In total, 48 adult male Sprague–Dawley rats (weighing 280–300 g) were purchased from the Southern Medical University Experimental Animal Centre (Guangzhou, China) and randomly divided into four groups: sham (n = 12), injury (n = 12), T50 (intraperitoneal administration of 50 mg/kg trihydroxyethyl rutin once daily, n = 12), and T100 (intraperitoneal administration of 100 mg/kg trihydroxyethyl rutin once daily, n = 12). Members of the sham group underwent laminectomy, as well as non-injurious contact of the injury probe to the exposed cord to control for the laminectomy; a “preparatory touch” was applied before actual injury in the other groups. Members of the spinal cord injury, T50, and T100 groups underwent laminectomy and spinal cord compression injury.

The cervical spinal cord hemi-contusion model was established using a previously published method (Liu et al., 2019a). An incision was made in the skin and muscle covering the C4–C6 cervical segment. A C5 laminectomy was performed without damaging the spinal cord during removal of the dorsal lamina; each rat was placed in a stereoscopic frame, with clamps on the spinous processes at the C4 and C6 segments to maintain stability during the injury procedure.

An electromagnetic probe (ElectroPuls E1000 testing machine; Instron, Canton, MA, United States) was oscillated in air and lowered to the surface of the exposed left spinal cord, targeted to the posterior median artery; upon activation, it caused a 1.2-mm displacement at a rate of 500 mm/s. Rats in the sham group were not subjected to the injury procedure. Subsequently, the rat was removed from the spinal frame and subjected to three layers of muscle closure, followed by skin closure. Each rat was administered intramuscular penicillin (400,000 units/day) to prevent infection.

Two weeks before rats were subjected to spinal cord injury, they were trained to comfortably grip a mechanical detector and perform a forelimb asymmetry task. These tasks were specifically designed to measure basic and advanced motion. After the rats exhibited proficiency in these tasks, they were filmed for pre-injury data collection.

Grip strength was assessed as described previously (Aguilar and Steward, 2010). Briefly, the grip strength was tested using a YISIDA mechanical detector (YISIDA Technology Co., Qingdao, China). The longitudinal part of a 0.15-cm-diameter “T” type stainless steel strip was attached to the detector, and the side part was gripped by the rats. The experimenter’s left hand fixed the rat’s tail and torso; the right hand was placed in front of the rat’s torso to avoid interference from the right paw, thereby allowing the left front paw to move autonomously and flexibly. After the rat had grasped the test rod, the rod was moved backward in a smooth manner; the detector showed the maximum tensile force. Each rat was tested three times, and the maximum value was taken as the maximum grip strength.

The forelimb asymmetry task, also known as the cylinder test, was used to measure changes in forelimb use during vertical exploration. Each rat was placed in a clear plastic cylinder and spontaneous exploratory behaviors were observed for 15 min. Slow-motion video playback was used to record instances in which the rat placed either its ipsilateral, contralateral, or bilateral limbs against the side of the cylinder during 20 weight-supported movements, in accordance with a previously published method (Schallert et al., 2000; Mondello et al., 2015). The proportion of ipsilateral limb placements relative to the total number of limb placements (averaged across all rats in each group) was calculated, and the values were compared across groups.

Forelimb somatosensory evoked potentials (SEPs) and motor evoked potentials (MEPs) were recorded preoperatively, and on postoperative days 3, 7, 14, 28, 42, 56, 70, and 84. Each rat was anesthetized with low-dose isoflurane; the head fur was then shaved and the skin was disinfected with 75% ethanol. A thermostatic heating pad was placed under the rat to maintain a body temperature of 37°C. Stimulation was delivered using a 16-channel neuromonitoring system (Yizhi Technology Co., Ltd., Hangzhou, China).

Continuous stimulation was used to generate a 5.3-Hz, 0.2-ms² wave current for continuous stimulation of the median nerve of the forelimb; the current intensity was sufficient to induce slight forepaw contraction. The recording electrode was placed on the Cz–Fz axis of the cortex for recording. During a single test, 300 SEP signals were obtained to reduce the signal-to-noise ratio. The signal strength was amplified 100,000-fold by two amplifiers, and the waveform was filtered into 2–2,000-Hz waves using a filter device. SEP peak latency was measured from stimulus initiation to the peak of the first negative peak. Amplitude was calculated as the voltage difference between the first positive and first negative peaks (Forgione et al., 2014).

Constant stimulation was used to generate MEPs. Electrodes were attached to the skull for anodal stimulation. A train of six constant current pulses (8–12 mA, 0.05 ms, 500 Hz) was delivered using a 16-channel neuromonitoring system. MEPs were recorded bilaterally from a single muscle group using two 1.2-cm needle electrodes placed approximately 1 cm apart. The recording electrodes were placed in muscle groups located on the mid-shaft of the forelimb (Morris et al., 2017).

Twelve weeks after cervical spinal cord contusion, five rats in each group were randomly selected for 7.0 T MRI of the cervical spinal cord (ClinScan; Bruker, Berlin, Germany). To prevent respiratory movement and other factors from interfering with the scans, in vitro scanning of the cervical spinal cord tissue was performed. After each rat had been deeply anesthetized with Beuthanasia (phenytoin/pentobarbital mixture, 200 mg/kg), cold (4°C) heparin-saline (50 mg/500 mL) and 4% paraformaldehyde solution was perfused through the left ventricle. The cervical spinal cord tissue was removed and placed in 4% paraformaldehyde solution, and then immediately subjected to MRI scanning. The scanning parameters were as follows: repetition time/echo time, 2,200 ms/126 ms; image matrix, 256 × 256; and slice thickness, 1 mm (Li et al., 2015; Moon et al., 2018). Cervical spinal cord tissue was scanned in the coronal, sagittal, and transverse planes. The scans were stored in DICOM format and viewed using Sante DICOM Viewer software (Santesoft, Nicosia, Cyprus). Pseudo-color was added to cervical spinal cord images to increase the contrast; the damaged area was calculated using ImageJ software (National Institutes of Health, Bethesda, MD, United States).

At the end of the trial (12 weeks after cervical spinal cord injury), one rat was randomly selected from each group for cervical spinal cord microvascular perfusion using a barium sulfate solution (medical gelatin, 5 g; barium sulfate, 20 g; physiological saline, 100 mL) (Liu et al., 2019a, b). Perfused cervical spinal cords were fixed to a small cystosepiment (U80810; outer diameter, 20 mm; inner diameter, 18.5 mm; height, 65 mm) for micro-computed tomography (CT) scanning (μCT 80; Scanco Medical AG, Switzerland) with an isotropic voxel length of 8 μm (55 kVp, 145 μA; integration time, 300 ms [averaged over two scans]). Each sample was embedded in the center of a sponge and then placed in a rotating focus X-ray tube for 360° scanning to obtain tomograms. Three-dimensional reconstruction images were obtained using the micro-CT analysis system.

CD34 immunohistochemical staining for cervical spinal cord microvessels was performed 12 weeks after spinal cord injury, using a previously published method (Schallert et al., 2000). Rats were perfused with cold (4°C) heparin-saline (50 mg/500 mL) and 4% paraformaldehyde through the left ventricle. The spinal cords were collected, fixed in 4% paraformaldehyde for 24 h, and cryopreserved in phosphate-buffered saline containing 30% sucrose. The specimens were incubated at room temperature in optimal cutting temperature compound for 1 h, and then placed in the freezer (embedded in optimal cutting temperature compound) in the coronal position and rapidly frozen with dry ice. Specimen blocks were stored at −80°C before further processing. Spinal cords were cut coronally at intervals of 20 μm. The resulting sections were soaked with 3% hydrogen peroxide to block endogenous peroxidase activity and then incubated overnight with rabbit polyclonal anti-CD34 antibody (1:1,000; Bioworld Technology, St. Louis Park, MN, United States). Microvessel density was calculated using Weidner’s microvascular counting method (Weidner, 1996). The number of positive cells in each group was quantified in three random fields of normal tissue around the epicenter on the injury side.

Spinal cord tissue was dehydrated using a sucrose gradient (10, 20, and 30%). One-centimeter-long cervical spinal cord tissue was harvested with C5 segments as described above, embedded in optimal cutting temperature compound. Each block was cut using a cryostat into 20-μm-thick coronal sections and stored in a −80°C freezer. Tissue staining was performed on 10 slices at 200-μm intervals. Sections were then subjected to an ethanol dehydration gradient (water for 5 min, 55% ethanol for 5 min, 95% ethanol for 5 min, 100% ethanol for 5 min and incubation in xylene for 10 min). The steps were reversed to gradually rehydrate the tissue. Sections were then incubated in Eriochrome cyanine (Sigma, St. Louis, MO, United States) for 10 min, washed in water, and differentiated in 1% ammonium hydroxide (Sigma) for 30 s. The slices were dehydrated using the gradual ethanol gradient described above. The resulting slides were sealed with neutral resin gel and observed via microscopy. The lesion area was calculated using ImageJ; the largest part of the lesion was defined as the damage center. The total area of the damaged side, residual area, and residual gray matter area were outlined; the lesion area was then divided by the healthy tissue area to calculate the relative lesion area.

Statistical analysis was performed using SPSS Statistics (version 20.0; SPSS Inc., Chicago, IL, United States). All measurement results are expressed as mean ± standard error of the mean. Student’s t-test was used for comparison of independent variables between two groups. Two-way analysis of variance, with time as a repeated measure, was used to evaluate significant differences among all groups. P < 0.05 was considered statistically significant for all tests.

The displacement depth and speed of the material testing machine were recorded by the engineer before creating the contusions, to ensure that the error was within 1%. In the injury group, the biomechanical results were as follows: displacement, 1.210 ± 0.003 mm; velocity, 507.700 ± 2.693 mm/s; and maximum load force, 1.000 ± 0.077 N. In the T50 group, the respective values were 1.189 ± 0.004 mm, 502.400 ± 2.408 mm/s and 1.027 ± 0.069 N; and in the T100 group, they were 1.197 ± 0.003 mm, 504.900 ± 1.366 mm/s and 0.953 ± 0.066 N. There were no significant differences in biomechanical parameters among the groups (Figure 1).

The maximum grip strength of the forepaw was measured by a mechanical tester. Two weeks before contusion, the rats were trained to grasp the strip. In a quiet environment, the rats were prompted to resist the pulling movement and the maximum tensile force was recorded. On the third day after contusion, a significant decrease in grip strength was observed in each group. At 12 weeks after contusion, grip strength was significantly lower in the injury (1.21 ± 0.07 N), T50 (1.64 ± 0.11 N), and T100 (1.59 ± 0.14 N) groups compared with the sham group. Grip strength recovery was significantly superior in the T50 and T100 groups than in the injury group (P = 0.01 and P = 0.0391, respectively) (Figure 2).

Normal rats generally use the bilateral forelimbs for climbing and touching. However, after cervical spinal cord injury, the injured side of the forelimb is used significantly less; the severity of cervical spinal cord injury and subsequent recovery were assessed herein based on this phenomenon. The frequency of initial limb placement in each group significantly decreased at 1 week after injury. At 12 weeks after contusion, the frequency of initial limb placement in the injury, T50, and T100 groups was 39.87 ± 3.56%, 41.5 ± 3.59%, and 40.56 ± 2.54%, respectively. The frequency of initial placement was significantly lower in the injury than sham group (P = 0.0479), and there was no significant pairwise difference between the T50 and injury group or T100 and injury group (P = 0.4901 and P = 0.8519, respectively). The frequency of total limb placement frequency in each group significantly decreased at 1 week after injury. At 12 weeks after contusion, the frequency of initial total placement in the injury, T50, and T100 groups was 36.63 ± 2.75%, 45.74 ± 2.26%, and 42.09 ± 2.23%, respectively. The frequency of total placement was significantly lower in the injury than sham group (P = 0.0113), and there was no significant pairwise difference between the T50 and injury group or T100 and injury group (P = 0.3254 and P = 0.8519, respectively) (Figure 3A). However, the total limb placement frequency was significantly higher in the T50 than injury group at 2, 4, 6, 10, and 12 weeks after contusion (P = 0.0428, P = 0.0217, P = 0.0069, P = 0.0336, and P = 0.0178, respectively), and the total limb placement frequency was significantly higher in the T100 than injury group at 2, 6, 8, and 10 weeks after contusion (P = 0.0280, P = 0.0452, P = 0.0498, and P = 0.0167, respectively) (Figure 3B).

The signal of a single SEP was superimposed on multiple signals room the flexor of the forelimb to reduce the signal-to-noise ratio. The signal intensity was amplified 100,000-fold by two amplifiers, and the waveform was filtered into 2–2,000-Hz waves using a filter device. At the preoperative baseline, the latency and amplitude of the left forelimb SEPs in each group were generally similar (approximately 9.1 ms and 9.2 μV, respectively) among the groups. The latency was prolonged in each group at 3 days after injury. The latency at 12 weeks after injury was 11.27 ± 0.38, 9.98 ± 0.39, and 9.92 ± 0.17 ms, respectively, in the injury, T50, and T100 groups. The SEP latency was significantly longer in the injury than sham group (P = 0.0375), and there was no significant pairwise difference between the T50 and injury group or T100 and injury group. However, the latency was shorter in the T50 than injury group (P < 0.05) at all-time points from week 1–12; the latency was also shorter in the T100 than injury group (P < 0.05) at 1, 4, 8, and 12 weeks. The SEP amplitude decreased on the third day after injury in all groups. The SEP amplitude in the injury, T50, and T100 groups at 12 weeks after injury was 5.07 ± 0.60, 7.64 ± 0.71, and 6.95 ± 1.04 μV, respectively. The SEP amplitude was significantly lower in the injury, T50, and T100 groups than in the sham group (P < 0.0001, P = 0.0311, and P = 0.0178, respectively), and there was no significant pairwise difference between the T50 and injury group or T100 and injury group. However, the SEP amplitude was higher in the T50 than injury group (P < 0.05) at all-time points from 3 days to 6 weeks, and at 12 weeks; the SEP amplitude was higher in the T100 than injury group (P < 0.05) at all-time points from 3 days to 8 weeks, and at 10 weeks (Figure 4).

MEPs arise from a single stimulation of the cerebral motor cortex, and are received by subcutaneous cortical electrodes placed on the left and right forelimbs. At the preoperative baseline, the latency and amplitude of left forelimb MEPs in each group were generally similar (approximately 12 ms and 15,35 μV, respectively) among the groups. The latency was prolonged in each group at 3 days after injury. The latency at 12 weeks after injury was 13.56 ± 0.74, 11.92 ± 0.07, and 12.11 ± 0.16 ms in the injury, T50, and T100 groups, respectively. The MEP latency was significantly longer in the injury than sham group (P = 0.0023), and significantly shorter in the T50 and T100 groups than in the injury group (P = 0.0471 and P = 0.0250, respectively). The MEP amplitude decreased on the third day after injury in all groups. The MEP amplitude in the injury, T50, and T100 groups at 12 weeks after injury was 801.20 ± 101.70, 1,376.00 ± 157.00, and 1,141.00 ± 80.77 μV, respectively. The MEP amplitude was significantly lower in the injury than sham group (P < 0.0001), and was significantly higher in the T50 and T100 groups than in the injury group (P = 0.0079 and P < 0.0001, respectively) (Figure 5).

At 12 weeks after injury, five rats were randomly selected from each group for cervical spinal cord MRI scanning. This examination revealed the boundary of the gray matter, which was useful for determining recovery from spinal cord injury in the coronal, sagittal, and transverse planes. The coronal lesion area in the injury, T50, and T100 groups was 4.19 ± 0.29, 1.03 ± 0.15, and 0.76 ± 0.12 mm², respectively (Figures 6A1,B4,C); the sagittal lesion area was 2.11 ± 0.16, 0.48 ± 0.08, and 0.55 ± 0.06 mm², respectively (Figures 6A2,B5,D); and the transversal lesion area was 1.37 ± 0.30, 0.53 ± 0.10, and 0.55 ± 0.06 mm², respectively (Figures 6A3,B6,E). The coronal, sagittal, and transversal lesion areas were significantly smaller in the T50 and T100 groups than in the injury group (P < 0.05) (Figure 6).

At the end of the experiment, 12 weeks after cervical spinal cord injury, one rat was randomly selected from each group for assessment of cervical spinal cord microvascular perfusion; it was placed in the micro-CT device to scan the central part of the injury. The results of the three-dimensional reconstruction of cervical spinal cord microvessels are shown in Figure 7. Notably, vascular density was significantly reduced in the injury group, particularly in terms of the microvessels. Vascular density was also significantly reduced in the T50 group, more obviously than in the injury group; however, the number of microvessels was not significantly reduced in the T100 group. Because of the small number of rats in each group, statistical analysis could not be performed. Therefore, microvascular changes were analyzed by microvascular immunohistochemical staining.

CD34 is a specific protein marker of vascular endothelial cells, and the density of blood vessels in tissues can be determined by CD34 immunohistochemical staining. We stained cervical spinal cord microvessels using this method to determine changes in microvessels among groups. At 12 weeks after spinal cord injury, the microvascular density in the injury, T50, and T100 groups was 23.50 ± 1.75, 31.33 ± 1.36, and 33.00 ± 1.27, respectively. The microvascular density was significantly lower in the injury, T50, and T100 groups compared with the sham group. Moreover, vascular density was significantly higher in the T50 and T100 groups than in the injury group (P = 0.0054 and P = 0.0013, respectively) (Figure 8).

Each set of slices was separated by 200 μm to observe continuous changes from the center to both ends of the lesion. Damage was mainly seen on the ipsilateral side, in the form of spinal cord atrophy, cavity formation, and a reduction in the gray matter area; the contralateral side was generally intact and exhibited normal morphology. Slice staining showed that the injured area was significantly larger in the T50 and T100 groups than in the injury group (Figure 9D). The lesion epicenter was largest in the injury group (Figure 9A), while the lesion area was significantly smaller in the T50 and T100 groups (Figures 9B,C) than in the injury group (Figure 9E). The areas of residual parenchyma and residual gray matter (Figures 9F,G) were significantly larger in the T50 and T100 groups than in the injury group.