin, and subjected to histological and immunohistochemical analyses.

The TTC-stained sections were used to determine the infarct volume in ischaemic rats. Non-infarcted ipsilateral areas and the areas contralateral to the occluded side were measured using Scion Image software BETA 4.0.3 (Scion Corp., Frederick, MD, USA). The total infarct area was multiplied by the thickness of the brain sections to obtain infarct volume. To minimize the error introduced by oedema and liquefaction after infarction, an indirect method for calculating infarct volume was used (^{Swanson et al. 1989}). The non-infarcted area in the ipsilateral hemisphere was subtracted from that in the contralateral hemisphere, and infarct volume was calculated using the following formula: corrected percentage of infarct volume = (contralateral hemispheric volume−ipsilateral non-infarcted volume)/contralateral hemispheric volume.

Coronal sections (5-μm-thick) were stained with haematoxylin and eosin (HE) for histological evaluation after MCA occlusion. Immunohistochemistry was performed by the indirect immunoperoxidase method. After deparaffinization and hydration, endogenous peroxidase was blocked with methanol containing 0.9% hydrogen peroxide for 10 min. After three rinses (10 min each) with 50 mM phosphate buffered saline (PBS, pH 7.6), the sections were blocked with 10% skimmed milk for 20 min at room temperature before being individually incubated at 4 °C overnight with affinity-purified rabbit anti-MK antibody (^{Yoshida et al. 1995}), rabbit anti-NGF antibody (1 : 300; Alomone, Jerusalem, Israel), and rabbit anti-caspase-3 antibody (1 : 200; Santa Cruz, Santa Cruz, CA, USA) for the evaluation of apoptosis and goat anti-platelet-endothelial cell adhesion molecule (PECAM-1) antibody (1 : 100; Santa Cruz) for the evaluation of angiogenesis. After three more rinses (10 min each) with PBS, the sections were reacted with goat anti-rabbit IgG conjugated to peroxidase-labelled dextran polymer (EnVision; Dako, Carpinteria, CA, USA) for 60 min. After rinsing the sections with PBS, immunoreactivity was visualized with diaminobenzidine/peroxide. In addition, the negative controls were rigorously examined to confirm that the observed MK and caspase-3 immunoreactivity were not the result of non-specific immunostaining.

Sections of rat brain tissues were double-immunostained as described before (^{Sakakima et al. 2004a}). The slides were then incubated with rabbit anti-NGF polyclonal antibody (1 : 300; Alomone) and mouse anti-glial fibrillary acidic protein (GFAP) monoclonal antibody (1 : 500; Chemicon, Billerica, MA, USA) or mouse anti-MAP-2 monoclonal antibody (1 : 200; Chemicon) overnight at 4 °C. After being washed with PBS, the slides were incubated with Alexa Fluor 568-labelled anti-rabbit IgG (Santa Cruz) and Alexa Fluor 488-labelled anti-mouse IgG diluted 1 : 200 in PBS for 1 h. After the slides were extensively washed with PBS, mounted with 90% glycerol, and examined under an Axioskope microscope (Carl Zeiss, Oberkochen, Germany).

The areas of immunoreactive cells in representative sections were measured in the motor cortex on the ischaemic side (Fig. 1). The same procedure was performed in the SE group. The areas of MK, NGF, PECAM-1 and caspase-3-positive cells were assessed quantitatively in three coronal sections (the sections began at intervals of 200 μm starting from the point 2 mm posterior to bregma) (2 mm × 2 mm in each field) within the frontoparietal cortex around the lesion. The areas of MK-, PECAM-1- and caspase-3-labelled cells were determined with a computer-assisted image analyzer using Scion Image software BETA 4.0.3 (Scion Corp.). The number of NGF-immunoreactive cells was counted with a computer-assisted image analyzer using Adobe® Photoshop® 5.0 (Adobe Systems, San Jose, CA, USA). All histological analyses were performed in a blind fashion.

Motor behaviour and neurological scores were expressed as median with quartiles. Values for other variables are presented as means and standard deviations. Statistical analyses were performed using STATVIEW version 5.0 software (StatView; SAS Institute, Cary, NC, USA). Motor behaviour was evaluated, and neurological assessments were performed using the score for each trial. Two-way factorial ANOVA analyses for different groups (IE, IC and SE groups) were used on training days (before, 1, 3, 5, 7, 14 and 28 days) to assess motor behaviour, neurological deficits, body weight, infarct volume, and areas of immunolabelling for MK, NGF, caspase-3, and PECAM-1, followed by one-way ANOVA and post hoc comparisons (with Dunnett’s test) when required. Statistical significance was accepted at the level of P<0.05.

Changes in motor behaviour and the results of neurological assessments are shown in Figure 2. In the evaluation of motor behaviour, the score of all animals before surgery was 5 (Fig. 2a). The rats of the IE and IC groups exhibited uniform, severe motor impairment at 1 day after the induction of ischaemia, as evaluated by beam-walking ability. Functional recovery was recorded at 1, 3, 5, 7, 14 and 28 days after the surgery. The rats of the IE group exhibited continuous functional recovery in the beam-walking test during the 28 days of the study. The score for the rats of the SE group was 5 throughout the 28-day period of post-operative examination, while that in the IC group was 0–1 for the first 7 post-operative days and 1–2 for the remaining 28 days. The improvement in motor behavioural score in the IE group occurred earlier than that in the IC group from 7 days after the induction of ischaemia. Two-factor factorial ANOVA revealed significant effects of day (F_5,65 = 15.125, P<0.0001), group (F_2,65 = 565.875, P<0.0001) and the interaction between day and group (F_10,65 = 5.452, P<0.0001). Subsequent one-way ANOVA (day) revealed a significant difference among groups at 28 days (F_2,14 = 31.969, P<0.0001) but not at other time points. In particular, at 28 days after ischaemia, Dunnett’s post hoc test revealed a significant difference between the IE and IC groups (P<0.05).

On neurological assessment, the scores for all rats before surgery were 0 (Fig. 2b). After MCA occlusion, the rats were observed for neurological deficits, and their neurological scores were found to have increased. In both the IC and IE groups, their neurological deficits improved, and their neurological score approached 0 over time. The improvement in neurological score in the IE group occurred earlier than that in the IC group. Two-factor factorial ANOVA revealed significant effects of day (F_5,65 = 17.750, P<0.0001), group (F_2,65 = 140.536, P<0.0001) and the interaction between day and group (F_10,65 = 5.054, P<0.0001). A subsequent one-way ANOVA (day) revealed a significant difference among the groups at 28 days (F_2,14 = 3.73, P<0.001) but not at other time points (Fig. 2b). In particular, the score in the IE group was significantly improved compared with that in the IC group at 28 days after surgery (P<0.01).

The body weights of the animals in the three groups gradually increased after the induction of ischaemia, without significant differences among the groups (1 day after: IE, 206.4 ± 29.5 g, IC, 222.4 ± 31.9 g, 3 days: IE, 187.9 ± 21.5 g, IC, 204.1 ± 34.1 g, 5 days: IE, 185.0 ± 25.6 g, IC, 202.0 ± 36.8 g, 7 days: IE, 208.1 ± 27.5 g, IC, 208.4 ± 40.8 g, 14 days: IE, 251.3 ± 35.7 g, IC, 245.5 ± 35.8 g, 28 days: IE, 303.3 ± 38.0 g, IC, 314.3 ± 41.7 g), suggesting that the animals did not suffer stress during exercise.

Histological analysis with TTC staining was performed to determine the extent of brain infarction at 1, 3, 5, 7, 14 and 28 days after ischaemia. Areas of damage were observed by HE staining, mainly in the cerebral cortex including the dorsolateral and lateral cortices, as well as in the lateral striatum (Fig. 3a,b). Although our analyses failed to detect any interaction between group and day (F_5,60 = 0.753, P=0.58), a subsequent one-way ANOVA (day) revealed a significant difference among groups in infarct volume at 28 days (F_1,10 = 4.60, P=0.002) but not at other time points (data not shown). At 28 days, the infarct volume in the IE group (12.4 ± 0.8%) was significantly decreased compared with that in the IC group (19.8 ± 4.2%) (Fig. 3c). This 38% reduction was highly significant (P<0.01). The sham-operated rats exhibited no infarction (data not shown).

To elucidate the endogenous neuroprotective effects of motor exercise after stroke, brain tissues from ischaemic rats in the IE and IC groups subjected to a maximum of 28 days of treadmill exercise were processed by immunocytochemistry to assess the cellular expression of neurotrophic factors and angiogenesis in the territory supplied by the MCA. One to five days after stroke, MK immunoreactivity was detected in the peri-infarct region in the acute stage in the rats of the IE and IC groups, but was not detected on the contralateral side or in the brains of the rats of the SE group. Three days after ischaemia, MK immunoreactivity that was significantly higher in intensity in the IE group than in the IC group was observed in the peri-infarct region (Fig. 4a,b). At 14–28 days, no MK signal was detected in the ipsilateral or contralateral cortex. Two-factor factorial ANOVA revealed significant effects of day (F_5,60 = 32.43, P<0.0001), group (F_1,60 = 11.61, P=0.003), and the interaction between day and group (F_10,60 = 10.4, P=0.0001). A subsequent one-way ANOVA (day) revealed a significant difference among groups in MK expression at 3 days (F_1,11 = 7.71, P=0.018) but not at other time points (data not shown). In particular, at 3 days after ischaemia, Dunnett’s post hoc test revealed a significant difference between the IE and IC groups with regard to MK expression (P<0.05) (Fig. 4c).

Nerve growth factor-immunopositive cells were consistently detected in the ipsilateral (Fig. 5a,b,d,e) and contralateral (Fig. 5c,f) cerebral hemispheres in the IE (Fig. 5a–c) and the IC (Fig. 5d–f) groups at 28 days after ischaemia. In particular, NGF-immunopositive cells were markedly increased in number over a widespread region around the infarct on the ipsilateral side (Fig. 5a,d). At 28 days, the number of NGF-immunoreactive neurons in the cortex in the IE group had increased more than that in the IC group (Fig. 5a). Double immunostaining with anti-NGF (Fig. 5g,j, red) and anti-MAP2 (Fig. 5h, green), as neuronal markers, or anti-GFAP (Fig. 5k, green), a marker of reactive astrocytes, antibodies using Axioskope microscope analysis showed NGF in neurons and astrocytes at 28 days after ischaemia (Fig. 5i,l). The majority of the labelled cells were, however, pyramidal neurons in the IE and IC groups, as the area of NGF-MAP-2 merged staining (indicative of neurons; IE: 70 ± 3.2%, IC: 65 ± 0.7%) was larger than that NGF-GFAP merged staining (indicative of reactive astrocytes; IE: 30 ± 3.2%, IC: 35 ± 0.7%) in both groups after 28 days. Figure 4m shows a quantitative comparison of immunolabelled cells in the counted areas of the motor cortex in the IE and IC groups. Two-factor factorial ANOVA revealed significant effects of day (F_5,60 = 23.98, P<0.0001), group (F_1,60 = 54.49, P=0.003) and the interaction between day and group (F_10,60 = 3.24, P=0.01). A subsequent one-way ANOVA (day) revealed a significant difference among groups in number of NGF-labelled cells at 28 days (F_1,11 = 4.84, P<0.001) but not at other time points (data not shown). ANOVA analysis demonstrated a significant increase in the number of NGF-labelled cells after exercise (Fig. 5m).

Platelet-endothelial cell adhesion molecule immunocytochemistry was used to label brain microvessels. PECAM-1-immunopositive cells were consistently detected in the ipsilateral (Fig. 6a,b) and contralateral (Fig. 6c,d) cerebral hemispheres in all groups. In particular, PECAM-1-immunopositive cells were markedly increased in area over a widespread region around the infarct on the ipsilateral side. Quantitative analysis of frontoparietal PECAM-1-immunopositive cells was performed by counting the area of PECAM-1-immunostained microvessels. Two-factor factorial ANOVA revealed significant effects of day (F_5,60 = 53.82, P<0.0001), group (F_1,60 = 147.25, P<0.0001), and the interaction between day and group (F_10,60 = 26.1, P<0.0001). Subsequent one-way ANOVA (day) revealed a significant difference among groups in the area of PECAM-1-immunopositive cells at 3 days (P<0.001), 7 days (P=0.014) and 14 days (P=0.041), but not at other time points (data not shown). In particular, the area of PECAM-1-immunopositive cells was significantly increased in the cells around the infarct in the IE group at 3, 7 and 14 days after surgery (Fig. 6e).

Photomicrographs of caspase-3-positive cells are shown in Figure 7a,b. Caspase-3-immunopositive cells were detected in the infarct and peri-infarct regions after the induction of ischaemia in rats of the IE and IC groups, but not on the contralateral side nor in the brains of rats of the SE group. The area of caspase-3-positive cells was nearly zero mm² in the SE group brain, but was markedly increased to 7.89 ± 3.68 mm² in the IC group (Fig. 7d) and reduced to 3.29 ± 2.04 mm² in the IE group by 14 days after surgery (Fig. 7c). Two-factor factorial ANOVA revealed a significant effect of group (F_1,60 = 16.38, P=0.0003) but no effect of day (F_5,60 = 1.71, P=0.17) or of the interaction between group and day (F_10,60 = 0.204, P=0.93). A subsequent one-way ANOVA (day) revealed a significant difference among the groups with regard to the number of caspase-3-positive cells at 5 days (P<0.01) and 14 days (P=0.04), but not at other time points (data not shown). These findings showed that infarction enhanced caspase-3 expression in the ipsilateral frontoparietal cortex including the motor cortex and that treadmill exercise after MCA occlusion significantly suppressed ischaemia-induced increases in caspase-3 expression (Fig. 7c).