Subjects

The experiments were performed on 45 female Wistar rats, 3-months old (250–300 g), at the beginning of the experiments. The animals were kept in separate cages in a room with 12:12 dark/light cycles. All of the procedures were conducted with the approval of the First Ethics Committee for Animal Experimentation in Poland according to the principles of experimental conditions and laboratory animal care of the European Union and the Polish Law on Animal Protection.

The experiments started from the time of spinal cord hemisection and lasted for 5 months. Locomotor investigation was carried out before spinal cord injury (SCI) and every day during the first 2 weeks after hemisection and then 3 times per week up to the moment when recovery of locomotor performance reach a plateau (about 1 month after the surgery). In the following 2 months, locomotion was tested once a week, and then every other week up to the end of experiments (5 months after the lesion). Three to five rats were taken at consecutive time-points (1, 2, 3 and 4 weeks and then 2, 3 and 5 months) after hemisection to investigate the lesion extent and the length of regenerated serotoninergic fibers below the level of hemisection.

Spinal cord hemisection

Spinal cord lateral hemisection was conducted under deep anesthesia (2 l/min of O₂ + 2% Isoflurane and s.c. Butomidor 0.05 mg/kg b.w.). Laminectomy was performed over the Th9 vertebral segment (approx. 1.5–2.0 mm) at the thoracic level. The left half of the spinal cord was dissected at the Th11 level using fine watchmakers forceps. After SCI, the overlying muscle layers were closed using sterile sutures (Mersilk 5–0). The skin was closed with stainless-steel surgical clips. After surgery, the animals received a non-steroidal anti-inflammatory and analgesic treatment (s.c. Tolfedine, 4 mg /kg b.w.), and during the following 7 days, the animals were given antibiotics (s.c. Baytril 5 mg/kg b.w. and Gentamycin 2 mg/ kg b.w.).

Implantation of EMG electrodes

For EMG recordings, bipolar electrodes were implanted in the flexor and extensor muscles of each limb (in the forelimbs: the Biceps Brachii (Bic) and Triceps Brachii (Tri) medial head; in hindlimbs: the Tibialis Anterior (TA) and Soleus (Sol) muscles) under Equithesin (0.35 ml/100 g b.w. administered i.p.) anesthesia (see [18] for details). The electrodes were made of Teflon-coated stainless steel wire (0.24 mm in diameter; AS633, Cooner Wire, Chastworth, CA, USA). The tips of the electrodes, with 1–1.5 mm of insulation removed, were pulled through a cutaneous incision on the back of the animal, and each of the hook electrodes was inserted into the appropriate muscle and secured by a suture [18,23]. The distance between the electrode tips was 1–2 mm. The ground electrode was placed under the skin on the back of the animal some distance from the hindlimb muscles. The connector with the other ends of the wires fixed to it was covered with dental cement (Spofa Dental, Prague, Czech Republic) and silicone (3140 RTV, Dow Corning) and secured to the back of the animal. During locomotion, both of the extensors (Tri and Sol) and one flexor (TA) have only one burst of activity in each step cycle (during the stance phase and the swing phase, respectively). The Bic muscle presented biphasic activity, and thus, only bursts present during the swing phase were analyzed.

The locomotion was examined using two methods: EMG recordings and the CatWalk Gait Analysis System. For the EMG recordings, the procedure for the examination of locomotion of freely moving rats was based on a shelter-seeking paradigm. Rats were pre-trained to walk on a horizontal runway 2.5 m long and 12 cm wide with an illuminated start platform at one end and a darkened goal box at the other (for details see [23]). For CatWalk analysis, the rats were taught to move on a 1.5 m long by 20 cm wide runaway in the dark. One experimental session using both methods consisted of 10 passes. Runs during which the animals stopped or changed their speed rapidly were not subjected to further analysis. EMG analysis was performed to observe changes of temporal parameters of locomotion, while the CatWalk analysis was performed to analyze changes in spatial parameters.

During the first and second week after hemisection, locomotor movements were very unstable, and runway and CatWalk trials were not possible. Therefore, locomotion was examined using a treadmill.

Analysis of temporal parameters of locomotor abilities using EMG recordings

EMG activity of selected muscles recorded during locomotion was filtered (0.1 to 1 kHz band pass), digitized and stored on a computer (2 kHz sampling frequency) using the Winnipeg Spinal Cord Research Centre Data Capture System. The locomotor EMG pattern was analyzed using custom software (http://www.scrc.umanitoba.ca/doc/).

Assessment of spatial gait parameters using the CatWalk Gait Analysis System

CatWalk Gait Analysis System allowed us to measure the following spatial parameters: 1) base of support (BOS) defined as the distance between the lines parallel to the direction of rat locomotion at points marked by the center of the paw prints for the fore- and hindlimbs; 2) hindlimb abduction (HA) defined as the distance between the lines that are parallel to the direction of the rat locomotion and across the center of the fore- and hind footprints and measured in a given pair of steps; 3) relative print distance (RPD) was defined as the distance between the lines that were perpendicular to the direction of rat locomotion and across the center of the fore- and hind footprints and measured in a given pair of steps. For all measurements, the center of the footprint was defined as the center of mass of the print. Moreover, with CatWalk the mean speed of a given run was calculated as a mean of stride length divided by step cycle duration. Speed was calculated individually for each limb and the rat velocity was calculated as a mean of four speeds.

Serotonin fiber distribution around motoneurons

After completing all of the behavioral and EMG data rats were euthanized with an overdose of pentobarbital (i.p., 150 mg/kg) and perfused transcardially with the 40 ml of cold (4°) prefixative solution (0.9% NaCl (Polfa), 0.1% sodium nitrite (Sigma-Aldrich), 0.2% heparine (Polfa) and 0.5 M PBS (Takara Bio Inc)) to washout the blood, followed by 300 ml of cold PFA solution (4% paraformaldehyde (PFA), Sigma-Aldrich) and 0.2% picric acid (Sigma-Aldrich) in 0.5 M sodium phosphate buffer (PBS) for tissue fixation. The spinal cords were dissected out, postfixed in PFA solution for 1.5 h and then cryoprotected with 30% sucrose (Poch) solution with 0.04% sodium azide (Sigma-Aldrich) and kept at 4°C for at least one week. Then, the L4-L6 level of each spinal cord was isolated, embedded in freezing medium (Jung Tissue, Leica) and frozen on dry ice. The specimen was then placed in a cryostat at -20°C for 30 min. Fifty μm coronal sections were cut and mounted onto microscope slides (SuperFrost Ultra Plus, Thermo Scientific). The specimen was then dried and rinsed 3 times for 10 min with PBST (0.5 M PBS, 0.3% TritonX-100 (Thermo Scientific)). The slices were then incubated with primary antibody (rabbit anti-serotonin, Sigma-Aldrich), diluted 1:500 in PBST with 10% NDS (Normal Donkey Serum, Jackson Immunoresearch), overnight at 4°C. Then the primary antibody was washed with PBST and the slides were incubated with the secondary antibody (DyLight, 549-coinjugated, Donkey anti-rabbit, Jackson ImmunoResearch), diluted 1:100 in PBST with 10% NDS, for 4h at room temperature in the dark and in a humid chamber. After rinsing in PBST (3 × 10min), the sections were stained with Neurotrace Fluorescent Nissl Stains 488 (Molecular Probes), diluted 1:100 in PBST, for 40 min at room temperature. After rinsing in PBST (3 × 5min), the specimens were dried and covered with Vectashield Medium (Hard Set Mounting Medium, Vector).

Anatomical verification of the lesion site

The level of the spinal cord containing the lesion (Th10/11) was carefully dissected, embedded with freezing medium and frozen on dry ice. The specimen was placed in a cryostat at -20°C for 30 min. Then, 40 μm coronal sections were cut, mounted on microscope glasses, dried and rinsed with 0.3% PBST (3 × 10min). Next, the samples were incubated with a primary antibody against neurofilament protein (Monoclonal Mouse Anti-Human Neurofilament Protein, Dako), diluted 1:200 in PBST with 5% NDS, at 4° overnight. The following day the slides were washed with PBST and then incubated with the secondary antibody (Donkey Anti-Mouse IgG, Invitrogen), diluted 1:200 in PBST with 5% NDS, for 2 h at room temperature in the dark. Nissl bodies were visualized by incubating the slides with Neurotrace Fluorescent Nissl Stains (Molecular Probes), diluted 1:100 in PBST, for 1h in the dark. The slides where then washed and the specimens dried and covered with Vectashield Medium. All of the results presented show data from hemisected animals that were verified to have no more than 1% of the spinal cord matter preserved in the ipsilateral half of the spinal cord, as well as no more than 1% of the contralateral half of the spinal cord affected by the surgical procedure.

The specimens were examined under a Nikon Optiphot-2 epifluorescence microscope equipped with a digital camera (QICAM, 12-bit, Qimaging). The photomicrographs were obtained using a 10× magnification with Image Pro Plus v. 4.0 software. The same camera and microscope settings were always used. Quantitative analysis was conducted using ImageJ software (http://rsb.info.nih.gov/ij/) with the relevant plugins necessary to perform skeletonization measurements (AnalyzeSkeleton, author: Ignacio Arganda-Carreras http://fiji.sc/AnalyzeSkeleton). In our study, skeletonization was used to measure the length of the serotonergic fibers. Using this method, the shape of a fiber was thinned to a width of a single pixel, which was positioned equidistant to the boundaries of the analyzed fiber. The total count of pixels represents the total length of the serotonergic fibers.

Statistical analysis

Statistical analyses were carried out using one-way ANOVA, two-way ANOVA or MANOVA tests. First, we confirmed that the data was characterized by a normal distribution (with a Shapiro-Wilk test) and verified the homogeneity of the variance (with respect to MANOVA using Box M tests with respect to ANOVA—Cochran C, Hartley, Barlett test). Tukey HSD or Fisher LSD post-hoc tests were used. If the assumptions of normality or homogeneity of variance or both were violated, then the Kruskal-Wallis test was used. For correlation analysis, a Pearson’s correlation test was used.

Spinal cord lesion

Inspection of the coronal sections of the spinal cord tissue under a light microscope revealed that a complete left hemisection was present in all of the animals included in this paper. Most of the preserved tissue was confined to the right side of the spinal cord. All of data presented in this paper were collected from hemisected animals verified to have no more than 1% of the spinal cord matter preserved on the ipsilateral half of the spinal cord, as well as no more than 1% of the contralateral half of the spinal cord affected by the surgical procedure.

General observations

During the first two days after hemisection, the rats were paraplegic, and able to move using their forelimbs to drag their body and hindquarters. The left hindlimb (LH), ipsilateral to the lesion side, was passively extended and mostly paralyzed with its foot lying on its dorsal side. The right hindlimb (RH), contralateral to the lesion side, was also passively extended, but more often it was bent at the knee and lying under the rat’s belly with the dorsal side of its foot down. Small movements were observed at the knee or hip joints. In subsequent days, hindlimb locomotor performance gradually recovered. In the second to fourth day after the lesion, progressive recovery of RH locomotor functions was present. The right foot was often placed on its plantar surface and occasional body weight support was seen. In the LH, locomotor recovery was more limited and only occasional movements in the hip, knee and ankle joints occurred. Around the fifth day after hemisection, a regular RH plantar foot position was observed in most of the rats, but body weight support was only present sporadically. LH plantar stepping occasionally occurred in more than half of the hemisected animals, but body weight support was still missing. On the seventh day after hemisection, RH plantar stepping was present in all rats, although difficulties with full body weight support present. Eleven days after hemisection, plantar stepping was present in both hindlimbs, but due to intermittent LH body weight support, locomotion was very unstable, and the animals often fell to the left. When plantar stepping was present, both hindlimbs were abducted and the hindlimb base of support was much wider than in intact animals. These abnormalities persisted to the end of the study and were quantified in the CatWalk test (see Base of support (BOS) and Hindlimb abduction (HA) in the Results). Regular and sustained locomotion was regained in the majority of animals 14 days after lesion and thereafter the animals were tested using the CatWalk. It is important to underline that, in contrast to another study, [8] after day 14, all of the hemisected animals did not need assistance during locomotion for the remainder of the experimental period, and the animals moved with stable body weight support, plantar stepping and toe clearance.

Changes in temporal locomotor parameters after spinal cord hemisection

To investigate the effects of spinal cord hemisection on locomotor performance at different time points after injury EMG activity of the forelimb and hindlimb extensor and flexor muscles was recorded. These measures were used to establish changes in the temporal parameters describing the quality of limb movements. Our analysis revealed that a dramatic alteration in hindlimb locomotor abilities, induced by spinal cord hemisection, was always associated with a substantial alteration of hindlimb muscle activity. Fig 1 presents examples of EMG activity recorded during locomotion along the runway from the hindlimb muscles (left/right Sol and TA muscles) before and after spinal cord hemisection. During regular locomotor activity in intact rats, Sol and TA were active in a rhythmic alternating pattern, according to a regular step cycle. The Sol muscle, as an extensor and antigravity muscle, was active during the stance phase of each step cycle; while the TA muscle, as a dorsiflexor of the ankle joint, was active during the swing phase (Fig 1A). During the first few days after injury, EMG activity of Sol and TA muscles was limited when the animals were moving around the experimental box or the runway (Fig 1B). In subsequent days, when hindlimb locomotor performance gradually recovered after the spinal cord lesion, a progressive improvement in EMG recordings was present. Thus, starting from the 14^th day after the lesion, regular and sustained locomotion was associated with a regular rhythmic EMG activity (Fig 1C and 1D). The effects of hemisection on intra- and interlimb coordination are illustrated in the right panels of Fig 1.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Comparison of EMG activity of the forelimb and hindlimb muscles.

EMG activity during locomotion of a representative rat before (A), 7 days (B), 14 days (C) and 42 days (D) after spinal cord hemisection. Interlimb coordination (onset of EMG activity in the left TA (l TA) with respect to the contralateral right TA (r TA)) and intralimb coordination (onset of EMG activity in the left extensor (l Sol) with respect to the left TA (l TA)) for these time points are shown in the left and right polar plots, respectively. Zero corresponds to the onset of activity in the right TA muscle, and the positions of the filled black circles indicate the times of onset of activity in the left TA—interlimb coordination (left panel) or the onset of activity in the left Sol -intralimb coordination (right panel) in the relation to the onset of activity in the left TA muscle. The black lines each represent data from one animal, and the red line represents the global mean for the group. l Tri—left Triceps; l/r Sol—left and right Soleus and l/r TA—left and right Tibialis Anterior.

https://doi.org/10.1371/journal.pone.0143602.g001

Fore-hindlimb coordination (coupling ratio—CR).

Within the first week locomotor movement is confined largely to the forelimbs. Therefore, we investigated movement of the forelimbs and coordination of this activity with hindlimbs. Various methods for scoring locomotor performance, described in the literature, rely on the achievement of fore-hindlimb coordination. Some of these methods, like the BBB [5]method, use video recordings that provide limited temporal precision. Thus, in our investigation, where spinal cord hemisection produced an obvious lack of coupling between fore- and hindlimb movement, we used EMG burst activity, which is present even when hindlimbs are not successfully stepping, to analyze limb coordination in a more precise way. Based on EMG burst activity, we calculated a parameter that we termed the “coupling ratio” (CR). This is the ratio of the number of bursts of a selected hindlimb muscle to the number of bursts of a forelimb muscle during a defined period of locomotion.

During unrestrained locomotion in intact animals there is a strict one-to-one step correlation in fore-hindlimb stepping which is expressed as an alternating pattern of EMG bursts (Fig 1A). Thus, the number of EMG bursts produced by the hindlimb muscles is similar to the number of corresponding bursts produced by forelimb muscles. As expected, during locomotion in intact rats CR was close to 1 (0.959 ± 0.013 and 0.981±0.09 for l TA-Bic and r TA-Bic, respectively) (Fig 2A; S1 Table).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Changes in temporal parameters induced by spinal cord hemisection.

(A) Time course showing the changes in fore-hindlimb coordination (coupling ratio represented as the mean±SEM) of fore- and hindlimb flexor (l and r TA-Bic) and extensor (l and r Sol-Tri) muscles post hemisection. (B) Time course showing the changes in interlimb (l TA—r TA; l Sol—r Sol) and intralimb (l Sol—l TA; r Sol—r TA) coordination post hemisection (r-values are represented as the mean±SEM). (C) Time course showing the changes in the phase shift in interlimb (l TA—r TA; l Sol—r Sol) and the intralimb (l Sol—l TA; r Sol—r TA) coordination post hemisection (r-values are represented as the mean±SEM). Abbreviations: d—days post operation, w—weeks post operation, m—months post operation, * − p<0.05, ** − p<0.01, *** − p<0.001.

https://doi.org/10.1371/journal.pone.0143602.g002

In the first two days after hemisection, the animals moved primarily using their forelimbs and showed minimal hindlimb EMG bursts during locomotion on the treadmill, either in the muscles of the hindlimb ipsilateral to the lesion side or in the muscles of the contralateral hindlimb. However, in the subsequent days, bursts of EMG activity of hindlimb muscles during locomotion gradually started to appear.

Our analysis showed a significant decrease in the CR values for all four of the muscle pairs as a result of spinal cord hemisection. Kruskal-Wallis analysis revealed that hemisection produced a significant decrease in the CR values for all four of the muscle pairs (l TA-Bic: H = 44.5, p<0.001; r TA-Bic: H = 39.6, p<0.001; l Sol-Tri: H = 33.5, p<0.001; r Sol-Tri: H = 33.5, p<0.001). Multiple comparisons analysis showed differences at 5 and 7 days post-operation with respect to the CR values of all four of the muscle pairs (see Fig 2A). On day 5 post hemisection, the number of hindlimb muscle EMG bursts was much smaller than before hemisection, especially for the LH muscle (CR values were 0.152±0.060 and 0.126±0.034 for l TA-Bic and l Sol-Tri, respectively). For the RH muscle, CR values were greater than for the LH muscle (0.495±0.034 and 0.529±0.038 for r TA-Bic and l Sol-Tri, respectively), but still much smaller compared to intact animals. These CR values show that the number of EMG bursts in hindlimb muscles was smaller compared to forelimbs. Seven days after the lesion, CR values increased but were still smaller than before hemisection (0.587±0.084 (l TA-Bic), 0.423±0.087 (l Sol-Tri), 0.660±0.058 (r TA-Bic) and 0.662±0.062 (r Sol-Tri)). There was no difference between the CR values for all four pairs of muscles. By day 11, the mean CR values increased and were not markedly different from what was calculated before hemisection (0.865±0.049, l TA-Bic; 0.814±0.101, l Sol-Tri; 0.784±0.061, r TA-Bic and 0.757±0.056, r Sol-Tri). Thirty-two days after hemisection, the CR values approached 1 (0.928±0.049, l TA-Bic; 0.944±0.038, l Sol-Tri; 0.947±0.049, r TA-Bic and 0.936±0.037, r Sol-Tri). The CR values remained at a similar level until the end of the study (5 months post hemisection).

Recovery and subsequent decrement of spatial locomotor parameters after hemisection—CatWalk data analysis

CatWalk analysis was possible two weeks after hemisection, when the rats were able to engage in plantar stepping with both hindlimbs. An example from a single animal is shown in Fig 3. Before hemisection, the print distance between footfalls for the forelimbs and hindlimbs is very small, the base of support (BOS) is relatively small and no hindlimb abduction (HA) is present (Fig 4A and 4B). After injury there is a permanent change in print distance, BOS and HA (Fig 4A, 4B and 4C), and locomotor speed is reduced (Fig 4D). After 3 months, a secondary deterioration occurs in speed and print distance, when the rats begin to walk more slowly, with an increased BOS, HA and RPD. This could be due to a voluntary adaptation with the rat adjusting to the deficits associated with hemisection by adopting a wide-based gait.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. CatWalk foot prints examples before and after spinal cord hemisection.

The left and right limb foot prints are shown in red and green, respectively. Darker shades represent hindlimbs and the lighter shades represent forelimbs. (A) intact rat, (B) 3 months after hemisection and (C) 5 months after hemisection.

https://doi.org/10.1371/journal.pone.0143602.g003

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Changes in speed and spatial parameters induced by spinal cord hemisection.

(A) Changes in the hindlimb base of support (BOS) before and at consecutive time points after hemisection (mean±SEM), (B) changes in hindlimb abduction (HA) before and at consecutive time points after hemisection (mean±SEM), (C) changes in the relative print distance (RPD) before and in the consecutive time points after spinal cord hemisection (mean±SEM), (D) changes in locomotor speed (V) before and at consecutive time points after hemisection (mean±SEM). Abbreviations: w—weeks post operation, m—months post operation, * − p<0.05, ** − p<0.01, *** − p<0.001.

https://doi.org/10.1371/journal.pone.0143602.g004

Changes in serotonin innervation after spinal cord hemisection.

We analyzed serotonin innervation in the ventral horn below the lesion, where the motoneurons supplying muscles of the hindlimbs are located, assuming that the concentration of the serotonin fibers in this region may be correlated with locomotor performance. In intact animals, the serotonergic fibers were evenly distributed in left and right ventral horns (Figs 5A1–5A4 and 6A). After hemisection, the number of 5-HT fibers reduced dramatically (Fig 6A). One week after lesion, serotonergic fibers on the left, ipsilateral side of the spinal cord were almost absent, while the number on the right side was reduced (Fig 5B4 and 5B5). In the following months, the number of the fibers increased and was almost equal to their numbers in intact animals on the right side contralateral to the lesion, while on the left side it remained reduced even 5–6 months after hemisection when the last verification of the 5-HT fibers was performed (Figs 5D1–5D4, 5E1–5E4 and 6; S8 Table).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Immunolabeling of transverse spinal cord sections showing 5-HT-positive staining in intact rats and at different time points after hemisection.

The representative images show the ventral horns below the lesion (L3-L5 level) in the ipsilateral (left) and contralateral (right) side to hemisection. (A) intact rat, (B) 1 week, (C) 2 weeks, (D) 3 months, (E) 6 months after the lesion. 1 and 3 –real images, 2 and 4 –images after skeletonization.

https://doi.org/10.1371/journal.pone.0143602.g005

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. Serotonin fiber length.

(A) Serotonin fiber length (in pixels) at the lumbar spinal cord level (below the area of hemisection) in the ipsi- and contralateral sides before and at consecutive time points after injury (mean±SEM). (B) Expanded scale of the diagram in (A) to show changes in serotonin fiber length (in pixels) in the lumbar spinal cord level (below the area of hemisection) in the ipsilateral side at consecutive time points after the injury (mean±SEM). Abbreviations: w—weeks post operation, m—months post operation. * − p<0.05, ** − p<0.01, *** − p<0.001.

https://doi.org/10.1371/journal.pone.0143602.g006

To evaluate changes in serotonergic innervation in the spinal cord at different time points after hemisection, tissue specimens were assessed using a measure of serotonin fiber length on the right and left side of the lumbar spinal cord (L4-L5) in the ventral horns. To estimate the length of the 5-HT fibers, a skeletonization algorithm was used (http://fiji.sc/AnalyzeSkeleton). Our analysis shows that, in intact animals, serotonin fibers in the ventral horn of the lumbar spinal level are abundant and prominent (Fig 6A). The measurement performed using the skeletonization algorithm showed that the length of the serotonin fibers was about 30,000 pixels in each selected part of the ventral horn. After hemisection, fiber length decreased on both sides of the spinal cord. A one-way ANOVA showed that hemisection significantly affected the length of serotonergic fibers in this area both on the contralateral side of the spinal cord (F = 8.42, p<0.001) and the ipsilateral side (F = 72.41, p<0.001) (Fig 6A).

One week after hemisection, there was almost complete loss of immunoreactive fibers on the ipsilateral side of the lumbar spinal cord (Fig 6A). The mean length was 306±87 pixels and was about 100 times less than in intact animals (p<0.001). In the consecutive weeks, the length of the serotonergic fibers in the sample area gradually increased, and starting from the 4^th week after hemisection, the mean length was higher than one week after hemisection (p<0.05). Six months after injury, the mean number of serotonergic fibers was equal to 1756±307 pixels. Although this value was very low compared to the length of serotonergic fibers in intact animals, it was 6 times greater than one week after hemisection (p<0.01) (Fig 6B).

The mean length of the serotonergic fibers on the side of the spinal cord contralateral to the lesion also decreased, and one week after hemisection it was 11700±3946 pixels, which was about 60% smaller than in intact animals (p<0.005, Tukey post-hoc test). Starting in the 2^nd week after hemisection, the number of serotonergic fibers gradually increased but was still lower than in intact animals 2 weeks (p<0.01) and 3 weeks (p<0.05) after surgery. Starting from 1 month after the lesion, the mean number of serotonergic fibers on the contralateral side of the spinal cord was 26608±2952 pixels and was similar to the intact animals up to 6 months after hemisection (Fig 6A).

Recovery of coordination

One of the locomotor deficits observed within the first two weeks after hemisection was the presence of unequal rhythms of the fore- and hindlimb movements, which could be expressed as an unequal number of fore- and hindlimb steps counted in the same period of time, or as an unequal number of fore- and hindlimb EMG bursts in the flexor and extensor muscles. This observation is consistent with previous reports concerning locomotor behavior after partial spinal cord injuries in rats [8,51–53]. Here, we propose a method of quantifying this phenomenon by establishing the fore-hindlimb coupling ratio based on muscle EMG activity recordings.

After hemisection the mean coupling ratio for the ipsilateral pair of limbs was more severely reduced than for the contralateral ones, and within several days both CRs increased and did not differ from the value obtained in intact animals (11 and 14 days post hemisection for contralateral and ipsilateral pairs of limbs, respectively). Over the next few days, the CR value was close to 1, which suggested that different fore- hindlimb rhythms were occasionally present. The exact time point when restoration of equal rhythms occurs in the fore- and hindlimbs remains unclear. Górska and coworkers [53] suggested that after partial spinal cord injury in rats, this process depends on the extent of the lesion. For example, rats in which only the ventral funiculi were preserved showed recovery of equal rhythms of the fore- and hindlimbs 3.5 months after surgery. Garcia-Alias and colleagues [1] observed that 6 weeks after thoracic hemisection, only the rats that received treatment (Chondroitinase ABC or NR2D vectors or NT3-sectreting plugs or combined) were able to restore equal stride length of fore- and hindlimbs. The rats that did not receive any treatment still showed abnormal fore- hindlimb rhythms at this time-point (as indicated by the different stride lengths in the fore and hindlimbs). This result is contradictory to ours since at this time point, we did not observe any rhythmic differences between the fore- and hindlimbs.

Episodes of different rhythms have also been observed in cats. For example, [51] in cats with an over-hemisected spinal cord at the thoracic level, 2 weeks after injury, the duration of the step cycle of the hindlimbs was about 1000 ms and for the forelimbs it was 800 ms. Clearly, unbalanced rhythms of the fore- and hindlimb episodes were present. Also, Bem and collaborators [52] showed that in cats with subtotal spinal cord injury there is a difference in the duration of the step cycles between the fore- and hindlimbs.

Important elements in the mechanism of motor control responsible for producing and maintaining coordination of locomotor movements of both limb girdles are the long ascending and descending proprospinal pathways [54–56]. These pathways connect the cervical and lumbar enlargements, where the motoneurons of fore- and hindlimb muscles are located. The cells of these pathways project from the cervical to the lumbar segments and vice versa. For example, the proprospinal neurons of the cervical segment send projections to both halves of the spinal cord at the lumbar level [54]. It is believed that the proprospinal pathways are in the lateral columns, as damage of these columns induces the appearance of unequal rhythms in the fore- and hindlimbs in cats [52]. According to Górska and colleagues [53], the observed uncoupling might occur due to destruction of these fibers. The recovery of fore-hindlimb coupling observed in our experiments within 2–3 weeks after the lesion might result from increased length of the 5-HT fibers in the ipsilateral half of the spinal cord (about a 5-fold increase) as well as in the contralateral half of the spinal cord (about a 50% increase). It is also likely that the recovery of fore-hindlimb coordination depends on the connection between descending tracts from the supraspinal structures and the priopriospinal fibers, which was revealed in relation to the reticulospinal tract after lateral cervical hemisection [45].

Two other parameters characterizing the coordination of limb movements also changed after unilateral hemisection. After the lesion, the r-value, which reflects the strength of coordination between the analyzed muscle burst onsets with respect to intra- and interlimb coordination dropped dramatically from about 0.9 to about 0.5, but after 2 weeks it returned to the value obtained in intact rats. The r-value is a very sensitive tool for estimating the quality of locomotor movements, especially when they are severely deteriorated.

The timing and the pattern of muscle activity in locomotion is produced by the spinal network (CPG, Central Pattern Generator) and controlled by descending commands and sensory information from peripheral receptors. The weak correlation between muscle activities after the lesion might result from disorganization of the spinal network activity evoked by spinal shock. It is also likely that the weak descending influence on the excitability of motoneurons decreases the muscle force and evokes the loss of body weight support. During locomotion, the CPG receives misleading information from the hindlimbs that are being dragged with the dorsal surface of the foot contacting the ground, which may produce changes in the pattern of locomotion. Once the animals recover body weight support and plantar stepping, the r-value returns to its value before the lesion, despite the fact that other locomotor parameters are still changed.

In intact animals, the phase shift of intralimb coordination (between l TA-Sol and r TA-Sol) was approximately 130±8°. Similar studies of intralimb coordination are rare, however [57], in cats, the phase shift of the antagonist muscles of the knee of the ipsilateral hindlimb, vastus lateralis (extensor) in relation to semitendinosus (flexor), was 0.35 before hemisection, which corresponds to 126°. This observation is consistent with our results. Moreover, Martinez and colleagues [57] did not observe any changes in this value 3 weeks after hemisection, which is also consistent with our findings.

Our results show that after hemisection the mean phase shift of interlimb coordination (between l-r TA and l-r Sol) decreased and no improvement was observed for the duration of the study. Barierre and coworkers [51] observed a similar phenomenon using a kinematic analysis of locomotion in cats after partial spinal cord injuries. These authors describe that in cats with a hemisection-like lesion, 20 days after the injury, the phase shift of the left-right hindlimb coupling decreased from 0.5 to a value of 0.4 (corresponding respectively to 180 and 144° on a 360° scale). Martinez and coworkers [57] also observed that, in cats, locomotion assessment after spinal cord hemisection showed that the right hip flexor sartorius, as well as the right knee extensor vastus lateralis, tended to contract earlier in the cycle of left sartorius and left vastus lateralis, respectively. They claim that after spinal cord hemisection the “cats probably developed behavioral strategies to compensate for the left hindlimb deficits by using the right hindlimb more.” Our results support this hypothesis.

Recovery and subsequent decrement of spatial parameters of locomotor abilities after spinal cord hemisection—Catwalk results

In our study, before hemisection rats moved along the CatWalk runway with a mean speed of about 57 cm/s. The mean speed of intact rats tested in our experiments is within the range observed by the other authors [58,59] and our previous studies [23,53,60,61]. Two weeks after lesion, when the animals regained quadrupedal locomotion, the mean speed was about 38 cm/s, which is about 35% slower than before hemisection. In the following months, the mean speed of locomotion was similar, but 5 months after hemisection, it dropped again and was more than 50% slower compared to intact animals. Several studies have shown that after a severe spinal cord lesion, the speed of locomotion is less than before the injury [53,60,62]. The drop in speed may be due to a reduction in the propulsive forces or the loss of the body stability, because these lesions deprived the LH spinal motor centers of the supraspinal drive transmitted to the lumbar cord by the reticulospinal and vestibulospinal tracts [60]. A similar mechanism has been suggested [40] for alterations in ground reaction forces recorded in rats after ventrolateral spinal injury. But, to our knowledge, none of these previous studies described a secondary decrease in locomotion speed.

Our investigations showed that the BOS did not change in the forelimbs after hemisection, in contrast to the hindlimbs, in which the BOS increased by about 40% two weeks after hemisection. One month after hemisection the BOS decreased and was not different compared to pre-lesion values. Starting from the 2^nd month, the BOS increased gradually, and 3 and 5 month afterwards it was significantly greater than before hemisection.

An increase in hindlimb BOS is a characteristic phenomenon observed after a large spinal cord lesion or contusion [62]. BOS was used as a distinctive locomotor parameter after a lateral hemisection in many studies [1,2,6,36,63], and in most of these studies, the BOS increased starting from 2 to 10 weeks after the lesion, but a detailed time course of the changes in the BOS have not been analyzed. The initial increase in the BOS 2 weeks after the lesion was associated with very unstable hindlimb locomotor movements, resulting from decreased facilitation of the spinal network from the supraspinal structures and might indicate a strategy to deal with unstable body balance after SCI. In the weeks following, the BOS decreased, reaching the lowest value one month after hemisection at the peak of recovery and then gradually increased reaching the highest value 5 months after hemisection. A similar time course of these changes was seen for the left and right HA. This parameter was not analyzed by other authors, but we consider it potentially important for estimating the symmetry of the hind leg position during locomotion after asymmetrical lesions such as lateral hemisection. Interestingly, the abduction of the left and right hindlimb was similar at each time point after the lesion. Another parameter that changed after hemisection was the relative print distance (RPD). At moderate velocities intact rats usually place their hind foot close to the previous position of the ipsilateral fore foot. This probably makes positioning of the hindlimb, which is not performed under the visual control, more secure. In intact animals, the RPD is close to 0, but after hemisection it increased, since both of the hind feet were placed behind the fore feet. RPD reached a plateau one month after hemisection but 4 months later we observed an increase. Similar to hindlimb abduction, the RPD of the left limbs was not different from the RPD of right limbs. The above dissociation between fore- and hindlimb movements may result from weakened muscle force as a consequence of decreased facilitation from supraspinal structures associated with damaged descending spinal tracts and/or reduced connection between the fore-hindlimb spinal networks due to destroyed proprospinal neurons.

Ipsi- and contralateral alterations of serotonergic innervation after spinal cord hemisection

Our analysis showed that after hemisection at the thoracic level, the length of the serotonergic fibers in the ventral horn of the lumbar spinal cord decreased. On the ipsilateral side, the decrease was substantial, and one week after lesion, the length of the serotonergic fibers did not exceed 1% of fiber length in the intact animals. In the following weeks their length increased reaching a plateau one month after the lesion, and over the next few months the fiber length stayed at a level of 5–6% of the total fiber length observed in the ventral horn of the spinal cords of intact animals. On the contralateral side, the decrease was less pronounced, and the length of the serotonergic fibers reduced to less than 50% of the fiber length measured in the spinal cords of the intact animals. However, it cannot be excluded that a transient decrease in the immunoreactivity on the contralateral side of the spinal cord was caused by a local decrease in 5-HT synthesis rather than fiber degeneration itself. In the following weeks the length of the fibers gradually increased, and 4 weeks after the lesion they were not different in comparison to intact animals. Three months after the hemisection, the mean serotonergic fiber length on the contralateral side was almost the same as in the intact animals, but 3 months later reductions in fiber length were observed.

The increase in serotonergic fiber density on the ipsilateral side we observed is lower than previous studies using rats [9,64] or mice [65,66]. In contrast, studies of reticulospinal projection after lateral thoracic hemisection show that there was no significant increase in the fiber density in the ventral horn at the lumbar spinal level [4]. However, it should be emphasized that, in many studies, the measures of fiber changes are mainly based on the immunoreactivity index or an optical density assessment. We found that this measurement gives only a general approximation of this process, rather than closely reflecting it. The determination of the nerve fiber density using optical density measurements has several disadvantages. For example, in the same image area, a smaller number of large diameter fibers can give the same density index as a larger number of fibers with a small diameter. The other is that immunofluorescence labeling is often non-uniform in a thick section (our case), and the image is often blurred [67]. Our observations indicate that the changes in serotonergic fiber patterns were very subtle, and therefore we decided to perform further analysis using a skeletonization algorithm. This method is used for investigating the length of longitudinal biostructures like blood vessels [68]. The initial reduction in serotonergic fiber length on the contralateral side of the spinal cord has not been described previously. Moreover, studies investigating the plasticity of other descending fibers after lateral hemisection did not show any changes on the contralateral side of the spinal cord [4,9].

Our results show that after lateral hemisection at the thoracic level, an increase in the length of serotonergic fibers in the ipsilateral lumbar (6%) was identified 6 months after hemisection. Despite the initial decrease in serotonergic fiber length in the ventral horn to about 40% of that in the intact animals on the contralateral side of the spinal cord, 4 weeks after hemisection, fiber length almost returned control values. However, it should be considered that in the intact animals serotonergic fibers are overlapping and their length can be underestimated. Therefore, the stain of the intact tissue (see Fig 5A3) appears more intense than in the ventral horn 3 months after the hemisection (see Fig 5D3) despite the similar number of pixels in the skeletonized image.

Serotonergic fiber regrowth and the recovery of locomotion

The pronounced role of serotonin, its receptor agonists or antagonists, as well as serotonin transplants in the recovery of locomotion after spinal cord injuries is well established [16,17,20]. However, it is unclear how the process of serotonin fiber restoration proceeds after partial spinal cord injury and the contribution of this process to locomotor function recovery is unknown. Saruhashi and coworkers [9] showed a correlation between changes in the density of serotonergic fibers or the 5-HT immunoreactive terminals in the lumbar spinal cord and locomotor recovery.

Our analysis showed a correlation between the length of serotonergic fibers in the ipsilateral and contralateral ventral horns and the r-values that describe the strength of the coupling between the burst onsets of EMG in a given pair of muscles, either with respect to intra- or interlimb coordination.

Serotonin exerts very complex effects on processes related to locomotion generation, such as increments of excitability of motoneurons or increased activity of the central pattern generator [13,28,69–71]. Upon application onto the isolated neonatal spinal rat cord, serotonin induces the generation of alternating, rhythmic discharges in the ventral roots (fictive locomotion) [69,72]. Therefore, it is probable that an increased serotonin influx via the regrowth of fibers caused the improvement in locomotion regularity as measured by the increased r-value.

Our observation of the deterioration of locomotor hindlimb performance in the chronic stage was related to several locomotor parameters: locomotor speed; intralimb coordination; hindlimb base of support (along with this parameter also hindlimb abduction) and relative foot print position. This deterioration was not significantly correlated with a decrease in spinal cord serotonergic fiber length, although we measured this parameter only within a selected sample of the ventral horn. There might be other sites of termination that could contribute to the deterioration. Moreover, it is worth pointing out some coincidences between changes in locomotor performance (outcome) and serotonergic fibers length. The recovery of several locomotor parameters (locomotor velocity, print distance, base of support etc.) within the first 4 weeks after the lesion occurred concomitantly with the increase in the length of serotonergic fibers; from 1% to more than 5% on the ipsilateral side and from about 40% to almost 100% of the fiber length in intact animals. Also, a late deterioration of these locomotor parameters was associated with a late decrease in serotonergic fiber length on the contralateral side. We believe that changes in locomotor performance after thoracic hemisection are not related exclusively to changes in the density of serotonergic fibers and that many other descending tracts are involved in the process of recovery and the late deterioration of locomotion.

This deterioration might be due to progressive changes in the physiology of spared fibers on the contralateral side of the spinal cord. Arvanian and coworkers [7] suggested that chronic hemisection induces a pathological state in the physiology of spared spinal fibers that starts about 1–2 weeks after hemisection. They observed that the transmission through the surviving axons was delayed. Moreover, in the subsequent weeks (up to week 14), the magnitude of the lumbar motoneuron responses to thoracic- (above the lesion) derived electrical stimuli (either on the ipsilateral or contralateral side) was reduced. Another possible mechanism is the accumulation of chondroitin sulfate proteoglycans in the neighborhood of the spared axons, which could reduce their ability to conduct impulses [2,73].

In conclusion, our findings suggest that after lateral hemisection rats recover substantial locomotor ability with body weight support and plantar stepping with toe clearance about 2–4 weeks after the lesion, but several locomotor parameters did not return to pre-lesion values. Moreover, we showed for the first time that a late deterioration of locomotion occurs 4–5 months after hemisection. We associated the temporal changes in locomotor function with the length of serotonergic fibers not only on the ipsilateral but also on the side contralateral to hemisection. Of all the locomotor measures used in our study, we only found a correlation between serotonergic fiber length and limb coordination. However, we have to consider that many other descending systems are involved in the changes in locomotion that occur after spinal cord injury. More studies are warranted to test this hypothesis, for instance, locally blocking not only 5-HT receptors but also receptors involved in other descending systems at the lumbar spinal level to evaluate their importance in locomotion after spinal injury (for the method describing intrathecal drug administration see [17,74]).