Experiments were carried out on male Sprague-Dawley rats (350–400 g) obtained from the Animal Center of Chang Gung University. The rats were housed in standard cages in a temperature (25°C) and humidity (50%) controlled facility with a 12 h light/dark cycle. All animal assessment and surgical procedures were approved by the guidelines of the Institutional Animal Care and Use Committee at Chang Gung University (IACUC Approval No. CGU107-104). All efforts were made to minimize the number of rats required in the present study.

To provide a stable and controllable environment and obtain detailed mechanical insights from the TBI animal model, researchers have widely used one of the typical TBI rodent models, known as the weight-drop model, to mimic diffuse axonal injury and concussion caused by falls or motor vehicle accidents in individuals with TBI (Foda and Marmarou, 1994; Marmarou et al., 1994). The weight drop model is the use of weights that are freely dropped or through a guiding tube to generate an impact on the head. The widely recognized weight drop induced TBI model is Marmarou’s impact acceleration model, which has been described as resulting in diffused brain injury in rats (Marmarou et al., 1994; Smith et al., 2015). In this model, weight-drop procedures can provide a secure and inexpensive method for producing different graded brain injuries in animals by adjusting the height during the weight drop (1–2 m; Foda and Marmarou, 1994; Marmarou et al., 1994; Hsieh et al., 2017). In the current study, for the induction of TBI in rats, the Marmarou’s impact acceleration model was modified and applied. Animal preparations for the induction of the TBI rat model were described previously (Hsieh et al., 2017). Briefly, to minimize animal suffering and distress during TBI lesions, we anesthetized the animals using an intraperitoneal administration of tiletamine-zolazepam (50 mg/kg, i.p.; Zoletil, Vibac, France) with xylazine (10 mg/kg; Rompun, Bayer, Leverkusen, Germany) 30 min before impact. A 2-cm incision was made, and the area was carefully cleared to expose the line of bregma. The stainless steel disc (10 mm in diameter, 3 mm in thickness) was fixed with the self-adherent wrap (1582, 3M, St. Paul, MN, USA) to the central portion of the rat skull vault between the bregma and lambdoid sutures. The rats were placed prone on flexible foam and secured by using two belts. A Plexiglas tube was then positioned vertically, and the lower end of the tube was centered directly above the stainless steel disc. TBI was induced using a 450-g brass weight falling from 2 m through a vertical transparent Plexiglas tube (Figure 1A). Based on our earlier study, the averaged response of impact force and acceleration are 9.43 ± 0.27 kg and 370.28 ± 9.98 g (Hsieh et al., 2017). Under this weight drop TBI model, the different graded severity of brain injury can be reproducibly and reliably induced. Following the TBI lesion, the body temperature was monitored throughout surgery, and the temperature was maintained at 37.0 ± 0.5°C using an adjustable heating pad during recovery from anesthesia.

After the TBI lesion, a CES electrode was implanted in the rat’s skull. Four burr holes were made using a dental drill (NE213, NSK-Nakanishi Inc, Tochigi, Japan) with a 1.5-mm burr for screw electrodes (1.6-mm-diameter pole; Plastics One, Inc., Roanoke, VA). According to the stereotaxic brain atlas of Paxinos and Watson, cortical electrodes were placed epidurally (A = + 4.0 mm, L = ± 2 mm for the frontal cortex; A = −3.6 mm, L = ± 4 mm for the parietal cortex; Paxinos and Watson, 2005; Figure 1B). All electrodes were inserted into a six-channel pedestal (MS363, Plastics One, Inc., Roanoke, VA, USA). The surgical incision was closed with three stitches. The screw electrodes and pedestal were secured to the skull surface with dental acrylic (Lang Dental Mfg., Wheeling, IL, USA; Figures 1C,D). Following TBI lesion and CES electrode implantation, the analgesia (Carprofen, 5 mg/kg; Pfizer Animal Health Inc., PA, USA) was administered subcutaneously every 24 h for 48 h postoperatively.

Twenty-three rats were used for the present experiment. Following TBI lesion, a 30.4% (7/23) mortality rate was observed following weight-drop induced TBI. 16 TBI rats were utilized to observe the efficacy of CES treatment. Animals were randomly divided into the sham CES treatment group and CES treatment group (n = 8 in each group). For the stimulation protocol of CES, we applied a popular and specific rTMS paradigm, the continue theta-burst stimulation (TBS; cTBS) and intermittent theta-burst stimulation (iTBS) protocol, have been proposed for inducing more efficient long-term potentiation (LTP) or long-term depression (LTD)-like plasticity in the motor cortex beyond the short period of stimulation and lower intensity (Huang et al., 2005; Fitzgerald et al., 2006; Khedr et al., 2007). The basic pattern of cTBS or iTBS consisted of three pulses at 50 Hz and repeated every 200 ms (Figure 2). In the cTBS paradigm, the stimuli were given in a continuous train lasting 40 s (i.e., 600 bursts; Huang et al., 2005). The iTBS scheme was given in a 2-s train and repeated every 10 s for 20 cycles (190 s, total 600 pulses). The stimulus intensity was set at 80% resting motor threshold (RMT). The RMT was defined as the minimal intensity of CES required for eliciting minimal forelimb muscle twitches. Under the intensity at 80% RMT for CES treatment, no obvious muscle twitches were observed during CES treatment.

In the CES treatment group, the CES intervention protocols were divided into two stages: the acute stage and the chronic stage. In the acute stage (24 h-1 week post-TBI), the CES treatment protocol using continuous theta-burst stimulation (cTBS) was designed for the suppression of the hyperexcitability cascade, which may prevent or minimize some of the disabling consequences of TBI and have a potential therapeutic effect (Demirtas-Tatlidede et al., 2012; Villamar et al., 2012). In the chronic stages (>1 week), the CES parameter using intermittent theta-burst stimulation (iTBS) was set to modulate or increase brain plasticity, which could be useful to reduce functionally maladaptive changes to counter disability (Demirtas-Tatlidede et al., 2012). One day after the TBI lesion, the TBI animals in the CES group received the CES-cTBS protocol at an intensity of 80% RMT for 40 s daily for five consecutive days on the first 7 days. In the subacute and chronic stage, from 7 days to 28 days post-TBI, the CES-iTBS protocol (1 session/day, five consecutive days/week, pulse intensity = 80% of the RMT) was carried out to evoke neural facilitation (Figure 2). In the sham CES treatment group, the TBI rats also experienced the same CES protocol but did not receive any electrical stimulation at the same time points. All TBI rats were allowed to move freely during the CES or sham CES intervention. Behavioral tests, including open field locomotor activity tests, adhesion removal tests, beam walking, and mNSS, were performed at baseline (pre-TBI) and 1, 3, 7, 14, 21, and 28 days post-TBI lesion. For the cognitive measure in TBI rats, because the exploratory behavior could be influenced by the impairment of locomotor ability during the novel object recognition test, to avoid this potential confound, the novel object recognition test was assessed pre-lesion and at 28 days post-TBI lesion. Immunohistochemistry analysis was applied on day 28 post-TBI lesion to identify the changes in neural inflammation levels following CES treatment.

A well-trained examiner was blinded to the type of intervention and performed all examinations before and after sham or CES treatment. All experimental animals were trained and pretested for these tasks at least 3 days before TBI lesion to record the baseline level (as pre-TBI data). After habituation and training, all behavioral test sessions were performed at our set time points under the same environmental conditions. Behavioral tests measured the time-course changes in sensorimotor and cognitive functions associated with TBI, i.e., mNSS, adhesion removal tests for sensory function, beam walking tests for balance function, and novel object recognition tests for short-term recognition memory. There was at least a 4-h break between behavioral tests to avoid possible interference.

After behavioral tests at 28 days post-lesion, TBI rats were sacrificed for glial fibrillary acidic protein (GFAP) staining. Briefly, brains were postfixed in a 4% paraformaldehyde fixative solution (PFA) and cytoprotected in 30% sucrose solution for 48 h at 4°C until the brain sank. The cerebral tissues between −3.00 and −3.36 mm to bregma were sectioned into coronal blocks at a thickness of 30 mm on a cryostat (Leica CM3050S Cryostat, FL, US), and the areas of the frontal cortex, corpus callosum, and hippocampus were selected. The sections were quenched with 0.3% H₂O₂/PBS for 10 min and 10% milk (Anchor Shape-up, New Zealand) for 1 h to block nonspecific antibodies and then incubated with rabbit primary anti-GFAP (1:1,000, AB7260, Millipore, USA) for 1 h at room temperature. After the sections were washed three times with PBS, they were incubated with the secondary anti-rabbit antibody (1:200, MP-7401, Vector Labs, USA) for 1 h at room temperature. The sections were developed by using a solution of 3.3-diaminobenzidine (DAB, SK-4105, Vector Labs, USA) for 5 min. Next, the sections were dehydrated in graded alcohols, cleared in xylene, and mounted with DPX. Mounted coronal sections were digitally imaged at 40x optical zoom (0.25 μM/pixel) using a digital pathology slide scanner (Aperio CS2, Leica Biosystems Inc. Buffalo Grove, IL, USA). For revealing the different expression levels of GFAP-positive cells between sham CES and CES rats, the higher magnification pictures were selected and captured from the Aperio ImageScope viewer software for further quantification and of GFAP-positive cells and observation of the morphology of GFAP-immunoreactive astrocytes. The consistent regions of interest for the frontal cortex, corpus callosum, and hippocampus were manually outlined according to the atlas of Paxinos and Watson (Paxinos and Watson, 2005). The obtained images with various degrees of GFAP expression were converted to binary (8-bit black-and-white) images. The binary threshold was determined to capture the GFAP-positive cells in the regions of interest while minimizing background staining and were kept constant for all images. The particle size used in particle analysis was set so that almost all astrocytes could be detected (Wakasa et al., 2009). The numbers of GFAP-positive cells in each region were counted by means of particle analysis under a computer-based image analysis system (Image-pro, Media Cybernetics, Bethesda, MD, USA) and then manually validated by two investigators in order to ensure the correct identification of immunoreactivity patterns. The density of GFAP-positive cells was calculated by individually counting the number of GFAP-positive cells within the region and was expressed as the mean numbers of cells per mm² (cells/mm²) for further statistical analysis.

Data were analyzed using SPSS version 21.0 with the significance level set as p < 0.05 for each analysis. All data are presented as mean ± SEM. The effect of CES on the behavioral tests (i.e., beam-walking test, adhesion-removal test, mNSS, locomotor activity) was evaluated by a two-way repeated-measures analysis of variance (ANOVA) with the group (CES and sham CES treatment) as the between-subject factor and time (pre, every week after sham or CES treatment over 4 weeks) as a within-subject factor. For the short-term recognition memory, a two-way repeated measures ANOVA was used with two-way with the group (real CES and sham CES treatment) as the between-subject factor and time (pre and after sham or CES treatment over 4 weeks) as a within-subject factor. Unpaired t-tests were performed to compare groups at each time point when the main effect of the group was significant. Furthermore, the post hoc Fisher’s LSD tests were also used to compare time points on behavioral and immunohistochemical data.

The neurological function of rats was rated by the mNSS, which is a multifunctional evaluation scale that comprises balance, sensory, motor, and reflex tests. Repeated measures ANOVA identified significant main effects of time (F_{6, 84} = 148.373, p < 0.001) and group (F_{1, 14} = 11.738, p = 0.004) and a significant time × group interaction (F_{6, 84} = 2.154, p = 0.04). When compared with baseline value, subsequent post hoc Fisher’s LSD tests demonstrate that the mNSS score was significantly increased at day 1 after TBI lesion (p < 0.001) and remained maintained in the high level for up to 28 days after TBI lesion (all p < 0.001 in both groups). For the comparison of the mNSS score between two groups, post hoc t-tests between the two groups showed that the scores reached significant differences at day 7 (p = 0.02), day 14 (p = 0.031), day 21 (p = 0.015), and day 28 (p = 0.049) after TBI lesion (Figure 3).

In the open field test, the overall distance traveled was calculated to investigate the general locomotor activity between the CES group and the sham CES group following the TBI lesion. Repeated measures ANOVA indicated a significant main effect of time (F_{6, 84} = 21.010, p < 0.001) and a time × group interaction (F_{6, 84} = 2.575, p = 0.024) but not an effect of group (F_{1, 14} = 3.799, p = 0.072). When compared with baseline value, subsequent post hoc Fisher’s LSD tests show that locomotor activity was significantly decreased at day 1 after TBI lesion (p < 0.001) and remained maintained in the low level for up to 28 days after TBI lesion in the sham treatment group (all p < 0.001). However, in the CES treatment group, when compared with baseline value, post hoc tests show that the locomotor activity was significantly decreased on day 1 and day 3 after TBI lesion (p < 0.01). No significant differences were found between baseline and the time points of observation after 7 days post TBI lesion in the CES treatment group. For the comparison of locomotor activity between the two groups, post hoc t-tests between the two groups showed that the distance traveled in the open field test was significantly different at day 21 (p = 0.031) and day 28 (p = 0.008) after TBI (Figure 4A).

Furthermore, the beam walking test was applied to investigate the balance and motor coordination of the TBI rats with or without CES invention (Figure 4B). Repeated measures ANOVA identified significant main effects of time (F_{6, 84} = 6.874, p < 0.001) and group (F_{1, 14} = 6.844, p = 0.02) but not a time × group interaction (F_{6, 84} = 1.722, p = 0.126). When compared with baseline value, subsequent post hoc Fisher’s LSD tests show that latency to traverse the beam was significantly increased at day 1 after TBI lesion (p < 0.001). No significant differences were found between baseline and the time points of observation after 3 days post TBI lesion in the sham treatment group. For the comparison of beam balance function between two groups, post hoc t-tests between the two groups showed that the latency in the beam walking test was significantly different at day 1 (p = 0.049) and at day 28 (p = 0.039) after TBI (Figure 4B).

The adhesive removal test was adopted in this study to observe sensory function in TBI rats. Repeated measures ANOVA applied to the removal time indicated significant main effects of time (F_{6, 84} = 7.012, p < 0.001) but not group (F_{1, 14} = 1.731, p = 0.209) or the time × group interaction (F_{6, 84} = 0.814, p = 0.562). When compared with baseline value, subsequent post hoc Fisher’s LSD tests show that sensory function was significantly decreased at day 1 after TBI lesion in both groups (p < 0.01). No significant differences were found between baseline and the time points of observation after 3 days post-TBI lesion in both groups (all p > 0.05). For the comparison of locomotor activity between the two groups, post hoc t-tests between the two groups showed that removal time showed a significant difference at day 14 (p = 0.019) and day 21 (p = 0.037) after TBI lesion (Figure 5).

Novel object recognition was used to evaluate short-term recognition memory based on the tendency of rats to discriminate between familiar and new objects. Repeated measures ANOVA indicated a significant main effect of time (F_{1, 14} = 5.28, p = 0.038) and time × group interaction (F_{1, 14} = 5.15, p = 0.04) but not an effect of group (F_{1, 14} = 2.92, p = 0.109). Independent t-tests between the CES and sham CES groups showed that the discrimination index was significantly different at day 28 after TBI lesion (Figure 6; t = 2.67, p = 0.018).

With regard to the effects of the 4-week CES intervention on GFAP positive cells, the results of GFAP immunohistochemistry in the frontal cortex, corpus callosum, and hippocampus are shown in Figure 7. The densities of GFAP-immunoreactive astrocytes in the frontal cortex and corpus callosum were similar between the two groups (Figures 7A,B,E,F). When compared with sham CES-treated rat, lower density of GFAP-positive cells in the hippocampus of CES-treated rat was found (Figures 7C,D). No morphological abnormalities or obvious hypertrophy in the GFAP-positive astrocytes were found between sham CES and CES rats. Independent t-tests between the CES and sham CES groups showed that the density of GFAP-positive cells was significantly lower in the CES treatment group compared with the sham CES group in the hippocampus (p = 0.033; Figure 7G).